Textbook of Clinical Trials
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 Textbook of Clinical Trials
Edited by
David Machin Division of Clinical Trials and Epidemiological Sciences, National Cancer Centre, Singapore and UK Children’s Cancer Study Group, University of Leicester, UK. Simon Day Medicines and Healthcare products Regulatory Agency, London, UK Sylvan Green Arizona Cancer Centre, Tucson, Arizona, USA
Section Editors Brian Everitt Institute of Psychiatry, London, UK Stephen George Department of Biostatistics and Informatics, Duke University Medical Centre, Durham, NC, USA Copyright 2004 John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England
Telephone (+44) 1243 779777
Email (for orders and customer service enquiries): [email protected] Visit our Home Page on www.wileyeurope.com or www.wiley.com
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP, UK, without the permission in writing of the Publisher. Requests to the Publisher should be addressed to the Permissions Department, John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England, or emailed to [email protected], or faxed to (+44) 1243 770620.
This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the Publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought.
Other Wiley Editorial Offices
John Wiley & Sons Inc., 111 River Street, Hoboken, NJ 07030, USA
Jossey-Bass, 989 Market Street, San Francisco, CA 94103-1741, USA
Wiley-VCH Verlag GmbH, Boschstr. 12, D-69469 Weinheim, Germany
John Wiley & Sons Australia Ltd, 33 Park Road, Milton, Queensland 4064, Australia
John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01, Jin Xing Distripark, Singapore 129809
John Wiley & Sons Canada Ltd, 22 Worcester Road, Etobicoke, Ontario, Canada M9W 1L1
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books.
Library of Congress Cataloging-in-Publication Data
Textbook of clinical trials / edited by David Machin ... [et al.]. p. cm. Includes bibliographical references and index. ISBN 0-471-98787-5 (alk. paper) 1. Clinical trials. I. Machin, David.
R853.C55T49 2004 610 .72 4–dc22 2003065147
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
ISBN 0-471-98787-5
Typeset in 10/12pt Times by Laserwords Private Limited, Chennai, India Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire This book is printed on acid-free paper responsibly manufactured from sustainable forestry in which at least two trees are planted for each one used for paper production. Contents
Contributors ...... vii
Preface ...... xi
INTRODUCTION ...... 1
1 Brief History of Clinical Trials ...... 3 S. Day and F. Ederer 2 General Issues ...... 11 David Machin 3 Clinical Trials in Paediatrics ...... 45 Johan P.E. Karlberg 4 Clinical Trials Involving Older People ...... 55 Carol Jagger and Antony J. Arthur 5 Complementary Medicine ...... 63 P.C. Leung
CANCER 85
6 Breast Cancer ...... 87 Donald A. Berry, Terry L. Smith and Aman U. Buzdar 7 Childhood Cancer ...... 101 Sharon B. Murphy and Jonathan J. Shuster 8 Gastrointestinal Cancers ...... 117 Dan Sargent, Rich Goldberg and Paul Limburg 9 Haematologic Cancers ...... 141 Charles A. Schiffer and Stephen L. George vi CONTENTS
10 Melanoma ...... 149 P.-Y. Liu and Vernon K. Sondak 11 Respiratory Cancers ...... 159 Kyungmann Kim
CARDIOVASCULAR ...... 173
12 Cardiovascular ...... 175 Lawrence Friedman and Eleanor Schron
DENTISTRY AND MAXILLO-FACIAL ...... 191
13 Dentistry and Maxillo-facial ...... 193 May C.M. Wong, Colman McGrath and Edward C.M. Lo
DERMATOLOGY ...... 209
14 Dermatology ...... 211 Luigi Naldi and Cosetta Minelli
PSYCHIATRY ...... 233
15 Overview ...... 235 B.S. Everitt 16 Alzheimer’s Disease ...... 243 Leon J. Thal and Ronald G. Thomas 17 Anxiety Disorders ...... 255 M. Katherine Shear and Philip W. Lavori 18 Cognitive Behaviour ...... 273 Nicholas Tarrier and Til Wykes 19 Depression ...... 297 Graham Dunn
REPRODUCTIVE HEALTH ...... 313
20 Contraception ...... 315 Gilda Piaggio 21 Gynaecology and Infertility ...... 337 Siladitya Bhattacharya and Jill Mollison
RESPIRATORY 357 22 Respiratory 359 Anders K¨all´en
Index ...... 401 Contributors
TONY ARTHUR School of Nursing, Faculty of Medicine and Health Sciences, University of Nottingham, Queens Medical Centre, Nottingham NG7 2UH, UK, Email: [email protected]
DONALD BERRY Department of Biostatistics, UTMD Anderson Cancer Centre, 1400 Holcombe Blvd, Box 447, Houston, TX 77030, USA, Email: [email protected]
S. BHATTACHARYA Department of Obstetrics and Gynaecology, Aberdeen Maternity Hospital, Cronhill Road, Aberdeen AB25 2ZD, UK, Email: [email protected]
ARMAN BUZDAR Department of Breast Medical Oncology, The University of Texas, M.D. Anderson Cancer Centre, Texas, USA
SIMON DAY Licensing Division, Medicines and Healthcare products Regulatory Agency, Room 13-205 Market Towers, 1 Nine Elms Lane, London SW8 5NQ, UK, Email: [email protected]
GRAHAM DUNN Biostatistics Group, School of Epidemiology and Health Sciences, Stopford Building, Oxford Road, Manchester M13 9PT, UK, Email: [email protected], [email protected]
FRED EDERER The EMMES Corporation, 401 North Washington Street, Suite 700, Rockville, MD 20850, USA, Email: [email protected] viii CONTRIBUTORS
BRIAN EVERITT Box 20, Biostatistics Department, Institute of Psychiatry, Denmark Hill, London SE5 8AF, UK, Email: [email protected]
LARRY FRIEDMAN National Heart, Lung, and Blood Institute, Building 31, Room 5A03, Bethesda, MD 20892 2482, USA, Email: [email protected]
STEPHEN GEORGE Department of Biostatistics and Bioinformatics, Duke University Medical Centre, Durham, NC 27710, USA, Email: [email protected]
RICH GOLDBERG Medical Oncology, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905, USA, Email: [email protected]
SYLVAN GREEN Arizona Cancer Centre, 1515 N Campbell Ave, PO Box 245024, Tucson, AZ 85724, USA, Email: [email protected]
CAROL JAGGER Trent Institute for HSR, Department of Epidemiology and Public Health, University of Leicester, 22–28 Princess Road West, Leicester LE1 6TP, UK, Email: [email protected]
ANDERS KALL¨ EN´ AstraZeneca R&D, Lund, Sweden, Email: [email protected]
JOHAN KARLBERG Clinical Trials Centre, Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, PR China, Email: [email protected]
KYUNGMANN KIM Department of Biostatistics & Medical Informatics, University of Wisconsin, 600 Highland Avenue, Box 4675, Madison, WI 53792, USA, Email: [email protected]
PHILIP LAVORI VACSPCC (151K), 795 Willow Road, Menlo Park, CA 94025 2539, USA, Email: [email protected]
P.C. LEUNG Department of Orthopaedics & Traumatology, The Chinese University of Hong Kong, Room 74026, 5th Floor, Clinical Sciences Building, Prince of Wales Hospital, Shatin, Hong Kong, Email: [email protected]
PAUL LIMBURG Gastroenterology and Hepatology, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905, USA, Email: [email protected] CONTRIBUTORS ix
P.Y. LIU SWOG Statistical Centre, MP 557, Fred Hutchinson Cancer Research Centre, 1100 Fairview Avenue North, PO Box 19024, Seattle, WA 98109 1024, USA, Email: [email protected]
EDWARD LO Dental Public Health, Faculty of Dentistry, The University of Hong Kong, 34 Hospital Road, Hong Kong, Email: [email protected]
DAVID MACHIN Division of Clinical Trials and Epidemiological Sciences, National Cancer Centre Singapore, 11 Hospital Drive, Singapore 169610, Singapore, Email: [email protected], [email protected]
COLAMN MCGRATH Dental Public Health, Faculty of Dentistry, The University of Hong Kong, 34 Hospital Road, Hong Kong, Email: [email protected]
COSETTA MINELLI Medical Statistics Group, Department of Epidemiology and Public Health, University of Leicester, 22–28 Princess Road West, Leicester LE1 6TP, UK, Email: [email protected]
JILL MOLLISON Department of Public Health, University of Aberdeen, Polwarth Building, Foresterhill, Aberdeen AB25 2ZD, UK, Email: [email protected]
SHARON B. MURPHY Children’s Cancer Research Institute, The University of Texas Health Science Centre at San Antonio, 7703 Floyd Curl Drive, MC 7784, San Antonio, TX 78229-3900, USA, Email: [email protected]
LUIGI NALDI Clinica Dermatologica, Ospedali Riuniti, L.go Barozzi 1, 24100 Bergamo, Italy, Email: [email protected]
GILDA PIAGGIO Special Programme of Research, Development and Research Training in Human Reproduction, Department of Reproductive Health and Research, World Health Organisation, 1211 Geneva 27, Switzerland, Email: [email protected]
DANIEL SARGENT Mayo Clinic, 200 First Street, SW, Kahler 1A, Rochester, MN 55905, USA, Email: [email protected]
CHARLES SCHIFFER Division of Oncology and Hematology/Oncology, Karmanos Cancer Institute, Wayne State University School of Medicine, Detroit, MI 48201, USA.
ELEANOR SCHRON National Heart, Lung, and Blood Institute, Bethesda, MD 20892 2482, USA, Email: [email protected] x CONTRIBUTORS
KATHERINE SHEAR Panic Anxiety and Traumatic Grief Program, Western Psychiatric Institute and Clinic, 3811 O’Hara Street, Pittsburgh, PA 15213, USA, Email: [email protected], [email protected]
JON SHUSTER PO Box 100212, College of Medicine, University of Florida, Gainesville, FL 32610 0212, USA, Email: [email protected]fl.edu
TERRY SMITH UTMD Anderson Cancer Centre, 1400 Holcombe Blvd, Box 447, Houston, TX 77030, USA, Email: [email protected]
VERNON K. SONDAK University of Michigan Medical Centre, Division of Surgical Oncology, 1500 E Medical Centre Drive, 3306 Cancer/Geriatrics Ctr Box 0932, Ann Arbor, MI 48109 0932, USA, Email: [email protected]
NICHOLAS TARRIER University of Manchester, Academic Division of Clinical Psychology Education & Research Building, Wythenshawe Hospital, Southmoor Road, Manchester M23 9LT, UK, Email: [email protected]
LEON THAL Department of Neurosciences, University of California, San Diego, 9500 Gilman Drive, San Diego, CA 92037, USA, Email: [email protected]
RONALD THOMAS Department of Family and Preventative Medicine, and Neurosciences, UCSD 9500 Gilman Drive, La Jolla, CA 92039 0645, USA, Email: [email protected]
MAY C.M. WONG Dental Public Health, Faculty of Dentistry, The University of Hong Kong, 34 Hospital Road, Hong Kong, Email: [email protected]
TIL WYKES Department of Psychiatry, De Crespigny Park, London SE5 8AF, UK, Email: [email protected] Preface
This Textbook of Clinical Trials is not a textbook needed from the concept stage, through design, of clinical trials in the traditional sense. Rather, it conduct, monitoring and reporting. catalogues in part both the impact of clinical tri- Some of the developments impacting on clini- als – particularly the randomised controlled trial cal trials have been truly statistical in nature, for –on the practice of medicine and allied fields example Cox’s proportional hazards model, while and on the developments and practice of medical others such as the intention-to-treat (ITT) princi- statistics. The latter has evolved in many ways ple are – in some sense – based more on expe- through the direct needs of clinical trials and the rience. Other important statistical developments consequent interaction of statistical and clinical have not depended on technical advancement, disciplines. The impact of the results from clin- but rather on conceptual advancement, such as ical trials, particularly the randomised controlled the now standard practice of reporting confidence trial, on the practice of clinical medicine and intervals rather then relying solely on p-values at the interpretation stage. Of major importance other areas of health care has been profound. In over this same time period has been the expansion particular, they have provided the essential under- in data processing capabilities and the range of pinning to evidence-based practice in many disci- analytical possibilities only made possible by the plines and are one of the key components for reg- tremendous development in computing power. ulatory approval of new therapeutic approaches However, despite many advances, the majority throughout the world. of randomised controlled trials remain simple in Probably the single most important contribu- design – most often a comparison between two tion to the science of comparative clinical trials randomised groups. was the recognition, more than 50 years ago, that On the medical side there have been many patients should be allocated to the options under changes including new diseases that raise consideration at random. This was the founda- new issues. Thus, as we write, SARS has tion for the science of clinical trial research and emerged: the final extent of the epidemic placed the medical statistician at the centre of the is unknown, diagnosis is problematical and process. Although the medical statistician may be no specific treatment is available. In more at the centre, he or she is by no means alone. established diseases there have been major Indeed the very nature of clinical trial research is advances in the types of treatment available, be multidisciplinary so that a ‘team’ effort is always they in surgical technique, cancer chemotherapy xii PREFACE or psychotropic drugs. Advances in medical the outcome cosmetic then a more conservative and associated technologies are not confined approach to treatment options would be justified. to curative treatments but extend, for example, In all this activity the choice of clinical trial to diagnostic methods useful in screening for design and its ultimate conduct are governed disease, vaccines for disease prevention, drugs by essential ethical constraints, the willingness and devices for female and male contraception, of subjects to consent to the trial in question and pain relief and psychological support and their right to withdraw from the trial should strategies in palliative care. they wish. Clinical trials imply some intervention affect- Thus the Textbook of Clinical Trials addresses ing the subjects who are ultimately recruited into some of these and many other issues as they them. These subjects will range from the very impact on patients with cancer, cardiovascular healthy, perhaps women of a relatively young age disease, dermatological, dental, mental and oph- recruited to a contraceptive development trial, thalmic health, gynaecology and respiratory dis- to those (perhaps elderly) patients in terminal eases. In addition, chapters deal with issues relat- decline from a long-standing illness. Each group ing to complementary medicine, contraception studied in a clinical trial, from unborn child to and special issues in children and special issues aged adult, brings its own constraint on the ulti- in older patients. A brief history of clinical tri- mate design of the trial in mind. So too does the als and a summary of some pertinent statistical relative efficacy of the current standard. If the issues are included. outcome is death and the prognosis poor, then bolder steps may be taken in the choice of treat- ments to test. If the disease is self-limiting or David Machin, Simon Day and Sylvan Green INTRODUCTION
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 1 Brief History of Clinical Trials
S. DAY1 AND F. EDERER2 1Licensing Division, Medicines and Healthcare products Regulatory Agency, London, UK 2The EMMES Corporation, Rockville, MD 20850, USA
The modern-day birth of clinical trials is usually Daniel lived around the period 800BC and although considered to be the publication in 1948 by it may not be possible to confirm the accuracy the UK Medical Research Council1 of a trial of the account, what is clear is that when this for the treatment of pulmonary tuberculosis with passage was written–around 150BC–the ideas streptomycin. However, earlier but less well certainly existed. documented examples do exist. The comparative The passage from Daniel describes not just a concept of assessing therapeutic efficacy has control group, but a concurrent control group. been known from ancient times. Slotki2 cites a This fundamental element of clinical research did description of a nutritional experiment involving not begin to be widely practised until the latter a control group in the Book of Daniel from the half of the twentieth century. Old Testament: Much later than the book of Daniel, but still very early, is an example from the fourteenth Then Daniel said to the guard whom the master century: it is a letter from Petrarch to Boccaceto of the eunuchs had put in charge of Hananiah, cited by Witkosky:3 Mishael, Azariah and himself, ‘Submit to us this test for ten days. Give us only vegetables to eat I solemnly affirm and believe, if a hundred and water to drink; then compare our looks with or a thousand of men of the same age, same those of the young men who have lived on the temperament and habits, together with the same food assigned by the king, and be guided in your surroundings, were attacked at the same time by the treatment of us by what you see.’ The guard listened same disease, that if one followed the prescriptions to what they said and tested them for ten days. At of the doctors of the variety of those practicing at the end of ten days they looked healthier and were the present day, and that the other half took no better nourished than all the young men who had medicine but relied on Nature’s instincts, I have no lived on the food assigned them by the king. So doubt as to which half would escape. the guard took away the assignment of food and the wine they were to drink, and gave them only The Renaissance period (fourteenth to sixteenth the vegetables. centuries) provides other examples including an
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 4 TEXTBOOK OF CLINICAL TRIALS unplanned experiment in the treatment of battle- operated, by way of gentle physic. Two others had field wounds. Packard4 describes how the surgeon each two oranges and one lemon given them every Ambroise Pare´ was using the standard treatment day. These they eat with greediness, at different of pouring boiled oil over soldiers’ wounds dur- times, upon an empty stomach. They continued but ing a battle to capture the castle of Villaine in 1537. six days under this course, having consumed the quantity that could be spared. The two remaining When he ran out of oil, he resorted to a mixture of patients, took the bigness of a nutmeg three times egg yolks, oil of roses and turpentine. The supe- a day of an electuary recommended by a hospital- riority of the new ‘treatment’ became evident the surgeon, made of garlic, mustard-feed, rad. raphan, next day: balsam of Peru, and gum myrrh; using for common drink barley-water well acidulated with tamarinds; I raised myself very early to visit them, when by a decoction of which, with the addition of beyond my hope I found those to whom I applied cremor tartar, they were greatly purged three or the digestive medicament feeling but little pain, four times during the course. their wounds neither swollen nor inflamed, and The consequence was, that the most sudden and having slept through the night. The others to whom visible good effects were perceived from the use I had applied the boiling oil were feverish with of the oranges and lemons; one of those who had much pain and swelling about their wounds. Then taken them, being at the end of six days fit for I determined never again to burn thus so cruelly by duty. The spots were not indeed at that time quite arquebusses. off his body, or his gums sound; without any other Perhaps the most famous historical example of medicine, than a gargle of elixir vitriol, he became a planned controlled, comparative, clinical trial quite healthy before we came into Plymouth, which was on the 16th June. The other was the best is from the eighteenth century: that where Lind5 recovered of any in his condition; and being now found oranges and lemons to be the most effective deemed pretty well, was appointed nurse, to the rest of six dietary treatments for scurvy on board ships. of the sick. He wrote: Pierre-Charles-Alexandre Louis, a nineteenth- On the 20th of May, 1747, I took twelve patients century clinician and pathologist, introduced the in the scurvy, on board the Salisbury at sea. Their 6 cases were as similar as I could have them. They all numerical aspect to comparing treatments. His in general had putrid gums, the spots and lassitude, idea was to compare the results of treatments on with weakness of their knees. They lay together groups of patients with similar degrees of disease, in one place, being a proper apartment for the i.e. to compare ‘like with like’: sick in the fore-hold; and had one diet common to all, viz. water-gruel sweetened with sugar in I come now to therapeutics, and suppose that you the morning; fresh mutton-broth often times for have some doubt as to the efficacy of a particular dinner; at other times puddings, boiled biscuit with remedy: How are you to proceed?...You would sugar etc. And for supper, barley and raisins, rice take as many cases as possible, of as similar a and currants, sago and wine, or the like. Two of description as you could find, and would count these were ordered each a quart of cyder a day. how many recovered under one mode of treatment, Two others took twenty-five gutts of elixir vitriol and how many under another; in how short a time three times a day, upon an empty stomach; using a they did so, and if the cases were in all respects gargle strongly acidulated with it for their mouths. alike, except in the treatment, you would have some Two others took two spoonfuls of vinegar three confidence in your conclusions; and if you were times a day, upon an empty stomach; having their fortunate enough to have a sufficient number of gruels and their other food well acidulated with it, facts from which to deduce any general law, it as also the gargle for their mouths. Two of the would lead to your employment in practice of the worst patients, with the tendons in the ham rigid method which you had seen oftenest successful. (a symptom none of the rest had) were put under a course of sea-water. Of this they drank half a ‘Like with like’ was an important step forward pint every day, and sometimes more or less as it from Lind’s treatment of scurvy. Note, although BRIEF HISTORY OF CLINICAL TRIALS 5 early in Lind’s passage he says that: ‘Their cases At the beginning of each year... students were were as similar as I could have them’, later assigned at random... to a control group or an he acknowledges that the two worst cases both experimental group. The students in the control received the same treatment: ‘Two of the worst groups...received placebos...All students thought patients, with the tendons in the ham rigid (a they were receiving vaccines... Even the physi- cians who saw the students... had no information symptom none of the rest had) were put under as to which group they represented. a course of sea-water’. It remained for Bradford Hill, more than a century later, to use a formal method for creating groups of cases that were ‘in However Gail14 points out that although this all respects alike, except in the treatment’. appears to be a randomised clinical trial, a further unpublished report by Diehl clarifies that this is another instance of systematic assignment: RANDOMISATION At the beginning of the study, students who The use of randomisation was a contribution by volunteered to take these treatments were assigned the statistician R.A. Fisher in agriculture (see, for alternately and without selection to control groups example, Fisher,7 Fisher and McKenzie8). Fisher and experimental groups. randomised plots of crops to receive different treatments and in clinical trials there were early Bradford Hill, in the study of streptomycin in schemes to use ‘group randomisation’: patients pulmonary tuberculosis,1 used random sampling were divided into two groups and then the numbers in assigning treatments to subjects, so treatment for each group was randomly selected. 9 that the subject was the unit of randomisation. This The Belgian medicinal chemist van Helmont study is now generally acknowledged to be the described an early example of this: ‘first properly randomised clinical trial’. Later Bradford Hill and the British Medical Let us take out of the hospitals, out of the Camps, Research Council continued with further ran- or from elsewhere, 200, or 500 poor People that domised trials: chemotherapy of pulmonary tuber- have Fevers, Pleurisies, &c, Let us divide them into 15 halfes, let us cast lots, that one half of them may culosis in young adults, antihistaminic drugs fall to my share, and the others to yours,...we shall in the prevention and treatment of the common 16 see how many funerals both of us shall have: But cold, cortisone and aspirin in the treatment let the reward of the contention or wager, be 300 of early cases of rheumatoid arthritis17,18 and florens, deposited on both sides. long-term anticoagulant therapy in cerebrovascu- lar disease.19 Amberson and McMahon10 used group randomi- In America, the National Institutes of Health sation in a trial of sanocrysin for the treatment started its first randomised trial in 1951. It was of pulmonary tuberculosis. Systematic assign- a National Heart Institute study of adrenocorti- ment was used by Fibiger,11 who alternately cotropic hormone (ACTH), cortisone and aspirin assigned diphtheria patients to serum treatment in the treatment of rheumatic heart disease.20 This or an untreated control group. Alternate assign- was followed in 1954 by a randomised trial of ment is frowned upon today because knowledge retrolental fibroplasia (now known as retinopathy of the future treatment allocations may selectively of prematurity), sponsored by the National Insti- bias the admission of patients into the treatment tute of Neurological Diseases and Blindness.21 group.12 Diehl et al.13 reported a common cold During the four decades following the pioneer- vaccine study with University of Minnesota stu- ing trials of the 1940s and 1950s, there was a dents as subjects where blinding and random large growth in the number of randomised trials assignment of patients to treatments appears to not only in Britain and the US, but also in Canada have been used: and mainland Europe. 6 TEXTBOOK OF CLINICAL TRIALS
BLINDING ETHICS
The common cold vaccine study published by Experimentation in medicine is as old as medicine Diehl et al.,13 cited earlier, in which University itself. Some experiments on humans have been of Minnesota students were alternately assigned conducted without concern for the welfare of the subjects, who have often been prisoners or to vaccine or placebo, was a masked (or blinded) 23 clinical trial: disadvantaged people. Katz provides examples of nineteenth-century studies in Russia and Ire- land of the consequences of infecting people with All students thought they were receiving vaccines... syphilis and gonorrhoea. McNeill24 describes Even the physicians who saw the students...had no information as to which group they represented. how during the same period in the US, physicians put slaves into pit ovens to study heat stroke, and poured scalding water over them as an experi- Blinding was used in the early Medical Research mental cure for typhoid fever. He even describes Council trials in which Bradford Hill was involved. how one slave had two fingers amputated in a Thus, in the first of those trials, the study of strep- 1 ‘controlled trial’, one finger with anaesthesia and tomycin in tuberculosis, the X-ray films were one without! viewed by two radiologists and a clinician, each Unethical experiments on human beings have reading the films independently and not knowing continued into the twentieth century and have if the films were of C (control, bed-rest alone) or been described by, for example, Beecher,25 S (streptomycin and bed-rest) cases. 26 24 22 Freedman and McNeil. In 1932 the US Pub- Bradford Hill (Reproduced with permission) lic Health Service began a study in Tuskegee, noted in respect of using such blinding and Alabama, of the natural progression of untreated randomisation: syphilis in 400 black men. The study continued until 1972, when a newspaper reported that the If [the clinical assessment of the patient’s progress subjects were uninformed or misinformed about and of the severity of the illness] is to be used the purpose of the study.26 Shirer,27 amongst oth- effectively, without fear and without reproach, the ers, describes how during the Nazi regime from judgements must be made without any possibility 1933 to 1945, German doctors conducted exper- of bias, without any overcompensation for any possible bias, and without any possible accusation iments, mainly on Jews, but also on Gypsies, of bias. mentally disabled persons, Russian prisoners of war and Polish concentration camp inmates. The Nazi doctors were later tried for their atrocities in Not simply overcoming bias, but overcoming 1946–1947 at Nuremberg and this led to the writ- any possible accusation of bias is an important ing, by three of the trial judges, of the Nuremberg justification for blinding and randomisation. Code, the first international effort to lay down In the second MRC trial, the antihistamine ethical principles of clinical research.28 Principle common cold study,16 placebos, indistinguishable 1 of the Nuremberg Code states: from the drug under test, were used. Here, Bradford Hill noted: The voluntary consent of the human subject is absolutely essential. This means that the person ... in [this] trial... feelings may well run high... involved should have legal capacity to give consent; either of the recipient of the drug or the clinical should be so situated as to be able to exercise observer, or indeed of both. If either were allowed free power of choice, without the intervention of to know the treatment that had been given, I believe any element of force, fraud, deceit, duress, over- that few of us would without qualms accept that reaching, or other ulterior form of constraint or the drug was of value–if such a result came out of coercion; and should have sufficient knowledge and the trial. comprehension of the elements of the subject matter BRIEF HISTORY OF CLINICAL TRIALS 7
involved as to enable him to make an understanding committee. In 1968 the first such committee was and enlightened decision. established, serving the Coronary Drug Project, a large multicentre trial sponsored in the United Other principles of the Code are that experiments States by the National Heart Institute of the should yield results for the good of society, National Institutes of Health.35,36 In 1967, after a that unnecessary suffering and injury should be presentation of interim outcome data by the study avoided, and that the subject should be free to co-ordinators to all participating investigators of withdraw from the experiment at any time and the Coronary Drug Project, Thomas Chalmers for any reason. addressed a letter to the policy board chairman Other early advocates of informed consent expressing concern: were Charles Francis Withington and William Osler. Withington29 realised, the ‘possible con- that knowledge by the investigators of early nonsta- flict between the interests of medical science and tistically significant trends in mortality, morbidity, those of the individual patient’, and concluded or incidence of side effects might result in some in favour of ‘the latter’s indefensible rights’. investigators–desirous of treating their patients in Osler30 insisted on informed consent in medi- the best possible manner, i.e., with the drug that is ahead–pulling out of the study or unblinding cal experiments. Despite this early advocacy, and the treatment groups prematurely. (Canner35,repro- the 1946–1947 Nuremberg Code, the application duced with permission from Elsevier) of informed consent to medical experiments did 31 not take hold until the 1960s. Hill, based on Following this, a data and safety monitoring com- his experience in a number of early randomised mittee was established for the Coronary Drug clinical trials sponsored by the Medical Research Project. It consisted of scientists who were not con- Council, believed that it was not feasible to draw tributing data to the study, and thereafter the prac- up a detailed code of ethics for clinical trials tice of sharing accumulating outcome data with the that would cover the variety of ethical issues that study’s investigators was discontinued. The data came up in these studies, and that the patient’s safety and monitoring committee assumed respon- consent was not warranted in all clinical tri- sibility for deciding when the accumulating data als. Gradually the medical community came to warranted changing the study protocol or termi- recognise the need to protect the reputation and nating the study. integrity of medical research and in 1955 a human The first formal recognition of the need for experimentation code was adopted by the Pub- 32 interim analyses, and the recognition that such lic Health Council in the Netherlands. Later, analyses affect the probability of the type I in 1964, the World Medical Assembly issued the 33 error, came with the publication in the 1950s Declaration of Helsinki essentially adopting the of papers on sequential clinical trials by Bross37 ethical principles of the Nuremberg Code, with and Armitage.38 The principal advantage of a consent being ‘a central requirement of ethical 34 sequential trial over a fixed sample size trial, is research’. The Declaration of Helsinki has been that when the length of time needed to reach an updated and amended several times: Tokyo, 1975; endpoint is short, e.g. weeks or months, the sample Venice, 1983; Hong Kong, 1989; Cape Town, size required to detect a substantial benefit from 1996; and Edinburgh, 2000. one of the treatments is less. In the 1970s and 1980s solutions to interim anal- ysis problems came about in the form of group se- DATA MONITORING quential methods and stochastic curtailment.39–41 In the group sequential trial, the frequency of In the modern randomised clinical trial, the accu- interim analyses is usually limited to a small mulating data are usually monitored for safety number, say between three and six. The Pocock and efficacy by an independent data monitoring boundaries42 use constant nominal significance 8 TEXTBOOK OF CLINICAL TRIALS levels; the Haybittle–Peto boundary43,44 uses 4. Packard FR. The Life and Times of Ambroise Par´e, stringent significance levels for all except the final 2nd edn. New York: Paul B. Hoeber (1921) 45 27,163. test; in the O’Brien–Fleming method, stringency 5. Lind J. A Treatise of the Scurvy. Edinburgh: Sands gradually decreases; in the method by Lan and Murray Cochran (1753) 191–3. DeMets,46 the total type I error probability is grad- 6. Louis PCA. The applicability of statistics to the ually spent in a manner that does not require the practice of medicine. London Med Gazette (1837) timing of analyses to be prespecified. More details 20: 488–91. 7. Fisher RA. The arrangement of field experiments. of these newer methods in the development of clin- JMinAgric(1926) 33: 503–13. ical trials are given in the next chapter. 8. Fisher RA, McKenzie WA. Studies in crop varia- Recent years have seen a huge increase in the tion: II. The manurial response of different potato numbers of trials carried out and published, and varieties. J Agric Sci (1923) 13: 315. in the advancement of methodological aspects 9. van Helmont JB. Oriatrike or Physik Refined (translated by J. Chandler). London: Lodowick relating to trials. Whilst many see the birth Loyd (1662). of clinical trials (certainly in their modern-day 10. Amberson B, McMahon PM. A clinical trial of guise) as being the Medical Research Council sanocrysin in pulmonary tuberculosis. Am Rev streptomycin trial,1 there remains some contro- Tuberc (1931) 24: 401. versy (see, for example, D’Arcy Hart,47,48 Gill,49 11. Fibiger I. Om Serum Behandlung of Difteri. 50 Hospitalstidende (1898) 6: 309–25, 337–50. and Clarke ). However, it is interesting to note 12. Altman DG, Bland JM. Treatment allocation in that one of the most substantial reviews of histor- controlled trials: why randomise? Br Med J (1999) ical aspects of trials is based on work for a 1951 318: 1209. M.D. thesis.51 Bull cites 135 historical examples 13. Diehl HS, Baker AB, Cowan DW. Cold vaccines: and other supporting references–but no mention an evaluation based on a controlled study. JAm Med Assoc (1938) 111: 1168–73. of Bradford Hill and the Medical Research Coun- 14. Gail MH. Statistics in action. J Am Stat Assoc cil. The modern-day story of clinical trials per- (1996) 91: 1–13. haps begins where Bull ended. 15. Medical Research Council. Chemotherapy of pul- monary tuberculosis in young adults. Br Med J (1952) i: 1162–8. ACKNOWLEDGEMENTS 16. Medical Research Council. Clinical trials of anti- histaminic drugs in the prevention and treatment This chapter is based heavily on work by of the common cold. Br Med J (1950) ii: 425–31. F. Ederer in: Armitage P, Colton T, eds, Ency- 17. Medical Research Council. A comparison of clopedia of Biostatistics. Chichester: John Wiley cortisone and aspirin in the treatment of early & Sons (1998). cases of rheumatoid arthritis – I. Br Med J (1954) i: 1223–7. 18. Medical Research Council. A comparison of REFERENCES cortisone and aspirin in the treatment of early cases of rheumatoid arthritis – II. Br Med J (1955) ii: 695–700. 1. Medical Research Council. Streptomycin treat- 19. Hill AB, Marshall J, Shaw DA. A controlled clin- ment of pulmonary tuberculosis. Br Med J (1948) ical trial of long-term anticoagulant therapy in 2: 769–82. cerebrovascular disease. QJMed(1960) 29(NS): 2. Slotki JJ. Daniel, Ezra, Nehemiah, Hebrew Text 597–608. and English Translation with Introductions and 20. Rheumatic Fever Working Party. The evolution Commentary Soncino Press: London (1951) 1–6. of rheumatic heart disease in children: five-year 3. Witkosky SJ. The Evil That Has Been Said of report of a co-operative clinical trial of ACTH, Doctors: Extracts From Early Writers (translated cortisone, and aspirin. Circulation (1960) 22: with annotations by T.C. Minor). The Cincinnati 505–15. Lancet-Clinic, Vol. 41, New Series Vol. 22 (1889) 21. Kinsey VE. Retrolental fibroplasia. AMA Arch 447–8. Ophthalmol (1956) 56: 481–543. BRIEF HISTORY OF CLINICAL TRIALS 9
22. Hill AB. Statistical Methods in Clinical and 36. Friedman L. The NHLBI model: a 25-year history. Preventative Medicine. Edinburgh: Livingstone Stat Med (1993) 12: 425–31. (1962). 37. Bross I. Sequential medical plans. Biometrics 23. Katz J. The Nuremberg Code and the Nuremberg (1952) 8: 188–295. Trial. A reappraisal. J Am Med Assoc (1996) 276: 38. Armitage P. Sequential tests in prophylactic and 1662–6. therapeutic trials. QJMed(1954) 23: 255–74. 24. McNeill PM. The Ethics and Politics of Human 39. Armitage P. Interim analysis in clinical trials. Stat Experimentation. Cambridge: Press Syndicate of Med (1991) 10: 925–37. the University of Cambridge (1993). 40. Friedman LM, Furberg CD, DeMets DL. Funda- 25. Beecher HK. Ethics and clinical research. New mentals of Clinical Trials, 2nd edn. Boston: Wright Engl J Med (1966) 274: 1354–60. (1985). 26. Freedman B. Research, unethical. In: Reich WT, 41. Pocock SJ. Clinical Trials: A Practical Approach. ed., Encyclopedia of Bioethics. New York: Free Chichester: John Wiley & Sons (1983). Press (1995) 2258–61. 42. Pocock SJ. Group sequential methods in the 27. Shirer WL. The Rise and Fall of the Third Reich. design and analysis of clinical trials. Biometrika New York: Simon and Schuster (1960). (1977) 64: 191–9. 28. US Government Printing Office. Trials of War 43. Haybittle JL. Repeated assessment of results of Criminals before the Nuremberg Military Tri- cancer treatment. Br J Radiol (1971) 44: 793–7. 44. Peto R, Pike MC, Armitage P, Breslow NE, Cox bunals under Control Council Law No. 10,Vol.2. DR, Howard SV, Mantel N, McPherson K, Peto J, Washington: US Government Printing Office Smith P. Design and analysis of randomized clin- (1949) 181–2. ical trials requiring prolonged observation of each 29. Withington CF. Time Relation of Hospitals to patient. 1. Introduction and design. Br J Cancer Medical Education. Boston: Cupples Uphman (1976) 34: 585–612. (1886). 45. O’Brien PC, Fleming TR. A multiple testing pro- 30. Osler W. The evolution of the idea of experiment. cedure for clinical trials. Biometrics (1979) 35: Trans Congr Am Phys Surg (1907) 7: 1–8. 549–56. 31. Hill AB. Medical ethics and controlled trials. Br 46. Lan KKG, DeMets DL. Discrete sequential bound- Med J (1963) 1: 1043–9. aries for clinical trials. Biometrika (1983) 70: 32. Netherlands Minister of Social Affairs and Health. 659–63. 4WorldMedJ(1957) 299–300. 47. D’Arcy Hart P. History of randomised controlled 33. World Medical Assembly. Declaration of Helsinki: trials (letter to the Editor). The Lancet (1972) i: recommendations guiding physicians in biomed- 965. ical research involving human subjects. World 48. D’Arcy Hart P. Early controlled clinical trials Medical Assembly, Helsinki (1964). (letter to the Editor). Br Med J (1996) 312: 378–9. 34. Faden RR, Beauchamp T, King NMP. A History 49. Gill DBEC. Early controlled trials (letter to the of Informed Consent. New York: Oxford Univer- Editor). Br Med J (1996) 312: 1298. sity Press (1986). 50. Clarke M. Early controlled trials (letter to the 35. Canner P. Monitoring of the data for adverse Editor). Br Med J (1996) 312: 1298. or beneficial treatment effects. Contr Clin Trials 51. Bull JP. The historical development of clinical (1983) 4: 467–83. therapeutic trials. J Chron Dis (1959) 10: 218–48. 2 General Issues
DAVID MACHIN National Cancer Centre, Singapore and United Kingdom Children’s Cancer Study Group, University of Leicester, UK
INTRODUCTION associated trial protocol should carefully describe (and in some detail) the elements essential for Just as in any other field of scientific and its conduct. Thus the protocol will describe medical research, the choice of an appropriate the rationale for the trial, the eligible group design for a clinical trial is a vital element. of patients or subjects, the therapeutic options In many circumstances, and for many of the and their modification should the need arise. It trials described in this text, these designs may will also describe the method of patient allo- not be overly complicated. For example, a cation to these options, the specific clinical large majority will compare two therapeutic assessments to be made and their frequency, or other options in a parallel group trial. In and the major endpoints to be used for eval- which case the analytical methods used for uation. It will also include a justification of description and analysis too may not be overly the sample size chosen, an indication of the complicated. The vast majority of these are analytical techniques to be used for summary described in basic medical statistics textbooks and comparisons, and the proforma for data and implemented in standard software packages. collection. Nevertheless there are circumstances in which Of major concern in all aspects of clinical trial more complex designs, such as sequential trials, development and conduct is the ethical necessity are utilised and for which specialist methods are which is written into the Declaration of Helsinki required. There are also, often rather complex, of 19641 to ensure the well-being of the patients statistical problems associated with monitoring or subjects under study. This in itself requires the progress of clinical trials, their interim that clinical trials are well planned and conducted analysis, stopping rules for early closure and the with due concern for the patient’s welfare and possibility of extending recruitment beyond that safety. It also requires that the trial is addressing initially envisaged. an important question, the answer to which will Although the clinical trial itself may not be bring eventual benefit to the patients themselves of complex design in the statistical sense, the or at least to future patients.
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 12 TEXTBOOK OF CLINICAL TRIALS
EVIDENCE-BASED MEDICINE Table 2.1. Objectives of the trials of different phases in the development of drug (after Day3) In many circumstances trials have been con- Phase Objective ducted that are unrealistically small, some unnec- essarily replicated while others have not been I The earliest types of studies that are published as their results have not been consid- carried out in humans. They are typically done using small numbers of ered of interest. It has now been recognised that healthy subjects and are to investigate to obtain the best current evidence with respect pharmacodynamics, pharmacokinetics to a particular therapy, all pertinent clinical trial and toxicity. information needs to be obtained, and if circum- II Carried out in patients, usually to find stances permit, the overview is completed by a the best dose of drug and to investigate meta-analysis of the trial results. This recogni- safety. tion has led to the Cochrane Collaboration and a III Generally major trials aimed at worldwide network of overview groups address- conclusively demonstrating efficacy. ing numerous therapeutic questions.2 In certain They are sometimes called situations this has brought definitive statements confirmatory trials and, in the context with respect to a particular therapy. For others of pharmaceuticals, typically are the studies on which registration of a new it has led to the launch of large-scale confirma- product will be based. tory trials. Although it is not appropriate to review the IV Studies carried out after registration of a product. They are often for marketing methodology here, it is clear that the ‘overview’ purposes as well as to gain broader process has led to many changes to the way experience with using the new in which clinical trial programmes have devel- product. oped. They have provided the basic information required in planning new trials, impacted on an appropriate trial size, publication policy and very be a Phase III study, as will a trial comparing importantly raised reporting standards. They are two surgical procedures for closing a cleft palate. impacting directly on decisions that affect patient In both these examples, any one of the Phase care and questioning conventional wisdom in I, II or IV trials would not necessarily be many areas. conducted. The objectives of each phase in a typical development programme for a drug are summarised in Table 2.1. TYPES OF CLINICAL TRIALS Without detracting from the importance of Phase I, II and IV clinical trials, the main focus In broad terms, the types of trials conducted in of this text is on Phase III comparative trials. human subjects may be divided into four phases. In this context, reference will often be made to These phases represent the stages in, for example, the ‘gold standard’ randomised controlled trial the development of a new drug which requires (RCT). This does not imply that this is the only early dose finding and toxicity data in man, type of trial worthy of conduct, but rather that it indications of potential activity, comparisons provides a benchmark against which other trial with a standard to determine efficacy and then designs are measured. (in certain circumstances) post-marketing trials. The nomenclature of Phase I, II, III and IV has been developed for drug development purposes PHASE I AND II TRIALS and there may or may not be exact parallels in other applications. For example, a trial to assess The traditional outline of a series of clinical the value of a health educational programme will trials moving sequentially through Phases I to GENERAL ISSUES 13
IV is useful to consider in an idealistic set- Although conducted in patients, Phase II trials ting, although in practice the sequential manner are typically still highly controlled and use is not always followed (for reasons that will highly defined (often narrow) patient groups so become clear). that extraneous variation is kept to a minimum. Pocock4 (pp. 2–3) also gives a convenient These are very much exploratory trials aimed summary of the four phases while Temple5 gives at discovering if a compound can show useful a discussion of them with emphasis from a clinical results. Although it is not common, regulatory perspective. Whether the sequential some of these trials may have a randomised nature of the four phases is adhered to or not, the comparison group. objectives of each phase are usually quite clearly defined. As we have indicated in Table 2.1, Phase I stud- NON-RANDOMISED EFFICACY STUDIES ies aim to investigate the metabolism of a drug and its pharmacodynamics and pharmacokinet- HISTORICAL CONTROLS ics. Typical pharmacokinetic data would allow, for example, investigation of peak drug concen- In certain circumstances, when a new treatment trations in the blood, the half-life and the time has been proposed, investigators have recruited to complete clearance. Such studies will assist in patients in single-arm studies. The results from defining what doses should be used and the dos- these patients are then compared with information ing frequency (once daily, twice daily, hourly) on similar patients having (usually in the past) for future studies. These Phase I studies (certainly received a relevant standard therapy for the the very first ones) are almost always undertaken disease in question. However, such comparisons in healthy volunteers and would naturally be the may well be biased in many possible ways, such very first studies undertaken in humans. However, that it may not be reasonable to ascribe the later in a drug development programme it may be difference (if any) observed to the treatments necessary to study its effects in patients with spe- themselves. Nevertheless it has been argued that cific diseases, in those taking other medications using regression models to account for possible or patients from special groups (infants, elderly, confounding variables may correct such biases,6 ethnic groups, pregnant women). but this is at best a very uncertain procedure Most of the objectives of a Phase I study can and is not often advocated. Similar problems often be met with relatively few subjects – many arise if all patients are recruited prospectively studies have fewer than 20 subjects. In essence, but allocation to treatment is not made at they are much more like closely controlled random. Of course, there will be situations laboratory experiments than population-based in which randomisation is not feasible and clinical trials. there is no alternative to the use of historical Broadly speaking, Phase II trials aim to set controls or non-randomised prospective studies. the scene for subsequent confirmatory Phase III One clear example of this is the early evaluation trials. Typically, although exceptions may occur, of the Stanford Heart Transplant Program in these will be the first ‘trials’ in patients. They which patients could not be randomised to are also the first to investigate the existence receive or not a donor heart. Many careful of possible clinical benefits to those patients. analyses and reviews of this unique data set However, although efficacy is important in Phase have been undertaken and these have established II it may often be in the form of surrogate, for the value of the programme, but progress would example, tumour response rather than survival have been quicker (and less controversial) had time in a patient with cancer. Along with efficacy, randomisation been possible. these studies will also be the first to give some In the era of evidence-based medicine, infor- detailed data on side effects. mation from non-randomised but comparative 14 TEXTBOOK OF CLINICAL TRIALS studies is categorised as providing weaker evi- equipoise, as described by Freedman9 and Weijer dence than randomised trials (see Altman,7 et al.10 to justify random allocation to treatment p. 3279). after due consent is given. Enkin,11 in a debate with Weijer et al.,10 provides a counter view. There are at least two aspects of the eligibility PHASE III CONTROLLED TRIALS requirements that are important. The first is that the patient indeed has the condition (here severe EQUIPOISE AND UNCERTAINTY burns) and satisfies all the other requirements. As indicated, the randomised controlled trial is There must be no specific reasons why the the standard against which other trial designs may patient should not be included. For example, in be compared. One such trial, and there are many some circumstances pregnant or lactating women other examples described in subsequent chapters, (otherwise eligible) may be excluded for fear compared conventional treatment, C, with a of impacting adversely either on the foetus or complementary medicine alternative in patients the newborn child. The second is that all (here with severe burns.8 The complementary medicine two) therapeutic options are equally appropriate was termed Moist Exposed Burns Ointment for this particular patient. Only if both these (MEBO). One essential difference between the aspects are satisfied should the patient be invited two treatments was that C covered the wounds to consent to participate in the trial. There will be (dressed) whilst MEBO left them exposed (not circumstances in which a patient may be eligible dressed). See Figure 2.1. for the trial but the attending physician feels (for In this trial patients with severe burns were whatever reason) that one of the trial options is emergency admissions requiring immediate treat- ‘best’ for the patient. In which case the patient ment, so that once eligibility was confirmed, should receive that option, no consent for the trial consent obtained, randomisation immediately fol- is then required and the randomisation would not lowed and treatment was then commenced. Such take place. In such circumstances, the clinician a trial is termed a two-treatment parallel group should not randomise the patient in the hope that design. This is the most common design for com- the patient will receive the ‘best’ option. Then, parative clinical trials. In these trials subjects are if he or she did not, withdraw the patient from independently allocated to receive one of several the trial. treatment options. No subject receives more than The consent procedure itself will vary from one of these treatments. trial to trial and will, at least to some extent, In addition there is genuine uncertainty as to depend on local ethical regulations in the country which of the options is best for the patient. It in which the trial is being conducted. The ideal is is this uncertainty which provides the necessary fully informed and written consent by the patient
A s Patients Eligible Random MEBO s presenting and allocation e s with partial consenting to s degree subjects treatment Conventional m burns e Dressing n t Source: Reproduced from Ang et al.,8
Figure 2.1. Randomised controlled trial to compare conventional treatment and Most Exposed Burns Ointment (MEBO) for the treatment of patients with partial degree burns GENERAL ISSUES 15 him or herself. However, departures from this STANDARD OR CONTROL THERAPY may be appropriate. For example, such departures may concern patients with severe burns who may In the early stages of the development of a new be unconscious at admission, very young children therapy it is important to compare this with the for whom a proxy must be used to obtain the current standard for the disease in question. In certain circumstances, the ‘new’ therapy may consent for them, or patients with hand burns that be compared against a ‘no treatment’ control. are so severe that they affect their ability to write For example, in patients receiving surgery for their signature. the primary treatment of head and neck cancer Clearly, during the period in which patients are followed by best supportive care, the randomised being assessed for eligibility and their consent controlled trial may be assessing the value of obtained, both the attending physician and the adding post-operative chemotherapy. In this case patient will be fully aware of the potential options the ‘control’ group are those who receive no being compared in the RCT. However, neither adjuvant treatment, whilst the ‘test’ group receive must be aware, at this stage, of the eventual chemotherapy. In certain circumstances, patients treatment allocation. It is important therefore may receive a placebo control. For example, that the randomisation list, for the current as in the randomised controlled trial conducted well as for future patients, is held by a neutral by Chow et al.12 in those with advanced liver third party. In most circumstances, this should cancer, patients are randomised to receive either be an appropriate trial office that is contacted placebo or tamoxifen. In this trial both patients by the responsible clinician once eligibility and and the attending physicians are ‘blinded’ to consent are obtained. This contact may be made the actual treatment given to individual patients. by telephone, fax, direct access by modem into Such a ‘double-blind’ or ‘double-masked’ trial a trial database, email or the web – whichever is is a design that reduces any potential bias convenient in the particular circumstance. It is to a minimum. Such designs are not possible then important that therapy is instituted as soon however in many circumstances and neither are as practicable after the randomisation is obtained. those with a ‘no treatment’ control. In many In specific cases, the randomisation can be con- situations, the ‘control’ will be the current best cealed within opaque and sealed envelopes which practice against which the new treatment will are distributed to the centres in advance of patient be compared. Should this turn out to be better recruitment. Once a patient is deemed eligible, than current practice then this, in its turn, may the envelope is taken in the order specified in a become standard practice against which future prescribed list, opened and the treatment thereby developments will be compared. revealed. Intrinsically, there is nothing wrong In general there will be both baseline and with this process but, because of the potential follow-up information collected on all patients. for abuse, it is not regarded as entirely satisfac- The baseline (pre-randomisation) information tory. However, in some circumstances it will be will be that required to determine eligibil- unavoidable; perhaps a trial is being conducted ity together with other information required to in a remote area with poor communications. In describe the patients recruited to the trial together such cases, every precaution should be taken to with those variables which are thought likely to ensure that the process is not compromised. influence prognosis. The key follow-up informa- The therapeutic options should be well des- tion will be that which is necessary to determine cribed within the trial protocol and details of the major endpoint(s) of the randomised con- what to do, if treatment requires modification trolled trial. Thus in the example of the burns or stopping for an individual patient should be patients these may be when the unhealed body given. Stopping may arise either when patients surface area finally closes or the size and sever- merely refuse to take further part in the trial or ity of the resulting scars. To establish the first of from safety concerns with a therapy under test. these, the burns areas may have to be monitored 16 TEXTBOOK OF CLINICAL TRIALS on a daily basis to determine exactly when the terms enormous. Similar types of trials have endpoint is achieved, whereas the latter may be a been conducted in patients with breast cancer, single assessment at the anniversary of the initial one in particular compared the three adjuvant burn accident. Pre-trial information on these end- treatment possibilities: tamoxifen, anastrozole or points, possibly from clinical experience or other their combination.14 studies, will usually form the basis of the antic- ipated difference between treatments (the effect size), help determine the trial size and be the INTERVENTION AND PREVENTION TRIALS variables for the statistical comparisons. The focus so far has been on randomised con- trolled trials in patients with medical conditions LARGE SIMPLE TRIALS requiring treatment or a medical intervention of some sort. Such designs do apply to situa- It has become recognised over time, particularly tions such as trials in normal healthy women in the fields of cardiovascular disease and cancer, in which alternative forms of contraception are 15 that there are circumstances where small ther- being tested. However, there are quite different apeutic advantages may be worthwhile demon- situations in which the object of a trial is to eval- strating. In terms of trial size, the smaller the uate alternative strategies for preventing disease potential benefit, essentially the effect size, then or for detecting its presence earlier than is rou- the larger the trial must be in order to be rea- tine. For example, intervention trials to encourage sonably certain that the small benefit envisaged ‘safe sex’ to prevent the spread of AIDS or breast really exists at all. screening trials to assess the value of early detec- Trials of this size, often involving many tion of the disease on subsequent treatment and thousands of patients, are a major undertaking. To prognosis. In such cases, it may be impossible to be practical, they must be in common diseases in randomise on an individual subject basis. Thus an order to recruit the required numbers of patients investigator may have to randomise communities in a reasonable time frame. They must be testing to test out different types of health promotion a treatment that has wide applicability and can be or different types of vaccines, when problems easily administered by the clinicians responsible of contamination or logistics, respectively, mean or the patients themselves. The treatments must that it is better to randomise a group rather than be relatively non-toxic else the small benefit will an individual. This is the approach adopted by 16 be outweighed by the side effects. The trials Donner et al. in a trial to compare a reduced must be simple in structure and restricted as antenatal care model with a standard care model. to the number of variables recorded, so that For this trial, because of the clustered randomisa- the recruiting clinicians are not overburdened by tion of the alternatives on a clinic-to-clinic basis, 17 the workload attached to large numbers of trial the Zelen single consent design was utilised. patients going through the clinic. They also need to be simple in this respect, for the responsible trial centre to cope with the large amounts of PHASE IV TRIALS – POST-MARKETING patient data collected. SURVEILLANCE One example of such a trial tests the value of aspirin in patients with cardiovascular disease.13 Within a regulatory framework, Phase IV trials This trial concerned a very common disease using are generally considered as ‘post-registration’ a very simple and low-cost treatment taken as trials: that is, trials of products that already have tablets with very few side effects. The resulting a marketing authorisation. However, within this estimates of absolute survival gain were (as post-registration period studies may be carried expected) small but the benefits in public health out for a variety of purposes, some within GENERAL ISSUES 17 their existing licence and others out-with that how is the size of a trial comparing three treat- licence. Studies may also be undertaken in ments, A, B and C, determined, since there are countries where a marketing authorisation has now three possible anticipated effect sizes that not been approved, in which case they are one can use for planning purposes? These corre- regulatory or Phase III-type studies, at least spond to the treatment comparisons A–B, A–C for that country. Studies may be undertaken to and B –C. The number of patients required for expand the indications listed on a marketing each of these comparisons may give three very authorisation either for a different disease or a different sample sizes. It is then necessary to different severity of the indicated condition. They decide which of these will form the basis for the may be undertaken to gain more safety data for final trial recruitment target, N. The trial will then newly registered products: this latter situation is randomise the patients equally into the three treat- more usually what is considered as a true Phase ment groups. In many circumstances, a three-arm IV study. trial will tend to require some 50% more patients Historically, pharmaceutical companies used to than a two-arm trial comparing two of the three carry out studies that were solely for market- treatments under consideration. ing purposes and answered very few (if any) Once the trial has been completed, the result- research questions. These were termed ‘seeding ing analysis (and reporting) is somewhat more studies’ although with tighter ethical constraints complex than for a simple two-group comparison. such studies are now very rare, if indeed they However, it is the importance of the questions take place at all. Certainly the ‘hidden’ objective posed, rather than the ease of analysis, which of many Phase IV studies carried out by pharma- should determine the design chosen. A good ceutical companies may be to increase sales, but example of the use of such a design is the previ- if the means of doing so is via answering a useful ously mentioned trial in post-menopausal patients scientific or medical question then this should be with breast cancer in which three options are of benefit both to society and to the company. compared.14 Many trials organised by academic depart- Nevertheless practical considerations may rule ments should also be considered as Phase IV out this choice of design. The design poses studies. Classic examples such as the RISC particular problems in data monitoring. For 13 Group trials looking at the cardiovascular ben- example, if an early advantage appears to favour efits of aspirin are studies of licensed products to one particular treatment and this suggests the trial expand their use. might be stopped early as a consequence. Then it may not be clear whether the randomisation between the other treatment groups should or ALTERNATIVE DESIGNS should not continue. Were the trial to stop early, then the questions relating to the other For illustrative purposes we have used the two- comparisons are unlikely to have been resolved arm parallel group RCT but these designs can be at this stage. Should the (reduced) trial continue generalised to compare three or more groups as then there may be very complex issues associated appropriate. In addition, there are other designs with its analysis and reporting. in common use which include 2 × 2 factorial and In addition, a potentially serious problem of crossover designs.18 bias can arise. At the onset of the trial, the clinician has to assess whether or not all the MORE THAN TWO GROUPS three options (A, B or C) available for treatment are suitable for the particular patient under Although not strictly a different design, a parallel examination. If any one of these were not thought group trial with more than two treatments to com- to be appropriate (for whatever reason), then the pare does pose some difficulties. For example, patient’s consent would not be sought to enter 18 TEXTBOOK OF CLINICAL TRIALS the trial and to be randomised. Suppose later in the trial, an interim analysis suggests recruitment I - Placebo to A is no longer necessary and that arm is Eligible and closed. For future patients, the clinician now has consenting Random II - Atorovastin alone to assess whether or not only the two options (B patients with allocation dyslipidaemia to or C) are suitable for the particular patient under in visceral treatment III - Fish oil alone examination. As a consequence, the patients now obsesity going into the trial are no longer potentials for A and hence may be somewhat different than those IV - Atorovastin entering at the earlier stages. Although this will and fish oil not bias the final comparison between B and C,it Figure 2.2. Randomised 2 × 2 factorial trial to does imply that there will be a bias if the patients determine the value of atorovastin, fish oil or both entering at this stage are compared with those in patients with dyslipidaemia in visceral obesity. receiving A. Reproduced from Chan et al.19 In the above, we have assumed that the three treatments A, B and C are, in a sense, unrelated albeit all suitable for the patients in In addition, the factorial design allows the mind. However, if they were related, for example, comparison of groups (I and IV) with (II and perhaps three doses of the same drug, then a III) and this estimates the so-called fish oil by note of this structure may change the approach atorovastin interaction. For example, suppose both to design from that outlined here. the main effect of fish oil alone and the main effect of atorovastin alone prove to be beneficial in this context, then an interaction would arise if the FACTORIAL DESIGNS combination treatment fish oil–atorovastin gives In a trial conducted by Chan et al.19 patients a benefit greater (or lesser) than the sum of these with dyslipidaemia in visceral obesity were constituent parts. As is the situation here, factorial randomised to either atorovastin alone, fish oil designs can be of a double placebo type, where alone, both or neither (placebo) to investigate subjects of Group I of Figure 2.2 receive double their influence on lipid levels. This trial may take placebo, one representing each treatment factor. the form of no treatment (1), atorovastin alone In principle, the 2 × 2or22 design can be (a), fish oil alone (f) or both atorovastin and fish extended to the 2k (k>2). Circumstances where oil (af). These alternative options are summarised this kind of design may be useful are if perhaps in Figure 2.2. the first two factors are major therapeutic or In contrast to the two-group parallel design of curative options and the third is a factor for Figure 2.1, where MEBO is contrasted with con- a secondary question not associated directly ventional treatment C, the factorial design poses with efficacy but (say) to relieve pain in such two questions simultaneously. Those patients subjects. However, the presence of a third factor assigned to groups I and II are compared with of whatever type increases the complexity of those receiving III and IV, to assess the value the trial design (not itself a particularly difficult of fish oil. This estimates the effect of fish oil statistical problem) which may have implications and is termed the ‘main effect’ of that treat- on the patient consent procedures and the timing ment. Those assigned to I and III are compared of randomisation(s). Nevertheless, a 23 design to those receiving II and IV to assess the main in a trial of falls prevention in the elderly effect of atorovastin. In most situations the final has been successfully conducted by Day et al.20 trial size will require fewer patients than would Piantadosi21 describes several examples of the use be necessary if the two questions were posed in of these designs. two distinct parallel group trials of the format of However, Green22 has issued a cautionary note Figure 2.2. that some of the assumptions behind the use of GENERAL ISSUES 19 factorial designs may not be entirely justified and ensures that the between-treatment comparison so any proposals for the use of such designs (A v B) within the patient remains unaffected should be considered carefully. by anything other than the change in treatment itself and random variation. These considerations tend to restrict the applicability of the design CROSSOVER TRIALS to patients with chronic conditions such as, for example, arthritis, asthma or migraine. Senn24 In contrast to the design of Figure 2.1, in the describes in careful and comprehensive detail the crossover trial each patient will receive both role of crossover designs in clinical studies. treatment options, one followed by the other in A clear advantage of the design is that the two periods of time. The two treatments will be patient receives both options and so the analysis given either in the order A followed by B (AB) includes within-patient comparisons which are or the reverse (BA). The essential features of a more sensitive than between-patient comparisons, crossover design are summarised in Figure 2.3 implying that such trials would require fewer for the trial conducted by Hong et al.23 in patients subjects than a parallel group design comparing with erectile dysfunction. the same treatment options. Typically, in the two-period, two-treatment crossover trial, for eligible patients there is a run- in stage in which the subject receives neither EQUIVALENCE TRIALS treatment. At the end of this randomisation to either AB or BA takes place. Following active In certain situations, a new therapy may bring treatment in Period I (in effect either A or certain advantages over the current standard, B depending on the randomisation), there is a possibly in a reduced side-effects profile, easier wash-out interval in which again no treatment administration or cost. Nevertheless, it may not is given, after which Period II commences and be anticipated to be better with respect to the the (other) treatment of the sequence is given. primary efficacy variable. Under such conditions, The characteristics of this design, for example the the new approach may be expected to be at possible run-in and the wash-out period, imply least ‘equivalent’ to the standard in relation to that only certain types of patients in which active efficacy. treatment can be withheld in this way are suitable Trials to show that two (or more) treatments are to be recruited. Further, there is an implication ‘equivalent’ to each other pose special problems that the patient returns to essentially the same in design, management and analysis.25 ‘Proving state at the beginning of Period II as he or she the null hypothesis’ in a significance testing was in at the commencement of Period I. This scenario is never possible: the strict interpretation
W Random Korean red a Korean red Patients allocation ginseng s ginseng with to sequence h erectile of - dysfunction treatments Placebo o Placebo u t
Source: Reproduced from Hong et al.,23 with permission.
Figure 2.3. Randomised placebo controlled, two-period crossover trial of Korean red ginseng in patients with erectile dysfunction 20 TEXTBOOK OF CLINICAL TRIALS when a statistically significant difference has there is no difference between the treatments and not been found is that ‘there is insufficient then look for evidence to refute that null hypoth- evidence to demonstrate a difference’. Small esis. In the case of equivalence we specify the trials typically fail to detect differences between range of equivalence, (25 g per week in the treatment groups but not necessarily because above example) and then test two null hypothe- no difference exists. Indeed it is unlikely that ses. We test that the observed difference is statis- two different treatments will ever exert truly tically significantly greater than − ; and that the identical effects. observed difference is statistically significantly A level of ‘therapeutic equivalence’ should be less than + . In practice it is much easier to defined and this is a medical question, not a consider a confidence interval for the difference statistical one. For example in a study of weight between the treatment means and draw this on gain in pre-term infants, if two treatments show a graph with the agreed limits of equivalence. mean increases in weight to within 25 g per week Figure 2.4 shows various scenarios. Some cases then they may be considered as therapeutically show equivalence, some fail to show equivalence; equivalent. Note that in this example 25 g is some cases show a statistically significant differ- not the mean weight gain that is expected per ence, others fail to show a difference. Note that week – we would hope that would be much it is quite possible to show a statistically signif- more. But if infants receiving one feeding icant difference between two treatments yet also regimen had a mean increase of 150 g per week demonstrate therapeutic equivalence. These are then we would consider an alternative treatment not contradictory statements but simply a real- to be equivalent if the mean weight gain were isation that although there is evidence that one between 125 and 175 g per week. treatment works better than another, the size of Conventionally, to show a treatment difference, the benefit is so small that it has little or no we would state the null hypothesis as being that practical advantage.
Statistical significance? Not equivalent Ye s Uncertain Ye s Equivalent Ye s Equivalent No Equivalent Ye s Uncertain Ye s Not equivalent Ye s Uncertain No
−∆ O +∆ True difference Examples of possible results of using the confidence interval approach: −∆ to +∆ is the prespecified range of equivalence; the horizontal lines correspond to possible trial outcomes expressed as confidence intervals, with the associated significance test result shown on the left; above each line is the decision concerning equivalence. Source: Reproduced from Jones et al.,86 with permission from the BMJ Publishing Group.
Figure 2.4. Schematic diagram to illustrate the concept of equivalence GENERAL ISSUES 21
The analysis and interpretation can be quite fixed sample size approach might have been straightforward but the design and management curtailed earlier had a sequential design been of equivalence trials is often much more complex. utilised. Fayers et al.27 describe the issues faced In general, careless or inaccurate measurement, when designing a sequential trial using α- poor follow-up of patients, poor compliance with interferon in patients with renal carcinoma. The study procedures and medication all tend to bias accumulating patient data from this trial crossed results towards no difference. Since we are trying an early termination boundary which inferred an to offer evidence of equivalence, poor study advantage to α-interferon.28 design and procedures may therefore actually A fully sequential design will monitor the trial help to hide treatment differences. In general, patient-by-patient as the information on the trial therefore, the quality of equivalence trials should endpoint is observed from them. Alternatively, be demonstrably high. a ‘group’ sequential trial will utilise information One special subset of equivalence trials is from successive groups of patients. Computer termed ‘non-inferiority’ trials. Here we only wish programs to assist in the implementation of these to be sure that one treatment is ‘not worse than’ designs are available29 and a review of some of or is ‘at least as good as’ another treatment: if it is the issues is given by Whitehead.30 better, that is fine (even though superiority would The solid lines of Figure 2.5 indicate stopping not be required to bring it into common use). boundaries for declaring a statistically significant All we need is to get convincing evidence that difference between treatments A and B.Ifthe the new treatment is not worse than the standard. broken boundary is crossed, the trial stops, We still have the same design and management concluding that no significant difference was considerations but here, looking at Figure 2.4, we found between the treatments. Potentially, the would only be concerned with whether or not the number of preferences could continue indefinitely lower limit of the confidence interval is greater between the upper solid and broken lines or − than the non-inferiority margin ( ). between the lower solid and broken lines; in such a case no conclusion would ever be reached. SEQUENTIAL TRIALS In contrast to the open sequential design, There has been an implicit assumption in the the closed design of Figure 2.6 will continue trial designs discussed above, that the total (and to recruit to a pre-stated maximum or until a final) sample size is determined at the design boundary is crossed. Thus this design has a finite stage and before recruitment commences to the size and a conclusion will always be drawn. trial in question. This fixed sample size approach Again the solid lines indicate stopping boundaries essentially implies that the data collected during for declaring a statistically significant difference the conduct of the trial will only be examined between treatments and if the broken line is for efficacy once the trial has closed to patient crossed, concluding that no significant difference accrual. However, the vast majority of Phase III was found between the treatments. trials will tend to recruit patients over perhaps For more complex designs implementing a an extended interval of time and so information sequential option remains possible. For example, on efficacy will be accumulated over this period. a factorial structure of the treatments does permit A sequential trial is one designed to utilise this a sequential approach to trial design as the two accumulating knowledge to better effect. Perhaps questions can be monitored separately. Neverthe- to decrease the final trial size if the data are less, if one of the factors, say the chemotherapy indicating an advantage to one of the treatments option in the example we discussed previously, and this can be firmly established at an early stage is stopped by crossing a boundary during the or to extend the trial size in other circumstances. sequential monitoring, then the recruiting clini- In fact, Donaldson et al.26 give examples cian will only have to put two rather than four where trials that had been conducted using a choices to the patients. Again, the subsequent 22 TEXTBOOK OF CLINICAL TRIALS
20
A significantly better than B 10
0 No significant difference
−10 Excess preferences (A − B) Excess preferences B significantly better than A −20 0 10 20 30 40 50 Number of preferences Open sequential design. The solid lines indicate stopping boundaries for declaring a statistically significant difference between treatments A and B. If the broken boundary is crossed, then the study stops, concluding that no significant difference was found between the treatments. Potentially the number of preferences could continue indefinitely between the upper solid and broken lines or between the lower solid and broken lines; in such a case no conclusion would ever be reached. Source: Reproduced from Day3 (Figure 24), with permission.
Figure 2.5. Open sequential design patients recruited may then be somewhat dif- terms the use of sequential designs is still some- ferent from those at the first stage of the trial. what limited. This will then impact on the form of the subse- quent analysis. ZELEN’S DESIGNS In contrast, the three-arm trial cannot be so easily designed in a sequential manner although Although not strictly an alternative ‘design’ 31 32 both Whitehead and Jennison and Turnbull in the sense of those of this section, Zelen’s address the problem. They do not however appear randomised consent design combines aspects of to give an actual clinical example of their use. design with problems associated with obtaining There are however several problems associ- consent from patients to participate in clinical ated with the use of sequential designs. These trials. They were motivated by the difficulties range from difficulties of financing a trial of expressed by clinicians in obtaining consent from uncertain size, making sure the data is fully up- women who they wished to recruit to trials to-date as the trial progresses, to the more tech- with breast cancer.17,33 Essentially subjects are nical concerns associated with the calculation of randomised to one of two treatment groups. the appropriate confidence intervals. However, Those who were randomised to the standard Whitehead,31 see also Jennison and Turnbull,32 treatment (conventional dressing in Figure 2.1) has argued very persuasively that all these objec- are all treated with it. For these patients no tions can be resolved. Nevertheless, in relative consent to take part in the trial is sought. On GENERAL ISSUES 23
20
A significantly better than B 10
0 No significant difference
−10 Excess preferences (A − B) Excess preferences B significantly better than A −20 0 10 20 30 40 Number of preferences Closed sequential design. The solid lines indicate stopping boundaries for declaring a statistically significant difference between treatments A and B. For example, if out of ten patients expressing a preference for one or other treatment, nine preferred treatment B and only one preferred A, then the study would stop, concluding that B is significantly better than A. If the broken boundary is crossed, then the study stops and the conclusion is drawn that no significant difference was found between the treatments. Source: Reproduced from Day3 (Figure 4), with permission.
Figure 2.6. Closed sequential design
the other hand, those who are randomised to Eligible patients the experimental treatment (MEBO dressing in Figure 2.1) are asked for their consent; if they agree they are treated with the experimental Randomise treatment; if they disagree they are treated with Standard (G1) New (G2) the standard treatment. This is known as Zelen’s single consent design. An alternative is that those Obtain consent Obtain consent randomised to the standard treatment may also Ye s No Ye s No be asked if they accept that treatment; again, Treat with Treat with Treat with Treat with they are actually given their treatment of choice. standard new new standard This latter double consent design is described in Figure 2.7; in either case, the analysis must Evaluate Evaluate Evaluate Evaluate be by intention-to-treat, that is based on the treatment to which patients were randomised, not the treatment they actually received. Compare
The properties of these designs have been Source: After Altman et al.34 (Figure 3). examined in some detail by Altman et al.34 who concluded that: ‘There are serious statistical argu- Figure 2.7. The sequence of events to follow in ments against the use of randomised consent Zelen’s double randomised consent design, seeking designs, which should discourage their use’. In consent in conjunction with randomisation 24 TEXTBOOK OF CLINICAL TRIALS any event, they have rarely been used in prac- may be of assistance in interpreting the results tice although they continue to be advocated.35 of a trial. Despite some technical difficulties, it is Nevertheless, in the large antenatal care trial fairly certain that Bayesian methods will become described by Donner et al.16 the Zelen single more prominent in Phase II trial methodology. consent design is utilised. However, this is a For example, Tan et al.37 suggest how such ideas cluster randomised trial and the issues are some- may be useful in a Phase II programme. what different.
BAYESIAN METHODS RANDOMISATION AND ALLOCATION TO TREATMENT The essence of Bayesian methodology in the con- text of the design of clinical trials is to incorpo- We have indicated above that randomisation of rate relevant information into the design process. patients to the treatment they receive is an At a later stage, Bayesian methods may assist important part of the ‘gold’ standard. In fact it is in data monitoring (as trial data accumulate) and the key element. The object of randomisation is to with the final analysis and interpretation of the help ensure that the final comparison of treatment (now complete) trial data. In theory the informa- options is as unbiased as possible, that is, that any tion available and which is pertinent to the trial in difference or lack of difference observed between question can be summarised into a prior distribu- treatments in efficacy is not due to the method by tion. This may include (hard) information from which patients are chosen for the options under the published literature and/or elicited clinical study. For example, if the attending clinician opinion. The mean of such a distribution would chose which of MEBO or C should be given to correspond to the possible effect size which can each patient, then any differences observed may then be assessed by the design team as clinically be due, at least in part, to the selection process worthwhile in their context and hence used for itself rather than to a true difference in efficacy. sample size estimation purposes. The same prior, Apart from the possible effect of the allocated or that prior updated from new external evidence treatments themselves, observed differences may accumulated during the course of the trial, may arise through the play of chance alone or possibly be used by the trial Data Monitoring Committee an imbalance of patients with differing prognoses (DMC) to help form the basis of recommenda- in the treatment groups or both. The object of tions with respect to future conduct of the trial. the statistical analysis will be to take account Perhaps to close the trial as efficacy has been of any imbalance and assess the role of chance. clearly established or increase the planned size of Some of the imbalance in the major prognostic the trial as appropriate. Finally once the trial data variables may be avoided by stratifying the is complete, these can be combined with the prior randomisation by prognostic group and ensuring (or updated prior) to obtain the posterior distribu- that an equal number of patients are allocated tion, from which Bayesian estimates of treatment within each stratum to each of the options. This effect and corresponding credibility intervals can may be achieved by arranging the randomisation be calculated. to be balanced within predetermined blocks of However, despite the feasibility of the above patients within each of the strata. Blocks are approach few trials to date have implemented usually chosen as neither too small nor too large, a full Bayesian approach. A review of articles four or six are often used. Sometimes the block in the British Medical Journal from 1996 to size, perhaps between these options, is chosen November 1999 found no examples.7 Never- at random for successive sequences of patients theless, Spiegelhalter et al.36 show convincingly within a stratum of patient types. how the concept of optimistic and sceptical prior Alternatively, the balance of treatments bet- distributions obtained from the clinical teams ween therapeutic options can be made using a GENERAL ISSUES 25 dynamic allocation procedure. In such schemes, so that they can be determined for each patient during the randomisation procedure a patient is recruited to the trial. It is good practice to identified to belong to a predetermined category define which endpoint is the major endpoint of according to certain covariates. This category the trial as this will be used to determine trial may, for example, be defined as those of a par- size and be the main focus for the efficacy ticular age, gender and tumour stage group. Once evaluation. In many situations, there may be the category is determined, then randomisation to several endpoints of interest, but in this case it the treatment options may proceed as described is important to order them in order of priority or above. One option, however, is to allocate the at least to identify those of primary or secondary next patient to the treatment with the fewest importance. If there are too many endpoints patients already assigned within that category. In defined, then the multiplicity of comparisons then which case, the allocation at that stage is deter- made at the analysis stage may result in spurious ministic. A better option in such circumstances is statistical significance. This is a major concern if to weight the randomisation, perhaps in the ratio endpoints for health-related quality of life and of 3:2 in favour of the option with the fewest health economic evaluations are added to the patients. Clearly, if numbers are equal, the ran- already established more clinical endpoints. domisation would revert to 1:1. We have implicitly assumed that, for two treat- SINGLE MEASURES ments, a 1:1 randomisation will take place. For all practical purposes, this will be statistically In some trials a single measure may be sufficient the most efficient. However, the particular con- to determine the endpoint in each patient. For text may suggest other ratios. For example, if example, the endpoint may be the diastolic the patient pool is limited for whatever reason, blood pressure measured at a particular time, say then the clinical team may argue that they should 28 days post-randomisation in each patient. In obtain more information within the trial from this case the treatment groups will be summarised the test treatment rather than the well-known by the respective means. In some situations the standard. Perhaps, there is a concern with the endpoint may be patient response, for example, toxicity profile rather than just the efficacy per the patient becomes normo-tensive following a se. In such circumstances, a randomisation ratio period of treatment. Those who respond are of say 2:3 or 1:2 in favour of the test therapy termed successes and those that do not failures. In this case, the treatment groups will be may be decided. However, some loss of statistical summarised by the proportion of responders. If, power will ensue and this loss should be quan- on the other hand, the patients are categorised tified before a decision on the allocation ratio is as: normo-tensive, still hypotensive but diastolic finally made. blood pressure (DBP) nevertheless reduced, or still hypotensive and DBP not improved, then ENDPOINTS this would correspond to an ordered categorical variable. Alternatively, the endpoint may be DEFINING THE ENDPOINT(S) defined as the time from randomisation and inception of treatment for the patient to become The protocol for every clinical trial will detail normo-tensive. In this situation repeated (say the assessments to be made on the patients daily) measures of DBP will be made until recruited. Some of these assessments may focus the value recorded is normo-tensive (as defined on aspects of the day-to-day care of the patient in the protocol). The interval between the date whilst others may focus more on those measures of randomisation and the date of recording the which will be necessary in order to determine the first occurrence of a normo-tensive recording is trial endpoint(s) for each subject. It is important the endpoint measure of interest. Such data are that these endpoints are unambiguously defined usually summarised using survival time methods. 26 TEXTBOOK OF CLINICAL TRIALS
A particular feature of time-to-event studies QUALITY OF LIFE occurs when the endpoint cannot be determined. For example, in the trial monitoring the DBP In many trials, endpoints such as the percent- it may be that a patient never becomes normo- age of patients responding to treatment, survival tensive during the trial observation period. In time or direct measures such as DBP have been which case the time from randomisation until used. In other situations, more psychosocial mea- the end of the trial observation period represents sures have been utilised such as pain scores, per- the time a patient has been under observation haps measured using a visual analogue scale, and but has not yet become normo-tensive. Such a emotional functioning scores, perhaps assessed survival time is termed censored and is often by patients completing a questionnaire them- + denoted by, say, 28 , which here means the selves. Such self-completed questionnaires have patient has been observed for 4 weeks but still also been developed to measure aspects of qual- remains hypotensive. In contrast, an observation ity of life (QoL) in patients undergoing treatment of 28 means the patient has been observed for for their disease. One such instrument is the SF- 4 weeks and became normo-tensive on the last 36 of Ware and Sherbourne,41 part of which is observation day. reproduced in Figure 2.8. The QoL domains measured by these instru- REPEATED MEASURES ments may then be used as the definitive end- points for clinical trials in certain circumstances. In the trial taking repeated DBP assessments, For example, in patients with terminal cancer these are recorded in order to determine a single the main thrust of therapy may be for palliation outcome – ‘time to becoming normo-tensive’. In (rather than cure) so that aspects of QoL may other situations, the successive values of DBP be the primary concerns for any comparison of themselves may be utilised in making the formal alternative approaches to management and care comparisons. If the number of observations made of such patients. If a single aspect of this QoL on each subject is the same, then the analysis measured at one time point is to be used for may be relatively straightforward, perhaps using comparison purposes, then no new principles are repeated measures analysis of variance. On the required either for trial design purposes or anal- other hand, if the numbers of observations ysis. On the other hand, and more usually, there recorded varies or if the intervals between may be several aspects of the QoL instrument successive observations vary from patient to that may need to be compared between treat- patient or if there is occasional missing data, then ment groups and these features will usually be the summary and analysis of such data may be assessed over time. This is further complicated quite complex. One option is to calculate the area by often-unequal numbers of assessments avail- under the curve (AUC) and use this as a single able from each patient caused either by missing measure for each patient, thus avoiding the use assessments in the series for a variety of reasons of more complex analytical methods.38 related or unrelated to their health status or per- However, the AUC method is now being haps in terminal patients by their death. Fayers superseded somewhat in Phase III trials by the and Machin42 and Fairclough43 discuss these fea- use of general estimating equations and multi- tures of QoL data in some detail. level modelling. The technical details are beyond As we have discussed previously, there is the scope of this book but most good statistics also a problem associated with the numerous packages39 now include facilities for these types statistical tests of significance of the multiple of analyses. Nevertheless, these methods have not QoL outcomes. These pose problems of inter- yet had much impact on the reporting of clinical pretation which have also been addressed by trials, although a good example of their use has Fayers and Machin42 (Chapter 14). In short, a been provided by Brown et al.40 cautious approach is needed to ensure apparently GENERAL ISSUES 27
The SF-36TM Health Survey
Instructions for Completing the Questionnaire
Please answer every question. Some questions may look like others, but each one is different. Please take the time to read and answer each question carefully by filling in the bubble that best represents your response.
EXAMPLE
This is for your review. Do not answer this question. The questionnaire begins with the section Your Health in General below.
For each question you will be asked to fill in a bubble, in each line.
1. How strongly do you agree or disagree with each of the following statements?
Strongly AgreeUncertain Disagree Strongly agree disagree
a) I enjoy listening to music. b) I enjoy reading magazines.
Please begin answering the questions now.
Your Health in General
1. In general, would you say your health is:
Excellent Very good Good Fair Poor
2. Compared to one year ago, how would you rate your health now ?
Much better Somewhat better About the Somewhat Much worse now than one now than one same as one worse now than now than one year ago year ago year ago one year ago year ago
Please turn the page and continue
© Medical Outcomes Trust and John E. Ware, Jr. —All Rights Reserved
For permission to use contact: Dr John Ware, Medical Outcomes Trust, 20 Park Plaza Suite 1014, Boston, MA 02116-4313, USA Source: Reproduced from Ware and Sherbourne,41 with permission.
Figure 2.8. 36-Item Short Form Health Survey (SF-36) 28 TEXTBOOK OF CLINICAL TRIALS
‘statistically significant’ differences are truly If we were to design a trial primarily to com- those. One way to overcome this problem is for pare costs associated with different treatments the clinical protocol to rank the domains of QoL we would follow the basic ideas of blinding and to be measured in terms of their relative impor- randomisation and then record subsequent costs tance and to confine the formal statistical tests incurred by the patient and the health provider. and confidence intervals to these only. A very careful protocol would be necessary to define which costs are being considered so that this is measured consistently for all patients. ECONOMIC EVALUATION A treatment that is not very effective might, for example, result in the patient needing more Most trials are intended primarily to address frequent consultations. The increased physician, questions of efficacy. Safety is frequently an nurse and other paramedical personnel contact important (though secondary) objective. Health time would then be recorded as a cost but it economics is becoming increasingly important needs to be clear whether patient travel costs, and is often now evaluated as part of a ran- for example (still direct costs, but not to the domised controlled trial. health service) are included, or not. However, There are four main types of cost analyses that most trials are aimed primarily at assessing effi- are usually considered: ciency and a limitation of investigating costs in • cost minimisation, simply to determine the best a clinical trial is that the schedule of, and fre- treatment to minimise the total cost of treating quency of, visits by the patient to the physician the disease; may be very different to what it would be in • cost effectiveness, a trade-off between the cost routine clinical practice. Typically patients are of caring for a patient and the level of efficacy monitored more frequently and more intensely offered by a treatment; in a trial setting than in routine clinical prac- • cost benefit, a trade-off between the cost of tice. The costs recorded, therefore, in a clinical caring for a patient and the overall benefit (not trial may well be different (probably greater but restricted to efficacy); possibly less) than in clinical practice. This is • cost utility, the trade-off between costs and sometimes put forward as a major objection to all measures of ‘utility’ which may include pharmacoeconomic analyses carried out in con- efficacy, quality of life, greater life expectancy junction with clinical trials. The same limitation or increased productivity. does, of course, apply to efficacy evaluations: the overall level of efficacy seen in clinical trials One of the big difficulties with pharmacoeco- is often not realised in clinical practice. How- nomic evaluations is determining what indirect ever, if we keep in perspective that, in a clini- costs should be considered. Direct costs are usu- caltrial,itistherelative efficacy of one treat- ally easier: costs of medication, costs of those ment over another (even if one of them is a giving the care (doctors, nurses, health visitors) placebo) then this limitation, whilst still impor- and the basic costs of occupation of a hospi- tant, can be considered less of an overall objec- tal bed. Indirect costs include loss of earnings tion. The same argument should be applied in and productivity, loss of earnings and produc- pharmacoeconomic evaluations and the relative tivity of spouses or other family members who increase/decrease in costs of one treatment over may care for a sick relative and contribution to another can be reported. hospital/pharmacy overhead costs. Because of the Recommendations on trials incorporating health ambiguity associated with these indirect costs, economics assessment have been given by the most pharmacoeconomic evaluations performed BMJ Economic Evaluation Working Party.44 as part of a clinical trial tend to focus solely on Neymark et al.45 discuss some of the method- direct costs. ological issues as they relate to cancer trials. GENERAL ISSUES 29
TRIAL SIZE these trials, some have an observed HR that is above the hatched horizontal line. This line When designing a new trial, a realistic assessment has been drawn at a level that is thought to of the potential benefit (the anticipated effect represent a clinically worthwhile advantage to the size) of the proposed test therapy must be made test treatment. These specific points would have at the onset. The history of clinical trials research been outside the funnel had they been estimated suggests that, in certain circumstances, rather from more observed deaths. Thus we might ambitious or over-optimistic views of potential conclude from Figure 2.9 that, had these trials benefit have been claimed at the design stage. been larger, the corresponding treatments may This has led to trials of inadequate size for the have been observed to bring worthwhile benefit, questions posed. rather than being dismissed as ‘not statistically’ The retrospective review by Machin et al.46 of significant. the published trials of the UK Medical Research However, it is a common error to assume that Council in solid tumour cancers is summarised the lack of statistical significance following a in the funnel plot of Figure 2.9. The benefit test of hypothesis implies no difference between observed, as expressed by the hazard ratio (HR) groups. Conversely, a statistically significant for the new treatment, is plotted against the result does not necessarily imply a clinically number of deaths reported in the trial publication. significant (important) result. Nevertheless, the Those trials within the left-hand section of the message of Figure 2.9 is that potentially useful funnel have relatively few deaths observed and therapies may be overlooked if the trials are so will be of correspondingly low power. Of too small.
30 Designed to demonstrate equivalence HR W Designed to demonstrate recurrence free survival O 2.0 R 1.8 S 1.6 E CE02 PE 1.4 BO02 ME LU02 1.2 LU06 CE01 LU03b BO01 LU02 LU09 CEB BAA LU09 BR01 1.0 LU08 HNA HNF LU04 LU07 0.9 BR03 CR01 OV01 CEA CR01 0.8 HNE PR01 BR02 B PR01 E 0.7 LU01 LU03a T HNB T 0.6 E NE R HNC 0.5 LU05 HND 0.4 100 200 300 400 500 600 Number of deaths Source: After Machin et al.46
Figure 2.9. Retrospective review of UK Medical Research Council trials in solid tumours published prior to 1996 30 TEXTBOOK OF CLINICAL TRIALS
ANTICIPATED (PLANNING) EFFECT SIZE and the alternative hypothesis a false negative rate. The former is variously known also as the A major factor in determining the size of a Type I error rate, test size or significance level, α. RCT is the anticipated effect size or clinically The latter is the Type II error rate β,and1− β worthwhile difference. In broad terms if this is the power. When designing a clinical trial it is not large then it should be of sufficient is often convenient to think in hypothesis-testing clinical, scientific or public health importance to terms and so set α and β and a specific effect warrant the consequentially large trial that will size for consideration. For determining the size be required to answer the question posed. If of a trial, α and β are typically taken as small, the anticipated effect is large, the RCT will be for example α = 0.05 (5%) and β = 0.1 (10%) relatively small in which case the investigators or equivalently the power 1 − β = 0.9 (90%) may need to question their own ‘optimistic’ view is large. of the potential benefit. In either case, a realistic If the trial is ultimately to compare the means view of the possible effect size is important. obtained from the two treatment groups, then In practice, it is usually important to calculate with randomisation to each treatment in equal sample size for a range of values of the effect numbers, the total sample size, N,isgivenby size. In this way the sensitivity of the resulting sample sizes to this range of values will provide 2 4(z − + z − ) options for the investigating team. N = 1 α/2 1 β ,(2.1) Estimates of the anticipated effect size may 2 be obtained from the available literature, formal meta-analyses of related trials or may be elicited where z1−α/2 and z1−β are obtained from tables from clinical opinion. In circumstances where of the standardised normal distribution for given there is little prior information available, Cohen47 α and β. = has proposed a standardised effect size, .Inthe If we set in equation (2.1) a two-sided α − = = case when the difference between two treatments 0.05 and a power of 1 β 0.9, then z1−α/2 = = = A and B is expressed by the difference between z0.975 1.96 and z1−β z0.9 1.2816, so that N = 42.028/ 2 ≈ 42/ 2. For large, moderate their means (µA − µB ) and σ is the standard deviation (SD) of the endpoint variable which is and small sizes of of 1, 0.5 and 0.1 the assumed to be a continuous measure, then = corresponding sample sizes are 42, 168 and 4200, respectively. More realistically these may be (µA − µB )/σ .Avalueof ≤ 0.1 is considered a small standardised effect, ≈ 0.5 as moderate rounded to 50, 200 and 4500. For a large simple = and ≥ 1 as large (see also Day3). Experience trial with 0.05, this implies 16 000 patients has suggested that in many clinical areas these may be recruited. can be taken as a good practical guide for design This basic equation has to be modified to purposes. However, for large simple trials, the adapt to the specific trial design (parallel group, equivalent of effects sizes as small as = 0.05 factorial, crossover or sequential), the type of or less may be clinically important. randomisation, the allocation ratio, as well as the particular type of endpoint under consideration. 48 SAMPLE SIZE Machin et al. provide examples for many different situations. Once the trial has been concluded, then a formal A good clinical trial design is that which will test of the null hypothesis of no difference answer the question posed with the minimum between treatments is often made. We emphasise number of subjects possible. An excessively large later that it is always important to provide an trial not only incurs higher costs but is also uneth- associated confidence interval for the estimate of ical. Too small a trial size leads to inconclusive treatment difference observed. The test of the null results, since there is a greater chance of missing hypothesis has an associated false positive rate the clinically important difference, resulting in a GENERAL ISSUES 31 waste of resources. As Pocock4 states, this too review. For example, the European Organization is unethical. for the Research on the Treatment of Cancer has a standing committee of three clinical and one statistical member, none of whom are MONITORING TRIAL PROGRESS involved in any way with the trials under review. This independent DMC reviews reports on trial DATA MONITORING COMMITTEES progress prepared by the data centre teams and makes specific recommendations to the relevant It is clear that a randomised controlled trial is a trial coordinating group. Early thoughts on the major undertaking, which clearly involves human structure of DMCs for the UK Medical Research subjects in the process. Thus, as we have stated, Council Cancer Therapy Committee are provided it is important that some form of equipoise in by Parmar and Machin.52 To emphasise the respect to the treatments under test is required importance of ‘independence’, such committees to justify the randomisation. However, once the sometimes choose the acronym IDMC. trial is in progress, information accumulates and as it does so it may be that the initial equipoise becomes disturbed. Indeed the very point of a SAFETY clinical trial is to upset the equipoise in favour of the best (if indeed one truly is) treatment. Although an IDMC will be concerned with Clearly there will be circumstances when the relative efficacy of the treatments under such early information may be sufficient to test, issues of safety will also be paramount in convincingly answer the question posed by the many circumstances. In many cases, safety issues trial. In which case the trial should close to may dominate the early stages of a trial when further patient entry. One circumstance when relatively new and untested treatment modalities this will arise is when the actual benefit far are first put into wider use, whereas in the exceeds that which the design team envisaged. later stages detailed review of safety may not For example, Lau et al.49 stopped a trial in be required as no untoward experiences have patients with resectable hepatocellular carcinoma been observed in the early stages. In contrast, after early results on 43 patients suggested serious safety issues may force a recommendation a substantial benefit to adjuvant intra-arterial for early closure of the trial even in situations iodine-131-labelled lipiodol. Their decision was where early indications of benefit in terms of subsequently criticised by Pocock and White,50 efficacy are present. Clearly the role of the who suggested the result was ‘too good to be IDMC or (in view of the ‘safety’ aspects) the true’ as early stopping may yield biased estimates IDMSC is to provide a balanced judgement on of the treatment effect. A confirmatory trial is these possibly conflicting aspects when making now in progress to substantiate or refute these their report. This judgement will derive from the findings.51 Essentially, although very promising, current evidence from the trial itself, external in this case the trial results as published provided evidence perhaps on new information since the insufficient evidence for other clinical teams to trial was inaugurated, and their own collective adopt the test therapy for their patients. experience. Nevertheless in this, and for the majority of clinical trials, it is clearly important to monitor INTERIM ANALYSIS AND EARLY the accumulating data. It has also been recognised STOPPING RULES that such monitoring should be reviewed (not by the clinical teams involved in entering patients At the planning stage of a clinical trial the design into the trial themselves) but by an independent team will be aware of the need to monitor the DMC. The membership and remit of a DMC will progress of the trial by reports to an IDMSC. On usually depend on the particular trial(s) under these occasions the data centre responsible for the 32 TEXTBOOK OF CLINICAL TRIALS conduct of the trial will expect to prepare reports published and negative studies tend not to be on many aspects of trial progress including published presents a distorted view of the true especially safety and efficacy. This requirement is situation. This approach to reporting is par- often detailed in the trial protocol. The detail may ticularly important for clinical trial overviews specify those aspects that are likely to be of major and meta-analysis where it is clearly impor- concern and also the timing (often expressed in tant to be able to include all relevant stud- terms of patient numbers or events observed) of ies (not just the published ones) in the overall such reports. synthesis. An interim report may include a formal The second aspect of reporting is the standard (statistical) comparison of treatment efficacy. of reporting, particularly the amount of neces- This comparison will then be repeated on the sary detail given in any trial report. The most accumulating data for each IDMSC and finally basic feature that has repeatedly been empha- following the close of the trial once the relevant sised is to give estimates (with confidence inter- data are to hand. These repeated statistical tests vals) of treatment effects and not just p-values. raise the possibility of an increased chance Guidelines for referees (useful also for authors) of falsely declaring a difference in efficacy have been published in several journals includ- between treatments. To compensate for this, ing the British Medical Journal.58 The Con- methods of adjusting for the multiple looks solidation of the Standards of Reporting Trials at the accumulating data have been devised. (CONSORT) statement is an international rec- Many of these are reviewed by Piantadosi53 ommendation adopted by many leading medical (Chapter 10). 59 Several of these methods of interim analy- journals. sis also include ‘stopping rules’; that is they One particular feature of the CONSORT state- incorporate procedures, or boundaries which once ment is that the outcomes of ‘all’ patients ran- crossed by the data under review, imply that the domised to a clinical trial are to be reported. trial should terminate. However, all these meth- Thus a full note has to be provided on those, ods are predicated on obtaining timely and com- for example, who post-randomisation refuse the plete data, very rapid analysis and report writing allocation and perhaps then insist on the competi- and immediate review by the IDMSC.54 They tor treatment. Two examples of how the patient also focus on only one aspect (usually efficacy) flow through a trial is summarised are given in and so do not provide a comprehensive view of Figure 2.10. the whole situation. It is of some interest to note that the writing The nature of the essential balance required team for Lau et al.49 were encouraged by the between a formal statistical approach to interim journal to include information on late (post- looks at the data and the less structured nature interim analysis) randomisations in their report. of IDMSC decision-making is provided by It is clear that no such stipulation was required Ashby and Machin,55 Machin56 and Piantadosi53 of the MRC Renal Cancer Collaborators.28 (Chapter 10.8). Parmar et al.57 describe an As indicated, the statistical guidelines referred approach for monitoring large trials using to, and the associated checklists for statistical Bayesian methods. review of papers for international journals,60 require confidence intervals (CIs) to be given REPORTING CLINICAL TRIALS for the main results. These are intended as an important prerequisite to be supplemented by The first rule after completing a clinical trial the p-value from the associated hypothesis test. is to report the results – whether they are pos- Methods for calculating CIs are provided in itive, negative or equivocal. Selective reporting many standard statistical packages as well as the whereby results of positive studies tend to be specialist software of Altman et al.61 GENERAL ISSUES 33
116 liver resections for diagnosis of hepatocelluar carcinoma
350 patients randomly 43 eligible patients randomised assigned treatment
174 assigned 176 assigned 21 patients in 22 patients in interferon-α MPA 131l-lipiodal group control group 7 excluded because 8 excluded because entered after interim entered after interim 18 patients received analysis analysis 131l-lipiodal 3 patients did not 167 included 168 included in interim in interim analysis analysis 21 patients followed up 22 patients followed up and completed trial and completed trial
Source: After MRC Renal Cancer Collaborators;28 Lau et al.49
Figure 2.10. Trial profiles following the CONSORT Guidelines
INTENTION-TO-TREAT (ITT) B, then the trial is no longer properly ran- domised and the resulting comparison may be As we have indicated, once patients have been seriously biased. randomised they should start treatment as spec- However, in certain circumstances, the ‘inten- ified in the protocol as soon as it is practi- tion-to-treat’ may be replaced or supplemented by cably possible. For the severely burnt patients a ‘per protocol’ summary.62 For example, if the either MEBO or C dressings can be immediately toxicity and/or side-effects profile of a new agent applied. On the other hand if patients, once ran- are to be summarised, any analysis including domised, have then to be scheduled for surgery, those patients who were randomised to the drug then there may be considerable delay before ther- but then did not receive it (for whatever reason) apy is activated. This delay may provide a period could seriously underestimate the true levels. If in which the patients change their mind about this is indeed appropriate for such endpoints, consent or indeed in those with life-threatening then the trial protocol should state that such an illness some may die before the scheduled date analysis is intended from the onset. of surgery. Thus, the number of patients actu- One procedure that used to be in widespread ally starting the protocol treatment allocated may use was, once the protocol treatment and follow- be less than the number randomised to receive up were complete, and all the trial-specific it. The ‘intention-to-treat’ principle is that once information collected on a patient, to review these randomised the patient is retained in that group data in detail. This review would, for example, for analysis whatever occurs, even in situations check that the patient eligibility criteria were where a patient after consent is randomised to satisfied and that there had been no important (say) A but then refuses and even insists on protocol deviations while on treatment. Any being treated by option B. The effect of such patients following this review, then found to be a patient is to dilute the estimate of the true dif- ineligible or protocol violators would then, in ference between A and B. However, if such a principle, be set aside and excluded from the patient was analysed as if allocated to treatment trial results. 34 TEXTBOOK OF CLINICAL TRIALS
One particular problem is one in which patients the texts we reference at the end of this chapter are recruited to a trial on the basis of clinical and many are covered in ICH E9 Expert Working examination during which a biopsy specimen is Group.65 However several aspects are fundamen- taken and sent for review. In the meantime the tal and these include the statistical significance patient is randomised and treatment commenced test, confidence intervals and analysis adjusted but once the report is returned the patient is for confounding (usually prognostic) variables. found not to comply with the eligibility criteria. The form of these techniques will depend on The above review process would automatically the design and especially the type of endpoint exclude this patient, whereas Freedman and variable under consideration. Thus if two means Machin63 argue otherwise. are to be compared then comparisons are made Usually, this review would not be blind to the using the Student t-test, for two proportions it treatment received – in fact even if the trial is is the χ 2 test which might be extended for an double blind – there may be clues once the data ordinal endpoint to the χ 2 test for trend. In are examined in this way as to which treatment is a survival time context, the difference between which. As a consequence, this process would tend treatments may, under certain conditions, be sum- to exclude more patients on the more aggressive marised by use of the hazard ratio (HR) and treatment. For example, the review conducted tested using the logrank test.66 Finally, whether by Machin et al.46 of some early randomised the endpoint variable is binary, ordered cate- trials in patients with cancer conducted by the gorical, continuous or a survival time, the cor- UK Medical Research Council showed that the responding between-treatment comparisons may earlier publications systematically reported on be adjusted for baseline characteristics of the fewer patients in the more aggressive treatment patients themselves using the appropriate regres- arm despite a 1:1 randomisation. The exclusion sion techniques. These are made by logistic of a larger proportion of patients receiving the regression, ordered logistic regression, linear more aggressive therapy would tend to bias the regression and Cox’s proportional hazards regres- results in its favour. Thus any patients who sion respectively. had ‘difficulties’ with the treatment, perhaps the Should the design include cluster randomisa- more sick patients, were not included in the tion, then Bland and Kerry67 state that the analy- assessment of its efficacy. This type of exclusion sis is done by group rather than on an individual was widespread practice, the consequences of subject basis. which included the development of ITT policies and standards for reporting clinical trials. The TESTS OF HYPOTHESES AND CONFIDENCE latter policy insisting that the progress of all the INTERVALS randomised patients be reported. In general, the application of ITT is con- If the data are continuous and can be summarised servative in the sense that it will tend to by the corresponding mean values in each of the dilute between-treatment differences. Piaggio and 64 two treatment groups A and B,thenasimple Pinol have pointed out that for equivalence tri- comparison is made using the difference d = als ITT will not be conservative but will tend to xA − xB and the test of the null hypothesis is favour the equivalence hypothesis. made using d z = (2.2) SE(d) SUMMARISING THE RESULTS Formally, this tests the null hypothesis that It is not relevant to review all the analytical the ‘true’ difference between treatments, δ,is techniques that may be utilised in summarising zero. For large trials z will have, under the clinical trial data. Many of these are described in assumption that the null hypothesis of equal GENERAL ISSUES 35 means is true, an approximately standard normal scientific conferences. There are many types of distribution from which the corresponding p- graphics that can be used but careful thought value can be obtained. should be given to the purpose of any graph. However, as has been pointed out it is very Graphs used for exploratory data analysis may important to report the observed difference d include histograms, scatter plots, etc. but these together with an associated confidence interval may not be appropriate for the final presentation for the true difference δ. In large samples, the of data. When presenting the results of clinical 100(1 − α)%CIisgivenby trials, the comparative nature of trials should be kept in mind and graphics produced that help in d − z1−α/2SE(d) to d + z1−α/2SE(d) (2.3) the communication of differences between treat- ments. In trials using time as an endpoint measure where z1−α/2 is obtained from tables of the the Kaplan–Meier survival curves provide an ele- 66 normal distribution. For a 95% CI z1−α/2 = gant summary (Figure 2.11). z0.975 = 1.96. Such a CI provides a sense of the precision NUMBER NEEDED TO TREAT (NNT) with which the observed difference between the two treatments is provided by the data. In broad Although many summary measures, for example terms, the width of the interval is determined by a difference in response rates or the hazard the number of subjects recruited, the larger the ratio, are utilised in clinical trials a measure number the narrower the corresponding CI. In unique to this context is the number needed general, the 95% CI will exclude the null value to treat. This is one very convenient way of (zero in this instance) if the corresponding p- assessing the treatment benefit from trials with a value <0.05. binary outcome. From the result of a randomised Although d provides a simple summary of trial comparing a new treatment with a standard the between-treatment group differences it is treatment, the NNT is the number of patients important to verify if this remains unchanged who need to be treated with the new treatment when taking full account of baseline character- rather than the standard (control) treatment in istics: sometimes termed confounding variables order for one additional patient to benefit. It or covariates. This is often achieved by using can be obtained for any trial that has reported regression techniques to adjust the observed dif- a binary outcome. ference, d, by the values of the baseline variables. The NNT is calculated as the reciprocal of In most circumstances, there will be some chance the difference between treatments where this is of imbalances in the values of the variables that expressed as a difference of two proportions may arise following randomisation. The adjust- (say) pT and pC for test and control treatments ment may affect the value of d itself as well as under study. Thus NNT = 1/(pT − pC) and a the associated standard error, SE(d), and hence large treatment effect thus leads to a small NNT. the CI. Such adjustments for important covari- A therapy that will lead to one life saved for ates affecting prognosis may result in the esti- every 10 patients treated is clearly better than a mate d being reduced, essentially unchanged or competing treatment that saves one life for every increased – which of these occurs will depend on 50 treated. When there is no treatment effects the the trial data itself. risk difference is zero and the NNT is infinite. A confidence interval for the NNT is obtained GRAPHS by taking reciprocals of the values defining the confidence interval for the treatment difference Graphical presentation of data is invaluable to itself. However, as Altman7 has pointed out there communicate results in published journal arti- are some difficulties if the treatment effect that cles or in presentations or posters presented at is not statistically significant and the confidence 36 TEXTBOOK OF CLINICAL TRIALS
1.0
0.9 n O E HR 95% CI for HR Placebo 130 113 131.4 - - 0.8 TMX60 74 69 69.5 1.16 (0.86, 1.38) TMX120 120 114 95.1 1.39 (1.07, 1.81)
0.7
0.6
0.5 Placebo
0.4 Cumulative survival Cumulative
TMX60 0.3
0.2
0.1 TMX120
0.0 0 2 4 6 8 10 12 14 16 18
Months from randomisation No. at risk (Death) Placebo 130 (0) 77 (52) 48 (27) 32 (15) 18 (9) 14 (3) 11 (3) 7 (2) 4 (1) 4 (0) TMX60 74 (0) 39 (35) 20 (19) 12 (5) 11 (1) 8 (3) 6 (2) 5 (1) 5 (0) 5 (0) TMX120 120 (0) 66 (53) 26 (40) 16 (9) 11 (5) 5 (5) 3 (2) 1 (0) 0 (0) 0 (0)
Source: After Chow et al.12
Figure 2.11. Kaplan–Meier estimates of survival in patients with inoperable hepatocellular carcinoma by double-blind treatment group interval includes the null value of 0 (see also the comparison has an associated two-sided test size comments by Hutton68). The NNT can also be α. This is set as the boundary below which obtained for survival analysis. For these studies the p-value, calculated from the data for the the NNT is not a single number, but varies primary endpoint of the trial, must fall to be according to time since the start of treatment. declared statistically significant. In this case the null hypothesis of no difference between groups MULTIPLE COMPARISONS is then rejected. When this approach is utilised in the analysis In a clinical trial in which two groups are of a clinical trial comparing two groups and there being compared, the formal statistical test of this is truly no difference between the two groups, GENERAL ISSUES 37 despite this ‘no difference’ there is a 100α% The problem of multiple comparisons is partic- probability of a statistically significant result and ularly acute in QoL assessments in clinical trials. the false rejection of the null hypothesis. Thus guidelines by Staquet et al.71 on reporting If more than one endpoint is measured for the such trials explicitly state: ‘...in the case of mul- two-group study in question then the situation tiple comparisons, attention must be paid to the becomes more complex. For example, if a total number of comparisons, to the adjustment, clinical trial is comparing two treatments (A if any, of the significance level, and to the inter- and B) but there are three different independent pretation of the results’. Perneger72 concludes outcomes being measured, then there are three that: ‘Simply describing what tests of significance comparisons to make between A and B and, have been performed, and why, is generally the in theory at least, three statistical tests. In best way of dealing with multiple comparisons’. this circumstance it can be shown that the In contrast the American Journal of Public Health false positive rate is no longer 100α% but recommend the use of a method of adjustment approximately 300α%. In fact for k (assumed due to Holm.73 From a clinical trials perspec- independent) outcome measures the false positive tive these issues are reviewed by Proschan and 74 rate is approximately 100kα%. Clearly, the Waclawiw. false positive rate increases as the number of comparisons made increases. SUBGROUP ANALYSIS In order to retain the false positive rate as In designing a RCT, sample size is usually deter- 100α% the Bonferroni correction is often sug- mined by considering a clinically worthwhile gested. This implies only declaring differences as effect which will be estimated from the trial data statistically significant at the 100α% level if the by a comparison of all patients randomised to one = observed p-value <α/k. In the case of α 0.05 group as compared to the other. It is recognised = and k 3, this implies p-value < 0.017. Equiv- that the precision with which this effect size is alently, and preferably, multiply the observed p- estimatedmaybeimprovedbyastratifiedanal- value by k and declare this significant if less ysis adjusted for baseline prognostic variables 69 than α. which are known to influence outcome irrespec- Similar considerations apply equally to CIs. tive of the treatment received. However, if treat- One approach that has been used to overcome this ments are compared within these strata (thereby difficulty is to quote 99% CIs rather than 95% CIs ignoring information on patients not in that stra- whenever more than a single outcome is regarded tum) it is clear that the patient numbers must be as primary. Thus the UK Prospective Diabetes less than the RCT as a whole. Thus any such Study Group70 report 21 distinct endpoints, comparisons will usually lack sufficient statistical ranging from fatal myocardial infarction to death power and hence may be unreliable. from unknown cause, and provide the 99% A common mistake is to observe a p-value confidence intervals for the corresponding 21 in excess of 0.05 (and hence judged naively as relative risks comparing tight with less tight not significant) but then to report subgroup analy- control of blood pressure. This is a ‘half-way ses – perhaps repeating the treatment comparison house’ proposal, since 0.01 (= 1 − 0.99) lies within each disease stage group. In some cir- between taking no account of the multiplicity cumstances, one of these subgroup comparisons and retaining 0.05 to define significance and the may be ‘significant’ (p-value < 0.05) which will Bonferroni corrected value of 0.05/21 = 0.0024. almost certainly imply that those within the other Similar considerations of multiplicity may also groups will not since the ‘all-data-combined’ apply in situations other than trials with multi- analysis is not significant. In extreme cases, sub- ple endpoints. For example, Green22 highlights group analysis can appear to favour one treatment this problem with respect to trials of a facto- in one subgroup and the other in the other sub- rial design. group(s). This may then lead to a false conclusion 38 TEXTBOOK OF CLINICAL TRIALS that the new treatment works for one group but COMPETING RISKS not for the other. If a subgroup analysis is planned for at the design stage, adjustment for this should In some situations, a patient may fail following be built into the sample size considerations. apparently successful treatment from one of C (≥2) so-called competing risk types, for example, a local recurrence or a distant metastasis, which BASELINE COMPARISONS are competing causes of ultimate mortality in can- Although the standard of reporting of randomised cer patients. If relapse is the outcome of interest controlled trials has improved in many medi- in the clinical trial, then usually it is the first event cal journals, there remain many who have not that is of primary importance to the clinician. yet adopted the full rigours of the CONSORT Thus, in a randomised trial of adjuvant therapy Guidelines. There are also many situations in in resected Dukes’ C colorectal carcinoma, 17- which inappropriate and substandard analyses are 1A monoclonal therapy was thought to be most conducted. Particular examples include statisti- effective against individually dispersed cells and less effective against local satellite tumour nod- cal significance tests of pre-randomisation (base- 78 line) variables, often describing demographic and ules or cell aggregates. Since the 17-1A anti- patient eligibility criteria in the different treat- body should be most effective in preventing or ment groups, despite the allocation to groups delaying distant metastases after surgery, distant having been made by randomisation so that any metastasis as a first event was thus a key endpoint observed differences are by definition random.75 in this trial. In the analysis of competing patterns of failure, the Kaplan–Meier method and the associated OTHER MISUSED APPROACHES logrank test are frequently used to estimate TO ANALYSIS the comparative rates of, for example, local recurrence and distant metastasis in patients Anderson76 catalogues some commonly misused receiving alternative treatments for their cancer in approaches used in the analysis of clinical trials. two distinct calculations. In one, local recurrence These include in, for example, cancer clinical as the first event is taken as the event of trials with dual endpoints of tumour response interest. In this situation, patients who do not and overall survival, the analysis of survival by have a local recurrence, or who have local tumour response itself. In these cases survival is recurrence as a second or subsequent event, compared between those patients who respond irrespective of whether or not they have distant and those who do not. Such analysis may lead to metastasis, are treated as censored observations in a false conclusion that response ‘caused’ longer the calculations. In the other calculations, distant survival. Anderson states categorically that such metastasis as the first event is taken as the event comparisons are wrong unless an appropriate of interest. methodology such as one involving a landmark is A preferred method, described in detail by Tai used. Similar considerations apply if comparisons et al.,79 is to use the cause-specific cumulative are made between groups established on the basis incidence functions and comparisons between of the amount of (protocol defined) treatment groups via the test developed by Gray.80 Gelman received, dose intensity or toxicity. Anderson et al.81 have pointed out that the Kaplan–Meier et al.77 provide details of the landmark approach. method produces estimates that are appreciably A common mistake is for investigators to higher than those of the competing risks method. provide treatment-specific CIs in their reports Substantial differences have also been noted for the endpoint(s) of concern, rather than for in data from patients with Ewing’s sarcoma.79 a relevant measure of treatment comparison such However a counter viewpoint on the use of the as their difference or a hazard ratio. two approaches is given by Farley et al.,82 using GENERAL ISSUES 39 illustrations from clinical trials in contraceptive However, the problems associated with ensuring development. this process is indeed in place in, for example, a large multinational trial involving prolonged SOFTWARE FOR STATISTICAL ANALYSIS follow-up of subjects are considerable. Indeed a whole new industry of Clinical Research There are many statistical packages available for Organisations (CROs) has developed to guarantee the summary and analysis of clinical trials and this process. Further, the Regulatory Agencies additional features are continually being added. rightly demand a guarantee of high quality data Pertinent features that distinguish the packages in any submission made to them for product are the flexibility and quality of graphical output registration. and the presentation of statistical tables or results. Information from randomised controlled trials Over the last decades, there have been many provides key information when the pharmaceu- statistical developments that have impacted on, tical and allied industries apply to register a for example, the endpoints that can be assessed, new drug or device with the relevant regulatory the design, size, analysis and summary. These authorities. The authorities themselves impose include the Kaplan–Meier method for summaris- certain constraints on the way in which trials ing survival time studies, the logrank test and are conducted – these will include basics with most influential of all the associated Cox pro- respect to a justification of a sample size for portional hazards model which allows between- the trial but will also specify standards.65 These treatment comparisons adjusted for potentially too have impacted on the ways that trials are confounding prognostic variables assessed at designed, conducted, analysed and reported. baseline. The Cox model can also accommodate time-dependent variables, that is variables that are assessed post-randomisation. These devel- PRINCIPLES OF QUALITY DATA opments would have remained theoretical in MANAGEMENT nature but for parallel developments in statisti- In clinical trials, subjects are usually entered cal software. one at a time, and their responses to treatment monitored sequentially. Regular monitoring of CLINICAL DATA MANAGEMENT trial progress, especially during the early stages, is advisable, and prompt attention to data errors, GOOD QUALITY DATA inconsistencies or missing items on the CRFs is required, so that corrections can be made Although it is rather outside the scope of this immediately. If inadequate control is exercised in text, a vital component to the eventual success of the management of clinical trials data, subsequent any clinical trial protocol is the quality of the analysis may be delayed or at worst wrong. associated clinical record forms (CRFs). Good The use of computers for data processing and CRFs will be pleasant to the eye, logical in statistical analysis requires careful planning and layout, comprehensive yet focused on the key execution by experienced personnel. Errors are information required, easy to complete and easy liable to happen in the recording of data in the to process. CRFs, as well as in the transfer of data to A key feature of any scientific study is the the computer using a database management implication that the data generated are of high software. quality. That is, that the observations made In a typical trial the sequence of data collection are carefully recorded at the point and time might be registration and randomisation, a record of observation, and then passed through to of the patient’s name or code, trial number, date the analysis stage without change. All this is of birth, date of randomisation and allocated equally true for clinical trials of whatever design. treatment (if the trial is not blind) will be 40 TEXTBOOK OF CLINICAL TRIALS entered into the database. The on-study form, transcription, electronic data entry or delay in which contains other demographic and baseline returns of CRFs, rather than the information not clinical information would be completed and having been collected. Queries should be raised added to the database in due course. Patients for any discrepancies identified by these checks. are then anticipated to return for follow-up The data should be edited accordingly after assessments often over an extended period of clarification with the investigator or comparison time. Since patients do not all enter the trial at with source documents. the same time, the amount of data collected at Although information in the CRF should be any one time may vary considerably as the trial checked manually and discrepancies resolved by progresses, and the management of the data on the data manager prior to data entry, on-line edit- a computer becomes increasingly complex and ing checks via computer provides an additional requires skilful intervention. means of detecting errors or missing data. The For repeated follow-up evaluations, a sepa- validation rules would normally be specified and rate record is normally kept for each evalua- the associated data checks programmed in paral- tion. For example, in a double-blind randomised lel with the building of the database. placebo controlled trial on inoperable hepatocel- lular carcinoma,12 information on ECOG perfor- SOFTWARE FOR CLINICAL DATA mance status, Okuda staging, serum albumin and MANAGEMENT bilirubin is collected. In addition, QoL is evalu- Several specialised software tools for managing ated not only on entry into the trial, but also at clinical trial data are commercially available. An monthly intervals. Since patients may drop out of early one written for the UK Medical Research a trial at any time for a variety of reasons this fur- Council was COMputer PAckage for Clinical ther complicates the management of the database. Trails (COMPACT)83 and this included many of Quality data management via computers entails the features necessary for handling ragged data careful planning and execution by well-trained sets which arise in clinical trails involving pro- data managers. As the processes of data tran- longed follow-up. However, current requirements scription and entry into computers are highly of GCP demand, for example, more intensive susceptible to producing errors, a series of checks audit trail facilities that were not part of the early for validity and completeness should be carried systems. The newer systems, unlike standard out immediately upon the return of CRFs. Items spreadsheets, incorporate all the basic features of that are commonly verified include range checks a good clinical data management system. for clinical parameters such as serum levels. An ideal clinical data management process is These may be excessively high or extremely low one that delivers valid and accurate data which (due to out-of-range or wrong units recorded), aid the maintaining of data integrity through or in the case of qualitative variables, a non- facilities for verifying and validating data, as well permitted code. as through the implementation of an audit trail Logical checks identify any inconsistency in to document database modifications. They also information that may be captured in different allow for automatic reporting of discrepancies, parts of CRFs. For instance, dates of randomisa- customised reports can be created and distributed tion, follow-ups and death carry important infor- to external sources such as the investigators mation in clinical trials where survival is the for resolution and have the ability to efficiently endpoint. Thus it is important to check that these handle repeated follow-up evaluations and track dates as well as other crucial ones have been status of patients throughout the trial. recorded and entered correctly. These systems also provide global libraries to Routine checks on missing items or forms store the definitions of standard code lists, and should also be undertaken, as any missing standard validation or derivation criteria, to opti- information could be due to oversight in data mise information input and access throughout the GENERAL ISSUES 41 operation. Standard codes can easily be com- Senn SJ. Statistical Issues in Drug Development. puted by defining appropriate derivation rules in Chichester: John Wiley & Sons (1997). the global libraries. These codes and rules can Williams CJ, ed. Introducing New Treatments for Cancer: Practical, Ethical and Legal Problems. be tailored to meet the varying requirements of Chichester: John Wiley & Sons (1992). individual trials and can be re-used in subse- quent databases. In multi-centre trials involving diverse and dis- DESIGN AND CONDUCT tant locations, it is possible with these clinical OF CLINICAL TRIALS data management systems to automatically propa- gate the study definitions, including amendments, Buyse ME, Staquet MJ, Sylvester RJ. Cancer Clinical Trials: Methods and Practice. Oxford: Oxford Uni- to all study sites. At each site, the local data versity Press (1984). can then be independently managed using the Crowley J, ed. Handbook of Statistics in Clinical same validation rules, derivation criteria and code Oncology. New York: Marcel Dekker (2000). lists. This is implemented by means of remote Jadad A. Randomised Controlled Trials: A User’s data entry. Guide. London: British Medical Journal Books (1998). With multiple sites and many users accessing Jennison C, Turnbull BW. Group Sequential Methods the same network, possibly performing different with Applications to Clinical Trials. Boca Raton: tasks, for example database design, entry and Chapman & Hall/CRC (2000). resolution by data managers, data retrieval, query Matthews JNS. Introduction to Randomised Controlled and analysis by statisticians, the security of the Clinical Trials. London: Arnold (2000). McFadden ET. Management of Data in Clinical Trials. system is of primary concern. Thus a clinical New York: John Wiley & Sons (1997). trial system should allow network-wide security Piantadosi S. Clinical Trials: A Methodologic Perspec- standards to be enforced by enabling system tive. New York: John Wiley & Sons (1997). administrators to assign, monitor and control Pocock SJ. Clinical Trials: A Practical Approach. access to sensitive clinical data records. Chichester: John Wiley & Sons (1983). Senn SJ. Cross-over Trials in Clinical Research. 2nd For the purpose of statistical analyses, there edn. Chichester: John Wiley & Sons (2002). should also be fast, flexible and easy extraction Whitehead J. Design and Analysis of Sequential Clini- of data into a variety of user-defined file formats, cal Trials, Revised 2nd edn. Chichester: John Wiley such as those of the more commonly used & Sons (1997). statistical packages. Wooding WM. Planning Pharmaceutical Clinical Tri- als. New York: John Wiley & Sons (1994). Ideally, a good clinical data management sys- tem should have facilities for randomising treat- ments, and registering all information captured at QUALITY OF LIFE this stage directly into the database. McFadden84,85 provides further details on Fairclough DL. Design and Analysis of Quality of many aspects of clinical data management and Life Studies in Clinical Trials. Andover: CRC Press (2002). associated areas. Fayers PM, Machin D. Quality of Life: Assessment, Analysis and Interpretation. Chichester: John Wiley & Sons (2000). SELECTED FURTHER READING Staquet MJ, Hays R, Fayers PM. Quality of Life Clin- ON CLINICAL TRIALS ical Trials: Methods and Practice. Oxford: Oxford Medical Publications (1998).
GENERAL SAMPLE SIZE Day S. Dictionary for Clinical Trials. Chichester: John Wiley & Sons (1999). Machin D, Campbell MJ, Fayers PM, Pinol APY. Redmond C, Colton T, eds. Biostatistics in Clinical Sample Size Tables for Clinical Studies. Oxford: Trials. Chichester: John Wiley & Sons (2000). Blackwell Science (1997). 42 TEXTBOOK OF CLINICAL TRIALS
EVIDENCE-BASED MEDICINE intravenous heparin in men with unstable coronary artery disease. Lancet (1990) 336: 827–30. Altman DG, Chalmers I, eds. Systematic Reviews. 14. The ATAC (Arimidex, Tamoxifen Alone or London: British Medical Journal Books (1995). in Combination) Trialists’ Group. Anastrozole Sutton AJ, Abrams KR, Jones DR, Sheldon TA, alone or in combination with tamoxifen versus Song F. Methods of Meta-Analysis in Medical tamoxifen alone for adjuvant treatment of post- Research. Chichester: John Wiley & Sons (2000). menopausal women with early breast cancer: first results of the ATAC randomised trial. Lancet (2002) 359: 2131–9. 15. Piya-Anant M, Koetsawang S, Patrasupapong N, REFERENCES Dinchuen P, d’Arcangues C, Piaggio G, Pinol A. Effectiveness of cyclofem in the treatment of depo medroxyprogesterone acetate induced amen- 1. World Medical Association. World Medical Asso- orrhea. Contraception (1998) 57: 23–8. ciation Declaration of Helsinki, 1996. Republished 16. Donner A, Piaggio G, Villar J, Pinol A, Al- in JAMA (1997) 277: 925–6. Mazrou Y, Ba’aqeel H, Bakketeig L, Belizan´ JM, 2. Chalmers I, Sackett D, Silagy C. Cochrane collab- oration. In: Redmond C, Colton T, eds, Biostatis- Berendes H, Carroli G, Farnot U, Lumbiganon P. tics in Clinical Trials. Chichester: John Wiley & Methodological considerations in the design of the Sons (2000) 65–7. WHO antenatal care randomised controlled trial. 3. Day S. Dictionary for Clinical Trials. Chichester: Paediat Perinatal Epidemiol (1998) 12(Suppl. 2): John Wiley & Sons (1999). 59–74. 4. Pocock SJ. Clinical Trials: A Practical Approach. 17. Zelen M. A new design for randomized clinical Chichester: John Wiley & Sons (1983). trials. New Engl J Med (1979) 300: 1242–5. 5. Temple R. Current definitions of phases of inves- 18. Wooding WM. Planning Pharmaceutical Clinical tigation and the role of the FDA in the conduct of Trials. New York: John Wiley & Sons (1994). clinical trials. Am Heart J (2000) 139: S133–5. 19. Chan DC, Watts GF, Mori TA, Barrett PH, Beilin 6. Gehan EA, Freireich EJ. Non-randomized con- LJ, Redgrave TG. Factorial study of the effects trols in cancer clinical trials. New Engl J Med of atorvastatin and fish oil on dyslipidaemia in (1974) 290: 198–203. visceral obesity. Eur J Clin Invest (2002) 32: 7. Altman DG. Statistics in medical journals: some 429–36. recent trends. Stat Med (2000) 19: 3275–89. 20. Day L, Fildes B, Gordon I, Fitzharris M, 8. Ang ES-W, Lee S-T, Gan CS-G, See PG-J, Flamer H, Lord S. Randomised factorial trial of Chan Y-H, Ng L-H, Machin D. Evaluating the falls prevention among older people living in their role of alternative therapy in burn wound manage- own homes. Br Med J (2002) 325: 128–31. ment: randomized trial comparing Moist Exposed 21. Piantadosi S. Factorial designs. In: Redmond C, Burn Ointment with conventional methods in the Colton T, eds, Biostatistics in Clinical Trials. management of patients with second-degree burns. Chichester: John Wiley & Sons (2000) 193–200. Medscape Gen Med (2001) 3(2): 3. 22. Green S. Factorial designs with time-to-event end 9. Freedman B. Equipoise and the ethics of clinical points. In: Crowley J, ed, Handbook of Statistics research. New Engl J Med (1987) 317: 141–5. in Clinical Oncology. New York: Marcel Dekker 10. Weijer C, Shapiro SH, Glass KC. Clinical equi- (2000) Chapter 9, 161–71. poise and not the uncertainty principle is the moral 23. Hong B, Ji YH, Hong JH, Nam KY, Ahn Ty. A underpinning of the randomised controlled trial. Br double-blind crossover study evaluating the effi- Med J (2000) 321: 756–8. cacy of Korean red gingseng in patients with 11. Enkin MW. Clinical equipoise and not the uncer- erectile dysfunction: a preliminary report. JUrol tainty principle is the moral underpinning of the (2002) 168: 2070–3. randomised controlled trial. Br Med J (2000) 321: 24. Senn SJ. Cross-over Trials in Clinical Research. 756–8. 2nd edn. Chichester: John Wiley & Sons (2002). 12. Chow PK-H, Tai B-C, Tan C-K, Machin D, John- 25. Simon R. Therapeutic equivalence trials. In: son PJ, Khin M-W, Soo K-C. No role for high- Crowley J, ed, Handbook of Statistics in Clinical dose tamoxifen in the treatment of inoperable Oncology. New York: Marcel Dekker (2000) hepatocellular carcinoma: an Asia–Pacific double- Chapter 10, 173–87. blind randomised controlled trial. Hepatology 26. Donaldson AN, Whitehead J, Stephens R, Ma- (2002) 36: 1221–6. chin D. A simulated sequential analysis based on 13. RISC Group. Risk of myocardial infarction and data from two MRC trials. Br J Cancer (1993) 68: death during treatment with low dose aspirin and 1171–8. GENERAL ISSUES 43
27. Fayers PM, Cook PA, Machin D, Donaldson AN, 43. Fairclough DL. Design and Analysis of Quality Whitehead J, Ritchie A, Oliver RTD, Yuen P. On of Life Studies in Clinical Trials. Andover: CRC the development of the Medical Research Council Press (2002). trial of α-interferon in metastatic renal carcinoma. 44. BMJ Economic Evaluation Working Party. Guide- Stat Med (1994) 13: 2249–60. lines for authors and peer reviewers of economic 28. Medical Research Council Renal Cancer Collab- submissions to the BMJ. Br Med J (1996) 313: orators. Interferon-α and survival in metastatic 275–83. renal carcinoma: early results of a randomised 45. Neymark N, Kiebert W, Torfs K, Davies L, Fay- controlled trial. Lancet (1999) 353: 14–17. ers P, Hillner B, Gelber R, Guyatt G, Kind P, 29. MPS Research Unit. PEST 4: Operating Manual. Machin D, Nord E, Osoba D, Revicki D, Schul- Reading: University of Reading (1993). man K, Simpson K. Methodological and statistical 30. Whitehead J. Use of the triangular test in sequen- issues of quality of life (QoL) and economic evalu- tial clinical trials. In: Crowley J, ed, Handbook of ation in cancer clinical trials: report of a workshop. Statistics in Clinical Oncology. New York: Marcel Eur J Cancer (1998) 34: 1317–33. Dekker (2000) Chapter 12, 211–28. 46. Machin D, Stenning SP, Parmar MKB, Fay- 31. Whitehead J. Design and Analysis of Sequential ers PM, Girling DJ, Stephens RJ, Stewart LA, Clinical Trials, revised 2nd edn. Chichester: John Whaley JB. Thirty years of Medical Research Wiley & Sons (1997). Council randomized trials in solid tumours. Clin 32. Jennison C, Turnbull BW. Group Sequential Oncol (1997) 9: 20–8. Methods with Applications to Clinical Trials.Boca 47. Cohen J. Statistical Power Analysis for the Behav- Raton: Chapman & Hall/CRC (2000). ioral Sciences, 2nd edn. New Jersey: Lawrence 33. Zelen M. Randomised consent trials. Lancet (1992) Earlbaum (1988). 340: 375. 48. Machin D, Campbell MJ, Fayers PM, Pinol APY. 34. Altman DG, Whitehead J, Parmar MKB, Sten- Sample Size Tables for Clinical Studies. Oxford: ning SP, Fayers PM, Machin D. Randomised con- Blackwell Science (1997). sent designs in cancer clinical trials. Eur J Cancer 49. Lau WY, Leung TWT, Ho SKW, Chan M, MachinD,LauJ,ChanATC,YeoW,MokTSK, (1995) 31A: 1934–44. Yu SCH, Leung NWY, Johnson PJ. Adjuvant 35. Machin D, Lee ST. The ethics of randomised intra-arterial iodine-131-labelled lipiodol for re- trials in the context of cleft palate research. Plastic sectable hepatocellular carcinoma: a prospective Reconstruct Surg (2000) 105: 1566–8. randomised trial. Lancet (1999) 353: 797–801. 36. Spiegelhaleter DJ, Freedman LS, Parmar MKB. 50. Pocock SJ, White I. Trials stopped early: too good Bayesian approaches to randomized trials. JR to be true? Lancet (1999) 353: 797–801. Statist Soc (1994) A 157: 357–416. 51. Tan SB, Machin D, Cheung YB, Chung YFA, 37. Tan S-B, Machin D, Tai B-C, Foo K-F, Tan E-H. Tai BC. Following a trial that stopped early: what A Bayesian reassessment of two Phase II trials of next for adjuvant intra-arterial iodine-131-labelled gemcitabine in metastatic nasopharyngeal cancer. lipiodol in resectable hepatocellular carcinoma? J Br J Cancer (2002) 86: 843–50. Clin Oncol (2002) 20: 1709. 38. Matthews JNS, Altman DG, Campbell MJ, Roys- 52. Parmar MKB, Machin D. Monitoring clinical tri- ton JP. Analysis of serial measurements in medical als: experience of, and proposals under consid- research. Br Med J (1990) 300: 230–5. eration by, the Cancer Therapy Committee of 39. StataCorp. STATA Statistical Software: Release the British Medical Research Council. Stat Med 7.0. Stata Corporation, College Station, TX (1993) 12: 497–504. (2001). 53. Piantadosi S. Clinical Trials: A Methodologic Per- 40. Brown JE, King MT, Butow PN, Dunn SM, spective. New York: John Wiley & Sons (1997). Coates AS. Patterns over time in quality of life, 54. Machin D. Discussion of ‘The what, why and how coping and psychological adjustment in late stage of Bayesian clinical trials monitoring’. Stat Med melanoma patients: an application of multilevel (1994) 13: 1385–9. models. Qual Life Res (2000) 9: 75–85. 55. Ashby D, Machin D. Stopping rules, interim anal- 41. Ware JE, Sherbourne CD. The MOS 36-Item ysis and data monitoring committees. Br J Cancer Short Form Health Survey (SF-36) I: conceptual (1993) 68: 1047–50. framework and item selection. Med Care (1991) 56. Machin D. Interim analysis and ethical issues 30: 473–83. in the conduct of trials. In: Williams CJ, ed, 42. Fayers PM, Machin D. Quality of Life: Assess- Introducing New Treatments for Cancer: Practical, ment, Analysis and Interpretation. Chichester: Ethical and Legal Problems. Chichester: John John Wiley & Sons (2000). Wiley & Sons (1992) Chapter 15. 44 TEXTBOOK OF CLINICAL TRIALS
57. Parmar MKB, Griffiths GO, Spiegelhalter DJ, Practice. Oxford: Oxford University Press (1998) Souhami RL, Altman DG, van der Scheuren E. 337–47. Monitoring of large randomised clinical trials: 72. Perneger TV. What’s wrong with Bonferroni a new approach with Bayesian methods. Lancet adjustments. Br Med J (1998) 316: 1236–8. (2001) 358: 375–81. 73. Holm . A simple sequentially rejective multiple 58. Altman DG, Gore SM, Gardner MJ, Pocock SJ. test procedure. Scand J Public Health (1979) 6: Statistical guidelines for contributors to medical 65–70. journals. In: Altman DG, Machin D, Bryant TN, 74. Proschan MA, Waclawiw MA. Practical guide- Gardner MJ, eds, Statistics with Confidence, 2nd lines for multiplicity adjustment in clinical trials. edn. London: British Medical Journal Books Contr Clin Trials, (2000) 21: 527–39. (2000) 171–90. 75. Altman DG, Dore´ CJ. Randomisation and base- 59. Begg C, Cho M, Eastwood S, Horton R, Moher D, line comparisons in clinical trials. Lancet (1990) Olkin I, Pitkin R, Rennie D, Schultz KF, Simel D, 335: 149–53. Stroup DF. Improving the quality of reporting ran- 76. Anderson JR. Commonly misused approaches in domized controlled trials: the CONSORT state- the analysis of cancer clinical trials. In: Crowley J, ment. J Am Med Assoc (1996) 276: 637–9. ed, Handbook of Statistics in Clinical Oncology. 60. Gardner MJ, Machin D, Campbell MJ, Altman New York: Marcel Dekker (2000) Chapter 24. DG. Statistical checklists. In: Altman DG, 77. Anderson JR, Cain KC, Gelber RD. Analysis of Machin D, Bryant TN, Gardner MJ, eds, Statistics survival by tumor response. J Clin Oncol (1983) with Confidence, 2nd ed. London: British Medical 1: 710–19. Journal Books (2000) 101–10. 78. Pugh RNH, Murray-Lyon IM, Dawson JL, Pietroni 61. Altman DG, Machin D, Bryant TN, Gardner MJ. MC, William R. Transection of the oesophagus for Statistics with Confidence, 2nd edn. London: bleeding oesophageal varices. Br J Surg (1973) 8: British Medical Journal Books (2000). 646–9. 62. Lewis JA, Machin D. Intention to treat: who 79. Tai B-C, Machin D, White I, Gebski V. Compet- should use ITT. Br J Cancer (1993) 68: 647–50. ing risks analysis of patients with osteosarcoma: a 63. Freedman LS, Machin D. Pathology review in comparison of four different approaches. Stat Med cancer research. Br J Cancer (1993) 68: 1047–50. (2001) 20: 661–84. 64. Piaggio G, Pinol APY. Use of the equivalence k approach in reproductive health. Stat Med (2001) 80. Gray R. A class of -sample tests for comparing 20: 3571–87. the cumulative incidence of competing risk. Ann 65. ICH E9 Expert Working Group. Statistical princi- Stat (1988) 16: 1141–54. ples for clinical trials: ICH harmonised Tripartite 81. Gelman R, Gelber R, Henderson IC, Coleman CN Guideline. Stat Med (1999) 18: 1905–42. and Harris JR. Improved methodology for ana- 66. Parmar MKB, Machin D. Survival Analysis: A lyzing local and distant recurrence. J Clin Oncol Practical Approach. Chichester: John Wiley & (1990) 8: 548–55. Sons (1995). 82. Farley TMM, Ali MM, Slaymaker E. Competing 67. Bland JM, Kerry SM. Statistics notes: trials ran- approaches to analysis of failure times with domised in clusters. Br Med J (1997) 315: 600. competing risks. Stat Med (2001) 20: 3601–10. 68. Hutton JL. Number needed to treat: properties and 83. Chilvers CED, Fayers PM, Freedman LS, Green- problems. J Roy Stat Soc A (2000) 163: 381–402. wood RM, Machin D, Palmer N, Westlake AJ. 69. Altman DG. Practical Statistics for Medical Re- Improving the quality of data in randomized clin- search. London: Chapman & Hall (1991). ical trials: the COMPACT computer package. Stat 70. UK Prospective Diabetes Study Group. Tight Med (1988) 7: 1165–70. blood pressure control and risk of macrovas- 84. McFadden ET. Management of Data in Clinical cular and microvascular complications in type Trials. New York: John Wiley & Sons (1997). 2 diabetes: UKPDS 38. Br Med J (1998) 317: 85. McFadden ET. Data management and coordina- 703–12. tion. In: Redmond C, Colton CT, eds, Biostatistics 71. Staquet MJ, Berzon RA, Osoba D, Machin D. in Clinical Trials, Chichester: John Wiley & Sons Guidelines for reporting results of quality of (2000) 158–67. life assessments in clinical trials. In: Staquet MJ, 86. Jones B, Jarvis P, Lewis JA, Ebbutt AF. Trials to Hays RD, Fayers PM, eds, Quality of Life assess equivalence: the importance of rigorous Assessment in Clinical Trials: Methods and methods. Br Med J (1996) 313: 36–9. 3 Clinical Trials in Paediatrics
JOHAN P.E. KARLBERG Clinical Trials Centre and Department of Paediatrics, The University of Hong Kong, Hong Kong SAR
WHY SHOULD WE DO CLINICAL TRIALS ‘Children are not simply Small Adults’ and its IN CHILDREN? Guidelines for the Ethical Conduct of Studies to Evaluate, published in 1995, reported that: Children are subject to many of the same diseases as adults, and are often treated with the same • In 1973, 78% of medications included a dis- drugs and biological products. However, many claimer or lack of dose information for children. drugs on the market used to treat children are • In 1991, 81% of listed drugs were restricted inadequately labelled for use with paediatric for certain age groups. patients; and many carry disclaimers stating that • In 1992, 79% of 19 new molecular entities safety and effectiveness in paediatric patients approved were not labelled for use in children. have not been established. Information about the safety and effectiveness of treatments for some As a result of effectively being denied access paediatric age groups is particularly difficult to well-studied drugs, paediatricians either don’t to find. Even today, no treatment is available treat children with potentially beneficial medica- for many of the thousands of rare and serious tions, or treat them with medications based either diseases that largely affect neonates, infants and on adult studies or anecdotal empirical experi- children. Most drugs used to treat common ence in children. Such non-validated administra- diseases in both children and adults have not been tion of medications may place more children at investigated in children at all. Over 50% of all risk than if the drugs were administered as part drugs prescribed in paediatric practice are either of well-designed, controlled clinical trials. There ‘unlicensed’ or ‘off label’. is therefore a moral imperative to formally study The paediatric medical community has for drugs in children, so they can enjoy equal access decades tried to persuade regulatory authorities to existing as well as new therapeutic agents. and the pharmaceutical industry to test new drugs The US National Institute of Health (NIH) in the paediatric population in parallel with the published regulations in 1999 clearly defining adult studies. The motto of the campaign has been what human studies could be funded by NIH
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 46 TEXTBOOK OF CLINICAL TRIALS to the exclusion of paediatric subjects. The REGULATORY ISSUES OF CLINICAL TRIALS exclusionary circumstances were: IN CHILDREN?
• Research topic irrelevant to children; Regulatory authorities in the US and Europe • Laws or regulations barring inclusion of chil- have in recent years taken important steps to dren; address the problem of inadequate paediatric test- • The knowledge is available for children or will ing and inadequate paediatric use information in be obtained from another ongoing study; drug and biological product labelling. But these • The relative rarity of the condition in children; efforts have, thus far, not substantially increased • The number of children is limited; the number of products entering the marketplace • Insufficient data are available in adults to judge with adequate paediatric labelling. The regulatory potential risk in children. authorities have therefore concluded that addi- tional steps are necessary to ensure the safety Not until recently have children been more reg- and effectiveness of drug and biological products ularly included in clinical studies to investi- for paediatric patients. Manufacturers of new and gate drugs. Considerable differences between marketed drugs and biological products must now the pharmacokinetics and pharmacodynamics evaluate the safety and effectiveness of the prod- of drugs in children and in adults frequently ucts in paediatric patients if the product is likely make it impossible to bridge conclusions from to be used in a substantial number of children, or data obtained in adults. Children cannot even provide a more meaningful therapeutic benefit to be considered a homogeneous group, since paediatric patients than existing treatments. age groups differ in their absorption, distri- Since 2000 in both the US and Europe, bution, metabolisation and excretion of drugs pharmaceutical companies have been obliged and their effect on developing organ sys- to include paediatric data in all new drug tems. The anatomical structure of children’s applications and licence extensions provided that organs differ from adults, causing different phar- substantial use of the drug in children and a macodynamic characteristics observed during meaningful therapeutic benefit is expected. The childhood. strength of this legislation is however different The lack of paediatric safety information in the two regions–and so is the extension of in product labelling exposes paediatric patients market exclusivity. to the risk of age-specific adverse reactions In recent years an independent ‘Orphan’ drug unexpected from adult experience. The absence regulation has been in force in the countries of of paediatric testing and labelling may also the European Community as well as in the US. expose paediatric patients to ineffective treatment This creates incentives for the development of through under-dosing, or may deny paediatric drugs for rare serious diseases, but is unlikely patients therapeutic advances. Failure to develop to achieve effective improvement in paediatric a paediatric formulation of a drug or biological drug therapy. The Food and Drug Administration product, where younger paediatric populations (FDA) Modernization Act established economic cannot take the adult formulation, may also incentives for pharmaceutical manufacturers to deny paediatric patients access to important conduct paediatric studies on drugs for which new therapies. patent protection or exclusivity is available Three conclusions can therefore be drawn under the Drug Price Competition and Patent about paediatric drug studies: studies must be Term Restoration Act or the Orphan Drug Act. made in different age groups; describing the phar- These provisions attach six additional months of macokinetics and pharmacodynamics is crucial; marketing exclusivity to any existing exclusivity and the safety of drugs must be studied to identify or patent protection of a drug for which FDA has potential severe side effects. requested paediatric studies. CLINICAL TRIALS IN PAEDIATRICS 47
However, there is likely to be a consensus development by Autumn 2002. This legislation during the coming years–at least in the ICH is considered by many to be pressing, creating GCP regions–over requirements for conducting the conditions needed to improve medicines for clinical trials on new drugs and other thera- children. Nearly all involved parties in Europe pies in children. But before this consensus can supported a legal and regulatory framework be reached, a number of points have to be for improving child health, especially regarding addressed and discussed, underlined by the fol- the labelling of medicines. The consultation lowing two examples. document concluded:
• EXAMPLE 1–ONGOING DISCUSSIONS A robust ethical framework for European pae- OF THE CONSENT PROCESS IN diatric research needs to be created, includ- PAEDIATRIC TRIALS ing guidance for informed consent, ethical review, recruitment of subjects, and safety The significant increase in the number of chil- and oversight. dren participating in clinical trials continues to • A robust paediatric clinical study infrastructure raise ethical and procedural concerns. The FDA needs to be created in Europe, since as a result addressed this issue in April 2001, calling on of reluctance to perform such studies up to institutional review boards to review study proto- now, there is a serious shortage of trained and cols that include children and ensure they adopt experienced people and centres of excellence. safeguards to protect young research participants. • Greater cooperation should be stimulated A group in the US is currently examining the between public sector research and private sec- ‘best practices’ related to research involving chil- tor research in paediatrics, in the interest of dren. The study will address: developing a European dimension to improv- ing medicines for children. • Process for obtaining informed consent and • A clear framework should be developed for assent from children and their parents or legal assembling international data and information representatives. regarding paediatric trials and medicines–to • How well participants in paediatric studies and ensure that unnecessary trials are not carried their guardians understand direct benefits and out in Europe, and that European paediatricians risks of study involvement? have the benefit of up-to-date and comprehen- • Definition of “minimal risk” related to healthy sive information regarding medicinal products and ill child study participants. for their patients, wherever in the world that • Whether regulations and policies should vary information has been generated. for children of different ages (for example, • A greater public dialogue is required in Europe teenagers and infants)? regarding the benefits and risks of paediatric • Appropriateness of payments to children, par- research for individual children participating ents, or legal representatives for participation in research, as well as for public health in research. in general. • Role of IRBs in monitoring compliance with regulations related to paediatric studies. ICH GUIDELINE ON PAEDIATRIC STUDIES EXAMPLE 2–ONGOING DISCUSSIONS OF THE LEGISLATION OF PAEDIATRIC TRIALS The International Conference on Harmonisa- tion (ICH) Guideline E11–Clinical Investigation Based on feedback from a consultation document, of Medicinal Products in the Pediatric Popu- the European Commission was expecting to lation–became operational in January 2001 in prepare draft legislation on paediatric medicinal the United States, Europe and Japan. The E11 48 TEXTBOOK OF CLINICAL TRIALS guideline outlines critical issues in paediatric • For medicinal products for diseases predom- drug development. In summary, this new and inantly or exclusively affecting paediatric important ICH document states that paediatric patients, the entire development programme patients should be given medicines that have will be conducted in the paediatric population, been appropriately evaluated for their use. It says except for initial safety and tolerability data, safe and effective therapy in paediatric patients which will usually be obtained in adults. requires the timely development of information • For medicinal products intended to treat seri- on the proper use of medicinal products in pae- ous or life-threatening diseases occurring in diatric patients of various ages and, often, paedi- both adults and paediatric patients, for which atric formulations of those products. The goal of there are currently no (or limited) therapeutic this guideline is to encourage and facilitate timely options, there is need for relatively urgent and paediatric medicinal product development inter- early initiation of paediatric studies. nationally. This guideline addresses five issues, • For medicinal products intended to treat other namely considerations when initiating a paedi- diseases and conditions there is less urgency. atric programme, the timing of its initiation, type Trials would usually begin at later phases of of studies, age categories and ethics. clinical development or, if a safety concern exists, even after a substantial post-marketing WHEN INITIATING A PAEDIATRIC period in adults. Testing of these medicinal PROGRAMME? (SUMMARY POINTS OF ICH products in the paediatric population would GCP E11) usually not begin until Phase II or III–since very early initiation of testing in paediatric The decision to proceed with a paediatric devel- patients might needlessly expose them to a opment programme for a certain medicinal prod- compound of no benefit. uct requires consideration of factors such as: TYPES OF STUDIES (SUMMARY POINTS • Prevalence of the condition in the paedi- OF ICH GCP E11) atric population; • Seriousness of the condition; Selection of the type of study should be on the • Availability and suitability of alternative treat- same principles as studies planned for adults. ments; However, several considerations are of specific • Unique paediatric indications; importance for paediatric studies. Some of the • Unique paediatric-specific endpoints; most important are: • Age ranges of paediatric patients likely to be treated; • When a medicinal product is to be used • Unique paediatric safety concerns; in the paediatric population for the same • Unique paediatric formulation development. indication(s) as those studied and approved in adults, the disease process is similar in adults The most common considerations when dis- and paediatric patients, and the outcome is cussing the need and timing of a paediatric pro- likely to be comparable, extrapolation from gramme are: adult efficacy can be appropriate. In such cases, pharmacokinetic (PK) studies in all the age • Most important is the presence of a serious or ranges of paediatric patients likely to receive life-threatening disease for which the medici- the medicinal product, together with safety nal product represents a potentially important studies, may provide adequate information. advance in therapy. This situation suggests rel- • When a medicinal product is to be used in atively urgent and early initiation of paedi- younger paediatric patients for the same indi- atric studies. cation(s) as those studied in older paediatric CLINICAL TRIALS IN PAEDIATRICS 49
patients, the disease process is similar, and the Special considerations should be taken when outcome is likely to be comparable, extrapola- blood is drawn more than once in paedi- tion of efficacy from older to younger paedi- atric subjects, such as in PK/PD studies. Sev- atric patients may be possible. In such cases, eral approaches can be used to minimise the pharmacokinetic studies in the relevant age amount of blood drawn and/or the number of groups of paediatric patients together with venipunctures: safety studies may be sufficient. • Many diseases in preterm and term newborn • Use of sensitive assays; infants are unique or have unique manifesta- • Use of laboratories experienced in handling tions precluding extrapolation of efficacy from small volumes; older paediatric patients and call for novel • Using routine clinical blood samples for phar- methods of outcome assessment. macokinetic analysis; • Where the disease course/outcome of therapy • Use of indwelling catheters; in paediatric patients is expected to be similar • Use of population pharmacokinetics and sparse to adults, but the appropriate blood levels sampling. are not clear, it may be possible to use measurements of a pharmacodynamic (PD) Efficacy effect related to clinical effectiveness. Thus, a PK/PD approach combined with safety and The principles in study design, statistical consid- other relevant studies could avoid the need for erations and choice of control groups are detailed clinical efficacy studies. in other ICH guidelines and apply to paediatric • When unique indications are being sought for efficacy studies. But there are also certain features the medicinal product in paediatric patients, unique to paediatric studies. or when the disease course and outcome of therapy are likely to be different in adults and • Extrapolation of efficacy from studies in adults paediatric patients, clinical efficacy studies in to paediatric patients, or from older to younger the paediatric population are needed. paediatric patients, as mentioned above. • For efficacy studies it may be important Pharmacokinetics to employ different endpoints for specific age groups. Pharmacokinetic studies generally should be per- • Measurement of subjective symptoms requires formed to support formulation development and different assessment instruments for patients of determine pharmacokinetic parameters in differ- different ages. ent age groups. Pharmacokinetic studies in the • The response to a medicinal product may vary paediatric population are generally conducted in among patients because of the developmental patients with the disease. Single-dose or steady- stage of the patient. state studies are the choice of pharmacoki- netic study: Safety • For medicinal products that exhibit linear phar- ICH guidelines (E2 and E6) describe adverse macokinetics in adults, single-dose pharma- event reporting and apply to paediatric studies. cokinetic studies in the paediatric population But there are certain safety aspects unique to may be sufficient. paediatric studies. • When there is a nonlinearity in absorption, distribution and elimination in adults and • Medicinal products may affect physical and difference in duration of effect between single cognitive growth and development, and the and repeated dosing in adults suggests steady- adverse event profile may differ in paediatric state studies in the paediatric population. patients, compared with adults. 50 TEXTBOOK OF CLINICAL TRIALS
• The dynamic processes of growth and devel- • Children(2to11years):Most pathways of opment may not manifest an adverse event drug clearance are exceeding adult values. at once, but at a later stage of growth Changes in clearance of a drug may be and maturation. dependent on maturation of specific metabolic • Long-term studies or surveillance data may pathways. The protocols should ascertain be needed to determine possible effects on assessment of the effect of the medicinal prod- skeletal, behavioural, cognitive, sexual and uct on growth and development. Recruitment immune maturation and development. of patients should ensure adequate represen- • Post-marketing surveillance may provide im- tation across the age range in this category. portant safety and/or efficacy information for Puberty can affect the activity of enzymes that the paediatric population. metabolise drugs, and dose requirements for • Age-appropriate, normal laboratory values and some medicinal products may decrease dramat- clinical measurements should be used in ically. adverse event reporting. • Adolescents (12 to 16–18 years (dependent on region)): This is a period of sexual maturation and medicinal products may interfere with the AGE CLASSIFICATION OF PAEDIATRIC actions of sex hormones. Medicinal products PATIENTS (SUMMARY POINTS OF ICH and illnesses that delay or accelerate the GCP E11) onset of puberty can have a profound effect Decisions on how to stratify studies and data by and may affect final height. Many diseases age need to take into consideration developmental are also influenced by the hormonal changes biology and pharmacology. The identification of around puberty and hormonal changes may which ages to study should be medicinal product- thus influence the results of clinical studies. specific and justified. Noncompliance is a special problem and compliance checks are important. • Preterm newborn infants: Preterm newborn infants have a unique pathophysiology and ETHICAL ISSUES IN PAEDIATRIC STUDIES responses to therapy. The complexity of and (SUMMARY POINTS OF ICH GCP E11) ethical considerations involved in studying preterm newborn infants requires a careful The paediatric population represents a vulnerable protocol development with expert input from subgroup. Therefore, the following special mea- neonatologists and neonatal pharmacologists. sures are needed to protect the rights of paediatric Only rarely can we extrapolate efficacy from study participants. studies in adults or in older paediatric patients to the preterm newborn infant. • Participants in clinical studies are expected to • Term newborn infants (0 to 27 days): Newborn benefit from the clinical study, except under infants are more mature than preterm newborn special circumstances. infants, but many of the physiologic and • When protocols involving the paediatric pop- pharmacologic principles for preterm infants ulation are reviewed, there should be IRB/IEC also apply to them. members or experts consulted by the IRB/IEC • Infants and toddlers (28 days to 23 months): who are knowledgeable in paediatric ethical, This is a period of rapid CNS maturation, clinical and psychosocial issues. immune system development and total body • Paediatric study participants are dependent growth. By 1–2 years of age, clearance of on their parent(s)/legal guardian to assume many drugs on a mg/kg basis may exceed adult responsibility for their participation in clinical values and then it may be dependent on specific studies. Participants of appropriate intellectual pathways of clearance. maturity should personally sign and date either CLINICAL TRIALS IN PAEDIATRICS 51
a separately designed, written assent form or First studies that promise no demonstrable ben- the written informed consent. efits to the child participating in the study or • Information that can be obtained in a less to children in general should not be conducted, vulnerable, consenting population should not irrespective of the minimal nature of the atten- be obtained in a more vulnerable population or dant risks. The risks include discomfort, incon- one in which the patients are unable to provide venience, pain, fright, separation from parents individual consent. or surroundings, effects on growth and develop- • Studies in handicapped or institutionalised ment of organs, and size or volume of biologi- paediatric populations should be limited to cal samples. diseases or conditions found principally in The proposed research must be of value to these populations, or when it is expected children in general and, in most instances, to the that the disease may alter the effects of a individual child subject: medicinal product. • To minimise risk in paediatric clinical studies, • The research design must take into consider- those conducting the study should be trained ation the unique physiology, psychology and and experienced in studying the paediatric pop- pharmacology of children and their special ulation, including the evaluation and manage- needs and requirements as research subjects. ment of potential paediatric adverse events. • The design should minimise risk while max- • In designing studies, every attempt should imising benefits. be made to minimise the number of partici- • The study design must take into account the pants and of procedures, consistent with good racial, ethnic, gender and socioeconomic char- study design. acteristics of the children and their parents. • To ensure that experiences of the study sub- • A placebo/observational control group may jects are positive and to minimise discomfort be acceptable when there is no commonly and distress. accepted therapy, or the commonly used ther- apy is of questionable efficacy, or the com- ETHICAL CONSIDERATIONS IN PAEDIATRIC monly used therapy has high frequency of side STUDIES effects, i.e. larger than the benefits.
