Postoperative Analgesia and Vomiting, with Special Reference to Day-Case Surgery: a Systematic Review

Home , Bentley compounds, Bicifadine, Ciramadol, Etofenamate, Felbinac, Tonazocine

Health Technology Assessment 1998; Vol. 2: No. 12 Review

Postoperative analgesia and vomiting, with special reference to day-case surgery: a systematic review

Henry J McQuay R Andrew Moore

Health Technology Assessment NHS R&D HTA Programme HTA Standing Group on Health Technology Chair: Professor Sir Miles Irving, Professor of Surgery, University of Manchester, Hope Hospital, Salford †

Dr Sheila Adam, Professor Howard Glennester, Mr Hugh Ross, Department of Health Professor of Social Science & Chief Executive, The United Bristol Professor Martin Buxton, Administration, London School of Healthcare NHS Trust † Professor of Economics, Brunel University † Economics & Political Science Professor Ian Russell, Professor Angela Coulter, Mr John H James, Department of Health, Sciences & Director, King’s Fund, London Chief Executive, Kensington, Chelsea & Clinical Evaluation, University of York Westminster Health Authority Professor Anthony Culyer, Professor Trevor Sheldon, Deputy Vice-Chancellor, University of York Professor Richard Lilford, Director, NHS Centre for Reviews & † Dissemination, University of York † Dr Peter Doyle, Regional Director, R&D, West Midlands Executive Director, Zeneca Ltd, Professor Michael Maisey, Professor Mike Smith, ACOST Committee on Medical Research Professor of Radiological Sciences, Director, The Research School † & Health UMDS, London of Medicine, University of Leeds Professor John Farndon, Dr Jeremy Metters, Dr Charles Swan, Professor of Surgery, University of Bristol † Deputy Chief Medical Officer, Consultant Gastroenterologist, † North Staffordshire Royal Infirmary Professor Charles Florey, Department of Health Department of Epidemiology & Mrs Gloria Oates, Dr John Tripp, Public Health, Ninewells Hospital & Chief Executive, Oldham NHS Trust Department of Child Health, Royal Devon & Exeter Healthcare NHS Trust † Medical School, University of Dundee † Dr George Poste, Professor John Gabbay, Chief Science & Technology Officer, Professor Tom Walley, Director, Wessex Institute for Health SmithKline Beecham † Department of Pharmacological Therapeutics, University of Liverpool † Research & Development † Professor Michael Rawlins, Professor Sir John Grimley Evans, Wolfson Unit of Clinical Pharmacology, Dr Julie Woodin, Department of Geriatric Medicine, University of Newcastle-upon-Tyne Chief Executive, Nottingham Health Authority † Radcliffe Infirmary, Oxford † Professor Martin Roland, Dr Tony Hope, Professor of General Practice, The Medical School, University of Oxford † University of Manchester † Current members

HTA Commissioning Board Chair: Professor Charles Florey, Department of Epidemiology & Public Health, Ninewells Hospital & Medical School, University of Dundee †

Professor Ian Russell, Professor Sir Miles Irving Dr Tim Peters, Department of Health, Sciences & (Programme Director), Professor of Reader in Medical Statistics, Department of Clinical Evaluation, University of York * Surgery, University of Manchester, Social Medicine, University of Bristol † † Dr Doug Altman, Hope Hospital, Salford Professor David Sackett, Director of ICRF/NHS Centre for Professor Alison Kitson, Centre for Evidence Based Medicine, † Statistics in Medicine, Oxford Director, Royal College of Oxford Mr Peter Bower, Nursing Institute † Professor Martin Severs, Independent Health Advisor, Professor Martin Knapp, Professor in Elderly Health Care, Newcastle-upon-Tyne † † Director, Personal Social Services Portsmouth University Ms Christine Clark, Research Unit, London School of Dr David Spiegelhalter, Honorary Research Pharmacist, Economics & Political Science MRC Biostatistics Unit, Institute of † Hope Hospital, Salford Public Health, Cambridge Professor David Cohen, Dr Donna Lamping, Senior Lecturer, Department of Public Dr Ala Szczepura, Professor of Health Economics, Health, London School of Hygiene & Director, Centre for Health Services Studies, University of Glamorgan Tropical Medicine † University of Warwick † Mr Barrie Dowdeswell, Chief Executive, Royal Victoria Infirmary, Professor Theresa Marteau, Professor Graham Watt, Department of General Practice, Newcastle-upon-Tyne Director, Psychology & Genetics Woodside Health Centre, Glasgow † Professor Martin Eccles, Research Group, UMDS, London Professor of Clinical Effectiveness, Professor Alan Maynard, Professor David Williams, Department of Clinical Engineering, University of Newcastle-upon-Tyne † Professor of Economics, University of York † University of Liverpool Dr Mike Gill, Professor Sally McIntyre, Director of Public Health and Health Policy, Dr Mark Williams, MRC Medical Sociology Unit, Glasgow Brent & Harrow Health Authority † Public Health Physician, Bristol Dr Jenny Hewison, Professor Jon Nicholl, Dr Jeremy Wyatt, Senior Lecturer, Department of Psychology, Director, Medical Care Research Unit, Senior Fellow, Health and Public Policy, † University of Leeds † University of Sheffield School of Public Policy, University College, † Dr Michael Horlington, Professor Gillian Parker, London Head of Corporate Licensing, Smith & Nuffield Professor of Community Care, * Previous Chair Nephew Group Research Centre University of Leicester † † Current members HTA

How to obtain copies of this and other HTA Programme reports. An electronic version of this publication, in Adobe Acrobat format, is available for downloading free of charge for personal use from the HTA website (http://www.hta.ac.uk). A fully searchable CD-ROM is also available (see below). Printed copies of HTA monographs cost £20 each (post and packing free in the UK) to both public and private sector purchasers from our Despatch Agents. Non-UK purchasers will have to pay a small fee for post and packing. For European countries the cost is £2 per monograph and for the rest of the world £3 per monograph. You can order HTA monographs from our Despatch Agents: – fax (with credit card or official purchase order) – post (with credit card or official purchase order or cheque) – phone during office hours (credit card only). Additionally the HTA website allows you either to pay securely by credit card or to print out your order and then post or fax it.

Contact details are as follows: HTA Despatch Email: [email protected] c/o Direct Mail Works Ltd Tel: 02392 492 000 4 Oakwood Business Centre Fax: 02392 478 555 Downley, HAVANT PO9 2NP, UK Fax from outside the UK: +44 2392 478 555 NHS libraries can subscribe free of charge. Public libraries can subscribe at a very reduced cost of £100 for each volume (normally comprising 30–40 titles). The commercial subscription rate is £300 per volume. Please see our website for details. Subscriptions can only be purchased for the current or forthcoming volume.

Payment methods Paying by cheque If you pay by cheque, the cheque must be in pounds sterling, made payable to Direct Mail Works Ltd and drawn on a bank with a UK address. Paying by credit card The following cards are accepted by phone, fax, post or via the website ordering pages: Delta, Eurocard, Mastercard, Solo, Switch and Visa. We advise against sending credit card details in a plain email. Paying by official purchase order You can post or fax these, but they must be from public bodies (i.e. NHS or universities) within the UK. We cannot at present accept purchase orders from commercial companies or from outside the UK.

How do I get a copy of HTA on CD? Please use the form on the HTA website (www.hta.ac.uk/htacd.htm). Or contact Direct Mail Works (see contact details above) by email, post, fax or phone. HTA on CD is currently free of charge worldwide.

The website also provides information about the HTA Programme and lists the membership of the various committees.

Postoperative analgesia and vomiting, with special reference to day-case surgery: a systematic review

Henry J McQuay R Andrew Moore

Pain Research & Nuffield Department of Anaesthetics, University of Oxford, Oxford Radcliffe Hospital Trust, Churchill Hospital, Oxford

Published December 1998

This report should be referenced as follows:

McQuay HJ, Moore RA. Postoperative analgesia and vomiting, with special reference to day-case surgery: a systematic review. Health Technol Assessment 1998; 2(12).