Of all the problems surrounding research in PAEDIATRIC INFORMED CONSENT children, the one that poses perhaps the most complex question is research ethics. Children Children are not legally able to provide consent are not legally able to provide consent and the and the extent to which children are able to under- extent to which children are able to understand stand the meaning, risks and potential benefits of the meaning, risks and potential benefits of participating in clinical trials varies enormously participating in clinical trials varies enormously according to age and background. Children are according to age and background. For this counted as members of a vulnerable population reason it may be appropriate to address some at risk for exploitation and are given special pro- points related to the IRB review, including the tection in clinical research. In paediatric trials, informed consent process, in paediatric trials just as in adult trials, materials in an understand- more specifically than outlined in the ICH GCP able language, opportunities to discuss the trial, E11 guideline. One document that addresses this and freedom to withdraw without penalty must topic at more depth is the Review and Award be provided to potential subjects. Codes for the NIH Inclusion of Children Policy Investigators are ultimately held responsible from 1999. The following partly originates from for ensuring adequate informed consent. More this document, but also incorporates sources than two decades of inquiry into the process listed at the end of this chapter. of consent have shown that adults are less than 52 TEXTBOOK OF CLINICAL TRIALS adequately informed about risks, benefits and par- understand; and the purpose, risks and bene- ticipation in research. The process is even more fits of a study should be explained to them. problematic for research involving individuals The following guideline has therefore been pro- with limited abilities in decision-making. The posed: evolving psychological and emotional develop- ment of children and adolescents presents chal- No greater than minimal risk lenges to paediatric investigators not encountered • Assent of the child and permission of at least when dealing with adult subjects. Unless oppos- one parent. ing evidence is identified, capacity to understand and provide informed consent has long been Greater than minimal risk and prospect of assumed in adults. Results from studies in healthy direct benefit and sick children suggest that also children have this capacity. Several investigators have evalu- • Assent of the child and permission of at least ated the degree to which minors from school one parent. age through adolescence are capable of provid- • Anticipated benefit justifies the risk. ing assent. Even very young children demonstrate • Anticipated benefit is as least as favourable as inquisitiveness about the proposed research. By alternative approaches. the age of nine, children can understand purpose, risk and the right to withdraw from the study. Greater than minimal risk and no prospect of Even seven-year-old children can understand the direct benefit purpose of a study. Such observations support • the requirement by most ethics boards that assent Assent of the child and permission of both parents. be obtained in children aged seven and older. • However, information regarding scientific versus Likely to yield generalisable knowledge about therapeutic study objectives for both research and the child’s disorder. alternative treatment is not well understood in seven to 20-year-old subjects. Paediatric subjects SPECIFIC PROBLEMS OF PAEDIATRIC can thus provide an informed agreement to partic- STUDIES ipate, but the assent process should be conducted using discussions that encourage questions. SUBJECT RECRUITMENT
Obtaining Informed Permission–Assent–to Insufficient enrollment of children is the most Participate common reason for discontinuing paediatric stud- ies. Creating and expanding networks for paedi- Regulations permit studies involving minimal atric pharmacology studies, such as in the US and risk in children, with the provision that permis- Europe, are steps in the right direction to recruit sion from parents and assent from subjects are enough subjects. Many reasons for this poor obtained. Research involving greater than min- recruitment rate for paediatric studies include: imal risk, but providing potential direct benefit to the child, is also permitted with the same • Strict inclusion and exclusion criteria; provision. There are some exceptions to the • Limited size of the paediatric population; requirement for assent and consent. Assent is • That each age group has to be consid- not necessary for research expected to directly ered separately; benefit the child. Assent must be an active • Inconvenience for the parents in having their affirmation from any child with an intellectual children participate in a clinical study; age of seven years or older. Assent should be • Fear of making one’s own child available as a obtained from children who are competent to ‘guinea pig for research’. CLINICAL TRIALS IN PAEDIATRICS 53
• Doctors are wary of jeopardising the doc- Just taking adult protocols, then changing the age tor–patient relationship, or losing the trust in the inclusion criteria and the dose, is not good of parents. enough. With a limited number of investigators and a limited number of potential subjects, study EARLY TESTING design is critical for successful development of new safe and life-saving therapeutic entities for There are no healthy paediatric volunteers. The paediatric usage. lack of volunteers for Phase I studies is a special problem and makes the planning of therapeutic studies in children difficult. The FURTHER READING requirements for paediatric study designs are for this and other reasons different from studies in PUBLICATIONS adults. Alternative study designs and alternative • Shirkey H. Therapeutic orphans (editorial com- statistical methods are required. ment). J Pediatr (1968) 72: 119–20. • Rogers LC, Cocchetto DM. Labeling prescription STUDY MANAGEMENT drugs for pediatric use in the United States. Appl Clin Trials (1997) 50–6. • Tarnowski KJ, Allen DM, Mayhall C, Kelly PA. To obtain a sufficient number of subjects requires Readability of pediatric biomedical research infor- a large number of study centres. Moreover, the med consent forms. Pediatr (1990) 85: 58–62. cost for each individual step of a paediatric study • Susman EJ, Dorn LD, Fletcher JF. Participation in is usually higher than for studies in adults–both biomedical research: the consent process as viewed to pharmaceutical companies, as sponsors of the by children, adolescents, young adults, and physi- cians. J Pediatr (1992) 121: 547–52. studies, and to the participating doctors. For • Ethical Conduct of Studies to Evaluate Drugs in instance, explaining the nature of a study–to Pediatric Populations. Committee on Drug. Pedi- obtain permission from parents and ensure their atrics (1995) 95: 286–94. cooperation during the course of the study–is • Wechsler J. Science, pediatric studies, and surro- a very time-consuming process. Explanatory gates. Appl Clin Trials (1999) 28–33. • Nahata MC. Lack of pediatric drug formulations. material and information has to be prepared not Pediatrics (1999) 104: 607–9. only for parents, but also adapted for the children. • Conroy S, McIntyre J, Choonara I. Unlicensed and Caring for the children during their visits to off label drug use in neonates. Arch Dis Childh: the study centre also requires creativity, patience Fetal Neonatal Edit (1999) 80: F142–5. • and time. Conroy S, McIntyre J, Choonara I, Stephenson T. Drug trials in children: problems and the way forward. Br J Clin Pharmacol (2000) 49: 93–7. • Kauffman RE. Clinical trials in children: problems FINAL COMMENTS and pitfalls. Paediatr Drugs (2000) 2: 411–18. • Simar MR. Pediatric drug development; the Interna- Faced with heavy workloads, paediatricians may tional Conference of Harmonization Focus on Clin- ical Investigation in Children. Drug Inform J (2000) often be reluctant to assume what looks like 34: 809–19. the extra work of clinical trials. But a shortage of investigators is not the only problem that OFFICIAL DOCUMENTS/GUIDELINES slows paediatric trials. It takes many subjects to satisfy the requirements for an adult drug to be • Review and award codes for the NIH inclusion of adequately studied in children–and frequently the children policy March 26, 1999. • population of paediatrics with a certain disease Regulation (EC) No. 141/2000 of the European Parliament and the Council of 16 December 1999 does not exist. So not only do studies need to be on orphan medicinal products. designed to use small populations efficiently; they • World Medical Association, Declaration of Helsinki also need to be designed with children in mind. (amended October 2000). 54 TEXTBOOK OF CLINICAL TRIALS
• Food and Drug Administration, Department of information may produce health benefits in the Health and Human Services, The Pediatric Exclusiv- paediatric population (May 20, 2001), Docket ity Provision January 2001 Status Report to Cong- No. 98N0056ICH Topic Ell, ‘Clinical Investiga- ress. tion of Medicinal Products in the Pediatric Popu- • Food and Drug Administration, Additional Safe- lation’. CPMP/ICH/2711/99, January 2001, adopted guards of Children in Clinical Investigations of July 2000. FDA-Regulated Products: Interim Rule, Federal • European Commission, Better Medicines for Chil- Register 66 (79) 20589 (24 April 2001). dren: Proposed regulatory actions on paediatric • Food and Drug Administration, Update of list medicinal products (European Commission, Brus- of approved drugs for which additional paediatric sels, Belgium, 2002). 4 Clinical Trials Involving Older People
CAROL JAGGER AND ANTONY J. ARTHUR Department of Epidemiology and Public Health, University of Leicester, Leicester, UK
As few diseases or conditions present for the first 80 years of age, found only 38% of studies time in later life, there are few treatments pre- included subjects over 75 years of age.3 scribed solely to older people. There is also little Trials of the efficacy of interventions should consensus on the definition of ‘the elderly’ since cover the age groups who are affected.4 Older ageing can be considered a continuous process people have been explicitly excluded through from birth to death. However the increasing like- the use of a maximum age for eligibility and lihood of illness other than that under treatment obviously such trials provide little information and greater mental and physical frailty with age- about the efficacy of treatments in older age ing means that older people can be inherently groups. However implicit exclusion is also com- different to younger adults and the numerous mon, through criteria such as the presence of co- physiological changes that accompany the ageing morbid conditions. In addition certain recruitment process may alter the way in which older people methods may result in study populations with respond to drugs.1 older people an under-representation of the gen- By 2010 in most of the developed countries, eral population likely to be treated. In these cases the 65+ age group will form over 15% of the it may be difficult for the clinician to be aware of total population and over 20% in Japan. In the the paucity of older people studied, resulting in UK, those aged 65 years and over make up 18% the late recognition of serious side effects when of the population but they receive nearly half of drugs tested on predominantly younger adult pop- all prescriptions.2 Despite this, few trials, unless ulations are finally released and prescribed to specially designed and conducted in this age larger numbers of older people. Perhaps the most group, have sufficient numbers of older people, famous, or infamous, case of this was benoxapro- particularly the ‘oldest-old’, to provide evidence fen, a non-steroidal anti-inflammatory drug mar- of efficacy, even for treatments of diseases keted as Opren, which was withdrawn after a and conditions that are seen predominantly in report of the death of five elderly patients who later life. A recent review of clinical trials in had taken the drug.5 Parkinson’s disease, where prevalence increases In this chapter we will explore, in more with age and incidence peaks between 70 and depth, the reasons why older people face a
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 56 TEXTBOOK OF CLINICAL TRIALS number of barriers at each stage of a trial: The process of patient selection and recruit- eligibility, recruitment, gaining informed consent ment mostly aims to produce an homogeneous and follow-up. In addition we will discuss strate- study population with the purpose of increas- gies for increasing the number of older people in ing the statistical power to detect the effects of clinical trials, so that in future, those responsible drugs.10 The resulting clinical trial, conducted in for the treatment of older people will be able to ‘sterile’ conditions, bears little resemblance to base their practice on high-quality evidence. practice and cannot be extrapolated to the gen- eral population. Indeed, although tight eligibility criteria may aim to produce very similar par- ELIGIBILITY ticipants, inter-patient variability is such that a truly homogeneous group of patients is difficult, Despite recommendations to the contrary, older if not impossible, to identify. Important prog- people are still being excluded from clinical nostic variables will be measured at baseline, research on the basis of age alone, shown by an but even if study participants are the same on analysis of studies reported in four leading jour- these criteria, they will still vary in the course nals (BMJ, Gut,theLancet and Thorax) which of their disease and on unmeasured prognostic 11 found 35% (170) excluded older people with factors. Thus the gain in attempting to study a no justifiable reason.6 Reviews of trials in acute group of homogeneous patients is outweighed by myocardial infarction7 and Parkinson’s disease,3 the loss in generalisability and clinical applica- conditions where prevalence and incidence are bility of the results. Even when treatment trials strongly related to advancing age, found that tri- are specifically designed for older people, overly als published later were more likely to exclude stringent exclusion criteria can produce highly older subjects. Moreover, since more women than skewed and non-representative patient popula- men survive to older age and in some cases, such tions. Many trials of treatments for Alzheimer’s as cardiovascular disease, women develop dis- disease have excluded patients with behavioural eases later in life than men, exclusion on the basis problems despite such problems being common of age disproportionately disadvantages women.8 with increasing cognitive impairment. Since there Operating an upper age limit for trials has is considerable scope for improving such symp- often been used to limit the problem of co- toms with drugs that enhance cognition, these morbid conditions and drug interactions that may trials may well be missing opportunities.12 occur with increasing age, in the belief that Studying a narrow group of patients also most adverse drug reactions in older people misses the potential to identify subgroups of are simply a consequence of advancing age. A patients who may respond particularly well to the review of pertinent studies suggests that this drug under test. A trial comparing the efficacy may be misguided since the physiological and of sertraline and nortriptyline in major depres- functional characteristics of the patient, rather sion included patients aged 60 years and over, than chronological age per se, appear to be the but a subgroup analysis of the 76 patients aged most important in drug interactions.9 70 years and over suggested that treatment with Even when age limits are not imposed, older sertraline may confer even greater benefit in this patients are often implicitly excluded because of older age group than patients aged 60 years and other exclusion criteria or because of investigator, over.13 In trials of intervention packages or ser- cultural or other biases in enrollment.8 In the vices rather than drugs, similar tensions exist review of acute myocardial infarction trials, between maximising the detection of a significant comparison of the age distributions of patients effect of the intervention in a population unen- in trials with and without age exclusions showed cumbered with concurrent illness, and a need to no differences, suggesting that factors other than assess effectiveness as close as possible to deliv- explicit age restrictions were at play.7 ery in practice after the trial. The advantages of CLINICAL TRIALS INVOLVING OLDER PEOPLE 57 wide eligibility criteria for entering patients into Although the experience of earlier trials on clinical trials are summarised in Box 4.1.11 strategies to maximise recruitment may not be immediately transferable across time and place, they may provide researchers with ideas. Box 4.1 Advantages of wide eligibility Experiences in recruiting older people to tri- criteria for entering patients into als have been described in the treatment of 15 clinical trials hypertension with both pharmacological and non-pharmacological interventions16 and in tri- 1. Easier screening and recruitment. Large als of exercise.17 Many of the reports simply trials are more feasible and economical. describe the experience of one or two particular 2. Large study sizes reduce random error, strategies, though mass mailing, media advertis- providing more reliable overall results. ing, community-based screening, clinical practice 3. Wider applicability of results. There- screening, participant referrals and other recruit- fore greater clinical and public health impact. ment methods have been compared in a trial of the efficacy of weight loss and sodium reduction 4. Greater opportunity to test sub- 16 group hypotheses. for preventing hypertension in the elderly. This study concluded that mass mailing of a brochure Source: Reproduced from Yusuf et al.11 or letter describing the study resulted in the great- est yield in terms of percent randomised (76%; N = 737) though it is less clear whether this applied to all subgroups of the population. Tri- RECRUITMENT als recruiting volunteers, although producing a population who may be more likely to remain throughout the length of the study, provide little The recruitment, in sufficient numbers and within evidence of applicability to the general elderly the desired time frame, of motivated and compli- population. Older volunteers tend to be more ant subjects, representative of the wider group likely than younger ones to be healthy and liv- ultimately receiving treatment, is the goal of all ing independently, of particular importance for who design and execute clinical trials. Recruit- trials of interventions involving exercise since ing motivated participants is a problem for all volunteers may not be the subjects most likely clinical trials but particular difficulties are evi- to benefit.17 dent when recruiting older patients. Clinical tri- als are likely to involve more regular monitoring Rarely does one single strategy succeed in and follow-up assessments than would routinely recruiting adequate numbers of representative take place in practice and this in itself may be patients. It is important therefore that the char- too burdensome for older people who may have acteristics of participants are regularly monitored other health problems, which they may perceive throughout the trial, and compared to the gen- as more important, or lack access to transport. eral population, so that, if necessary, specific A longer enrollment phase, although deterring demographic groups, such as the oldest-old or some older subjects, may allow a more com- particular ethnic groups may be targeted. Such plete collection of data and more accurate pre- mixed-mode recruitment has produced represen- diction of patient compliance, again highlighting tative samples of high-risk older people for a trial the tension between pragmatic and explanatory of geriatric evaluation and management.18 The trials.14 If possible, assessments should be offered final sample should aim to be as representative at home or, where this is impossible, transporta- as possible and a list of strategies that could be tion to clinics provided at times convenient to used if shortfalls occur during recruitment should the subject. be developed at the design stage of the trial. 58 TEXTBOOK OF CLINICAL TRIALS
GAINING INFORMED CONSENT recalled, subjects found the concept of randomisa- tion difficult to accept and were confused by terms Obtaining informed consent from patients before such as ‘trial’ and ‘random’ which have different randomisation is a universal requirement, although meanings to lay and professional groups.20 legal requirements across countries may differ. Studies examining significant predictors of As with eligibility and recruitment, the means of enrollment into trials are equivocal in their find- gaining informed consent from subjects enrolling ings. A systematic review of literature on informed should be addressed at the design stage of the trial consent found evidence of impaired understand- and the information that the patient requires to give ing of the informed consent information in older informed consent is listed in Box 4.2. A synopsis subjects and those with less formal education,21 of the practicalities in obtaining informed consent whilst the Recruitment and Enrollment Assess- for clinical trials has been reported, stressing that ment in Clinical Trials study, part of the Car- gaining informed consent ‘should not be seen as diac Arrythmia Suppression Trial (CAST), did not an exercise in bureaucratic form filling, but as an find education differences in enrollers and non- essential part of the trial requiring time, insight and enrollers, although enrollers were more likely to 19 communication skills’. have read the informed consent themselves and to have understood it.22 The ability to understand information about clinical trials, particularly the Box 4.2 Patient information necessary for randomisation process, may well be correlated informed consent with level of education.23 An instrument to assess 1. Diagnosis. understanding of information given to ascertain 2. Available treatments and treatment informed consent for ambulatory trials has been on trial. developed, but its disadvantages are that it is study- 3. Potential risks and benefits of treat- specific and it was tested on relatively young and ment. well-educated subjects.24 4. Concept of a clinical trial (includ- Family members have also been found to play ing randomisation, use of placebos, an important role in the informed consent pro- double-blind procedures). cess, approval by family members, particularly 5. Discomforts or inconveniences associ- spouses, being associated with successful enroll- ated with assessments. ment in a cardiovascular trial.25 This study also 6. Number of follow-up visits or extra found that the majority (96%) of those approached travel for trial. by a physician agreed to enrol, compared to 66% of those approached by an experienced cardio- vascular nurse. Reasons given by a subsample of Clinicians may see relaying the concept of a those enrolled by the physician were predomi- randomised controlled trial as admittance of igno- nantly the trust and respect subjects had for their rance about the best treatment for the patient, or doctor, though a small number admitted to agree- may make ageist assumptions concerning the abil- ing through fear. ity of older people to consent to a trial. Further- Much healthcare provision is imperialistic and more, the clinical trial design is complex and even this may re-enforce the belief, held by some older if explained carefully, patients may not under- people, that all decisions relating to their treatment stand fully enough to give true informed consent. rest with their doctor. It should be remembered that A qualitative study, as part of a set of trials of not all older people however want active treatment the effectiveness of treatments for men with uri- in all cases and there may be reluctance to take nary retention and benign prostatic disease, found medication for certain conditions. A recent trial that, although information given was accurately of selective serotonin reuptake inhibitors in the CLINICAL TRIALS INVOLVING OLDER PEOPLE 59 treatment of depression and anxiety in community- gaining informed consent generally for trials, not dwelling older people found that, whilst 11 of just those specifically for dementia treatments. 67 people with clinically diagnosed depression Currently, informed consent is usually gained and/or a phobic anxiety disorder fulfilled one from proxies on behalf of dementia patients, or more of the exclusion criteria, 89% of the although technically only the subject may provide remaining subjects eligible refused medication,26 consent to be entered into a trial. Within clinical inferring that the process by which older people care there has been encouragement for patients to make decisions to participate in clinical trials is a prepare advanced directives or living wills to cover complex one, meriting further research. the eventuality that they may not have the capacity The nature of the trial may well be another to agree to treatment being given or withdrawn. At important factor in the decision by older peo- first sight this might appear a solution for dementia ple to enrol. Trials including invasive procedures research also, but the strong motivational factors such as venepuncture, which may be necessary to for individuals with clinical care are unlikely to determine compliance may not be seen as nec- be present for dementia research.30 In addition, essary to the subject and may therefore be more the number of people preparing living wills is likely to lead to refusal.27 Our experience of non- still very small and tends to be the well-educated, randomised studies of the health of older peo- higher social class groups. A more realistic future ple suggests that, on the whole, older people will goal might be that people are encouraged to name participate in lengthy interviews and are keen to proxies and state broad beliefs about research in assist with research that they perceive will help advanced directives. others,28 confirming findings from cardiovascular Rather than immediately approaching a proxy trials.25 When the study involves an assessment for consent with dementia patients, it may be at an outpatients’ clinic or an invasive procedure best to promote the pragmatic view of decision- such as providing a blood sample, refusal rates can making capacity that if an individual appears com- increase noticeably, making it necessary for trial petent then they are.31 Dementia patients have been personnel to explain clearly not only the need for shown to be capable of understanding and differ- randomisation but also the importance of subse- entiating the risk/benefit ratio between different quent assessments to monitor health. This should, treatments and of expressing their contentment of course, be balanced by any inconvenience the with having a proxy make decisions on involve- patient might incur by extra visits. ment in research although the proxies themselves After World War II, the Nuremberg code tended to be more protective with their relatives required the ‘voluntary consent of the human sub- than with themselves.31 A more pressing problem ject’ in experimental research and that ‘the per- is the lack of suitable proxies to provide informed son should have legal capacity to give consent’. consent on behalf of patients, one trial of palliative Ethicists argue that to make a properly informed care of patients with advanced dementia who had choice to enter a clinical trial, subjects should been hospitalised finding that almost half (72/146) not only have been provided with the necessary of eligible patients could not be enrolled.32 In only information about the trial but should also have four cases was this due to the proxy refusing con- understood it. The two aspects of having sufficient sent, the proxies themselves lacking the capacity understanding to give consent without coercion are to understand the protocol in 18% of cases and in key, but if taken literally, for example by being almost one-third no functional proxy being found. required to pass a test of competency, research on None of the patients for whom a proxy could not the efficacy of treatments and management strate- be found had made a living will. gies for dementia patients, particularly those with It is when older people with dementia are res- advanced dementia, would be impossible.29 The ident in nursing homes–with their dependency increasing prevalence and incidence of dementia and vulnerability–that the process of informed with advancing age may also pose problems for consent is most complex.33 Certainly the high 60 TEXTBOOK OF CLINICAL TRIALS prevalence of dementia within long-term care set- accommodate incomplete data. Finally, outcomes tings exacerbates the problems already stated, such as mortality that may be easy to measure and although the more immediate barrier from a important for younger populations may, in older researcher’s point of view may be the home- people, be valued less than quality of life and the owner acting as gatekeeper. Experience of con- ability to function independently.37 ducting censuses of older residents in all types of long-term care within Leicestershire over the last 20 years has shown, from the falling response CONCLUSIONS rates of homes, the increasing difficulty over 34,35 time in gaining access to residents or staff. If clinicians and other professionals caring for It is important that the inherent problems in older people are to provide optimal treatment and conducting research within long-term care set- the elderly are to benefit from new advances in tings are recognised by researchers at the out- treatments, decisions need to be based on firm set and that the shortage of staff time and often evidence of efficacy in older people. At present this rigid regimes in homes are taken into consid- is noticeably lacking in many aspects of care. We eration when designing trials, particularly those have discussed the reasons why older people have 36 of interventions. In these care settings, careful been and are still being excluded both explicitly explanation to home-owners and staff of the pur- and implicitly from trials. We cannot give any pose of the trial is likely to be vital to the success definitive solutions to ensure that older people of the study. are recruited and enrolled in large numbers into trials, since the setting for the trial (community, FOLLOW-UP nursing home, outpatient clinic) will influence the feasibility of different design options as well as Even when older people have been enrolled the country in which the research is conducted. into trials, greater risk of the development of However we list below important features that co-morbid conditions, cognitive impairment and should be considered when designing trials of polypharmacy will mean that they are more likely future therapies that may ultimately be used by to withdraw before the final outcome assessment. large numbers of older people: To some extent this can be planned for in advance by allowing for a realistic withdrawal rate in the • Aim for as wide eligibility criteria as possible sample size calculation. Provision of information to ensure smaller random error, a wider appli- about the trial should not be considered a ‘once and cability of results and a greater opportunity to for all’ activity at the commencement of the trial, test pre-planned subgroup hypotheses. and opportunities to re-enforce the importance of • At the design stage, agree a list of strategies the participants’ role in the success of the trial for recruiting specific subgroups (the very (for example at interim assessments) should be elderly, ethnic minorities) if these become exploited. More flexible timing of follow-up visits under-represented during recruitment. may also help and these days should not pose any • Regularly monitor the characteristics of subjects problems for analysis. enrolled to ensure good representation of the Missing data is still likely to be present and general population. more complex data analysis techniques should be • Give careful thought to the information to be used to maximise the use of the data that are given to subjects and the method by which it will present. Some statistical packages for repeated be given, to gain informed consent. Consider measures data analysis–a common analysis for whether and when consent will need to be trials with regular follow-ups–ignore cases with obtained from a proxy. data missing. Newer techniques such as multi- • If possible offer home assessments or, where level modelling and random effects models can this is impossible, provide transportation to CLINICAL TRIALS INVOLVING OLDER PEOPLE 61
clinics at times convenient to the subject and 10. Yastrubetskaya O, Chiu E, O’Connell S. Is good their caregivers. clinical research practice for clinical trials good • Design a realistic withdrawal rate into the clinical practice? Int J Geriat Psychiatry (1997) 12: 227–31. sample size calculation. 11. Yusuf S, Held P, Teo K. Selection of patients for randomized controlled trials: implications of wide Finally, many clinical trials fail because of poor or narrow eligibility criteria. Stat Med (1990) 9: recruitment, lack of adherence to protocols and 73–86. contamination between trial arms. The problems 12. Schneider LS, Olin JT, Lyness SA, Chui HC. Eli- outlined in this chapter may mean that this is a gibility of Alzheimer’s disease clinic patients for clinical trials. J Am Geriatr Soc (1997) 45: particular issue for trials involving older people. 923–8. There are useful lessons to be learnt from these 13. Finkel SI, Richter EM, Clary CM. Comparative experiences, yet by definition, these are rarely efficacy and safety of sertraline versus nortriptyline shared in the published literature. Methodological in major depression in patients 70 and older. Int issues that arise from others’ mistakes in carrying Psychogeriat (1999) 11: 85–99. out clinical trials involving older people need to 14. Applegate WB, Curb JB. Designing and executing be aired and discussed in journals. If the quality randomized clinical trials involving elderly per- sons. J Am Geriatr Soc (1990) 38: 943–50. of the evidence is improved then older people can 15. Petrovich H, Byington R, Bailey G, et al. Systolic expect to see an improvement in the quality of their Hypertension in the Elderly Program (SHEP) health care. Part 2: Screening and recruitment. Hypertension (1991) 17(Suppl II): 16–23. 16. Whelton PK, Bahnson J, Appel LJ, et al. Recruit- REFERENCES ment in the Trial of Nonpharmacologic Intervention in the Elderly (TONE). J Am Geriatr Soc (1997) 45: 185–93. 1. Stewart RB. Clinical research and the geriatric 17. Halbert JA, Silagy CA, Finucane P, Withers RT, patient: an assessment of needs. Clin Res Pract Hamdorf PA. Recruitment of older adults for a Drug Regulat Affairs (1985) 3: 477–500. randomized, controlled trial of exercise advice in a 2. Statistical Bulletin. Statistics of prescriptions dis- general practice setting. J Am Geriatr Soc (1999) pensed in the Family Health Service Authorities: 47: 477–81. England 1984–1994 (1995). 18. Boult C, Boult L, Morishita L, Pirie P. Soliciting 3. Mitchell SL, Sullivan EA, Lipsitz LA. Exclusion defined populations to recruit samples of high-risk of elderly subjects from clinical trials for Parkin- older adults. J Gerontol: Med Sci (1998) 53(5): son’s Disease. Arch Intern Med (1997) 157: M379–84. 1393–8. 19. Wager E, Tooley PJH, Emanuel MB, Wood SF. 4. ICH Harmonised Tripartite Guidelines. Studies in How to do it. Get patients’ consent to enter clinical support of special populations: Geriatrics (1993). trials. Br Med J (1995) 311: 734–7. 5. Taggart HMcA, Allardice JM. Fatal cholestatic 20. Featherstone K, Donovan JL. Random allocation jaundice in elderly patients taking benoxaprofen. Br Med J (1982) 284: 1372. or allocation at random? Patients’ perceptions of 6. Bugeja G, Kumar A, Banerjee AK. Exclusion of participation in a randomised controlled trial. Br elderly people from clinical research: a descriptive Med J (1998) 317: 1177–80. study of published reports. Br Med J (1997) 315: 21. Sugarman J, McCrory DC, Hubal RC. Getting 1059. meaningful informed consent from older adults: a 7. Gurwitz JH, Col NF, Avorn J. The exclusion of structured literature review of empirical research. the elderly and women from clinical trials in J Am Geriatr Soc (1998) 46: 517–24. acute myocardial infarction. JAMA (1992) 268: 22. Gorkin L, Schron EB, Handshaw K et al. Clini- 1417–22. cal trial enrollers vs. nonenrollers: the Cardiac 8. Wenger NK. Exclusion of the elderly and women Arrhythmia Suppression Trial (CAST) Recruit- from coronary trials. Is their quality of care ment and Enrollment Assessment in Clinical Trials compromised? JAMA (1992) 268: 1460–1. (REACT) project. Contr Clin Trials (1996) 17: 9. Gurwitz JH, Avorn J. The ambiguous relation 46–59. between aging and adverse drug reactions. Ann Int 23. Pucci E, Belardinelli N, Signorino M, Angeleri F. Med (1991) 114: 956–66. Patients’ understanding of randomised controlled 62 TEXTBOOK OF CLINICAL TRIALS
trials depends on their education. Br Med J (1999) 30. Sachs GA. Advance consent for dementia research. 318: 875. Alzh Dis Assoc Disord (1994) 8: 19–27. 24. Miller CK, O’Donnell DC, Searight HR, Barbarash 31. Sachs GA, Stocking CB, Stern R, Cox DM, Hou- RA. The Deaconess Informed Consent Comprehen- gham G, Sachs RS. Ethical aspects of dementia sion Test: an assessment tool for clinical research. research: informed consent and proxy consent. Clin Pharmacotherapy (1996) 16(5): 872–8. Res (1994) 42: 403–12. 25. DeLuca SA, Korcuska LA, Oberstar BH, Rosen- 32. Baskin SA, Morriss J, Ahronheim JC, Meier D, thal ML, Welsh PA, Topol EJ. Are we promoting Morrison RS. Barriers to obtaining consent in true informed consent in cardiovascular clinical tri- dementia research: implications for surrogate als? J Cardiovasc Nurs (1995) 9(3): 54–61. decision-making. J Am Geriatr Soc (1998) 46: 26. Stevens T, Katona C, Manela M, Watkin V, Liv- 287–90. ingston G. Drug treatment of older people with 33. Becker D. Informed consent in demented patients: affective disorders in the community: lessons from a question of hours. Med Law (1993) 12: an attempted clinical trial. Int J Geriat Psychiatry 271–6. (1999) 14: 467–72. 34. Campbell Stern M, Jagger C, Clarke M et al.Resi- 27. Ameer B, Burlingame MB. Difficulties enrolling dential care for elderly people: a decade of change. elderly patients in pharmacokinetic studies. J Br Med J (1993) 306: 827–30. Geriatr Drug Ther (1990) 4: 61–7. 35. Lindesay J, Jagger C, Parker G et al.TheTrent 28. Clarke M, Jagger C, Anderson J, Battcock T, Kel- census of older people in residential care. Report ly F, Campbell Stern M. The prevalence of demen- to NHS Executive Trent (1999). tia in a total population – the comparison of two 36. Eisch JS, Colling J, Ouslander J, Hadley BJ, screening instruments. Age Ageing (1991) 20: Campbell E. Issues in implementing clinical 396–403. research in nursing home settings. J New York State 29. Agarwal MR, Ferran J, Ost K, Wilson KCM. Nurses Assoc (1991) 22: 18–21. Ethics of ‘informed consent’ in dementia 37. Grimley Evans J. Evidence-based and evi- research – the debate continues. Int J Geriat dence-biased medicine. Age Ageing (1995) 24: Psychiatry (1996) 11: 801–6. 461–3. 5 Complementary Medicine
PING-CHUNG LEUNG Institute of Chinese Medicine, The Chinese University of Hong Kong, Hong Kong
INTRODUCTION What followed must have been more and more observations on more and more grasses Medicine has been an art of healing. Although and leaves which became considered as ‘herbs’. there is no absolute account on its history of Herb taking as a means to remove symptoms development in the prehistorical and extreme and ailments is, therefore, the standard early 1,2 primitive days, it must be closely related to the stage of the healing art in human history. The very ancient people’s eating habits and animal valuable observations and experiences were kept behaviour. Ancient people fallen sick must prefer until today. light meals with plenty of drinks. The latter might All primitive tribal populations today are still mean fruits and plant-related products, which are using herb treatment, as the standard popular the forerunners of medicinal herbs. method of healing. The practice does not rule out Ancient people lived with animals: either trial uses of new herbs and their combinations, keeping them as domestic friends or observing but the major practice depends on past experience them closely in the fields. Animal instincts and and documentations. These early clinical trials behaviour lent the ancient people much wisdom were not the result of imagination but were ini- of healing. When dogs and cats bit and ate up tiated after observations on the anecdotal effects 3,4 special grasses and leaves when they fell sick, of different herbs. followed by vomiting or diarrhoea, sometimes Traditionally there was no real need for large- bringing out special unwanted ingested food or scale clinical trials for complementary medicine. worms, the ancient people noted the special The need came only when scientific healers grasses and leaves. When they desired to clear became interested in complementary medicine up their guts under difficult circumstances, they and started making use of herbs and other meth- recalled those grasses and leaves their animals ods in their attempts to supplement modern ate and hence they imitated the animals, hoping medicine. They wanted to know whether, by util- to achieve the same healing. In this way, the ising the same logic of analysis commonly prac- primitive art of healing started. tised in modern medicine, they could prove that
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 64 TEXTBOOK OF CLINICAL TRIALS herbal treatment constituted a logical substitution 3. Massage or supplement to scientific medicine. 4. Chiropractice This chapter explores the promises and fal- 5. Spiritual healing by others lacies of clinical trials in herbal medicine and 6. Megavitamins some other complementary medicine; identifies 7. Self-help groups the similarities and difficulties, the developments 8. Imagery and limitations. 9. Commercial diets 10. Folk remedies 11. Lifestyle diets TYPES OF COMPLEMENTARY MEDICINE 12. Energy healing 13. Homeopathy The current main stream of medicine is the 14. Hypnosis scientific variety. Other forms of health care 15. Biofeedback outside the main stream fall into the category of 16. Acupuncture complementary or alternative medicine. 17. Self-prayer If one uses history as the criterion of iden- tification and considers ancient medicine was Of these varieties, the one that commanded the equivalent to complementary medicine, one sees highest popularity was acupuncture as a form of four main systems of ancient healing. They are: pain control. Chinese, Indian, Greek and Egyptian. Geograph- The author cannot possibly be knowledge- ically, the four systems are separated and yet, able about all the varieties of complementary the nearby areas do carry similarities. China and medicine, and is not able to discuss all their India certainly did communicate. So did Greece clinical trials. Rather, he will concentrate on the and Egypt. China probably also obtained infor- two varieties that he is familiar with, herbal mation from Greece, i.e. Europe, later in history medicine and acupuncture. While discussions are 5 through the ‘silk route’. being made, examples of clinical trials will be The four different systems have two main given, based on personal interests and experi- unique features. The Greek and Egyptian systems ences. concentrate on the use of single herbs, while the Chinese and Indian systems use multiple combi- nations. Combined formulae are most commonly FUNDAMENTAL CONSIDERATIONS FOR prescribed in Chinese herbal medicine. CLINICAL TRIALS ON CHINESE MEDICINE After thousands of years, the four ancient systems of medicine still survive well. Greek How should clinical trials of Chinese medicine be medicine in Europe has established itself as a conducted? Are there differences between such homeopathic healing art, while the other three trials and others designed for modern medicine? systems enjoy persistent but varying popularities. We have explained earlier that originally, In the modern sense, alternative/complemen- complementary medicine and its practitioners did tary medicine includes not only the herbal not demand clinical trials. However, clinical trials streams, but any other form of medicine that is are indicated for modern scientists because once unrelated to the modern scientific stream. When the efficacy is proven, an alternative methodology the American Medical Association did a survey in of treatment can be endorsed.7 the USA aimed at the revelation of the populari- If modern medicine were not totally successful, ties and users of alternative medicine, 17 modali- there would be a real need for supplementing ties were targeted.6 These included the following: with alternative medicine. Generally speaking, the successes of modern medicine are well 1. Relaxation techniques known in most areas. It is therefore necessary 2. Herbal medicine to look to complementary medicine only in COMPLEMENTARY MEDICINE 65 those areas where the scientific main stream has INDICATIONS AND PHILOSOPHY OF deficiencies. APPLYING COMPLEMENTARY MEDICINE
WHERE ARE THE DEFICIENCIES? Current medical treatment emphasises the effec- tiveness and statistical chances of obtaining good The deficient areas lie where modern medicine, results. Modern medicine has developed as direct in spite of recent advances, yet fails to get corrective measures. Hence when it is effective, good solutions. the probability of arriving at good results is Modern medicine has developed from the logic very high. Unless not available, there is there- of modern science which follows the deductive fore no reason why modern medicine should not approach. The problem is first thoroughly under- be endorsed as the primary mainline treatment. stood by identifying the cause. The cause could Although there are still confident herbal prac- then be removed by working out an effective titioners who believe and declare that whatever means. In the situation of a disease when the the modern scientific practitioners can do, they cause is simple and straightforward, removing it could substitute with other herbal remedies, the would be easy. On the other hand, when the cause number who remain that committed is getting less is either complicated, not well understood or mul- and less. Indeed today, most herbal practitioners tiple, removal becomes difficult or impossible. accept the role of functioning as supplementary or alternative healers in a combined effort of cure Simple disease-inducing causes include straight- 11 forward infections and simple hormonal deficien- and care. cies. The former is easily tackled with an efficient In this context, complementary herbal treat- antibiotic while the latter could be treated with ment is seldom used as the only healing modality. hormonal replacement. Instead, it is often given as an adjuvant treatment, When the causative agent is not yet thor- either together with the mainline or after comple- oughly known, e.g. viral infections, treatment tion of the mainline. Users of herbal preparations, becomes difficult. moreover, frequently look forward to a tonic sup- When the cause is complicated, e.g. in allergic portive effect, rather than a curative effect. conditions, treatment does not guarantee effec- tive results. HOW DOES HERBAL MEDICINE REALLY When the cause is complicated, e.g. involv- WORK? ing many factors such as physiological, social and psychological, modern scientific medicine Traditionally the system of herbal medicine was becomes obviously deficient or incapable.8–10 built on the rich experience of herb users or Therefore the deficient areas in modern medi- herbalists, accumulated over more than two thou- cine that deserve contributions from complemen- sand years in China since the early Chinese tary medicine include: culture. For some reason, while basic medical sciences, e.g. anatomy and physiology, devel- • Allergic conditions oped gradually in European territories around the • Autoimmune diseases Renaissance period, Chinese healers never felt • Cancers the need to explore these basic medical sciences. • Chronic pain Without sound knowledge of anatomy and phys- • Chronic derangements iology, i.e. biological structure and function of • Degenerative diseases the human body, it would not be possible to • Nerve damages explore abnormal structures and functions, i.e. • Viral infections pathology. Without understanding the pathology, • Other areas that modern conventional ther- it would not be possible to apply a direct means apy fails. of removing the pathology. Herbal practitioners 66 TEXTBOOK OF CLINICAL TRIALS therefore try to heal, not by direct confrontation superficial and deep, empty and solid. The causes with the pathological problem, but by indirectly of imbalance could be traced to a lack of bal- supporting the individual to overcome his own ance of any pair of opposing forces. In the difficulties.12,13 actual treatment, therefore, all efforts are spent on the maintenance of balance, by a supple- HOW DOES THE INDIVIDUAL OVERCOME ment of the deficient force, or a decrement of HIS OWN PATHOLOGICAL PROBLEMS? the excessive. Since the pathological causes of the symptoma- Firstly by surviving the harmful disturbances tology are unclear to the herbal expert, he would imposed by the pathological processes. Secondly need to observe the changes of symptoms and by supporting the unaffected organs and systems adjust his day-to-day protocol accordingly. This in their proper functions. Thirdly by preventing approach differs very much from conventional future pathological mishappenings while the modern medicine, which successfully identifies a current problem is being solved. pathological cause of disease, chooses a method The herbal practitioner has means to suppress of cure with good chance of success, then admin- the symptoms which are manifestations of the isters it with all effort and commitment, until the pathology. Suppression of symptoms like cough, total removal of the pathology is achieved. diarrhoea or dyspnoea helps the sick individual While the aetiology, epidemiology and natural to survive. course of a disease affect the design of clinical While waiting for the pathological damages trials for modern medicine, it is now clear that to heal naturally, the unaffected organs and in Chinese medicine, there is little analogy of systems need to be supported to maintain their aetiological and epidemiological considerations. efficient function, which in turn will support the The course of events in a disease, for a herbal overall function and metabolic harmony of the expert, is the appearance of the symptoms: the living individual. loss of biological well-being due to the lack of balance between the vital forces. The aim Prevention in the modern biological sense of treatment is the re-establishment of balance; frequently refers to an immunological mecha- once balance is re-established, either naturally or nism through which the individual becomes more through herbal intervention, well-being will be resistant to future attacks of similar pathologi- re-established. Treatment consists of a dynamic cal nature. application of symptomatic relief with the goal The main focus of disease management for of re-establishing the balance.14 Chinese medicine is often the control of adverse Clinical trials for Chinese medicine or herbal symptoms. The ultimate goal is maintaining the medicine therefore could follow the line of well-being of the biological system. The aeti- thought for scientific planning on data collection ological consideration is therefore not directed and subsequent data meta-analysis. However, the towards the actual cause of the disease (of which pre-treatment data would be confined mainly to the herbal expert has no idea), but a general symptomatology. Other parameters would carry conceptual state of the biological balance of the little weight for the herbal expert; but could human bodily functions. The ancient healers cor- still be included for more scientific knowledge related this conceptual state with the Taoist phi- in clinical trials. losophy and imagined that bodily function was kept at a balanced state between Yin and Yang (i.e. negative and positive). Any loss of balance GENERAL CONSIDERATIONS FOR CLINICAL led to ailment and disease. TRIALS ON HERBS The aim of treatment is therefore to main- tain the balance. Yin and Yang includes other In the modern scientific world, only up-to-date contrasting opposing forces like cold and heat, methodologies should be adopted. The set of COMPLEMENTARY MEDICINE 67 common methodologies for conducting clinical and metabolic pathways before clinical tests be trials on modern medicine has been logical, conducted. What is the chemistry of specific useful and has made wonderful contributions to herbs? What are the pathways of action and the clinical testing of new drugs and new meth- metabolic degradation? Are there adverse effects ods of clinical treatment. The proper analysis in the process of metabolism? A lot of work of data and the use of statistics have revealed has been done in the past 50 years on this basic the trustworthiness of certain accumulated expe- understanding and not much has come out. Each rience, while at the same time the fallacies of and every herb contains so much complicated some even well accepted and widely practised chemistry that many years of research might methods.15 not yield much fruit. Actually at least four The common methodology of random selec- hundred herbs are popular and possess records tion, blinding and placebo control, followed by of therapeutic action and impressive efficacy. To statistical analysis, should be adopted. In the demand thorough knowledge on just this popular design of the trial, good clinical practice should proportion of herbs is not practical, not to speak be the aim. However, due to the nature of the of the less commonly used extra one to two herbs, which come from different origins and thousand varieties.17 carry different species, it is not uncommon to Uniformity is another difficult area. Strictly encounter situations where basic principles that speaking, since herbs are agricultural products, 16 cannot be strictly kept. uniformity should start with the sites of agricul- tural production. The sites of production have THE OLD APPROACHES different weather, different soil contents, and the methods of plantation are also different. At the The herbal experts fervently respect case reports moment, maybe over 50% of popular Chinese and anecdotal reports, particularly when results herbs are produced on special farms in China. appear promising. Of course the reason behind However, these farms are scattered over different it is that they don’t make use of statistics. provinces, which have widely different climatic Moreover, they believe that treatment results and soil environments. Good agricultural practice are different with different patients. Once good demands that environmental and nurturing proce- results are known to be possible, the expert could dures be uniformly ensured. Procedures include try to achieve equally good results by wisely soil care, watering, fertilisers, pest prevention and manipulating the varieties of treatment. harvests. When such procedures are not uniform In this chapter, we do not endorse this and there are no means to ensure a common prac- traditional approach. We do want to apply modern tice, good agricultural practice is not possible. assessment tools for a better understanding of Not only is there a lack of uniformity in the herbal or Chinese medicine treatment while at the mode of herb production, but different species same time we need not argue against the value of the same herb are found or planted in dif- of anecdotal observations in Chinese medicine. ferent regions and provinces. These different After all, the development of this system of species may have different detailed chemical con- healing depends solely on anecdotal analysis. tents. Herbal experts have extensive experience Good clinical practice insists that the pre- and knowledge about some special correlations scribed drug for the clinical trial should be between the effectiveness of particular herbs and thoroughly known and uniform. However, using their sites of production. Some commonly used herbal preparations for clinical trials faces diffi- herbs are even labelled jointly with the best sites culties of thorough technical knowledge and uni- of production. With the development of molec- formity. ular biology, coupled with modern means of Pharmaceutical tests demand that details be assessment for active ingredients within a chem- known about the chemistry, the mode of action ical product, species-specific criteria could be 68 TEXTBOOK OF CLINICAL TRIALS identified, using the ‘finger-printing’ technique. We need to acquire an updated understanding on Uniformity today should include screening using the effectiveness of the herbal preparations on ‘finger-printing’ techniques. disease entities. That is why we could not be sat- When we consider the other 50% of herbs that isfied with records on efficacy alone, but should are only available from the wilderness, i.e. around start a series of clinical trials to further prove the mountains, highlands or swamps, and cannot be efficacy of the herbs.19 grown on agricultural farms, the insistence on The National Institutes of Health of the United product uniformity becomes even more difficult. States have openly endorsed the approach of Putting together what we have discussed, to accepting traditional methods of healing as safe strictly insist on good clinical practice in clinical measures and then putting them to proper clini- trials for herbal medicine is largely impossible. cal trials.20 The recognition of acupuncture as a We have to accept a compromise. Indeed in the practical effective means of pain control started past 50 years, many attempts have been made at in 1998.21 The subsequent formation of a spe- a proper analytical study of herbal preparations. cial section devoted to research on complemen- The intention was: to put the herb to processes tary/alternative treatment followed. The National of extraction, analyse the important ingredients, Centre for Complementary, Alternative Treat- then try to work out the chemical formulae which ment (NCCAM) was properly formed and given could be responsible for the clinical effects. a substantially large budget. Extraction eliminates the useless components Clinical trials to be discussed within this and concentrates the effective components, which chapter follow the efficacy-driven principle. They not only cuts down the volume of herbs used but are planned strictly according to the principles set also intensifies the biological actions. Knowing out under the modern philosophy of clinical tri- the actual effective ingredients and working als aimed at the production of objective evidence out the chemical formulae would be ideal for of the effectiveness of the methods used. It is modernisation of herbal preparations with the however understood that product uniformity and aim of converting the preparations into proper quality could not be absolutely guaranteed and pharmaceuticals. that although GMP (Good Manufacturing Pro- In spite of the efforts and resources put into cess) could be assured, GCP (Good Clinical Prac- herbal extractions and chemical analyses in the tice) could not be absolutely ensured because of past 50 years, successful examples have not been the lack of guarantee for any herbal preparations. impressive. The results of such efforts certainly In our discussion full reference will be given do not match the resources put in.18 to what is being recommended in China, which The unsatisfactory outcome has initiated a new undoubtedly harbours most activity in Chi- approach. Instead of following the scientific path- nese medicine. way already taken by pharmaceuticals, which has shown too many difficulties rather than promises, HERBAL DRUGS IN CHINA a more practical line has been endorsed. Since In 1999 the National Bureau on Drug Control most, if not all, the herbs have been used for hun- defined new drugs as ‘a manufactured product dreds of years, there should be sufficient amount for medical treatment that is produced for the first of reliability on the safety and efficacy of the time or an old product reproduced with different herbs. The safety and efficacy of the herbs are formulation and different indications’. already well documented, but their practical util- New drugs are divided into five categories: isation in specific clinical circumstances needs to be further established. The traditional use of I. Group 1 the herbs had been focused on symptomatic con- Artificial derivatives from Chinese herbs trol. Nowadays, the aim of clinical management Newly discovered Chinese herbs and deriva- is directed towards curing of a disease entity. tives COMPLEMENTARY MEDICINE 69
Extracts from Chinese herbs and derivatives laboratory research and clinical trials. When Extracts from decortions of herbs studies on the pharmacology, pharmacodynamics II. Group 2 and pharmacokinetics are carried out after the Herbal injections clinical trials, positive values, in support of the Herbal preparations processed inside animals clinical observations, might not be impressive. Extracts from complex decortions The possible explanation to this observation III. Group 3 may lie in the fact that the clinical consump- New preparations of decortions tion of the herbal preparations involves multiple, Combined herbal and chemical preparations complex in vivo biological interactions, whereas Imported herbal preparations laboratory tests consist of only simple unidirec- IV. Group 4 tional biological interactions. Converted formulary Cultivated herbal and domestic animal prepa- HOW DO CONCEPTS OF TRADITIONAL rations HEALING AFFECT CLINICAL TRIALS ON V. Group 5 CHINESE MEDICINE? Herbal preparations with extended uses Earlier in this chapter, the author has already STAGES RECOMMENDED FOR HERBAL touched on the unique concepts in Chinese RESEARCH medicine, which are different from modern sci- entific medicine. The application of modern con- The usual four stages are recommended. cepts in the direction of clinical trials leads to the Phase I : Study on the general acceptance of inevitable sacrifice of some of the fundamental the human being after consumption of the herbal principles. Experienced herbal experts, therefore, preparation. Normally Phase I refers to a toxicity might not like to participate. study. The code of practice given under ‘Code The following list includes the important of Practice for the Scrutiny of New Drugs’ in concepts in Chinese medicine practice being China, however, recommends that the general sacrificed: well-being of the individual after consumption be observed.22 The logic of skipping toxicity tests is 1. SymptomandsyndromeIdentification Principle probably based on an assumption that Chinese Following this principle, the herbal expert herbal preparations have been used safely for adjusts details of his treatment according to centuries, therefore a special toxic screening is observations on the day-by-day changes of not necessary. The author has strong reservations symptomatology. He may then use different on this attitude and would recommend that drugs for the same symptoms or use the same toxicity clearance should remain the first phase drug for different symptoms. Proper clinical of clinical trials. trials could only use a uniform choice of Phase II : Study on the safety and efficacy treatment modality. This violates the symptom while working out the effective dosage. and syndrome identification principle. Phase III : Expands on Phase II study, col- 2. Holistic Approach lecting more reliable confirmation on the safety Chinese medicine emphasises holistic care and efficacy. and holistic response, whereas clinical trials Phase IV : Further study on the safety and prefer objective, specific data as endpoints. efficacy after the new drug is put to market. More The inclusion of specific data into herbal observations on adverse effects are expected. research probably does not invite objection It has been pointed out that, as far as herbal from the herbal expert, as long as general data medicine is concerned, it is not unusual to like different aspects of well-being, i.e. quality find that correlation does not exist between of life, are included. 70 TEXTBOOK OF CLINICAL TRIALS
3. Response to Pathological Processes 6. Establish unique outcome studies. Chinese medicine emphasises the responses 7. Establish unique quality of life assessments. of healthy organs to diseases. The ability of 8. Insist on using modern statistics. the healthy organs to respond to pathological changes ensures that the individual would be able to better resist adversities. Modern ADVERSE EFFECTS clinical trials aim mostly at diseased organs. 4. Old System of Clinical Observation Historically, great herbal masters in China in Herbal experts utilise a system of clinical the ancient days did produce records on adverse observations which today might be considered effects and toxic problems of some herbs. As obsolete and over-subjective. This system of early as the Han dynasty (second century) documents were produced on herbs that need clinical signs includes tongue observation, 25 pulse detection and a collection of subjective to be utilised with care or extreme care. This 23 tradition was followed closely in the subsequent feelings. Modern clinical trials insist on 26 objective data that could be monitored. We centuries. therefore have to either develop means to More reports were available on methods and means with which toxicities and adverse effects objectively assess the subjective signs in the 27 tongue and the pulse or we sacrifice the whole could be reduced. system of observations. Herbal experts might In spite of the good past experience, the preva- lent belief is that Chinese medicinal herbs are not appreciate either choice. safe. On the other hand, more and more reports 5. Strong Tradition appeared on adverse effects and toxicities, and Herbal experts place genuine confidence on non-users of herbs tend to exaggerate the reports. anecdotal observations and experience of sin- When new preparations come into the market, gle patients. Insisting on the need to investi- the innovative processes of extraction and/or gate collective observations and condemning production might have produced or initiated new single-case experience would not be wel- possibilities of adverse effects or toxicity. This comed by herbal experts. This conceptual experience is already well recorded in a number difference directly affects the participation of modernised preparations, particularly those for and cooperation of the traditional and mod- injection.28 Among the adverse effects, allergic ern experts. reactions are commonest. To date, standard instructions on clinical trials While thoroughly recognising the unique nature for Chinese medicine define adverse drug reac- of Chinese medicine and having pointed out the tion in exactly the same way as modern scientific lack of harmony between the old tradition and clinical trials, and explanations of the reactions modern science, one realises that the current have been identically identified.29 compromise adopted in China is to insist on a Categories of adverse reactions include the modern scientific approach as far as possible. following: Hence in standard textbooks, the following are 24 advocated as standard instructions for clinical 1. Reactions to herbs trials: Reactions are defined as harmful and unex- pected effects while the standard dosages are 1. Use the principles of randomisation, blinding used in certain drug trials. It is specially and repetition. pointed out that for Chinese medicine, the 2. Adopt good protocols for clinical trials. harmful reactions could be due to the qual- 3. Avoid bias at all cost. ity of the herb and poor choice of indica- 4. Eliminate chance factors. tion. These reactions do not include aller- 5. Establish new standards of clinical assessment. gic responses. COMPLEMENTARY MEDICINE 71
2. Dosage-related adverse effects for Chinese medicine trials is that of Naranjo.31 Using an unnecessarily high dose could induce Naranjo’s system of grading adverse effects excessive effects, side effects or even toxic according to fact-finding results is shown in effects. Secondary effects like electrolyte Table 5.2. Overall assessment: imbalance might also be observed. 3. Dosage-unrelated adverse effects ADR confirmed ≥9 These adverse effects could be the result ADR likely 5–8 of unfavourable preparation, contaminants in ADR possible 1–4 the herbs, sensitivity of the consumer, aller- ADR unlikely ≤0 gic reactions or specific inductive effects of the herb. Detection and recording of adverse effects 4. Drug interactions should bear different emphases at different phases Classically, records are available in old Chi- of the trial, e.g. Phase I trial aims at detection of nese medicinal literature on combined effects adverse effects in relation to dosage, Phase II and of herbs, their facilitatory and antagonistic III collect details, whereas Phase IV is concerned effects. Nowadays, not only drug interactions mainly with marketed drugs. between herbs are important, but possible Whatever is the situation, detection of adverse interactions between herbs and commonly effects should include both clinical observations used pharmaceutical preparations are becom- and laboratory data, and detection should be ing issues of great concern since users of followed with follow-up observations. The sum- herbal preparations are greatly increasing. In mation of observations should be thoroughly the area of anaesthesia, drug interactions analysed so that explanations of the adverse between herbs and modern medicine could effects may eventually be worked out. induce life-threatening reactions. Table 5.1 illustrates some studies currently done on this REPORTING OF ADVERSE EFFECTS issue.30 5. Delayed adverse effects It is currently required in China that adverse Adverse effects of delayed nature include effects should be reported to the relevant mon- induction to cancer formation, foetal abnor- itoring body as soon as possible. Once a drug is malities and even blockage of bacterial sensi- marketed, adverse effects should be continuously tivities. reported to the National Control Bureau, within 6. Drug dependence the first five years. There might be suspicions that herbal prepa- Adverse effects detected at the post-market rations might lead to drug dependence. Apart Phase IV might be particularly important for Chi- from a few opium-related herbs, Chinese herbs nese medicine trials. Since herbal preparations in fact are well known to be non-addictive do not have clear, definite information about the because of their gross lack of specificities. effective contents of the herbs, bias and chances might be more likely than other trials on simpler From the above account it might appear drugs at the early phases. The large trial popu- obvious that adverse effects in clinical trials lation during Phase IV gives a better chance of using Chinese medicine in fact follow closely the elimination of bias and allows a better oppor- experience encountered in other drug trials. tunity of objective detection of adverse effects. As far as the grading of adverse effects is During the Phase IV trial, the following aspects concerned, it would be appropriate to categorise would deserve particular attention: the effects as mild, moderate and severe. With regard to the overall assessment of 1. Actual danger level of the adverse effects. adverse effects, a convenient recommendation Degree of danger of course depends on the 72 TEXTBOOK OF CLINICAL TRIALS
Table 5.1. Examples of Herb-Drug Interaction Herb Drug Interaction Mechanism
Radix Salviae Warfarin Increased INR Danshen decreases Miltiorrhizae Prolonged PT/PTT elimination of Warfarin (Danshen) in rats Radix Angelicae Warfarin Increased INR and Danggui contains Sinensis (Danggui) widespread bruising coumarins Ginseng (Radix Ginseng) Alcohol Increased alcohol Ginseng decreases the clearance activity of alcohol dehydrogenase and aldehyde dehydrogenase in mice Garlic Warfarin Increased INR Post-operative bleeding and spontaneous spinal epidural haemorrhage Herbal ephedrae (Ma Pargyline, Isoniazid, Headache, nausea, Pargyline, Isoniazid, and Huang) Furazolidone vomiting, bellyache, Furazolidone interfere blood pressure with the inactivation of increase noradrenalin and dopamine; ephedrine in herbal ephedrine can promote the release of noradrenalin and dopamine Ginkgo Biloba Aspirin Spontaneous hyphema Ginkgolides are potent inhibitors of (PAF) Cornu cervi adrenomimetic Strengthens the effect of Natural MAOIs in Cornu pantotrichum Fructus increasing blood cervi pantotrichum, crataegi pressure Fructus crataegi and Radix polygoni multiflori inhibited the metabolism of adrenomimetic, levodopa and opium Radix polygoni multiflori Levodopa Increased blood pressure and heart rate Opium Central excitation Bitter melon Chlorpropamide Decreased urea glucose Bitter melon decreased the concentration of blood glucose Liquorice Oral contraceptives Hypertension, oedema, Oral contraceptive may hypokalaemia increase sensitivity to glycyrrhizin acid St. John’s wort Warfarin Decreased INR Decreased the activity of Warfarin Cyclosporin Decreased concentration in serum COMPLEMENTARY MEDICINE 73
Table 5.1. (continued) Herb Drug Interaction Mechanism
Radix Isatidis Trimethoprin (TMP) Significantly increase (Banlangen) anti-inflammation effect Liu Shen pill Digoxin Frequent ventricular premature beat Tamarind Aspirin Increased the bioavailability of aspirin Yohimbine Tricyclic antidepressants Hypertension
Note : ACE angiotension-converting enzyme INR international normalised ratio PT prothrombin time PTT partial thromboplastin time PAF platelet-activating factor AUC area under the concentration/time curve MAOIs monoamine oxidase inhibitors
Source: Reproduced from De Smet and D’Arcy,30 with permission from Springer-Verlag.
Table 5.2. Questions to be asked about adverse Drug Reactions Yes No Not clear
1. Are there decisive records about the ADR? +10 0 2. Are the ADRs found after consumption of +2 −10 other drugs? 3. Are the ADRs improved after consumption +10 0 of antidote? 4. Are there repeated ADRs or repeated +2 −10 administration? 5. Are there other predisposing factors? −1 +20 6. Are there ADRs after placebo? −1 +10 7. Has the blood level of drug giving ADRs +10 0 been investigated? 8. Do the ADRs correlate to dosages? +10 0 9. Is there past history? +10 0 10. Is there objective proof +10 0
incidence of occurrence. The requirement for of comparisons before results can be instruc- treatment and the financial implications are tive. Case reports might still be useful, but also important. might function as special warning signals to 2. More thorough studies at Phase IV should be calls for more serious studies.32 considered according to epidemiological prin- ciples. Randomised controlled trials should QUALITY OF LIFE be insisted on. Cohort studies might be con- venient and useful, but there need to be While clinical trials aim at a thorough scien- markedly obvious differences between series tific understanding of the effectiveness of specific 74 TEXTBOOK OF CLINICAL TRIALS forms of treatment, endpoints of measurement rehabilitation underway. Different specialties and are set to give objective standards of evaluation. special areas of concern have created charts of Primary endpoints are unique, focused, specific their own and all these provide valuable infor- criteria which indicate the situation of the target mation when equivalent studies come up. Usu- against which the trial is directed. Changes of ally they are adopted right away or after vali- primary endpoints illustrate the efficacy directly. dation. Hence there are charts already developed Secondary endpoints are supplementary crite- for children and the elderly, and different med- ria created to support observations on changes ical specialties and subspecialties likewise have and efficacy. Secondary endpoints become more created their own charts. Just to mention a few, important when, predictably, primary endpoints special QoL charts are available for the men- do not give clear-cut, impressive results. Sec- tally ill, cardiovascular diseases, rheumatologi- ondary endpoints become more important when cal disease, respiratory problems, gynaecological primary endpoints are expected to change slowly problems and special infections.36 QoL charts for and would be particularly important when chronic Chinese medicinal studies need to be developed. problems are being faced. Since Chinese medicine, under most circum- stances, does not operate via a direct confronta- WHAT ARE THE RECOMMENDATIONS FOR tion route but rather acts indirectly to support CLINICAL TRIALS OF CHINESE MEDICINE? the healthy organs and helps to maintain vitality and prevent functional deterioration, critical and The simplest thing to do is to make reference to detailed assessment of the secondary endpoints is the available charts in whichever clinical trial is therefore of utmost importance. being conducted and think about amendments to Quality of life (QoL) is an important aspect make them even more suitable. on the assessment of care given to the chron- ically ill. QoL often measures the competency ARE THERE UNIQUE FEATURES THAT NEED of the care and the ethical standard of the TO BE OBSERVED? society in mental disorders and other disor- ders that demonstrate strong social orientations. There are features related to health which are Not infrequently, using technical endpoints as derived from the philosophy of Chinese medicine results of clinical trials a reasonable outcome ever since its initial development. Chinese people is observed, and yet patients might not be sat- in all walks of life are influenced by this isfied with their QoL. QoL is therefore multi- philosophy without being aware of it at all stages focal: it differs between developed and under- of their life. The belief that health depends on developed areas, it also differs under differ- a harmony between contrasting forces prompts ent cultural circles.33 Different countries and the individual to feel either ‘hot or cold’, regions therefore try to develop their own data ‘light’ or ‘heavy’, ‘sick inside’ or ‘sick outside’. to be included within their own studies on After treatment, the feeling might remain, might QoL.34 Meanwhile global, generally acceptable reverse or might get balanced. The feeling is QoL charts are also being planned, examined and subjective, but in any clinical trial including the validated.35 data of QoL, could one ignore the outcome of the Before an acceptable general data chart is philosophical guideline responsible for the whole ready, one has to accept the achievements already system of healing art? revealed in different fields. Generally speaking, Henceforth, it is obvious the QoL studies QoL data sheets take in information about phys- are particularly important for clinical trials of iological well-being, psychological well-being, Chinese medicine and research should be done social well-being and the individual’s subjec- on special inclusions of data which are unique tive feeling on the treatment received and the for Chinese medicine. COMPLEMENTARY MEDICINE 75
EXAMPLES OF CLINICAL TRIALS ON CHINESE MEDICINE parallel study. Patients will be randomised To give more solid information about clinical to one of the four treatment groups and trials on Chinese medicine being done in Hong treated for a duration of 6 months. Kong, the following paragraphs are devoted to summaries of such trials. Study Population: A minimum of 85 hepatitis B patients will be enrolled, 25 Synopsis I subjects per treatment group, 10 subjects in the control group, a total of four groups.
Name of Study Medication: Phyllanthus Definition of Endpoints: SP. Compound • The primary safety endpoint is tolerabil- Title of Study: A Prospective Randomised, ity. Double-Blind, Placebo-Controlled, Parallel • Tolerability failure is defined as a per- Study to Evaluate the Effect of Phyllanthus manent discontinuation of Phyllanthus SP. Compound in the Treatment of Chronic PLUS as the result of an adverse event. Hepatitus B Virus Infection. • The primary endpoint is a reduction in HBV DNA level from the baseline. Study Centre: Single-centre • The secondary endpoint is HbeAg nega- Objective: tive, anti-Hbe positive and a decrease in ALT level from baseline. Primary Study Regimen: Subjects will be randomly • To evaluate the efficacy of normalisa- and alternatively assigned to receive Phyl- tion of liver enzyme, seroconversion of lanthus PLUS or placebo for 6 months HbeAg and disappearance of HBV DNA prospective parallel study. in serum. • To evaluate the safety of Phyllanthus SP. Duration of Treatment: 6 months Compound in patients with hepatitis B. Statistical Methods: Efficacy: Summary Secondary statistics for the change of HBV DNA, • Proportion of patients with end-of-treat- HbsAG, HbeAg and ALT from baseline ment HbeAg seroconversion (HbeAg will be generated and provided for each to anti-Hbe, normalisation of ALT and treatment group. disappearance of HBV DNA at the end of treatment). Safety • Proportion of patients with HbeAg to anti-Hbe. • The incidence of adverse events and • Proportion of patients with sustained laboratory toxicity will be summarised normalisation of ALT. by treatment group and severity. Change • Proportion of patients with undetectable from baseline in vital signs will be HBV DNA. summarised by treatment group. • Group differences with an error proba- Design: A single-centre, prospective ran- bility of less than 5% (p < 0.05) will be domised, double-blind, placebo-controlled, considered statistically significant. 76 TEXTBOOK OF CLINICAL TRIALS
• The statistical analyses will be made a permanent discontinuation of Danggui with SPSS for Windows 10.0. Buxue Tang as the result of an adverse event. • The primary efficacy endpoint is the change in severity and frequency of hot Synopsis II flushes and night sweats. • The secondary efficacy endpoint is the Name of Study TCM: Danggui Buxue changes in score for the domains mea- Tang sured in the Menopause Specific Quality of Life Questionnaire. Title of Study: A Randomised, Double- Blind, Comparison Study of the Effect of Study Regimen: Subjects will be randomly Danggui Buxue Tang with Oestradiol on and alternatively assigned to receive Dang- Menopausal Symptoms and Quality of Life gui Buxue Tang or placebo for 6 months. in Hong Kong Chinese Women. Duration of Treatment: 6 months treat- Study Centre: Single-centre ment period and 18-month follow-up.
Objective: Statistical Methods:
Primary • Data will be processed to give group • To compare the effects of Danggui mean values and standard deviations Buxue Tang with Oestradiol on meno- where appropriate. • pausal symptoms of hot flushes and Mann–Whitney U-test will be used to sweating. compare the differences between the two • To evaluate the safety of Danggui Buxue groups of treatment. Group differences Tang in patients with menopausal symp- with an error probability of less than 5% toms. (p < 0.05) will be considered statisti- cally significant. Secondary • The statistical analyses will be made with SPSS 10.0 for Windows. • To evaluate the quality of life of the patients with menopausal symptoms.
Design: A single-centre, randomised, dou- ble-blind and comparison study. Subjects Synopsis III will be randomised to one of the two treat- ment groups and treated for a duration of Name of Study TCM: Danggui Buxue 6 months with follow-up of 18 months. Tang
Study Population: A minimum of 100 Title of Study: A Randomised Comparison patients with menopausal symptoms will be Study of the Effect of Danggui Buxue Tang enrolled, 50 subjects per treatment group. with Tranexamic Acid on Dysfunctional Uterine Bleeding and Quality of Life in Definition of Endpoints: Hong Kong Chinese Women. • The primary safety endpoint is tolera- bility. Tolerability failure is defined as Study Centre: Single-centre COMPLEMENTARY MEDICINE 77
Objective: Danggui Buxue Tang or Tranexamic acid for 6 months treatment and 24 months Primary follow-up. • To compare the effects of Danggui Buxue Tang with Tranexamic acid on Duration of Treatment: 6 months treat- menstrual blood loss per month. ment and 24 months follow-up. • To compare the patient’s satisfaction between using Danggui Buxue Tang and Statistical Methods: Tranexamic acid. • To evaluate the safety of Danggui Buxue • Data will be processed to give group Tang in patients with dysfunctional uter- mean values and standard deviations ine bleeding. where appropriate. • All the data in the outcome measure are Secondary continuous variables. Mann–Whitney U- test will be used to compare the differ- • To evaluate the improvement of anaemia. ences between the two groups of treat- • To evaluate the status of iron deficiency. ment. The level of significance will be • To evaluate the unwanted side effects. set at p<0.05. • Design: A single-centre, randomised com- The statistical analyses will be made parison study. Subjects will be randomised with SPSS 10.0 for Windows. to one of the two treatment groups and treated for a duration of 6 months with follow-up of 24 months. Synopsis IV Study Population: A minimum of 125 patients with dysfunctional uterine bleeding Name of Study TCM: Formula A and will be enrolled, 63 subjects in the Danggui Formula B Buxue Tang group and 62 subjects in the Tranexamic acid group. Title of Study: A Randomised, Double- Blind, Placebo-Controlled Study on the Definition of Endpoints: Clinical Effects of Integrated Western Medicine and Traditional Chinese Medicine • The primary safety endpoint is tolera- for Diabetic Foot Ulcer. bility. Tolerability failure is defined as a permanent discontinuation of Danggui Study Centre: Multi-centre Buxue Tang as the result of an adverse event. Objective: • The primary efficacy endpoint is change in frequency and severity of menstrual Primary bleeding. • To evaluate the wound healing effect of • The secondary efficacy endpoint is im- Formula A and Formula B in patients proving anaemia and iron deficiency. with diabetic foot ulcer. • To evaluate the safety of Formula A Study Regimen: Subjects will be ran- and Formula B in patients with diabetic domly and alternatively assigned to receive foot ulcer. 78 TEXTBOOK OF CLINICAL TRIALS
Synopsis V Secondary Name of Study TCM: Relieve Wheezing • To evaluate the effect of control of the Tablet local infection. Title of Study: A Randomised, Double- Design: A multi-centre, randomised, dou- Blind, Placebo-Controlled Parallel Study of ble-blind, placebo-controlled study. Sub- the Effect of Relieve Wheezing Tablet in jects will be randomised to one of the two the Treatment of Childhood Asthma. treatment groups and treated for a duration of 6 months. Study Centre: Single-centre
Study Population: A minimum of 80 Objective: diabetic foot ulcer patients will be enrolled, 40 subjects per treatment group. Primary • To evaluate the medication score, includ- Definition of Endpoints: ing daily use of inhaled steroids. • To evaluate the symptom score, includ- • The primary safety endpoint is tolerabil- ing cough on daytime and nighttime, ity. Tolerability failure is defined as a wheeze/chest tightness on daytime and permanent discontinuation of formula A nighttime, degree of shortness of breath and Formula B as the result of an adverse on exertion. event. • The primary efficacy endpoint is diabetic Secondary foot ulcer healing and to avoid leg • amputation. To evaluate the spirometry lung function • result. The secondary efficacy endpoint is the • control of local infection. To evaluate the breakthrough attacks requiring medical attention from A/E doctors, family physicians on hospital- Study Regimen: Subjects will be randomly isation. and alternatively assigned to receive For- • To evaluate the degree of skin allergy. mula A and Formula B or placebo of a • To evaluate the changes in peripheral 6-month prospective parallel study. blood and Eosinophilic Cationic Pro- tein (ECP). Duration of Treatment: 6 months Design: A single-centre, randomised, dou- Statistical Methods: ble-blind, placebo-controlled, parallel study. Subjects will be randomised to one of the • Results will be presented as the mean ± two treatment groups and treated for a SE per group. duration of 6 months. • Group differences with an error proba- bility of less than 5% (p < 0.05) will be Study Population: A minimum of 80 considered statistically significant. patients with moderate to severe perennial • The statistical analysis will be made with asthma will be enrolled, 40 subjects per SPSS 10.0 for Windows. treatment group. COMPLEMENTARY MEDICINE 79
Definition of Endpoints: Secondary • • The primary safety endpoint is tolera- To evaluate the lipid and homocysteine- bility. Tolerability failure is defined as lowering effect of Danshen and Radix a permanent discontinuation of Relieve Puerariae Compound. Wheezing Tablet as the result of an adverse event. Design: A single-centre, prospective ran- • The primary efficacy endpoint is a domised, double-blind, placebo-controlled, change in improving the symptoms of parallel study. Patients will be randomised asthmatic children and use of inhaled to one of the two treatment groups and steroids. receive Danshen and Radix Puerariae Com- • The secondary efficacy endpoint is im- pound or placebo for a duration of 24 provement of lung function. weeks.
Study Regimen: Subjects will be ran- Study Population: A total of 100 patients domly and alternatively assigned to receive with Coronary Artery Disease (CAD) will Relieve Wheezing Tablet or placebo for be enrolled, 50 subjects treated with Dan- 6 months. shen and Radix Pruerariae Compound and 50 treated with placebo.
Definition of Endpoints:
Synopsis VI • The primary safety endpoint is tolerabil- ity. Name of Study Medication: Danshen and • Tolerability failure is defined as a per- Radix Puerariae Compound manent discontinuation of Danshen and Radix Puerariae Compound as the result Title of Study: A Prospective Randomised, of an adverse event. Double-Blind, Placebo-Controlled, Parallel • The primary endpoint is improving car- Study to Evaluate the Effect of a Herbal diovascular function (endothelial func- Preparation with Compound Formula of tion and carotid intima-medial thickness) Danshen and Radix Puerariae as Cardio- from the baseline. vascular Tonic in Cardiac Patients. • The secondary endpoint is decrease of Study Centre: Single-centre plasma lipid and homocysteine lev- els. Objective: Study Regimen: Subjects will be randomly Primary assigned to receive Danshen and Radix • To evaluate the safety of Danshen and Puerariae Compound (TCM) or placebo Radix Puerariae Compound as adjunc- for 24 weeks in a prospective parallel tive therapy in patients with coronary study. artery disease. • To evaluate the efficacy of treatment Duration of Treatment: 24 weeks and secondary prevention of cardiovas- Duration of Project: 30 months cular diseases. 80 TEXTBOOK OF CLINICAL TRIALS
surface, to varying depths of soft tissue, then Statistical Analyses allowing the needles to stay for some minutes. While the needles stay inside the soft tissue, • The statistical significance of changes the puncturist may give regular or occasional between TCM and placebo groups will rotary movements on the nail. In recent years, be assessed by one-way analysis of acupuncturists have applied direct electrical cur- variance. rent stimulation, so as to unify the stimulations, • Group differences with an error proba- widen the effects and save manpower. Acupunc- bility of less than 5% (p < 0.05) were ture is an invasive process directly aiming at the considered statistically significant. removal of symptoms. It is easy to imagine then, • The statistical analyses were made with that patients would not agree to participate in a SPSS for Windows 10.0. study where they would not be able to enjoy the puncture treatment and function as recipients of ‘sham’ puncture. Likewise, if there were other placebo punctures which fulfil the requirement ACUPUNCTURE of randomisation and placebo control, very few patents would be willing to participate.37,38 Acupuncture is a practical procedure using a However, studies on placebo acupuncture have special needle to enter special regions of the started and the varieties included the following: human body surface by which symptoms suffered by the patient are removed. Like other aspects of 1. Placebo points–entry points are sites outside Chinese medicine, it aims at symptom control, the acupuncture meridians. not at the treatment of a specific disease entity. 2. Sham puncture–puncture lightly then with- The most popular use must be in the field of draw. pain control. 3. Hiding entry points–while entry points are In 1998, the National Institutes of Health in the hidden, it might be possible to achieve real USA held a consensus conference on the use of placebo effect. Hiding of entry points may be acupuncture for pain control. The conclusion was achieved by puncturing through plastic tubes that acupuncture should be accepted as an effec- or soft plastic blocks. tive means of pain control for musculo-skeletal 4. Camouflage puncture by which a needle is just 21 problems and under other specific situations. taped to the skin. Since then, interest in the use of acupuncture in the United States grew and many clinical studies None of these methods could be endorsed were started. because the requirements for placebo in the strict Of course, acupuncture has been used for the sense are far from being satisfied; most recipi- control of other symptoms. Examples include ents could differentiate right away whether it is nerve damages, allergic conditions like rhinitis, true or false puncture. The conventional appli- asthmatic attacks and general feelings of being cation of acupuncture depends on a subjective ‘unwell’, often labelled as ‘derangement’. feeling of ‘numbness’ felt within the punctured How are clinical trials on acupuncture being area. Puncturing without checking this feeling conducted? Could the clinical trials on acupunc- is not considered appropriate. This requirement ture be put in line with modern epidemiological makes ‘placebo’ puncture impossible. The use requirements? Or would it be even more difficult of electrical stimulation is a means to enhance compared with herbal medicine? We have first the effects in modern situations where there is of all, to look at the procedures involved and the insufficient experience on acupoint identification explanations given to the effects produced. and actual puncturing techniques. It is also con- Acupuncture involves the insertion of thin nee- sidered as a method of modernising acupunc- dles, through specific acupoints on the body ture. When electrical stimulation is used, placebo COMPLEMENTARY MEDICINE 81 becomes absolutely impossible because the elec- the spinal cord, loss of neurological and sec- trical stimulation is always felt. ondary muscular functions becomes permanent. The considerations are further complicated by Acupuncture is widely used under such difficult the theories of acupuncture. There are two accept- situations. Although many reports of impressive able theories–the neurological and the humoral. results are available, it is difficult to realise the The neurological theory observes that since some real value. Scientific data coming from well- of the meridians and most of the acupunc- planned cohort studies for the observation of ture points are related to the peripheral nerves, functional restoration is still difficult to inter- stimulation of these points leading to physio- pret, since the damage could not be uniform and logical effects could be working via neurolog- the factors affecting the different aspects of reha- ical pathways, possibly through proprioceptive bilitation and functional return are multiple and receptors.39 The humoral theory assumes that complicated. needle stimulation produces humoral (hormonal) We are therefore still relying on careful reactions, manifested as serological appearance case studies. However, one does not need to of functional factors which possess pain control get upset with the obvious limitations. After effects and other regulatory functions.40 all, in the field of rehabilitation, experience in Whichever theory is at work, it specifies the last three decades has already shown that that the tiny area of puncture is producing qualified, broad, trustworthy clinical trials are chain reactions, either directly or indirectly. An not possible.41 Although meta-analysis has ruled apparently harmless, non-productive action on out the absolute scientific justification for all the skin and soft tissue, imitating acupuncture, the rehabilitation attempts like different forms might trigger off similar effects and would be far of physiotherapy, massage, bracing and even from being a placebo. invasive means like injections and surgery, we Therefore standard epidemiological planning could still rely on them because we have to for clinical trials in Chinese medicine including release our patients from suffering. We are all acupuncture would be very difficult, if at all aware of the fact that we are not certain which possible. Randomisation would not be acceptable patient is the best candidate to receive which 42 to patients, whereas in situations of acupuncture, treatment. if sham puncture is insisted on, it is both unacceptable to patients and falls short of the CONCLUSION placebo requirements. Carefully planned cohort studies aiming to Complementary medicine does not have any compare the effects of different means of pain history of modern scientific development. It control and other treatment expectations are built up its knowledge relying on observations therefore the only reasonable means to objec- and experience. Now that we try to make tively look at the clinical effects of acupuncture. use of this traditional stream of medicine in Many cohort studies of this nature are being done a scientific world, we need to explain why it for the study of back pain, neck pain, arthritis of works in the area of our concern. Very often, the knee, etc. The effects of puncture were com- these areas are not well served by scientific pared with conventional techniques using phys- medicine. This makes the scientific explanations iotherapy and other means. even more important. Another common application for acupunc- The way to go about giving scientific expla- ture is the restoration of nerve function. Dam- nations to healing processes involves the appli- aged neurological tissues suffer from a real lack cation of methodologies that are well known and of regeneration. Peripheral damage feasible for accepted by all clinical scientists. The standard repair carries reasonable promise. When cell way to start a scientific approach to clinical bodies are involved, either intra-cranially or in trials using traditional Chinese medicine would 82 TEXTBOOK OF CLINICAL TRIALS just be an application of the same methodolo- and patterns of use. New Engl J Med (1998) gies. However this approach is not feasible and 328(4): 246–52. would probably remain unfeasible in spite of 11. Fair WR. The role of complementary medicine in urology. JUrol(1999) 162: 411–20. enthusiasms. We are barred from a smooth appli- 12. Leung PC. Traditional Chinese Medicine in China. cation of the scientific methodology, basically Seminar on Health Care in Modern China, because of the different philosophy behind the Yale–China Association (2001). traditional Chinese way of healing. Moreover, the 13. Leung PC. Development of traditional Chinese lack of knowledge about the exact chemistry of medicine in Hong Kong and its implications for orthopaedic surgery. Hong Kong J Orthopaed surg the active component of the herbal remedy when (2002) 6(1): 1–5. herbal drug trials are being carried out further 14. Ho T. An anthropologist’s view on traditional jeopardises the validity of the clinical trials. chinese medicine. In: A Practical Guide to Chinese In spite of the essential difficulties, efficacy- Medicine. Singapore: World Scientific (2002). 15. Chalmers I. The Cochrane Collaboration: prepar- driven trials are still carried out, utilising prin- ing, maintaining and disseminating systemic re- ciples of evidence-based medicine. As long as views of the effects of health care. Ann NY Acad the scientific gap is successfully narrowed, the Sci (1993) 203: 166–72. practical use of complementary medicine will 16. Tyler VE. Herbs of Choice: The Therapeutic use become safer, more logical and would deserve of Phytomedicinals. New York: Pharmaceutical Product Press (1994). wider applications. At the same time, workers 17. Ernst E. Harmless herbs? A review of the recent on complementary medicine should workout a literature. Am J Med (1996) 104: 170–8. unique, relevant system of assessment for the 18. Bisset Ng. Herbal Drugs and Phytopharmaceuti- clinical effects. cals. Boca Raton: Medpharm Scientific Publisher (1982). 19. Leung PC. Clinical trials in Chinese medicine – REFERENCES the efficacy driven approach. Hospital Authority Convention, Hong Kong (2001). 20. NIH. National Centre for Complementary and 1. Bensky D, Gamble A. Chinese Herbal Medicine. Alternative Medicine Five Year Strategic Plan Materia Medica. Seattle: Eastland Press (1993). 2001–2005. NIH Document (2000). 2. Tang W, Eisenberg G. Chinese Drugs of Plant 21. NIH. Acupuncture. NIH Consensus Statement Origin: Chemistry, Pharmacology and Use in (1997) 15(5): 1–34. Traditional and Modern Medicine.NewYork: 22. Wang PY. New Chinese Medicine: Research, Pro- Springer-Verlag (1992). duction and Registration. China: Chinese Medi- 3. Su WT. History of Drugs. Beijing: United Medical cine Press (1995). University Press (1992). 23. Chen KC. Research in Chinese Medicine. Beijing: 4. McGuire MB. Ritual Healing in Suburban Amer- China Health Publisher (1996). ica. Newbrunswick: Rutgers University Press 24. Lai SL, ed., Clinical Trials of Traditional Chi- (1988). nese Materia Medica, Chapter 2, 3, 4. Guang- 5. Kong YC. Formulation’s from South-West Asia. dong: People’s Publishing House (2001). Beijing: Chinese University Press (1998). 25. Chang CK (Han Dynasty). Discussions of Fever. 6. Eisenberg DM, Roger DB. Trends in alternative Classics. medicine use in the USA 1990–1997. JAMA 26. Li ZC (Ming Dynasty). Materia Medica. Classics. (1998) 280(18): 1569–75. 27. Suen TM (Tang Dynasty). Important Prescrip- 7. Lai SL. Clinical Trials of Traditional Chinese tions. Classics. Materia Medica, Chapter 1. Guangdong: People’s 28. National Bureau on Drug Supervision. Regulations Publishing House (2000). on Clinical Trials Using Drugs and Herbs. Bureau 8. Leung PC. Editorial – Seminar series on evi- Publication (1999). dence-based alternative medicine. Hong Kong 29. Suen TY, Wang SF, Wang KL. Adverse Effects of Med J (2001) 7(4): 332–4. Drug Treatment. Beijing: People’s Health Press 9. Kaptchuk TJ, Eisenberg DM. The persuasive ap- (1998). peal of alternative medicine Ann Int Med (1998) 30. De Smet P, D’Arcy PF, eds. Drug interactions 129(12): 1061–5. with herbal and other non-toxic remedies. In: 10. Eisenberg DM, Kessler RC, Foster C. Unconven- Mechanisms of Drug Interactions. Berlin: Sprin- tional medicine in the USA – prevalence, costs ger-Verlag, (1996). COMPLEMENTARY MEDICINE 83
31. Sackett DL. Clinical Epidemiology, 2nd edn. 37. Park J, White AD, Lee H, Ernst ED. Develop- Boston: Little, Brown & Co. (1991) 297–9. ment of a new sham needle. Acupunct Med (1999) 32. Uppsala Monitoring Centre, WHO. Safety moni- 17(2): 347–52. toring of medicinal products: guidelines for set- 38. Streiberger K, Kleinhenz J. Introducing a placebo ting up and running a pharmacovigilance centre needle into acupuncture research. The Lancet (2000). (1998) 352: 364–5. 33. Jaeschke R, Singer J. Measurement of health sta- 39. Godfrey CM, Morgan P. A controlled trial of the tus. Ascertaining the minimally important differ- theory of acupuncture in musculoskeletal pain. J ence. Contr Clin Trials (1989) 10: 407–15. Rheumatol (1978) 5: 121–4. 34. Wyrwick KW, Nienaber NA. Linking clinical rel- 40. Lewith GT, Machin D. On the evaluation of the evance and statistical significance in health related clinical effects of acupuncture. Pain (1983) 16: quality of life. Med Care (1999) 37: 469–78. 111–27. 35. Stewart AL, Greenfield GS, Hayo RD. Functional 41. Nachemson E. Causes and treatment of low back status and well-being of patients with chronic pain. American Academy of Orthopaedic Surgery conditions. JAMA (1989) 262: 907–13. Lecture Series (1989). 36. Keininger DL. Why develop IQOD? International 42. Spitzer WO, Leblanc FE. Scientific approach to Health Related Quality of Life Outcome Database. the assessment and management of activity-related QoL News (1999) Sept/Dec (23): 2. spinal disorders. Spine (1987) 12(Suppl. 1): 51–9. CANCER
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 6 Breast Cancer
DONALD A. BERRY, TERRY L. SMITH AND AMAN U. BUZDAR Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, Houston, TX 77030, USA
INTRODUCTION in situ (DCIS) increased by about sixfold, from about 5 cases per 100 000 women in 1980 to more More than 200 000 women in the United States than 30 per 100 000 in 1998. are diagnosed annually with breast cancer. About Despite the increasing incidence of breast 40 000 women die from the disease each year. cancer among US women since the early 1980s, Among women, it is the most common malig- and perhaps indirectly because of this increase, nancy, and is exceeded only by lung cancer as the the annual age-adjusted rate of breast cancer leading cause of cancer death. Although the risk mortality has decreased by almost 2% per year of breast cancer is substantially higher in older in the 1990s. Researchers generally attribute women, many cases occur in young women. Of this improvement in breast cancer survival to cases diagnosed in the US in 1998, 5% occurred increased use of screening mammography and to in women under the age of 35, 30% in women improvements in the treatment of breast cancer. aged 35 to 49, 31% in women aged 50 to 64, and Efforts are underway to better delineate the 33%inwomenaged65orolder.1 For US women, relative impacts of factors influencing breast the lifetime risk of developing breast cancer is cancer survival (see CISNET2 for additional about 13%. About 0.1% of US women carry an information). inherited mutation of a breast/ovarian cancer sus- An important development for breast cancer ceptibility gene, BRCA1 or BRCA2, and so have research in the 1980s and 1990s was not directly a lifetime risk in excess of 50%. related to science. These years saw the for- From 1940 to the early 1980s, breast cancer mation of strong advocacy groups that worked incidence in the US increased by a fraction to promote research in breast cancer. Federal of a percent per year when adjusted for age. funding has increased more than sixfold since Chiefly because of the widespread dissemination 1990, and grass-roots action has resulted in of screening mammography beginning in the an unprecedented programme of breast cancer early 1980s, invasive breast cancer incidence has research funding administered by the Depart- increased by 3–4% per year into the 1990s. Over ment of Defense. In addition, patient advocates the same period, the rate of ductal carcinoma have become highly educated about research
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 88 TEXTBOOK OF CLINICAL TRIALS issues and many serve regularly alongside profes- 0 through Stage IV disease are 21%, 42%, 29%, sional scientists on various governmental boards 5% and 4%, respectively.1 An additional tumour guiding the direction of research expenditures classification method based on histopathologic and treatment recommendations. Patient advo- examination has limited discrimination ability cates also serve on cooperative group committees because 70–80% of tumours are of a single type: that plan clinical trials in breast cancer, institu- infiltrating ductal carcinoma. tional review boards, and data safety monitor- ing boards. PROGNOSIS Advocacy groups have worked to increase the number of women who participate in clinical tri- Breast cancer is heterogeneous. Many breast als. The Clinical Trial Initiative of the National cancers are slowly growing and their carriers Breast Cancer Coalition Fund (NBCCF) main- survive for many years and die of other causes. tains a registry of clinical trials and urges women Other tumours are very aggressive and may with breast cancer to participate (see NBCCF3). have spread to distant sites by the time the Before a clinical trial can be included in their primary tumour is diagnosed. This heterogeneity registry, experts from the NBCCF ascertain that has implications for research in all phases of the it addresses an important, novel research ques- disease, beginning with screening and diagnostic tion related to breast cancer, and that its design methods through the evaluation of treatments for is scientifically rigorous and employs appropriate advanced disease. and meaningful outcomes. Stage is the most widely recognised deter- minant of patient outcome. Stage IV disease is generally regarded to be incurable, with median STAGING survival in the range of 18 to 24 months, although a small fraction of patients with Stage IV disease Breast cancer is staged using a system developed achieve complete remission following systemic by the American Joint Committee on Cancer, chemotherapy, and survive for many years.6 On and based on the size and other characteristics the other hand, patients with Stage I disease, con- of primary tumour (T), the status of ipsilateral sisting of a small primary tumour and no involved lymph nodes (N), and the presence or absence lymph nodes, have at least a 90% probability of distant metastases (M) (AJCC4 and Singletary of being disease-free after five years. Lymph et al.5). The stage of disease, ranging from 0 node involvement is associated with a worse to IV, is based on combinations of these TNM prognosis, with five-year disease-free rates rang- rankings. ing from 50–75%. Tumour grade, proliferative Stage 0 consists of ductal and lobular carci- activity and menopausal status play relatively noma in situ (DCIS, LCIS), non-invasive and minor roles. possibly non-malignant forms of the disease. Although stage is an important prognostic Stages I to III are invasive stages in which the factor, it is of limited use as a determinant tumour is confined to the breast or its immedi- of treatment outcome. The relative benefits ate vicinity. Higher stage indicates larger primary of treatment are reasonably consistent across tumours or greater locoregional tumour involve- stages – although the absolute benefit can be ment. Patients having evidence of distant metas- much greater for higher stage disease. Much cur- tasis are classified as Stage IV. rent research focuses on factors that may predict Distribution of disease stage at the time clinical benefit from certain treatment approaches of breast cancer diagnosis varies by country, (‘predictive factors’) in contrast to the more con- depending on the health care system’s approach ventional ‘prognostic factors’ which are regarded to diagnosis and reporting. In the US, the approxi- as indicators of general tumour aggressiveness, mate proportions of women diagnosed with Stage irrespective of type of therapy. BREAST CANCER 89
The best-studied predictive factor is oestrogen- breast cancer throughout the first half of the cen- receptor (ER) status, which is an important tury, sometimes combined with radiotherapy. indicator of whether a tumour will respond to When the concept of large-scale randomised hormonal treatment. Tamoxifen and other selec- clinical trials to investigate alternative therapies tive oestrogen-receptor modulators (SERMs) are was proposed in the 1960s, controversy arose highly effective in patients with hormone- among breast cancer researchers as well as in sensitive breast cancer, but they have no other medical fields. In a heated exchange, a benefit in patients whose tumours are ER neg- prominent breast cancer surgeon denounced such ative and progesterone-receptor negative (Early studies as ‘a great leap backward in the treat- Breast Cancer Trialists’ Collaborative Group ment of breast cancer’.17 Despite such opposition, [EBCTCG]7). Patients who benefit from SERMs pioneers in the field persisted in designing tri- may also benefit from aromatase inhibitors.8 als to address important therapeutic questions of HER-2 (also referred to as HER-2/neu, ErbB2, the time, and moreover, were able to persuade c-erbB-2) is a member of the epidermal growth patients to participate in this novel idea of assign- factor receptor family that is overexpressed in ing treatment by randomisation. These early tri- 20% to 40% of breast tumours, and has been als compared various surgical and radiotherapy cited in numerous reports as conveying poor approaches. In a trial of almost 1700 women prognosis.9 Studies in early breast cancer have implemented in 1971, there were no significant suggested that patients with HER-2 positive survival differences between conventional radi- tumours are more likely to benefit from anthra- cal mastectomy, total mastectomy with radiation, cycline therapy.10–12 Trastuzumab, a monoclonal and total mastectomy with removal of axillary antibody against HER-2, is effective in delay- nodes.18,19 Results of this and other trials of the ing progression in Stage IV disease that over- era challenged long-held views of the disease and expresses HER-2,13 and is being evaluated for gradually convinced researchers that their con- its efficacy in treating HER-2-overexpressing pri- cept of breast cancer as a local disease which mary tumours. could best be treated by radical local treatment techniques was incorrect. Rather, breast cancer came to be understood as a systemic disease that HISTORICAL PERSPECTIVE ON CLINICAL could benefit from systemic therapy, and radi- TRIALS IN BREAST CANCER cal local therapies were no longer regarded as essential for prolonging survival. SURGERY AND RADIOTHERAPY CHEMOTHERAPY Scientific understanding of the biology of breast cancer has changed radically in the past 50 years. Cytotoxic agents for treatment of solid tumours Results of large randomised trials have played a were first developed in the 1950s. Breast cancer major role in this transition. From the nineteenth proved to be highly sensitive to several of these, century and up into the 1970s, breast cancer was when used as single agents in small trials. Subse- understood to be a local/regional disease that quently, combinations of these cytotoxic agents spread by direct extension along lymphatic path- were evaluated, one of the earliest being the ways to distant sites. This concept gave rise to the Cooper regimen (cyclophosphamide, methotrex- surgical methods promoted by W.S. Halsted14–16 ate, 5-fluorouracil, vincristine and prednisone).20 around the turn of the twentieth century, i.e., With the understanding of breast cancer as a sys- extensive resection of the breast, regional lym- temic disease and the proven sensitivity of breast phatics, lymph nodes and muscle. This surgi- cancer cells to cytotoxic agents, the stage was set cal technique, known as radical mastectomy, for the rapid development of adjuvant chemother- remained the principal approach to treatment of apy once this concept was introduced in the 90 TEXTBOOK OF CLINICAL TRIALS
1970s. A randomised trial comparing surgery fol- most important advance in treating breast cancer. lowed by combination chemotherapy to surgery Questions remain about the optimum treatment alone demonstrated that disease recurrence could duration even though a trial conducted by the be significantly reduced using this adjuvant ther- National Surgical Adjuvant Breast and Bowel apy approach.21 Project (NSABP) comparing 5 and 10 years of The introduction of doxorubicin for treatment tamoxifen therapy concluded there was little or of breast cancer is illustrative of the series of no advantage to longer therapy.26 clinical trials typically undertaken for the devel- opment of new agents. Small trials conducted HIGH-DOSE CHEMOTHERAPY WITH BONE in solid tumours in the early 1970s established MARROW TRANSPLANT OR STEM safety and dosing, and these were quickly fol- CELL SUPPORT lowed by Phase II trials of the agent in metastatic breast cancer. Subsequently, doxorubicin was An unresolved question in therapy of breast can- evaluated in combination with other agents, and cer that has presented an unusual challenge for randomised trials established that higher response the conduct of clinical trials is that of high- rates could be achieved in metastatic disease with dose chemotherapy supported by autologous bone combinations that included doxorubicin. These marrow transplant or peripheral blood progenitor successes prompted the introduction of various cells. Ten trials addressing the question of high- doxorubicin and other anthracycline-containing dose versus standard-dose chemotherapy have combinations as adjuvant therapy for primary been reported. Two of these were subsequently breast cancer. Known by such acronyms as discredited following an international investiga- ‘FAC’ = ‘CAF’, ‘FEC’, ‘AC’, these combina- tion. Only two of the remaining eight trials tions continue to play a prominent role in the entered more than 200 patients. Financial issues, treatment of breast cancer.22,23 Anthracycline- patient and physician acceptance and competing containing therapies further reduce the risk of treatment strategies have compromised accrual, recurrence and favourably impact survival in and it is unclear if ongoing trials can be com- early breast cancer.24 pleted. The available evidence suggests that high- dose therapy provides little or no benefit for 27,28 HORMONAL THERAPY patients regardless of their disease stage.
Hormonal therapy is a key component of therapy MAMMOGRAPHY when tumours are hormone-receptor positive. Early trials focused on ovarian ablation by Eight large randomised trials conducted since surgery or chemical means. The anti-oestrogen 1963 assessed the value of screening mam- agent tamoxifen was introduced in the 1970s, mography for reducing breast cancer mortal- at a time when there was high regard for the ity. These are of particular interest for the potential of cytotoxic agents, but little interest scrutiny they have undergone in recent years. in hormonal therapies. Early small trials in The preponderance of evidence from the ran- metastatic breast cancer were equivocal and could domised trials indicates a benefit associated with have led to abandoning the agent. However, the screening mammography.29–31 However, a meta- weight of evidence from laboratory studies and analysis concluded that six of the eight trials several small trials pointed to superior efficacy were seriously flawed and the remaining two with prolonged administration in ER positive trials showed insignificant breast-cancer mortal- disease. After a series of large randomised trials, ity differences between the screened and non- tamoxifen is now regarded as standard therapy screened groups.32 The National Cancer Institute for pre- and post-menopausal women with ER recommends screening mammography every 1 positive tumours.25 Tamoxifen may be the single to 2 years for women aged 40 and older, while BREAST CANCER 91 recognising that there are risks associated with Oxford University, serves as a centre for data false-positive results. synthesis rather than actual conduct of clinical trials. Beginning in 1983, this group has col- PREVENTION lected data from virtually all major randomised trials conducted in early breast cancer, published Beginning in the 1990s, coinciding with the or not. Data from more that 200 000 women detection of methods for identifying women at have been analysed, using statistical techniques high risk of breast cancer, the first large-scale for meta-analysis, with results published at the trials were mounted to determine if the incidence end of each five-year analysis cycle, beginning in of breast cancer could be reduced in targeted 1985. These publications have addressed the role high-risk groups. These trials established that of radiation, ovarian ablation, polychemotherapy, breast cancer incidence could be greatly reduced tamoxifen and quality of life; these ‘overview’ by daily doses of tamoxifen.33,34 This reduction articles are frequently cited in support of treat- was due entirely to a lower incidence of ER ment approaches.7,24,25,37–42 The weaknesses of positive tumours with no change in the incidence meta-analysis have been widely discussed in the of ER negative tumours. This suggests that statistical literature, chief among these being the prophylactic tamoxifen will not have as great issue of heterogeneity among the trials being an impact on survival as it does on incidence, combined. For example, the overviewers com- although none of the prevention trials address bine various therapeutic regimens under the sin- survival as an endpoint. An ongoing trial (STAR: gle rubric ‘polychemotherapy’. However, these the Study of Tamoxifen and Raloxifene) will overview reports have allowed researchers to reli- recruit 22 000 women at high risk of breast cancer ably assess moderate-size treatment effects which in order to compare the effects of tamoxifen and could not have been detected in individual trials. another SERM, raloxifene, on the incidence of Treatments causing even moderate reductions in primary breast cancer.35 mortality, if implemented widely among women with breast cancer, could prevent or delay thou- sands of deaths due to the disease. The meta- MAJOR TRIAL GROUPS analysis has also addressed questions of treatment One of the largest cooperative groups conduct- efficacy within subsets, for example the confir- mation of benefit of adjuvant tamoxifen in ER ing trials in breast cancer in the US is the 7 NSABP. Trials from this group are often referred positive pre-menopausal women as well as in to by their ‘B’ numbers, e.g., B-06, which estab- post-menopausal women. lished the equivalence of lumpectomy to total mastectomy.36 Other major cooperative groups conducting clinical trials in breast cancer are the TIME-DEPENDENT HAZARDS Eastern Cooperative Oncology Group (ECOG), Cancer and Leukemia Group B (CALGB), South- In this section we address a methodological west Oncology Group (SWOG), Breast Cancer issue that arises quite generally in survival International Research Group (BCIRG), Euro- analysis. Consider disease-free survival. This is pean Organisation for Research and Treatment of the usual primary outcome measure in evaluating Cancer (EORTC), North Central Cancer Treat- adjuvant therapies, with results presented in ment Group (NCCTG), and the National Cancer the form of Kaplan–Meier survival curves and Institute of Canada (NCIC). compared using statistical tests that take into An important information resource regarding account the entire survival distributions. The the benefits of treatment for early breast cancer simple hazard function, which is in effect the is the Early Breast Cancer Trialists’ Collabo- derivative of the survival curve, can serve as an rative Group (EBCTCG). This group, based at effective graphical aid to understanding treatment 92 TEXTBOOK OF CLINICAL TRIALS and covariate effects. Moreover, it can reveal then the second-year hazard is 10/90 = 11%. important effects that are not apparent in the When calculating hazards from survival plots survival curves, themselves. such as those in Figure 6.1 (which incorporate We will use trial CALGB 8541 as an censored observations), we subtract the current example.43 This trial considered three differ- year’s survival proportion from the previous ent dose schedules of CAF (cyclophosphamide, year’s survival proportion and divide by the doxorubicin and 5-fluorouracil) in node positive previous year’s survival proportion. The resulting breast cancer. The schedules consisted of four yearly values are shown in Figure 6.2. cycles of CAF at 600, 60, 600 mg/m2 (high dose), Some authors like to smooth hazard estimates six cycles at 400, 40, 400 mg/m2 (moderate dose) over time. We prefer to show the raw estimates. or four cycles at 300, 30, 300 mg/m2 (low dose). The reason is that each time period provides a The primary endpoint was disease-free survival, ‘nearly independent’ trial of therapeutic compar- which is shown in Figure 6.1 for the three dose isons. Depending on what assumptions are made groups using Kaplan–Meier plots. Details of the about the underlying survival distribution, these comparison are provided in the original report, trials may not be truly independent, but events and will not be repeated here. that have occurred previously are set aside and Time-to-event curves such as those in Figure a ‘new trial’ is begun. Each time period has 6.1 do not tell the whole story regarding any the potential for confirming observations made benefit of increasing dose and dose intensity. A in other time periods. clearer picture is contained in plots of hazard over A striking observation from Figure 6.2 is that time. The hazard in any particular time period all three hazards decrease over time (after year is the proportion of events occurring during that 2). This is a reflection of the heterogeneity of time period in comparison with the number of breast cancer. The most aggressive tumours recur patients who are at risk at the beginning of the early, yielding the high hazards evident in the period. For example, if there are 100 patients first few years. Once their tumours have recurred, in a group and 10 of these recur in the first patients are removed from the at-risk population. year, then the first-year hazard is 10%. Going The remaining tumours tend to be less aggressive into the second year, only 90 patients are at and so they recur at a lower rate. risk. If another 10 recur in the second year,
Low 1.0 0.15 0.9 High 0.8 Mod 0.7 Mod 0.10 0.6 0.5 Low 0.4 High 0.05 0.3 0.2 0.1 0.0 0.00 0 2 4 6 8 10 12 14 16 18 0 2.5 5 7.5 10 12.5 15 Years Years
Figure 6.1. Disease-free survival proportion for the Figure 6.2. Hazards for the three CAF dose groups of three CAF dose groups of CALGB 8541 CALGB 8541, derived from Figure 6.1 BREAST CANCER 93
As regards treatment-arm effect, the apparent 0.35 benefit of a regimen of high-dose chemotherapy is restricted to the first five years or so. Actually, 0.30 the hazard for a high dose is lower than those 10+ of the other two arms in each of the first six 0.25 years (although it is not much lower in years 0.20 4, 5 and 6, and it is not much lower than that for a moderate-dose regimen in any of the six 0.15 * years). In view of the ‘near independence’ of the 4−9 six time periods, this observation is impressive. 0.10 Another important observation from Figure 6.2 is − that after five years the risks of all three groups 0.05 1 3 come together, with the annual risk of recurrence 0.00 being approximately 5% in all three groups. 0 2.5 5 7.5 10 12.5 15 The reduction in hazard of recurrence for high Years versus low doses is 14% over the 18 years of follow-up (95% confidence interval: 6–22%). Figure 6.3. Hazards for the three categories of This is an average over these years (weighted positive lymph nodes (1–3, 4–9 and 10 or more) over time because of differences in at-risk sample for CALGB 8541. There are few patients at risk in the sizes over time), but since there is no reduction later years, especially in the 10+ group, and for two at all in the later years, the overall reduction is reasons. One is that this was the smallest group to being carried by the early years. Restricting to the start with (174 of the 1550 patients in the trial), and first three years, the reduction is 24% (13–33%). the other is that most recurred early. For example, the asterisk at 13 years indicates a time point at which A benefit of chemotherapy that is restricted to the there were only 24 patients at risk, and three of these first few years is typical in breast cancer trials. An recurred in the 13th year implication is that a hazard reduction seen early in a trial, say one with a median of three years of follow-up, will deteriorate over time. This is no positive nodes. The effects of both the number because the comparison will eventually involve of positive nodes and dose of CAF have elapsed averaging over periods where there is no longer after five years. a treatment benefit. An important aspect of CALGB 8541 is the In the later years, the hazards of about 5% role of tumour HER-2/neu expression and in are very similar to the annual hazard for node particular its interaction with dose of CAF.10 negative breast cancer patients. Interestingly, HER-2/neu assessment was carried out for a convergence to about 5% applies irrespective of subset of 992 patients from the original study. Its the number of positive lymph nodes. Figure 6.3 interaction with dose was shown to be significant shows this effect. It gives hazard plots for three in a multivariate proportional hazards model. categories of positive nodes: 1–3, 4–9 and 10 or But the manner of interaction is easiest to more (for the three dose groups combined). Early understand using hazards. Figure 6.4 shows the in the trial, patients with 10+ positive nodes have effect of dose of CAF separately for patients a very high annual recurrence rate of 20–30%. with HER-2/neu negative tumours (n = 720) and However, after five years or so, the annual hazard HER-2/neu positive tumours (n = 272).HER- is about 5% in all three groups. A patient with 2/neu negatives show no dose effect. The entire a large number of positive nodes who has not benefit of high dose over moderate dose and experienced recurrence in the first five years or high dose over low dose that is observed in so has the same updated prognosis as a patient these patients is concentrated in patients whose with a small number of positive nodes, including tumours are HER-2/neu positive. Moreover, this 94 TEXTBOOK OF CLINICAL TRIALS
0.20 0.20 Low
0.15 Mod 0.15
Low Mod 0.10 0.10 High
High 0.05 0.05
0.00 0.00 0123456789101112 0123456789101112 Year Year
Figure 6.4. Annual disease-free survival hazards for a subset of patients (n = 992) in CALGB for whom expression of HER-2/neu in the patient’s tumour was assessed. Patients in the left-hand panel had tumours that were HER-2/neu negative and the tumours of those in the right-hand panel were HER-2/neu positive benefit occurs through a reduction in hazard in The information available about these hazards each of the first three to four years. Again, is shown in Figure 6.2. For predicting when each year is a separate study and so each of and whether a patient recurs, hazards should be these years provides a separate confirmation of considered one year at a time, and based on the the overall conclusion. The hazard reduction in current year of follow-up. the first three years for high dose as compared with the other two groups combined was 65% among patients whose tumours were HER-2/neu ASSESSING LONG-TERM IMPACTS positive. HER-2/neu overexpression apparently OF THERAPY conveys a poor prognosis for lower doses but not for a high dose – it might even provide a Showing that a cancer therapy is beneficial using favourable prognosis for a high dose. logrank tests or proportional hazards regression Many of the above conclusions would have models or whatever other analysis one uses, does been difficult or impossible without considering not allow for concluding the nature of the benefit. hazards over time. A final comment regarding It may be that some patients are cured of their hazards relates to the common problem of disease; or the therapy may delay the disease’s predicting survival results into the future for progress in some patients; or the effect may be patients already accrued to a trial. Consider a mixture of the two. Deciding among these Figure 6.1. Some patients have as little as possibilities may be possible when all or almost 10 years of follow-up information. As more all events occur in a modest amount of follow- follow-up information becomes available, there up time. In primary breast cancer, a goodly will be no change in these curves prior to the 10- proportion of patients never recur. Therefore, year time point, but they may change subsequent such a decision is difficult or impossible to make. to 10 years. Because the focus is on patients who Figure 6.5 illustrates the difficulty in discrim- have not yet recurred, the way the curves will inating between cure and prolonging survival in change depends on the hazards beyond 10 years. breast cancer trials. Consider a clinical trial that BREAST CANCER 95
1.0 long. Moreover, information beyond 20 years is relatively sparse because earlier events and 0.9 Prolong competing risks (such as cardiovascular disease) 0.8 will have removed patients from the at-risk 0.7 population. To make inferential matters worse, 0.6 there is an enormous array of possible curves that are similar to the two shown in the figure, with 0.5 Cure some having cure rates and others not. Finally, 0.4 the ‘current’ survival distribution assumes that Survival Prob 0.3 Current all breast cancer is fatal (although survival times 0.2 vary). More realistically, some breast cancer (including some invasive as well as in situ breast 0.1 cancer) will never kill the patient. Deciding 0.0 whether a new therapy cures some patients is 0 5 10 15 20 even more difficult if a proportion of patients is Years assumed to have non-fatal disease. This inability to distinguish between curing Figure 6.5. Hypothetical survival curves comparing patients and prolonging survival has further cure and prolonging survival implications in the evaluation of screening and diagnostic methods. It is possible that breast is designed to evaluate a new therapy, one that cancers become lethal or not in their very may improve survival. Suppose further that in the early development, as suggested by studies of population of interest, the current annual rate of tumour markers purported to identify especially breast cancer mortality is 8%. The correspond- aggressive tumours. If this is so, then early ing survival distribution is shown by the dashed detection may not help, and the observed benefits survival curve labelled ‘current’ in Figure 6.5. of therapy, however substantial, may be the result It assumes exponential survival (although the of slowing the progress of the disease rather than same effect holds for any parametric form) and curing it. Such slowing may be beneficial whether so the current median survival is 8.66 years. it comes early or late in the disease. The goal of the new therapy is to improve this by 1/3 to 11.55 years. If this happens then it may be through prolonging every patient’s sur- ADAPTIVE DESIGNS OF CLINICAL TRIALS vival by reducing the annual mortality rate to 6% – the curve labelled ‘prolong’ – or by leav- Adaptive designs have the dual goals of efficient ing the annual rate unchanged (at 8%) for most learning from all relevant results and effective of the patients but curing a fraction of them – the treatment of patients. They are more flexible than curve labelled ‘cure’. To have the same median conventional designs, and have application in all (11.55 years) as ‘prolong’ implies a cure rate phases of drug development. Such designs can of 17.1%. be implemented using Bayesian methodology as In view of the sampling variability present in a means to incorporate new information into the empirical survival information, it is impossible trial design. to discriminate between the ‘prolong’ and ‘cure’ Designs of clinical trials for breast cancer curves shown in Figure 6.5 on the basis of are usually static in the sense that the sample results of even impossibly large clinical trials. size and any prescription for assigning treatment, Indeed, the critical part of the follow-up period including the randomisation of patients, are fixed for this discrimination is 20 years and beyond, in advance. While designs may include stopping and few trials have followed patients for this rules, such as the two-stage Phase II trial design 96 TEXTBOOK OF CLINICAL TRIALS of Simon,44 or the interim comparisons in Phase possibilities are considered in light of the accu- III designs,45,46 the criteria for early stopping mulating information. are very conservative and therefore few trials Adaptive designs also include unbalanced ran- actually stop early. The simplicity of trials domisation, in which the degree of imbalance with static design makes them solid inferential depends on the accumulating data. For example, tools. The sample sizes tend to be large, with arms that give more information about the a straightforward treatment comparison as the hypothesis in question or that are performing objective. Despite their virtues, static trials result better than other arms can be weighted more in slow and unnecessarily costly development of heavily.47 Current (Bayesian) probabilities that new therapeutic agents. each of several doses or agents surpass standard The tradition of drug development is to evalu- or placebo therapy are calculated. These calcu- ate a single drug at a time. Given the fast pace of lations use all information from patients treated current new drug discovery (there are hundreds to date. A new patient is then assigned to of known experimental drugs with potential bene- treatment randomly, with weights proportional fits in breast cancer), these inefficient evaluation to these probabilities. The assignments involve methods are no longer adequate. In addition to some degree of randomisation, but all patients the traditional focus on false-positive and false- are more likely to receive treatments that are per- negative errors in standard drug testing, another forming better. Those that are doing sufficiently kind of error applies to drugs not under investi- poorly become inadmissible in the sense that their gation. Every such drug is a false neutral. Given assignment weight becomes 0. When and if we the limited resources available to the medical learn that a new agent is effective (or ineffective), establishment to develop new therapies, resource we stop the trial. Patients in the trial benefit from allocation must be approached in a more rational data collected in the trial. The explicit goal is way. This is as true in breast cancer, for which a to treat patients more effectively, but in addition relatively large number of women are willing to we learn about the new agents more efficiently. participate in clinical trials, as it is for other forms Initially we evaluate each design’s frequentist of cancer. Pharmaceutical companies and medi- operating characteristics using Monte Carlo sim- cal researchers generally must be able to consider ulation, possibly modifying the parameters of the hundreds of drugs for development at the same assignment algorithm to achieve the desired char- time. Static trials inhibit the simultaneous pro- acteristics. cessing of many drugs. They cannot efficiently Adaptive designs are being used increasingly address dose–response questions or prioritisation in cancer trials. This is true for trials sponsored by of similar agents when many drugs are under con- pharmaceutical companies, and more generally. sideration. Dynamic designs that are integrated A variety of trials at The University of Texas with the drug development process are necessary M.D. Anderson Cancer Centre (MDACC) are for reasonable progress in medical research. prospectively adaptive. For example, we are Using an adaptive design means examining building the foundation for a Phase II trial the accumulating data periodically – or even con- for evaluating drugs for breast cancer that is tinually – with the goal of modifying the trial’s more a process than a trial. The idea is an design. These modifications depend on what extension of more general adaptive assignment the data show about the unknown hypotheses. strategies. We start with a number of treatment Among the modifications possible are stopping arms plus a control – possibly a standard therapy. early, restricting eligibility criteria, expanding We randomise to the arms and learn about their accrual to additional sites, extending accrual relative efficacy as the trial proceeds. Arms that beyond the trial’s original sample size if its perform better get used more often. An arm that conclusion is still not clear, dropping arms or performs sufficiently poorly gets dropped. An doses, and adding arms or doses. All of these arm that does well enough graduates to Phase III, BREAST CANCER 97 and if it does sufficiently well it might even yet fully evaluable. Other improvements to dose- replace the control. As more treatments become finding methods are underway. These include available, they are added to the mix and the the simultaneous incorporation of efficacy results process can continue indefinitely. into the design, and the use of toxicity severity A trial of a new agent for treatment of rather than the usual assumption that toxicity is metastatic breast cancer is being compared to the dichotomous. current standard therapy in a dynamic manner that allows the incorporation of newly available CONCLUSION treatments in the randomisation process, as well as the elimination of treatments when a lack of Breast cancer clinical trials are not fundamentally improved efficacy can be established. Patients are different from those of other cancers. However, randomised to treatments with weights propor- breast cancer stands out for several reasons. First, tional to the probability that a treatment is better it is common, and it is becoming even more than the standard therapy. The result is that supe- common with the improvements in and greater rior therapies move through quickly and poorer use of detection methods. That implies a greater therapies get dropped. Patients in the trial are pro- ability to investigate the potential for therapeutic vided with better treatment (when the arms are agents and combinations. As a consequence, there not equally good). Patients outside the trial get have been hundreds of randomised clinical trials access to better treatments more rapidly. conducted in breast cancer, more by far than in Dose-finding trials of new agents are also con- any other cancer. Second, it is a disease that is ducted adaptively at MDACC, with dose assign- fatal in only a minority of cases. Third, patient ment based on Bayesian updating of a model advocates in the breast cancer community have been very influential, both as a research force which relates dose and toxicity, using results and a political force, in lobbying for research from preceding patients. The model is the con- funding. Fourth, breast cancer has been shown tinual reassessment method or CRM.48,49 Each to be sensitive to a number of chemotherapies patient is assigned to the dose having a prob- and hormonal therapies. The advances that have ability of toxicity closest to some predetermined been made in breast cancer therapy are more target value. This is the Bayesian posterior proba- impressive than for any other type of cancer, bility calculated from the data available up to that except for testicular cancer and some forms of point (and so it is based on sufficient statistics). leukaemia that commonly affect children. These The CRM more effectively finds the maximum advances have been built on a foundation of tolerated dose (MTD) than does the conventional clinical trials. 3 + 3 design.50 A way in which both the 3 + 3 and CRM designs are crude is the need REFERENCES to pause accrual while waiting for toxicity 51 information. Such pauses are inefficient and 1. Surveillance, Epidemiology, and End Results they cause logistical problems. Trials should be Program (SEER) Public-Use Data (1973–1998), paused or stopped if there are safety concerns, National Cancer Institute, DCCPS, Surveillance not because the design cannot get out of its own Research Program, Cancer Statistics Branch, Bethesda, MD, August 2000 submission; released way. In getting information about toxicity (or April 2001 [online database, available at http:// efficacy), there is seldom a magical dose that the seer.cancer.gov]. next patient must get. All doses are potentially 2. Cancer Intervention and Surveillance Modeling informative. Rather than stopping, one should Network (CISNET) [online information site]. Available through the National Cancer Institute, use a design that models dose–response (toxicity US government, at http://cisnet.cancer.gov/about. and efficacy) and is able to assign a next dose 3. National Breast Cancer Coalition Fund (NBCCF). even though patients previously treated are not A patient guide to quality breast cancer care 98 TEXTBOOK OF CLINICAL TRIALS
[online information site]. Available at http://www. 14. Halsted WS. The results of operations for the cure natlbcc.org/nbccf. of cancer of the breast performed at the Johns 4. American Joint Committee on Cancer (AJCC) Hopkins Hospital from June 1889 to January 1894. [online information site]. Available at http://www. Ann Surg (1894) 20: 497–555. cancerstaging.org. 15. Halsted WS. The results of radical operations for 5. Singletary SE, Allred C, Ashley P, Bassett LW, the cure of carcinoma of the breast. Ann Surg Berry D, Bland KI, Borgen PI, Clark G, Edge SB, (1907) SLVI: 1–19. Hayes DF, Hughes LL, Hutter RBP, Morrow M, 16. Buchanan EB. A century of breast cancer surgery. Page DL, Recht A, Theriault RL, Thor A, Weaver Cancer Invest (1996) 14: 371–7. DL, Wieand HS, Greene FL. Revision of the 17. Haagensen CD. A great leap backward in the American Joint Committee on Cancer Staging treatment of carcinoma of the breast. JAMA (1973) System for Breast Cancer. J Clin Oncol (2002) 224: 1181–3. 20: 3628–36. 18. Fisher B, Montague E, Redmond C, Barton B, 6. Greenberg PA, Hortobagyi´ GN, Smith TL, Ziegler Borland D, Fisher ER, Deutsch M, Schwarz G, LD, Frye DK, Buzdar AU. Long-term follow-up Margolese R, Donegan W, Volk H, Konvolinka C, of patients with complete remission following Gardner B, Cohn I, Lesnick G, Cruz AB, Law- combination chemotherapy for metastatic breast rence W, Nealon T, Butcher H, Lawton R. Com- cancer. J Clin Oncol (1996) 14: 2197–205. parison of radical mastectomy with alternative 7. Early Breast Cancer Trialists’ Collaborative Group treatments for primary breast cancer: a first report (EBCTCG). Tamoxifen for early breast cancer: an of results from a prospective randomized clinical overview of the randomised trials. Lancet (1998) trial. Cancer (1977) 39: 2827–39. 351: 1451–67. 19. Fisher B, Jeong J-H, Anderson S, Bryant J, Fisher 8. Buzdar A, Howell A. Advances in aromatase ER, Wolmark N. Twenty-five-year follow-up of a inhibition: clinical efficacy and tolerability in the randomized trial comparing radical mastectomy, treatment of breast cancer. Clin Cancer Res (2001) total mastectomy, and total mastectomy followed 7: 2620–35. by irradiation. N Engl J Med (2002) 347: 567–75. 9. Yamauchi H, Stearns V, Hayes DF. When is a 20. Cooper RG. Combination chemotherapy in hor- tumor marker ready for prime time? A case study mone-resistant breast cancer. Proc Am Assoc of c-erbB-2 as a predictive factor in breast cancer. Cancer Res (1969) 10: abstr 57. J Clin Oncol (2001) 19: 2334–56. 21. Bonadonna G, Brusamolino E, Valagussa P, 10. Thor AD, Berry DA, Budman DR, Muss HB, Rossi A, Brugnatelli L, Brambilla C, DeLena M, Kute T, Henderson IC, Barcos M, Cirrincione C, Tancini G, Bajetta E, Musumeci R, Veronesi U. Edgerton S, Allred C, Norton L, Liu ET. erbB-2, Combination chemotherapy as an adjuvant p53, and efficacy of adjuvant therapy in lymph treatment in operable breast cancer. N Engl J Med node-positive breast cancer. J Natl Cancer Inst (1976) 294: 405–10. (1998) 90: 1346–60. 22. Hortobagyi´ GN. Anthracyclines in the treatment 11. Paik S, Bryant J, Tan-Chiu E, Yothers G, Park C, of cancer. Drugs (1997) 54(Suppl): 1–7. Wickerham DL, Wolmark N. HER2 and choice of 23. Weiss RB. The anthracyclines: will we ever find adjuvant chemotherapy for invasive breast cancer: a better doxorubicin? Semin Oncol (1992) 19: National Surgical Adjuvant Breast and Bowel 670–86. Project Protocol B-15. J Natl Cancer Inst (2000) 24. Early Breast Cancer Trialists’ Collaborative Group 92: 1991–8. (EBCTCG). Polychemotherapy for breast cancer: 12. Ravdin PM, Green S, Albain KS, Boucher V, an overview of the randomised trials. Lancet Ingle J, Pritchard K, Shepard L, Davidson N, (1998) 352: 930–42. Hayes DF, Clark GM, Martino S, Osborne CK, 25. Early Breast Cancer Trialists’ Collaborative Group Allred DC. Initial report of the SWOG biologi- (EBCTCG). Tamoxifen for early breast cancer cal correlative study of c-erbB-2 expression as a (Cochrane Review) [online database, available at predictor of outcome in a trial comparing adju- http://cochrane.de/cochrane]. Cochrane Database vant CAF T with tamoxifen (T) alone. Proc Am Syst Rev (2001) 1: CD000486. Soc Clin Oncol (1998) 17: 97a. 26. Fisher B, Dignam J, Bryant J, Wolmark N. Five 13. Slamon DJ, Leyland-Jones B, Shak S, Fuchs H, versus more than five years of tamoxifen for Paton V, Bajamonde A, Fleming T, Eiermann W, lymph node-negative breast cancer: updated find- Wolter J, Pegram M, Baselga J, Norton L. Use ings from the National Surgical Adjuvant Breast of chemotherapy plus a monoclonal antibody and Bowel Project B-14 randomized trial. JNatl against HER2 for metastatic breast cancer that Cancer Inst (2001) 93: 684–90. overexpresses HER2. N Engl J Med (2001) 344: 27. Farquhar C, Basser R, Hetrick S, Lethaby A, 783–92. Marjoribanks J. High dose chemotherapy and BREAST CANCER 99
autologous bone marrow or stem cell transplanta- cancer: an overview of 61 randomized trials tion versus conventional chemotherapy for women among 28,896 women. N Engl J Med (1988) 319: with metastatic breast cancer (Cochrane Review) 1681–91. [online database, available at http://cochrane.de/ 38. Early Breast Cancer Trialists’ Collaborative Group cochrane]. Cochrane Database Syst Rev (2003) 1: (EBCTCG). Treatment of Early Breast Cancer, CD003142. Vo l . I, Worldwide Evidence 1985–1990. Oxford: 28. Farquhar C, Basser R, Marjoribanks J, Lethaby A. Oxford University Press (1990). High dose chemotherapy and autologous bone 39. Early Breast Cancer Trialists’ Collaborative Group marrow or stem cell transplantation versus con- (EBCTCG). Systemic treatment of early breast ventional chemotherapy for women with early cancer by hormonal, cytotoxic, or immune ther- poor prognosis breast cancer (Cochrane Review) apy: 133 randomised trials involving 31,000 recur- [online database, available at http://cochrane.de/ rences and 24,000 deaths among 75,000 women. cochrane]. Cochrane Database Syst Rev (2003) 1: Lancet (1992) 339: 1–15; 71–85. CD003139. 40. Early Breast Cancer Trialists’ Collaborative Group 29. Miller AB, Baines CJ, To T, Wall C. Canadian (EBCTCG). Effects of radiotherapy and surgery in National Breast Screening Study: 1 – Breast can- early breast cancer: an overview of the randomized cer detection and death rates among women aged trials. N Engl J Med (1995) 333: 1444–55. 40 to 49 years. Can Med Assoc J (1992) 147: 41. Early Breast Cancer Trialists’ Collaborative Group 1459–76. (EBCTCG). Ovarian ablation in early breast 30. Miller AB, Baines CJ, To T, Wall C. Canadian cancer: overview of the randomised trials. Lancet National Breast Screening Study: 2 – Breast can- (1996) 348: 1189–96. cer detection and death rates among women aged 42. Early Breast Cancer Trialists’ Collaborative Group 50 to 59 years. Can Med Assoc J (1992) 147: (EBCTCG). Favourable and unfavourable effects 1477–88. on long-term survival of radiotherapy for early 31. Nystrom L, Andersson I, Bjurstam N, Frisell J, breast cancer: an overview of the randomised Nordenskjold B, Rutqvist LE. Long-term effects trials. Lancet (2000) 355: 1757–70. of mammography screening: updated overview of 43. Budman DR, Berry DA, Cirrincione CT, Hender- the Swedish randomised trials. [erratum in Lancet son IC, Wood WC, Weiss RB, Ferree CR, Muss (2002) 360: 724]. Lancet (2002) 359: 909–19. HB,GreenMR,NortonL,FreiIIIE.Doseand 32. Gøtzsche PC, Olsen O. Is screening for breast dose intensity as determinants of outcome in the cancer with mammography justifiable? Lancet J Natl Cancer (2000) 355: 129–34. adjuvant treatment of breast cancer. 33. Fisher B, Costantino JP, Wickerham DL, Red- Inst (1998) 90: 1205–11. mond CK, Kavanah M, Cronin WM, Vogel V, 44. Simon R. Optimal two-stage designs for phase II Robidoux A, Dimitrov N, Atkins J, Daly M, Wie- clinical trials. Control Clin Trials (1989) 10: 1–10. and S, Tan-Chiu E, Ford L, Wolmark N. Tamox- 45. O’Brien PC, Fleming TR. A multiple testing pro- ifen for prevention of breast cancer: report of cedure for clinical trials. Biometrics (1979) 35: the National Surgical Adjuvant Breast and Bowel 549–56. Project P-1 Study. J Natl Cancer Inst (1998) 90: 46. Lan KKG, DeMets DL. Discrete sequential bound- 1371–88. aries for clinical trials. Biometrika (1983) 70: 34. IBIS investigators. First results from the Interna- 659–63. tional Breast Cancer Intervention Study (IBIS-I): 47. Berry DA, Fristedt B. Bandit Problems: Sequen- a randomised prevention trial. Lancet (2002) 360: tial Allocation of Experiments. London: Chapman- 817–24. Hall (1985). 35. Vogel VG. Follow-up of the breast cancer preven- 48. O’Quigley J, Pepe M, Fisher L. Continual tion trial and the future of breast cancer prevention reassessment method: a practical design for phase efforts. Clin Cancer Res (2001) 7: 4413s–18s. I clinical trials in cancer. Biometrics (1990) 52: 36. Fisher B, Anderson S, Bryant J, Margolese RG, 673–84. Deutsch M, Fisher ER, Jeong J-H, Wolmark N. 49. Goodman S, Zahurak M, Piantadosi S. Some prac- Twenty-year follow-up of a randomized trial tical improvements in the continual reassessment comparing total mastectomy, lumpectomy, and method for phase I studies. Stat Med (1995) 14: lumpectomy plus irradiation for the treatment of 1149–61. invasive breast cancer. N Engl J Med (2002) 347: 50. Ahn C. An evaluation of phase I cancer clinical 1233–41. trial designs. Stat Med (1998) 17: 1537–49. 37. Early Breast Cancer Trialists’ Collaborative Group 51. Thall PF, Lee JJ, Tseng C-H, Estey E. Accrual (EBCTCG). Effects of adjuvant tamoxifen and strategies for phase I trials with delayed patient of cytotoxic therapy on mortality in early breast outcome. Stat Med (1999) 18: 1155–69. 7 Childhood Cancer
SHARON B. MURPHY1 AND JONATHAN J. SHUSTER2 1Children’s Cancer Research Institute, University of Texas Health Science Centre, San Antonio, TX 78229-3900, USA 2College of Medicine, University of Florida, Gainesville, FL 32610-0212, USA
INTRODUCTION tumours, such as Wilm’s tumour, retinoblastoma, rhabdomyosarcoma, neuroblastoma, Ewing’s sar- coma and osteogenic sarcoma, for example, that There are substantial differences in the conduct are either exclusively or predominantly paediatric of clinical cancer research in children com- in nature. In contrast, carcinomas of differenti- pared to adults. First, childhood cancer is com- ated epithelial tissues, like the aerodigestive tract paratively rare. According to statistics released or breast or prostate, do not occur in children. by the National Cancer Institute SEER pro- Thanks to the usual lack of co-morbid conditions 1 gramme in 1999, it is estimated that approxi- and concomitant illnesses, children usually have mately 12 400 children and adolescents, younger a greater tolerance to cancer therapy than adults. than 20 years of age, are diagnosed annually Taken together with the differing spectrum of with cancer. Stated another way, the average cancer seen, the host differences related to age annual age-adjusted incidence rate for all child- necessitate that paediatric studies of anti-cancer hood cancers is 150 per million persons, aged drug dosage, efficacy and safety are needed. <20. This is under 2% of the total cancers Recognising that children are not just small adults diagnosed in the USA. Despite the rarity and and that special considerations apply, the FDA notwithstanding the spectacular success in treat- has issued regulations mandating the testing of ment of paediatric cancer, compared to inci- new drugs in paediatric patients. dence and mortality rates of cancer occurring Given the fact that modern treatments result in among adults, cancer is the leading cause of cure of 75–80% of all children and adolescents death from disease among children and adoles- with cancer who are managed appropriately, the cents. Only accidents and firearms kill more chil- long-term consequences of therapy for children dren than cancer. Further, the distribution of can- are also potentially much greater than in adults, cer diagnoses in children is very different from as the therapy can interfere with normal growth that in adults. There are a number of major and development, leaving them exposed for
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 102 TEXTBOOK OF CLINICAL TRIALS decades at risk for serious sequelae, major Table 7.1. Major categories of paediatric cancer and organ disturbance, such as cardiac damage and projected annual accrual of the Children’s Oncology cognitive dysfunction, or second malignancies. Group Children diagnosed with cancer generally also Leukaemia have less of a problem with competing mortality Infant ALLa:50 risks, when compared to adult cancer. ‘Standard Risk’ B-Precursor ALLa: 1100 Given these factors, to adequately size child- ‘High Risk’ B-Precursor ALLa: 600 hood cancer research studies, a substantial pro- T-Cell ALLa: 240 Philadelphia Chromosome Positive (Ph+) ALLa:20 portion of the incident cases must be enrolled. B-Cell (SIg+)ALLa:45 In fact, in some situations, such as randomised ANLLa: 300 Phase III trials of new regimens compared to already effective front-line treatments, nation- Lymphoma wide multi-institutional trials are a necessity. Hodgkin’s Disease: 250 These trials will need to enroll nearly every child ‘Early Stage’ NHLa:90 ‘Advanced Stage’ Lymphoblastic NHLa:90 in the target population with the disease being ‘Advanced Stage’ Large Cell NHLa:80 studied for three to five years, with many years of ‘Advanced Stage’ Small Non-Cleaved Cell NHLa: 125 follow-up needed to assess long-term outcomes. Accrual duration in childhood trials may be con- Brain Tumours siderably longer than a corresponding adult trial. Astrocytoma: 140 However, since clinical practice closely approx- Medulloblastoma: 140 imates that of the ongoing study, it is rare that Glioma: 70 Ependymoma: 35 progress from an external source ever renders the study question obsolete. Finally, given the rar- Sarcomas ity of childhood cancer, and the desirability of Ewing’s Sarcoma: 80 study designs of maximal efficiency, it is often Osteosarcoma: 170 desirable to conduct ‘2 × 2 factorial studies’, ‘Low Risk’ Rhabdomyosarcoma: 64 where two interventions are used in the same trial ‘Intermediate Risk’ Rhabdomyosarcoma: 99 ‘High Risk’ Rhabdomyosarcoma: 21 (Standard vs. Standard + A vs. Standard + Bvs. + + Standard A B). Such designs carry some risk Kidney where there is a qualitative interaction between the two interventions. For example, this would Wilm’s Tumour: 480 occur if the impact of A is highly dependent Embryonal upon whether B is given or not. Hence the choice of randomised interventions needs to take this ‘Low Risk’ Neuroblastoma: 200 ‘Intermediate Risk’ Neuroblastoma: 110 pitfall into account when such factorial designs ‘High Risk’ Neuroblastoma: 150 are considered. Hepatoblastoma: 65 Based on previous trials and internal registry Germ Cell Tumours: 25 data of the major national paediatric coopera- aALL = Acute Lymphoblastic Leukaemia AML = Acute Non-Lymphoblastic Leukaemia tive oncology groups, Table 7.1 provides esti- NHL = Non-Hodgkin’s Lymphoma. mates of the potential accrual to the Children’s Oncology Group (COG), a consortium of about 230 American, Canadian, European and Aus- Study Group (NWTSG). Virtually every US or tralasian medical centres. COG was formed in Canadian hospital with a childhood paediatric 2000 by the merger of the Pediatric Oncol- haematology/oncology division belongs to COG. ogy Group (POG), the Children’s Cancer Group Note that the anticipated annual accrual to COG (CCG), the Intergroup Rhabdomyosarcoma Study trials does not mirror incidence figures, as there is Group (IRSG) and the National Wilm’s Tumor a substantial gap between incident rates and rates CHILDHOOD CANCER 103 of referral of newly diagnosed paediatric cancer are presented. A special section dealing with the patients to member institutions of COG and par- ethical aspects and unique considerations affect- ticipation in clinical trials. Ross et al.2 analysed ing the conduct of clinical trials in children is 21 026 incident paediatric cancer cases, diag- also included. The final section is devoted to a nosed from 1989–91, and compared observed look into the future. to expected numbers of cases seen at member institutions of POG and CCG and found vastly HISTORY AND PERSPECTIVES different ratios (observed/expected) depending on ON IMPORTANT PAEDIATRIC age and site, and to a much less extent, geo- CANCER CLINICAL TRIALS graphic region. According to this survey, 92% of children aged less than 15 years in the US Statistically and clinically significant improve- received their care at a CCG or POG institu- ments have been achieved in all major forms tion, thereby providing at least a mechanism of childhood cancers through conduct of well- approximating population-based studies. How- organised single institution and cooperative group ever, the ratio (O/E) for 15–19 year olds was clinical trials which have resulted in sequen- only 0.21, pointing to an adolescent gap in tial and steady improvement in survival rates access to national cancer clinical trials at qual- since the 1960s when curative treatments were ified institutions. first devised. SEER data document that the over- As seen in Table 7.1, the Children’s Oncology all childhood cancer mortality rates have con- Group runs Phase III clinical trials in a wide sistently declined throughout the 1975–95 time variety of tumours, with accrual ranging from period.1 Documentation of the overall progress as few as 20 patients per year to over 1000 per achieved by POG investigators has been reported, year. The Children’s Oncology Group is also demonstrating significant improvements in over- heavily involved in correlative science, pilot all survival (OS) and event-free survival (EFS) studies of potential Phase III interventions, as for 8 of 10 disease areas, in a sample of over 7000 well as standard Phase I (dose escalation) studies children and adolescents treated between 1976 and Phase II (early efficacy) studies. The group and 1989.4 Similar results have been achieved by places special emphasis on translational research CCG and by European national paediatric coop- (biologic correlation studies) and cancer control erative clinical trials organisations. There is also (supportive care studies to limit long-term side evidence that children and adolescents with acute effects and epidemiologic studies to learn about lymphocytic leukaemia (ALL), non-Hodgkin’s the aetiology of childhood cancer). Due to the lymphoma (NHL), Wilm’s tumour, medulloblas- presumably genetic origin of most forms of toma and rhabdomyosarcoma enjoy a significant childhood cancer and the short lag time between survival advantage when treated according to symptoms and diagnosis, prevention trials and well-defined protocols, compared to paediatric screening trials are difficult to do in paediatric patients not enrolled on protocols and treated out- cancer, with neuroblastoma,3 which is based side of paediatric cancer centres.5 Most probably on urinary screening for elevated levels of the inclusion benefit related to participation in catecholamines, a notable exception. clinical trials is a result of a number of factors, This chapter is organised into several sections. including the rigorous process of protocol devel- In the first section, major accomplishments in opment, incorporation of rapid pathology review the area of childhood cancer treatment are dis- and reference laboratories, defined staging prac- cussed. In the following section, examples are tices and procedures, on-study review of radio- cited where translational research has affected therapy port films, and close monitoring for toxi- the design of paediatric cancer trials. In the suc- city and efficacy. Some of the important advances ceeding section, the typical methods of design- achieved in treatment of paediatric cancers are ing trials for the Children’s Oncology Group listed in Table 7.2. 104 TEXTBOOK OF CLINICAL TRIALS
Table 7.2. Examples of important advances resulting Hospital in Memphis who pioneered a ‘Total Ther- from paediatric cancer clinical trials apy’ curative approach beginning in the 1960s. It is beyond the scope of this chapter to review the • Adjuvant chemotherapy improves survival from 20% to 70% in non-metastatic osteosarcoma of treatment advances achieved through clinical trials the extremity.50 for ALL by the BFM (Berlin–Frankfurt–Munster)¨ • Doxorubicin improves outcome when added to Group, POG, CCG, the Dana Farber Consor- 51 other chemotherapy for Ewing’s sarcoma and tium, the Medical Research Council/UKALL, the the addition of ifosfamide and etoposide to vincristine, adriamycin, cyclophosphamide and Dutch Childhood Leukaemia Study Group, the actinomycin results in greater benefit.52 French Acute Lymphoblastic Leukaemia Coopera- • Radiation therapy does not improve survival for tive Group (FRALLE) and the Italian Association patients receiving chemotherapy with Stage I and of Paediatric Haematology–Oncology (AIEOP), 53,54 II, Wilm’s tumour, Stage I but the interested reader may consult reviews rhabdomyosarcoma55 or localised non-Hodgkin’s lymphoma.56 summarising the spectacular progress achieved in 6 • Demonstration of improved event-free survival in treatment of ALL. high-risk neuroblastoma receiving myeloablative The lymphomas, Hodgkin’s disease (HD) and therapy in conjunction with autologous bone non-Hodgkin’s lymphomas (NHL), are the third marrow transplantation and subsequent treatment most common form of paediatric malignancy, with 13-cis-retinoic acid compared to chemotherapy alone.57 next in frequency behind leukaemias and tumours • Attainment of 80% 4-year event-free survival rates of the central nervous system. Currently 80–90% for standard risk B-precursor ALL.58 of all children and adolescents with malignant • Achievement of 78% EFS for patients with lymphomas are curable with optimal multidis- loco-regional embryonal rhabdomyosarcoma ciplinary management, based on immunopatho- through intensification of chemotherapy in Intergroup Rhabdomyosarcoma Study (IRS)-IV.59 logic classification, staging for determination of disease extent, and design and selection of risk-adapted therapies. Paediatric investigators at Success in treatment of the most common Stanford, beginning in 1970, first pioneered com- form of paediatric malignancy, acute lymphoblas- bined modality treatment for children with HD tic leukaemia (ALL), has been most gratifying. and demonstrated that low-dose involved field Indeed, a major reason for improvements in over- radiotherapy combined with multiple cycles of all survival for childhood cancer in general is due chemotherapy (MOPP or MOPP/ABVD) resulted to improvement in survival rates for ALL, which in cure of 90% of paediatric patients.7 Similarly accounts for roughly a third of paediatric cancer.1 outstanding rates of disease control with com- With modern chemotherapy, 97–99% of children bined modality management of paediatric HD can be expected to attain complete remission, and have since been reported by others, establishing it is not inconceivable to predict that modifica- the curability of HD in nearly all cases, such that tions of the currently most successful protocols the thrust of current trials in paediatric HD is will boost long-term leukaemia-free survival rates towards reduction of serious late effects of HD to as high as 85–90%. Treatment success has treatments, such as secondary malignancies, par- been achieved through post-induction intensifi- ticularly leukaemia, infertility, pulmonary fibrosis cation/consolidation and re-induction treatments, and restrictive lung disease, serious cardiac prob- effective treatments (‘prophylaxis’) for subclinical lems and premature death. central nervous system leukaemia, and prolonged The non-Hodgkin’s lymphomas occurring anti-metabolite-based continuation treatments of among children and adolescents are virtually 24–36 months duration. Advances have been all high-grade, diffuse malignancies, differing achieved by many single institutions and coopera- markedly from the distribution of histologic types tive groups treating childhood leukaemias, includ- typically seen among older adults. Staging systems ing investigators at St. Jude Children’s Research in use for childhood and adult NHL also differ.8 CHILDHOOD CANCER 105
Ninety percent of localised NHLs, regardless of specifically for biologically defined, risk-adapted histology, are readily cured by nine weeks of subsets of patients. For example, the National chemotherapy without radiation.9 Wilms’ Tumor Study-5, a therapeutic trial and Progress in the treatment of paediatric solid biology study, was designed to reduce treatment tumours has been equally striking in the last intensity for the subgroup(s) of patients with the 30 years as the progress in treating child- most favourable prognosis and intensify treat- hood leukaemias and lymphomas, and may be ment for the patients with the least favourable attributable to development of accurate diag- prognosis, based on stage, histology (favourable nostic methods and systems of disease staging or unfavourable, anaplastic, rhabdoid and clear and effective multimodal treatments combin- cell types), tumour size and bilaterality, and to ing surgery, chemotherapy and radiation. Cure investigate the impact of loss of heterozygos- rates for rhabdomyosarcoma have increased from ity (LOH) of chromosome 16q and 1p on two- approximately 25% in 1970 to greater than 75% year relapse-free survival through collection of currently, to 60–70% for non-metastatic bone tumour and normal kidney tissue for DNA anal- sarcomas, to over 80% for Wilm’s tumour, over ysis and banking. 90% for retinoblastoma, over 90% for infants Perhaps the best example of important trans- and children with localised neuroblastoma, and lational research that has led directly to thera- to over half of all children with brain tumours. peutic implications is in the collection of bone Aims of current trials are to increase or marrow specimens for cytogenetic studies in preserve high cure rates, decrease acute toxicity childhood acute lymphocytic leukaemia (ALL).10 and long-term adverse sequelae of treatment, While classical karyotype analysis is typically decrease costs and improve the quality of life informative in 60–70% of the patients, impor- for children with readily curable cancers. Patients tant genetic markers can now be identified by with high risk or metastatic disease at diagnosis probes, using FISH (fluorescence in situ hybridi- or those who recur after front-line therapies sation) or PCR (polymerase chain reaction) in continue to pose challenges and should properly virtually all patients. Translocations, such as the benefit from pilot trials and Phase I or II studies t(4;11),11–13 t(9;22)14–16 and t(1;19),17,18 confer of new treatments. an adverse prognosis and lead to targeting the patients for more aggressive therapy. On the other hand, patients with the cryptic t(12;21) genetic PROGNOSTIC FACTORS, TRANSLATIONAL lesion encoding the TEL-AML1 transcript,19–21 RESEARCH AND THERAPEUTICALLY with hyperdiploid leukaemia identified by flow RELEVANT RISK GROUPS cytometric measurement of DNA index (typically 53+ chromosomes in their primary clone),22,23 or Successful childhood cancer research is in large with specific trisomies detected by FISH, such as part dependent upon its translational research 4, 10 and 17,23,24 have a more favourable out- programme. Over the past three decades, initial come and can be targeted for less intensive treat- diagnosis and classification of childhood cancer ment. As an example of the latter, POG investi- has become far more sophisticated, as laboratory gators designed a trial (#9201) with less intense scientists have collaborated closely with clinical chemotherapy for ALL patients with lesser risk of investigators. In addition, special biological and relapse, defined by initial white blood cell counts pharmacological studies, conducted during and <50 000, age between 1 and 10 years, absence of after treatment, offer tools to clinical investigators CNS disease, and presence of one or both of the that were never previously available. As a result, following: DNA index >1.16, and/or trisomies of paediatric oncologists, surgeons, pathologists and chromosomes 4 and 10 by FISH. collaborating statisticians have the opportunity In addition to the well-recognised prognos- and the obligation to design and stratify trials tic importance of initial white blood cell count, 106 TEXTBOOK OF CLINICAL TRIALS age at diagnosis, extramedullary disease and assignments for the Intergroup Rhabdomyosar- blast cell genetic features in B-precursor ALL coma Study V (shown in Table 7.3) are based on and their significance for stratification and trial favourable prognostic factors identified in Stud- design, the early response to therapy, presence ies I–IV, conducted from 1972 through 1991, of minimal residual disease (MRD), and pharma- and include (1) undetectable distant metastases at cologic and pharmacokinetic variables are also diagnosis; (2) primary sites in the orbit and non- predictive of outcome. Slow early response to parameningeal head/neck and genitourinary non- induction treatment is predictive of an adverse bladder/prostate regions; (3) grossly complete outcome and can be defined in several ways: surgical removal of localised tumour at diagno- slow clearance of circulating blast cells to one sis; (4) embryonal/botryoid histology; (5) tumour week of prednisone or multiagent induction, or size ≤5 cm; and (6) age younger than 10 years greater than 25% marrow blasts on day 7 (or at diagnosis. Patients defined as shown into day 14) of treatment. Quantitation of MRD by low, intermediate and high risk are predicted to immunologic methods or PCR assay of rear- have an estimated three-year EFS rate of 88%, ranged T-cell-receptor or immunoglobulin heavy- 55–76% and <30%, respectively.31 chain genes of leukaemic blasts as clonal markers Similarly, significant advances in translational of leukaemia in patients in clinical remission research is neuroblastoma, which accounts for 8 has been shown to identify patients at elevated to 10% of all childhood cancers, have resulted in risk for relapse, a factor which (arguably) should a refined risk-related approach to therapy based be taken into account in assigning alternative on the age of the patient, the stage of the tumour treatment.25,26 Wide variability in absorption of according to the International Neuroblastoma orally administrated chemotherapy, such as 6- Staging System (INSS), histopathologic features, mercaptopurine, and inter-patient variability in the number of N-myc copy numbers and the systemic exposure to both methotrexate and 6- ploidy of tumour cells (Table 7.4). mercaptopurine are important determinants of Because childhood cancer is rare, national ref- outcome in ALL.27,28 Graham et al.29 uncovered erence laboratories have been established to anal- a pharmacologic interaction between methotrex- yse and store samples from the membership of ate and cytosine arabinoside when given simulta- the Children’s Oncology Group as well as other neously, and demonstrated a correlation between large institutions and other international paedi- host drug levels and adverse outcome. Individ- atric clinical trials organisations. Such laborato- ual variability in response to cancer treatment is ries help the research programme in terms of sci- surely related to genetic polymorphisms in drug- entific expertise, quality control and correlative metabolising enzymes, transporters, receptors and science. Few institutions can afford to maintain other drug targets, and suggests that these genetic such laboratories solely for their own paediatric differences may form a solid scientific basis for cancer patients, and web-based informatics appli- optimising therapies within the context of clinical cations afford access to the most sophisticated trials.30 on-line resources and information even in smaller Given the plethora of prognostic factors now remote centres. known for most paediatric malignancies, a prag- matic and rational approach to clinical trials STUDY DESIGN FOR CHILDHOOD design and stratification consists of risk assign- CANCER TRIALS ment by a combination of clinical and biological factors identified through multivariate analysis PHASE I STUDY DESIGN to be of prognostic significance. Treatment is then tailored to risk status, commonly consider- Because childhood cancer is rare and the response ing variables such as patient age, extent of dis- to conventional treatment good, most children ease and tumour biology. For example, the risk never experience recurrent disease and are thus CHILDHOOD CANCER 107 XRT XRT XRT XRT XRT XRT XRT + + + + + + + VAC XRT, Gp II) XRT, Gp II) XRT XRT XRT XRT Topo Topo Topo Topo Topo XRT XRT + + + + + + ± ± ± ± ± + + C VA VA VA VAC VAC VAC ( VAC VAC VAC rNX VAC( rNX VAC M1 N0 or N1 or NX CPT-11, VAC M0 N0 M0M0 N0 N0 or NX VA M0 N0 M0 N1 M0 N1 M0 N0 or NX VAC M0 N1 M0 N0 or NX VAC M0 N1 irinotecan. = biliary tract. S. Histology Metastasis Nodes Treatment Group; CPT-11 = 21 EMB 21 EMB 2121 EMB EMB 21 EMB 21 EMB 21 EMB 21 EMB M0 N0orN1orNX 21 EMB 21 EMB M0 N0orN1o 21 EMB 21 EMB 21 EMB M0 N0orN1o 21 ALV/UDS M0 N0 or N1 or NX 10 EMB M1 VA N0orN1orNX 21 ALV/UDS M1 N0 or N1 or NX CPT-11, 10 EMB ≥ < < < < < < < < < < < < < < < < node status unknown. = undifferentiated sarcomas. = topotecan; Gp aorb aorb aorb aorb aorb = oneal, pelvis, other. (excluding orbit) unfavourable unfavourable unfavourable unfavourable radiotherapy; Topo with alveolar RMS, UDS ingeal), genitourinary (not bladder or prostate) and trunk, retroperit Orbit only a or b Orbit only a or b Unfavourable a Unfavourable a Favourable or Favourable or Unfavourable b Unfavourable a Favourable Favourable or Unfavourable a Unfavourable b Favourable a or b Favourable a or b Unfavourable a Favourable a or b = omas or ectomesenchymomas with embryonal RM 5cmindiameter. rgroup Rhabdomyosarcoma Study Group study V > regional nodes clinically involved; NX = or IV (p. 218), with permission. tumour size = 31 .
et al n D, cyclophosphamide; XRT 1 III 12 III I 1 III 2 II 3 I or II 3 I or II 3 III 2 III 4 I or II or III 4 IV 4 IV 3 III 1 II 1 or 2 or 3 I or II or III Favourable or 5 cm in diameter; b bladder, prostate, extremity, parameningeal, ≤ = orbit/eyelid, head and neck (excluding paramen = vincristine, actinomyci embryonal, botryoid, or spindle-cell rhabdomyosarc alveolar rhabdomyosarcomas or ectomesenchymomas : Reproduced from Raney regional nodes clinically not involved; N1 = = = tumour size = = VAC ALV EMB a N0 Unfavourable Low, subgroup B 1 II (D9602) Table 7.3. Risk group assignments for inte Risk (protocol)Low, subgroup A Stage Group 1 I Site Size Age Favourable (D9602) Intermediate (D9803) High (D9802) Source 108 TEXTBOOK OF CLINICAL TRIALS
Table 7.4. International Neuroblastoma Staging System (INSS)
Stage 1: Localized tumour continued to the area of origin; complete gross resection, with or without microscopic residual disease; identifiable ipsilateral and contralateral lymph node negative for tumour. Stage 2A: Unilateral with incomplete gross resection; identifiable ipsilateral and contralateral lymph node negative for tumour. Stage 2B: Unilateral with complete or incomplete gross resection; with ipsilateral lymph node positive for tumour; identifiable contralateral lymph node negative for tumour. Stage 3: Tumour infiltrating across midline with or without regional lymph node involvement; or unilateral tumour with contralateral lymph node involvement; or midline tumour with bilateral lymph node involvement. Stage 4: Dissemination of tumour to distant lymph nodes, bone marrow, liver or other organs except as defined in stage 4S. Stage 4S: Localized primary tumour as defined in stage 1 or 2, with dissemination limited to liver, skin or bone marrow Risk group and protocol assignment schema: POG and CCG
INSS N-myc Shimada DNA Risk stage Age (y) status histology ploidy group/study
1 0–21 Any Any Any Low 2A and 2B <1 Any Any Any Low ≥1–21 Nonamplifieda Any NA Low ≥1–21 Amplifiedb Favourable NA Low ≥1–21 Amplified Unfavourable NA High 3 <1 Nonamplified Any Any Intermediate <1 Amplified Any Any High ≥1–21 Nonamplified Favourable NA Intermediate ≥1–21 Nonamplified Unfavourable NA High ≥1–21 Amplified Any NA High 4 <1 Nonamplified Any Any Intermediate <1 Amplified Any Any High ≥1–21 Any Any NA High 4S <1 Nonamplified Favourable >1Low <1 Nonamplified Any 1 Intermediate <1 Nonamplified Unfavourable Any Intermediate <1 Amplified Any Any High aN-myc copy number ≤10. bN-myc copy number >10.
POG, Pediatric Oncology Group; CCG, Children’s Cancer Group; INSS, International Neuroblastoma Staging System; NA, not applicable.
Source: Reproduced from Castleberry,60 (pp. 926, 930), with permission from Elsevier. not eligible for trials of new agents. Phase I are accomplished as multi-institutional collab- trials are designed to estimate the maximal tol- orations. Paediatric drug development requires erated dose of a drug, to determine the nature separate Phase I studies (i.e., separate and dis- and frequency of toxicities, and to define the tinct from studies done in adults) because pae- drug pharmacokinetics. While eligibility varies, diatric patients may tolerate either higher or patients have typically failed front-line therapy lower levels of drugs and may exhibit toxici- and usually they will also have failed second-line ties unique to children. Separate trials warranting therapy. Because of the small number of pae- emphasis may also reflect unique agents active diatric patients eligible for Phase I trials, most in paediatric tumours, differing from agents that CHILDHOOD CANCER 109 are of the highest priority for cancers common paediatric Phase I trials have been established.32 among adults. A problem recently identified is the determination The basic design is to begin at about 80% of MTDs in paediatric trials that are lower of the adult maximal tolerated dose. Patients are than those defined in adult patients, which may entered in cohorts and treated at increasing doses. relate to differences in the intensity of prior At each level, three patients are typically accrued. therapy between adult and paediatric patients If there is no dose-limiting toxicity amongst the entered onto Phase I trials. There is a well- three patients, the dose is raised to the next established association between prior therapy level (usually a 20–30% escalation), in succes- and reduced tolerance to myelotoxic drugs. sive cohorts of patients with no intrapatient dose If current paediatric Phase I trials in heavily escalation. If two or all three of these initially pretreated patients define MTDs that tend to be accrued patients experience dose-limiting toxic- lower than those determined in adult patients ity (DLT), the maximum tolerated dose (MTD) with minimal prior therapy, then application of will have been deemed exceeded. Finally, if one the paediatric MTD to less heavily pretreated patient amongst the initial three patients experi- paediatric patients, e.g., in Phase II trials, may ences dose-limiting toxicity, an additional three be problematic. patients are accrued. If six patients are needed, a dose escalation will occur if a total of one in PHASE II STUDY DESIGN six (i.e. zero of the next three) has dose-limiting toxicity. If two or more (i.e. one or more of The specific purpose of a Phase II trial is to the next three) experience dose-limiting toxic- determine activity, i.e., to develop estimates of ity, the maximal tolerated dose will be deemed the response rate of patients with specific tumour to have been exceeded. The MTD is defined types to a particular drug or novel combination. as the dose level immediately below the level Eligible patients typically will have relapsed on at which two patients in three to six experi- a front-line therapy, and the prospect of a cure ence DLT. The definition of dose-limiting toxicity is unlikely. Typically, the dependent variable can vary from study to study, but it generally is an objective all or none response variable falls into two categories: (a) Grade 3, 4 or 5 such as achievement of a complete or partial non-haematologic toxicity other than (1) Grade (>50%) response. Interim results are masked 3 nausea/vomiting; (2) Grade 3 transaminase from the participants until the study closes to elevation; and (3) Grade 3 fever/infection and accrual and response information for all patients (b) Grade 4 myelosupression, that lasts more has been established. There are three types of than 7 days, which requires transfusions twice in Phase II trial designs that depend upon the 7 days, or causes a delay in therapy exceeding study objectives. 14 days. While the study is temporarily closed after accrual of each set of three patients in order Testing Activity to assess patient-specific responses and toxicities, a patient reservation system is used to obtain The most common is ‘proving activity’. For places when and if the study reopens. Phase I these studies, a fixed objective response rate is trials often require the evaluation of many dose specified for activity (null hypothesis), and the levels. At a given dose level, the probabilities of goal is to reject the hypothesis in favour of declaring that the MTD has been exceeded are the alternate hypothesis that the response rate is 9.3%, (50%) and [83%], when the true probabil- greater than this fixed figure. Generally, since the ities of dose-limiting toxicities are respectively number of Phase II agents that can be tested 0.1, (0.3) and [0.5]. is large in comparison to patient availability, Consensus guidelines established by American sequential designs are preferred. However, as and European investigators for the conduct of Simon33 pointed out, it is rarely advantageous to 110 TEXTBOOK OF CLINICAL TRIALS go beyond two stages. Two excellent references PHASE III DESIGN with regard to Phase II design are Simon33 and Shuster34 The designs of Simon33 stop at the first These studies typically ask a randomised question stage only if lack of activity is demonstrated. about either survival or event-free survival (the His argument is that patients should benefit from time from study entry to the earliest of induction failure, relapse, second cancer, or death of any active drugs. However, in paediatrics, due to 40 the relative scarcity of patients with recurrent cause). Intent-to-treat is the analysis of choice disease, designs that stop early for either lack of for efficacy, with other analysis done as sec- activity or proven activity are preferred. ondary supportive inference. For treatment ques- tions where the randomised divergence is con- siderably after study entry or where a significant Historical Comparison number of failures are expected to occur before divergence, a delayed randomisation is typically Another strategy for defining efficacy would be done as close to the divergence point as possible. to prove a response rate is superior to that For these randomisations, the dependent variable seen in an historical control study. The response would be event-free survival from the randomi- rate of the new study is statistically compared sation date. to that of the control therapy. Makuch and 35 Phase III studies are typically designed assum- Simon have provided methods to determine ing either proportional hazards or the cure model the sample size requirements for these studies. 41 36 of Sposto and Sather. In either case, the designs Chang et al. have extended this to two-stage are group sequential in nature with planned designs (i.e., a sequential approach that could interim analyses. In the case of proportional haz- save patient resources). ards, the O’Brien–Fleming method42 is used. The reader is referred to Shuster43 for specific details. Randomised Phase II Comparison Nearly all Phase III childhood cancer trials are run either as two-armed studies or as 2 × 2 fac- Due to a limited availability of patients, it is torial studies. It is rare that sufficient numbers of exceedingly rare that a randomised comparison paediatric cancer patients are available to conduct of a new agent to a control is feasible in a three-armed studies, except perhaps in ALL, the paediatric Phase II study. However, such studies most commonly occurring malignancy. The type have been done. See McWilliams et al.37 for of questions utilised in 2 × 2 factorial studies an example from childhood neuroblastoma. As must be such that the expectation is for no ‘quali- above, two-stage or group sequential designs are tative interaction’ between the two interventions. the preferred method. The programme EAST38 A qualitative interaction between treatments A can be used for designs that allow for both and B would occur if a standard regimen plus A early acceptance and early rejection of the null is superior to the standard regimen alone, but the hypothesis that the new treatment is equivalent standard plus A plus B is inferior to the standard to the control treatment. plus B. For example, if a study is to randomise In paediatric oncology, with limited patient leukaemia patients to receive or not receive regi- numbers, only one or two cooperative Phase II menA,designedtohaveanimpactontheCNS, trials are conducted with each new agent, and while at the same time to receive or not receive all malignancies refractory to standard therapy regimen B, designed to have an impact on mar- are typically combined into a single paediatric row remission, a factorial design would seem Phase II trial, usually stratified by histology. Not appropriate. Essentially, we can run two studies surprisingly, Phase II trials of novel multiagent for the price of one. If the two interventions have regimens provide greater evidence of activity much in common, this would be a contraindi- than single agent Phase II trials and offer cation for a factorial design. In contrast, if we considerable possibility of therapeutic benefit.39 wished to ask if the same drug had an impact in CHILDHOOD CANCER 111 induction therapy (first intervention) and in main- ANCILLARY STUDIES tenance therapy (second intervention), there is, at least intuitively, the plausibility that the advan- In paediatric cancer, there is considerable activity tage of both interventions over just one may be in translational research (see above). This can zero or even harmful. take the form of biologic studies, late effects, or Phase III studies done in cooperative groups in controlling acute side effects. These studies are required by the NCI to have a Data Safety are designed on a case-by-case basis. Examples and Monitoring Board which reviews the study at include the conduct of case–control ‘tissue a minimum of every six months for toxicity and bank’ studies to establish a promising prognostic at planned intervals for efficacy, until it releases marker. Cases are defined as patients failing the study to the study committee. The release a protocol (typically a relapse) and controls can occur no sooner than the earlier of (1) all are long-term successes. These studies can be subjects have completed the planned intervention done using sequential designs, typically two- stage designs. Other typical studies might look or (2) the study was closed early and a new at cognitive impairment (multivariate analysis intervention is needed for patients on one or of variance of neuropsychological variables), both arms. Any release prior to the planned date acute toxicity of a specified type (typical Chi- of final analysis requires approval of the board. square test), the prognostic significance of serial Double-blind Phase III studies are rarely feasible pharmacologically measured drug levels (time- due to the toxic nature of cancer treatment. dependent covariate in survival analysis), or However, they are encouraged for studies of exploratory analysis (e.g. microarrays). supportive care, as long as the intervention is given in a pill form, and has no major known side effects requiring special medical monitoring. ETHICAL AND OTHER SPECIAL Negative questions are often posed for paedi- CONSIDERATIONS AFFECTING CONDUCT atric cancer. For such studies, a very high cure OF TRIALS IN CHILDREN WITH CANCER rate of at least 85% has been shown possible on a conventional regimen. The question posed is can Children and adolescents constitute a special we do ‘almost as well’ with reduced therapy? To vulnerable population of research subjects, often answer such questions with confidence requires grouped with other special classes, like the large numbers, and it is rare that even the entire mentally retarded, mentally ill and prisoners. patient resources of COG are sufficient to address There are special federal protections which apply this in a randomised manner. For example, if a to all research involving children as subjects disease has a historical 4-year remission rate of which are covered by Subpart D of Part 46 of 90%, and an accrual rate of 200 patients per year, Title 45 of the Code of Federal Regulations a randomised study would take 6 years of accrual (45 CFR 46), requiring that institutional review (10-year duration) to have 95% power to detect boards (IRBS) give consideration to the degree a degradation to 85% under reduced therapy at of risk, the benefit to child subjects, the nature p = 0.20, one-sided. (Note that the typical values of the knowledge to be gained, permission of of type I error and power are reversed.) A single- the parent or guardian, and the concurrence of arm study would require 315 patients to ask the the child subjects, known as assent. A child’s same question of a fixed standard of 90% vs. a capacity to give assent is conditioned by his or reduction to 85% (nearly a 75% reduction in sam- her developmental level.44,45 ple size). While the benefits of reduced therapy Subsequent to the promulgation of the original may be obvious, such studies carry considerable rules, adopted in 1983 and modified in 1991, risk and must be carefully monitored for early there has been nearly continuous debate and evidence that the reduction in therapy is unsafe controversy surrounding safeguards for all human and is associated with an inferior outcome. subjects of research and for children especially. 112 TEXTBOOK OF CLINICAL TRIALS
The tragic death of an 18-year-old research certain to increase paediatric clinical trials, partic- subject in 1999 in a gene-transfer trial at ularly drug trials. Federal NIH policies promul- a major research university in which human gated in 1998 were aimed at increasing the par- subjects were not protected, adverse events had ticipation of children in research so that adequate not been reported and financial conflicts of data would be developed to support the treatment interest were involved, served to trigger several for disorders affecting adults which also affect new federal initiatives to further strengthen children, and rules mandated that children (i.e., protections of human research subjects in clinical individuals under age 21) must be included in trials,46 including the imposition of sanctions on all human subjects research unless there are sci- investigators who fail to adhere to regulations. As entific and ethical reasons not to include them. 48 this chapter goes to press, the federal Office for The FDA rules and regulations require phar- Protection from Research Risks (OPRR) has been maceutical manufacturers to assess the safety reorganised, expanded and renamed the Office and effectiveness of new drugs and biologics in for Human Research Protections (OHRP) and paediatric patients and established powerful eco- transferred to the Office of the Secretary, Health nomic incentives for manufacturers (six-months’ and Human Services (HHS) and the National extension of market exclusivity) on any drug Biothetics Advisory Commission, at the request for which FDA requested paediatric studies (see of the President, has undertaken a sweeping www.fda.gov/cder/cancer for further information examination of the ethical and policy issues in on regulatory aspects of paediatric oncology drug the oversight of human research in the United development). In addition to ethical and regulatory issues States (see www.bioethics.gov). As a result, the which impact the conduct of paediatric trials, ethical and regulatory framework within which there are also practical problems associated with paediatric cancer clinical trials are conducted, clinical cancer research in children. Due to an now and in the future, will continue to evolve, understandably greater concern for long-term and investigators must remain abreast. adverse consequences of treatment in a popula- Specific ethical issues impacting statisticians tion of patients, the majority of whom are likely involved in collaborative research include ensur- to be cured and alive for decades at risk for late ing confidentiality, data and safety monitoring, effects, it is absolutely essential that long-term and problems and pitfalls in interpretation of follow-up and serial surveillance of survivors is interim analyses and planning studies to answer built into the studies. While follow-up is essen- 47 negative questions. A negative question, e.g., tial, it is also exceedingly difficult and expensive what is the minimum therapy needed to produce to maintain, as children and adolescents grow up, cure?, has particular relevance for paediatric can- go away to school, leave home, marry, change cer trials which are (often) aimed at reduction of name, etc. The frequency and severity of late the acute or delayed effects of cancer treatment effects also tend to progress with time off treat- on the growing child. ment, making follow-up beyond 15 or 20 or Notwithstanding the strict ethical guidelines 30 years critical and identification of risk factors and regulations surrounding research in children, for the development of these late consequences there is substantial and even increasing pressure of treatment essential. For example, Lipshultz to enroll children in clinical trials as a result et al.49 studied 120 survivors of childhood ALL of other federal policies and recent legislation, or osteogenic sarcoma who had been treated with including the Food and Drug Administration’s doxorubicin a mean of 8.1 years earlier (range (FDA’s) 1998 paediatric rule, the paediatric pro- 2–14 years) and compared their cardiac func- visions of the FDA Modernization Act (FDAMA) tion to a control population, and evaluated the of 1997, and the sweeping Children’s Health Act impact of gender, age at diagnosis, length of of 2000 (PL 106–310), the sum of which is time since completion of therapy, and dosage and CHILDHOOD CANCER 113 cumulative dose of doxorubicin on cardiac sta- agents, many of which will likely be orphan drugs tus. Calculating sex-specific standardised scores for orphan diseases. or z scores (expressed as the number of standard With advances in translational research, the deviations above or below the value for the nor- pie (universe of childhood cancer patients) will mal controls) for cardiac contractility, wall thick- be divided into smaller but more homogeneous ness and afterload, the results of univariate and slices than ever before. International collabora- multivariate analysis showed that female sex and tion will probably be required in a substantial higher cumulative dose of doxorubicin were asso- segment of cancer types in order to obtain suf- ciated with depressed contractility, that there was ficient patient numbers to conduct randomised an association between younger age at diagnosis trials. Enlightened partnerships between industry and reduced left ventricular wall thickness and and academia, with the assistance of the FDA and increased afterload, and that the prevalence and NCI, will be needed for efficient development of severity of abnormalities increased with longer new agents. follow-up.49 Such studies typify the challenge of Finally, the skill sets necessary to conduct methodologic and statistical issues in the study paediatric cancer research are expanding. Tradi- of late effects of childhood cancer, the greatest tionally the field involved paediatric haematolo- challenge being data collection. gist/oncologists, surgeons, radiation oncologists, pathologists, nurses, clinical research associates, pharmacologists, epidemiologists and biostatisti- A LOOK INTO THE FUTURE cians. Today, diagnostic imagers, bench scien- OF CHILDHOOD CANCER RESEARCH tists, geneticists, pharmacists, clinical psychol- ogists, health economists and others also play Despite the progress of the last half century there significant roles in the research. In the future, remain a number of challenges in childhood can- other fields of expertise will surely need to be cer. The focus of research in certain patient sub- added to the team. The cooperation of a multi- sets with very high cure rates will be on quality disciplinary team and prompt referral of patients of life endpoints. For example, retinoblastoma is to paediatric cancer centres participating in clini- curable in nearly 100% of cases, so preservation cal trials will be critical to achieving future goals of sight and reduction of second malignancies of refining and improving therapy. (not survival) are now considered to be the pri- mary goals and endpoints, and trials avoiding REFERENCES enucleation and eliminating external beam ther- apy are now the norm. One would hope that future therapies for child- 1. National Cancer Institute, SEER Program. Cancer Incidence and Survival among Children and Ado- hood cancer will be developed which would be lescents: United States SEER Program 1975–1995. more rational, less empirical and less toxic, rely- NIH Publ. No. 99–4649. Bethesda, MD: National ing more on strategies for growth control (e.g., Cancer Institute (1999). anti-angiogenesis) and regulation of gene expres- 2. Ross JA, Severson RK, Pollock BH, Robison LL. sion and cell proliferation, and/or induction of Childhood cancer in the United States. A geo- graphic analysis of cases from the Pediatric Coop- apoptotic pathways or blocking of anti-apoptotic erative Clinical Trials Groups. Cancer (1996) 77: signals, than on cytotoxic or ablative treatments. 201–7. Assuming that deregulated and/or mutated cellu- 3. Woods WG, Tuchman M, Robison LL et al.A lar proto-oncogenes or loss of tumour suppressor population-based study of the usefulness of genes are the proximate cause(s) of most forms screening for neuroblastoma. Lancet (1996) 348: 1682–7. of childhood cancer, then the genes and/or their 4. Pediatric Oncology Group. Progress against child- protein products will very likely be the targets hood cancer: the Pediatric Oncology Group expe- for the next generation of paediatric anti-cancer rience. Pediatrics (1992) 89(4): 597–600. 114 TEXTBOOK OF CLINICAL TRIALS
5. Murphy SB. The national impact of clinical coop- of contemporary therapies: a report from the erative group trials for pediatric cancer. Med Children’s Cancer Group. J Clin Oncol (1998) 16: Pediat Oncol (1995) 24: 279–80. 527–35. 6. Pui CH. Childhood Leukemias. Cambridge: Cam- 19. Shurtleff SA, Buijs A, Behm FG et al. TEL/AML1 bridge University Press (1999). fusion resulting from a cryptic t(12;21) is the 7. Donaldson SS, Link MP. Combined modality most common genetic lesion in pediatric ALL and treatment with low-dose radiation and MOPP defines a subgroup of patients with an excellent chemotherapy for children with Hodgkin’s prognosis. Leukemia (1995) 9: 1985–9. disease. J Clin Oncol (1987) 5(5): 742–9. 20. Rubnitz JE, Shuster JJ, Land VJ et al.Case– 8. Murphy SB. Classification, staging, and end control study suggests a favorable impact of results of treatment of childhood non-Hodgkin’s TEL rearrangement in patients with B-lineage lymphomas: dissimilarities from adults. Sem Oncol acute lymphoblastic leukemia treated with anti- (1980) 7: 332–9. metabolite-based therapy: a Pediatric Oncology 9. Link M, Shuster J, Donaldson S, Berard CW, Group study. Blood (1997) 89: 1143–6. Murphy SB. Treatment of children and young 21. Raimondi SC, Shurtleff SA, Downing JR et al. adults with early-stage non-Hodgkin’s lymphoma. 12p abnormalities and the TEL gene (ETV6) in New Engl J Med (1997) 337(18): 1259–66. childhood acute lymphoblastic leukemia. Blood 10. Shuster JJ, Carroll AJ, Look TA et al. Manage- (1997) 90: 4559–66. ment of cytogenetic data in multi-center leukemia 22. Trueworthy R, Shuster J, Look T et al. Ploidy trials. Comput Methods Programs Biomed. (1993) of lymphoblasts is the strongest predictor of 40: 269–77. treatment outcome in B-progenitor cell acute 11. Pui CH, Frankel LS, Carroll AJ et al. Clinical lymphoblastic leukemia of childhood: a Pediatric characteristics and treatment outcome of child- Oncology Group study. J Clin Oncol (1992) 10: hood acute lymphoblastic leukemia with the 606–13. t(4;11)(q21;q23): a collaborative study of 40 cases. 23. Heerema NA, Sather HN, Sensel MG et al. Prog- Blood (1991) 77: 440–7. nostic impact of trisomies of chromosomes 10, 17, 12. Pui CH, Carroll AJ, Raimondi SC et al. Child- and 5 among children with acute lymphoblastic hood acute lymphoblastic leukemia with the leukemia and high hyperdiploidy (>50 chromo- t(4;11)(q21;q23): an update. Blood (1994) 83: somes). J Clin Oncol (2000) 18: 1876–87. 2384–5. 24. Harris MB, Shuster JJ, Carroll A et al.Trisomy 13. Heerema NA, Sather HN, Ge J et al. Cytogenetic of leukemic cell chromosomes 4 and 10 identifies studies of infant acute lymphoblastic leukemia: children with B-progenitor cell acute lymphoblas- poor prognosis of infants with t(4;11) – a report tic leukemia with a very low risk of treatment of the Children’s Cancer Group. Leukemia (1999) failure: a Pediatric Oncology Group study. Blood 13: 679–86. (1992) 79: 3316–24. 14. Ribeiro RC, Broniscer A, Rivera GK et al. Phila- 25. Cave H, van der Werff ten Bosch J, Suciu S delphia chromosome-positive acute lymphoblas- et al. Clinical significance of minimal residual dis- tic leukemia in children: durable responses to ease in childhood acute lymphoblastic leukemia. chemotherapy associated with low initial white European Organization for Research and Treat- blood cell counts. Leukemia (1997) 11: 1493–6. ment of Cancer – Childhood Leukemia Cooper- 15. Uckun FM, Nachman JB, Sather HN et al. Poor ative Group. N Engl J Med (1998) 339: 591–8. treatment outcome of Philadelphia chromosome- 26. Campana D, Pui CH. Detection of minimal resid- positive pediatric acute lymphoblastic leukemia ual disease in acute leukemia: methodologic despite intensive chemotherapy. Leukemia Lym- advances and clinical significance. Blood (1995) phoma (1999) 33: 101–6. 85: 1416–34. 16. Arico M, Valsecchi MG, Camitta B et al.Out- 27. Evans WE, Crom WR, Abromowitch M et al. come of treatment in children with Philadel- Clinical pharmacodynamics of high-dose metho- phia chromosome-positive acute lymphoblastic trexate in acute lymphocytic leukemia. Identifica- leukemia. N Engl J Med (2000) 342: 998–1006. tion of a relation between concentration and effect. 17. Crist WM, Carroll AJ, Shuster JJ et al. Poor prog- New Engl J Med (1986) 314: 471–7. nosis of children with pre-B acute lymphoblastic 28. Koren G, Farrazini G, Sulh H et al.Systemic leukemia is associated with the t(1;19)(q23;p13): exposure to mercaptopurine as a prognostic factor a Pediatric Oncology Group study. Blood (1990) in acute lymphocytic leukemia in children. New 76: 117–22. Engl J Med (1990) 323: 17–21. 18. Uckun FM, Sensel MG, Sather HN et al. Clinical 29. Graham ML, Shuster JJ, Kamen BA et al. significance of translocation t(1;19) in childhood Changes in red blood cell methotrexate phar- acute lymphoblastic leukemia in the context macology and their impact on outcome when CHILDHOOD CANCER 115
cytarabine is infused with methotrexate in the 45. Leikin SL. Minors’ assent or dissent to medical treatment of acute lymphocytic leukemia in chil- treatment. JPediat(1983) 102(2): 169–76. dren: a Pediatric Oncology Group study. Clin Can- 46. Shalala D. Protecting research subjects – what cer Res (1996) 2: 331–7. must be done. N Engl J Med (2000) 343(11): 30. Evans WE, Relling MV. Pharmacogenomics: 808–10. translating functional genomics into rational ther- 47. Shuster J et al. Ethical issues in cooperative apeutics. Science (1999) 286: 487–91. cancer therapy trials from a statistical viewpoint, 31. Raney RB, Anderson JR, Barr FG et al. Rhab- II. Specific issues. Am J Ped Hematol/Oncol domyosarcoma and undifferentiated sarcoma in (1985) 7(1): 64–70. the first two decades of life: a selective review 48. Hirschfeld S et al. Pediatric oncology: regulatory of intergroup Rhabdomyosarcoma Study Group initiatives. The Oncologist (2002) 5: 441–4. experience and rationale for intergroup rhab- 49. Lipshultz SE, Lipsitz SR, Mone SM, Goorin AM domyosarcoma study V. J Ped Hematol/Oncol et al. Female sex and higher drug dose as risk (2001) 23: 215–20. factors for late cardiotoxic effects of doxorubicin 32. Smith MD, Ho P, Ungerleider R et al. The con- therapy for childhood cancer. New Engl J Med duct of Phase I trials in children with cancer. J Clin Oncol (1998) 16: 966–78. (1995) 332(26): 1738–43. 33. Simon R. Optimal two-stage designs for phase II 50. Link MP, Goorin AM, Miser AW, Green AA, clinical trials. Contr Clin Trials (1989) 10: 1–10. Pratt CB, Belasco JB, Pritchard J et al. The effect 34. Shuster JJ. Optimal two-stage designs for single of adjuvant chemotherapy on relapse-free survival arm Phase II cancer trials. J. Pharm. Statist (2002) in patients with osteosarcoma of the extremity. 12: 39–51. New Engl J Med (1986) 314: 1600–6. 35. Makuch R, Simon R. Sample size considerations 51. Smith M. The impact of doxorubicin dose inten- for non-randomized comparitive studies. J Chron sity on survival of patients with Ewing’s sarcoma. Dis (1980) 33: 175–81. J Clin Oncol (1991) 9: 889–91. 36. Chang M, Shuster J, Kepner J. Group sequential 52. Grier HE, Krailo MD, Tarbell NJ et al. The addi- designs for phase II trials. Contr Clin Trials (1999) tion of ifosfamide and etoposide to standard 20: 353–64. chemotherapy in Ewing’s sarcoma/primitive neur- 37. McWilliams N, Hayesm F, Green A et al.Cyclo- oectodermal tumor of bone: a Children’s Cancer phosphamide/doxorubicin vs. cisplatin/teniposide Group/Pediatric Oncology Group study. New Engl in the treatment of children older than 12 months JMed(2003) 348: 694–701. of age with disseminated neuroblastoma: a Pedi- 53. D’Angio G, Breslow N, Beckwith B, Evans A, atric Oncology Group randomized Phase II study. Baum E, DeLorimier A et al. Treatment of Wilms’ Med Pediat Oncol (1995) 24: 176–80. tumor: Results of the Third National Wilms’ 38. Mehta C. EaST (Early Stopping in Clinical Trials). Tumor Study. Cancer (1989) 64: 349–60. Cambridge, MA: Cytel Software Corp. (1992). 54. D’Angio G, Evans A, Breslow N, Beckwith B, 39. Weitman S, Ochoa S, Sullivan J et al. Pediatric Bishop H, Feigl P et al. The treatment of Wilms’ Phase II cancer chemotherapy trials: a Pediatric tumor: Results of the National Wilms’ Tumor Oncology Group study. J Ped Hematol/Oncol Study. Cancer (1976) 38: 633–46. (1997) 19: 187–91. 55. Maurer H, Beltangady M, Gehan E et al.The 40. Armitage P. Controversies and achievements in intergoup rhabdomyosarcoma study – 1. Cancer clinical trials. Contr Clin Trials (1984) 5: 67–72. (1988) 61: 1215. 41. Sposto R, Sather H. Determining the duration of 56. Link M, Donaldson S, Berard C, Shuster J, Mur- comparative clinical trials while allowing for cure. J Chronic Dis (1985) 38: 683–90. phy SB. Results of treatment of childhood local- 42. O’Brien PC, Fleming TR. A multiple testing pro- ized non-Hodgkin’s lymphoma with combination cedure for clinical trials. Biometrics (1979) 35: chemotherapy with or without radiotherapy. New 549–56. Engl J Med (1990) 322: 1169–74. 43. Shuster J. Power and sample size for phase 57. Matthay KK, Villablanca JG, Seeger RC, Stram III clinical trials of survival. Chapter 7. In: DO, Harris RE, Ramsay NK, Swift P et al. Treat- Crowley J, ed., Handbook of Statistics in Clinical ment of high-risk neuroblastoma with inten- Oncology. New York: Marcel Dekker (2001). sive chemotherapy, radiotherapy, autologous bone 44. Jonsen AR. Research involving children: recom- marrow transplantation, and 13-cis retinoic acid. mendations of the National Commission for the New Engl J Med (1999) 341: 1165–73. Protection of Human Subjects of Biomedical and 58. Smith M, Arthur D, Camitta B, Carroll AJ, Behavioral Research. Pediatrics (1978) 62(2): Crist W, Gaynon P, Gelber R, Heerema N, 131–6. Link M, Murphy SB et al. Uniform approach to 116 TEXTBOOK OF CLINICAL TRIALS
risk classification and treatment assignment for local or regional embryonal rhabdomyosarcoma: children with acute lymphoblastic leukemia. J Clin results from the Intergroup Rhabdomyosarcoma Oncol (1996) 14(1): 18–24. study IV. J Clin Oncol (2000) 18: 2427–34. 59. Baker KS, Anderson JR, Link MP, Grier HE et al. 60. Castleberry R. Biology and treatment of neurob- Benefit of intensified therapy for patients with lastoma. Ped Clin North Am (1997) 44: 919–37. 8 Gastrointestinal Cancers
DAN SARGENT, RICH GOLDBERG AND PAUL LIMBURG Mayo Clinic, Rochester, MN 55905, USA
INTRODUCTION fallen substantially. For example, in 1930, gastric cancer was the most common cancer diagnosis. Cancers of the gastrointestinal tract account for By 1994, gastric cancer had fallen to 12th in approximately 20% of all new cancer cases in incidence among cancers. In contrast, the rates the United States, and the same proportion of of colon and rectal cancer have remained very cancer-related deaths. In this discussion we will stable. Incidence rates for GI cancers also vary use a broad definition of GI cancer, including greatly worldwide: gastric cancer is tenfold more any cancer of a digestive organ. In this def- prevalent in Asia than in the US. inition we include cancers of the oesophagus, One common feature in all GI cancers is gastro-oesophageal junction, stomach, pancreas, the prognostic importance of staging. The TNM gallbladder, bile duct, liver, small and large intes- system has been widely adopted to describe the tine, rectum and anus. Incident cancers of the patient’s disease status at the time of detection, oesophagus, stomach, pancreas, liver, large intes- and has great relevance to the choice of therapy tine and rectum all exceed 10 000 a year in the and eventual outcome in all GI cancers. The United States. In addition to the high prevalence importance of early detection is clear, and some and the large number of cancer sites within the GI cancers are sufficiently frequent and amenable GI tract, the prognosis of patients with GI cancers to detection to allow cost-effective screening. varies greatly. For example, patients with can- In this chapter we will review, for the major cers of the large intestine, when discovered early sites of the GI tract, the important clinical trials in the course of disease, have 5-year survival that have been conducted. Whenever possible, rates exceeding 90%. In contrast, the prognosis we will highlight the methodological and design for patients with pancreatic cancer is very poor, issues of these trials, in an effort to provide with a 5-year survival rate of less than 5% across insight into their results. We will describe, all stages. through this review, how the current standard Incidence rates for GI cancers show a similar treatments in each disease site have evolved, as diversity. In the past 50 years, the incidence rates well as presenting some of the most pressing for liver and gastric cancers in the US have issues for future research.
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 118 TEXTBOOK OF CLINICAL TRIALS
OESOPHAGEAL CANCER advantage in disease-free survival in the treated group. In smaller trials, Le Prise et al.6 and Urba 7 Oesophageal cancer is an area where contro- et al. reached the same conclusion based on versy as to the appropriate and optimal ther- 86 and 100 randomised patients, respectively. 8 apy exists in almost every aspect. In patients In contrast, Walsh et al., in a trial of 113 with localised disease (stage 1–3), the roles of patients, found a striking survival advantage for surgery, radiotherapy and chemotherapy, alone or the combined modality pre-operative approach, in combination, have all been both advocated and with a median survival of 16 months in the questioned. In advanced disease, it seems clear multimodality arm compared to 11 months in that chemotherapy regimes have provided some the surgery alone arm (p = 0.01). However, degree of progress, albeit limited. the Walsh study has been criticised for several factors, including the small sample size, poorer than expected survival for the surgery alone LOCALISED DISEASE control group, and the fact that the study was In the past two decades, a large number of stopped early at an unplanned interim analysis. randomised clinical trials, involving thousands In an effort to resolve this controversy, a large of patients, have investigated the contributions multicentre randomised trial was mounted in of radiotherapy or chemotherapy, both alone the US, with an accrual goal of 500 patients. and in combination, in the pre-operative and Unfortunately, accrual to the trial was very slow, post-operative settings or as definitive therapy and the trial was closed early, far short of its without surgery. Pre-operative radiotherapy, as a accrual goal. Currently, the combined modality single modality, has been shown in two relatively pre-operative approach has been widely adopted, small randomised trials to provide no additional despite the conflicting evidence of benefit. benefit compared to surgery. These two trials, Additional controversy exists in this setting reported by Launois et al.1 and Gignoux et al.,2 as to whether surgery itself is beneficial. The randomised 124 and 208 patients, respectively. Radiation Therapy Oncology Group (RTOG) has Chemotherapy as a single modality added to conducted two randomised trials that have not surgery was investigated in 440 patients by included surgery as part of the treatment. Her- Kelsen et al.3 and shown to have no advantage skovic et al.9 randomised 129 patients to radi- over surgery alone. The larger sample size of ation alone versus combined chemoradiotherapy. this study lends credence to this result. The two The study was stopped early (planned sample size modalities have also been compared to each other of 150 patients) when the first planned interim as single agents,4 andnodifferenceinpatient analysis showed a significant survival advantage outcomes were observed. Based on these results to the combined modality group. The RTOG it seems clear that single modality therapy has then followed that study with a study comparing limited if any impact on patient outcome. two doses of radiotherapy, both combined with Recently, interest has focused on combined chemotherapy.10 This study was also stopped radiochemotherapy regimens in the pre-operative early, in this case due to a lack of any additional setting. The results in this regard have been benefit in the high-dose radiation arm. No tri- conflicting. Four studies have been conducted, als to date have compared a surgical approach to three with negative results and one with a a non-surgical approach, such a trial would sci- positive conclusion. Bosset et al.5 randomised entifically be highly desirable but the practical 297 patients to pre-operative chemoradiation feasibility of such a trial is questionable. followed by surgery versus surgery alone, and Based on these results, it is clear that there found no evidence of a difference in overall is no consensus as to a ‘standard of care’ survival (a relative risk for survival between the for patients with localised oesophageal cancer, two arms of 1.0), though they did observe an and that there is a great need for additional GASTROINTESTINAL CANCERS 119 clinical trials. Historically, trials in this setting conclusion that the most important surgical prin- have tended to be small and underpowered for ciple is achievement, when possible, of a patho- detecting moderate effects on outcome. Larger, logically negative resection (an R0 resection). more definitive trials should be conducted. However, patients have improved post-operative quality of life if some of the stomach is retained, ADVANCED DISEASE and most surgeons resect only as much of the Trials in advanced oesophageal cancer have stomach as is needed to achieve pathologically been plentiful, though attention in this setting free margins. The rich lymphatic networks of has focused more on Phase II trials than ran- the stomach can sometimes result in apparently domised Phase III trials. A multitude of agents clear margins, yet residual intralymphatic disease have been investigated, alone and in combina- may be present in ‘skip areas’. This has impli- tion. It is clear that progress has been made; cations regarding post-operative treatment, and over the last 20 years median survival for suggests a potential role for adjuvant radiation advanced oesophageal cancer has increased from to the tumour bed and regional structures. 3 months to 6–9 months or greater. The empha- Many surgeons, particularly those in Japan, sis on Phase II trials, in an attempt to find a advocate extended lymph node dissections as promising new approach, is certainly appropriate a means to improve outcome due to the cen- given the modest results available from current tral location of the stomach with many lymph chemotherapies. node-bearing areas at risk for metastatic spread. In a landmark study the Dutch Gastric Cancer Group employed a single Japanese surgeon to GASTRIC CANCER train participating Dutch surgeons to perform the classical Japanese extended lymphadenectomy.12 While the incidence of gastric cancer has declined These investigators randomised 711 eligible in the United States over several decades, 21 600 patients to resection of the primary tumour with new diagnoses and 12 400 deaths were still 11 clear gastric margins and either standard (D1) expected in 2002. The nearly 40% cure rate or extended lymphadenectomy (D2). Three-year that these numbers imply likely results from survival rates were 56% and 58% respectively for a better natural history than oesophageal or the two cohorts, suggesting no advantage to more pancreatic cancer, early detection via endoscopy, aggressive surgery. The British Medical Research improvements in surgery, and the post-operative Council conducted a similar, albeit smaller (400 use of chemotherapy with radiation for patients patients) trial that confirmed this finding.13 with resected disease. While gastric cancer is The adjuvant therapy of gastric cancer, mainly unusual among GI primary sites because of the using 5-FU based regimens, has been a mat- large number of antineoplastic agents that show ter of investigation for many years. Many ran- some activity (as measured by tumour response domised trials of chemotherapy versus surgery rate), in the advanced disease setting even the most alone have been reported and these individual active combination chemotherapy regimens result trials have generally been negative. A meta- in remissions that generally last for only a few analysis of 21 randomised controlled trials con- months and median survivals of less than one year. ducted worldwide, that included 3962 patients with 1840 allocated to surgery alone and 2122 LOCALISED DISEASE allocated to adjuvant chemotherapy, did show a The ideal operation for gastric cancer, includ- modest potential benefit for treatment.14 The odds ing the issues of limited versus total gastrectomy ratio (OR) in favour of chemotherapy was 0.84 and extended versus more limited lymph node overall, but the principal benefit was confined to dissection, has been a matter of controversy. Tri- patients enrolled in trials done in Asia (n = 888 als done in the 1980s and 1990s have led to the patients, OR 0.58) as opposed to Western patients 120 TEXTBOOK OF CLINICAL TRIALS
(n = 3074, OR 0.96). This finding lends some with supportive care alone is around three support to the possibility of a geographically or months. One or more agents from virtually all ethnicity based difference in the natural history classes of chemotherapy drugs have demonstra- of this disease, a finding supported by some epi- ble activity, and median survivals approaching demiologic evidence. Studies of post-operative one year have been reported with several com- radiation versus surgery alone have not shown bination chemotherapy regimens. One example any advantages, although interpretation of the representative of modern Phase III trials ran- limited data addressing this issue is problematic. domised patients to epirubicin, cisplatin and 5-FU Adjuvant radiation and chemotherapy used (ECF) versus 5-FU, doxorubicin and methotrex- in combination has recently been shown to be ate (FAMtx).16 The overall response rate was advantageous in North American patients. In a 45% compared to 21% (p = 0.0002) and the 603 patient study, patients were randomised to overall survival was 8.9 months compared to either surgery alone or to surgery followed by 5.7 months (p = 0.0009) forECFoverFAMtx. combined modality therapy.15 In the treatment Despite the intensive nature of these two reg- arm patients were given 5-FU plus leucovorin imens, and other combinations tested to date, before and after 4500 cGY to the gastric bed the beneficial effects in terms of improved (with radiosensitising 5-FU + leucovorin admin- patient longevity have been modest. Earlier istered for four consecutive days at the beginning detection, improvements in the management of and three days near the end of the radiation). local regional disease, and the testing of new Adjuvant chemoradiotherapy led to a significant agents seem to provide the best avenues towards median survival advantage of 36 compared to better outcomes. 27 months (p = 0.005) and a reduction in local regional relapse rate to 67% compared to 82%. In addition to these outcome improvements, two PANCREATIC CANCER important patterns of care findings were noted. The trial recommended but did not demand at Pancreas cancer has a very poor prognosis. It least a D1 resection and noted that a D2 resec- affects approximately 27 000 new patients each tion was preferred. However, when operative year in the US, and is fatal in approximately reports were analysed, only 10% of patients had 95% of cases. As in all GI cancers, therapy D2 resections, 36% D1 resections, with the bal- includes surgery, radiotherapy and chemotherapy, ance having less aggressive surgery. Secondly, depending on the disease stage. pretreatment radiation field review by a single radiation oncologist indicated that 35% of sub- LOCALISED DISEASE mitted treatment plans contained major or minor deviations from the protocol, indicating a need In the setting of resectable or locally advanced for education of surgeons and radiation oncol- disease, both radiotherapy and chemotherapy, and ogists as to the preferred procedures in these the combination, have been tested extensively. settings. Some readers have raised the possibility Studies conducted prior to the mid-1990s tended that the chemotherapy and radiation were benefi- to be small and underpowered, which has led to cial mainly because of suboptimal surgery in this a variety of conflicting results. cohort of patients. In locally advanced disease, the Gastrointesti- nal Tumor Study Group (GITSG) randomised ADVANCED DISEASE 227 patients to three arms: radiotherapy alone, or radiotherapy at two different dose levels Palliative therapy does make a meaningful dif- given with chemotherapy (5-FU).17 Accrual to ference to many patients with advanced dis- the radiotherapy alone arm was stopped early due ease in gastric cancer, whose median survival to poor results. Two studies have investigated the GASTROINTESTINAL CANCERS 121 need for chemoradiotherapy versus chemotherapy justified by the occasional tumour response that alone, with conflicting results. Klaassen et al.,18 was observed. Single agent therapy with 5-FU in a two-arm randomised study of 191 patients, has been used as the control arm for multiple found no advantage for combined therapy ver- randomised trials, with the assumption that 5-FU sus chemotherapy alone, while GITSG19 reported was at worst a toxic placebo, thus if a new that overall survival was improved with the addi- experimental regimen were shown superior to tion of radiation to chemotherapy in a two-arm 5-FU, it would indeed have improved efficacy study of 43 patients. The small sample sizes of when compared to no treatment. Burris et al.21 all these trials make definitive conclusions diffi- reported a Phase III randomised trial with 126 cult, but there is little evidence to support a role patients that showed an improved overall survival for radiation alone in this setting. for gemcitabine compared to 5-FU alone (median In the setting of a complete surgical resec- survivals of 5.7 versus 4.4 months respectively, tion, several small randomised studies have sug- p = 0.003). The Burris trial established gemc- gested a benefit to post-operative chemotherapy itabine as a new standard of care in this setting. or chemoradiotherapy. None of these trials Ongoing and future trials will likely use gemc- enrolled greater than 114 patients, limiting the itabine as a base, comparing gemcitabine alone ability to draw conclusions. The recent report to a multi-drug chemotherapy regimen including by Neoptolemos et al.20 provided much more gemcitabine. conclusive evidence in this regard. In this A recently completed trial in pancreatic cancer study, 541 eligible patients were randomised to can be used to illustrate the need for careful con- receive post-operative chemotherapy (6 months sideration of an agent prior to Phase III testing. of post-surgical treatment), chemoradiotherapy (a Due in large part to the dismal prognosis and lim- 10-day course of radiotherapy accompanied by ited treatment options available for patients with chemotherapy), both or neither (the design was pancreatic cancer, pressure has been applied to not a true 2 × 2 factorial because clinicians were rapidly introduce novel agents into Phase III tri- allowed to choose to participate in either one or als. The goal is to seek to speed the process of both randomisations). In this study, there was no testing a new agent by avoiding the Phase II stage benefit to the chemoradiotherapy, while a clear of testing. Such was the case in a randomised benefit was observed for the chemotherapy group Phase III trial reported by Moore et al.,22 where a compared to the no treatment group (median sur- novel agent (a matrix metalloproteinase inhibitor vivals of 19.7 versus 14.0 months). Interestingly, (MMPI)) was tested against gemcitabine in 277 the authors of that study did not conclude that patients. In this trial the MMPI had significantly a no treatment arm was inappropriate for future inferior outcome compared to gemcitabine. The trials, in fact they are currently sponsoring a trial was carefully and appropriately designed to three-arm trial in 990 patients of two different allow early stopping if the results were extreme, chemotherapy approaches (5-FU with folinic acid whichinthiscasetheywere.APhaseIItrialmay or gemcitabine) versus surgery alone. This trial have identified the lack of efficacy of this agent has been criticised for including patients with prior to its large-scale testing. involved margins after surgery and also primaries arising in the ampulla and bile ducts. COLORECTAL CANCER ADVANCED DISEASE Colorectal cancer is the most common malig- Chemotherapy has been considered the standard nancy in the GI tract. Not surprisingly, it is also of care in the US for advanced pancreatic cancer, the GI cancer that has been the most extensively despite the lack of any randomised trial demon- investigated in clinical trials. strating a survival benefit for chemotherapy ver- Likely as the direct result of these intensive sus no treatment. The use of chemotherapy was research efforts, considerable progress has been 122 TEXTBOOK OF CLINICAL TRIALS made in many facets of colorectal cancer, effect. Nonetheless, the existing data do not including chemoprevention, early detection and support a major role for this agent in colorectal treatment. cancer chemoprevention. Antioxidant vitamins such as the retinoids, CHEMOPREVENTION carotenoids, ascorbic acid and alpha-tocopherol Cancer chemoprevention can be defined as the may prevent carcinogen formation by neutral- use of nutritional or pharmaceutical agents to ising free radicals within the intestinal lumen. prevent, inhibit or reverse carcinogenesis at a Although somewhat inconsistent, the prepon- pre-invasive stage of disease. Candidate agents derance of data from case–control and cohort are often identified through a combination of studies support an inverse association between epidemiological and laboratory-based research. antioxidant vitamin intake and colorectal can- Since most subjects enrolled onto chemopre- cer risk. Four colorectal cancer chemopreven- vention trials are generally healthy (except for tion trials have investigated antioxidant vitamins their increased cancer risk), minimal toxicity at different doses and in various combinations. represents an important criterion for select- One relatively small study found that recur- ing candidate agents. Colorectal adenomas are rent adenomas were less common among sub- commonly employed as intermediate endpoint jects treated with vitamin A (30 000 IU per biomarkers to facilitate more rapid comple- day), vitamin C (1 g per day) and vitamin E tion of colorectal cancer chemoprevention trials. (70 mg per day) over a mean intervention period 28 To date, several colorectal cancer chemopre- of 17.8 months. Another three-year chemopre- vention agents have been investigated, includ- vention trial reported a 69% reduction in the ing fibre, antioxidant vitamins, calcium and number of recurrent colorectal polyps among sub- nonsteroidal anti-inflammatory drugs (NSAIDs). jects randomised to receive multiple antioxidants Selective cyclooxygenase-2 enzyme inhibitors, (beta-carotene, selenium, vitamin C, vitamin E) which may have a better safety profile than tradi- plus calcium versus placebo compounds.29 Con- tional NSAIDs, have received considerable atten- versely, two larger trials of vitamin C and vita- tion in this context as well. Further discussion min E yielded unremarkable results.30,31 Thus, regarding the current status of these agents is pro- definitive evidence for a protective benefit from vided below. antioxidant vitamins on colorectal cancer risk Dietary fibre represents a heterogeneous mix- remains to be demonstrated. ture of complex materials derived primarily from Calcium may serve as a colorectal cancer plant cell walls. Extensive observational data chemoprevention agent through at least two collected over more than three decades suggest mechanisms: functionally removing toxic bile that fibre might help to prevent colorectal neo- acids from the faecal stream and decreasing cellu- plasia by diluting or adsorbing faecal carcino- lar proliferation in the large bowel mucosa. Data gens, reducing colonic transit time, altering bile compiled from 24 observational studies yielded acid metabolism, or increasing short-chain fatty a summary risk estimate of 0.86 (95% CI = acid production. However, high fibre interven- 0.74–0.98) for colorectal cancer among sub- tions have not been associated with a reduced jects with high versus low calcium intakes.32 In risk for recurrent colorectal adenomas in five addition to encouraging data from the relatively clinical trials.23–27 In fact, one small randomised small clinical trial of multiple antioxidants plus study observed a higher adenoma recurrence rate calcium mentioned above,29 the Calcium Polyp among subjects in the active fibre intervention Prevention Study found that treatment with cal- group.25 It remains possible that administration of cium carbonate at 3 g per day for 4 years was dietary fibre at an earlier point in tumourigenesis associated with a statistically significant 15% (for example, prior to first adenoma formation) reduction in recurrent colorectal adenoma risk, might have a more appreciable anti-carcinogenic as compared to placebo.33 Further data regarding GASTROINTESTINAL CANCERS 123 the chemopreventive potential of calcium (and subjects with familial adenomatous polyposis.37 vitamin D) are being collected from the large Additional COX-2 inhibitor trials are ongoing to Women’s Health Initiative Clinical Trial,34 which confirm the initial findings and to evaluate the should help to clarify whether or not application effect of these agents on sporadic colorectal can- of this agent to average-risk subjects has measur- cer risk. able value. A number of other candidate agents, includ- NSAIDs are a structurally diverse class of ing oestrogen compounds, ursodeoxycholic acid, pharmaceutical agents that appear to reduce difluoromethylornithine and Bowman–Birk in- proliferation, delay cell cycle progression and hibitor, have shown promising results in cell cul- induce apoptosis in epithelially-lined tissues. ture experiments, animal model systems and/or Extensive data from rodent models suggest that observational studies. Further data regarding NSAID administration can reduce gastrointesti- these (and other) potential colorectal cancer nal tumour incidence and/or multiplicity by up chemopreventive agents are anticipated in the to 80%. In human populations, regular NSAID near future as new Phase I, II and III clinical use has been associated with decreased col- trials are organised and completed. orectal cancer risk in numerous observational studies. Despite a consistent demonstration of EARLY DETECTION probable benefit, NSAIDs have not been rig- orously evaluated in colorectal cancer chemo- Due to a variety of factors, colorectal cancers prevention trials until recently. The Physicians’ are very amenable to early detection. First, the Health Study (n = 22 071 subjects), which was biology of colorectal carcinogenesis is becom- a randomised, double-blind, placebo-controlled ing increasingly well understood, as evidenced trial of aspirin 325 mg every other day to prevent by continued expansion of knowledge regard- cardiovascular disease, did analyse colorectal ing the molecular events associated with different cancer incidence rates as a secondary endpoint. stages in the adenoma–carcinoma sequence. This After a mean follow-up period of 5 years, no relatively slow process typically requires sev- statistically significant difference was observed eral years to progress from normal mucosa to between the active and placebo groups (RR = advanced neoplasia, which affords a clear oppor- 1.15; 95% CI = 0.80–1.65).35 Extension of the tunity for detecting lesions at an asymptomatic follow-up period to 12 years did not apprecia- stage. Second, there are a variety of possible bly alter the risk estimate (RR = 1.03; 95% CI = screening methods that range from non-invasive 0.83–1.28).36 However, certain limitations of the stool tests or imaging studies to invasive endo- Physicians’ Health Study trial design, such as scopic evaluations. Third, due to the high inci- the relatively low aspirin dose and lack of uni- dence of colon cancer, such screening may be form colorectal cancer surveillance guidelines, cost-effective in terms of screening costs versus may have hindered its ability to detect a protec- years of life saved. Fourth, the high incidence of tive association. colon cancer provides a motivation for many indi- The chemopreventive effects of traditional viduals to seek out screening. Based on these and NSAIDs are thought to result primarily from other considerations, several randomised trials of inhibition of cyclooxygenase-2 (COX-2). Selec- various screening methods have been conducted. tive COX-2 inhibitors like celecoxib and rofe- With respect to faecal occult blood testing, coxib are therefore being aggressively pursued as three large clinical trials have shown that regular potential colorectal cancer prevention agents. In screening may reduce colorectal cancer mortal- the first trial to be reported, celecoxib 400 mg ity by 13–33%.38–40 In two trials from Europe, twice per day was associated with statistically subjects (n = 61 933 and n = 150 251) were ran- significant reductions in both the mean num- domised to undergo screening every other year ber and total burden of colorectal polyps among versus usual care. In the Minnesota Colon Cancer 124 TEXTBOOK OF CLINICAL TRIALS
Study, subjects (n = 46 551) were randomised the absence of symptoms or other known risk to annual screening, biennial screening or usual factors). However, screening colonoscopy has not care. Follow-up in these studies ranged from yet been investigated in a randomised clinical 11–18 years. Interestingly, only one trial found trial, with the exception of one ongoing feasi- that programmatic screening was associated with bility study.46 Two novel methods of colorectal a statistically significant reduction in colorectal cancer screening, CT colonography and DNA- cancer incidence.40 These data suggest that pre- based stool assays, are currently being tested in invasive adenomas (arguably the most relevant population-based clinical trials as well. Results screening target) are poorly detected by faecal from these studies are anticipated in the near occult blood testing. Thus, despite the inclusion future and may necessitate further modification of faecal occult blood testing in widely endorsed of current early detection algorithms. colorectal cancer screening guidelines,41,42 fur- ther pursuit of more sensitive and specific stool TREATMENT: COLON CANCER biomarkers is needed. Direct examination of the distal colorectum by Localised Disease flexible sigmoidoscopy represents another option for colorectal cancer screening. However, this Surgery is the primary modality for the treat- procedure is at least moderately invasive and ment of localised colon cancer. Depending on may be associated with transient discomfort. As disease stage, surgery alone produces 5-year sur- such, adherence to recommendations for initial vival rates of 50% to greater than 90%. As and repeat flexible sigmoidoscopies was recently opposed to gastric and rectal cancer, however, evaluated in the Prostate, Lung, Colorectal and the surgical technique for colon cancer resec- Ovarian (PLCO) Cancer Screening Trial. Among tion has been the subject of limited investigation subjects randomised to the screening intervention in randomised clinical trials. One large surgical arm (n = 17 713), 83% completed the baseline trial recently completed accrual of approximately flexible sigmoidoscopy. Additionally, 87% of 900 patients. In this study, patients were ran- subjects who were eligible for repeat testing domised to either the standard ‘open’ colectomy after three years complied with the follow-up or a ‘minimally invasive’ laproscopically assisted evaluation.43 At present, the effects of flexible colectomy.47 The trial’s primary endpoint was sigmoidoscopy screening on colorectal cancer cancer recurrence, and it was designed to demon- incidence in the PLCO trial cohort remain strate equivalence of the two approaches. The unknown. An even larger flexible sigmoidoscopy trial also included extensive quality of life and screening trial is underway at 14 centres in cost-effectiveness assessments. the United Kingdom and Wales (n = 170 432 The value of adjuvant treatment for patients randomised subjects).44 When available, data with stage 3 colon cancer (cancer able to be from these two trials should be highly informative completely resected, but with positive lymph regarding the utility of flexible sigmoidoscopy nodes in the resection specimen) is well accepted. screening to reduce colorectal cancer incidence The first trial conducted with a positive result was rates in the general population. conducted by the North Central Treatment group, Colonoscopy is currently the gold standard initiated in 1978.48 This was a three-arm trial, for structural evaluation of the large intes- with a sample size of approximately 135 patients tine. Cost-effectiveness models suggest that per arm, comparing no post-surgical treatment to one-time screening colonoscopy between ages adjuvant treatment with either levamisole alone 50–54 years may be a rationale colorectal cancer or 5-FU plus levamisole. The initial results of prevention approach.45 Existing early detection this trial indicated a moderate but statistically guidelines support a slightly more conserva- significant benefit for the 5-FU plus levamisole tive strategy (i.e. colonoscopy every 10 years, in arm compared to control. Given the novelty of GASTROINTESTINAL CANCERS 125 this result, in a decision that likely would never infusion. Based on promising results in the be made in the current day, the investigative advanced disease setting (as discussed below), team decided to embark on a larger, confirmatory multiple clinical trials have been conducted using trial prior to the release of the results to the regimens based on a long-term infusion with oncology community. This confirmatory trial, 5-FU. Intergroup trial 0153 directly compared known as Intergroup trial 0035, enrolled over a bolus to an infusional 5-FU based regimen 1200 patients to the same three arms as the in a randomised Phase III trial of 1078 patients initial trial. Intergroup 0035 clearly demonstrated (terminated early at an interim analysis–original improved overall survival in patients treated with planned sample size of 1800 patients).56 In this 49 adjuvant 5-FU and levamisole. These findings trial no difference in efficacy was observed lead in part to the 1990 consensus statement between the arms, although the toxicity profile from the National Cancer Institute that patients did differ substantially. Based on these results, with stage 3 colon cancer who are unable to two recent Phase III randomised trials in the enter a clinical trial should be offered adjuvant United States have used control arms of 6 months treatment with 5-FU plus levamisole unless there of bolus 5-FU plus leucovorin. However in 50 are contraindications. Europe, regimens using short-term 24–48 hour A number of clinical trials were in progress 5-FU infusions are more popular. at the time of the publication of the beneficial Future efforts in the adjuvant treatment of stage results from the use of adjuvant 5-FU plus 3 colon cancer will likely be directed in two levamisole. Several of these trials included a areas – first, to improving the treatment options no post-surgical treatment control arm, and thus available, and second, and relatedly, to tailoring these trials were closed prior to reaching their therapy to the individual patient. In the treatment accrual goals due to ethical reasons. Included in area, for the time being, new studies will ran- this list of trials that were closed prematurely domise patients to treatments based on adding a were five Phase III randomised trials testing new treatment to a 5-FU and leucovorin regimen. 5-FU plus leucovorin versus no post-surgical For example, the two recent large Phase III trials treatment control. The results from three of these trials were pooled for analysis;51 the other two in the United States compared 5-FU and leucov- were reported separately.52,53 In each of these orin to either a 5-FU, leucovorin and irinotecan analyses, adjuvant 5-FU plus leucovorin showed (trial C89803) or 5-FU, leucovorin and oxali- a survival advantage compared to control. In platin (trial C-06). In somewhat of a leap of faith, subsequent studies, throughout the 1990s, various the trial currently being planned by the US Gas- investigative groups conducted trials comparing trointestinal Intergroup will compare 5-FU, leu- various different schedules and combinations covorin and irinotecan to 5-FU, leucovorin and of 5-FU combined with either leucovorin or oxaliplatin. This leap is necessitated by the new levamisole. None of these trials demonstrated a realities of conducting trials in the adjuvant colon statistically significant improvement in survival setting – namely, that patient outcome has been between study arms, although through such trials sufficiently improved that significant follow-up it did become clear that 6 months of 5-FU plus is required in order to obtain sufficient events to leucovorin was at least as effective as 12 months power a study. Both the C89803 and C-06 trials of 5-FU plus levamisole.54,55 As a result, the will require a minimum of three years of follow- current standard of care in the United States (as up prior to any formal analysis. The options for of 2002) for stage 3 colon cancer is 6 months of conducting a follow-up study are thus to wait at 5-FU plus leucovorin. least three years, or to push ahead, assuming that In the discussion in the preceding paragraph, at least one of the experimental regimens will all of the regimens discussed were based on prove superior to the current standard. A dis- the delivery of 5-FU as a short-term bolus cussion of the second area, tailoring therapy to 126 TEXTBOOK OF CLINICAL TRIALS the individual patient, will be deferred until the Mamounas et al.,59 pooled data from four ran- next section. domised trials conducted by the NSABP. In none One additional insight into the conduct of clin- of these four trials was there a direct randomised ical trials in GI cancers may be gained by exam- comparison between treatment and control. In ining the steady increase in the sample sizes that their analysis, the authors estimated the magni- has occurred in stage 3 colon clinical trials over tude of the difference in outcome between the two the past two decades. In trials conduced in the study arms in each of the four studies. They then early 1980s, sample sizes of 100–200 per arm compared whether this difference in outcome dif- were typical,48,52 with some exceptions (such as fered by patient stage. The authors concluded that the NSABP C-01 trial, with approximately 380 the treatment effect within each study was similar patients per arm).57 With such a sample size, the between stage 2 and stage 3 patients, and since study provided adequate power to detect only a it had been previously demonstrated that treat- relatively large effect. Fortunately, 5-FU, when ment is beneficial in stage 3 patients, they con- combined with either levamisole or leucovorin, cluded that treatment is also beneficial in stage did provide a rather large effect, with a reduction 2 patients. in the hazard of death by approximately 25%.58 The second investigation60 used a more direct However, the likelihood of a subsequent treat- approach. In this analysis, the study team pooled ment advance by such a magnitude is unlikely, the data from stage 2 patients who had partic- and smaller advances may indeed be clinically ipated in five randomised trials of 5-FU plus relevant. Therefore, more modern trials in stage 3 leucovorin versus control. They found no statis- disease have included sample sizes of 1600 (trial tically significant benefit of treatment, based on C89803), 2400 (trial C-06) and 4900 patients for a pooled sample size of just over 1000 patients. a four-arm trial (the QUASAR trial).54 As ther- Due to the excellent outcome of stage 2 patients, apy continues to improve, the sample size neces- with an approximately 80% 5-year survival in sary to detect further incremental advances will untreated patients, even this pooled sample size continue to grow. One possible strategy for prac- had poor power to detect a small but possi- tically conducting such large trials is discussed in bly important improvement in patient outcome the next section. (only 60% power to detect an 85% 5-year sur- As opposed to the adjuvant treatment of stage vival in treated patients). Thus, the benefit of 3 disease, the benefit of adjuvant treatment for 5-FU based adjuvant therapy in stage 2 disease stage 2 (node negative) disease is unclear. In remains unclear, and further pooled analyses will many previous trials, patients with stage 2 disease likely be necessary. A large randomised trial of have been pooled together with stage 3 patients. a monoclonal antibody in the setting of stage 2 The sample size for such trials has typically been disease has recently completed accrual of over based on an analysis pooling the data from both 1700 patients (trial C9581); the results of this patient groups. For a variety of reasons, patients trial should help clarify the appropriate treatment with stage 3 disease have typically constituted for patients with stage 2 disease. a majority of the enrollment to such trials, thus each individual trial has been underpowered to Advanced Disease detect a moderate benefit of treatment in stage 2 patients. Due to the limited sample size in It is likely that more clinical trials have been each trial, two attempts have been made to pool conducted in advanced colon cancer than in any data from several trials in order to gain a suffi- other GI disease site. This is due to the high cient sample size to draw a definitive conclusion incidence of the cancer, and the fact that it is to at regarding the value of adjuvant therapy in stage 2 least some degree sensitive to chemotherapeutic disease. However, the two analyses have reported agents. Trials in advanced colon cancer typically differing conclusions. One analysis, reported by include patients with advanced rectal cancer, as GASTROINTESTINAL CANCERS 127 the response to chemotherapy has not been shown At almost the same time as the introduction to depend on the precise site of the patient’s of oral 5-FU based agents for advanced colorec- disease within the colorectum. tal cancer, new chemotherapeutic agents have The drug 5-FU has been the mainstay of treat- been added to 5-FU with promising results. Based ment for colorectal cancer for over 40 years. on results with the agent irinotecan in patients From 1950–1990, a multitude of trials were con- who had failed a 5-FU based regimen,66,67 tri- ducted in an effort to improve the efficacy of als with irinotecan were performed in the setting 5-FU based regimens, by changing methods of of patients with previously untreated advanced administration, combining it with various supple- disease. As reported by Saltz et al.68 and Douil- mental agents (such as leucovorin or levamisole), lard et al.,69 irinotecan, when added to 5-FU and or changing the dose and schedule. Regarding leucovorin, resulted in improved time to progres- the timing of administration, the clear result of sion and overall survival when compared to 5- multiple studies is that, among regimens where FU and leucovorin alone in first-line treatment 5-FU is delivered as a bolus injection, the partic- of advanced disease. These two relatively large ulars of the administration have a definite impact trials (683 and 387 patients, respectively) estab- on toxicity, some impact on tumour response, lished a new standard of care in this setting. In but little impact on patient survival. The addi- the Saltz trial irinotecan was added to a bolus tion of leucovorin to 5-FU has been demon- 5-FU schedule, while the Douillard trial added strated in a meta-analysis to provide increased irinotecan to an infusional 5-FU regimen, thus efficacy in terms of response rate compared to the optimal method in which to give 5-FU with 5-FU alone.61 In another meta-analysis, a sched- the new agent remained unclear. Recently, the ule where the 5-FU is delivered by a continuous drug oxaliplatin has shown promising activity infusion has been shown to provide an advantage when combined with 5-FU and leucovorin in sev- in both toxicity and overall survival compared to eral studies,70,71 with reported median survivals bolus schedules.62,63 However, the improvement equaling or exceeding 18 months. in median survival was modest at 0.8 months, The proven efficacy of irinotecan as second- thus many practitioners (at least in the United line therapy in patients who fail 5-FU based States) have continued to administer the bolus 5- therapy has complicated the design of first- FU based regimens based on perceived benefits line advanced disease trials. Traditionally, overall of patient and physician convenience. survival has been used as the primary endpoint After 40 years of testing variations on a 5-FU for such studies, and extending the patient’s theme, two more recent developments have added longevity remains the ultimate goal. However, excitement to the advanced colorectal cancer given that there are second- and even third-line clinical trials arena. The first is the introduction therapies with proven benefit, the relative merits of oral 5-FU based regimens. The oral method of of overall survival as the primary outcome for delivery offers clear benefits in terms of patient a trial warrant a reconsideration. Consider the preference. However, an oral approach would not design used in the Saltz trial,68 where irinotecan likely be accepted if it did not provide at least plus 5-FU and leucovorin was compared to 5- equivalent efficacy to an IV approach. Therefore, FU and leucovorin. In this trial, patients who two large equivalence trials have been conducted progressed on the 5-FU and leucovorin arm comparinganoraltoanIVregimen.Thesetwo were able to receive irinotecan off study, as it trials, one reported by Hoff et al.64 and the other was approved for the second-line indication. The by van Cutsem et al.,65 enrolled 605 and 602 availability of this effective second-line agent patients respectively, and were formally designed provided at least the theoretical possibility that to test the equivalence of the oral regimen to the the two primary study arms could show no IV approach. In both cases, formal equivalence difference in terms of overall survival, even was declared. though irinotecan was beneficial to patients on 128 TEXTBOOK OF CLINICAL TRIALS both arms of the study. For this reason, time be optimised by using continuous infusion 5-FU to tumour progression was specified as the based therapy as opposed to bolus. All patients primary endpoint for the trial. In retrospect, in this study received radiation. This study of the addition of irinotecan as a component of 680 patients concluded that (1) semustine is not the initial treatment resulted in both improved necessary, and (2) infusional 5-FU based ther- time to progression and overall survival, making apy during the radiotherapy provides a survival the result clear. However, these factors must be advantage compared to bolus therapy.73 taken into consideration for future trials, where at Two studies conducted by the National Sur- minimum data on the use of second- and third- gical Breast and Bowel Project (NSABP) have line therapy should be collected. questioned the value of radiation in the post- operative setting. In the Krook and O’Connell TREATMENT: RECTAL CANCER studies mentioned above, all patients received radiation, and the studies focused on the rela- Rectal cancer is second to colon cancer among tive benefit of different chemotherapy regimens. GI malignancies in the number of new cases In contrast, NSABP study R-01 tested three arms per year. When the initial diagnosis for rectal in a randomised manner in 574 patients: no post- cancer is as advanced disease, i.e. not surgically surgical treatment, post-operative radiation and completely resectable, its primary treatment is in post-operative chemotherapy. A survival benefit the same manner as for advanced colon cancer. was observed for the chemotherapy arm com- However, the optimal adjuvant treatment for pared to the no treatment arm, but this advan- rectal cancer is the issue of considerable study. tage was not observed in the radiation alone Questions abound as to the importance of surgical arm.74 The NSABP followed this study with a technique, the value of radiation therapy, the two-arm randomised trial of 741 patients com- optimal chemotherapy regimen and the timing of paring chemotherapy alone to chemotherapy plus therapy, either pre- or post-resection. radiation.75 The results of this trial showed no improvement in overall survival for the combined Surgery/Adjuvant Therapy modality arm, although there was a statistically significant improvement in the rate of local recur- Prior to 1990, external beam radiotherapy in the rence associated with radiation. Despite these two post-operative setting was considered by many as consistent results, radiation continues to be com- the standard of care in the United States, based monly used in the post-operative treatment of primarily on an observed benefit in lowering the stage 2 and 3 rectal cancer. risk of local recurrence. In particular, radiation Increasingly, practitioners are turning to deliv- as a single agent added to surgery had never ering radiotherapy for rectal cancer in the pre- been shown to improve overall survival com- operative setting. There are several theoretical pared to surgery alone. In a randomised study of advantages to the pre-operative approach. Per- 204 patients, Krook et al.72 demonstrated a bene- haps most importantly, from the patient’s per- fit in overall survival of post-operative combined spective, pre-operative therapy may shrink the therapy with radiation and 5-FU and semus- tumour sufficiently to allow a sphincter-sparing tine compared to radiation alone. The 1990 NIH resection. Pre-operative radiotherapy has been consensus statement concluded that ‘Combined shown to improve outcome compared to no treat- postoperative chemotherapy and radiation ther- ment in a large randomised trial of the Swedish apy improves local control and survival in stage II Rectal Cancer Trial group. This trial randomised and III patients and is recommended’.50 In a sub- 1168 patients to a two-arm trial of a short sequent study conducted by the US GI Intergroup, course (25 Gy in 1 week) of pre-operative radia- two questions were asked in a 2 × 2 factorial tion compared to no pre-operative treatment, and design: is semustine necessary, and can therapy showed a statistically significant improvement in GASTROINTESTINAL CANCERS 129 both local recurrence rate and overall survival.76 radiation therapy have been shown to reduce the In the United States, the standard pre-operative local recurrence rate and improve overall survival regimen is to deliver the 5-week post-operative based on large randomised trials. It is also clear course of 50.4 Gy pre-operatively. The efficacy that considerable work remains to define the opti- of this approach has never been established in mal timings and combinations of the different a randomised trial. A comparison of these two treatment modalities. pre-operative approaches is clearly warranted. Regardless of the specifics of the pre-operative approach, a burning question concerns whether SELECTED METHODOLOGIC ISSUES the pre- or post-operative approach provides the WITH EXAMPLES best outcome. Two randomised trials have been attempted in the United States, and both were The number and variety of large Phase III closed early far short of their accrual goals due randomised trials that have been conducted in GI to poor accrual. However, an ongoing trial in cancers brings to light a number of clinical trials Europe has been more successful, with accrual of methodology issues that have direct relevance to over 600 patients to one of the two approaches.77 the conduct and conclusions that can be drawn The 50.4 Gy long course radiation is being used from such trials. In this section we will highlight in both arms, and both arms are receiving the three such issues: subgroup analysis, the study of same chemotherapy regimen in combination with prognostic and predictive factors, and monitoring the radiation. of clinical trials for safety. In addition to the controversies present in chemotherapy and radiation therapy, there is SUBGROUP ANALYSIS considerable interest in the optimal surgical management of this disease. In particular, the Studies are often conducted in multiple possible surgical approach of total mesorectal excision heterogeneous groups of patients in order to (TME) has been promoted as an important achieve a necessary sample size, or to complete surgical advance. Based on case-series and other a trial more quickly. The conduct subgroup historical data, proponents of TME have claimed analyses, that is, to examine the result of a significant reductions in local recurrence rates trial separately in different groups of randomised and improved overall survival compared to patients, is extremely important. In fact, such standard surgery.78 However, TME has never analyses can help demonstrate the robustness of been tested against non-TME surgery in a a result: if a study finding can be shown to be randomised trial, and such a trial is unlikely to consistent in different patient groups, such as ever be conducted. In a large randomised trial of patients of different ages, or patients enrolled 1861 patients conducted by the Dutch Colorectal from different countries, then the overall result Cancer Group, pre-operative radiation was shown cannot be explained by an extreme result in to reduce the rate of local recurrence compared one subset of patients. Additionally, subgroup to no radiation when all patients received TME analyses are important in generating hypotheses surgery.79 In this early report, with a median for future study. However, caution must be follow-up of 2 years in living patients, there was exercised when examining the results of subset no improvement in overall survival for patients analyses, particularly if they were not pre- receiving radiation. specified in advance. In summary, it is clear that rectal cancer is One example of a subgroup analysis of great an area where randomised clinical trials have controversy in the setting of GI cancers is made several important contributions to improv- whether post-surgical adjuvant treatment is ben- ing patient outcomes. Post-operative chemother- eficial for patients with stage 2 colon cancer, apy and chemoradiotherapy, and pre-operative as discussed previously. Historically, most trials 130 TEXTBOOK OF CLINICAL TRIALS in adjuvant colon cancer have included patients treatment benefit in both subgroups of patients; with both stage 2 and 3 disease. The majority of the absence of a significant interaction implies patients in such trials have been stage 3 patients, that there is no evidence that the benefit of treat- and subset analyses within these trials have con- ment differs by patient subgroup. sistently shown 5-FU based chemotherapy to be Two adjuvant rectal trials conducted by the beneficial in this group. However, because the NSABP, trials R-0174 and R-02,75 can be used majority of patients in such trials have not been to demonstrate a different feature of subgroup stage 2, and because the prognosis for the stage analyses. The first trial, R-01, demonstrated a 2 patients is overall more favourable, with fewer significant benefit in terms of overall survival patient deaths and thus less statistical power, for the addition of chemotherapy following the individual studies have not shown consistent resection compared to no post-surgical treatment. results in the subset of B2 patients. As mentioned However, in subgroup analyses, this benefit above, two pooled analyse have been conducted, seemed to be limited to the male patients, with conflicting results. Both analyses are to be which could not be explained. The NSABP is commended for obtaining and using the individ- to be commended for treating this finding as ual patient data from each of the included trials, hypothesis generating, and testing the hypothesis and for having very complete follow-up for the in their next study R-02. In study R-02, the studies used. However, both analyses could be randomisation scheme differed for men and criticised for other methodologic issues. The first, women, with females being randomised to two by the NSABP investigators,59 pooled data from arms and males to four arms. The results of R- multiple trials with different treatment arms, none 02 did not demonstrate the need for different of which compared directly a no treatment arm treatment for the two genders, which put the to what would be considered a standard treat- issue to rest after being tested as appropriate in ment by current standards. The second, by the a randomised trial. This experience demonstrates IMPACT investigators,60 did pool results from the value of confirming a finding that results from randomised trials of no post-surgical treatment to a subgroup analysis prior to accepting the result therapy with 5-FU and leucovorin, the current US into clinical practice. standard of care. However, these investigators did not include data from two large trials testing 5-FU TUMOUR MARKER STUDIES and levamisole versus control, which despite hav- A second area of considerable interest and debate ing the 5-FU modulated by a different agent did in the GI cancer community regards the use indeed test a very similar question, with 5-FU and of putative prognostic and predictive markers to levamisole shown in large randomised trials to guide the choice of therapy for an individual give results indistinguishable from 5-FU and leu- patient. Hundreds if not thousands of studies have covorin. Additionally, the IMPACT investigators been done on markers based on immunohisto- used a less powerful analysis than might be pos- chemistry, flow cytometry, chromosomal markers sible. The original trials included in the IMPACT such as allelic loss and microsatellite instability, analysis included patients with both stage 2 and pathologic features, and many others. Unfortu- 3 cancers. In these trials, treatment was proven nately, few if any of these markers have made beneficial overall. Therefore, the most relevant their way into clinical practice. The reasons for question is whether the effect of treatment dif- this lack of progress are many,80 we will focus fered between the stage 2 patients and the stage 3 here on three that are directly related to the patients. This question could be tested by obtain- later stages of clinical trials: analyses confined ing all of the data from the trials, examining to patient subgroups, inadequate sample size and the degree of benefit overall, and then testing improper design. for a stage by treatment interaction. This is the In any report investigating tumour markers most statistically powerful method for testing the based on patients from clinical trials, the rate GASTROINTESTINAL CANCERS 131 of sample collection is an important issue. In several Phase III trials that are in development the case of retrospective trials conducted on tis- in the US. sue obtained from a clinical trial, often tissue is available on only a subset of the patients who SAFETY MONITORING entered the initial trial. In these retrospective marker studies, comparisons are frequently made As a final clinical trials methodologic issue, between characteristics such as baseline demo- we consider the process of monitoring patient graphics and/or tumour stage for the patients safety in clinical trials. Clinical trials have whose tumours were used in the marker study been conducted for many years, and detailed and those whose tissues were not used. How- and effective methods have been developed for ever, even if the characteristics for the patients ensuring the safety of participants. One of the who were used in the analysis and those who fundamental tenants of clinical trials is the were not appear similar, the results of such stud- progression of an agent or regimen through ies could still be biased. Such an example has a series of trials, starting in small, typically 81 been described by Pajak et al., who reported on single-centre Phase I studies, on to somewhat 82 a study originally conducted by Grignon et al. larger, possibly multi-centre Phase II trials, then In their study, in patients with advanced prostate on to Phase III trials. The sequential nature of adenocarcinoma, Grignon et al. studied the rela- such trials, among other features, allows for tionship between tumour p53 status and progno- considerable experience with an agent or drug sis in 129 of 456 patients entered onto a trial combination prior to a large, multi-centre trial. conducted by the Radiation Therapy Oncology The recent pressure towards developing and Group (RTOG study 86-10). When reanalysed testing agents more rapidly, although beneficial by Pajak et al., it was shown that the distribu- in many ways, does challenge this established tion of treatment received and combined Glea- mechanism for establishing the safety of an agent. son Score of patients who had p53 assessed More Phase I/II, Phase II/III, or even Phase I/III was identical to those who did not have p53 trials are being conducted, the result of which assessed. However, the survival of those patients is that agents or combinations are being pushed who had a p53 determination performed on their into the multi-centre setting more rapidly that in tumour, and were thus included in the study, was the past. This warrants an examination of why significantly worse than those not included in caution is warranted as agents are taken from the study. Phase I testing to larger trials. This example demonstrates the possible pitfalls Phase I trials have a successful history as the of performing marker studies on patient subsets. first step in testing new cancer therapies. How- Despite such examples, the approach of collect- ever, several factors must be considered as a ing a set of patients with tissue available, testing new agent or combination of agents is taken a possible prognostic or predictive marker, and from a Phase I trial to a Phase II or Phase III reporting the results continues to be the method trial. These factors relate to possible differences by which most markers are examined. The rea- between patients entered onto Phase I trials and sons for this are many: expediency, ethical issues those entered on later trials. First, Phase I tri- of mandatory tissue submission, and policies of als often include patients with any type of solid informed consent and institutional review boards. tumour. This speeds the completion of the trial, It does raise the issue of whether for future and is justified because in many cases the tolera- prospective studies, tissue submission should be bility of an agent is not related to the location of mandatory prior to patient entry on a therapeu- the patient’s primary tumour. However, in some tic clinical trial. Such a discussion raises ethical instances the tolerability of an agent may dif- and legal issues that are beyond the scope of fer in patients with different tumour types. Sec- this work, but such discussions are ongoing for ond, Phase I trials are frequently conducted at 132 TEXTBOOK OF CLINICAL TRIALS highly specialised cancer centres, where medical report was prompted by the finding of an personnel experienced with detailed monitoring unexpected number of early deaths on two GI and rapid intervention reduce the consequences cancer trials, one in advanced colorectal cancer of and future risk for toxicity whenever possi- and the other in adjuvant cancer. In one of the ble. In contrast, Phase II and III trials are often trials in the review, 23 patients were reported conducted in the community setting, where the to have died within 60 days of initiation of clinical staff may be using a new agent or com- therapy. When these 23 events were reported bination of agents for the first time. to the group operations office for this trial, A final reason for caution is that patients are only 10 were deemed to have been related to enrolled onto Phase I trials only after failing all treatment. However, upon independent review, standard therapies. This has several implications. 16 of the 23 events were deemed treatment- Patients enrolled on Phase I trials have typically related or treatment-induced. been previously exposed to cytotoxic agents, thus Several conclusions can be drawn regard- they may be physically or emotionally more ing toxicity in clinical trials. First, as the pace robust and tolerate a new therapy better than a of drug development continues to quicken, it patient in the first-line setting. Additionally, only is likely that there will continue to be agents patients that have tolerated their initial therapy pushed into large trials prior to full and exten- acceptably are likely to choose enrollment, or sive Phase I and II testing. Second, effective be considered candidates, for a Phase I trial. A systems must be established to monitor toxic- final issue with respect to patients who have ity in trials of all phases. Third, an independent been previously treated is that patients and their assessment of the attribution of an event may physicians may have a sense of the rate of disease be beneficial, as local investigators may be hes- progression such that patients with the most itant to attribute a poor event to a treatment. virulent disease are often not even considered for In addition, new metrics (such as 60-day all- 85 enrollment in a Phase I trial. cause mortality ) and new terminology (such In part due to the combination of these various as treatment-induced, treatment-exacerbated and 84 factors, study sponsors, including the National treatment-unrelated deaths ) may be helpful in Cancer Institute, have developed sophisticated standardising the reporting of adverse events in systems to aid in the timely identification of clinical trials. toxicity problems in ongoing studies. Despite these systems, gaps remain. For example for CASE STUDY: 5-FU PLUS LEUCOVORIN agents that are commercially available, expedited IN COLON CANCER reporting of severe but expected events may not be required. If such an event were occurring at As is clear, the history of clinical trials in GI a greater frequency than expected, and expedited cancer is long and has been very successful. As reporting was not required, a multi-centre trial an example illustrating several facets of both may not detect such an incidence for weeks or the past history of GI clinical trials and issues months due to the nature of the process of data that will likely be faced again in future studies, collection, editing, entry and analysis. This has here we present a case study of the development, led some groups to propose supplements to the establishment and replacement of what was once standard systems to collect data on all severe the US standard of care for advanced colorectal events in a timely manner.83 cancer and, as of 2002, remains the standard for Even when data is reported in a timely manner, adjuvant stage 3 colon cancer, the ‘Mayo Clinic’ care must be taken in interpreting the data bolus regimen of 5-FU and leucovorin delivered received. As an example, consider the experience for 5 consecutive days every 4 or 5 weeks. in two large Phase III randomised trials reported The activity of fluorinated pyrimidines in the on by Rothenberg et al.84 The Rothenberg et al. treatment of GI cancers has been reviewed GASTROINTESTINAL CANCERS 133 extensively. 5-Fluorouracil (5-FU) is the most explore the efficacy of 5-FU and leucovorin in ubiquitous of the fluorinated pyrimidines, which inhibiting tumour growth. at least in part exert their antineoplastic effect by Early investigators studying the biochemical inhibiting the activity of the enzyme thymidylate modulation of 5-FU with leucovorin in the treat- synthase (TS), which in turn interferes with DNA ment of colorectal and gastric cancers included synthesis in dividing cells. Often agents designed Machover and colleagues.94,95 The Machover to improve the efficacy of fluorinated pyrimidines regimen consisted of administering high-dose are combined with these agents in an effort leucovorin at 200 mg/m2/d prior to 5-FU at a to preferentially sensitise tumour cells relative dose of 370 mg/m2/d, with both drugs given to host cells to the agent(s). Leucovorin is an consecutively for 5 days. With this dose of agent commonly used in such a setting. The leucovorin, the blood level is approximately Mayo regimen of 5-FU and leucovorin is thus a 10–20 µmol/L.96 In large part to lower the cost combination of an active chemotherapy agent, 5- of the regimen (leucovorin was very expensive FU, with a ‘biochemical modulator’ leucovorin. at the time), the ‘Mayo’ regimen was devised to Prior to the early 1980s, 5-FU was primarily use the identical 5-FU schedule to the Machover administered as a single agent. Administered regimen, but to use low-dose leucovorin at a dose in this fashion, it was associated with limited of 20 mg/m2/d, which resulted in blood levels of activity and moderate toxicity. Response rates 1–2 µmol/L. for metastatic colorectal cancer were low, in the This regimen was first tested as part of a ran- neighbourhood of 10%, and these responses were domised Phase II study in advanced unresectable short-lived, lasting on average a few months.61 colorectal cancer.97 Three of the treatment arms Based on pre-clinical laboratory studies,86–88 are relevant for this discussion: (1) 5-FU as a sin- the addition of leucovorin to cell culture with one gle agent administered at a dose of 500 mg/m2/d of the metabolites of 5-FU, fluorodeoxuridylate by IV bolus for 5 consecutive days every monophosphate (FdUMP), resulted in enhanced 5 weeks; (2) the Machover regimen repeated at binding to and inhibition of TS as compared 4 weeks, 8 weeks and every 5 weeks thereafter; to the binding when FdUMP was used alone. and (3) the Mayo regimen repeated at the same This improved inhibition of thymidylate synthase frequency as the Machover regimen. In this trial, resulted in inhibited DNA synthesis and resulted provision was made in the protocol to escalate in enhanced tumour shrinkage. Depending on the the 5-FU dose on any treatment arm if there was model systems, optimal concentration of leucov- no observed myelosuppression or significant non- orin ranged from leucovorin 1–20 mmol/L.89–93 haematologic toxicity during the previous treat- These studies supported the use of leucovorin ment course. When the toxicity was analysed doses ranging from 10 to 600 mg/m2 in clini- after treatment of the first 100 patients, the start- cal trials where leucovorin was added to 5-FU ing dose of 5-FU for the Mayo regimen was in an effort to improve on 5-FU’s single agent increased to 425 mg/m2/d in order to produce activity. While such laboratory studies provided definite but tolerable toxicity that was of simi- basic information on the modulation of 5-FU lar magnitude between the six treatment arms.98 using leucovorin, the applicability of these results The original combination of low-dose leucovorin to humans with colorectal cancer was unclear. with 370 mg/m2/d of 5-FU for 5 consecutive Based on clinical experience, individuals with days was empiric; no formal Phase I trial of this colorectal cancer clearly exhibit significant het- regimen had ever been performed. In the 208 eli- erogeneity in their response to treatment. The gible patients entered on the three study arms of sequence of administration of 5-FU and leucov- interest, the overall response rates were 10% for orin, the optimal concentration of leucovorin, and 5-FU alone, 26% for the Machover regimen, and the appropriate interval of 5-FU and leucovorin 43% for the Mayo regimen. Both leucovorin reg- administration all were variables to be studied to imens demonstrated significant improvement in 134 TEXTBOOK OF CLINICAL TRIALS response rate and overall survival compared to were similar between the Mayo and RPMI 5-FU 5-FU alone. plus leucovorin programmes, and the 5-FU plus Concurrent to the previously mentioned study, both leucovorin and levamisole regimen.55 Based investigators at the Roswell Park Memorial on the essentially identical activity profiles of Cancer Institute (RPMI) began testing a reg- these regimens, the choice between the two 5- imen of leucovorin 500 mg/m2/d with 5-FU FU and leucovorin regimens (Mayo and RPMI) 600 mg/m2/d given for 6 consecutive weeks fol- has been based on issues related to schedule lowed by a 2-week rest period.99 In a small study, (some patients preferred weekly therapy over the RPMI regimen was shown to significantly five consecutive days of treatment), cost (at the improve the tumour response rate compared to time of these studies leucovorin was expensive), single agent 5-FU. Shortly thereafter, the RPMI toxicity profile and clinician’s preference. and Mayo regimens were compared in a ran- From the late 1980s until the year 2000, domised trial of 366 patients.100 In this trial, the the Mayo regimen of 5-FU and leucovorin was objective response rates and overall survival was regarded as the standard of care for advanced similar between the two arms. The toxicity profile colon cancer. As discussed previously, in the late of the two regimens did differ, but no clear win- 1990s and early 2000s, several randomised tri- ner was identified. Based largely on cost consid- als were conducted in both the US and Europe in erations, investigators from the Mayo Clinic and which infusion-based 5-FU regimens or regimens the North Central Cancer Treatment group chose that combine 5-FU with CPT-11 or oxaliplatin to pursue the Mayo regimen for future testing. have demonstrated improved patient outcomes The activity seen with the combination of compared to those seen with the Mayo regi- leucovorin and 5-FU in the advanced disease men. In addition, the oral agent capecitabine setting naturally led to the evaluation of several has been approved as an alternative to IV 5-FU of these regimens in the adjuvant treatment of in advanced disease. Thus it appears that in patients with stage 2 and 3 colon cancer. In the advanced disease setting, the Mayo 5-FU a study that was suspended after accrual of + leucovorin regimen has been replaced as the 317 patients (based on the results of a large standard of care, indeed a welcome advance. trial that demonstrated 5-FU plus levamisole was In the adjuvant setting, no randomised trial has an effective treatment in this setting49), patients been completed that demonstrates improved over- with resected stage 2 or 3 colon cancer were all survival for a multiple drug combination, or randomised to the Mayo 5-FU plus leucovorin equivalence for an oral regimen, although trials in regimen for 6 months or to a no treatment control both areas have completed accrual and are await- arm.53 The 5-year survival for treated patients ing results. Increased toxicity has clearly been was 74%, compared to 63% in the control group demonstrated for multiple drug combinations in 84 (p = 0.02). This result established the efficacy the adjuvant setting, demonstrating the value of of the Mayo 5-FU plus leucovorin regimen in waiting for the results from these definitive tri- the adjuvant setting. als before adopting a promising new therapy as Following this small study, a large trial was a standard of care in the community. conducted to test four different combinations of 5-FU with leucovorin and/or levamisole in patients with stage 2 and 3 colon cancer. REFERENCES The regimens included the Mayo 5-FU plus leucovorin regimen for 6 months, 5-FU plus 1. Launois B, Delarue D, Campion JP, Kerbaol M. levamisole for 12 months, 5-FU with high-dose Preoperative radiotherapy for carcinoma of the esophagus. Surg, Gynecol Obstet (1981) 153: leucovorin (the RPMI regimen) for 8 months, 690–2. or 5-FU plus leucovorin plus levamisole for 2. Gignoux M, Roussel A, Paillot B, Gillet M, 12 months. In this study of 3759 patients, results Schlag P, Favre J-P, Dalesio O, Buyse M, GASTROINTESTINAL CANCERS 135
Duez N. The value of preoperative radiotherapy 13. Cuschieri A, Weeden S, Fielding J, Bancewicz J, in esophageal cancer: results of a study of the Craven J, Joypaul V, Sydes M, Fayers P. Patient E.O.R.T.C. World J Surg (1987) 11: 426–32. survival after D1 and D2 resections for gastric 3. Kelsen DP, Ginsberg R, Pajak TF, Sheahan DG, cancer: long-term results of the MRC random- Gunderson L, Mortimer J, Estes N, Haller DG, ized surgical trial. Surgical Cooperative Group. Ajani J, Kocha W, Minsky BD, Roth JA. Che- Br J Cancer (1999) 79: 1522–30. motherapy followed by surgery compared with 14. Janunger KG, Hafstrom L, Nygren P, Gli- surgery alone for localized esophageal cancer. melius B. A systematic overview of chemother- New Engl J Med (1998) 339: 1979–84. apy effects in gastric cancer. Acta Oncol (2001) 4. Kelsen DP, Minsky BD, Smith M, Beitler J, 40: 309–26. Niedzwiecki D, Chapman D, Bains M, Burt M, 15. Macdonald JS, Smalley SR, Benedetti J, Hun- Heelan R, Hilaris B. Preoperative therapy for dahl SA, Estes NC, Haller DG, Ajani JA, Gun- esophageal cancer: a randomized comparison of derson LL, Jessup JM, Martenson JA. Chemora- chemotherapy versus radiation therapy. J Clin diotherapy after surgery compared with surgery Oncol (1990) 8: 1352–61. alone for adenocarcinoma of the stomach or gas- 5. Bosset J-F, Gignoux M, Triboulet J-P, Tiret E, troesophageal junction. New Engl J Med (2001) Mantion G, Elias D, Lozach P, Ollier J-C, Pavy 345: 725–30. J-J, Mercier M, Sahmoud T. Chemoradiotherapy 16. Webb A, Cunningham D, Scarffe JH, Harper P, followed by surgery compared with surgery alone Norman A, Joffe JK, Hughes M, Mansi J, Find- in squamous-cell cancer of the esophagus. New lay M, Hill A, Oates J, Nicolson M, Hickish T, Engl J Med (1997) 337: 161–7. O’BrienM,IvesonT,WatsonM,UnderhillC, 6. Le Prise E, Etienne PL, Meunier B, Maddern G, Wardley A, Meehan M. Randomized trial com- Hassel MB, Gedouin D, Boutin D, Campion JP, paring epirubicin, cisplatin, and fluorouracil ver- Launois B. A randomized study of chemother- sus fluorouracil, doxorubicin, and methotrexate apy, radiation therapy, and surgery versus surgery in advanced esophagogastric cancer. J Clin Oncol for localized squamous cell carcinoma of the (1997) 15: 261–7. esophagus. Cancer (1994) 73: 1779–84. 17. Moertel CG, Frytak S, Hahn RG, O’Connell MJ, 7. Urba SG, Orringer MB, Turrisi A, Iannettoni M, Reitemeier RJ, Rubin J, Schutt AJ, Weiland LH, Forastiere A, Strawderman M. Randomized trial Childs DS, Holbrook MA, Lavin PT, Livs- of preoperative chemoradiation versus surgery tone E, Spiro H, Knowlton A, Kalser M, Bar- alone in patients with locoregional esophageal carcinoma. J Clin Oncol (2001) 19: 305–13. kin J, Lessner H, Mann-Kaplan R, Ramming K, 8. Walsh TN, Noonan N, Hollywood D, Kelly A, Douglas HO, Thomas P, Nave H, Bateman J, Keeling N, Hennessy TPJ. A comparison of Lokich J, Chaffey J, Corson JM, Zamcheck N, multimodal therapy and surgery for esophageal Noval JW. Therapy of locally unresectable pan- adenocarcinoma. New Engl J Med (1996) 335: creatic carcinoma: a randomized comparison of high dose (6000 Rads) radiation alone, moder- 462–7. + − 9. Herskovic A, Martz K, Al-Sarraf M, Leich- ate dose radiation (4000 Rads 5 fluorouracil), + − man L, Brindle J, Vaitkevicius V, Cooper J, and high dose radiation 5 fluorouracil. Cancer Byhardt R, Davis L, Emami B. Combined che- (1981) 48: 1705–10. motherapy and radiotherapy compared with 18. Klaassen DJ, MacIntyre JM, Catton GE, Eng- radiotherapy alone in patients with cancer of the strom PF, Moertel CG. Treatment of locally esophagus. New Engl J Med (1992) 326: 1593–8. unresectable cancer of the stomach and pancreas: 10. Minsky BD, Berkey B, Kelsen DK, Ginsberg R, a randomized comparison of 5-fluorouracil alone Pisansky T, Martenson J, Komaki R, Okawara G, with radiation plus concurrent and maintenance Rosenthal, S. Preliminary results of intergroup 5-fluourouracil – an Eastern Cooperative Oncol- INT 0123 randomized trial of combined modality ogy Group Study. J Clin Oncol (1985) 3: 373–8. therapy (CMT) for esophageal cancer: standard 19. Gastrointestinal Tumor Study Group. Treatment vs. high dose radiation therapy. Proc Am Soc Clin of locally unresectable carcinoma of the pan- Oncol (2000) 19: 239a (abstr. 927). creas: comparison of combined-modality therapy 11. Jemal A, Thomas A, Murray T, Thun M. Cancer (chemotherapy plus radiotherapy) to chemother- statistics, 2000. CA: a Cancer J Clin (2002) 52: apy alone. J Natl Cancer Inst (1988) 80: 751–5. 23–47. 20. Neoptolemos JP, Dunn JA, Stocken DD, Al- 12. Bonenkamp JJ, Hermans J, Sasako M, van de mond J, Link K, Beger H, Bassi C, Falconi M, Velde JH. Extended lymph node dissection for Pederzoli P, Dervenis C, Fernandez-Cruz L, gastric cancer. New Engl J Med (1999) 340: Lacaine F, Pap A, Spooner D, Kerr DJ, Friess H, 908–14. Buchler MW, Members of the European Study 136 TEXTBOOK OF CLINICAL TRIALS
Group for Pancreatic Cancer. Adjuvant chemora- 29. Hofstad B, Almendingen K, Vatn M, Ander- diotherapy and chemotherapy in resectable pan- sen SN, Owen RW, Larsen S, Osnes M. Growth creatic cancer: a randomised controlled trial. and recurrence of colorectal polyps: a double- Lancet (2001) 358: 1576–85. blind 3-year intervention with calcium and 21. Burris HA, Moore MJ, Andersen J, Green MR, antioxidants. Digestion (1998) 59: 148–56. Rothenberg ML, Modiano MR, Cripps MC, Port- 30. McKeown-Eyssen G, Holloway C, Jazmaji V, enoy RK, Storniolo AM, Tarassoff P, Nelson R, Bright-See E, Dion P, Bruce WR. A randomized Dorr FA, Stephens CD, Von Hoff DD. Improve- trial of vitamins C and E in the prevention ments in survival and clinical benefit with gem- of recurrence of colorectal polyps. Cancer Res citabine as first-line therapy for patients with (1988) 48: 4701–5. advanced pancreas cancer: a randomized trial. J 31. Greenberg ER, Baron JA, Tosteson TD, Free- Clin Oncol (1997) 15: 2403–13. man DH Jr, Beck GJ, Bond JH, Colacchio TA, 22. Moore MJ, Hamm J, Eisenberg P, Dagenais M, Coller JA, Frankl HD, Haile RW. A clinical trial Hagan K, Fields A, Greenberg B, Schwartz B, of antioxidant vitamins to prevent colorectal ade- Ottaway J, Zee B, Seymour L. A comparison noma. Polyp Prevention Study Group. New Engl between gemcitabine (GEM) and the matrix JMed(1994) 331: 141–7. metalloproteinase (MMP) inhibitor BAY12-9566 32. Bergsma-Kadijk JA, van’t Veer P, Kampman E, (9566) in patients (pts) with advanced pancreatic Burema J. Calcium does not protect against cancer. Proc Am Soc Clin Oncol (2000) 19: 240a colorectal neoplasia. Epidemiology (1996) 7: (abstr. 930). 590–7. 23. McKeown-Eyssen GE, Bright-See E, Bruce WR, 33. Baron JA, Beach M, Mandel JS, van Stolk RU, Jazmaji V, Cohen LB, Pappas SC, Saibil FG. A Haile RW, Sandler RS, Rothstein R, Summers randomized trial of a low fat high fibre diet in the RW, Snover DC, Beck GJ, Bond JH, Green- recurrence of colorectal polyps. Toronto Polyp berg ER. Calcium supplements for the prevention Prevention Group. J Clin Epidemiol (1994) 47: of colorectal adenomas. Calcium Polyp Preven- 525–36. tion Study Group. New Engl J Med (1999) 340: 24. MacLennan R, Macrae F, Bain C, Battistutta D, 101–7. Chapuis P, Gratten H, Lambert J, Newland RC, 34. The Women’s Health Initiative Study Group. Ngu M, Russell A. Randomized trial of intake of Design of the women’s health initiative clinical fat, fiber, and beta carotene to prevent colorec- trial and observational study. Contr Clin Trials tal adenomas. The Australian Polyp Prevention 19 Project. J Natl Cancer Inst (1995) 87: 1760–6. (1998) : 61–109. 25. Bonithon-Kopp C, Kronborg O, Giacosa A, 35. Gann PH, Manson JE, Glynn RJ, Buring JE, Rath U, Faivre J. Calcium and fibre supplemen- Hennekens CH. Low-dose aspirin and incidence tation in prevention of colorectal adenoma recur- of colorectal tumors in a randomized trial. JNatl rence: a randomised intervention trial. European Cancer Inst (1993) 85: 1220–4. Cancer Prevention Organisation Study Group. 36. Sturmer T, Glynn RJ, Lee IM, Manson JE, Bur- Lancet (2000) 356: 1300–6. ing JE, Hennekens CH. Aspirin use and colorec- 26. Alberts DS, Martinez ME, Roe DJ, Guillen- tal cancer: post-trial follow-up data from the Rodriguez JM, Marshall JR, van Leeuwen JB, Physicians’ Health Study. Ann Int Med (1998) Reid ME, Ritenbaugh C, Vargas PA, Bhatta- 128: 713–20. charyya AB, Earnest DL, Sampliner RE. Lack 37. Steinbach G, Lynch PM, Phillips RK, Wallace of effect of a high-fiber cereal supplement on MH, Hawk E, Gordon GB, Wakabayashi N, the recurrence of colorectal adenomas. Phoenix Saunders B, Shen Y, Fujimura T, Su LK, Levin Colon Cancer Prevention Physicians’ Network. B. The effect of celecoxib, a cyclooxygenase- New Engl J Med (2000) 342: 1156–62. 2 inhibitor, in familial adenomatous polyposis. 27. Schatzkin A, Lanza E, Corle D, Lance P, Iber F, New Engl J Med (2000) 342: 1946–52. Caan B, Shike M, Weissfeld J, Burt R, Cooper 38. Jorgensen OD, Kronborg O, Fenger C. A ran- MR, Kikendall JW, Cahill J. Lack of effect of domised study of screening for colorectal cancer a low-fat, high-fiber diet on the recurrence using faecal occult blood testing: results after of colorectal adenomas. Polyp Prevention Trial 13 years and seven biennial screening rounds. Study Group. New Engl J Med (2000) 342: Gut (2002) 50: 29–32. 1149–55. 39. Scholefield JH, Moss S, Sufi F, Mangham CM, 28. Ponz de Leon M, Roncucci L. Chemoprevention Hardcastle JD. Effect of faecal occult blood of colorectal tumors: role of lactulose and of screening on mortality from colorectal cancer: other agents. Scand J Gastroenterol Suppl (1997) results from a randomised controlled trial. Gut 222: 72–5. (2002) 50: 840–4. GASTROINTESTINAL CANCERS 137
40. Mandel JS, Church TR, Bond JH, Ederer F, 51. International Multicentre Pooled Analysis of Geisser MS, Mongin SJ, Snover DC, Schu- Colon Cancer Trials (IMPACT) Investigators. man LM. The effect of fecal occult-blood screen- Efficacy of adjuvant fluorouracil and folinic acid ing on the incidence of colorectal cancer. New in colon cancer. Lancet (1995) 345: 939–44. Engl J Med (2000) 343: 1603–7. 52. Francini G, Petrioli R, Lorenzini L, Mancini S, 41. Winawer SJ, Fletcher RH, Miller L, Godlee F, Armenio S, Tanzini G, Marsili S, Aquino A, Stolar MH, Mulrow CD, Woolf SH, Glick SN, Marzocca G, Civitelli S, Mariani L, De Sando D, Ganiats TG, Bond JH, Rosen L, Zapka JG, Olsen Bovenga S, Lorenz M. Folinic acid and 5- SJ, Giardiello FM, Sisk JE, Van Antwerp R, fluorouracil as adjuvant chemotherapy in colon Brown-Davis C, Marciniak DA, Mayer RJ. Col- cancer. Gastroenterology (1994) 106: 899–906. orectal cancer screening: clinical guidelines 53. O’Connell MJ, Mailliard JA, Kahn MJ, Mac- and rationale. Gastroenterology (1997) 112: donald JS, Haller DG, Mayer RJ, Wieand HS. 594–642. Controlled trial of fluorouracil and low-dose 42. Smith RA, Cokkinides V, von Eschenbach AC, leucovorin given for 6 months as postoperative Levin B, Cohen C, Runowicz CD, Sener S, adjuvant therapy for colon cancer. J Clin Oncol Saslow D, Eyre HJ. American Cancer Society (1997) 15: 246–50. guidelines for the early detection of cancer. CA: 54. QUASAR Collaborative Group. Comparison of a Cancer J Clin (2002) 52: 8–22. fluorouracil with additional levamisole, higher- 43. Weissfeld JL, Ling BS, Schoen RE, Bresalier dose folinic acid, or both, as adjuvant chemother- RS, Riley T, Prorok PC. Adherence to repeat apy for colorectal cancer: a randomised trial. screening flexible sigmoidoscopy in the Prostate, Lancet (2000) 355: 1588–96. Lung, Colorectal, and Ovarian (PLCO) Cancer 55. Haller DG, Catalano PJ, Macdonald JS. Fluo- Screening Trial. Cancer (2002) 94: 2569–76. rouracil (FU), leucovorin (LV) and levamisole 44. UK Flexible Sigmoidoscopy Screening Trial (LEV) adjuvant therapy for colon cancer: five- Investigators. Single flexible sigmoidoscopy year final report of INT-0089. Proc Am Soc Clin screening to prevent colorectal cancer: baseline Oncol (1998) 17: 256a (abstr. 982). findings of a UK multicentre randomised trial. 56. Poplin E, Benedetti N, Estes D, Haller DG, Lancet (2002) 359: 1291–300. Mayer R, Goldberg R, Macdonald J. Phase III 45. Ness RM, Holmes AM, Klein R, Dittus R. Cost- randomized trial of bolus 5-FU/leucovorin/ utility of one-time colonoscopic screening for levamisole versus 5-FU continuous/infusion/ colorectal cancer at various ages. Am J Gastroen- terol (2000) 95: 1800–11. levamisole as adjuvant therapy for high risk 46. Winawer SJ, Zauber AG, Church TR, Mandel- colon cancer (SWOG 9415/INT-0153). Proc Am son M, Feld A, Bond JH. National Colonoscopy Soc Clin Oncol (2000) 19: 240a. Study (NCS) preliminary findings: a random- 57. Wolmark N, Fisher B, Rockette H. Postoperative ized clinical trial of general population screening adjuvant chemotherapy or BCG for colon cancer: colonoscopy. Gastroenterology (2002) T1560. results from NSABP protocol C-01. JNatl 47. Stocchi L, Nelson H. Laparoscopic colectomy Cancer Inst (1988) 80: 30–6. for colon cancer: trial update. J Surg Oncol 58. Sargent DJ, Goldberg RM, Jacobson SD, Mac- (1998) 68: 255–67. donald JS, Labianca R, Haller DG, Shepherd LE, 48. Laurie JA, Moertel CG, Fleming TR, Wieand Seitz JF, Francini G. A pooled analysis of adju- HS, Leigh JE, Rubin J, McCormack GW, Gerst- vant chemotherapy for resected colon cancer in ner JB, Krook JE, Mailliard J, Twito DI, Morton elderly patients. New Engl J Med (2001) 345: RF, Tschetter LK, Barlow JF. Surgical adjuvant 1091–7. therapy of large-bowel carcinoma: an evalua- 59. Mamounas E, Wieand S, Wolmark N, Bear HD, tion of levamisole and the combination of lev- Atkins JN, Song K, Jones J, Rockette H. Com- amisole and fluorouracil. J Clin Oncol (1989) 7: parative efficacy of adjuvant chemotherapy in 1447–56. patients with Dukes’ B versus Dukes’ C colon 49. Moertel CG, Fleming TR, Macdonald JS, Haller cancer: results from four National Surgical Adju- DG, Laurie JA, Goodman PJ, Ungerleider JS, vant Breast and Bowel Project Adjuvant Stud- Emerson WA, Tormey DC, Glick JH, Veeder ies (C-01, C-02, C-03, and C-04). J Clin Oncol MH, Mailliard JA. Levamisole and fluorouracil (1999) 17: 1349–55. for adjuvant therapy of resected colon carcinoma. 60. International Multicentre Pooled Analysis of B2 New Engl J Med (1990) 322: 352–8. Colon Cancer Trials (IMPACT B2) Investigators. 50. NIH Consensus Conference. Adjuvant therapy Efficacy of adjuvant fluorouracil and folinic acid for patients with colon and rectal cancer. JAm in B2 colon cancer. J Clin Oncol (1999) 17: Med Assoc (1990) 264: 1444–50. 1356–63. 138 TEXTBOOK OF CLINICAL TRIALS
61. Advanced Colorectal Cancer Meta-analysis Pro- fluorouracil with or without oxaliplatin as first- ject. Modulation of fluorouracil by leucovorin line treatment in advanced colorectal cancer. J in patients with advanced colorectal cancer: Clin Oncol (2000) 18: 2938–47. evidence in terms of response rate. J Clin Oncol 71. Goldberg RM, Morton RF, Sargent DJ, Fuchs C, (1992) 10: 896–903. Ramanathan S, Williamson S, Findlay B, 62. Meta-Analysis Group in Cancer. Efficacy of NCCTG, CALGB, ECOG, SWOG, NCIC. intravenous continuous infusion of fluorouracil N9741 oxaliplatin (Oxal) or CPT-11 + 5- compared with bolus administration in advanced fluorouracil (5FU)/leucovorin (LV) or oxal + colorectal cancer. J Clin Oncol (1998) 16:301–8. CPT-11 in advanced colorectal cancer (CRC). 63. Meta-Analysis Group in Cancer. Toxicity of Initial toxicity and response data from a GI fluorouracil in patients with advanced colorectal intergroup study. Proc Am Soc Clin Oncol (2002) cancer: effect of administration schedule and 21: 128a. prognostic factors. J Clin Oncol (1998) 16: 72. Krook JE, Moertel CG, Gunderson LL, Wie- 3537–41. and HS, Collins RT, Beart RW, Kubista TP, 64. Hoff PM, Ansari R, Batist G. Comparison of Poon MA, Meyers WC, Mailliard JA, Twito DI, oral capecitabine versus intravenous fluorouracil Morton RF, Veeder MH, Witzig TE, Cha S, Vid- plus leucovorin as first-line treatment in 605 yarthi SC. Effective surgical adjuvant therapy patients with metastatic colorectal cancer: results for high-risk rectal carcinoma. New Engl J Med of a randomized phase III study. J Clin Oncol (1991) 324: 709–15. (2001) 19: 2282–92. 73. O’Connell MJ, Martenson JA, Wieand HS, 65. Van Cutsem E, Twelves C, Cassidy J. Oral Krook JE, Macdonald JS, Haller DG, Mayer RJ, capecitabine compared with intravenous flu- Gunderson LL, Rich TA. Improving adjuvant orouracil plus leucovorin in patients with therapy for rectal cancer by combining protracted- metastatic colorectal cancer: results of a large infusion fluorouracil with radiation therapy after New Engl J Med 331 phase III study. J Clin Oncol (2001) 19: curative surgery. (1994) : 502–7. 4097–106. 74. Fisher B, Wolmark N, Rockette H, Redmond C, 66. Rougier P, Van Cutsem E, Bajetta E. Randomised Deutsch M, Wickerham DL. Postoperative adju- trial of irinotecan versus fluorouracil by continu- vant chemotherapy or radiation therapy for rectal ous infusion after fluorouracil failure in patients cancer: results from NSABP Protocol R-01. J with metastatic colorectal cancer. Lancet (1998) Natl Cancer Inst (1988) 80: 21–9. 352: 1407–12. 75. Wolmark N, Wieand HS, Hyams DM, Colan- 67. Cunningham D, Pyrhonen S, James RD. Ran- gelo L, Dimitrov NV, Romond EH, Wexler M, domised trial of irinotecan plus supportive care Prager D, Cruz AB Jr, Gordon PH, Petrelli versus supportive care alone after fluorouracil NJ, Deutsch M, Mamounas E, Wickerham DL, failure for patients with metastatic colorectal can- Fisher ER, Rockette H, Fisher B. Random- cer. Lancet (1998) 352: 1413–18. ized trial of postoperative adjuvant chemotherapy 68. Saltz LB, Cox JV, Blanke C, Rosen LS, Fehren- with or without radiotherapy for carcinoma of the bacher L, Moore MJ, Maroun JA, Ackland SP, rectum: National Surgical Adjuvant Breast and Locker PK, Pirotta N, Elfring GL, Miller LL. Bowel Project Protocol R-02. J Natl Cancer Inst Irinotecan plus fluorouracil and leucovorin for (2000) 92: 388–96. metastatic colorectal cancer. New Engl J Med 76. Swedish Rectal Cancer Trial. Improved survival (2000) 343: 905–14. with preoperative radiotherapy in resectable rec- 69. Douillard JY, Cunningham D, Roth AD, tal cancer. New Engl J Med (1997) 336: 980–7. Navarro M, James RD, Karasek P, Jandik P, 77. Sauer R, Fietkau R, Wittekind C, Martus P, Iveson R, Carmichael J, Alakl M, Gruia G, Rodel C, Hohenberger W, Jatzko G, Sabitzer H, Awad L, Rougier P. Irinotecan combined with Karstens JH, Becker H, Hess C, Raab R. Adju- fluorouracil compared with fluorouracil alone vant versus neoadjuvant radiochemotherapy for as first-line treatment for metastatic colorectal locally advanced rectal cancer. A progress cancer: a multicentre randomised trial. Lancet report of a phase-III randomized trial (protocol (2000) 355: 1041–7. CAO/ARO/AIO-94). Strahlenther Onkol (2001) 70. De Gramont A, Figer A, Seymour M, Home- 177: 173–8. rin M, Hmissi A, Cassidy J, Boni C, Cortes- 78. Enker WE, Thaler HT, Cranor ML, Polyak T. Funes H, Cervantes A, Freyer G, Papamichael Total mesorectal excision in the operative treat- D, Le Bail N, Louvet C, Hendler D, de Braud F, ment of carcinoma of the rectum. J Am Coll Surg Wilson C, Morvan F, Bonetti A. Leucovorin and (1995) 181: 335–46. GASTROINTESTINAL CANCERS 139
79. Kapiteijn E, Marijnen CAM, Nagtegaal ID, Put- covorin administration, plasma concentrations of ter H, Steup WH, Wiggers T, Rutten HJT, Pahl- reduced folates, and pools of 5,10 methylenete- man L, Glimelius B, Han J, van Krieken HJM, trahydrafolates and tetrahydrafolates in human Leer JWH, van de Velde JH. Preoperative radio- colon adenocarcinoma xenografts. Cancer Res therapy combined with total mesorectal excision (1990) 50: 3493–502. for resectable rectal cancer. New Engl J Med 91. Houghton JA, Williams LG, Loftin SK. Fac- (2001) 345: 638–46. tors that influence the therapeutic activity of 80. Hammond MEH, Taube SE. Issues and barri- 5-fluorouracil [6RS] leucovorin combinations ers to development of clinically useful tumor in colon adenocarcinoma xenografts. Cancer markers: a development pathway proposal. Semin Chemother Pharmacol (1992) 30: 423–32. Oncol (2002) 29: 213–21. 92. Houghton JA, Williams LG, Cheshire PJ. Influ- 81. Pajak TF, Clark GM, Sargent DJ, McShane LM, ence of the dose of [6RS] leucovorin on Hammond MEH. Statistical issues in tumor reduced folate pools and 5-fluorouracil-mediated marker studies. Arch Pathol Lab Med (2000) 124: thymidylate synthase inhibition in human colon 1011–15. adenocarcinoma xenografts. Cancer Res (1990) 82. Grignon D, Caplan R, Sarkar F. p53 status and 50: 3940–6. prognosis of locally advanced prostatic adenocar- 93. Cao S, Frank C, Rustum YM. Role of fluo- cinoma. J Natl Cancer Inst (1997) 89: 158–65. ropyrimidine scheduling and (6R,S) leucovorin 83. Sargent DJ, Goldberg RM, Mahoney MR, Hill- dose in a preclinical animal model of colorec- man DW, McKeough T, Hamilton SF, Darcy JM, tal carcinoma. J Natl Cancer Inst (1996) 88: Anderson VL, Krook JE, O’Connell MJ. Rapid 430–6. reporting and review of an increased incidence 94. Machover D, Goldschmidt E, Chollet P. Treat- of a known adverse event. J Natl Cancer Inst ment of advanced colorectal and gastric ade- (2000) 92: 1011–13. nocarcinomas with 5-fluorouracil and high-dose 84. Rothenberg ML, Meropol NJ, Poplin EA, Van folinic acid. J Clin Oncol (1986) 4: 685–96. Cutsem E, Wadler S. Mortality associated with 95. Machover D, Schwartenberg L, Goldschmidt E. irinotecan plus bolus fluorouracil/leucovorin: Treatment of advanced colorectal and gastric summary findings of an independent panel. J Clin adenocarcinomas with 5-FU combined with high- Oncol (2001) 19: 3801–7. dose folinic acid: a pilot study. Cancer Treat Rep 85. Goldberg RM, Sargent DJ, Morton RF, Mahoney (1982) 66: 1803–7. MR, Krook JE, O’Connell MJ. Sensing toxicity 96. Rustum YM, Trave F, Zakrzewski SF. Biochem- and adjusting ongoing clinical trials protocols ical and pharmacologic basis for potentiation of to enhance patient safety: the history and per- 5-fluorouracil action by leucovorin. NCI Monogr formance of the NCCTG ‘Real Time Toxicity (1987) 5: 165–70. Monitoring Program’. J Clin Oncol (2002) 20: 97. Poon MA, O’Connell MJ, Moertel CG. Bio- 4591–6. chemical modulation of fluorouracil: evidence of 86. Ullman B, Lee M, Martin DW Jr, Santi DV. significant improvement in survival and quality Cytotoxicity of 5-fluoro-2-deoxyuridine: require- of life in patients with advanced colorectal car- ment for reduced folate co-factors and antago- cinoma. J Clin Oncol (1989) 7: 1407–18. nism by methotrexate. Proc Natl Acad Sci (1978) 98. O’Connell MJ. A phase III trial of 5-fluorouracil 75: 980–3. and leucovorin in the treatment of advanced 87. Evans RM, Laskin JD, Hakala MT. Effect of colorectal cancer: a Mayo Clinic/North Central excess folates and deoxyinosine on the activity Cancer Treatment Group study. Cancer (1989) and site of action of 5-fluorouracil. Cancer Res 63: 1026–30. (1981) 41: 3288–95. 99. Petrelli N, Herrara L, Rustum Y. A prospective 88. Waxman S, Bruckner H. The enhancement of 5- randomized trial of 5-fluorouracil vs. 5-flu- fluorouracil antimetabolic activity by leucovorin, orouracil and high dose leucovorin +5−fluoro- menadione, and alpha-tocopherol. Eur J Cancer uracil and methotrexate in previously untreated Clin Oncol (1982) 18: 685–92. patients with advanced colorectal carcinoma. J 89. Keyomarsi K, Moran RG. Folinic augmentation Clin Oncol (1987) 5: 1559–65. of the effects of fluoropyrimidines on murine and 100. Buroker TR, O’Connell MJ, Wieand HS. Ran- human leukemic cells. Cancer Res (1986) 46: domized comparison of two schedules of flu- 5229–35. orouracil and leucovorin in the treatment of 90. Houghton JA, Williams LG, Loftin SK. Rela- advanced colorectal cancer. J Clin Oncol (1994) tionship between the dose rate of [6RS] leu- 12: 14–20. 9 Haematologic Cancers
CHARLES A. SCHIFFER1 AND STEPHEN L. GEORGE2 1Karmanos Cancer Institute, Wayne State University School of Medicine, Detroit, MI 48201, USA 2Department of Biostatistics and Bioinformatics, Duke University Medical Centre, Durham, NC 27710, USA
INTRODUCTION of haematopoietic growth factors, stem cell transplantation in first remission and modula- tion of various mechanisms of intrinsic drug The story of therapeutic research for acute resistance.1–7 Some of these trials have, in fact, myeloid leukaemia (AML) is one of mixed changed the standard care of the disease but success. AML has been studied extensively both most have unfortunately been negative in that in the clinic and in the laboratory, in part because the newer approaches failed to improve remission of the accessibility of cells for in vitro testing, and rates and overall survival. Intensive treatment has served as a model for the elucidation of many of the principles of anti-neoplastic therapy and regimens have improved outcome in younger translational research in infectious disease and patients, but clinical trials in patients 60 years transfusion medicine supportive care. Complete and older by the Cancer and Leukemia Group B remission after initial therapy is achieved in (CALGB) and others have shown remarkably about two-thirds of patients, a significant fraction consistent poor results, with complete remis- of whom can be cured with additional post- sion (CR) rates of around 50%, and only 10% 1,8,9 remission treatment. Unfortunately, however, the of patients surviving four years. Indeed, it great majority of patients eventually relapse and is arguable that the prognosis of older patients succumb to complications of the disease and with AML has not changed in the last 15 years. its treatment. The incidence of AML spans the Therefore, it is imperative that new therapies are entire age spectrum but is most common in adults evaluated in as efficient a fashion as possible. greater than 60 years of age and the prognosis is There are a number of issues which can serve particularly poor in these individuals. as impediments to new drug development, some Large numbers of randomised trials have of which are idiosyncratic to AML. This chapter been performed in patients with AML includ- will review some of these problems and sug- ing comparative evaluations of different doses gest and discuss some of the statistical issues in and types of chemotherapeutic agents, the use trial design.
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 142 TEXTBOOK OF CLINICAL TRIALS
BACKGROUND cured of their leukaemia; however, less than 10% of patients greater than 60 years of age remain in 1,6,8,9 AML occurs as a consequence of an acquired long-term remission. Multiple randomised mutation in a haematopoietic stem cell which trials evaluating high-dose therapy with either results in a failure of normal maturation and dif- autologous or allogeneic stem cell rescue have ferentiation of myeloid cells and an accumulation produced results similar to those achieved with of juvenile leukaemia cells or ‘blasts’. By mecha- chemotherapy alone, although the causes of treat- nisms which are not well understood, the expan- ment failure vary somewhat, with lower rates sion of this malignant clone suppresses normal of relapse with transplant approaches offset by 6 blood cell formation and patients usually present higher treatment associated mortality. In addi- with symptoms related to absence of normal tion, transplant approaches are generally suitable blood cells including weakness and fatigue due to only for younger patients. anaemia, infection because of decreased number AML is also an extremely heterogeneous dis- of normal white blood cells, and bleeding because order biologically. Multiple subtypes can be of marked decreases in the number of platelets. identified by cytogenetic or molecular studies Because this is a systemic disease, initial treat- of the leukaemia cells. Some of these sub- ment is with chemotherapy, generally including types (predominantly found in younger patients) an anthracycline and cytarabine (ara-c), adminis- have an excellent prognosis with cure rates of tered for three and seven days respectively. This approximately 60%, whereas other subtypes, gen- induces a period of low blood counts for three erally characterised by chromosome loss and to four weeks at which time the patient is at risk duplication, have almost no patients cured with 11 for bleeding and infection and invariably requires chemotherapy alone. The latter is much more transfusions of red blood cells, platelets and common in older patients, particularly those in the use of systemic antibiotics. Should therapy whom AML developed as a progression of a be successful, a complete remission is obtained prior bone marrow disorder either of a myelo- which is defined as normal blood counts and bone proliferative nature or more commonly following marrow with no evidence of leukaemia.10 It is a myelodysplastic syndrome. known that small amounts of leukaemia remain, Cytogenetic and molecular characterisation of which cannot be detected morphologically with the type of AML can be critical as evidenced the microscope, because without further therapy, by the remarkable results achieved in recent leukaemia invariably relapses, generally within years in acute progranulocytic leukaemia (APL), six months. Remission rates are approximately a subgroup representing about 10% of patients 75% in younger patients but are only 50% in with AML, predominantly in younger adults older patients (greater than 60 years of age). and children.12 All patients with APL have a There have been major improvements in infec- balanced translocation between chromosomes 15 tious disease and transfusion supportive care so and 17, resulting in a mutation of the gene that, today, the major cause of initial treatment coding for the retinoic acid nuclear receptor. failure is drug-resistant leukaemia (i.e. persis- By complex mechanisms, this confers unique tence of leukaemia after treatment). Randomised sensitivity to an oral retinoid, all-trans retinoic trials comparing different types of anthracyclines, acid (ATRA), which has appreciably fewer side different doses and schedules of administration effects than traditional chemotherapy. A series of ara-c, and the use of growth factors for sup- of randomised trials have elucidated the optimal portive care, have not improved these induction means of combining ATRA with chemotherapy results. With appropriate post-remission therapy, such that more than 70% of patients with approximately 35% of patients less than 50 years this previously devastating leukaemia can be of age remain disease-free after three years, with cured.13 Interestingly, APL also has a unique almost all of these patients being functionally sensitivity to arsenical compounds. It is hoped HAEMATOLOGIC CANCERS 143 that similar strategies with different compounds However, there are a number of both practical can be discovered for other AML variants with and biologic issues complicating the conduct of discrete activating mutations, as has recently also such trials: been achieved in patients with chronic myeloid leukaemia (CML) with the use of the tyrosine • Evaluation of post-remission manipulations kinase inhibitor imatinib mesylate (STI571), is made more difficult by the low com- which specifically targets the abnormal enzyme plete response rate, so that less than 50% produced by the bcr/abl mutation characteristic of patients initially entered on trial are eli- of CML.14 gible for post-remission treatment. In addi- Some studies have suggested differential res- tion, many such patients are not candidates ponsiveness of AML subtypes to different types for intensive therapy because of compromised of chemotherapy. In particular, patients with organ function from toxicities encountered dur- more favourable balanced translocations seem ing induction, and because many older patients to benefit from high-dose ara-c-based consoli- do not recover fully normal blood counts even dation therapy. In contrast, patients with poor after a significant anti-leukaemic response dur- prognosis chromosomal changes do not bene- ing induction. fit from these more intensive chemotherapeutic • In addition, many older individuals decline approaches,15 although a fraction may be cured post-remission treatment, preferring to spend with the graft versus leukaemia effect induced their remaining time outside of the hospital, by allogeneic stem cell transplantation. Stem cell as far from aggressive medical ministrations transplant is currently not a possibility for older as possible. patients. Because of this, many treatment coop- erative groups have devised different therapeutic Thus, randomised studies of new therapies intro- approaches for older and younger patients, with duced post-remission need larger numbers of manipulations of stem cell transplantation being patients to account for this drop-off in patients evaluated in the latter group. as the study progresses. This represents a major issue since only a small fraction of such patients are captured for clinical trials. IMPROVING THERAPY FOR OLDER Furthermore: PATIENTS WITH AML • AML in older individuals is extremely hetero- Rates of complete remission are much lower and geneous. Some therapies might be appropriate remission duration more abbreviated in patients only for certain AML subtypes and positive greater than the age of 60 years as a consequence effects can be missed when tested in the over- of more intrinsic drug resistance and more base- all AML population. This may be particularly line organ dysfunction than are encountered in true for newer targeted therapies. younger individuals. New therapeutic approaches • A focus on patients with highly resistant should focus both on increasing remission rates disease represents a particularly high hurdle as well as on prolonging remission and enhanc- for new therapies and treatments. Modest, but ing the cure fraction of such patients. Many AML nonetheless important, benefits which could studies have focused on older patients because be of value to other patient groups could be of the large numbers of such patients available missed by studying only patients in very poor for studies as well as the feeling that the overall prognostic groups. results of therapy are so poor that it would be pos- sible to identify truly active agents very rapidly These problems are particularly relevant, be- because differences with historical or randomised cause there is no shortage of new agents which controls would be obvious. could and should be evaluated in AML. In 144 TEXTBOOK OF CLINICAL TRIALS addition to a continued supply of cytotoxic on Phase II data alone which showed benefit in drugs, there will be large numbers of anti- patients with resistant disease and otherwise few angiogenesis compounds, immune modulators, therapeutic options. signal transduction inhibitors (either with specific or more generic enzymatic targets), as well as new and less acutely toxic approaches to stem cell STATISTICAL ISSUES IN DESIGN transplant. Many of the non-cytotoxic therapeutic AND ANALYSIS approaches also have the allure of oral treatment with potentially much less toxicity. Because of the nature of AML and its treat- If an agent can be safely added to the ment, several statistical issues in the design and usual dose of conventional therapy, it might analysis of clinical trials need special attention. be most efficient to utilise the new therapy in Four of these are discussed: factorial designs, both induction and consolidation, thereby perhaps outcome measures, competing risks and statisti- maximising the chance to detect anti-leukaemic cal modelling. activity. Possible study designs for trials of new post-remission therapies are shown in Table 9.1, FACTORIAL DESIGNS where conventional therapy might refer to a The treatment phases for AML are conventionally few courses of low-dose ara-c which results in divided into an initial phase of induction therapy a median remission duration of approximately and, for those achieving a complete remission eight months and less than 10% long-term during this phase, a post-remission or mainte- disease-free survival.9 This is slightly better than nance therapy phase. The post-remission phase observation without treatment which produces is sometimes further divided into earlier consol- very few if any long-term disease-free survivors idation therapy and later maintenance therapy, and shorter CR durations. The choice among but for our purposes here, two phases are suf- the various randomised approaches might be ficient. It is natural to design studies to compare influenced by the unique features of the agent therapies in each of these two phases, leading to being tested. Also, given the very poor results factorial designs, in which patients are randomly observed with standard therapy, it could be assigned to one of two or more induction thera- argued that a straightforward Phase II trial in pies (the first factor) and then to one of two or which the new agent is evaluated alone could more maintenance therapies (the second factor). have merit, although the usual problems with With two possible treatment assignments in each historical controls and patient selection would be phase, this is a 2 × 2 factorial design, a common issues. However, a number of anti-cancer agents and well-known statistical design. Much has been have been approved by the FDA in recent years written about this design applied in the clinical under an accelerated approval mechanism, based trials setting.16 The twist in the current situation is that the second randomisation is applicable only for patients who respond to the induction ther- Table 9.1. Possible study designs evaluating new apy. As noted above, in the case of older AML agents in post-remission therapy patients, only about 50% of all patients entered Phase II studies on study may respond and, thus, be eligible and New agent alone medically suitable for the second randomisation. Randomised Phase III studies It is typical to separate the objectives of such Observation vs. new agent studies into a comparison of induction regi- Conventional vs. new agent mens with respect to response rates and, sep- + Conventional vs. conventional new agent arately, a comparison of maintenance regimens Conventional followed by observation vs. new agent New agent in both induction and consolidation with respect to the length of remission, disease- free survival or overall survival. For example, HAEMATOLOGIC CANCERS 145
CALGB 8923 was a randomised clinical trial attention is focused on identifying activity of of this type involving AML patients at least an agent, no matter how small. Achievement of 60 years old.2,9 The induction phase involved a a complete response is sine qua non for long- randomisation between GM-CSF, a haematopoi- term control of disease. However, the CR rate etic growth factor, and placebo following initial is a very imperfect surrogate for more mean- chemotherapy. The hypothesis was that the GM- ingful clinical outcome measures described CSF would reduce infectious complications and below, and has been defined differently by dif- perhaps increase the response rate. Responding ferent leukaemia treatment groups, and should patients were to be randomised to receive one never be used as a substitute for them, espe- of two post-remission regimens, cytarabine alone cially in Phase III clinical trials. The primary or a combination of cytarabine and mitoxantrone. role for the CR rate is as a measure of clinical Overall, 388 participants were randomised to the activity in Phase II trials. induction therapies; 205 (53%) achieved a CR, • Event-free survival (EFS) – this is the time but only 169 (44%) were randomised in the post- from the start of study until a failure to remission phase. respond, relapse (for those achieving a CR) One of the problems with the usual approach to or death, whichever occurs first. This outcome these designs is that there is no direct estimation measure is a good measure of the overall or testing of the four possible treatment poli- control of disease from the start of therapy cies implied in the design. The policies are and combines the effects of induction and defined by selecting one of two induction ther- post-remission therapies. In a Phase III trial, apies followed by one of two post-remission all randomised patients contribute to any therapies, if a response is obtained and the analysis of EFS under the usual intent-to- patient consents to continue. One paper deals treat approach.18 Standard techniques for time- directly with this issue, making efficient use of to-event data (survival methods) are used in data from all patients.17 There are also method- design and analysis for EFS, and for the other ologic issues about when the randomisation to time-to-event endpoints defined next.19 the post-remission therapy should be done. For • Disease-free [or relapse-free] survival (DFS) – example, if both randomisations are done at the this is a standard outcome measure in trials time of study entry with a planned intent to of adjuvant therapy for solid tumours, but in treat analysis, then the inevitable (and antici- AML trials, DFS refers to the survival time pated) large patient drop-out can substantially spent free of disease. Thus, DFS is applicable complicate evaluation of the second therapeutic only to patients who achieve a CR. It is defined manoeuvre. as the time from achieving a CR to relapse or death, whichever occurs first. Since patients OUTCOME MEASURES who fail to achieve a CR are excluded, this measure is unsuitable as an overall assessment There are various choices for outcome measures of therapy. However, it is useful for compar- in clinical trials involving AML patients. The ing two or more post-remission therapies as primary ones are: long as it is recognised that the distribution of DFS is not representative of the result to be • Response rate – the proportion of patients who expected for all patients. achieve an initial clinical response to the induc- • Length of remission (LR) – the length of tion therapy is referred to as the response rate. remission is ordinarily defined as the time In older AML patients, as in all leukaemia from achieving CR to the time of relapse, with patients, the critical category is the complete deaths in remission counted as censored obser- response (CR) rate, although partial responses vations. This measure suffers from the same are sometimes included in Phase II trials where problems as DFS and, in addition, the usual 146 TEXTBOOK OF CLINICAL TRIALS
Kaplan–Meier estimation is no longer valid which treats other risks as independent censoring (see discussion below on competing risks). mechanisms, are inherently flawed. One way to • Overall survival (OS) – the time from the start properly account for the dependence is through of study to death is an obviously critical out- the use of the cumulative incidence curve, a come measure for any generally fatal disease topic that has been extensively explored in recent like AML in older adults. It has the virtue of years.23 being unambiguously defined and captures a result of obvious significance. However, there STATISTICAL MODELS are often difficulties in interpretation, partic- ularly if multiple therapies are given, or if Statistical models are heavily used in AML patients cross over to the alternative therapy trials. The usual time-to-event measures (EFS, after relapse. Nevertheless, the importance of DFS, OS) are often handled non-parametrically overall survival is so fundamental that it should in the primary analysis (e.g., Kaplan–Meier always be analysed, even if it is not used as the estimates, logrank tests, etc.), but the semi- primary outcome measure. parametric proportional hazards regression model • Other outcome measures – there are some is surely the most commonly used method to other measures occasionally used in AML adjust for covariates in the analysis.19 In addition, studies, particularly measures of quality of life because of the nature of AML, increasing interest (QOL).20 Some attempt to measure survival is being focused on so-called cure models, or related time to event measures adjusted for in which it is hypothesised that an unknown quality of life. For example, the Q-TWiST fraction (c) of patients are cured (or at least method discounts survival time spent with an will have long-term control of disease) and unacceptable level of adverse symptoms due the rest (1 − c) are not.24 Interest then focuses to treatment.21 Such methods attempt to quan- on estimating c, comparing the cure rates of tify the generally accepted notion that sim- various treatments, identifying factors predictive ply prolonging survival is not a sufficient of c, and identifying prognostic factors for the objective. Improved quality of life is equally time-to-failure in the patients not cured. In important. older patients with AML, this model has not been used as much due to the obviously low value of c, but it has been important in other COMPETING RISKS leukaemias. For some purposes in designing and analysing tri- als of therapy for older AML patients, it is infor- SUMMARY mative to use the techniques of competing risks analysis.22 That is, rather than using a composite Acute myeloid leukaemia in the older patient measure such as EFS, one can break this measure is a common and important disease which is into its constituent parts by considering the time relatively resistant to current therapies. Careful to each outcome separately. The term ‘compet- consideration of the particular characteristics of ing risks’ refers to various risks of failure, each this disorder is required when designing clinical competing with the others. This terminology orig- trials. There will be a large number of compounds inally arose in the context of analysing various available for evaluation in upcoming years and causes of death, but it is applicable more gener- it is desirable that such studies be conducted ally. The fundamental problem from a statistical using the most efficient and informative designs perspective is that the risks cannot be assumed to rapidly identify therapies which lengthen to be operating independently from each other. survival and increase the fraction of patients who Thus, methods which assume such independence, are cured. such as the Kaplan–Meier life table analysis, HAEMATOLOGIC CANCERS 147
REFERENCES RA. Phase III study of the multidrug resis- tance modulator PSC-833 in previously untreated patients 60 years of age and older with acute 1. Mayer RJ, Davis RB, Schiffer CA, Berg DT, myeloid leukemia: Cancer and Leukemia Group Powell BL, Schulman P, Omura GA, Moore JO, B study 9720. Blood (2002) 100: 1224–32. McIntyre OR, Frei E. 3. Intensive postremission 8. Schiffer CA, Dodge R, Larson RA. Long-term chemotherapy in adults with acute myeloid follow-up of Cancer and Leukemia Group B leukemia. Cancer and Leukemia Group B. N Engl studies in acute myeloid leukemia. Cancer (1997) 80 JMed(1994) 331: 896–903. : 2210–14. 2. Stone RM, Berg DT, George SL, Dodge RK, Pa- 9. Stone RM, Berg DT, George SL, Dodge RK, Pa- ciucci PA, Schulman PP, Lee EJ, Moore JO, Pow- ciucci PA, Schulman P, Lee EJ, Moore JO, Po- ell BL, Baer MR, Bloomfield CD, Schiffer CA. well BL, Schiffer CA. Granulocyte-macrophage Postremission therapy in older patients with de colony-stimulating factor after initial chemother- novo acute myeloid leukemia: a randomized trial apy for elderly patients with primary acute myel- comparing mitoxantrone and intermediate-dose ogenous leukemia. N Engl J Med (1995) 332: cytarabine with standard-dose cytarabine. Blood 1671–7. (2001) 98: 548–53. 3. Rowe JM, Andersen JW, Mazza JJ, Bennett JM, 10. Cheson BD, Cassileth PA, Head DR, Schiffer Paietta E, Hayes FA, Oette D, Cassileth PA, Sta- CA, Bennett JM, Bloomfield CD, Brunning R, dtmauer EA, Wiernik PH. A randomized placebo- Gale RP, Grever MR, Keating MJ. Report of controlled phase III study of granulocyte-macro- the National Cancer Institute-sponsored work- phage colony-stimulating factor in adult patients shop on definitions of diagnosis and response in (> 55 to 70 years of age) with acute myeloge- acute myeloid leukemia. J Clin Oncol (1990) 8: nous leukemia: a study of the Eastern Coopera- 813–19. tive Oncology Group (E1490). Blood (1995) 86: 11. Byrd JC, Mrozek K, Dodge RK, Carroll AJ, Ed- 457–62. wards CG, Arthur DC, Pettenati MJ, Patil SR, 4. Lowenberg B, Suciu S, Archimbaud E, Ossenkop- Rao KW, Watson MS, Koduru PR, Moore JO, pele G, Verhoef GE, Vellenga E, Wijermans P, Stone RM, Mayer RJ, Feldman EJ, Davey FR, Berneman Z, Dekker AW, Stryckmans P, Schou- Schiffer CA, Larson RA, Bloomfield CD, Can- ten H, Jehn U, Muus P, Sonneveld P, Dardenne cer and Leukemia Group. Pretreatment cytoge- M, Zittoun R. Use of recombinant GM-CSF dur- netic abnormalities are predictive of induction ing and after remission induction chemother- success, cumulative incidence of relapse, and apy in patients aged 61 years and older with overall survival in adult patients with de novo acute myeloid leukemia: final report of AML-11, acute myeloid leukemia: results from Cancer and a phase III randomized study of the Leukemia Leukemia Group B (CALGB 8461). Blood (2002) Cooperative Group of European Organisation 100: 4325–36. for the Research and Treatment of Cancer and 12. Tallman MS, Andersen JW, Schiffer CA, Appel- the Dutch Belgian Hemato-Oncology Cooperative baum FR, Feusner JH, Woods WG, Ogden A, Group. Blood (1997) 90: 2952–61. Weinstein H, Shepherd L, Willman C, Bloom- 5. Godwin JE, Kopecky KJ, Head DR, Willman CL, field CD, Rowe JM, Wiernik PH. All-trans reti- Leith CP, Hynes HE, Balcerzak SP, Appelbaum noic acid in acute promyelocytic leukemia: long- FR. A double-blind placebo-controlled trial of term outcome and prognostic factor analysis from granulocyte colony-stimulating factor in elderly the North American Intergroup protocol. Blood patients with previously untreated acute myeloid (2002) 100: 4298–302. leukemia: a Southwest oncology group study 13. Fenaux P, Chastang C, Chevret S, Sanz M, Dom- (9031). Blood (1998) 91: 3607–15. bret H, Archimbaud E, Fey M, Rayon C, Huguet 6. Cassileth PA, Harrington DP, Appelbaum FR, F, Sotto JJ, Gardin C. A randomized compari- Lazarus HM, Rowe JM, Paietta E, Willman C, son of all transretinoic acid (ATRA) followed Hurd DD, Bennett JM, Blume KG, Head DR, by chemotherapy and ATRA plus chemother- Wiernik PH. Chemotherapy compared with autol- apy and the role of maintenance therapy in ogous or allogeneic bone marrow transplantation newly diagnosed acute promyelocytic leukemia. in the management of acute myeloid leukemia The European APL Group. Blood (1999) 94: in first remission. N Engl J Med (1998) 339: 1192–200. 1649–56. 14. Druker BJ, Sawyers CL, Kantarjian H, Resta DJ, 7. Baer MR, George SL, Dodge RK, O’Loughlin Reese SF, Ford JM, Capdeville R, Talpaz M. KL, Minderman H, Caligiuri MA, Powell BL, Activity of a specific inhibitor of the BCR- Kolitz JE, Schiffer CA, Bloomfield CD, Larson ABL tyrosine kinase in the blast crisis of 148 TEXTBOOK OF CLINICAL TRIALS
chronic myeloid leukemia and acute lymphoblastic 19. Hosmer DW, Lemeshow S. Applied Survival Ana- leukemia with the Philadelphia chromosome. N lysis. New York: John Wiley & Sons (1999). Engl J Med (2001) 344: 1038–42. 20. Chow SC, Ki FY. Statistical issues in quality-of- 15. Bloomfield CD, Lawrence D, Byrd JC, Carroll A, life assessment. J Biopharm Stat (1996) 6: 37–48. Pettenati MJ, Tantravahi R, Patil SR, Davey FR, 21. Gelber RD, Goldhirsch A, Cole BF, Wieand HS, Berg DT, Schiffer CA, Arthur DC, Mayer RJ. Schroeder G, Krook JE. A quality-adjusted time Frequency of prolonged remission duration after without symptoms or toxicity (Q-TWiST) analysis high-dose cytarabine intensification in acute mye- of adjuvant radiation therapy and chemotherapy loid leukemia varies by cytogenetic subtype. Can- for resectable rectal cancer. J Natl Cancer Inst cer Res (1998) 58: 4173–9. (1996) 88: 1039–45. 22. Gooley TA, Leisenring W, Crowley J, Storer BE. 16. Peterson B, George SL. Sample size requirements Estimation of failure probabilities in the presence and length of study for testing interaction in a × of competing risks: new representations of old 2 k factorial design when time-to-failure is the estimators. Stat Med (1999) 18: 695–706. outcome. Contr Clin Trials (1993) 14: 511–22. 23. Klein JP, Rizzo JD, Zhang MJ, Keiding N. Sta- 17. Lunceford JK, Davidian M, Tsiatis AA. Estima- tistical methods for the analysis and presentation tion of survival distributions of treatment policies of the results of bone marrow transplants. Part I: in two-stage randomization designs in clinical tri- unadjusted analysis. Bone Marrow Transpl (2001) als. Biometrics (2002) 58: 48–57. 28: 909–15. 18. Lachin JM. Statistical considerations in the intent- 24. Ghitany ME, Maller R, Zhou S. Estimating the to-treat principle. Contr Clin Trials (2000) 21: proportion of immunes in censored samples: a 167–89. simulation study. Stat Med (1995) 14: 39–49. 10 Melanoma
P.-Y. LIU1 AND VERNON K. SONDAK2 1Fred Hutchinson Cancer Research Centre, Seattle, WA 98109 1024, USA 2University of Michigan Comprehensive Cancer Centre, Ann Arbor, MI 48109 0932, USA
INTRODUCTION and death from melanoma. Clinically localised melanomas are grouped into three prognostic Randomised Phase III clinical trials are the categories based on the thickness of the pri- gold standard for medical decision making, mary tumour as measured by the pathologist particularly in terms of adjuvant therapies where using a micrometer built into the microscope eye- a modest incremental benefit is sought. However, piece (Breslow’s thickness). Melanomas less than there is sometimes marked disagreement among 1.0 mm in thickness have an overall excellent clinicians in their interpretation of Phase III trial prognosis with relatively minimal intervention results. Nowhere is this more evident than in and are considered ‘low-risk’ lesions. Melanomas the arena of adjuvant therapy of resected ‘high- between 1.0 and 3.9 mm are considered to be risk’ melanoma. In this chapter, we will review intermediate risk, while melanomas 4.0 mm or the results of several key randomised trials and greater are considered ‘high-risk’ tumours. The attempt to reconcile at times conflicting clinical presence of ulceration of the primary tumour interpretations. increases the risk of metastasis and death within any given thickness category.1 The thickness is highly predictive of the risk BACKGROUND of regional lymph node metastasis, with nodal involvement in <5% of melanomas that are A basic familiarity with malignant melanoma is <1.0mmversus>30% in melanomas ≥4.0mm. required in order to understand the statistical and Intermediate-thickness melanomas have an inter- clinical issues presented herein. mediate risk of nodal spread, on the order of 20%. The prognosis of localised cutaneous melanoma The prognostic significance of the presence of is based on several well-defined factors. Patho- nodal metastasis far outweighs the significance of logic analysis of the primary tumour can predict tumour thickness: a thin or intermediate-thickness the likelihood of regional and distant metastasis melanoma with nodal metastases generally has
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 150 TEXTBOOK OF CLINICAL TRIALS a worse prognosis than a thick melanoma with from reactive nodes, but is still not able to identify negative nodes. Once nodal metastasis has been microscopic foci of melanoma in normal nodes.4 documented, the number of involved nodes is the Currently, neither PET nor CT are routinely rec- strongest predictor of subsequent outcome, along ommended for clinical staging. with the manner of detection of the metastasis. For patients with low-risk melanomas, i.e., Melanoma in clinically enlarged nodes portends those that are <1 mm in Breslow’s depth and a worse prognosis than melanoma in clinically have no evidence of ulceration or significant normal nodes.1 regression, clinical staging by physical exami- The mainstay of treatment for localised or nation is standard practice. Currently, surgical regionally metastatic melanoma is surgery. Ade- staging is used in the majority of patients with quate wide excision of the primary tumour site higher-risk lesions. For any patient with clinically (generally taking a margin of 1 to 2 cm of nor- evident nodal involvement, a complete therapeu- mal skin around the visible edge of the melanoma tic lymph node dissection is associated with cure or biopsy scar) is highly efficacious in controlling in about 20% to 40% of patients. disease at the primary site.2,3 Three main options are available for stag- SURGICAL STAGING BY COMPLETE ing regional nodes in patients with cutaneous (ELECTIVE) LYMPH NODE DISSECTION melanoma: clinical staging, surgical staging by complete (elective) lymph node dissection, and Elective removal of clinically normal regional surgical staging by sentinel lymph node biopsy. nodes identifies evidence of metastasis about 20% of the time, and is clearly a more accurate deter- CLINICAL STAGING minant of nodal status than clinical staging. Ret- rospective reviews suggested a survival advan- Physical examination is the mainstay of clini- tage for elective node dissection compared to cal staging of the regional nodes. Any palpa- clinical staging with subsequent therapeutic node ble lymph nodes that are ≥1cminmaximum dissection at the time of nodal recurrence.5 To diameter or very hard or fixed to adjacent struc- date, however, no prospective study has demon- tures must be considered highly suspicious for strated an overall survival advantage for elective metastatic involvement. Unfortunately, both the node dissection.3,6 Although the lack of a demon- specificity and sensitivity of physical examina- strated benefit is not the same as the demonstra- tion for detecting melanoma nodal metastases are tion of no benefit, elective dissection of clinically low. In muscular or obese patients, even rela- normal nodes is not considered standard practice tively large lymph node metastases can be missed for cutaneous melanoma at the present time. It on physical examination. Lymph nodes may be is clear, however, that elective node dissection enlarged after a biopsy procedure due to reactive results in durable regional disease control in the hyperplasia without containing metastasis. Most vast majority of patients, and failures within the importantly, metastatic involvement of normal- dissected nodal basin are quite uncommon. sized lymph nodes cannot be identified by phys- ical examination. SURGICAL STAGING BY SENTINEL LYMPH Radiologic studies–computed tomography (CT) NODE BIOPSY and positron emission tomography (PET)–are also available to clinically stage the regional nodes. Sentinel lymph node biopsy is based on the CT shares many of the deficiencies of physical concept that lymphatic fluid from an area of examination: enlarged nodes may not be malig- skin drains specifically to an initial node or nant, and normal-sized nodes harbouring metas- nodes (‘sentinel nodes’) prior to disseminating tases will be deemed normal. PET is more sensitive to other nodes in the same or nearby basins. than CT for differentiating melanoma-containing Morton et al. described a reliable method for MELANOMA 151 identification and removal of the sentinel node while others will relapse regardless of adjunc- draining the site of a cutaneous melanoma.7 They tive measures. Currently there are no predictive showed conclusively that the pathologic status of methods to distinguish one group of patients the sentinel node accurately determines whether from another, therefore it is necessary to treat melanoma cells have metastasised to that spe- all patients in hopes of gaining an incremental cific lymph node basin.8 An important aspect benefit for a select few. Hence, in addition to of sentinel node biopsy is a detailed histologic the overall level of efficacy, clinicians evaluate examination of the sentinel lymph nodes. Gen- toxicity, convenience, cost-effectiveness and the erally, this examination is more thorough than prospects of post-relapse salvage therapy when is practical to perform on the larger number deciding whether to employ adjuvant therapy. of nodes obtained during elective node dissec- Virtually all of these factors can be determined tion. This more detailed pathologic analysis, com- accurately only in randomised trials. bined with the ability to identify sentinel nodes In 1995, high-dose interferon-α2b (IFN-α2b) that are outside the defined boundaries of a was approved by the United States Food and regional basin, makes sentinel node biopsy the Drug Administration, based on the positive most sensitive and specific test for nodal metas- results of a single, randomised Phase III clinical tasis currently available. The prognostic value of trial, E1684. The FDA’s decision was considered sentinel node status has been demonstrated in controversial at the time. Subsequent randomised multiple studies. In published multivariate anal- trials involving the same basic interferon regimen yses, histologic status of the sentinel nodes is have not only failed to put this controversy to the most powerful predictor of disease-specific rest, but have in fact enhanced it. survival.9 Overall, 5-year disease-specific sur- vival is >80% for patients with negative sentinel ADJUVANT INTERFERON CLINICAL TRIALS nodes, compared to about 50% for patients with one or more positive sentinel nodes. Importantly, patients with positive sentinel nodes go on to E1684 elective complete lymph node dissection. Among Eastern Cooperative Oncology Group (ECOG) patients with negative sentinel nodes, only 4% trial E1684, with 280 eligible patients with thick or fewer ultimately experience a clinically evi- primary (≥4.00 mm) or node-positive melanoma dent relapse within the nodal basin. Thus, sentinel who were randomly assigned after surgery to node biopsy matches the excellent regional con- observation or post-operative adjuvant treat- trol achieved by elective node dissection while ment with IFN-α2b for one year, demonstrated subjecting fewer patients to the morbidity of the statistically-significant improvements in relapse- complete node dissection procedure. free and overall survival for patients randomised to the interferon arm. IFN-α2b therapy increased ADJUVANT THERAPY FOR MELANOMA the median relapse-free survival by 9 months (1.72 years for IFN-α2b patients versus 0.98 years The development of effective adjuvant therapy for observation patients) and produced a relative has been a long-standing goal of melanoma 42% improvement in the 5-year relapse-free sur- researchers, and the subject of over 100 ran- vival rate (37% for IFN-α2b patients versus 26% domised clinical trials involving a host of dif- for observation patients). In addition, IFN-α2b ferent agents.10 Adjuvant therapy is the systemic therapy significantly increased median overall sur- or regional administration of drugs or radiation vival by 1 year (3.82 years for IFN-α2b patients to patients after apparently successful surgery, versus 2.78 years for observation patients) and in an effort to minimise the risk of subsequent produced a 24% relative improvement in the recurrence. Although many patients are cured by 5-year overall survival rate (46% for IFN-α2b surgery, some benefit from adjuvant treatment patients versus 37% for observation patients).11 152 TEXTBOOK OF CLINICAL TRIALS
Side effects were common and frequently severe, CLINICAL CONSIDERATIONS but even when adjusted for time with toxicity, the results favoured adjuvant IFN-α2b therapy.12 RELAPSE-FREE SURVIVAL VERSUS OVERALL SURVIVAL E1690 It has been the authors’ experience that clinicians A subsequent Intergroup adjuvant therapy trial, tend to view clinical trial results as dichotomous, E1690, also compared the high-dose IFN-α2b to that is, ‘positive’ or ‘negative’. Moreover, par- observation after complete resection of all known ticularly for adjuvant therapy trials, the accep- disease.13 This was a three-arm trial involving tance of a clinical trial as ‘positive’ is often 608 eligible patients. The eligibility criteria were restricted to trials demonstrating a statistically the same as for E1684, except for the fact that significant benefit in overall survival. From this elective node dissection was not required for perspective, there seems to be an obvious discrep- patients entered onto E1690 with thick primary ancy among the two observation-controlled trials: melanomas and clinically negative nodes. Results E1684 demonstrated seemingly striking benefits of this trial confirmed the relapse-free survival from the high-dose interferon regimen in both advantage seen in E1684 but with no survival relapse-free and overall survival, whereas E1690 advantage observed. validated only the relapse-free survival benefit with no survival difference. However, the impor- E1694 tance of relapse-free survival may be worth closer examination in the current setting. In light of the discordant survival results in Statistically it is commonly known that, com- E1684 and E1690, the initial results of another pared to overall survival, disease relapse is a less Intergroup trial, E1694, have received intense objective endpoint because it depends on the def- scrutiny. This trial compared one year of high- inition of relapse as well as the frequency and dose interferon not to an observation control method of detection. Defining relapse is less of as in the two earlier studies, but rather to two an issue in the adjuvant setting since patients years of a ganglioside vaccine called GMK. enter the study with no detectable disease and This was the largest of the three trials, with thereafter any new disease found is considered 774 eligible patients between two study arms. a relapse. In a well-conducted clinical trial the For the first time, staging of the lymph nodes interval and method of disease assessment are by sentinel node biopsy was performed in a specified in the protocol and generally complied significant fraction of patients. Gangliosides are with by trialists, thereby rendering relapse-free carbohydrate antigens found on the surface of survival a more reliable endpoint than in other melanoma cells, as well as normal cells of neural situations. From the purely clinical viewpoint, crest origin and tumour cells of other types. A patients have made clear that they are willing pilot randomised trial suggested a relapse-free to accept even toxic adjuvant therapies that pro- survival benefit in patients who were treated vide improvements in relapse-free survival, even with purified ganglioside GM2 (the specific if they do not result in any prolongation of over- ganglioside in the GMK vaccine) plus BCG all survival. This observation has been directly compared to those treated with BCG alone.14 In validated in melanoma patients,16 and represents May 2000, the E1694 trial’s independent Data the perception that time spent without signs or Safety Monitoring Committee concluded that the symptoms of recurrent cancer is inherently of high-dose interferon arm was associated with value even in the absence of prolongation of total highly significantly improved relapse-free and lifespan. In addition, relapse-free survival often overall survival, and mandated that the study represents a truer reflection of the biologic activ- results be disclosed early.15 ity of an adjuvant therapy since randomised trials MELANOMA 153 rarely include rigorous controls on post-relapse arm received subsequent interferon-α-containing salvage therapy. The confounding effect of such regimens (31% vs. 15%) and/or biochemotherapy treatment on overall survival is unknown and (17% vs. 6%). not assessable. While there is some evidence of differential post-relapse treatment received, concluding that RECONCILING THE STUDY RESULTS BASED the lack of interferon survival benefit observed ON CLINICAL CONSIDERATIONS in E1690 is due to these differences is not justified. Making this conclusion presupposes Two of the three randomised Phase III trials of survival efficacy from these salvage therapies, high-dose interferon, E1684 and E1690, demon- which cannot be substantiated with currently strate a relapse-free survival advantage. The third available data. In addition, comparing outcomes trial, E1694, also shows a relapse-free survival by post-relapse treatment groups provides little benefit but with GMK vaccine and not obser- useful information because patients were not vation as the control treatment. The implication randomised to salvage treatment strategies upon of this design difference is discussed in detail relapse. As is inherent in observational data, below. Nevertheless, many consider there is uni- unknown patient selection factors cannot be formity of evidence that high-dose interferon has accounted for by analysis techniques and their biologic activity in at least delaying relapse after impact can easily remain even after adjusting surgical therapy. This fact alone, combined with for known prognostic factors. Therefore, although the lack of proven alternatives, is enough for available data appear compatible with the notion many patients to choose interferon therapy in the that initial observation after surgery followed by absence of consensus regarding the overall sur- high-dose interferon in case of resectable relapse vival benefit. presents an alternative strategy to routine use of Crossover to interferon therapy upon relapse adjuvant high-dose interferon, this study offers might have partially affected the outcome of at no proof for the conjecture. The conservative least one study. The original trial, E1684, was conclusion is that salvage treatment difference unlikely to have been affected by crossover for is a possible confounding factor that limits the two reasons. Surgical staging of the regional confidence regarding the lack of overall survival nodes by complete (elective or therapeutic) node benefit of high-dose interferon from study E1690. dissection was required. Hence, few patients were likely to experience regional relapse or other resectable recurrence, where secondary STATISTICAL CONSIDERATIONS resection and delayed adjuvant interferon could be employed. Most relapses occurred in non- Although clinical factors clearly impact on the resectable distant sites. In recent medical practice, interpretation of the three trials, our main goal is interferon is rarely employed for the treatment of to examine the statistical aspects of these trials measurable metastatic disease. to determine the extent to which they actually In contrast, the E1690 trial required only clin- present ‘conflicting’ information. We focus first ical staging of the regional nodes, and surgery on E1684 and E1690. was not required for patients with thick primary tumours and clinically negative nodes. Among STATISTICAL TESTS EMPLOYED AND all relapsed patients (n = 114 in the high-dose PRESENTATION OF RESULTS interferon arm and n = 121 in the observation control arm), 54% on high-dose interferon and One source of confusion could be due to the fact 45% on observation experienced regional recur- that one-sided p-values were presented for E1684 rence only. Retrospective data collection indi- but two-sided p-values were presented for E1690. cated more patients relapsing on the observation Since all comparisons involved were one-sided 154 TEXTBOOK OF CLINICAL TRIALS in nature (i.e., is high-dose interferon superior to contributed to an exaggerated impression of the observation after surgery), we use all one-sided overall survival benefit from E1684. p-values (p1) in this discussion. In addition, all hazard ratios are expressed as observation TRIAL SIZE, OVERALL RESULTS arm versus treatment arm ratios. Thus, a hazard AND OTHER ASPECTS ratio >1 indicates an excess of hazard in the observation arm, or treatment advantage. To interpret the combined results E1684 and Another possible source of confusion could E1690, it is useful to compare the study param- be the fact that, in E1684, statistically signifi- eters and overall results. Tables 10.1–10.3 are cant p-values for relapse-free and overall sur- extracted mainly from Ref. 13. Since there was vival differences by the stratified logrank test not a low-dose interferon arm in E1684, only (adjusted for disease burden and presentation at the high-dose interferon and observation arms of initial diagnosis versus recurrent nodal disease E1690 are included in the tables. Due to the status) were reported (Table 2 of Ref. 11). But limitations of data availability, all randomised when Cox regression analysis was performed, patients regardless of eligibility determination are further adjusting for age, time from diagnosis to presented for consistency. randomisation and ulceration status of the pri- The tables indicate that when E1690 results mary tumour, a significant interferon over obser- became available, the study had 50% more vation benefit was presented only for those with patients than E1684, reflecting wider participa- nodal disease (Table 4 of Ref. 11). The haz- tion from the US Melanoma Intergroup. The ard ratio was 1.64 for relapse-free survival and patient enrollment periods were non-overlapping. Although the updated data for E1684 had longer 1.49 for overall survival with p = 0.01 in both 1 follow-up at the time of E1690 publication, more cases. However, these hazard ratios (presented in their reciprocals as interferon over observa- tion ratios, 0.61 and 0.67, in actuality) were Table 10.1. E1684 and E1690 study characteristics labelled ‘Treatment with IFN’ without reference ∗ Study E1684 E1690 to the positive nodal disease subset. An interac- tion term between the interferon treatment and Participating ECOG ECOG, SWOG, the thick primary, no nodal disease patient cat- groups CALGB, egory was actually included in the Cox mod- MDACC Patient accrual 1984–1990 1991–1995 els and the results were presented in the same period table with the label ‘CS1/PS1 + IFN’. The haz- N (all 286 427 ard ratios were 0.36 and 0.34 respectively for randomised) relapse-free survival and overall survival. These Median 6.9 4.3 interaction hazard ratios translated into observa- follow-up (years) tion over interferon hazard ratios of 0.60 and 0.50 Event count: RFS 197 241 for relapse-free and overall survival in patients Event count: OS 175 190 with thick primary tumours and pathologically ∗High-dose interferon and observation arms only. negative nodes, reflecting the occurrence that interferon-treated patients fared worse than the observation patients in this subset. For the readers Table 10.2. E1684 and E1690 patient disease stage distribution who did not appreciate these details of the Cox modelling, the hazard ratios for the nodal disease Disease T4 T1-4 N+ T1-4 N+ N+ subset could have been over-interpreted as the stage N0 (occult) (overt) Recurrent Cox model treatment effects for the study as a E1684 11% 12% 14% 63% whole, which were not presented in the original E1690 26% 11% 12% 50% publication. Such misinterpretation might have MELANOMA 155
Table 10.3. E1684 and E1690 results other) with a one-sided p-value of 0.0125 for each comparison to maintain an overall one-sided Hazard ∗ type I error rate of 0.025 for the study. When Study ratio 95% CI p-Value the results were presented, however, one-sided p- Relapse-free survival values less than 0.025 were treated as statistically ∗∗ E1684 1.43 (1.08, 1.89) 0.002 significant for each comparison, representing a E1690 1.28 (1.0, 1.65) 0.03 study-wide, one-sided type I error rate of 0.05 or Overall survival a two-sided error rate of 0.10. Also, per design ∗∗ E1684 1.32 (0.98, 1.77) 0.03 the study was sized so that the power for each E1690 1.0 (0.75, 1.33) 0.50 individual comparison was 0.83. In other words, ∗One-sided p-value by stratified logrank test. the type II error rate for each comparison was ∗∗Ref. 21. 0.17 for an approximate study-wide type II error rate of 0.34. Should the true magnitude of benefit events were analysed for E1690 from the larger from both interferon regimens be the same, the sample size and the fact that few events occurred power to detect both effects in the same study after 5 years. The main known patient char- was close to 0.66. With the inflated type I error acteristic difference was in the distribution of rate in the end, the overall power would increase disease stage. There were more node-negative somewhat but would likely remain less than ade- patients (26% vs. 11%) and fewer recurrent dis- quate for detecting reasonable effects from both ease patients (63% vs. 50%) in E1690, repre- treatment arms. Hence, the question about the senting a somewhat more favourable prognosis. low-dose interferon regimen’s treatment effect It may be worth pointing out that, among those was essentially unanswered in this study, yet clin- with nodal disease, there did not appear to be sur- icians seem to have uniformly concluded that vival differences between newly diagnosed and low-dose interferon is inactive in E1690. recurrent disease patients. The more favourable relapse and survival experiences of the obser- WHAT DOES E1694 TELL US? vation patients in E1690 compared to those in E1684 (5-year relapse-free survival of 35% vs. E1694 was designed to detect a GMK vaccine 26% and overall survival of 54% vs. 37%) remain benefit over interferon as the contemporary largely unexplained by known factors. Regard- treatment standard. As is often practiced with ing the treatment outcome, the magnitude of the superiority designs, the trial would be stopped interferon benefit was smaller in E1690 than at planned interim analyses if the hypothesised in E1684 for both relapse-free survival (hazard vaccine benefit could be definitively ruled out. ratio 1.43 vs. 1.28) and overall survival (haz- This provision was incorporated in the study ard ratio 1.32 vs. 1.00). The larger event counts design in the following manner. Instead of in E1690 resulted in narrower confidence inter- the typical, highly stringent interim p-value vals. As offered by the authors as one plausible requirements, the GMK vaccine needed only conclusion,13 the combined evidence from these to be inferior to interferon at a fixed, one- two trials seems to indicate that, for node posi- sided p-value of 0.05 for relapse-free survival tive and thick primary, node-negative melanoma in order to consider study termination at interim patients, treatment of high-dose interferon pro- analyses. Such evidence might not establish the longs relapse-free survival. Survival benefit, if it vaccine inferiority but would certainly rule out exists, may be more limited. its superiority. It is worth pointing out that E1690 was Considering the substantially more favourable designed with not one but two primary compar- vaccine toxicity profile, a more appropriate trial isons, comparing high-dose interferon and low- design might have sought to demonstrate the dose interferon to observation (but not to each equivalence of the two agents in their efficacy 156 TEXTBOOK OF CLINICAL TRIALS rather than the superiority of the vaccine. In fact reconcile a trend in favour of antibody respon- the Data and Safety Monitoring Committee in ders with speculations of a deleterious effect of this case seemed to have followed the equiva- the vaccine resulting from production of ‘block- lence principle and disclosed the study results ing’ antibodies. However, it is known that effects only when there was decisive evidence that the of prognostic factors such as disease stage can GMK vaccine was inferior to high-dose inter- easily overwhelm any treatment effects. feron in both relapse-free survival (p1 = 0.0015) and overall survival (p = 0.009).15 Because no 1 DID ANY SUBSET OF PATIENTS BENEFIT observation control arm was incorporated in the MORE FROM INTERFERON? study design, the clinical interpretation of E1694 in this respect is subject to debate. Obviously, The predominant subcategories of high-risk if it were known that the GMK vaccine had melanoma patients are those having thick primary some level of clinical efficacy, the finding that tumours with clinically or pathologically nega- high-dose interferon was significantly better in tive nodes and those having documented involve- both disease-free and overall survival would be ment of the nodes. Among the node-positive of great clinical significance and would substan- patients, subsets include those with 1, 2 to 3 and tiate the benefits identified in the initial E1684 ≥4 nodes; patients with clinically evident ver- trial. Without this knowledge, some have main- sus microscopic nodal involvement; and patients tained the possibility of a deleterious vaccine found to have nodal involvement at the time of effect and insisted that the study cannot be used initial presentation versus those developing recur- to give information on the non-design comparison rent disease in the nodes. of interferon versus observation. The initial findings of E1684 indicated that the Unfortunately, no credible evidence exists that subset of patients with thick primary tumours and the GMK vaccine is either beneficial or delete- pathologically negative nodes had no benefit, and rious. It is likely that the GMK vaccine acted perhaps even a detrimental effect, from adjuvant essentially as placebo and the study provided interferon.11 The veracity of this finding was further validation that high-dose interferon was called into question from the outset, because efficacious over no treatment in both relapse-free of the small number of node-negative patients and overall survival. But we do not know this (a total of 31 out of 280 eligible patients, or for certain. As the dramatic survival difference 11%) and an imbalance in a major prognostic between E1684 and E1690 observation patients factor (ulceration of the primary tumour) biasing amply illustrates,13 comparison of patient out- the results in favour of the observation arm. comes in the GMK vaccine arm to historical In contrast, subset analysis of the results of controls in the other two trials offers few clues trial E1690 found that the relapse-free survival to the efficacy of the vaccine. benefit for patients with thick primary tumours Data were presented that, among the vaccine- and clinically negative nodes (making up 25% of treated patients, those displaying antibody respon- the eligible patient population) was identical to ses had a trend towards favourable outcomes that for the study population as a whole.13 Subset Ref. 17. Even assuming that the analyses cor- analysis of E1694 showed the greatest interferon rected for the inherent responder versus non- over vaccine benefit for the subset of thick, node- responder bias,20 the results still cannot be used negative patients.15 to establish a causal relationship between vaccine Indeed, in each of the three clinical trials, response and favourable outcome. As pointed out subset analysis indicated a different group as in numerous publications, response to treatment obtaining the most benefit from high-dose inter- could simply serve as a selection mechanism feron: the subset with one single positive node wherein responders represented a better progno- in E1684; the subset with two to three positive sis group. One may contend that it is difficult to nodes in E1690; and the node-negative subset MELANOMA 157 in E1694. The authors properly suggested that, of a far more homogeneous patient population taken together, there was no indication of prefer- than any prior clinical trial, potentially enhancing ential treatment effect in any one subset.15 These the scientific validity. Of note, this group now results exemplify the lack of reliability of sub- constitutes by far the largest fraction of ‘high- set results, a phenomenon previously discussed in risk’ melanoma patients being seen and treated regard to other melanoma clinical trials.18 With- in the United States today, yet less than 10% out appropriate study size for adequate power of participants in the three prior trials combined within subsets, and control for inflated type were from this category. Unfortunately, this trial errors stemming from multiple testing, post hoc is likely to be small compared to the most recent subset analyses suffer both high false-positive Intergroup trials and, regardless of the results, it and high false-negative rates. will not directly address the role of interferon in all of the other high-risk categories. It is now nearly 20 years since the design CONCLUSIONS of clinical trial E1684, and 8 years since the FDA’s approval of high-dose interferon-α for Three randomised trials evaluating high-dose the adjuvant therapy of high-risk melanoma, and interferon, involving over 1600 patients, have we may never fully know to what extent this been conducted, yet its treatment benefit remains toxic and inconvenient regimen improves overall controversial. The combined evidence indicates survival. The implications of that statement are that, for high-risk melanoma patients, treatment profound, and the burden they place on clinical of high-dose interferon prolongs relapse-free trialists is clear: design and analyse our trials survival. Survival benefit is less certain. There carefully to have the greatest probability of a is no credible evidence to suggest that interferon clear and unambiguous result. exerts a differential effect in different subsets of ‘high-risk’ patients. There are many reasons why high-dose inter- REFERENCES feron has not been uniformly embraced by physi- cians and patients around the world, even though 1. Balch CM, Soong SJ, Gershenwald JE, et al.Pro- it is the only adjuvant therapy yet shown to have gnostic factors analysis of 17,600 melanoma patients: validation of the American Joint Com- any sustained impact on relapse-free survival. mittee on Cancer melanoma staging system. J Clin When the three trials are looked at in the light Oncol (2001) 19: 3622–34. of statistical principles, what seem to be glaring 2. Veronesi U, Cascinelli N. Narrow excision (1- differences are more plausibly regarded as under- cm margin): a safe procedure for thin cutaneous standable variations reflecting trial design and melanoma. Arch Surg (1991) 126: 438–41. 3. Balch CM, Soong S, Ross MI, et al. Long-term analysis, combined with the fluctuations inher- results of a multi-institutional randomized trial ent in human clinical trials conducted over time comparing prognostic factors and surgical results in similar yet subtly different patient populations. for intermediate thickness melanomas (1.0 to While it is easy to conclude that further 4.0 mm). Intergroup Melanoma Surgical Trial. research is necessary to determine if high-dose Ann Surg Oncol (2000) 7: 87–97. 4. Wagner JD, Schauwecker D, Davidson D, et al. interferon α-2b improves overall survival, there is Prospective study of fluorodeoxyglucose-positron in fact little chance that definitive further research emission tomography imaging of lymph node will take place. Only one current clinical trial, basins in melanoma patients undergoing sentinel the Sunbelt Melanoma Trial, is comparing one node biopsy. J Clin Oncol (1999) 17: 1508–15. year of high-dose interferon to a control group. 5. Balch CM. The role of elective lymph node dissection in melanoma: rationale, results, and This study includes only patients with a single controversies. J Clin Oncol (1988) 6: 163–72. positive sentinel node identified at the time of 6. Cascinelli N, Morabito A, Santinami M, MacKie initial presentation.19 As such, it is comprised RM, Belli F. Immediate or delayed dissection 158 TEXTBOOK OF CLINICAL TRIALS
of regional nodes in patients with melanoma of 14. Livingston PO, Wong GYC, Adluri S, et al.Im- the trunk: a randomised trial. WHO Melanoma proved survival in stage III melanoma patients Programme. Lancet (1998) 351(9105): 793–6. with GM2 antibodies: a randomized trial of 7. Morton D, Wen D, Wong J, et al. Technical details adjuvant vaccination with GM2 ganglioside. J Clin of intraoperative lymphatic mapping for early Oncol (1994) 12: 1036–44. stage melanoma. Arch Surg (1992) 127: 392–9. 15. Kirkwood JM, Ibrahim JG, Sosman JA, et al. 8. Morton DL, Thompson JF, Essner R, et al. Vali- High-dose interferon alfa-2b significantly prolongs dation of the accuracy of intraoperative lymphatic relapse-free and overall survival compared with mapping and sentinel lymphadenectomy for early- the GM2-KLH/QS-21 vaccine in patients with stage melanoma: a multicenter trial. Multicen- resected stage IIB–III melanoma: results of ter Selective Lymphadenectomy Trial Group. Ann Intergroup trial E1694/S9512/C509801. J Clin Surg (1999) 230: 453–65. Oncol (2001) 19: 2370–80. 9. Gershenwald JE, Thompson W, Mansfield PF, 16. Kilbridge KL, Weeks JC, Sober AJ, et al. Patient et al. Multi-institutional melanoma lymphatic preferences for adjuvant interferon alfa-2b treat- mapping experience: the prognostic value of ment. J Clin Oncol (2001) 19: 812–23. sentinel lymph node status in 612 stage I or 17. Kirkwood JM, Ibrahim J, Lawson DH, et al. II melanoma patients. J Clin Oncol (1999) 17: High-dose interferon alfa-2b does not diminish 976–83. antibody response to GM2 vaccination in 10. Sondak VK, Wolfe JA. Adjuvant therapy for patients with resected melanoma: results of the melanoma. Curr Opin Oncol (1997) 9: 189–204. multicenter Eastern Cooperative Oncology Group 11. Kirkwood JM, Strawderman MH, Ernstoff MS, phase II trial E2696. J Clin Oncol (2001) 19: et al. Interferon alfa-2b adjuvant therapy of high- 1430–6. risk resected cutaneous melanoma: the Eastern 18. Sondak VK. Multi-institutional melanoma vac- Cooperative Oncology Group trial EST 1684. J cine trial [Letter]. Ann Surg Oncol (1996) 3: Clin Oncol (1996) 14: 7–17. 588–9. 12. Cole BF, Gelber RD, Kirkwood JM, et al.Qua- 19. McMasters KM, Sondak VK, Lotze MT, Ross lity-of-life – adjusted survival analysis of inter- MI. Recent advances in melanoma staging and feron alfa-2b adjuvant treatment of high-risk therapy. Ann Surg Oncol (1999) 6: 467–75. resected cutaneous melanoma: an Eastern Cooper- 20. Anderson JR, Cain KC, Gelber RD. Analysis of ative Oncology Group study. J Clin Oncol (1996) survival by tumor response. J Clin Oncol (1983) 14: 2666–73. 1: 710–19. 13. Kirkwood JM, Ibrahim JG, Sondak VK, et al. 21. Kirkwood JM, Manola J, Ibrahim J, et al. A pooled High- and low-dose interferon alfa-2b in high-risk analysis of ECOG and Intergroup trials of adjuvant melanoma: first analysis of Intergroup trial E1690/ high-dose interferon for melanoma. Accepted for S9111/C9190. J Clin Oncol (2000) 18: 2444–54. publication in Clinical Cancer Research (2004). 11 Respiratory Cancers
JOAN H. SCHILLER1 AND KYUNGMANN KIM2 Departments of 1Medicine and 2Biostatistics and Medical Informatics, University of Wisconsin, Madison, WI 53792, USA
INTRODUCTION late 1970s, the incidence of lung cancer is still ris- ing because of long latency and a steady increase Carcinoma of the lung and bronchus is estimated in smoking among the female population. to account for almost 13% of all new cancer With the litigation and subsequent settlement cases, but to be responsible for about 157 200 between the tobacco industry and the state gov- 1 deaths in the United States in 2003. This ernments in the US, the marketing effort of the represents more than 28% of all deaths due to US tobacco industry has shifted to the emerging cancer, exceeding the number of deaths due to the markets in Asia and Eastern Europe. As a conse- next four leading cancers, colon, breast, prostate quence, the smoking-related public health problem and pancreas, all combined. is predicted to pose a serious threat to the national The incidence of the disease continues to rise, security in a country such as China, in which smok- particularly in women and blacks, and thus is likely ing has increased dramatically in recent years. to present a significant public health problem for The best investment for prevention of smoking- years to come. Cigarette smoking is attributed as related cancer incidence and death, as well as the cause of 80% to 90% of lung cancer cases, with other diseases such as cardiovascular and other the risk for lung cancer among smokers being 20 pulmonary diseases, appears to be in smoking to 30 times that among non-smokers. Other risk cessation and prevention of taking up the smok- factors include exposure to asbestos and radon. ing habit among teenagers and females. Asbestos exposure, known to cause malignant mesothelioma, increases the risk for lung cancer, CLASSIFICATIONS especially among smokers. There are limited data on molecular and genetic profile as a risk factor, Lung cancer consists of four major histological and familial predisposition to lung cancer. Diet’s types: adenocarcinoma, squamous cell carcinoma, role in lung cancer is even less obvious. large-cell carcinoma and small-cell carcinoma. Despite the significant reduction in smoking, Because of the unique biological features of small- especially among the male population since the cell lung cancer (SCLC), its staging and treatment
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 160 TEXTBOOK OF CLINICAL TRIALS differ radically from the other three types of lung the TNM staging.2 It is this classification that cancer, which are collectively called non-small- forms the basis for management and treatment of cell lung cancer (NSCLC). patients with NSCLC. Besides histological classification, lung can- A staging system entirely different from that cer is also classified according to the Tumour, for NSCLC is used for patients with small-cell Node and Metastasis (TNM) staging and the carcinoma of the lung. Small-cell lung cancer International Staging Classification. The TNM is clinically categorised into two stages: limited staging system is applied primarily to NSCLC and extensive. Limited-stage SCLC is defined and consists of three components, each accord- as tumours confined to one hemithorax and its ing to primary tumour (T), nodal involvement regional lymph nodes that can be encompassed (N) and distant metastasis (M) as summarised in a tolerable irradiation field. Extensive-stage in Table 11.1. The International Staging Classi- SCLC is defined as any extent of disease beyond fication summarised in Table 11.2 is based on that classification.