Health Technology Assessment is indexed in Index Medicus/MEDLINE and Excerpta Medica/ EMBASE. Copies of the Executive Summaries are available from the NCCHTA web site (see overleaf). NHS R&D HTA Programme

he overall aim of the NHS R&D Health Technology Assessment (HTA) programme is to T ensure that high-quality research information on the costs, effectiveness and broader impact of health technologies is produced in the most efficient way for those who use, manage and work in the NHS. Research is undertaken in those areas where the evidence will lead to the greatest benefits to patients, either through improved patient outcomes or the most efficient use of NHS resources. The Standing Group on Health Technology advises on national priorities for health technology assessment. Six advisory panels assist the Standing Group in identifying and prioritising projects. These priorities are then considered by the HTA Commissioning Board supported by the National Coordinating Centre for HTA (NCCHTA). This report is one of a series covering acute care, diagnostics and imaging, methodology, pharmaceuticals, population screening, and primary and community care. It was identified as a priority by the Pharmaceutical Panel and funded as project number 94/11/04. The views expressed in this publication are those of the authors and not necessarily those of the Standing Group, the Commissioning Board, the Panel members or the Department of Health. The editors wish to emphasise that funding and publication of this research by the NHS should not be taken as implicit support for the recommendations for policy contained herein. In particular, policy options in the area of screening will, in England, be considered by the National Screening Committee. This Committee, chaired by the Chief Medical Officer, will take into account the views expressed here, further available evidence and other relevant considerations. Reviews in Health Technology Assessment are termed ‘systematic’ when the account of the search, appraisal and synthesis methods (to minimise biases and random errors) would, in theory, permit the replication of the review by others.

Series Editors: Andrew Stevens, Ruairidh Milne and Ken Stein Assistant Editors: Jane Robertson and Jane Royle

The editors have tried to ensure the accuracy of this report but cannot accept responsibility for any errors or omissions. They would like to thank the referees for their constructive comments on the draft document.

ISSN 1366-5278

Enquiries relating to copyright should be addressed to the NCCHTA (see address given below).

Published by Core Research, Alton, on behalf of the NCCHTA. Printed on acid-free paper in the UK by The Basingstoke Press, Basingstoke.

Copies of this report can be obtained from:

The National Coordinating Centre for Health Technology Assessment, Mailpoint 728, Boldrewood, University of Southampton, Southampton, SO16 7PX, UK. Fax: +44 (0) 1703 595 639 Email: [email protected] http://www.soton.ac.uk/~hta Health Technology Assessment 1998; Vol. 2: No. 12

Contents

List of abbreviations ...... i 8 Paracetamol with and without codeine in acute pain ...... 61 Executive summary ...... iii Summary ...... 61 1 Introduction ...... 1 Introduction ...... 61 Background and key questions ...... 1 Methods ...... 61 Conclusion ...... 4 Results ...... 62 References ...... 4 Comment ...... 69 References ...... 77 2 Finding all the relevant trials ...... 7 Developing a citation database ...... 7 9 Oral ibuprofen and diclofenac in Trends in the numbers of RCTs in pain postoperative pain ...... 81 relief research published, 1950–90 ...... 9 Summary ...... 81 Conclusions ...... 9 Introduction ...... 81 References ...... 9 Methods ...... 81 Results ...... 82 3 Judging the quality of trials ...... 11 Comment ...... 91 Developing and validating a quality scale ..... 11 References ...... 94 Comments on the scale ...... 12 References ...... 14 10 Comparison of the analgesic efficacy of NSAIDs given by different routes 4 Pain measurement, study design in acute and chronic pain ...... 97 and validity ...... 17 Summary ...... 97 Pain measurement ...... 17 Introduction ...... 97 Using pain measurement data for Methods ...... 98 systematic reviews ...... 18 Results ...... 98 Study design and validity ...... 19 Comment ...... 104 References ...... 21 References ...... 106 5 Estimating relative effectiveness ...... 23 11 Topically-applied NSAIDs ...... 109 Placebo responses in analgesic trials ...... 23 Summary ...... 109 Deriving dichotomous outcome measures Introduction ...... 109 from continuous data in RCTs Methods ...... 109 of analgesics ...... 25 Results ...... 110 Verification from independent data ...... 31 Comments ...... 119 Use of pain intensity and VAS ...... 33 References ...... 122 Comment ...... 37 References ...... 38 12 Injected morphine in postoperative pain ...... 127 6 Combining data and interpreting results ... 39 Summary ...... 127 Combining data: quantitative Introduction ...... 127 systematic reviews ...... 40 Methods ...... 127 How well does an intervention work? ...... 43 Results ...... 128 Is it safe?...... 44 Comment ...... 128 Using NNTs ...... 45 References ...... 134 References ...... 46 13 Dihydrocodeine in postoperative pain ..... 137 7 Existing systematic reviews ...... 47 Summary ...... 137 Introduction ...... 47 Introduction ...... 137 Methods ...... 47 Methods ...... 137 Results ...... 48 Results ...... 138 Comments ...... 51 Comment ...... 140 References ...... 51 References ...... 141 Contents

14 Dextropropoxyphene in postoperative Methods ...... 191 pain ...... 143 Results ...... 192 Summary ...... 143 Comment ...... 193 Introduction ...... 143 References ...... 196 Methods ...... 143 Results ...... 144 20 TENS in labour pain ...... 199 Comment ...... 148 Summary ...... 199 References ...... 149 Methods ...... 199 Results ...... 200 15 Oral tramadol codeine and combination Comment ...... 200 analgesics in postoperative pain ...... 151 References ...... 204 Summary ...... 151 Introduction ...... 151 21 Efficacy and harm of anti-emetic Methods ...... 152 interventions in the surgical setting ...... 207 Results ...... 153 Data source ...... 207 Comment ...... 156 Efficacy of anti-emetic interventions ...... 207 References ...... 159 Potential for harm of anti-emetic interventions ...... 210 16 Pain relief from intra-articular morphine Prophylaxis versus treatment ...... 211 after knee surgery ...... 161 Research agenda ...... 213 Summary ...... 161 References ...... 213 Introduction ...... 161 Methods ...... 161 22 Cost-effectiveness of ondansetron: Results ...... 162 prophylaxis compared with treatment Comment ...... 166 in PONV ...... 215 References ...... 168 Summary ...... 215 Introduction ...... 215 17 Analgesic efficacy of peripheral opioids ..... 171 Methods ...... 215 Summary ...... 171 Results ...... 218 Introduction ...... 171 Comment ...... 221 Methods ...... 171 References ...... 223 Results ...... 172 Comment ...... 176 23 Conclusion ...... 225 References ...... 177 Interventions ...... 225 18 Pre-emptive analgesia: a systematic Clinical settings and recommendations ...... 229 Conclusion ...... 230 review of clinical studies to 1994 ...... 181 References ...... 230 Summary ...... 181 Introduction ...... 181 The clinical evidence ...... 182 Acknowledgements ...... 231 Comment ...... 187 References ...... 188 Health Technology Assessment reports published to date ...... 233 19 TENS in acute postoperative pain ...... 191 Summary ...... 191 Health Technology Assessment Introduction ...... 191 panel membership ...... 235 Health Technology Assessment 1998; Vol. 2: No. 12

List of abbreviations

AUC area under a curve* CATPR categorical verbal rating scale of pain relief CER control event rate CI confidence interval EER experimental event rate ESR erythrocyte sedimentation rate NNH number-needed-to-harm NNT number-needed-to-treat NSAID non-steroidal anti-inflammatory drug NSD no significant difference* PCA patient-controlled analgesia PI pain intensity* PID pain intensity difference* PO postoperative* PONV postoperative nausea and vomiting PR pain relief* RB relative benefit* RCT randomised controlled trial RR relative risk* RWJ Robert Wood Johnson Pharmaceutical Research Institute, Spring House, PA, USA SD standard deviation* SEM standard error of the mean* SPID sum of pain intensity differences TENS transcutaneous electrical nerve stimulation TFA time to first analgesic* TOTPAR total pain relief VAS visual analogue scale VRS verbal rating scale*