Table 11.1. TNM staging
Primary tumour (T) TX Tumour proven by the presence of malignant cells in bronchopulmonary secretions but not visualised roentgenographically or bronchoscopically, or any tumour that cannot be assessed as in a retreatment staging T0 No evidence of primary tumour Tis Carcinoma in situ T1 A tumour that is 3.0 cm or less in greatest dimension, surrounded by lung or visceral pleura, and without evidence of invasion proximal to a lobar bronchus at bronchoscopy T2 A tumour more than 3.0 cm in greatest dimension, or a tumour of any size that either invades the visceral pleura or has associated atelectasis or obstructive pneumonitis extending to the hilar region. At bronchoscopy, the proximal extent of demonstrable tumour must be within a lobar bronchus or at least 2.0 cm distal to the carina. Any associated atelectasis or obstructive pneumonitis must involve less than an entire lung T3 A tumour of any size with direct extension into the chest wall (including superior sulcus tumours), diaphragm, or the mediastinal pleura or pericardium without involving the heart, great vessels, trachea, oesophagus or vertebral body, or a tumour in the main bronchus within 2 cm of the carina without involving the carina T4 A tumour of any size with invasion of the mediastinum or involving the heart, great vessels, trachea, oesophagus, vertebral body for carina or presence of malignant pleural effusion; a satellite nodule within the same lobe
Nodal involvement (N) NX Regional lymph nodes cannot be assessed N0 No demonstrable metastasis to regional lymph nodes N1 Metastasis to lymph nodes in the peribronchial or the ipsilateral hilar region, or both, including direct extension N2 Metastasis to ipsilateral mediastinal lymph nodes and subcarinal lymph nodes N3 Metastasis to contralateral mediastinal lymph nodes, contralateral hilar lymph nodes, ipsilateral or contralateral scalene or supraclavicular lymph nodes
Distant metastasis (M) M0 No distant metastasis M1 Distant metastasis, including pulmonary nodule not in the same lobe as the primary tumour RESPIRATORY CANCERS 161
Table 11.2. International Staging Classification for First, it has a more rapid clinical course and lung cancer natural history, with the rapid development of metastases, symptoms and eventually death. Left Five-year survival (%) untreated, the median survival time is typically Clinical Pathological 12–15 weeks for patients with local disease and Stage TNM subset stage stage 6–9 weeks for those with advanced disease. IA T1, N0, M0 61 67 Second, it exhibits features of neuroendocrine IB T2, N0, M0 38 57 differentiation in many patients, which may IIA T1, N1, M0 34 55 be distinguishable histopathologically and is IIB T2, N1, M0; T3, N0, 24 39 associated with paraneoplastic syndromes. Third, M0 unlike NSCLC, SCLC is exquisitely sensitive to IIIA T3, N1, M0; T1–3, 925 N2, M0 both chemotherapy and radiotherapy, although IIIB T4, any N, M0; any 13 23 resistant disease often develops. T, N3, M0 Due to sensitivity of patients with SCLC to IV Any T, any N, M1 1 – chemotherapy, it can pose challenges in design of clinical trials for drug development as will be discussed in some detail. INCIDENCE
In the year 2002, 1 284 900 new cases of invasive CLINICAL TRIALS IN LUNG CANCER cancer were expected in the United States, excluding carcinoma in situ of any site except Clinical trials have resulted in significant seminal the urinary bladder and also excluding basal and trials which have led to changes in the man- squamous cell cancers of the skin. Lung cancer agement of these patients. Those seminal studies is estimated to account for 13% (169 400 cases) in screening, chemoprevention and treatment are of all new cancer cases, 14% (90 200) in males outlined. and 12% (79 200) in females. The annual age-adjusted incidence rate of lung SCREENING AND EARLY DETECTION cancer in the male population has been in a steady decline since its peak in the early 1980s. Three US randomised screening studies failed to However, that in the female population appears detect an impact of screening high-risk patients to be still increasing, although the rate of increase with chest radiographs or sputum cytology on has slowed in the late 1990s. mortality, although earlier stage cancers were detected in the screened groups.3–5 These studies PROGNOSIS have been criticised for a number of potential methodological and statistical problems, such as Prognosis for patients diagnosed with lung cancer over-diagnosis and analysing data by survival is dismal, with less than 15% surviving longer rather than mortality.6 than 5 years from the time of diagnosis, and it Recently, several clinical studies have demon- is highly dependent on stage of the disease as strated that early stage lung cancers can be indicated in Table 11.2. detected with the use of spiral CT that would not have been detected by routine chest X-ray.7 SALIENT FEATURES OF SMALL-CELL Spiral CT is a CT scan which does not evaluate LUNG CANCER the mediastinum and thus does not use contrast or require the presence of a radiologist, employs low Small-cell lung cancer, which makes up a quarter doses of radiation and can be completed within to a third of all lung cancer at diagnosis, differs one patient ‘breath’. Because it can be done from NSCLC in a number of important ways. rapidly and does not require a radiologist to be 162 TEXTBOOK OF CLINICAL TRIALS present, it is being used in some centres to screen The selenium group had fewer total carcino- for lung cancers in high-risk populations. How- mas, including lung cancer with a relative risk ever, it has not been determined whether there of 0.54 and a 95% confidence interval of (0.30 is a survival benefit with this technique.6,7 Given to 0.98) (p = 0.04).11 This has formed the basis the availability of this scanning technique in the for an intergroup chemoprevention trial which is community, it is imperative that clinical trials be now ongoing. completed to determine if the early detection of small tumours results in improved survival that Stage II and ‘Non-Bulky’ IIIA Disease is not a result of lead time or length bias. Treatment of locally advanced NSCLC is one TREATMENT: NON-SMALL-CELL of the most controversial issues in the manage- LUNG CANCER ment of lung cancer. Treatment options include Treatment of NSCLC is dependent primarily on surgery for less-advanced disease, or radiother- stage of disease at the time of diagnosis and stage, apy, either of which has been given with or in turn, is dependent upon the size of the tumour without chemotherapy for control of micrometas- (T), location of nodal involvement (N), if any, and tases. Interpretation of the results of clinical tri- presence or absence of distant metastases (M). The als involving patients with locally advanced dis- current TNM staging classification is shown in ease has been clouded by a number of issues, Table 11.1 and the stage grouping in Table 11.2. including changing diagnostic techniques, differ- ent staging systems and heterogeneous patient populations that may have disease that ranges Stage I Disease from ‘non-bulky’ stage IIIA (clinical N1 nodes, A lobectomy is the treatment of choice for stage I with N2 nodes discovered only at the time of NSCLC, with cure rates of 60–80% reported. surgery or mediastinoscopy), to ‘bulky’ N2 nodes Within stage I, patients with T2, N0 disease do (enlarged adenopathy clearly visible on chest not fare as well as those with T1, N0 cancers. X-ray films, or multiple nodal level involvement), In approximately 20% of patients with medical to clearly inoperable stage IIIB disease. contraindications to surgery but with adequate pulmonary function, high-dose radiotherapy will Post-operative Thoracic Radiotherapy:The result in cure. No role of adjuvant chemotherapy treatment for stage II and selected IIIA NSCLC for stage I NSCLC has been identified. patients is surgical resection. However, many of these patients will relapse, prompting numer- Chemoprevention: Patients with a resected stage ous trials evaluating the role of post-operative I NSCLC are at high risk of approximately 1% radiotherapy or chemotherapy. A meta-analysis per year for the development of second lung examining the role of post-operative radiother- cancers, prompting a number of ongoing clini- apy (PORT) found that patients randomised to cal trials looking at the role of chemoprevention. receive PORT actually had an inferior survival Surprisingly, several randomised studies have to those randomised to observation alone.12 In demonstrated that the use of vitamin A or one a meta-analysis of 2128 patients in nine clini- of its derivatives at best, does not prevent lung cal trials of post-operative radiotherapy, a 7% cancer in smokers and at worst, may increase survival decrement from radiation was identi- the risk of developing it.8–10 Preliminary stud- fied. However, this particular analysis included ies have suggested that selenium may reduce the a number of trials from the 1960s and 1970s incidence of lung cancer and total cancer mortal- when staging was highly inaccurate and relatively ity. In a multi-centre, double-blind, randomised, outmoded radiation therapy technologies were placebo-controlled trial, 1312 patients were ran- utilised. In addition, several of the trials included domised to receive either selenium or placebo. in this report aggressively treated patients with RESPIRATORY CANCERS 163 no evidence of nodal involvement or those with trials. In the European trial, the median survival early nodal involvement only, a group that by time was 26 months for patients receiving pre- today’s standards would not be subjected to post- operative chemotherapy plus surgery, compared operative radiation therapy. More recent studies to 8 months for patients treated with surgery looking at the role of PORT have concluded alone.16 In the MD Anderson trial, the median that PORT does not prolong survival, but does survival of the 32 patients randomised to the enhance local control. The most comprehensive surgery-alone group was 11 months compared randomised trial in this regard was performed to 64 months in the 28 patients randomised to by the Lung Cancer Study Group and it demon- the combined-modality arm.17 Of note, how- strated major improvement in intrathoracic dis- ever, is the fact that updated results of the MD ease control.13 For those patients receiving tho- Anderson trial, while still statistically significant, racic radiotherapy, the intrathoracic failure rate showed a narrowing of the survival curves, with a was only 3%, compared to 43% for patients not median survival of 14 months and 21 months for receiving post-operative radiotherapy, although the surgery alone and combined modality arms, no significant survival advantage was identified. respectively.18 A larger trial has recently been reported.19 Adjuvant Chemotherapy: Given the propensity Three hundred and fifty-five patients with stage I, of these resected patients to relapse with distant II or IIIA disease were randomised to three disease, adjuvant post-operative chemotherapy cycles of chemotherapy followed by surgery or has been of significant interest. A meta-analysis to surgery alone. Median survival (37 months published in 1995 found a small improvement in vs. 26 months) and 2-year survival (52% vs. survival with post-operative adjuvant chemother- 59%) were not statistically different between apy that borderlined on statistical significance the two groups. However, a subset analysis in 14 (p = 0.08), leading some clinicians to conclude which patients who died within 150 days of peri- that adjuvant chemotherapy was of benefit. How- operative problems were excluded revealed a ever, a randomised intergroup study has been 0.77 reduction in risk which was statistically sig- completed in which patients were randomised nificant (p = 0.03). Other subset analysis looked to receive either radiotherapy plus chemotherapy at outcome by patient stage and found that (cisplatin and etoposide for four cycles) or radio- the patients with N0/N1 disease who received therapy alone. The median and long-term survival chemo/surgery had a hazard ratio of 0.68, com- 15 of the two arms was nearly identical. Once pared to patients with N2 disease, where the haz- again, the fact that the meta-analysis included ard ratio was 1.04. older studies, in which chemotherapy regimens, Despite the results of the Depierre trial, staging and other clinical characteristics were many clinicians continue to use pre-operative different, may account for this discrepancy. chemotherapy for patients with stage IIIA dis- Although the role of neoadjuvant chemotherapy ease. An intergroup study evaluating chemo/RT is under investigation, it cannot be routinely rec- vs. chemo/RT surgery has recently been com- ommended until the results of randomised clinical pleted; these results are eagerly awaited. trials confirm clinical benefit.
Pre-operative Chemotherapy plus Surgery: Locally Advanced ‘Bulky’ There have been two small randomised stud- Stage IIIA/IIIB Disease ies involving surgery with or without pre- operative chemotherapy which popularised this The optimal treatment for bulky stage IIIA and approach. Both involved 60 patients and both stage IIIB disease is also controversial. Current report response rates of 35–62% following induc- investigational efforts are directed at identify- tion chemotherapy. Both have also reported pro- ing the optimal combined-modality approach, longed survival, prompting early closure of both involving treatments directed at local control 164 TEXTBOOK OF CLINICAL TRIALS of the disease, i.e., surgery or radiotherapy, combined with cisplatin, although there are clear and micrometastatic disease, i.e., chemother- differences in toxicity and cost.31,32 apy. Possibilities include radiotherapy only, pre- operative chemotherapy, or chemotherapy plus Second-Line Chemotherapy radiotherapy. Docetaxel was recently approved for the second- :Chemo- Chemotherapy plus Radiation Therapy line treatment of NSCLC, based upon two therapy plus radiotherapy is the treatment of clinical trials. One trial compared two doses of choice for patients with bulky or inopera- docetaxel with best supportive care, and found an ble stage III disease. Two randomised studies improvement in median and long-term survival, have demonstrated an improvement in median despite a low response rate of 7%.33 The other and long-term survival with chemotherapy fol- trial compared docetaxel to either vinorelbine or lowed by radiation therapy versus radiotherapy ifosfamide (the treatment physician was allowed alone.20,21 More recently, two randomised trials to choose) and found an improvement in long- have shown that concurrent chemoradiotherapy term, although not median survival.34 results in prolonged survival, albeit at the expense of enhanced toxicity, compared to sequential treatment.22,23 Other active areas of investiga- ‘Targeted’ Therapy tion include choice of chemotherapy, fractiona- tion and treatment fields. Given the overall poor results with standard cyto- Recently, weekly, low-dose ‘sensitising’ che- toxic therapies and the number of advances that motherapy plus radiation therapy has become have been made recently in our understanding popular, primarily due to lower toxicities when of the biology of cancer, a strong interest has administered with radiotherapy than ‘standard’ emerged in targeting pathways unique to neo- dose chemotherapy.24 However, this schedule has plastic cells. One such example is the epidermal never been looked at in a formal phase III setting, growth factor receptor (EGFr), which has been so its relative efficacy compared to standard dose found to be expressed in the majority of patients chemotherapy has not been rigorously assessed. with lung cancer. Based upon two phase II tri- als in previously treated NSCLC patients, in which response rates of 10–20% were found,35,36 Stage IV Disease two phase III trials were initiated comparing Several meta-analyses have demonstrated that chemotherapy plus an EGFr inhibitor, ZD1839, chemotherapy improves survival in patients with with chemotherapy in untreated NSCLC. Some- metastatic NSCLC (approximately 10% 1-year what surprisingly, no benefit was observed in survival untreated vs. 35–40% 1-year sur- these trials.37,38 These unexpected findings have vival with treatment),25,26 particularly if the resulted in clinical researchers, statisticians and chemotherapy is platin-based.14 In the past the pharmaceutical industry re-aiming the princi- 10 years, numerous different cytotoxic drugs ples of study design. have become available for the treatment of lung cancer patients. These include, among TREATMENT: SMALL-CELL LUNG CANCER others, vinorelbine, the taxanes (docetaxel and paclitaxel), gemcitabine and the topoisomerase Small-cell lung cancer differs from NSCLC in I inhibitors (irinotecan and topotecan). Ran- a number of important ways: (1) it has a more domised studies have shown that these agents rapid clinical course and natural history, with the improve survival when combined with cisplatin, rapid development of metastases, symptoms and as compared to cisplatin alone,27,28 or the other death; (2) it exhibits features of neuroendocrine agent alone.29,30 However, there is probably lit- differentiation in many patients which may tle difference in outcome between agents when be distinguishable histopathologically and is RESPIRATORY CANCERS 165 associated with paraneoplastic syndromes; and phase II studies is only between 3% and 11%. (3) unlike NSCLC, SCLC is exquisitely sensitive Median survival is about 20 weeks.40 Results of a to both chemotherapy and radiotherapy, although randomised trial comparing topotecan with CAV resistant disease often develops. Because of the (cyclophosphamide, adriamycin and vincristine) rapid development of distant disease and its as second-line therapy revealed no difference in extreme sensitivity to the cytotoxic effects of response rates, duration of response, or survival chemotherapy, this mode of therapy forms the between the two groups.41 backbone of treatment for this disease. Chemotherapy plus Chest Irradiation First-Line Therapy A number of combination chemotherapeutic reg- Numerous studies have been done with chemo- imens are available for SCLC. With these therapy and thoracic radiotherapy for patients chemotherapy regimens, overall response rates of with limited-stage SCLC. Conflicting results have 75–90% and complete response rates of 50% been attributed to differences in chemother- for localised disease can be anticipated. For apy regimens and different schedules integrat- extensive-stage disease, overall response rates ing chemotherapy and thoracic radiation, con- of about 75% with complete response rates of current, sequential and ‘sandwich’ approaches. 25% are common. Despite these high response Two recent meta-analyses concluded that thoracic rates, however, the median survival time remains radiation does result in a small but significant about 14 months for limited-stage disease and improvement in survival and major control of 7–9 months for extensive-stage disease. Less the disease in the chest, although no conclusions could be made regarding the optimal sequencing than 5% of extensive-stage patients have long- 25,42 term survival of greater than 2 years. of chemotherapy and thoracic radiation. A phase III randomised trial has been reported : For limited-stage in abstract form, in which patients with SCLC Fractionation of Radiotherapy SCLC, thoracic radiotherapy has been known were randomised to the control arm of etoposide to improve survival, but the best ways of inte- and cisplatin, versus cisplatin and the topoiso- grating chemotherapy and thoracic radiotherapy merase I inhibitor, irinotecan.39 Median survival are uncertain. In order to settle this question, a and 1-year survival was 420 days and 60% in phase III randomised clinical trial was conducted the cisplatin/irinotecan arm and 300 days and in which 417 patients with limited SCLC were 40% in the cisplatin/etoposide arm. If ongo- ing phase III studies confirm these results, cis- randomised to receive a total of 45 Gy of radio- platin/irinotecan would become the first combi- therapy, either twice-daily over a 3-week period or once-daily over a 5-week period, concurrently with nation of chemotherapy to improve survival over 43 cisplatin/etoposide in SCLC patients in decades. four 21-day cycles of cisplatin plus etoposide. Twice-daily radiotherapy improved median sur- vival as compared with once-daily radiotherapy Second-Line Therapy (23 months vs. 19 months, p = 0.04). However, No curative regimens for patients with recur- grade 3 or 4 oesophagitis was significantly more rent disease have been identified. Topotecan has frequent with twice-daily than with once-daily a 20–40% response rate in patients with ‘sen- fractionation (32% vs. 16%, p<0.001). sitive’ SCLC, those patients who relapsed two or more months after their first-line therapy, Prophylactic Cranial Irradiation with a median survival of 22–27 weeks. For patients with ‘refractory’ disease which pro- Numerous trials have demonstrated that prophy- gressed through or within 3 months of comple- lactic brain irradiation (PCI) does not enhance tion of first-line therapy, the response rate in survival, but does decrease the risk of brain 166 TEXTBOOK OF CLINICAL TRIALS metastases without a decrease in mental func- standard methods have been proposed.49,50 These tion.44 However, a recent meta-analysis demon- methods typically include initial dose-escalation strated a small but statistically and clinically in a single patient and a later switch-over to significant improvement in survival with PCI.45 standard dose-escalation at the earliest indication of dose-limiting toxicity, and may also include intra-patient dose-escalation.50 METHODOLOGIC ISSUES In order to address the failure of the standard dose-escalation designs to provide an estimate With the traditional cytoreductive and cytotoxic of the probability of toxicity at the recom- chemotherapy, both as single agents and in com- mended dose level, several new approaches bination, there are well-established and accepted have been proposed, including the continual designs for phases I, II and III clinical tri- reassessment method.51 This method is primar- als. In general these designs are based on the ily based on Bayesian statistical modelling of the paradigm that with the increased myelosuppres- dose–toxicity relationship with a targeted toxicity sion, tumour cells are more likely to be killed, probability for the MTD. With radiotherapy con- leading to shrinkage of tumours, and that there cerned with late-onset toxicities as the primary is a monotonically increasing dose–response and endpoint, the standard dose-escalation design for dose–toxicity relationship. It is also assumed that phase I clinical trials is inadequate because of the tumour shrinkage will eventually lead to clinical long-term follow-up required for late-onset toxic- benefit such as prolonged survival or improved ities associated with radiotherapy. With late-onset quality of life. In essence, tumour shrinkage has toxicities, the continual reassessment method has served as a surrogate for clinical benefit. been extended by statistical models for the distri- bution for time to toxicity and by pooling toxicity information across patients receiving the same PHASE I CLINICAL TRIALS dose level.52 In typical phase I clinical trials with acute dose- PHASE II CLINICAL TRIALS limiting toxicities as the primary endpoint, a standard dose-escalation scheme with a cohort In phase II clinical trials with cytotoxic chemo- of fixed number of patients treated at each dose therapy, multi-stage designs with objective tumour level is used to estimate the so-called maximum response defined as shrinkage of tumour by more tolerated dose (MTD) or safe dose46,47 to be than 50% as the primary endpoint are widely used in subsequent phase II studies. However, used.53–55 These are essentially sequential designs the choice of the initial dose and dose levels in the sense that a decision to treat additional have been rather ad hoc. Worse yet, the standard patients for establishment of clinical efficacy is dose-escalation design does not provide a well- predicated by the observed clinical efficacy or defined basis for estimation of the MTD and safety with the patients from the previous stages. is known to have several shortcomings such This is primarily to avoid treating patients with as slow dose escalation at the beginning and seemingly ineffective therapy. underestimation of the targeted dose-limiting Typically these designs are based on tests toxicity.48 Most critically, the standard design of statistical hypotheses with specific minimally does not provide an estimate of the probability acceptable and maximally unacceptable tumour of toxicity at the recommended dose level. response rates associated with type I and II Nevertheless this standard dose-escalation design error probabilities. Given these design parame- for phase I clinical trials has served a useful ters available from historical data, there are many function in this setting. candidate designs. In order to select a design, In order to avoid slow dose-escalation and one may use either the minimax or the opti- underestimation, a number of variations on the mality criterion.56 Subject to type I and II error RESPIRATORY CANCERS 167 probabilities, the minimax designs minimised the considered simultaneously in a parametric mix- maximum sample size required, while the opti- ture model. mal designs minimised the expected sample size under the null hypothesis that the true response SMALL-CELL LUNG CANCER rate is less than or equal to the maximally unacceptable response rate. Oftentimes they are As was noted earlier, small-cell lung cancer is very disparate, causing confusions to those not known to be biologically distinct from other his- so statistically sophisticated. A graphical search tologic subtypes of lung cancer in both laboratory method may be used to search for what appears and clinical studies. It is the most chemosensitive to be a compromise between the minimax and type of lung cancer and as a consequence it poses the optimal designs with more desirable practi- some difficulties in development of investiga- cal features such as having much smaller max- tional cytotoxic drugs. For example, there is eth- imum sample size than the optimal design and ical concern for testing investigational cytotoxic much smaller expected sample size than the min- drugs in previously untreated small-cell lung can- imax design.57 cer patients. As a result of these observations, it has PHASE III CLINICAL TRIALS been suggested that different phase II designs be used depending on whether patients had Overall survival typically being the ultimate been previously treated with cytotoxic drugs or criterion for evaluation of the efficacy of cancer have relapsed following treatment with cytotoxic treatment in phase III clinical trials, a traditional drugs.63 Also depending on whether patients are randomised, controlled design with time to death refractory to or have relapsed during previous due to all causes as the primary endpoint has treatment, different values for minimally accept- become recognised as a golden standard for able and maximally unacceptable response rates establishment of standard therapies in cancer. should be used in phase II clinical trials. Different However, depending on the disease setting, other considerations should be given to elderly patients endpoints such as time to disease progression, or patients with poor prognosis as well. time to treatment failure, etc. may be appropriate as a surrogate endpoint despite the problems associated with the surrogate endpoint.58 TARGETED THERAPY AND CYTOSTATIC It has been argued that the traditional way of DRUGS moving to phase III trials based on the results of phase II trials perhaps was the cause of fail- Advances in molecular biology and cancer genet- ure of many experimental therapies including the ics coupled with biotechnology are bringing forth recent failure of experimental therapies including a number of new novel agents which appear novel targeted therapies such as matrix metallo- to target molecular pathways such as cancer proteinase inhibitors and epidermal growth factor initiation, angiogenesis, invasion or metastasis. receptor inhibitors.59 One approach is to com- Examples include antiangiogenesis agents, epi- bine phase II and III trials into two-stage designs dermal growth factor receptor inhibitors, pro- involving selection and testing based on accept- tein kinase inhibitors, matrix metalloproteinase able primary endpoints or in combination with inhibitors and other so-called molecular targeted auxiliary endpoints for phase III trials.60,61 A therapies. These new agents are not expected sequential Bayesian phase II/III design has been to shrink tumours. Instead they are expected to proposed for a non-small-cell lung cancer involv- inhibit tumour growth or prevent metastasis as ing an adjuvant adenovirus for p53.62 In this they have demonstrated in a number of animal Bayesian design, local control of unresectable models. With the emergence of these different stage II or III NSCLC and overall survival are classes of agents with entirely different mode of 168 TEXTBOOK OF CLINICAL TRIALS action and expected therapeutic effects, the tradi- REFERENCES tional designs for phase I, II and III clinical trials 64–66 appear no longer adequate. 1. Jemal A, Murray T, Samuels A, Ghafoor A, With these cytostatic agents, it is unclear Ward E, Thun MJ. Cancer statistics, 2003. CA whether there is a clear dose–toxicity and Cancer J Clin (2003) 53: 5–26. dose–response relationship to help guide us 2. Mountain CF. Revisions in the international sys- tem for staging of lung cancer. Chest (1997) 111: in determining the most appropriate dose for 1710–17. phase II and III clinical trials. Indeed the 3. Frost JK, Ball WC, Levin M, Tockman MS, Baker paradigm for dose-escalation designs for cyto- RR, Carter D, Eggelston JC, Erozan YS, Gupta toxic agents for phase I clinical trials appears no PK, Khouri NF, Marsh BR, Stitik FP. Early lung cancer detection: results of the initial (prevalence) longer relevant as acute toxicities may not be radiologic and cytologic screening in the Johns meaningful with such agents. Hopkins study. Am Rev Respir Dis (1984) 130: This obviously calls for new methods for esti- 549–54. mating a safe, but effective dose in phase I clini- 4. Flehinger BJ, Melamed MR, Zaman MB, Hee- lan RT, Perchick WB, Martini N. Early lung can- cal trials. In such a setting with cytostatic drugs, cer detection: results of the initial (prevalence) it was suggested that a biological endpoint other radiologic and cytologic screening in the Memo- than toxicity be used in phase I trials to define the rial Sloan–Kettering study. Am Rev Respir Dis dose for subsequent phase II trials.64 For phase II (1984) 130: 555–60. 5. Fontana RS, Sanderson DR, Taylor WF, Wool- preliminary efficacy screening trials, single-arm ner LB, Miller WE, Muhm JR, Uhlenhopp MA. designs can be used in which comparisons are Early lung cancer detection: results of the initial made with historical control data. Sequentially (prevalence) radiologic and cytologic screening in measured times to disease progression within the Mayo Clinic Study. Am Rev Respir Dis (1984) 130: 561–5. each patient who have failed previous treatment 6. Patz EF, Goodman PC, Bepler G. Screening for may be used in phase II designs where statisti- lung cancer. N Engl J Med (2000) 343: 1627–33. cal hypotheses regarding a hazard ratio of times 7. Henschke CI, Naidich DP, Yankelevitz DF, Mc- to disease progression before and after treatment Guinness G, McCauley DI, Smith JP, Libby DM, with cytostatic drugs can be tested.65 Considering Pasmantier MW, Vazquez M, Koizumi J, Flieder D, Altorki NK, Miettinen OS. Early Lung the heterogeneity of cancer, one may wish to dis- Cancer Action Project: initial findings on repeat tinguish antiproliferative activity attributable to screening. Cancer (2001) 92: 153–9. cytostatic drugs from less aggressive disease in 8. van Zandwijk N, Dalesio O, Pastorino U, de phase II screening trials. In that setting, one may Vries N, van Tinteren H. Euroscan, a randomized trial of vitamin A and N-acetylcysteine in patients use a randomised discontinuation design in which with head & neck cancer or lung cancer. JNatl all patients are treated initially with the cytostatic Cancer Inst (2000) 92: 977–86. drug and only those whose disease is stable are 9. Omenn GS, Goodman GE, Thornquist MD, randomised in a double-blind fashion to the same Balmes J, Cullen MR, Glass A, Keogh JP, Mey- 66 skens FL, Valanis B, Williams JH, Barnhart S, cytostatic drug, active vs. placebo. Hammar S. Effects of a combination of beta As illustrated above, these new classes of carotene and vitamin A on lung cancer and drugs will challenge the existing paradigm for cardiovascular disease. N Engl J Med (1996) 334: design, conduct and analysis of phase I, II and 1150–5. 10. The Alpha-Tocopherol, Beta Carotene Cancer III clinical trials in cancer. These challenges are Prevention Study Group. The effect of vitamin E certainly not unique to lung cancer clinical trials. and beta carotene on the incidence of lung cancer Clinical investigators and statisticians need to and other cancers in male smokers. N Engl J Med work more closely to address these challenges (1994) 330: 1029–35. 11. Clark LC, Combs GF, Turnbull BW, Slate EH, in developing most efficient and relevant clinical Chalker DK, Chow J, Davis LS, Glover RA, Gra- trial designs to help advance the care of patients ham GF, Gross EG, Krongrad A, Lesher JL, Park with lung cancer. HK, Sanders BB, Smith CL, Taylor JR. Effects of RESPIRATORY CANCERS 169
selenium supplementation for cancer prevention in 20. Dillman RO, Herndon J, Seagren SL, Eaton WL, patients with carcinoma of the skin: a random- Green MR. Improved survival in stage III non- ized controlled trial. J Am Med Assoc (1996) 276: small cell lung cancer: seven-year follow-up of 1957–63. cancer and leukemia group B (CALGB) 8433. J 12. PORT Meta-analysis Trialists Group. Postopera- Natl Cancer Inst (1996) 88: 1210–15. tive radiotherapy in non-small-cell lung cancer: 21. Sause W, Kolesar P, Taylor S, Johnson D, Liv- systematic review and meta-analysis of individ- ingston R, Komaki R, Emami E, Curran W, ual patient data from nine randomised controlled Byhardt R, Fisher B. Final results of phase III trial trials. Lancet (1998) 352: 257–63. in regionally advanced unresectable non-small cell 13. The Lung Cancer Study Group. Effects of post- lung cancer: Radiation Therapy Oncology Group operative mediastinal radiation on completely (RTOG) 88-08 and Eastern Cooperative Oncology resected stage II and stage III epidermoid cancer Group (ECOG) 4588. Chest (2000) 117: 358–64. of the lung. N Engl J Med (1986) 315: 1377–81. 22. Furuse K, Fukuoka M, Kawahara M, Nishikawa H, 14. Non-small Cell Lung Cancer Collaborative Group. Takada Y, Kudoh S, Katagami N, Ariyoshi Y. Chemotherapy in non-small cell lung cancer: a Phase III study of concurrent versus sequential meta-analysis using updated data on individual thoracic radiotherapy in combination with mit- patients from 52 randomised clinical trials. Br Med omycin, vindesine, and cisplatin in unresectable J (1995) 311: 899–909. stage III non-small-cell lung cancer. J Clin Oncol 15. Keller SM, Adak S, Wagner H, Herskovic A, (1999) 17: 2692–9. Komaki R, Brooks BJ, Perry MC, Livingston RB, 23. Curran W, Scott C, Langer C, Komaki R, Lee J, Johnson DH for the Eastern Cooperative Oncol- Hauser S, Movsas B, Wasserman T, Russell A, ogy Group. A randomized trial of postopera- Byhardt R, Sause W, Cox J. Phase III comparison tive adjuvant therapy in patients with completely of sequential vs. concurrent chemoradiation for resected stage II or IIIA non-small cell lung can- patients with unresected stage III non-small cell cer. N Engl J Med (2000) 343: 1217–22. lung cancer (NSCLC): initial report of the Radia- 16. Rosell R, Gomez-Codina J, Camps C, Maestre J, tion Therapy Oncology Group (RTOG) 9410. Proc Padille J, Canto A, Mate JL, Li S, Roig J, Olaz- Am Soc Clin Oncol (2000) 18: 484a. abal A, Canela M, Ariza A, Skacel Z, Morera- 24. Choy H, Akerley W, Safran H, Graziano S, Prat J, Abad A. A randomized trial comparing Chung C, Williams T, Cole B, Kennedy T. preoperative chemotherapy plus surgery with Multiinstitutional phase II trial of paclitaxel, surgery alone in patients with non-small-cell lung cancer. N Engl J Med (1994) 330: 153–8. carboplatin, and concurrent radiation therapy for 17. Roth JA, Fossella F, Komaki R, Ryan MB, Put- locally advanced non-small-cell lung cancer. J nam JB, Lee JS, Dhingra H, DeCaro L, Chasen M, Clin Oncol (1998) 16: 3316–22. McGavran M, Atkinson EN, Hong WK. A ran- 25. Soquet PJ, Chauvin F, Boissel JP, Cellerino R, domized trial comparing perioperative chemother- Cormier Y, Ganz PA, Kaasa S, Pater JL, Quoix E, apy and surgery with surgery alone in resectable Rapp E, Tunarello D, Williams J, Woods BL, stage IIIA non-small-cell lung cancer. J Natl Can- Bernard JP. Polychemotherapy in advanced non- cer Inst (1994) 86: 673–80. small cell lung cancer: a meta-analysis. Lancet 18. Roth JA, Atkinson EN, Fossella F, Komaki R, (1993) 342: 19–21. Ryan MB, Putnam JB, Lee JS, Dhingra H, De- 26. Marino P, Pampallona S, Preatoni A, Cantoni A, Caro L, Chasen M, Hong WK. Long-term follow- Inveinezzi F. Chemotherapy vs. supportive care in up of patients enrolled in a randomized trial advanced non-small cell lung cancer: results of a comparing perioperative chemotherapy and sur- meta-analysis of the literature. Chest (1994) 106: gery alone in resectable stage IIIA non-small-cell 861–5. lung cancer. Lung Cancer (1998) 21: 1–6. 27. Sandler A, Nemunaitis J, Denham C, von Pawel J, 19. Depierre A, Milleron B, Moro-Sibilot, Chevret S, Cormier Y, Gatzemeier U, Mattson K, Mane- Quoix E, Lebeau B, Braun D, Breton J-L, Le- gold C, Palmer M, Gregor A, Nguyen B, Niyik- marie E, Gouva S, Paillot N, Brechot J-M, Jan- iza C, Einhorn L. Phase III trial of gemcitabine icot H, Lebas F-X, Terrioux P, Clavier J, plus cisplatin versus cisplatin alone in patients Foucher P, Monchatre M, Coetmeur D, Level M- with locally advanced or metastatic non-small cell C, Leclerc P, Blanchon F, Rodier J-M, Thiber- lung cancer. J Clin Oncol (2000) 18: 122–30. ville L, Villeneuve A, Westeel V, Chastang C. 28. Wozniak AJ, Crowley JJ, Balcerzak SP, Weiss Preoperative chemotherapy followed by surgery GR, Spiridonidis CH, Baker LH, Albain KS, compared with primary surgery in resectable Kelly K, Taylor SA, Gandara DR, Livingston RB. stage I (except T1N0), II, and IIIa non-small cell Randomized trial comparing cisplatin with cis- lung cancer. J Clin Oncol (2002) 20: 247–53. platin plus vinorelbine in the treatment of advanced 170 TEXTBOOK OF CLINICAL TRIALS
non-small-cell lung cancer: a Southwest Oncology Crawford J, Lutzker SG, Lilenbaum R, Helms L, Group study. J Clin Oncol (1998) 16: 2459–65. Wolf M, Averbuch S, Ochs J, Kay A. A phase II 29. Lilenbaum RC, Herndon J, List M, Desch C, Wat- trial of zd1839 (‘Iressa’) in advanced non-small son D, Holland J, Weeks JC, Green MR. Sin- cell lung cancer patients who had failed platinum- gle agent (SA) versus combination chemotherapy and docetaxel-based regiments (IDEAL 2). Proc (CC) in advanced non-small cell lung cancer: a Am Soc Clin Oncol (2002) 20: 292a. CALGB randomized trial of efficacy, quality of 37. Johnson D, Herbst R, Giaccone G, Schiller J, life (QoL), and cost effectiveness. Proc Am Soc Natale R, Miller V, Wolf M, Holton A, Aver- Clin Oncol (2002) 20:1a. buch S, Grous J. Zd1839 (‘Iressa’) in combination 30. Sederholm C. Gemcitabine (G) compared with with paclitaxel and carboplatin in chemotherapy- gemcitabine plus carboplatin (GC) in advanced naive patients with advanced non-small-cell lung non-small cell lung cancer (NSCLC): a phase III cancer: results from a phase III clinical trial study by the Swedish Lung Cancer Study Group (Intact2). Ann Oncol (2002) 13(Suppl 5): 127–8. (SLUSG). Proc Am Soc Clin Oncol (2002) 20: 38. Giaccone G, Johnson DH, Manegold C, Scagliotti 1162a. GV, Rosell R, Wolf M, Rennie P, Ochs J, Aver- 31. Kelly KJ, Crowley J, Bunn PA, Presant CA, buch S, Fandi A. A phase III clinical trial of Grevstad PK, Moinpour CM, Ramsey SD, Woz- zd1839 (‘Iressa’) in combination with gemcitabine niak AJ, Weiss GR, Moore DF, Israel VK, Liv- and cisplatin in chemotherapy-naive patients with ingston RB, Gandara DR. Randomized phase III advanced non-small-cell lung cancer (Intact 1). trial of paclitaxel plus carboplatin versus vinorel- Ann Oncol (2002) 13(Suppl 5): 2. bine plus cisplatin in the treatment of patients with 39. Giaccone G, Johnson DH, Manegold C, Scagliotti advanced non-small-cell lung cancer: a Southwest GV, Rosell R, Wolf M, Rennie P, Ochs J, Aver- Oncology Group trial. J Clin Oncol (2001) 19: buch S, Fandi A. Irinotecan plus cisplatin com- 3210–18. pared with etoposide plus cisplatin for extensive 32. Schiller JH, Harrington D, Belani CP, Langer C, small cell lung cancer. N Engl J Med (2002) 346: Sandler A, Krook J, Zhu Z, Johnson DH for the 85–91. Eastern Cooperative Oncology Group. Compari- 40. Schiller J, Kim K, Hutson P, DeVore R, Glick J, son of four chemotherapy regimens for advanced Stewart J, Johnson D. Phase II study of topotecan non-small cell lung cancer. N Engl J Med (2002) in patients with extensive-stage small cell carci- 346: 92–8. noma of the lung: an Eastern Cooperative Oncol- 33. Shepherd F, Dancey J, Ramlau R, Mattson K, ogy Group trial (E1592). J Clin Oncol (1996) 14: Gralla R, O’Rourke M, Levitan N, Gressot L, 2345–52. Vincent M, Burkes R, Coughlin S, Kim Y, Be- 41. von Pawel J, Schiller J, Sheperd F, Kleisbauer J, rille J. Prospective randomized trial of docetaxel Chrysson N, Steward D, Fields S, Clark P, Pal- versus best supportive care in patients with mer M, Depierre A, Carmichael J, Krebs J, non-small cell lung cancer previously treated Ross G, Gralla R. Topotecan versus CAV for the with platinum-based chemotherapy. J Clin Oncol treatment of recurrent small cell lung cancer. J (2000) 18: 2095–103. Clin Oncol (1999) 17: 658–67. 34. Fossella FV, DeVore R, Kerr RN, Crawford J, 42. Pignon J, Arriagada R, Ihde D, Johnson D, Natale RR, Dunphy F, Kalman L, Miller V, Lee Perry M, Souhami R, Brodin O, Joss R, Kies M, JS, Moore M, Gandara D, Karp D, Vokes E, Lebeau B, Onoshi T, Osterlind K, Tattersall M, Kris M, Kim Y, Gamza F, Hammershaimb L. Wagner H. A meta-analysis of thoracic radio Randomized phase III trial of docetaxel ver- therapy for small-cell lung cancer. N Engl J Med sus vinorelbine or ifosfamide in patients with (1992) 327: 618–24. advanced non-small cell lung cancer previously 43. Turrisi AT, Kim K, Blum R, Sause WT, Liv- treated with platinum-containing chemotherapy ingston RB, Komaki R, Wagner H, Aisner S, regimens. J Clin Oncol (2000) 18: 2354–62. Johnson DH. Twice-daily compared with once- 35. Fukuoka M, Yano S, Giaccone G, Tamura T, Nak- daily thoracic radiotherapy in limited small-cell agawa K, Douillard J-Y, Nishiwaki Y, Van- lung cancer treated with concurrently with cis- steenkiste JF, Kudo S, Averbuch S, Macleod A, platin and etoposide. N Engl J Med (1999) 340: Feyereislova A, Baselga J. Final results from a 265–71. phase II trial of zd1839 (‘Iressa’) for patients with 44. Arriagada R, Le Chevalier T, Borie F, Riviere A, advanced non-small cell lung cancer (IDEAL 1). Chomy P, Monnet I, Tardivon A, Viader F, Proc Am Soc Clin Oncol (2002) 20: 298a. Tarayre M, Benhamou S. Prophylactic cranial 36. Kris MG, Natale RB, Herbst RS, Lynch TJ, Pra- irradiation for patients with small-cell lung cancer ger D, Belani CP, Schiller JH, Kelly K, Spiri- in complete remission. J Natl Cancer Inst (1995) donidis C, Albain KS, Brahmer JR, Sandler A, 87: 183–90. RESPIRATORY CANCERS 171
45. Auperin A, Arriagada R, Pignon J, Pechoux C, 55. Fleming TR. One-sample multiple testing proce- Gregor A, Stephens R, Kristjansen P, Johnson B, dures for phase II trials. Biometrics (1982) 38: Ueoka H, Wagner H, Aisner J. Prophylactic cra- 143–51. nial irradiation for patients with small cell lung 56. Simon R. Optimal two-stage designs for phase II cancer in complete remission. N Engl J Med clinical trials. Contr Clin Trials (1989) 10: 1–10. (1999) 341: 476–84. 57. Jung SH, Carey M, Kim K. Graphical search for 46. Schneiderman MA. Mouse to man: statistical two-stage designs for phase II clinical trials. Contr problems in bringing a drug to clinical trial. Proc Clin Trials (2001) 22: 367–72. Fifth Berkeley Symp Math Stat Prob (1967) 4: 58. Fleming TR. DeMets DL. Surrogate endpoints in 855–66. clinical trials: are we being misled? Ann Int Med 47. Geller NR. Design of phase I and II clinical trials (1996) 125: 605–13. in cancer: a statistician’s view. Cancer Invest 59. Scher HI, Heller G. Picking the winners in a sea (1984) 2: 483–91. of plenty? Clin Cancer Res (2002) 8: 400–404. 48. Storer B, DeMets DL. Current phase I/II designs: 60. Thall PF, Simon R, Ellenberg SS. Two-stage are they adequate? J Clin Res Drug Dev (1987) 1: selection and testing designs for comparative 21–30. clinical trials. Biometrika (1988) 75: 303–10. 49. Storer B. Design and analysis of phase I clinical 61. Schaid DJ, Wieand S. Therneau TM. Optimal trials. Biometrics (1989) 45: 925–37. two-stage screening designs for survival compar- 50. Simon R, Freidlin B, Rubinstein L, Arbuck SG, isons. Biometrika (1990) 77: 507–13. Collins J, Christian MC. Accelerated titration de- 62. Inoue LYT, Thall PF, Berry DA. Seamlessly ex- signs for phase I clinical trials in oncology. JNatl panding a randomized phase II trial to phase III. Cancer Inst (1997) 89: 1138–47. Biometrics (2002) 58: 823–31. 51. O’Quigley J, Pepe M, Fisher L. Continual re- 63. Moore TD, Korn EL. Phase II trial design consid- assessment method: a practical design for phase I erations for small-cell lung cancer. J Natl Cancer clinical trials in cancer. Biometrics (1990) 46: Inst (1992) 8: 150–4. 33–48. 64. Korn EL, Arbuck SG, Pluda JM, Simon R, Kaplan 52. Cheung YK, Chappell R. Sequential designs for RS, Christian MC. Clinical trials designs for phase I clinical trials with late-onset toxicities. cytostatic agents: are new approaches needed? J Biometrics (2000) 56: 1177–82. Clin Oncol (2001) 19: 265–72. 53. Gehan E. The determination of the number of 65. Mick R, Crowley JJ, Carroll RJ. Phase II clinical patients required in a preliminary and a follow-up trial design for noncytotoxic anticancer agents for trial of a new chemotherapeutic agent. J Chronic which time to disease progression is the primary Dis (1961) 13: 346–53. endpoint. Proc Am Assoc Clin Res (1999) 59: 228a. 54. Lee YJ, Staquet M, Simon R, Catane R, Mug- 66. Rosner GL, Stadler W, Ratain MJ. Randomized gia F. Two-stage plans for patient accrual in discontinuation design: application to cytostatic phase II cancer clinical trials. Cancer Treat Rep antineoplastic agents. J Clin Oncol (2002) 20: (1979) 63: 721–61. 4478–84. CARDIOVASCULAR
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 12 Cardiovascular
LAWRENCE FRIEDMAN AND ELEANOR SCHRON National Heart, Lung, and Blood Institute, Bethesda, MD 20892 2482, USA
INTRODUCTION public benefits, particularly if the treatment is simple and inexpensive, such as aspirin. Again Most of the principles in developing, managing because the condition is common, when there are and analysing clinical trials in cardiovascular potential treatments that are simple to administer, diseases are the same as for other conditions. trials may be conducted in communities outside There are some special aspects, however, that we of academic health centres, with their special will review here, and then provide examples for expertise and facilities. in the subsequent sections. Another important consideration is that athero- A major point that influences many of the clini- sclerotic and hypertensive cardiovascular dis- cal trials is that in developed countries, and unfor- eases are chronic conditions, often taking decades tunately more and more in developing nations, to develop and lasting many years after being cardiovascular disease is common. Atherosclero- first diagnosed. Clinical trials, therefore, may be sis and hypertension are the primary causes of initiated well before the development of risk fac- cardiovascular disease in adults, although there tors (sometimes called primordial prevention), are many contributing factors to these (often after the development of risk factors, but before termed risk factors). Although clinical trials have the occurrence of organ damage (primary pre- been conducted in other causes of cardiovas- vention), or after organ damage has occurred cular disease, including congenital conditions, (secondary prevention). Some interventions are there are far fewer trials in these areas. There- potentially useful in all three settings, but others, fore, because in developed countries most heart particularly expensive or invasive approaches, disease, stroke and peripheral vascular diseases may be best suited for secondary prevention. are due to atherosclerosis and hypertension, and The relative importance of risk factors and clin- because most of the cardiovascular disease clini- ical findings may also differ, affecting the likely cal trials have been conducted in developed coun- impact of the interventions. After a heart attack, tries, this chapter will emphasise those. for example, how well the surviving myocardium Because cardiovascular diseases are common, functions in pumping blood may be more impor- small treatment benefits may yield important tant in determining longevity than cholesterol
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 176 TEXTBOOK OF CLINICAL TRIALS level (though the latter has been clearly shown the body, such as the kidneys or the legs (as to affect recurrent infarctions and death). Often, in intermittent claudication), may be affected. treatments for atherosclerotic or hypertensive car- Even within a single organ such as the heart, diovascular diseases do not cure the underlying the presentation of cardiovascular disease may conditions. Rather, they may reduce the likeli- take various forms, such as angina pectoris, hood of having clinical sequelae or control the myocardial infarction, cardiac arrhythmias of serious consequences of disease. different sorts, heart failure and sudden death. As noted, the common causes of cardiovas- The interventions studied in clinical trials may cular diseases (atherosclerosis and hypertension) be directed at any of those, though some are multifactorial in origin. Atherosclerosis may interventions, such as blood pressure lowering be influenced by cholesterol level, blood pres- drugs, may affect more than one outcome. sure, cigarette smoking, obesity, physical activ- One issue in designing and interpreting the ity, inflammatory processes, diabetes, age and results of cardiovascular disease trials is choice of genetics, plus other factors. Hypertension may be the outcome. As noted, interventions may affect influenced by things such as diet (intake of salt only one aspect of the disease. Therefore, inves- and various nutrients), obesity, physical activ- tigators would prefer to use as the outcome of ity, stress or emotion, and genetics. With regard interest only that most likely to be modified by to genetic influences, it is not thought that the the treatment. But for outcomes such as cause- common cardiovascular diseases or their risk fac- specific mortality, that is not easy. Even a rather tors are influenced by single genes. Rather, there broad outcome such as death due to cardiovascu- are likely to be many genes that interact with lar disease has limitations, because many deaths environmental conditions to effect most common are unwitnessed and autopsies much less common cardiovascular diseases. Because of the multiple than in the past. When finer splits are used, such risk factors, clinical trial interventions that alter as death due to arrhythmia or myocardial infarc- individual factors might yield only modest reduc- tion, the difficulties mount.2 Similar problems tions in clinical outcomes such as myocardial exist for non-fatal events, such as myocardial infarction or death from cardiovascular causes. infarction. In the Framingham Heart Study, more Antihypertensive drugs, though, that lower blood than 25% of myocardial infarctions were ‘silent’, pressure regardless of the reason for the hyper- that is, occurred without symptoms and were only tension, have been shown to give impressive recognised on electrocardiographic examination.3 reductions in stroke,1 much of which is due to Extra efforts need to be made to identify these, hypertension. On the other hand, treating hyper- as they convey considerable risk of death, even tension has led to only modest reductions in heart though they are asymptomatic. disease, probably because the other risk factors Because many of the cardiovascular disease for heart disease were unchanged. The multifac- conditions are common, there is rarely a short- torial nature of much cardiovascular disease has age of people with the condition of interest led some to design trials that have attempted to for most clinical trials. Particularly with mul- intervene on several factors simultaneously. ticentre, in fact, multinational, trials, there are Whether as part of a clinical trial, or because adequate numbers of potential participants so of usual clinical care, many participants in car- that studies using clinical outcomes are feasi- diovascular disease clinical trials are on multi- ble. It is usually not necessary to resort to trials ple interventions. This can affect adherence to with surrogate endpoints on account of partici- study protocol and may lead to various drug pant unavailability. However, when a trial seeks interactions. special subtypes of cardiovascular disease, sub- We typically think of cardiovascular disease as ject availability becomes more of an issue. It is affecting the heart, as in coronary heart disease, still usually unnecessary to conduct trials with or the brain, as in stroke, but other parts of surrogate endpoints, but extra efforts do have CARDIOVASCULAR 177 to be made to identify and enroll the partic- TRIALS OF PHARMACOLOGIC AGENTS ipants. Connected with this, many physicians sub-specialise in particular types of cardiovas- Pharmaceutical agents are the most common cular disease. Involving the kind of physician interventions tested in clinical trials of cardio- most likely to have knowledge of and access vascular disease. Most trials of drugs are sim- to the relevant patient population is therefore ilar in structure and design to trials in any 4 essential. other field of medicine. A few points, how- The large size of many cardiovascular ever, should be made. Because, as noted above, clinical trials means that stratification to ensure most cardiovascular disease takes decades to balance among key baseline factors is usually develop, there is a long period when people unnecessary, with the notable exception of site, have few if any symptoms. The likely exis- in multicentre trials. Site is almost always a tence of atherosclerosis, for example, is usu- stratification variable. Beyond that, investigators ally determined by the presence of risk factors, should stratify on at most a very few variables. such as hypertension, hyperlipidaemia or fam- For example, for studies of heart failure, we ily history, advanced age in people who have often stratify by left ventricular ejection fraction; the typical lifestyle of most Western countries, for studies of arrhythmia, we might stratify by and the use of sophisticated imaging methods. type of arrhythmia and ejection fraction; for Prevention of the sequelae of atherosclerosis is primary prevention trials, we might stratify by therefore quite feasible. Because participants in age and sex; and for trials of blood pressure trials of primary prevention are asymptomatic, lowering, we would stratify by prior use of several principles apply. First, serious or trou- antihypertensive drugs. blesome adverse events due to interventions in One feature of trials in cardiovascular dis- people who are generally healthy are unaccept- ease that always needs to be considered is the able and not tolerated by the participants. There- dramatic reduction in mortality from heart dis- fore, only drugs that are well-characterised (and ease and stroke over the past few decades in are presumably safe and well-tolerated) are gen- most developed countries.5 As a result of a erally studied in prevention trials. Second, the combination of improved prevention and much rate of clinical events is likely to be low. Unless better medical care, death rates in developed the trials use surrogate outcomes, they need to countries have decreased to a level that makes be very large (thousands and sometimes tens of mortality outcome studies less feasible. From a thousands of participants) and long (often five public health and a patient standpoint, this is years or more). Third, people who are asymp- certainly a happy state to be in, but it means tomatic and consequently notice no obvious ben- that clinical trials must be designed with the efit from treatment may have trouble adhering to expectation that the event rates may be consider- the regimen, especially in a long trial. Therefore, ably less than expected. The trials need either a considerable ‘drop-out’ rate must be built into to be much larger, or else the outcome needs the sample estimate, increasing the size of the to be a combination of death and other clini- trial even more. cal events. Among the first large clinical trials in car- The remainder of this chapter will consider diovascular disease were trials of lipid lower- issues in specific trials. It is divided into trials ing. The Coronary Drug Project, which began of drugs or biologics, trials of devices and in the 1960s, tested five interventions (clofibrate, surgical procedures, and trials of lifestyle or nicotinic acid, dextrothyroxine, and two doses of other non-pharmacologic interventions. For fuller equine estrogen) against a placebo in men with a discussions of various cardiovascular disease history of a myocardial infarction.7,8 These early clinical trials, see the book, Clinical Trials in efforts at lipid modification were only modestly Cardiovascular Disease.6 successful. The interventions had major adverse 178 TEXTBOOK OF CLINICAL TRIALS events and three were stopped before the sched- reductions in all-cause mortality and in coronary uled end of the trial. In addition to the adverse heart disease events. events, the amount of lipid lowering was rela- The 4S trial compared simvastatin against tively small, in general, so the lack of improve- placebo in 4444 participants with known coro- ment in mortality, the primary outcome, was nary heart disease and elevated serum cholesterol. perhaps not surprising. Nicotinic acid was shown The intervention lowered low-density lipoprotein to reduce non-fatal reinfarction, however, and cholesterol by 35% and mortality by 30%.11 The post-trial follow-up disclosed a significant reduc- CARE trial compared pravastatin against placebo tion in mortality.9 A key finding in the Coronary in 4159 post-myocardial infarction patients. The Drug Project was that the mortality rate in the baseline serum cholesterol level was somewhat control group was only two-thirds of that pre- lower in this trial than in the 4S trial. As in dicted when the study was started. This proba- 4S, there was greater than a 30% reduction in bly reflected selection of better risk participants, low-density lipoprotein cholesterol and a 24% but improved care may also have played a role. relative benefit in the primary outcome of coro- This phenomenon is one that many cardiovas- nary heart disease death plus non-fatal myocar- 12 cular trials conducted since the Coronary Drug dial infarction. The MRC/BHF Heart Protection Project have had to take into account in the sam- Study extended these finding to those who had coronary heart disease or were otherwise at high ple size estimates. 13 The next large lipid-lowering trial was the risk, regardless of the baseline cholesterol level. Lipid Research Clinics Coronary Primary Preven- As a result of these and other trials of choles- terol lowering and trials of blood pressure reduc- tion Trial.10 This trial compared cholestryramine tion, new evidence-based guidelines for treatment resin versus placebo to see if there would be a dif- of risk factors such as hyperlipidaemia and hyper- ference in the primary outcome of coronary heart tension have been developed and widely dissem- disease death or non-fatal myocardial infarction inated. Therefore, regardless of whether the trial in 3806 men free of prior evidence of heart is one of primary prevention or in people with disease, but with hyperlipoproteinaemia. There known end-organ damage, the control group must were 155 events in the intervention group and be adequately treated. This means that the event 187 in the placebo group (one-sided p<0.05). rate in the control group will be less than it has This study was one of the first to show bene- been in past years, making it even more difficult fits from lipid lowering, even though some ques- to detect benefit from a new intervention. tioned the significance of the results, given the An example of a recent trial designed to mimic one-sided test. clinical practice illustrates some of these issues. More definitive outcomes from cholesterol The Antihypertensive and Lipid-Lowering Treat- lowering had to wait for the development of ment to Prevent Heart Attack Trial, or ALLHAT, agents that were more effective in lowering compared treatment, in a blinded fashion, begin- lipids and, importantly, better tolerated. The clin- ning with three different antihypertensive agents ical trials of hydroxymethylglutaryl coenzyme A against the control, which was thiazide diuretic (HMG-CoA) reductase inhibitors (‘statins’) were treatment, in more than 40 000 people aged 55 primarily conducted in the 1990s. Trials such or over who had hypertension and at least one as the Scandinavian Simvastatin Survival Study other risk factor. Thus, the hypertension compo- (4S),11 the Cholesterol and Recurrent Events nent of ALLHAT used an ‘active control’ arm. (CARE) trial12 and the MRC/BHF Heart Pro- The primary outcome of this component was tection Study13 have clearly demonstrated that fatal coronary heart disease or non-fatal myocar- cholesterol lowering in both people with known dial infarction. ALLHAT also included a lipid- heart disease and in those at high risk, but with- lowering agent in over 10 000 of the enrolled out evidence of heart disease, leads to impressive participants in a factorial design. The primary CARDIOVASCULAR 179 outcome for the lipid component was all-cause Rates of death in people who have had a mortality; this part of ALLHAT was an open, or myocardial infarction used to be quite high. Mod- non-blinded study.14–16 ern therapy has reduced those rates remarkably. One of the antihypertensive agents, an alpha- Thus, trials using mortality alone as an end- adrenergic blocker, was stopped early because point may no longer be feasible even in sur- although there was little difference in the primary vivors of a heart attack, unless a very high outcome, there was a significant increase in risk group is studied. This has led to increased heart failure in the alpha-adrenergic blocker arm, use of combination endpoints, such as car- compared with the diuretic arm. The other two diovascular mortality plus non-fatal myocardial antihypertensive treatments, a calcium channel infarction. The Heart Outcomes Prevention Eval- blocker and an angiotensin-converting enzyme uation (HOPE) study compared the angiotensin- inhibitor, continued to the scheduled end of the converting enzyme inhibitor ramipril against trial. There were no differences between either placebo in 9297 people with either known vas- of these arms and the active control thiazide cular disease or diabetes plus another risk factor. diuretic arm for the primary outcome. There were The primary outcome was myocardial infarction, some differences in secondary outcomes, with the stroke or death from cardiovascular causes. Thus, diuretic being superior to the other agents for even though this was a high-risk sample, and heart failure, for example. The blood pressure the sample size was considerable, it was neces- component of ALLHAT showed that in an active sary to have a combination endpoint to achieve control trial, a very large sample size needed to be adequate power. There was a highly significant used to achieve adequate power, even when the and clinically impressive reduction in the primary primary outcome was a combination of events. outcome from ramipril.18 Also contributing to the need for a large sample A similar study is the Prevention of Events was the fact that about 30% of the participants with Angiotensin-Converting Enzyme Inhibitor who had a follow-up visit at five years had Therapy, or PEACE.19 This trial is not yet discontinued the study drug. completed, but it too compares an angiotensin- With respect to the lipid component, there converting enzyme inhibitor (trandolapril) against was no significant difference in total mortality, standard therapy in over 8000 people with despite the fact that many other studies have documented coronary heart disease and a left shown benefit from lipid-lowering treatments.17 ventricular ejection fraction of at least 40%. One explanation may be that there was only The reason for the ejection fraction criterion a modest difference in low-density lipoprotein is that angiotensin-converting enzyme inhibitors cholesterol between the two groups. Almost 30% have been shown to be beneficial in those of the control group participants were receiving with heart failure or low ejection fraction. This non-study lipid-lowering therapy by the end eligibility criterion, however, also means that of the trial. The non-blinded design probably the event rate is lower than if those with helped foster that. The public campaigns aimed impaired ejection fraction were included. So in at getting people to reduce their cholesterol levels order to have adequate power, even in this undoubtedly played a role. Also, people who relatively large study, a combination of events were thought to require lipid-lowering therapy is necessary as the primary outcome. Originally, and those already on such therapy were not the sample size was set at 14 000, and the eligible to be enrolled. Because only those primary outcome was cardiovascular death and already entered in ALLHAT for the hypertension non-fatal myocardial infarction. Early in the trial, component were candidates for the lipid-lowering primarily for feasibility reasons, the sample size component, the originally expected number of was reduced to 8100 and the primary outcome about 20 000 enrollees turned out to be 10 355, expanded to include the need for coronary further limiting the study power. revascularisation procedures. Procedures such as 180 TEXTBOOK OF CLINICAL TRIALS need for revascularisation are often included as use of second, third and even fourth choice part of the endpoint. This can be appropriate, but drugs is built into the protocol. For example, is subject to considerable bias if the trial is not in the Systolic Hypertension in the Elderly blinded, which PEACE is. program,20 a multicentre, randomised, double- Antihypertensive agents and lipid-lowering blind, placebo-controlled, community-based trial, drugs have generally been approved by drug 4736 participants with isolated systolic hyperten- regulatory agencies on the basis of their effects sion were randomised to receive either chlorthali- on blood pressure and cholesterol, rather than on done, 12.5 mg daily, or matching placebo. The their effects on clinical outcomes. The clinical goal systolic blood pressure differed for each outcomes, however, are so important that many participant depending upon initial systolic blood trials have successfully tested their effects on pressure. If the blood pressure remained above death, myocardial infarction and stroke.11–16,18,20 the goal at two consecutive monthly visits, the Cardiac ventricular arrhythmias are known to dose was increased to 25 mg of chlorthalidone correlate with total and sudden death. Therefore, or matching placebo daily. If the participants for years, it was thought that drugs that reduced were still above the goal at two consecutive vis- cardiac arrhythmias should be approved on its, 25 mg of atenolol daily or matching placebo the basis of their antiarrhythmic effect, on was added. In participants who still did not the assumption that they would be clinically reach the goal systolic blood pressure, the dose beneficial. However, when the trials were done was increased to 50 mg of atenolol or matching that looked at clinical outcomes, it was seen that placebo. If atenolol was contraindicated, 0.05 to arrhythmia suppression was not a good surrogate 0.1 mg of reserpine or matching placebo daily for mortality. was substituted. Blood pressure above apri- The Cardiac Arrhythmia Suppression Trial ori established escape levels, despite maximal (CAST) tested whether suppression of ventric- stepped-care therapy or corresponding placebo, ular arrhythmias by any of three antiarrhythmic was an indication for prescribing open-label drugs would reduce the incidence of sudden car- active drug therapy.20 diac death. In the first part of this trial, over Some trials of pharmaceutical agents compare 1700 patients whose ventricular arrhythmias were strategies, rather than drugs. Recently, the Atrial suppressed by encainide, flecainide or moricizine Fibrillation Follow-up Investigation of Rhythm were randomly assigned to the drug that was Management (AFFIRM) evaluated which of two most effective in suppressing the arrhythmia or approaches for treating patients with atrial fib- matching placebo. However, two of the drugs, rillation was better.23,24 Participants had to have encainide and flecainide, were soon seen to sig- atrial fibrillation, be over age 65 (or under 65 and nificantly increase both sudden death and all- at high risk for stroke) and the enrolling physi- cause mortality and they were discontinued early cian had to deem it appropriate to treat patients in the study.21 The study was continued with as part of the assigned strategy for up to five moricizine as the only antiarrhythmic drug. This years. AFFIRM included 4060 people, enrolled at too was stopped ahead of schedule because of over 200 sites in Canada and the United States, adverse trends in mortality.22 As a result of who were randomised to either rhythm or rate CAST, the use of surrogate outcomes in many control strategies for managing their atrial fib- clinical trials has been seriously questioned. rillation. Investigators could select from various Another feature of many drug trials in car- options on an approved menu of pharmacologic diovascular disease is that a ‘stepped care’ and non-pharmacologic therapies. Cardioversion approach is used. This is common in trials and antiarrhythmic drugs were used to maintain of blood pressure lowering. Because a sin- sinus rhythm (called ‘rhythm control’). Agents gle drug is often either insufficiently effec- such as digitalis, calcium channel blockers and tive or not well tolerated by the participant, β-blockers, or ablation of the atrioventricular CARDIOVASCULAR 181 junction and pacemaker implantation, were used experience in implanting a device, an interven- to control the ventricular response rate from the tion might be claimed to be not beneficial, or atrial fibrillation (called ‘rate control’). The trial even harmful, when in the hands of a more skilled showed that there was no significant difference in operator it would be beneficial. This was seen in the primary outcome (all-cause mortality), though the Department of Veterans Affairs trial compar- there was a trend favouring the rate control group, ing surgical and medical management of angina and there were fewer adverse effects in the rate pectoris.25 Thirteen hospitals participated in this control group. trial. Three of the hospitals had surgical mortality considerably greater than the other 10. The results comparing surgery against medical care were TRIALS OF DEVICES AND SURGICAL favourable for surgery among patients at high risk PROCEDURES of death from their disease, even when all 13 hos- pitals were included in the analysis. However, for Devices and surgical procedures are commonly lower risk patients, only the comparison involv- used in patients with heart disease. Examples of ing the 10 hospitals with better surgical results devices are replacements for heart valves, stents showed benefit from surgery. These data may that help keep coronary vessels that have been reflect normal variation, but they raise the issue opened patent, cardiac pacemakers and cardiac of requiring a certain level of experience from defibrillators. Examples of surgical procedures the surgeons before they participate in a trial. are corrections of congenital abnormalities, coro- The issue of skill of the operators who insert a nary artery bypass grafts (CABG) and aneurysm device also needs to be emphasised. Most recent resection. Trials of various surgical procedures clinical trials of device implantation require that are usually surgery versus medical treatment or the operators have experience with a certain min- surgery versus device implantation. Examples imum number of devices before being allowed are coronary artery bypass graft procedures in to participate in the trial. This does not guar- patients with ischaemia that are compared against antee that only highly skilled operators will be use of thrombolytic agents or against implanta- involved, but it means that the trial is a better tion of coronary artery stents, or coronary bypass test of how the device will perform in close to graft procedures in patients with stable angina optimal circumstances. An example is the expe- pectoris that have been compared against best rience required of investigators and the establish- medical therapy. Less often, there are trials com- ment of minimum standards for the device and paring one surgical procedure against another. lead systems used in the Antiarrhythmics Versus Devices may be compared against surgery, as Implantable Defibrillators (AVID) trial.26 noted, or against medical care, or sometimes, A second, related issue is how broadly the against another device. study results can be generalised if only the Obviously, as with all clinical trials, the ques- most experienced surgeons participate in the tions in these kinds of trials need to be important trial. After a drug study shows benefit from and the answer relevant, the study needs to be a new pharmaceutical agent, presumably most appropriately designed and carried out, and the practitioners are able to administer the drug in data must be properly analysed. In addition, there a safe, effective manner. Transferring surgical are certain features of such studies that need to technique and skill from investigators in the trial to be considered. First, there is an unavoidable inte- others is less straightforward. Similarly, if a device gration of the intervention being employed and is shown to be beneficial, and then used more the technique with which it is done. The skill widely by less well-trained operators, the results of the investigator is far more important than will not be as positive as in the clinical trial setting. in, for example, drug trials. Unless the inves- Trials of both devices and surgical procedures tigator has considerable surgical competence or affect the way in which the primary trial outcome 182 TEXTBOOK OF CLINICAL TRIALS is assessed. Because of the invasive nature of Changes in surgical technique or modifications the intervention, it is likely, indeed expected, that in devices while the study is being conducted there will be an early adverse experience associ- can cause difficulties in interpreting the results of ated with the procedure. The trauma involved; the clinical trials. If, partway through a study, there is consequences of anaesthesia, particularly if gen- an important change in the intervention, depend- eral anaesthesia is used; and the risks of infection ing on the outcome, it may be hard to reach a will almost inevitably lead to morbidity and per- clear conclusion about the possible benefits of haps mortality early after the intervention. There- the intervention. In the past, implantable car- fore, the study needs to be designed such that dioverter defibrillators required a thoracotomy. it lasts long enough for any hoped-for benefit This carried considerable risk that needed to to overcome the early unfavourable experience. be considered when inserting the device. Leads Sometimes, the expected benefit does not appear that could be inserted transvenously were sub- for quite some time. Not only the investigators, sequently developed, reducing the early com- but institutional ethics committees and prospec- plication rate. The AVID trial and the Cana- tive study participants need to understand this dian Implantable Defibrillator Study (CIDS) trial, implication. both of which compared implantable cardioverter An example is the Program on the Surgical defibrillators against medical therapy in peo- Control of Hyperlipidemia.27,28 This trial com- ple resuscitated from cardiac arrest, were in the pared partial ileal bypass surgery against medical process of enrolling patients when the switch therapy in patients with a prior myocardial infarc- in practice from primarily using thoracotomy- based defibrillators to transvenous defibrillators tion. The goal was to decrease the absorption of occurred.26,29 Because both types of defibrilla- lipids, thereby reducing serum cholesterol. For tor performed similarly, there was no problem in the first two years of the trial, there was lit- combining the results. This was particularly the tle difference in the primary outcome, all-cause case because it was shown significantly in AVID mortality, with the surgical group doing slightly and with a strong trend in CIDS that the defib- worse than the control group. The curves crossed rillator was more effective in reducing mortality after about three years, and at the scheduled end than antiarrhythmic drug therapy. If there were of the trial, there was a non-significant trend in no difference, or if drug therapy turned out to be favour of surgery. The study investigators fol- superior overall, questions about the validity of lowed the participants after the formal end of the the trial might have been raised, given the mid- trial. The trend in favour of surgery continued and study switches in device use. five years after the end, the mortality difference Trials of cardiovascular devices involve sev- 28 was statistically significant. eral other issues that are not common to other This study shows that data monitoring com- kinds of trials. One is the decision during study mittees need to consider how long to wait to design as to the primary question. The device is see if the benefit appears and counterbalances the developed to meet certain specifications. These known risks. Even though the primary outcome include minimising the possibility of rejection by was not sufficiently adverse early in the trial to the patient, reducing the likelihood of develop- justify stopping, other factors combined with lack ment of thrombi and emboli, physical character- of benefit might have influenced a monitoring istics such as size and weight, and, importantly, committee to do so. For example, in POSCH, does it do what it is designed to do. Does a there were side effects such as diarrhoea and, defibrillator detect and convert life-threatening more seriously, a higher rate of kidney stones rhythm disturbances? Does a pacemaker detect and gallstones.27 The slight early adverse trend in and correct severe bradycardia? Can a stent be mortality plus the increased morbidity could have easily employed and will it retain its structural led to a decision to stop the study prematurely. integrity? These are engineering questions that CARDIOVASCULAR 183 should be addressed and satisfactorily answered in 196 patients with heart failure, a prior myocar- before a clinical trial is conducted. The clinical dial infarction, left ventricular ejection fraction trial should be designed to answer the questions less than or equal to 35%, a documented episode posed by the clinician. Will the device reduce of asymptomatic unsustained ventricular tachy- mortality and/or morbidity, what is the resteno- cardia, and inducible, non-suppressible ventricu- sis/occlusion rate, and what are the risks and side lar tachyarrhythmia on electrophysiologic testing. effects? The answers to these questions incor- In this very high-risk group of patients, the defib- porate the structural and functional aspects of rillator led to highly significant reductions in the device, the skill of the person inserting the all-cause and cardiac mortality.34 A subsequent device, and the often unknown biologic interac- study by the same group of investigators (MADIT tion between patients and the device. II) assessed whether the implantable defibrilla- The fact that only devices designed and tor would reduce mortality in patients with a fully expected to be mechanically functional prior myocardial infarction and left ventricular are used raises a serious ethical issue. If the ejection fraction less than or equal to 30%. Elec- device defibrillates, for example, how can it trophysiologic testing was not used to identify be withheld from someone with known life- high-risk patients. Because the risk of mortality threatening cardiac arrhythmias? This was faced was lower in this study than in the prior study, in the AVID trial.30,31 The rationale for the 742 patients were randomised to receive either trial was that the balance between expected the implantable defibrillator or conventional med- reduction in death due to arrhythmias versus ical therapy. Here too, there was a significant death from other causes, plus adverse events such reduction in mortality in the defibrillator group.35 as infection, and the possible seriously impaired One trial of a device that raised considerable quality of life, was uncertain. questions about ethics was the Randomised Eval- A key issue is the risk level of the patients uation of Mechanical Assistance for the Treat- being enrolled. If the patients are at truly very ment of Congestive Heart Failure (REMATCH), high risk of arrhythmic death, even though which was conducted from 1997 to 2001. This optimal medical therapy is being used, then it trial compared use of a left ventricular assist might be inappropriate to randomise them to device versus medical therapy in 129 patients medical therapy if a possibly useful device or with end-stage heart failure who were not can- surgical procedure exists. The defibrillator trials didates for cardiac transplantation. The one-year that were done showed that in moderately high- survival was 52% in the group receiving the left risk patients, the use of the defibrillator saved ventricular assist device and 25% in the medical lives with an acceptable number of adverse therapy group, a highly significant difference. At events. If the risk level is less, however, as might two years, the rates of survival were 23% and 8%. be the case in patients with better left ventricular There were over twice as many serious adverse ejection fraction, the balance between benefit events (infection, bleeding, device malfunction) and potential harm might remain unknown, and in the device group as in the medical arm.36,37 the use of a defibrillator might not be justified, Because the expected survival rate in these absent clear demonstration of benefit in a clinical patients was so low, there were many questions trial.32,33 about the ethics of randomisation. It was known Trials have looked at various ways of identi- that the device was mechanically sound, and fying patients at sufficiently high risk to see if worked in the short-term as a bridge to trans- defibrillators are beneficial, but not at so high plantation. The justification for the trial was that risk that it would be unethical to randomise. The long-term benefit, either for survival or quality Multicentre Automatic Defibrillator Implantation of life, was unknown. For a more extensive dis- Trial (MADIT) compared use of an implanted cussion of the design issues and ethics of this defibrillator versus conventional medical therapy kind of trial, using REMATCH as an example, 184 TEXTBOOK OF CLINICAL TRIALS see the report of a conference on mechanical car- To the extent that the devices in the ‘strategy- diac support.38 approach’ trial are similarly effective, the results It is uncommon (and often impractical) for can be more broadly generalised to a class of trials of devices or surgical procedures to be devices. There is a risk, however, that devices blinded. Occasionally, however, this can be made by one manufacturer will be better than done. One such trial was Mode Selection Trial others, blurring the outcome of the trial. in Sinus Node Dysfunction (MOST).39 Dual- Another feature of device and surgical trials chamber (atrioventricular) pacing was compared is that unlike most drugs (vaccines being an with single-chamber (ventricular) pacing in 2010 exception) that need to be administered regularly, patients with clinically important bradycardia. devices are implanted and expected to work The primary outcome was death or non-fatal for a long time, and surgery, unless reversed, stroke. All patients received a dual-chamber can be life-long. Batteries and other components device, but in those randomised to single- may need to be replaced, but unless there are chamber pacing, only one lead was activated, problems, they last for years. This is generally a therefore mode of pacing was randomised rather strength of such trials. There is less problem with compliance to protocol and long-term follow- than type of device. Physician investigators and up is not only feasible, but desirable. The patients were blinded regarding whether the several coronary artery bypass graft surgery trials patient was in the dual- or single-chamber arm; assessed outcome 10, and in some cases more cross-over at the last follow-up was 31.4% for than 20 years after the initial procedure.41 those assigned to ventricular pacing, almost half If a drug trial turns out not to show benefit of these due to pacemaker syndrome. This was in from the drug, simply stopping administration contrast to another study that inserted only single- is usually sufficient. But what if the device or chamber devices in those randomised to that surgery trial turns out not to show benefit? What group and dual-chamber devices only in those is the obligation of the investigator, especially if randomised to the dual-chamber group. Here the the device or procedure is shown by the trial to 40 cross-over rate was 2.7%. This latter trial was be harmful? The Coronary Artery Bypass Graft not blinded, but cross-over from single- to dual- (CABG) Patch trial42 compared transthoracic chamber pacing would have required another implantation of cardioverter defibrillators against procedure, accounting for the low cross-over rate. control in patients undergoing CABG surgery. As with many drug trials, trials of devices At the end of an average 32 months follow- can look at either single devices (or upgrades up, there was no significant mortality difference of these devices as they become available) or between the groups. All patients were given the classes of devices. The AVID26 trial compared results of the trial and subsequent therapy was the use of advanced-generation units with tiered individualised. All patients were urged to have therapy capable of antitachycardia pacing, car- electrophysiologic testing to see if they were at dioversion and defibrillation, as well as brady- high risk of serious arrhythmia, and thus possibly cardia pacing, made by more than one company, in need of the defibrillator in the future. About against any of several drugs (though primarily 40% of the patients in the intervention group amiodarone), thus testing whether the strategy of elected to have the device turned off or removed using implantable defibrillators was preferable to [J.T. Bigger, personal communication]. a strategy of pharmacologic approach. This kind of trial is more likely to be done by public organ- isations, such as the National Institutes of Health, TRIALS OF BEHAVIOUR CHANGE than by industry. Industry-supported trials, such as the previously discussed MADIT,34 almost Trials of behaviour change are fairly common always compare a single manufacturer’s device. in heart disease. They include trials aimed at CARDIOVASCULAR 185 smoking prevention or cessation; diet change and pressure can affect the trials in major ways. for weight loss, blood pressure reduction or If there are changes in restaurant or workplace cholesterol reduction; and exercise for risk factor smoking regulations during the time of a trial reduction and better outcomes in people with a looking at ways to get people to stop smoking, heart attack or heart failure. the likely trends in the control group as a result of These sorts of trials have several aspects that the new regulations will make detection of benefit make them different from most other trials. First, from the intervention more difficult. there is considerable cross-over or recidivism by Examples of successful trials of diet inter- the study participants. The interventions are ones vention are the Dietary Approaches to Stop that many can either implement on their own Hypertension (DASH) trial43 and a subsequent or stop because they are difficult to maintain. trial of the DASH diet with different lev- Second, partly as a consequence of the first, the els of sodium intake.44 The first DASH study trials are quite resource intensive on the part of enrolled 459 adults with systolic blood pressure the interventionists. Implantation of devices or under 160 mm Hg and diastolic pressure 80 to surgical procedures require major efforts by the 95 mm Hg. Participants were randomly allocated investigator or surgeon, but usually only on a one- to one of three groups: a diet rich in fruits and time basis. Trials of behaviour change require vegetables; a diet rich in fruits and vegetables considerable effort on a continuing basis to help plus low-fat dairy products and reduced satu- participants stay with a change from what may rated fat (‘combination’ diet); or a control diet have been a life-long habit of smoking, eating with fruit, vegetable and fat content similar to poorly or sedentary lifestyle. Third, again as a that commonly eaten in the United States. At the consequence of the first two factors, the study end of eight weeks, both of the intervention diets duration is often much shorter than with other reduced blood pressure, with a greater reduction types of trials. Getting people to adhere to an from the diet containing low-fat dairy products. exercise programme for months, let alone years, The second DASH study enrolled 412 partici- is extraordinarily difficult. And of course volun- pants in a factorial design trial. Participants were teers willing to be randomised to exercise or no assigned to either the DASH combination diet exercise will have more of an interest in exercise or the control diet and to any of three levels of than the general public. Therefore, those allo- sodium intake. The moderate and low sodium cated to the no-exercise programme group will intake diets reduced blood pressure in both the have more of a tendency to cross-over. Fourth, DASH diet and control diet groups. The DASH because of the generally shorter duration of the diet led to lower blood pressure than the control trials, surrogate outcomes are more often used diet at each sodium level. than in other trials in heart disease. Rather than In the DASH studies, the food was spe- assessing clinical outcomes such as heart dis- cially prepared and provided to the partici- ease or stroke, the behaviour trials will often use pants. Whether or not the DASH-type diet can weight change, biochemical measures, or attitude be maintained, over time, in people obtain- or knowledge assessed by questionnaires or inter- ing their food in the usual way, was stud- views. Fifth, standardisation of the intervention ied in PREMIER, a trial of 810 participants and measurement of the degree of compliance are whose blood pressure is greater than opti- more complicated. Unless highly controlled feed- mal or who have mild hypertension.45 Partic- ing studies are performed, we have only modestly ipants were randomised to one of three arms: good ways of assessing overall food and nutri- advice only; comprehensive lifestyle interven- ent intake. This is particularly so if maintenance tion using behavioural approaches; and a com- of caloric intake is one of the objectives of the bined comprehensive lifestyle intervention plus trial, as weight would not be able to serve as a the DASH diet. The behavioural approaches marker of change in diet. Sixth, societal changes consisted of 18 counselling sessions. Unlike 186 TEXTBOOK OF CLINICAL TRIALS
DASH, the participants were not provided spe- enough magnitude to alter mortality. This is what cially prepared foods. The primary outcome was happened in some of the early trials of cholesterol systolic blood pressure six months after randomi- lowering that failed to show improvement in sation. At six months, the systolic blood pressure mortality. The early lipid-lowering drugs were had decreased by 11.1 mm Hg in the combined not as effective as the current ones in reducing group, 10.5 mm Hg in the behavioural interven- cholesterol. Third, the measures of depression tion group, and 6.6 mm Hg in the advice only and social support, particularly when obtained group. The largest reductions in diastolic blood shortly after a major event such as a heart attack, pressure and in the percentage of people with might not reflect true ‘baseline’. Fourth, however, hypertension were also seen in the combined even though mortality and recurrent infarction group, with the behavioural intervention group a were unchanged, the apparent improvements in close second. Thus, a combination of behavioural depression and social support are not trivial approaches and dietary changes can result in findings. Unlike surrogate outcome variables that meaningful blood pressure reduction. Even so, have little clinical meaning, these outcomes are the DASH diet, unlike in the previous feed- clinically important in their own right. ing studies,43,44 did not contribute significantly Often, it is more appropriate to conduct beyond the comprehensive lifestyle intervention behaviour change trials in community settings, alone. The adoption of the diet in PREMIER with one group of communities compared against was not as intensive as in the feeding studies. another. The changes, in order to be effective, Whether the changes observed in PREMIER per- need to be community-wide. An example of sist at 18 months is being assessed.45 efforts to improve response time to symptoms of A different kind of behavioural intervention a heart attack was the Rapid Early Action for trial was the Enhancing Recovery in Coronary Coronary Treatment (REACT) trial. This study Heart Disease (ENRICHD)46 study. This trial involved 10 matched pairs of cities. One group enrolled 2481 patients at 73 hospitals who had of cities received intervention through the media, had a myocardial infarction within the previous community organisations, and professional and 28 days. In addition, all participants had depres- patient education in an effort to improve the sion, low social support, or both. Because depres- response time in the event of symptoms of an sion and poor social support are associated with acute myocardial infarction. The other group of increased mortality after a heart attack, it was cities served as the control. The primary outcome thought that intervening on those factors might was time from symptom onset to arrival in the lead to improved survival. Those randomised hospital emergency department. REACT showed to intervention received counselling; the control increased use of the emergency medical system, group received usual medical care. Both groups but no difference between the groups in time to received information on heart disease risk factors. arrival at the hospital.48 Although the intervention decreased depression There have been several trials aimed at chang- and improved social support, there was no dif- ing multiple risk factors for heart disease in com- ference at three years in the primary outcome of munity settings.49–51 These trials showed small death or recurrent myocardial infarction (24.1% differences between the intervention communities vs. 24.2%).47 and the control communities in selected variables. The ENRICHD study raises several issues. In general, however, they were not particularly First, despite the association between depression successful in achieving large differences despite and heart disease, treatment of depression may intensive education efforts. The reasons are prob- not lead to change in mortality from heart disease. ably multiple. Among them are that the interven- That is, depression may not be a causative factor. tions were not delivered in sufficiently persua- Second, the observed improvement in depression sive manners and that the control communities and social support may not have been of great showed changes because of national attention to CARDIOVASCULAR 187 the need to stop smoking, improve diet and get Woosley RL, Singh SN, eds, Arrhythmia Treat- better preventive medical attention. ment and Therapy. New York: Marcel Dekker The less than outstanding results from these (2000) 323–42. 5. National Institutes of Health and National Heart, community-wide efforts at behaviour change Lung and Blood Institute. Morbidity & Mortality: illustrate both the difficulties in achieving behav- 2002 Chart Book on Cardiovascular, Lung, and iour change and the problems in community, as Blood Diseases. U.S. Department of Health and opposed to individual, interventions. Human Services, Public Health Service, National Institutes of Health (2002). 6. Clinical Trials in Cardiovascular Disease. A Com- SUMMARY panion to Braunwald’s Heart Disease. Philadel- phia: W.B. Saunders (1999). There are certain features that need to be con- 7. Canner PL, Klimt CR. The Coronary Drug Project. Experimental design features. Contr Clin Trials sidered in cardiovascular disease trials, whether (1983) 4: 313–32. they are primary prevention or secondary preven- 8. The Coronary Drug Project Research Group. tion. Those features include the chronic nature Clofibrate and niacin in coronary heart disease. of the common cardiovascular diseases, the mul- JAMA (1975) 231: 360–81. tiple risk factors responsible for those diseases, 9. Canner PL, Berge KG, Wenger NK, Stamler J, Friedman L, Prineas RJ, et al. Fifteen year mor- and the fact that huge numbers of people develop tality in Coronary Drug Project patients: long-term cardiovascular disease of one form or another. In benefit with niacin. J Am Coll Cardiol (1986) 8: this chapter, we have reviewed these and other 1245–55. factors, and described selected trials of drugs, 10. Lipid Research Clinics Coronary Primary Preven- devices and surgical procedures, and behavioural tion Trial. The Lipid Research Clinics Coronary interventions that exemplify these features. Over- Primary Prevention Trial results. I. Reduction in incidence of coronary heart disease. JAMA (1984) all, however, trials of cardiovascular diseases are 251: 351–64. designed, conducted and analysed in ways similar 11. The Scandinavian Simvastatin Survival Study. to trials of other conditions. Randomised trial of cholesterol lowering in 4444 Future trials may include targeting and ‘opti- patients with coronary heart disease: the Scandi- mising’ interventions based on genotype and the navian Simvastatin Survival Study (4S). Lancet (1994) 344: 1383–9. use of new diagnostic and imaging techniques, 12. Sacks FM, Pfeffer MA, Moye LA, Rouleau JL, potentially yielding better characterisation of the Rutherford JD, Cole TG, et al. The effect of disease or condition pathology. pravastatin on coronary events after myocardial infarction in patients with average cholesterol levels. Cholesterol and Recurrent Events Trial REFERENCES investigators. N Engl J Med (1996) 335: 1001–9. 13. MRC/BHF Heart Protection Study of cholesterol 1. Chobanian AV, Bakris GL, Black HR, Cushman lowering with simvastatin in 20,536 high-risk WC, Green LA, Izzo JL, et al. Seventh report individuals: a randomised placebo-controlled trial. of the Joint National Committee on prevention, Lancet (2002) 360: 7–22. detection, evaluation, and treatment of high blood 14. ALLHAT Collaborative Research Group. Major pressure. Hypertension (2003) 42: 1206–52. cardiovascular events in hypertensive patients 2. Greene HL, Richardson DW, Barker AH, Roden randomized to doxazosin vs chlorthalidone: the DM, Capone RJ, Echt DS, et al. Classification of antihypertensive and lipid-lowering treatment to deaths after myocardial infarction as arrhythmic prevent heart attack trial (ALLHAT). JAMA (2000) or nonarrhythmic (the Cardiac Arrhythmia Pilot 283: 1967–75. Study). Am J Cardiol (1989) 63: 1–6. 15. The ALLHAT Officers and Coordinators for the 3. Kannel WB. Silent myocardial ischemia and ALLHAT Collaborative Research Group. Major infarction: insights from the Framingham Study. outcomes in moderately hypercholesterolemic, Cardiol Clin (1986) 4: 583–91. hypertensive patients randomized to pravastatin 4. Bardy GH, Lee KL, Mark DB, Poole JE, Fish- vs usual care. The Antihypertensive and Lipid- bein DP, Singh SN, et al. The Sudden Car- Lowering Treatment to Prevent Heart Attack Trial diac Death-Heart Failure Trial (SCD-HeFT). In: (ALLHAT-LLT). JAMA (2002) 288: 2998–3007. 188 TEXTBOOK OF CLINICAL TRIALS
16. The ALLHAT Officers and Coordinators for simple clinical variables. Circulation (1981) 63: the ALLHAT Collaborative Research Group. 1329–38. Major outcomes in high-risk hypertensive patients 26. The AVID Investigators. Antiarrhythmics Ver- randomized to angiotensin-converting enzyme sus Implantable Defibrillators (AVID) – rationale, inhibitor or calcium channel blocker vs diuretic. design, and methods. Am J Cardiol (1995) 75: The Antihypertensive and Lipid-Lowering Treat- 470–5. ment to Prevent Heart Attack Trial (ALLHAT). 27. Buchwald H, Varco RL, Matts JP, Long JM, Fitch JAMA (2002) 288: 2981–97. LL, Campbell GS, et al. Effect of partial ileal 17. National Institutes of Health and National Heart, bypass surgery on mortality and morbidity from Lung and Blood Institute. Third Report of the coronary heart disease in patients with hyperc- National Cholesterol Education Program (NCEP) holesterolemia. Report of the Program on the Sur- Expert Panel on Detection, Evaluation, and Treat- gical Control of Hyperlipidemia (POSCH). N Engl ment of High Blood Cholesterol in Adults (Adult JMed(1990) 323(14): 946–55. Treatment Panel III). U.S. Department of Health 28. Buchwald H, Varco RL, Boen JR, Williams SE, and Human Services, Public Health Service Hansen BJ, Campos CT, et al. Effective lipid (2001). modification by partial ileal bypass reduced long- 18. Yusuf S, Sleight P, Pogue J, Bosch J, Davies R, term coronary heart disease mortality and mor- Dagenais G. Effects of an angiotensin-converting- bidity: five-year posttrial follow-up report from enzyme inhibitor, ramipril, on cardiovascular the POSCH. Program on the Surgical Control of events in high-risk patients. The Heart Outcomes the Hyperlipidemias. Arch Intern Med (1998) 158: Prevention Evaluation Study Investigators. N Engl 1253–61. JMed(2000) 342: 145–53. 29. Connolly SJ, Gent M, Roberts RS, Dorian P, 19. Pfeffer MA, Domanski M, Rosenberg Y, Verter J, Roy D, Sheldon RS, et al. Canadian implantable Geller N, Albert P, et al. Prevention of events defibrillator study (CIDS): a randomized trial of with angiotensin-converting enzyme inhibition the implantable cardioverter defibrillator against (the PEACE study design). Prevention of Events amiodarone. Circulation (2000) 101: 1297–302. 30. Epstein AE. AVID necessity. PACE Pacing Clin with Angiotensin-Converting Enzyme Inhibition. Electrophysiol (1993) 16: 1773–5. Am J Cardiol (1998) 82: 25H–30H. 31. Fogoros RN. An AVID dissent. PACE Pacing Clin 20. SHEP Cooperative Research Group. Prevention of Electrophysiol (1994) 17: 1707–11. stroke by antihypertensive drug treatment in older 32. The AVID Investigators. A comparison of anti- persons with isolated systolic hypertension. Final arrhythmic-drug therapy with implantable defib- results of the Systolic Hypertension in the Elderly rillators in patients resuscitated from near-fatal Program (SHEP). JAMA (1991) 265: 3255–64. ventricular arrhythmias. N Engl J Med (1997) 337: 21. Echt DS, Liebson PR, Mitchell LB, Peters RW, 1576–83. Obias-Manno D, Barker AH, et al. Mortality and 33. Domanski MJ, Epstein A, Hallstrom A, Sak- morbidity in patients receiving encainide, fle- sena S, Zipes DP. Survival of antiarrhythmic cainide, or placebo. The Cardiac Arrhythmia Sup- or implantable cardioverter defibrillator treated pression Trial. N Engl J Med (1991) 324: 781–8. patients with varying degrees of left ventricular 22. The Cardiac Arrhythmia Suppression Trial II dysfunction who survived malignant ventricular Investigators. Effect of the antiarrhythmic agent arrhythmias. J Cardiovasc Electrophysiol (2002) moricizine on survival after myocardial infarction. 13: 580–3. N Engl J Med (1992) 327: 227–33. 34. Moss AJ. Background, outcome, and clinical 23. The Planning and Steering Committee of the implications of the Multicentre Automatic Defib- AFFIRM Study for the NHLBI AFFIRM Inves- rillator Implantation Trial (MADIT). Am J Cardiol tigators. Atrial Fibrillation Follow-up Investiga- (1997) 80: 28F–32F. tion of Rhythm Management – the AFFIRM study 35. Moss AJ, Zareba W, Hall WJ, Klein H, Wilber design. Am J Cardiol. (1997) 79: 1198–202. DJ, Cannom DS, et al. Prophylactic implantation 24. The Atrial Fibrillation Follow-up Investigation of of a defibrillator in patients with myocardial Rhythm Management (AFFIRM) Investigators. A infarction and reduced ejection fraction. N Engl comparison of rate control and rhythm control JMed(2002) 346: 877–83. in patients with atrial fibrillation. N Engl J Med 36. Rose EA, Moskowitz AJ, Packer M, Sollano JA, (2002) 347: 1825–33. Williams DL, Tierney AR, et al.TheREMATCH 25. Detre K, Peduzzi P, Murphy M, Hultgren H, trial: rationale, design, and end points. Ran- Thomsen J, Oberman A, et al. Effect of bypass domised Evaluation of Mechanical Assistance for surgery on survival in patients in low- and the Treatment of Congestive Heart Failure. Ann high-risk subgroups delineated by the use of Thorac Surg (1999) 67: 723–30. CARDIOVASCULAR 189
37. Rose EA, Gelijns AC, Moskowitz AJ, Heitjan DF, Approaches to Stop Hypertension (DASH) diet. N Stevenson LW, Dembitsky W, et al. Long-term Engl J Med (2001) 344: 3–10. mechanical left ventricular assistance for end- 45. Writing Group of the PREMIER Collaborative stage heart failure. N Engl J Med (2001) 345: Research Group. Effects of comprehensive lifestyle 1435–43. modification on blood pressure control. Main 38. Stevenson LW, Kormos RL, Bourge RC, Geli- results of the PREMIER clinical trial. JAMA (2003) jns A, Griffith BP, Hershberger RE, et al. Mechan- 289: 2083–93. ical cardiac support 2000: current applications and 46. The ENRICHD Investigators. Enhancing recovery future trial design, June 15–16, 2000, Bethesda, in coronary heart disease (ENRICHD) study Maryland. J Am Coll Cardiol (2001) 37: 340–70. intervention: rationale and design. Psychosom Med 39. Lamas GA, Lee KL, Sweeney MO, Silverman R, (2001) 63: 747–55. Leon A, Yee R, et al. Ventricular pacing or dual- 47. Writing Committee for the ENRICHD Investi- chamber pacing for sinus-node dysfunction. N gators. Effects of treating depression and low Engl J Med (2002) 346: 1854–62. perceived social support on clinical events after 40. Connolly SJ, Kerr CR, Gent M, Roberts RS, myocardial infarction. The Enchancing Recovery Yusuf S, Gillis AM, et al. Effects of physiologic in Coronary Heart Disease Patients (ENRICHD) pacing versus ventricular pacing on the risk of randomized trial. JAMA (2003) 289: 3106–16. stroke and death due to cardiovascular causes. 48. Luepker RV, Raczynski JM, Osganian S, Gold- Canadian Trial of Physiologic Pacing Investiga- berg RJ, Finnegan JR Jr, Hedges JR, et al. Effect tors. N Engl J Med (2000) 342: 1385–91. of a community intervention on patient delay and 41. Yusuf S, Zucker D, Peduzzi P, Fisher LD, emergency medical service use in acute coronary Takaro T, Kennedy JW, et al. Effect of coronary heart disease: the Rapid Early Action for Coro- artery bypass graft surgery on survival: overview nary Treatment (REACT) trial. JAMA (2000) 284: of 10-year results from randomised trials by the 60–7. Coronary Artery Bypass Graft Surgery Trialists 49. Farquhar JW, Fortmann SP, Flora JA, Taylor CB, Collaboration. Lancet (1994) 344: 563–70. Haskell WL, Williams PT, et al. Effects of com- 42. Bigger JT Jr. Prophylactic use of implanted car- munitywide education on cardiovascular disease diac defibrillators in patients at high risk for ven- risk factors. The Stanford Five-City Project. JAMA tricular arrhythmias after coronary-artery bypass (1990) 264: 359–65. graft surgery. Coronary Artery Bypass Graft 50. Luepker RV, Murray DM, Jacobs DR, Mittel- (CABG) Patch Trial Investigators. N Engl J Med mark MB, Bracht N, Carlaw R, et al. Community (1997) 337: 1569–75. education for cardiovascular disease prevention: 43. Appel LJ, Moore TJ, Obarzanek E, Vollmer WM, risk factor changes in the Minnesota Heart Health Svetkey LP, Sacks FM, et al. A clinical trial of Program. Am J Public Health (1994) 84: 1383–93. the effects of dietary patterns on blood pressure. 51. Carleton RA, Lasater TM, Assaf AR, Feldman DASH Collaborative Research Group. N Engl J HA, McKinlay S, Pawtucket Heart Health Pro- Med (1997) 336: 1117–24. gram Writing Group. The Pawtucket Heart Health 44. Sacks FM, Svetkey LP, Vollmer WM, Appel LJ, Program: community changes in cardiovascular Bray GA, Harsha D, et al. Effects on blood risk factors and projected disease risk. Am J Public pressure of reduced dietary sodium and the Dietary Health (1995) 85: 777–85. DENTISTRY AND MAXILLO-FACIAL
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 13 Dentistry and Maxillo-facial
MAY C.M. WONG, COLMAN MCGRATH AND EDWARD C.M. LO Dental Public Health, Faculty of Dentistry, The University of Hong Kong, Hong Kong
INTRODUCTION Aside from oral diseases there are a number of conditions or disorders of the oral cavity that SCOPE OF THE CHAPTER are of concern. Malalignment and malocclusion of teeth (crooked teeth) is prevalent and severe Dentistry is concerned with the prevention and in many countries and most report a growing treatment of diseases and disorders of the teeth, demand for orthodontic treatment to correct the gums (periodontium) and oral cavity. The two malocclusion. In the US, it is estimated that most common oral diseases are dental caries around half of the population are in need of some (tooth decay) and periodontal (gum) diseases. kind of orthodontic treatment to improve their Data from the Global Oral Health Databank occlusion.2 Another problem has been the need of the World Health Organization (http://www. for replacement of missing teeth; congenitally whocollab.od.mah.se/index.html) reports that at absent or lost because of caries or periodontal least half of the children and nearly all of the disease. Removable prosthesis (dentures or false adults in most countries throughout the world teeth) and fixed prosthesis (bridges) as well as have been affected by dental caries. In addition, implants (screw in teeth) have been used to findings from epidemiological surveys throughout address these problems. the world have reported that less than 10% of As the scope of dentistry is very wide, it their adult population have no periodontal disease would not be possible to include all kinds of (completely healthy gums). One of the more clinical trials in dental research in this chapter. life-threatening diseases of the oral cavity is Instead, the discussion here will focus on the oral cancer, primarily cancer involving the oral more common oral diseases and conditions. mucosa (lining of the oral cavity). The prevalence First, the disease aetiology and measurements of oral cancer varies from country to country, in of dental caries and periodontal disease will most countries it accounts for less than 1% of be presented. Second, clinical trials methods the total cancer incidence whereas in the Indian used in dentistry will then be outlined and subcontinent it can account for 30–50% of the illustrated with examples. Third, the designs of total cancer incidence.1 clinical trials conducted in the areas of dental
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 194 TEXTBOOK OF CLINICAL TRIALS caries, oral rehabilitation, periodontal disease clinical signs.10 For more advanced periodon- and orthodontics will be discussed. Fourth, the tal disease, usually the depth of the periodontal current issues of evidence-based dentistry and pocket and/or the attachment loss are measured hierarchical data analysis will also be discussed. in millimetres using marked probes. Last, the impact of clinical trials on dental Following the developments in medical re- practice will be summarised. search, patient-based outcomes measures have also been developed and employed in dental DISEASE AETIOLOGY AND MEASUREMENTS research. These have focused primarily on the concept of patient satisfaction and health-related Evidence from animal and epidemiological stud- quality of life measures. A number of question- ies shows that dental caries arise from dem- naires have been developed recently to measure ineralisation of tooth hard tissue due to organic the oral health-related quality of life of peo- acids produced by plaque bacteria on the tooth ple, for example, the Oral Health Impact Profile surface.3 Frequent intake of fermentable carbo- (OHIP) and the General Oral Health Assess- hydrates, especially sugars, has been shown to ment Index (GOHAI).11 Some of these have been be related to the development of dental caries.4 used in clinical studies to assess the outcome of The most common measure used in clinical tri- 12,13 als to quantify the severity of dental caries is to dental treatment, to supplement the clinical count the number of decayed, filled and missing assessments. (due to caries) teeth or tooth surfaces, the DMFT 5 or DMFS index. The current treatment approach CLINICAL TRIALS METHODS for dental caries emphasises prevention of the dis- ease by strengthening the tooth, such as the use of fluorides and fissure sealants, modification of In dental research, the clinical trials methods the diet, such as the use of sugar substitutes, and used mainly follow those developed and adopted appropriate health behaviours. Cavities produced in medical research. The basic design principles by the caries process can be filled by various and considerations are very similar, thus the direct and indirect restorative materials. following discussion on clinical trials methods Periodontal disease is characterised by the can be kept short and precise for general areas, inflammatory response of the gums and its sequel with specific areas unique in dentistry discussed to the toxic substances produced by the plaque in more detail. A few references on clinical trial bacteria.6 The current treatment approach for methodology and some from the dental literature 14–17 periodontal disease emphasises primary and sec- are recommended for general reading. ondary prevention of the disease through the removal of plaque by mechanical and chemi- RANDOMISED CONTROLLED TRIALS cal means, e.g. toothbrushing and the use of mouthrinses. There are also various surgical and As with the developments in medical research, non-surgical ways to treat the periodontal pock- randomised controlled trials (RCTs) have become ets that are formed in more advanced periodontal the gold standard in conducting clinical trials disease states. Many indices have been used in in dentistry. The key features of RCTs are clinical trials to quantify the amount of plaque on treatment modalities being assigned randomly to the tooth surfaces, ranging from a simple dichoto- the subjects and the existence of a control group. mous scale of presence or absence of visible The controls can either be concurrent controls as plaque7 to recording the thickness of the plaque at in parallel studies or self-controls as in crossover the gum margin8 or the area of tooth surface cov- studies. The treatment received in the control ered by plaque.9 Gingivitis is usually measured group can be placebo or standard treatment. In by the presence or absence of bleeding after gen- the perfect setting, RCTs should also be double- tle probing7 or in an ordinal scale using various blinded which requires that both the subjects and DENTISTRY AND MAXILLO-FACIAL 195 the examiners/observers involved in the trials are subjects as their own controls prevents confound- not aware of the assignment of the treatment ing by many characteristics that may influence modalities to the subjects, thus reducing any the outcome. In a crossover study each subject is biases in the comparison of the groups besides given the different treatments (or treatment and randomisation. placebo) under comparison, one after another. Informed consent, ethical consideration, data Each subject is his/her own control. The sequence monitoring and pre-study sample size calculation of assignment is generally randomised, so that are also important issues in conducting RCTs. this is in effect a type of RCT. A ‘washing-out Subjects should be informed about the research period’ may be required between treatments, to protocols, their roles as participants in the permit the effects of the previous treatment to studies, the different treatment modalities, the disappear. However, since subjects who partici- possibility of any side-effects or risks arising pate in clinical trials with crossover design need from receiving the treatments, and their rights of to receive all treatment regimens and undergo discontinuing participation. After the explanation ‘washing-out periods’ between treatments, it can of the above details to the subjects, the subjects make the investigation periods of the clinical tri- should be given ample opportunity and time to als very long and not feasible.19 ask any questions and to discuss the trial with Crossover trials are frequently employed in family and friends. Written consent is normally oral hygiene studies where treatment effects are required, however under special circumstances, reversible. An example is a single-blind, short- verbal consent may be employed.18 It is good term crossover clinical trial where the plaque practice that clinical trials only be conducted after removal performance of three commercially approval from local ethical committees in order available manual toothbrushes was compared.20 to protect human rights. A sample of 25 dental hygiene students partici- Data monitoring is especially important in pated (age 19 to 42 years old). The participants large-scale, multicentre RCTs and usually a were instructed to refrain from toothbrushing or data monitoring committee is established to flossing for 24 hours before the trial. A pre- monitor the data collected during the study. brushing plaque index using disclosing solution The committee needs to monitor the data for was performed on each participant. One of the patient safety and statistical significance, while three test brushes was then randomly assigned to keeping their findings confidential to prevent the each participant, and they were allowed to brush introduction of bias.15 for 90 seconds without the aid of a mirror. A post-brushing plaque index was then performed PARALLEL AND CROSSOVER on each participant. This procedure was repeated STUDY DESIGNS twice more at 2-week intervals so that each par- ticipant was tested with all three toothbrushes. In the parallel study design, concurrent groups Crossover design will not be applicable if of subjects are involved in the study and the the treatment has protracted ‘carry-over’ effect, comparison of the different modalities is the com- i.e. the effect of the treatment is not reversible. parison of between-subject variation. When the In this situation, either parallel or split-mouth number of treatment modalities increases, the design should be adopted. For example in the corresponding sample size required in order to case of dental caries, since most carious lesions achieve a particular level of power and signifi- occur in the pit and fissure on the occlusal cance needs to be increased greatly. Crossover surfaces of the posterior teeth, the effectiveness study design is a self-controlled study design, of sealing these pits and fissures in order to subjects serve as their own controls. Thus, the prevent dental caries is studied. In studies for comparison of the different modalities is the com- comparing the effectiveness of fissure sealant parison of within-subject variation. The use of the compared to non-sealed teeth or sealant with 196 TEXTBOOK OF CLINICAL TRIALS different active ingredients to prevent caries, receiving all the treatment modalities at the crossover design is not applicable as once the different parts of the mouth concurrently, the teeth are sealed, the process is not reversible. study period of the investigation could be shorter. Thus for these studies, either parallel design The study period could then be of the same or split-mouth design would be used. In the duration as if the parallel design was used, but setting of parallel design, subjects are assigned the number of subjects used could be reduced. to different concurrent test groups; while in In an investigation of the efficiency of split- the setting of split-mouth design, different teeth mouth design compared to whole-mouth design of the same group of subjects are assigned to (with the use of parallel study design), it was different test groups at the same study period.21 concluded that when disease characteristics are symmetrically distributed over the within-patient SPLIT-MOUTH DESIGN experimental units, the split-mouth design could provide moderate to large gains in relative Split-mouth design is one of the self-controlled efficiency. For periodontal disease, division of study designs, that is unique in dentistry. This the mouth into two experimental units, either left design is characterised by subdividing the mouth and right or upper and lower sides, provided the 22 of the subjects into homogeneous within-patient greatest symmetry of the disease characteristics. experimental units such as quadrants (upper left, Besides the distribution of the disease, one upper right, lower left and lower right sides of should also consider the carry-across effects the mouth), sextants (upper left posterior, upper arising from the split-mouth designs. Carry- anterior, upper right posterior, lower left poste- across effects occur where treatment performed rior, lower anterior and lower right posterior), in one experimental unit can affect the treat- contralateral (left and right) or ipsilateral (upper ment response in other experimental units of and lower) quadrants or sextants or a symmetrical the mouth. These effects cannot be estimated combination of these. With these within-patient from split-mouth data, thus unless prior knowl- experimental units, a range of two to six different edge indicates that no carry-across effects exist, treatment modalities can be randomly assigned to reported estimates of treatment efficacy are the experimental units.22 The number of treat- potentially biased. Thus, researchers should ment modalities usually equals the number of weigh the potential gain in precision against a within-patient experimental units. For instance, potential decrease in validity when using split- in a study where two treatment modalities are mouth designs in clinical trials.23 In a study to compared, the within-patient experimental units compare the effectiveness of two electric tooth- would usually be the left or right sides of the brushes in plaque removal, a split-mouth design mouth. In a study where four treatment modali- was used in which either the first (upper right) ties are compared, the within-patient experimen- and third (lower left) quadrants or the second tal units could be the four quadrants of the mouth. (upper left) and fourth (lower right) quadrants The split-mouth design has been the principal of 84 subjects were brushed with one or the research tool in periodontal clinical trials to com- other toothbrush in a random assignment.24 In pare different treatment modalities. In the peri- this situation, since the distribution of plaque odontal literature, at least 11 different types of inside the mouth is symmetrical, and the use of split-mouth design have been described.22 different toothbrushes in brushing different parts The major advantage of using the split-mouth of the mouth should not affect the other parts, design, like the crossover design, is that subjects i.e. no carry-across effect, the use of a split- serve as their own controls, thus this design may mouth design should be appropriate. In studies be more efficient than designs with between- to compare the effectiveness of fluoride tooth- subject comparisons. However, in contrast to paste in preventing dental caries, split-mouth the crossover design, since the subjects are design is not advisable as the fluorides from the DENTISTRY AND MAXILLO-FACIAL 197 toothpaste could go freely within the mouth, i.e. to a common examination site demonstrates an with a carry-over effect. For these studies, paral- innovative way for researchers to maximise the lel design should be more appropriate. degree of blindness.
BLINDING CLINICAL TRIALS IN DENTISTRY In order to achieve double-blinding, the subjects CARIES PREVENTION AND TREATMENT and examiners/observers should not be able to tell STUDIES which treatment modalities have been assigned to the subjects. This can be done, for example, in a The aims for these prevention studies are to mouthrinse study comparing the test mouthrinse investigate the effectiveness of different ways of with placebo; the placebo mouthrinse is made preventing dental caries. These include different with the same appearance and taste as the test methods of strengthening the teeth (such as the mouthrinse, so that the subjects would not be able use of fluorides in different forms), modification to distinguish the two by sight and they are not of diet (such as the use of sugar substitutes), told which mouthrinse they have been assigned. or modification of health behaviours (such as On the other hand, the examiners/observers tooth brushing techniques and habits, oral edu- are also not able to reach the information cation programmes). The target populations for for the assignment of the mouthrinse to the these studies are mainly children, the elderly and subjects.25 However, in many other situations, we special needs groups. Most of the clinical trials can only blind the examiners/observers but not are phase I or phase II types. For those studies the subjects (single-blinding). In the study we investigating the effectiveness of different forms mentioned above concerning the comparison of of fluorides (in the form of toothpaste, topical the effectiveness of three electric toothbrushes in fluorides, sealant), randomisation of the assign- plaque removal, it is inevitable that the subjects ment of groups with different regimes (including will know which toothbrush they were using, the control group) can be done at the individual thus in this situation, the best that one can level with parallel design. In a study to com- achieve is to blind the investigator from knowing pare the effectiveness of two toothpastes with which toothbrushes have been assigned to which different concentration of fluoride to arrest root quadrants of the subjects.24 There are situations carious lesions, 201 subjects with at least one where even single-blinding is not feasible. In a root carious lesion were recruited from dental study comparing the performance of two dental school patients. They were randomly assigned to filling materials, amalgam (metal) versus resin- use either Prevident 5000 Plus (5000 ppm F) or modified glass ionomer cement (tooth colour), it Colgate Winterfresh Gel (1100 ppm F), both con- would be very difficult to blind the investigator as taining sodium fluoride in the same silica base. one can distinguish the two by their appearance.26 Measurements of lesion hardness, area, distance In conducting clinical trials, one should use the from the gingival margin, cavitation and plaque maximum degree of blindness that is possible. In were recorded at baseline and 3 months later by studying the prevalence of caries and fluorosis a single examiner.28 of children from a water-fluoridated site and For those studies involving the modification of a nearby non-fluoridated site, children were diet and behaviour, field studies were used more transported to a common examination site so often because randomisation of the assignment of as to blind the investigators from knowing the groups with different regimes would be easier to residence of the children.27 Insteadofhavingthe achieve at the school or community level. In a investigators examining the children at the sites 3-year community intervention trial to determine and then recognising the fluoride content in the the caries preventive effect of sugar-substituted water, this method of transporting the children chewing gum among Lithuanian school children, 198 TEXTBOOK OF CLINICAL TRIALS a total of 602 children, aged 9–14 years, from and 0.619% sodium fluoride (2800 ppm fluo- five secondary schools in Kaunas, Lithuania ride ion). All products were formulated with the were recruited. Baseline clinical and radiographic same fluoride-compatible silica abrasive. Sub- caries examinations were given. The schools jects were examined by visual–tactile and radio- were randomly allocated to receive one of the five graphic examination at baseline and after 1, 2 and interventions: sorbitol/carbamide gum; sorbitol 3 years of using the sodium fluoride dentifrices.31 gum; xylitol gum; control gum; and no gum. The aims of the caries treatment studies were Children in the four active intervention groups to investigate the performance of the different were asked to chew at least five pieces of gum materials or different approaches used to fill up per day, preferably after meals. The children were the decayed cavities in the mouth in terms of re-examined clinically after 1, 2 and 3 years, and bond strength and retention of the materials. radiographically after 3 years.29 As the treatments being delivered cannot be In both the above examples, parallel design reversed, thus crossover design is not applicable was adopted. However, studies like the com- for these studies. Among these studies, the use of parison of two different fissure sealant materials split-mouth design was more common. In a study could be carried out using the split-mouth design. to compare the clinical performance of two glass In a study to compare the retention and the caries ionomer cements, ChemFlex (Dentsply DeTrey) preventive effect of a glass ionomer developed and Fuji IX GP (GC), when used with the for fissure sealing (Fuji III) and a chemically atraumatic restorative treatment (ART) approach polymerised resin-based fissure sealant (Delton), in China, 89 schoolchildren aged between 6 179 7-year-old children with at least one pair of and 14 years who had bilateral matched pairs permanent first molars that were caries-free or of carious posterior teeth were included. A only had incipient lesions were recruited from split-mouth design was used in which the two schools. A split-mouth design was adopted by materials were randomly placed on contralateral assigning the two sealing materials randomly to sides. The performance of the restorations was the contralateral teeth. Follow-up examinations assessed directly and also indirectly from die- for sealant retention were done after 6 months, stone replicas at baseline and after 6, 12 and 1 year, 2 years and 3 years.30 24 months.32 In order to determine the level of fluoride that should be used (dose finding), some studies have focused on comparing the effectiveness of differ- ORAL REHABILITATION STUDIES ent concentrations of fluorides in caries preven- tion. For example, in a randomised, double-blind One concern of oral rehabilitation studies is study comparing the anti-caries effectiveness of to compare a range of treatment modalities to sodium fluoride dentifrices containing 1700 ppm, replace missing teeth including removable and 2200 ppm and 2800 ppm fluoride ion relative to fixed dental prostheses, and the use of implants. an 1100 ppm fluoride ion control, a population Depending on the treatment modalities, parallel, of 5439 schoolchildren, aged 6–15 years, was split-mouth and crossover designs have been recruited from an urban central Ohio area with applied in these studies. In a 5-year parallel a low fluoride content water supply (<0.3 ppm). study to compare implant-retained mandibular Subjects were stratified according to gender, overdentures (IRO) with complete dentures (CD), age and baseline DMFS scores derived from 61 and 60 patients were randomly assigned to the visual–tactile baseline examination. Random the IRO and CD groups. The clinical aspects and allocation of the four treatment groups was done: patient satisfaction were measured.33 0.243% sodium fluoride (1100 ppm fluoride ion), In a split-mouth study, sandblasted and acid- 0.376% sodium fluoride (1700 ppm fluoride ion), etched (SLA) implants (recently introduced to 0.486% sodium fluoride (2200 ppm fluoride ion) reduce the healing period between surgery and DENTISTRY AND MAXILLO-FACIAL 199 prosthesis) were compared to titanium plasma- using both the Visual Analogue Scale and the sprayed (TPS) implants under loaded condi- Category Scale.35 In another study, an oral- tions 1 year after placement. Thirty-two healthy specific quality of life measure, the Oral Health patients with comparable bilateral edentulous Impact Profile,36 wasusedtomeasurethe sites and no discrepancies in the opposing denti- impact of the clinical intervention on quality of tion were recruited. The surgical procedure was life. Three groups of subjects were compared, performed by the same operator and was iden- edentulous/edentate subjects who requested and tical for all the test and control sites. Abut- received complete implant-stabilised oral pros- ment connection was carried out at 35 Ncm theses (IG, n = 26), edentulous/edentate subjects 6 weeks post-surgery for test sites and 12 weeks who requested implants but received conven- for the controls by the same dentist who was tional dentures (CDG1, n = 22), and edentulous blinded to the type of surface of the implant. subjects who had new conventional complete Provisional restoration was fabricated and a new dentures (CDG2, n = 35).13 OHIP is one of tightening was performed after 6 weeks. Simi- the most comprehensive instruments available in lar gold–ceramic restorations were cemented on evaluating oral health-related quality of life. the same type of solid abutments on both sites. Clinical measures and radiographic changes were TRIALS RELATED TO PERIODONTAL DISEASE recorded by the same operator who was blinded In order to prevent periodontal disease, plaque to the type of surface of the implant, 1 year post- 34 removal and prevention of calculus deposit surgery. was one of the key concerns. Thus ways to In a study to compare two designs of maxil- improve toothbrushing (mechanical means to lary implant overdentures, a crossover trial was remove plaque) or the use of mouthrinse or designed to measure differences in patient satis- toothpaste with active ingredients like chlorhex- faction with maxillary long-bar implant overden- idine and pyrophosphate ion (chemical means to tures with and without palatal coverage opposed remove plaque and to prevent calculus deposi- by a fixed mandibular implant-supported prosthe- tion) were investigated. Two main streams of sis. A mandibular fixed prosthesis was inserted therapy existed: non-surgical and surgical ther- in 13 total edentulous participants, who were apy. The aims for these studies were to investi- = then divided into two groups. One group (n 7) gate the effectiveness of the treatment modalities received maxillary long-bar overdentures with in reducing the depth of the periodontal pocket palate, then long-bar overdentures without palate. or improving periodontal attachment level. The other group (n = 6) received the same treat- Similar to the caries research, different study ments in the reverse order. For each overdentures designs in RCT have been applied to periodon- design, mastication tests and patient satisfaction tal research; one particular design issue that has were assessed 2 months after the fitting of the been discussed more in periodontal research com- overdenture to allow adaptation.35 pared to other areas in dental research is the Besides clinical outcomes, very often patient- consideration of therapeutic equivalency. This is based outcomes such as patient satisfaction important because clinical trials for testing supe- and quality of life measures were also mea- riority and for testing equivalency should have sured in these oral rehabilitation studies. As in different designs and there has been a tendency the above quoted example, patients were asked by the clinicians in periodontal research to con- to rate (1) their general satisfaction with the fuse the difference between the two.37 Clinical upper prosthesis; (2) satisfaction on the physi- trials whose purposes are to show equivalence of cal functions of the prosthesis such as reten- two or more treatments have traditionally utilised tion, stability, comfort, ease of cleaning, etc.; methods for demonstrating superiority. Thus if no and (3) satisfaction on the psychosocial func- statistical differences are found, this only demon- tions such as a esthetics, self-confidence, etc. strates that the treatments are not superior to the 200 TEXTBOOK OF CLINICAL TRIALS others, they may not be equivalent.19 The sam- monitoring of the trials are all that is needed ple size requirements for both equivalence and for conducting RCTs. He has also urged journal superiority studies investigating products used in editors to promote the publication of RCTs. periodontal regeneration have been investigated. Hopefully, there should be a paradigm shift It was found that since equivalence clinical trials in conducting RCTs in orthodontic research in require much smaller differences between groups, future years. thus much larger sample sizes are required.38 CURRENT ISSUES ORTHODONTIC STUDIES EVIDENCE-BASED DENTISTRY In orthodontic studies, two main concerns are the effectiveness of early orthodontic treatment for One of the major implications of conducting Class II malocclusion and the value of maxillary clinical trials is to provide scientific evidence arch expansion for the treatment of posterior for the clinicians when they are making clinical crossbite. Relatively speaking, fewer RCTs are decisions. Evidence-based medicine was defined done in this research area. Some researchers have as ‘the conscientious, explicit and judicious use argued that even though RCT has become the of current best evidence in making decisions gold standard of conducting medical research, about the care of individual patients’.41 Den- it could only apply to a very narrow spectrum tal researchers started addressing the issues of of orthodontic questions. One quoted example evidence-based dentistry (EBD) in the 1990s is ‘it would be nearly impossible to enroll and a series of articles has been published to fully-informed subjects into any study whose address the concepts of EBD, the misunderstand- alternatives are of markedly different morbidity: ings of EBD, the barriers to EBD and the pro- extraction versus non-extraction or orthodontics cesses of finding, evaluating and applying the versus surgery’.39 Three confusions (or inertia) evidence.42–48 Various studies making use of the have been summarised in explaining why the EBD approach have been applied to different move to conduct RCTs in orthodontic research areas in dental research and the journal Evidence- has been slow. First, there is a remarkable Based Dentistry has also been published since level of non-acceptance that the highest level of November 1998. evidence is derived from RCTs and retrospective Systematic literature review is the foundation investigation is regarded as more useful. Second, of the EBD approach. It differs from narrative there is the argument that it is not ethical review as in the latter case the authors or experts to subject patients to a random allocation of do not use standardised ways to retrieve articles treatments if it is already known that one of the and summarise information. Thus different writ- treatments is superior. Third, there is a feeling ers may arrive at different conclusions for the that RCTs are very large and difficult to manage same question. In systematic review, standards in and then they are expensive and a large amount finding, evaluating and synthesising evidence are of funding is required.40 O’Brien has provided reported and thus the conclusion is more reliable. solutions to the above confusions: (1) one should The Cochrane Collaboration is an international accept that RCTs derive the highest level of organisation that ‘has developed in response to evidence and retrospective investigation still has Archie Cochrane’s call for systematic, up-to-date a great value in generating questions for RCTs; reviews of all relevant RCTs of health care’ (2) one should not simply believe that most (http://www.cochrane.org). In an influential book, treatments are superior to another without being Cochrane49 drew attention to the importance tested in an unbiased manner, thus it is actually of organising critical summaries of all relevant unethical to provide treatment that has not RCTs by specialty or subspecialty areas of health been properly evaluated; (3) careful planning and care, so that people can make more informed DENTISTRY AND MAXILLO-FACIAL 201 decisions. The first ‘Cochrane Centre’ was of the subjects will be examined and in periodon- opened and funded in the UK in October 1992, to tal research, usually six sites of each tooth will facilitate systematic reviews of randomised con- be examined; all these measurements are possi- trolled trials across all areas of health care. Cur- bly correlated or clustered. More examples have rently, there are 15 Cochrane Centres around the been given by Macfarlane and Worthington50 world. However, the Cochrane Centres are not and Gilthorpe et al.51 With these correlated or directly responsible for preparing and maintain- clustered data, conventional statistical methods, ing systematic reviews. This is the responsibil- which assume observations being independent, ity of international collaborative review groups, are not appropriate for the analysis. Thus spe- which also maintain registers of systematic cial statistical analysis is required when data have reviews currently being prepared or planned, so a hierarchical structure. Analysing data without that unnecessary duplication of effort can be recognising the hierarchical structure and treat- minimised and collaboration promoted. Currently ing the observations as independent will lead to there are about 50 review groups covering all a spurious increase in the sample size, and corre- of the important areas of health care and den- sponding spurious significant relationship.52 For tistry included in the Cochrane Oral Health Group example in periodontal research, one might take (http://www.cochrane-oral.man.ac.uk). The prin- up to six site measurements for each tooth and cipal output of the collaboration is the Cochrane with 28 teeth (not including the wisdom teeth) for Reviews which are published electronically in an individual, the total number of observations successive issues of The Cochrane Database of for one individual could then be 168. Thus it is Systematic Reviews. Ten Cochrane Reviews have easy to have thousands of observations with only been finished in the Cochrane Oral Health Group, 20 subjects. In a study, the total number of sites including: fluoride gels for preventing dental assessed was 2236 distributed on 559 teeth in 22 caries in children and adolescents; guided tissue subjects.53 One way to overcome this problem is regeneration for periodontal infra-bony defects; to aggregate the raw observations to the highest interventions for preventing oral mucositis or level of the data structure, for example, site/teeth oral candidiasis for patients with cancer receiving level measurements can be aggregated to the level chemotherapy; interventions for treating burn- of an individual subject like the DMFT/DMFS ing mouth syndrome, oral mucositis, oral leuko- index, mean probing pocket depth, etc. Thus one plakia, oral candidiasis and oral lichen planus; measurement was taken from each individual and orthodontic treatment for posterior crossbites; and conventional statistical analysis could then be potassium nitrate toothpaste for dentine hyper- applied. However, aggregation has several draw- sensitivity. Twenty-seven protocols are registered backs, the most obvious of which is the loss of currently to review various fluoride products in detailed information. Therefore, the aggregated preventing caries, various regimes of interven- measure differentiates poorly both trait and sever- tions for replacing missing teeth, and various ity of the measures made at the disaggregated orthodontic treatments protocols, etc. level. Furthermore, aggregation is of little value when the focus of interest lies at a lower level of HIERARCHICAL AND MULTILEVEL the data hierarchy, for example sites with peri- 54 DATA ANALYSIS odontal pockets. ‘Multilevel modelling’ (MLM)55 or equiv- Hierarchical data (or clustered data) is common in alently ‘hierarchical linear modelling’56 is a dental research, as people may have up to 32 teeth class of techniques developed to analyse hier- and taking measurements from multiple teeth of archical data. It was first adopted to anal- the same individual is very typical in the data col- yse educational data. These techniques are the lection of clinical trials in dentistry. For example, modified version of statistical methods available in caries prevention studies, the individual teeth for the analysis of single-level data structures 202 TEXTBOOK OF CLINICAL TRIALS
(e.g. multiple regression, logistic regression, log- studies. Human studies have largely been of the linear modelling) for the analysis of data with observational type: world-wide epidemiological hierarchical structure. One could carry out the studies, ‘before and after’ studies, and studies multilevel data analysis through the use of among people with both high and low sugar software specially written for MLM, like MLwiN consumption. Very few interventional studies on (http://multilevel.ioe.ac.uk), or one could write up human subjects have been conducted,4,60 and are macros in other statistical softwares like SAS, unlikely to be undertaken in the future given S-plus.57 In carrying out MLM, one should spec- the difficulties of placing groups of people on ify the number of levels in the model and then rigid dietary regimes for long periods of time and the variables incorporated at each of the levels. because of ethical issues. The main conclusion Back to the example of the periodontal study in of studies relating to sugar and dental caries which the number of sites assessed was 2236 dis- has been that (1) consumption of sugar, even tributed on 559 teeth in 22 subjects.53 One of at high levels, is associated with only a small the analyses performed by the researchers was increase in caries increment if the sugar is taken a multilevel analysis on the factors affecting the up to four times a day and none between meals; change in probing pocket depth at the sites over (2) consumption of sugar both between meals and the course of the treatment. A three-level model at meals is associated with a marked increase was built with site as level 1, tooth as level 2 in caries increment.61,62 These conclusions have and subject as level 3. At the site level, 12 vari- shaped key dental education messages of oral ables were constructed (e.g. presence of plaque health promotion campaigns relating to diet at the site, treatment received at the site); at the and dental health around the world and also tooth level, three variables were included (e.g. formed the basis of dentist–patient dental health tooth type) and at the subject level, 19 vari- education relating to diet and dental caries.63 ables were constructed (e.g. age, gender, smoking Other trials have provided evidence of variations habit). The above analysis was performed using in caries incidence with different types of sugars, the software MLn (the non-WindowsTM version notably, the low caries rate associated with the of MLwiN). With the use of MLM, it is then pos- use of sugar alcohols like xylitol.60 This has led to sible to investigate the change in probing pocket more widespread use of non-carcinogenic sugar depth at the site with the consideration of the alternatives in drinks and foodstuffs.64 However, effects from the site itself, the tooth that it belongs this may be attributed more to their low caloric to and the subject overall. In order to evaluate value than their low cariogenicity. whether a three-level model is necessary, one should test the ‘null model’ that no indepen- dent variables were included and then check the WATER FLUORIDATION significance of the variance at each level and Evidence of the effectiveness of water fluorida- fit the model accordingly. Several other studies tion has largely been based on cross-sectional using multilevel modelling in analysing caries, and ecological studies, ‘before and after’ stud- periodontal and orthodontics data have also been ies, and fewer cohort or case–control studies. No published.54,57–59 randomised controlled trials have been reported in the dental literature. Systematic reviews of IMPACT OF TRIALS ON DENTAL PRACTICE the effectiveness of water fluoridation have con- cluded that it is an effective, efficient and safe DIET AND DENTAL CARIES method of preventing dental caries and possibly promotes equity in oral health in society.65 The Evidence of the role of diet, particularly sugars, studies have examined the relationship between in relation to dental caries has largely been dental caries experience and the fluoride content collated from animal experiences or in vitro of the water supply and have shown clearly the DENTISTRY AND MAXILLO-FACIAL 203 association between increased fluoride concentra- that salt fluoridation offers the consumer a choice tion in the drinking water and decreased dental to use fluoride supplements or not, there are caries experience in the population. However, the only a handful of countries where it is widely studies have suggested that there is little ben- available and consumed. Concerns about the efit where water fluoride concentrations exceed appropriate dosage (a minimum of 200 mg/l F is 1 ppm. These findings have resulted in the imple- recommended) and safety for general health may mentation of water fluoridation in many indus- impede its widespread implementation.70 trialised and developing countries where central Fluoride supplements in the form of tablets water supplies have made it feasible to do so. It to drops have long been considered an alterna- remains a key World Health Organization goal tive to water fluoridation. Although the effec- for oral health. tiveness of fluoride supplements was endorsed However, a number of studies relate to the by many small clinical studies, closer examina- negative influences of water fluoridation on tion of the experimental conditions of these, their dental and general health, primarily on the effects methods and the analysis of their results under- of water fluoridation in producing dental fluorosis mined confidence in their findings. More modern, (tooth mottling) among the population.66 Fluoride well-conducted clinical trials of supplements sug- at a concentration of 1 ppm is likely to produce gest that today, in children also exposed to flu- a small increase in dental mottling. However, in oride from other sources such as toothpaste, the the most part such mottling is unlikely to be marginal effect of fluoride supplements is very of aesthetic concern. Despite strong evidence of small and there is substantial risk of fluorosis if the effectiveness and safety of water fluoridation supplements are used by young children.71 This some communities have ceased to fluoridate their has resulted in changes to recommended fluoride water supplies because of legal issues, social dosage schedules and deferment of the age com- acceptance and concern about the additional mencing the use of supplements, implemented benefits of fluoridated water where other sources in many countries. Overall, poor compliance and of fluoride are readily accessible. potential risks of fluorosis make fluoride supple- ments a poor public health measure and they are infrequently prescribed in dental practice.72 ALTERNATIVES TO WATER FLUORIDATION
A wide range of alternative methods for admin- FLUORIDATED TOOTHPASTE istering systemic fluoride have been suggested in the literature, particularly milk fluoridation and The daily use of fluoride toothpaste is a highly salt fluoridation.67 Extensive literature describ- effective method of delivering fluoride to the ing the study of fluoride compounds administered tooth surface and has proved to be a major aid with calcium-rich food, as well as clinical tri- to caries prevention.73 The concentration of flu- als and laboratory experiments with fluoridated oride at 1000 ppm F has been suggested as a milk, have demonstrated the effectiveness of milk safe and effective means of preventing caries.74 fluoridation in caries prevention.68 However, the Although evidence suggests that toothpastes with criticism of decreased bioavailability of the fluo- higher fluoride concentrations are more effective ride, the cost and administrative burden involved, in preventing dental caries, because of safety and conflicting evidence of efficacy has resulted concerns dentifrices exceeding 1500 ppm F are in few community milk fluoridation programmes. only sold by prescription in many countries.75 Salt fluoridation has also been advocated as However, for children trials have suggested that an alternative to water fluoridation. Evidence of a lower concentration of fluoride in dentifrices the effectiveness of salt fluoridation has largely (250–500 ppm F) be used and that only a mini- come from test and control community studies mum amount (less than 5 mm) should be placed in several different countries.69 Despite the fact on the brush to minimise risk.76 Some trials have 204 TEXTBOOK OF CLINICAL TRIALS suggested that combining more than one fluoride clinical grounds, on patients with special needs, a agent is more effective than using one source of history of extensive caries in the primary deten- fluoride agent in preventing dental caries. How- tion or caries involving one or more molars.81 ever, different formulations of toothpaste appear It is important that they are reviewed at regu- to have similar effectiveness.77 To some extent lar intervals. the use of dentifrice has removed the need for professionally applied fluoride agents, except in TREATMENT OF CARIES LESIONS special circumstances. A key focus of research has been the perfor- mance of direct posterior restorations (fillings), OTHER FORMS OF TOPICAL FLUORIDE the longevity and reasons for failure of direct resin-based composite (RBC), amalgam and glass Many forms of professionally applied fluoride ionomer cement (GIC) restorations in stress- have been studied, including solutions, gels or bearing posterior cavities. Predominantly studies foams of sodium fluoride, stannous fluoride, have been either of the longitudinal or retro- organic amine fluoride, acidulated phosphate flu- spective cross-sectional type, with few controlled oride and non-aqueous fluoride varnishes in an clinical studies. GIC perform significantly worse alcoholic solution of natural resins and difluorosi- compared with amalgam and RBC.82 However, lane agents covered by a polyurethane coating. reasons for placement and replacement of direct All of these professionally applied topical agents restorations in dental practice relates to many have anti-caries benefits, although the benefits 78 factors, and aesthetic and safety concerns have and ease of application vary. However, a recent resultedinanincreaseduseofRBCorGIC systematic review of the periodic scientific liter- restorations in posterior teeth.83 The handling and ature undertaken to determine the strength of the fluoride leaching properties of GIC have made evidence for the efficacy of professional caries them popular in general practice.84 preventive methods applied to high-risk individ- uals, and the efficacy of professionally applied methods to arrest or reverse non-cavitated carious REPLACEMENT OF MISSING TEETH lesions, concluded that the strength of the evi- A range of treatment modalities to replace miss- dence was judged to be fair for fluoride varnishes ing teeth have been studied, including remov- 79 and insufficient for all other methods. In dental able and fixed dental prostheses and the use of practice professionally applied fluoride is infre- implants. Increasingly these studies have incorpo- quently employed owing to more widespread use rated patients’ perceptions of outcomes. Evidence of other fluoride sources, reports of inconclusive has largely been collated from longitudinal or evidence and because of health care reimburse- case–control studies with relatively few RCTs. ment for such preventive procedures. Implant-retained overdentures are reported to be superior to complete dentures.33 In addition, FISSURE SEALANTS implants are useful in the treatment of par- tial edentulism. However, the widespread use of Most carious lesions occur in the pit and fis- implants in practice has been limited by a number sure on the occlusal surface of posterior teeth. of factors including health care cover and costs, Over the years clinical trials have demonstrated operator experience and appropriateness for indi- the effectiveness of sealing these fissures and vidual cases. pits in preventing dental caries.21 Light-curing Another contentious issue has been the use and auto-polymerising sealants are equally effec- of resin-bonded bridges (RBBs) which provide tive. However, the cost-effectiveness of fissure a greater degree of conservation of tooth struc- sealants in clinical trials remains questionable.80 ture of abutments compared with designs of con- Thus, fissure sealants should be employed on ventional fixed prostheses in the treatment of DENTISTRY AND MAXILLO-FACIAL 205 partial edentulism.85 A key concern has been therapy is more appropriate than surgical peri- the longevity of RBBs, however studies suggest odontal therapy, and that surgical therapy should that with appropriate case selection, preparation only be considered when sites fail to respond design and cementation they are a viable treat- to non-surgical methods despite adequate oral ment option compared to conventional bridges. hygiene.92 State-funded and third-party payers of Increasingly RBBs are being employed in den- oral health care usually require detailed justifica- tal practice in the treatment of short edentu- tion for surgical periodontal care. lous spaces. ORTHODONTICS TRIALS RELATING TO PERIODONTAL While there has been considerable growth in DISEASE the practice of orthodontics there is a dearth of evidence-based research, particularly RCTs.40 A key focus of periodontal trials has been the A contentious issue has been the timing of need for plaque control to prevent periodon- orthodontics and the need for early orthodon- tal diseases and for the maintenance of peri- tic intervention. The evidence relating to early odontal health.86 Primarily studies have focused orthodontic treatment is inconclusive, with the on mechanical methods of plaque control. Stud- result that many clinicians decide, on a case- ies have shown that the most important plaque by-case basis, when to provide orthodontic control method is toothbrushing; precise tech- treatment.93 Another key concern has been the nique is less important than the result, which is value of maxillary arch expansion for the treat- removal of plaque without causing damage to ment of posterior crossbite. A Cochrane Review the teeth or gums.87 It is widely promoted to on the subject was unable to propose recom- establish toothbrushing practice as a daily routine mendations based on inadequate trials.94 Lack from childhood. of evidence relating to the value of orthognathic Additional methods of mechanical plaque con- surgery versus orthodontic camouflage in the trol include interdental cleaning aids such as den- treatment of mandibular deficiency, and also as tal floss. While such aids have been shown to be to the need for extraction of teeth for orthodon- effective in plaque control with minimum damage tic purposes, has resulted in clinicians decid- if used correctly, they are generally prescribed ing on a case-by-case basis without any clinical depending on the individual’s periodontal health guidelines.95 and their ability to use them appropriately.88 Chemical antimicrobial agents may be a useful CONCLUSION adjunct measure to managing periodontal health. In particular the use of chlorhexidene in the Currently the lack of high-quality research within chemical control of plaques has widely been dentistry, namely the lack of RCTs, has impeded advocated, particularly in acute phases and in the identification of best dental practice and preventing post-surgical infection.89 However, the implementation of evidence-based practice with the long-term use of chlorhexidene there within dentistry. There is however widespread is a tendency for it to stain (extrinsic) teeth. In recognition of these problems and concerted more recent times the addition of antimicrobial efforts to undertake more collaborative high- agents to dentifrices to aid plaque control has quality research that can inform policies and become commonplace.90 The use of chemical guidelines to be implemented in dental practice. agents in the removal of plaque, while effective, is not recommended over the use of mechanical REFERENCES agents.91 In the treatment of periodontal disease many tri- 1. Stjernsward J, Stanley K, Eddy D, Tsechkovski M, als have concluded that non-surgical periodontal Sobin L, Koza I, Notaney KH. National cancer 206 TEXTBOOK OF CLINICAL TRIALS
control programs and setting priorities. Cancer 19. Burns DR, Elswick RK Jr. Equivalence testing Detect Prevent (1986) 9: 113–24. with dental clinical trials. J Dental Res (2001) 80: 2. Proffit WR, Fields HW, Morav LJ. Prevalence of 1513–17. malocclusion and orthodontic treatment need in 20. McDaniel TF, Miller DL, Jones RM, Davis MS, the United States: estimates from the NHANES Russell CM. Effects of toothbrush design and III survey. Int J Adult Orthodont Orthognat Surg brushing proficiency on plaque removal. Compend (1998) 13: 97–106. Contin Educ Dent (1997) 18: 572–7. 3. Bowden GHW, Edwardsson S. Oral ecology and 21. Morphis TL, Toumba KJ, Lygidakis NA. Fluoride dental caries. In: Thylstrup A, Fejerskov O, eds, pit and fissure sealants: a review. Int J Paediat Textbook of Clinical Cariology. Copenhagen: Dent (2000) 10: 90–8. Munksgaard (1994) 45–70. 22. Hujoel PP, Loesche WJ. Efficiency of split-mouth 4. Gustafsson BE, Quensel CE, Lanke LKS. The designs. J Clin Periodontol (1990) 17: 722–8. Vipeholm dental caries study. The effect of 23. Hujoel PP, DeRouen TA. Validity issues in split- different levels of carbohydrate intake on caries mouth trials. J Clin Periodontol (1992) 19: 625–7. activity in 436 individuals observed for five years 24. Dorfer¨ CE, Berbig B, von Bethlenfalvy ER, (Sweden). Acta Odontol Scand (1954) 11: 232–64. Staehle HJ, Pioch T. A clinical study to compare 5. World Health Organization. Oral Health Surveys: the efficacy of 2 electric toothbrushes in plaque Basic Methods, 4th edn. Geneva: WHO (1997). removal. J Clin Periodontol (2001) 28: 987–94. 6. Zambon JJ. Periodontal diseases: microbial fac- 25. Corbet EF, Tam JO, Zee KY, Wong MC, Lo EC, tors. Ann Periodontol (1996) 1: 879–925. Mombelli AW, Lang NP. Therapeutic effects of 7. Ainamo J, Bay I. Problems and proposals for supervised chlorhexidine mouthrinses on untreated recording gingivitis and plaque. IntDentalJ gingivitis. Oral Dis (1997) 3: 9–18. (1975) 25: 229–35. 26. Donly KJ, Segura A, Kanellis M, Erickson RL. 8. Silness J, Loe H. Periodontal disease in preg- Clinical performance and caries inhibition of nancy. II. Correlation between oral hygiene and resin-modified glass ionomer cement and amalgam J Am Dental Assoc 130 periodontal condition. Acta Odontol Scand (1964) restorations. (1999) : 1459–66. 22: 112–35. 27. Stephen KW, Macpherson LM, Gilmour WH, 9. Turesky S, Gilmore ND, Glickman I. Reduced Stuart RA, Merrett MC. A blind caries and plaque formation by the chloromethyl analogue of fluorosis prevalence study of school-children in Victamine C. J Periodontol (1970) 41: 41–3. naturally fluoridated and nonfluoridated townships 10. Loe H, Silness J. Periodontal disease in preg- of Morayshire, Scotland. Commun Dent Oral nancy. I. Prevalence and severity. Acta Odontol Epidemiol (2002) 30: 70–9. Scand (1963) 21: 533–51. 28. Lynch E, Baysan A, Ellwood R, Davies R, Peters- 11. Slade GD, ed. Measuring Oral Health and Quality son L, Borsboom P. Effectiveness of two fluoride of Life. Chapel Hill: University of North Carolina dentifrices to arrest root carious lesions. Am J Dent (1997). (2000) 13: 218–20. 12. Dolan TA. The sensitivity of the geriatric oral 29. Machiulskiene V, Nyvad B, Baelum V. Caries health assessment index to dental care. JDental preventive effect of sugar-substituted chewing Educ (1997) 61: 37–46. gum. Commun Dent Oral Epidemiol (2001) 29: 13. Allen PF, McMillan AS, Locker D. An assess- 278–88. ment of sensitivity to change of the Oral Health 30. Poulsen S, Beiruti N, Sadat N. A comparison of Impact Profile in a clinical trial. Commun Dent retention and the effect on caries of fissure Oral Epidemiol (2001) 29: 175–82. sealing with a glass-ionomer and a resin-based 14. Pocock SJ. Clinical Trials: A Practical Approach. sealant. Commun Dent Oral Epidemiol (2001) 29: Chichester, UK: John Wiley & Sons (1983). 298–301. 15. Dennison DK. Components of a randomized clin- 31. Biesbrock AR, Gerlach RW, Bollmer BW, Faller ical trial. J Periodont Res (1997) 32: 430–8. RV, Jacobs SA, Bartizek RD. Relative anti-caries 16. Koch GG, Paquette DW. Design principles and efficacy of 1100, 1700, 2200, and 2800 ppm statistical considerations in periodontal clinical fluoride ion in a sodium fluoride dentifrice over trials. Ann Periodontol (1997) 2: 42–63. 1 year. Commun Dent Oral Epidemiol (2001) 29: 17. Friedman LM, Furberg CD, DeMets DL. Fun- 382–9. damentals of Clinical Trials, 3rd edn. Mosby: 32. Lo EC, Luo Y, Fan MW, Wei SH. Clinical inves- Springer (1998). tigation of two glass-ionomer restoratives used 18. Lau G. Ethics. In: Kalberg J, Tsang K, eds, Intro- with the atraumatic restorative treatment approach duction to Clinical Trials. Hong Kong: The Uni- in China: two-years results. Caries Res (2001) 35: versity of Hong Kong (1998) 73–112. 458–63. DENTISTRY AND MAXILLO-FACIAL 207
33. Meijer HJ, Raghoebar GM, Van’t Hof MA, Geert- 48. Sutherland SE, Walker S. Evidence-based den- man ME, Van Oort RP. Implant-retained mandibu- tistry: Part III. Searching for answers to clinical lar overdentures compared with complete dentures; questions: finding evidence on the Internet. JCan a 5-years’ follow-up study of clinical aspects and Dental Assoc (2001) 67: 320–3. patient satisfaction. Clin Oral Implant Res (1999) 49. Crochane AL. Effectiveness and Efficiency. Ran- 10: 238–44. dom Reflections on Health Services. London: 34. Roccuzzo M, Bunino M, Prioglio F, Bianchi SD. Nuffield Provincial Hospital Trust (1972). Early loading of sandblasted and acid-etched 50. Macfarlane TV, Worthington HV. Some aspects (SLA) implants: a prospective split-mouth com- of data analysis in dentistry. Commun Dental parative study. Clin Oral Implant Res (2001) 12: Health (1999) 16: 216–19. 572–8. 51. Gilthorpe MS, Maddick IH, Petrie A. Introduction 35. de Albuquerque RF Jr, Lund JP, Tang L, Larivee to multilevel modelling in dental research. Com- J, de Grandmont P, Gauthier G, Feine JS. Within- mun Dental Health (2000) 17: 222–6. subject comparison of maxillary long-bar implant- 52. Altman DG, Bland JM. Units of analysis. Br Med retained prostheses with and without palatal J (1997) 314: 1874. coverage: patient-based outcomes. Clin Oral 53. Axtelius B, Soderfeldt¨ B, Attstrom¨ R. A multi- Implant Res (2000) 11: 555–65. level analysis of factors affecting pocket probing 36. Slade GD, Spencer AJ. Development and evalua- depth in patients responding differently to peri- tion of the Oral Health Impact Profile. Commun odontal treatment. J Clin Periodontol (1999) 26: Dental Health (1994) 11: 3–11. 67–76. 37. Duke SP, Garrett S. Equivalence in periodontal 54. Gilthorpe MS, Griffiths GS, Maddick IH, Zamzuri trials: a description for the clinician. J Periodontol AT. The application of multilevel modelling (1998) 69: 650–4. to periodontal research. Commun Dental Health 38. Gunsolley JC, Elswick RK, Davenport JM. Equiv- (2000) 17: 227–35. alence and superiority testing in regeneration clin- 55. Goldstein H. Multilevel Statistical Models, 2nd J Periodontol 69 ical trials. (1998) : 521–7. edn. London: Edward Arnold (1995). 39. Johnston LE Jr. Let them eat cake: the struggle 56. Bryk AS, Raudenbush S. Hierarchical Linear between form and substance in orthodontic clin- Models: Applications and Data Analysis Methods. ical investigation. Clin Orthodont Res (1998) 1: Newbury Park: Sage Publications (1992). 88–93. 57. Hannigan A, O’Mullane DM, Barry D, Schafer¨ F, 40. O’Brien K. Editorial: Is evidence-based orthodon- tics a pipedream? J Orthodont (2001) 28: 313. Robers AJ. A re-analysis of a caries clinical trial 41. Sackett DL, Rosenberg WM, Gary JA, Haynes by survival analysis. J Dental Res (2001) 80: RB, Richardson WS. Evidence based medicine: 427–31. what it is and what it isn’t. Br Med J (1996) 312: 58. Albandar JM, Goldstein H. Multi-level statistical 71–2. models in studies of periodontal diseases. J 42. Sutherland SE. The building blocks of evidence- Periodontol (1992) 63: 690–5. based dentistry. J Can Dental Assoc (2000) 66: 59. Gilthorpe MS, Cunningham SJ. The application of 241–4. multilevel, multivariate modelling to orthodontic 43. Sutherland SE. Evidence-based dentistry: Part I. research data. Commun Dental Health (2000) 17: Getting started. J Can Dental Assoc (2001) 67: 236–42. 204–6. 60. Scheinin A, Makinen KK. Turku sugar studies. 44. Sutherland SE. Evidence-based dentistry: Part II. An overview. Acta Odontol Scand (1976) 34: Searching for answers to clinical questions: how 405–8. to use MEDLINE. J Can Dental Assoc (2001) 67: 61. Rugg-Gunn AJ. Nutrition, diet and dental public 277–80. health. Commun Dental Health (1993) 10: 47–56. 45. Sutherland SE. Evidence-based dentistry: Part IV. 62. Sheiham A. Dietary effects on dental diseases. Research design and levels of evidence. JCan Public Health Nutr (2001) 4: 569–91. Dental Assoc (2001) 67: 375–8. 63. Duggal MS, van Loveren C. Dental considera- 46. Sutherland SE. Evidence-based dentistry: Part V. tions for dietary counselling. Int Dental J (2001) Critical appraisal of the dental literature: papers 51: 408–12. about therapy. J Can Dental Assoc (2001) 67: 64. Linke HA. Sugar alcohols and dental health. 442–5. World Rev Nutr Diet (1986) 47: 134–62. 47. Sutherland SE. Evidence-based dentistry: Part VI. 65. Tresaure ET, Chestnutt IG, Whiting P, McDon- Critical appraisal of the dental literature: papers agh M, Kleijnen J. The York review–a systematic about diagnosis, etiology and prognosis. JCan review of public water fluoridation. Br Dental J Dental Assoc (2001) 67: 582–5. (2002) 192: 495–7. 208 TEXTBOOK OF CLINICAL TRIALS
66. McDonagh MS, Whiting PF, Wilson PM, Sutton 81. Smallridge J. UK National Clinical Guidelines in AJ, Chestnutt I, Cooper J, Misso K, Bradley M, Paediatric Dentistry. Management of the stained Treasure E, Kleijnen J. Systematic review of fissure in the first permanent molar. Int J Paediat water fluoridation. Br Med J (2000) 321: 855–9. Dent (2000) 10: 79–83. 67. O’Mullane DM. Systemic fluorides. Adv Dental 82. Hickel R, Manhart J, Gracia-Godoy F. Clinical Res (1994) 8: 181–4. results and new developments of direct posterior 68. Marino R. Should we use milk fluoridation? A restorations. Am J Dent (2000) 13: 41D–54D. review. Bull Pan Am Health Org (1995) 29: 83. Wilson NH, Burke FJ, Mjor IA. Reasons for 287–98. placement and replacement of restorations of 69. Stephen KW. Fluoride prospects for the new direct restorative materials by a selected group of millennium – community and individual patient practitioners in the United Kingdom. Quintess Int aspects. Acta Odontol Scand (1999) 57: 352–5. (1997) 28: 245–8. 70. Bergmann KE, Bergmann RL. Salt fluoridation 84. Yip HK, Smales RJ. Glass ionomer cements used and general health. Adv Dental Res (1995) 9: as fissure sealants with the atraumatic restorative 138–43. treatment (ART) approach: review of literature. Int 71. Riordan PJ. Fluoride supplements for young chil- Dent J (2002) 52: 67–70. dren: an analysis of the literature focusing on ben- 85. Botelho M. Resin-bonded prostheses: the current efits and risks. Commun Dental Oral Epidemiol state of development. Quintess Int (1999) 30: (1999) 27: 72–83. 525–34. 72. Burt BA. The case for eliminating the use of 86. Loe H. Oral hygiene in the prevention of caries dietary fluoride supplements for young children. and periodontal disease. IntDentalJ(2000) 50: J Public Health Dent (1999) 59: 269–74. 129–39. 87. Westfelt E. Rationale of mechanical plaque con- 73. Clarkson JJ, McLoughlin J. Role of fluoride in trol. J Clin Periodontol (1996) 23: 263–7. oral health promotion. IntDentalJ(2000) 50: 88. Iacono VJ, Aldrege WA, Luckus H, Schwartz- 119–28. stein S. Modern supragingival plaque control. Int 74. Stephen KW. Dentifrices: recent clinical findings Dental J (1998) 48: 290–7. and implications for use. IntDentalJ(1993) 43: 89. Jones CG. Chlorhexidine: is it still the gold 549–53. standard? Periodontology (2000) 15: 55–62. 75. Ripa LW. Clinical studies of high-potency flu- 90. Sheen S, Pontefract H, Moran J. The benefits of oride dentifrices: a review. J Am Dental Assoc toothpaste – real or imagined? The effectiveness (1989) 118: 85–91. of toothpaste in the control of plaque, gingivitis, 76. Warren JJ, Levy SM. A review of fluoride den- periodontitis, calculus and oral malodour. Dental tifrice related to dental fluorosis. Paediat Dent Update (2001) 28: 144–7. (1999) 21: 265–71. 91. Sheiham A. Is the chemical prevention of gin- 77. Holloway PJ, Worthington HV. Sodium fluoride givitis necessary to prevent severe periodontitis? or sodium monofluorophosphate? A critical view Periodontology 2000 (1997) 15: 15–24. of a meta-analysis on their relative effectiveness 92. Greenstein G. Nonsurgical periodontal therapy in in dentifrices. Am J Dent (1993) 6: S55–8. 2000: a literature review. J Am Dental Assoc 78. Newbrun E. Topical fluorides in caries prevention (2000) 131: 1580–92. and management: a North American perspective. 93. Kluemper GT, Beeman CS, Hicks EP. Early ortho- J Dental Educ (2001) 65: 1078–83. dontic treatment: what are the imperatives? JAm 79. Bader JD, Shugars DA, Bonito AJ. A systematic Dental Assoc (2000) 131: 613–20. review of selected caries prevention and man- 94. Harrison JE, Ashby D. Orthodontic treatment for agement methods. Commun Dent Oral Epidemiol posterior crossbites. Cochrane Database System (2001) 29: 399–411. Rev (2001) 1: CD000979. 80. Deery C. The economic evaluation of pit and 95. Berg R. Orthodontic treatment – yes or no? A fissure sealants. Int J Paediat Dent (1999) 9: difficult decision in some cases. JOrofacial 235–41. Orthopaed (2001) 62: 410–21. DERMATOLOGY
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 14 Dermatology
LUIGI NALDI1 AND COSETTA MINELLI2 1Department of Dermatology, Ospedali Riuniti di Bergamo, Bergamo, Italy 2Unit of Clinical Epidemiology, Pharmacology Research Institute, M.Negri-Villa Camozzi, Ranica, Italy
WHAT IS DERMATOLOGY? event occurring with conditions such as exten- sive bullous disorders or exfoliative dermatitis. The most usual health consequence of skin dis- Dermatology deals with disorders affecting the orders is connected with the discomfort of symp- skin and associated specialised structures such as toms, such as itching and burning or pain, which hair and nails. The skin is a biological barrier frequently accompany skin lesions and interfere between ourselves and the outside world con- with everyday life and sleeping. Moreover, vis- sisting of a stratified epithelium, an underlying ible lesions may result in a loss of confidence connective tissue, i.e., dermis, and a fatty layer and disrupt social relations. Feelings of stigmati- usually designed as ‘subcutaneous’. The skin is sation and major changes in lifestyle caused by a not a simple inert covering of the body but chronic skin disorder such as psoriasis have been a sensitive dynamic boundary. It offers protec- repeatedly documented in population surveys.1 tion against infections, ultraviolet radiation and Additional problems may arise under diverse cir- trauma. It is essential for controlling water and cumstances: the exudation or loss of substances heat loss and contributes to the synthesis of sub- that interfere locally with the barrier function stances such as vitamin D. The skin is also an (and dressing); the shedding of scales whenever important organ of social and sexual contact. excessive desquamation occurs; the need to pre- Body perception, which is deeply rooted within vent contact dissemination in the case of trans- the culture of any given social group, is largely missible diseases. affected by the appearance of the skin and its associated structures. Extensive disorders affecting the skin may dis- A LARGE NUMBER OF DIFFERENT SKIN rupt its homeostatic functions resulting in a prop- DISEASES erly speaking ‘skin failure’, needing intensive care with hydration, maintenance of caloric bal- Unlike most other organs that usually count ance and temperature. However, this is a rare around 50 to 100 diseases, the skin has a
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 212 TEXTBOOK OF CLINICAL TRIALS complement of 1000 to 2000 conditions and over have an impact in terms of physical disability or 3000 dermatological categories can be found in even mortality, are rare or very rare. They include, the International Classification for Disease ver- among others, autoimmune bullous diseases, such sion 9 (ICD-9). This is partly justified by the skin as pemphigus, severe pustular and erythrodermic being a large and visible organ. Beside disorders psoriasis, generalised eczematous reactions, and primarily affecting the skin, there are cutaneous such malignant tumours as malignant melanoma manifestations with most of the major systemic and lymphoma. The disease frequency may show diseases (e.g., vascular and connective tissue dis- variations according to age, sex and geographic eases). The classification of skin disorders is far area. Eczema is common at any age while acne from satisfactory (Table 14.1). Currently, there is is decidedly more frequent among male adoles- a widespread use of symptom-based or purely cents. Skin tumours are particularly frequent in descriptive terms, such as parapsoriasis or pytiri- aged white populations. Infestations and infec- asis rosea, which reflects our limited understand- tions such as scabies, pyoderma and dermato- ing of the causes and pathogenetic mechanisms phytosis predominate in developing countries and of a large number of skin disorders. some urban pockets of developed countries. In Skin diseases as a whole are very common many cases, skin diseases are minor health prob- in the general population. A limited number of lems, which may be trivialised in comparison with prevalence surveys have documented that skin other more serious medical conditions. However, disorders may affect 20–30% of the general popu- as mentioned above, skin manifestations are vis- lation at any one time. The most common diseases ible and may cause more distress to the public are also the most trivial ones. They include such than more serious medical problems. The issue is conditions as mild eczematous lesions, mild to complicated by the fact that many skin disorders moderate acne, benign tumours and angiomatous are not present in the population as a yes or no lesions. More severe skin disorders, which may phenomenon, but as a spectrum of severity. The
Table 14.1. Operational classification of skin diseases
Anatomical Pathogenetic distribution Morphology Pathology process Aetiology
Genodermatoses XX Nevi and other development defects XX Mechanical and thermal injuries X Photodermatoses XX Eczemas XX XX Lichenoid disorders XX Disorders of keratinisation XX Psoriasis XX Infections and infestations XX Disorders of skin colour X Bullous eruptions XX X Disorders of sebaceous glands XX X Disorders of sweat glands XX Immune-related diseases XX Urticaria XXX Vascular disorders XXX Disorders of hair XXX Disorders of nails XXX Disorders of subcutaneous fat XXX Tumours XXX DERMATOLOGY 213 public’s perception of what constitutes a ‘disease’ A peculiar aspect of dermatology is the pos- requiring medical advice may vary according to sible option for topical treatment. This treatment cultural issues, the social context, resources and modality is ideally suited to localised lesions, the time. Minor changes in health policy may have a main advantage being the restriction of the effect large health and financial impact simply because to the site of application and the limitation of a large number of people may be concerned. For systemic side effects. A topical agent is usually example, most of the campaigns conducted to described as a vehicle and an active substance, increase the public awareness of skin cancer have the vehicles being classified as powder, grease, led to a large increase in the number of benign liquid or combinations such as pastes and creams. skin conditions such as benign melanocytic nevi Much traditional topical therapy in dermatol- being evaluated and excised. ogy has been developed empirically with so- Large variations can be documented among called magistral formulations. Most of these different countries in terms of health service products seem to rely on physical rather than organisation for treating skin disorders. A rough chemical properties for their effects and it may indicator of these variations is the density of be an arbitrary decision to appoint one spe- dermatologists ranging, in Europe, from about cific ingredient as the ‘active’ one. Physical 1:20 000 in Italy and France to 1:150 000 in the effects of topical agents may include detersion, United Kingdom. hydration and removal of keratotic scales. The In general, only a fraction of those with border between pharmacological and cosmetolog- skin diseases are expected to seek medical help ical effects may be imprecisely defined and the while an estimated large proportion opt for term ‘cosmeceuticals’ is sometimes employed.3 self-medication. Pharmacists occupy a key role It should be noted that the evaluation of even the in advising the public on the use of over- most recent cosmetic products is far from being the-counter products. Primary care physicians satisfactory. In addition to pharmacological treat- seem to treat the majority of people among ment, a number of non-pharmacological treat- those seeking medical advice. Primary care of ment modalities exists including phototherapy dermatological problems seems to be imprecisely or photochemotherapy and minor surgical proce- defined with a large overlap with specialist dures such as electrodessication and criotherapy. activity. Most of the dermatologist’s workload Large variations in treatment modalites for the around the world is concentrated in the outpatient same condition have been documented, which 4,5 department. In spite of the vast number of mainly reflect local traditions and preferences. dermatological diseases, it has been documented that just a few categories account for about 70% ACNE of all dermatological consultations. Brief, more detailed descriptions of the most frequent skin The term acne refers to a group of disorders categories are given below while skin cancer is characterised by abnormalities of the sebaceous dealt with in another section. glands. Acne vulgaris is the most common Generally speaking, dermatology requires a condition and is characterised by polymor- low technology clinical practice. Clinical exper- phous lesions, including comedones (black- tise is mainly dependent on the ability to recog- heads), inflammatory lesions such as papules or nise a skin disorder quickly and reliably which, pustules, and scars, affecting the face and less fre- in turn, depends to a large extent on the aware- quently the back and shoulders. A combination of ness of a given clinical pattern, based on pre- factors are considered as pathogenetic, including vious experience and on the exercised eye of a the hormonal influence of androgens, seborrhea, visually literate physician.2 Complementary diag- abnormalities in the bacterial flora with over- nostic procedures include skin biopsy, patch test- growth of Propionibacterium Acnei, and plugging ing and immunopathology. of pilosebaceous openings. Mild degrees of acne 214 TEXTBOOK OF CLINICAL TRIALS are extremely common amongst teenagers (more two to threefold over the last 30 years. There is than 80%) and decrease in later life. The preva- some evidence that it may be possible to prevent lence of moderate to severe acne has been esti- atopic dermatitis in high-risk children born to par- mated at about 14% in 15–24 year-olds, 3% in ents with atopic disease by restricting maternal 25–34 year-olds and about 1% in 35–54 year- allergens and reducing house dust mite levels.8,9 olds. It is likely that the vast majority of suf- No treatment has been shown to alter the natural ferers of mild acne do not seek medical advice. history of established eczema and the mainstay Around 70% of those affected with acne experi- of treatment is emollients, which moisturise the ence shame and embarrassment because of their skin, and topical steroids. acne. Criteria for treatment include clinical sever- ity, as judged by the extension and presence of inflammatory lesions, and the degree of psycho- PSORIASIS logical distress from the disease. The aim of treatment is to prevent scarring, limit disease This is a chronic inflammatory disorder charac- duration and reduce psychological stress. Mild terisedbyredscalyareas,whichtendtoaffect acne is usually treated by topical modalities such extensory surfaces of the body and scalp. Its as benzoyl peroxide or tretinoin, while moderate overall prevalence is about 1–3% and males are severity acne is treated by systemic antibiotics or affected more frequently than females. Several antiandrogens in women. Oral isotretinoin is used varieties have been described including guttate, under specialist supervision for severe unrespon- pustular and erythrodermic psoriasis. In about sive disease. There are a number of published 3% of cases it may associate with a peculiar 6 systems for measuring the severity of acne. arthritis. Significant disability has been docu- These vary from sophisticated systems with up mented with psoriasis. Multifactorial heredity is to 100 potential grades to simple systems with 4 usually considered for disease causation. This or 5 grades. A specially designed acne disability implies interaction between a genetic predispo- index has also been devised to assess the psycho- sition and environmental factors. Heritability, a logical impact of the disease and disability, and measure that quantifies the overall role of genetic has been found to correlate well with severity as factors, ranges from 0.5 to 0.9. Acute infections, measured by an objective grading system, even physical trauma, selected medications and psy- if a small group experiences disability which is chological stress are usually viewed as triggers. 7 out of proportion with their severity. Pustular varieties of psoriasis have been strongly linked with smoking. Sun exposure usually tem- ATOPIC DERMATITIS porarily improves the disease. Altered kinetics of epidermal cells has been repeatedly documented. Typically, this condition is characterised by itch- The lesions are visible and may itch, sting and ing, dry skin and inflammatory lesions especially bleed easily. The aim of treatment is to achieve involving skin creases. Patients suffering from short-term suppression of symptoms and long- atopic dermatitis may also develop IgE-mediated term modulation of disease severity, improving allergic diseases such as bronchial asthma or the quality of life with minimal side effects. allergic rhinitis. An overall cumulative preva- Topical agents such as vitamin D derivatives, lence of between 5% and 20% has been suggested dithranol and steroids can be used for short-term by the age of 11. Around 60–70% of children are control of the disease. Ultraviolet B phototherapy, clear of significant disease by their mid-teens. psoralen plus ultraviolet A phototherapy (PUVA) Even if genetic factors seem to play a major and systemic agents such as methotrexate or role, environmental factors such as allergens and cyclosporine are employed to control extensive irritants are important and there is reasonable evi- lesions that fail to respond to topical agents. dence to suggest that the prevalence has increased Relapse is common upon withdrawal. Outcomes DERMATOLOGY 215 that matter to the patient include disease suppres- the UK proved to offer advantages over home sion and duration of remission, patient satisfac- treatment.12 The overwhelming rates of recur- tion and autonomy and disease-related quality of rence clearly suggest that more attention should life. A number of clinical activity scores have be paid to prevention. been developed, the most popular being the Pso- riasis Area and Severity Index–PASI.10 There is no documented evidence that such indexes are a SKIN DISORDERS AND CLINICAL TRIAL reliable proxy for the above-mentioned outcomes. METHODS: ADAPTING STUDY DESIGN TO In the long term, a simple measure such as the SETTING AND DISEASE number of patients reaching complete or nearly complete stable remission appears as the most As for other disciplines, the last few decades relevant outcome variable. have seen an impressive increase in the number of clinical trials carried out in dermatology. LEG ULCERS However, there are indications that the upsurge of clinical research has not been paralleled by a Venous and arterial leg ulcers are recognised refinement in clinical trial methodology and the as the most common chronic wounds in West- quality of randomised control trials (RCTs) in ern populations. A skin ulcer has been defined dermatology falls well below the usually accepted as a loss of dermis and epidermis produced by standards.13–15 In this section we would like sloughing of necrotic tissue. Ulcers persisting for to mention some issues which deserve special 4 weeks or more have been rather arbitrarily attention when designing a randomised clinical classified as chronic ulcers. Based on popula- trial in this speciality area. There is a need tion surveys, the point prevalence of leg ulcers for innovative thinking in dermatology to make ranges from 0.1% to 1.0% and increases with clinical research address the important issues and age. Venous ulcers are the end result of super- not simply ape the scientific design. ficial or deep venous insufficiency and a venous origin is diagnosed in about 80% of cases. Arte- RANDOMISATION rial ulceration may be regarded as a multistep process, starting, in general, with a systemic vas- It can be estimated that there are at least cular derangement such as atherosclerosis. The a thousand rare or very rare skin conditions prognosis of leg ulcers is less than satisfactory, where no single randomised trial has been with about one-quarter of subjects not healing in conducted. These conditions are also those which over 2 years and the majority of patients hav- carry a higher burden in terms of physical ing recurrence. In a large-scale clinical study, disability and mortality. The annual incidence the healing time varied according to the dimen- rate of many of them is lower than 1 case sion of the ulcers, their duration and the mobil- per 100 000 and frequently less than 1 case ity of the patient. The quality of life of ulcer per 1 000 000. International collaboration and patients may be severely affected. Social iso- institutional support is clearly needed. There are lation, depression and negative self-image have no examples of such an effort. been associated with ulcers in a high percent- For quite different reasons, there are also com- age of patients.11 A number of studies point to mon skin conditions where randomised clinical the less than satisfactory management of ulcer trials have been rarely performed. These con- patients in the community, including the lack of ditions include several varieties of eczematous any clinical assessment leading to long periods of dermatitis (e.g., nummular eczema), psoriasis ineffective or inappropriate treatment and delays (e.g., guttate psoriasis) and urticaria (e.g., pres- in instituting effective pain-relieving strategies. sure urticaria), a number of exanthematic reac- Ulcer clinics in vascular surgical services in tions (e.g., pytiriasis rosea), rosacea, and common 216 TEXTBOOK OF CLINICAL TRIALS infections such as warts and molluscum con- approach with the aim of clearing existing pso- tagious. One alleged difficulty with mounting riasis lesions, and a maintenance phase, with the randomised clinical trials in dermatology is the main aim of preventing disease relapse. The dif- visibility of skin lesions and the consideration ferent phases are not necessarily well separated that much more than in other areas, patients in time. Long-term disease-modifying strategies self-monitor their disease and may have precon- can be adopted at the same time when a treat- ceptions and preferences about specific treatment ment modality for reaching clearance has been modalities.16 The decision to treat is usually dic- started. An example is the treatment of atopic tated by subjective issues and personal feelings. dermatitis by topical steroids and diet. Most ran- As we will consider below, there is a need to edu- domised clinical trials in dermatology use a sim- cate physicians and the public about the value of plified approach to evaluating treatment effects randomised trials to assess interventions in der- and most of them analyse the effect of a single matology. The need to evaluate the attitudes of manoeuvre over a limited time span. One as yet patients and to educate should be clearly con- not fully explored issue is the potential for com- sidered when planning a study and developing bining different treatment approaches in a simul- modalities to obtain an informed consent from taneous or subsequent order. There are examples the patient. of combinations of such treatment modalities as Randomised clinical trials are usually designed calcipotriol and ultraviolet B radiation in psoria- in dermatology with an expected large effect from sis treatment, but other rational combinations are the test treatment and most trials do not recruit not fully explored. A way of addressing the issue more than a few dozen patients. In small tri- is by relying on factorial design. An example als there may be substantial differences in group of such a design would be a randomised clin- sizes that will reduce the precision of the esti- ical trial of the effect of a low-allergen diet mated differences in treatment effect and hence compared with an unrestricted diet in atopic the efficiency of the study. As a consequence, women during pregnancy and breast-feeding on block randomisation may be preferable. On the the subsequent development of atopic disorders other hand, a substantial imbalance may persist in in children where women are randomised to prognostic characteristics, and minimisation can all the possible combinations of restricted and be used to make small groups more similar with unrestricted dietetic measures during the peri- respect to major prognostic variables.17 There is ods examined. some evidence that the group sizes of clinical trials, apparently based on simple randomisation BLINDING and published in a number of leading dermatolog- ical journals, tend to be much more similar than There are several reasons for considering blind- expected by chance (unpublished data from the ing as a major bias-reducing procedure in ran- European Dermatoepidemiology Network psori- domised clinical trials of skin disorders. Firstly, asis project). The cluster around equal sample it is expected that physicians and patients are sizes may be due to publication bias, failure to subject to strong, though difficult to document, report blocking, or even to the rectification of an hopes and prejudices about the optimal care of unsatisfactory imbalance by adding extra patients skin disorders. This is reflected, for example, to one treatment. in the large variations of treatment procedures In many instances, the management of a for the same condition which have been repeat- chronic skin disorder is a multiple step pro- edly documented in different areas of derma- cess where different phases can be identified. tology. Secondly, most outcome measures are For example, at least two phases are usually soft end points involving subjective judgement, considered when treating psoriasis: a clearance which may be influenced to a significant extent phase, which involves a more intensive treatment by the previous knowledge of the treatment DERMATOLOGY 217 adopted. Thirdly, the visibility of lesions may STANDARDISATION OF TREATMENT influence the decision to rely or not on a given MODALITIES AND ACCESSORY CARE treatment to a larger extent as compared with situations where disease variables are not so obvi- Independently of the ‘active’ intervention admin- ous. On the other hand, there may be prob- istered, accessory non-pharmacological treatment lems with blinding which may be difficult or and skin care seem to play a significant role impossible to solve, like with trials comparing in the outcome of most skin disorders. It is complex procedures such as ultraviolet light radi- common sense that emollients may improve dry ation and drug regimens. An issue which war- skin and wet soaks may help to dry exudat- rants more attention than it is often given in ing lesions. As a consequence, accessory care requires careful standardisation. However, while randomised trials is the possibility that certain it is relatively easy to ensure that the pharma- ‘marker variables’ occur, together with obvious cological treatment is conducted in an appro- side effects. These variables, observable during priate way (particularly timing and administra- treatment, may in part unblind the trial, even at tion route), non-pharmacological accessory care a subliminal level.18 This is an issue with the is prone to a larger variability that is affected use of topical retinoids and the associated mild by social and cultural factors among others. To cutaneous irritation, which may be noticed but a greater extent, variability may affect topical not reported at all as a ‘side effect’. It is quite treatment as compared with systemic treatment. common to find randomised clinical trials where Topical treatment is usually more cumbersome the authors claim blindness in situations where in comparison with systemic treatment and may treatment modalities are responsible for frequent well depend on the physician’s and patient’s con- and obvious side effects. In 1982, for example, sistency. As documented in randomised clinical a trial was published examining three different trials of the retinoid derivative tazarotene in pso- therapeutic strategies for psoriasis: oral etretinate riasis, the modalities of application may play a associated with topically applied betamethasone, significant role in tolerability and side effects.20 oral etretinate associated with topically applied Once again a well-informed patient as well as an placebo, and oral placebo associated with topi- 19 active and supportive clinician are vitally impor- cal betamethasone. Systemic retinoids such as tant. The issue of standardisation is also impor- etretinate are responsible for common side effects tant for assessing compliance when the treatment which are reminiscent of vitamin A overdosage, is self-applied by the patient. If indeed there are including dryness of the skin and mucous mem- limitations with such methods as tablet count for branes, while topical steroids commonly produce assessing compliance with systemic agents, the a transitory blanching effect. It is difficult to limitations are even greater when similar methods accept blindness in the trial when there is no addi- are used to monitor the consumption of topical tional information on how blinding was actually agents in the absence of strict rules to define a assured. It is worth mentioning that the dropout ‘single dose’. The amount consumed cannot be rates showed large variations among the differ- monitored if patients are not carefully instructed ent trial arms because of alleged side effects on how to apply the topical agent. The observed of treatment. compliance behaviour may range from compliant, One way to overcome the problem of an overusing, erratic using and dropping out. unachievable blindness and avoid the influence of the researcher’s subjective judgement is to DIFFERENT STUDY SETTINGS AND plan the study so that the clinician who treats the DISPARATE DISEASE SEVERITY patient is not the same one who judges the effect of the therapy. This way the second clinicians We have already mentioned that the contents of can be blind to the treatment assigned to the primary care for many skin disorders are impre- patient. cisely defined as opposed to specialist clinical 218 TEXTBOOK OF CLINICAL TRIALS practice, with possibly large overlapping areas. example, the category of ‘steroid responsive der- In addition, it has been noticed that there may be matoses’ or ‘itching disorders’. wide variations in terms of severity within any given diagnostic category, with conditions rang- OUTCOME MEASURES ing from subclinical manifestations, e.g., psori- atic ‘markers’, to skin failure, e.g., erythrodermic There are obvious analogies between the prob- or generalised pustular psoriasis. Moreover, it lems implied in the development of severity cri- teria and those implied in outcome measures. should be noticed that for any given disease They both consist of measures that must have there might be clinical variants, which may have the properties of validity and reliability. In addi- peculiar prognostic features and responses to tion, outcome measures must be responsive to treatment, e.g., guttate psoriasis versus plaque change, i.e., they must have the ability to iden- psoriasis. As a consequence, it is of the out- tify what may be small but nevertheless clini- most importance that entry criteria in RCTs of cally important changes. On the other hand, with skin disorders are defined as precisely as possi- severity criteria the intent is more to discriminate ble. This should include as a major requirement between individuals. If an instrument is useful the definition of the study setting, clinical variety, to discriminate between severity levels of psori- disease severity and duration, previous treatments asis, it does not necessarily mean that it will be and concomitant systemic disorders. able to detect changes which are important as a It should be noticed that the severity assess- result of treatment within these categories. The ment of most skin disorders implies an under- distinction between the two aims (i.e., describing standing of the many influences of the disease variations between individuals and assessing vari- on the patient’s life, including disease-associated ations within individuals) is frequently blurred discomfort, level of disability and social disrup- when developing measurement systems for skin tion. Most of these influences are better expressed disorders. In spite of the fact that, from a clini- as a continuum of severity rather than a yes or no cal point of view, distinguishing between disease phenomenon. On the other hand, there are prac- severity levels may represent a different issue tical advantages in trying to translate the contin- as compared with assessing clinically important uum into a limited number of workable severity changes in individual patients, the two issues are categories. The main advantage is a better com- usually dealt with by relying on identical scale pliance with the discrete nature of most clinical systems in dermatology. decisions where thresholds are usually required There are indications that many score sys- for implementing interventions (examples of cat- tems employed in dermatology lack the basic egorical classifications of a severity continuum requirements for reliability and validity. Even are tumour staging and arterial hypertension a simple measure such as the approximate per- definition). Unfortunately, for many inflamma- centage of area involved in a skin disease is tory skin disorders no reliable severity criteria prone to wide inter- and intra-observer varia- have been developed. Even when such criteria tions if the evaluation methods are not clearly are available, there is uncertainty about sever- specified.22 In spite of their lacking basic require- ity thresholds. Consequently, large variations are ments, a large number of different scales have expected to occur among different RCTs and, in been developed for such common disorders as fact, have been documented on several occasions. psoriasis or atopic dermatitis (Table 14.2). One A rather common attitude in published RCTs example is the ‘Psoriasis Area Severity Index’ is the lack of entry criteria and severity defini- (PASI).10 This index is obtained by summing tions, so that the patient population appears to up the scores concerning three features of pso- be recruited in a vacuum.15,21 One habit which riasis, namely the body district affected, the should be discouraged is to include broad diag- severity of the condition (judged by the degree nostic categories that lack specificity like, for of erythema, infiltration and desquamation) and DERMATOLOGY 219
Table 14.2. Measures used in the outcome evaluation of selected skin diseases Disease-specific Quality of Life Disease Clinical scales measures
Acne6,37,38 • Lesion counting (papule, pustule and • ADI (Acne Disability Index) comedone counts) • CADI (Cardiff Acne Disability Index) • Plewing and Kligman grading system • APSEA (Assessment of the • Cunliffe score (Leeds technique) Psychological and Social Effects of • Cook’s photonumeric method Acne) • American Academy of Dermatology • AQL (Acne Quality of Life) index (AAD) classification • Allen and Smith photographic method • Fluorescence photography • GAGS (Global Acne Grading System)
Atopic • SCORAD (severity Scoring of Atopic • EDI (Eczema Disability Index) dermatitis38–40 Dermatitis) • SASSAD (Six-Area, Six-Sign Atopic Deramtitis) severity index • ADASI (Atopic Dermatitis Area and Severity Index) • EASI (Eczema Area and Severity Index) • Rajka and Langerland scoring system • SSS (Simple Scoring System) • BCSS (Basic Clinical Scoring System) • ADSI (Atopic Dermatitis Severity Index) • SIS (Skin Intensity Score) • ADAM (Assessment Measure for Atopic Dermatitis) • Nottingham Eczema Severity Score Psoriasis22,41 • Severity scores based on individual • PDI (Psoriasis Disability Index) signs (involved Body Surface Area, • PLSI (Psoriasis Life Stress Inventory) erythema, induration, desquamation) • PASI (Psoriasis Area and Severity Index) • SAPASI (Self-administered PASI) • Ultrasound evaluation of the thickness of psoriasis
Leg ulcers42,43 • Clinical skin score • Simple wound measurements • Planimetric wound area measurements Dermatological • DIDS (Dermatology Index of • DLQI (Dermatology Life Quality diseases as a Disease Severity) Index) class38,44,45 • CDLQI (Children’s Dermatology Life Quality Index) • IMPACT (Impact of Skin Disease Scale) • SKINDEX 220 TEXTBOOK OF CLINICAL TRIALS the extension of the disease. The last two are if the PASI index accurately quantifies disabil- judged according to the body district analysed. ity or discomfort, then it may be of value as a Although the PASI score has been widely used, it surrogate outcome measure for psoriasis. What is largely unsatisfactory.22 It has never been stan- may be a relevant outcome variable is a mat- dardised and there is limited testing for inter- and ter of judgement, based on the knowledge of the intra-rater reliability. Validity is another issue. It disease, the patients’ requirements and the val- has never been demonstrated that the weights ues of society. The outcome of skin disorders arbitrarily attributed to each item in the PASI that affect the quality rather than the quantity of score actually reflect the clinical severity of life is expected to be largely culture-dependent. lesions. PASI is only relying on the derma- It is our conviction that the development of a tologist’s judgement of a few clinical features ‘gold standard’ requires a deep understanding of psoriasis and there is increasing awareness of patients’ requirements and expectations from that the patient’s judgement is equally impor- treatment. In the lack of reliable scales, trials with tant. An additional drawback of PASI is that the simplest and most objective outcome vari- similar scores can be attributed to varieties of ables are preferable. Such measures as complete psoriasis which differ clinically and in terms of remission or recurrence should be preferred, pro- response to treatment.23 The ‘Self-Administered vided that these categories are clearly defined. PASI’ (SAPASI), which asks the patient to make Clearly, remission or recurrence are events which the same evaluation as the physician for PASI, occur with a lower frequency as compared with does not escape the limitations we have pointed less dramatic variations in disease activity mea- sured by clinical scales. This, in turn, affects the to for PASI as an outcome measure.24 sample size calculation. To overcome the problems arising from subjec- There are at least two different choices when tive judgement, more ‘objective’ measures have analysing outcome measures expressed by any been repeatedly advocated, such as the use of given score system. We might compare the ultrasound to evaluate the thickness of psoriasis difference between the initial and final scores in plaques.23 In fact, any measurement is fully justi- the treatment and control groups or, alternatively, fied only when it represents a good surrogate for ignore the pre-test scores and simply analyse the clinically important outcomes, such as the patient 25 scores after treatment. There are two important disability and quality of life. analytical reasons to consider in the use of The notion of responsiveness to change expres- change scores.27 The first is that the subtraction ses the idea that any measure used in a of scores before treatment has the effect of trial should be sensitive to ‘clinically important removing stable individual differences between 26 changes’ in response to therapy. A conceptual subjects, thereby increasing the power of the difficulty arises in specifying what a clinically statistical test. The second reason, which is of important change is. With most scales developed relatively minor importance in a randomised in dermatology the issue remains fraught since trial, is that there may be overall differences no ‘gold standard’ has gained wide acceptance. between the two groups at the baseline, and the It should be considered that the ‘outcome’ of use of change scores can potentially correct for the treatment refers to ‘all the possible results these differences. The usual presentation of score that stem from preventive or therapeutic interven- data over time (e.g., PASI score) is to build tions’ and consists of several separate dimensions up a curve based on the mean score values of (e.g., discomfort and disability), which may be the treatment and control groups. A common broken down into components and simple mea- but inappropriate analysis of these curves is surable items. Any given measure achieves its to apply two separate sample tests on mean value only to the extent to which it serves as a score values at several time points (Figure 14.1). proxy for an outcome component. For example, The means may not represent a good descriptor DERMATOLOGY 221
10 Calcipatriol 9 Betamethasone * p < 0.001 8
7
6 * 5 PASI score PASI 4 *
* 3
2
1
0 0 2 4 6 Time (wk) The figure shows a common but unsatisfactory modality of analysing PASI scores collected serially in an RCT. The mean score is calculated at different time points and a graph is presented with lines joining the means at the different time points for the experimental and control group. In the graph, ‘errors’ bars are attached at each time point and an indicator of statistical significance is placed by each time point to summarise the results of separate significance tests. The curve joining the means may not be a good descriptor of a typical curve for an individual and no account in the analysis is taken of the fact that measurements at different time points are from the same subjects and are likely to be correlated. The number of statistical tests performed and the choice of time points to be tested are additional problematic issues. Further, dividing the results into ‘significant’ and ‘not significant’ introduces an artificial dichotomy into serial data. Source: Reproduced from Kragballe et al.,46 with permission from Elsevier.
Figure 14.1. Problematic analysis of PASI score over time of a typical curve for an individual and the better the median, of indexes such as PASI in separate analyses of different time points does not different treatment groups. In addition, the score convey information on how individual subjects of patients who leave the study prematurely and respond over time. Moreover, this practice can are lost to follow-up cannot be evaluated and be criticised on statistical grounds because of the ‘intention-to-treat’ analysis may be difficult multiple potentially data-driven statistical tests to perform. In this respect, the use of simple and because the values over time are not clinical variables (e.g., the number of total or independent and one time point is likely to partial remissions) could be more informative. influence successive time points.28 It should be A remedy has been proposed for the analysis of noticed that the information from each patient serial measurements. In the first stage, a suitable might be diluted when comparing the mean, or summary of the response in each individual, 222 TEXTBOOK OF CLINICAL TRIALS such as a rate of change or an area under the or the use of minoxidil in androgenetic alopecia. curve, is identified and calculated. Subsequently, It is widely accepted that a phase I study is these summary measures are analysed by simple one that examines the initial introduction of a statistical techniques.28 drug in human beings with the treatment tested In conclusion, the situation of outcome eval- either in normal volunteers or in patients. The uation for many skin disorders could be briefly main issues are the pharmacokinetics, pharmaco- summarised as follows. (1) Doctor-centrism: con- dynamics and tolerability of the drug being tested siderably more attention has been given to the with a focus on assessing inter-patient variability. dermatologists’ views on treatment effects than While problems with systemic drugs in dermatol- to those of the patients. (2) Biologic reduction- ogy do not differ from those usually encountered ism: most of the interest has been concentrated on in other speciality areas, some peculiarities exist skin involvement. Only recently have psycholog- with the assessment of topical drugs. Penetration ical and social factors relevant to outcome been within the deep epidermal layers and dermis is appraised. (3) Process, i.e., what happens along a parameter of particular interest since it clearly the way, rather than outcome: assessment of affects the local activity of the drug itself. On the involved areas by ‘objective’ quantitative indexes other hand, pharmacokinetic parameters describ- has been preferred to hazarding definition of syn- ing such a penetration are less stringent as com- thetic relevant outcomes (e.g., clearing). pared with systemic drugs. The assessment can be performed on normal or diseased skin. Relevant methods are those which allow measurements of PHASE I AND PHASE II STUDIES the concentration of the drug in the skin, in a given time, after topical application, while con- The ordered development of treatment modal- centration gradients are formed. Such profiles are ities according to well-identifiable phases29 is usually obtained by direct invasive techniques the exception rather than the rule in dermatol- (e.g., skin biopsy) using topically applied radi- ogy. There are several reasons for this situation. olabelled drugs. In some instances, a close corre- Many treatments are non-pharmacological inter- lation has been documented between the barrier ventions (e.g., ultraviolet phototherapy) which do function of the horny layer, its reservoir function not need to comply with the regulatory require- and the resulting penetration into the skin. Pen- ments for drug development, and there are no etration into human skin can thus be predicted strict guidelines on how to assess them at an from drug quantification in horny layer strip- early clinical phase. Secondly, in spite of being pings. This allows non-radioactive methods of so common, the resources allocated to the study drug dosage, like high-performance liquid chro- of skin disorders are limited as compared with matography, to be applied. Indirect measurements other clinical areas. As a consequence, our under- such as urinary excretion or blood levels are also standing of pathomechanisms is limited, as it analysed as parameters indicative of the systemic is the development of disease-specific therapy. adsorption of the drug and possible toxicity. In Until the causation of the main skin disorders is many instances, it may be of interest to per- unravelled, disparate therapies with imprecisely form penetration studies in the same patient with defined biological activities will continue to be the drug being applied on the involved versus available and many treatments will enter the ther- the uninvolved skin. Whenever the horny layer apeutic arena serendipitously. It was the case of barrier is disrupted, penetration within the dis- a renal-transplant recipient with psoriasis whose eased area is usually facilitated. In addition to skin lesions cleared with cyclosporine that led to adsorption, tolerability of a locally applied drug studies demonstrating the efficacy of that drug.30 may be of interest. This is usually evaluated by Similar considerations can be made for such treat- studying local reactions with increasing concen- ment modalities as topical vitamin D in psoriasis trations of the drug. All the above-mentioned DERMATOLOGY 223 studies are usually conducted on a few healthy of psoriasis (unpublished data), a self-controlled subjects or voluntary patients and in an uncon- design accounted for one-third of all the studies trolled way. Measurement error is a crucial issue, examined and was relied on at any stage in which needs standardisation and careful evalua- drug development. Crossover studies are studies tion at the design level. where patients are randomly allocated to study For a limited number of topical drugs phar- arms, where each arm consists of a sequence macodynamic parameters have been developed. of two or more treatments given consecutively. An example is the blanching or vasoconstric- These trials allow the response of a subject to tion assay, which has been employed to screen a given treatment A to be contrasted with the new topical steroids for clinical efficacy. The same subject’s response to treatment B. There are bioavailability of steroids from topical formula- some inconsistencies with the definition of self- tion has been rather improperly defined as the controlled studies provided by different authors. relative absorption efficiency of a drug, as deter- We consider as self-controlled studies those mined by the release of the steroid from its for- clinical trials where patients act simultaneously mulation. Its subsequent penetration through the as their own control. A prerequisite for this kind stratum corneum and viable epidermis into the of study is the existence of pair organs, e.g., eyes, dermis would produce the characteristic blanch- which can be treated by a locally applied drug ing effect. This effect is measured through scores in the lack of any significant systemic effect. that have a subjective component and need care- From our definition we exclude either those ful standardisation. There have also been some studies where a single treatment is administered attempts to identify biologic pharmacodynamic to patients and a ‘before–after’ comparison is markers of some chronic skin disorders like pso- carried out, and the so-called ‘N-of-one’ RCTs, riasis to be used at an early stage of drug where different time periods are randomised in a development.31 However, these indexes are based single patient to different treatment. on cross-sectional studies and there is still lim- The main advantage of a within-patient study ited information on their modifications with dis- over a parallel concurrent study is a statistical ease activity. one. A within-patient study obtains the same sta- According to the Food and Drug Administra- tistical power with far fewer patients, while at tion (FDA) regulations, a phase II study is the the same time reducing problems of variabil- first controlled clinical study that evaluates the ity between the populations confronted. Within- effectiveness of a drug for a given specific thera- patient studies may be useful when studying peutic use in patients. It is also the first study to conditions that are uncommon or show a high evaluate the risks of a drug’s side effects. Such a degree of patient-to-patient variability. On the study is typically a well-controlled, very closely other hand, within-patient studies impose restric- monitored trial that tests a relatively small, nar- tions and artificial conditions, which may under- rowly defined patient population, usually num- mine validity and generalisability of results and bering no more than a few hundred patients. If may also raise some ethical concern. The washout the criterion is the number of patients recruited, period of a crossover trial as well as the treatment then most randomised clinical trials in dermatol- schemes of a self-controlled design, which entails ogy would come under this definition. applying different treatments to various parts of Study designs that are frequently employed at a the body, do not seem to be fully justifiable from preliminary stage in drug development are within- an ethical point of view. In fact, they don’t satisfy patient control studies, i.e., crossover and self- the principle of providing the patients enrolled controlled studies or simultaneous within-patient in clinical trials with the best-proven diagnostic control studies. In dermatology they are also and therapeutic method. By necessity these stud- used, albeit improperly, at a more advanced stage. ies are restricted to the evaluation of short-term In a survey of more than 350 published RCTs outcomes. A higher degree of collaboration from 224 TEXTBOOK OF CLINICAL TRIALS
half body quarter of body left-right limbs
left-right arms four limbs two patches Source: Reproduced from Naldi et al.,32 with permission.
Figure 14.2. Design of intra-patient comparison in 26 trials concerning dithranol short-contact therapy of psoriasis the patients is requested as compared with other 26 self-controlled trials on short-contact dithranol study designs. Clearly, the impractical treatment in psoriasis (Figure 14.2), which had a median modalities in self-controlled studies or washout number of 16 patients (range 5 to 63), half of period in crossover studies may be difficult for the trials experienced dropout.32 Dropouts may the patient to accept. In this kind of study the have more pronounced effects in a within-patient number of dropouts may be higher when com- study as compared to other study designs because pared to parallel group designs. In a survey of each patient contributes a large proportion of DERMATOLOGY 225 the total information, and the design is sensi- opposed to those driven by surrogate end points) tive to departure from the ideal plan. The sit- and last longer than phase II trials. The distinction uation is compounded in self-controlled studies between phase II and phase III trials is blurred in where the dropping out from the study may be dermatology, where most randomised trials are caused by observing a difference in treatment small and, being short-term, employ surrogate effect between the parts in which the patient has measures in well-selected groups of patients. A been ‘split up’. In this case, given that dropouts few points are worth mentioning when discussing are related to a difference in treatment effect the design of phase III studies in the area of between interventions, the estimate of the effect dermatology. of the intervention could be incorrect and falsely equalised. There are several more problems to be PATIENT MOTIVATION AND PREFERENCE considered. Contamination of treated areas and systemic absorption may complicate the interpre- It has already been mentioned that one of the tation of self-controlled studies, while crossover main concerns of patients suffering from a skin disorder is the visibility of lesions and, much studies require that the disease lasts long enough more than in other areas, the patient self-monitors to allow the investigator to expose patients to his/her disease. Patients’ motivations and previ- each of the experimental treatments and mea- ous experience are obvious crucial points when sure the response. Also the treatment must be entering a trial. Motivations and expectations are one that does not permanently alter the disease likely to influence clinical outcome of all treat- or process under study. Carry-over and period ments, but they may have a more crucial role in effects may clearly compound the analyses.33 situations where ‘soft end points’ are of concern Generalisability is an issue of concern in within- as in dermatology. Commonly, more than 20% patient controls. Not only entry criteria are usu- of the patients entering randomised clinical trials ally greatly restricted, e.g., symmetrical lesions, of psoriasis experience improvement on placebo but also outcome measures need to be selected independently of the initial disease extension. among those reflecting short-term changes in dis- Motivations are equally important in pragmatic ease activity. Such issues as patient satisfaction trials where different packages of management and quality of life are obviously beyond the scope are evaluated, such as in the comparison of a self- of a self-controlled design. It is surprising that administered topical product for psoriasis with self-controlled designs have been the preferred hospital-based therapy like phototherapy. Tradi- design in situations like topical immunotherapy tionally, motivation is seen as a characteristic of of alopecia areata or short-contact therapy of the patient that is assumed not to change with the psoriasis where patient satisfaction and mainte- nature of the intervention. However, it has been nance of effects over time (e.g., maintenance of argued that it is more realistic to view motivation the hair restoration to an acceptable extent) are in terms of the ‘fit’ between the nature of the vitally important. treatment and the patient’s wishes and percep- tions, especially with complex interventions that require the patient’s active participation. We have PHASE III TRIALS already mentioned that the boundary between disease and non-disease is particularly shady in From phase III studies we request randomised dermatology. On the other hand, the public is trials that gather additional information regarding confronted with a great deal of uncontrolled and the effectiveness and safety of a treatment, under sometimes misleading or unrealistic messages on conditions which are closer to the usual clinical how to improve the body’s appearance. All in all, practice as compared with phase II trials. They there is a need to ensure that patient information should study those clinical outcomes that are and motivations are taken into proper considera- of major interest to physicians and patients (as tion when designing and analysing clinical trials 226 TEXTBOOK OF CLINICAL TRIALS on skin disorders. The issue is not only a mat- with variations in disease severity over time. This ter of ‘informed consent’. There is a need to is commonly observed with chronic inflamma- study the influences that determine patients’ pref- tory skin diseases characterised by a relapsing erences and to understand how these may affect course such as atopic dermatitis or psoriasis. in the outcome of clinical trials. A distinction should situations where a variable time-course of the be drawn between an informed choice based on clinical condition is expected, it may be advis- factual data–such as a reliable estimate of the able to proceed with sequential evaluations using risks and benefits of interventions–and attitudes standardised criteria to judge the stability of the towards treatment based on emotional aspects disease over time. Quite surprisingly, informa- and preconceptions. In recent years, a number tion about the stability of the clinical condition is of design variants on the traditional randomised often neglected in clinical trial reports. A review trial have been proposed to take into account that focused on the selection of patients with patients’ preferences. They include the partially psoriasis examined more than 60 clinical trials randomised patient preference design and the between 1988 and 198921 and documented that so-called randomised consent or Zelen design.34 information about the stability of the condition These designs have never gained wide acceptance was missing in more than 70% of the studies. and none have ever been used in dermatology. Exclusion criteria have the function of select- The shift from a paternalistic attitude, whereby ing the ‘more suitable’ patients among all possi- enrollment decisions are made by doctors, to the ble candidates (e.g., excluding patients in whom choice freely exercised by individual patients is the treatment under investigation is contraindi- likely to affect the composition of populations in cated). This selection also has the aim of reducing clinical trials. However, when agreement to enrol factors of variability in the study population, in is based on patients’ preferences for individual order to maximise the chance of detecting and treatments, as in the Zelen design, the group quantifying the treatment effect (e.g., excluding assembled is unlikely to mirror the target popula- patients who are too young or too old). The tion of all the eligible patients. There is a need to best way to provide an account of the selec- study the influences that determine patients’ pref- tion process is a log that lists the included and erences and understand how these may influence excluded patients and specifies the reasons for the final outcome of a trial. In a recent survey, exclusion. This is rarely found in clinical trial Dutch patients affected by psoriasis considered reports concerning skin disorders. An example of the safety issue and long-term management as how far exclusion criteria may operate and limit more important than fast clearing.16 It was also the possibility of generalising the study results important to them to have a vote in the selection is offered by a clinical trial on the effectiveness of the treatment. It is worth mentioning that the of a Chinese herbal extract called ‘Dabao’ in 35 large majority of RCTs in psoriasis are short-term the treatment of alopecia androgenetica. Among studies dealing with short-term clearance rates the 3000 patients available to take part in the that are assessed by the treating physicians. There trial, only 396 were eventually selected to be is room for testing study designs that allow for randomised in the treatment or placebo group. different preference assessment strategies. Such exclusion must be a warning when interpret- ing the actual effectiveness of Dabao on males affected by alopecia androgenetica. It is quite ENTRY CRITERIA plausible that a similar selection process operates in many RCTs concerning skin disorders. The definition of the study population is of par- ticular importance in dermatology where large PLACEBO USE variations in disease severity and different clini- cal subgroups may exist–e.g., plaque versus gut- There are still controversies about the use of tate psoriasis. In addition, there may be problems placebo in randomised control trials. It is widely DERMATOLOGY 227 accepted that ‘in any medical study every patient but questionable claim that justifies this practice should be assured of the best proven diagnostic from an ethical point of view is that withhold- and therapeutic method’. As a consequence, ing the active therapy does not necessarily affect the use of placebo should be proscribed when the long-term prognosis. The above-mentioned a ‘proven’ therapeutic method exists. In spite issues of symptomatic relief and moderate sever- of these principles, studies which breach the ity disorders are commonly encountered in der- ethical principle are still commonly conducted matology and, in fact, a large number of placebo- with the approval of regulatory agencies and controlled RCTs are conducted in this area even institutional review boards. It is widely accepted when alternative therapies exist. The results of that placebo-controlled trials have high internal delaying or withholding the treatment may not validity, but they may be difficult to apply to be straightforward in dermatology. However, clinical practice in situations where alternative there is no question that an extraordinary large interventions of proven efficacy already exist. In number of similar molecules employed for the these circumstances, the information of clinical same clinical indications can be found in this value is the effect size of the new intervention as area. These molecules are mostly assessed in compared with the alternative treatment strategy. placebo-controlled RCTs rather than in compar- The use of placebo may sometimes undermine the ative RCTs. Examples include topical steroids, validity of the study if the treatment falls short oral antihistamines, antifungal drugs and topi- of patients’ expectations, resulting in reduced cal antibacterial drugs. More than 200 treatment compliance and a large dropout rate. Some years modalities were identified in a recent survey of ago a placebo-controlled trial was published on published clinical trials of psoriasis with only a the effect of ebastine, an H1 receptor antagonist, few comparative trials. There is a need to estab- in chronic urticaria.36 A number of other non- lish criteria for the use of placebo in dermatology. sedative antihistamine drugs of proven efficacy They should be developed with the active and were available when the trial was conducted. One informed participation of the public and should might argue that it is unethical to deprive the be considered by review boards and regulatory patients in the placebo group of any effective agencies. Pragmatic randomised trials contrast- therapy, even if only for a limited time (14 days ing alternative therapeutic regimens are urgently in this study). As a matter of fact, the authors needed to inform clinical decisions. reported a high number of dropouts due to the lack of effect in the placebo group. A remark TIME FRAME FOR EVALUATION AND on the possible misinterpretation of the results of OUTCOME MEASURES IN CONTEXT placebo-controlled trials comes from this study. The authors’ conclusion that ‘ebastine represents This discussion will focus on chronic inflam- an effective and well tolerated alternative to other matory skin disorders like psoriasis or atopic non-sedative antihistamine drugs in the treatment dermatitis. There is a necessary link between of chronic urticaria’ is likely to be true but far the time frame for evaluation and the measures from proven. adopted to assess clinical response; therefore the Researchers may have a number of different two issues should be dealt with together. Many options for their choice of placebo or compari- chronic inflammatory skin diseases do not neces- son intervention in randomised clinical trials but, sarily have a progressive deteriorating course, but in practice, many regulatory agencies still con- they may vary in severity over time causing prob- sider placebo controls as the ‘gold standard’. lems that are similar to those encountered with Placebo controls are usually required for the eval- many psychiatric disorders and some rheumatic uation of symptom relief or short-term modifica- diseases. Whenever a definite cure is not rea- tion of disorders of moderate severity even when sonably attainable, it is common to distinguish an alternative treatment is available. The usual between short, intermediate (usually measurable 228 TEXTBOOK OF CLINICAL TRIALS within months) and long-term outcomes. We have complete assessment at withdrawal and are fol- already mentioned that clearing the disease in the lowed up. If some categorical outcome variable short term is different from maintaining clearance is also considered, e.g., hospital admission, the over time, and long-term results are not simply relation between the score value at withdrawal predictable from short-term outcomes. Most of and the final outcome may be explored. the score systems available for skin disorders seem to fit best with the clearance issue. On the other hand, it is not easy to define what repre- OTHER ISSUES sents a clinically significant long-term change in the disease status. This is an even more diffi- The most precise definition of the profile of an cult task than defining outcome for other clinical intervention requires a perspective on the risks conditions, such as cancer or ischaemic heart dis- and benefits, which is wider than the one usu- eases, where mortality or major hard clinical end ally provided by any single RCT. For many points (e.g., myocardial infarction) are of particu- chronic skin diseases, efficacy data are derived lar interest. In the long-term, the way the disease from short-term RCTs, whereas patients tend is controlled and the treatment side effects are to be treated over years. The main issues of vitally important. It has been documented that safety and long-term effectiveness are usually compliance with the duration of the treatment is addressed in the context of observational studies, limited and is worst with topical treatments.16 i.e., phase IV studies. One of the best examples Measures of the quality of life appear rather of such a study is the PUVA follow-up study, attractive. However, what represents an important a cohort study of more than 1400 patients who change for most quality of life measures is impre- had received a first course of PUVA-treatment in cisely defined especially if one considers a long- 1977. These patients are still being followed up term time frame for evaluation. Clearly, treatment and provide information on disease associations effects can be seen from different perspectives and prognostic factors. The study pointed to a and several dimensions can be taken into account. dose-related increased risk of non-melanoma skin cancer in PUVA-treated patients. We lack simi- However, in view of the limitations of the avail- lar studies for many other systemic treatments of able measures in the long term, simple and cheap psoriasis, including methotrexate, retinoids and outcome measures applicable in all patients seem cyclosporine. The safety profiles of most sys- to be preferable. These may include the number temic antihistamines are also imprecisely defined. of patients in remission, the number of hospi- Observational studies may represent the most fea- tal admissions or ambulatory consultations and sible way to study the usefulness of long-term major disease flare ups. Dropouts merit special treatment strategies for chronic inflammatory skin attention. In chronic inflammatory skin diseases diseases, when disease modification rather than that lack hard end points, they may strongly symptom control becomes a desired outcome. As reflect dissatisfaction with treatment. Whatever has been proposed for some rheumatologic disor- the outcome measure adopted, dropouts cannot ders, e.g., rheumatoid arthritis, drug survival, etc. simply be ignored because the patients who do the interval individual patients remain on an agent not provide PASI, Disability or Quality of Life may offer an indication for long-term acceptabil- scores might be different from those who do. ity that takes into account adverse effects, lack or Analysis by randomised group irrespective of loss of effect and patients’ preference. subsequent changes is the method recommended A final mention should be made of those for the analysis of clinical trials. This analysis activities that aim at summarising the results of poses special problems when relying on quan- several RCTs on the same issue. There is a large titative scores. It is suggested that every effort burden of small RCTs13 addressing disparate should be made to ensure that patients have a clinical questions, as well as a lack of consensus DERMATOLOGY 229
Table 14.3. List of the systematic reviews on skin conditions already available, or in an advanced stage of development, in the Cochrane Library (Cochrane Skin Group, August 2000)
Completed reviews Surgical treatments for ingrowing toenails Topical treatments for fungal infections of the skin and nails of the foot Minocycline for acne vulgaris: efficacy and safety Interventions for guttate psoriasis Systemic treatments for metastatic cutaneous melanoma Antistreptococcal interventions in the treatment of guttate and plaque psoriasis Reviews undergoing the editorial process Drug treatments for discoid lupus erythematosus (DLE) Laser resurfacing for the improvement of facial acne scarring Protocols under conversion to reviews Systemic treatments for fungal infections of the skin of the foot Antihistamines for atopic eczema Interventions for toxic epidermal necrolysis (TEN) Complementary therapies for acne Local treatments for common warts Interventions for photodamaged skin Interventions for chronic palmoplantar pustular psoriasis
Source: Reproduced from the Cochrane Library. on the management of many skin disorders. This 3. The limitations of available outcome measures creates an increasing emphasis on systematic that neglect patients’ needs and expectations. reviews, and a Cochrane Skin Group has been 4. Problems with external generalisability like established within the Cochrane Collaboration the lack of adequate description of the study in 1997. A list of systematic reviews already populations and study settings. available within the Cochrane Library is reported 5. The lack of comparative RCTs. in Table 14.3. 6. The overwhelming role of pharmaceutical In the light of the increasing role system- industries with defining priorities. atic reviews may play with informing clinical practice, special care should be devoted to set- ting priorities so that the most important ques- REFERENCES tions are addressed first. Otherwise, they would risk amplifying the irrelevant issues. The impor- tance of the involvement of consumers cannot 1. McKenna KE, Stern RS. The outcomes movement be underestimated. The first analyses from sys- and new measures of the severity of psoriasis. J Am Acad Dermatol (1996) 34: 534–8. tematic assessment of published RCTs point to 2. Webster GF. Is dermatology slipping into its anec- some ‘peculiarities’ of dermatology that have dotage? Arch Dermatol (1995) 131: 149–50. already been discussed in the previous sections, 3. Vermeer BJ, Gilchrest BA. Cosmeceuticals – a and include among others: proposal for rational definition, evaluation, and regulation. Arch Dermatol (1996) 132: 337–40. 4. Peckham PE, Weinstein GD, McCullough JL. The 1. The ‘moving boundary’ between cosmetology treatment of severe psoriasis: A national survey. Arch Dermatol (1987) 123: 1303–7. and medicine. 5. Farr PM, Diffey BL. PUVA treatment of psoriasis 2. The need to develop study designs that address in the United Kingdom. Br J Dermatol (1991) 124: questions posed by chronic recurrent diseases. 365–7. 230 TEXTBOOK OF CLINICAL TRIALS
6. Doshi A, Zaheer A, Stiller MJ. A comparison of 20. Weinstein GD. Safety, efficacy and duration of current acne grading systems and proposal of a therapeutic effect of tazarotene in the treatment novel system. Int J Dermatol (1997) 36: 416–18. of plaque psoriasis. Br J Dermatol (1996) 135 7. Motley RJ, Finlay AY. Practical use of a disability (Suppl): 32–6. index in the routine management of acne. Clin Exp 21. Petersen LJ, Kristensen JK. Selection of patients Dermatol (1992) 17: 1–3. for psoriasis clinical trials: a survey of the recent 8. Zeigher RS, Heller S, Mellon MH, Forsythe AB, dermatological literature. JDermatolTreat(1992) Hamburger RN, Schatz M. Effect of combined 3: 171–6. maternal and infant food-allergen avoidance on 22. Ashcroft DM, Li Wan Po A, Williams HC, Grif- development of atopy in early infancy: a ran- fiths CEM. Clinical measures of severity and out- domized study. J Allerg Clin Immunol (1989) 84: come in psoriasis: a critical appraisal of their qual- 72–89. ity. Br J Dermatol (1999) 141: 185–91. 9. Arshad SH, Matthews S, Gant C, Hide DW. Ef- 23. Marks R, Barton SP, Shuttleworth D, Finlay AY. fect of allergen avoidance on development of Assessment of disease progress in psoriasis. Arch allergic disorders in infancy. Lancet (1992) 339: Dermatol (1989) 125: 235–40. 1493–7. 10. Frederiksson AJ, Peterssonn DC. Severe psoria- 24. Feldman SR, Fleischer AB Jr, Reboussin DM, sis: oral therapy with a new retinoid. Dermato- Rapp SR, Exum ML, Clark DR, Nune L. The logica (1978) 157: 238–44. self-administered psoriasis area and severity index 11. Phillips T, Stanton B, Provan A, Lew R. A study is valid and reliable. J Invest Dermatol (1996) 106: of the impact of leg ulcers on quality of life: 183–6. financial, social, and psychological implications. 25. Krueger GG, Feldman SR, Camisa C, Duvic M, J Am Acad Dermatol (1994) 31: 49–53. Elder JT, Gottlieb AB, et al. Two considerations 12. Moffatt CJ, Franks PJ, Oldroyd M, Bosanquet N, for patients with psoriasis and their clinicians: Brown P, Greenhalgh RM, McCollum CN. Com- what defines mild, moderate and severe psoriasis? munity clinics for leg ulcers and impact on heal- What constitutes a clinically significant improve- ing. BMJ (1992) 305: 1389–92. ment when treating psoriasis? J Am Acad Derma- 13. Williams HC, Seed P. Inadequate size of ‘nega- tol (2000) 43: 281–3. tive’ clinical trials in dermatology. Br J Dermatol 26. Husted JA, Cook RJ, Farewell VT, Gladman DD. (1993) 128: 317–26. Methods for assessing responsiveness: a critical 14. Lehmann HP, Robinson KA, Andrews JS, Hol- review and recommendations. J Clin Epidemiol loway V, Goodman SN. Acne therapy: a method- (2000) 53: 459–68. ologic review. J Am Acad Dermatol (2000) 47: 27. Norman GR. Issues in the use of change scores 231–40. in randomized trials. J Clin Epidemiol (1989) 42: 15. Naldi L, Svensson A, Diepgen T, Elsner P, Grob 1097–105. JJ, Coenraads PJ, Bavinck JN, Williams H. Euro- 28. Matthews JNS, Altman DG, Campbell MJ, Roys- pean Dermato-Epidemiology Network. Random- ton P. Analysis of serial measurements in medical ized clinical trials for psoriasis 1977–2000: the research. Br Med J (1990) 300: 230–5. EDEN survey. J Invest Dermatol (2003) 120: 29. Temple R. Current definitions of phases of inves- 738–41. tigation and the role of the FDA in the con- 16. Van de Kerkhof PCM, De Hoop D, De Korte J, duct of clinical trials. Am Heart J (2000) 139: Cobelens SA, Kuipers MV. Patient compliance S133–5. and disease management in the treatment of 30. Krueger GG. Psoriasis therapy–observational or psoriasis in the Netherlands. Dermatology (2000) 200: 292–8. rational? New Engl J Med (1993) 328: 1845–6. 17. Day S. Commentary: Treatment allocation by the 31. Gottlieb AB. Product development for psoria- method of minimisation. Br Med J (1999) 319: sis: Clinical challenges and opportunities. In: 947–8. Roenigk HH, Maibach HI, eds, Psoriasis, 3rd edn. 18. Boateng F, Sampson A, Schwab B. Statistical New York: Marcel Dekker (1998) 421–33. analysis of possible bias of clinical judgements 32. Naldi L, Carrel CF, Parazzini F, Cavalieri d’Oro due to observing an on-therapy marker variable. L, Cainelli T. Development of anthralin short- Stat Med (1996) 15: 1747–55. contact therapy in psoriasis: survey of pub- 19. Christiansen JV, Holm P, Moller R, Reymann F, lished clinical trials. Int J Dermatol (1992) 31: Schmidt H. Etretinate (Tigason) and betametha- 126–30. sone valerate (Celeston valerate) in the treatment 33. Louis TA, Lavori PW, Bailar JC, Polansky M. of psoriasis. A double-blind, randomized, multi- Crossover and self-controlled designs in clinical center trial. Dermatologica (1982) 165: 204–7. research. New Engl J Med (1984) 310: 24–31. DERMATOLOGY 231
34. Lambert MF, Wood J. Incorporating patient pref- a critical appraisal of their quality. J Clin Pharm erences into randomized trials. J Clin Epidemiol Ther (1998) 23: 391–8. (2000) 53: 163–6. 42. Kantor J, Margolis DJ. Efficacy and prognostic 35. Kessels AG, Cardynaals RL, Borger RL, Go MJ, value of simple wound measurements. Arch Der- Lambers JC, Knottnerus JA, Knipschild PG. The matol (1998) 134: 1571–4. effectiveness of the hair-restorer ‘Dabao’ in males 43. Lindholm C, Bjellerup M, Christensen OB, Zed- with alopecia androgenetica. A clinical experi- erfeldt B. Quality of life in chronic leg ulcer ment. J Clin Epidemiol (1991) 44: 439–47. patients. An assessment according to the Notting- 36. Peyri J. Ebastine in chronic urticaria: a double- ham Health Profile. Acta Dermato-Venereol (1993) blind placebo-controlled study. JDermatolTreat 73: 440–3. (1991) 2: 51–3. 44. Faust HB, Gonin R, Chuang TY, Lewis CW, 37. Gupta MA, Johnson AM, Gupta AK. The devel- Melfi CA, Farmer ER. Reliability testing of the opment of an Acne Quality of Life scale: reli- dermatology index of disease severity (DIDS). ability, validity, and relation to subjective acne An index for staging the severity of cutaneous severity in mild to moderate acne vulgaris. Acta inflammatory disease. Arch Dermatol (1997) 133: Dermato-Venereol (1998) 78: 451–6. 1443–8. 38. Finlay AY. Quality of life measurement in der- 45. Chren MM, Lasek RJ, Quinn LM, Mostow EN, matology: a practical guide. Br J Dermatol (1997) Zyzanski SJ. Skindex, a quality of life measure 136: 305–14. for patients with skin disease: reliability, validity, 39. Charman C, Williams H. Outcome measures of and responsiveness. J Invest Dermatol (1996) 107: disease severity in atopic eczema. Arch Dermatol 707–13. (2000) 136: 763–9. 46. Kragballe K, Gjertsen BT, De Hoop D, Karls- 40. Emerson RM, Charman CR, Williams HC. The mark T, van de Kerkhof PC, Larko O, Nieboer C, Nottingham Eczema Severity Score: preliminary Roed-Petersen J, Strand A, Tikjob G. Double- refinement of the Rajka and Langeland grading. blind, right/left comparison of calcipotriol and Br J Dermatol (2000) 142: 288–97. betamethasone valerate in treatment of psoriasis 41. Ashcroft DM, Li Wan Po A, Williams HC, Grif- vulgaris. Lancet (1991) 337: 193–6. fiths CEM. Quality of life measures in psoriasis: PSYCHIATRY
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 15 Overview
B.S. EVERITT Biostatistics and Computing Department, Institute of Psychiatry, King’s College, London, UK
The mentally ill have always been with us–to largely abandoned in favour of viewing the insane be feared, marvelled at, laughed at, pitied or as wild beasts who should be kept constantly in tortured, but all too seldom cured. fetters. Indeed according to Foucalt,2 ‘madness Alexander and Selesnick, The History of Psychi- borrowed its face from the mask of the beast’. In atry, 1967. (Allen & Unwin, Sydney, Australia) early medieval times beating, incarceration and restraint were the ‘treatments’ endured by the INTRODUCTION majority of the mentally ill. Insanity was almost universally regarded as a spiritual trial which one In his dictionary of psychology, the late Pro- had to undergo as a punishment for vice, a test fessor Stuart Sutherland defines psychiatry as of faith, or a method of purging sin–a form of ‘the medical speciality that deals with mental Purgatory on Earth–which could be dealt with disorders’. An almost equally brief definition only by spiritual remedies such as exorcism or appears in Campbell’s Psychiatric Dictionary, being locked up in a church overnight. Gradually namely, ‘the medical speciality concerned with other approaches to treatment were introduced the study, diagnosis, treatment and prevention of although most were equally harsh, bleeding, behaviour disorders’. In terms of either defini- vomiting and purging for mentally ill patients tion it would appear that psychiatry has a long were common, as were more whimsical forms history; Pythagoreans, for example, employed of treatment such as whirling or spinning a a form of music therapy with emotionally ill madman round on a pivot. These treatments were 1 patients, and Aretaeus (AD 50–130) observed in addition to the continued use of manacles and mentally ill patients and did careful follow-up chains for restraint. Apart from their harshness, studies on them. As a result, he established that what these treatments also had in common was manic and depressive states often occur in the that they were mostly ineffective. same individual and that lucid intervals generally It was not until the seventeenth century that exist between manic and depressive periods. the tide of opinion seems to have turned against But a thousand years on such a seemingly rough treatment. For example, on 18 July 1646 enlightened approach to the mentally ill had been the Court of Governors of Bethlem Hospital
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 236 TEXTBOOK OF CLINICAL TRIALS ordered ‘that no officer or servant shall give obscenities at anybody who came near. A scene of any blows or ill language to any of the mad human degradation. folks on pain of loosing his place’ and at the same hospital in 1677 the Governors propounded And, as we shall see in the next section, a rule that ‘No Officer or Servant shall beat several new twentieth century treatments were or abuse any Lunatik, nor offer any force to equally as harsh as those used 200 years earlier, them, but upon absolute, Necessity, for the better and in the main, almost equally ineffective in governing of them’. As a substitute for coercion, producing a cure. One positive change from some institutes housing the insane began to earlier times, however, was that now some offer kindness, attention to health, cleanliness clinicians began to take the first small steps to and comfort. Reformers such as John Monro evaluating treatments scientifically by making pioneered the introduction of ‘moral treatment’ qualitative and quantitative observations and which stressed the value of occupation to combat measurements. Empiricism was, at last, about to the dangers of idleness, and the need for patients play a role in psychiatric practice. to be dealt with tenderly and with affection. Such an approach was now considered to be PSYCHIATRIC TREATMENT AND ITS more likely to restore reason than harshness EVALUATION: THE EARLY TWENTIETH or severity. CENTURY But although there was an increasing desire for caring to replace constraint in dealing with In the 1920s Dr Henry A. Cotton3 proposed a the- the mentally disturbed, drugs such as corium, ory relating focal infection to mental disorders, in digitalis, antimony and chloral were still used particular the functional psychoses. According to to quieten patients, replacing physical fetters Dr Cotton: with pharmacological ones. And despite the best efforts of the advocates of the moral treatment The so called functional psychoses we believe approach, asylums housing the insane often today to be due to a combination of many remained depressing and degrading places until factors, but the most constant one is the intra- well into the twentieth century, as is illustrated cerebral, bio-chemical cellular disturbance arising by the following account of a visit by a newly from circulating toxins originating in chronic appointed psychiatrist in 1953 to the chronic foci of infection, situated anywhere in the body, ward of a mental hospital in Cambridge in the associated probably with secondary disturbance United Kingdom: of the endocrin system. Instead of considering the psychosis as a disease entity, it should be considered as a symptom, and often a terminal I was taken in by someone who had a key to unlock symptom of a long continued masked infection, the the door and lock it behind you. The crashing of toxaemia of which acts directly on the brain. keys in the lock was an essential part of asylum life then just as it is today in jail. This led into a big Dr Cotton identified infection of the teeth and bare room, overcrowded with people, with scrubbed tonsils as the most important foci to be consid- floors, bare wooden tables, benches screwed to the ered, but the stomach and in female patients, the floor, people milling around in shapeless clothing. cervix could also be sources of infection responsi- There was a smell in the air of urine, paraldehyde, ble, according to Dr Cotton’s theory, for the men- floor polish, boiled cabbage and carbolic soap – the asylum smell. Some wards were full of tousled, tal condition of the patient. The logical treatment apathetic people just sitting in a row because for for the mentally ill resulting from Dr Cotton’s twenty years the nurses had been saying ‘sit down, theory was surgical elimination of the chronically shut up’. Others were noisy. At the back of the infected tissue, all infected teeth and tonsils cer- ward were the padded cells, in which would be tainly and for many patients, colectomies. Addi- one or two patients, smeared with faeces, shouting tionally female patients might require enucleation OVERVIEW 237 of the cervix, or in some cases complete removal became popular in the 1930s and 1940s. Insulin of fallopian tubes and ovaries. Such treatment coma, for example, required patients to be given was, according to Dr Cotton, enormously suc- large doses of insulin which, by lowering the cessful with, out of 1400 patients treated, only blood sugar, induced a comatose state from which 42 needing to remain in hospital. they would be rescued by a large dose of glucose The focal infection theory of functional psy- (if they were amongst the lucky ones–some choses was not universally accepted, neither were patients died). According to Sargant and Slater,5 the striking results said to have been obtained (reproduced with permission of Elsevier Science) by the removal of these infections. So in 1922 ‘reliable statistics are mostly in favour of the Drs Kopeloff and Cheney4 of the New York State treatment’, although this claim needs to be Psychiatric Institute undertook a study to investi- considered alongside their recommendation as to gate Dr Cotton’s proposed treatment in the spirit how to select patients for treatment: of, in their own words: ‘an approach free from prejudice and without preconceived ideas as to It is rarely indeed that facilities will exist for the treatment by a full course of insulin of all the possible results’. To achieve this laudable schizophrenics coming under observation, and it is if somewhat pious aim, Kopeloff and Cheney therefore important not to waste the treatment on planned their study in the form of an experiment. patients not very likely to respond while denying it All the patients were divided into two groups to the favourable cases. as nearly identical as possible. All members of one group received operative treatment for foci Perhaps the most severe of the physical therapies of infection in teeth and tonsils, while members was a lobotomy, where the brain was cut with of the other group received no such treatment and a knife. The operation was pioneered by Egas consequently could be regarded as controls. No Moniz, a Lisbon neurologist, and later taken up doubt Kopeloff and Cheney’s study would have enthusiastically by psychiatrists such as William been hard pressed to have gained ethical approval Sargant of St. Thomas’s Hospital in the United today, but despite its ethical and probable sci- Kingdom. Evaluation of the effectiveness of entific limitations it did produce results (sum- the therapy was largely anecdotal, and even marised here in Table 15.1) that cast grave doubts an enthusiast such as Sargant knew that the over removal of focal infections as a treatment operation was often performed at a price: for some types of mental illness, and indirectly It is probable that the highest powers of the intellect at least, drove a nail into the coffin of Dr Cotton’s are affected detrimentally, and if the patient shows theory as to the cause of these conditions. little sign of this in his day-to-day behaviour it may Dr Cotton’s suggested treatment for patients be because the daily routine of existence makes with functional psychoses was severe, but not little call on his best powers. We recognise too that more so than other ‘physical therapies’ which temperamental qualities also are not unaffected, that
Table 15.1. Results from Kopeloff and Cheney’s study4 Demential praecox Manic depressive Controls Operated Controls Operated
Number of cases 15 17 15 9 Recovered – – 5 4 Improved 5 5 8 1 Total benefited 5 5 13 5 Unimproved 10 12 2 4 Left hospital 3 5 6 3 238 TEXTBOOK OF CLINICAL TRIALS
the reduction in self-criticism may lead to tactless an approach. That by Karagulla,9 for example, and inconsiderate behaviour, and that the more compared results for six groups of patients. Two immediate translation of thought and feeling into groups, men and women, had been treated at action can show itself in errors of judgement. The the Royal Edinburgh Hospital for Mental and damage, once done, is irreparable... Sargant and Slater5 Nervous Disorders in the years 1900–39 (before the advent of ECT). The other four groups Both insulin therapy and lobotomies were slowly had been treated in the years 1940–48, two phased out as treatments for the mentally ill, but (men and women) by ECT and two others (men another of the physical therapies introduced in and women) not using ECT. It requires little the mid-twentieth century, electric shock (ECT), imagination to suppose that the historical controls remains in use to this day largely because it has seen during the period 1900–39 are of little been found to be effective in a number of studies use in evaluating ECT; any difference between (see next section). This treatment, introduced the recovery rates for the periods 1900–39 by Cerletti and Bini, consists of producing and 1940–48 in favour of the latter could be convulsions in a patient by means of passing explained by many other factors than treatment an electric current through two electrodes placed with ECT. The differences between the ECT on the forehead. The idea that such convulsions groups and the concurrent controls are also might help the mentally ill patient was not new; virtually impossible to assess since the decision as long ago as 1798, for example, Weickhardt had to use ECT on a patient was a subjective one by the clinicians involved. There is no way of recommended the giving of camphor to the point of producing vertigo and epileptic fits. knowing whether the treated and untreated groups are comparable. ECT was (and is) used primarily in the treat- But the evaluation of treatments in medicine ment of patients with severe depression. Early in general and psychiatry in particular was about claims for its effectiveness bordered on the mirac- to be placed on a scientifically far firmer footing, ulous. Batt,6 for example, reported a recovery by the introduction and then the increasing use rate of 87%. Fitzgerald7 was only slightly less and acceptance of the controlled clinical trial. optimistic, suggesting the figure was 78%. In nei- ther report however was there any attempt to gather data on recovery rates in concurrent con- PSYCHIATRIC TREATMENTS AND THEIR trols. Despite this, other psychiatrists accepted EVALUATION: THE 1950s ONWARDS the quoted recovery rates as an indication of the effectiveness of ECT. Typical is the following quotation from Napier:8 Kopeloff and Cheney’s study of removal of focal infections as a treatment for particular forms of It is a remarkable advance that a type of case in mental illness has many aspects of a modern which the outlook was formerly so problematical clinical trial, although it is missing that most can now be offered with some confidence the essential component, random allocation. Kopeloff prospect of restoration in a matter of weeks. and Cheney decided themselves in regard to each (Reproduced with permission of the British patient whether they should be operated on or Journal of Psychiatry) whether they should be a control. It was Fisher who recognised the need for Some researchers attempted to evaluate ECT by randomisation to treatment groups in medical, comparing their results with those from historical biological and agricultural experiments, and the controls or from concurrent patients who for eventual adoption of the principle into the eval- one reason or another had not been offered the uation of treatments has led to what Sir David treatment of choice (ECT). But such studies Cox has called ‘the most important contribu- largely only illustrated the weaknesses of such tion of 20th Century statistics’, the randomised OVERVIEW 239 controlled clinical trial. In such trials patients differences, the use of randomisation represented are assigned to treatment groups according to a great improvement over earlier studies. chance. Prior to Fisher’s randomisation principle Other small trials of ECT were reported in the being adopted, most of the early studies to com- 1950s but it was not until 1965 that the MRC pare competing treatment for the same condition published the results of the first large-scale trial involved arbitrary, non-systematic schemes for of the treatment. This was a multicentre trial deciding which treatment a patient would receive. involving 55 clinicians and 269 patients. As well The first trial with a properly randomised control as demonstrating the effectiveness of ECT in group was that for streptomycin in the treatment the treatment of depression, the MRC trial also of pulmonary tuberculosis, carried out by Brad- showed that a large multicentre trial in psychiatry ford Hill in 1947. The first psychiatrist to advo- was feasible. The trial did, however, have some cate the use of Fisher’s experimental approach critics. In a letter to the British Medical Journal, in the evaluation of psychiatric treatments, par- Sargant12 wrote: ticularly the physical treatments, appears to have been Lewis.10 (Reproduced with permission from There is no psychiatric illness in which bedside Oxford University Press) In his paper he criticises knowledge and long clinical experience pays better past studies and of a controlled trial he concludes: dividends; and we are never going to learn how to treat depressions properly from double-blind sampling in an MRC statistician’s office. An organised experiment would demand much (Sargant W Antidepressant drugs. Br Med J. that has not hitherto been practicable, includ- (196) 1; 1495. Reproduced with permission from ing voluntary acceptance by independent hospi- the BMJ Publishing Group) tals and clinics of an agreed procedure for the selection, management, evaluation of mental state, and follow-up investigation of treated, as well as At the end of the 1940s and the beginning of of control cases. Such an experiment, as R.A. the 1950s, the physical treatments introduced into Fisher (1942) has demonstrated, requires much psychiatry 30 years earlier still formed the core of forethought and self-discipline on the part of those most psychiatrists’ treatment armoury. But mat- who carry it out. ters were about to change; in the 1950s several entirely new types of drugs were to be intro- For many psychiatrists practising at the time duced in psychiatric practice. In the main the this was not altogether welcome news. The discovery of these drugs was not based on a physical therapies, such as insulin coma and scientific knowledge of brain chemicals, rather psychosurgery remained in use, with advocates their discovery was for the most part serendip- of these treatments retaining their enthusiasm, ity, resulting from acute observations made by apparently untroubled by the usual requirements clinicians such as Henri Laborit (the effects of rational scientific scepticism. Demands that of the antihistamine promethazine, from which clinical trial methodology be adopted to evaluate developed chlorpromazine), and John Cade who treatments whose effectiveness most psychiatrists first described the value of lithium in manic already took for granted, fell largely on deaf ears. depression by observing its effect on a num- But slowly matters did improve. In the early ber of patients. The tricyclic antidepressants 1950s Miller and his colleagues randomly allo- and the Selective Serotonin Reuptake Inhibitors cated ten schizophrenic patients to each of (SSRIs), which had fewer side effects in treating three alternative treatments, ECT, Pentothal and depression, were also discovered in the 1950s. Pentothal plus non-convulsive stimulation, and Finally, almost by accident, Leo Sternback in assessed them before treatment began, after the 1957 identified the benzodiazepines for treating cessation of treatment, and then again two weeks mild anxiety. later.11 Although the sample size was totally inad- The need to establish whether or not these equate to demonstrate any significant treatment newly discovered compounds were effective 240 TEXTBOOK OF CLINICAL TRIALS in treating mentally disturbed patients greatly SUMMARY increased most psychiatrists’ appreciation of the need to use clinical trials for evaluating treat- Treatment of the mentally ill has made giant ments. And after 1960 the increasing need to strides in the last 50 years. Drug treatment of satisfy regulatory authorities (prior to 1960 only schizophrenia, depression and anxiety disorders the USA had such a body overseeing the intro- have, in randomised clinical trials, been found duction of new drugs into general use, but to be effective and have done much to alleviate the thalidomide tragedy changed the situation the misery of these conditions. Drug treatment dramatically) meant that the randomised con- of mental illness works by altering in some way trolled clinical trial has now become established the chemistry of the body. Chlorpromazine, for as the ‘gold standard’ for evaluating compet- example, has been shown to interfere with the ing therapies, although even as late as 1963 action of the neurotransmitter dopamine. But the some psychiatrists were unwilling to accept modern view of mental illness, that it has both that such an approach was necessary; this is psychological and physical dimensions, implies from the preface of a 1963 edition of Sar- 5 that effective treatment must aim to ease the suf- gant and Slater commenting on controlled clin- fering of the mind as well as correcting possi- ical trials: ble abnormalities of chemistry. And so, in the 1970s, behavioural psychotherapy began to be If they fail to demonstrate any differences between used to treat particular disorders. More recently a placebo and a drug which everybody knows to cognitive therapy has been introduced. This pro- be effective, this means only that the work has not vides a simple, straightforward treatment regi- been done well enough. men which lasts weeks rather than years, and above all permits the patients to make sense of, Fortunately it is difficult to imagine such a view and thus hopefully control, their psychological being expressed in a major psychiatric text nowa- problems. days. Over the last 40 years the use of clinical Clinical trials in psychiatry initially involved trials in psychiatry, particularly for evaluating the evaluation of drug treatments. More recently, new drugs, has become widespread. A quotation however, psychological therapies have been sub- from one of the psychiatric champions of this jected to the rigours of the randomised clini- approach, Michael Shepherd,13 remains almost cal trial, and there has been a growing aware- the perfect model for the modern scientific view ness that the theoretical and logistical prob- that psychiatrists should have in the evalua- lems of such trials differ from those of the tion of psychotropic drug therapies in particular, average drug trial. Consequently three of the and in the evaluation of psychiatric treatments chapters in this section concentrate on clinical in general: trials of psychological treatments as now used in psychiatry. In Chapter 17 Katherine Shear The clinician is compelled to hold the balance and Philip Lavori discuss the many problems between the scales of laboratory data on the one hand and stochastic theory on the other. associated with assessing treatments for anx- Though his experience and judgement are essential iety, in particular how interventions work in it will be necessary for him to adopt a more the community settings where they will even- experimental role in the future if he is to co-operate tually be used. In Chapter 18 Nicholas Tar- fully with the pharmacologist and the statistician rier and Til Wykes give a masterly overview whose techniques he should understand if full of how clinical trials have been used to eval- weight is to be given to observations made in the uate the effectiveness of cognitive behavioural clinical setting. treatments of psychosis. Finally in Chapter 19 (Shepherd, 1959, reproduced with permission) Graham Dunn considers the many issues that arise in applying clinical trial methodology to OVERVIEW 241 the use of psychotherapy for treating depression. 5. Sargant W, Slater E. An Introduction to Physical The difficulties of undertaking clinical trials Methods of Treatment in Psychiatry. Edinburgh: in this area are clearly identified in all the Livingstone (1944). 6. Batt JC. One hundred depressive psychoses treated papers, but these difficulties should be seen with electrically induced convulsions. J Mental Sci as a challenge to psychiatrist and statistician (1943) 89: 289–96. alike, in what must been seen as the long term 7. Fitzgerald OWS. Experiences in the treatment of goal of alleviating the misery that is mental depressive states by electrically induced convul- illness. sions. J Mental Sci (1943) 89: 73–80. 8. Napier FJ. Death from electrical convulsion ther- apy. J Mental Sci (1944) 90: 875–8. REFERENCES 9. Karagulla S. Evaluation of electrical convulsion therapy as compared with conservative methods of treatments of depressive states. J Mental Sci 1. Gordon BL. Medicine Throughout Antiquity. (1950) 96: 1060–91. Philadelphia: Davies (1949). 10. Lewis AJ. On the place of physical treatment in 2. Foucalt M. Madness and Civilization: A History psychiatry. Br Med Bull (1946) 3: 22–34. of Insanity in the Age of Reason. London: Tavis- 11. Miller DH, Chang J, Cumming E. A com- tock (1967) (transl. from the French by Richard parison between unidirectional current non- Howard; originally published as Histoire de la convulsive electrical stimulation given with Folie). Reiter’s machine, standard alternating current 3. Cotton HA. The etiology and treatment of the so- electron shock (Cerletti method) and pentothal called functional psychoses. Summary of results in chronic schizophrenia. Am J Psychiat (1953) based upon the experience of four years. Am J 189: 226–40. Psychiat (1922) 2: 156–91. 12. Sargant W. Antidepressant drugs. Br Med J (1965) 4. Kopeloff N, Cheney CO. Studies in focal infec- 1: 1495. tion: its presence and elimination in the func- 13. Shepherd M. Evaluation of drugs in the treatment tional psychoses. Am J Psychiat (1922) 2: of depression. Can Psychiat Assoc J (1959) 4: 139–56. 120–8. 16 Alzheimer’s Disease∗
LEON J. THAL AND RONALD G. THOMAS Department of Neurosciences, University of California San Diego, San Diego, CA 92037, USA
BACKGROUND years were considered to not have AD. Thus, AD was considered to be a rare disorder causing only HISTORY presenile dementia. Over the succeeding decades, advances were Descriptions of patients with dementia appear made largely based on pathological studies of in the Bible, Roman and Greek writings, and AD. Amyloid-like staining was noted by Dirvy.3 Shakespeare. However, it was not until the early Terry4 and Kidd5 described the ultrastructure portion of the 20th century that Alzheimer’s dis- of the neurofibrillary tangle as a paired helical ease (AD) was identified as a specific disease filament. In 1968, Blessed et al.6 carried out entity. The index case, Frau Auguste D, a 50- autopsies in an elderly cohort and found cerebral year-old woman, was admitted to the Frankfurt atrophy as well as plaques and tangles. These Insane Asylum in 1901 (see Table 16.1, time- authors concluded that these elderly subjects had line). She was found to be suffering from a pre- AD and that there was no difference between senile dementia with memory loss, generalised early-onset and late-onset AD. cognitive impairment, delusions and hallucina- At that time, the prevailing opinion in the tions. Upon her death in 1906, brain exam- United States was that dementia after age 65 ination revealed generalised cerebral atrophy was primarily due to cerebral atherosclerosis. It and two microscopic lesions called plaques and was not until the mid-1970s that physicians and tangles.1 The disease was subsequently named scientists in the US recognised that most elderly ‘Alzheimer’s disease’ by Kraepelin.2 However, individuals with dementia suffered from AD. 7 Kraepelin, the dominant figure of the day, defined Katzman noted the high prevalence of dementia AD as a presenile dementia occurring between and pointed out that AD was the fourth most the ages of 40 and 60 so that individuals who common cause of death in the US. At the same had similar clinical presentations in their later time, modern neurochemical techniques were applied to the study of AD brain tissue. Three ∗ Supported by Alzheimer’s Disease Cooperative Study Grant laboratories independently described the loss of #AGO 10483. cholinergic markers in AD.8–10 At the same time,
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 244 TEXTBOOK OF CLINICAL TRIALS
Table 16.1. AD timeline
1907 Index case, Frau Auguste D described by Alzheimer 1910 Kraepelin calls AD a rare presenile dementia 1927 Dirvy identified amyloid in the plaque 1963 Terry and Kidd describe the ultrastructure of the neurofibrillary tangle 1975 NIA established–AD is its primary focus 1976/77 Selective loss of cholinergic markers reported in AD 1976 Katzman reports that AD is the fourth leading cause of death in the US 1980 Alzheimer’s Association formed 1984 Glenner sequences amyloid and postulates that it causes AD 1984 NIA initiated Alzheimer Centers programmes 1987 First gene for AD on chromosome 21 1991 Alzheimer’s Disease Cooperative Study funded by NIA 1992 Presenilin 1 mutation on chromosome 14 for early-onset AD 1993 Tetrahydroaminoacridine (Cognex), first cholinesterase inhibitor is marketed 1993 Apo E4 allele identified as a risk factor for late-onset AD 1995 Presenilin 2 mutation on chromosome 1 for early-onset AD 1996 Donepezil (Aricept), second AChE inhibitor, marketed 1997 Vitamin E and selegiline delay time to endpoints in AD including progression of disease 2000 Rivastigmine (Exelon), third AChE inhibitor, marketed
the National Institute on Aging (NIA) was formed this same time, two additional genes coding for in 1975 and its first director, Dr Robert Butler, presenilin 1, located on chromosome 14, and pre- identified AD as the most important disease senilin 2, located on chromosome 1, were found of aging. In 1980, the Alzheimer’s Association to be responsible for additional families with was formed to represent public interest in this early-onset AD.14,15 Additional cholinesterase disorder. It has subsequently grown to more than inhibitors were approved in 1996 and 2000. In 250 national chapters. 1997, the first trial demonstrating a delay in time Initially, the NIA focused on the funding to endpoints was published using the antioxidant, of individual investigator grants and some pro- vitamin E, and the monoamine oxidase inhibitor, gramme project grants. Developments occurred selegiline.16 rapidly, and in 1984, Dr George Glenner iso- By 1984, the NIA recognised the need for lated amyloid from the blood vessel wall of the study of well-characterised brain tissue for Alzheimer tissue and postulated that amyloid patients with AD. It initiated the Alzheimer’s was the causative agent of AD.11 By 1987, the disease centres programme by funding five cen- first gene for early-onset familial AD, which tres in 1984 with gradual growth of the cen- later was found to be the β-amyloid precur- tres’ programme to 30 by 2000. In 1991, the sor protein gene, was linked to chromosome NIA also funded the Alzheimer’s Disease Coop- 21.12 In 1993, an association was reported erative Study (ADCS) with a mandate to both between the presence of the E4 allelic variant of develop new instrumentation for the testing of apolipoprotein E and the development of AD.13 patients with AD and to carry out clinical tri- This was the first genetic risk factor described als for promising agents that would not be tested for late-onset AD. In the same year, the first by the pharmaceutical industry, such as com- acetylcholinesterase (AChE) inhibitor, tetrahy- pounds lacking patent protection, compounds in droaminoacridine (Cognex) was approved and the public domain for the treatment of other marketed. This drug was developed based on the disorders, and molecules derived from individ- empirical observation of a decrease in choliner- ual investigator laboratories or small biotechnol- gic functioning originally made in 1976. During ogy companies. ALZHEIMER’S DISEASE 245
Over the past four decades, interest in AD Epidemiological studies have also identified has markedly increased. Funding has increased the use of non-steroidal anti-inflammatory drugs to approximately $400 million per year by the (NSAIDs) and oestrogens as potential protective Federal government. Citations in Index Medicus factors (see Kawas and Katzman17 for a review). have increased from fewer than 100 in 1970 to These epidemiologic observations plus studies more than 3000 per year by 2000. on the basic biological mechanisms of NSAIDs and oestrogens led to the development of several EPIDEMIOLOGY clinical drug trials designed to determine whether or not NSAIDs or oestrogens can either slow the Epidemiological studies of AD have been extre- rate of decline in patients with established AD or mely useful in defining prevalence, incidence prevent AD in normal elderly. and risk factors. Numerous prevalence studies Epidemiological studies have also been useful have been carried out in the US, Europe and in determining that exposure to aluminium and China. Overall, these studies indicate that the cerebrovascular disease are not associated with prevalence of AD is under 1% below the the development of AD. age of 65. After age 65, the prevalence rises rapidly, doubling with every five-year epoch. The GENETICS prevalence exceeds 30% by age 90 (see Kawas and Katzman17 for a review). Genetic inheritance of AD and familial clustering Epidemiological studies have also been useful was first reported in the 1950s21 and 1960s,22 in defining both risk and protective factors then confirmed in a pathological series in the for the development of AD (Table 16.2). The 1970s.23 Inheritance was complex but appeared most important risk factor is age. A second to be autosomal dominant in a small proportion important risk factor is family history. This was of cases, inherited as a complex trait in up recognised by Heston et al.18 who noted that to 50% of the population and sporadic in the the cumulative risk for AD was significantly remainder. Since these initial studies were carried higher for parents and siblings of patients than for out, inheritance for early-onset AD has been the general population as a whole. This finding confirmed in patients carrying mutations for APP, has been subsequently confirmed by many other presenilin 1 and presenilin 2. In addition, late- 19,20 individuals. The discovery of early-onset AD onset AD has been linked to the apo E, E4 families further strengthened the importance of allele as a risk factor gene. Thus, at present, genetics as a risk factor. Finally, the association approximately one-half of known AD cases may of the apo E, E4 allele and AD represents a have some underlying genetic component. genetic risk for late-onset AD. In addition, a number of more minor risk factors have been PATHOLOGY identified including female sex, head injury and low education. The macroscopic pathology of AD consists primarily of cerebral atrophy and dilatation of the ventricles. The hallmarks of the disease are Table 16.2. Risk factors for AD two microscopic lesions. The first, neurofibrillary Major risk factors Protective factors tangles, are intraneuronal paired helical filaments Age Higher education consisting predominantly of hyperphosphorylated Positive family history NSAIDs tau protein. The second is the senile plaque, a AD genes Oestrogens pathological inclusion in the neuropil consisting Minor risk factors Non-risk factors of damaged neuritic endings, compact and diffuse Low education Aluminium amyloid, and other proteins. The number of Head trauma Vascular disease Female sex both plaques and tangles correlates moderately well with the degree of dementia. There is 246 TEXTBOOK OF CLINICAL TRIALS also a significant neuronal loss and loss of metabolic-enhancing agents known as nootrop- synapses. The loss of synapses correlates best ics. None of these agents were proven to be with the severity of dementia (see Terry et al.24 efficacious for the treatment of AD, although the for a review). use of ergoloid alkaloids, a class of agents with both cerebral metabolic-enhancing and vasodi- PATHOGENESIS AND AETIOLOGY lating activity, did produce a minor degree of improvement on subjective rating scales. With At the present time, the cause of AD remains the demonstration of a cholinergic deficiency in unknown. However, the best hypothesis at AD, development of these compounds in the US present is that misprocessing of amyloid and its largely ceased. deposition in the brain results in nerve cell dam- age followed by loss of neurons and the demen- GENERAL ISSUES IN AD CLINICAL TRIALS tia syndrome. Data supporting this hypothesis is derived from studies of early-onset familial AD families. Individuals carrying obligate genes for DIFFICULTY OF DIAGNOSIS AD on chromosomes 1, 14 and 21 all develop There are many problems associated with the AD at an early age. In addition, fibroblasts trans- development of drugs and the conduct of clin- fected with DNA from these individuals either ical trials in AD. At present, there is no bio- produce excess amounts of both a-beta 1-40 and logical marker for the disease during life. Thus, 1-42 or undergo a shift in the ratio of a-beta pro- clinicians are never 100% certain of the diag- duction favouring the production of more a-beta nosis. However, recent clinicopathological series 1-42. This shift is important since a-beta 1-42 is reveal that the diagnostic accuracy for AD highly insoluble and represents the predominant now generally exceeds 80% and approaches form of a-beta present in the senile plaque. Indi- 90%, especially for cases selected for clinical viduals carrying one or two apo E4 alleles also drug trials.26–29 More recently, approximately appear to deposit more a-beta in their brains than 15–20% of AD patients have also been found individuals lacking an E4 allele. Finally, individ- to have extrapyramidal features. Examination of uals with Down’s syndrome carry an extra copy their brains reveals the presence of neocortical of chromosome 21, have three genes for amy- and subcortical Lewy bodies as well as abundant loid, and overproduce this protein. All develop plaques but few tangles indicating the presence the pathological features of AD if they live long of dementia associated with Lewy bodies.30–32 enough. Thus, the deposition of amyloid appears Thus, in any contemporary trial, approximately to be the central feature in the pathogenesis of 10% of individuals will turn out to have another AD. However, this hypothesis awaits formal test- disease at autopsy and approximately 20% will 25 ing (see Selkoe for a review). have Lewy bodies accompanying the neuropatho- logical changes of AD. HISTORY OF CLINICAL AD TRIALS PRIOR TO 1976 PATIENTS Patients with AD always have memory impair- Prior to the discovery of a cholinergic defi- ment as the core feature. However, many other ciency in the brains of patients with AD, drugs features may be present such as difficulties chosen for clinical testing in AD were chosen with language, praxis, visuospatial relations and based on the premise that cerebrovascular insuf- behaviour. There is substantial heterogeneity in ficiency caused AD. Thus, numerous therapeu- the clinical presentation of the patient popula- tic modalities were tried including: vasodilators, tion. In addition, patients decline at varying rates anticoagulants, hyperbaric oxygen and cerebral over time which leads to increased variability in ALZHEIMER’S DISEASE 247 response measures. These differences in patient of several prespecified endpoints or who termi- population characteristics and change over time nates from the study provides useful data. Drop- are largely responsible for the need to include outs for advancing disease in a longitudinal study reasonably large samples in AD clinical trials. are less of a statistical problem because infor- mation concerning whether or not that individual ENDPOINTS reached an endpoint may be available. Third, the use of survival analysis allows patients who reach The endpoints studied in AD clinical trials an endpoint (usually the diagnosis of AD) to exit depend primarily on the question being asked in the study and seek alternative treatments with- the trial. Early trials of cholinesterase inhibitors out impacting the statistical analysis. This feature were designed to detect treatment–placebo differ- may potentially enhance recruitment for long- ences in cognition over relatively short periods term, placebo-controlled survival trials. Fourth, of time. For these trials, the primary endpoints survival analysis allows for comparison of the consisted of a cognitive measure to determine entire group despite varying lengths of follow-up, the specificity of the agent on important cogni- i.e., no imputation. Fifth, it is usually more infor- tive endpoints and a clinical global impression mative unless the incidence is low. The potential to make certain that the overall effect was suf- disadvantage of survival analysis in AD trials is ficiently robust to be clinically significant. Trials that the time to reach certain endpoints (such as examining agents designed to alter the rate of institutionalisation) is likely to be more variable decline have generally used a difference in slope and affected by social support systems than the or a difference at endpoint in cognitive and global rate of change on a cognitive measure. Also, if measures. One recent trial used the time to devel- large numbers of patients drop out of the study opment of functional endpoints such as insti- without reaching the defined study endpoint, the tutionalisation, death, loss of activities of daily validity of the study may be open to question. living (ADL) or progression to a more severe Some examples of useful endpoints or milestones stage of dementia as a composite endpoint.16 in AD patients include death, institutionalisation, Finally, in primary and secondary prevention tri- loss of basic ADLs, loss of instrumental ADLs als where enrolled patients are either normal or and decline in the disease stage. mildly cognitively impaired upon enrollment, the time from enrollment to diagnosis of AD is often used as the endpoint. PHASE 1 TRIALS
SURVIVAL ANALYSIS IN AD Phase 1 trials for AD are carried out to deter- mine the general tolerability of the agent and While the use of endpoint differences and maximum tolerated dose. These trials commonly changes in the rate of decline are currently the utilise fewer than 100 subjects exposed to drug. most frequent approach being used in many AD Early phase 1 trials are generally single dose studies, the use of survival analysis is becom- exposure carried out in normal elderly sub- ing increasingly frequent. There are a number of jects to determine the maximum tolerated dose. inherent advantages to survival analysis in AD. Subsequently, the tolerability of multiple daily First, endpoints can be real-life events rather than doses is evaluated in brief trials lasting for artificial constructs such as the amount of change one to two weeks. More recently, trials involv- on a psychometric test. Events such as death and ing multiple daily dosing have been carried out institutionalisation require little interpretation and in early AD patients rather than normal con- clearly possess face validity. Second, survival trols in so-called ‘bridging studies’. The advan- analysis naturally allows the combination of mul- tage of this approach is that if the metabolism tiple endpoints; also, any patient who reaches one of the drug differs between AD patients and 248 TEXTBOOK OF CLINICAL TRIALS healthy normal controls, the doses tolerated by Few phase 2 trials are designed to examine the AD patients will be found early in the drug ability of the agent to slow decline in AD. Such development process. Early phase 1 studies focus efficacy-oriented studies are infrequently carried on tolerability, side effects and pharmacokinet- out because of the need for a large sample size ics. Subsequent phase 1 studies are often carried and long duration. In general, studies designed to out to look for food interactions and interac- slow decline are carried out in phase 3 clinical tions with other commonly used pharmaceuti- trials. A few phase 2/3 trials have been carried cal products. out in an attempt to look for both cognitive and global improvement and for alteration in the rate of decline over the course of one year. For PHASE 2 TRIALS one-year trials designed to slow decline in AD, using a typical outcome measure in which the Phase 2 trials are classically designed to explore standard deviation of the rate of change is equal the dose range of an agent and to establish an to the one-year decline, a typical study using initial determination of efficacy. They generally 80% power and two-sided testing with an alpha utilise 100–500 subjects. Due to the time and (type 1 error) of 5% would require 63 subjects cost involved in the drug development process, per group (assuming no drop-outs) comparing many sponsors are currently carrying out com- drug to placebo for significance to detect a 50% bined phase 2/3 studies. Most phase 2 studies decrease in rate of decline (Table 16.3). Most are carried out as multi-arm, parallel, placebo- studies are powered to detect 25–40% decreases controlled trials. The maximum dose used in such in rate of decline and therefore require larger a trial is approximately one-half to two-thirds sample sizes. of the maximum tolerated dose determined dur- ing phase 1 testing. Two, three or four doses are generally employed and compared to placebo. In PHASE 3 TRIALS some trial designs, an arm of an already-approved agent may be added as a positive control. Most Many AD trials are currently carried out as com- phase 2 trials designed for symptomatic treatment bined phase 2/3 studies. Depending on the num- are approximately six months in duration in order ber of arms, these trials generally utilise 300–600 to meet both European and US regulatory guide- subjects per trial. A complete phase 3 pro- lines. Endpoints in these trials are generally treat- gramme (for an FDA new drug application) will ment–placebo differences on a cognitive scale, include at least two pivotal trials and will utilise most commonly the Alzheimer’s Disease Assess- 1000–3000 subjects to both test efficacy and to ment Scale–Cognitive Component (ADAS-Cog), determine the side-effect profile of the agent. In and a Clinician’s Global Impression (CGI). recent years, several different trial designs have The number of subjects needed to demonstrate been utilised. For short-term trials designed to efficacy in clinical trials with continuous response improve cognition, three trial designs have been measures depends on the relationship between the used: crossover designs, randomised control par- effect size sought, the standard deviation of the allel designs (RCPD) and enrichment designs outcome measure, and other parameters such as type 1 error, type 2 error, drop-out rate, drop- in rate, and base rate for the control group. For Table 16.3. Same size calculations for AD slope trial example, for a treatment trial seeking to detect Reduction in rate Subjects per Total sample a four-point difference on the ADAS-Cog at six of decline (%) group (N) size (N) months assuming a standard deviation of nine, 100 subjects per group would be required in a 25 251 502 50 63 126 two-arm trial with a power of 90% and an alpha 75 28 56 (type 1 error) of 5%. ALZHEIMER’S DISEASE 249
(a notable variant of RCPD). The use of the used neuropsychological instruments, the annual crossover design presents a number of problems rate of change is approximately equal to the one- for the study of AD patients. This design assumes year standard deviation of change. Thus, for the that there are no carry-over or period effects of ADAS-Cog, the rate of change is approximately the drug and that the treatment response is the 7.0 + / − 7.8 points per year.34 Examination of same in both periods. This is rarely the case. The rate of change studies indicates that the rate of major advantage of this design is in the economy decline on common instruments is reasonably of subjects because each subject acts as its own constant during the middle stages of dementia but control. However, because AD is not a static dis- is slower in early and more severe dementia.35 order but the change is over time, carry-over and The rate of change is predictable for groups but period effects are often present and this design quite variable for individual patients. Rate of is now rarely used. Randomised, controlled, par- decline is likely to be influenced by the distribu- allel design studies have two major advantages: tion of disease subtypes such as the presence of the control population is concurrent (thus any the Lewy body variant of AD.36 The gene dosage period effects are balanced) and uncontaminated of apo E4 does not appear to significantly alter by drug exposure and there are no period effects. the rate of decline over one year.37 Knowledge The disadvantage of this design is that more sub- of the rate of decline allows for the accurate com- jects are required to answer the central question. putation of sample sizes for studies designed to An enrichment design combines features of both slow cognitive decline in AD. Most contempo- crossover and parallel designs. It was used in sev- rary phase 3 studies examining rate of decline in eral cholinesterase inhibitor trials including one AD are designed to detect group differences of of the US multicentre tacrine trials.33 Patients approximately 25–35% since most clinical inves- demonstrating improvement after initial tacrine tigators and AD advocacy groups believe that treatment were withdrawn from drug, washed out, finding smaller effect sizes would not result in and subsequently randomised to drug or placebo clinically useful drugs. and analysed for response in a double-blind par- In addition to trials designed to slow decline allel phase. Non-responders were dropped from in AD, one trial of vitamin E and selegiline16 the trial prior to the double-blind phase, thereby utilised survival analysis in a 2 × 2 factorial resulting in enrollment of an enriched popula- design to examine the time to important endpoints tion in the double-blind phase. While this design in patients with AD. One advantage of the 2 × has many advantages including individual dose 2 factorial design is that two agents can be titration for each patient, its major disadvan- tested simultaneously. In addition, interactions tage is that all subjects are exposed to drug at between the two agents can be examined. A some point during the study allowing for possi- third advantage of this design is that 75% of ble carry-over effects. Also, there is no placebo- subjects are randomised to treatment thereby treated population throughout the study on which enhancing enrollment. The disadvantage of the to base an estimate of adverse events in the con- 2 × 2 factorial design is that negative interactions trol population. Thus, parallel, placebo-controlled (sub-additivity) can occur thereby reducing the designs currently predominate in phase 3 symp- 2 × 2 factorial design to a four-arm study which tomatic studies. These studies are similar to results in a significant loss of power in the the description of phase 2/3 symptomatic studies analysis. In the 2 × 2 factorial design study of described above. selegiline and vitamin E, the primary endpoint Slowing decline in AD remains an important was the time to reach any of the following: goal since if slowing can actually be accom- death, institutionalisation, loss of basic ADLs and plished, the treatment effect size will continue to progression from moderate to severe dementia. increase with the duration of treatment. Natural Although this study was useful in determining history studies indicate that for most commonly that treatment could cause a delay in the time to 250 TEXTBOOK OF CLINICAL TRIALS these endpoints, this trial design could not resolve Table 16.4. Age-specific incidence of AD the issue as to whether or not the delay in the time to endpoints resulted as a consequence of drug Age Incidence (%) Cases/1000/yr treatment alone or a change in brain structure 65–69 0.6 6 as a consequence of the potential neuroprotective 70–74 1.0 10 effect of these two agents. 75–79 2.0 20 In an attempt to intervene earlier in the 80–84 3.3 33 course of the disease, individuals with mild 85–89 4.0 40 cognitive impairment (MCI) are currently being studied. Patients with MCI generally present with a memory complaint. When evaluated, they Thus, several strategies can be considered for demonstrate the poor recall and rapid forgetting primary prevention trials. In the first strategy, but are otherwise generally normal with respect healthy individuals would be treated in an attempt to cognitive functioning and ADLs. However, to delay the onset of disease. The main advantage patients identified with MCI convert to AD at a of this design is that the results would be gener- rate of 12–15% per year. In contrast, unselected alisable to other healthy individuals. Because of normal subjects over the age of 65 develop AD the large sample size and high costs associated at a rate of only 1% per year.38 This high rate of with this strategy, enrichment strategies should conversion has allowed for the development of also be considered. These include enrolling: older clinical trials of a reasonable size using the MCI subjects, subjects with a positive family history patient population. The endpoint of such clinical of AD, and subjects at risk because of the pres- trials is the time to development of clinically ence of an apo E4 allele. Several primary pre- diagnosable AD analysed with survival analysis. vention trials for AD are currently preparing to Several such trials have been initiated recently get underway utilising some of these strategies. examining the effects of vitamin E, cholinesterase A second strategy would be to find subjects who inhibitors and NSAIDs to delay the time to the are already randomised to compounds of interest development of AD in patients with MCI. in trials for other indications to which cognitive endpoints could be added. This has already been successfully accomplished within the framework PRIMARY PREVENTION TRIALS of the Women’s Health Initiative (WHI) where approximately 8000 non-demented women over Numerous studies have indicated that AD is the age of 65 have been enrolled and randomised uncommon before the age of 65. However, to hormone replacement therapy (HRT) to deter- after the age of 65, the prevalence doubles mine the effects of these agents on coronary approximately every five years reaching an artery disease and osteoporotic fractures. Cog- average prevalence of over 30% by 90 years nitive endpoints have now been added to this of age.39,40 Given that the prevalence doubles trial to determine the effect of HRT on incidence every five years, delaying the onset of appearance of dementia. A third approach would be to ran- of the disease by five years would result in a domise to drug treatment a non-demented popula- 50% reduction of prevalence in one generation. tion being studied for another purpose such as an Delaying onset by 10 years would halve the epidemiological study of cardiovascular disease. prevalence again for a total reduction of 75%. If individuals are entered into a study entirely Thus, primary prevention would yield the greatest on the basis of age, a large sample size will cost benefit. be required using population-based, non-enriched The earliest pathological changes of AD may entry criteria. For example, for individuals at occur 20–30 years before the clinical expres- a mean age of 75 who have normal cognition sion of disease.41 Age-specific incidence of AD on initial evaluation, dementia will appear at a increases with aging (Table 16.4). rate of approximately 1.5% per annum. Over ALZHEIMER’S DISEASE 251
five years, the cumulative incidence of dementia clinical symptoms of disease appear. This will be would be 7.5%. For a two-arm study with the ultimate goal in treating AD. alpha = 0.05 and power = 80%, 2000 subjects per group would be required to detect a 30% SYMPTOMATIC VS STRUCTURAL CHANGE effect size, not accounting for losses secondary to death. With such low incidence, even these large Treatments that slow progression in AD may samples will yield only 150 cases of dementia produce underlying structural change within the in the untreated group. If a drug reduced the brain. One could logically reason that if treatment incidence of AD by 50%, 75 cases would exist in caused slowing of the rate of decline, it may the treated group. To allow for losses to follow-up have a permanent underlying effect on the brain. and death, approximately 2500 subjects would Unfortunately, a change in slope in a measure be needed per group for a total sample size of of cognition is insufficient evidence to prove 5000. These sample sizes would only allow for that an underlying brain structural effect has an 80% probability of detecting a 30% increase occurred. Additional trial design manoeuvres in disease incidence. must be utilised. The most widely utilised Several primary prevention trials to prevent manoeuvre is that of withdrawal. If the effect dementia or AD have been initiated with most induced by the drug is purely symptomatic, even utilising an enrichment strategy. though individuals are better at the end of the trial, they should decay back to the curve of 1. Women’s Health Initiation Memory Study untreated subjects when the drug is withdrawn (WHIMS)–8000 normal women >65 years of (Figure 16.1). This was clearly demonstrated age randomised to HRT or placebo. in the 26-week Aricept trial in which patients 2. Preventing postmenopausal memory loss and were removed from Aricept and cognitive scores Alzheimer’s disease with Replacement Estro- dropped to placebo levels after a six-week gens (PREPARE) study–900 normal elderly wash-out.42 An alternative clinical manoeuvre women with a family history of AD ran- to demonstrate the same effect would be a domised to HRT or placebo. randomised start design. In this experiment, half 3. Gingko study–3000 normal elderly subjects the group is started on drug and half started on (age >75) treated with Ginkgo or placebo. placebo. If the drug has a purely symptomatic 4. AD Anti-Inflammatory Prevention Trial (AD- effect, individuals begun on placebo then crossed APT)–2600 normal elderly subjects with a over to active drug will ‘catch up’ to those started positive family history of AD randomised to on drug at the beginning of the trial. If the drug one of two anti-inflammatory drugs or placebo. has a structural effect on the brain, individuals started on drug later in the course of the treatment MISCELLANEOUS ISSUES period will fail to ‘catch up’. In addition to various trial manoeuvres, the PRIMARY, SECONDARY AND TERTIARY demonstration of a structural change within the TREATMENTS brain could be used to support a structural effect for a therapeutic agent. For example, in a rate Treatment of existing symptoms represents ter- of decline trial, the maintenance of hippocampal tiary treatment and is representative of all of volume or the maintenance of a synaptic num- the currently approved drugs for AD. Secondary ber, as demonstrated on an imaging study, would treatment refers to treatment when minimal but serve as a direct demonstration that a pharma- not full-blown disease is present. This is best cological agent produced a difference in brain exemplified by the treatment of patients with MCI structure. Finally, a biochemical marker could be who have minimal symptomatology. Finally, pri- utilised. For example, in Parkinson’s disease if mary prevention refers to intervention before any a ‘dopaminergic sparing’ agent were developed, 252 TEXTBOOK OF CLINICAL TRIALS
Randomized withdrawal design Randomized start design
Randomized Placebo Active phase phase Randomized phase phase Active Disease-modifying Active
effect Performance
Placebo Performance Placebo Symptomatic Symptomatic effect effect Disease-modifying effect
Time Time (a) (b)
Figure 16.1. Randomised start and randomised withdrawal it might be reflected in higher homovanillic acid drug and looking for improvement. In reality, this levels in the cerebrospinal fluid after drug with- manoeuvre is often carried out in the clinic with drawal at the end of the treatment period. For AD, simple withdrawal of the agent, retesting, rein- a clear-cut biochemical marker does not yet exist. troduction and retesting but without the use of placebo. While still useful, the lack of a blinded N-OF-ONE DESIGN crossover to placebo limits the interpretation of the results of this manoeuvre. This manoeuvre is not particularly useful for drug development but is often used in the clinical set- ting to determine continued response to drug. ETHICAL ISSUES An example of an N-of-one design in an ide- alised setting would be to answer the question of At present, available drugs to treat AD are symp- whether or not a patient on a cognitive-enhancing tomatic. Several agents are currently approved agent, such as a cholinesterase inhibitor, is still and before enrolling patients in any clinical trial, responding to drug. This question is an impor- disclosure and discussion of these agents with the tant one since AD patients continue to decline patient and their caregiver is necessary. In addi- while on treatment. After one or two years on tion, vitamin E and selegiline have been reported treatment, the clinician is faced with the decision to delay the time to endpoints in moderately as to whether or not to continue treatment. In advanced AD patients. The results of the vitamin an idealised version, a patient could be blindly E study also need to be discussed with patients crossed over to placebo to examine for a with- before enrolling them in clinical trials since the drawal effect. If the patient was on a symp- use of both cholinesterase inhibitors and vitamin tomatic treatment and continuing to respond, a E is currently the standard of care in the US. decline in cognition should be observed follow- A vigorous debate has emerged in the US ing drug withdrawal. This manoeuvre could then regarding the ethics of placebo-controlled tri- be followed by blindly restarting the patient on als. Some have argued that placebo-controlled ALZHEIMER’S DISEASE 253 trials are unethical since they deny patients the and of senile change in the cerebral grey mat- use of approved agents while enrolled in the ter of elderly subjects. Br J Psychiat (1968) 114: trial. Individuals espousing this viewpoint claim 797–811. 7. Katzman R. The prevalence of malignancy of that drug development can continue using add- Alzheimer’s disease. Arch Neurol (1976) 33: on studies or by demonstrating equivalence of 217–18. a new agent to an existing agent. Others argue 8. Davies P, Maloney AJF. Selective loss of cen- that placebo-controlled trials are ethical since cur- tral cholinergic neurons in Alzheimer’s disease. rently available agents are entirely symptomatic Lancet (1976) 2: 1403. 9. Bowen DM, Smith CB, White P, Davison AN. and do not alter the underlying course of the dis- Neurotransmitter-related enzymes and indices of ease. In addition, patients are free to not enter hypoxia in senile dementia and other abiotrophies. placebo-controlled trials if they wish to receive Brain (1976) 99: 459–96. currently approved medications. There is con- 10. Perry EK, Perry RH, Blessed G, Tomlinson BE. Necropsy evidence of central cholinergic deficits cern that comparator studies may result in the in senile dementia. Lancet (1977) 1: 189. development and licensing of many marginal or 11. Glenner GG, Wong CW. Alzheimer’s disease: ini- ineffective agents since currently approved agents tial report of the purification and characteriza- are of marginal efficacy and may produce nega- tion of a novel cerebrovascular amyloid pro- tive results in some clinical trials. It is possible tein. Biochem Biophys Res Commun (1984) 120: 885–90. that if a new agent were compared to an approved 12. St. George-Hyslop PH, Tanzi RE, Polinsky PJ, agent and the true effect size in the approved et al. The genetic defect causing familial Alzhei- agent was minimal, the newly tested drug would mer’s disease maps on chromosome 21. Science be approved even though the effect size may (1987) 235: 885–90. not have been sufficient to warrant approval in 13. Corder EH, Saunders AM, Strittmatter W, et al. Gene dose of apolipoprotein E type 4 allele and a placebo-controlled trial. A major drawback of the risk of Alzheimer’s disease in late onset using active control groups is the requirement families. Science (1993) 261: 921–3. of larger sample sizes. At present, most new 14. Schellenberg GD, Bird TD, Wijsman EM, et al. agents being tested in the US are compared to Genetic linkage evidence for a familial Alzhei- placebo. If and when agents are developed that mer’s disease locus on chromosome 14. Science (1992) 258: 668–71. can clearly alter the course of the disease, long- 15. Levy-Lahad E, Wijsman EM, Nemens E, et al.A term, placebo-controlled trials will become both familial Alzheimer’s disease locus on chromo- unethical and socially unacceptable. some 1. Science (1995) 269: 970–3. 16. Sano M, Ernesto C, Thomas RG, et al. A con- trolled trial of selegiline, α-tocopherol, or both REFERENCES as treatment for Alzheimer’s disease. New Engl JMed(1997) 336: 1216–22. 17. Kawas CH, Katzman R. Epidemiology of demen- 1. Alzheimer A. Uber eine eigenartige Erkrankung tia and Alzheimer’s disease. In: Terry RD, Katz- der Hirnrinde. Allgem Z Psychiatr-Gerich Med man R, Bick KL, Sisodia SS, eds, Alzheimer Dis- (1907) 64: 146–8. ease. Philadelphia: Lippincott Williams & Wilkins 2. Kraepelin E. Psychiatrie: ein Lehrbuch fur Studi- (1999) 95–114. erende und Artze. Leipzig: Verlag v. Johann 18. Heston LL, Mastri AR, Anderson VE, White J. Ambrosius Barth (1910). Dementia of the Alzheimer type. Clinical genetics, 3. Dirvy P. Etude histochemique des plaques seniles. natural history, and associated conditions. Arch J Belge Neurol Psychiat (1927) 27: 643–57. Gen Psychiat (1982) 38: 1085–90. 4. Terry RD. Neurofibrillary tangles in Alzheimer’s 19. Heyman A, Wilkinson WE, Stafford JA, Helms disease. J Neuropathol Exp Neurol (1963) 22: MJ, Sigmon AH, Weinberg T. Alzheimer’s dis- 629–42. ease: a study of epidemiological aspects. Ann Neu- 5. Kidd M. Paired helical filaments in electron rol (1984) 15: 335–41. microscopy of Alzheimer’s disease. Nature (1963) 20. Amaducci LA, Fratiglioni L, Rocca WA, et al. 197: 192–3. Risk factors for clinically diagnosed Alzheimer’s 6. Blessed G, Tomlinson BE, Roth M. The associa- disease: a case–control study of an Italian popu- tion between quantitative measures of dementia lation. Neurology (1986) 36: 922–31. 254 TEXTBOOK OF CLINICAL TRIALS
21. Sjogren T, Sjogren H, Lindgren GH. Morbus Alz- 33. Davis K, Thal L, Gamzu E, et al. A double-blind, heimer and morbus Pic: a genetic, clinical, and placebo-controlled multicenter study of tacrine for patho-anatomical study. Acta Psychiat Neurol Alzheimer’s disease. New Engl J Med (1992) 327: Scand (1952) 8: 9–152. 1253–9. 22. Larsson T, Sjogren T, Jacobsen G. Senile demen- 34. Thal LJ, Carta A, Clarke WR, et al. A one-year tia: a clinical, sociomedical and genetic study. multicenter placebo-controlled study of acetyl-l- Acta Psychiat Scand (1963) 39(Suppl 167): 1–259. carnitine in patients with Alzheimer’s disease. 23. Heston LL. The genetics of Alzheimer’s disease. Neurology (1996) 47: 705–11. Arch Gen Psychiat (1977) 34: 976–81. 35. Stern RG, Mohs RC, Davidson M, et al. A lon- 24. Terry RD, Masliah E, Hansen LA. The neuro- gitudinal study of Alzheimer’s disease: mea- pathology of Alzheimer’s disease and the struc- surement, rate, and predictors of cognitive tural basis of its cognitive alterations. In: Terry deterioration. Am J Psychiat (1994) 151: 390–6. RD, Katzman R, Bick KL, Sisodia SS, eds, Alz- 36. Klauber MR, Hofstetter CR, Hill LR, Thal LJ. heimer Disease. Philadelphia: Lippincott Williams Patterns of decline in the Lewy body variant & Wilkins (1999) 188–203. of Alzheimer’s disease. Arch Psychiat Shanghai 25. Selkoe DJ. Molecular pathology of Alzheimer’s News (1992) 4(Suppl): 50–3. disease: the role of amyloid. In: Growdon JH, 37. Growdon JH, Locascio JJ, Corkin S, Gomez- Rossor MN, eds, The Dementias Boston: Butter- Isla T, Hyman BT. Apolipoprotein E genotype worth Heinemann (1998) 257–84. does not influence rates of cognitive decline 26. Wade JPH, Mirsen TR, Hachinski VC, Fish- in Alzheimer’s disease. Neurology (1996) 47: man M, Lau C, Merskey H. The clinical diagno- 444–8. sis of Alzheimer’s disease. Arch Neurol (1987) 44: 38. Petersen RC, Smith GE, Waring SC, Ivnik RJ, 24–9. Tangalos EC, Kokmen E. Mild cognitive impair- 27. Morris J, McKeel D, Fulling K, Torack R, Berg L. ment: clinical characterization and outcome. Arch Validation of clinical diagnostic criteria for Alz- Neurol (1999) 56: 303–8. heimer’s disease. Ann Neurol (1988) 24: 17–22. 39. Rocca WA, Hofman A, Brayne C, et al.Fre- 28. Burns A, Luthert P, Levy R, Jacoby R, Lantos P. quency and distribution of Alzheimer’s disease Accuracy of clinical diagnosis of Alzheimer’s in Europe. A collaborative study of 1960–1990 disease. Br Med J (1990) 301: 1026. prevalence findings. Ann Neurol (1991) 30: 29. Galasko D, Hansen L, Katzman R, Wiederholt W, 381–90. Masliah E, Terry R, Hill LR, Lessin P, Thal LJ. 40. Evans DA, Funkenstein HH, Albert MS, et al. Clinical–neuropathological correlations in Alz- Prevalence of Alzheimer’s disease in a community heimer’s disease and related dementias. Arch population of older persons. Am Med Assoc J Neurol (1994) 51: 888–95. (1989) 262: 2551–6. 30. Dickson D, Davies P, Mayeux R, et al.Diffuse 41. Braak H, Braak E. Diagnosis of Alzheimer’s dis- Lewy body disease. Acta Neuropathol (1987) 75: ease: development of cytoskeletal changes and 8–15. staging of Alzheimer related intraneuronal pathol- 31. Hansen L, Salmon D, Galasko D, et al.TheLewy ogy. Neurobiol Aging (1994) 15: S141. body variant of Alzheimer’s disease: a clinical and 42. Rogers SL, Farlow MR, Doody RS, Mohs R, Fri- pathological entity. Neurology (1990) 40: 1–8. edhoff LT. A 24-week, double-blind, placebo- 32. Perry R, Irving D, Blessed G, Fairbairn A, Per- controlled trial of donepezil in patients with ry E. Senile dementia of Lewy body type. JNeurol Alzheimer’s disease. Donepezil Study Group. Sci (1990) 95: 119–39. Neurology (1996) 50: 136–45. 17 Anxiety Disorders
M. KATHERINE SHEAR1 AND PHILIP W. LAVORI2 1Western Psychiatric Institute and Clinic, Pittsburgh, PA 15213, USA 2VA Medical Center, Menlo Park, CA 94025 2539, USA
INTRODUCTION specified and explained in manualised format. Treatment training and adherence measures are Anxiety disorders are the most prevalent psychi- available. These methodological advances mean atric conditions in the community with a life- that studies of the efficacy of new interventions 1 time community prevalence of 20–30%. These can be conducted efficiently and with confidence. disorders can be seriously impairing, reducing Given the availability of efficacious treatments, quality of life and causing disability. Recent researchers are now turning their attention to studies suggest some forms of anxiety are asso- studies that test these interventions in the commu- ciated with early mortality. Many who suffer nity settings where they will be used, and in clin- from anxiety disorders have other serious medical ical contexts (such as maintenance of response) problems, such as depression, pulmonary disease, that go beyond the phase of acute illness that is cardiovascular illness and neurological condi- the focus of most efficacy studies. With this shift tions. Prevalent and debilitating, anxiety disor- in focus, new methodological problems appear. ders are serious, persistent illnesses that warrant treatment. Clinical trials are needed to establish Generic problems that need to be addressed in efficacy of promising interventions and to deter- designing such studies, often known as effec- mine the best ways to deliver efficacious treat- tiveness studies, have been described in the 2 ments in different contexts. literature. In this chapter we discuss method- Methods for conducting efficacy trials in anx- ological issues pertaining to effectiveness stud- iety disorders have evolved over the past few ies of anxiety disorders. We identify some key decades. Reliable diagnostic instruments and features of these disorders and consider the prob- symptom severity scales have been developed lems they create for study questions and study and tested. Strategies for medication adminis- design. Solutions to methodological problems in tration have been identified and manuals writ- clinical trials often require trade-offs, and the ten to standardise these procedures. Cogni- problems we discuss are posed in this way. We tive behavioural treatment methods have been provide our view of the best way to manage these
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 256 TEXTBOOK OF CLINICAL TRIALS problems, and in some cases, make suggestions given question, is obvious. Given this ambigu- for methodological innovations. ity, we suggest anxiety disorder researchers might Clinical researchers regularly make method- be guided by some of the key features of these ological choices regarding subject recruitment, disorders (Table 17.1). We discuss the method- selection and characterisation of subjects, proce- ologic relevance of five such features: (1) anxiety dures for enrollment, assignment to experimen- disorders are characterised by high community tal group, experimental manipulation, outcome prevalence; (2) diagnostic boundaries are ambigu- assessment and follow-up process. Methods cho- ous, both between pathological and normal anx- sen will place specific limits upon what can be iety and among the different anxiety disorders; learned from a study. Thus, it is fundamental that (3) phobic fear and avoidance is prevalent in these study methodology be driven by the question the disorders; (4) anxiety disorders are treatable using researcher seeks to answer. However, unlike effi- either medication or cognitive–behavioural inter- cacy studies, in effectiveness studies, the question ventions; and (5) anxiety disorders frequently co- is not always clear. Defining the study question is occur with other disorders. Each of these features the first problem for the effectiveness researcher. will affect decisions about the research question Most experienced clinical researchers are expert and the choice of methods. in conducting efficacy studies to answer the ques- tion ‘Does a new treatment produce better results FEATURES OF ANXIETY DISORDERS THAT than a control condition for a well-defined con- IMPACT STUDY METHODOLOGY dition, under tightly controlled circumstances of use?’ Both psychosocial and pharmacological HIGH COMMUNITY PREVALENCE treatment researchers have successfully under- taken such studies, and thus are poised to test Anxiety disorders are highly prevalent in the efficacy hypotheses for new interventions. community. The high prevalence means there are The field of effectiveness research is far many patients in need of treatment. Epidemiolog- less developed. Investigators move forward in ical studies document that most of these patients unmarked terrain as they decide upon the most do not present for care in a specialty mental important next questions. For example, a study of setting.3 Instead, they can be found in a range of Long Term Strategies in the Treatment of Panic community service settings. Even among those Disorder (MH045963-6) currently in progress who do seek specialty mental health treatment, under the direction of the authors is designed to only a subset will be enticed to an academic med- answer the question ‘Should non-responders to ical clinic, regardless of the incentives provided. an initial trial of CBT receive medication or an For those who seek treatment at a community additional dose of CBT?’ This important ques- mental health setting, usual practice diagnostic tion is not addressed by efficacy studies of either procedures cannot be relied upon to identify anx- medication or psychotherapy. Having articulated iety disorders.4 It is clear that we need to know such a question, decisions must be made about how to recognise and treat the people with anx- what methodological approach should be used, iety disorders who most researchers never see. and what problems to anticipate. For example, Put another way, we need to study those who Principal Investigators of the Long Term Strate- do not participate in studies. This obvious para- gies Panic study had to confront the issue of dox underscores the principle that effectiveness what the right duration of the initial CBT trial studies will not be straightforward. might be, and what level of non-response to ini- The job is not simply a matter of running tial trial should be chosen to define intake into the an efficacy study in one or more community randomised maintenance trial. Decisions such as settings. Doing so would be important only if these are not trivial, since neither the most impor- there are serious questions about whether patients tant questions, nor the best way to approach a in such settings respond to proven treatments. ANXIETY DISORDERS 257
If this is the case, it is important to frame research will be conducted and in how many the specific questions the study should answer, different kinds of settings. based on the reasons for predicting response A different kind of research question might differences. For example, if researchers suspect be driven by the subject paradox (how to severity is an important treatment moderator, study patients who do not participate in stud- it might be important to conduct a standard ies), for example ‘What is the most success- randomised efficacy trial in settings with patients ful way of recruiting and engaging individuals of varying severity. Likewise, some patients have who do not seek treatment in a research clinic?’ co-occurring symptoms or syndromes, such as The investigator might compare a public edu- serious medical illness, along with an anxiety cation programme to a professional educational disorder such as panic disorder. It might make intervention. Or, the research aim might focus on sense to recruit patients from medical clinics into evaluating alternative screening strategies in dif- an efficacy trial in order to study the influence ferent settings. Another example might be ‘How of the medical illness on the treatment of the much diversity of setting should be incorporated target condition. Other examples of parameters into a study?’ In addressing this question the that might be predicted to render uncertain researcher might investigate the variation of clin- outcome with a proven efficacious treatment ical presentation, treatment acceptance, or out- include organisational features of the setting or come across different ethnic or socio-economic socio-economic status of the patient. Specific groups. Alternatively, the investigators might considerations like these ought to drive the examine the effect of different organisational important design decisions such as where the structures or the impact of the organisational
Table 17.1. Implications of features of anxiety disorders for research design Feature Research issues