* Used only in figures and tables

Health Technology Assessment 1998; Vol. 2: No. 12

Executive summary

Background • the use of dihydrocodeine, 30 mg, in acute postoperative pain (it is no better than placebo). Day-case surgery is of great value to patients and the health service. It enables many more patients Interventions of doubtful value to be treated properly, and faster than before. Some interventions may be effective but the size of Newer, less invasive, operative techniques will allow the effect or the complication of undertaking them many more procedures to be carried out. confers no measurable benefit over conventional methods. Such interventions include: There are many elements to successful day-case surgery. Two key components are the effective- • injecting morphine into the knee joint after ness of the control of pain after the operation, surgery: there is a small analgesic benefit which and the effectiveness of measures to minimise may last for up to 24 hours but there is no clear postoperative nausea and vomiting. evidence that the size of the benefit is of any clinical value • manoeuvres to try and anticipate pain by using Objectives pre-emptive analgesia; these are no more effective than standard methods To enable those caring for patients undergoing • administering non-steroidal anti-inflammatory day-case surgery to make the best choices for their drugs (NSAIDs) by injection or per rectum in patients and the health service, this review sought patients who can swallow; this appears to be no the highest quality evidence on: more effective than giving NSAIDs by mouth and, indeed, may do more harm than good • the effectiveness of the control of pain after • administering codeine in single doses; this has an operation poor analgesic efficacy. • the effectiveness of measures to minimise postoperative nausea and vomiting. Interventions of proven value These include a number of oral analgesics including (at standard doses): Methods • dextropropoxyphene Full details of the search strategy are presented in • tramadol the report. • paracetamol • ibuprofen • diclofenac. Results Diclofenac and ibuprofen at standard doses give Analgesia analgesia equivalent to that obtained with 10 mg The systematic reviews of the literature explored of intramuscular morphine. Each will provide at whether different interventions work and, if they least 50% pain relief from a single oral dose in do work, how well they work. A number of patients with moderate or severe postoperative conclusions can be drawn. pain. Paracetamol and codeine combinations also appear to be highly effective, although Ineffective interventions there is little information on the standard doses There is good evidence that some interventions are used in the UK. The relative effectiveness of ineffective. They include: these analgesics is compared in an effectiveness ‘ladder’ which can inform prescribers making • transcutaneous electrical nerve stimulation choices for individual patients, or planning in acute postoperative pain day-case surgery. Dose–response relationships • the use of local injections of opioids at sites show that higher doses of ibuprofen may be other than the knee joint particularly effective. iii Executive summary

Topical NSAIDs (applied to the skin) are effective Research recommendations in minor injuries and chronic pain but there is no obvious role for them in day-case surgery. • To extend the number of systematic reviews to include other analgesics, including the Postoperative nausea and vomiting newer NSAIDs. This would provide a more The proportion of patients who may feel nauseated comprehensive ladder of relative efficacy. or vomit after surgery is very variable, despite It is unlikely that large ‘head-to-head’ similar operations and anaesthetic techniques. comparisons of analgesics in randomised Systematic review can still lead to clear estimations controlled trials would provide more of effectiveness of interventions. Whichever useful information. anti-emetic is used, the choice is often between prophylactic use (trying to prevent anyone • To establish pilot audits of the implementation vomiting) and treating those people who do of the information included in this report; feel nauseated or who may vomit. both before and after audits are needed to put the existing evidence into clinical practice Systematic reviews of a number of different anti- to good effect. emetics show clearly that none of the anti-emetics is sufficiently effective to be used for prophylaxis. • To investigate the effect of randomness in Moreover, a cost-effectiveness analysis shows that clinical trials. Because there are substantial prophylaxis, especially with newer anti-emetics, not numbers of analgesic trials and they are only does not prevent any more people from vomit- usually performed using standard methods ing or feeling nauseated than treating established and including patients with similar entry nausea or vomiting, but exposes patients to con- criteria (moderate or severe pain), they may siderably more drug at considerably higher cost. be usefully studied to examine the effects of randomness in clinical trials. Variability between trials is large and understanding the effects Conclusions of chance would help to inform us of how large trials need to be to give an accurate This report has focused on two elements of clinical feel for a new drug. day-case care. It is clear that the economics of day-case work require that the vast majority of • To investigate how to minimise postoperative patients are fit to go home and that, once home, nausea and vomiting. This varies considerably they do not have to return to hospital or seek between trials and may be the result of random advice from the primary care team. To date, audits chance, but it is just as likely that components at local level have identified both pain and of the overall package of care other than nausea and vomiting as problems. Providing anaesthesia or anti-emetics are important. There adequate analgesia may be easier than guaran- is an obvious and important research agenda teeing minimal nausea and vomiting. The package here in understanding how best to minimise of care in day-case surgery needs to be revisited postoperative nausea and vomiting. However, regularly lest surgical interventions are the cause this is a complex area which will not easily of increased hospitalisation and returns. be understood.

iv Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 1 Introduction

simple way of looking at day-case surgery is that providing better pain control produced a A that both pain relief and postoperative nausea worthwhile reduction in both types of problem. and vomiting (PONV) should be well controlled. Patients should neither have to stay in hospital The authors hope that this report will enable a (because of poor pain relief or PONV) nor have higher quality of evidence to be used for future to contact healthcare professionals after they guidelines, for pain relief both in and out of have left hospital. hospital, and for prevention of PONV.

We have therefore concentrated on developing two league tables, or rank ordering. One is for Background and key questions which pain-killer or analgesic works best by mouth after surgery. This has produced surprises – the The rational approach to postoperative care is to use ‘standard’ take-home analgesic package often the highest quality evidence available and, in this contains analgesics which do poorly in the ranking. context, this comes from systematic reviews of valid Also, the efficacy of oral pain-killers has been randomised trials. The results will still have to be compared with injections of morphine or non- adapted to the circumstance of the individual but steroidal anti-inflammatory drugs (NSAIDs). The our chosen policies will be more discerning. Treat- net has been thrown quite wide and information ments which are simple, clinically appropriate and on transcutaneous electrical nerve stimulation evidence-based are focused on in this study. The (TENS) and injections of morphine into the knee authors were fortunate that there is now a steady joint and other peripheral sites has been included, supply of systematic reviews available in the pain together with information on whether the timing world3 (an updated listing is available on the Internet of the analgesic (before or after surgery) makes at ).

Producing this ranking has involved developing To achieve the best outcome it has been necessary new methods and extending existing ones. The to pick our way through the evidence. There is first part of the report documents the methods a complicated relationship between evidence, because, although the required information is guidelines, research and legal considerations, that in the league-table, the credibility of the and the patients’ outcomes as assessed by audit league table rests on the credibility of the (see Figure 1). methods used to derive it. High-quality postoperative care needs effective Similar methods have been used to tackle pain management. While we would like to believe treatment of PONV. A preliminary league-table of that we practice good pain control, a survey4 of various prophylactic measures is presented in the report. None of them appear to work particularly well. There is much less information on treatment (rather than prevention) of PONV but some Guidelines useful results are presented.

It is hard to be precise about the effect that poorly ResearchAudit cycle Legal controlled pain or poorly controlled PONV have (patient outcomes) on the incidence of patients having to stay in hospital, or on the incidence of consultations after leaving hospital. Ideal targets seem to be that less Postoperative care Marketing than 1% of patients should have to stay in, and less than 1% should have to consult.1 Audits show that poor pain control can certainly result in rates greater than 1% for both categories,2 and FIGURE 1 Influences on postoperative care 1 Introduction

TABLE 1 Inpatient survey4 • The next relates to prophylaxis versus treat-as-necessary: are the arguments for Number Percentage prophylaxis convincing? of patients • A third is whether or not injections of local Pain was present all or anaesthetic, with or without opioids, are most of the time 1042/3162 33 useful in the day surgery setting.

Pain was severe or moderate 2755/3157 87 For the future, we need to know why, in some patients, acute pain becomes chronic pain. Pain was worse than expected 182/1051 17 Can the patient swallow? Had to ask for drugs 1085/2589 42 Most postoperative pain is managed solely with medication. Perhaps because anaesthetists work Drugs did not arrive with injected drugs, there is a natural belief that immediately 455/1085 41 drugs which are injected are more powerful than drugs taken by mouth. 5150 recently discharged patients from 36 UK hospitals showed that, in the 3163 who responded Some of the questions which need to be to questions on pain, practice was far from answered include: ideal (Table 1). The questions asked are useful for audit. • which classes of drugs are the most effective postoperative analgesics? (or which are Postoperative care is more than a collection of least effective?) interventions. It is a package of care that needs • within a class of drugs, does the same dose work to be examined as a whole as well as in its parts. better when injected or when taken orally? Publications which analyse the process of postoperative care provision are rare, perhaps because It is important to know which oral analgesics to they attract few academic plaudits. There is good recommend to patients because so much post- evidence that the risk of adverse events is increased operative care is now in the home. It is biased to when high-technology approaches are used for think of patients after major surgery but they too drug administration.5 need oral analgesics when they can swallow. The evidence from trials in which drugs are compared Pain charts with placebo may be used to build a ranking of Pain charts used as part of normal practice will relative efficacy. improve quality of care.6,7 The fact of a chart is more important than its form, with pain measure- For this study, all the trials of a particular drug ments recorded at the same time as sedation, compared with placebo in postoperative pain were respiratory frequency and nausea, and as part obtained. The drug’s performance in the trials was of ongoing audit. An example is the Burford then converted into a common currency, viz. the chart.8 There are special scales for children.9 proportion of patients with moderate or severe postoperative pain who achieved at least 50% pain Acute pain services relief compared with placebo over 6 hours. One remedy for poor management is the provision of an acute pain service.10 The dispute about what The most effective drugs have a low number- should be provided ranges from a full ‘menu’, needed-to-treat (NNT) of about 2, meaning that including all the high-technology options,11 to for every two patients who receive the drug one a service limited to supervision of good practice will achieve at least 50% relief because of the guidelines for low-technology approaches and treatment (the other patient may obtain relief staff education.6,7 Training and education should but it does not reach the 50% level). The NNT is be the main tasks of an acute pain service. treatment-specific, which is useful for comparison of relative efficacy but, because these NNT com- Key questions parisons are against placebo, the best NNT of Several key questions for pain relief are obvious. 2 means that while 50 of 100 patients will get at least 50% relief because of the treatment, another • The simplest observation is that if patients 20 patients will have a placebo response which gives can swallow then the oral route should them at least 50% relief; hence, with ibuprofen 2 be preferred. 70 patients from 100 will have effective pain relief. Health Technology Assessment 1998; Vol. 2: No. 12