1. High community prevalence Settings: which and how many Recruitment: reaching the unstudied patient Assessment: measuring outside the research clinic Awareness: bridging the patient’s knowledge gap Human subjects protection: make or buy Technology: the right machine for the setting Comorbidity: adjusting to increased variability 2. Poorly defined nosologic boundaries Normal vs disordered: a question of excess Differential diagnosis: core symptoms overlap Pluripotency: treatments with broad efficacy Double counting: symptoms Endpoints: ranking outcomes Aiming low: focus on preventing relapse Stability: the time frame for outcome 3. Phobic fear and avoidance Evasion: measuring the avoidant subject Fear: recruiting the anxious subject Identification: personal choice vs avoidance 4. Discordant models of the disorders Acknowledgement: both models have treatment successes Control: paying attention to the other intervention Comparing: accommodating preference for modality Targets: agreeing on the goals Dissemination: thinking ahead about the audience 258 TEXTBOOK OF CLINICAL TRIALS climate5 on outcomes or study the implementa- non-traditional research settings will tax the inge- tion of organisational interventions to optimise nuity of the next generation of effectiveness the likelihood of dissemination of a treatment.6 researchers. Recent work using adaptive testing Whatever studies are done, it is clear that for methods holds promise as a technique. Given anxiety disorders, researchers need to extend their these challenges, it is tempting to suggest that reach if they seek to make an impact on the great researchers concentrate on one research setting, majority of individuals who suffer from these and hope or assume that results will generalise. conditions. Methods need to be devised to study But the decision to limit the setting has uncer- patients in primary care and specialty medical tain implications for interpreting and generalising settings, dental offices, churches, schools, com- results. There is a trade-off between generalis- munity centres, and a range of other community ability and the cost of dealing with heterogeneity service or support settings (e.g. domestic vio- of setting. These costs must be borne, and the lence or homeless shelters, or even highly utilised methodological challenges met, in order to pro- commercial operations such as supermarkets7 or duce research-grade answers to the question of department stores). The use of such settings to effectiveness. deliver care may be particularly relevant for In working in almost any non-mental health patients with anxiety disorders who have pho- community setting, the investigator must address bic restrictions, and are unable to travel outside stigma and self-criticism that can be associ- a restricted area. ated with the idea of having a mental disorder. Designing a new clinical trial for an anxi- Researchers need to take steps to minimise the ety disorder outside of an established research difficulties that may be caused by identifying a centre raises other problems. Investigators make person as ill, especially when the person in ques- deliberate decisions about where to recruit, assess tion has not already identified their symptoms as and treat patients, as well as whether to carry problematic. In such a situation, the news may out any of these activities in more than one come as an unwelcome surprise, or may be per- kind of setting. Existing clinical research meth- ceived as insulting or embarrassing. The newly ods for recognising and recruiting affected indi- diagnosed individual may feel suddenly stigma- viduals may be too cumbersome to work in a tised and this may lead to a rejection of the setting where research activities are not custom- diagnosis and/or the researcher bearing the news. ary. For example, a busy primary care practice or There may be anger or discouragement towards even a mental health facility may not be oriented the community setting in which the person sought towards identifying and tracking individuals who help. The researcher needs to be sensitive to these meet criteria for anxiety disorders. Frequently possibilities and proactive in dealing with unto- staff in such places are very busy, very dedi- ward reactions associated with identification of cated and sometimes opinionated about what is an anxiety disorder. For example, if there is a best for their patients. The researcher who comes decision to recruit subjects in a non-psychiatric to study usual practice may be seen as challeng- setting such as a church or supermarket, the ing the skills, competence or even integrity of the researcher would need to include an introductory staff. Still, recent studies in primary care8 have phase of the work that addresses fears and stigma succeeded in overcoming these barriers and have associated with a diagnosis. This can be done in a done much to provide information to inform pro- variety of ways. A community educational phase cesses to optimise care. might be undertaken prior to initiation of recruit- Protocol-driven treatments face additional bar- ment. Individual or group consciousness raising riers to acceptance in settings other than the might be offered. Focus groups are a very useful research clinic. Assessment of outcomes is hard strategy being increasingly used by researchers. enough in a research clinic; assuring good In this situation, small groups of individuals with follow-up and reliability of measurement in different types of anxiety might be invited to ANXIETY DISORDERS 259 participate in a focus group. Participants are paid kind of work.10 Partnering with administrators in and the group leader might guide the group in different service agencies to find ways to improve discussion of topics such as the meaning of hav- their efforts is likely to provide easier access to ing an anxiety disorder, the perceived response subjects and better support for implementation of of others, including family, friends and/or the study procedures. Careful attention to such issues community at large. A focus group might also can determine the feasibility of the study. be asked to discuss how researchers might best approach undiagnosed people in the community Assessment Strategies in Community Settings who suffer from these disorders, or the group might be asked how to best present treatment The standard research diagnostic interview and options, or how to explain and encourage par- follow-up batteries were designed to achieve ticipation in research. Thus armed, the researcher careful, reliable descriptions of different well- will be more successful in recruiting and retaining specified phenotypes of psychiatric illness. While subjects for a clinical trial. An example of a very highly successful in meeting this goal, such innovative approach to the problem of commu- instruments have not been designed to maximise 9 nity recruitment utilised an intensive telephone efficiency and minimise patient and staff burden. engagement strategy in which mothers of inner These extensive and time-consuming inventories city minority children were invited to identify will not survive transplantation into a primary and problem-solve an important difficulty they care setting, a hospital emergency room or a were experiencing. Only after the intake recruiter dental office, let alone a church or school. Instead, had successfully helped with this practical prob- radically simplified tools must be developed that lem did they explain the availability of services utilise innovative statistical and psychometric for other kinds of problems. This approach was methods (e.g. adaptive testing) and/or study shown to significantly increase attendance at the sample sizes must be increased to compensate first clinic appointment. Whatever the approach, for extra variance. it is clear that the prospective patient research The assessment strategy used in a community volunteer must be given opportunities to under- setting may need to be altered in other ways as stand their anxiety symptomatology as a treat- well. No matter how prevalent an anxiety disor- able condition underlying what may be just an der may be in the community, it will be lower in awareness of limitation or fear. These individu- the community setting, compared to the preva- als further need to decide for themselves which lence in the enriched intake stream of a specialty treatment programme they wish to access. The anxiety clinic. The odds on a disease may easily researcher needs to present a clinical trial in vary fivefold or more from clinic to commu- this context. nity. Given that the specificity and sensitivity of Sometimes stigma can be best addressed at the the diagnostic instruments will be no better in level of the service provider–such as a primary the community setting, and may well be worse, care physician, or administrators and service the ‘Bayes factor’ of the test (sensitivity divided providers in different kinds of agencies. In order by 1 − specificity) will be smaller in the commu- to access patients in a given facility it may be nity setting. For example, if the sensitivity and very important to first understand the headaches specificity both decline from 90% to 80% then of the facility administrators. A researcher who the Bayes factor declines from 0.9/(1 − 0.9) = 9 takes the time to both identify and respond to to 0.8/(1 − 0.8) = 4. If both the prior odds of the problems faced by those attempting to deliver disease and the Bayes factor are lower, the posi- care will be rewarded with a much higher level of tive predictive value will also be lower (the odds enthusiasm and support for the research project. of disease given a positive test are just the prior Researchers in the field of services research have odds of disease times the Bayes factor). In the understood and successfully accomplished this numerical example above, given only a fivefold 260 TEXTBOOK OF CLINICAL TRIALS difference in prior odds, the posterior odds on dis- Practical and Administrative Issues ease given a positive test may vary by an order of magnitude from clinic to community, mak- Human subjects review may need to be coordi- ing interpretation of intake diagnosis problematic. nated among several kinds of organisations. Some Multi-step diagnostic procedures may be neces- may be willing to enter into agreements to accept sary to avoid over-diagnosis.11 the investigator’s home institutional review, oth- ers may need to develop their own review pro- cesses and obtain Federal-Wide Accreditation. Research Recruitment Template agreements that have been shown to work would be a valuable resource. Study subjects volunteer to participate in research. The investigator must choose methods of data In a clinical trial, the manner of presentation of capture and processes for the data edit cycle that treatment options may make or break a study. work in diverse settings at sites that are distant The highly selected population of patients who from the coordinating institution. Data monitor- present to an academic centre often come pre- ing, correction of errors and tracking of follow-up conditioned to the value of research protocols, are all affected. Technological limitations need and may have specifically sought out the clinic to be respected. For example, fax-based meth- because of its reputation as a research centre. ods may be more easily deployed than internet- The potential research volunteer in a community based methods, especially in settings where a fax setting has not voted for research ‘with his machine is already in use. On the other hand, feet’, and may need a gradual, informative and as personal digital assistants become ubiquitous, upbeat approach, to accept the idea of protocol- patient follow-up may be individualised, remote driven treatment and randomisation. Institutional and remotely cued. We can imagine technol- review boards may well regard placebo control ogy that rings a telephone number or sends an as especially unattractive in such a context, and instant message, asking for self-report follow- may also be concerned about ‘overselling’ the up, and that can schedule and connect the sub- potential benefits of research to patients. Yet, ject with a live interviewer, all implemented on there is reason to believe that the patient in the the same small wireless device, that might be community context may be the one with the most cheap enough to give away as a free incentive to gain from participation in research, because to participation. of the likelihood that her illness would otherwise Despite the difficulties associated with export- go unrecognised, or the equally disadvantageous ing clinical trials to the community, it is clear likelihood of inadequate treatment. that the next generation of clinical research in anxiety disorders needs to be rolled out into the settings where individuals with these disorders Context-Relevant Treatment Protocols live and work. In addition to the many issues related to the setting of a study in the commu- To optimise study results, strategies must be nity, there are many design considerations related developed for providing protocol treatments in to which patients should be included in a given a context-relevant manner. This may include study. Patients in different community settings adjusting to the absence of third-party payers, or are likely to be heterogeneous in different ways, making use of setting-specific para-professional and to differ from patients who seek treatment at personnel for some of the interventions. Or, it traditional research clinics. Existing studies doc- may mean incorporating ‘escalation’ strategies ument a high rate of comorbidity among anxiety into the treatment protocol, so that subjects disorders, between anxiety disorders and depres- identified with substantial needs are transferred sion, and between anxiety disorders and medical to a more traditional setting. illnesses. There is also comorbidity of anxiety ANXIETY DISORDERS 261 disorders with for example psychotic disorders12 have had little exposure to proven efficacious and substance abuse.13,14 Such comorbidity may treatment. Often such patients have sought help increase the likelihood that a patient seeks treat- from clergy or other informal sources. In the case ment at research clinics, and therefore it is pos- of anxiety disorders, the awareness of the ‘irra- sible that studies in the community will have tionality’ of symptoms often means these individ- to deal with less comorbidity than studies in uals suffer in silence, embarrassed to reveal their the research clinic. Nevertheless, an effectiveness self-perceived defects. Such patients are often researcher must decide how to manage comorbid- enormously relieved when they learn that their ity. There are many consequences of decisions disorder is understood. Even when treatment has to include or exclude comorbidities from study apparently been offered, it may be less vigorous eligibility criteria. There are a variety of assess- than the versions that have been proven effi- ment considerations that are different in comorbid cacious in clinical trials. Inadequate doses and versus non-comorbid subjects. Symptoms of co- durations of pharmacotherapy may be the rule, occurring depression or substance abuse may be and specific psychotherapies may be offered in difficult to disentangle from anxiety symptoms. name only. It may be particularly important not Many medical disorders produce symptoms of to assume (for example) that a patient has demon- autonomic nervous system activation, as do anx- strated a lack of response to treatment, and there- iety disorders. Such medical comorbidity may be fore be ruled out as ineligible. especially likely in primary care or medical clinic If patients are identified in settings other than settings. The trade-off between heterogeneity and the specialty clinic, they may not view their its attendant increase in measurement variance, anxiety disorder (which may be news to them) and homogeneity and its attendant restrictions as the main problem they should be concerned on generalisability, must be carefully considered. with (along with their hypertension, macular In addition, rigid exclusion criteria may be less degeneration, current spousal abuse or arthritis). acceptable in community settings than in the spe- They may be unwilling to make accommodations cialty research clinic; patients who are surprised in schedules and may have needs for unusual by a diagnosis may be disappointed if they are availability of research staff in time and space. ruled out from studies by being ‘too compli- Some patients may not understand the usual cated’. An alternative for the researcher is to standard procedures for treatment in a mental simply accept comorbidity and heterogeneity of health clinic. They may need to be approached the population and evaluate a treatment that tar- in an accommodating way. gets a specific symptom, behavioural pattern or symptom cluster, without regard to the context POORLY DEFINED NOSOLOGIC in which it occurs. To make this decision the BOUNDARIES researcher accepts the ‘noise’ this will cause in the system and powers the trial accordingly. A second feature of anxiety disorders is that Other considerations related to patient het- the boundaries between normal and pathologi- erogeneity include the fact that illness severity cal anxiety and among the pathological disorders and typical background treatment history may are ill defined. Unlike most psychiatric disorders, vary across settings. Patients in some settings the symptoms that comprise the diagnostic cri- may have already received multiple treatments, teria for anxiety disorders are recognisable in while in other settings they may be treatment normal people every day. The pathological state na¨ıve. Given the findings from multiple studies is defined by excess. However, the definition of that have documented that affective and anxi- excess is not precise. Because anxiety is a nor- ety disorders are under-recognised and under- mal emotion, it is not always clear where the treated in the community, it is likely that patients boundary between normal and pathological lies. recruited from non-mental health settings will This is particularly true in the context of stressful 262 TEXTBOOK OF CLINICAL TRIALS life events and ongoing difficulties. The bound- is useful to have some anxiety symptoms in ary with normal may arise in defining a clinical order to keep coping functions operative and/or population in need of treatment. Boundary issues provide ‘toughening up’ experiences. Perhaps are also relevant to treatment targets and defini- some continued symptomatology is a good idea tion of remission. In general, there is no consen- to encourage continued exposure. The continued sus on what comprises remission of an anxiety presence of low-level symptoms may increase disorder. We discuss this problem and suggest the chances that the patient does not become some ways it might be addressed. The problem complacent16 and/or provide opportunities to of the boundary between normal and pathological confirm the absence of more severe symptoms. is not a question raised only in the area of anx- On the other hand, since anxiety disorders are iety disorders, but rather is a continued question clearly debilitating, perhaps it is best to eliminate in the ongoing discussion related to definitions symptoms as fully as possible. Perhaps if we of psychopathology. A recent paper15 provides a leave residual symptoms, this indicates that we good summary of current issues. As these authors have not eliminated the underlying vulnerability point out, it is also relevant to consider the rela- and relapse will be more likely. Ideally, we tionship between mental and physical disorders. would like to eliminate pathological anxiety These considerations are important for clinical while leaving ‘normal’ anxiety intact. Yet this researchers to keep in mind but a detailed dis- distinction may be difficult to define. If we have cussion of the various issues is beyond the scope a pharmaceutical compound that reduces anxiety, of this chapter. might we overshoot the mark? If so, would that However, as noted above, anxiety is a normal be as problematic as undertreatment? Common emotion, and so its pathological state must be sense, and the results of a famous study,17 suggest distinct from normal variation. It is best to that a moderate level of anxiety is associated with experience anxiety in moderation. While anxiety optimal performance in situations like test-taking. can be disabling in excess, a deficiency of anxiety Laurence Olivier is known to have suffered, as can also be impairing. The question of how many actors do, from tremendous stage fright. much anxiety is optimal is not a philosophical His view of this was that this fear was an one. Rather, it is one of the conundrums that essential motivator that ensured his performance currently face the clinical trials investigator. would be undertaken with the highest possible Namely, the investigator must decide how much focus and concentration. Every researcher knows symptom relief is optimal and how much is that the approach of the deadline for grant sufficient to declare a meaningful response to submission generates substantial anxiety which treatment. Given that anxiety is normal, is again motivates the highest possible level of there some expected floor for the intensity of energy and productivity. anxiety symptoms, or is symptomatic anxiety Threshold issues relate to the decision to qualitatively different? begin as well as the decision to end treatment. Another design question relates to the level At what point do we declare anxiety to be of anxiety that results in optimal long-term at a clinically significant level that warrants outcome. Still another relates to the definition intervention? If Laurence Olivier experienced of remission of a given disorder. The field intense anxiety at each performance, should he be has not reached consensus on how to define treated? The goal of treatment of an unhappy but remission for any of the anxiety disorders. This successful person should be first and foremost to is a critical methodological problem that needs prevent failure (inability to perform, because of to be addressed. Investigators need to consider paralysing fear or shoddy performance, because whether there is a way to overshoot the mark or of cavalier attitude) while, if possible, reducing is less always more? This is a serious question, the discomfort of unhappiness. In this context we as researchers are not agreed upon whether it echo a famous quote of Freud, concerning the ANXIETY DISORDERS 263 goals of psychoanalysis vis-a-vis` unhappiness. mitigate the perception of likelihood of danger Anxiety clearly exists on a continuum yet a and/or the perception of consequences of the dan- treatment decision is a binary one. We do not ger. Anecdotally, some anxiety disorder patients attempt to administer a partial treatment. The are thought to have unusually good interpersonal decision of who to treat, of the minimal level skills. Turning to others may be one way a patient of symptomatology an eligible subject may have with panic disorder copes with a world perceived will have implications for interpreting and acting as persistently and unpredictably frightening. For on study results. It is likely that there is a distance other individuals with anxiety disorders, anxiety from the boundary with normality associated may be exacerbated by relationships with oth- with optimal effect of treatment. The closer to ers. A patient with social phobia fears scrutiny the boundary, the more likely the study will by others and this may motivate them to avoid show non-specific or placebo effects. The farther relationships or to concentrate on developing a from the boundary (i.e. the more severe and small group of ‘safe’ people. Someone with OCD complicated the symptoms are) the less likely may fear contamination from others, or an OCD that the treatment will be fully effective. One patient may fear harming other people. Either consideration in deciding who to treat in a may lead them to have reduced social contacts research study of anxiety disorders is the life and less overall sense of safety. It is also pos- context and the personal context in which the sible that deficits in internal representations of anxiety disorder symptoms arise. other people can lead to problems in regula- tion of emotions. This can contribute to anxiety Considerations Related to Life Context symptoms and perhaps even to the onset of anx- and Individual Psychology iety symptomatology. There is some indication that relationships with others help regulate neu- Because of the salience of environmental stimuli roendocrine and autonomic nervous system func- as a trigger for normal anxiety, and the impor- tioning. Whether to include measures of social tance of coping mechanisms and social supports support and attachment into clinical trials in anx- as responses to anxiety, it might be argued that iety disorders is a decision researchers must begin these context measures are of particular impor- to consider. Such information is an additional tance in anxiety studies. Little is known about the patient burden. However, it may be difficult to relationship between onset, course and treatment determine optimal interventions for patients in the of anxiety disorders and these external factors. community if researchers do not begin to address There is a need to examine what the nature of some of these issues. these relationships may be. For example, it is not Also ill defined are boundaries between dif- known whether faulty coping mechanisms play a ferent diagnostic categories with similar symp- role in the vulnerability to one or another of the tomatology, and frequent comorbidity. Given the anxiety disorders. If so, perhaps this should be fact that fears, worries, somatic symptoms and a target of a treatment intervention. If not, per- behavioural manifestations are shared across dis- haps coping skills are variable across individuals orders, distinctions can sometimes be blurred. and/or across stressors and may act as a modera- There have been changes in diagnostic criteria tor of treatment response. In this case, improving sets since DSM III, especially for panic disor- coping may be a strategy for treatment of non- der and generalised anxiety disorder (see below). responders. There has also been a change in the relationship Strong social support is well known to be an between panic and agoraphobia, and between important contributor to a sense of safety. Anx- these disorders and other DSM IV anxiety disor- iety disorder patients experience the world as ders. These changes reflect growing recognition more dangerous. Safety is not necessarily the of the occurrence of panic and phobic symptoms opposite of danger, but a sense of safety can across disorders. In addition, the core generalised 264 TEXTBOOK OF CLINICAL TRIALS anxiety disorder (GAD) symptom, worry, is also part, these are beyond the scope of this paper. frequently found in other anxiety disorders and Researchers have developed tools to use for in depression. The content of worry apparently screening, diagnosis and severity ratings of anxi- differs in depression and GAD.18 However, this ety symptoms. Ensuring that clinicians are aware is a nuance for an assessment strategy, and again, of these and that the instruments are user-friendly time is required to tease this apart. is a focus of current work. There are many exist- The diagnostic groups that comprise the anx- ing publications related to different assessment iety disorders share cognitive manifestations of instruments, so we will not provide this infor- fear or worry, behavioural manifestations of mation here. Instead, we suggest that even with efforts to cope with the anxious thoughts (pho- a good instrument, there are some difficulties in bia or compulsions), and somatic symptoms that establishing clearly a single target condition, and often accompany the anxiety. Yet each disorder that the attempt may be a useless diversion of has a different ‘flavour’ of symptoms, and each effort if discriminatory precision is less important may reflect different aetiological underpinnings. than inclusiveness of intake. The similarities have implications for clinicians The high rate of co-occurrence of anxiety dis- and researchers alike and will have a bearing orders is an area of confusion that concerns on the design of new treatments and dissemina- the diagnostic nomenclature. For this and other tion of efficacious interventions. As study intake more theoretical reasons, there is controversy in moves into the community, we can expect the the field with regard to whether different DSM diagnostic overlap to increase, if only because IV diagnoses describe truly distinct illness cat- mild versions of disorders are harder to separate egories. Moreover, even if they are different, than severe ones. Efficacy studies have focused their co-occurrence creates measurement prob- on specific diagnostic groups, rather than on anx- lems. For example, if a patient in treatment for a iety as a loose complex of symptoms. This means panic disorder has a co-occurring specific phobia that the ostensible usefulness of study results of heights, should avoidance of bridges be rated depends on clear, reliable identification of the a symptom of panic disorder, of height phobia specific disorders in patients, at best a dubious or both? proposition in the community setting. However, As noted above, it may be possible to aim cur- it should be recognised that while the efficacy rent established treatments on the generic symp- studies have been carried out in carefully con- toms that occur across the disorders: fear or worry, somatic symptoms and behavioural changes such structed ‘pure cultures’, the results of those stud- as avoidance or compulsive ritualising. The broad ies show a startling uniformity of options for spectrum effects of serotonin-active medications treatment across the spectrum of anxiety dis- lend themselves to such an approach, as do orders. It may be that the careful and expen- the psychosocial treatments which may reduce sive nosologic dissection characteristic of the first generic cognitive, somatic and behavioural symp- generation of clinical trials is added to the pre- toms across disorders. An investigator planning a cision and power of those studies, but may be community study needs to decide whether to test relaxed in the next generation of effectiveness treatments in the classical disorder-specific trial or studies, making a virtue of necessity. in a more broadly-based group of patients defined We now know how to reliably identify anxiety by the common symptoms and behaviours of the disorders and we know how to provide effica- anxiety disorders. We think that the latter choice cious acute interventions, but these demonstra- deserves serious consideration. tions have occurred only in the research clinic. The traditional decision point for clinical inter- Defining Outcomes and Measuring Results vention, i.e. clinical (DSM IV) diagnosis, is fairly clear, though there are remaining contro- Katschnig and Amering19 point to the consider- versies about psychiatric diagnosis. For the most able complexity of symptoms in panic disorder, ANXIETY DISORDERS 265 suggesting that spontaneous and situational panic levels while remission entails a return to func- attacks, anticipatory anxiety, phobic avoidance, tioning with symptoms at a level that they cause disabilities, comorbid depression and substance no noticeable distress. Studies are underway to abuse must be considered. One might add to this identify precise markers of these important clin- the presence of other anxiety disorders, personal- ical transitions. ity disorders and physical illnesses. These authors The fact that the symptoms of a single dis- list methodological difficulties that emerge from order do not necessarily travel together creates this complexity. They raise questions such as difficulties in defining the endpoint for a treat- which of the phenomena are most important in ment. Such a ‘carousel course’ (Figure 17.1) of assessing course and outcome, what are the rel- symptoms leads to assessment quandaries that evant time intervals for symptom assessment, at can be daunting. Again, taking the case of panic what point should the clinician consider that the disorder with agoraphobia, if a patient starts treat- illness has transitioned to a remitted state? Such ment with several full panic attacks per week, considerations are relevant for each anxiety dis- andthenhasamarkedreductioninpanicattacks, order. All are comprised of multiple domains but continues to have frequent limited symptom of symptoms, including panic, anticipatory anx- episodes, and remains moderately agoraphobic, iety, worry, phobia, obsessions or compulsions. how much improved is the patient compared to Since the diagnostic criteria do not require the baseline? How should life context be factored presence of each of these, it is possible to meet into assessment of illness severity? If a social criteria for one or another anxiety disorder with phobic gets a new job which requires less pub- prominent symptoms in one domain and none in lic performance than previously, but the job is another. Over time this may change. In some situ- below his or her level of competence, social anx- ations different domains within a disorder may be iety symptoms may diminish noticeably but is the negatively correlated. For example, an individual patient really better? may experience a reduction in panic attacks while Several authors have drawn attention to these becoming increasingly phobic. Is this improve- problems and the general recommendations have ment or worsening of the overall condition? A been to use composite measures of severity over patient with OCD may experience a decrease in an extended period of time. Such composite mea- obsessions as the compulsive behaviours grow sures are available for most of the disorders, and and become instantiated. Should this be con- most are quite user-friendly: The Yale–Brown sidered a change in severity? A person with a Obsessive Compulsive Scale has been widely phobia may experience lower overall impairment used and is available in a self-report format. and/or fear as their phobic behaviour become The same is true for the more recently devel- more fixed, and they begin to accommodate the oped Panic Disorder Severity Scale. The Social phobia in their lives. Does this mean the phobia Phobia Inventory (SPIN) is also brief and com- is in partial remission? What if the intensity of prehensive. There is little agreement in the field symptoms is actually worse than when the dis- about the one or two best measures for each disor- order was first diagnosed, and yet there is less der. The same measures can sometimes be used impairment? Similarly, if an individual with OCD for screening diagnosis and outcome though it has prominent obsessions and intermittent com- makes sense to pick the instrument most rele- pulsions are they better off, worse off, or the same vant to the goal of the assessment. The use of a as if the opposite is true? What is the role of composite measure presupposes that it is possi- impairment and/or quality of life in determining ble in principle to rank order the outcomes of the outcome? What criteria should we use for ill- patients, although there may be many outcomes ness severity? What about treatment response or that are distinct but not ordered. From a statistical remission? It is clear that response entails a clin- point of view, the ability to reliably order patient ically significant, noticeable change in symptom outcomes into as few as four or five categories 266 TEXTBOOK OF CLINICAL TRIALS
100%
80%
60%
40%
20%
0% Time 0 Time 1 Time 2 Time 3 Time 4
Anticipatory Anxiety Panic Agoraphobia
Figure 17.1. Panic disorder as an example of a ‘carousal’ symptom pattern provides considerable increase in the power to sizes, in order to detect modest effects on low detect treatment differences, compared to a sim- probability outcomes. ple dichotomy of response/no response. There are diminishing returns even to perfectly reli- Choosing a Time Frame for Outcome able orderings with more than five levels. Given Assessment even modest unreliability, it may not pay to push composite measures beyond a few levels of dis- The specifics of time frame are also contro- crimination. The challenge posed by the ‘carousel versial. While a group of senior panic disorder course’, and the pleiotropic outcomes, of anxiety researchers achieved consensus on a recommen- disorder is fundamental: the ability to rank patient dation of an optimal period of four weeks for outcomes is the most basic feature of scientific assessment of symptoms, and a minimum of two measurement and study. weeks, these recommendations are not always As studies extend into the community, they followed. In fact, frequently symptom status is will explore the less severe forms of the disor- reported without specification of the time frame der, and may be even more vulnerable to the of the assessment. The issues around time course problem discussed above. This raises the possi- are further complicated by variability between bility that the target of measurement should not domains and within a domain, depending on be improvement (alone) but prevention of signif- life circumstances. Some domains of symptoms, icant worsening. The advantage of this approach such as phobic symptoms, are very stable, and a is that it may move the measurement into a more change in them, even over a fairly brief period, reliable regime, in which there is less controversy e.g. a few weeks, can be an indicator of change about the meaning of the outcome. The disad- in illness severity or course. (A caveat here is the vantage is that it may also require large sample change in life context.) However, other symptoms ANXIETY DISORDERS 267 are very unstable. For example, it is typical for both a coping mechanism and a symptom. By its panic attacks to occur in clusters and then to nature, it can be difficult to measure, since many subside. The problem is further compounded by anxiety disorder patients try to avoid thinking difficulties inherent in rating panic frequency. about anxiety-provoking situations, in addition Anticipatory anxiety can be far worse if there to avoiding confronting these situations. This is a specific environmental demand to confront means that asking a person if there is anything an anxiety-provoking situation. For example, if a they are avoiding often results in under-reporting. social phobic must go to a wedding. This raises It is necessary to inquire about avoidance by the question of the time frame over which differ- asking specific questions, and this can be time- ent types of symptoms should be assessed, and consuming. Some behaviour therapists argue that the situations in which the symptoms should be phobias can only be assessed using a behavioural evaluated. Again, a focus on long-term preven- challenge protocol. However they are measured, tion of serious worsening may help. it is clear that phobic symptoms are important as We do not have definitive answers to these they are among the strongest and most consistent myriad questions, but suggest that we should be predictors of long-term outcome. paying attention to these assessment challenges. Avoidance can also play a role in silencing anx- It may be possible to undertake secondary data iety symptoms and reducing the impetus to seek analyses that target these questions. In the mean- help. This may be one way that phobic symptoms time, we suggest that outcome assessment must act to worsen the course of illness. Silencing of take into consideration multiple domains to make symptoms is also reminiscent of the hypertension a meaningful judgment of response or remission. analogy where serious consequences result from Further, it makes sense to establish and publish lack of awareness of symptoms and difficulty conventions for raters so that others can under- adhering to treatment regimens. In fact, phobic stand clearly results of studies. Reports of study avoidance has now been found, like hyperten- results rarely describe conventions for rating pho- sion, to be a predictor of cardiovascular mortality, bia, including changes in life context and/or situ- at least for men. A further issue related to the ational demands. Many published panic disorder silencing of distress is that it can be difficult to studies use panic attack frequency as the only distinguish pathological from normal avoidance outcome. Reporting conventions should be broad- behaviours. Phobic symptomatology may become ened to address these issues. so integrated into the patient’s life that it seems normal. Avoidance of some situations may be Phobic Fear and Avoidance treated as though they are simply life choices. The patient may say that he or she simply does A third issue specific to anxiety disorders is not enjoy shopping in a mall when the fact is that the occurrence of phobic symptomatology. One they are afraid to go to a mall because they may of the trickiest problems in anxiety disorders have a panic attack. The problem of distinguish- treatment is the assessment and management of ing normal from pathological anxiety is broader avoidance. Avoidance is a natural reaction to fear than the issues related to phobia. and is usually successful in reducing anxiety in This realm of symptoms causes methodolog- the short run. However, the longer-term effect is ical problems that involve both assessment and virtually always to increase anxiety. Avoidance treatment. Phobic symptoms entail avoidance of also causes substantial functional impairment. cues that evoke fear, anxiety or other dyspho- Avoidance may lead a social phobic to seriously ric affects. An individual with phobic avoidance curtail his or her education, or resist career will make every effort to evade exposure to the development because of fear of speaking in a feared situation. Evasion often extends to think- group. The net result can be highly significant ing or talking about the situation. This means to income and productivity. Thus, avoidance is that the phobic individual cannot be counted on 268 TEXTBOOK OF CLINICAL TRIALS to talk about their symptoms spontaneously. In presents some especially challenging method- fact, to obtain a clear picture of the extent of ological issues for which there is no simple behavioural avoidance often requires a detailed solution. The practice of ignoring the findings inquiry. Such an assessment takes time and is of the other modality when conducting stud- not desirable in many community settings. An ies is increasingly problematic. If not specif- investigator must decide whether this time is ically instructed, pharmacotherapists may vary worth the trade-off of information. The answer widely in their knowledge and use of effi- to this question will be influenced by the type cacious behavioural interventions. This varia- of study and the population being studied. How- tion can be highly problematic for a treatment ever, it is important for researchers to be aware study. On the other hand, much effort must go that more co-occurring phobia has been consis- into controlling the interaction of the pharma- tently associated with poorer response to treat- cotherapists with the patient. Researchers must ment and greater likelihood of relapse.20,21 The decide how much behavioural intervention the clinical significance of phobic symptoms under- medication therapist should administer. Compli- scores the need to attend to this component of cated and time-consuming procedures are often anxiety symptomatology. required to ensure that such interventions are pro- vided uniformly. THE TWO CULTURES: DISCORDANT On the other hand, patients in the community MODELS OF THE ANXIETY DISORDERS often receive medication that can be efficacious for treating anxiety disorders, even before pre- A fourth characteristic of anxiety disorders is senting to the cognitive behavioural therapist for based upon the fact that there are two powerful treatment. Investigators must decide how to man- models of these disorders that are not yet fully age this situation. Should such patients be elimi- integrated. Specifically, neurobiological (gener- nated from a CBT trial? Should they be eligible ally biomedical) and learning theory (generally and left on medication that is not fully effective? academic psychological) researchers use different Or should all medication be discontinued? Each paradigms to explain symptom onset and to guide of these decisions is problematic since a partially treatment. When studying treatment in commu- effective medication can affect outcome whether nity settings, it is important that neither group it is continued or discontinued. Omission of this ignore the other. In anxiety disorders, perhaps increasingly large group of patients, on the other more than any other conditions, there is a need hand, can also be an important threat to general- to build on information obtained from both of isability of the study. these academic disciplines, given that each field Another problem for researchers is how to can claim clinical results. decide whether to compare medication and psy- chosocial treatments, and if so, to decide how best Incorporating Information from Biomedicine to do so. There is clearly a need in the field to and Academic Psychology in Study Designs address this problem, but the solution to the prob- lem is not trivial. Among the problems are that Anxiety disorders are unusual in that they have patients often have treatment preferences. Many been the focus of intensive and more or less will simply refuse to participate in a study in independent study by both biomedical/psychiatric which they must agree to be assigned to a treat- and behavioural/psychological researchers. Effi- ment modality at random. Others will agree and cacious treatments have been devised by each drop out when they receive an unwanted treat- group. Yet, most treatment studies test inter- ment assignment. There are a number of possible ventions in one, but not both areas. The exis- designs for a comparison treatment study. These tence of two very different types of efficacious include a full factorial design (Figure 17.2) in intervention for each of the anxiety disorders which each active treatment is compared to an ANXIETY DISORDERS 269
Active medication Placebo medication No pill
Active X X X psychotherapy
‘Placebo’ X X X psychotherapy
No X X X psychotherapy
Figure 17.2. Full factorial design inactive (placebo) control treatment and no treat- a more comfortable meeting ground for both ment, and the two treatments are combined in groups. Still, there are disagreements. all possible combinations. While this is the most In addition to the scientific differences, there complete design it is often impractical because of are social and political differences between the the treatment combinations (or lack of treatment) two groups of researchers that can complicate or because of the number required. It is difficult methodological decisions in treatment trials. The to undertake such a study at a single site and then investigators need to be clear about who the audi- there are problems with multiple sites in equiva- ence for their results will be. Design decisions lence of providing all treatments and in minimis- may influence who will listen to their results. ing patient heterogeneity. An alternative class of Ideally, a study can be designed so that it will be designs has been described recently22 that allows convincing to any treatment researcher. However, patients and investigators to describe preferences there are turf issues that may influence the mutual in advance of randomisation and then be ran- acceptance. Clinicians and researchers from one domised within their preference set (‘equipoise camp may feel the other is poaching on their turf. stratum’). This is more likely to occur if there is insufficient Other design issues are related to the differ- attention to the issues of efficacy of the alterna- ent putative underpinnings of symptoms as con- tive treatments. ceptualised from different points of view. These Guild issues are prominent in this field, and different viewpoints sometimes translate to dif- few pharmacotherapists understand the princi- ferent treatment targets. For example a CBT ples and techniques of administering cognitive approach to panic disorder focuses on underly- behavioural treatment. Similarly many psychoso- ing fear of bodily sensations, while the pharma- cial researchers are not well informed about phar- cotherapist targets bursts of autonomic arousal. macotherapy. Investigators from each group tend Pharmacotherapists and psychosocial researchers to have strong allegiance to the unique validity of typically use different assessment strategies, and their own methods. At times, there has been ran- may or may not accept those of the other camp. In cour and contentiousness between them, though recent years, a series of multisite studies under- this has improved in recent years. In the few taken as a collaboration between neurobiologic instances that there has been a head-to-head com- and cognitive behavioural scientists has produced parison of medication and cognitive behavioural 270 TEXTBOOK OF CLINICAL TRIALS treatment, they have been similar in efficacy. It long run? Do patients with complex comorbid is not yet clear when or how combination treat- conditions respond to treatment in a way that is ments might be best administered. There is a similar or different than those with less comor- need to take into consideration both biomedi- bidity? Can a clinician be confident that proven cal and behavioural–psychological perspectives efficacious treatments are appropriately utilised in designing treatment studies. in a patient whose symptoms meet criteria for the target disorder, but who differs in demo- graphic characteristics, social supports, or other CONCLUSIONS ways from those seen in efficacy studies? How closely must procedures in the community follow those used in research studies in order to achieve Although proven efficacious treatments have the same results? been identified for each of the anxiety disorders, These and other questions like them are often the work of clinical trials remains unfinished. broadly grouped under the rubric of ‘effective- There are many unanswered questions, and much ness’ studies. We focus especially on characteris- left to study in order to inform clinicians about tics of anxiety disorders that make these decisions how to optimise treatment decisions for patients complicated and that comprise methodological with these debilitating conditions. In spite of challenges for researchers. We confess that we achievements in documenting treatment efficacy may raise more questions than we can answer. over the past decades, treatment research has just However, where possible, we will at least pro- scratched the surface. Innovations are needed in vide suggestions about possible ways to address treatment development and in dissemination of the problems. proven interventions. To accomplish this there is Decisions about primary, secondary and tertiary a need for innovations in methodology. Efficacy prevention interventions are not so well specified. studies, designed to meet US FDA regulatory 23 There is accumulating evidence for psychological needs, will continue to have a role in the clinical and neurobiological precursor states for anxiety research pipeline. But, there is a need for new disorders and for psychosocial risk factors. This clinical methods to support studies before and information raises hopes for the possibility of pri- after efficacy. It is not our purpose to provide a mary prevention. Clearly it would be advantageous comprehensive review of such methods. Rather, to be able to intervene early, before the devel- we have focused on a few key issues in anxiety opment of the disorder and, ideally, even before disorders that require special consideration. the onset of a precursor state. Once established, Existing clinical trials in anxiety disorders, like anxiety disorders are chronic relapsing conditions. those in other areas of psychiatry, have provided With or without treatment, patients are likely to information telling us which treatments are active experience symptoms that wax and wane, to mean- in reducing target symptoms. Unanswered are a der in and out of full-fledged symptom states in myriad of critical questions that relate to daily life different configurations, to experience temporary, decisions in the clinic. For example, do impair- partial or incomplete states of remission. We need ments as well as symptoms respond to efficacious more information about how to manage anxiety treatments? If so, what is the time course of disorders, once established, in order to best pre- response? What is the optimum dose and dura- vent complications and recurrence of full symp- tion of treatment to achieve maximal results? tomatic episodes. How often can we produce remission with exist- ing treatments? What is the best way to define remission? Is maintenance treatment needed after REFERENCES remission is achieved? If so, how long? What if 1. Kessler RC, McGonagle KA, Zhao S, Nelson CB, a patient does not achieve symptom remission? et al. Lifetime and 12-month prevalence of DSM- How should such a patient be managed over the III-R psychiatric disorders in the United States: ANXIETY DISORDERS 271
Results from the National Comorbidity Survey. with disparities in treatment of patients with Arch Gen Psychiat (1994) 51(1): 8–19. schizophrenia and comorbid mood and anxiety 2. Hohmann A, Shear MK. Community based inter- disorders. Psychiat Serv (2001) 52(9): 1216–22. vention research: coping with the noise of real 13. Kandel DB, Huang F-Y, Davies M. Comorbidity life in study design. Am J Psychiatr (2002) 159: between patterns of substance use dependence 201–7. and psychiatric syndromes. Drug Alcohol Depend 3. Kessler RC, Soukup J, Davis RB, Foster DF, (2001) 64(2): 233–41. Wilkey SA, Van Rompay MI, Eisenberg DM. The 14. Brady KT. Comorbid posttraumatic stress disorder use of complementary and alternative therapies to and substance use disorders. Psychiat Ann (2001) treat anxiety and depression in the United States. 31(5): 313–19. Am J Psychiat (2001) 158(2): 289–94. 15. Widiger TA, Sankis LM. Adult psychopathology: 4. Shear MK, Greeno C, Kang J, Ludewig D, issues and controversies. Ann Rev Psychol (2000) Frank E, Swartz HA, Hanekamp M. Diagnosis of 51: 377–404. non-psychotic patients in community settings. Am 16. Fava GA, Rafanelli C, Ottolini F, Ruini C, Caz- J Psychiat (2000) 157(4): 581–7. zaro M, Grandi S. Psychological well-being and 5. Glisson C, Hemmelgarn A. The effects of organi- residual symptoms in remitted patients with panic zational climate and interorganizational coordina- disorder and agoraphobia. J Affect Disord (2001) tion on the quality and outcomes of children’s ser- 65(2): 185–190. vice systems. Child Abuse Neglect (1998) 22(5): 17. Yerkes RM, Dodson JD. The relation of strength 401–21. of stimulus to rapidity of habit formation. JComp 6. Rosenheck R. Stages in the implementation of Neurol Psychol (1908) 18: 459–82. innovative clinical programs in complex organiza- 18. Diefenbach GJ, McCarthy-Larzelere A, William- tions. JNervMentDis(2001) 189(12): 812–21. son DA, Mathews A, Manguno-Mire GM, Bentz 7. Swartz H, Shear MK, Frank E. Supermarket Psy- BG. Anxiety, depression, and the content of chotherapy. Psych Services (in press). worries. Depress Anxiety (2001) 14: 247–50. 8. Roy-Byrne PP, Katon W, Cowley DS, Russo JE, 19. Katschnig H, Amering M. The long-term course Cohen E, Michelson E, Parrot T. Panic disorder in of panic disorder and its predictors. J Clin primary care biopsychosocial differences between Psychopharm (1998) 18(6, Suppl 2) 6S–11S. recognized and unrecognized patients. Gen Hosp 20. Scheibe G, Albus M. Predictors and outcome in Psychiat (2000) 22(6): 405–11. panic disorder: a 2-year prospective follow-up 9. McKay M, McCadam K, Gonzales J. Addressing study. Psychopathology (1997) 30: 177–84. the barriers to mental health services for inner 21. Cowley DS, Flick SN, Roy-Byrne PP. Long-term city children and their caretakers. Commun Mental course and outcome in panic disorder: a naturalis- Health J (1996) 32(4): 353–61. tic follow-up study. Anxiety (1996) 2: 13–21. 10. Essock S, Goldman H. Outcomes and evaluation: 22. Lavori PW, Rush AJ, Wisniewski SR, Alpert J, system, program and clinician level measures. Fava M, Kupfer DJ, Nierenberg A, Quitkin FM, In: Minkoff K, Pollack D, eds, Managed Mental Sackeim HA, Thase ME, Trivedi M. Strength- Health Care in the Public Sector: A Survival Man- ening clinical effectiveness trials: equipoise- ual. Chronic Mental Illness,Vol.4.Amsterdam: stratified randomization. Biol Psychiat (2001) Harwood Academic (1997) 295–307. 50(10): 792–801. 11. Dohrenwend BP, Shrout PE. Toward the develop- 23. Rubinow DR. Practical and ethical aspects of ment of a two-stage procedure for case identifica- pharmacotherapeutic evaluation. In: Ginsburg BE, tion and classification in psychiatric epidemiology. Carter BF, eds, Premenstrual Syndrome: Ethical Res Commun Mental Health (1981) 2: 295–323. and Legal Implications in a Biomedical Perspec- 12. Dixon L, Green-Paden L, Delahanty J, Luck- tive. New York: Plenum Press (1987) 47–55. sted A, Postrado L, Hall Jo. Variables associated 18 Cognitive Behaviour Therapy
NICHOLAS TARRIER1 AND TIL WYKES2 1School of Psychiatry and Behavioural Sciences, University of Manchester, Manchester M23 9LT, UK 2Institute of Psychiatry, De Crespigny Park, London SE5 8AF, UK
BACKGROUND levels of financial independence and little social fulfillment. There is some underlying variation We have chosen in this chapter to provide an in the disorder,2 which is probably affected overview of the difficulties for the investigation by interactions with other clinical, social and of psychological therapies using the methodology environmental demands and supports such as of randomised control trials. In order to do so life events (death of parent), absence of a we have selected studies of a new treatment for supportive family (or presence of a critical one) psychosis, cognitive behaviour therapy (CBT). and economic conditions (high unemployment). This is a new therapy that, following a period Several different sorts of psychological thera- of development, has now resulted in four large pies have been developed to address the follow- randomised control trials. ing outcomes: Cognitive behaviour therapy is a therapy that targets the symptoms of one disorder, schizophre- • Total number of symptoms nia. The lifetime risk is 1%. This disorder is char- • Distress caused by symptoms acterised by a cluster of specific symptoms that • Relapse are typically divided into two categories, positive • Social functioning and negative. Positive symptoms include auditory • Family engagement hallucinations and delusions, both of which pro- • Quality of life duce much distress. Negative symptoms include • Skills/thinking style, e.g. problem solving, lack of drive, emotional apathy as well as poverty coping skills. of speech and social withdrawal. In many, if not most, cases the disorder follows The currently accepted treatment for the posi- a relapsing course.1 A significant proportion, tive symptoms of schizophrenia is medication. but not all, people suffering from the disorder This has been shown to reduce them significantly have poor outcomes, i.e. with high levels of and does reduce relapse. However, it also has dependence on continuing psychiatric care, low costs as well as benefits in that there is the risk
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 274 TEXTBOOK OF CLINICAL TRIALS of developing side effects such as tremor, rest- proportions by different groups of professionals lessness and uncontrollable mouth movements. and for different individuals within a single Most side effects disappear on stopping the med- service. Below is a list of the ones that we have ication but there is the chance that the mouth identified as being used by most groups: movements will develop into a condition known as tardive dyskinesia that is irreversible. Some • Engagement with the client. patients, about one-third, also continue to expe- • Problem identification. rience positive symptoms despite adequate doses • Agreeing on a collaborative formulation of the of medication. It is these symptoms that were the problems to be assessed. targets for change in this further development of • Use of alternative explanations to challenge a psychological therapy, CBT. delusional and dysfunctional thoughts. Because of the potential risks of long-term • Establishing the link between thoughts and medication and the unpleasant side effects also emotions. experienced on short-term treatment, consumers • Encouraging the patient to examine alternative of mental health services have been extremely views of events. positive about the development of psychological • Encouraging the patient to examine the link treatment. This has led to further pressure on between thoughts and behaviour. funders of health service research to provide • Use of behavioural experiments to reality test. more data on acceptable alternatives or adjuncts • The setting of behavioural goals and targets. to medication treatment. Hence the recent trials • Developing coping strategies to reduce psy- of CBT in the UK sponsored by either the chotic symptoms. UK Department of Health directly, government • Development and acquisition of relapse pre- research agencies or large UK research charities. vention strategies. Some groups have also included: WHAT IS COGNITIVE BEHAVIOUR THERAPY? • Improvement in self-esteem. • Increasing social support and social networks. The main developmental roots for CBT have been • Schema focused therapy. in depression and anxiety. This began over 20 years ago but more recently the approach has TREATMENT DEVELOPMENT been applied to people with schizophrenia. This later development produced changes in the way New treatments usually evolve through a number the intervention is presented, although the under- of stages. Initially the problem is identified and lying model of change may be similar to that suggestions, involving theoretical and pragmatic adopted for the other disorders. The main aim of elements in varying degrees, are advanced for its the intervention is to reduce distress, disability solution. Innovative case studies are carried out. and emotional disturbance as well as the relapse Replications and developments in other case stud- of the acute symptoms.3 Cognitive behaviour ies and case series follow. The next stage consists therapies are active and structured therapeutic of uncontrolled and small exploratory controlled methods and should be distinguished from psy- trials. These are often innovative but methodolog- choeducation which tends to be simple, didactic ically weak. Finally the ‘gold standard’ of evalua- and educational. Brief educational packages have tion, the large randomised controlled trial (RCT), been shown to be ineffective either with families4 is carried out if the new treatment is showing suf- or with individual patients.5 ficient promise. RCTs are increasingly large and Although there are specific components of methodologically rigorous and therefore more CBT that would be accepted by all its proponents, expensive, often now involving numerous sites these ingredients may be given in different and large numbers of patients. A further theme COGNITIVE BEHAVIOUR THERAPY 275 is that of identifying what is responsible for the not random. Drury et al.12,13 evaluated cognitive improved outcome following the treatment; that therapy with acutely ill patients, but the treat- is, the trial includes an explanatory element. Tri- ment included individual and group treatment als that identify the key components responsible of patients and families while assessments were for the changes are essential to the further devel- neither independent nor blind. However, three opment of treatment and to the dissemination medium size methodologically robust trials of of the treatment package into the wider health CBT variants have been carried out with chronic service. However, this lack of an accepted the- schizophrenic patients,14–16 and one large multi- oretical base does not (and should not) prevent site trial with recent onset acute patients (the a number of different and successful treatment SoCRATES Trial17). It is therefore appropriate to innovations from being introduced into health review not only these trials but also the changes services. Pragmatic trials, which address the issue in clinical trial methods in this field in order to of whether or not a new treatment works within begin to define the most optimal strategy for the a routine service setting, are usually large, simple future evaluation of this and other psychologi- and multi-centred and evaluate a small number of cal therapies. outcomes. These trials do not address the ques- tion of why a treatment works. But this under- WHY CARRY OUT CLINICAL TRIALS standing is essential because the costs of care OF PSYCHOLOGICAL TREATMENTS? might be reduced considerably if it is discov- ered that a rather simple and cheap component of treatment is responsible for the majority of PURPOSES AND OUTCOMES the variance in treatment outcome. These treat- There are a number of different beneficiaries ment extensions, although important, rarely get from clinical trials. From the health services adequate funding following the initial innova- perspective there is an increase in knowledge tive RCTs. about what treatments are likely to provide The development and evaluation of CBT for the most benefits (see section on Evidence- psychosis is no different from the development Based Practice below). In addition, for clinical of other treatments in mental health and has academics there may be elements of the design of followed a characteristic path. Numerous case a trial that will allow certain models of aetiology studies were published, some as far back as or treatment efficacy to be tested which can the 1950s. For example Dr A.T. Beck initially inform theories of the disorder as well as leading worked with psychotic patients and published to improvements in treatment. For therapists a case study of the cognitive treatment of a the trial may produce clinical improvements patient suffering from delusional disorder6 before that mean that the participants can make health moving to start his seminal work on depres- gains and for the patients the treatments may sion. Other case studies were published in the provide them with changes that are valued, such 1970s and 1980s (see Tarrier7,8 and Haddock as increased social inclusion. So it cannot be et al.9 for reviews), but it was not until the recent assumed that there is a single purpose for carrying decade that randomised controlled trials were car- out a clinical trial. These different purposes ried out. Small trials with methodological weak- change the type of trial performed, particularly in nesses were initially published. For example, relation to how outcomes are defined. We have set Tarrier et al.10 compared coping training with out a number of different outcomes below which problem solving but assessments were not blind may be variously valued by different groups and drop-outs were not included in the analy- (in the list respectively health service, clinical sis. Garety et al.11 compared cognitive behaviour academic, therapist and participant) and which therapy to treatment-as-usual but again assess- could be targets in CBT trials. It cannot be ments were not blind and group allocation was assumed that all groups will value all outcomes 276 TEXTBOOK OF CLINICAL TRIALS to the same extent, or that the same outcome at the Institute of Psychiatry in London. The would be measured in the same way from the involvement of service users in clinical trials in different viewpoints. For instance, symptoms can the UK is now defined in guidelines provided be measured as a simple change over treatment, by the Consumers in Research Unit within the by a threshold amount or by the effects on the Department of Health. This new undertaking does emotional life of the patient, for example the not seem to be prevalent in other countries. distress caused by the symptom. The difficulty for research into psychological Possible outcomes of treatment: treatments is that studies are usually funded from public resources even at the early stages. This • The occurrence or frequency of a particular is in contrast to trials of medications where event: e.g. number of relapses, time to relapse particular companies are not only required to (survival functions). carry out specific research for licensing but are • The use of services or other resources: e.g. also likely to benefit financially from the results days spent in hospital, use of community of trials. Unlike drugs, psychological treatments mental health care. do not have a specific product champion and • Improvement in symptoms at a level assumed therefore have to compete with other health care to be of clinical significance: e.g. at least trials for scarce resources. 20% or 50% improvement, return to within normal range. THERAPEUTIC RELEVANCE • Change in a single symptom or other continu- In addition to the list of possible outcomes above ous outcome that is considered central or pri- there are other measures that may be essential mary to the disorder: e.g. severity of delusions in the assessment of outcome in a trial. For or hallucinations. instance, if one of the hypotheses is that the • Change in psychopathology that is general or therapy works through a specific mechanism then secondary to the disorder: e.g. scores on a a sole outcome measure without recourse to either standard measure of psychopathology, severity qualitative and/or process measures would not of distress or anxiety. provide a test of this hypothesis. This is an • Changes to other important aspects of the extension of the sorts of questions stipulated person’s life: e.g. social functioning, number by a clinical academic but is also essential to of friends, quality of life. the health services. It may be that the treatment provides its effects through a simple mechanism Trials are also expensive and so the chances which could be provided in a less sophisticated of funding are dependent on the types of trials way, that is not requiring high levels of training wanted by the funding agencies. The main and supervision. beneficiary (and funding) of clinical trials is the It has been suggested that psychological ther- health service who would prefer pragmatic rather apies may all work through a common pathway: than model testing trials. But in the UK there that the non-specific effects of psychotherapy has also been a new trend that may also affect may account for much of the effect of treatment the type of trial–the inclusion of mental health outcome.20,21 This is hardly surprising as psycho- service users (consumers) and, where appropriate, logical therapies have much in common with each carers on the trial management committees. There other–they involve, for example, an interaction, are examples of this; users and carers were negotiation of goals, an agenda for each session. represented in this way on a trial of CBT in dual The improvement could be produced by these diagnosis patients18 and of effectiveness of family commonalities and not through the specific model interventions,19 and also on the research steering of therapy adopted. For example, treatments that group which generates the research designs for were designed as non-specific placebo controls the Centre for Recovery in Severe Psychosis (e.g. befriending in Sensky et al.16 and supportive COGNITIVE BEHAVIOUR THERAPY 277 counselling in Tarrier et al.15) performed much • Acute antibiotic treatment which kills off better than expected, although never better than the bacteria causing the disease (intensive CBT. Therefore, the choice of a comparison psychological treatment which changes a key group is extremely important. If psychological factor in the psychological make-up of the therapy is compared to treatment-as-usual (TAU), individual, e.g. cognitive behaviour therapy for which includes less individual attention than the panic disorder22). psychological therapy, its effectiveness may be • Acute treatment of symptom exacerbations due to shared common themes of psychological and maintenance treatment, e.g. asthma treat- therapy not to ingredients of a particular therapy. ment with steroids followed by maintenance Tarrier et al.10 investigated the effects of expecta- with Salbutamol and/or sodium chromoglycate tion of therapeutic benefit by the use of a demand (CBT for chronic depression23). and counter-demand manipulation. Half of the • Prophylactic treatment for malaria (e.g. de- participants were told that they should expect briefing treatments for possible post-traumatic therapeutic benefit to accrue with their progress stress disorder24). through treatment (demand condition) and the other half were told that they should expect ben- Psychological therapy often sets itself the same efit but that it would not be apparent until after target as treatment using antibiotics, with an acute the end of treatment and post-treatment assess- phase followed by a follow-up during which ment (counter-demand condition). This manipu- there is no active treatment. This protocol mainly lation of expectancies had no effect on clinical resulted from the lack of specialist input in the measures, suggesting that at least the anticipa- health services, making it imperative to ration tion of treatment benefit was not influential in services. It also follows a set of expectations this patient group. that come from the behavioural tradition in the Alternatively, as psychological therapies in- treatment of psychological problems and recent clude specific attributes in common it may be CBT interventions for anxiety disorders where wrong to conclude, in a comparison of two types interventions are brief and the effects durable. of psychological therapy, that CBT is not the For example, Figure 18.1 shows the effects of best form of therapy when the two treatments imipramine and CBT for panic disorder from a do not differ significantly from each other. trial by Clark et al.22 In their trial the drug and There is always the danger that the study will the psychological treatment had similar effects at be underpowered to demonstrate an advantage the end of treatment, but psychological treatment of CBT when the non-specific control group had a more permanent effect and the differences does better than expected. However, CBT may between the two treatments were significant at be significantly better than TAU, whereas the follow-up. The improvement was predicted by alternative may not give such an advantage. the change in cognitions following treatment. In When the health services have to decide which of other words, the psychological treatment changed several forms of psychological therapy to choose a maintenance factor for the disorder. to add to their therapeutic armantarium, selecting This expectation of intensive treatment pro- CBT would be their best choice. ducing durable gains may not be appropriate for schizophrenia as it is a relapsing condition ACUTE CARE, MAINTENANCE THERAPY that, in some cases, may have a deteriorating AND DURABLE EFFECTS course. Furthermore, residual symptoms may be Schizophrenia is most often a chronic relapsing present between episodes of exacerbation. Resid- condition. If we take the metaphors from treat- ual positive symptoms at discharge are a risk ments with medication then psychological ther- factor for relapse.5 Many maintenance and causal apy could be provided in a number of differ- factors have been proposed and are in multiple ent ways. domains, such as social, biological as well as 278 TEXTBOOK OF CLINICAL TRIALS
45
40
35 CBT 30 Imipramine
25
20 baseline post- 15 month treatment follow-up Source: Reproduced from Clark et al.,22 with permission.
Figure 18.1. Psychological treatment for panic disorder psychological. It is possible that CBT could have as time passes, although it would not be clear a successful effect on one of these factors but to the research team how this latter improve- fail later when other factors become crucial in the ment came about. This poses the question of progress of the disorder. This would be shown as how do we explain increases in effect size a successful outcome at post-treatment but a lack post-therapy which cannot be explained merely of durability of gains at follow-up. However, the by the loss to follow-up of those people for usual interpretation of this pattern of results is whom the therapy conferred hardly any benefit that the effect on the disorder was only tempo- at all. rary. If gains were ‘only temporary’ in a group Trials of acute CBT have shown significant of patients who were chosen because their symp- effects of therapy mostly at the cessation of toms were ‘residual’ then even this ‘temporary’ treatment but always after a follow-up period. gain would be welcome. This set of results, rather Figure 18.2 provides data from Gould et al.25 of than dismissing the treatment effect, actually gen- the effect sizes of seven trials calculated from the erates a further question–how do we maintain the following equation: gains made during treatment when treatment is withdrawn? Effect size = (Mt − Mc)/SDc The therapeutic protocol adopted for schizo- phrenia with medication is to provide medication where Mt is the mean of the treatment group, intensively at the acute stage that is followed by Mc is the mean of the control group and this is maintenance treatment at lower dose of similar divided by the standard deviation of the control drugs. It may be that psychological treatment group of participants. needs to be provided in an equivalent way. The mean effect size for the trials studied An alternative mechanism and pattern of by Gould et al. is 0.65 (95% CI 0.56–0.71). results for CBT could be improvement in one This average is called a medium to large effect factor, such as self-esteem, which then allows fur- size according to Cohen26 (p. 40). Patients con- ther improvements in other factors to occur, such tinued to improve over the follow-up period as increased social support through the exten- with the combined effect size cited by Gould sion of a support network by increased social et al.25 of 0.93 for the four studies reporting a contact. This would produce an improvement at follow-up period. These results are encouraging post-treatment and even greater gains at follow- given that schizophrenia is a relapsing condi- up. However, it would appear that CBT was tion where life events and other stressors may not only durable but conferred greater benefits trigger new episodes of illness. However, the COGNITIVE BEHAVIOUR THERAPY 279 interpretation of the results of individual trials should include the number assessed for eligibility has since changed, mainly because the accepted for the trial, reasons for exclusion, who was standards for trials have changed. Several trials randomised and what happened to them prior to that make up this figure are methodologically final assessment and analysis of the trial results. weak with difficulties in random assignment, These standards on reporting, by implication, blindness of ratings, adequate outcome assess- provide strong recommendations to researchers ment and problematic or unsophisticated analyses about what they need to consider and action (see below). when designing and managing a trial. Table 18.1 contains a list of those points of the design or FAIRNESS AND CHANGING STANDARDS analysis that can seriously bias the interpretation IN TRIAL DESIGN AND REPORTING of the results. The majority of the current CBT trials do Standards have changed and what was reported not conform to the reporting guidelines as set in papers a number of years ago would have out in here. For some trials the significant been adequate and satisfactory for the times. discrepancies between Table 18.1 and the trial However, there are now clear guidelines on may lead to a biased interpretation of the results. how trials should be reported, formalised in For example, in Drury et al.12,13 it is not clear the CONSORT Statement.27,28 CONSORT is a what specific therapy is provided as a variety checklist and flow diagram that were designed of different components were being tested at to improve the quality of reports of randomised the same time. In Garety et al.11 the participants controlled trials. The checklist gives detailed were not randomly allocated to treatment groups. instruction on describing the study’s method and Kuipers et al.14 had no blind assessment of design, assignment and randomisation, masking treatment outcomes. Sensky et al.16 recruited by (blinding), participant flow and follow-up, and repeatedly canvassing local services for referrals. analysis. The flow diagram provides readers with However, the current meta-analyses do show a clear picture of the progress of all participants that despite these methodological difficulties in the trial, from the time they are referred to there seem to be significant changes in overall the trial until the end of their involvement. It symptoms following treatment with CBT.
2 Post-treatment effect size Follow-up effect size 1. 5 (where available)
1 Studies cited 1. Drury et al, 1996 2. Garety et al, 1994 3. Kuipers et al, 1997, 1998 0. 5 4. Milton et al, 1978 5. Sensky et al, 2000 6. Tarrier et al, 1993 7. Tarrier et al, 1998, 1999 0 1234567
Source: Reproduced from Gould et al.,25 with permission from Elsevier.
Figure 18.2. Effect sizes of CBT trails 280 TEXTBOOK OF CLINICAL TRIALS
Table 18.1. Items that should be included in reports of randomised trials Heading Subheading Descriptor
Title Identify the study as a randomised trial Abstract Use a structured format Introduction State prospectively defined hypothesis, clinical objectives, and planned subgroup or covariate analyses Methods Protocol Describe Planned study population with inclusion or exclusion criteria Planned interventions: their nature, content and timing Primary and secondary outcome measure(s) and the minimum important difference(s), and indicate how the target sample size was estimated Reasons for statistical analyses chosen, and whether these were completed on an intention-to-treat basis Mechanisms for maintaining intervention quality, adherence to protocol and assessment of fidelity Prospectively defined stopping rules (if warranted) Assignment Describe Randomisation (e.g. individual, cluster, geographic) Allocation schedule method Method of allocation concealment Masking (blinding) Describe Mechanism for maintaining blind and allocation schedule control Evidence for successful blinding Results Participant flow and Provide a trial profile summarising participant flow, numbers and follow-up timing of randomisation assignment, interventions, and measurements for each randomised group Analysis State estimated effect of intervention on primary and secondary outcome measures, including a point estimate and measure of precision (confidence interval) State results in absolute numbers when feasible (for example, 10/20, not 50%) Present summary data and appropriate descriptive and interferential statistics in sufficient detail to permit alternative analyses and replication Describe prognostic variables by treatment group and any attempt to adjust Describe protocol deviations Discussion State specific interpretations of study findings, including sources of bias and imprecision (internal validity) and discussion of external validity, including appropriate quantitative measures when possible State general interpretation of the data in light of the available evidence
Source: Modified from Begg et al.27 and Moher et al.28
EVIDENCED-BASED PRACTICE implemented in routine settings both in the UK29 and abroad.30,31 There is considerable current enthusiasm for Evidence-based practice is the delivery of evidence-based health care in general and mental interventions for which there is strong scientific health care, but there is a current debate on evidence that they improve relevant patient how evidence-based practice can be consistently outcomes. Although the type of scientific COGNITIVE BEHAVIOUR THERAPY 281 evidence does vary, the gold standard for treat- environments, usually by sophisticated university ment outcome is the RCT. Where several trials based research teams, and often involve highly exist they can be considered together through expert therapists. For CBT trials the outcomes meta-analysis. Knowledge concerning evidence- of interest are a reduction in overall psychotic based practice accrues through the accumulating symptoms, reductions in relapse or reduced rates results of efficacy and effectiveness studies. on admission to hospital, reduced psychopathol- Thus the purpose of evidence-based practice is ogy and improvements in functioning. These tri- (a) to ensure that the wealth of research evidence als may also include various control groups and informs clinical practice so that those who are process measures to help understand why the in receipt of treatment will receive the treatment treatments work. An effectiveness trial attempts that is the best available and represents the to more closely resemble the real world of rou- current knowledge base, and (b) to ensure that tine services, inclusion criteria are wider so planning and policy is determined by empirical the sample treated is more heterogeneous and evidence, for those purchasing services to be able includes the atypical patients, and the therapists to make informed choices and for those receiving are recruited from the routine services. The mea- services to be empowered by such knowledge. sured outcomes are reduced to the minimum and Furthermore, the establishment of an evidence- tend to be gross measures that are clinically sig- based practice knowledge base of what works nificant such as relapse or hospital admission; allows the practice of mental health services health economic measures to assess cost are also and individual clinicians to be compared to desirable. In special cases an equivalence trial the evidence base. This increases accountability may be designed, in which a new treatment is and establishes guidelines for good practice and expected to match the clinical efficacy of an improves the quality of mental health services. established treatment but may have other bene- Limitations in evidence also set the research fits, for example in terms of acceptability or cost. agenda for the future. These trials have special methodological features There are, however, critics of the colla- that distinguish them from simple comparative tion of data for evidence-based practice. This trials.32 mainly focuses around the use of specific meta- analytic techniques that have very limited entry criteria. The Cochrane database, for example, PARTICIPANT SAMPLES provides valuable searches and evaluations of Recruitment Bias randomised control trials with strict criteria for entry. Although the evidence may be strong for Figure 18.3 shows how the patient flow in a study a particular practice it may be based on a very should be described. The box of particular inter- small number of studies. The main criteria for est is the one at the very top that describes those exclusion are the lack of randomisation of the who have been assessed for eligibility. In order participants within the trial and the lack of data to prevent bias in recruitment the best method on all those participants who entered the trial. for ascertaining samples of patients for a trial Although clearly the results of such trials should is to recruit them from a cohort of patients in be less weighted in the final evaluation, such contact with a service that covers a geographic information may be valuable when few other data area (as in Tarrier et al.15 and Lewis et al.17). are available. This ensures that the people who are in the trial TRIALS METHODOLOGY AND AIMS do represent those who have the disorder. In the UK it is largely assumed that those patients Efficacy trials are devised to test whether the ther- with schizophrenia in contact with the services apy has an effect overall on the outcomes of inter- will represent those with the disorder requiring est. They are carried out in relatively controlled clinical intervention. For example, Tarrier et al.15 282 TEXTBOOK OF CLINICAL TRIALS
Assesssed for eligibility (n = …)
Excluded (n = …)
Not meeting inclusion criteria (n = …)
Refused to participate Enrollment (n = …)
other reasons (n = …)
Randomised (n = …)
Allocated to intervention Allocated to intervention (n = …) (n = …)
Received allocated Received allocated intervention (n = …) intervention (n = …)
Allocation Did not receive allocated Did not receive allocated intervention intervention (give reasons) (n = …) (give reasons) (n = …)
Lost to follow up (n = …) Lost to follow up (n = …) (give reasons) (give reasons)
Discontinued intervention Discontinued intervention
Follow up Follow (n = …) (give reasons) (n = …) (give reasons)
Analysed (n = …) Analysed (n = …)
Excluded from analysis Excluded from analysis = = Analysis (give reasons) (n …) (give reasons) (n …)
Figure 18.3. Participant recruitment and flow screened all patients who might have a diagno- sample was representative (see Sensky et al.16). sis of schizophrenia in a number of NHS trusts, What a comparison of samples allows is just selected those who achieved predetermined crite- that–if the samples are similar then the results ria and examined their notes further. All putative of the trials can be usefully compared, but this candidates following this procedure were inter- information cannot be used as evidence of sample viewed to ascertain whether they satisfied the representativeness. entry criteria. This method has been used as a Convenience samples which recruit from clinic gold standard and other trials of CBT have used attenders or, even more problematically, patients the data from Tarrier et al.15 to compare with referred to the project by their clinicians are at their study sample in order to conclude that their risk of selection bias. The referrer may only select COGNITIVE BEHAVIOUR THERAPY 283 those possible participants who they view as Bentall et al.35) would prefer the adoption of good candidates for the treatment or conversely symptomatic entry criteria as schizophrenia is a patients who are difficult or treatment refractory. term covering a group of people with a wide vari- Recruitment of referred patients is unfortunately ety of abnormal experiences. So some trials have the norm.14,16 Even though it may be possible as their entry criteria a specific symptom experi- to compare the recruited sample to the whole enced as distressing rather than membership of a population of patients who may be eligible in single diagnostic category.11,36 However, even in terms of socio-demographic and clinical service these studies some patients were excluded on the contact, this will not be enough data to rule basis of diagnosis because of not fulfilling other out a systematic bias. In the treatment of panic criteria (see below), and it was certainly the view disorder, Klein33,34 has argued vigorously that in of one of the authors (TW) that in feasibility stud- comparisons of psychotherapy verses drug, a pill ies of group CBT some patients with diagnoses placebo–drug comparison is necessary to ensure other than schizophrenia, e.g. personality disor- that the sample is not atypical since the efficacy der or bipolar affective disorder, did not respond of the drug (in this case imipramine) is well in similar ways to the patients with diagnoses established. This is largely an argument about of schizophrenia or schizoaffective disorder. Cur- how representative or typical any sample is, given rent CBT studies have generally included patients a reliance on convenience samples. from the schizophrenia spectrum and it is cer- tainly the view of some CBT therapists that the Selection type of therapy offered to people with bipolar affective disorder is different from that designed There are a number of different factors that need for schizophrenia.37 to be considered as part of the recruitment pro- Even when diagnosis is used there are too cess such as service delivery system, academic many different systems to choose from (e.g. support, socio-economic status of the area and clinical case note diagnosis, research diagnoses geography (urban, suburban and rural area). It is (RDC), DSMIV, ICD10, etc.). The choice of a unlikely that these will have a specific interac- different system will change the characteristics tion with the outcome from therapy, but as these of the sample. For instance, if people are drawn factors will affect the generalisation of the trial on RDC criteria they will not necessarily be as results it is probably important for the sample to chronic as those fulfilling the DSMIV criteria. represent a variety. But ethnicity and cultural mix may potentially Exclusion Criteria affect therapy outcomes. As we know very little about how to target psychological therapy to As well as criteria for inclusion into trials most different cultural groups, it seems reasonable studies also exclude people on the basis of to start investigating a new treatment with a specific issues. In trials of psychological therapy culturally homogeneous group and in later trials for psychosis one usual criterion is that the modify to accommodate cultural diversity, if people who enter the trial are those whose such modification would be a requirement of symptoms have remained despite adequate doses effectiveness in cultural subgroups. of medication. The group chosen on this basis is extremely chronic and refractory and provides Diagnosis an extremely stringent test of the efficacy of psychological treatment. In psychological therapies, especially in the field A further thorny issue is that of co-morbid of psychosis, there has been a dilemma about substance abuse. Most studies will exclude indi- whether to adopt medical diagnosis as entry cri- viduals when the abuse is severe, but the cri- teria to studies. Some clinical psychologists (e.g. teria for severity are rarely set out clearly so 284 TEXTBOOK OF CLINICAL TRIALS that it is impossible to compare between trials. might never have achieved any change following Patients who are recruited from inner city areas therapy. Clearly if a treatment produces high levels are unlikely to be free of recreational drug use. of drop-outs this might imply something about the A small consumption of cannabis may not affect acceptability of treatment. A precise description of the therapy efficacy, but it is not clear whether drop-outs is required but, from the trials submitted any use of class A drugs affects the therapeu- so far this is missing in all but a few cases. tic effect of psychological treatment. A more More research on drop-outs is clearly required. recent trial has been designed to test the effi- But in the area of mental health in particular, the cacy of CBT and family intervention to treat research is difficult, if not impossible, to carry dual diagnosis patients (those diagnosed as suf- out. The new guidelines on research governance40 fering from schizophrenia and substance abuse) do not allow for the harassing of people who in which the substance abuse is thought to have dropped out of trials for their reasons increase the risk of poor outcomes in the primary for dropping out or for data on their current disorder.18 In this case severe substance abuse health status. However, it is not only of interest was an entry criterion. academically, as it provides some information on Again some people may have a co-morbid the veracity of the theory underlying the disorder, organic condition such as epilepsy that may war- but also essential to inform the health services. rant exclusion, although most trials again would For example, Tarrier et al.41 reported that patients evaluate whether the organic condition is primar- who dropped out of treatment tended to be male, ily responsible for the symptoms of the disorder unemployed and unskilled, single, with a low which they are trying to alleviate. Deteriorating level of educational attainment and a low pre- brain disorders such as Alzheimer’s disease may morbid IQ. They had a lengthy duration of illness be a reasonable exclusion criterion as CBT relies although at the time of discontinuation they were on the carry-over of changes in one session to not severely ill and functioned at a reasonable subsequent sessions. Similarly, people who have level. They were likely to be paranoid but not learning disabilities may also have some difficul- suspicious of the therapist. They were unlikely ties with CBT as it is currently devised, although to be grandiose. They did not understand the therapists have extended treatment for depression rationale for therapy or the potential for benefit to the learning disabilities field. Current trials also but feared it could make them worse. do not support the idea that lower IQ prevents It is not clear whether it is appropriate or 38 therapeutic changes. But all current trials do ethical to collect personal information that is have a lower cut-off for IQ, usually around 65. kept for routine monitoring purposes for a person who has dropped out of a trial. This information Drop-out or Lost to Follow-up may consist of health service contacts kept on health care databases such as case notes as Two main issues affect the inferences about the well as information from third parties. For trials trial results. The first is the effect of those people involving people with severe disorders, third- who drop out of the therapy and the second party information from key workers is nearly is those people who are lost to contact at any always included as part of the measurement of stage of the trial. Different systems of dealing outcome. The lack of data on drop-out may affect with drop-outs can be adopted. Some systems the relevance and benefit of the trial results to the assume that the person would not have changed wider community. It may therefore be unethical at all since leaving the trial (LOCF), but this not to collect as much of it as possible. The approach has its problems.39 But assuming that the interpretation of new legal rights such as the new group who drop out would have performed in the UK Human Rights Act should make the position same way as those who remained also produces of researchers clearer, but it is also possible that difficulties. Drop-outs may be those people who the idiosyncratic interpretations made by local COGNITIVE BEHAVIOUR THERAPY 285
Research Ethics committees will lead to further ways. Recruitment may be affected by what ser- confusion in this already complicated area. vices are already available and who has access to them. For example, recruitment is likely to be different if there is free universally available PLATFORM AND ORDNANCE health care provided by a service committed to research and development. A large proportion of A naval military analogy between the vehicle the population will use this service and poten- of delivery, the platform (e.g. battleship, frigate, tially be available for recruitment and eligible for etc.) and what is delivered, the ordnance (e.g. the trial. This is essentially what happens in the shell, missile, etc.) is helpful in understanding UK National Health Service. This case is very the difference between service organisation and different when health care is provided, funded by 42 therapy. In terms of this analogy the platform reimbursements in a fragmented manner to cer- would be aspects of the mental health service, tain groups of the population by private services such as assertive outreach, case management who are unlikely to be committed to research. In and so on; whereas ordnance would consist this case the proportion of the population avail- of different types of therapeutic intervention, able for recruitment will be much reduced and such as CBT and family interventions. This biased towards certain subgroups who may, for distinction is useful in clarifying what is being whatever reasons, have no access to private care. tested. For example a trial of different service Here recruitment is likely to be highly selective organisations (platforms) would be the UK 700 and potentially biased. The provision of different 43 trial, in which 708 psychotic patients in four services to different income groups or other popu- centres were randomly assigned to standard or lation subgroups mitigates against representative- intensive case management. In this trial the only ness of trial populations in the USA, Australia specific difference between the two trial limbs and some European countries44,45 and compro- was the number of patients the case managers mises the value of such trials. had in their case loads. No investigation was made about the therapeutic input that the case RANDOMISATION managers implemented. The results indicated that there was no advantage in clinical or social Purpose outcomes of intensive case management. In contrast there are examples of therapy trials in To give an equivalent chance of a recruit being which a comparison was made between CBT in any of the groups in the trial design. Some plus routine care, supportive counselling plus researchers think that one of the purposes is routine care, and routine care alone for chronic to balance the groups on every factor that patients15 and acute patients.17 In these trials may be relevant to the treatment response, but patients are recruited across a number of sites so purely random allocation will not provide such that variations in routine care and service delivery matching. If there is strong evidence that a are accommodated. It may be questioned whether particular factor may affect the outcome then trials of services are of much value if they do this should be included as a factor in the not include effective therapies. A battleship is analysis. In the past researchers have said they unlikely to perform well in a naval engagement have provided evidence of the equivalence of firing a bow and arrow! their samples in analyses of pre-treatment group comparisons and on the basis of finding no statistically significant differences on factors BACKGROUND SERVICES pertinent to the treatment response have then not included these factors in their analyses of The background mental health services and their therapeutic outcome. The current advice is not accessibility may affect trials in a number of to carry out such pre-treatment comparisons 286 TEXTBOOK OF CLINICAL TRIALS but to include pertinent factors in the outcome to significant treatment effects. Although one assessment. However, there is a need for a clear recent study has not found this to be true,38 description of the people who dropped out of trialists should consider this factor in their treatment in relation to those who remained as analysis if only to counter such criticisms. this may bias the interpretation of the results. • Interactions with other treatments need to be For studies of psychological treatments in psy- considered where these are variables within the chosis the most relevant factors are listed below. group of patients entered into the trial. The most pertinent for CBT studies is the issue of • The chronicity of the illness, measured in the use of medication. Medication is now often months or years since first diagnosis, is likely divided into two main types, typical antipsy- to affect treatment outcome because it is well chotics which have been available for a number known in the field of psychiatry that those of years and atypical antipsychotic medications with longer illnesses may be less likely to which have become available recently. Most change. Tarrier et al.15 found that this modestly published CBT studies were carried out before predicted treatment outcome in an intensive the wide availability of these newer medica- CBT trial. tions. However, medication was not a predic- • tor of outcome in Garety et al.38 or Tarrier The duration of untreated psychosis (DUP) is 15 47 the time spent experiencing symptoms prior et al. Kuipers et al. comment that in their to the diagnosis and treatment of the disor- CBT group, because symptomatic improve- der. Several studies suggest that DUP affects ment would be achieved, these patients would the success of other treatments, particularly be less likely to be prescribed clozapine (an medication, and it may be that this is also a atypical antipsychotic) and would generally be factor in the efficacy of psychological disor- prescribed lower doses of medication. These predictions, although only a trend towards sig- ders. nificant, were shown in their data. Pinto et al.48 • The severity of the symptoms has been shown chose their sample on the basis of a failure of at to affect treatment outcome.15,38 In the Lon- least two medications to reduce positive symp- don–East Anglia trial the best outcomes fol- toms. In their study, which was not method- lowing CBT were found for those people who ologically strong, the effect size was extremely said they were not absolutely certain that their large with the combined effect of CBT and delusions were true. The effect of this same the new medication (clozapine) producing an factor was also alluded to several years ago effect size of 2.18.49 This suggests that med- in a small trial by Watt et al.46 But, although ication should be taken into account in ran- the outcome at post-treatment was affected by domisation or at least in subgroup analyses. this factor there was no measurable influence at follow-up nine months after the end of Entry to the study may be stratified if the 47 therapy. variable has a known interaction with treatment • 25 Gender was investigated by Gould et al. in or the variable can be used as a covariate in the their meta-analysis of CBT trials. They found analyses. Being clear about which variables may no relationship between effect size and the interact with psychological treatment is essential proportion of men in the trial but as they at the outset of the trial because the trial must be themselves point out no data were available for defined and have a sufficiently large sample to be the specific outcomes for men and women that adequately powered to test for these effects. precludes a definitive evaluation. However, young men are usually thought to have a poorer Details outcomeandaremorelikelytodropout.41 • Intellectual status has been suggested by Details of the process of randomisation must be critics of psychological therapy to be a bar supplied in the paper. For instance in Kuipers COGNITIVE BEHAVIOUR THERAPY 287 et al.14 a randomised permuted blocks allocation an increased estimate of benefit of 34%. This was adopted in each centre which contributed replicated a similar earlier finding of Schulz participants to the trial. Other studies with et al.51 who also reported exaggerated treatment multiple centres (e.g. Sensky et al.16) randomised efficacy of 30–40% in trials with inadequate participants at each centre, this they then argued concealment. There is good evidence that the allows them to control for within-centre effects poorer the trial methodology the better, and more and allows them to test between-centre effects of inflated, the treatment results obtained. treatment efficacy. The assessment and treatment procedures must be separate and independent, in other words the person who carries out the assessment should be Blindness different from the person who delivers the treat- In clinical trials blindness usually refers to ment. This is not always the case in published tri- 52,53 two aspects of the trial. The first relates to als, for example Brooker et al. trained mental the allocation of participants to the different health nurses in family intervention and assessed treatment limbs so that the allocation process is the effectiveness of the intervention by having independent and concealed from those involved the nurses perform the assessments. It could in the assessment or treatment. This prevents be argued that any problem of bias could be people from choosing who to put into the trial avoided in cases such as these by performing on the basis of the patient’s own preference, the assessments through patient self-report. How- resulting in more enthusiastic people being in ever, this does not address the problem of social the treatment arm, or the research worker’s approval that may introduce bias where patients preference which may result in those with more give results they think their therapist would want favourable prognoses being allocated to the to receive. experimental treatment. Independent assessors and therapists will not The second use of the word blindness relates ensure that assessors remain naive to treatment to the concealment of which treatment the allocation. Accidental knowledge of allocation participant received from those involved with can be minimised by using separate adminis- assessment, especially of outcome. This is an trative procedures and geographically separating extremely important issue and one that is difficult therapists from assessors in terms of office loca- to ensure. The aim is to prevent any bias, tion and administrative procedure. This should conscious or otherwise, entering the assessment prevent assessors bumping into patients about process through knowledge of which treatment to receive therapy and such similar accidents. the participant received. For example, knowledge Patient allocation should be multiply coded so that the hypothesis to be tested was that CBT that learning of one patient’s allocation does not would be better that treatment-as-usual because break the whole trial code. Patients should be previous studies had demonstrated this may bias instructed not to reveal any detail of their treat- an assessor to rating the patient as more improved ment or who has treated them to the assessors at if they knew the patient had received CBT. the start of the trial and before each assessment. The importance of adequate concealment was It is unlikely that this will be fool-proof but it demonstrated in a study by Moher et al.50 who will minimise revelations. See Tarrier et al.15 for examined the quality of concealment in treatment further details of efforts to maintain blindness in trials in circulatory and digestive disease, mental a clinical trial of CBT. health, obstetrics and childbirth. The examined Opinions differ as to whether verification trials had already passed a number of quality of maintenance of blindness is desirable. It is assessments and been included in a number possible to ask assessors to guess the allocation of meta-analyses. They found that trials with of trial participants. This can be used as evidence poorer quality blinding were associated with of successful blindness.15,54 Assessors should not 288 TEXTBOOK OF CLINICAL TRIALS be informed that they will be asked to guess as of identifiers. To be successful interviews need this would prime them to the task. Guessing is to follow a protocol as to the procedure of less likely to be successful when there are more the interview so that adequate and sufficient than two treatment groups. With two groups, or information is available to make ratings. In most even more than two, an assessor could adopt a studies of CBT in general and for psychosis strategy that patients who improved should be in particular, the process for blind allocation is in the experimental treatment group because this rarely described, for example Kuipers et al.14 would be in line with the study hypothesis. If the In contrast, Sensky et al.16 and Tarrier et al.15 trial had been successful this strategy would have both describe the method for ensuring blindness been correct and the assessor would most likely and the maintenance of allocation of subjects have guessed right in many cases although for the to groups. wrong reasons. This would not be an indication that the assessor knew of the treatment allocation PROTOCOL and was hence biased in their assessment but that they knew who improved which aided them Design Protocol in guessing group allocation. The problem for the trial investigators here would be that their There are various ways of testing whether assessors appear not to have been blind. If the a particular treatment is efficacious but the assessors were not able to guess correctly using accepted method is to compare the treatment with this strategy it would probably mean that the a placebo control that allows for a comparison experimental treatment had not been effective and of client expectations of improvement during the trial was a failure anyway. Having assessors therapy with the active ingredient itself. We guess allocation holds the investigators hostage to have discussed above the importance of these fortune, although with multiple treatment groups non-specific factors in psychological treatments. it can be effective in demonstrating blindness. Social contact, social support and the modelling Even if assessors do maintain blindness to of interpersonal behaviour are all an integral treatment allocation they will still be aware of the part of psychological therapy. Tests of the timing of the assessment, pre-, post-treatment or effectiveness of individual CBT have used a follow-up. Thus the only way to ensure blindness variety of designs. Table 18.2 below gives the of both treatment allocation and assessment time outline of the main recent trials. is to separate the gathering of information from There are a variety of designs that will allow its rating. Thus all assessment interviews should the examination of both the effectiveness and be audio-taped independently of their rating specificity of the effect of CBT above the and rating should be carried out by a different effects found for psychological interventions in assessor who is unaware of the allocation or general in this group. The results show that assessment. This would also allow the audio- there are significant effects over treatment-as- tapes to be edited of any accidental revelation usual. There is also one study16 that shows a
Table 18.2. Designs for randomised control trials of CBT for psychosis
Comparison with Comparison with ‘placebo’ Comparison with alternative therapy and treatment-as-usual ‘placebo’ therapy treatment-as-usual
Garety et al.11 Sensky et al.16 Tarrier et al.10 Kuipers et al.14,47 Drury et al.12,13 Tarrier et al.15,73 Barrowclough et al.18 SOCRATES17 COGNITIVE BEHAVIOUR THERAPY 289 difference between CBT and a ‘placebo’ therapy aspects of therapy to be assessed. However, (befriending), but only at follow-up. However, the scale allocates equal weighting to all items. Tarrier et al.15 found a significant difference There is, as yet, no empirical evidence to sup- between CBT and TAU but no overall difference port such equal weightings, and it may well between the two therapies at any stage of the be, for example, that ‘agenda setting’ is less study.55 Analysis of specific symptoms found important than the ‘choice of intervention’. that there was a significant advantage of CBT • Does the progression of therapy cover all the over supportive counselling in the treatment of key topics of the manual? This requires that hallucinations.56 several sessions of therapy are recorded at different times and that the content of these Treatment Protocol is scored for the timing of the interventions in the treatment programme. This is probably It is essential to have a clear and unambiguous the most important part of rating CBT trials treatment protocol for psychological treatments. because although it can be clear how many However, even when a manual is available it sessions are provided to a patient it is not is much harder to evaluate exactly whether the clear whether the content given is the same. protocol has been adhered to. In treatments with CBT researchers (personal communications, medication this process is relatively easy as the including one of the authors, NT) observe dose and timing of the treatment can be veri- that some patients are able to travel through fied using simple procedures. For psychological the whole manual whereas others cover much treatment the verification process relies on taped less. So although the therapy duration may be interviews of treatment sessions that are then equivalent, exposure to the complete protocol rated later for fidelity with the treatment pro- can be different. This dosage of treatment may tocol. However, there are several problems that be an important factor in defining treatment may interfere with this process. Firstly the patient outcome as some patients are clearly getting must agree to the recording of the session and in more treatment than others. However, the some studies, e.g. Chadwick et al.,57 the patients number of treatment sessions is not related to refused to have any sessions taped. Once taped effect size as presented in Gould et al.25 sessions have been collected the independent rat- ing must answer a number of questions: Progression through therapy may be affected by a • Does the session represent the treatment to number of factors such as the level of disturbance be provided? In other words is it possible to or cognitive impairment of the patient. Therapists differentiate the experimental treatment from will also differ in their ability to progress therapy the placebo treatment. Sensky et al.16 and and their skills in different aspects–determined 15 by skills, training, profession, trial provided Tarrier et al. were able to show that their 59 independent assessors was able to assign 100% training (see Tarrier et al. ). So far in CBT and 97% of the tapes rated to the appropriate trials these factors have not been investigated in treatment arm. any detail. • Is the experimental treatment manual being adhered to? This requires that the researchers INDIVIDUALISED TREATMENTS have a specific rating scale that will allow the rating of key areas of their treatment. Had- Turkington and Siddle60 claim that a case formu- dock et al.58 has developed a rating scale (the lation is essential to treating psychotic patients Cognitive Therapy Scale for Psychosis–CTS- successfully. We do not disagree that a case Psy) to assess quality of therapy. This allows formulation is desirable but with psychotic assessment of general (e.g. interpersonal effec- patients it is not always possible and a purely tiveness) and specific (e.g. guided discovery) symptomatic approach has to be adopted on 290 TEXTBOOK OF CLINICAL TRIALS occasions.61 In spite of our support for case for- the numbers in each group required to show a sig- mulation the evidence from general adult mental nificant difference for the five outcome measures health that a case formulation is necessary is poor would be between 25 and in excess of 15 000, and the results equivocal. Schulte et al.62 treated with a median of 800 patients in each treatment a mixed group of 120 phobic patients with a stan- group. The issue of case formulation-based treat- dardised treatment and individualised behaviour ment versus protocol-based treatment is unlikely therapy based on functional behaviour analysis. to be resolved by a direct head-to-head compari- They also included a yoked control group in son, which would be too large and costly. which treatment was based not on the individ- ual’s assessment but that of the yoked patient. TREATMENT COMPONENTS The standardised treatment group showed the CBT treatments also differ in other ways from most improvement and patients who acted as each other. Although they have a basic set of yoked controls improved as well as the other 63 ingredients the emphasis may be placed differ- patients. Similarly Emmelkamp et al. allocated ently. For example, the different emphases on 22 obsessive–compulsive patients to either tailor- behavioural activation and cognitive schema in made cognitive behavioural therapy or standard- the changes in thinking thought to be the cause ised exposure in vivo therapy. There were no of the treatment effect. Changing behaviour can significant differences between groups but the have an effect on thinking as studies of CBT for = group sizes were small (n 11 in each group). panic disorder have discovered.22 Patients in the 64 Jacobson et al. treated 30 distressed marital Clark study were treated with behavioural acti- couples with either a manual-based version of vation programmes that are embedded in CBT. marital therapy or a clinically flexible version They showed that the prediction of outcome was of the same treatment in which treatment plans dependent on one main factor, cognitions about were individually based and the number of treat- their bodily sensations. The behavioural experi- ment sessions were not specified. Both treat- ments seemed to have an effect on cognition. But, ments resulted in significant improvements at other groups in the field of psychosis emphasise post-treatment but at six-month follow-up the more distal stimuli such as the developmental couples treated with the structured format were path of the delusion. The particular component more likely to have deteriorated and flexibly of CBT that accounts for most of the variance in treated couples were more likely to have main- outcome has not yet been differentiated and these tained their treatment gains. There appears little more subtle differences are not used in meta- advantage of case formulation-based treatment analyses of the treatment studies. Figure 18.4 over a standard package. This result is not sur- shows the effect sizes taken from Gould et al.25 prising given the sample sizes of these stud- on a scale devised by ourselves on the amount of ies. Standard treatment programmes are effective behavioural activation that the treatment empha- for a wide range of psychological disorders and sises. As can be seen from the graph the effect even if an individualised treatment was superior size is increased when more behavioural activa- the difference in effect sizes will most proba- tion is included. It may be that a simple change in bly be small and the sample size to significantly behaviour via a behavioural experiment may pro- demonstrate such a difference would necessarily vide enough evidence to reduce delusional con- be large. Therefore, the studies that have been viction. For instance Birchwood and Chadwick66 done are massively underpowered. To substanti- suggest that the perceived powerfulness of an ate this Tarrier and Calam65 have estimated the auditory hallucination directly predicts the dis- sample sizes required to show significant differ- tress experienced. Adopting one successful cop- ences with 80% power and 0.05 significance level ing strategy may provide enough evidence to based on the data provided in the published report reduce the perceived power of an auditory hal- of Emmelkamp et al.63 On the basis of their data lucination and increase the amount of perceived COGNITIVE BEHAVIOUR THERAPY 291
0.8
et al. 0.7 0.6 0.5 0.4 Effect size
(2001) 0.3 0.2 0.1 0 Effect size from Gould size Effect Sensky Kuipers Garety Tarrier Tarrier 00 97 94 98 93 Amount of behavioural activation in CBT
Figure 18.4. How much ‘B’ in CBT for psychosis control the patient has over their symptoms. the effects of behavioural activation alone (see Wykes et al.67 provides some evidence of this Figure 18.5). relationship in a waiting list control trial of group CBT for auditory hallucinations. If this is true OUTCOMES then successful behavioural experiments should always be included and should predict successful Current thinking from methodologists on trials treatment outcome. As yet no study has attempted in psychiatry is that designs should be simplified to measure this process. and that outcome measures should be kept to a Turkington and Siddle60 maintain that cogni- minimum. In particular the use of rating scales tive therapy with psychotic populations will result should be restricted to ‘one or two which are in long-term improvement because it involves best understood’ (Johnson,69 p. 229). Follow-up schema change whereas cognitive behaviour ther- should be carried out on ‘few occasions rather apy (as carried out by Tarrier et al.15) will result than many’ and entry criteria should be as broad in short-term change only. They go on to say as possible.69 This advice is rather in conflict that ‘schema change seems vital in terms of with that given previously which suggests that, the durability of any achieved benefits’ (p. 302). in the treatment of psychosis, multiple outcomes However, there is little evidence that schema are which reflect the complex nature of the disorder causative in psychosis or that change is important and its effects should be used70 and that data on for treatment effects (see Figure 18.4). The direct the process of therapy are essential. The multiple transport of Beck’s model of depression to psy- effects of therapy may, but not necessarily, have chosis in the absence of evidence for its explana- a common outcome in relapse prevention or total tory value in this population has been criticised.68 symptoms. For instance cognitive therapy should The evidence for the effect of schema work on affect cognition, CBT should affect cognitions outcome is anyway sparse even in the area of and behaviour, and family interventions should depression for which it was designed. Jacobson affect families’ interactions. All these effects randomised 150 people to three treatment arms: could in some way change the outcome of the (i) behavioural activation, (ii) behavioural activa- disorder but via multiple pathways. Because of tion and work on dysfunctional thoughts, and these multiple effects, multiple outcomes may be (iii) total CBT with work on cognitive schema. recommended in order to differentiate the route The results of this trial showed no differences to effectiveness in explanatory trials. in outcome either at post-treatment or follow- All current trials measure overall symptoms. up between the three groups. The effect of the However, they all do so in different ways and other components of CBT was no different to even when the measure appears to be a standard 292 TEXTBOOK OF CLINICAL TRIALS
35 30 Behavioural activation 25 20 Activation and Automatic thoughts 15 10 Complete CBT, including schema 5 work 0 Pre Post 6 month follow-up Source: Data from Jacobsen et al.75
Figure 18.5. Evidence of the effect of schema work in depression measure (e.g. BPRS71) it is often adapted. For CONCLUSIONS instance Kuipers et al.14 added items to the BPRS making it difficult to compare their results with Because psychological therapy has no product others adopting the conventional version of the champion as found in the drug industry, prag- same instrument. The use of non-standard instru- matic trials are needed initially to convince peo- ments to measure awareness of stigma, coping ple that therapy is worthwhile. However, these skills, etc. prevents comparisons being made and need to be followed by explanatory trials that does need replication with standardised rating can establish the specificity of the treatment. Cur- scales with no known psychometric properties. rently the trials in the field of psychosis have In all medical trials a statistically significant mainly been pragmatic and these have shown that difference in outcome may provide little benefit the therapy is worthwhile with improvements in to the patients. What needs to be defined for trials positive and negative symptoms at post-treatment of CBT for psychosis is the clinical significance and follow-up in some studies. However, the tri- of outcomes. This was alluded to earlier in this als that have been designed to test the specificity chapter. Clinically significant outcomes may be of treatment have not been so successful. Very reductions in the distress associated with the few differences emerge between CBT for psy- disorder. Currently clinical significance is defined chosis and alternative therapies are shown at post- as the sorts of improvements that are achieved treatment. Of the two studies with long follow- in drug trials–20% change in symptoms. This ups one showed an advantage for CBT over the may be a low threshold for what could be alternative therapy16 but the other showed equiv- achieved through psychological therapy. Many alent benefits of both therapies (CBT and sup- trials of psychological therapy adopt only a portive counselling) over routine care.55,73 One statistically significant test of effectiveness, but resolution of this conflict may be the nature of Tarrier et al.15 and Sensky et al.16 adopt a 50% the therapy chosen. In the Sensky et al.16 study change criterion for their measures, although the comparison condition was befriending which Kuipers et al.14 use the lower threshold of 20%. is described as a therapy with equivalent amounts Where trials use such a threshold of achieving of therapist contact but where the content was a clinical significance or not, comparisons can be discussion focusing on neutral topics such as hob- made by comparing the Numbers Needed to Treat bies, sports and current affairs where the therapist (NNT), which represents the number of patients is instructed to be empathic but non-directive. that need to be treated to achieve one clinically Even this therapy was associated with reductions significant outcome.72 in symptoms at the end of therapy that may be COGNITIVE BEHAVIOUR THERAPY 293 a testament to the paucity of the social contact will be possible to establish routes to change that and lack of warm relationships of people with can then inform the development of therapy. It is continuing active psychosis. However, the effects only by these later developments that it will be of this therapy were not sustained to a follow-up possible to develop training and therefore provide where the patients in the comparison therapy con- larger numbers of people with psychosis with dition actually got worse. In the Tarrier et al.15 effective psychological therapy. study the comparison therapy chosen was sup- portive counselling in which the therapist tried to achieve a supportive relationship which fos- REFERENCES tered rapport and provided emotional support. This therapy proved to be successful in reduc- 1. Wiersma D, Nienhuis FJ, Sloof CJ, Geil R. Nat- ural course of schizophrenic disorders: a 15 year ing symptoms over the course of the therapy and follow-up of a Dutch incidence cohort. Schizophr follow-up. This result suggests that some of the Bull (1998) 24: 75–85. essential ingredients of CBT encompassed within 2. British Psychological Society. Recent Advances in the counselling framework may be either shared Understanding Mental Illness and Psychotic Expe- with supportive counselling or are as effective riences. Leicester: British Psychological Society (2000). as these other ingredients. It is of course possi- 3. Fowler D, Garety P, Kuipers E. Cognitive therapy ble that within the model of schizophrenia which for psychosis: formulation, treatment, effects and encompasses a vulnerability stress model that the service implications. J Mental Health (1998) 7: two forms of therapy may work via different 123–33. pathways. Supportive counselling may work by 4. Tarrier N, Barrowclough C, Vaughn C, Bam- rah JS, Porceddu K, Watts S, Freeman HL. The emphasising self-esteem through rapport within community management of schizophrenia: a con- the therapeutic relationship. Unlike befriending trolled trial of a behavioural intervention with fam- this support produces more stable changes that ilies. Br J Psychiat (1988) 153: 532–42. are durable to follow-up. However, it should be 5. Cunningham-Owens DG, Carroll A, Fattah S, Clyde Z, Coffey I, Johnstone EC. A randomised, noted that supportive counselling did significantly controlled trial of a brief educational package worse at treating hallucinations when compared for schizophrenic outpatients. Acta Psychiat Scand to CBT on this symptom alone.56 (2001) 103: 362–9. Currently there is no evidence that cognitive 6. Beck AT. Successful outpatient psychotherapy behaviour therapy works via a cognitive system, with a schizophrenic with delusion based on borrowed guilt. Psychiatry (1952) 15: 305–12. although training in coping skills has been shown 7. Tarrier N. Psychological treatment of schizo- 74 to improve coping. None of the studies have phrenic symptoms. In: Kavanagh D, ed., Schizo- yet produced analyses showing that the cognitive phrenia: An Overview and Practical Handbook. change established during therapy is the key London: Chapman & Hall (1992). to later improvement. In fact few have even 8. Tarrier N. Modification and management of resid- ual psychotic symptoms. In: Birchwood M, Tar- provided analyses which test this possibility. rier N, eds, Psychological Approaches to Schizo- Because of the complex nature of the aetiology phrenia: Innovations in Assessment, Treatment and of psychosis with its multiple causal processes it Services. Chichester: John Wiley & Sons (1992). may be impossible to identify a single route to 9. Haddock G, Tarrier N, Spaulding W, Yusupoff L, change. There may be many idiosyncratic routes Kinney C, McCarthy E. Individual cognitive- behaviour therapy in the treatment of schizophre- that will only be established in large trials with nia: a review. Clin Psychol Rev (1998) 18: several hundred patients included. One such trial 821–38. is taking place in Manchester, the Socrates trial.17 10. Tarrier N, Beckett R, Harwood S, Baker A, In this trial hundreds of acute patients who are at Yusupoff L, Ugarteburu I. A trial of two cognitive behavioural methods of treating drug-resistant the beginning of their illness have been provided residual psychotic symptoms in schizophrenic with therapy and followed up over a long period. patients. I: Outcome. Br J Psychiat (1993) 162: It is only through these sorts of studies that it 524–32. 294 TEXTBOOK OF CLINICAL TRIALS
11. Garety P, Kuipers E, Fowler D, Chamberlain F, empirically, “all must have prizes”. Psychol Bull Dunn G. Cognitive-behaviour therapy for drug (1997) 122: 203–15. resistant psychosis. Br J Med Psychol (1994) 67: 22. Clark DM, Salkovskis PM, Hackmann A, Mid- 259–71. dleton H, Anastasiades P, Gelder M. A compar- 12. Drury V, Birchwood M, Cochrane R, Macmil- ison of cognitive therapy, applied relaxation and lan F. Cognitive therapy and recovery from acute imipramine in the treatment of panic disorder. Br psychosis. I. Impact on psychotic symptoms. Br J J Psychiat (1994) 164: 759–69. Psychiat (1996) 169: 593–601. 23. Paykel ES, Scott J, Teasdale JD, Johnson AL, 13. Drury V, Birchwood M, Cochrane R, Macmil- Garland A, Moore R, Jenaway A, Cornwall PL, lan F. Cognitive therapy and recovery from acute Hayhurst H, Abbott R, Pope M. Prevention of psychosis. II. Impact on recovery time. Br J Psy- relapse in residual depression by cognitive ther- chiat (1996) 169: 602–7. apy. Arch Gen Psychiat (1999) 56: 829–35. 14. Kuipers E, Garety P, Fowler D, Dunn G, Beb- 24. Yammey G. Psychologists question debriefing for bington P, Freeman D, Hadley C. London–East traumatised employees. Br Med J (2000) 320: 140. Anglia randomised controlled trial of cognitive- 25. Gould RA, Mueser KT, Bolton E, Mays V, behavioural therapy for psychosis: 1. Effects of Goff D. Cognitive therapy for psychosis in schizo- the treatment phase. Br J Psychiat (1997) 171: phrenia: an effect size analysis. Schizophr Res 319–27. (2001) 48: 335–42. 15. Tarrier N, Yusupoff L, Kinney C, McCarthy E, 26. Cohen J. Statistical Power Analysis for the Behav- Gledhill A, Haddock G, Morris J. A randomised ioral Sciences. Hillsdale, NJ: Lawrence Erlbaum controlled trial of intensive cognitive behaviour (1988). therapy for patients with chronic schizophrenia. 27. Begg C, Cho M, Eastwood S, Horton R, Moher D, Br Med J (1998) 317: 303–7. Olkin I, Pitkin R, Rennie D, Schulz KF, Simelk D, 16. Sensky T, Turkington D, Kingdon D, Scott JL, Stroup DF. Improving the quality of reporting Scott J, Siddle R, O’Carroll M, Barnes TRE. A of randomized controlled trials: The CONSORT randomised controlled trial of cognitive-behavioral Statement. J Am Med Assoc (1996) 276: 637–9. therapy for persistent symptoms in schizophrenia 28. Moher D, Schulz K, Altman D. The CONSORT resistant to medication. Arch Gen Psychiat (2000) Statement: revised recommendations for improv- 57: 165–72. ing the quality of reports of parallel group 17. Lewis SW, Tarrier N, Haddock G, Bentall R, randomisation. J Am Med Assoc (2001) 285: Kinderman P, Kingdon D, Siddle R, Drake R, Everitt J, Leadley K, Benn A, Glazebrook K, 1987–91. Haley C, Akhtar S, Davies L, Palmer S, Fara- 29. Geddes J, Reynolds S, Streiner D, Szatmari P, gher B, Dunn G. Randomised controlled trial of Haynes B. Evidence-based practice in mental cognitive-behaviour therapy in early schizophre- health. Evidence-Based Mental Health (1998) 1: nia: acute phase outcomes. Br J Psychiat (2002) 4–5. 181(Suppl 43): 91–7. 30. Barlow DH. Health care policy, psychotherapy 18. Barrowclough C, Haddock G, Tarrier N, Lewis S, research and the future of psychotherapy. Am Moring J, O’Brian R, Schofield N, McGovern J. Psychol (1996) 51: 1050–8. Randomised controlled trial of motivational inter- 31. Drake RE, Goldman HH, Leff HS, Lehman AF, viewing and cognitive behavioural intervention Dixon L, Mueser KT, Torrey WC. Implementing for schizophrenia patients with associated drug evidence-based practices in routine mental service or alcohol misuse. Am J Psychiat (2001) 158: settings. Psychiat Serv (2001) 52: 179–82. 1706–13. 32. Jones B, Jarvis P, Lewis JA, Ebbutt AF. Trials to 19. Barrowclough C, Tarrier N, Sellwood W, Quinn J, assess equivalence: the importance of rigorous Mainwaring J, Lewis S. A randomised controlled methods. Br Med J (1996) 313: 36–9. effectiveness trial of a needs based psychosocial 33. Klein DF. Preventing hung juries about therapy intervention service for carers of schizophrenic studies. J Consult Clin Psychol (1996) 64: 81–7. patients. Br J Psychiat (1999) 174: 506–11. 34. Klein DF. Discussion of ‘methodological contro- 20. Horvath A, Symonds D. Relationship between versies in the treatment of panic disorders’. Behav working alliance and outcomes in psychotherapy: Res Ther (1996) 34: 849–53. a meta-analysis. J Counsel Psychol (1991) 38: 35. Bentall RP, Jackson HF, Pilgrim D. Abandoning 139–49. the concept of schizophrenia: some implications of 21. Wampold BE, Mondin GW, Moody M, Stich F, validity arguments for psychological research into Benson K, Ahn H. A meta-analysis of outcome psychotic phenomena. Br J Clin Psychol (1989) studies comparing bona fide psychotherapies: 27: 303–24. COGNITIVE BEHAVIOUR THERAPY 295
36. Wykes T, Parr A-M, Landau S. Group treatment 50. Moher D, Pham B, Jones A, Cook DJ, Jadad AR, for auditory hallucinations – a waiting list con- Moher M, Tugwell P, Klassen TP. Does quality trolled study. Br J Psychiat (1999) 175: 180–5. of reports of randomised trials affect estimates of 37. Lam D, Bright J, Jones S, Hayward P, Schuck N, intervention efficacy reported in meta-analyses? Chisholm D, Sham P. Cognitive therapy for bipo- Lancet (1998) 352: 609–13. lar illness – a pilot study of relapse prevention. 51. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Cognit Ther Res (2000) 24: 503–20. Empirical evidence of bias: dimensions of method- 38. Garety P, Fowler D, Kuipers E, Freeman D, ological quality associated with estimates of treat- Dunn G, Bebbington P, Hadley C, Jones S. Lon- ment effect in controlled trials. J Am Med Assoc don–East Anglia randomised controlled trial of (1995) 280: 178–80. cognitive-behavioural therapy for psychosis. II 52. Brooker C, Tarrier N, Barrowclough C, Butter- Predictors of outcome. Br J Psychiat (1997) 171: worth A, Goldberg D. Training community psy- 319–27. chiatric nurses for psychosocial intervention: 39. Everitt BS. The analysis of longitudinal data: report of a pilot study. Br J Psychiat (1992) 160: beyond MANOVA. Br J Psychiat (1998) 172: 836–44. 7–10. 53. Brooker C, Falloon I, Butterworth A, Goldberg D, 40. Department of Health. Research Governance Graham-Hole V, Hillier V. The outcome of train- Framework for Health and Social Care. Depart- ing community psychiatric nurses to deliver psy- ment of Health, UK (2001). chosocial intervention. Br J Psychiat (1994) 165: 41. Tarrier N, Yusupoff L, McCarthy E, Kinney C, 222–30. Wittkowski A. Some reason why patients suffer- 54. Basoglu M, Marks I, Livanou M, Swinson R. ing from chronic schizophrenia fail to continue in Double-blindness procedure, rater blindness, and psychological treatment. Behav Cognit Psychother ratings of outcome. Arch Gen Psychiat (1997) 54: (1998) 26: 177–81. 744–8. 42. Marshall M. A systematic review of the effec- 55. Tarrier N, Kinney C, McCarthy E, Humphreys L, tiveness of acute day hospitals versus admission. Wittkowski A, Morris J. Two year follow-up of Paper presented at The Making Mental Health Ser- cognitive-behaviour therapy and supportive coun- vices Effective: Now and Tomorrow Conference, selling in the treatment of persistent positive 7–8 September 2000, Manchester, UK (2000). symptoms in chronic schizophrenia. J Consult Clin 43. Burns T, Creed F, Fahy T, Thompson S, Tyrer P, Psychol (2000) 68: 917–22. White I, & the UK 700 Group. Intensive versus standard case management for severe psychotic 56. Tarrier N, Kinney C, McCarthy E, Wittkowski A, illness: a randomised trial. Lancet (1999) 353: Yusupoff L, Gledhill A, Morris J. Are some types 2185–9. of psychotic symptoms more responsive to CBT? 44. Wykes T, Tarrier N, Lewis S. Outcome and Inno- Behav Cognit Psychother (2001) 29: 45–55. vation in the Psychological Treatment of Schizo- 57. Chadwick P, Sambrooke S, Rasch S, Davies E. phrenia. Chichester: John Wiley & Sons Challenging the omnipotence of voices: group (1998). cognitive behavior therapy for voices. Behav Res 45. Tarrier N. What can be learned from clinical Ther (2000) 38: 993–1003. trials? Reply to Devilly and Foa (2001). J Consult 58. Haddock G, Devane S, Bradshaw T, McGovern J, Clin Psychol (2001) 69: 117–18. Tarrier N, Kinderman P, Baguley I, Lancashire S, 46. Watt F, Powell G, Austin S. Modification of delu- Harris N. An investigation into the psychometric sional beliefs. Br J Med Psychol (1983) 46: properties of the Cognitive Therapy scale for 359–63. psychosis (CTS-Psy). Behav Cognit Psychother 47. Kuipers E, Fowler D, Garety P, Chisholm D, (2001) 29: 93–106. Freeman D, Dunn G, Bebbington P, Hadley C. 59. Tarrier N, Barrowclough C, Haddock G, McGov- London–East Anglia randomised controlled trial ern J. The dissemination of innovative cognitive- of cognitive-behavioural therapy for psychosis. III behavioural treatments for schizophrenia. JMental Follow-up and economic evaluation at 18 months. Health (1999) 8: 569–82. Br J Psychiat (1998) 171: 319–27. 60. Turkington D, Siddle R. Improving understanding 48. Pinto A, La Pia S, Mennella R, Giorgio D, DeSi- and coping in people with schizophrenia by mone L. Cognitive-behavioral therapy and cloza- changing attitudes. Psychiat Rehab Skills (2000) pine for clients with treatment-refractory schizo- 4: 300–20. phrenia. Psychiat Serv (1999) 50: 901–4. 61. Tarrier N, Haddock G. Cognitive-behaviour ther- 49. Rector NA, Beck AT. Cognitive-behavioral ther- apy for the treatment of schizophrenia: a case apy for schizophrenia: an empirical review. JNerv formulation approach. In: Hofman SG, Thompson Ment Dis (2001) 189: 278–87. MC, eds, Handbook of Psychological Treatments 296 TEXTBOOK OF CLINICAL TRIALS
for Severe Mental Disorders. New York: Guilford 69. Johnson T. Clinical trials in psychiatry: back- Press (2001). ground and statistical perspective. Stat Meth Med 62. Schulte D, Kunzel R, Pepping G, Schulte- Res (1998) 7: 209–34. Bahrenberg T. Tailor-made versus standardised 70. Barrowclough C, Tarrier N. Psychosocial inter- therapy of phobic patients. Adv Behav Res Ther ventions with families and their effects on the (1992) 14: 67–92. course of schizophrenia: a review. Psychol Med 63. Emmelkamp PMG, Bouman TK, Blaauw E. Indi- (1984) 14: 629–42. vidualised versus standardised therapy: a com- 71. Ventura J, Green M, Shaner A, Liberman R. parative evaluation with obsessive–compulsive Training and quality assurance with the Brief patients. Clin Psychol Psychother (1994) 1: Psychiatric Rating Scale: the drift busters. Int J 95–100. Meth Psychiat Res (1993) 3: 221–44. 64. Jacobson NS, Schmaling KB, Holtzworth-Munroe 72. Sackett DL, Richardson WS, Rosenberg W, A, Katt JL, Wood LF, Follette VM. Research- Haynes RB. Evidence-Based Medicine. London: structured vs clinically flexible versions of social Churchill Livingstone (1998). learning-based marital therapy. Behav Res Ther 73. Tarrier N, Wittkowski A, Kinney C, McCarthy E, (1989) 27: 173–80. Morris J, Humphreys L. The durability of the 65. Tarrier N, Calam R. New developments in effects of cognitive behaviour therapy in the cognitive-behavioural case formulation. Epidemi- treatment of chronic schizophrenia: twelve months ological, systemic and social context: an integra- follow-up. Br J Psychiat (1999) 174: 500–4. tive approach. Cognit Behav Psychother (2002) 74. Tarrier N, Sharpe L, Beckett R, Harwood S, 30: 311–28. Baker A, Yusupoff L. A trial of two cognitive 66. Birchwood M, Chadwick P. The omnipotence of behavioural methods of treating drug-resistant voices: testing the validity of cognitive model. residual psychotic symptoms in schizophrenic Psychol Med (1997) 21: 1345–53. patients. II: Treatment-specific changes in coping 67. Wykes T, Reeder C, Corner J, Williams C, and problem-solving skills. Social Psychiat Everitt B. The effects of neurocognitive remedi- Psychiat Epidemiol (1993) 28: 5–10. ation on executive processing in patients with 75. Jacobson NS, Dobson KS, Truax PA, Addis ME, schizophrenia. Schizophr Bull (1999) 25: 291–308. Koerner K, Gollan JK, Gortner E, Prince SE. J 68. Patience DA. Cognitive-behaviour therapy for Consult Clin Psychol (1996) 64: 295–304. schizophrenia. Br J Psychiat (1994) 165: 266–7. 19 Depression
GRAHAM DUNN Biostatistics Group, School of Epidemiology & Health Sciences, University of Manchester, Manchester, UK
INTRODUCTION analysing means, or do not account for individual differences, are simply revealing their ignorance The purpose of the present chapter is to explore of statistics and of recent developments in statis- the pitfalls in and challenges to the valid estima- tical methodology’.2 tion of the effects of psychotherapies. Although A belief in the fundamental role of randomi- ‘depression’ appears in the title, the discussion sation, however, does not imply that the na¨ıve will be relevant to psychological treatments for implementation of RCTs in outcome research any mental illness (including psychotic disorders cannot lead to some invalid or unsafe conclu- such as bipolar depression and schizophrenia). sions. The design of the trial and statistical Most of the illustrative examples, however, will analysis of the results have to be appropriate refer to the treatment of depression. Right from to the setting. Psychotherapy involves complex the start, despite pointing out all of its poten- interactions between patient and therapist and tial problems, I will assume that by far the best sometimes (as in group therapy) involves the way of trying to estimate treatment effects is via interaction of a group of patients with each other the use of a randomised controlled trial (RCT). as well as with their therapist. It is not as sim- I have little sympathy with the increasingly pop- ple as taking a tablet! A psychotherapy trial is ular view that we can learn much of real value likely to be far more complex, both in its imple- about treatment effects from systematically col- mentation and in the analysis and interpretation lected outcome data in routine clinical practice of the subsequent results, than most drug trials. (see Dunn1 for a critique of this view). Nor do I There are also far more opportunities for invalid have any sympathy for the often-heard view that inferences concerning treatment effects. RCTs and the use of statistical methods are inap- First, and often primarily, we are concerned propriate vehicles for the evaluation of something with internal validity: the valid estimation of a as complex as psychotherapy. As we have writ- causal effect of treatment from the data actually ten elsewhere: ‘Clinicians who claim that statis- collected (given set of patients, therapists, treat- tical methods are inappropriate for the evaluation ment centres and psychotherapy actually deliv- of psychotherapies because they are limited to ered). Are the group differences we see the causal
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 298 TEXTBOOK OF CLINICAL TRIALS effects of treatment? Or can they be explained by In the above paragraph we are trying to find a other factors? Later we may be concerned with way to estimate what may be called the causal external validity: generalisation of the inferred effect of a treatment. The essence of the solution causal effects to other patients, therapists, treat- to the problem is a comparison. For each of ment centres and, perhaps, other forms of psy- the three patients, Mr Smith, Mrs Jones and chotherapy. To help clarify the discussion I have Mr Adams, there are two possible outcomes of made much use of Rubin’s counterfactual model the referral to see a psychotherapist. The first of causality and its use in the estimation of treat- is to receive therapy and have the severity of ment effects.3,4 The kernel of the present discus- depression measured after a given interval after sion, applied to the problems of patient choice the onset of the course of treatment. The second and non-adherence to treatment, can be found is to fail to get the offered help, but again in Dunn.5 have the severity of depression measured after the allotted time. Let the variable T represent treatment received. It has two possible values, THE CAUSAL EFFECT OF TREATMENT T = t (therapy) and T = c (no therapy). I use ‘c’ for no therapy to indicate that it can be regarded as a control condition. Let i indicate Mr Smith has suffered from severe depression, = on and off, for several years. Six months ago his the identity of the patient (i 1 for Mr Smith, 2 family doctor advised him to undergo a course for Mrs Jones and 3 for Mr Adams). Finally, let Y of psychotherapy. He accepted this advice, has T (i) indicate the final BDI score for patient i T had several what he thinks were very helpful after receiving treatment option .Therearetwo potential sessions with the psychotherapist, and now is outcomes for each of the three patients, as indicated in the following table: feeling considerably better. Let’s assume, for the sake of argument, that he has a total score of 10 points on the Beck Depression Inventory (BDI),6 Patient BDI with BDI without having started with a score of 20 six months ago. therapy therapy What is the effect of psychotherapy? How do we measure this effect? Putting it another way, Mr Smith Yt(1)Yc(1) what proportion, if any, of the drop from 20 Mrs Jones Yt(2)Yc(2) to 10 points might be attributed to the receipt Mr Adams Yt(3)Yc(3) of therapy? Or perhaps I should have written: ‘What proportion might validly be attributed to We define the causal effect of the receipt of the receipt of therapy?’ But, before we attempt therapy as the difference between the BDI to answer this question, let us consider another score for the ith patient after therapy and patient, Mrs Jones, who also suffers from chronic the corresponding BDI score after receiving no depression. Like Mr Smith, she was advised to therapy. That is, by the difference Yt(i) − Yc(i). have psychotherapy by her family doctor six It is a random variable that varies from one months ago but, for various reasons, she never patient to another. Unfortunately, it can never managed to keep any of her appointments and be observed. The obvious problem is that each has not received any help from the therapist. A patient receives one of the treatment conditions, third patient, Mr Adams, refused outright to have or the other, but not both. Either the patient anything to do with the therapist. Mrs Jones’ receives psychotherapy or he/she does not. That present BDI score is 12 and that for Mr Adams is, the ith patient provides a value for either is 15. For Mrs Jones and Mr Adams one might Yt(i) or Yc(i), but not both. Mr Smith provides ask what would have been the effect of therapy Yt(1) but not Yc(1), Mrs Jones provides Yc(2) offered if they had actually received it? What but not Yt(2), and so on. We provide a statistical might their BDI scores have been? solution to this problem in the following section, DEPRESSION 299 but we will re-emphasise the point that the causal estimates of E[Yt(i)]andE[Yc(i)], respectively. effect of psychotherapy is the comparison of the In general, however, the average of the Yt(i)’s outcome actually observed with that which would for the whole of the population (i.e. all possi- have been observed if, contrary to fact, the other ble i’s) is not thesameastheaverageofthe treatment option had been taken. Yt(i)’s for those patients who have happened to Similar arguments apply to the comparison of receive the treatment (psychotherapy). Expressed the effects of different types of psychotherapy, mathematically: or to the comparison of a specific type of psy- = | = = | = chotherapy with, for example, a psychopharma- E[Yt(i)] E[Yt(i) T t] E[Yt(i) T c] cological intervention such as a tricyclic antide- (3) pressant. The essence is always to try to get an and estimate of the difference between the patient’s E[Y (i)] = E[Y (i)|T = t] = E[Y (i)|T = c] observed response with that which would have c c c (4) been observed if the patient had received the where ‘|’ means ‘given that’. To summarise, alternative treatment. the ACE is defined by the difference between E[Yt(i)]andE[Yc(i)] but what we actually COMPARISON OF GROUP AVERAGES observe are the estimators of E[Yt(i)|T = t] and AND THE ROLE OF RANDOMISATION E[Yc(i)|T = c]. How do we ensure that our observed averages are also valid estimators of Now let’s assume that we have access to a E[Yt(i)]andE[Yc(i)]? If we are able to do this large population of eligible patients – the target then we have replaced an impossible-to-observe population about which we wish to draw causal causal effect on an individual patient with a inferences about the value of psychotherapy possible-to-estimate average of the causal effects or counselling. And let us concentrate on the for our target population.4 average causal effect (ACE)3,4 of the therapy If both E[Yt(i)] = E[Yt(i)|T = t] and E[Yc(i)] for this target population. The average for the = E[Yc(i)|T = c] then the potential outcomes population is called an expected value in statistics (both the Yt(i)’s and the Yc(i)’s) are statistically and the ACE can therefore be written as: independent of the mechanism of assigning (or choosing) treatment options. Otherwise, they are ACE = E[Y (i) − Y (i)] (1) t c not. If either the patient’s family doctor, or the where the expectation E[·] is over all values of i. patient himself, or both, were to decide which From the mathematical properties of expectations treatment option to choose then it is almost (averages) it follows that: certain that this choice will not be statistically independent of the potential outcome. This is the
ACE = E[Yt(i)] − E[Yc(i)] (2) familiar problem of confounding. The difference in observed outcomes may arise from the fact that This simple formula shows us that information the patients with the best (or worst) prognosis, on different patients can be used to estimate on average, might be the ones that opt for E[Yt(i)]andE[Yc(i)] separately and the differ- therapy. The observed outcomes in this situation ence between these two expectations (averages) might tell us something about the selection can be used to estimate the average of the dif- mechanism (treatment choice) but are not very ferences (i.e. the ACE). We can observe the informative about the causal effect of therapy. Yt(i)’s in patients receiving therapy and we can Knowing the values of all of the prognostic also observe the Yc(i)’s for those in the con- variables, together with a little knowledge of trol condition. All that we need is to be sure experimental design, might lead us to match that the observed averages for the treated (ther- or stratify the patients prior to estimation of apy) and untreated (control) patients are unbiased the treatment effects. But we cannot guarantee 300 TEXTBOOK OF CLINICAL TRIALS that we are aware of all possible confounders. randomly allocated groups) provides us with an There is always the possibility that we have not estimate of the causal effect of offering treatment thought of, or forgotten, something that is vitally (i.e. randomisation) rather than the effect of important. Although we may be able to convince actually receiving it. It is a valid estimator of a ourselves that we have not missed an important causal effect but many investigators (particularly confounder, the only way we can ensure that we psychotherapists!) might claim that it is an can convince a sceptical reviewer is to allocate estimator of the wrong effect. As an estimator treatment options randomly. Random allocation of the causal effect of receiving therapy the ensures that both E[Yt(i)] = E[Yt(i)|T = t] and ITT estimate is likely to be biased. However, E[Yc(i)] = E[Yc(i)|T = c] providing that t and many other investigators might be convinced caretheallocated treatments (not, necessarily, that this is the estimator of real interest – it those actually received). Randomisation is the measures the effect of a decision to treat in only sure way of coping with all confounders, a given way and is therefore vitally important and it copes with them irrespective of whether for people involved in making these decisions we are aware of them or not. Randomisation (or, at the very least, those paying for them!). does not guarantee that treatment groups will It is the standard approach to the analysis of be exactly comparable in any given comparison, drug trials and that usually expected by the but it does ensure that on average there will be regulatory authorities such as the US Food and comparability. Our conclusion is that if we wish Drug Administration (FDA) and the UK National to be sure that we are estimating the desired ACE Institute for Clinical Excellence (NICE). we need an RCT. An essential corollary of randomisation is that CHOICE OF, AND ADHERENCE TO, AN we obtain outcome data on all of the randomised APPROPRIATE FORM OF PSYCHOTHERAPY patients and that we calculate our group averages from the patients as they were randomised and What constitutes the active treatment for our not according to whether they actually received required comparison? There are several com- or adhered to the treatment option that they were mon forms of psychotherapy that are regu- allocated to. This is the intention-to-treat (ITT) larly used for patients with depression, includ- 7 principle (see, for example, Sheiner and Rubin ). ing behaviour therapy, cognitive behaviour ther- If we do not use ITT then the fundamental apy (CBT),8 interpersonal psychotherapy (IPT),9 assumptions concerning our estimates of the brief dynamic psychotherapy.10 Usually the ther- causal effect of treatment (ACE) no longer hold. apy involves the treatment of individual patients, Loss to follow-up (i.e. a failure of the patient but there is also the possibility of working with to provide outcome data) is a major threat to all groups of patients with similar problems, or with RCTs but, in this part of the discussion we will the patient and his or her family. For a gen- simplify matters by assuming that outcome has, eral review, see for example, Scott11 or Roth and indeed, been obtained for all patients entering Fonaghy.12 If our aim is to evaluate the efficacy the trial. But what if some of the patients of one of these forms or models of treatment, choose a treatment option other than the one they or to compare its efficacy with another model were randomly assigned to? Or perhaps some of psychotherapy or even pharmacotherapy, then patients adhere to the allocated treatment much it must be self-evident that we need to be able less than others – they turn up to the occasional to describe explicitly and precisely what treat- session of therapy, for example, but not all of ment using any of the specified models actually those which had been planned. This will clearly involves; i.e. they must be standardised. Crits- dilute (attenuate) the effect we wish to estimate. Christoph and Gladis13 give two main reasons In fact, our ACE estimator (the difference for standardisation. First, from a clinical view- between the observed mean outcomes for the two point, there is a need to be able to describe what DEPRESSION 301 actually seems to work (or does not) so that good as the therapy on offer. The test therapy, for clear treatment recommendations can be made to example, might involve supportive counselling in other potential therapists. Second, from a research addition to the specific elements implied by the viewpoint, therapies need to be replicable. Stan- psychotherapeutic model, and the natural control dardisation of psychotherapies, however, is not condition would be the receipt of the same level easy, and it is a topic beyond both the scope of support in the absence of the psychotherapeutic of the present chapter and the competence of elements under test. If, however, we wish to evalu- the present writer. Briefly, it involves the cre- ate supportive counselling itself, then we still have ation of a detailed treatment manual, the selection a problem. Trialists often refer to ‘equipoise’ in and subsequent training of appropriate therapists justification of randomisation in an RCT. To main- in the use of the manual, certification of thera- tain equipoise we need to be convinced that the pists based on adherence to the treatment model, control group patients are at least provided with and continued assessment of therapist adherence the best-available routine care and that they are and competence during a clinical trial.13 Clearly, not allocated to a condition that might cause harm. when critically appraising the results of a particu- I will illustrate the choice of control groups by lar RCT, we need to be able to convince ourselves referring to a particularly well-known and influ- that the therapy has been undertaken as intended, ential psychotherapy trial. The National Institute and that the therapy as given was exactly what of Mental Health (NIMH) Treatment of Depres- it is said to be. For this we need a published sion Collaborative Research Program (TDCRP)14 treatment manual and a well-validated method of involved the use of four treatment arms. Groups 1 measuring adherence to the therapy as described and 2 received CBT and IPT, respectively. Group in the manual. 3 received pharmacotherapy with imipramine (administered double blind) together with a care package called ‘clinical management (CM)’. CHOICE OF AN APPROPRIATE Finally, Group 4 received a pill placebo (again CONTROL GROUP administered double blind) and CM. Elkin et al.14 state that ‘The CM component of both phar- Standardisation of psychotherapy might be macotherapy conditions was introduced into the thought to be a difficult problem but it is often study to ensure standard clinical care, to max- far more difficult to come up with a valid and imize compliance, and to address ethical con- convincing control condition. Crits-Christoph and cerns regarding the use of a placebo on depressed Gladis13 consider this as perhaps the single most patients. The CM component provided guide- vexing problem for research into the outcome of lines, not only for the management of medication psychotherapy. Too often, we see that researchers and side effects and review of the patient’s clini- have used ‘no treatment’ or ‘waiting-list’ controls. cal status, but also for providing the patient with Too often we see the phrase ‘routine care’ used support and encouragement and direct advice if for the control condition when, in many circum- necessary. Although specific psychotherapeutic stances it implies routine neglect. It is important interventions were proscribed (especially those that when patients are invited to take part in an that might overlap with the two psychotherapies), RCT they are convinced that they will receive ade- the CM component approximated a “minimal quate levels of advice, support and care if they are supportive therapy” condition.’ allocated to the nominal control condition. Other- The essence of the imipramine–CM and wise, why should they consent to randomisation? placebo–CM conditions was to provide a fully Otherwise, why should an ethics committee grant standardised package of clinical care, either of its approval for the trial? It is also important that which could be used as a control group for the the test psychotherapy is being compared to a care evaluation of the efficacy of the psychotherapies. package that might be regarded as potentially as So, the two major questions addressed by the 302 TEXTBOOK OF CLINICAL TRIALS
NIMH TDCRP study were ‘(1) Is there evidence to maintain blindness. The therapists themselves of the effectiveness of each of the psychothera- should not undertake the assessment of outcome. pies, as compared both with the standard refer- One should always bear in mind, however, that ence treatment of imipramine–CM and with the irrespective of who carries out the assessment, placebo plus CM (PLA–CM) control condition? there is always the possibility of subjective biases (2) Are there any differences in the effective- in the assessments. The Hamilton Rating Scale ness of the two psychotherapies?’ These ques- for Depression, HRSD15 and the Beck Depres- tions emphasise the comparative nature of this sion Inventory, BDI6 are the two most commonly and any other well-designed RCT. When one asks used measures of depressive symptoms in RCTs questions about the effectiveness of psychother- for the treatment of depression. In fact, they are apy one should always add the rider ‘relative frequently both used within the same RCT to to what?’ A valid and well-standardised control assess different aspects of symptomology. The condition is as vital to the comparison as is the HRDS is a clinician-rated measure, based on standardised package of therapy. Crits-Christoph an interview with the patient, which gives more and Gladis,13 referring to the TDCRP trial, com- weight to the ‘biological’ or somatic symptoms of ment that whilst the placebo–CM is perhaps not depression, whilst the BDI is a patient-completed the ideal control condition for psychotherapy, it questionnaire which concentrates more on the serves a practical function. That is, if a spe- cognitive aspects. There have been suggestions cific psychotherapy can do no better than the that different forms of therapy (drugs as opposed placebo–CM control should the psychotherapy to psychological treatments, for example) might be pursued as a treatment option? Beware of have a differential effect on these outcome mea- authors who make claims about the improvement sures (drugs doing better according to the HRDS of patients in a particular treatment group with- and the BDI favouring CBT, for example). The out reference to that in other comparison groups. expected treatment group by outcome measure Roth and Fonaghy’s12 (p. 64) comment that the interaction needs to be specified (and preferably small differences between the four TDCRP trial published) as part of the trial protocol and, if it groups, in terms of their outcome, is due to the is regarded as being important, the trial needs to unexpectedly good outcome under placebo–CM be powered accordingly. In reality, it is hard to (explained by the fact that it contains non-specific imagine a convincing justification for a trial of elements of psychotherapy) seems to be missing the size and expense needed for such a test. the point. What about missing outcome assessments? Drop-outs and other sources of missing data lead to real problems for the valid estimation of the CHOICE OF ASSESSMENT METHOD effects of treatment. A detailed discussion of this AND OUTCOME MEASURES topic is beyond the scope of the present chapter, but it must be stressed that the only effective way It is very difficult to see how one could possi- of dealing with missing data is to ensure that bly design a so-called double blind RCT in the there are none. Investigators should make every field of psychotherapy evaluation. The patients effort to ensure that outcome data (however brief) are likely to know what is going on, unless they are collected on all of the patients randomised, have been deceived by their therapists, and it irrespective of their subsequent treatment history. would be rather bizarre if the therapists were In particular, data collection should not be unaware of what treatment was being offered! abandoned simply because the patient has not Blind assessment by a third party (a clinician or taken up the offer of therapy or has not adhered research worker not involved in the provision of to the prescribed course of treatment.16,17 There therapy or clinical support) is often the preferred is no logical reason why patients should refuse option, but even here it is frequently difficult to be assessed even though they have decided DEPRESSION 303 that therapy is not for them, although, in practice, These include training and experience of the the two types of protocol violation are likely to therapist, degree of adherence to the therapeutic go together. model (use of a manual, for example) and the But what if you have drop-outs and hap- capacity to develop a therapeutic alliance with hazardly missing data? What is the best way the patient (see, for example, Crits-Christoph of dealing with them in the statistical analy- et al.22 and Roth and Fonaghy12). In a multi- sis? If we use a na¨ıve complete-case analysis centre RCT there are also likely to be differences (that is, base the inferences concerning causal in the effectiveness of the collaborating clinical effects on those patients with complete data) centres. Therapists in some centres may have then we are likely to have two problems: lack considerably more experience in the use of a of statistical efficiency (low statistical power) given treatment approach than in others, reflected, and bias. Bias may be caused, for example, by for example, in their degree of adherence to a the drop-outs not occurring completely at ran- given therapeutic model. In addition, if patients dom. Ideally, analyses should be available on all are treated as groups rather than individually available data (and should include where pos- there are also likely to be differences between sible all patients randomised to the competing groups arising not only from the characteristics treatment arms) and should compensate for the of the patients and of the therapist but also from observed patterns of drop-out. Possibilities for 23 dealing with the missing values include impu- interactions between the patients. If they get tation (ranging from rather crude and unsatisfac- on well together the group might thrive. If, on tory methods – at least from the point of view the other hand, there is a particularly disruptive of estimation – such as last-observation-carried- or difficult patient within a particular group then forward to the much more realistic stochastic the group as a whole may not do as well as it imputation methods including hot-decking and might otherwise have done. multiple imputation), the use of inverse probabil- Consider a hypothetical single-centre RCT with ity weighting and, finally, a full likelihood analy- two treatment arms. In Group A all the patients sis based on statistical models for both the miss- receive CBT individually from an internationally ing data process and for the outcome given that it respected pioneer of CBT, Dr Garner. In Group has been assessed. For further details, readers are B all the patients, again individually, receive sup- referred to a series of reviews in Statistical Meth- portive counselling (SC) from a recently-trained ods in Medical Research, 8(1), particularly the community psychiatric nurse, Mr Martin. Let’s primer on multiple imputation by Schafer.18 The assume that Dr Garner’s patients do considerably use of inverse probability weights is widespread better than those of Mr Martin. What can we infer in survey statistics but has only occasionally been about the causal effect of CBT from such a trial? used to allow for drop-outs in RCTs. The use What are the threats to the validity of the trial? Dr of this technique in longitudinal clinical trials is Garner is likely to be a very experienced, highly- 19 explained and illustrated in Everitt and Pickles. skilled and highly-motivated ‘brand champion’. Its use is also illustrated in a recent depression Mr Martin, on the other hand, lacks experience. 20 trial by Dowrick et al. Finally, a discussion of Is the observed difference due to the difference in the analysis of longitudinal data with drop-outs, abilities and experience of the two therapists, or paying particular attention to the NIMH TDCRP is it an effect of CBT? We cannot tell. The two et al 21 trial, is provided by Gibbons . effects are completely confounded in this simple design. This is a severe threat to the internal valid- CENTRE, GROUP AND THERAPIST EFFECTS ity of the trial. If, however, we believe that the observed differences are an effect of CBT, then It is clear that the outcome of psychotherapy is what? We still cannot be sure that there is not dependent upon characteristics of the therapist. some attribute of Dr Garner that enables him to 304 TEXTBOOK OF CLINICAL TRIALS be particularly successful in delivering this par- Here yijk is the outcome of the ith patient of the ticular variant of CBT. Could other clinicians use kth therapist within the jth treatment arm of the the same model and achieve the same or, at least, trial. Assuming that λj is zero in the control arm, comparable results? We do not know. This is a λj is the effect of the treatment effect for the threat to the external validity or generalisability jth arm of the trial. Each xijkp is the baseline of the trial’s findings. measurement of the pth patient characteristic In practice, many RCTs, which otherwise have (such as a demographic or other prognostic admirable design characteristics and quality con- variable, including treatment centre) and βp is the trol procedures, involve the use of only two or corresponding regression coefficient. The term three highly-skilled and experienced therapists. ujk is the average effect of the kth therapist They are often the academic clinicians who have within the jth treatment arm of the trial. It is been involved in the development or modification a random variable (i.e. randomly varying from of the therapy under evaluation. Does this inval- one therapist to another) with an assumed mean 2 idate the findings of these trials? No, but it does of zero and variance of σj . This variance may limit their generalisability. They should, perhaps, vary from one arm of the trial to another (there be regarded as the equivalent of the pharmacother- is no apriori reason why the variation between apist’s ‘Phase II’ drug trials, being a necessary therapists within different arms of the trial should preliminary to the design and conduct of a full be the same) and, in particular, if the control arm ‘Phase III’ evaluation using a large and repre- does not involve the use of therapists at all, then, sentative sample of therapists. It would be inap- 2 for the controls σj = 0 (i.e. in this situation there propriate and certainly difficult to justify a large are no therapist effects in the control group). In multi-centre trial involving large numbers of ther- the case of the possibility of one or more arms of apists without first being able to establish that the the trial involving group therapies, the statistical ‘experts’ or ‘brand champions’ are able to achieve model would be even more complex. promising results. If the latter cannot demonstrate Models such as that described in equation (5) worthwhile effects then it would be pointless to are called random effects, random regression or move on to the larger trial. If they can, however, multilevel models.27 Technical details of their we then (but only then) need to ask how well the use are beyond the scope of this chapter and therapy might work in routine clinical practice. interested readers are referred to Roberts26 for In a large multi-centre trial we need to involve an illustration of their use in the context of as many therapists as possible. Each therapist is RCTs involving therapist effects. What readers likely to be based in only one of the centres and should note, however, is that failure to allow to be delivering treatment in only one arm of the for appropriate therapist effects in the statistical trial (i.e. therapists are nested within both centres analysis (assuming that they are present in the and treatments), but it is possible for a therapist to deliver more than one of the forms of therapy data) is likely to lead to spurious statistical in a comparative trial (i.e. therapists, like centres, significance (i.e. the stated p-values will be are crossed with treatments). These designs have too low) and estimated confidence intervals or implications for the statistical analysis and for the standard errors for treatment effects that are too validity of statistical inferences based on these optimistic (i.e. smaller than they should be). A analyses.24–26 Both centre and therapist effects corollary of this is that, even when the analysis should be incorporated into an appropriate statis- is correct, a trial whose sample size has not been tical model. Using the notation of Roberts,26 such determined after allowance for the possibility of a model, for a quantitative outcome measure, for therapist effects is very likely to be underpowered example, will have the form: (too small!). This is the same problem as those faced by the designers of cluster randomised trials 28 26 yijk = α + λj + βpxijkp + ujk + εijk (5) (see, for example, Donner ). Again, Roberts p provides details of the required adjustments to DEPRESSION 305 sample size calculations, on the assumption that not the difference in psychotherapeutic approach. therapist variation is the same for each of the Consider therapeutic competence, for example. arms of the trial. The CBT therapists might, for example, be But there is more to the problem of therapist either more or less competent than their SC effects than can be solved by the technical counterparts. But how could we assess this? How device of allowing for them in an appropriate could we possibly compare the competence of statistical model. Nor is the main problem one Dr Garner as a cognitive therapist, for example, of generalising from the impact of therapists in a with that of Mr Martin as a counsellor. It is given trial to the wider community of therapists. akin to asking whether I am more competent as We started the discussion in the present section a statistician than my scientific colleague is as by comparing the outcome of CBT as delivered a laboratory worker. And moving to a crossed by Dr Garner with that of SC as delivered by design (both types of therapy being offered by Mr Martin. We pointed out that the required every therapist) does not solve our problem. If Dr treatment effect is fully confounded with the Garner, for example, were to be experienced and difference between the two therapists. Now let highly competent as both a cognitive therapist us move on to a larger trial in which each of and an interpersonal therapist we still would not the patients in Group A receives CBT from a be able to compare the competencies in the two randomly selected therapist from a team of, say approaches. Elkin29 (Reproduced with permission 50, experienced and highly competent cognitive of Oxford University Press) concluded that ‘We therapists. Each of the patients in Group B, on may never be able to truly “disentangle” the the other hand, receives SC from a randomly effects due to the therapist from those due selected therapist from a team of experienced to the therapy, because they may often be and highly competent interpersonal therapists. inherently intertwined and also very interactive We still have a problem. Again, the required with particular patient attributes.’ treatment effect (the difference between outcomes ‘The treatment conditions being compared for CBT and SC) cannot be disentangled from ... are, in actuality, “packages” of particular the difference between the average effects of therapeutic approaches and the therapists who 30 the two groups of therapists. In general, the chose and are chosen to administer them.’ The interpretation of the results of RCTs should λj in equation (5) can either be interpreted as the average of the within-Group j therapist explicitly acknowledge this fact. As well as effects or as an effect of therapy j –that very carefully defining both the treatment and is, the two interpretations are equivalent. ‘At the control conditions, authors should provide the present time, researchers and consumers of critical information about the therapists carrying psychotherapy research findings are left with a out the treatments, and the information should basic dilemma when interpreting the findings be included in the dissemination concerning 29 of studies focusing on the efficacy of specific supposedly empirically validated results. The treatments: how to disentangle the effects due latter is particularly important when we come to the therapeutic approach from those due to systematic review and/or meta-analysis of the to the particular therapists who have carried results from a disparate collection of individual out the approach. It is particularly pressing trials, and in the formulation of any subsequent when different therapists carry out each of the clinical guidelines based upon the results of treatments in a comparative outcome study.’29 these trials. The cognitive therapists and counsellors in the above hypothetical trial (or even in a real one WHAT WORKS FOR WHOM? such as the NIMH TDPRC study) might differ in lots of ways and these therapist differences may The question implies a belief that there is no be the causal effects of the treatment difference, constant treatment effect. That is, it implies that 306 TEXTBOOK OF CLINICAL TRIALS a given form of treatment has a greater effect on methodologically sound work in this area. Crits- some patients than it does on others; that the receipt Christoph and Gladis13 comment that two of the of psychotherapy A will be more beneficial for largestrandomisedclinicaltrialseverundertakento Mr Smith than the receipt of psychotherapy B, evaluatepsychotherapies(althoughnotspecifically for example, but that B might be better than A for depression) failed to provide much support for for Mrs Jones. Mr Smith has a particular attribute specified patient – treatment interactions.22,32 (presenting symptoms, clinical or family history, for example) that indicates therapy A. Mrs Jones, ESTIMATION OF CAUSAL EFFECTS on the other hand, has characteristics that indicate IN AN RCT WITH PATIENT CHOICE therapy B. In the epidemiological literature this is called ‘effect modification’ – a particularly useful Consider a hypothetical RCT in which 200 eligi- term as it should remind us that ‘causal effect’ ble depressed patients have been randomly allo- implies comparison of observed outcome with that cated to receive either counselling plus routine which would have been observed under different care (T = t) or routine care alone (T = c).For circumstances. In terms of statistical modelling simplicity, assume that all of the 100 patients (analysis of variance, or covariance, for example) allocated to routine care receive exactly that (they it will provide an example of a treatment group do not have access to counselling unless they by patient attribute interaction, where the attribute have been allocated to that treatment arm of the could be one of a potentially vast range of measures trial). Of the 100 patients offered counselling, made on the patients at or prior to randomisation. however, only 70 accept the offer. A fixed time Supposed examples of such interactions are rarely interval after randomisation (six months, say) convincing. Even if based on a valid statistical the patients’ clinical status (improved versus not analysis (i.e. a test of an appropriate two-way improved) is assessed and used as the primary interaction) they are usually ‘discovered’ as part outcome of the trial. The effects of either treat- of a post hoc ‘fishing trip’. More frequently their ment allocation or treatment actually received existence has been based on an invalid analysis. are to be estimated from the differences between average outcomes as before, the only difference All too often the investigators are looking for a being that we are averaging binary outcomes so-called ‘predictor of outcome’ by searching in (1 = improved; 0 = not improved, say) to obtain the relevant treatment group for patient attributes observed proportions. The results of this hypo- associated with good outcome. This tells us nothing thetical trial are summarised in Table 19.1 (note about effect modification – the same attributes that we have simplified the issue by assuming might lead to the better outcomes within the control there are no drop-outs). group(s). One should always remember that valid The estimate of the ITT effect is both simple inferences from an RCT involve comparison of and familiar. The proportion of those receiving the randomised groups. Here we are concerned counselling who improve is 0.70 (i.e. 70/100) with the question ‘Does the treatment effect (e.g. and the corresponding proportion for the control comparison of outcomes in Groups A and B) depend on, say, patient attribute C?’ The identity of attribute C should be clearly specified in the trial Table 19.1. Results of a hypothetical trial of coun- selling protocol, together with a prior estimate of the size of the proposed interaction. The sample size for the T = t T = c trial should then be determined such that there is Improved Total Improved Total sufficient power to detect this interaction through the use of an appropriate statistical significance Comply 60 70 test. One good candidate for attribute C might Do not comply 10 30 Overall 70 100 50 100 be patient preference,31 but there is little, if any, DEPRESSION 307 group is 0.50 (i.e. 50/100). The difference (the refuse the offer of counselling whether or not they ACE for being offered counselling) is 0.20. For are in the group actually offered counselling. The readers who prefer NTT (the reciprocal of the offer, in itself, is not beneficial. difference between the two proportions), this is Assumption 1 allows us to estimate the propor- 5 (i.e. 1/0.20). But what about estimating the tion of potential compliers in the control group. causal effect of receiving treatment? There are In our example it is 70/100. The estimated num- two commonly used, but invalid, methods of ber of non-compliers in the control group is 30 analysis–analysis per protocol or analysis as and the number of compliers is 70. treated. There is also the correct (correctness, of Assumption 2 allows us to estimate the propor- course, being vitally dependent on the validity tion (number) of patients who improve amongst of a few key assumptions) but much less famil- the non-compliers in the control group. In our iar estimator – the complier average causal effect example the number of patients who improve in (CACE).33–35 The per protocol analysis com- this group is estimated to be 10 (the propor- pares the outcome in those people in the coun- tion is 10/30). Now, there were a total of 50 selling group who actually receive counselling patients who were observed to improve in the with that in the control group (i.e. it excludes control group and therefore the estimated number patients who have violated the treatment proto- of potential compliers who improve in the control col from the analysis). Here the difference is group must be 40 (that is 50 − 10). Otherwise 60/70 − 50/100(= 0.36). The as treated analysis the numbers do not add up! So, the proportion compares outcome in those patients who receive of patients improving in the counselling group counselling with that in those who do not receive amongst those who actually receive counselling is it (all patients are included in this analysis). Here 60/70. The proportion in the corresponding con- it is 60/70 − 60/130(= 0.40). The problem with trol group (i.e. those who would have accepted both of these estimators is that it is impossible the offer) is estimated to be 40/70. The CACE to interpret them as a causal effect in the sense estimator is the difference between these two of comparing potential outcomes on the same proportions, 60/70 − 40/70(= 0.29). The corre- patient. The patient groups are not comparable. sponding NNT is 3.5. The estimated effects are merely associations, Note that in the above example the potential subject to confounding. And association, as you compliers did better than the non-compliers, all know, does not imply causality! irrespective of which treatment arm they were What about the CACE? This is an estimate allocated to. This is not unexpected and not of the difference between the outcome in the too difficult to rationalise. But now consider a compliers (i.e. those who accepted and received second, more ‘difficult’ example. The results of a the offered counselling) with that which would second hypothetical trial are shown in Table 19.2. − have been expected in the same patients if they The ITT effect (ACE) is estimated by 50/100 = had not been offered counselling. This is where 30/100( 0.20). The corresponding NTT is 5. − = we need two key assumptions.36,37 The first one is The CACE estimate is 35/70 15/70( 0.29). easy to defend for a randomised trial. The second needs a bit more careful thought. Table 19.2. Results of a second hypothetical trial of Assumption 1: the proportion of patients who counselling are potential compliers is the same in the two randomly allocated groups. This follows directly T = t T = c from the random allocation mechanism. Improved Total Improved Total Assumption 2: the proportion of potential non- compliers who improve is independent of treat- Comply 35 70 Do not comply 15 30 ment allocation. In other words, it makes no dif- Overall 50 100 30 100 ference to the outcome of a patient who would 308 TEXTBOOK OF CLINICAL TRIALS
The corresponding NNT is again 3.5. But note actually means or whether, strictly speaking, it that this time the potential compliers in the is ever possible. In the context of our example control group are doing a lot worse than the illustrating the effect of patient compliance to a non-compliers (15/70 vs 15/30). Again, this is treatment offer, is it ever ethically justified to reasonably straightforward to rationalise. The randomise and then only seek consent to treat patients who accept the offer of counselling are in the group allocated to receive therapy? This those with the worst prognosis or, equivalently, is an example of Zelen’s40 original form of the those that turn it down (or would turn it down randomised consent design. All patients in the if offered it) are those who are getting better trial are asked to provide outcome data, of course, anyway. The latter do not need treatment. But but those in the control group may never know what this should do is prompt the data analyst that they had taken part in a trial. Is this really to ask whether Assumption 2 is really justified. a serious ethical problem? This design would Might the offer of help on its own be of benefit? circumvent the potential problem of resentful And if so, of how much benefit? Or perhaps demoralisation amongst the controls. I will not those patients in the control group who would attempt to answer the question raised. have accepted the offer feel let down (resentful In an attempt to solve some of the serious prob- demoralisation) and do worse than they would lems surrounding the issue of patient preference, otherwise have done if they had known nothing Brewin and Bradley41 (see also Bradley42 and Sil- about the possible offer of help. verman and Altman43) have proposed what they In practice, we do not have to work right describe as the patient preference design.Inthis through the estimation from first principles in the design, eligible patients are told about the reasons above way. It can be shown that the required for the trial and the treatments on offer. Patients estimates can be obtained from the following who do not have a strong preference (that is, simple formula:33,38,39 they are prepared to be randomised) are entered into a conventional RCT. Those patients with ITT estimate for outcome CACE = a strong preference are offered the treatment of ITT estimate for receipt of treatment their choice. So, for the comparison of two treat- (6) ments, A and B, for example, the patient prefer- This formula applies in situations where both ence trial finishes up with four groups: those who of the measures of outcome and treatment prefer A; those without preference who are ran- received are binary (i.e. both the ITT effects are domly allocated to A; those who prefer B; those differences between proportions) or where both without preference who are randomly allocated are quantitative (i.e. both the ITT effects are to B. In the context of the present discussion, the differences between means), or one is binary and comparison of the randomly allocated groups can the other quantitative. So, for the first example lead to an ACE or CACE estimate as described = − − above, CACE (70/100 50/100)/(70/100 above. But what can the two patient preference = 0) 0.29, as before. groups provide? Merely an estimate of associa- tion. Like per protocol or as treated estimators, they do not appear to be able to provide estimates RANDOMISED CONSENT AND PATIENT of causal effects. And for this reason they can- PREFERENCE DESIGNS not be used to check the (external) validity of the estimates of causal effects provided by the ran- A serious issue in the design of RCTs concerns domised groups. Whether the difference between the amount of information given to the patient the two preference groups is the same as or com- about the aims of the trial. So-called informed pletely different from that provided within the consent is a prerequisite for most trials but it core RCT, so what? What does it tell us? That is not always obvious what ‘informed consent’ there are selection effects? The treatment effect DEPRESSION 309 may, indeed, be different in those patients with- of two people (the patient and the therapist) out a strong preference (i.e. those prepared to be and it is the involvement of these two people randomised) when compared to the rest, but the that is the essence of the complexity. Added to rest cannot provide the valid information from this are the problems of the choice of adequate which we can test whether it is true. But per- control groups (in particular, the absence of haps readers should see the results of such a trial a convincing placebo) and the impossibility and decide for themselves. An example of the of conducting a trial that is double blind. In use of a patient preference design is provided the critical appraisal of such trials we should by a recently published trial of counselling for not, perhaps, be searching for methodological depression.44,45 perfection but, instead, be aware of the pitfalls Another design possibility which, in my view, to valid inferences concerning treatment effects has much more promise is to simply ask the and temper our judgements accordingly (and this participants of a conventional RCT what their applies just as much to the trials that we have preferences are prior to randomisation. The aim been involved in as it does to the trials of other here is not to allow patient preference to influ- investigators). ence treatment received (but in the presence of This chapter has not considered systematic non-compliance this will be inevitable) but to review and meta-analysis of trials of psychother- incorporate patient preferences into the analysis apies but the authors (and appraisers) of such of the resulting data. This has been tried by Torg- studies should be fully aware of all the method- erson et al.31 Although the latter authors do not ological pitfalls of the individual trials. A meta- pursue all of the possibilities in terms of esti- analysis of a series of trials that have na¨ıvely mating treatment effects, the design offers ways, ignored random therapist effects, for example, at least partially, of testing the validity of the or ignored the structure of a group therapy assumptions necessary for the above CACE esti- trial, simply summarises the faulty analyses of mator, or, equivalently, looking for a poor prog- the originals. Unfortunately, the consumers of nosis/demoralising effect in the potential com- meta-analyses (particularly if they are produced pliers of the control group. Getting preference under the auspices of such august bodies as the information prior to randomisation would also Cochrane Collaboration) seem to place far too improve the precision of the estimates of the much faith in their findings. Consumers need to CACE, but this is well beyond the scope of be aware that the authors of systematic reviews the present chapter – for further information, see are capable of missing subtle (or not so subtle) 39 Fisher-Lapp and Goetghebeur. The latter article methodological flaws in the original trials. Con- will also provide a suitable entry to the literature sumers should resist taking the conclusions of the on adjustment for partial compliance (i.e. regres- authors of these meta-analyses, and the clinical sion models for the response in terms of levels guidelines that result from them, on trust. In order of compliance). to critically appraise a published systematic view one needs to know about not only the mechanism (and quality) of the review itself, but also have a CONCLUSIONS detailed knowledge of what the reviewers should have been looking for in the way of method- The design and analysis of a convincing RCT ological problems in the original trials. Reporting for the estimation of the effects of psychotherapy guidelines such as CONSORT46,47 are having a are difficult. It is not safe to simply assume that substantial impact on the quality of clinical trials, the theoretical and logistical problems are similar and on the appraisal methodologies of system- to those of the average drug trial. Life here atic reviewers. In the case of psychotherapy trials, is much more complex. Psychotherapy (at least however, the CONSORT recommendations only in its individual form) involves the interaction cover a small part of the key components of the 310 TEXTBOOK OF CLINICAL TRIALS trial. Sticking to CONSORT guidelines is neces- 14. Elkin I, Shea T, Watkins JT, et al. National Insti- sary for a good trial report, but is not sufficient.I tute of Mental Health Treatment of Depression hope the present chapter succeeds in stimulating Collaborative Research Program (1989). Arch Gen Psychiat (1989) 46: 971–81. readers to think of other aspects of such trials that 15. Hamilton M. A rating scale for depression. J. need to be equally well reported. Neurology Neurosurgery & Psychiatry (1960) 23: 56–61. 16. Lavori PW. Clinical trials in psychiatry: should REFERENCES protocol deviation censor patient data? Neuropsy- chopharmacology (1992) 6: 39–48. 17. Siqueland L, Chittams J, Frank A, et al. The pro- 1. Dunn G. Statistical methods for measuring out- tocol deviation patient: characterization and impli- comes. In: Tansella M, Thornicroft G, eds, Men- cations for clinical trials research. Psychother Res tal Health Outcome Measures. London: Gaskell (1998) 8: 287–306. (2001) 5–18. 18. Schafer D. Multiple imputation: a primer. Stat 2. Garety P, Dunn G, Fowler D, Kuipers E. The Meth Med Res (1999) 8: 3–15. evaluation of cognitive behaviour therapy for psy- 19. Everitt BS, Pickles A. Statistical Aspects of the chosis. In: Wykes T, Tarrier N, Lewis S, eds, Out- Design and Analysis of Clinical Trials. London: come and Innovation in Psychological Treatment Imperial College Press (1999). of Schizophrenia. Chichester: John Wiley & Sons 20. Dowrick C, Dunn G, Ayuso-Mateos J-L, et al. (1998) 101–18. Problem solving treatment and group psychoed- 3. Rubin DB. Estimating causal effects of treatments ucation for depression: a multicentre randomised in randomized and nonrandomized studies. J Educ controlled trial. Br Med J (2000) 321: 1450–5. Psychol (1974) 66: 688–701. 21. Gibbons RD, Hedecker D, Elkin I, et al.Some 4. Holland PW. Statistics and causal inference (with conceptual and statistical issues in analysis of discussion). J Am Stat Assoc (1986) 81: 945–70. longitudinal psychiatric data. Arch Gen Psychiat 5. Dunn G. The challenge of patient choice and non- (1993) 50: 739–50. adherence to treatment in randomised controlled 22. Crits-Christoph P, Siqueland L, Blaine J, et al. trials of counselling or psychotherapy. Understand Psychological treatments for cocaine dependence: Stat (2002) 1: 19–29. National Institute for Drug Abuse Collaborative 6. Beck AT, Ward CH, Mendelson M, et al.An Cocaine Treatment Study. Arch Gen Psychiat inventory for measuring depression. Arch Gen (1999) 56: 493–502. Psychiat (1961) 4: 561–71. 23. Burlinghame G, Kircher J, Taylor S. Methodologi- 7. Sheiner LB, Rubin DB. Intention-to-treat and the cal considerations in group psychotherapy research: goals of clinical trials. Clin Pharmacol Ther past, present and future practices. In: Fuhriman A, (1995) 57: 6–15. Burlinghame G, eds, Handbook of Group Psy- 8. Fennell MJV. Cognitive-behaviour therapy for chotherapy and Counseling. New York: John Wiley depressive disorders. In: New Oxford Textbook & Sons. (1994) 41–80. of Psychiatry. Oxford: Oxford University Press 24. Martindale C. The therapist-as-fixed-effect fallacy (2000) 1394–405. in psychotherapy research. J Consult Clin Psychol 9. Markovitz JC, Weissman MM. Interpersonal psy- (1978) 46: 1526–30. chotherapy for depression and other disorders. 25. Crits-Christoph P, Mintz J. Implications of ther- In: New Oxford Textbook of Psychiatry. Oxford: apist effects for the design and analysis of com- Oxford University Press (2000) 1411–20. parative studies of psychotherapies. J Consult Clin 10. Ursano RJ, Ursano AM. Brief individual psycho- Psychol (1991) 59: 20–6. dynamic psychotherapy. In: New Oxford Textbook 26. Roberts C. The implications of variation in out- of Psychiatry. Oxford: Oxford University Press come between health professionals for the design (2000) 1421–32. and analysis of randomized clinical trials. Stat Med 11. Scott J. Psychological treatments for depression: (1999) 18: 2605–15. an update. Br J Psychiat (1995) 167: 289–92. 27. Goldstein H. Multilevel Statistical Models, 2nd 12. Roth A, Fonaghy P. What Works for Whom? A edn. London: Arnold (1995). Critical Review of Psychotherapy Research.New 28. Donner A. Design and Analysis of Cluster Ran- York: The Guilford Press (1996). domization Trials in Health Research. London: 13. Crits-Christoph P, Gladis M. The evaluation of Arnold (2000). psychological treatment. In: New Oxford Textbook 29. Elkin I. A major dilemma in psychotherapy out- of Psychiatry. Oxford: Oxford University Press come research: disentangling therapists from ther- (2000) 1259–69. apies. Clin Psychol: Sci Pract (1999) 6: 10–32. DEPRESSION 311
30. Elkin I,Parloff MB,Hadlet SW,Autry JH.National of compliance in randomized trials. Contr Clin Tri- Institute of Mental Health Treatment of Depression als (1999) 20: 531–46. Collaborative Research Program: background and 40. Zelen M. A new design for randomized clinical research plan. Arch Gen Psychiat (1985) 42: trials. New Engl J Med (1979) 300: 1242–5. 305–16. 41. Brewin C, Bradley C. Patient preferences and 31. Torgerson DJ, Klaber-Moffett J, Russell IT. Patient randomised clinical trials. Br Med J (1989) 299: preferences in randomised trials: threat or opportu- 313–15. nity? J Health Serv Res Policy (1996) 1: 194–7. 42. Bradley C. Designing medical and educational 32. Project MATCH Research Group. Matching alco- intervention studies. Diabet Care (1993) 16: holism treatments to client heterogeneity: Project 509–18. MATCH posttreatment drinking outcomes. J Stud- 43. Silverman WA, Altman DG. Patient’s preferences ies Alcohol (1997) 58: 7–29. and randomised trials. Lancet (1996) 347: 171–4. 33. Angrist JD, Imbens GW, Rubin DB. Identification 44. Ward E, King M, Lloyde M, et al. Randomized of causal effects using instrumental variables (with controlled trial of non-directive counselling, discussion). J Am Stat Assoc (1996) 91: 444–55. cognitive-behaviour therapy, and usual general 34. Little R, Yau LHY. Statistical techniques for ana- practitioner care for patients with depression. lyzing data from prevention trials: treatment of I: Clinical effectiveness. Br Med J (2000) 321: no-shows using Rubin’s causal model. Psychol 1383–8. Meth (1998) 3: 147–59. 45. Bower P, Byford S, Sibbald B, et al. Random- 35. Little R, Rubin DB. Causal effects in clinical and ized controlled trial of non-directive counselling, epidemiological studies via potential outcomes: cognitive-behaviour therapy, and usual general concepts and analytical approaches. Ann Rev practitioner care for patients with depression. Public Health (2000) 21: 121–45. II: Cost effectiveness. Br Med J (2000) 321: 36. Bloom HS. Accounting for no-shows in exper- 1389–92. imental evaluation designs. Eval Rev (1984) 8: 46. Moher D, Schultz KF, Altman DG. The CON- 225–46. SORT statement: revised recommendations for 37. Sommer A, Zeger SL. On estimating efficacy improving the quality of reports of parallel- from clinical trials. Stat Med (1991) 10: 45–52. group randomized trials. Ann Int Med (2001) 134: 38. Cuzick J, Edwards R, Segnan N. Adjusting for 657–62. non-compliance and contamination in randomized 47. Altman DG, Schultz KF, Moher D, et al.The clinical trials. Stat Med (1997) 16: 1017–29. revised CONSORT statement for reporting ran- 39. Fisher-Lapp K, Goetghebeur E. Practical proper- domized trials: explanation and elaboration. Ann ties of some structural mean analyses of the effect Int Med (2001) 134: 663–94. REPRODUCTIVE HEALTH
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 20 Contraception
GILDA PIAGGIO∗ Department of Reproductive Health and Research, World Health Organisation, Geneva, Switzerland ∗The views expressed in this paper are solely those of the author and do not necessarily reflect the views of the World Health Organization.