For paracetamol, 1 g, the NNT is 4 (see chapter 8). Pain which persists: prophylaxis or Combination of paracetamol with codeine, 60 mg, wait till it happens? improves the NNT to 3. Ibuprofen is better at 2. The intriguing questions in acute pain are The clear message is that, of the oral analgesics, the following. NSAIDs perform best and that paracetamol alone or in combination is also effective. The strongest • Is there a link between bad, acute postoperative oral analgesic regimen would be an oral NSAID pain and perseveration of this pain into a supplemented as necessary with paracetamol and chronic status? opioid. As pain wanes, then the prescription should • Can anything be done to prevent this? be paracetamol-based, supplemented if necessary by an NSAID. When used in day surgery, a regimen What remains unexplained is why some patients like this results in high-quality pain relief without end up with chronic pain after surgery when recourse to general practitioner visits.2 others do not. A simplistic explanation is that the chronic pain results from nerve damage at Even if patients can swallow, is it best to give surgery. An alternative explanation is that it is drugs by injection or suppository? those patients with severe postoperative pain There is no evidence that NSAIDs given rectally or who develop the chronic pain. The easy linkage by injection perform any better than the same drug is then to propose that if the acute pain was at the same dose given by mouth (see chapter 10); better controlled then chronic pain would two randomised, double-blind placebo-controlled not develop.21,22 comparisons of oral ibuprofen arginine, 400 mg, failed to distinguish any difference from intra- Pre-emptive short-term analgesia muscular ketorolac, 30 mg.12,13 These other non- The evidence for a clinical advantage in giving oral routes become appropriate when patients an intervention before pain as opposed to giving cannot swallow. the same intervention after pain is still uncon- vincing23 (see chapter 18). Certainly by far the The patient can’t swallow analgesics majority of trials of pre-pain versus post-pain There is, as yet, no information on the relative medication has failed to show any clinically efficacy of injected opioids or NSAIDs and regional meaningful benefit. anaesthetic techniques. However, some statements can be made: Peripheral opioids For intra-articular peripheral opioids at least, the • injecting morphine at a dose of 10 mg provides story becomes a little clearer. A systematic review similar analgesia to oral NSAID14 of valid trials of intra-articular morphine in knee • injecting morphine at doses of 10–20 mg surgery has shown that morphine in the knee joint provides similar analgesia to injected NSAID15 can indeed provide analgesia24 (see chapter 17). • injecting NSAID provides similar analgesia to This analgesia can continue for up to 24 hours, oral NSAID12,13 although there is no dose response. It is the long • injecting 20 mg of morphine provides greater duration of action which suggests this technique analgesia than injecting 10 mg, and greater might have practical application beyond its analgesia than the best performers in the oral research interest.25 league table.15 Postoperative nausea and vomiting Other techniques Managing PONV well is part of quality post- TENS operative care. Recent evidence from systematic TENS is not effective in postoperative pain16,17 reviews of PONV shows that prophylactic anti- (see chapter 19) and is of limited value in labour emetics are less effective than might have been pain18 (see chapter 20). hoped (see chapter 20). Use of propofol,26 omitting nitrous oxide,27 and different anaesthetic Psychological methods techniques28 all have similar efficacy, preventing There is evidence that psychological approaches about one patient vomiting for every four or five are beneficial.19 Cognitive behavioural methods can treated. This is also the case in high-risk settings reduce pain and distress in patients with burns. Pre- such as strabismus surgery.29 There is much less paration before surgery can reduce postoperative data for treatment of established PONV than for analgesic consumption. The evidence for the use prophylaxis but treatment with ondansetron of relaxation and music on postoperative pain is appears to have a similar level of efficacy confounded by the poor quality of trials.20 to prevention.30 3 Introduction

12. Laveneziana D, Riva A, Bonazzi M, Cipolla M, Conclusion Migliavacca S. Comparative efficacy of oral The availability of high-quality evidence gives a firm ibuprofen arginine and intramuscular ketorolac in patients with postoperative pain. Clin Drug Invest foundation for building better postoperative care. 1996;11:8–14. It enables informed decisions to be made about drugs and route of administration for individual 13. Pagnoni B, Vignali M, Colella S, Monopoli R, patients and services, both for pain and for nausea Tiengo N. Comparative efficacy of oral ibuprofen and vomiting. Bringing this together into an arginine and intramuscular ketorolac in patients efficient and effective service will be the challenge, with postcaesarean section pain. Clin Drug Invest 1996;11:15–21. so that audit or controlled trials can demonstrate the effectiveness of our postoperative care. 14. Mansfield M, Firth F, Glynn C, Kinsella J. A comparison of ibuprofen arginine with morphine sulphate for pain relief after orthopaedic surgery. References Eur J Anaesthesiol 1996;13:492–7. 15. Norholt S, SindetPedersen S, Larsen U, Bang U, 1. Millar JM, Rudkin GE, Hitchcock M. Practical Ingerslev J, Nielsen O, et al. Pain control after dental anaesthesia and analgesia for day surgery. Oxford: surgery: a double-blind, randomised trial of lornoxi- Bios, 1997. cam versus morphine. Pain 1996;67:335–43. 2. Haynes TK, Evans DEN, Roberts D. Pain relief after 16. Reeve J, Menon D, Corabian P. Transcutaneous day surgery: quality improvement by audit. J 1-day electrical nerve stimulation (TENS): a technology Surg 1995;(Summer):12–15. assessment. Int J Technol Assess Health Care 1996;12:299–324. 3. Jadad AR, McQuay HJ. Meta-analyses to evaluate 17. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. analgesic interventions: a systematic qualitative Randomization is important in studies with pain review of their methodology. J Clin Epidemiol outcomes: systematic review of transcutaneous 1996;49:235–43. electrical nerve stimulation in acute postoperative 4. Bruster S, Jarman B, Bosanquet N, Weston D, Erens pain. Br J Anaesth 1996;77:798–803. R, Delbanco TL. National survey of hospital 18. Carroll D, Tramèr M, McQuay H, Nye B, Moore patients. BMJ 1994;309:1542–6. A. Transcutaneous electrical nerve stimulation in labour pain: a systematic review. Br J Obstet 5. Bates DW, Cullen DJ, Laird N, Petersen LA, Small Gynaecol 1997;104:169–75. SD, Servi D, et al. Incidence of adverse drug events and potential adverse drug events. JAMA 19. Justins DM, Richardson PH. Clinical management 1995;274:29–34. of acute pain. Br Med Bull 1991;47:561–83. 20. Good M. Effect of relaxation and music on post- 6. Gould TH, Crosby DL, Harmer M, Lloyd SM, Lunn operative pain: a review. J Adv Nurs 1996;24:905–14. JN, Rees GAD, et al. Policy for controlling pain after surgery: effect of sequential changes in manage- 21. Eija K, Tiina T, Pertti NJ. Amitriptyline effectively ment. BMJ 1992;305:1187–93. relieves neuropathic pain following treatment of breast cancer. Pain 1996;64:293–302. 7. Rawal N, Berggren L. Organization of acute pain 22. Katz J, Jackson M, Kavanagh B, Sandler A. Acute services: a low-cost model. Pain 1994;57:117–23. pain after thoracic surgery predicts long-term post- 8. Burford Nursing Development Unit. Nurses and thoracotomy pain. Clin J Pain 1996;12:50–5. pain. Nurs Times 1984;18:94. 23. McQuay HJ. Pre-emptive analgesia: a systematic review of clinical studies. Ann Med 1995;27:249–56. 9. McGrath PJ, Ritchie JA, Unruh AM. Paediatric pain. In: Carroll D, Bowsher D, editors. Pain management 24. Kalso E, Tramèr M, Carroll D, McQuay H, Moore and nursing care. Oxford: Butterworth Heinemann, RA. Pain relief from intra-articular morphine after 1993: 100–23. knee surgery: a qualitative systematic review. Pain 1997;71:642–51. 10. Royal College of Surgeons of England and the 25. Kanbak NM, Akpolat N, Öcal T, Doral NM, Ercan College of Anaesthetists. Report of the working M, Erdem K. Intra-articular morphine adminis- party on pain after surgery. London: Royal College tration provides pain relief after knee arthroscopy. of Surgeons, 1990. Eur J Anaesthesiol 1997;14:153–6. 11. Ready LB, Oden R, Chadwick HS, Benedetti C, 26. Tramèr M, Moore A, McQuay H. Propofol Rooke GA, Caplan R, et al. Development of an anaesthesia and postoperative nausea and vomiting: anesthesiology based postoperative pain manage- quantitative systematic review of randomized 4 ment service. Anesthesiology 1988;68:100–6. controlled studies. Br J Anaesth 1997;78:247–55. Health Technology Assessment 1998; Vol. 2: No. 12

27. Tramèr M, Moore A, McQuay H. Omitting nitrous 29. Tramèr M, Moore A, McQuay H. Prevention of oxide in general anaesthesia: meta-analysis of vomiting after paediatric strabismus surgery: a intraoperative awareness and postoperative emesis systematic review using the numbers-needed-to- in randomized controlled trials. Br J Anaesth treat method. Br J Anaesth 1995;75:556–61. 1996;76:186–93. 30. Tramèr MR, Moore RA, Reynolds DJM, McQuay HJ. 28. Tramèr M, Moore A, McQuay H. Meta-analytic A quantitative systematic review of ondansetron in comparison of prophylactic anti-emetic efficacy for treatment of established postoperative nausea and postoperative nausea and vomiting. Propofol vomiting. BMJ 1997;314:1088–92. anaesthesia vs. omitting nitrous oxide vs. total i.v. anaesthesia with propofol. Br J Anaesth 1997;78:256–9.

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 2 Finding all the relevant trials

elevant and valid evidence is necessary for In practice, constrained by time and cost, reviewers R effective care. The randomised controlled trial have to compromise, and hope that what they have (RCT) is the most reliable way to estimate the effect found is a representative sample of the unknown of an intervention. The principle of randomisation total number of eligible trials. The more compre- is simple. With randomisation, patients taking part hensive the searching the more trials will be found, in a randomised trial have the same probability of and any conclusions will then be stronger. Compre- receiving any of the interventions being compared. hensive searching can be very time-consuming and Randomisation abolishes selection bias by prevent- costly; again, this emphasises the necessary ing the investigators influencing allocation of the compromise, where the target is the highest interventions. It also helps to ensure that other possible yield for given resources. factors, such as age or sex distribution, are equivalent for the different treatment groups. Inade- ‘Retrieval bias’ is the failure to identify reports quate randomisation, or inadequate concealment which could have affected the results of a syste- of randomisation, lead to exaggeration of matic review or meta-analysis.4 The failure may be therapeutic effect.1 because trials are still ongoing, or completed but as yet unpublished (publication bias) or because For reviews of evidence to be valid, they need to be although published the search did not find them. systematic; to be systematic, qualitative or quanti- Trying to identify unpublished trials by surveying tative, they need to include all relevant RCTs. researchers had a very low yield,5 and is not cheap. Identifying all relevant trials is a ‘fundamental Registers of ongoing and completed trials are challenge’,2 which is easily underestimated. another means of finding unpublished data but such registers are rare. The first obstacle faced by any reviewer is finding out how many eligible RCTs exist. Commonly the In this chapter we describe: total is unknown. Only for newer interventions are reviewers likely to be sure that they have found all • the methods used to identify eligible reports of relevant RCTs. Otherwise the total number of trials RCTs published from 1950 to date can only be identified by scanning every record in • information management. each of the available bibliographic databases, by searching manually all non-indexed journals, theses, proceedings and textbooks, by searching the Developing a citation database reference lists of all the reports found, and by asking the investigators of previous RCTs for other The process had three phases: definition of published or unpublished information (Figure 2).3 inclusion criteria, identification of reports, and information management.

Inclusion criteria A report was regarded as eligible if both the following criteria were fulfilled. yes old intervention whole area • Allocation of patients to the intervention was novel intervention widespread research single topic all reports described as randomised (no precise description few research groups limited time of the method of randomisation was required), published reports limited funds or as double-blind, or as both, or if it was no deadline unlimited funds suggested that the interventions were given at unlikely random and/or under double-blind conditions. • Analgesic interventions with pain or adverse effects as outcomes, and/or any intervention using pain as an outcome measure, FIGURE 2 Is a complete search possible? were compared. 7 Finding all the relevant trials

Reports were excluded which investigated TABLE 2 Refined high-yield MEDLINE search strategy analgesic effectiveness during (as opposed to after) diagnostic or surgical procedures. Step Request number

Identification of reports 1 PAIN* Details of the process are presented elsewhere.6 Since that publication, the major changes are in 2 explode PAIN / all subheadings in MeSH the use of other databases as well as MEDLINE. Searching EMBASE, the Cochrane Library, 3 ANALG* CINAHL and PsycLIT is now part of our standard 4 explode ANALGESIA / all subheadings operating procedures (see Figure 3). in MeSH

5 explode ANALGESICS / all subheadings in MeSH

searches on: 6 CLINICAL MEDLINE EMBASE 7 TRIALS Cochrane Library 8 CLINICAL TRIALS CINAHL RCT PsycLIT citation 9 EXPLODE CLINICAL-TRIALS / database all subheadings in MeSH handsearching references from other papers 10 RANDOM*

11 RANDOM-ALLOCATION (term allows no subheadings) in MeSH

FIGURE 3 Finding the citations 12 RANDOMIZED-CONTROLLED-TRIALS / all subheadings in MeSH

MEDLINE search for RCTs published 13 DOUBLE from 1966 to date The records identified by the optimised 14 BLIND MEDLINE search strategy (Table 2) were 15 DOUBLE BLIND downloaded (Biblio-Link version 1.1, Personal Bibliographic Software, Inc.) and transferred 16 DOUBLE-BLIND-METHOD (term allows no to a reference management program (Pro-Cite, subheadings) in MeSH Personal Bibliographic Software, Inc., version 17 META-ANALYSIS 2.1). The records were then sorted in alphabetical order and each downloaded record was checked 18 META-ANALYSIS (term allows no on-screen for definite eligibility, probable eligibility subheadings) in MeSH or ineligibility and coded accordingly within each Pro-Cite record. Hard copies of eligible 19 HUMAN and probable documents were obtained and, 20 (#1 or #2 or #3 or #4 or #5) and (HUMAN if necessary, translated, and eligibility was in MeSH) then confirmed. 21 HUMAN Hand-searching of journals published from 1950 to date 22 (#8 or #9 or #10 or #11 or #12 or #15 or A Pro-Cite file of all the records regarded as #16 or #17 or #18) and (HUMAN in MeSH) eligible and probably eligible (1950–90) was 23 #20 and #22 created. This file was used to produce a list of the 50 journals with the highest yield. These journals were then searched by hand to find RCTs. These RCTs, either missed by MEDLINE indexing, or in Management of the information non-indexed journals, were then added to the The citation database is maintained as a Pro-Cite citation database if perusal of the hard copy file. The number in that database is used as the 8 confirmed that they were indeed RCTs. unique identifying number for the hard copy. Health Technology Assessment 1998; Vol. 2: No. 12