INTRODUCTION very safe so that it does not offset the benefits obtained from their use, and this emphasizes the Contraception deals with the prevention of preg- importance of addressing the safety concerns. nancy. The basic pillar of family planning is Both contraceptive efficacy and risks should be a wide spectrum of contraceptive methods that well defined to enable the user and the prescriber enables men and women to make informed to make the best choice of a contraceptive choices about timing and size of family. Effec- method.1 tive and safe methods should be available such The development of effective and safe methods that they fit the needs of women and men in of contraception poses special challenges. First, very diverse social and cultural settings world- to achieve an understanding of the complex phys- wide. There should be reversible methods that pro- iology of reproduction. Second, the effectiveness tect against sexually transmitted infections, can be of many methods depends on a successful inter- controlled by women, can be used by adolescents action between men and women. Some methods and by breast-feeding women. The choice of a have to be used by the man or the couple but contraceptive method involves personal decisions failure (pregnancy) is always observed in the and depends on the stage of life, family situation woman. Third, for some methods, behavioural or civil status, age, preferences and health profile and social factors are critical, determining com- of individuals and couples. Contraceptive research pliance, which is very difficult to assess. and development has resulted in a variety of avail- Contraceptive methods can in general be clas- able methods and continues to address important sified into hormonal and non-hormonal meth- issues to fill gaps in the available portfolio and in ods. Hormonal methods used by women include knowledge about method safety and effectiveness. oral contraceptive pills (OCs), injectable prepara- Contraceptives are generally used by healthy tions, implants, hormone-releasing devices (vagi- individuals to prevent pregnancy and some also nal rings and progestogen-releasing IUDs) and serve prophylactic purposes. They need to be post-coital oral pills (visiting pills and emergency
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 316 TEXTBOOK OF CLINICAL TRIALS contraceptive (EC) pills). Non-hormonal meth- Modern low-dose COCs contain a combination ods used by women include intrauterine devices of oestrogen and a progestin (20 to 35 mcg of (IUDs), barrier methods (diaphragm and female oestrogen and 150 mcg or less of levonorgestrel, condom), spermicides, natural methods (calendar or 200 to 300 mcg of norgestrel or 400 to and lactational amenorrhoea) and sterilisation, 1000 mcg of norethindrone or the equivalent of as well as immunocontraceptives, that are being another progestin). There are monophasic formu- developed. Hormonal male methods consist of lations, with constant daily doses of oestrogen injectable preparations and implants, still under and progestin, biphasic ones, in which the dose development. Non-hormonal ones comprise con- of progestin changes in each of two periods and doms, withdrawal and sterilisation (vasectomy triphasic ones, in which the dosages change in and vas occlusion), while immunocontraceptives each of the three seven-day periods of pill intake or vaccines are under development. during the 21 days of pill cycle. These broad classes of contraceptive methods COCs prevent conception through the suppres- differ in the length of the acting period, in the sion of ovulation via hypothalamic and pituitary mechanism of action, in the interval and way of effects and progestin-mediated alterations in the administration or insertion, in the possibility of consistency and properties of cervical mucus. It control by the woman, in their effectiveness and is still unconfirmed if the mechanism of action in their possible effects on health and indications also includes alterations in the endometrial lining for their use. They also differ in the way they and of tubal transport mechanisms. meet the interests of men and women in different POPs have a lower dose of progestin than social and cultural settings. do COCs (75 mcg of norgestrel or 350 mcg of norethindrone). They prevent conception through CONTRACEPTION METHODS: a combination of mechanisms including suppres- AN OVERVIEW sion of ovulation, alteration of cervical mucus, of the endometrium and of the fallopian tubes. Table 20.1 presents a list of currently used con- Synthetic oestrogens were first developed in traceptive methods with the interval of action the early 1930s and the more potent ethinyl required, the pregnancy rates under typical and perfect use and the main safety concerns. Exten- oestradiol in 1938, while synthetic orally active sive and detailed descriptions of old and new progestins were first produced in the early contraceptive methods are available.2,3 Acom- 1950s. In this decade the first generation pro- prehensive review of the literature on contracep- gestins, like ethynodiol and lynesterol were tive efficacy was done by Trussell and Kost.4 developed. OCs became available in the United States in 1959. A major breakthrough in the HORMONAL CONTRACEPTIVES development of OCs was the finding that the FOR WOMEN oestrogen and progestin acted synergistically to inhibit the pituitary. This allowed the transi- Hormonal methods prevent conception by inhibit- tion from high-dose to low-dose, of both the ing ovulation or preventing implantation or oestrogen and the progestin. Low-dose oestro- changing the quality of cervical mucus and thus gen COCs have less frequent complaints about preventing sperm access to the cervix. Oral meth- 5 ods exert their action if administered within a breast tenderness, nausea and leg cramps. Infor- cycle, while injectable preparations, implants and mation on efficacy and common side-effects hormone-releasing devices are long-acting. was obtained from randomised clinical trials (RCTs),6,7 one of them comparing COCs and POPs.7 The progestins that have been most Oral Contraceptives widely studied are norethindrone (or norethis- OCs comprise combined oral contraceptives terone) and levonorgestrel (second-generation). (COCs) and progesterone-only pills (POPs). Around the 1990s, three new progestins have CONTRACEPTION 317
Table 20.1. Contraceptive methods available, their characteristics, typical failure and discontinuation rates and safety concerns
Typical one year Perfect one year pregnancy pregnancy Duration of rate per 100 rate per 100 Type Method action woman-years woman-years Safety concerns
Hormonal for women OCs COCs Daily 6–8 0.1 Cardiovascular diseases, depression, hepatic adenomas POPs Daily 1 (breastfeeding) 0.5 Unknown Injectables DMPA 3-month 0.3 0.3 None NET-EN 2-month 0.3 0.3 None Combined once-a-month 0.3 0.3 Same as for COCs for severe pathologies Implants Norplant 5-year 0.1 0.1 Infection at implant site Jadelle 5-year Infection at implant site Implanon 3-year Infection at implant site Vaginal rings Combined 3–12 months <3 <3 Vaginal irritation (at insertion) Lesions Low-dose Lng 3–12 months 4.5 3.2 Vaginal irritation (at insertion) Lesions Non-hormonal for women IUD Copper 8–10 yrs 0.8 0.6 Increased menstrual bleeding, anaemia Uterine perforation PID 20–30 days post-insertion Lng-releasing 5–7 yrs Not available Not available STD risk Barrier Female condom Coitus-related 21 5 None Diaphragmw/ Coitus-related 20 6 Toxic shock syndrome spermicide Urinary tract infections Cervical cap Coitus-related 20 (nulliparous) 9 (nulliparous) Toxic shock syndrome 40 (parous) 26 (parous) Spermicides Spermicides Coitus-related 26 6 Vaginal infection and irritation Natural Periodic Daily 20 1–9 None abstinence Lactational Duration of 2 0.5 None amenorrhoea breastfeeding Sterilisation Female Permanent 0.5 0.5 Infection, anesthesia complications Non-hormonal for men Barrier Male condom Coitus-related 14 3 Natural Withdrawal Coitus-related 19 4 Sterilisation Male (vasectomy) Permanent 0.2 0.1 Cancer of prostate and testes been introduced (third-generation): norgestimate, of the ovary and the endometrium.12 However, desogestrel and gestodene. a possible link between the use of OCs for a OCs are likely to affect lipid and carbohydrate long period and breast cancer risk among young metabolism and the coagulation system, which women and cancer of the cervix is still a con- seem to be predictive of cardiovascular problems. cern. Side-effects of COCs are nausea, headaches, The effect of low-dose OCs on these physiologi- dizziness, spotting, weight gain, breast tenderness cal functions has been shown to be non-existent and chloasma. For POPs, the main side-effect is or small.8–11 Another safety concern regards can- menstrual irregularities. cer. Some studies have reported that the use of The association between OCs and cardiovas- hormonal contraceptives is protective of cancer cular diseases, namely venous thrombosis (VTE), 318 TEXTBOOK OF CLINICAL TRIALS ischaemic heart disease and cerebrovascular dis- mucus and the endometrium. Combined injectable ease, has been the object of a controversy. preparations seem to have a mechanism of action There seems to be a small increase in the risk similar to COCs. of VTE, larger for the OCs containing third- DMPA was first used as a contraceptive generation progestins compared to those con- in the 1960s. Subsequently other alternative taining second-generation progestins. However, injectable contraceptives were developed among ‘modern, low oestrogen dose OCs are extremely which NET-EN gained widespread use. Once- safe if used appropriately in young women’.13 a-month injectables were developed with the The most recent evidence suggests that myocar- purpose of overcoming the inconvenience of the dial infarction and stroke are rare and limited to disruption of the menstrual bleeding pattern of women who smoke cigarettes or have hyperten- progesterone-only preparations. WHO undertook sion or other cardiovascular risk factors. the evaluation and optimisation of the dose and OCs constitute the most common form of oestrogen/progesterone ratio of Cyclofem and steroidal hormonal contraception and it is also Mesigyna.18 Also, the Chinese Injectable No. the most common method of reversible contra- 1, with a complicated administration schedule, ception in countries other than China. It is esti- was developed in China. A multicentre trial mated that 60 to 80 million women are OC users was important to decide between the 100 mg 14 worldwide. COCs are a safe method of contra- or 150 mg dose for DMPA.19 A number of ception, only not to be recommended for women clinical trials showed that NET-EN was highly with cardiovascular diseases, diabetes mellitus, effective.20 Other trials determined that NET-EN 15 cancer or smoking. POPs can be taken by lac- needs to be administered every two months and tating women, but are not recommended in cases also compared DMPA and NET-EN.21 of thromboembolism or vein thrombosis. Regarding safety concerns, a large multicentre OCs require daily attention by the woman study provided reassurance that the use of DMPA and they have a high discontinuation rate: in was not associated with cancer22 and thus DMPA programmes, it has been reported that less than was registered in the United States as a long- 50% of the women who start use continue acting contraceptive.23 Headache is a common treatment after one year.16,17 Both COCs and POPs 15 complaint, side-effects are weight gain and delay are very effective under perfect use, and under in the return of fertility. Menstrual irregularities typical use they are still effective (Table 20.1). are frequent, including prolonged and heavy bleeding, mostly during the first months of use, Injectable Preparations and long periods of amenorrhoea. The most common injectable contraceptive is About 16 million women worldwide are users the progestin-only preparation depot-medroxy- of injectable contraceptives: 13 million DMPA progesterone acetate (DMPA), that provides con- users in 90 different countries, 1 million NET- traceptive protection for three months. Nor- EN users and 2 million once-a-month injectables ethisterone enantate (NET-EN) is also a progestin- users in Latin America and China, and introduc- only preparation that provides protection for two tory studies are being conducted in several other months. Combined oestrogen–progestin once-a- countries.24 month injectable contraceptives are Cyclofem, Progesterone-only contraceptives can be taken which combines DMPA with an oestrogen, and by breastfeeding women when they do not have Mesigyna, which combines NET-EN with an access to other methods. It is not recommended oestrogen. Injectable preparations are long-acting for women with multiple risk factors for arte- and require the intervention of health care profes- rial cardiovascular disease or with unexplained sionals to administer the injection. vaginal bleeding. A theoretical concern is the The mechanism of action of DMPA is sup- effect on the neonate for breastfeeding women pression of ovulation and changes in the cervical <6 weeks postpartum. CONTRACEPTION 319
Injectable contraceptives are very effective: 0.3 use. Common complaints are headache, weight pregnancies per 100 woman-years in the first gain, mood change and depression. The safety of year of use for DMPA, NET-EN and combined Norplant has been confirmed.28 injectables15 (Table 20.1). It is estimated that one million women are Norplant users. Patterns of use are similar to other progestogen-only contraceptives. It is very Implants effective, with 0.1 pregnancies per 100 woman- 15 Implants used for contraception in women consist years in the first year of use. of a silicone rubber tube or capsule inserted subdermally in the arm, containing a steroid Vaginal Rings or progestin released through it at a constant rate for several years. Implants are thus long- Vaginal rings are devices releasing either a acting and require the intervention of health care combination of a progestin and an oestrogen or professionals for insertion and removal. Norplant only a progestin, of which the most common are is the most widely used implant, consisting of six levonorgestrel and progesterone. levonorgestrel-releasing rods with contraceptive The mechanism of action of levonorgestrel- action during five years. Jadelle is a two-rod releasing rings is similar to that of POPs, but levonorgestrel implant with a five-year duration. with an increased effect on cervical mucus. Implanon is a single implant releasing 3-keto- The first ring was progesterone-only, then the desogestrel during three years. Another single progesterone dose was reduced and combined implant releases ST 1435, a progestin rapidly rings were developed. Several designs were metabolised, making the implant suitable for studied before an active core ring surrounded lactating women. by an active silastic membrane was developed, The mechanism of action, similarly to that of leading to multi-compartment rings. A low-dose POPs, includes a combination of effects, there levonorgestrel contraceptive ring (20 mcg/day) being indications that ovulation suppression is not was studied in WHO multicentre trials.29,30 the only one,25 since only about 50% of women Safety concerns related to the levonorgestrel show suppression of progesterone levels.26 ring are menstrual disturbances, vaginal symp- Norplant became available in the United States toms, lesions and repeated expulsion. in 1991, after regulatory approval based on large Vaginal rings are attractive because they can clinical trials, which provided information on be discontinued easily by the woman herself and discontinuation rates and side-effects.27 Norplant are thus under her control, they do not require II is a two-rod levonorgestrel implant easier daily attention like the OCs, and they are not to insert and remove and less conspicuous coitus-related like the condoms. The one year than Norplant, with a modified manufacturing pregnancy rates of combined rings were less design, but its development was stopped after than 3 pregnancies per 100 woman-years in a trials comparing it with Norplant showed lower multicentre trial31 (Table 20.1). efficacy. The pregnancy rates were found to depend on the type of tubing used to manufacture Emergency Contraception the implant, the soft tubing being an improvement over the hard tubing. Emergency contraception (EC) based on pills is The main safety concern of implants is infec- a post-coital method that is recommended up to tion at the implant site, otherwise they are consid- three to five days after an act of unprotected inter- ered safe. The main side-effect observed among course. The standard EC method was the Yuzpe Norplant users is disturbances in the menstrual regimen of COCs (ethinylestradiol 100 mcg and bleeding pattern, with episodes of prolonged and levonorgestrel 0.5 mg or dl-norgestrel 1.0 mg heavy bleeding, mostly during the first months of repeated 12 hours later). A superior regimen 320 TEXTBOOK OF CLINICAL TRIALS consists of two 0.75 mg doses of levonorgestrel doses ranging from 10 mg to 600 mg.38 With per- 12 hours apart32 or a single 1.5 mg dose.33 Single fect use, the corresponding figures in the cited doses of the anti-progestin mifepristone, ranging trials were 0.8% after levonorgestrel, 1.9% after from 10 mg to 600 mg is another method. EC is Yuzpe and 0.3% to 1% after mifepristone. a back-up method and cannot be used regularly. Regarding the mechanism of action, if unpro- NON-HORMONAL CONTRACEPTIVES tected intercourse occurs within a few days of FOR WOMEN ovulation, the only time when fertilisation can occur, ECs will exert their effect prior to implan- IUDs tation being completed (day 6–7 after fertilisa- tion) and thus an established pregnancy would Intrauterine devices (IUDs) are inert intrauterine not be disrupted.34,35 If EC is administered when rings or plastic devices with or without drug a woman is already pregnant there is evidence loading (copper or levonorgestrel). They are long- from a study with pregnant women that ‘ethinyl acting and require the intervention of health care oestradiol is not a reliable abortifacient ... and professionals for insertion and removal. IUDs that its efficiency as a postcoital contraceptive inserted after an unprotected coitus are also may be limited to a relatively short period fol- effective as EC. lowing ovulation prior to implantation’.36 The mechanism of action of IUDs is to inhibit Although EC started in the 1980s in Europe sperm and ovum transport and fertilisation. with the Yuzpe regimen, it was only in 1997 IUDs were first introduced for contraceptive that the FDA in the United States declared purposes at the beginning of the 1900s. The first the regimen safe and effective.37 The standard IUD consisted of a loop of silk thread. Then a EC method until the late 1990s was the Yuzpe metal copper-releasing ring was developed. In regimen, when levonorgestrel was shown in a the 1960s plastic coils became popular, like the trial to be more effective and have a better side- Lippes Loop. The IUD used in the 1970s, the effect profile.32 It has been registered in 1999 plastic Dalkon Shield, was associated with high pregnancy rates and high infection rates. Safety in the United States and is available now in at problems with old devices included the risk of least 80 countries around the world. Mifepristone contracting pelvic inflammatory disease, and that started to be used as an EC method at the dose of septic abortion and infertility, with consequent of 600 mg. RCTs have shown that 10 mg can high discontinuation rates. Randomised clinical be used instead of higher doses, with a reduced trials (RCTs) published in 1975 compared the effect of delay of menses observed with higher 38 Dalkon Shield with the Lippes Loop D and the doses. new Copper-7 and Copper-T 200. The superiority EC pills are relatively benign and they pose of collared Copper-T was thus established.39–41 no serious safety concerns. Nausea and vomiting In the early 1980s trials were conducted including are common with high-oestrogen regimens. Lev- NOVA T, MLCu250, Copper-T 220C and MLCu onorgestrel and mifepristone regimens have a bet- 375.42 In other trials the Copper-T 380 showed 32,38 ter side-effect profile than the Yuzpe regimen. superiority over the MLCu 375.43–44 Many IUD A concern with mifepristone is the delay of trials were conducted in China to try to design menses, mainly with high doses.38 copper IUDs adapted to Chinese women. A major In women receiving EC up to 72 hours after development was that of steroid-releasing IUDs. unprotected intercourse, 1.1% pregnancy rates Devices releasing 20 mcg/day of levonorgestrel have been observed after levonorgestrel with typ- have been shown to be very effective.47 ical use and 3.2% after Yuzpe.32 With the coitus- High efficacy and low risk have been observed to-treatment interval extended to 120 hours, preg- for copper IUDs and the levonorgestrel-releasing nancy rates were 1.1% to 1.3% after mifepristone IUD in large trials47–50 and there has been CONTRACEPTION 321 a progressive increase in effectiveness with The diaphragm is an elastic membrane with continued research.26 cavity rim, which may be attractive to poten- IUDs are the most commonly used contra- tial users but it lacks the advantage of protec- ceptive methods after sterilisation, and the most tion against STDs. A new microbicide-releasing commonly used reversible method in China. It diaphragm is being developed to address this is estimated that about 120 million women are concern. Safety concerns for the diaphragm are IUD users worldwide.51 The method is not rec- a history of toxic shock syndrome and allergy ommended for pregnant women and those with to latex.15 It is effective under perfect use current sexually transmitted infections or at risk (Table 20.1). of acquiring them. Spermicides are in the form of creams, jellies, Compliance is not a problem with IUDs, foams in pressurised containers, foaming tablets but there could instead be discontinuation for or suppositories. They are not very effective when several reasons. New copper devices and the new used by themselves, but can be used in combi- hormone-releasing IUD combine low pregnancy nation with other methods. Effectiveness: 6 preg- rates with low discontinuation rates. One-year nancies per 100 woman-years in the first year of pregnancy rates for the most efficient IUDs are use if used correctly and consistently (effective), <1 per 100 woman-years for the copper IUDs, 26 under typical use (only somewhat effective).15 0.1–0.2 for the levonorgestrel-releasing IUD and Self-administered topical preparations with sper- 0.6–0.8 for the TCu-380A IUD.15 micidal and microbicidal activities are being stud- ied, that provide both contraceptive and anti- infection protection and are under the control of Barrier Methods the woman. The cellulose sulphate gel is one such preparation. Barrier methods used by women are the dia- The cervical cap or sponge is a mushroom phragm, the female condom and spermicides. cap-shaped device releasing a spermicide and The importance of developing effective dual whose concave side is applied over the cervix. Its protection barrier methods that provide protection maximal insertion time is 24 hours. It is effective against sexually transmitted diseases (STDs) has under perfect use (Table 20.1). increased in the last years with the spread of the HIV/AIDs epidemic. Condoms are barrier methods providing this feature of dual protection. Natural Methods Barrier methods prevent conception by avoid- ing contact between sperm and the ovum. They Periodic abstinence restricts intercourse to the act as a mechanical barrier (condom, diaphragm) infertile phase of the woman’s cycle, which or by inactivating the sperm (spermicides) or both depends on the ability of the woman to identify (diaphragm with spermicide and cervical cap). the fertile period.52 It acts through prevention The female condom has become an important of fertilisation. It is effective under perfect use alternative because it is under the woman’s (Table 20.1). control and can provide women with the ability of The lactational amenorrhoea method is an protecting themselves against STDs in situations accepted method of contraception when the inter- where men refuse to use condoms. It is coitus- est of the woman is birth-spacing, since it has related and thus pregnancy can be the result of been observed that breastfeeding women within either method failure or inconsistency of use. six months of delivery who are amenorrhoeic Effectiveness: 5 pregnancies per 100 woman- have <2% risk of becoming pregnant.53 Effec- years in the first year of use under perfect use tiveness: 0.5 pregnancies per 100 woman-years (effective), 21 under typical use (only somewhat in the first year of perfect use (very effective), 2 effective).15 under typical use (effective).15 322 TEXTBOOK OF CLINICAL TRIALS
Sterilisation membrane 0.06 to 0.07 mm thick.55 Most are lubricated, and some contain spermicides. The Sterilisation in women is an effective surgical feature of dual protection and the mechanism of procedure involving the blockage of the fallopian action are the same as those described for the tubes, which transport mature ova from the female condom. ovaries to the uterus. The most widely practised Research on the male condom has dealt with techniques are minilaparotomy and laparoscopy. efficacy and acceptability issues. Old condoms Sterilisation is the only permanent contraceptive were made of hard material, acceptability was method and the most prevalent, 180 million low and they were not very resistant to adverse couples have been reported to be sterilised storage conditions. Improvement has been made (male or female). Research is in progress to with the latex condoms. A new polyurethane develop a non-surgical procedure. Effectiveness: condom was compared with latex condoms in 0.5 pregnancies per 100 woman-years in the first RCTs.56 year of use (always very effective).15 Male condom is the most widely used of bar- rier methods but its use is not more widespread ImmunoContraceptives because it is often not accepted, mainly by the male partner. Condoms have practically no risk Research is in progress for the development of of side-effects. The only concern has been allergy a female vaccine based on the human chorionic to latex in latex condoms. gonadotrophin molecule (hCG), for action after Effectiveness: 3 pregnancies per 100 woman- fertilisation and before implantation. years in the first year of use if used correctly and consistently (effective), 14 under typical use HORMONAL CONTRACEPTIVES FOR MEN (only somewhat effective).15 Hormonal injectable methods for men that reduce the production of spermatozoa are based on Natural Methods weekly injections of testosterone. Research is in progress for the development of a longer- Withdrawal, or coitus interruptus,isalow acting injectable preparation, namely at four- effectiveness method (Table 20.1) which depends week intervals. Preparations based on DMPA and on the man successfully withdrawing the penis testosterone and on NET-EN and testosterone are from the vagina before ejaculation starts, and thus under study. preventing fertilisation. Implants for men are also under investiga- tion. A depot androgen/progestin combination Sterilisation has recently demonstrated high contraceptive efficacy with satisfactory short-term safety and Vasectomy in the male is a simple surgical recovery of spermatogenesis.54 In this trial, a hor- procedure, but it is questionable in some cultural monal implant was given every four months to settings due to the incision and non-reversibility. replace testosterone and the progestin DMPA was Research is in progress for the development of a injected every three months. reversible procedure. A possible association between vasectomy and prostate cancer was a safety concern, but obser- NON-HORMONAL CONTRACEPTIVES FOR vational studies have shown that if there is MEN such an association, the increased risk in vasec- Barrier Methods tomised men compared to non-vasectomised men is small.57 Condoms used by men are tubes closed spher- Sterilisation is the only permanent contra- ically on one side, normally made of a latex ceptive method for men. Effectiveness: 0.1–0.2 CONTRACEPTION 323 pregnancies per 100 woman-years in the first year WHO trial under preparation, two types of of use (always very effective).15 implants will be compared with regard to efficacy in preventing pregnancies, allocating at random ImmunoContraceptives the type of implant to women choosing implants. To assess the effect of hormones on the bleeding Research on vaccines for men is in progress based pattern, IUD users or sterilised women will on antibodies that neutralise the biological effect constitute a control group. of the gonadotrophin hormone-releasing hormone Clinical trials generally include an insufficient (GnRH) and follicle-stimulating hormone (FSH), number of women to provide conclusive informa- with a resulting oligospermia or azoospermia. tion on rare events, like cancer and cardiovascular diseases.26 However, a careful documentation of serious adverse events and predisposing risk fac- CLINICAL TRIAL METHODS tors in the conduct of clinical trials should always IN CONTRACEPTION be provided.1 General principles applying to the conduct Observational studies constitute the source of of clinical trials and post-registration assess- ment of steroidal contraceptive drugs have been information for comparisons of efficacy, discon- 1,61 tinuation rates or safety across broad classes established. The development of a new con- of contraceptive methods, for example implants traceptive method involves a long process until it and IUDs. Women cannot usually be randomised is registered and reaches the market. The method- to different broad classes of methods because ology used depends on the stage of development the woman’s choice of contraceptive is deter- and will be treated separately for Phase I/II and mined by social, cultural and personal reasons. Phase III trials. An exception to this was one large RCT which allocated women at random to OCs or to vagi- PHASE I/II TRIALS nal methods (consisting of diaphragms, jellies, creams or foams).58 However, results were diffi- Objectives cult to interpret because there were many women switching methods and lost to follow-up.59,60 Phase I trials deal with drug safety and aim RCTs, on the other hand, have been an to determine an acceptable drug dosage, and important tool to find new safe and efficient also study drug metabolism and bioavailability. regimens or devices within each broad class and In contraceptive research, Phase I trials are improve existing ones by answering questions conducted to investigate the pharmacology of about the best compound, the best dose, the best steroidal contraceptive or other contraceptive interval or route of administration (compounds) drugs in healthy volunteers.61 Phase I trials must or the best physical properties (devices). be preceded by initial toxicity studies in rodents Sometimes partially randomised trials are used and pharmacokinetic studies in primates, which to compare two types of hormonal contraceptives give an indication for the dose used in the first within the same broad class with a non-hormonal clinical study.61 one, used as a placebo control group. For When a contraceptive has been assessed to example, in a WHO trial (data not published) be safe in Phase I trials, its research progresses on the effect of two injectable contraceptives to Phase II trials, using the optimal dose and (DMPA and NET-EN) on lipid and lipoprotein administration schedule. Contraceptive Phase II metabolism, women requesting an injectable trials are small-scale investigations into the effec- contraceptive were allocated at random to the tiveness and safety of a contraceptive method, two preparations, and a group of non-hormone- carried out on closely monitored patients. They releasing IUD users was the control. In another have the goal to establish its mechanisms of 324 TEXTBOOK OF CLINICAL TRIALS action, metabolic effects and provide prelimi- and conducting ultrasound of ovaries. Thus, nary estimates of the frequency of common side- information is obtained on the extent to which effects, its effectiveness and its acceptability. It ovarian function is suppressed. Time to onset is recommended to previously conduct repeated- of action and to return to normal ovarian func- dose toxicity and reproduction studies in animals. tion after discontinuation should be studied. Phase II studies are conventionally subdivided 3. Other pharmacological effects on the repro- into Phase IIA, studies on the pharmacology of ductive system: effects on the endometrium the drug in patients and Phase IIB, definitive and on the cervical mucus. dose-finding studies.61 4. Effects on other endocrine systems: pituitary, Since steroidal contraceptives are used by adrenal, thyroid. healthy people, it is desirable to assess the mini- 5. Metabolic effects: effects on hemostatic vari- mum effective dose at the initial stages of clinical ables, plasma lipids and carbohydrate metabolism. testing.61 This can be done already in Phase I For products not containing an oestrogen trials instead of in later stages. The direct assess- and suppressing oestrogen secretion, effect on ment of efficacy in small trials is not possible bone mineral density and bone metabolism. because with reasonably effective contraceptives pregnancy is a rare event. There exist surro- Design gate variables for efficacy of a steroid drug for pregnancy prevention. Phase I trials on contra- For contraception in women, recruitment into ceptives, therefore, are often also used to look Phase I trials is conducted among volunteers of at these surrogates of efficacy in addition to reproductive age, not pregnant or lactating, regu- safety issues, so that Phase I and Phase II tri- larly menstruating, identified in family planning als are combined to evaluate both safety and clinics or selected community groups, who are endocrinological endpoints. Examples of surro- IUD users or sterilised, and therefore, not at risk gates of efficacy are hormone levels as indica- of becoming pregnant. Users of other hormonal tors of inhibition of ovulation in contraceptives contraceptives than the one being studied are not for women, sperm concentration in long-acting acceptable because the method might interfere androgen–progestogen formulations for men as with the assessment of clinical and laboratory an indicator of inhibition of spermatogenesis, and parameters. Other selection criteria depend on amount of serum hCG antibodies in immunocon- the contraceptive being studied. For example, for traceptive trials for women. Serological and clin- contraceptive vaccines, acute hypersensitivity to ical diagnoses of pregnancies are also conducted. the carrier should be excluded, and if reversibil- In the case of hormonal contraceptives for ity cannot be assured, participants should be women, the following clinical pharmacological no younger than 25 years and have had chil- parameters should be studied to assess the dren previously. inhibition of ovulation:1 A series of sequential studies using different dose levels are conducted to assess the minimum 1. Hormonal activity: the nature and hormonal effective dose. These studies involve doses that activity of a steroid contained in the contra- are two or three times the initial dose. For each ceptive and its principal metabolites should dose level, a study is conducted in 10–20 healthy be described. volunteers.61 2. Pharmacological action on ovarian function: Selection criteria for Phase II trials on women, the mechanism by which the contraceptive different to those mentioned for Phase I trials, are effect is attained should be described. This being currently exposed to the risk of pregnancy involves, in the first place, a description of and have proven fertility. At this stage, if the ovarian function in women with normal volunteers participating in Phase I studies are ovulatory function, measuring plasma concen- IUD users and they are willing to continue, then tration of ovarian steroids and gonadotrophins the IUD should be removed to assess efficacy. CONTRACEPTION 325
Phase IIB trials require about 50–100 subjects once-a-month preparations. The crossover design to assess efficacy and side-effects of the dosage has been used in a metabolic study to compare determined in early trials (Phase I–IIA). three different progestogens (norgestrel, norethis- terone and medroxyprogesterone) in treatment Examples periods of 3-week duration immediately preceded by 3 weeks of ‘wash out’.64 Examples of Phase I and Phase II clinical trials are the ones conducted with injectable prepa- Ethical Considerations rations to evaluate well-known potent synthetic progestins in combination. A Phase I trial tested When conducting Phase I/II trials, the fact that the use of progesterone as an alternative.62 contraceptive methods are used by healthy indi- Several examples for injectable contraceptives viduals implies a different risk/benefit assess- are summarised in a review by Newton et al.24 ment compared with therapeutic drugs for life- An early pharmacological trial on Cyclofem with threatening diseases. This justifies the assessment 11 women involved one pre-treatment cycle, a of the minimum effective dose at early stages of 3-month treatment phase with an injection of development. Cyclofem every 28 days and then a 3-month When volunteers are advised on the risks and recovery phase. It confirmed that ovulation was benefits of the study in order to seek their inhibited and that inhibition of luteal activity per- informed consent, the specific risks of receiving sists after the last injection for several cycles.63 a steroidal contraceptive should be explained. A comparative non-randomised study of Cyclo- fem and Mesigyna with 15 women, 8 receiv- PHASE III TRIALS ing Cyclofem and 7 Mesigyna, involved one Objectives pre-treatment cycle, three treatment cycles of 28 days and a 90-day follow-up period. The After a contraceptive is shown to be reasonably results showed that the suppressive effect of effective in Phase II trials, it is essential to com- Cyclofem was greater than Mesigyna.64 pare it with the current standard contraceptive(s) A four-arm trial of reduced dose of medrox- within the same broad class in a large trial involv- yprogesterone acetate and oestradiol cypionate in ing a substantial number of patients, with the goal Cyclofem recruited 88 women into the follow- of establishing its efficacy.26 Phase III trials per- ing groups: Cyclofem full dose, Cyclofem half mit more refined estimates of safety, effectiveness dose, DMPA full dose, DMPA half dose. All four and acceptability in comparison with a standard. preparations were found to be effective in inhibit- In contraceptive research, this information pro- ing ovulation for at least one month after the vides the basis for introducing a hormonal contra- injection, and the combined preparations showed ceptive into family planning practice in field trials more regular bleeding patterns.65 in various settings, as a prerequisite for registra- tion with drug regulatory authorities (see introduc- Metabolic Studies tory trials in the other Issues section later). Specific Phase II studies on biochemical variables Design and Trial Size are conducted when required. These variables include lipid and lipoprotein metabolism, coagu- The most common design to compare meth- lation, fibrinolysis and platelet function as well as ods within each broad class of contracep- other physiological events such as vaginal blood tives has been the parallel group design, loss.61 The parallel group designs have been the with simple randomisation in single-centre tri- most common design in this type of study, but fac- als, and stratified by centre in multicentre torial designs have also been used. Newton et al.24 trials. This was the case for the develop- describe examples of these studies for injectable ment of OCs,5–7 injectables,18,19,22,32,33,38,55,67 326 TEXTBOOK OF CLINICAL TRIALS implants,27 IUDs,40–46 condoms68 and EC hand, the Committee for Proprietary Medicinal regimens.32,38,56 Products (CPMP) recommends that 20 000 cycles The control used in RCTs to compare effi- be observed, which at 13 28-day cycles per cacy of methods is typically an active control, year is equivalent to 1540 women-years or 770 since a placebo control would not be ethical. women followed completely for two years. This Examples of comparisons of new versus standard, calculation is based on the criterion that the respectively, are the following: NET-EN ver- difference between the upper 95% confidence sus DMPA (injectables), Norplant II versus Nor- limit for the Pearl index (number of pregnancies plant (implants), steroid-releasing versus copper per 100 women-years) and the point estimate IUDs, polyurethane condom versus latex con- does not exceed 1.1 dom, Yuzpe versus levonorgestrel regimens (EC). Placebo controls have been used to assess effi- Recruitment cacy of a treatment to improve the bleeding pattern disrupted by the use of progestin-only Participants in Phase III contraceptive trials are contraceptives. usually recruited in family planning clinics. A In contraceptive trials, the main end-point for majority of attendants requesting contraception in efficacy is based on pregnancies, a rare event. family planning clinics (other than STD clinics) The number of subjects required per group to are healthy. On arrival, subjects (women or men) detect as significant a difference between groups or couples requesting or using the method under corresponding to a doubling of the rate, in a two- study are screened for eligibility. An eligibility sided 5% level test, with 80% power, is usually criterion common to contraceptive efficacy trials large (1140 for a control rate of 2%, 4700 for a is good general health, but others are specific control rate of 0.5%). to the contraceptive method, depending on the When the effect of two factors is of interest and corresponding safety concerns and eligibility if an interaction is likely to be present, the sample criteria.15 size needed is approximately double in a four-arm Trial participants are not therefore a random 2 × 2 trial than in a two-arm trial comparing the sample from women in reproductive age, and two doses of only one component. This might be their particular characteristics affect external a reason for which factorial designs have not been validity, making difficult the generalisation of commonly used in contraceptive efficacy trials. results to wider populations.4,70 First, women In the study of bleeding patterns among users of choosing a particular broad class of contraceptive progestogen-only contraceptives, an example of are likely to be self-selected. For example, a factorial design is provided by a trial compar- implants are often selected by older women. ing the effect of low-dose aspirin and vitamin Second, clinicians are likely to select different E alone or in combination on Norplant-induced types of women for different contraceptives. prolonged bleeding.69 Third, women who are long-term users of a For registration of a steroidal contraceptive, method and are happy with it do not come to some regulatory agencies require clinical trials the clinic and are less likely to be enrolled. with 200 (FDA) or 400 subjects completing two According to current ethical principles, all eli- years of observation, while some others require gible subjects have to provide informed consent even less.26 It is clear that this number does not before being enrolled into the trial. In contra- provide sufficient power to detect a difference ceptive trials, obtaining this consent from ado- in rare events with the control. Nor does it lescents is problematic because some countries provide sufficient precision for a confidence require a minimum legal age to provide consent. interval estimation of a rare event: 5 events Consent from relatives is not always possible due observed in 200 subjects gives a rate of 2.5% to the need to maintain confidentiality in sensitive with 95% CI of 1% to 10%. On the other issues like contraception. CONTRACEPTION 327
Randomisation, Allocation Concealment trials due to unblinding caused by ancillary and Blinding information, like differential side-effects from the treatments being compared. For example, in EC Randomisation in contraceptive RCTs is achieved trials, higher doses of a compound might cause in a similar way to RCTs in other areas, by the use nausea more frequently than lower doses. of a computer-generated randomisation sequence. In multicentre trials the randomisation is usually Effectiveness and Efficacy of Contraceptive stratified by centre, done in blocks, and prepared Methods: Theoretical Model centrally. Treatments are allocated by assigning Effectiveness of a method can be defined as ‘the the next consecutive subject number in the ran- proportionate reduction in fecundability caused domisation sequence to the next enrolled subject. by the use of a method’.4 As such, it is Allocation concealment strategies include not measurable because one would have to sealed opaque envelopes (unblinded trials) and compare the rate of conception under use of packing of drugs by a central company (blinded the method with that in the same population trials). Many multicentre multinational RCTs not practising contraception nor lactating. The have included settings with poor telecommunica- common use of effectiveness is to denote how tion systems, in which central telephone randomi- well a method works. Sometimes efficacy is used sation as a strategy for allocation concealment with this meaning. was not a feasible option. Steiner et al.71 proposed a theoretical model Most clinical trials comparing implants, IUDs (see also Refs. 4 and 51) in which the couple’s and other devices cannot be blinded because ability to conceive and the timing and frequency insertion or placement of the device usually of intercourse determine the unobservable preg- implies that both the administrator and the nancy rate in the absence of contraception. In the user will see it, touch it or smell it. The presence of (perfect or imperfect) contraceptive situation is similar in sterilisation trials in which protection this pregnancy rate is reduced, deter- surgical procedures are compared. Some trials mining the ‘typical’ pregnancy rate. This typical comparing IUDs or sterilisation techniques can rate is composed of the perfect use pregnancy rate be blinded to the woman but not to the device and the imperfect use pregnancy rate, weighted or procedure administrator. Depending on the by the proportion of each type of user. treatments being compared, many clinical trials A measure of efficacy that implies a com- comparing injectable preparations, drugs for EC parison with the same treated population under and possibly spermicides can be blinded to users, placebo is the proportion of pregnancies pre- treatment administrators and outcome evaluators vented out of the expected pregnancies, or pre- (‘double blind’). ventable fraction, given by 1− observed preg- Blinding in contraceptive trials can avert bias nancy rate/expected pregnancy rate. after treatment allocation by preventing the The contraceptive method efficacy is the pre- following causes of bias. First, it is possible that ventable fraction under conditions of perfect use, the health care provider or the user will tend to and the effectiveness is the preventable fraction discontinue one treatment more than the other. under conditions of typical use. The difference Second, ascertainment bias could be introduced between these two rates depends on both the in the evaluation of subjective outcomes, like pregnancy rates under each condition and the pro- lesions in contraceptive rings trials. The delay portions of the two types of users.52 in the recognition of pregnancy, the imprecision in the estimate of the date of conception and Estimation of Pregnancy Rates and Efficacy the occurrence of chemical pregnancies noted in Non-coitus Related Methods above are sources of uncertainty which also pave the way for the introduction of ascertainment Sterilisation, which acts continuously and is per- bias. Bias could still be present even in blinded manent, and methods which act continuously but 328 TEXTBOOK OF CLINICAL TRIALS are reversible, like IUDs and long-acting hor- introduced in comparative trials by the failure to monal methods, are non-coitus related methods observe all subjects through the completion of the in the sense they do not require any particular study. The magnitude of the bias depends on the action by the user to be effective. proportion of subjects lost to follow-up and the If there is no daily monitoring of follicu- outcome mean or proportion in this group. lar growth or urinary metabolites, the estimates The estimation of the pregnancy rate using the of efficacy by the preventable fraction are con- Pearl index has been shown to be not appropriate structed with external estimates of probabilities due to the decline in fertility of the cohort of conception72 and with data on pattern and tim- being followed with duration of the contraceptive ing of intercourse. The latter is difficult to obtain use. This decline in fertility has been illustrated in large trials comparing regular use contracep- by Sivin and Schmidt42 from long-term studies, tives, therefore the common measure of how well where a progressive increase in the effectiveness a contraceptive method works in preventing preg- of each device with age was observed, as well as nancy is failure, or the occurrence of pregnancy a wider difference in failure rates among devices in the period of time during which the contracep- and a progressive increase in effectiveness with 4 tive is used. Thus, efficacy in this loose sense continued research. is evaluated in the woman and (inversely) mea- Life table techniques have been the recom- sured in number of pregnancies per woman-time mended analysis technique for years, using the of observation (typically per 100 woman-years). single decrement method, in which women who For reversible methods (for example IUDs and exit for other reasons than pregnancy are cen- long-acting hormonal methods), the assessment sored at the time of exit.52,70,73 The estimation of of the pregnancy status might be difficult due the pregnancy rate is given by the cumulative life to the following sources of uncertainty:52,70 table rate (net rate). The daily life table method, (1) when the decision to stop using a method is using the Kaplan–Meier product-limit estimate made, the pregnancy might be recognised after of the cumulative pregnancy rate (net rate) gives the method is stopped; (2) imprecision in the similar results and leads naturally to the logrank estimate of the date of conception when the 73 estimate is based on the date of start of the test to compare groups. last menstrual period; (3) occurrence of early A difficulty in the estimation of pregnancy rates ‘chemical’ pregnancies, of which a considerable is the presence of other reasons of discontinua- percentage is lost before reaching the stage of tion. For IUDs, the commonly analysed discon- clinical pregnancy and (4) early foetal losses, tinuation reasons are expulsion, medical removal which might be unnoticed by the woman. (due to pain, bleeding or pelvic inflammatory dis- In clinical trials comparing regular contracep- ease), non-medical removal (wish to become preg- tives, women are usually required to return to nant, no further need of contraception) and loss the clinic at specified intervals during a follow- to follow-up. The use of net rates from life table up period. The timing of reporting pregnancies techniques deals with competing causes by cen- varies among women. It is important that the soring. This approach has been criticised by Tai method of counting pregnancies does not depend et al.74 because it assumes independence of the on this timing. The ‘active follow-up’ prevents different reasons for discontinuation. They argue this problem by defining a cut-off date and con- that the estimates obtained by this approach are tacting women three months later to learn their overestimates of the rate for each of the reasons, pregnancy status at the cut-off date.70 as can be seen by the fact that the sum of the prob- One of the main problems affecting data quality abilities of discontinuation due to all the reasons is the loss to follow-up in trials comparing regular is greater than one. They are hypothetical rates use contraceptives, that require long periods of that would occur if discontinuations for other rea- follow-up. Bardin and Sivin26 discuss the bias sons could not take place. Tai et al. recommend CONTRACEPTION 329 the use of cumulative incidence rates, that esti- and perfect use. The comparison of the effective- mate the pregnancy rate in the presence of com- ness between treatments for all cycles (whether peting causes. They also argue in favour of the perfect or imperfect use took place) provides the competing risk methodology to compare cumula- treatment effect under conditions of typical use. tive incidences between two methods (two IUDs). The comparison of the efficacy between treat- Farley discussed these recommendations, con- ments is given by a subgroup analysis with cycles cluding that ‘Kaplan–Meier estimates are more of perfect use. appropriate when estimating the effectiveness of a contraceptive method, ... (and the) cumulative Estimation of Pregnancy Rates, Effectiveness incidence estimates are more appropriate when and Efficacy of Emergency Contraceptives making programmatic decisions regarding contra- In trials comparing EC methods, it is feasible to ceptive methods’.75 know the pattern and timing of intercourse: pro- tocols usually require a single unprotected act of Behavioural Patterns and the Estimation intercourse, and its date is reported by the woman, of Efficacy of Contraceptive Methods as well as the date of start of the last menstrual Methods that are used around the time of inter- period. A crude estimate of efficacy is the pro- course, like barrier methods and spermicides, are portion of observed pregnancies (number of preg- coitus-related and require a high degree of user nancies divided by the number of women treated), compliance with the correct way of using the but this measure is affected by the distribution of method in order to prevent pregnancy reliably.52 timing of intercourse with respect to the woman’s For these methods, a pregnancy can be the result cycle. The number of pregnancies occurring in the of either a method failure or lack of use or incor- same population under no use of method is unob- rect use. OCs are not coitus-related but have to servable. It is estimated by the expected number of be taken daily by women, posing similar types of pregnancies, calculated by multiplying the number problems. Similarly, periodic abstinence relies for of women having unprotected intercourse on each its effectiveness on rules of when to abstain from day of the menstrual cycle by the probability of sexual intercourse in order to avoid pregnancy, conception on that day, using external probabilities and users may depart from these rules. of conception.72 The proportion of pregnancies In order to separate a method failure from prevented, or preventable fraction, is given by (1− a lack of use or an incorrect use of a method [observed pregnancies/expected pregnancies]). A which is coitus-related, investigators denoted technique for the construction of confidence inter- pregnancies in which the method had not been vals for the preventable fraction is available, using used or had been incorrectly used as ‘user variance–covariance matrices from the external failures’. Pregnancies in which the method had estimates of conception probabilities.72 The day of been correctly used were denoted as ‘method ovulation, and thereof the day of the cycle in which failures’. Trussell70 illustrated the inadequacy intercourse took place, is usually estimated from of computing pregnancy rates corresponding to the date of the last menstrual period as reported by these two sources using the same denominator the women, and thus subject to imprecision. that includes all exposure from both ‘perfect’ The success of EC depends on not having and ‘imperfect’ use. He proposed to collect further unprotected acts of intercourse during information on ‘imperfect’ use for all months of the same cycle, since the EC treatment does exposure, or alternatively obtain information on not prevent pregnancies resulting from these correct and consistent use at the end of the trial, acts.32,38 Therefore user compliance can affect and calculate separate rates for ‘perfect’ users and the effectiveness of the method. If the EC for ‘imperfect’ users. treatment includes two doses, its success also For comparative trials, this issue is addressed depends on the treatment compliance, i.e., on by conducting a stratified analysis by imperfect the woman taking the second dose (at home) 330 TEXTBOOK OF CLINICAL TRIALS at the correct interval. The estimation of typical use and under typical use are likely to be of dif- and perfect use is therefore relevant in EC trials. ferent magnitude. In the second place, unless the When comparing groups, this has been achieved trial was designed to have sufficient power at the by two analyses, one including all subjects and a subgroup level, a relevant treatment effect in the subgroup analysis with perfect users. subgroup will not be detected, or the confidence interval estimate of the effect will be imprecise. Caveats in Comparing Efficacy and Stratification into perfect and imperfect users is Effectiveness Between Groups: another way of reporting results.32,38 Intention-to-Treat and Subgroup Analysis Assessment of Side-effects and Acceptability In RCTs, the comparison of estimates of effective- of Contraceptive Methods ness obtained with two treatments corresponds to an intention-to-treat (ITT) analysis (in the absence In women using contraceptives regularly, infor- of lost to follow-up), while that of efficacy corre- mation on side-effects and complaints such as sponds to a subgroup analysis of perfect users. In nausea, vomiting, diarrhoea, fatigue, dizziness, large RCTs, the proportions of perfect and imper- headache, lower abdominal pain and breast ten- fect users are likely to be similar in the treatments derness, as well as adverse events, can be col- being compared, so that differences in effective- lected in follow-up visits. Differences between ness between two treatments will depend mainly groups in events which have a rate of 5 or more on differences between the pregnancy rates under per 100 can be detected with small trials. Rates the two treatments. Thus, the comparisons of effec- of 1 to 5 per 100 require larger trials. Detection tiveness between treatments within the RCT are of differences for lower rates would require even not biased (internal validity). larger trials.26 On the other hand, the comparison of efficacy Side-effects of EC are complaints (or signs between treatments has the limitations of a sub- and symptoms) of nausea, vomiting, diarrhoea, group analysis. In the first place, the advantages fatigue, dizziness, headache, lower abdominal of randomisation are lost, since imperfect users pain and breast tenderness, and are treated as are excluded from the analysis. When the sub- binomial proportions. Delay in the return of next group analysis is based on subject characteristics menses, a time-to-event outcome, is undesirable that are not affected by treatment, like baseline because it gives the opportunity for more acts of variables, each smaller subgroup is like a smaller unprotected intercourse and is a source of anxiety randomised trial.76 But when the subgroup is for the woman. It has been found to be a concern defined by a variable observed after randomi- with mifepristone, mainly with high doses. sation and potentially affected by the treatment, Acceptability of a contraceptive method then the treatment effect may influence classi- depends not only on the characteristics of the fication into the subgroup. The treatment effect method but on the service delivery setting and observed in the subgroup would then be biased. the socio-demographic and economic factors of a This caveat is illustrated by an RCT to compare particular country.24 It can be assessed by ques- mifepristone and levonorgestrel for EC. The main tions to the user regarding satisfaction, willing- variable to define perfect use is adhering to the ness to recommend the method to others and to protocol requirement of not having further acts pay to have access to the method. Many side- of unprotected intercourse before the start of next effects of regular use contraceptives are reflected menses. Mifepristone is known to delay ovulation in discontinuation. Some of these discontinua- and thus is associated with a delay in the start tion reasons are related to the acceptability of the of menses, while levonorgestrel is not.33,38 This method by the user. For long-acting hormonal provides women under both regimens with a dif- methods, for example, the main discontinuation ferential opportunity to violate the requirement, reason is disturbances in the menstrual bleeding and then the effects of treatment under perfect pattern, largely determined by cultural and social CONTRACEPTION 331 factors. An Egyptian study on the acceptability of pattern was contained in four indices: number once-a-month injectable contraceptives found dif- of bleeding/spotting episodes, mean length of ferences between women discontinuing and those episodes, mean length of bleeding-free intervals continuing in all measures of acceptability.77 Fac- and the range of bleeding-free interval lengths. tors important in determining acceptability were: Based on the indices, the following clinically age, contraceptive history, learning about injecta- important patterns are derived:80 no bleeding bles, the husband’s attitude and knowing about (amenorrhoea), prolonged, frequent, infrequent another user’s satisfaction. and irregular bleeding. The 90-day reference period method was OTHER ISSUES applied to diary data collected from women treated with Cyclofem, Mesigyna, a low-dose Vaginal Bleeding Patterns levonorgestrel-releasing ring and DMPA taking Hormonal contraception is often associated with part in Phase III WHO clinical trials. Among disturbances in the vaginal bleeding pattern. women using once-a-month injectable and the These disorders are common with the use of levonorgestrel-releasing ring the incidence of progestogen-only hormonal methods and they do acceptable patterns was higher than among DMPA users, although the patterns were different not imply a health risk per se, since it has been 82 shown that the amount of blood loss is not a from those of normally menstruating women. problem. They may be tolerated by the woman, Several placebo controlled RCTs have been and this depends on cultural and behavioural conducted to investigate the therapeutic effec- patterns. The measurement of bleeding patterns tiveness of one or more treatments for bleeding can be achieved by direct questions to women, irregularities. An example is given by a trial com- by their completing menstrual diaries or by paring the bleeding pattern of untreated DMPA 26 users with groups treated with ethinyl oestradiol measuring blood loss. 83 The most used method of analysis of men- or oestrone sulphate. strual diaries is the reference period method,78 Equivalence Trials which was standardised by WHO using a 90-day reference period.79 It consists in analysing bleed- In equivalence trials, the objective is to show ing/spotting records in women’s menstrual diary that a new contraceptive or contraceptive device cards by taking fixed-length segments of time (the that has advantages with regard to side-effects, reference period, for which a 90-day segment has user preference or cost has equivalent efficacy been used as a convention) and deriving measures to that of the standard one. Failure to detect a of bleeding pattern, or indices. The following difference in a conventional significance test does 10 indices have been recommended:80 number not imply equivalence and a significant difference of bleeding/spotting days, number of spotting may correspond to equivalence within a margin days, number of bleeding/spotting episodes, num- of clinical relevance (or margin of equivalence, ber of spotting-only episodes, mean, range and denoted by ). A confidence interval for the maximum value of lengths of bleeding/spotting difference between the methods, on the absolute episodes, mean, range and maximum value or relative scale, on the other hand, is meaningful of lengths of bleeding-free intervals. These because it contains all likely values for the effect. indices have been analysed using box–whisker If these are all smaller than , then the methods plots and non-parametric analysis techniques. can be considered equivalent. To summarise the information provided, Belsey For trials in the reproductive field, it has been and Carlson81 conducted a principal component reported that the first difficulty encountered with analysis with data from women using different the corresponding trials was that the ‘equiva- contraceptives, and concluded that most of the lence’ nature of the trial had not been recognised essential information about a woman’s bleeding by the design teams. Therefore no statement of 332 TEXTBOOK OF CLINICAL TRIALS an equivalence hypothesis with the specification to detect uncommon but important reactions, can- of the margin of equivalence had been formu- not last long enough to identify long-term effects lated, nor had the sample size been calculated and the experimental group cannot be compared with the objective to demonstrate equivalence. to a placebo.61,85 This last limitation implies that This resulted in underpowered trials to demon- when comparing two active treatments through strate equivalence within a clinically relevant dif- an RCT, the absence of effect does not mean that ference. In only a few cases was an equivalence either has no risk compared to a placebo. Another hypothesis stated with a clear specification of the limitation of RCTs as a strategy at this stage of margin of equivalence.84 development of the contraceptive is that RCTs An example of an equivalence trial is given are conducted on healthy women and the risk by the WHO Yuzpe-levonorgestrel trial.32 The of adverse reactions might be relevant in women Yuzpe regimen using combined oral contracep- with risk conditions.60 tives had been used in EC as an effective method to prevent unwanted pregnancy. However, like Systematic Reviews other regimens containing oestrogen, it is asso- Systematic reviews on contraceptive methods are ciated with side-effects like nausea and vom- available in the Cochrane Library.86 A search was iting. The progestogen regimen levonorgestrel done using the word ‘contraception’, obtaining was shown to be better tolerated and equally or 45 hits out of 1456 total. A systematic review more effective, and it was recommended as a was included in the list if it included comparisons better alternative to the Yuzpe regimen. Another of efficacy, side-effects or acceptability of these example is a trial establishing the equivalence methods or effectiveness of treatments for bleed- between a single dose and a split dose of 1.5 mg ing irregularities induced by them. Trials includ- levonorgestrel for EC.33 ing contraceptives as treatment for complications or diseases and those comparing treatments for Introductory Trials and Phase IV Trials complications due to contraceptive use other than bleeding irregularities were not included. Subfer- Introductory trials are field studies to assess tility trials were not included. The title and if acceptability, effectiveness, continuation of use, necessary the abstract were examined to assess side-effects and service-related needs of a method whether the review was eligible. The 22 reviews in specific populations, in the context of family satisfying these criteria are listed in Table 20.2. planning services.61 They are expanded Phase III As an example, the systematic review ‘Inter- trials. Some countries may require to conduct ventions for emergency contraception’ included these trials in a network of 5–10 centres, includ- 33 trials, most of which were conducted in China. ing an acceptability component. Such studies The authors conducted 46 meta-analyses with dif- might involve 1000 to 5000 subjects. ferent comparisons and various outcomes com- Phase IV trials are those conducted after a prising efficacy (pregnancies) and side-effects, drug has been approved for marketing, to fur- including delay of menses. For the mifepristone ther investigate adverse effects of the drug. dose-comparisons they grouped the doses used in Very rare events cannot be rigorously assessed different trials in low, mid and high doses.87 before the contraceptive drug reaches the mar- ket because even Phase III trials do not have ACKNOWLEDGEMENTS sufficient power. Strategies for post-registration surveillance of contraceptive drugs are reports The author is greatful to Dr TMM Farley for of adverse reactions, large-scale experimental useful discussions regarding the structure of the studies, formal epidemiological studies and indi- chapter, and to Dr P D Griffin for proof-reading rect correlational studies.61 The most commonly the section on Phase I/II trials. used consists of epidemiological studies. Post- The UNDP/UNFPA/WHO/World Bank Spe- registration RCTs are costly, lack sufficient power cial Programme of Research, Development CONTRACEPTION 333
Table 20.2. Systematic reviews in The Cochrane Database of Systematic Reviews addressing efficacy or side-effects of contraceptive methods Method Stage Review
OCs Complete review Biphasic versus monophasic oral contraceptives for contraception Complete review Biphasic versus triphasic oral contraceptives for contraception Protocol Triphasic versus monophasic oral contraceptives for contraception Protocol Skin patch and vaginal ring versus combined oral contraceptives for contraception Protocol Comparison of acceptability of low-dose oral contraceptives containing norethisterone Injectables Protocol Treatment of vaginal bleeding irregularities induced by progestin-only contraceptives Implants Protocol Subdermal implantable contraceptives versus other forms of reversible contraceptives as effective methods of preventing pregnancy EC Complete review Interventions for emergency contraception IUDs Complete review Frameless versus classical intrauterine device for contraception Complete review Hormonally impregnated intrauterine systems (IUSs) versus other forms of reversible contraceptives as effective methods of preventing pregnancy Barrier Complete review Condom effectiveness in reducing heterosexual HIV transmission Complete review Diaphragm versus diaphragm with spermicides for contraception Complete review Sponge versus diaphragm for contraception Protocol Cervical cap versus diaphragm for contraception Protocol Female condom for preventing heterosexually transmitted HIV infection in women Protocol Non-latex versus latex condoms for contraception Lactational Protocol Lactational amenorrhoea for family planning amenorrhoea Sterilisation Complete review Minilaparotomy and endoscopic techniques for tubal sterilisation and Research Training in Human Reproduc- 6. WHO Task Force on Oral Contraceptives. A tion, Department of Reproductive Health and randomized, double-blind study of six combined Research, WHO, Geneva, Switzerland, supported oral contraceptives. Contraception (1982) 25: the author during her work on this chapter. 231–41. 7. WHO Task Force on Oral Contraceptives. A randomized, double-blind study of two combined REFERENCES and two progesterone-only oral contraceptives. 1. Committee for Proprietary Medicinal Products Contraception (1982) 25: 243–52. (CPMP). Clinical investigation of steroid contra- 8. Fotherby K. Update on lipid metabolism and oral ceptives in women. The European Agency for the contraception. Br J Fam Plann (1990) 15: 23–6. Evaluation of Medicinal Products (2000). 9. Gaspard UJ. Metabolic effects of oral contracep- 2. Van Look PFA, Perez-Palacios´ G. Contraceptive tives. Am J Obstet Gynecol (1987) 157: 1029–41. Research and Development 1984 to 1994. Oxford: 10. Runnebaum B, Rabe T. New progestogens in oral Oxford University Press (1994). contraceptives. Am J Obstet Gynecol (1987) 157: 3. Rabe T, Runnebaum B. Fertility Control – Update 1059–63. and Trends . Berlin: Springer-Verlag (1999). 11. Sabra A, Bonnar J. Hemostatic system changes 4. Trussell J, Kost K. Contraceptive failure in the United States: a critical review of the literature. induced by 50 mcg and 30 mcg estrogen/proges- Stud Fam Plann (1987) 18(5): 237–83. togen oral contraceptives. J Reprod Med (1983) 5. Dionne P, Vickerson F. A double-blind compari- 28: 85–91. son of two oral contraceptives containing 50 mu 12. WHO. Oral contraceptives and neoplasia. Report g. and 30 mu g. ethinyl estradiol. Curr Ther Res of a WHO Scientific Group, 817. WHO Technical Clin Exp (1974) 16(4): 281–8. Report Series (1992). 334 TEXTBOOK OF CLINICAL TRIALS
13. Farley TMM, Meirik O, Collins J. Cardiovascular in women during longterm use of Norplants. disease and combined oral contraceptives: review- Contraception (1980) 22(6): 583–96. ing the evidence and balancing the risks. Human 26. Bardin CW, Sivin I. Lessons learnt from clinical Reprod Update (1999) 5(6): 721–35. trials. In: Michal F, ed. Safety Requirements for 14. Rabe T, Runnebaum B. The future of oral hor- Contraceptive Steroids. WHO (1989) 400–13. monal contraception. In: Rabe T, Runnebaum B, 27. Sivin I. International experience with Norplant eds. Fertility Control – Update and Trends. Berlin: and Norplant-2 contraceptives. Stud Fam Plann Springer-Verlag (1999) 73–89. (1988) 19: 81–94. 15. RHR/WHO. Improving access to quality care 28. International Collaborative Post-Marketing Sur- in family planning: medical eligibility criteria, veillance of Norplant. Post-marketing surveil- second edition. WHO/RHR/00.02 (2000). lance of Norplant((R)) contraceptive implants: II. 16. Hagenfeldt K. Contraceptive research and devel- Non-reproductive health(1). Contraception (2001) opment today: an overview. In: Van Look PFA, 63(4): 187–209. Perez-Palacios´ G, eds. Contraceptive Research 29. Koetsawang S, Ji G, Krishna U, Cuadros A, and Development 1984 to 1994. Oxford: Oxford Dhall GI, Wyss R, et al. Microdose intravaginal University Press (1994) 3–22. levonorgestrel contraception: a multicentre clinical 17. Rivera R. Oral contraceptives: the last decade. trial. I. Contraceptive efficacy and side effects. In: Van Look PFA, Perez-Palacios´ G, eds. Contra- World Health Organization. Task Force on Long- ceptive Research and Development 1984 to 1994. Acting Systemic Agents for Fertility Regulation. Oxford: Oxford University Press (1994). Contraception (1990) 41(2): 105–24. 18. WHO Task Force of Long-Acting Systemic 30. Koetsawang S, Ji G, Krishna U, Cuadros A, Agents for Fertility Regulation. A multicentred Dhall GI, Wyss R, et al. Microdose intravaginal phase III comparative study of two hormonal con- levonorgestrel contraception: a multicentre clinical traceptive preparations given once-a-month by trial. II. Expulsions and removals. World intramuscular injection. II. The comparison of Health Organization. Task Force on Long- Contraception 40 bleeding patterns. (1989) (5): Acting Systemic Agents for Fertility Regulation. 531–51. Contraception (1990) 41(2): 125–41. 19. WHO Task Force of Long-Acting Systemic 31. Sivin I, Mishell DR Jr, Victor A, Diaz S, Alvarez- Agents for Fertility Regulation. A multicentred Sanchez F, Nielsen NC, et al. A multicenter phase III comparative trial of 150 mg and 100 mg study of levonorgestrel–estradiol contraceptive of depot-medroxyprogesterone acetate given every three months: Efficacy and side-effects. Contra- vaginal rings. I-Use Effectiveness. An international ception (1986) 34: 223–35. comparative trial. Contraception (1981) 24(4): 20. Hall P. Long-acting formulations. In: Dicz- 377–92. falusy E, Bygdeman M, eds. Fertility Regulation 32. WHO Task Force on Postovulatory Methods of Today and Tomorrow. New York: Raven Press Fertility Regulation. Randomised controlled trial (1987) 119–41. of levonorgestrel versus the Yuzpe regimen of 21. WHO Task Force of Long-Acting Systemic combined oral contraceptives for emergency con- Agents for Fertility Regulation. A multinational traception. Task Force on Postovulatory Methods comparative clinical trial of long-acting injectable of Fertility Regulation. Lancet (1998) 352(9126): contraceptives: norethisterone enantate given in 428–33. two dosage regimens and depot-medroxyproges- 33. Von Hertzen H, Piaggio G, Ding J, Chen J, Sonj S, terone acetate, final report. Contraception (1983) Bartfai´ G, et al. Low dose mifepristone and two 28: 1–20. regimens of levonorgestrel for emergency con- 22. WHO Collaborative Study of Neoplasia and traception: a WHO multicentre randomised trial. Steroid Contraceptives. Breast cancer and depot- Lancet (2002) 360: 1803–10. medroxyprogesterone acetate: a multinational 34. Wilcox AJ, Weinberg CR, Baird DD. Timing of study. Lancet (1991) 338: 833–8. sexual intercourse in relation to ovulation: effects 23. WHO Collaborative Study of Neoplasia and on the probability of conception, survival of the Steroid Contraceptives. Depot-medroxyproges- pregnancy, and sex of the baby. N Engl J Med terone acetate (DMPA) and risk of epithelial ovar- (1995) 333(23): 1517–21. ian cancer. Int J Cancer (1991) 49: 191–5. 35. Wilcox AJ, Weinberg CR, Baird DD. Post-ovula- 24. Newton JR, d’Arcangues C, Hall PE. Once- tory ageing of the human oocyte and embryo a-month combined injectable contraceptives. J failure. Human Reprod (1998) 13(2): 394–7. Obstet Gynaecol (1994) 14: S1–34. 36. Bacic M, Wesselius dC, Diczfalusy E. Failure of 25. Croxatto HB, Diaz S, Miranda P, Elamsson K, large doses of ethinyl estradiol to interfere Johansson ED. Plasma levels of levonorgestrel with early embryonic development in the human CONTRACEPTION 335
species. Am J Obstet Gynecol (1970) 107(4): and 7 years of use. Results from three randomized 531–4. trials. Contraception (1990) 42: 141–58. 37. Grimes DA. Emergency contraception – expand- 50. WHO. Long-term reversible contraception. Twelve ing opportunities for primary prevention. N Engl years of experience with the TCu380A and JMed(1997) 337(15): 1078–9. TCu220C. Contraception (1997) 56(6): 341–52. 38. WHO Task Force on Postovulatory Methods of 51. Wagner H. Intrauterine contraception: past, present Fertility Regulation. Comparison of three single and future. In: Rabe T, Runnebaum B, eds. Fertil- doses of mifepristone as emergency contraception: ity Control – Update and Trends. Berlin: Springer- a randomised trial. Task Force on Postovulatory Verlag (1999) 151–71. Methods of Fertility Regulation. Lancet (1999) 52. Farley TM. Reproduction. In: Armitage P, Col- 353(9154): 697–702. ton T, eds.Encyclopedia of Biostatistics (1998). 39. Tatum HJ. Comparative experience with newer 53. Labbok M, Jennings VH. Advances in fertility models of the Copper T in the United States. In: regulation through ovulation prediction during Hefnawi F, Segal SJ, eds. Analysis of Intrauterine lactation (lactational amenorrhoea method) and Contraception. New York: American Elsevier during the menstrual cycle. In: Van Look PFA, Publications (1975) 155–63. Perez-Palacios´ G, eds. Contraceptive Research 40. Jain AK. Safety and effectiveness of intrauterine and Development 1984 to 1994. Oxford: Oxford devices. Contraception (1975) 11(3): 243–59. University Press (1994). 41. Sivin I, Stern J. Long-acting, more effective cop- 54. Turner L, Conway AJ, Jimenez M, Liu PY, For- per T IUDs: a summary of U.S. experience, bes E, McLachlan RI, Handelsman DJ. Contra- 1970–75. Stud Fam Plann (1979) 10(10): 263–81. ceptive efficacy of a depot progestin and androgen 42. Sivin I, Schmidt F. Effectiveness of IUDs: a combination in men. J of Clinical Endocrinology review. Contraception (1987) 36(1): 55–84. and Metabolism (2003) 88(10): 4659–67. 43. Champion CB, Behlilovic B, Arosemena JM, 55. Rabe T, Vladescu E, Runnebaum B. Contracep- Randic L, Cole LP, Wilkens LR. A three-year tion: historical development, current status and evaluation of TCu 380 Ag and multiload Cu 375 future aspects. In: Rabe T, Runnebaum B, eds. intrauterine devices. Contraception (1988) 38(6): Fertility Control – Update and Trends. Berlin: 631–9. Springer-Verlag (1999) 29–72. 44. Cole LP, Potts DM, Aranda C, Behlilovic B, 56. Population Council Center for Biomedical Re- Etman ES, Moreno J, et al. An evaluation of the search. Annual Report 1991.NewYork:The TCu 380Ag and the Multiload Cu375. Fertil Steril Population Council (1991). (1985) 43(2): 214–7. 57. Farley TMM, Meirik O, Metha S, Waites GMH. 45. Rowe P. Clinical performance of copper IUDs. The Safety of vasectomy: recent concerns: Bul- Proceedings of a meeting: a new look at IUDs. letin of the World Health Organization (1993) In: Bardin CW, Mishell DR, eds. Proceedings 71(3-4): 413–9. from the Fourth International Conference on IUDs, 58. Fuertes-de la Haba A, Bangdiwala IS, Pelegrina I. Newton, MD. Butterworth Heinemann (1994) Success of randomization in a controlled contra- 13–31. ceptive experiment. J Reprod Med (1973) 11(4): 46. Sastrawinata S, Farr G, Prihadi SM, Hutapea H, 142–8. Anwar M, Wahyudi I, et al. A comparative clin- 59. Potts M, Feldblum PJ, Chi I, Liao W, Fuertes-de ical trial of the TCu 380A, Lippes Loop D and La HA. The Puerto Rico oral contraceptive study: Multiload Cu 375 IUDs in Indonesia. Contracep- an evaluation of the methodology and results of a tion (1991) 44(2): 141–54. feasibility study. Br J Fam Plann (1982). 47. Sivin I, Stern J, Coutinho E, Mattos CE, el Mah- 60. Skegg DCG. The epidemiological assessment of goub S, Diaz S, et al. Prolonged intrauterine con- safety of hormonal contraceptives: a methodolog- traception: a seven-year randomized study of the ical review. In: Michal F, ed. Safety Requirements levonorgestrel 20 mcg/day (LNg 20) and the Cop- for Contraceptive Steroids. WHO (1989) 21–37. per T380 Ag IUDS. Contraception (1991) 44(5): 61. Special Programme of Research, Development 473–80. and Research Training in Human Reproduc- 48. Luukkainen T, Allonen H, Nielsen NC, Nygren tion. Guidelines for the toxicological and clini- KG, Pyorala T. Five years’ experience of intrauter- cal assessment and post-registration surveillance ine contraception with the Nova-T and the Copper- of steroidal contraceptive drugs. In: Michal F, ed. T-200. Am J Obstet Gynecol (1983) 147(8): Safety Requirements for Contraceptive Steroids. 885–92. WHO (1989) 415–54. 49. WHO Task Force on the Safety and Efficacy of 62. Garza-Flores J, Fatiikun T. Future directions in Fertility Regulation. The TCu 380A, the TCu fertility regulation. In: Negro-Vilar A, Perez-´ 220C, Multiload 250 and Nova T IUDs at 3, 5 Palacios G, eds. Reproduction, Growth and 336 TEXTBOOK OF CLINICAL TRIALS
Development. New York: Raven Press (1991) 75. Farley TM, Ali MM, Slaymaker E. Compet- 383–97. ing approaches to analysis of failure times with 63. Fotherby K, Benagiano G, Toppozada HK, et al. competing risks. Stat Med (2001) 20(23): A preliminary pharmacological trial of the monthly 3601–10. injectable contraceptive Cycloprovera. Contracep- 76. Yusuf S, Wittes J, Probstfield J, Tyroler HA. Anal- tion (1982) 25: 261–72. ysis and interpretation of treatment effects in sub- 64. Aedo AR, Landgren BM, Johannisson E, Dicz- groups of patients in randomized clinical trials. falusy E. Pharmacokinetic and pharmacodynamic JAMA (1991) 266(1): 93–8. investigations with monthly injectable contra- 77. Hassan EO, el Nahal N, el Hussein M. Accept- ceptive preparations. Contraception (1985) 31: ability of the once-a-month injectable contracep- 453–69. tives Cyclofem and Mesigyna in Egypt. Contra- 65. WHO Task Force of Long-Acting Systemic ception (1994) 49(5): 469–88. Agents for Fertility Regulation. A multicen- 78. Rodriguez G, Faundes-Latham A, Atkinson LE. tred pharmacokinetic, pharmacodynamic study of An approach to the analysis of menstrual patterns once-a-month injectable contraceptives. I. Differ- in the critical evaluation of contraceptives. Stud ent doses of HRP112 and of DepoProvera. Con- Fam Plann (1976) 7(2): 42–51. traception (1987) 36(4): 441–57. 79. Belsey EM, Farley TM. The analysis of menstrual 66. Silfverstolpe G, Gustafson A, Samsioe G, Svan- bleeding patterns: a review. Contraception (1988) borg A. Lipid metabolic studies in oophorec- 38(2): 129–56. tomized women: effects of three different 80. Belsey EM, Machin D, d’Arcangues C. The anal- progestogens. Acta Obstet Gynecol Scand (1979) ysis of vaginal bleeding patterns induced by fertil- Suppl 88: 89–95. ity regulating methods. World Health Organization 67. WHO Task Force of Long-Acting Systemic Special Programme of Research, Development and Agents for Fertility Regulation. A multicentred Research Training in Human Reproduction. Con- phase III comparative study of two hormonal con- traception (1986) 34(3): 253–60. traceptive preparations given once-a-month by 81. Belsey EM, Carlson N. The description of men- intramuscular injection: I. Contraceptive efficacy strual bleeding patterns: towards fewer measures. and side effects. World Health Organization. Con- Stat Med (1991) 10(2): 267–84. traception (1988) 37(1): 1–20. 82. Fraser IS. Vaginal bleeding patterns in women 68. Family Health International. Contraceptive tech- using once-a-month injectable contraceptives. nology and family planning research. Family Contraception (1994) 49(4): 399–420. Health International (1992). 83. Said S, Sadek W, Rocca M, Koetsawang S, Kir- 69. d’Arcangues C, Piaggio G, Brache V, Ben wat O, Piya-Anant M, et al. Clinical evaluation Haissa R, Hazelden C, Mossai R et al. Effective- of the therapeutic effectiveness of ethinyl oestra- ness of Vitamin E and low-dose aspirin, alone or diol and oestrone sulphate on prolonged bleed- in combination, on Norplant induced prolonged ing in women using depot medroxyprogesterone bleeding. Contraception (2004) (in press). acetate for contraception. World Health Organiza- 70. Trussell J. Methodological pitfalls in the analysis tion, Special Programme of Research, Develop- of contraceptive failure. Stat Med (1991) 10(2): ment and Research Training in Human Reproduc- 201–20. tion, Task Force on Long-acting Systemic Agents 71. Steiner M, Dominik R, Trussell J, Hertz-Picciott I. for Fertility Regulation. Human Reprod (1996) 11 Measuring contraceptive effectiveness: a con- Suppl 2: 1–13. ceptual framework. Obstet Gynecol (1996) 88(3 84. Piaggio G, Pinol AP. Use of the equivalence Suppl): 24S–30S. approach in reproductive health clinical trials. Stat 72. Trussell J, Rodriguez G, Ellertson C. Updated Med (2001) 20(23): 3571–7. estimates of the effectiveness of the Yuzpe reg- 85. Petitti DB, Shapiro S. Strategies for post-registra- imen of emergency contraception. Contraception tion surveillance of contraceptive steroids. In: (1999) 59(3): 147–51. Michal F, ed. Safety Requirements for Contracep- 73. Farley TM. Life-table methods for contraceptive tive Steroids. WHO (1989) 335–47. research. Stat Med (1986) 5(5): 475–89. 86. The Cochrane Library (Issue 4). Oxford: Update 74. Tai BC, Peregoudov A, Machin D. A compet- Software (2002). ing risk approach to the analysis of trials 87. Cheng L, Gulmezoglu AM, Ezcurra E, Van Look of alternative intra-uterine devices (IUDs) for PFA. Interventions for emergency contraception. fertility regulation. Stat Med (2001) 20(23): The Cochrane Library. 2001. Oxford: Update 3589–600. Software (2001). 21 Gynaecology and Infertility
SILADITYA BHATTACHARYA1 AND JILL MOLLISON2 1Department of Obstetrics and Gynaecology, University of Aberdeen, Aberdeen, UK 2Department of Public Health, University of Aberdeen, Aberdeen, UK
INTRODUCTION are drawn from general gynaecology, infertility and fertility control. Trials in obstetrics, con- The randomised clinical trial is widely accepted traception and gynaecological oncology, which as the gold standard for scientific evaluation of are discussed elsewhere, will be excluded from treatments. In the current climate of clinical gov- our discussion. ernance, data from clinical trials are considered to represent the highest level of evidence that can be used to inform effective treatment strategies. TAXONOMY OF CLINICAL TRIALS Yet, there are fewer trials in gynaecology in com- parison with other disciplines. Those reported Clinical trials may be classified in a number in the literature account for a minority of pub- of ways (see Table 21.1). Some of these are 1 lished papers in major journals. Gynaecologi- discussed below. cal trials incorporate a wide spectrum of clinical conditions and proposed interventions. Women PHASE I CLINICAL TRIALS can differ substantially in terms of age, physi- cal and psychological disability; while treatments These preliminary studies generally address drug can range from drug therapy to surgical proce- safety rather than efficacy, and may be performed dures, from information provision to physiother- on healthy volunteers. Examples include stud- apy and dietary advice. The aim of this chapter is ies of drug metabolism and bio-availability of to examine, test and explore the basic principles recombinant gonadotrophins in infertile women.2 of clinical trials in gynaecology. An overview Most phase I trials are either directly or indi- of different types of trials is provided and refer- rectly supported by the pharmaceutical industry ence will be made to specific challenges, includ- and involve relatively small numbers of subjects. ing identification of sample populations, choice Women are required to adhere to a strict proto- of appropriate outcomes and tools, randomisa- col and agree to fairly extensive evaluation often tion and arrangements for follow-up. Examples involving multiple investigations such as blood
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 338 TEXTBOOK OF CLINICAL TRIALS
Table 21.1. Taxonomy of clinical trials current standard management for the same con- dition in a large trial involving a substantial Phased trials Phase I number of patients. This is also the design used Phase II Phase III for non-pharmacological interventions, which are Phase IV increasing in number. The majority of the tri- als referred to in this chapter are phase III tri- Conduct Pragmatic Explanatory als. This is the point of evaluation following which interventions are introduced into clini- Design Parallel group cal practice. Crossover Factorial Patient preference PHASE IV CLINICAL TRIALS Cluster randomisation Even after a treatment finds general acceptance, Randomisation True Quasi-randomisation unanswered questions about its safety and long- term effectiveness continue to be addressed in the context of phase IV trials. The long-term counts, biochemistry, endocrine profile and liver implications of new methods of treatment of and kidney function tests. In this context, it may menorrhagia such as endometrial ablation are still be useful to be aware of the fact that finding under evaluation a decade after the results of ‘normal’ subjects for such trials in reproductive the first phase III trials were reported. Medium- medicine can sometimes be challenging as a large term data have been presented in a number of 4,5 proportion of young, fit, healthy women may publications. either be on oral contraception or actively trying for a pregnancy. PRAGMATIC AND EXPLANATORY TRIALS
PHASE II CLINICAL TRIALS In terms of design, clinical trials are often described as either explanatory or pragmatic. These are also fairly small-scale investigations Explanatory trials measure efficacy–the benefit a into the efficacy and safety of a drug and require treatment produces under ideal conditions. Prag- close monitoring of each patient. Sometimes matic trials measure effectiveness–the benefit the they can be employed as a screening process treatment produces in routine clinical practice.6 to screen drugs, which are either potentially Examples of the former include evaluation of inactive or toxic. They may also be used drugs used to treat menorrhagia or those used to determine the most appropriate dose and to undertake medical termination of pregnancy. route of administration of a drug. Examples The aim is to assess the outcome of a new drug of these types of trials include those involving under controlled conditions using a homogeneous the use of misoprostol for medical termination group of patients. Eligibility criteria are strict, 3 of pregnancy. Where patient acceptability of and protocol violations are not allowed. In an the route of administration, i.e. vaginal, oral explanatory trial comparing recombinant follicle or sublingual, is an important outcome, these stimulating hormone with a urinary preparation, trials may need to break out of the traditional any woman who fails to receive the appropriate mould of strictly controlled explanatory trials and drug in the prescribed dose will be excluded from assume the pragmatic approach associated with the study on grounds of protocol violation. phase III trials. In contrast, a pragmatic trial aims to mirror the normal variations between patients that occur PHASE III CLINICAL TRIALS in real life. For example, a pragmatic trial of After a drug has been shown to be reasonably medical versus surgical treatment for menorrha- effective it is essential to compare it with the gia will include all women with a subjective GYNAECOLOGY AND INFERTILITY 339 complaint of heavy menstrual loss. Women ran- i.e. an experimental versus a control group. As domised to drug treatment, who find the interven- this type of trial is most easily understood by tion unacceptable and elect for surgery, do not researchers as well as patients, examples abound. face disqualification from the trial. This some- Occasionally a trial may have three arms, e.g. what relaxed policy is justified on the grounds intrauterine insemination (IUI) versus ovarian that women’s decisions to reject their allocated stimulation + intrauterine insemination versus treatment are likely to reflect real-life situa- in vitro fertilisation (IVF) for the treatment of tions and can actually be interpreted as a mea- unexplained infertility.7 Sample sizes for such sure of dissatisfaction. Furthermore, the treatment trials will be dictated by the minimum significant offered to patients in the surgery arm may not difference in outcome between any two arms. be identical, as operations may be performed by Due to the nature of the interventions, expected more than one surgeon, each with a slightly dif- differences between the arms will vary. The ferent technique. A similar attitude would apply number of women required to show a clinically to a pragmatic trial of physiotherapy for prolapse significant difference in pregnancy rates between or counselling for premenstrual syndrome, where IUI and IVF is smaller than that necessary to identical interventions cannot be guaranteed by show a difference between IUI and stimulated different physiotherapists or counsellors. IUI, and will ultimately be the one chosen for There are other differences between the two this trial. types of trials. Blinding is more likely to be used in an explanatory trial such as one comparing CROSSOVER oral metformin with placebo in women with polycystic ovarian syndrome. Pragmatic trials Crossover trials have the advantage of using may also be blinded, but this is often not feasible women as their own controls, thus reducing (for example, in surgical versus medical trials), the sample size required. Women are randomly nor always desirable. There is also less of a exposed to either the control or the intervention compulsion to use placebos, as the objective is arm first, followed by the other. Often a ‘washout to compare the new intervention, not with a period’ is introduced between the two arms to placebo, but with the ‘gold standard’ or best reduce the risk of contamination. Unfortunately of the existing treatments. Clinician and patient this design is more suited for medical treatments biases caused by the absence of blinding may of chronic conditions as opposed to surgical not necessarily be detrimental to the trial, but trials or infertility trials. In the first group the could actually be seen to be part of the response practicalities of the situation render such a design to treatment. The outcome in a pragmatic trial impossible. In the second, a definite outcome such as one comparing oral clomiphene citrate such as pregnancy has the natural effect of (a drug treatment) versus expectant management preventing women from completing later phases 8 in unexplained infertility incorporates the total of the trial. In such situations, exaggerated difference between the interventions that are estimates of treatment effect can occur, leading being evaluated. This may include the effect of to erroneous clinical decisions. From a practical the treatment as well as the associated placebo point of view, only data from the first phase effect as this best reflects the likely clinical of the crossover trial may be valid. However, response in practice.6 this design may well be suitable for exploring drug treatment of chronic conditions such as premenstrual syndrome or sexual function. TRIAL DESIGN
SIMPLE PARALLEL GROUP FACTORIAL This is the simplest and commonest trial design Factorial designs are often efficient as they can involving a comparison between two groups, address two questions within the context of a 340 TEXTBOOK OF CLINICAL TRIALS
Women with unexplained infertility
Randomisation
A B
Expectant management Clomiphene
C D
Intra-uterine insemination Clomiphene + intra-uterine (IUI) insemination (IUI)
Figure 21.1 single trial. Women with unexplained infertility is not targeted at individual patients, but at can be randomised to receive either expectant groups of patients. This can happen where the management (no treatment), insemination alone, intervention is an information package for the clomiphene alone or clomiphene and insemina- management of menorrhagia in primary care9 tion treatment as shown in Figure 21.1. or a clinical guideline for the management of When the two treatments (factors) do not infertility.10 In these studies, randomising patients interact with one another, this design has the to receive management using the information advantage of requiring half the number of patients package or clinical guidelines would have intro- that would be required if two parallel RCTs were duced contamination, since GPs would have to be conducted. The advantage is self-evident. In been expected to manage both study (information this case, the effect of IUI alone can be assessed leaflet, clinical guidelines) and control patients. by comparing groups C and D with A and B, Potentially this could underestimate the true effec- while the effect of clomiphene can be evaluated tiveness of the intervention. Therefore, clusters by comparing B and D versus A and C. The of patients (e.g. general practices) were randomly advantage is self-evident. In this case, the effect allocated to receive intervention (e.g. information of IUI alone can be assessed by comparing groups leaflets, clinical guidelines) or control. Cluster C and D with A and B, while the effect of randomisation should only be carried out when clomiphene can be evaluated by comparing B and there is a strong justification for doing so. D versus A and C. The primary implication of cluster randomised CLUSTER RANDOMISED TRIALS IN trials is that the measurements on individuals GYNAECOLOGY AND INFERTILITY are not statistically independent of one another; that is measurements from individuals within the A potential problem that can occur with ran- same cluster will be correlated to one another. domised controlled trials is where the intervention This has implications in the design (e.g. sample GYNAECOLOGY AND INFERTILITY 341 size), conduct (e.g. informed consent), analysis embryo. However, they noted that the differ- and reporting. Cluster randomised trials should ence in implantation rates was likely to be wider adjust for this clustering when determining the than that reported due to failure to adjust for number of patients required. The sample size the clustering of embryos/oocytes transferred to that would be required if patients were to be each woman. Studies where oocytes have been randomised must be inflated by a factor which randomised have no clustering implications since takes into account the extent of the clustering oocytes retrieved from the same women are ran- and the size of the cluster.11 The extent of domly allocated to receive ICSI or IVF. the correlation is measured by the intra-cluster When conducting randomised controlled trials correlation coefficient (ICC)12 and researchers are in infertility, consideration should be given to the required to have some indication of this, in order unit of randomisation and the outcome measures that the study can be adequately powered. Studies to be applied. When there is implicit clustering that fail to adequately inflate the sample size will in the data, the statistical analysis should account suffer from a type II error. for this using the methods described above.13 Similarly, the correlated responses obtained from each cluster have an implication for the QUASI-RANDOMISED TRIALS statistical analysis, since standard statistical tests (e.g. t-test) assume that observations are inde- These are controlled experimental studies where pendent of one another. There are a number of treatment allocation is performed on the basis of approaches to analysing cluster randomised tri- patient unit numbers or days of the week when als and these are detailed elsewhere.13 Failure to the patients are recruited. Although this design of account for the correlated responses in the anal- treatment allocation affords an element of chance, yses will result in an increased type I error. it cannot be considered to be genuine randomi- Clustering of outcomes can also occur in infer- sation. This type of design may still appeal to tility trials where alternative treatments are being those involved in laboratory trials involving incu- compared. For example, in randomised controlled bation or cryopreservation of human embryos. In these cases, it may be easier and cheaper to trials comparing IVF with ICSI the unit of allo- use a certain protocol for all embryos on alter- cation varies between patients,14 oocytes15 and nate days or alternate weeks rather than change cycles.16 Often, outcomes such as implantation the protocol or a freezer setting for each embryo rate and fertilisation rate are considered. These or each woman. The consequent loss of allo- are both expressed as percentages out of the total cation concealment will lead to serious inclu- number of oocytes retrieved. Hence, in trials that sion bias as some patients may be deliberately randomise patients (couples) or cycles and report excluded. This, is turn, can exaggerate treatment implantation or fertilisation rates, there will be effects. clustering of the outcome since oocytes are clus- tered within patients or cycles.14,16 Hence, in PATIENT PREFERENCE TRIAL these studies adjustment should be made in the analysis to adjust for the correlated outcomes A potential problem in some randomised trials assessed (on each oocyte) within patients (or arises when patients or their clinicians refuse to cycles). In trials that randomise by patients and be randomised on grounds of strong treatment report fertilisation of implantation rates, some preferences. Exclusion of these patients may adjustment is required. However, for outcomes affect the generalisability of the results as par- such as live birth rate or pregnancy rate no ticipants may not be representative. Yet recruit- adjustment is required since the percentages are ment of these patients may introduce substantial expressed out of the total number of patients ran- bias especially when it is impossible to blind domised. Bhattacharya et al.16 randomised cycles them. In addition, compliance and satisfaction and reported implantation rates per transferred may be higher with the preferred intervention.17 342 TEXTBOOK OF CLINICAL TRIALS
This is particularly so when the ‘new’ treatment numbers quickly add up to generate a total sam- is only available within the context of the trial ple size double that for a conventional trial. This or when, as in trials in unexplained infertility, has the predictable effect of adding to the cost and one of the arms comprises a ‘no treatment’ or duration of the trial. Entry of a disproportionate ‘expectant management’ group. Dissatisfaction number of patients into either the randomised or with the allocation may lead to differential com- the preference arms is also a problem, as the trial pliance and follow-up resulting in groups which will not be completed unless the appropriate num- cannot then be assumed to be similar. The out- ber have been recruited into the two components of come measures could also be affected by how the trial. The situation may be further complicated satisfied patients are with their allocated treat- by patients favouring one treatment over another, ment. The effect of patient preference on outcome making comparison of the two groups in the pref- would depend to a great extent on the specific erence arm more difficult. outcomes being assessed. If the principal out- A further problem with this approach lies in come is death or live birth, then the effect of the analysis. Any comparison using the non- patient preferences is likely to be small. If the randomised groups is unreliable because of principal outcome is satisfaction with care, then unknown and uncontrolled confounders. Patient the effect of patient preference is large.18 Under preference designs have been used in trials of , such circumstances the conventional randomised termination of pregnancy20 21 and menorrhagia.22 trial will underestimate the relationship between The evidence to support the use of PRPP trials the intervention and the outcome, i.e. show the compared with conventional randomised trials is minimum effect size. Conversely a comparison limited. A randomised comparison of the two 22 between two groups of patients who have chosen strategies by Cooper et al. suggested that the their treatment and thus optimised their treatment extra cost and complexity were not justified in choice will be considered to represent the max- the context of medical versus surgical treatment imum effect size. An intervention in question of menorrhagia. will have an effect size between the minimum A conventional randomised trial could address and maximum as derived from the randomised the effect that patient preference has on outcome 23 and the preference part of a partially randomised by recording this information before allocation. patient preference trial.18 This would allow resources to be concentrated on To deal with patient preferences within a recruiting as many patients as possible into the trial, the use of a partially randomised patient randomised comparison group but would allow preference (PRPP) trial has been suggested.19 stratification of the results by initial preference. Patients with strong preferences are allowed their desired treatment. Those without such views are EQUIVALENT TRIALS subjected to randomisation. For example in a trial Often in reproductive health care the aim is to of medical and surgical termination of pregnancy show that one treatment is as effective (equiv- we end up with four groups–randomised to alence), or no less effective (non-inferior), as medical, randomised to surgical, prefer medical another. The methodology for equivalence tri- and prefer surgical. als differs to that of superiority trials in design, Potential disadvantages with PRPP trials include analysis and interpretation. In designing equiva- effects of the trial size. The size of a total PRPP lence trials, attention must be given to defining cohort will need to be much larger than for a con- an equivalence margin. This is the difference ventional randomised controlled trial. As already in effect that would be deemed to be ‘clini- mentioned, the size of the randomised cohort needs cally insignificant’.24 In comparison with supe- to be the same as in a conventional trial. In addi- riority trials, larger sample sizes are needed tion, the number of patients in the non-randomised to demonstrate equivalence. In the analysis of preference arms needs to be of equivalent size. The equivalence trials, conventional statistical testing GYNAECOLOGY AND INFERTILITY 343 has little relevance and interpretation of results to generate debate amongst clinicians. Disagree- should be conducted though use of confidence ment about the definition of a particular condition intervals in relation to the predefined equiva- can lead to dismissal of the conclusions of a trial lence margin.25 Statistical significance is demon- as irrelevant. Certain terms continue to pose par- strated if the upper and lower limits of the 95% ticular problems. For example, infertility which confidence interval do not cross the equivalence is defined as ‘the lack of pregnancy after one margin.25 In superiority trials, the most conser- year of regular unprotected coital exposure’ refers vative analysis is by intention to treat (ITT). In to the lack of an exposure, i.e. pregnancy rather an equivalence trial, however, a ‘per protocol’ than a particular disorder. There may be contrib- (PP) analysis is usually considered statistically utory factors from both sexes, and definitions of more conservative. This is because an ITT anal- subgroups such as unexplained infertility or poly- ysis may blur the comparison between groups and cystic ovarian syndrome vary widely. Variation lead to an increased chance of declaring the two in laboratory procedures (such as semen analy- treatments as equivalent when they are not. The ses) may affect the diagnosis of male infertility decision about which should be the primary form while the investigations used for tubal patency of analysis (ITT or PP) in an equivalence study (laparoscopy versus hysterosalpingogram) may is not straightforward.26 It depends on the par- have an effect on the identification of endometrio- ticular characteristics of the study, including the sis in infertile women. definitions adopted for the ITT and PP analyses With menorrhagia, the problem is different. and the risk of bias.27 The conventional textbook definition of menor- rhagia (menstrual blood loss >80 mls) is imprac- PERFORMING AN RCT tical and the pragmatic approach is to include all women with a subjective complaint of heavy SYSTEMATIC REVIEW menstrual loss. This encourages a situation where critics can question the external validity of tri- A systematic review of the literature is an essen- als where women have been included either tial component of the pre-trial work-up. It enables on the basis of objective measurement of men- the researcher to define the clinical question in strual blood loss or on pragmatic grounds with the light of work that has gone before and assess increased self-reported bleeding. Some purists the need for a trial. It also provides vital infor- will argue that efficacy of treatments for men- mation about the limitations of previous trials, orrhagia cannot be evaluated accurately without outcome measures used and nature of follow-up. patients with ‘genuine’ pathology. However, the This is useful in planning the design of the pro- 28 outcome of such trials may not necessarily be posed study. A recent systematic review has applicable to the vast majority of clinical situa- identified typical problems associated with pre- tions where menstrual loss is not routinely mea- vious trials in unexplained infertility including sured. From a clinical point of view, it is probably small sample sizes, inappropriate outcome mea- more useful to use a pragmatic approach and sures (pregnancy rates per cycle) and lack of cost recruit all women on the basis of a subjective data. Similar information relating to other top- complaint of menorrhagia. ics in gynaecology can also be obtained from Studies on urinary incontinence also require Cochrane reviews. appropriate definition of the population group, as it is well known that women who attend DEFINING THE STUDY POPULATION urodynamic clinics constitute a small proportion Definition of the study sample is a vital part of of the total number of women in the community any clinical trial. Unfortunately this aspect of trial suffering from urinary incontinence. Although design can be contentious. The diagnostic crite- there is no way of completely avoiding selection ria of many gynaecological conditions continue bias, explicit description of the eligibility criteria 344 TEXTBOOK OF CLINICAL TRIALS allows the readers to draw their own conclusions Table 21.2. Types of interventions subjected to regarding the applicability of the data to their own clinical trials in gynaecology specific contexts. Those performing secondary Intervention Examples of trials research can also use these data to assess heterogeneity between trials. Packages of care Information packages in use A specific problem associated with infertility in general practice for trials is the question of how to deal with the appropriate treatment and referral in menorrhagia male partner. Conventionally it is the woman The value of guidelines in who undergoes treatment, and it is she who infertility for general is considered to be the participant in trials practitioners and subjected to recruitment, randomisation and Surgical techniques Hysterectomy versus follow-up. However, in trials where satisfaction, endometrial ablation acceptability and costs are outcomes, it is perhaps Different types of appropriate to seek the male partners’ views endometrial ablation, e.g. as well. TCRE versus Laser An important aspect of the choice of the study Drug trial Placebo versus tranexamic population involves the effect of the participants acid for menorrhagia on generalisability of the findings. Although Comparison of Medical versus surgical study participants may meet eligibility criteria, different treatment termination of pregnancy modalities participation is voluntary and volunteers may dif- Mirena IUS versus fer from the general population in terms of gen- endometrial ablation for eral health, co-interventions, educational level, menorrhagia motivation and ability to follow a protocol. Eth- Expectant treatment versus nic minorities may be missed on account of IVF for unexplained unfamiliarity with the language of the question- infertility naires used. Laboratory In vitro fertilisation (IVF) techniques versus intra-cytoplasmic INTERVENTIONS sperm injection Alternative methods of Due to its unique mix of medical and surgi- cryopreservation of human cal workload, gynaecology offers a number of embryos diverse interventions that need to be tested in Place of care One-stop specialist clinic the context of clinical trials. Some examples are versus general clinic Out-patient versus in-patient shown in Table 21.2. endometrial ablation Investigations Effectiveness of methods of DEFINING OUTCOMES screening for chlamydia trachomatis For any trial, it is crucial to have an apriori Hysterosalpingography as a hypothesis and a clinically relevant primary out- test of tubal patency come on which the power calculation is based. The post-coital test in the diagnosis of infertility This essential primary step prevents the sacri- Hysteroscopy in the fice of quality for expediency and discourages diagnosis of menorrhagia opportunistic and inappropriate manipulation of collected data. Outcomes of choice include those that are purely clinical, as well as others which its clinical context. This may involve differ- may be patient-centred or economic. The pre- ent levels of observation and analysis, incor- cise nature of the primary and secondary out- porating the individual, the family and the comes will depend on the type of trial and community.29 GYNAECOLOGY AND INFERTILITY 345
CLINICAL OUTCOMES incidence of hip fracture may be chosen as a prin- cipal outcome in trials of hormone replacement Clinical outcomes are essential components of therapy.30 any clinical trial. They should be meaningful Pragmatic trials usually require the evaluation to other clinicians and have a direct bearing on of more than one outcome measure in order to the women’s appreciation of the burden of dis- come to a decision about the effectiveness, risks, ease. Generally speaking, outcomes which are costs and acceptability of an intervention. For represented as discrete or categorical variables example, in surgical trials of menorrhagia out- are often more meaningful than continuous out- comes should include satisfaction with treatment, comes. For example, the proportion of women menstrual flow, pain, premenstrual syndrome and requiring blood transfusion after hysterectomy is period of recovery. Sometimes when the impact seen to be a more clinically relevant outcome than of a disease spreads beyond the individual to a the volume of blood lost during the procedure. wider group such as the family, GPs or carers, The proportion of women in whom the uterus is outcomes may need to be expanded to include a empty at 24 hours may be a more meaningful out- wider group. This may be relevant in trials of uri- come than the mean number of hours required for nary incontinence or HRT. It is however impor- medical termination. In reproductive medicine, tant to remember that it is the primary outcome on many clinical trials have tended to choose surro- which the power calculation is usually based, and gate markers such as number of oocytes retrieved the one that the trial is best designed to address. or fertilised as primary outcome rather than live birth or pregnancy. The reason for such a choice is not difficult to guess. Fewer participants are QUALITY OF LIFE required to achieve a ‘significant’ difference if the outcome is represented as a continuous vari- Quality of life (QOL) is now accepted by most able (such as number of oocytes) instead of a clinicians as an important outcome in clinical dichotomous variable such as live birth. This in trials.31 However the term is sometimes used turn reduces the sample size and costs, allowing loosely and without a clear understanding of what a single centre to perform a trial that would oth- it means.32 Since QOL is considered to be a erwise require far higher levels of funding and a complex concept comprising physical, emotional multi-centre approach. and other dimensions, most questionnaires in Explanatory trials usually rely on a single clin- common use not only assess the detailed aspects ical outcome. For example, in a trial comparing of QOL but also provide a summary score for drug treatments for menorrhagia, menstrual blood overall health status.33 Generic measures such loss in millilitres may well be an appropriate pri- as short form health survey (SF-36)34 broadly mary outcome. Other physiological or biochemi- assess physical, mental and social health and can cal outcomes such as haemoglobin level, volume be used to compare conditions and treatments. urinary loss, extent of endometriosis visualised Although the number of such instruments in by laparoscopy, number of ovarian follicles seen current use is rapidly increasing, there is a on ultrasound scan and serum estradiol levels fol- remarkable level of consistency between them.33 lowing ovarian stimulation may also be used in Other methods include tools focusing on a sin- different situations. Unfortunately, they may not gle aspect such as pain, anxiety as well as indi- always correlate well with the clinically relevant vidualised measures in which patients themselves outcomes–certainly from the patients’ perspec- define and rate the most important aspects of tive. In certain cases, costs and a need to opt their quality of life.35 A number of condition- for a modest and realistic sample size dictate specific tools, which can be used either indepen- the need for surrogate outcomes in preference dently or to supplement generic measures, have to more robust but less common substantive out- been developed.36 Examples include the King’s comes. Thus bone mineral density rather than the College Questionnaire for Urinary Incontinence37 346 TEXTBOOK OF CLINICAL TRIALS and Menstrual Distress questionnaire.38 The the sum total of a number of patient-related fac- Endometriosis Health Profile-3039 and The Meno- tors, including expectations, characteristics and pause Rating Scale (MRS).40 psychosocial determinants.44 Over the past few A systematic review by Sanders et al.41 showed years, patient satisfaction has become increas- that despite the plethora of instruments, the ingly accepted as a measure of quality in health prevalence of reporting on quality of life remains services and a valid outcome in randomised clin- low, increasing from 1% in 1980 to 4% in 1997. ical trials.45 This is particularly significant in the There is also a general unwillingness to ask current climate of delivery of health care which patients to supplement questionnaire-based data aims to provide a patient-centred service with with personal responses, and lack of appreciation greater public involvement in planning.46 The about the critical importance of response rates. purpose of patient satisfaction measurement is to Patients themselves can find it difficult to dis- describe health care services from the patient’s tinguish between quality of life and health status point of view, measure the ‘process’ of care and or to rate their health without a point of reference. evaluate health care.44 The particular strength of At the same time, the effects of age and chang- using satisfaction as an outcome is related to the ing expectations need to be adjusted for when unique circumstances of certain gynaecological interpreting QOL scores. Overall, QOL offers trials such as those used for menorrhagia where a superior way of assessing treatment success not only the interventions but also the clinical in trials involving general gynaecology (such as outcomes may be dissimilar. In a trial of hys- menorrhagia, urinary incontinence, menopause, terectomy versus endometrial ablation, women pre-menstrual tension) where interventions are would be expected to be amenorrhoeic follow- targeted at women with benign but debilitating ing hysterectomy but not after ablation. Here, illnesses that compromise several key areas of comparison of amenorrhoea rates is unlikely to day-to-day life. On the other hand, women seek- be helpful in comparing the two groups, while ing fertility treatment or abortion services are satisfaction is not only a robust measure of treat- not necessarily unwell. The aim of treatment ment success, but also incorporates the sum total is to enhance their physical and mental well- of a woman’s experience of the alternative treat- being rather than correct a pre-existing deficit ment arms including discomfort, recovery time in health status. Existing instruments do not dis- and side-effects. A similar argument can be used criminate between these two broad groups and to justify the use of the same outcome for trials further refinements are needed with respect to comparing surgical and non-surgical treatment of assessing positive aspects of general and sexual urinary incontinence. health as opposed to the conventionally used neg- Despite their widespread use in clinical trials, ative aspects.42 Meanwhile, simple global ques- assessment of patient satisfaction has been criti- tions on self-reported health or QOL continue to cised on theoretical and methodological grounds be useful as prognostic measures for stratification and their practical use questioned.47 Relatively of treatment allocation and as important outcome few patients express open dissatisfaction with measures alongside purely clinical outcomes. treatment.48 Indeed satisfaction rates of 80% or more are reported by most hospital-based 49 PATIENT SATISFACTION studies. There is also little evidence to indi- cate that expressions of satisfaction result from There continues to be a general lack of agreement the fulfillment of expectations; in some situ- about the mechanisms which produce satisfac- ations it is difficult to establish the fact that tion, as well as the meaning of the word ‘satisfac- expectations exist at all. High satisfaction rat- tion’ itself which has been defined as an ‘evalu- ings do not necessarily mean that women have ation based on the fulfillment of expectations’.43 had good experiences in relation to the ser- The conventional view is that satisfaction reflects vice as satisfaction may well make allowances GYNAECOLOGY AND INFERTILITY 347 for mitigating circumstances. If the aim is to not only for evaluating interventions, but also provide women with a voice, it is important not comparing costs.53 The costs of treatments are to rely on satisfaction with treatment as a sin- usually estimated using information about the gle outcome but to prioritise methods of access- quantities of the resources used. For example ing women’s experience of interventions and the the resources used for hysterectomy include the meaning and value they attach to them.47 There staff time involved, the consumables used and are no off-the-shelf questionnaires that are com- the length of the subsequent in-patient stay. pletely satisfactory50 and qualitative studies have To estimate the cost of treatment, information demonstrated that high satisfaction rates cannot about this resource use is combined with unit be taken as proof of positive experience. Many cost estimates, which provide a fixed monetary tools mentioned in the literature are not val- value to each cost-generating item.54 The total idated, while many expressions of satisfaction cost is then the weighted sum of quantities of may not be evaluations at all.51,52 Dissatisfac- resources used where weights are unit costs. tion may be more useful as a minimum level of Carrying out an economic evaluation alongside negative experience and may be of potential use a randomised trial allows detailed information to in benchmarking exercises. At the moment most be collected about the quantities used by each clinical trials in gynaecology attempt to mea- patient in the study. Such information allows a sure satisfaction using a number of direct and cost for each patient, producing a patient-specific indirect questions. Some of these questions have cost data. This is turn reduces the extent to been repeated at various points during follow-up which comparison between the groups is based to assess change in satisfaction rates over time. on assumptions about resource use. However Despite the obvious shortcomings of the existing randomised trials are not necessarily the only way system, there has been an opportunity to refine or necessarily the best way to address economic and validate some of these questionnaires through questions.55 There is an important role for other 4 repeated use in a series of related trials. Accept- methods, including modelling. ability has been measured by direct questions and In the context of RCTs however there is an by other tools such as the Semantic Differen- urgent need to revise the way in which health eco- tial technique in the context of menorrhagia and nomic outcomes are addressed within a clinical 20–22 termination. trial. While cost outcomes are generally regarded In other areas such as infertility, satisfaction as secondary outcomes, the rationale for a for- with treatment is more difficult to assess as the mal sample size calculation with adequate power effect of the desired outcome (live birth) is pre- for the planned analysis is still relevant given the dominant even where treatment is invasive or large variability in costs between individuals.56 unpleasant. Conversely there is dissatisfaction This is even more relevant where subsets are with treatment where the outcome is failure to used for cost data for practical reasons. A recent fall pregnant. Some attempts have been made review has identified an urgent need to improve in recent trials to specifically address separately the statistical analysis and interpretation of cost satisfaction with ‘treatment’ as opposed to satis- data in RCTs.54 This is particularly relevant to the faction with ‘outcome’. This area is deserving of provision of descriptive statistics relating to costs. further study. As cost data are typically skewed, the median can be interpreted as the typical cost for individuals. ECONOMIC EVALUATION However, it is the mean cost that is important for With the emergence of new methods of treatment policy decisions as it is this value, multiplied by comes an increasing awareness of the need to the number of patients, which gives an estimate study not just the clinical effectiveness but also of the total cost of an intervention. the cost-effectiveness of alternative treatments. Table 21.3 provides some examples of outcome Pragmatic clinical trials are the standard approach measures used in different types of gynaecological 348 TEXTBOOK OF CLINICAL TRIALS
Table 21.3. Outcomes in gynaecological trials Clinical area Outcomes Comments
Infertility • Live birth rate per couple Although live birth per couple is the most • Live birth rate per treatment robust outcome, it demands large sample • Clinical pregnancy rate per couple sizes and a longer duration of follow-up. • Clinical pregnancy rate per Live birth/clinical pregnancy rate per treatment treatment is still used in many trials. • Biochemical pregnancy rate Multiple pregnancy and its effect on • Fertilisation rate maternal and perinatal morbidity is • Implantation rate increasingly being acknowledged as an • Multiple pregnancy important outcome of fertility trials. • Morbidity (e.g. ovarian hyperstimulation) • Costs Menorrhagia • Satisfaction Satisfaction and QOL are clinically more • Acceptability useful than objective measurement of • Quality of life menstrual blood loss or amenorrhoea • Menstrual blood loss rates, especially when trials compare • Bleeding and pain scores treatments such as hysterectomy which • Morbidity guarantees amenorrhoea versus the • Repeat surgery Mirena intrauterine system or • Haemoglobin level endometrial ablation which do not. • Amenorrhoea rates Satisfaction with treatment may not • Costs correspond to amenorrhoea rates. Long-term follow-up is important in the evaluation of all new technologies. Urogynaecology • Satisfaction Symptom relief and objective assessment of • Acceptability bladder function may not necessarily • Quality of life correspond with quality of life or • Symptom relief satisfaction. • Objective measurement of urinary Long-term follow-up is necessary for loss effective evaluation of treatments. • Surgical morbidity, repeat surgery • Length of hospital stay • Urodynamic assessment • Costs Hormone replacement • Hip fracture Historically, surrogate outcomes like lipid therapy • Cardiovascular disease profile and bone density have been more • Menopausal symptoms popular than rates of cardiovascular • Quality of life disease or fracture. • Satisfaction • Acceptability • Bone mineral density • Serum lipid profile • Side-effects and morbidity Termination of • Efficacy: evacuation of the uterus Quality of life is difficult to assess in the pregnancy • Acceptability context of termination. • Morbidity Long-term follow-up difficult. • Quality of life • Costs GYNAECOLOGY AND INFERTILITY 349 trials. A crude list such as this is useful, if only to between groups of a certain magnitude. It is illustrate the specific demands of different clinical important to ensure that the study is designed areas. Overall, due to the limitations of using ‘pure’ to detect significant differences if they exist. clinical outcomes in benign gynaecology, ‘satis- Conversely if the statistical power is low, the faction’ and ‘quality of life’ (however defined) results of the trial will be questionable as the have found widespread acceptance as appropri- numbers may have been too small to detect ate outcomes. In other areas such as infertility, genuine differences. In general, a clinically ‘satisfaction’ is meaningless without the promise significant difference in the primary outcome of live birth while even the most invasive and should be identified as the point of reference for uncomfortable treatment may be perceived to be a sample size calculation. Intimate knowledge entirely acceptable if it leads to pregnancy. In gen- of the clinical area is crucial for this. For eral, even when relevant, purely clinical outcomes example a 20% difference in satisfaction rate may lead to potential conflicts between the clin- between two forms of treatment for incontinence icians’ and patients’ points of view. A number may be considered to be clinically important. of health state measures incorporating validated Conversely, against a background of low live and reliable scales have been developed to address birth rates, a difference of 5% to 10% may this very issue.57 These may be generic or disease- be enough to change clinical practice in an specific. Most pragmatic trials will use a number infertility trial. of outcomes from the above categories. At the In determining the sample size adequate atten- same time, it is best, in very large trials, to con- tion should also be paid to the possibility of centrate on a few simple outcomes–for reasons of sample attrition and the need for any future sub- convenience and efficiency.58 There is also a sta- group analysis. For example, in abortion trials, a tistical drawback to the use of multiple outcomes. high non-response to follow-up should be antici- The greater the number of outcomes, the higher pated and the sample size inflated accordingly.21 is the possibility of one of them reaching statisti- In infertility trials, where it may be clinically cal significance on the basis of chance alone. It is important to assess the effect of the interven- important to consider relevance of outcome mea- tion in different clinical groups, a similar exercise sures to the stakeholders. It is thus important to will ensure meaningful subgroup analysis. At the predefine primary and secondary outcomes. The same time, aiming for unrealistically large sample extent to which a trial changes practice will depend sizes is counterproductive and should be avoided. on the outcomes chosen. With a large sample size it is almost always pos- sible to reject any null hypothesis (type I error). Conversely samples which are too small have a SAMPLE AND SAMPLE SIZE high risk of failing to demonstrate a real differ- ence (type II error). The latter is more frequent in The sample size refers to the number of women gynaecological trials. In small trials, a subgroup needed to provide adequate power (usually 80% analysis based on tiny numbers of patients should to 90%) in order to show that the findings of the be perceived as a hypothesis generating exercise. trial are not merely due to chance. The sample size for each trial is usually calculated with the RANDOMISATION primary outcome in mind. Although secondary outcomes are often investigated and subgroup Randomisation involves allocating women to analyses performed, the power of an RCT to groups such that individual characteristics do provide conclusive answers to these may be not influence the nature of the intervention. For limited. The statistical approach to determining example in a trial of treatments for menorrhagia, sample size is the power calculation, which the aim is to avoid bias by distributing factors that determines how likely the study is to produce may influence outcome, such as age, parity, dys- a statistically significant result for a difference menorrhea, premenstrual syndrome and uterine 350 TEXTBOOK OF CLINICAL TRIALS
fibroids, randomly between treatment groups. It internet-based randomisation systems continue to is anticipated that any difference in outcome is generate concerns about ensuring security and purely due to the treatment and not influenced by confidentiality of patient details. Practical prob- one or more of these other characteristics. Ran- lems with randomisation may arise in surgical dom allocation does not guarantee that the groups and laboratory-based trials where randomisation will be identical but it does ensure that any dif- may need the assistance of nursing or techni- ferences between them are due to chance alone. cal staff. Randomisation also facilitates the concealment While simple randomisation techniques will, of the type of treatment from the researchers on average, allocate equal numbers to each arm, and subjects to further reduce bias in treatment occasionally, even in large trials, groups of dif- comparison. Thus it ensures that women with a ferent sizes can result. Block randomisation can higher BMI (body mass index) are not prefer- be used to keep the numbers in each group very entially allocated to endometrial ablation rather close at all times. In a trial of two alternative sur- than hysterectomy. In addition, it leads to treat- gical treatments for menorrhagia we might want ment groups which are random samples of the to ensure that each surgeon treats similar num- population sampled and thus makes valid the use bers of women by either method. Stratified ran- of standard statistical tests based on probabil- domisation produces a separate randomisation list ity theory. for each surgeon (stratum) so that we get very While the simplest method of randomisation similar numbers of patients receiving each treat- is tossing a coin, in practice, this is not an ment within each stratum. If envelopes are used, accepted method of treatment allocation. The this may involve separate lists of random num- main reason for this is the lack of an audit trail bers and separate piles of sealed envelopes for that makes it difficult to confirm that the ran- each surgeon. We may additionally use blocks to dom allocation was done correctly. For these ensure that there is a balance of treatments within reasons the random allocation should be deter- each stratum. While stratified randomisation can mined in advance, preferably by using pseu- be extended to two or more stratifying variables, dorandom numbers generated by a mathemat- we have to be careful to include only a few strata, ical process. After the randomisation list has to prevent generating extremely small subgroups. been prepared (by someone who will not be Stratification by centre is standard practice in involved in recruitment), it must be made avail- multi-centre trials. able to researchers. Although the process of ran- In small studies with several important prog- domisation can occur at the recruitment point nostic variables such as infertility trials, ran- this is preferably done at long range, by tele- dom allocation may not provide adequate bal- phone or even the internet. If envelopes are used, ance within the groups. The lack of numbers may these must be opaque, as researchers could the- make it difficult to stratify for all the important oretically hold envelopes to a lamp in order variables. Here, it is still possible to achieve bal- to read what is written inside. For the same ance using minimisation, which is based on the reason these envelopes should be sequentially concept that the next patient to enter the trial numbered so that the recruiter has to take the is allocated to whichever treatment would min- next envelope. Differences in outcome between imise the overall imbalance between groups at treatment groups are considerably larger in tri- any stage of the trial. Even in small trials this als where allocation concealment is not strictly provides groups that are comparable across sev- enforced as this produces a clear bias. Tele- eral prognostic factors. It is important to specify phone randomisation, either by means of an oper- exactly which prognostic variables are to be used ator or a computer-operated 24-hour phone line, and to say how they are to be grouped. For is ideal for large trials and especially multi- example age, previous pregnancy and duration centre trials. Although potentially more efficient, of infertility are important prognostic factors for GYNAECOLOGY AND INFERTILITY 351 fertility. Minimisation in this context will require costs. Often there is an imperative to provide a statement about the actual age groups, for ‘treatment’ at the request of the couples. This example <30 years and ≥30 years. Minimisation makes it difficult to recruit couples into a clin- is crucial in infertility trials where a clinically sig- ical trial where one of the options is ‘expectant nificant difference in live birth rates associated management’. with alternative treatments is small and easily overpowered by the effect of prognostic factors CONCEALMENT OF ALLOCATION such as age, parity and previous pregnancy. Occasionally we allocate a group of subjects The unpredictability of the randomisation pro- together rather than individuals to treatments. For cess can only be successful if followed by example, in a health promotion study carried allocation concealment, i.e. concealment of the out in general practices, we might need to sequence until patients have been assigned to 59 apply the intervention to all patients in the their groups. This ensures strict implementa- practice. This may involve display of publicity tion of a random allocation sequence without material in the waiting room, for example. In foreknowledge of treatment assignments. Aware- this situation, we may need to keep groups of ness of the next treatment allocation could lead patients separate in order to avoid contamination. to exclusion of certain women based on their In a different setting, if we are providing a prognosis because they would have been allo- special physiotherapist to advise patients in a cated to the perceived inappropriate group. For ward, it would be difficult for the nurse to visit example, in a trial of unexplained infertility, some patients and not others. If we are providing women with a prolonged duration of infertility training to the patients or their carers, we do not could be excluded if the next treatment allocation want the subjects receiving training to pass on were known to be a ‘no treatment’ arm. Adequate what they have learned to controls. This might be concealment would ensure that the decision to desirable in general, but not in a trial. A group of accept or reject a participant should be made and informed consent obtained without prior knowl- subjects allocated to a treatment together is called edge of the nature of the assignment. a cluster. Clusters must be taken into account Trials that use inadequate or unclear allocation in the design as the use of clusters reduces the concealment have tended to yield 40% larger esti- power of the trial and so requires an increase in mates of effect compared to those which used sample size. adequate concealment.60–62 Trials with poorly A practical problem relating to randomisation concealed allocation also generated greater het- concerns the emotive nature of some of the con- erogeneity in results, i.e. the results fluctuated ditions under evaluation such as infertility or extensively above and below the estimates from termination of pregnancy. Some women may be better studies.60 unwilling to accept the extra stress of participat- ing in a trial over and above what is already a BLINDING complex and psychologically challenging experi- ence. There may also be compelling social rea- Double blinding seeks to prevent ascertainment sons why women undergoing termination are less bias, protects the sequence after allocation and likely to opt for randomisation, comply with trial cannot always be implemented.1 As in the case protocols and follow-up arrangements. Infertile with allocation concealment, lack of blinding couples may be required to fund their treatment may lead to exaggerated estimate effects of treat- themselves. This could influence their decision to ment. A survey of trials in gynaecology found refuse to participate in a trial where the experi- that investigators could have used double blind- mental arm (such as assisted hatching) is substan- ing more often.1 When used, methods of double tially more expensive than standard treatment, blinding were poorly reported and rarely eval- unless the trial organisers offer to absorb the extra uated. It is recommended that authors provide 352 TEXTBOOK OF CLINICAL TRIALS adequate information about the methods used treatment irrespective of randomised assignment to ensure double blinding. This should include (analysis by treatment received). Secondary anal- details such as the type of intervention (cap- yses are acceptable, as long as researchers label sules/tablets), and efforts made to duplicate the them as such, and as non-randomised compar- characteristics of the treatment (taste, appear- isons. The advantage of randomisation is entirely ance, route of administration). In addition it is lost when investigators exclude participants and important to be explicit about the methods used in effect present a non-randomised comparison as to control the allocation schedule, such as loca- the primary result, i.e. similar to a cohort study. tion of the schedule during the trial, details of Exclusions of participants can lead to misleading when the code was broken for analysis and the results.64 Researchers sometimes exclude patients circumstances under which the code could be on the basis of outcomes that happen before treat- broken for individual cases (adverse reactions). ment has begun such as pregnancy in a couple Finally there should be a statement about the with infertility. Although this may seem sensi- perceived success or failure of the double blind- ble in as much as the event of interest occurred ing efforts. independent of the treatment, the same argument could be used for excluding pregnancies in a no EXCLUSIONS intervention arm of the trial. It is important to attempt to minimise exclu- Exclusions can occur due to eventual discov- sions and be explicit about those cases where ery about ineligibility, deviations from protocol, exclusions occurred. This can be enforced by withdrawals or losses to follow-up. Exclusions minimising the delay between randomisation and before randomisation do not affect the internal initiation of treatment. This can be particularly validity of the trial but can compromise general- relevant to infertility trials where couples could isability. For most pragmatic trials it is important fall pregnant before treatment can start or where to keep the eligibility criteria to a minimum. In the intervention is conditional on a set of clini- practice it is unusual to find significant qualitative cal criteria. For example, in couples randomised differences between women in trials and those to IVF or ICSI it may be more efficient to in the general population. Exclusions after trial delay randomisation until after oocyte recovery entry represent a further source of bias within so that women who have failed to respond to an RCT as any erosion over the course of the gonadotrophin stimulation are not included. trial from those initially randomised groups is not likely to be random in nature. The accepted FOLLOW-UP method of primary analysis in all cases is by ‘intention to treat’, i.e. analysis of patients in It is important to pre-determine the length and the originally assigned groups regardless of any type of follow-up for each trial. The precise breaches of protocol.63 This can prove unnerving circumstances and the time interval will depend for clinicians especially in the context of surgical on the nature of the trial. In fertility trials trials. For example in a trial comparing hysterec- the traditional method was to express outcomes tomy versus endometrial ablation many clinicians as pregnancy rates per cycle. This meant the would find it difficult to accept results of anal- duration of follow-up was brief. For more robust ysis of amenorrhoea rates by intention to treat outcomes like pregnancy rate per woman, it may arguing that it is inappropriate to include hys- be necessary to extend the follow-up for three terectomised women in the ablation group as this to six cycles depending on the nature of the would lead to an overestimation of amenorrhoea treatment. A further 9 months need to be added rates. Investigators can also do secondary analy- on to allow live birth per couple to be used ses, preferably pre-planned based on only those as an outcome. For menorrhagia trials, 80% of participants who fully complied with the trial pro- re-treatments occur within 2 years, making this tocol (per protocol) or who received a particular an acceptable duration for follow-up in the first GYNAECOLOGY AND INFERTILITY 353 instance. A prolonged period of follow-up of reminders put into place. In pragmatic trials it up to 5 years would be ideal as many women is often important to distinguish those women could expect the effects of their treatment to wane who no longer wish to continue with the allo- over time and long-term complications of therapy cated treatment from those who wish to terminate to surface. This would appear to be equally their involvement with the trial and do not wish to true for urogynaecology trials. For termination of be contacted for follow-up or have questionnaires pregnancy, follow-up has to be kept short as the sent to them. Hopefully the numbers in this lat- loss to follow-up is high and many women may ter group should be small but their wishes should not wish to be contacted at a later date. For HRT be respected. trials, which genuinely wish to address crucial outcomes such as rates of fracture, cardiovascular Data Entry and Analysis disease or Alzheimer’s disease, follow-up may need to be extended to tens of years. This This is an important aspect of the trial and errors obviously raises significant ethical, logistic and here can lead to significant bias. As mentioned financial issues which may well need to be taken above, analysis should be by intention to treat. into account whilst planning such trials. Each woman should be analysed as though she had received the intervention to which she had DATA COLLECTION been randomised. This minimises any bias due Data in a trial are usually collected from sources to non-random removal of participants from such as case notes, local clinic databases and the trial. The exception is explanatory trials, patient questionnaires. Occasionally interviews usually phase I and II drug trials, where strict may be used to explore areas which are not capa- rules of exclusion for protocol violation apply. ble of being probed adequately with question- Occasionally it may be important from a clinical naires. General practitioners, local and national point of view to perform as separate analysis databases may also be accessed to obtain clini- by treatment received. This should be clearly cal information such as retreatment rates or seri- described as such and should be used to assess ous complications about patients who are lost to the primary outcome. Intention to treat can cause follow-up. much consternation among clinicians particularly in surgical trials where some outcomes may seem CONDUCT absurd–for example continuing menstrual loss in women allocated to hysterectomy who did Recruitment not undergo the operation but were analysed by intention to treat. To avoid recruitment bias, it is important to target all eligible women and record all refusals. It may be helpful to obtain some baseline Presenting Results clinical details about them in order to explore any major differences between participants and Analysis should follow the original plan set out non-participants, which could affect the external in the protocol and the CONSORT recommen- validity of the trial. dations should be observed. Particularly helpful is a trial chart which sets out in an explicit manner any exclusions or loss to follow-up. Trial Co-Ordination Results of subgroup analyses should be treated Following informed consent, it is important to with caution and used mainly as hypothesis- obtain baseline information by filling in datasheets generating exercises in most modest-sized trials. or questionnaires prior to randomisation. Subse- There should be a conscious attempt to limit quent data collection should occur at the pre- discussion to the results generated by the trial specified times and an efficient system of timely and avoid speculation. 354 TEXTBOOK OF CLINICAL TRIALS
ETHICS OF TRIALS arm demonstrates a significantly higher rate of infection.66 Alternatively, in a trial comparing a The scientific rationale for conducting trials is policy of single versus double embryo transfer collective equipoise. Clinicians need to be gen- (in order to prevent twin pregnancies) it may be uinely uncertain about the best treatment. In such appropriate to stop if the pregnancy rate in the a clinical situation, there should be no conflict single embryo group becomes unacceptably low. between the interests of those participating in a trial and those who stand to gain in the future. The important issue is that participants are also CONCLUSION in personal equipoise and give informed consent. Despite awareness of its importance, there is Clinical trials in gynaecology have lagged behind evidence that some doctors do not seem to take those in other disciplines in terms of overall informed consent as seriously as they should.65 numbers as well as quality. There are few large This may well be because participants seem to multi-centre trials, particularly surgical trials. The be less willing to be randomised, when they are clinical population is heterogeneous and interven- given more preliminary data, and made aware of tions under scrutiny diverse. Some treatments, any accumulating evidence of effectiveness. In such as those for infertility and unwanted fer- many trials, a significant number of participants tility target women (and their partners) who have emerge from consultations expecting to benefit specific reproductive health needs but are oth- personally by their participation. erwise in good health. Trials also need to be Some infertility-related procedures are de- able to compare interventions that cross different scribed as ‘licensed treatment’ under the aegis of treatment boundaries. Trialists in this field need the Human Fertilisation and Embryology Author- to design more pragmatic trials with clinically ity (HFEA) in the United Kingdom. Clinical meaningful outcome measures. In gynaecology data pertaining to licensed treatments (including these should be quality of life and satisfaction; donor insemination, IVF and ICSI) are confi- in infertility, live birth rates per couple/woman. dential and may not be revealed to researchers Finally, the importance of collecting cost data (including clinicians) who are not covered by cannot be overstated in terms of planning gynae- the institutional HFEA treatment licence, without cological services which are effective, acceptable the explicit permission of the couple. This can and affordable. create problems in accessing data, particularly follow-up data from notes or databases. Further- more, trials involving manipulation of gametes REFERENCES and embryos need separate approval from the HFEA in addition to approval from the local 1. Schulz KF, Grimes DA, Altman DG, Hayes RJ. Blinding and exclusions after allocation in ran- ethics committee. domised controlled trials: survey of published par- For all clinical trials, it is sensible from an ethi- allel group trials in obstetrics and gynaecology. cal and financial point of view to have clear stop- BMJ (1996) 312(7033): 742–4. page rules as part of the original study design. An 2. Out HJ, Schnabel PG, Rombout F, Geurts TB, independent data monitoring committee should Bosschaert MA, Coelingh Bennink HJ. A bioe- quivalence study of two urinary follicle stimulat- be available to review the results of an interim ing hormone preparations: Follegon and Metrodin. analysis. Early stopping should only occur under Human Reprod (1996) 11(1): 61–3. pre-planned, well-specified circumstances such a 3. El-Refaey H, Rajasekar D, Abdalla M, Calder L, marked superiority or toxicity of one arm of Templeton A. Induction of abortion with Mifepri- the study which is greater than that originally stone (RU 486) and oral or vaginal Misoprostol. N Engl J Med (1995) 332: 983–7. hypothesised. Examples include stopping a trial 4. The Aberdeen Endometrial Ablation Group. A evaluating the use of prophylactic antibiotics randomised trial of hysterectomy versus endome- during hysteroscopic surgery where the control trial ablation for the treatment of dysfunctional GYNAECOLOGY AND INFERTILITY 355
bleeding: clinical psychological and economic out- intra-cytoplasmic sperm injection (ICSI) in non- come at four years. Br J Obstet Gynaecol (1999) male factor infertility. Lancet (2001) 357: 2075–9. 106: 360–6. 17. Torgerson D, Sibbald B. Understanding clinical 5. Cooper KG, Jack SA, Parkin DE, Grant AM. Five- trials: what is a patient preference trial? BMJ year follow up of women randomised to medical (1998) 316: 360. management or transcervical resection of the 18. Brocklehurst P. Br J Obstet Gynaecol (1997) 104: endometrium for heavy menstrual loss: clinical 1332–5. and quality of life outcomes. BJOG: Int J Obstet 19. Brewin CR, Bradley C. Patient preferences and Gynaecol (2001) 108(12): 1222–8. randomised clinical trials. BMJ (1989) 299: 6. Roland M, Torgerson DJ. Understanding controlled 313–15. trials: what are pragmatic trials? BMJ (1998) 316: 20. Henshaw RC, Naki SA, Russell IT, et al.Com- 285. parison of medical abortion with vacuum aspi- 7. Goverde AJ, McDonnell J, Vermeiden JPW, Schats ration; women’s preferences and acceptability of R, Rutten FFH, Schoemaker J. Intrauterine insem- treatment. Br Med J (1993) 307: 714–17. ination or in-vitro fertilisation in idiopathic sub- 21. Ashok PW, Kidd A, Flett GM, Fitzmaurice A, fertility and male subfertility: a randomised trial Graham W, Templeton A. A randomised compari- and cost-effectiveness analysis. Lancet (2000) 355: son of medical abortion and surgical vacuum aspi- 13–18. ration at 10–13 weeks gestation. Human Reprod 8. Khan KS, Daya S, Collins JA, Walter SD. Empir- (2002) 17(1): 92–8. ical evidence of bias in infertility research: over- 22. Cooper KG, Parkin DE, Garret AM, Grant AM. estimation of treatment effect in crossover trials A randomised comparison of medical and hystero- using pregnancy as the outcome measure. Fertil scopic management in women consulting a gynae- Steril (1996) 65(5): 939–45. cologist for treatment of heavy menstrual loss. Br 9. Fender GRK, Prentice A, Gorst T, Nixon RM, J Obstet Gynaecol (1997) 104: 1360–6. Duffy SW, Day NE, et al. Randomised controlled 23. Torgerson DJ, Klaber-Moffett J, Russell I. Patient trial of educational package on management preferences in randomised trials: threat or oppor- of menorrhagia in primary care: the Anglia tunity? J Health Serv Policy (1996) 1(4): 194–7. menorrhagia education study. BMJ (1999) 318: 24. Piaggio G, Pinol APY. Use of the equivalence 1246–50. approach in reproductive health clinical trials. Stat 10. Morrison J, Carroll L, Twaddle S, Cameron I, Med (2001) 20: 3571–87. Grimshaw J, Leyland A, et al. Pragmatic ran- 25. Jones B, Jarvis P, Lewis JA, Ebbutt AF. Trials to domised controlled trial to evaluate guidelines for assess equivalence: the importance of rigorous the management of infertility across the primary methods. BMJ (1996) 313: 36–9. care–secondary care interface. BMJ (2001) 322: 26. Rohmel¨ J. Therapeutic equivalence investigations: 1–7. statistical considerations. Stat Med (1998) 17: 11. Kerry SM, Bland JM. Sample size in cluster 1703–14. randomisation. BMJ (1998) 316: 549. 27. Ebbutt AF, Frith L. Practical issues in equivalence 12. Kerry SM, Bland JM. The intracluster correlation trials. Stat Med (1998) 17(15 & 16): 1691–1701. coefficient in cluster randomisation. BMJ (1998) 28. Pandian Z, Bhattacharya S, Nikolaou N, Vale L, 316: 1455. Templeton A. In-vitro fertilisation for unexplained 13. Donner A, Klar N. Design and analysis of cluster subfertility (Cochrane Review). In: The Cochrane randomization trials in health research, 1st edn. Library, Issue 4. Oxford: Update Software (2002). London: Arnold (2000). 29. Roland M, Torgerson DJ. Understanding controlled 14. Aboulgar MA, Mansour RT, Serour JI, Amin YM, trials: what outcomes should be measured? BMJ Kamal A. Prospective controlled randomised study (1998) 317: 1075–80. of in-vitro fertilisation versus intracytoplasmic 30. Cranney A, Welch V, Adachi JD, Guyatt G, sperm injection in the treatment of tubal factor Krolicki N, Griffith L, Shea B, Tugwell P, infertility with normal semen parameters. Fertil Wells G. Etidronate for treating and preventing Steril (1996) 66: 753–6. postmenopausal osteoporosis (Cochrane Review). 15. Ruiz A, Remohi J, Minguez Y, Guanes P, Simon In: The Cochrane Library, Issue 2. Oxford: Update C, Pellicer A. The role of in vitro fertilisation Software (2002). and intracytoplasmic sperm injection in couples 31. Patrick DL, Bergner M. Measurement of health with unexplained infertility after failed intrauterine status in the 1990s. Ann Rev Public Health (1990) insemination. Fertil Steril (1997) 68: 171–3. 11: 1165–83. 16. Bhattacharya S, Hamilton MPR, Shabban M, Kha- 32. Gill TM, Feinstein AR. A critical appraisal of laf Y, Seddler M, Ghobara T, et al. A randomised the quality of life measurements. JAMA (1994) controlled trial of in-vitro fertilisation (IVF) and 272(8): 619–26. 356 TEXTBOOK OF CLINICAL TRIALS
33. Fayers PM, Sprangers MAG. Understanding self- 50. McIver S, Meredith P. There for the asking: can rated health. Lancet (2002) 359: 187–8. the government’s planned annual survey really 34. Ware JE, Sherbourne CD. The MOS 36-item short measure patient satisfaction? Health Serv J (1998) form health survey (SF-36). Conceptual frame- 19: 26–7. work and item selection. Med Care (1992) 30: 51. Leimkuhler A, Muller U. Patient satisfaction– 473–83. artifact or social fact. Nervenarzt (1996) 67: 35. McGee H, Hannah M, O’Boyle CA, Hickey A, 765–73. O’Malley K, Joyce CRB. Assessing the quality of 52. Owens DJ, Batchelor C. Patient satisfaction and life of the individuals: the SEIQOL with a health the elderly. Social Sci Med (1996) 42: 1483–91. and gastroenterology unit population. Pschol Med 53. Drummond MF, Davies L. Economic analysis (1991) 21: 749–59. alongside clinical trials. Revisiting the method- Int J Technol Assess Health Care 36. Guyatt GH, Bombardier C, Tugwell PX. Measur- ological issues. (1991) 7: 561–73. ing disease specific quality of life in clinical trials. 54. Barber JA, Thompson SG. Analysis and interpre- Can Med Assoc J (1986) 134: 889–95. tation of cost data in randomised controlled tri- 37. Kelleher CJ, Cardozo LD, Khullar V, Salvatore S. als: review of published studies. BMJ (1998) 317: A new questionnaire to assess the quality of life of 1195–200. urinary incontinent women. Br J Obstet Gynaecol 55. Fayers PM, Hand DJ. Generalisation from phase (1997) 104(12): 1374–9. III clinical trials: survival, quality of life and health 38. Moos RH. Typology of menstrual cycle symp- economics. Lancet (1997) 350: 1025–7. toms. Am J Obstet Gynecol (1969) 103: 390–402. 56. Briggs A. Economic evaluation and clinical trials: 39. Jones G, Kennedy S, Barnard A, Wong J, Jenkin- size matters. BMJ (2000) 321: 1362–3. son C. Development of an endometriosis quality- 57. Bowling A. Measuring Disease: A Review of of-life instrument: The Endometriosis Health Disease Specific Quality of Life Measurement Profile-30. Obstet Gynecol (2001) 98(2): 258–64. Scales. Mitton Keynes: Open University Press 40. Schneider HP, Heinemann LA, Rosemeier HP, (1994). Potthoff P, Behre HM. The Menopause Rating 58. Peto R, Collins R, Gray R. Large scale randomised Scale (MRS): comparison with Kupperman index evidence: large sample trials and overviews of and quality-of-life scale SF-36. Climacteric (2000) trials. J Clin Epidemiol (1995) 48: 23–40. 3(1): 50–8. 59. Schulz KF and Grimes DA. Generation of alloca- 41. Sanders C, Egger M, Donovan J, Tallon D, Frankel tion sequences in randomised trials: chance not S. Reporting on quality of life in randomised choice. Lancet (2002) 359: 515–19. controlled trials: bibliographic study. BMJ (1998) 60. Schulz KF, Chalmers I, Hayes RJ, Altman DG. 317: 1191–4. Empirical evidence of bias: dimensions of method- 42. Graham WJ. Outcomes and effectiveness in repro- ological quality associated with estimates of treat- ment effects in controlled trials. JAMA (1995) 273: ductive health. Social Sci Med (1998) 47(12): 408–12. 1925–36. 61. Moher D, Pham B, Jones A, et al. Does quality 43. Williams B. Patient satisfaction: a valid concept? of reports of randomised trials affect estimates Social Sci Med (1994) 38: 509–16. of intervention efficacy reported in meta-analysis. 44. Sitzia J, Wood N. Patient satisfaction: a review of Lancet (1998) 352: 609–13. issues and concepts. Social Sci Med (1997) 45(12): 62. Juni P, Altman D, Egger M. Assessing quality of 1829–43. controlled trials. BMJ (2001) 323: 42–6. 45. Bury M. Doctors, patients and interactions in 63. Pocock SJ. Clinical Trials: A Practical Approach. health care. In: Health and Illness in a Changing Chichester: John Wiley & Sons (1983). Society. London: Routledge (1997). 64. Schulz KF, Grimes DA. Sample size slippages in 46. Department of Health. The New NHS: Modern and randomised trials: exclusions and the lost and Dependable. London: HMSO (1998). wayward. Lancet (2002) 359: 781–5. 47. Williams B, Coyle J, Healy D. The meaning 65. Edwards SJ, Lilford RJ, Hewison J. The ethics of of patient satisfaction: an explanation of high randomised controlled trials from the perspectives reported levels. Social Sci Med (1998) 47(9): of patients, the public and health care profession- 1351–9. als. BMJ (1998) 317: 1209–12. 48. Hopton JL, Howie JGR, Porter MD. The need to 66. Bhattacharya S, Parkin DE, Reid TMS, Abramo- look at the patient in General Practice satisfaction vich DR, Mollison J, Kitchener HC. A prospective surveys. J Fam Prac (1993) 10: 82–7. randomised study of the effects of prophylactic 49. Williams SJ, Calnan M. Key determinants of con- antibiotics on the incidence of bacteraemia follow- sumer satisfaction in general practice. J Fam Prac ing hysteroscopic surgery. Eur J Obstet Gynaecol (1991) 8: 237–42. Reprod Biol (1995) 63: 37–40. RESPIRATORY
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 22 Respiratory
ANDERS KALL¨ EN´ AstraZeneca, Lund, Sweden
AIRWAYS DISEASES MEDICAL BACKGROUND Asthma Airway obstruction is a common and impor- tant feature of some respiratory diseases. It can From a clinical point of view, asthma presents be acute, ‘semi-chronic’ (e.g. due to cancer) itself by recurrent breathlessness, cough or or chronic. The chronic obstructive airway dis- wheeze caused by variable or intermittent nar- eases can be divided into whether the obstruc- rowing of the intra-pulmonary airways. The tion is reversible or not. In the former case the severity of these symptoms has a wide range, patient usually has asthma, in the latter case from very mild intermittent with symptoms only chronic obstructive pulmonary disease, abbrevi- upon provocation, to severe persistent with large ated COPD. These disease concepts lack precise impact on daily life. There is no precise definition definitions, and the division is only meant as of asthma, and therefore the prevalence is hard to a first approximation. Both diseases are inflam- establish. We know, however, that it is commoner matory diseases mainly of the lower respiratory in children than in adults, and more common in tract: in asthma there is an inflammatory pro- boys than in girls.1 A figure for children around cess mainly in the central airways, whereas the 10% and half that for adults is probably close inflammation of COPD is predominantly periph- to reality in most of the western world. There is eral with progressive destruction of lung tis- however a definite regional inhomogeneity with sue. Inflammation in the upper respiratory tract, regions with much higher prevalence and regions i.e. rhinitis, is characterised by both acute and where the disease is rare. Most epidemiological chronic conditions, the most distinctive being studies seem, however, to agree that the preva- seasonal hayfever. lence is rising.2 This chapter will discuss clinical trials in the The high prevalence of asthma gives a poor three diseases asthma, COPD and rhinitis, with prediction of the impact of the disease on the the focus on the first of these. community, because the overwhelming majority
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 360 TEXTBOOK OF CLINICAL TRIALS of asthmatics are mild sufferers with symptoms year, with long periods of absolute or relative confined to wheezing after exercise or breath- relief between attacks of varying severity. lessness in association with an upper respiratory In general, asthma carries a favourable prog- tract infection. At the other end of the spectrum, nosis because the bronchial inflammation does asthma is a crippling, life-threatening disease not usually cause permanent tissue damage. How- with acute severe attacks requiring emergency ever, in a subgroup of subjects, irreversible room treatment. In the western world, about 1% bronchial obstruction develops later in life.6 of adults and 2% of children need medical atten- tion for asthmatic symptoms.3,4 COPD Many factors are known to cause narrowing of intra-pulmonary airways. The sensitivity to Chronic obstructive pulmonary disease is char- such stimuli varies between individuals but under acterised by long-term, in general progressive, normal circumstances the concentration of such irreversible obstruction of the flow of air out of substances is too low to produce symptoms in the lungs. To a large extent it is comprised of two healthy subjects. Asthmatic patients are more related disease: or less characterised by a high sensitivity to 1. Chronic bronchitis, whose clinical definition such stimuli, a phenomenon called non-specific is productive cough (from bronchial secre- bronchial hyperresponsiveness. The most com- tion) on most days for 3 months/year for mon cause of non-specific bronchial irritation is two consecutive years. The mucus hyper- exercise and, for many, this may be the only man- secretion comes from hypertrophied bronchial ifestation of their asthma. glands and increases the risk of bacterial It is essential to make a clear distinction lung infections. between this non-specific hyperresponsiveness 2. Emphysema, which has a pathological defini- and the allergic reactions. Allergy is an immuno- tion with enlargement of the alveoli due to the logical reaction to a specific environmental destruction of the walls between them. These agent. Hyperresponsive bronchi, in addition to walls contain elastic fibres, so their destruc- responding in an exaggerated fashion to exoge- tion reduces the elasticity of the lung, leading nous stimuli, will also respond in an enhanced to collapse, and thus obstruction, of airways. fashion to inflammatory mediators released in the bronchial wall as a result of an allergic The disease entities asthma, chronic bronchi- reaction. Thus a trivial allergic reaction in a tis and emphysema are in no way mutually hyperresponsive bronchus may provoke a large exclusive: a given patient can have symptoms bronchoconstrictive response. There is little, if from more than one. The definitions of the last any, relationship between the degree of atopy two does not imply that the patient has airway and non-specific hyperresponsiveness. Instead the obstruction, so not everyone with these diseases degree of non-specific hyperresponsiveness is has COPD. associated with the degree of inflammation in the COPD is believed to affect more than half respiratory tract.5 a billion people worldwide, causing perhaps Asthma may start at any age. When starting 3 million deaths annually. When diagnosed, this during childhood and adolescence it is likely is often in a relatively late stage of the disease, to be associated with atopy, as compared to with less than 50% of lung function remaining, so when symptoms start later in life. Most asthmatic the majority of cases are at any specified point in patients have perennial symptoms, but a minority time undiagnosed. The prevalence of diagnosed shows a seasonal variation, sometimes confined COPD is about 5%, and is increasing. to periods with air-borne pollen, sometimes to Pathologically COPD is a disease with per- the winter months. Thus different asthmatics may iferal inflammation (thus rather a bronchioli- have symptoms during different periods of the tis than bronchitis) with progressive lung tissue RESPIRATORY 361 destruction. In the western world, by far the sneezing, which is the cleaning reflex of the upper most important factor responsible for COPD is airways corresponding in a way to coughing, smoking; it has been said to be responsible for which is the cleaning reflex of the lower airways. up to 90% of COPD patients.7 However, only Inflammation in the upper respiratory tract, about 15–20% of all cigarette smokers develop rhinitis, presents as one or more of the symptoms COPD. The mechanism seems to be that cigarette nasal congestion, rhinorrhea (i.e. runny nose), smoke attracts cells (neutrophils, macrophages sneezing and itching. Chronic inflammatory con- and cytotoxic T-cells as opposed to eosinophils ditions can in predisposed individuals result in and T-helper cells in asthma) to the lungs that benign protrusions of nasal polyps into the nasal promote inflammation, and these are stimulated cavity, polyposis. to release elastase, an enzyme that breaks down Rhinitis can be allergic or non-allergic, where the elastic fibres in lung tissue. Normally the the latter is characterised by presence of symp- lungs are protected against this enzyme by the toms of varying severity. Allergic rhinitis can be elastase inhibitors, among them α1-antitrypsin, seasonal as hay fever (SAR = Seasonal Aller- which is produced in the liver (congenital defi- gic Rhinitis), or perennial. In the latter case the ciency of this enzyme is another, but rare, cau- symptoms can be due to continuous exposure sation for emphysema). Air pollution has been to allergens like the house dust mite, or may suspected to have similar effects as smoking, but present themselves intermittently as episodes trig- it is unclear to what extent that is an important gered by allergens like, e.g., grass pollen. Despite aetiological factor for COPD. Also, there is a high the common inflammatory denominator for aller- COPD incidence in women in Asia attributed to gic rhinitis and polyposis, there is no evidence cooking fumes.8 that the two conditions are closely linked, or that The typical COPD patient has been smok- allergy plays a major role in the aetiology of ing 20 or more cigarettes a day for more than nasal polyposis. 20 years and presents with a chronic cough, Rhinitis is a very common disease, but sur- shortness of breath (dyspnea) and frequent res- prisingly little is known about its epidemiology. piratory infections. If the underlying disease is The nose has a filter function, and is therefore mainly emphysema, shortness of breath may be exposed to a much larger amount of inhaled aller- the only symptom. Initially the dyspnea only gens per square centimetre than bronchi, espe- comes during physical exercise, but as the disease cially when the allergens are large. SAR is due progresses it occurs already on minimal exertion. to air-borne plant pollen. From a clinical point of For the patient with chronic bronchitis dominat- view the most widely distributed ones are those ing, the major symptoms are chronic cough and of grasses, but some tree pollen, including birch sputum production. The sputum may be clear but and olive tree, are also important, as is ragweed. is usually coloured and thick as bacterial coloni- It is important to note that the pollen season for sation is common. an individual plant species varies from one coun- try or region to another. Also, whereas the season Rhinitis and Nasal Polyposis for air-borne pollen is limited to perhaps half a year in temperate zones, in warmer climates it The upper respiratory tract is to some extent is so long that what seems to be perennial symp- like terminal bronchioli without smooth muscles. toms may be provoked by multiple and sequential Instead the nose has venous sinusoids and the seasonal allergies. major reason for obstruction of the upper airway Patients with nasal polyposis suffer from a tract is vasodilation of capacitance vessels and series of symptoms, just as in rhinitis but in oedema while secretion can contribute. Another particular nasal blockage and an absence of smell. difference to the lower tract is that stimulation The prevalence is not known, there are few of nervous irritant receptors in the nose results in epidemiological studies, but it is probably in the 362 TEXTBOOK OF CLINICAL TRIALS range of 1–4%. The diagnosis of polyps requires cardiac stimulation and intestinal inhibition, appropriate inspection of the nasal cavities by a whereas stimulation of the β2-receptor results in trained physician. bronchodilatation, vasodilatation, stimulation of skeletal muscles and uterine contractile inhibi- 11 CURRENT TREATMENTS tion. A third type, β3-receptors, cause lipolysis. The development within this drug class has The respiratory tract has a limited repertoire of been towards more and more potent and selective responses to irritation or other stimulation. In the β2-agonists. The first generation of drugs were nose vasodilation leads to decreased airway cali- short-acting with a duration of action of, at most, bre and nasal blockage. The bronchi may change 4–6 hours. Lately a few long-acting drugs, with their calibre or alter the amount of glandular duration of action superseding 12 hours, have secretion produced, leading to obstruction. There been introduced.12 is oedema, hyperaemia and cellular infiltration of β2-agonists are of benefit to the majority of the wall of the tract. Afferent nerves may signal asthmatic patients because of the bronchodilator information to the brain stem to produce sneezing property; rapid-acting ones are often given as (upper tract) or cough or the sensation of breath- rescue medication for relief of symptoms. The lessness (lower tract). The relative importance of drug class does however have actions other than these factors varies between individuals, and dif- smooth muscle relaxation that may contribute ferent drugs interfere with different factors. to their long-term therapeutic effect in asthma Drugs for chronic obstructive respiratory dis- and motivate their use in COPD: they stimulate eases are given either systemically, as tablets, mucociliary function in the airways, restore or by local administration using an inhalation normal clearance of bronchial secretion and device. When it comes to inhaled products, it inhibit microvascular permeability in the airways is important to note that a treatment consists of leading to decreased mucosal oedema. two objects, a drug to be delivered to the body Side-effects are a consequence of the binding and an inhalation device used for this deliver- to receptors in tissues and organs outside the ance. We will not discuss devices here, only drug lung: tremour by binding to receptors in skeletal classes. It is however important to understand that muscle, tachycardia by binding to receptors in the the amount of drug delivered to the airways may heart (this problem has been reduced as the drugs 9 vary considerably from one inhalator to another. became highly selective, but there are some β2- The same might be true of the distribution pattern receptors in the heart as well) and hypokalemia, within the lungs, with potential consequences for due to a redistribution from extra- to intracellular the effectiveness of the treatment. spaces. In general tolerance develops rapidly to the extra-pulmonary effects, so these are usually Bronchodilator Drugs mild or absent in patients, though individual variation in the sensitivity can make the use of There are three basic groups of bronchodilator these drugs impossible in the occasional patient. drugs–β2-agonists (today by far the largest), xanthines and anticholinergics.10 Their modes of Xanthines , the most well-known member of action differ somewhat. We discuss each class of which is caffeine but the most widely used drugs separately. one as treatment for airway obstruction is theo- phylline, are potent smooth muscle relaxants β2-Agonists bind to the β-adrenergic receptor by acting directly in the intracellular messenger and stimulate the intracellular accumulation of cAMP. Thus they have about the same phar- the signal substance cAMP (cyclic adenosine macological actions as β2-agonists. But since monophosphate). There are now three known they act intracellularly and not by binding to a types of β-adrenergic receptors in the human receptor on the cell surface, the effect is more body: stimulation of the β1-receptor causes generalised and the side-effects are somewhat RESPIRATORY 363 different and potentially more serious than those own remedy for inflammation: if we remove of β2-agonists. The most important ones relate the adrenal glands inflammatory reactions are to the gastrointestinal, cardiovascular and central greatly exacerbated. Regulation of endogenous nervous systems. At the start of treatment with cortisol is complex, involving the hypothala- oral theophylline, most patients will experience mic–pituitary–adrenal (HPA) axis. During a some caffeine-like symptoms including irritabil- severe inflammatory response, elevated levels of ity and nausea, symptoms which usually fade cytokines stimulate centres in the brain, leading away after a few days. For that reason, however, to an increase of cortisol in the circulation thereby treatment is usually initiated in subtherapeutic attenuating the inflammatory response. It is now doses and progressively increased over a period believed that even at normal levels, endogenous 13 of 1–2 weeks. hormones will regulate inflammation. The GCS The serious side-effects, in contrast to the mode of action is by binding to a glucocorticoid caffeine-like ones, are well correlated to plasma receptor within the cell’s cytosol. When used for concentrations. In clinical practice theophylline treating, e.g., asthma GCSs lead to a reduction of concentration in plasma has to be monitored and airway inflammation, mucous hypersecretion and dose adjusted so that the plasma concentration airway reactivity while restoring the integrity of lies within a therapeutic window. Because of the airways.15 this the use of xanthines has diminished over Originally GCSs were given systemically as the last 10 years as alternative treatment has become available. asthma treatment. There are however well known side-effects of high does of oral GCSs over a Anticholinergic drugs have been used since long time that limits that usage. These include, ancient times for the treatment of asthma. The but are not limited to, osteoporosis, hypertension, use of various plant derivatives has evolved adrenal insufficiency and Cushingoid features as through synthetic atropine to more selective well as growth retardation in children. Concern bronchodilating anticholinergic agents with fewer about these side-effects diminished the use of oral 14 side-effects than atropine. GCS as an asthma treatment. Inhaled GCSs have The bronchodilating effect of this drug class improved the benefit/risk ratio. Since adminis- is due to their antagonism of the binding of tration is aimed directly at the site of inflam- acetylcholine (from the vagal nerve) to the mation, lower doses can be used, giving lower muscarinic receptors of bronchial smooth muscle. GCS concentrations in plasma with largely neg- These drugs are particularly used in treating ligible systemic side-effects as a result. Inhaled reversible airway obstruction in COPD. GCSs are now widely accepted as first-line anti- The side-effects of anticholinergic agents are inflammatory therapy for asthma.16 due to blockade of muscarinic M -receptors in 2 To evaluate the long-term side-effects of other organs and include dryness of the mouth, inhaled GCSs is difficult. They are rare at doses blurred vision, urine retention and difficulty in given in asthma treatment, so large numbers of micturition, tachycardia, flushing and lighthead- edness. patients and long-term clinical studies are needed. Some information can be gained by studying the endogenous cortisol levels. As already men- Corticosteroids tioned, the endogenous cortisol level is controlled That glucocorticosteroids (GCSs) have a thera- by the highly complex HPA axis. Introduction of peutic effect on asthma, rhinitis and other anti- exogenous GCS in the plasma will affect this axis inflammatory diseases has been known for a and lead to a suppression of the endogenous cor- long time and is due to their being manmade tisol levels, the degree of which is determined analogues of an endogenous anti-inflammatory by the plasma concentrations and the potency of steroid–cortisol. Cortisol is in a way nature’s the drug. Thus, the degree of suppression is a 364 TEXTBOOK OF CLINICAL TRIALS measure of the amount of active (on the HPA production, airway wall oedema and causing axis) exogenous GCS in the body. bronchospasm. Oral leukotriene receptor antago- nists, to be administered once or twice daily, are Other Drugs available along with an oral leukotriene synthe- sis inhibitor, which has to be administered four Vasoconstrictors are used extensively in rhinitis. times daily. Topical α-agonistic sympatomimetics effectively Leukotriene modifiers improve airway func- and promptly alleviate the nasal blockage. They tion and decrease the need for additional mainte- have no effect on rhinorrhea, nasal itch or nance and rescue asthma therapies. Leukotriene sneezing.17 modifiers also attenuate bronchospasm induced by allergens, exercise, cold air, salicylates and Antihistamines are used for rhinitis, mainly as exercise. In patients with chronic, persistent rescue medication. Their main effect is to block asthma, results from clinical studies indicate peripheral H1-receptors which limits vasodilata- that inhaled corticosteroids have a more con- tion in the nasal mucosa. They have an effect on sistent and greater average effect than antileu- nasal itching, sneezing and discharge, but little or cotriene drugs.20 no effect on nasal congestion and blockage.18 The long-term safety of leukotriene modi- fiers is still not clear. Some patients reducing Disodium cromoglycate (DSCG) and nedo- their oral corticosteroids when treatment with cromil sodium DSCG has been used as a pro- antileukotriene drugs has been initiated have phylactic anti-asthma drug, mainly by children developed a special type of vasculitis called and young adults.19 To be effective it should be Churg–Strauss syndrome. However, it might be administered four times per day. Originally its the unmasking of a pre-existing condition and not mechanism of action was proposed to be sta- induced by the leukotriene modifier per se. bilisation of mast cells, though that is probably not the case. Taken immediately before exposure, MEASUREMENT SCALES DSCG affects the asthmatic reactions induced by various stimuli. However, after discontinuation of When measuring the status of the chronic long-term treatment, DSCG seems not to have obstructive airways disease in a subject we can modulated the bronchial hyperresponsiveness or rely either on data obtained at a visit to the clinic, the underlying inflammatory reaction. or we can monitor the patient over a longer period Nedocromil sodium is another non-steroidal by daily recordings at home. substance with anti-inflammatory properties in At a visit to the clinic the primary focus is vitro. It acts as an inhibitor at several levels usually to obtain an objective, indirect, measure of neurogenic inflammation in asthma. Clinical of airways narrowing – a lung function test. Such studies have demonstrated improvements in air- a test measures some functional index of the way functions, including a reduction of bronchial airway calibre in some kind of experimental hyperreactivity, but it does not protect against setting. We will discuss various such indices and maximal airway narrowing, which is an impor- experimental settings and what they measure. tant feature of inhaled corticosteroids. In the latter case, long-term daily recordings, Both DSCG and nedocromil are remarkably we provide the patient with a diary card and usu- free from side-effects. They are also used for ally ask him/her to record twice daily information rhinitis, with effects similar to antihistamines. relating to symptoms of the disease under study. In addition patients are often given a device to Leukotriene modifiers The cysteinyl leukotrienes obtain an objective measurement of lung function are products of the arachidonic acid metabolism at home, traditionally a peak flow meter. with effects that mimic many features of asthma, These two approaches are in no way mutually e.g. by increasing eosinophil migration, mucus exclusive – in a long-term study we can make RESPIRATORY 365 experimental manoeuvres of the first kind. As an possible. The spirometer records the exhaled example there is virtually no long-term asthma volume as a function of time, V (t). From this trial that does not measure FEV1 on visits to the curve (Figure 22.1) a number of spirometric clinic. However, for the present discussion we indices can be obtained. The most widely used consider experimental approaches and diary card measure is the forced expiratory volume in approaches separately, except that single FEV1 one second (V (1), denoted FEV1), followed measurements at the clinic will be discussed by the forced vital capacity (V(∞), denoted along with diary cards. FVC). If the expiratory effort has been markedly inadequate it is usually obvious from the trace. Lung Function Measurements By calculating numerically the derivative of V (t) we obtain the expiratory airflow. Its maxi- Airway narrowing leads to an increased resis- mum value, which usually occurs within 100 ms, tance to the airflow. The airway resistance can is the peak expiratory flow (PEF). Often the be measured directly with body plethysmogra- spirometric result is shown by plotting the flow phy (in a ‘body-box’), an expensive and rather against the volume exhaled. From this curve complicated procedure. Another way of measur- we can identify both PEF and FVC (but not ing lung function is by flow measurements, which FEV1), but can also define new measurements, uses much more inexpensive apparatus, a spirom- like FEF25%, which is the flow when 25% of eter. However spirometric measurements, to be the FVC has been exhaled. Another measure discussed below, depend not only on airways of current interest is FEF25–75%, which is the resistance, but also on lung volumes. amount of volume expired per second when the When doing a spirometric manoeuvre the exhaled volume increased from 25% to 75% of patient takes a maximum deep breath and then FVC. It is considered to measure effects in the exhales as rapidly as possible as much as small airways.