Trends in the numbers of RCTs interventions, and 2% each for psychological in pain relief research published, and complementary treatments. 1950–90 Conclusions For 1956–80, there were twice as many reports of RCTs published in each successive 5-year The importance of basing systematic reviews on period. For 1980–90 the number of reports the highest quality evidence (randomised trials) is increased by more than 1000 per 5-year period. obvious from our experience in the pain field,7 and More than 85% of the reports identified were from the experience of others. This means that very published in the last 15 years. This is illustrated considerable time and effort has to be spent to by the trend in the number of RCTs published gather all the relevant material for each review. in the journal Pain over the past 20 years (Figure 4). The process described here gives an outline of what is a laborious task. The addition of another year’s A simple breakdown (Table 3) showed that 54% citations, maintaining the existing database (now of the reports were in acute pain, 43% in chronic 15,000 citations), and the associated chores are a non-cancer and 3% in chronic cancer. Pharmaco- full-time job. To make the information accessible to logical reports were commonest (75%), with 14% others we have contributed our citations of known classified as invasive, 7% as reports of physical RCTs to the Cochrane Library, and to the compilers of MEDLINE (National Library of Medicine), to ensure that all the RCTs found only by hand- Number of RCTs searching are tagged. 120 100 References 80 1. Schulz KF, Chalmers I, Hayes RJ, Altman DG. 60 Failure to conceal treatment allocation schedules in trials influences estimates of treatment effects. 40 Control Clin Trials 1994;15:63–4S. 20 2. Chalmers I, Dickersin K, Chalmers T. Getting to 0 grips with Archie Cochrane’s agenda. BMJ 1975–79 1980–84 1985–89 1990–94 1992;305:786–7. Publication years 3. Jadad AR, McQuay HJ. A high-yield strategy to identify randomized controlled trials for systematic reviews. Online J Curr Clin Trials 1993; Doc no 33. FIGURE 4 RCTs in the journal Pain from 1975 to 1994 4. Simes RJ. Confronting publication bias: a cohort TABLE 3 Pain trials database 1950–94 design for meta-analysis. Stat Med 1987;6:11–29. 5. Hetherington J, Dickerson K, Chalmers I, Meinert Acute Chronic Cancer Total % C. Retrospective and prospective identification of unpublished controlled trials: lessons from a survey Complementary 112 223 10 345 2 of obstetricians and pediatricians. Pediatrics 1989;84:374–80. Invasive 1697 336 34 2067 14 6. Jadad AR, Carroll D, Moore A, McQuay H. Pharmacological 5390 4978 337 10,705 75 Developing a database of published reports of randomised clinical trials in pain research. Pain Physical 402 501 36 939 7 1996;66:239–46. Psychological 100 191 10 301 2 7. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. Randomization is important in studies with pain Total 7701 6229 427 14,357 outcomes. Systematic review of transcutaneous %54433 electrical nerve stimulation in acute postoperative pain. Br J Anaesth 1996;77:798–803.

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 3 Judging the quality of the trials

nce all the reports of the trials relevant to What makes a trial worthy of the label ‘high O your question have been found, there is quality’? Quality could refer to the clinical rele- another stage in the process. This is to confirm vance of the study, to the likelihood of biased that, first, these reports meet certain quality results, to the appropriateness of the statistical standards and, second, even though a report may analysis, to the presentation of the data, or to the pass those quality standards, whether the trial is ethical implications of the intervention or to the valid. Imagine a situation where 40 reports of literary style of the manuscript. We consider that relevant trials were found. Twenty of the reports quality must primarily indicate the likelihood that say that the intervention is terrific, and 20 conclude the study design reduced bias. Only by avoiding that it should never be used. Delving deeper, the bias is it possible to estimate the effect of a given 20 ‘negative’ reports are found to score highly intervention with any confidence. on your quality standards scale, whereas the 20 ‘positive’ reports score poorly. The quality scale The purpose of our scale is to assess the likelihood should include measures of bias. Bias is the simplest of the trial design to generate unbiased results and explanation why poor quality reports give more approach the ‘therapeutic truth’. This has also positive conclusions than high quality reports. been described as ‘scientific quality’.17 Other trial characteristics, such as clinical relevance of the The quality standards which are required cannot be question addressed, data analysis and presentation, absolute, because for some clinical questions there literary quality of the report or ethical implications may not be any RCTs. Setting RCTs as a minimum of the study, are not included in our definition. absolute standard would therefore be inappropriate for some of the questions to which we might The aims of the scale are as follows. want answers. In the study of pain, however, there are two reasons for setting this high standard, and 1. To assess the scientific quality of any clinical requiring trials to be randomised. The first reason trial in which pain is an outcome measure is that there are, particularly for drug interventions, or in which analgesic interventions are quite a number of RCTs. The second, we would compared for outcomes other than pain argue, is that that it is even more important to (e.g. a study looking at the adverse effect stress the minimum quality standards of random- profile of different opioids). isation and double-blinding when the outcome 2. To allow consistent and reliable assessment of measures are subjective. quality by raters with different backgrounds, including researchers, clinicians, professionals This chapter describes briefly the development of a from other disciplines, and members of the quality scale which was then used for the systematic general public. reviews which follow. A detailed description of the way the scale was developed and tested has been The judges published.1 The chapter concludes with our A multidisciplinary panel of six judges was current views on this and other quality scales. assembled (a psychologist, a clinical pharma- cologist, a biochemist, two anaesthetists, and a research nurse), all with an interest in pain Developing and validating a research. The definition of quality and the quality scale purposes of the scale were discussed. Each judge then had to produce a list of suggested items to be Previous methods to measure the ‘quality’ of included on the scale. To generate the items, the clinical reports and incorporate the results in judges used both criteria published previously and systematic reviews may all be criticised because of their own judgement. The suggestions were then failure to define quality and because they were not combined in a single list of 49 items. validated.2–16 The danger is that using these scales might lead to conclusions in the review as incon- Using a modified nominal group approach to sistent and unreliable as the component studies. reach consensus,18 the judges assessed the face 11 Judging the quality of the trials

validity of each of the items, according to estab- without clinical or research experience in pain lished criteria.19 Items associated with low face relief, it should be possible to score the quality validity were deleted. An initial instrument was of research reports consistently. Our particular created from the remaining items. purpose was to allow differential analysis to be undertaken within our systematic reviews based The initial instrument was pre-tested by three on the quality of the individual primary studies – raters on 13 study reports and problems in clarity but the scale may have much wider use. and/or application of each of the items were identified. The panel of judges then modified the Chalmers suggested many years ago that the quality wording of the items accordingly and produced of clinical reports should be assessed blind.3 In our detailed instructions describing how each of the study, such blinded assessment was found to produce items should be assessed and scored. The items significantly lower scores. This may be very important were classified by their ability to reduce bias if absolute cut-off scores are imposed by systematic (direct or indirectly), and individual scores were reviewers, and if quality scores are used to weight allocated to them by consensus. The frequency the results of primary studies in subsequent meta- of endorsement, consistency and validity of each analysis.16,20 The results of open evaluations are good item were then assessed (see Figure 5). enough for busy readers. The improved reliability with blind testing is of more relevance to journal editors, for manuscript selection, and to systematic reviewers. Quality scales without clinimetric 6 judges Variables considered (49 items) evaluation have already been used in pain work to Consensus meeting 11,13,14 (criteria for face validity) support the conclusions of systematic reviews. Pilot instrument (11 items) None of the items on the scale are specific to pain 14 raters studies. The three items are very similar to the 36 reports components of a scale used extensively to assess the effectiveness of interventions during pregnancy and childbirth,8 and also appear in most other Frequency of endorsement Inter-observer agreement scales. Control of selection bias and rater bias is Construct validity obviously regarded as crucial to quality. Refined instrument (3 items) Selection bias is best controlled by allocating patients at random to the different study groups. Each patient should have the same probability of FIGURE 5 Developing the scale being included in each comparison group, and the allocation should be concealed until after the Final version of the scale patient has consented to take part. Methods of The final version of the scale contains the three allocation based on alternation, date of birth or items with highest frequency of endorsement (see hospital record number cannot be regarded as Table 4). Advice on using the scale is presented in random. Failure to secure proper randomisation Table 5, and the method of scoring RCTs is shown increases the likelihood that potential participants in Table 6. in a ‘randomised’ study will be admitted to the study selectively because of prior knowledge of Open versus blind assessments the group to which they would be allocated or A chastening finding during the development of excluded selectively before formal admission to the scale was that blind assessment (not knowing the study.21 Ideal methods of randomisation are authors, journal, year, etc.) of reports produced those in which individuals with no direct relation- significantly lower and more consistent quality ship to the study participants are in charge of the scores than open assessments.1 This has important allocation (e.g. allocation by telephone from a implications, because the cost of organising truly central coordinating office, concealed from the blind assessment is very considerable. investigators). Appropriate simpler alternatives are coin tossing, tables of random numbers and numbers generated by computers, but these Comments on the scale carry a higher risk of selective selection.

The three-point scale is simple, short, valid and All of these methods are regarded as appropriate 12 reliable. The results suggest that even for those for the purposes of our scale, although we are Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 4 Scale (3-point) to measure the likelihood of bias in pain research reports

This is not the same as being asked to review a paper. It should not take more than 10 minutes to score a report and there are no right or wrong answers.

Please read the article and try and answer the following questions (see attached instructions).

1. Was the study described as randomised (this includes the use of words such as randomly, random and randomisation)? 2. Was the study described as double-blind? 3. Was there a description of withdrawals and drop-outs?

Scoring the items Award a score of 1 point for each ‘yes’ and 0 points for each ‘no’. There are no in-between marks.

Give 1 additional point if: On question 1, the method of randomisation was described and it was appropriate (table of random numbers, computer-generated, coin tossing, etc.)

and/or if: On question 2, the method of double-blinding was described and it was appropriate (identical placebo, active placebo, dummy, etc.)

Deduct 1 point if: On question 1, the method of randomisation was described and it was inappropriate (patients were allocated alternatively, or according to date of birth, hospital number, etc.)

and/or if: On question 2, the study was described as double-blind but the method of blinding was inappropriate (e.g. comparison of tablet vs. injection with no double-dummy)

TABLE 5 Advice on using the scale

1 Randomisation If the word randomised or any related words such as random, randomly, or randomisation are used in the report but the method of randomisation is not described, give a positive score to this item. A randomisation method is regarded as appropriate if it allows each patient the same chance of receiving each treatment but investigators could not predict which treatment was next. Methods of allocation using date of birth, date of admission, hospital numbers or alternation should not be regarded as appropriate.

2 Double-blinding A study must be regarded as double-blind if the term double-blind is used (even without describing the method) or if it is implied that neither care giver nor patient could identify the treatment being assessed.

3 Withdrawals and drop-outs Patients included in the study but who did not complete the observation period or who were not included in the analysis must be described.The number and the reasons for withdrawal must be stated. If there are no withdrawals, it should be stated. If there is no statement on withdrawals, a negative score (0 points) must be given. aware that selective selection is still possible TABLE 6 Scoring RCTs (maximum 5, minimum 1) even when the group allocation is concealed until after consent has been obtained. The Score randomisation method is rated as inappropriate Randomised? if the potential participants did not have the yes 1 same chance of being included in any of the comparison groups (methods based on date of Appropriate? birth, hospital number or alternation). Even with yes (table) 1 no (alternative) –1 excellent randomisation, selection bias may still be introduced if biased and selective withdrawals Double-blind? and drop-outs occur after the allocations have yes 1 22 been made. This is why an adequate description Appropriate? of withdrawals and drop-outs is included in the yes (double-dummy) 1 scale. With this information it is possible to analyse no –1 on an intention-to-treat basis (that is, all those randomised whether or not they were exposed Withdrawals described? to the study interventions23). yes 1 13 Judging the quality of the trials

Rater bias can be minimised by blinding the person 4. Tyson JE, Furzan JA, Reisch IS, Mize SG. An receiving the intervention, the individual adminis- evaluation of the quality of therapeutic studies in tering it, the investigator measuring the outcome perinatal medicine. J Pediatr 1983;102:10–13. and the analyst. Blinding can be tested by asking 5. Andrew A. Method for assessing of the reporting the study patients and the researchers which inter- standard of clinical trials with roentgen contrast vention they had. This is not often done. The usual media. Acta Radiol Diag 1984;25:55–8. ‘best’ level of blinding is of the study subject and 6. Evans M, Pollock AV. A score system for evaluating those making the observations (double-blinding). random control clinical trials of prophylaxis of Double-blinding is often achieved by using control abdominal surgical wound infection. Br J Surg interventions with similar physical characteristics 1985;72:256–60. to those of the intervention under evaluation, or 7. Poynard T. Evaluation de la qualité méthodologique by the use of dummies when two or more inter- des essais thérapeutiques randomisés. Presse Med ventions have to be given by different routes. 1988;17:315–18.

Sometimes, however, one of the interventions 8. Chalmers I, Hetherington J, Elbourne D, Keirse MJNC, Enkin M. Materials and methods used in may produce effects which make blinding very synthesizing evidence to evaluate the effects of care difficult to sustain. Then the use of active placebos during pregnancy and childbirth. In: Chalmers I, or active controls may decrease the likelihood of Enkin M, Keirse MJNC, editors. Effective care in rater bias. All these precautions are relatively easy pregnancy and childbirth. Oxford: Oxford to achieve in pharmaceutical studies. In non-drug University Press 1989;38–65. studies, testing under blind conditions is either 9. Gøtzsche PC. Methodology and overt and hidden difficult or inappropriate (e.g. surgical procedures) bias in reports of 196 double blind trials of non- or impossible (e.g. acupuncture or TENS). The steroidal anti-inflammatory drugs in rheumatoid risk of rater bias limits the confidence with which arthritis [published erratum appears in Control Clin conclusions can be reached. Studies which are Trials 1989;50:356]. Control Clin Trials 1989;10:31–56. not double-blind are known to risk an average 10. Reisch JS, Tyson JE, Mize SG. Aid to the evaluation exaggeration of treatment effect of 17% (Box 1).24 of therapeutic studies. Pediatrics 1989;84:815–27. 11. ter Riet G, Kleijnen J, Knipschild P. Acupuncture BOX 1 Effect of randomisation and blinding and chronic pain: a criteria-based meta-analysis. on treatment effect24 J Clin Epidemiol 1990;43:1191–9.

Overestimation 12. Brown SA. Measurement of quality of primary studies for meta-analysis. Nurs Res 1991;40:352–5. Inadequate Unclear 13. Koes BW, Bouter LM, Beckerman H, van der Randomisation: concealment Heijden, Knipschild PG. Physiotherapy exercises of treatment allocation 41% 30% and back pain: a blinded review [see comments]. BMJ 1991;302:1572–6. Blinding 14. Koes BW, Assendelft WJ, van der Heijden, Bouter If not double-blind 17% LM, Knipschild PG. Spinal manipulation and Is this important in every setting? (unconscious) mobilisation for back and neck pain: a blinded Is this practical in every setting? (surgery) review [see comments]. BMJ 1991;303:1298–303. 15. Detsky AS, Naylor CD, O’Rourke K, McGeer AJ, l’Abbé KA. Incorporating variations in the quality of individual randomized trials into meta-analysis. References J Clin Epidemiol 1992;45:255–65. 1. Jadad AR, Moore RA, Carroll D, Jenkinson C, 16. Nurmohamed MT, Rosendaal FR, Buller HR, Reynolds DJM, Gavaghan DJ, et al. Assessing the Dekker E, Hommes DW, Vandenbroucke JP, et al. quality of reports of randomized clinical trials: is Low molecular-weight heparin versus standard blinding necessary? Control Clin Trials 1996;17:1–12. heparin in general and orthopaedic surgery: a meta-analysis. Lancet 1992;340:152–6. 2. Reiffenstein RJ, Schiltroth AJ, Todd DM. Current standards in reported drug trials. Can Med Assoc J 17. Oxman AD, Guyatt GH. Validation of an index 1968;99:1134–5. of the quality of review articles. J Clin Epidemiol 1991;44:1271–8. 3. Chalmers TC, Smith H, Blackburn B, Silverman B, Schroeder B, Reitmand D, et al. A method for 18. Fink A, Kosecoff J, Chassin M, Brook RH. Consensus assessing the quality of a randomized control methods: characteristics and guidelines for use. 14 trial. Control Clin Trials 1981;2:31–49. Am J Public Health 1984;74:979–83. Health Technology Assessment 1998; Vol. 2: No. 12

19. Feinstein AR. Clinimetrics. Newhaven: Yale 22. Sackett DL, Gent M. Controversy in counting and University Press, 1987. attributing events in clinical trials. N Engl J Med 1979;301:1410–12. 20. Fleiss JL, Gross AJ. Meta-analysis in epidemiology, with special reference to studies of the association 23. Peto R, Pike MC, Armitage P, Breslow NE, Cox DR, between exposure to environmental tobacco smoke Howard SV, et al. Design and analysis of randomised and lung cancer: a critique. J Clin Epidemiol clinical trials requiring prolonged observation of 1991;44:127–39. each patient. 1. Introduction and design. Br J Cancer 1976;34:585–612. 21. Chalmers I. Evaluating the effects of care during pregnancy and childbirth. In: Chalmers I, Enkin M, 24. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Keirse MJNC, editors. Effective care in pregnancy Empirical evidence of bias: dimensions of methodo- and childbirth. Oxford: Oxford University Press, logical quality associated with estimates of treatment 1989;1–37. effects in controlled trials. JAMA 1995;273:408–12.