14 5.0 FVC 4.5 PEF 12 4.0 3.5 FEV1 3.0 10 FEF 25% 2.5 2.0 8 1.5 Volume Exhaled (L) Volume 1.0 0.5
Flow (L/s) Flow 6 0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 Time (s) 4
2 FEF 75% FVC 0 0123456 Volume Exhaled (L)
Figure 22.1. Illustration of some spirometric measurements 366 TEXTBOOK OF CLINICAL TRIALS
A full spirometric manoeuvre consists of various predicted formulas have been obtained measurement of the inspiratory part also. The for different lung function parameters. Thus, inspiratory vital capacity (IVC) is a measure of e.g., a key disease severity parameter is the the functional residual capacity (FRC) and is an FEV1 in percent of predicted normal, both for important measure in COPD patients. asthma and COPD. There are a number of If performed correctly, the spirometric test is such formulae available, generally depending on highly reproducible but somewhat effort-depen- demographic variables like race, gender, age dent. Different parameters are effort-dependent and height.22 It should be emphasised, however, to different degrees: e.g. FEV1 is less dependent that these measures cannot be anything but than FVC, since it only needs maximum effort for rather approximative ones, since the predicted 1 s. The direct measurement of airway resistance normal values are not exact counterparts to (Raw), which is done in the body box, is effort- the unknown lung function without disease! If the independent, but has a poor reproducibility. Since lung function is between 80% and 120% of the the spirometry has a good reproducibility, and predicted normal value, it is in general considered uses fairly simple and portable equipment, it to be ‘normal’. is most useful for clinical purposes. In special Another disease characteristic obtained from situations, however, the assessment of resistance lung function measurements is the reversibility. might be preferable. This is an index obtained from a very simple PEF is much more effort-dependent than FEV1, single-dose monitoring experiment: we measure but it can also be measured by a much cheaper FEV1, give a rapid-acting β2-agonist and wait apparatus than a spirometer. Such a peak flow 30 min (typically) to measure FEV1 again. The meter is often provided to the patients for classical reversibility is then obtained as self-monitoration at home. Instructions are then reversibility given to fill in a diary card and to contact the − healthcare service when PEF has dropped for a FEV1(after) FEV1(before) = 100 × few consecutive days below prespecified levels. FEV1(before) In the same way, PEF can be monitored with this simple device in a long-term study by recording, A value in excess of 15% was previously consid- ered indicative of reversible airways obstruction, often twice daily, in a diary card. 21 There is a diurnal variation in FEV and though later guidelines use 12%. The basis for 1 these numbers is somewhat unclear – it is prob- other lung function measurements. It is therefore ably chosen in order to be ‘certain’ that there is important when comparing different such mea- an effect: the variability in FEV is such that a surements obtained at different visits to the clinic 1 numerical increase per se is not a definite proof for the same patient, that these measurements are of an improved lung function. taken at approximately the same time of the day. Lung function measurements can be followed Upper Airway Function Tests in order to assess effects, but also to characterise disease severity. However lung function is a There are also upper airway function tests simi- function of both gender, age and ‘size of patient’. lar to the lung function indices discussed above. Therefore a lung function parameter cannot be They are however much less used, since symptom judged on an absolute scale – an FEV1 of 2.4 L scores are considered of overriding importance in means different things for a young, tall boy rhinitis studies. Resistance can be estimated by and an old, tiny woman. A measure of disease two different techniques: posterior rhinomanome- severity would be the ratio of the actual FEV1 try in which values are obtained by probes placed and the would-be, and unmeasurable, FEV1 in the mouth, and anterior rhinomanometry in the patient should have without the obstructive which a device in the nose is used. Less compli- airway disease. As a remedy for the latter cated, and expensive methods for assessing nasal RESPIRATORY 367 patency rely on the measurement of peak nasal is declared at a timepoint t if there is a 15% flow either on inspiration (PNIF) or expiration increase compared to baseline at that time. Based (PNEF). We do not discuss these methods in any on such a concept, we can define: the time of further detail. onset as the timepoint (if any) at which the polygonal curve cuts the line E = 1.15 × E0 for DESIGNS FOR EXPERIMENTAL ASTHMA the first time (for rapid-acting bronchodilators TRIALS one usually added the restriction that this should occur within 30 min). The ending of effect occurs For asthma studies, there are a number of at the timepoint on the polygonal approximation experimental designs to measure various aspects which is followed by at least two observations of the therapeutic effect based on objective lung below the line E = 1.15 × E0, provided that two function measures. For this section, let E denote measurements were taken. If only one was taken, an index of lung function. In most real-life cases that will suffice and if no measurement was found this is FEV1, but the discussion is not restricted below the line, censor the end of effect to the last to this case. measured time. The duration is then the time from We can group the designs in two groups: either onset of action to end of effect. the response after administration(s) of a study The main problem with these definitions of drug is followed, which can be done by time or onset and duration of action is not the arbitrary by increasing doses, or the protective effect of number 15%, but the fact that effect is measured the study drug to some provocation is assessed. by relating to baseline. This is not appropriate, since lung function has a clear diurnal variation. Single Dose Monitoring It might be a reasonable approximation for a few hours, the perceived time of clinical efficacy of This type of experiment is simple. Consider one a short-acting β2-agonist, but will produce an individual on one occasion when this experiment incorrect result if used for a longer period. In is performed. We first take a baseline measure- fact, there are studies in which a patient receiving ment, E0, give the study drug and then follow placebo as treatment has had a definite increase lung function at predetermined timepoints after in lung function already on the first measurement study drug administration. This provides us with after treatment administration (changes in the an approximation of a response curve E(t), where means – not individual spurious events), so the we use E(0) = E0 (though technically it was use of baseline as a reference when declaring obtained at a timepoint t<0). From this curve effects should very much be questioned. a number of measures can be obtained for fur- A related problem is to define responders. ther analysis. The two most important measures As the name suggests, a responder is a subject derived from the curve E(t)are who responds to the treatment. Traditionally this has been decided based on the maximal 1. The average level, defined as area under the increase from baseline. The discussion above curve (of the polygonal approximation we implies that this is not necessarily a good way have observed to the response curve) divided to go. To actually measure effect, we need to by observational time. This we denote by Eav. relate the measurements to the measurements 2. The maximal level, Emax. obtained without drug administration. However, since asthma is not a stable disease, these must be We can also compute tmax, the time at which taken simultaneously. And this is impossible! In Emax occurred. This is sometimes useful. Other clinical trials we do not really need this concept potential measures are related to the concept at all, except for descriptive purposes. We will responders ‘onset of action’ and ‘duration of return to the question of duration in the discussion effect’. Tradition has it that for FEV1, effect of an example. 368 TEXTBOOK OF CLINICAL TRIALS
One lesson, however, from the discussion common lung function measurement here is again is that effect for many of these variables is FEV1. It should be noted that a better definition of often clinically measured as percent change. This the index would be IndexEIB = 100 × Emin/E0, means that since then the analysis could be done on the mul- tiplicative scale as discussed above! effect = E/E ≈ ln(E) For allergen challenge test we are more inter- ested in the whole response for 10–12 hours, which by integration motivates why many lung in order to study both the immediate reac- function indices should be analysed on a loga- tion and the late reaction. The immediate rithmic scale. We analyse these types of trials reaction (EAR = early asthmatic reaction) is an with multiplicative models, which is justified by episode of acute bronchoconstriction which peaks this observation. between 10 and 20 min after inhalation and resolves within 1.5–2 hours. The late reac- tion (LAR = late asthmatic reaction) is probably Challenge Tests an inflammation mediated bronchoconstriction A challenge test is similar to the single dose mon- which starts about 3 hours after allergen inhala- itoring test, except that most of the monitoring tion and does not resolve for many hours. Aller- takes place after a provocation of some kind. gen challenge tests are potentially dangerous, and Two important cases of challenge are exercise, are therefore not much favoured as a mode of either a treadmill test or using a cycle ergometer, studying asthma. and an allergen to which the patient is allergic. If they are, we need to measure FEV1 repeat- edly during the first hour, and then more sparsely A baseline measurement E0 is taken, often after administration of study drug. Then the provo- during the next 7–8 hours (perhaps once an cation is done and lung function followed. In hour). The EAR is most often defined as the most cases there are two phases in the reac- maximum percent reduction in FEV1 (from base- tion found. First there is an immediate reaction line) occurring in the first hour after challenge, with bronchoconstriction within minutes which whereas the LAR is defined as the maximum lasts 1–2 hours. Several hours later there is a percent reduction in FEV1 (again from baseline) delayed reaction with a much slower and sus- occurring between 3 and 7 hours after challenge. tained time course. Alternatively we compute the area under the Typically an exercise test is followed only dur- curve for the first hour and for the period between ing the immediate reaction, the actual existence 3 and 7 hours after challenge and use that as an of a delayed reaction is controversial. The protec- efficacy measure in much the same way as for tive effect of the study drug can be measured by the single dose monitoring experiment. maximal decrease in lung function from baseline: Hyperresponsiveness Studies IndexEIB = 100 × (E0 − Emin)/E0 The level of airway responsiveness to a non- and we only need to follow the patient until specific stimulus is measured by exposing sub- we know he has attained the low turning point, jects to the stimulus and measuring the response. whereafter he is given a high dose of a bron- There are a number of dialects of this test, chodilator in order to restore lung function. by varying the selected stimulus, the mode of Because of the intrinsic variability in lung func- administration of it and the method of assessing tion measurements spurious local minima can the response. occur in the measurement series – it is important The most commonly used stimuli are metha- that the investigator has certified that the global choline and histamine, though small doses of minima has occurred before stopping. The most allergens can also be used. Methacholine and RESPIRATORY 369 histamine produce similar responses, but the The actual algorithm for estimation of PDx can latter has more side-effects and can only be vary. The following suggestion is justified by this administered safely in concentrations up to description of the dose–response curve. 32 mg/ml, whereas methacholine can be used safely in concentrations up to 256 mg/ml (these 1. If there is a dose with less than x% decrease numbers should be compared to the clinical followed by a dose with more than x%, definition of hyperresponsiveness which is that loglinear interpolation (of log D vs. response) the provocation dose (PD20, see below) is ≤ is done. 8 mg/ml). The stimuli is administered from an 2. If the first dose provoked a fall in excess of aerosol which can be done in different ways. x%, we cannot do loglinear interpolation. In Suffice it to note that one can either do it that case we do a linear interpolation back to with or without a dosimeter which controls the baseline and obtain a dose corresponding to a dose. Response is generally measured either as fall of x% from this. However, we never go FEV1 or as airway resistance (or its inverse, back more than to half the first dose given. conductance). 3. If the last dose produced a fall of less than Technically the subject first inhales saline x%, we extrapolate loglinearly, but only up to and then inhales progressively increasing, often twice the highest dose given. doubled, doses of the stimuli from the aerosol at 3-min intervals. There is a measurement after What to use as dose can also be discussed. each dose administration, so we can consider If we do not control the dose, we must use the response to be a function of the last the concentration given. If we control the dose, concentration or dose given. In both cases the we can choose to use cumulative doses or last saline inhalation produces the baseline value. dose without much difference in the final results, From this dose–response curve (I call it that, when provocation doses are compared (because though sometimes it is a concentration–response for a geometric series, the sum is essentially curve) different characteristics can be computed. proportional to the next dose, and we compare A general dose–response curve is sigmoidal ratios). In general the use of the cumulative dose in shape which is well approximated with a seems to be favoured. loglinear portion over most of its response ThemeasurePD20 is not limited to the possible range. We can, however, not clinically obtain interpretation discussed above (as the relative information on much more than the lower part dose potency ρ of the stimuli). If the effect is of this dose–response curve, which means that due to changes in both position and shape, the traditional measures for dose–response curves measure can still be used. For epidemiological (ED50 and slope) are not usually estimatable. We purposes another index has been introduced, the can think of the effect of the drugs as a parallel two-point slope, which is the percent decline shift of the response curve so that if a given from baseline to last dose, divided by last dose. response is obtained with dose D without the Though this index has a clear interpretation (as drug, it takes dose ρD (with, hopefully, ρ>1) percent decrease per unit dose), this interpretation to get that response with the drug. To estimate is wrong since the decrease is not linear with ρ in this type of study we fix a level, expressed dose – instead it is virtually zero until it becomes as percent decrease in the response, and estimate linear with log-dose. the dose of stimuli needed to obtain that level. Note: It has been suggested that you cannot The dose which gives a decrease of x%inthe estimate PDx if there has not been a fall of x%. response variable is denoted PDx (or PCx if we do Technically this means that we should note it as not control doses). For FEV1 we usually compute missing. This might be sensible for the caring of PD20, whereas for Raw a higher percentage can the patient, where this is perceived as no hyper- be used. responsiveness. However, for a clinical study, 370 TEXTBOOK OF CLINICAL TRIALS where treatments are compared, it is imperative these experiments is an endurance time, though to do an estimation. Setting it to missing means for some cycle-ergometry tests you could alter- that the analysis loses the information that a high natively use the total workload (but these should dose is needed to achieve the specified decrease! be heavily correlated). For a comparison of some exercise tests in EXERCISE TESTS IN COPD COPD, see Ref. 23. In conjunction with these tests measurements Since a progressive decline in physical fitness is of breathlessness are usually done. There are the main characteristic of COPD, exercise tests different tests available. A much used dyspnoea are useful for a proper evaluation of treatment score is the Modified Borg scale,24 in which effects in these patients. In these tests exercise dyspnoea is scored on a 0–10 scale before and can be either walking, running (treadmill tests) after the exercise test. Alternatively one can use or bicycling. The basic design of the test can be, a visual analogue scale with the same effect. in its purest form An alternative scale is the Transitional Dyspnoea Index,25 for which we first rate three factors 1. to determine the maximal workload sustain- (functional impairment, magnitude of task and able, or magnitude of effort) on baseline, each on scales 2. to fix the workload at some level, and 0–4 (well, 4–0 actually – the scale is reversed), determine the endurance time. and then rate the changes over the exercise directly on a scale from −3to+3. An example of the first kind is to measure the distance walked in a prespecified time, like 6 or 12 min. The second kind counterpart EXPOSURE STUDIES ON RHINITIS would be to fix (individually) the pace which For allergic rhinitis there are two study designs at walking should be done and then measure of the experimental type available. Both are time walked. It is believed that the second exposure studies, one in the natural season, one kind of experiment is more relevant in the in an artificial season: study of COPD – that it correlates better with breathlessness and disability. The first kind of 1. The Park study. In this study the subjects are test is probably much influenced by attitude and exposed to pollen over a 1–2 day period by expectation. We should also note that the second walking around in a park. There are two main kind of test should provide a lower metabolic problems with this type of study – it is highly and respiratory stress than the first one and that dependent on season and the patients often the limiting factor in an exercise test does not find it very boring. have to be the physical fitness – COPD patients 2. The experimental Nasal Allergen Challenge may well fail due to muscular fatigue before Artificial Season model. In this type of study general fatigue. the subjects are artificially exposed to pollen In practice many tests used constitute a com- for some period. This can be either as spray promise between these two approaches: a spe- application for a few consecutive days, or cific time schedule is designed so that the work in an Environmental Exposure Unit (EEU) load is held fixed for some fixed time, then in which subjects are exposed to pollen in increased for ‘a step’ for another period of time, a special room for, typically, 3 hours on a etc. Typical cycle-ergometry and treadmill tests number of consecutive days. In this room there have this design, as has the so-called shuttle is a flow of air to which the pollen is added walking test in which the patient walks at a and evenly distributed in the air by fans. given pace for one minute, then increases the pace for another minute, etc., all according to In both cases we measure nasal symptoms as a well-defined protocol. The natural outcome of outcome variable. Both these studies are parallel RESPIRATORY 371 groups in design, but effects can often be such an study might be, e.g., budesonide Tur- demonstrated with rather small patient numbers. buhaler 200 µg b.i.d. 2. Studies in which the treatment is not fixed throughout the period under investigation. In LONG-TERM CLINICAL STUDIES WITH such studies we can either vary the dose DIARY CARDS of the investigational product, or vary the In a diary card study, the patient is provided with dose of some concomitant treatment. One a diary card to fill in various information about typical such study has an arm in which the status of his disease under investigation, often treatment is initiated with a high dose of a twice daily. For most asthma/COPD studies, given GCS, which is then reduced according the patients also measure PEF. It is important to some scheme until the patient is no that the patient uses the same peak flow meter longer controlled on the present dose. A throughout the study, since different brands have variant are the steroid sparing studies, in different scales, and there is a considerable which a fixed dose of some investigational within-brand variability as well. In addition treatment is given throughout the study period to this, some symptom scoring is requested. and concomitant with this treatment some This can be either an overall assessment of GCS is given, the dose of which is then symptoms, or assessments of specific symptoms, reduced in steps. For inhaled GCS, oral steroid like wheezing, shortness of breath and cough sparing studies have been performed in this for asthma. Finally, for asthma/COPD studies, way, for other anti-inflammatory drugs like the use of rescue medication, usually a short- leukotriene modifiers inhaled steroid sparing acting β2-agonist, should be entered into the diary studies are relevant. card. With the increased use of IT, paper-based diary cards are more and more replaced with The usage of the diary card data varies between electronic counterparts, which has the potential these two types of data. In studies with fixed benefit of monitoring when the recordings are treatments they define the primary efficacy vari- done. Some such devices can also contain a ables, whereas in studies with varying treatment spirometer, which makes it possible to replace doses, dose changes are conditioned on the diary the somewhat variable PEF measurement with card variables and these therefore act only as con- the more accepted FEV1 measurement. The fact trol variables. that the electronic device can be programmed so Diary card data in a long-term clinical study that it only accepts data obtained at the timepoint often represents a considerable amount of data, when it should be obtained, increases the validity as measured in megabytes on disk. The number of this type of data. The FEV1 measurements of megabytes, however, does not truly reflect its recorded with a portable spirometer should be information value. Data is not obtained in a very more valid than PEF data obtained by a peak controlled fashion. Morning values are generally flow meter and manually recorded on a paper- considered slightly sharper than evening values, based diary card. Our discussion will primarily since sleep is comparatively similar among relate to the old paper-based diary cards with a patients and data should be obtained and recorded concomitant peak flow meter. We leave it to the immediately after waking up. reader to assess potential changes that occur for For that reason, peak flow obtained in the electronically fetched data. morning is often considered the primary variable In terms of basic design we have two types of of interest in a long-term diary card study on asth- long-term clinical studies in asthma: matics. The day-to-day variability, for a symp- tomatic asthmatic, can be considerable. However, 1. Studies in which treatments are fixed through- using the mean of all values over a prespecified out the period under investigation. An arm in period, as long as possible, generally provides 372 TEXTBOOK OF CLINICAL TRIALS us with a measure that has proven to be useful also of symptom scores and of the use of rescue in many clinical studies. An alternative efficacy medication. Because of the intrinsic variability in measure is to use FEV1 measurements obtained at the underlying disease it is important to compute visits to the clinic. Though each individual FEV1 means over long periods, preferably the full measurement so obtained is much more reliable treatment period. This means that, for some drugs than a single PEF measurement, the overall mean at least, the mean will contain data from a period over a treatment period of daily recorded PEF of onset of action, though the effect of that will measurements obtained in the morning is, in our be minor in long-term studies. experience, a more efficient variable for demon- Mean values of symptom scores do not seem strating differences between treatments in lung very meaningful to most clinicians in assessing function. Since most treatments are mainly symp- the actual response. An alternative is to compute tomatic, integrated measures over time are the the percentage of symptom-free days, which is relevant ‘endpoint’ measures. somewhat simpler to interpret clinically. Simi- When using FEV1 obtained at visits to the larly it might be useful to compute the percentage clinic as primary variable in a long-term clinical of days with no rescue medication. trial, we must take the diurnal variation of lung When many symptoms are recorded individu- function into account. Thus it is important that ally, one approach to the analysis of the data is to FEV1 is measured at approximately the same time compute the sum of the symptoms (as the nasal of day on each visit. To obtain maximal efficiency index score), but an alternative is to analyse them we also need to schedule the patients for visits simultaneously in a multivariate analysis. to the clinic early in the morning (around 8 Even more useful, sometimes, is to collect data a.m.), with approximately the same argument within a day, or adjacent days, to form new as given for peak flow morning measurements measures. One simple such measure for asthma is above. The possibility of enforcing this will very to define a patient to have control of the asthma, if much determine the effectiveness of the FEV1 there are no symptoms and the patient did not use variable in discriminating between treatments. rescue medication. The percentage of such days In COPD studies lung function is also of with asthma control can be a useful summary interest, but for this disease the symptomatic measure for some patient populations, typically benefit is stressed more. For COPD, rating of rather mild ones. A variant of this is to define night sleep, breathlessness, coughing and chest mild exacerbations, or episodes, of asthma from tightness seem to be accepted symptoms to diary cards by looking for worsening of lung include in diary cards. function and/or increase in rescue consumption For rhinitis, the symptoms recorded are nasal and/or symptoms. The exact criteria for such blockage, rhinorrhea, sneezing and/or itchy nose episodes probably need to be adjusted to the which sometimes are combined into the nasal patient population under study, and to the study index score, which is the sum of them. In design. In order to avoid spurious events, it might addition to this, eye symptoms are recorded as be a good idea to define an event to have occurred a secondary variable. The most readily available for two consecutive days in order to be labelled objective measure in the clinical trial setting is an episode. From an analysis point of view we either the Peak Nasal Inspiratory Flow (PNIF) or can analyse time to first such exacerbation or Peak Nasal Expiratory Index. Of these the PNIF the percentage of episode-free days. Analysis of parameter seems to be the most discriminative. exacerbation is even more relevant in COPD The data in diary cards can be used in studies than in asthma studies. different ways to compute variables for use One final note on response data in studies on in statistical comparisons. As already indicated, asthmatic patients, especially PEF, and the dis- in fixed dose studies period mean values are often ease asthma in general. When we interpret diary computed, not only of PEF measurements, but card data, obtained over a longer period, we must RESPIRATORY 373 interpret it on a group mean level. A discussion General health status scales such as the Sick- on individual responders is virtually meaning- ness Impact Profile with 136 items26 have been less. A responder refers to a patient that responds proposed. A compromise between lengthy ques- to the investigational treatment. This cannot be tionnaires and single-item measures of health assessed on the basis of diary card data, since has also been proposed. The Nottingham Health the underlying disease is, by definition, vary- Profile with 45 items and SF-36 (a Measures ing – what seems to be a clear response could of Sickness short-form general health survey) well be a period of good asthma control totally are now widely used and validated. The SF-36 unrelated to drug effect (in some cases a study Health Status Questionnaire is based on 36 items effect) and the converse. This is obvious once selected to represent eight health concepts (phys- one has inspected placebo data in a long-term ical, social and role functioning; mental health; study. However this does not exclude that one can health perception; energy/fatigue; patin; and gen- define responders according to some criteria and eral health).27 Its quality of life scales have been compare percent responders between treatments, shown to correlate with the severity of asthma, since that is a comparison on group level. but it has yet to demonstrate any superiority over So far we have considered studies with fixed the simpler, diary card-based symptom scores for treatments during the investigational period. In a demonstrating effect in clinical trials. study which tries to identify the minimal dose For COPD the St. George’s Respiratory Ques- on which the patient is controlled, the obvious tionnaire has gained importance in later years. endpoint for analysis is this minimal dose. This is It is perhaps the most comprehensive question- true irrespective of whether the dose in question naire for evaluation of quality of life in airways refers to the investigational drug or to some diseases and allows for direct numerical compar- concomitant drug (as e.g. in the oral sparing isons to be made among patients, study popu- studies alluded to above). lations and therapies, and has sensitivity when More explicit examples of this will be demon- applied to mild as well as severe disease.28 It strated later. was developed by Paul Jones at St. George’s Hospital in London in 1990 and is designed to QUALITY OF LIFE measure impact on overall health, daily life and perceived well-being. The measure consists of 50 Asthma and COPD are chronic disorders that (76 responses) items that produce three domain can place considerable restrictions on the physi- scores and one overall score. The domains are: cal, emotional and social aspects of the lives of symptoms (severity and frequency), activity (that patients. Assessments of the patient’s own per- cause or are limited by breathlessness) and impact ception of the impact of asthma on their life, (on social life and psychological disturbances of general well-being, is known as measurement caused by the airways disease). of quality of life. Quality of life may be use- ful for assessing the degree of morbidity, e.g. in order to evaluate the health economic impact of CLINICAL TRIAL METHODS the diseases in the community. It is assessed by questionnaires that include a large set of phys- HOW TO AVOID BIAS ical and psychological characteristics assessing the general functioning and well-being in the con- Blinding text of lifestyle. Quality of life scales are either general and not specifically designed for patients Most effect measurements of the respiratory with asthma or COPD, or they are more specific diseases discussed here are influenced to a disease-related but, as of today, in general not non-negligible degree by the patients’ expec- applicable to the general population due to cul- tations. One typical example of this is that tural differences. in some double-blind, placebo-controlled single 374 TEXTBOOK OF CLINICAL TRIALS dose trials measuring bronchodilation, there is the group have one device and the other half the an immediate response in lung function also in other device. That way, at least, the patient does the placebo group, which probably is due to not know whether he gets active drug or not. (false) expectations. The classical methods to Open labelled studies might be acceptable for avoid expectation bias, blinding and randomisa- some systemic effects studies where the outcome tion, are therefore important. A clinical study in variable is the plasma concentrations of some this area should follow a double-blind approach marker, or in long-term safety studies. in which study drugs are prepacked in accordance with a suitable randomisation schedule, and sup- Randomisation plied to the trial centre(s) labelled only with the subject number and the treatment period so that Studies in respiratory diseases must also be no one involved in the conduct of the trial is properly randomised, so that prognostic fac- aware of the specific treatment allocated to any tors are distributed between groups by chance 29 particular subject, not even as a code letter. alone, which minimises selection bias. For many The code should not be broken until all deci- outcome variables some prognostic factors are sions concerning data validity have been taken known, and it is important for the credibility of and documented. the result that the choice of group for individ- Many studies in the respiratory area concern ual patients has not been made by the inves- inhalation products, where there are not only, tigator. The observed outcome can still be due say, two different drugs involved, but also two to an imbalance of prognostic factors between different inhalers (or perhaps one drug in two groups, but randomisation at least means that this different inhalers). To maintain blinding in those was produced by chance alone, not by the par- situations one often needs to resort to the ‘double- ties involved in the study (provided the study is dummy’ technique. This means that for each also blinded). inhaler there has to be two clones: one with active By the way, this last point is why it is drug and one with placebo. On each inhalation meaningless to do group comparability testing at occasion, the subject has to inhale not only from baseline. The p-value computed from a statistical the inhaler with active substance, but also from test is a measure of how certain one is that the the other inhalers, but containing placebo. This ‘null hypothesis’ is wrong, how improbable the might lead to a large number of inhalations per result is assuming that the null hypothesis is occasion, which in turn might lead to incomplete true. For a test comparing baseline data (i.e. data compliance. Note that the use of different inhalers obtained before randomisation), say the mean, for implies a consideration on the order in which two treatments, the p-value is a measure of how these should be taken. Carry-over effects from unlikely the observed group mean difference is. one type of inhaler might dictate which should If this is small, say p<5%, this means that an be taken first/last, whereas in other situations a unlikely event has occurred, or that the groups balanced scheme might be called for. are not equal on average. However, if we trust Rhinitis studies pose a special problem in terms our randomisation procedures, there is no factor of blinding because the double dummy technique that can explain why the two groups should not is not considered appropriate – there is a fear that be equal except for chance alone, so we must additional placebo material may clear the airways conclude that an unlikely event has occurred. of drug so that the response is different with The unlikely event, i.e. the difference observed at and without simultaneous placebo administration. baseline, might have consequences on the results. This is a problem mainly when two different But the extent of that effect does not depend on drugs inhaled through different devices are to be the p-value of a test at baseline, it depends on the compared. The partial remedy that is most often correlation of the effect variable with the baseline used is to include a placebo group, and let half variable in question. This means that a large RESPIRATORY 375 baseline difference might have no consequences problem, since they should occur with similar at all, or a small difference at baseline might frequency in different groups. However, a more have large consequences. In the respiratory area effective treatment is expected to have less con- it is very rare that the latter is the case. If sumption of rescue medication. As a consequence observed differences at baseline causes concern, there is a bias towards no effect by includ- the robustness of the results should be checked ing those measurements when computing period with respect to this issue, not a separate test means for instance. If we ignore them (i.e. con- at baseline. sider them to be missing) we introduce a bias in the same direction, since we only count the Other Sources of Bias days when the patient was, relatively speaking, symptom-free. There seems to be no easy way Another way to risk selection bias, also in a out of this dilemma, and the approach we have randomised, double-blind study, is to exclude taken is to ignore the additional information on data obtained on treatment. To exclude patients recently taken rescue medication for the analy- on data obtained prior to first dose cannot sis, but instead plot, descriptively only, for each in itself produce bias. However the prognostic day the proportion of patients that takes rescue close enough to peak flow measurement. Hope- factors for respiratory trials, like FEV1 in percent of predicted normal and reversibility, are only fully, and this is usually the case, the main result estimates of time-varying entities and are not and this graph gives the same message on effect. precise enough that we can actually claim that At least this approach should be conservative. a patient violating some inclusion criteria is not necessarily an appropriate patient for the TRIAL DESIGN CONSIDERATIONS trial. They are essential in order to focus the investigator on the appropriate patient population, From a bird’s-eye perspective, there are two types but once a violation to the protocol criteria has of responses for respiratory diseases: emerged it might be appropriate to use the patient in the statistical analysis. With Senn,30 I consider 1. Immediate responses that disappear within the protocol a guide to the physician, not the a short period of time. This includes the statistician. fast bronchodilating effect of β2-agonists, Protocol violations after the first dose should responses to various provocations and spe- not in general invalidate the patient for the cific systemic effects that are measured by statistical analysis. If such a protocol violation is markers in the blood (like plasma cortisol for confounded with treatment effects, their omission glucocorticosteroids and serum potassium for might bias the result. However, the fact that they β2-agonists). Many of these studies are single are protocol violations might in itself imply that dose studies. the measurements are improper measurements, 2. Long term studies addressing effects on symp- which is another type of bias. This problem toms or average lung function of the underly- is illustrated in respiratory trials by the use of ing condition. rescue medication. During an asthma or COPD trial patients are Crossover Trials usually provided with short-acting β2-agonists to use as rescue medication. This means that some For the first type of studies, the crossover design measurements of lung function and symptoms is well suited. In such a study each subject is will be influenced by this add-on therapy, and the randomised to a sequence of treatments, and acts validity of those measurements (as direct treat- as his own control for treatment comparisons. ment measures) will therefore be questioned. If In many cases this is attractive, because the treatments have the same effect they pose no within-subject variability is smaller than the 376 TEXTBOOK OF CLINICAL TRIALS between-subject variability which means that a Sequential Trials smaller study is required for the same power, as compared to a parallel group study. Numerous Sequential trials are not much used in the variations exist, e.g. trials in which each subject respiratory area. A long-term study can obviously receives only a subset of the treatments studied not be done this way, since the total study period (incomplete design), and trials where the same would be enormous. For experimental studies, treatment is repeated within a subject. on the other hand, the setup is often of such However, there are caveats with crossover complexity that the clinic need to have clear studies. The primary caveat is the possible specifications on patient numbers before the start presence of carry-over effects (in fact, non-equal of the study for their planning. carry-over effects), which might bias treatment comparisons. When deciding if a crossover Interim Analysis design is appropriate for a particular study, we therefore must convince ourself, beforehand, that Interim analyses should not be needed in trials we can get rid of possible carry-over effects in the respiratory area, except possibly for first by separating the various treatment periods with dose in man studies on new drugs – but then it washout periods during which no treatment is a drug issue and not a therapeutic area issue. is given. For a single-dose short-acting β2- Interim surveys of safety data might well be agonist trial a washout period need in general justified in the beginning of clinical programmes, only be a few days. A trial which studies but for efficacy issues they are seldom needed. cortisol depression after short periods of GCS administration should have washout periods of 1–2 weeks, though shorter ones are acceptable HANDLING OF MISSING VALUES in single dose studies. The experimental type of studies are often very When periods in crossover trials contain difficult to perform clinically. To get good quality repeated dosing over a few weeks, and the actual data, it is extremely important that the investiga- experiment is performed at the end of such a tor with staff has a good knowledge and experi- period, it is often unnecessary to have drug-free ence in this type of study. It is in general better to washout periods between periods. But to take that do such studies in one or two experienced cen- step, one must make plausible that taking a new tres with many patients, as compared to using treatment directly after another does not by itself many centres with fewer patients – despite the have any effects on the variables to analyse. fact that the study might take longer to perform. With those premises missing values are, in our Parallel Group Trials experience, a negligible problem since in gen- eral experimental sequences are complete. When Since the treatment for respiratory diseases in missing values occur, they are due to discontin- general is to achieve a prolonged improvement uations between experimental sessions and these of the underlying condition, the most important are few and there is no problem in analysing the trials need to extend over longer periods. The resulting unbalanced study. most natural design for them is the parallel group It is a completely different issue with the design, where the subjects are randomised to one long-term clinical studies. Here patients not only of a number of arms, each arm being allocated a discontinue treatment, but there are also missing different treatment. These treatments will include observations during the period. For the fixed the investigational product at one or more doses, treatment trial the purpose is in general to achieve possibly including a placebo (dose = 0) arm, and a steady state, on group level, on the treatment possibly some active control treatments at one or and then compare the level of the measured more doses. variable in steady state between groups. For RESPIRATORY 377 patients reaching steady state we therefore in from the analysis means that there might be a general have a number of data points measuring potential bias in the end result. To understand steady state level, in the case of diary cards quite this, assume that there are no withdrawals in a few. The efficacy variable is then usually a group A, but half the patients withdraw from summary statistic of these data points, like a group B because of insufficient efficacy. The period mean of a diary card variable. Missing remaining patients in group B are then the ones values during this period does therefore not who needed less treatment. That patient group constitute a major problem – we take the mean has a corresponding subgroup in group A of of available data. approximately the same size (as a consequence of Sometimes a long-term clinical study contains proper blinding and randomisation procedures), experimental procedures, like a methacholine but group A also contains another subgroup of provocation test. Or just spirometry at the clinical patients, corresponding to the ones that dropped visit, at least FEV1. This is then done at least out from group B. The remaining groups are once pre-randomisation and then again only a therefore not really comparable, and inference few times on treatment, in particular on the last drawn from available data might be misleading! protocol visit. The effect variable should not However, there is no simple, trustworthy, be defined as the change from baseline to last remedy for this. Our approach is to use available protocol visit, but as the change from baseline data for the analysis, hoping that the potential to the last visit on treatment the patient attended. bias is conservative (which it probably is in most Specifically, instead of analysing the change in cases for respiratory trials). However, if there is a log PD20 from visit 2 to visit 8, we define the large difference in withdrawal rates between the efficacy variable as the change from visit 2 to groups, it is logical to do the primary analysis on the last visit on treatment, which might be visit withdrawal data to assert group differences. 4, 6 or 8 in a particular study. Technically this is When describing diary card data, daily mean equivalent to what is called the last value carried value curves by treatment are useful. When forward, or the last value extended, principle, but computing these mean values, missing values there is no need to use that label if we define the pose great problems in that raw mean values efficacy variable appropriately. can produce very misleading impressions on This still leaves us with one key prob- group behaviour. To see why, consider a placebo lem – what if we do not have any efficacy mea- arm in a diary card study in asthma in which surements on treatment to use. To avoid that the patients with worsening of symptoms drop problem in diary card studies, it is often better out progressively (the worse the symptoms, the to define the full treatment period as the period earlier they drop out). As the patients with over which to compute summary statistics. At low response values drop out, the group mean least that provides an effect measurement for each will increase, so the temporal behaviour of the individual who has started to fill in the diary mean values will indicate that the placebo group cards. The drawback is that the period mean for increases in effect with time. However, this effect one patient can be the mean of 90 data points, is solely due to withdrawals! whereas for another it is the mean of only a few To avoid this culprit when plotting the tempo- data points. The next step is in general to analyse ral behaviour of variables some kind of impu- these period means with an ANOVA, and then tation of data is needed, in order to keep the the information of the precision of the computed denominator the same when computing mean val- mean is lost. On balance, it seems better to have ues within a treatment group. The simplest such an effect measurement on each patient. imputation is to use the last value extended (LVE) For data obtained on clinical visits, the risk approach, in which the last value for a withdrawal of having no measurement at all on treatment is extended to later timepoints. Using this prin- is not negligible. The omission of such patients ciple, the mean values plotted can be interpreted 378 TEXTBOOK OF CLINICAL TRIALS as follows: the mean at time t is the mean of SAMPLE SIZE DETERMINATIONS the last recorded measurements up to and includ- ing time t. When using this principle for diary In order to certify that a proposed study is of card variables like PEF it is often better not an appropriate size, a sample size justification is to take only the last measurement, but rather needed in the protocol. It cannot be justified eth- a mean of a few measurements. More sophisti- ically to succumb a number of patients to a study cated approaches based on some kind of multiple protocol if there is no hope whatsoever to demon- imputation technique for missing data can also strate what you want to demonstrate. Similarly, if be considered, but the add-on value of doing that the study is heavily overdimensionalised we have is probably very small for the average study in put an unnecessary number of patients at what- respiratory diseases. ever risk the study can carry with it, and that is not ethical too. However, sample size deter- mination is there to ethically justify the study MULTIPLE COMPARISONS in advance – it has no consequences when the results are obtained. A respiratory trial usually contains a number In the respiratory area many test hypotheses of effect variables, and often also a number are stated in terms of mean values, and for of different treatments. Thus there are multiple such variables the sample size is (essentially) comparisons to be done. This poses a major proportional to the ratio (σ/)2,whereσ is the problem, because of the risk of over-emphasising residual standard deviation and is the mean fluke significances because of many comparisons. difference we do not want to miss. When using a To handle the many effect variables we there- multiplicative model for a variable, these entities fore have to predefine which one is to be con- refer to the logarithm of the variable in question. sidered the primary one. It is from the result on Note that σ means different things in a crossover this variable that the overall statistical conclusion trial and in a parallel group trial – in the former from the study can be drawn. In general one study case it refers√ to a within-patient variability (more can have a few different objectives that are not exactly 2× the residual standard deviation of closely related (like efficacy and safety), and then the ANOVA) and in the latter to a between- a primary variable for each objective should be patient variability. Also what is relevant is the appropriate. However, it is probably a too sta- residual standard deviation from the proposed tistical approach to focus only on the primary analysis of variance, which might contain a variable when trying to understand the results of baseline adjustment. a clinical trial. No variable fetches all aspects of The following table shows some typical values a respiratory disease, and the approach should be of the sample size parameters that can be used for to select the most sensitive variable as primary asthma trials. Each example will be discussed in variable, to decide on the overall conclusion, but more detail below. then a number of secondary variables should be so described that one gets an overall picture of 2 what is ‘going on’. Increase in Design σ(σ/) When it comes to the problem of multiple PEF morning PG 40–45 10–20 4–20 (L/min) treatment comparisons, the study logic should be Symptom score PG 0.4–0.5 0.05–0.15 6–100 structured in terms of well spelled out objectives. (0–3 scale) To prove efficacy might mean one comparison, FEV1 (L) PG 0.4–0.5 0.05–0.1 15–100 ln(AUC FEV1) XO 0.07–0.10 0.05–0.10 0.5–4 to estimate a relative dose potency another (L) 2 analysis. With precisely formulated questions the log(PD20) XO 0.9–1.1 1–2 0.8–4 multiplicity problem here should at least diminish substantially. This approach will be illustrated in Here the range is not a range – the lower number what follows. for σ represents an optimistic number, the larger RESPIRATORY 379 number a conservative one. Similarly, for inevitably presents itself, as in so many areas of the range is more of a typical range for which medical statistics. It is however no more sensible to dimensionalise, not a range on what can be to do such analysis on data on lung function obtained. To obtain numbers (per group in the than in the general case.30 Ultimately it has to parallel group case) from this we should multiply do with the generalisability of the results of by approximately 25. atrial. For the crossover measurements of the table, In airways diseases, asthma in particular, the we just note that the AUC refers to AUC-based disease severity varies among patients. Thus average over the full period and that for that the magnitude of the response attainable will variable the pre-dose FEV1 value is used as vary between patients. How much will partly covariate in the analysis. For the PD20 case no depend on what we measure. If we measure baseline covariate is used. lung function, patients with small lungs, like lit- For the parallel group measurements we use tle women, will be expected to have smaller baseline covariate. For PEF morning a baseline numerical effects than patients with large lung is obtained as the mean value over a number volumes, like tall men. This does not mean of measurements, typically 1–2 weeks, and then that the actual benefit to the patient is less, the effect variable as the mean of 1–3 months only that the outcome variable suffers from of data. The shorter the periods, the larger the residual standard deviation. Similarly, for this deficiency. We could remedy this partially by measuring lung function in percent of pre- FEV1 the table refers to a measurement both at baseline and end of treatment, but the treatment dicted normal, which tries to capture size dif- value could well be a mean of a number of ferences, instead. But that is only a partial remedy to a larger problem – that there is a measurements. Moreover, the FEV1 data refers to the situation when visits to the clinic are spread large heterogeneity in response sizes for some out over the morning, the European style, as outcome variables, which does not necessarily discussed earlier. With visits scheduled at 8 a.m. reflect different clinical responses. Compliance to precisely larger effects can be expected. study procedures might also influence the mea- Concerning symptoms scores, these too are sured responses. obtained from period means of diary card data, This heterogeneity in the disease population and relate to a typical asthma study. Changes is not a problem for the proper conclusions of in symptom scores are often small in studies in a clinical trial. Consider a randomised parallel asthmatics with mild–moderate severity, since group study in which treatments A and B they do not have many symptoms on entry. In are compared. If properly conducted observed rhinitis studies a combination of symptom scores treatment differences should be explained by is often done. If we use the TNS discussed earlier different treatment effects alone, and any claim we typically have a standard deviation of about from the study should relate to the relative 2 1.3 and effect sizes of 0.5–1, giving a (σ/) of effect of the two treatments. If it turns out 2–8. Typically, therefore, rhinitis studies can be that the effects differ, say, between men and smaller than asthma studies. women, that should not jeopardise the outcome For COPD, exacerbation rates are more impor- of the trial, since a proper randomisation implies tant as outcome variable. A rate of one exac- that the sex distribution is similar in the two erbation per year can be used in sample size groups. We get problems only when we try to calculations. compare effect sizes between studies. Differences in effect sizes could well be explained not only SUBGROUP ANALYSIS by different patient populations, including gender When doing statistics on trials in respiratory distribution, but also by different compliance to medicine, the question of subgroup analysis study procedures in the trials. 380 TEXTBOOK OF CLINICAL TRIALS
This discussion implies that estimated treat- PHASE I/II STUDIES ment effects, at least for lung function variables, should be interpreted with caution. It is often EFFICACY STUDIES much better to try to transform the information In terms of efficacy, not much can be done in on the effect scale to a dose scale, as will be a phase I trial. These trials, mainly concerned extensively discussed in sections to come. with tolerability and pharmacokinetics, give no Another implication of the discussion is that real clue to whether a new drug actually works if we start to compare different subgroups in or not. Note that in general a respiratory drug an asthma trial we can expect to find different must be very well tolerated to be useful, since responses. One such example is the multi-centre there are so many efficacious and safe drugs on trial, in which we have many centres, often the market. from different countries, with different patient When trying to establish that a new drug is populations and medical traditions. Different effective in asthma or not and to estimate clini- effect sizes should not come as a surprise, cally relevant doses, the approach differs between and do not necessarily indicate interpretational the drugs that have more or less immediate problems in terms of overall effects. We would effect on lung function, typically bronchodila- also expect that if we compare the effects in a tors, and the ones that work more indirectly on subgroup of mild asthmatics to a subgroup of lung function, via the inflammatory process, as more severe asthmatics within the trial, they will glucocorticosteroids. For bronchodilators we can not be identical. use small-scale experimental studies, whereas for Subgroup analysis will by definition be less anti-inflammatory drugs we typically need long- powered than the analysis of the full analysis data term studies from the very start. set. As a consequence, subgroup analysis is not To establish efficiency is conceptually simple: a sensible thing to do in a single trial, unless all we need to do is to show that a given well specified in advance, and then the study dose of the drug is superior to placebo. There size should be large enough for this subgroup is however no true placebo treatment in long- analysis. If not specified in advance, the risk of term asthma or COPD trials – at a minimum a fluke significance because of the ever-present the patients need to be provided a short-acting multiplicity problem is too large, leading to false β2-agonist to be used as rescue medication. All new drugs are therefore studied on top of some conclusions. baseline treatment, which in most cases is not However, subgroup analysis is in general very constant. For example, a GCS treatment totally unnecessary. If we want to see if the is taken in addition to rescue medication. It response differs between men and women, as an potentially becomes a problem when you want example, the proper approach is not to analyse to introduce a new rescue treatment, which is to the subgroup of men and the subgroup of women be taken when needed, and you need to document separately. It is to analyse the full data set with long-term effects. an interaction factor for treatment and sex. If The next question, possibly posed in the same this interaction test is statistically significant, we study, is which is the relevant dose-span for fur- know that there is a difference in the response for ther investigation? For a dose–response study men and women. We can even quantify it, if we (including the simple one with only placebo and so choose. Similar adjustments can be done for one dose) the choice of patients is important. The disease indices like lung function in percent of majority of asthmatics, the mild ones, do not need predicted normal or reversibility. But the add-on much therapy and because of the relative impre- value of doing this remains to be proven, except cision in the measurement tools, many patients that it might reduce the residual variance in the will have no measurable response in many of analysis of variance model. the tests discussed. For a bronchodilator study RESPIRATORY 381 of the crossover type, where the design contains effect). This does not rule out (if the slope is a number of experimental days when response is positive) that small doses have a negative effect followed after a dose, some bronchoconstriction and larger doses a positive effect, something is needed in order to see an effect. By defini- that should be borne in mind when interpreting tion the bronchoconstriction varies with time for the result. an asthmatic, so the measured response will not What can be done in a placebo-controlled only depend on the dose of the bronchodilator, dose–response study without an active control but also on the degree of airway narrowing on is to estimate the minimal effective dose (MED). the particular day the experiment is done. Thus This addresses the question: ‘Which is the lowest it is difficult to assess efficacy for the individ- dose, of those studied, that is proven effective’ ual patient, which we handle in clinical studies andtests,inarecursivemannertocontrol by considering means. But more importantly, in significance level, the doses from highest to order to achieve dose–response we need suffi- lowest with placebo. There are a number of ciently many patients with sufficiently many days algorithms available that can be used, but the key on which they can respond. The selection of such point about MED is that it is the lowest dose that patients is not easy! we from this study can claim is effective. Thus For long-term diary card studies we have a the result depends heavily on the size of the study similar problem. The effect measures are rather and choice of patient population, a property the noisy, and we generally need somewhat large average physician probably would not like MED studies to measure a signal through all the noise. to have. Thus there is a great danger in using In parts of the world with a widespread health MED as defined here for decision making. In care system most asthmatics are rather well my view MED is more likely to lead to false treated. In particular the use of GCS already decisions than correct ones. in fairly mild asthma in Europe, Australia and The information we actually want from a Canada seems to have made the majority of dose–response study is the shape of the dose–re- asthmatics more or less symptom-free for most of sponse curve, to allow us to pick the ‘best’ dose. the time. That in turn means that the traditional Not a detailed shape, a simple approximation diary card might have difficulty in catching which can be used to derive insights from. As any responses. long as there is a monotonous dose–response From the foregoing discussion it should be curve a more complicated description than the clear that the magnitude of response in a one provided by the formula particular variable is very difficult to assess: if γ it is small, is it because the patients studied E = E0 + Emax/(1 + (ED50/D) )(1) did not have much room for improvement or was it because the drug was of minor efficacy? is rarely needed. This formula contains four The only way, it seems, to actually assess the parameters: E0 is the baseline level of the degree of response is to compare it to something response variable, corresponding to placebo. we know, by experience, to work – to include Emax is the maximal effect attainable, ED50 is the an active control in the clinical trial. In my dose required to obtain 50% of this and finally, personal view, a clinical dose – response trial γ is a sensitivity parameter which measures how in asthmatics without an active control has very much the response changes with changes in dose. little information value. Also note that we should The shape of this function is a sigmoidal curve not need placebo in order to prove efficacy – it with the extremely important property that over should suffice if we could prove that there is much of the range (say from E0 + 0.2Emax to a dose–response relationship (this is in fact the E0 + 0.8Emax) it can be well approximated by point with the expression ‘show dose–response’: a straight line E = a + b ln D. A description to prove that the drug has a pharmacological of such a dose–response curve should be the 382 TEXTBOOK OF CLINICAL TRIALS purpose of the dose–response trial, not to discuss studied in a randomised, double-blind, double- the individual doses that were actually chosen to dummy crossover study: 6, 12 and 24 µgofdrug be used in the study. A,50µgofdrugB and placebo. Identifying the dose–response curve, however, In Figure 22.2 we see the geometric mean does not give you a hint on how well the values, expressed as percent increase from the treatment compares to competing treatments. For measurement taken prior to treatment adminis- that purpose it is wise to include an active control tration. The reason for plotting geometric, and in the trial. We can then use the dose–response not arithmetic, means is that results are often to curve to estimate the dose of the new drug that be expressed as percent increases, and then data produces the same effect as the active control should be analysed on a logarithmic scale and does, hopefully with confidence limits. geometric means are therefore more natural than arithmetic ones. As a consequence, differences Example: Bronchodilation are unnatural entities to discuss and should be replaced by ratios. The bronchodilating effects of two long-acting To actually analyse the data we want some β2-agonists, we call them A and B, each with overall summary statistic that includes both the its own inhalation device, were compared by maximal effect and the duration of response and giving single dose administrations, followed by we use the area-based average FEV1,av over repeated measurements of FEV1 over a 12- 12 hours. If we want to keep the idea of analysing hour period. The following five treatments were on a multiplicative scale, we need to compute the
26
22
18
14
10
6 (% increase from baseline)
1 2 FEV −2 A 6 µg A 12 µg A 24 µg −6 B 50 µg Placebo −10 024681012 Time since treatment administration (hours)
Figure 22.2. Geometric mean values, expressed as percent increase from the baseline measurement, of FEV1 measurements over 12 hours for individual treatments RESPIRATORY 383 area all the way down to zero. Alternatively, we drug B? To do this, we fit (weighted linear could integrate over the baseline measurement, regression to keep track of the uncertainties of but then the area could be negative and we would the means31) a straight line to drug A means be forced to do the final analysis on the original vs. log-dose and estimate the dose that has the scale. We have chosen the former approach. same mean effect as the reference treatment. The ANOVA uses the model This is illustrated in Figure 22.3. The actual dose estimate was 9.3 µg with 95% confi- µ ln(FEV1,av) = patient + treatment dence limits 3.4–19 g. As a consequence we find that 24 µgofdrugA as a single dose has + period + ln(FEV ) 1,base greater bronchodilating effect over 12 hours than 50 µgofdrugB. To have baseline as a covariate in a single dose 3. What about duration of effect? Looking at study is rather essential. If we observe lung Figure 22.2 we see from the placebo curve function over a short time period, baseline is indications of the diurnal variation that is very important and we could probably just as known for lung function. To define duration well use FEV1,av/FEV1,base as effect variable. by identifying when individual curves cross a However, when we observe over a longer time line, say 15% above baseline, does not seem period, the influence of baseline should diminish appropriate – if the placebo curve drops, you and after many hours could probably just as might still have a good response even when well be ignored. By using it as a covariate, you are back to baseline. A more statistically we get a reasonable compromise between these sound approach would be to rephrase the two extremes. question as ‘is there still an effect after Based on the results from this analysis we can 12 hours’ and then compare the treatments to address various questions: placebo at that point in time. This is done in the following table: 1. Which doses of drug A can we claim to be effective? To find this out we compare them, from highest to lowest dose, with placebo. Mean 95% Confidence Here is the result in tabular form: Treatment ratio limits 24 µgofdrugA 1.246 1.187, 1.309 Mean 95% Confidence 12 µgofdrugA 1.176 1.120, 1.234 Treatment ratio limits 6 µgofdrugA 1.168 1.112, 1.226 50 µgofdrugB 1.200 1.144, 1.260 24 µgofA 1.214 1.177, 1.252 12 µgofA 1.176 1.140, 1.212 6 µgofA 1.154 1.119, 1.190 Again mean ratio relates to placebo and we have responses that are between 17% and 25% larger than that after 12 hours. Thus all active Mean ratio relates to placebo. We see that treatments clearly have a duration in excess the mean effect is 15–21% larger than it of 12 hours. is for placebo, and the confidence intervals clearly show that all these comparisons were SYSTEMIC EFFECTS statistically significant. So we can claim that 6 µg is an effective dose of drug A, without Effects of anti-asthma drugs are in general not compromising the significance level (see the confined to the lungs. Both β2-agonists and discussion on MED). GCSs have receptors on/in cells throughout the 2. Which dose of drug A has the same mean body. Since the drugs are cleared through the effect as the reference treatment, 50 µgof bloodstream they will therefore have systemic 384 TEXTBOOK OF CLINICAL TRIALS
2.70
2.65 (L) 1 2.60 mean FEV
2.55
2.50 101 Dose (µg)
Figure 22.3. Treatment mean values for 12-hour average FEV1 with fitted log-linear dose–response curve for drug A and estimation of Deq relative to 50 µgofdrugB effects (albeit perhaps not measurable). In con- With this model in mind we can use cortisol trast to the anti-asthmatic effects, these effects in plasma as an index of the systemic burden can be measured both in healthy volunteers and of therapeutically given GCS. By measuring in patients. the effect on cortisol we get a rather direct GCSs are synthetic cousins to an endogenous, measure of the overall potency and concentration anti-inflammatory, substance, cortisol. Given this, in plasma of the GCS. We can however not we can compare the pharmacodynamic systemic measure it timepoint by timepoint and compare effects of different GCSs by comparing their to measurements without drug, since the level effects on endogenous cortisol levels. This has the of cortisol is determined as a balance between added advantage over drug plasma concentrations production and elimination (with a half-life of that it accounts for differences in potency in about 1.5 hours) and the GCS acts on the decreasing plasma levels of cortisol (which is production side. We therefore need to study a done by negative feedback on the HPA axis). longer period. The cortisol levels in plasma have It is important to state at this point that we a diurnal rhythm which is very pronounced, so do not study endogenous cortisol levels because the most appropriate study to do is to give they themselves represent a dangerous side- repeated doses of the GCS until a new steady effect. They are studied because they are sensitive state has been reached and then measure p- markers for the ‘amount’ of exogenous GCS in cortisol for 24 hours. A typical schedule is every the body! other hour. The most useful variable to study is RESPIRATORY 385 the area under the curve for those 24 hours. In steady state, when there is a 24-hour periodicity, Mean 95% Confidence this is proportional to the amount produced Treatment ratio (%) limits p during 24 hours. 200 µgofA 97.4 73.3, 129.3 0.85 The actual clinical consequences of the levels 400 µgofA 94.1 70.9, 125.0 0.67 attained are very hard to assess. They are 1000 µgofA 70.0 52.7, 92.9 0.014 µ surrogate measures of the long-term effects, 200 gofB 76.4 57.6, 101.5 0.063 375 µgofB 53.5 40.3, 71.1 0.00003 but as such they should provide useful relative 1000 µgofB 9.1 6.9, 12.1 <0.00001 information on different GCSs, though we can never expect effects on p-cortisol to be a perfect Here the mean ratio is presented as a percentage. predictor of say relative risk of osteoporosis. The For instance, 76.4% for 200 µgofdrugB means distribution pattern in the body of the GCS might that there is a suppression of 100 − 76.4% = be of some consequence for this. 23.6%. This result does not tell much about how the Example: Comparison of Plasma Cortisol drugs compare. To do that we can fit parallel Dose–Response Curves non-linear dose–response curves to the mean effect data, adjusting for precision by using a We want to compare two inhaled steroids (with weighted non-linear regression.31 We assume for inhalation devices), call them A and B in this analysis that given enough steroids, the terms of their degree of suppression on the cortisol levels go down to zero. The result is plasma cortisol level. The comparison is done in graphically shown in Figure 22.4. healthy volunteers in a randomised, open seven- With the appropriate parametrisation here, we way crossover study. Each treatment period obtain that the relative dose potency is estimated consists of 4 days, and there was a washout to be 3.7, with 95% confidence limits 2.7 and 6.4. period of at least 2 days between each such Thus, in terms of depressing cortisol levels, B is treatment period. Each steroid was given in estimated to be about four times more potent than three doses: 200, 400 and 1000 µg b.i.d. for A (remember that a letter stands for a GCS plus A and 200, 375 and 1000 µg b.i.d. for B. a device – change device and this relation might The seventh treatment was a placebo treatment. change). Or put in other words: to achieve the Blood samples were measured every second hour same average depression in cortisol, we can use during the last 24 hours in each treatment period a four times larger dose of A than of B. (10 p.m. to 10 p.m.). In all 21 healthy volunteers Having obtained this result, the immediate participated in the study and all completed all question is: ‘How relevant is this result to the treatment periods. target group of the drug – the asthmatic patient?’. The effect of the fact that the study is open We cannot extrapolate these results to patients is hard to assess. If the administration of one ‘as is’. There is a basic difference between a of the drugs is associated with more stress than healthy volunteer and an asthmatic – the latter the administration of the other, this might bias has an ongoing inflammatory process. This means the result. However, this seems unlikely, and that the dynamic system regulating cortisol is doing the study open has the benefit that fewer disturbed (compared to healthy) and we can inhalations are required on each occasion. expect smaller absolute effects of a given dose To analyse the study, we first do an ANOVA. of the GCS. So a typical patient might have a It is done on the logarithm of the concentrations larger ED50 than a typical healthy volunteer. By with standard factors for a crossover study: sub- the same token we can expect different patients ject, treatment and period. As a first presentation to vary considerably in this respect. of the results we can compare all active treat- However, there is no reason to expect that ments to placebo: different GCS should act differently in patients 386 TEXTBOOK OF CLINICAL TRIALS
160
140
120
100
80
60
40 Mean cortisol level 24 hours (nmol/L) Mean cortisol level
20
0 102 103 Dose (µg)
Figure 22.4. Estimated dose–response mean value curves for treatments A (to the right) and B (to the left) and in healthy volunteers, i.e. there is no reason not be an issue of a new drug. Thus the to claim that the relative effect of two GCS, pivotal confirmatory trials for the airway diseases as measured by the potency ratio, ρ, should asthma, COPD and rhinitis are parallel group, differ between patients and healthy volunteers, or diary card studies typically spanning from a between different categories of patients for that month up to a year. matter. If such differences turn out to be the case, the reason must be that the systemic dose differs Asthma Trials between healthy volunteers and patients, and For asthma the typical study length for an efficacy then, most likely, between patients of different study seems to be 3 months, though occasionally degrees of severity in their disease. longer studies are needed. As already discussed, there is a continuous scale of severity of each of PHASE III STUDIES the respiratory diseases. The severity of asthma can be classified into a few groups, each of which DOCUMENTING EFFICACY AND SAFETY has its recommended medical treatment. Both the classification and the recommended treatment Most drugs for obstructive airway diseases are are, like the disease under study, somewhat given for maintenance treatment, and the main varying with time. The following classification point to document is the level of disease control is meant only to be indicative on what type of of the proposed treatment. At the same time criteria are used for such classification. it is important to document the adverse event profile, since most of the drugs in this therapeutic Intermittent: These patients have normal lung area are considered very safe, and safety must function between occasional exacerbations RESPIRATORY 387
with symptoms at most once a week. FEV1 in show little improvement in lung function. Instead percent of predicted normal should be ≥ 80%. the focus for studies in severe patients might be Mild persistent: Now symptoms appear weekly, on how a new treatment can substitute for an old but not many times a day and exacerbations treatment without compromising the patient: e.g. may affect both activity and sleep. FEV1 in how much oral steroids can be spared by tak- percent of predicted normal should be ≥ 80%. ing inhaled GCS, or how much inhaled steroids Moderate persistent: Symptoms appear daily, can be spared when taking a concomitant leu- and affect both activity and sleep. There are cotriene modifier. night-time symptoms weekly and FEV1 is The main objective of the confirmatory trials within 60–80% of predicted normal. is to show efficacy, and thus require a placebo Severe persistent: These patients have contin- control. This was discussed earlier. Concerning uous symptoms, frequent exacerbations and doses, for the majority of drugs for airway night-time symptoms, which limits physical diseases, there is not one dose that is appropriate activity. FEV1 in percent of predicted normal for all patients. Instead a range of doses has is ≤ 60%. to be justified and documented in the clinical programme. The general discussion on MED and This classification borrows from the GINA dose–response is relevant here too. classification.32 However, that classification also uses a concept of variability which we do Example: A Diary Card Study with Fixed not discuss. Treatment Arms To describe the patient’s disease severity in a clinical trial we use data obtained at The main outcome variable in a diary card a visit to the clinic before randomisation. In trial with fixed treatment arms is some kind of addition, a diary card is provided for a run-in summary measure (typically a mean value) over a period to assess symptoms and use of rescue longer period, presumed to represent, on a group medication. Inclusion into a study is often based level, for most part a steady state situation. We on these measurements. A more practically, also need a corresponding measure during the oriented classification of the severity of asthma is baseline/run-in period for the statistical analysis. based on the use of GCS in the patient’s regular Figure 22.5 shows the estimated daily means treatment: no GCS (intermittent–mild), inhaled of a 3-month clinical trial in asthma. The study GCSupto400µg/day (mild persistent), inhaled was a multi-centre, double-blind, double-dummy GCS in the range 400–1000 µg/day (moderate) study of 3 months’ treatment with investigational andinhaledGCS≥ 1000 µg/day (severe). So drugs. There were 52 centres in seven different the daily dose of background GCS treatment countries worldwide and the randomised treat- can be used as an indicator of asthma severity. ments were placebo, two dose levels, 100 and 33 Another classification is based on PC20. 600 µg daily dose, of the GCS A and one dose The best way to use the information in diary level, a daily dose of 200 µgoftheGCSB.In cards might vary between patient groups. As all 547 patients were enrolled, of which 472 were already explained, the traditional use of diary randomised and 383 completed the study. There card data is to assess changes in period means. were twice as many withdrawals in the placebo This is expected to work best in patients with group as in the other groups. The mean age moderate asthma. In the intermittent group one was 44 years, the mean FEV1 in percent of pre- should not expect effects of any considerable dicted normal was 70% and the mean reversibility magnitude because the lung function is close to was 24%. All patients were on inhaled steroids normal, and the patient is, for most of the time, when entering the study – the mean daily dose symptom-free. In the group of the most severe was 850 µg ranging from 500 to 1500. Thus the patients, patients often have obtained an irre- population must be characterised as being mod- versible component to the disease and therefore erate–severe. 388 TEXTBOOK OF CLINICAL TRIALS
To plot the temporal behaviour of the effect 2. If the patient withdrew from the study before of the four treatments, simple mean values are the 90th treatment day, the mean of the last expected to produce a bias towards no effect. At three recorded days is extended from the least when comparison is done to placebo: in this last recorded day plus one to day 90 post- group there were more withdrawals and many of randomisation. these can be expected to be due to reasons that are correlated to (lack of) treatment effects. Some Then daily means are computed by treatment. weeks into the treatment period, raw mean values In Figure 22.5 an additional operation has been for this group will therefore mainly include made: for each treatment we compute a baseline patients that are not in desperate need of GCS by taking the average of the run-in means and treatment. This will introduce a selection bias subtract that from all daily means. This is done which will partly hide the effect of the treatments. in order to highlight changes since that is what To avoid this problem, that different days will the analysis focuses on. contain different patients, we have pre-processed The statistical analysis of this data uses the data when plotting Figure 22.5 to make sure all period means from the individual patients: the days contain data from all patients. The way this mean is computed for the run-in period, which is is done is as follows: used as a baseline, and then for the treatment period (from the day of randomisation and 1. Linear interpolation is done in order to impute onwards). The change from baseline is used as all missing values between the first and the last effect variable, and the analysis is an ANOVA recorded day for each patient. with treatments and centre as factors and baseline
15
10
5
0
−5
−10
−15 A 100 µg A 600 µg Change in PEF Morning (L/min) from baseline −20 Placebo B 200 µg
−25 −20 0 20 40 60 80 100 Days since randomisation
Figure 22.5. Estimated daily mean values of the change from baseline in PEF morning. See text for computational details RESPIRATORY 389 as covariate. The following table shows the no asthma symptoms and no rescue medication adjusted mean values for the effect variable was needed. The percentage of such days is for the four treatment groups, adjusted to a often a useful variable, at least in studies on common baseline value (the mean over the full mild–moderate asthmatics. Modification to a study population): ‘well-controlled’ day for more severe patient populations is possible. Another approach to diary card data is to mea- 95% Confidence sure time to first exacerbation. In a population Treatment Mean SEM limits like this, this is not expected to produce better placebo −10.5 3.3 −17.0, −4.0 results than the ones presented but, as already dis- A 100 µg −0.3 3.3 −6.8, 6.2 cussed, this approach might be useful in milder A 600 µg 7.4 3.3 1.0, 13.9 populations. For illustration, we have used the B 200 µg2.03.4 −4.7, 8.8 following definition of an exacerbation for the present study (which was in moderate–severe patients): for there to be a mild exacerbation one From this analysis we also find that there is a of the following three criteria should be fulfilled statistically significant, negative, dependence of on two consecutive days: the change in PEF on the baseline PEF and that the estimated residual standard devotion was 35.9 L/min. As is common for this kind of 1. Morning or evening PEF should be more than data, the explanatory power of the analysis is 20% below the period mean during run-in. small: only 8% of the variability is explained by 2. Rescue medication should be up at least four the model. inhalations above the period mean during run- Other diary card variables are often analysed in. the same way as was shown for PEF morning, 3. The patient woke up during night-time and by first computing individual period means. took rescue medication. Symptom scores and rescue medication are however variables for which the average value By scanning the diary cards we can, for each is not necessarily easy to interpret. Symptom patient, compute the time to first exacerbation, if scores are really ordered categorical data, and there is one, otherwise the patient is censored. even though the average gives a hint of the With this definition, we can present results amount of symptoms, it is not clear that e.g. as Kaplan–Meier plots for the time to event, (2 + 2)/2 means the same as (1 + 3)/2. For as in Figure 22.6. We see a picture similar to rescue medication, the problem is mainly that that conveyed by the period means. Statistically we use the mean as a measure of location for we can compare groups either by log-rank tests a distribution which, for some patients, might be or semiparametric Cox regression. In the table skew. Also the distribution of period means over below treatments are compared to placebo based patients may well be skew. on a Cox regression model: For symptom scores and rescue medication it is therefore often useful to compute the percentage of symptom-free days, or the percentage of days Hazard 95% Confidence with no rescue, instead of period means. At ratio limits p-Value least the former is often for mild patients a A 200 µgvs 0.7813 0.5642, 1.082 0.137 more efficient measure than the corresponding placebo period mean. And it is clinically easier to A 600 µgvs 0.6731 0.4812, 0.9414 0.0207 interpret. This idea can be carried one step placebo further: we can introduce the concept of an B 200 µgvs 0.717 0.5127, 1.003 0.0519 ‘asthma controlled day’, as one in which there are placebo 390 TEXTBOOK OF CLINICAL TRIALS
1.0
0.9
0.8
0.7
0.6
0.5
0.4 A 100 µg Fraction patients with no exacerbation Fraction A 600 µg 0.3 Placebo B 200 µg
0.2 0 102030405060708090 Days since randomisation
Figure 22.6. Kaplan–Meier plot of the time to a mild exacerbation
We see that the hazard for a mild exacerbation to the last one). This we call the MED (Minimal for the high dose of A, e.g., is estimated to be Effective Dose) for the patient. about 67% of that of placebo. As expected in this Such a study needs a detailed definition of what patient group, the results are, from a statistical is meant by asthma control. There is no such point of view, somewhat weaker than the ones universal definition, but lack of asthma control is obtained from the analysis of period means. essentially the same thing as a mild exacerbation, as discussed above. So the working definition A Dose Reduction Study above could be used, and the algorithm is to step down until a mild exacerbation occurs. To compare the efficacy of two GCSs, a ran- In such a study the diary card variables per se domised, double-blind parallel group study with are not of independent use, they should not be two treatment arms (one for each GCS) was compared between groups, except possibly the designed. The objective is to estimate the rela- data on the highest dose. Instead it is expected tive dose potency by starting each arm on a high that the mean values are similar in the two dose of the GCS, treating for some weeks, step- arms over the treatment period, what varies is ping down the dose and treating for some weeks, the underlying dose producing those effects. The making a further step down, etc. This is done until effect variable of interest is the MED, which is the asthma is no longer controlled. This way we to be compared between the groups. obtain for each patient the lowest dose on which The nature of MED is such that the best way of the patient had asthma control (the one previous analysing it is not immediate. On the one hand RESPIRATORY 391 it will be rather discrete in nature, with only a Moderate: The patient has breathlessness on few possible levels. On the other hand the most exertion and FEV1 in the range 40–60% of informative way of expressing the result is to predicted normal. say how much more was needed on average for Severe: The patient has breathlessness on every- one as compared to the other (like that the MED day activities and FEV1 < 40% of predicted for A was on average 125% that for B). We normal. advocate for that reason that MED is analysed under a multiplicative ANOVA model, i.e. after What is to be proved in a clinical programme for log transformation of the dose, provided that a COPD drug depends on what the claim of the MED > 0 in all cases. The most appropriate way drug is. We can crudely divide effects on COPD into two groups: symptomatic effects and disease to do this is to regard the data as interval censored modifying effects. The natural history of COPD data for the analysis. is one of an accelerated progressive decline One final comment on the design of long-term in lung function leading up to a distressful, asthma trials. In many instances, especially for premature, death. This decline in lung function dose–response studies, it is informative for the leads to progressive symptoms and diminished interpretation of the results to relate the observed exercise endurance. Symptomatic effects relate effects to the ‘highest effects attainable’. This to the alleviation of symptoms and improvement can be done within a study, so that the patients of quality of life, whereas disease modifying are put on a heavy treatment, consisting of a effects are effects that lessen the decline rate in high dose of a GCS and a long-acting β -agonist 2 lung function. (or whatever is considered necessary to get the A drug with symptomatic effects on COPD patient in as good a condition as possible), either should work on a rather short time-scale. It during run-in, in a period after a run-in period should lead to improved symptoms, fewer exacer- or by adding on a period at the end of the bations and better performance on exercise tests. study with a similar treatment. The purpose of Many drugs that were originally anti-asthma this is to be able to quantify the response in drugs have been tried, and licensed, for the terms of what can actually be achieved in the COPD indication. Part of their effect may well patients under study. If we put this reference be due to the reversible component that many period at the end of the study, we must make COPD patients have in their disease – in other certain that all patients, including withdrawals, words an anti-asthma effect within the COPD. In pass it in order to avoid having problems with order to claim effects above this, studies have a selection bias. If we put this reference period been performed in which one tries to exclude before randomisation, we might have carry-over patients with reversible components by using effects into the randomised treatments with their exclusion criteria on patients with a reversibility potential problems. But having such a period test above 15% of baseline. To claim that short- as reference often helps in the interpretation of term effects seen in the population are due to the results. non-anti-asthma effects because of such exclu- sion criteria is obviously not correct – a patient COPD Studies with reversible airways obstruction can well have a reversibility of less than 15% on a particular For COPD the classification of disease severity occasion. Since COPD is a disease affecting small can be made as follows:34 airways, it seems more logical to base short- term effects on measurements related to these Mild: This is what is called the smoker’s cough. airways, as opposed to FEV1. However, regula- The patient has FEV1 > 60% of predicted tory requirements make FEV1 the primary effi- normal, no breathlessness and is in general cacy variable in COPD studies – at least as of unknown to the health care system. this writing – though emphasis is made on the 392 TEXTBOOK OF CLINICAL TRIALS symptoms and/or exercise tests also. In fact the intensity of the rhinitis is dependent on pollen CPMP guidelines require two primary efficacy counts in the air, and lack of treatment effects variables in COPD studies – one should be FEV1 can well be due to insufficient pollen exposure. and the other a symptom score.35 Therefore concomitant collection of pollen data is A COPD study for a drug with primarily not only useful but almost necessary when trying symptomatic effects is in general a long-term to understand lack of effect in such studies. study, probably with diary cards, not very different from the asthma trial discussed above. THERAPEUTIC EQUIVALENCE Prevention of exacerbations is perhaps the most One of the challenges for drug development is important aspect of COPD treatment, so a 6- to prove that a new treatment is therapeutically month study is the minimum. equivalent to a reference treatment. In the area of A COPD drug which claims disease modifying respiratory diseases this problem appears in two properties has a heavier burden of proof on different settings – when we want to register a it. The effect of disease modifying is that the new formulation, most often a new inhaler, and rate of decline in lung function is reduced. To in market claims of equality of two treatments. prove this, we need to do long-term studies The background and motivation of these differ over 3–5 years, or perhaps more, in which lung somewhat, so we discuss them separately. function is measured repeatedly. The statistical analysis should focus on the rate of decline, Bioequivalency of Two Devices which could be done using a linear mixed Bioequivalency refers to a specific problem. effects model. Assume that a drug is delivered as a tablet or in Rhinitis Trials some other form, such that it must pass through the bloodstream before reaching its site of action. Classification of rhinitis patients into groups Then the plasma concentration profiles of the according to severity is lacking. The accepted drug define the clinical effect. This reasoning division is between occasional and continuous is the rationale for the bioequivalence concept: expression of symptoms, i.e. between seasonal to prove that two formulations are bioequiva- allergic rhinitis and perennial rhinitis. The rhinitis lent, show that the plasma concentration profiles symptoms are the same, so the measurements, are sufficiently similar. From that we could then notably symptom scores, are the same. As already logically infer that the therapeutic effects are suf- indicated symptoms are often recorded in diary ficiently similar to have the same therapeutic cards for blockage, runny nose, sneezing/itchy effect. This is in general a rather straightfor- and eye symptoms, and the sum of the first three ward problem, requiring only small pharmacoki- make up the Combined Nasal Symptom score netic studies. or Nasal Index Score, which is a useful primary For many years there has been a well-defined variable for clinical trials. method to establish bioequivalency in this situ- The difference between seasonal and perennial ation. We reduce the general question of similar rhinitis lies more in the study design/conduct. For plasma concentration curves to key measures of perennial rhinitis the situation is similar to that rate and extent of absorption, including AUC. for asthma or COPD in that the patient can start The ratio of two AUCs measures the relative the trial almost any time. For hay fever, however, bioavailability of the two formulations and the the study must be conducted over a rather short requirement is that the ratio of the means (anal- period of pollen exposure. What makes these ysed under a multiplicative model) should have trials more difficult is that ideally the patients confidence limits within 80–125%. should be included during the onset of the pollen For inhaled respiratory products, however, the season to get baseline data, then followed over site of action, the lungs, lie prior to the blood- one to several pollen peaks with treatment. The stream (i.e. when the drug appears in the blood RESPIRATORY 393 it has in general had its desired pharmacolog- of each inhaler, in order to see not only that the ical effect). Thus plasma concentrations cannot response is similar, but also that there is a similar predict effect by pure logic! Equal delivered sensitivity to changes in dose. These are, rela- dose does not by logic imply the same effect tively speaking, simple studies to perform. for different inhalers, because different inhalers The next question is, which should the deci- could deposit the drug in different parts of the sion rule be for bioequivalency for such a lungs. For that reason, to bridge from one inhala- study – when are two inhalers considered to be tion device to another is not necessarily a sim- similar? There seems to be two approaches used ple case of measuring plasma concentrations. As over the last few years. One is to use the word of this writing there is substantial confusion on comparability. This is, for good reasons, not well how to proceed with bioequivalence studies for defined and essentially means that there is a inhalers. We will discuss some aspects of the dose–response relationship on each device and problem here. that, numerically, the mean on each dose level is The first aspect is that what you inhale are similar in the eye of the regulator. Thus there is particles, and these will be deposited differently no true statistical decision plan associated with depending on size. To give an equivalent effect, the study and it is not used as proof per se, we therefore need equivalent in vitro performance only as supportive information to what in vitro of the two inhalers. For the rest of the discussion and PK data provides. The other approach is we assume that in vitro data is similar for the strictly statistical. In this case we should com- two inhalers. pute the relative dose potency with confidence The basic assumption is that since what limits. At present, in the case of bioequivalency appears in the bloodstream does not have to for pMDIs for albuterol (salbutamol in the US), have passed the site of action, systemic exposure, FDA requires that the 90% confidence limits for as measured by drug concentrations in the this parameter should be contained in the inter- circulation, is not necessarily enough to conclude val 2/3–1.5. The justification for these limits is efficacy. However, similar systemic exposure not clear, but they imply that the mean effect is should be sufficient to deduce similar systemic so similar for the two pMDIs that they could be effects, and therefore reduce much of the question switchedonthemarket. of bioequivalent side-effects to the classical When it comes to decide bioequivalency bioequivalence method. for orally inhaled anti-inflammatory drugs the Logically, if we measure blood concentrations problem is even more difficult, since there are, as and conclude that the systemic exposure is the of today, no designs available that can provide the same, the outstanding question is if the drug has kind of answers discussed for bronchodilators, to been delivered to the site of action. For a nasal a reasonable price. The problem is that in most spray it is hard to see how it can fail to do so. studies the dose–response appears rather flat, so For a nasal spray, therefore, to require anything very large studies are needed for the type of beyond in vitro data and pharmacokinetic data decision plan that was indicated above. might be an overkill. For an orally inhaled drug the situation is more complicated. Marketing Therapeutic Equivalence If we consider a bronchodilator as an example, we need the drug to hit the receptors of the The other aspect of therapeutic equivalence is contracted muscles. To check that the drug has to show that a new treatment is as effective as hit these, we can therefore do a pharmacody- an old one, whereas it has some other benefits namic study, e.g. a single dose study in which compared to it. FEV1 is followed for a number of hours, or a Proving that two treatments are equivalent bronchoprovocation study if that is preferred. A has, however, a long history in the context of suggested design is to study two or three doses medicine. The traditional way was to misuse the 394 TEXTBOOK OF CLINICAL TRIALS p-value technology – if we could not demonstrate dose of A that has the same effect as treatment a difference (p > 5%) the treatments are equal. B lies between a/2 and 2a it is reasonable that This is obviously wrong (it is similar to ‘not dose a of A is equivalent from a clinical point proven guilty’ in court, which means only that, of view to treatment B from an efficacy point not that one is proven innocent), which is by of view. This is because half the dose has less now acknowledged by most, but not all, work- effect and twice the dose more effect. Implic- ers in the field. A theoretically valid approach itly this assumes that the clinical response to a became legitimised by the ICH guidelines,29 suboptimal dose is to double this, which is what which defines an algorithm borrowed from the is done with most drugs in the respiratory area. original bioequivalency concept for plasma con- In general, to draw the conclusion of therapeutic centrations. First you prespecify some limits (cor- equivalence, large studies might be needed.36 responding to the 80–125% above) and if your Often one tries to establish the therapeutic 95% (sic!) confidence interval for the mean dif- equivalence in one clinical study, and compare ference is contained within this prespecified inter- benefits in the other. Basically I do not think this val, you can declare therapeutic equivalence. This is the way to go about this kind of problem – in approach is however complicated, when your most cases it is probably a problem that should effect scale has no obvious interpretation, like a be discussed in terms of therapeutic ratio, as lung function scale or a symptom scale. To be discussed in the next section. a sensible approach, the prespecified limits must imply that clinicians do consider such a small The Real Issue – The Therapeutic Ratio difference to be of no clinical consequence – the Dose–response studies provide us, at best, with predetermined limits must be agreed on. It is hard doses for further investigation. However, whether to forsee that this can actually be done in the a drug ends up being superior or inferior to field of respiratory medicine, since many effect what is on the market is not determined by what changes mean different things depending on what dose it is given in. What is the point in halving population is studied. the nominal dose, if you get twice as much There is however a sensible, though expensive, adverse effects? approach available – the one used for demon- The appropriate measure here is the therapeutic strating bioequivalency of bronchodilators. The ratio. The therapeutic ratio for a drug relates key there is to translate from the effect scale the positive effect to the negative effect. To to the dose scale, by studying dose–response. understand it, assume that effects, both positive Almost all drugs in the respiratory area (though and negative, are measured on a scale from there are exceptions) are available in multi- 0 to 100. Also assume, for the time being, ple doses, and when two treatments (drug plus that the dose–response curves for positive and inhaler) are to be compared, at least one of negative effects are parallel. Then a therapeutic them can in general be varied on some kind of ratio for this drug of 2 means that twice as dose scale. large a dose is needed to get the same negative The simplest such design is as follows. To effect as positive effect. Thus we can define prove that dose a of a treatment A is therapeuti- the therapeutic ratio for drug A as TR(A)= cally equivalent to a treatment B (dose need not ED50(side-effect)/ED50(effect). If we have two be specified), we can study doses a/2 and 2a of A drugs, we can define the relative therapeutic (not dose a!) and treatment B. By assuming a lin- index of A to B as TR = T R(A)/T R(B),which ear dose–response relationship (versus log-dose) is equivalent to the ratio ρ(effect)/ρ(side-effect), for A, we can estimate the dose of A that has the where ρ is the relative dose potency for the same effect as treatment B. Illustrations of this, two drugs with respect to the indicated effect. with more than two doses of A were illustrated This means that we can estimate not only the earlier. Now, if the 90% confidence limit for the relative therapeutic index, but also confidence RESPIRATORY 395 intervals to the estimate, by assessing the relative means together with straight line approximations dose potencies. Moreover, we can define the to the dose–response curves for each variable. therapeutic index as a ratio of relative dose In order to make the results more interpretable potencies without assuming that the effect and the mean values (and lines) are expressed as a side-effect curves are parallel. To be meaningful percentage of the mean value for placebo. From it only requires that the dose–response curves these straight lines we can estimate the relative for efficacy are parallel and the dose–response dose potency, as discussed earlier, for each curves for the side-effect are parallel. We can variable separately: estimate the therapeutic index by combining effects from two studies – one on efficacy and 95% Confidence one on side-effect, but better still is to obtain all Variable ρ limits the information in one study. The first problem to solve is the precise FEV1 147 65, 534 definition of outcome variables, both positive and S-potassium 60 41, 91 negative. Different results can be obtained by using different outcome variables. It is therefore Thus we see that in terms of efficacy, the long- important that the precise objective is spelt out acting drug A is almost 150 times more efficient and the outcome variables related to this. The than the short-acting B. On the side-effect side, following example is an illustration. A is 60 times more potent, so from this data we see that the relative therapeutic index is estimated Example: Estimating the Relative Therapeutic to be 146.6/59.75 = 2.4. To obtain confidence Index limits is somewhat involved since we need to take In order to assess the relative usefulness of a into account the covariation of the two variables. 38 long-acting β2-agonist, call it A, and a short- How to do this is outlined in Ref. The result acting one, which we call B, we want to compare is that the relative therapeutic index is estimated one topical effect, bronchodilation, and one to be 2.5 with (approximative) 95% confidence systemic effect, suppression of serum potassium. limits 1.02 and 9.0 (p = 0.046). The study was of crossover design with single- The conclusion from this is, that in terms of the dose administrations and serial measurement of variables of this analysis treatment A is estimated both variables taken.37 In order to be able to get to be 2.5 times ‘better’, but it is certified that it meaningful estimates of the relative therapeutic is ‘better’ than treatment B. Of course, this result index we need many patients, relatively speaking, is not better than data. From Figure 22.7 we see and in order to obtain a simpler study we choose that the lowest dose of each drug has a very small to measure the maximal effect on each parameter average effect on serum potassium. It could (and and compare that. should) be questioned if these really are on the Thus a randomised, double-blind, six-period log-linear part of the dose–response curve. We crossover study was designed with the following can repeat the analysis by only incorporating the single dose treatments: placebo, 6 µg, 24 µg doses 24 and 72 µgofA and 1800 µgofB for and 72 µgofdrugA and 200 µg and 1800 µg serum potassium. of drug B. Each treatment period consisted of a single dose administration which was OTHER ISSUES followed for 4 hours and from each experimental sequence the maximal FEV value and the 1 PHASE IV minimal S-potassium value was extracted for statistical analysis. Much phase IV work focuses on comparisons Figure 22.7 demonstrates the main result. We with competitor products in order to demonstrate see (period and baseline adjusted) treatment the advantages of the new product. This has been 396 TEXTBOOK OF CLINICAL TRIALS
124 100
98
120 96
94 116
92 (percent of placebo) 1 112 90
maximal FEV 88
108 (percent of placebo) minimal S–Potassium
86
104 84 100 101 102 103 100 101 102 103 Dose Dose
Figure 22.7. Adjusted mean values for each treatment and outcome variable discussed in previous chapters and will not be devices, consultations with physicians, emer- repeated here. In addition to that, special safety gency room visits and hospitalisations. issues might have to be addressed, which might 2. Indirect costs, defined as lost productivity, call for large-scale studies in order to study some include time off work or school, either patient rare event. or relative, premature retirement and death. Indirect costs may account for up to 50% of PHARMACO-ECONOMICS the cost of asthma. Asthma is a costly illness – its costs can account 3. Intangible costs, contain factors related to for as much as 1–1.5% of all resources in quality of life (grief, fear, unhappiness, the health care sector. The costs are, however, pain, etc.). unevenly distributed among the asthmatics; it is not uncommon that 10% of the most severe Within the direct costs, drugs and general prac- cases (usually patients with uncontrolled asthma) titioner visits can, crudely, be considered costs account for over 50% of the total cost. There of managing controlled asthma, whereas emer- should thus be room for a significant cost gency room and hospital costs can be assumed reduction by improving disease control. to relate to treatment failure. Assuming this to The costs can, basically, be divided into be about 75% of the total cost, the major part three categories (usually only the first, although of the costs of asthma appear to be a result of sometimes also the second category is measured): inadequately controlled disease. Thus, the goal is to get control of the asthma, which (for example) 1. Direct costs, defined as health care resources can be done by patient education39 and prophy- consumed, include costs associated with drugs, lactic therapy. As an example of the latter, it RESPIRATORY 397 has been demonstrated that the introduction of support in answering the question on whether high-dose inhaled steroids in patients with severe the additional effectiveness (or quality of life or asthma reduced the number of days of hospital- asthma control) can justify the extra costs. isation by 80%.40 It should also be noted that Furthermore, it is important to keep in mind a part of the drug costs consists of rescue med- that costs often are country-specific, e.g. in the ication, like short-acting β2-agonists, and is in case of an international clinical trial, and some itself a sign of the disease not being adequately adaptation is needed before translating results under control, and that an increase in prophylac- from one country to another. This is so because tic therapy, like inhaled steroids, might decrease the outcome of a new treatment will depend on these costs. the local medical tradition, drug pricing and the So, the economics of asthma inform us that unit costs of other health care resources, and a large population of mild-to-moderate asthmat- a number of social conditions like the labour ics has a low daily cost. For some of these market. In general, though, as the interest for patients, the disease becomes uncontrolled and ‘value for money’ and cost-containment in the costly, with hospitalisations and time off work or health care sector grows, the importance of these school, possibly progressing into early retirement kind of evaluations is likely to increase. From or death. Some of these cases are probably due to clinical trials it is thus important to measure these bad compliance with treatment regimens. Interna- kind of variables (both costs and effectiveness), tional guidelines therefore stress that prophylactic which then can be used as the basis for a treatment should be introduced at an early stage cost–effectiveness analysis. in asthma treatment, resulting in an increase in drug and general practitioner costs, but hopefully META-ANALYSIS leading to reduced hospitalisation costs and indi- Meta-analysis is in general useful when the focus rect costs. As an observation on this topic, it can is on an issue for which data is available in be mentioned that the cost of the avoidance of one a number of studies, but each individual study admission to hospital will pay for about 3 years is underpowered for that particular issue. Such of treatment with inhaled steroids.41 issues might be questions related to special Apart from collecting data on costs, it is also subgroups or issues of rare events. Obvious important to measure the individual’s quality examples of the latter are special, and rare, of life and the effectiveness of various inter- adverse events. In this respect the respiratory area ventions (where effectiveness, as compared to does not differ much from other areas. The results efficacy, ideally should measure the effects of obtained so far within the respiratory area from an intervention in clinical practice and also in meta-analysis are not impressive.42 units the patients care about). This is perhaps especially important if a new medical treatment increases the total costs, as it then becomes REFERENCES important to relate these extra costs to the additional effectiveness gained. Currently recom- 1. Beasley R. Worldwide variation in the prevalence mended effectiveness variables include the num- of symptoms of asthma, allergic rhinoconjunc- tivitis and atopic eczema. Lancet (1998) 351: ber of symptom- and episode-free days. In cost- 1225–32. effectiveness analyses, if both costs and effec- 2. Chadwick DJ, Cardew G. The Rising Trends in tiveness are higher for one of the treatments, the Asthma, Ciba Foundation. Chichester: John Wiley difference in costs is divided by the difference in & Sons (1997). effectiveness to obtain a cost–effectiveness ratio. 3. Peat J, Li J. Reversing the trend: reducing the prevalence of asthma. J Allergy Clin Immunol This ratio is, for example, expressed as: com- (1999) 103: 1–10. pared to treatment A, treatment B costs $x per 4. Sears MR. Epidemiology of childhood asthma. symptom-free day gained. That ratio thus gives Lancet (1997) 350: 1015–20. 398 TEXTBOOK OF CLINICAL TRIALS
5. Barnes PJ. The changing face of asthma. QJMed 20. Barnes NC. Effects of antileukotrienes in the (1987) 63: 359–65. treatment of asthma. Am J Respir Crit Care Med 6. Lange P, Parner J, Vestbo J, Schnohr P, Jensen G. (2000) 161: S73–6. A 15-year follow-up study of ventilatory function 21. ATS guidelines. in adults with asthma. N Engl J Med (1998) 339: 22. Quanjer PH, Tammeling GJ, Cotes JE, Peder- 1194–200. sen OF, Peslin R, Yernault JC. TI: lung volumes 7. Sherman CB. The health consequences of cigarette and forced ventilatory flows. Report Working smoking. Pulmonary diseases. Med Clin N Am Party: Standardization of Lung Function Tests. (1992) 2: 355–75. European Community for Steel and Coal. Eur 8. Buist SA. Smoking and other risk factors. In: Mur- Respir J (1993) 6(Suppl 16): 5–40. ray JF, Nadel JA, eds. Textbook of Respiratory 23. Oga T, Nishimura K, Tsukino M, Hajiro T, Ikeda Medicine, 2nd edn. Philadelphia: WB Saunders A, Izumi T. The effects of oxitropium bromide (1994) 1259–87. on exercise performance in patients with stable 9. Selroos O, Pietinalho A, Riska H. Delivery de- chronic obstructive pulmonary disease. A compar- vices for inhaled asthma medication Clinical ison of three different exercise tests. Am J Respir implications of differences in effectiveness. Clin Crit Care Med (2000) 161: 1897–901. Immunother (1996) 6: 273–99. 24. Borg GAV. Psychophysical bases of perceived 10. Selroos O. Bronchial asthma, chronic bronchi- exertion. Med Sci Sports Exercise (1982) 14: tis and pulmonary parenchymal diseases. In: 377–81. Moren´ F, Dolovich MB, Newhouse MT, New- 25. Mahler DA, Wells CK, Feinstein AR. The mea- man SP, eds. Aerosols in Medicine. Principles, surement of dyspnea. Contents, interobserver Diagnosis and Therapy, 2nd edn. Amsterdam: agreement, and physiologic correlates of two new Elsevier Science (1993) 261–89. clinical indexes. Chest (1984) 85: 751–8. 11. Bartnes PJ. Beta-adrenergic receptors and their 26. Berger M, et al. The Sickness Impact Profile: regulation. Am J Respir Crit Care Med (1995) 152: development and final revision of a health status 838–60. measure. Med Care (1981) 19: 787–805. 12. Lofdahl¨ CG, Chung KF. Long-acting β2-adreno- 27. Stewart AL, Hays RD, Ware JE Jr. The MOS ceptor agonists: a new perspective in the treatment short-form general health survey. Reliability and of asthma. Eur Respir J (1991) 4: 218–24. validity in a patient population. Med Care (1988) 13. Rall TW. Drugs used in the treatment of asthma: 26: 724–35. the methylxanthines, cromolyn sodium, and other 28. Jones PW, Quirk FH, Baveystock CM. The St agents. In: Gilman AG, Rall TW, Nies AS, Tay- George’s Resp questionnaire. Resp Med (1991) 85: lor P, eds. Goodman and Gilman’s The Pharmaco- 25–31. logical Basis of Therapeutics, 8th edn. New York: 29. ICH Harmonised Tripartite Guideline. Statistical Pergamon Press (1990) 618–37. principles in clinical trials. Stat Med (1999) 18: 14. Gandevia B. Historical review of the use of 1905–42. parasympatholytic agents in the treatment of 30. Senn S. Statistical Issues in Drug Development. respiratory disorders. Postgrad Med J (1975) Statistics in Practice. Chichester: John Wiley & 51(Suppl 7): 13–20. Sons (1997). 15. Brattsand R, Selroos O. Glucocorticosteroids. In: 31. Kall¨ en´ A, Larsson P. Dose response studies: how Page CP, Metzger WJ, eds. Drugs and the Lung. do we make them conclusive? Stat Med (1999) 18: New York: Raven Press (1994) 101–220. 629–41. 16. The British Guidelines on asthma management. 32. Global Initiative for Asthma. National Institutes of Thorax (1997) 52(Suppl 1): S1–21. Health. National Heart, Lung, and Blood Institute 17. Howarth PH. The medical treatment of chronic Publication 95-3659 (1995). rhinitis. In: Busse WW, Holgate ST, eds. Asthma 33. Woolcock AJ. Therapies to control the airway and Rhinitis. Boston: Blackwell Scientific (1995) inflammation of asthma. Eur J Respir Dis (1986) 1415–28. 69(Suppl 147): 166–74. 18. Simons FER. Antihistamines. In: Mygind N, Nac- 34. BTS Guidelines for the management of chronic lerio R, eds. Allergic and Non-allergic Rhini- obstructive pulmonary disease. Thorax (1997) tis. Clinical Aspects. Copenhagen: Munksgaard 52(Suppl 5): S1–28. (1993) 123–36. 35. Committee for proprietary medicinal products 19. Bernstein JA, Bernstein IL. Cromones. In: Barnes (CPMP). Points to consider on clinical investi- PJ, Grunstein MM, Leff AR, Woolcock AJ, eds. gation of medicinal products in the treatment of Asthma. Philadelphia: Lippincott-Raven (1997) patients with chronic obstructive pulmonary dis- 1647–66. ease (COPD). London: EMEA (1998). RESPIRATORY 399
36. Kall¨ en´ A, Larsson P. On the definition of ther- 39. Liljas B, Lahdensuo A. Is asthma self-manage- apeutic equivalence. Drug Inform J (2000) 34: ment cost-effective? Patient Educ Counsel (1997) 349–54. 33: 97–104. 37. Rott Z, Bocskei¨ C, Poszi M, Juhasz G, Larsson P, 40. Adelroth E, Thompson S. Advantages of high- Rosenborg J. On the relative therapeutic index dose inhaled budesonide. Lancet (1988) i: 476. between formoterol Turbuhaler and salbutamol 41. Blainey D, Lomas D, Beale A, Partridge M. The pressurized metered dose (pMDI) inhaler in asth- cost of acute asthma: how much is preventable? matic patients. EurRespirJ(1998) 12(Suppl 28): Health Trends (1991) 22: 151–3. 324s. 42. Jadad RJ, Moher M, Brownman GP, Booker L, 38. Kall¨ en´ A. A note on therapeutic equivalence and Sigouin C, Fuentes M, Stevens R. Systemic re- therapeutic ratio with application to studies in views and meta-analyses on treatment of asthma: respiratory diseases. Drug Inform J (2001) 35: critical evaluations. BMJ (2000) 320: 537–40. 1495–505. Index
Note: page numbers in italics refer to figures and tables and boxes
α-blockers 179 GM-CSF 145 reporting acetylcholinesterase (AChE) induction regimens 144–5 in Chinese medicine inhibitors 244 Kaplan–Meier survival curves 71, 73 acne 213–14 146 in gastrointestinal cancer disability index 214 length of remission 145–6 trials 132 outcome measures 219 maintenance regimens 144–5 age limit, upper for trials 56 treatment 214 molecular characterisation 142 air pollution 361 acupuncture 80–1 older patients 143–4, 146 airway cohort studies 81 outcome measures 145–6 diseases 359–73 humoral theory 81 overall survival 146 narrowing 360 neurological theory 81 post-remission therapy 142, upper, function tests 366–7 pain control 68 143, 144, 145 airway resistance 365 placebo 80–1 QoL 146 measurement 366–7 acute lymphoblastic leukaemia response rate 145 albuterol (salbutamol) 393 (ALL) statistical issues in allergic disease childhood 103, 104, 112–13 design/analysis of trials atopic dermatitis 214 translational research 105 144–6 respiratory 360 acute myeloid leukaemia (AML) subtypes 142 rhinitis 361, 370–1 141–6 treatment failure 142 allocation to treatment 24–5 competing risks 146 treatment phases 144 cognitive behavioural therapy complete response (CR) rate acute progranulocytic leukaemia 287, 288 144, 145 (APL) 142–3 contraception trials 327 cure models 146 adaptive design of trials 95–7 depression 300 cytogenetic characterisation adolescents 50 dynamic procedure 25 142 cancer trials 103 gynaecology 350 disease-free survival 145, 146 adverse events concealment 351 drug-resistant 142, 143 herbal medicine 70–1, 72–3, infertility 341, 350 event-free survival 145, 146 73–4 concealment 351 factorial design of trials monitoring in paediatric clinical random 14 144–5 trials 49–50 alopecia androgenetica 226
Textbook of Clinical Trials. Edited by D. Machin, S. Day and S. Green 2004 John Wiley & Sons, Ltd ISBN: 0-471-98787-5 402 INDEX
Alzheimer’s disease 243–53 anticholinergic drugs 363 outcome aetiology 246 anticipated effect size 30 assessment 266–7 amyloid 243, 244 antidepressants, tricyclic 239 definition 264–6 deposition 246 antihistamines 364 pathological state 262 apo E4 allele 246, 250 Antihypertensive and phobic symptoms 267–8 apolipoprotein E 244 Lipid-Lowering Treatment to prevalence 259 bridging studies 247–8 Prevent Heart Attack Trial primary care setting 258 cholinergic marker loss 243 (ALLHAT) 178–9 recruiting individuals not decline slowing 248, 249, 251 antihypertensive drugs seeking treatment 257 diagnosis 246 176, 180 research recruitment 260 endpoints 247 antimicrobial agents in toothpaste result measurement 264–6 epidemiology 245 205 self-criticism 258 ethics 252–3 antioxidant vitamins, colorectal severity 257 familial clustering 245 cancer prevention 122 composite measures 265 genetics 245, 246 anxiety disorders 255–70 social support 263 mild cognitive impairment avoidance 267–8 specialty clinics 259 studies 250 Bayes factor of test 259 stigma 258, 259 N-of-one design 252 co-occurring symptoms pathogenesis 246 symptoms/syndromes complexity 264–5 pathology 245–6 257, 264 elimination 262 patients 246–7 community recruitment 259 severity scales 255 Phase I trials 247–8 community settings 256–7, silencing 267 Phase II trials 248 258, 260, 266 stability 266–7 Phase III trials 248–50 assessment strategies treatment placebo-controlled trials 259–60 background history 261 252–3 efficacious medication 268 cognitive behavioural 255, prevalence 245, 250 comorbidity 260–1 256, 268 primary prevention trials data capture/edit cycle 260 decisions 262–3 250–1 diagnostic category definitions efficacious 268 randomised start/withdrawal 263–4 medication 268–9 studies 252 diagnostic groups 264 protocol-driven 258 rate of change studies 249 diagnostic instruments 255 protocols 260 secondary prevention 251 discordant models 268–70 psychosocial 268–9 structural change 251–2 effectiveness studies 255–6 apo E4 allele 246, 250 survival analysis 247, 249–50 efficacy trials 257, 264 apolipoprotein E 244 symptomatic change 251–2 escalation strategies 260 area under the curve (AUC) 26 synapse loss 246 exclusion criteria 261 bioequivalency 392 tertiary prevention 251 features 257 glucocorticosteroids 385 trials 56, 246–51 full factorial study design Aricept trial 251 Alzheimer’s Disease Assessment 268–9 aromatase inhibitors 89 Scale–Cognitive Component generalised 263–4 asbestos exposure 159 (ADAS-Cog) 248 heterogeneity/homogeneity aspirin, cardiovascular disease Alzheimer’s Disease Cooperative trade-off 261 15, 16, 175 Study (ADCS) 244 high community prevalence assent to participate, paediatric amalgam 204 256–61 clinical trials 52 anastrozole 15 illness severity 261 asthma 359–60 aneurysm resection 181 individual psychology anticholinergic drugs 363 angiotensin converting enzyme 263–4 β2-agonists 362 (ACE) inhibitors 179 life context 263–4 challenge tests 368 anthracycline therapy 89, 90 measurement targets 266 control 390–1, 396–7 AML 142 multi-step diagnostic controlled days 389 antiangiogenesis agents 167 procedures 260 cost–effectiveness ratio of antiarrhythmic drugs 180 normal state 262 treatment 397 Antiarrhythmics Versus nosological boundary definition crossover studies 381 Implantable Defibrillators 261–8 data pre-processing 388 (AVID) trial 181, 182, 183, optimal long-term outcome diary card studies 371–3, 184 262 387–90 INDEX 403
disease severity 379 β-blockers 180 continual reassessment method dose–response curves 369, β2-agonists 362, 380 (CRM) 97 381–2 bronchodilatation 382–3 curing 94 dose–response study 381–2, relative therapeutic index 395, dose-finding trials 97 391 396 doxorubicin 90 duration of action 367 systemic effects 383–4 HER-2 89 effectiveness variables 397 baseline comparisons 38 HER-2/neu expression 93–4 efficacy studies 380–3, baseline information 15 hormonal therapy 90 386–91 Bayesian methods 24 incidence 87 experimental trials 367–70 Beck Depression Inventory (BDI) long-term therapy impact glucocorticosteroids 363, 371, 298, 302 94–5 381 benoxaprofen (Opren) 55 mammography 87, 90–1 meta-analysis 91 diary card studies 387–90 benzodiazepines 239 new agent trials 97 dose reduction study 390–1 betamethasone 217 oestrogen-receptor (ER) status level of use 387 bias 6 clinician 339 89, 90, 91 highest effects attainable 391 ovarian ablation 91 hyperresponsiveness studies patient 339 recruitment 281–3 patient advocates 87–8 368–70 polychemotherapy 91 Kaplan–Meier survival curves respiratory disease studies 373–5, 377 prevention 91 389, 390 prognosis 88–9 minimal effective dose (MED) bioequivalency 392–3 blinding 6 prolonging survival 94 381, 390–1 quality of life 91 cardiovascular device trials missing data 388 radiotherapy 89, 91 184 onset of action 367 raloxifene 91 cognitive behavioural therapy prophylactic treatment 397 recurrence 92, 93 studies 287–8 QoL 373 research promotion 87–8 contraception trials 327 reference period 391 staging 88 dentistry 197 rescue medication 375, 389 stopping rules 95–6 explanatory trials 339 responders 367, 373 surgery 89 gynaecology studies 351–2 response curve 367 tamoxifen 15, 89, 90, 91 infertility studies 351–2 time-dependent hazards 91–4 safety studies 386–91 pragmatic trials 339 sample size parameters 378–9 treatment benefits 88 respiratory diseases 373–4 trial groups 91 severity classification 386–7 rhinitis studies 374 single dose monitoring 367–8 breastfeeding, contraception 321 skin disorders 216–17 breathlessness measurements 370 subgroup analysis 379–80 blood taking, paediatric clinical symptom scores 389 bridges, dental 193 trials 49 resin-bonded 204–5 asthmatic reactions, early/late Boccaceto 3 bronchial carcinoma 159 368 body plethysmography 365 bronchial hyperresponsiveness atenolol 180 Bonferroni correction 37 non-specific 360 atherosclerosis 175, 176, 177 Bowman–Birk inhibitor 123 studies 368–70 atopy 216 brain tumour, childhood 105 bronchitis, chronic 360, 361 atorvastatin 18 breast cancer 87–97 bronchodilator drugs 362–3 atrial fibrillation, rate control 181 adaptive designs of trials bioequivalency 393 Atrial Fibrillation Follow-up 95–7 efficacy studies 380–3 Investigation of Rhythm adjuvant therapy 90 bronchoprovocation studies 393 Management (AFFIRM) aggressive 92 burns, severe 14 180–1 anastrozole 15 Butler, Robert 244 atrioventricular junction ablation anthracycline therapy 89, 90 180–1 CAF treatment regime 92–4 χ 2 test 34 audit trail facilities 40 chemotherapy 89–90 Cade, John 239 auditory hallucinations, perceived high-dose with bone marrow calcipotriol 216 powerfulness 290–1 transplant/stem cell calcium, colorectal cancer average causal effect (ACE) 299, support 90 prevention 122–3 307–8 low-dose versus high dose calcium channel blockers 179, avoidance behaviours 267–8 92–4 180 404 INDEX
Canadian Implantable Defibrillator behaviour change trials 184–7 endpoints 74 Study (CIDS) 182 causes 175, 176 herbal drugs 68–71, 72, 73–4 cancer community settings for trials holistic approach 69 adaptive design of trials 96–7 186 Naranjo’s grading of adverse aggressive treatment 34 congenital abnormality effects 71 chemoprevention 122–3 correction 181 old system of clinical distant metastases 38 control group observation 70 local recurrence 38 active 178–9 quality of life 73–4 solid tumours, retrospective treatment 178 recommendations for clinical review 29 depression 186 trials 74–5, 76 –9, 80–1 see also named cancers genetic factors 176 response to pathological cancer, childhood 101–13 left ventricular ejection fraction processes 70 accrual duration in trials 102 179, 183 strong tradition 70 ancillary studies 111 multifactorial 176 symptom identification categories 101, 102 multiple interventions 176 principle 69 clinical research problems 112 oral contraceptives 317–18 traditional healing concepts control 103 pharmacological agent trials 69–70 cure rate 101 177–81 unique features 74 doxorubicin 112–13 presentation 176 chlorhexidine 205 ethics 111–12 primary prevention 175, 178 chlorthalidone 180 incidence 101 primordial prevention 175 Cholesterol and Recurrent Events minimal residual disease 106 rhythm control 180–1 (CARE) trial 178 national reference laboratories secondary prevention 175 cholesterol levels 175–6 106 social support 186 lowering 178, 182 negative questions 111 stepped care approach 180 cholestyramine resin 178 outcome prediction 106 strategy comparisons 180–1 cholinesterase inhibitors 244, prognostic factors 105–6 stratification of trials 177 252 reduced therapy 111 surrogate outcomes 185 chronic myeloid leukaemia (CML) risk assignment 106 ventricular arrhythmias 180 143 risk groups 106 cardioversion 180, 184 chronic obstructive pulmonary solid tumour treatment 105 care settings, older people trials study design 108, 110–11 59–60 disease (COPD) 359, survival improvement 103 causal effects 306–8 360–1 translational research 103, complier average 307, 308, anticholinergic drugs 363 105, 106, 111 309 β2-agonists 362 treatments 101–2 causality, counterfactual model diary card studies 371–3, 392 advances 103, 104 298 disease modifying effects 391, effects 112 cause-specific cumulative 392 success 104 incidence functions 38 exacerbation rates 379 Cancer and Leukemia Group B CAV (cyclophosphamide, exercise tests 370 (CALGB), breast cancer trial adriamycin and vincristine) measurement scales 364–7 91, 92–4 165 QoL 373 Cardiac Arrhythmia Suppression celecoxib 123 rescue medication 375 Trial (CAST) 180 cervical cap/sponge 321 reversibility 391 cardiac defibrillators, implantable children 50 severity classification 391 182 informed consent 111 symptomatic benefit 372, cardiovascular device trials see also cancer, childhood; 391–2 181–4 paediatric clinical trials cigarette smoking 159 blinding 184 Children’s Health Act (2000, US) cessation 159 risk levels of patients 183 112 COPD 361 strategy approach 184 Children’s Oncology Group habit change 185 cardiovascular disease 175–87 (COG) 102–3 prevention of uptake 159 adherence to trial treatments Chinese medicine 64–5 pustular psoriasis 214 177 adverse effects 70–1, 72–3, cisplatin aspirin 15, 16, 175 73–4 non-small-cell lung cancer atrial fibrillation rate control alopecia androgenetica 226 164 181 disease management 66 small-cell lung cancer 165 INDEX 405 clinical observation, old system gender 286 colorectal cancer 121–9 (herbal medicine) 70 intellectual status 286 chemoprevention 122–3 clinical record forms (CRFs) 39 IQ cut-off 284 early detection 123–4 data transfer 40 lost to follow-up 284–5 combined oral contraceptives Clinical Research Organisations mean effect size 278–9 (COCs) 316–18 (CROs) 39 multiple effects 291 common cold vaccine 6 clinical trials number needed to treat (NNT) COMPACT (Computer Package alternative designs 17–24 292 for Clinical Trials) 40 designs 339–40 outcomes of treatment 276, complementary medicine 63–82 types 12–24, 337–9 281, 291–2 indications 65–6 Clinician’s Global Impression clinical significance 292 philosophy 65–6 (CGI) 248 prediction 290 severe burns 14 clofibrate 177–8 panic disorders 269 types 64 clomiphene citrate 339 participant samples 281–5 complier average causal effects clozapine 286 patient allocation 287, 288 (CACE) 307, 308, 309 cluster randomisation 340–3, progression through therapy computed tomography (CT), spiral 351 289 for lung cancer detection Cochrane Collaboration 12 protocol 288–9 161–2 Cochrane Skin Group 229 randomisation 285–8 computers 39 contraception 332, 333 randomised controlled trials data management 40 dentistry 200–1 274–5 security of system 41 evidence-based practice 281 rating scales 291 software 40–1 cognitive behavioural therapy recruitment 285 condoms 273–93 bias 281–3 female 321 acute 278 reporting guidelines 279 male 322 anxiety disorders 255, 256, selection of patients 283 confidence intervals 32, 35, 37 268 specificity 288–9 treatment-specific 38 assessors 287–8 standardised exposure therapy confounding background services 285 290 depression studies 299–300 benefits 278 substance abuse 283–4 variables 13 blindness 287–8 symptoms consent case formulation 290 measurement 291–2 randomised controlled trials changing standards in trial severity 286 14–15 design/reporting 279, therapeutic relevance 276–7 see also informed consent 280 therapists 287 Consolidation of the Standards of choice of 300–1 effects 303–4 Reporting Trials chronicity of illness 286 treatment 289 (CONSORT) guidelines 32, clinical trial purpose/outcome components 290–1 33, 279, 353 275–6 individualised 289–90 continual reassessment method co-morbid conditions 284 interactions 286 (CRM) 97 substance abuse 283–4 cognitive impairment, mild contraception 315–33 components 274 250 acceptability of methods development 274–5, 275 cognitive therapy, schema change 330–1 diagnosis 283 291 barrier methods 322 dropouts from study 284–5 cohort studies, acupuncture 81 women 321 durability 278 coitus interruptus 322 controls 326 duration of untreated psychosis colon cancer discontinuation 330–1 286 adjuvant treatment 124–5, dose levels 324–5 effectiveness trials 281, 132–4 efficacy 288–9 advanced disease end-point 326 efficacy trials 281 126–8 surrogates 324 entry criteria 283 localised disease 124–6 emergency 319–20 evaluation 275 stage 2 disease 126 effectiveness/efficacy studies evidence-based practice stage 3 disease 126 329–30 280–5 surgery 124 pregnancy rate 329–30 exclusion criteria 283–4 treatment 124–8 side-effects 331 funding for trials 276 colonoscopy 124 equivalence trials 331–2 406 INDEX contraception (continued) non-hormonal methods Danggui Buxue Tang (Chinese ethics of trials 325, 326–7 317, 320–2 herbal medicine) 76–7 hormonal methods 315, 316 Phase I trials 324–5 Daniel, Book of (Old Testament) men 322 controls 3 vaginal bleeding patterns contraception 326 Danshen and Radix Puerariae 331 no treatment 15 Compound (Chinese herbal women 316–20, 324 Phase III trials 15–16 medicine) 79–80 implants 319, 322 placebo 15 data informed consent for trials coping skills/strategies 263, accumulation in adaptive design 326–7 290–1 96 injectable 318–19, 322 copper IUDs 320–1 aggregation 201 intention to treat (ITT) 330 coronary artery bypass graft 181, checking 40 introductory trials 332 184 collection sequence 39–40 men 316 Coronary Artery Bypass Graft cost 347 hormonal 322 (CABG) Patch trial 184 diary cards 371, 372 non-hormonal 317, 322–3 Coronary Drug Project 7, 177–8 graphical presentation 35 methods for clinical trials corticosteroids see hierarchical analysis 201–2 323–33 glucocorticosteroids imputation 303, 377–8 natural methods 321, 322 cortisol 363, 384 inverse probability weighting non-hormonal methods dose–response curves 385–6 303 315–16 cost analysis 28 likelihood analysis 303 men 317, 322–3 cost benefit 28 management 39–41 women 317, 320–2 cost effectiveness 28 quality 39–40 Phase I trials 323–5 ratio 397 software 40–1 Phase II trials 323–5 cost minimisation 28 missing metabolic studies 325 cost utility 28 asthma studies 388 depression trials 302–3 Phase III trials 325–31 costs older people trials 60 acceptability of methods direct 28, 396 respiratory disease trials 330–1 indirect 28, 396 376–8 allocation concealment 327 intangible 396 monitoring 7–8, 31 behavioural patterns 329 Cotton, Henry A 236–7 Cox, David 238 committees 31 blinding 327 dentistry studies 195 design 325–6 Cox proportional hazards model 39 multilevel analysis 201–2 effectiveness/efficacy 327, melanoma 154 pre-processing for asthma 329–30 Cox regression models 389 studies 388 objectives 325 cranial irradiation, prophylactic in processing 39 pregnancy rate estimation small-cell lung cancer quality 39–40 328–9 165–6 transcription/entry to computers randomisation 327 crossover study design 223, 339 40 recruitment 326–7 Alzheimer’s disease 249 Declaration of Helsinki 7, 11 side-effects assessment asthma 381 defibrillators 183, 184 330–1 caveats 376 dementia trial size 325–6 dentistry 195–6 associated with Lewy bodies Phase IV trials 332 respiratory diseases 381 246 progesterone-only 316–18 crossover trials 19 patients and informed consent randomised controlled trials respiratory diseases 375–6 59 323, 325–31 cyclo-oxygenase 2 (COX-2) see also Alzheimer’s disease side-effects 330–1 inhibitors 123 dental caries 193 sterilisation 322–3 Cyclofem 325, 331 aetiology 194 subgroup analysis 330 cyclosporine 214 diet 202 systematic reviews 332–3 cytarabine 145 measurement 194 vaginal bleeding patterns 331 cytostatic drugs 167–8 prevention 197–8 women 315–16 split-mouth study design barrier methods 321 196–7 hormonal methods 316–20, Dabao (Chinese herbal medicine) treatment 197–8, 204 324 226 dental filling materials 197 INDEX 407 dental floss 205 randomised trials 215–16 Early Breast Cancer Trialists’ dental fluorosis 203 score systems 218, 220 Collaborative Group dental prostheses 193 topical treatment 213 (EBCTCG) 91 dentistry 193–205 see also skin disorders early stopping rules see stopping blinding of studies 197 desogestrel 317 rules clinical trials 197–200 dextrothyroxine 177–8 Eastern Cooperative Oncology Cochrane Collaboration diaphragm, contraceptive 321 Group (ECOG) 151–2 200–1 diary card studies 377 ebastine 227 control groups 196 asthma 387–90 economic evaluation 28 crossover study design 195–6 COPD 392 gynaecology studies 347, 348, dental practice impact of trials 349 202–5 long-term 371–3, 381 data analysis 372 infertility studies 349 evidence-based 200–1 respiratory diseases 396–7 multilevel modelling 201–2 diet effectiveness studies parallel study design 195–6 dental caries 202 anxiety disorders 255–6 patient satisfaction 199 intervention trials 185 QoL 199 low-allergen 216 cognitive behavioural therapy randomised controlled trials Dietary Approaches to Stop 281 194–5 Hypertension (DASH) trial contraception 327, 329–30 split-mouth study design 185–6 pragmatic trials 338–9 196–7, 198 dietary change variables 397 systematic literature review cardiovascular disease studies efficacy 13 200–1 185 documenting in respiratory dentures, complete 198, 204 dentistry studies 197–8 disease studies 386–92 depot-medroxyprogesterone dietary fibre, colorectal cancer herbal medicine 68, 69 interim report 32 acetate (DMPA) 318–19, prevention 122 older people trials 55 322 difluoromethylornithine 123 trials 325, 331 paediatric clinical trials 49 digitalis 180 depression 297–310 relative 28 disease-free survival 145 causal effect of treatment efficacy studies disodium cromoglycate 364 298–9 asthma 380–3, 386–91 dithranol 224 estimator 300 cognitive behavioural therapy confounding 299–300 docetaxel 164 281 external validity of trials 298 dose–response curves contraception 327, 329–30 group averages comparison asthma 369, 381–2 explanatory trials 338–9 299–300 cortisol 385–6 non-randomised 13–14 internal validity of trials dose–response studies 394–5 respiratory diseases 380–3 297–8 asthma 381–2, 391 electric shock treatment 238, 239 loss to follow-up 300 double-blind trial 15 eligibility missing data 302–3 double-dummy technique 374 older people trials 56–7 observed outcomes 299 Down’s syndrome 246 requirements for randomised older people trials 56 doxorubicin controlled trials 14, 15 patient preference designs breast cancer 90 wide criteria 57 308–9 childhood cancer 112–13 psychotherapy 298–9 emphysema 360 drugs encainide 180 random allocation 300 herbal medicine interactions randomisation 299–300 endometrial ablation 346 71, 72–3 randomised consent 308 exclusions from trials 352 randomised controlled trials metabolism 13 Endometriosis Health Profile 346 297 potency ratio 386 endpoints 25–6 dermatitis, atopic 214 psychiatric 236 Chinese medicine 74 outcome measures 219 therapeutic equivalence defining 25 dermatology 211–29 392–4 dual 38 classification 211–12 therapeutic ratio 394–5 measures 35, 36–7 non-pharmacological treatments dyslipidaemia in visceral obesity multiple 37 213 18 single measures 25–6 primary care 213 dyspnoea score 370 statistical analysis 39 408 INDEX
Enhancing Recovery in Coronary false-neutral errors 96 gangliosides 152 Heart Disease (ENRCHD) false-positive errors 96 gastric cancer 119–20 study 186 fertilisation rates 341 adjuvant therapy 119–20 enrichment trial designs 248–9 fish oil, dyslipidaemia in visceral advanced disease 120 enrolment predictors 58 obesity 18 extended lymph node dissection epidermal growth factor receptor Fisher, Ronald A 5, 238–9 119 (EGFr) 164, 167 flecainide 180 localised disease 119–20 equipoise 14 fluoridation of water 197, 198, palliative therapy 120 equivalence trials 19–21 202–3 gastrointestinal cancers 117 contraception 331–2 alternatives 203 adverse event reporting 132 infertility 342–3 fluoride incidence 117 periodontal diseases 199–200 professional application to teeth post-surgical adjuvant treatment see also therapeutic equivalence 204 129–30 prognosis 117 erectile dysfunction, Korean red supplements 203 ginseng 19 safety monitoring in trials topical 204 escalation strategies, anxiety 131–2 5-fluorouracil disorders 260 staging 117 colon cancer adjuvant treatment ethical necessity 11 subgroup analysis 129–30 ethics 6–7 125, 126 toxicity in trials 132 Alzheimer’s disease 252–3 with leucovorin 132–4 tumour marker studies 130–1 childhood cancer studies colon cancer treatment 127–8 see also named cancers 111–12 gastric cancer adjuvant therapy gemcitabine contraceptive trials 325, 119–20 non-small-cell lung cancer 326–7 leucovorin combination 164 dentistry studies 195 132–4 pancreatic cancer 121 gynaecology trials 354 pancreatic cancer 120–1 general estimating equations 26 infertility trials 354 post-surgical adjuvant treatment General Oral Health Assessment paediatric clinical trials 50–1, for gastrointestinal cancers Index (GOHAI) 194 51–2 130 gestodene 317 randomisation 183–4 rectal cancer adjuvant therapy gingivitis 194 small-cell lung cancer trials 128 glass ionomer cements 198, 204 167 focal infection theory 236–7 Glenner, George 244 ethinyl oestradiol 316, 331 focus groups 258–9 glucocorticosteroids 363–4 emergency contraception follicle-stimulating hormone asthma 363, 371, 381 319–20 (FSH) 323 diary card studies 387–90 etoposide 165 follow-up information 15 dose reduction study 390–1 etretinate 217 repeated evaluation 40 level of use 387 event-free survival 145, 146 Food and Drug Administration plasma cortisol dose–response evidence-based practice 12, (FDA, US) curves 385–6 13–14 consent in paediatric clinical side-effects 363–4 cognitive behavioural therapy trials 47 systemic effects 383–5 GMK ganglioside vaccine 152, 280–5 Modernization Act 46, 112 153, 155–6 dentistry 200–1 forced expiratory volume in one gold standard 12 Ewing’s sarcoma, competing risks second (FEV ) 365–6 38 1 gonadotrophin hormone-releasing asthma challenge tests 368 hormone (GnRH) 323 exclusion criteria 56 diary cards 371, 372 exercise programmes 185 graft-versus-leukaemia effect electronic devices 371 exercise tests 370 143 forced vital capacity 365 explanatory trials 338–9 granulocyte–macrophage colony clinical outcomes 345 Formula A and Formula B stimulating factor (GM-CSF) external validity of trials 298 (Chinese herbal medicine) 145 77–8 graphs 35 Framingham Heart Study 176 group sequential methods 7 factorial design 18–19, 339–40 full factorial study design, anxiety gynaecology faecal occult blood testing disorders 268–9 blinding 351–2 123–4 functional residual capacity (FRC) cluster randomised trials false-negative errors 96 366 340–3, 351 INDEX 409
data collection 353 safety 68, 69 infants 50 economic evaluation 347, stages recommended for infertility 348, 349 research 69 blinding 351–2 ethics 354 toxicity clearance 69 cluster randomised trials exclusions from trials 352 treatment aim 66 340–3, 351 follow-up 352–3 herbs data collection 353 interventions 344 chemistry 67 dropouts from study 349 outcome finger-printing techniques 68 economic evaluation 349 definition 344–5 growing 67 equivalence trials 342–3 measures 347, 348, 349 species 67 ethics 354 patient satisfaction 346–7 uniformity 67, 68 exclusions from trials 352 QoL 345–6 hierarchical linear modelling follow-up 352–3 randomisation 349–51 201–2 intention to treat (ITT) 343 recruitment to trials 353 Hill, A Bradford 5, 6, 239 minimisation 350–1 results presentation 353 histamine 368–9 outcome sample size 349 history of clinical trials 3–8 clusters 341 stopping rules 354 Hodgkin’s disease, childhood definition 344–5 study population 343–4 104 measures 348, 349 subgroup analysis 349, 353 holistic approach, Chinese patients treatment allocation 350 medicine 69 preference trials 341–2 concealment 351 hormone replacement therapy satisfaction 347 250, 345 per protocol analysis 343 follow-up of subjects 353 QoL 345–6 haematological cancers 141–6 human chorionic gonadotrophin Hamilton Rating Scale for quasi-randomised trials 341 (hCG) 322 randomisation 349–51 Depression (HRSD) 302 Human Fertilisation and hay fever 361, 392 recruitment to trials 341–2, Embryology Authority 353 Haybittle–Peto boundary 8 (HFEA, UK) 354 hazard ratio (HR) 29, 34 results presentation 353 Human Research Protections sample size 349 melanoma 154 (OHRP, US) 112 health economics 28 stopping rules 354 Human Rights Act (UK) 284–5 study population 343–4 Heart Outcomes Prevention hydroxymethylglutaryl coenzyme Evaluation (HOPE) study subgroup analysis 349, 353 A (HMG-CoA) reductase systematic review 343 179 inhibitors 178 treatment allocation 341, 350 Helmont, Jan Baptista van 5 hyperlipidaemia 177 concealment 351 hepatocellular carcinoma treatment guidelines 178 trials 339, 340 iodine-131-labelled lipiodol hyperlipoproteinaemia 178 informed consent 7 31, 33 hypertension 175, 176, 177 children 111 Kaplan–Meier survival curves treatment guidelines 178 36 hypothesis testing 34–5 contraception trials 326–7 HER-2 89 hysterectomy 345, 346 dentistry studies 195 HER-2/neu expression in breast exclusions from trials 352 information understanding cancer 93–4 instrument 58 herbal medicine 63, 65–6 older people trials 58–60 adverse effects 70–1, 72–3, ifosfamide 164 paediatric clinical trials 47, 73–4 ileal bypass surgery 182 51–2 alopecia androgenetica 226 imatinib mesylate (STI571) 143 proxies 59 China 68–71, 72, 73–4 imipramine randomised controlled trials clinical trials 66–8 panic disorders 277 308 disease concept 66 psychotherapy study 301–2 inspiratory vital capacity (IVC) drug interactions 71, 72–3 immunocontraceptives 322, 323 366 efficacy 68, 69 implantation rates 341 insulin coma 237 extraction 68 imputation 303, 377–8 intention to treat (ITT) 33–4, methodologies 67 linear interpolation 388 300 old approaches 67–8 multiple 378 contraception 330 post-marketing surveillance in vitro fertilisation (IVF) 339, infertility 343 69 341 psychotherapy 306–7, 308 410 INDEX interferon α (IFN-α), renal leucovorin non-small-cell 160 carcinoma 21, 33 colon cancer adjuvant treatment adjuvant adenovirus for p53 interferon α2b (IFN-α2b) 125, 126 167 melanoma 154, 155, 156–7 with 5-FU 132–4 adjuvant chemotherapy adjuvant trials 151–2, 153 colon cancer treatment 127 163 interim analyses 7 gastric cancer adjuvant therapy chemoprevention 162 internal validity of trials 297–8 120 chemotherapy 164 International Conference on post-surgical adjuvant treatment chemotherapy plus radiation Harmonisation (ICH), for gastrointestinal cancers therapy 164 guideline on paediatric 130 locally advance bulky stage studies 47–51, 51 leukaemia IIIA/IIIB disease intervention trials 15 childhood 103, 104 163–4 intra-cluster correlation coefficient drug-resistant 142 metastatic 164 (ICC) 341 see also acute lymphoblastic non-bulky stage IIIA disease intracytoplasmic sperm injection leukaemia (ALL); acute 162–3 (ICSI) 341 myeloid leukaemia (AML) post-operative thoracic intrauterine contraceptive device leukotriene modifiers 364 radiotherapy 162–3 (IUD) 320–1, 328–9 levamisole 126, 134 pre-operative chemotherapy intrauterine insemination (IUI) levonorgestrel 316, 319 plus surgery 163 339 emergency contraception stage I disease 162 inverse probability weighting 319–20 stage II disease 162–3 303 equivalence trial 332 stage IV disease 164 iodine-131-labelled lipiodol 31, IUDs 320–1 targeted therapy 164 33 vaginal ring 331 treatment studies 162–4 irinotecan Lewy bodies 246 optimality criterion 166–7 colon cancer treatment 127–8 life table techniques 328 prognosis 161 non-small-cell lung cancer lifestyle change 185 screening 161–2 164 likelihood analysis 303 small-cell 159–60, 161 small-cell lung cancer 165 Lind, James 4, 5 chemotherapy 165 isotretinoin 214 lipid lowering trials 177–8, 180 chemotherapy plus chest Lipid Research Clinics Coronary irradiation 165 Primary Prevention Trial ethics of trials 167 Jadelle (contraceptive implant) 178 first-line therapy 165 319 living wills 59 methodology of trials 167 lobotomy 237 prophylactic cranial Kaplan–Meier survival curves logrank test 34, 39 irradiation 165–6 35, 38, 39 breast cancer 94 radiotherapy fractionation AML 146 Long Term Strategies in the 165 asthma 389, 390 Treatment of Panic Disorder recurrent disease 165 breast cancer 91–2 256 second-line therapy 165 pregnancy rate estimation Louis, Pierre-Charles-Alexandre treatment 164–6 328, 329 4 staging 160, 161 King’s College Questionnaire for low density lipoprotein (LDL) targeted therapy 167–8 Urinary Incontinence 345 cholesterol 178, 179 TNM staging system 160 Korean red ginseng 19 lung cancer 159–68 lung function measurements Kraepelin E 243 chemotherapy 166–7 365–6 classifications 159–60, 161 clinical trials 161–6 Laborit, Henri 239 cytostatic drugs 167–8 mammography, breast cancer 87, Lan and DeMets method 8 dose escalation 166 90–1 large simple trials 15 dose-limiting toxicity 166 mandibular deficiency 205 last value extended (LVE) 377–8 early detection 161–2 mastectomy, radical 89 left ventricular assist device 183 incidence 161 matrix metalloproteinase inhibitor leg ulcers 215 latency 159 (MMPI) 167 outcome measures 219 maximum tolerated dose 166 pancreatic cancer 121 legislation, paediatric clinical trials methodology of trials 166–8 maxillary arch expansion 200, 47 minimax trial design 166–7 205 INDEX 411 maximum tolerated dose (MTD) reporting 32 National Bureau on Drug Control lung cancer 166 respiratory diseases (China) 68 paediatric clinical trials 109 397 National Centre for measurement scales in respiratory methacholine 368–9 Complementary, Alternative diseases 364–7 provocation test 377 Treatment (NCCAM, US) medroxyprogesterone acetate, methodologies in herbal medicine 68 depot (DMPA) 318–19, 67 National Institute on Aging (NIA, 322 methotrexate 214 US) 244 trials 325, 331 mifepristone 319, 330 National Institutes of Health (NIH, melanoma 149–57 milk fluoridation 203 US) 5 adjuvant therapy 151 minimal effective dose (MED) acupuncture for pain control antibody responses 156 381, 390–1 80 Breslow’s thickness 149–50, minimisation in infertility trials complementary medicine 68 156 350–1 Inclusion of Children Policy Cox models 154 mitoxantrone 145 51 GMK ganglioside vaccine Mode Selection Trial in Sinus paediatric subject exclusion 152, 153, 155–6 Node Dysfunction (MOST) 45–6 hazard ratio 154 184 Nazi regime (Germany) 6 high-risk subcategories 156 modern medicine 64 nedocromil sodium 364 IFN-α2b 151–2, 153, 154, deficiencies 65 neuroblastoma 155, 156–7 Modified Borg scale 370 childhood 105 lymph node dissection 150, molecular targeted therapies staging system 108 151 167 translational research 106 metastases 149–50, 153 monitoring progress of trials neurofibrillary tangles 245 overall results of trials 154–5 31–2 newborn infants 50 overall survival 152–3, 154, Moniz, Egas 237 nicotinic acid 177–8 156 Monro, John 236 non-Hodgkin’s lymphoma, p-values 153–4, 155 Monte Carlo simulation 96 childhood 104–5 patient characteristics 155 more than two group design non-inferiority trials 21 post-relapse salvage therapy 17–18 non-randomised studies 13–14 151, 153 moricizine 180 non-steroidal anti-inflammatory prognosis 149, 156 mouthrinse trials 197, 199 drugs (NSAIDs) regional node staging 150–1 MRC/BHF Heart Protection Study Alzheimer’s disease 245 clinical 150 178 colorectal cancer prevention surgical 150–1, 153 Multicentre Automatic 123 relapse-free survival 152–3, Defibrillator Implantation norethindrone 316 154, 156 Trial (MADIT) 183, 184 norethisterone enantate (NET-EN) sentinel lymph node biopsy multilevel modelling 26, 60 318–19, 322 150–1 dentistry 201–2 norgestimate 317 statistical aspects of trials multiple comparisons 36–7 norgestrel 316 153–7 myelotoxic drug tolerance 109 dl-norgestrel 319 subset analysis 156–7 myocardial infarction Norplant (contraceptive implant) surgery 150 176, 183 319 trial size 154–5 nortriptyline, older people trials type I error rate 155 Naranjo’s grading of adverse 56 Menopause Rating Scale 346 effects in Chinese medicine Nottingham Health Profile 373 menorrhagia 345 71 null hypothesis 30, 34–5 patient satisfaction 346 Nasal Allergen Challenge number needed to treat (NNT) randomisation 350 Artificial Season model 35–6, 292 study population 343 370–1 psychotherapy 307, 308 menstrual diary analysis 331 nasal patency assessment 366–7 Nuremburg Code 6–7 Menstrual Distress questionnaire nasal polyposis 361–2 346 National Bioethics Advisory O’Brien–Fleming method mental illness 235 Commission (US) 112 8, 110 Mesigyna 325, 331 National Breast Cancer Coalition oesophageal cancer 118–19 meta-analysis fund (NBCCF, US) Clinical advanced 118 breast cancer 91 Trial Initiative 88 localised 118–19 412 INDEX oesophageal cancer (continued) implant-retained mandibular patient satisfaction pre-operative 198, 204 gynaecology 346–7 radiochemotherapy 118 maxillary implant 199 infertility 347 survival 119 overview groups 12 skin disorders 225 oestradiol cypionate 325 overviews, reporting 32 peak expiratory flow (PEF) 365, oestrogens oxaliplatin 127, 134 366 Alzheimer’s disease 245 diary cards 371, 372, 389 compounds 123 p-value 35, 37 electronic devices 371 equine 177–8 p53 tumour marker 131 peak nasal flow 367, 372 oral contraceptives 316 adjuvant adenovirus 167 Pearl index 326, 328 synthetic 316 pacemaker implantation 181, 184 Pentothal 239 oestrone sulphate 331 paclitaxel 164 per protocol analysis 343 older people, clinical trials paediatric clinical trials 45–6 per protocol summary 33 55–61 age classification of patients periodontal diseases 193 AML 143–4, 146 50 aetiology 194 atherosclerosis 177 assent to participate 52 equivalence trials 199–200 decision making 58–9 dose-limiting toxicity 109 measurement 194 eligibility 56–7 drug development 108–9 randomised controlled trials follow-up 60 early testing 53 199 informed consent 58–60 efficacy 49 split-mouth study design 196 nature of trial 59 ethics 50–1, 51–2 trials 199–200, 205 nursing home residents 59–60 ICH Guideline 47–51 Petrarch 3 understanding 59 informed consent 47, 51–2 pharmaceutical companies, oocytes 341 initiation 48 paediatric data requirements retrieval 345 legislation 47 46 oral cancer 193 maximum tolerated dose 109 pharmacists, skin disorders 213 oral contraceptives 316–18 recruitment 52–3 pharmacodynamics 13 cardiovascular disease regulatory issues 46–7 in children 46 317–18 safety 49–50 paediatric clinical trials 49 safety concerns 317 safety information lack 46 pharmacoeconomics 396–7 Oral Health Impact Profile (OHIP) study management 53 see also economic evaluation 194, 199 types of studies 48–50 pharmacokinetics 13 oral hygiene studies 195 see also cancer, childhood paediatric clinical trials 46, oral rehabilitation studies 198–9 pancreatic cancer 120–1 48–9 orphan drug regulation 46 panic attacks 267 Phase I trials 12–13, 337–8 orthodontic treatment 193, 200 panic disorders Alzheimer’s disease 247–8 randomised controlled trials cognitive behavioural childhood cancer 108, 110–11 200, 205 treatments 269 contraception 323–5 timing 205 imipramine 277 lung cancer 166 Osler, William 7 placebo–drug comparison 283 respiratory diseases 380–6 osteogenic sarcoma, childhood parallel study design safety monitoring 131–2 105, 112–13 contraceptives 325–6 skin disorders 222–5 outcome measures dentistry 195–6 Phase II trials 12–13, 338 clusters 341 randomised control 248–9 Alzheimer’s disease 248 gynaecology 347, 348, 349 respiratory diseases 376, 379 childhood cancer 109–10 infertility 348, 349 simple 339 contraception 323–5 pragmatic trials 349 Pare,´ Ambroise 4 historical comparison 110 skin disorders 218, 219, Park study 370–1 lung cancer 166–7 220–2, 227–8 partially randomised patient oesophageal cancer 119 outcomes preference (PRPP) 342 proving activity 109 clinical 345 pathological process response 70 randomised comparison 110 definition in patient motivation in skin respiratory diseases 380–6 gynaecology/infertility disorders 225–6 safety monitoring 131–2 344–5 patient preference trials skin disorders 222–5 primary 349 infertility 341–2 testing activity 109–10 secondary 349 psychotherapy 308–9 within-patient control studies overdentures skin disorders 225–6 223–5 INDEX 413
Phase III trials 12, 14–16, 338 emergency contraception assessment method 302–3 Alzheimer’s disease 248–50 329–30 brief dynamic 300 childhood cancer 102, 103, estimation 328–9 causal effects 306–8 110–11 PREMIER trial 185–6 centre effects 303–4 colon cancer adjuvant treatment presenilin 244 choice of 300–1 125 preterm newborn infants 50 clinical management 301–2 contraception 325–31 Prevention of Events with comparison of types 299 control therapy 15–16 Angiotensin-Converting components 288 cure model 110 Enzyme Inhibitor Therapy control group 301–2 double-blind for supportive (PEACE) trial 179–80 depression 298–9 cancer care 111 prevention trials 15 dropouts from study 302–3 lung cancer 167 primary care effect modification 306 proportional hazards 110 anxiety disorders 258 efficacy evaluation 300 respiratory diseases 386–95 dermatology 213 equipoise 301 safety monitoring 131–2 primary prevention 178 intention to treat (ITT) 306–7 skin disorders 225–8 Alzheimer’s disease 251 interpersonal 300, 305 standard therapy 15–16 cardiovascular disease 175 missing data 302–3 Phase IV trials 12, 15–16, 338 primordial prevention 175 number needed to treat (NNT) respiratory diseases 395–6 progesterone 319 307 photochemotherapy 213 progesterone-only pills (POPs) outcome features 302–3 phototherapy 213 316–18 patient preference designs PUVA 214, 228 progestins 316–17 308–9 ultraviolet B 214, 216 implants 319 randomisation 301 Phyllanthus SP (Chinese herbal Program on the Surgical Control randomised consent 308 medicine) 75–6 of Hyperlipidaemia standardisation 300–1 pilosebaceous opening plugging (POSCH) 182 therapist effects 303–5 213 Propionibacterium acne 213 treatment manual 301 placebo 6 proportional hazards regression treatment protocol 289 acupuncture 80–1 models 94 Alzheimer’s disease 252–3 protein kinase inhibitors 167 Q-TWiST method 146 controls 15 psoralen plus ultraviolet A quality of life (QoL) 26, 27, 28 psychotherapy study 301–2 phototherapy (PUVA) 214 AML 146 rhinitis studies 374 follow-up study 228 asthma 373 skin disorder trials 226–7 psoriasis 214–15 breast cancer 91 plaque control 205 dithranol treatment 224 Chinese medicine 73–4 plaque index, pre-/post-brushing outcome measures 219 COPD 373 195 randomised controlled trials data sheets 74 plaque removal 199 225, 226 dentistry 199 blinding of study 197 treatment 216 evaluation 40 mechanical 205 trials 217 gynaecology 345–6 split-mouth study design 196 Psoriasis Area and Severity Index infertility 345–6 Pocock boundaries 7–8 (PASI) 215, 218, 220, 221 leg ulcers 215 pollen allergy 361, 370–1 psychiatry 235 multiple comparisons 37 season onset 392 early twentieth century post-marketing surveillance treatment 236–8 skin disorders 225, 228 quasi-randomised trials 341 15–16 late twentieth century 238–40 27, adverse effects in Chinese physical therapies 237–8, 239 questionnaires 26, 28 medicine 71, 73 randomisation 239 herbal medicine 69 removal of foci of infection Radiation Therapy Oncology paediatric clinical trials 50 236–7 Group (RTOG) 118 potency ratio 386 see also named conditions radiochemotherapy, pre-operative pragmatic trials 338–9 psychosis 118 clinical outcomes 345 chronicity 286 radiotherapy, post-operative outcome measures 349 duration of untreated 286 thoracic for non-small-cell pravastatin 178 functional 236–7 lung cancer 162–3 pre-randomisation variables 38 psychotherapy 276–7 radon exposure 159 pregnancy rate adherence to 300–1 raloxifene 91 414 INDEX random effects models 60 Rapid Early Action for Coronary measurement scales 364–7 randomisation 5, 13, 24–5, 297 Treatment (REACT) trial meta-analysis 397 block 216, 350 186 methods for trials 373–80 cluster 340–3 recruitment 34 missing data 376–8 cognitive behavioural therapy bias in cognitive behavioural multiple comparisons 378 285–8 therapy 281–3 parallel study design 376, 379 concealed 15 gynaecology trials 353 pharmacoeconomics 396–7 contraception trials 327 infertility trials 353 Phase I/II studies 380–6 depression studies 299–300 older people trials 56, 57 Phase III trials 386–95 ethics 183–4 paediatric clinical trials 52–3 Phase IV studies 395–6 gynaecology studies 349–51 strategies 57 protocol violations 375 infertility studies 349–51 rectal cancer randomisation 374–5 list 15 postoperative radiotherapy safety documenting 386–92 psychiatry 239 129 sample size determinations psychotherapy 301 preoperative radiotherapy 378–9 ratio 25 128–9 sequential trials 376 respiratory disease studies surgery/adjuvant therapy subgroup analysis 379–80 374–5 128–9 systemic effects of drugs skin disease trials 215–16 total mesorectal excision 383–6 stratification 24, 350 (TME) 129 systemic exposure to drugs unbalanced 96 treatment 128–9 393 understanding by subjects 58 referees, guidelines 32 therapeutic equivalence randomised consent 226, 308 regression models 13 392–5 randomised control parallel trial regression techniques 35 treatment costs 396 designs 248–9 Regulatory Agencies 39 respiratory tract, upper, randomised controlled trials regulatory authorities, paediatric inflammation 361 clinical trials 46–7 (RCTs) 14, 238–9 retinoblastoma, childhood 105 Relieve Wheezing Tablet (Chinese causal effects 306–8 all-trans retinoic acid (ATRA) herbal medicine) 78–9 cluster 340–3 142 remission, length of 145–6 cognitive behavioural therapy retinopathy of prematurity renal carcinoma, IFN-α 21, 33 274–5 (retrolental fibroplasia) 5 reporting of clinical trials 32, 33 retrospective review, solid tumour comparison of randomised rescue medication 375 groups 306 cancers 29 asthma 389 reversibility of lung function 366 consent 14–15 β2-agonists 380 rhabdomyosarcoma contraception 323, 325–31 resin-based composite (RBC) childhood 105 cost data 347 204 risk assignment 106, 107 dentistry 194–5 respiratory cancers see lung cancer rheumatic heart disease 5 depression 297 respiratory diseases 359–97 rhinitis 359, 361–2 eligibility requirements 14, 15 airway diseases 359–73 blinding of studies 374 equipoise 301 bias 377 exposure studies 370–1 informed consent 308 bioequivalency 392–3 glucocorticosteroids 363 multicentre 303 blinding 373–4 perennial 392 orthodontics 200, 205 bronchodilatation 382–3 placebo groups 374 periodontal diseases 199 crossover trials 375–6, 381 seasonal 392 placebo-controlled 227 current treatments 362–4 studies 392 skin disorders 218, 228 design of trials 375–6 rhinomanometry, posterior 366 psoriasis 225, 226 diary card studies 371–3, 377 risks, competing 38–9 summarising results of dropout from study 377 rofecoxib 123 several trials 228–9 efficacy systematic review 343 documenting 386–92 therapist effects 303–5 studies 380–3 safety Randomised Evaluation of inhalation products 374, asthma studies 386–91 Mechanical Assistance for 392–3 documenting in respiratory the Treatment of Congestive interim analysis 376 disease studies 386–92 Heart Failure (REMATCH) lung function measurements herbal medicine 68, 69 183–4 365–6 paediatric clinical trials 49–50 INDEX 415 safety monitoring 7, 31 skin disorders 211–15 statins 178 gastrointestinal cancer trials accessory care 217 statistical analysis 24 131–2 blinding 216–17 computer use 39 paediatrics 46 clinical trial methods 215–18, guidelines 32 St George’s Respiratory 219, 220–2 software 39 Questionnaire 373 disease-modifying strategies statistical significance 29 salbutamol (albuterol) 393 216 stem cell transplantation, salt fluoridation 203 disease severity 217–18 allogeneic 143 sample size 30–1 dropouts from studies 228 sterilisation 322–3, 328 dentistry studies 195 entry criteria to studies 226 Sternback, Leo 239 determination 37 management 216 stochastic curtailment 7 respiratory diseases 378–9 measurement error 223 stopping rules 31–2 Sargant, William 237–8, 239 outcome measures 218, 219, breast cancer 95–6 Scandinavian Simvastatin Survival 220–2, 227–8 infertility/gynaecology trials Study (4S) 178 patient motivation/preference 354 schema change 291 225–6 streptomycin, use in pulmonary schizophrenia patient satisfaction 225 tuberculosis 5, 6, 239 acute care 277–9 pharmacists 213 stringency 8 cognitive behavioural therapy Phase I and II studies 222–5 stroke 176 273–93 Phase III trials 225–8 Student t-test 34 maintenance therapy 277–9 placebo use 226–7 subgroup analysis 37–8 medication 278 prevalence 212 contraception 330 recruitment bias 281–3 QoL 225, 228 gastrointestinal cancers residual symptoms 277–8 randomisation 215–16 129–30 scurvy 4, 5 randomised controlled trials infertility/gynaecology trials seasonal allergic rhinitis 361 218, 228 349, 353 secondary prevention summarising results of respiratory diseases 379–80 Alzheimer’s disease 251 several trials 228–9 sugar alcohols 202 cardiovascular disease 175 score systems 228 sugar consumption 202 selective oestrogen-receptor severity superiority trials 343 modulators (SERMs) 89, criteria 218 surrogates 13 91 spectrum 212 survival selective serotonin reuptake study settings 217–18 analysis 38 inhibitors (SSRIs) 239 time frame for evaluation disease-free 145, 146, 152–3 selegiline, Alzheimer’s disease 227–8 event-free 145, 146 249–50, 252 treatment overall 146, 152–3, 154, 156 selenium 162 modality standardisation relapse-free 145, 152–3, 154, Self-Administered Psoriasis Area 217 156 and Severity Index (PASI) penetration 222 survival time methods 25, 26 220 topical 213 symptoms self-controlled study design 223 variation 213 identification 69 senile plaque 245 skin failure 211 measurement in cognitive sequential trials 21–2 Social Phobia Inventory (SPIN) behavioural therapy closed 21, 23 265 291–2 open 21, 22 social phobics 267 pathological causes 66 respiratory diseases 376 SoCRATES Trial 275 suppression 66 sertraline, older people trials 56 spermicides 321 syphilis, Tuskegee study 6 sexual intercourse spirometry 365–6 systematic review of infertility coitus interruptus 322 split-mouth study design, dentistry 343 periodic abstinence 321 196–7, 198 Systolic Hypertension in the sexually transmitted diseases 321 sputum production 361 Elderly program 180 Shepherd, Michael 239 ST 1435 (progestin) 319 Short Form Health Survey (SF-36) standard codes 40–1 26, 27, 345, 373 standard error (SE) 35 tacrine 249 Sickness Impact Profile 373 standardised effect size 30 tamoxifen, breast cancer 15, 89, sigmoidoscopy, flexible 124 Stanford Heart Transplant 90, 91 simvastatin 178 Program 13 taxanes 164 416 INDEX teeth small-cell lung cancer vaginal bleeding patterns 331 extraction for orthodontics 165 vaginal rings 319 205 topotecan validity fissure sealant 195–6, 198, non-small-cell lung cancer external 298 204 164 internal 297–8 fluoride application 204 small-cell lung cancer 165 vasectomy 322–3 implants 193, 198–9, 204 tradition, strong 70 vasoconstrictors 364 malalignment/malocclusion traditional healing 69–70 venous thromboembolism 193, 200 trandolapril 179 317–18 replacement 204–5 Transitional Dyspnoea Index 370 ventricular arrhythmias 180 tertiary prevention, Alzheimer’s translational research 103, 105 ventricular ejection fraction, left disease 251 ancillary studies 111 179, 183 testosterone, injectable neuroblastoma 106 vinorelbine 164 contraceptives 322 trastuzumab 89 vitamin A 162 tetrahydroaminoacridine (Cognex) treatment vitamin E in Alzheimer’s disease 244 efficacy interim report 32 249–50, 252 theophylline 362–3 random allocation 14 therapeutic advantage, small Treatment of Depression 15 Collaborative Research Western lifestyle 177 therapeutic equivalence 20, Program (TDCRP) 301 Wilms’ tumour, childhood 105 392–4 trial size 29–31 within-patient studies 223–5 marketing 393–4 tuberculosis, pulmonary and comparisons 19 therapeutic index, relative streptomycin use 5, 6, 239 Withington, Charles Francis 7 394–5, 396 tumour markers, gastrointestinal Women’s Health Initiative (WHI) 250 therapeutic ratio 394–5 cancers 130–1 thoracotomy 182 Tuskegee (Alabama, US), syphilis time-to-event studies 26 study 6 xanthines 362–3 TNM staging system 117 Type I error 7, 8 xylitol 202 lung cancer 160 Type I error rate 30 toddlers 50 melanoma trials 155 toothbrushing 205 Type II error rate 30 Yale–Brown Obsessive toothpaste trials 199 Compulsive Scale 265 antimicrobial agents 205 Yin and Yang, balance fluoridation 203–4 maintenance 66 topical drugs ulceration, leg 215, 219 Yuzpe regimen 319, 320, 332 pharmacodynamic parameters ultraviolet B phototherapy 214, 223 216 skin penetration 222 urinary incontinence 345 ZD1839 (EGFr inhibitor) 164 topoisomerase I inhibitors study population 343–4 Zelen consent design 308 non-small-cell lung cancer ursodeoxycholic acid 123 double 23, 226 164 urticaria 227 single 15, 22–4