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 4 Pain measurement, study design and validity

he efficacy of analgesic interventions is judged • pain intensity – none = 0, mild = 1, moderate = 2, T by the change they bring about in the patient’s severe = 3 report of pain. A brief description of methods of • pain relief – none = 0, slight = 1, moderate = 2, pain measurement follows. In the second part of good or lots = 3, complete = 4. this chapter, the problems and some solutions when using pain measurement data for systematic Data from different subjects is then combined reviews are discussed. to produce means (rarely medians) and measures of dispersion (usually standard errors of means). The validity of converting categories into numerical Pain measurement scores was checked by comparison with concurrent VAS measurements. Good correlation was found, Pain is a personal experience which makes it especially between pain relief scales using cross- difficult to define and measure. It includes modality matching techniques.2–4 Results are both the sensory input and any modulation by usually reported as continuous data, mean or physiological, psychological and environmental median pain relief or intensity. Few studies factors. Not surprisingly, there are no objective present results as discrete data, giving the number measures – there is no way to measure pain directly of participants who report a certain level of pain by sampling blood or urine, or by performing intensity or relief at any given assessment point. neurophysiological tests. Measurement of pain must The main advantages of the categorical scales therefore rely on recording the patient’s report. The are that they are quick and simple. However, the assumption is often made that because the measure- limited number of descriptors may force the ment is subjective it must be of little value. The scorer to choose a particular category when reality is that, if the measurements are made prop- none of them describes the pain satisfactorily. erly, remarkably sensitive and consistent results can be obtained. There are contexts, however, in which The VAS, a line with left-hand end labelled it is not possible to measure pain at all, or when ‘no relief of pain’ and right-hand end labelled reports are likely to be unreliable. These include ‘complete relief of pain’ (Figure 6), seem to over- impaired consciousness, young children, psychiatric come this limitation. Patients mark the line at the pathology, severe anxiety, unwillingness to cooper- point which corresponds to their pain. The scores ate, and inability to understand the measurements. are obtained by measuring the distance, usually Such problems are deliberately avoided in trials. in millimetres, between the no relief end and the patient’s mark. The main advantages of VAS are Measurement scales that they are simple and quick to score, avoid Most analgesic studies include measurements of imprecise descriptive terms and provide many pain intensity and/or pain relief, and the common- points from which to choose. More concentration est tools used are categorical and visual analogue and coordination are needed, however, which scales (VAS). can be difficult postoperatively or with neurological disorders. Categorical and visual analogue scales Categorical scales use words to describe the Pain relief (PR) scale magnitude of the pain and were the earliest pain measure.1 The patient picks the most appropriate NO COMPLETE relief relief word that describes a pain. Most research groups of pain of pain use four words (none, mild, moderate and severe). Scales to measure pain relief were developed later. Pain intensity (PI) scale The most common is the five category scale (none, LEAST WORST slight, moderate, good or lots, and complete). possible possible pain pain For analysis purposes, numbers are given to the verbal categories: FIGURE 6 Visual analogue scales 17 Pain measurement, study design and validity

Pain relief scales are perceived as more con- nn venient than pain intensity scales, probably ∑ ∑ because patients have the same baseline relief SPID = PIDt TOTPAR = PRt (zero) whereas they could start with different t = 0–6 t = 0–6 baseline intensity (usually moderate or severe). Relief scale results are thus easier to compare. They may also be more sensitive than intensity Where at the tth assessment point (t = 0, 1, 2,..., n), 4,5 scales. A theoretical drawback of pain relief Pt and PRt are pain intensity and pain relief mea- scales is that the patient has to remember what sured at that point, respectively, P0 is pain intensity the pain was like to begin with. at t = 0 and PIDt is the pain intensity difference calculated as (P0 – Pt). Other tools Verbal numerical scales and global subjective These summary measures reflect the cumulative efficacy ratings are also used. Verbal numerical response to the intervention. Their disadvantage is scales are regarded as an alternative or comple- that they provide no information about the onset mentary to the categorical and VAS scales. Patients and peak of the analgesic effect. If onset or peak choose a number for the pain intensity or relief are important, then time to maximum pain relief (for pain intensity, 0 usually represents no pain (or reduction in pain intensity) or time for pain to and 10 the maximum possible, and for pain relief, return to baseline are necessary. 0 represents none and 10 complete relief). They are very easy and quick to use, and correlate well with conventional VASs.6 Using pain measurement data for systematic reviews Global subjective efficacy ratings, or simply global scales, are designed to measure overall treatment Standardising the summary measures performance. Patients are asked questions such as, The method used to standardise TOTPAR values, ‘How effective do you think the treatment was?’ and derived from a categorical verbal rating scale of answer using a labelled numerical or a categorical relief (CATPR), is shown in Figure 7. The actual scale. Although these judgements probably include TOTPAR value is divided by the maximum possible adverse effects they can be the most sensitive in TOTPAR score (maximum duration in hours discriminating between treatments. One of the multiplied by the maximum pain relief score) oldest scales was the binary question, ‘Is your pain and converted to a percentage. half gone?’. It has the advantage that it has a clearer clinical meaning than a 10 mm shift on a VAS. The This calculation presumes that categorical relief disadvantage, for the small trial intensive measure score data is available. One major problem we pundits at least, is that all the potential intermediate faced is that not all trials use this classic scale. In information (1–49% or > 50%) is discarded.

Analgesic requirements (including patient- maxTOTPAR controlled analgesia, PCA), special paediatric 4 scales, and questionnaires (such as McGill) are also used. The limitation to guard against is that 3 they usually reflect other experiences as well as CATPR 2 or instead of pain.7 1 TOTPAR

Judgement by the patient rather than by the carer 0 is the ideal. Carers tend to overestimate the pain 02 4 6 relief compared with the patient. Time (hours)

Analysis of scale results – Categorical verbal rating scale pain relief (CATPR): summary measures 0 = none, 1 = slight, 2 = moderate, 3 = good, 4 = complete In the research context, pain is usually assessed TOTPAR before the intervention is made and then on x 100 = % maxTOTPAR maxTOTPAR multiple occasions. Ideally, the area under the time–analgesic effect curve for the intensity (sum of pain intensity differences, SPID) or relief (total FIGURE 7 Calculating percentage of maximum possible pain 18 pain relief; TOTPAR) measures is derived. relief score Health Technology Assessment 1998; Vol. 2: No. 12

order to include trials which used different scales, scale and a VAS pain intensity scale. The distribution ways of converting those different scales back to the of the VAS scores was checked for 736 patients re- common denominator of % maxTOTPAR had to porting moderate pain and for 344 reporting severe be developed. The development and validation of pain. The VAS scores corresponding to moderate or these methods is discussed in chapter 5. We still severe pain were also checked by gender. cannot include trials which use analgesic drug consumption (e.g. PCA), or trials which use non- Baseline VAS scores for patients reporting standard scales. moderate pain were significantly different from those of patients reporting severe pain (see Table 7 The hazard for meta-analysis is that if many papers and Figure 8). Of the patients reporting moderate have to be discarded because they do not use stand- pain, 85% scored over 30 mm on the correspond- ard scales, are the remaining trials representative? ing VAS, with a mean score of 49 mm. For those For most drug interventions this has not proved reporting severe pain, 85% scored over 54 mm with to be a major problem, because the majority of a mean score of 75 mm. There was no difference trials used standard methods. However, there are between the corresponding VAS scores for men exceptions. For some older drugs, such as dihydro- and women. These results indicate that if a patient codeine, remarkably few trials were found which records a baseline VAS score in excess of 30 mm used standard methods. For academically inspired they would probably have recorded at least investigations, as opposed to trials required for moderate pain on a four-point categorical scale. drug registration, many trials which use non- standard methods have had to be excluded. TABLE 7 Descriptive statistics for the distribution of VAS pain Restricting to moderate and severe intensity scores initial pain intensity Baseline Pain The trail blazers of analgesic trial methodology found that if patients had no pain to begin with, it (using a 4-point categorical scale) was impossible to assess analgesic efficacy, because Moderate Severe there was no pain to relieve. To optimise trial sensitivity, a rule was developed – only those patients n 736 344 with moderate or severe pain intensity at baseline would be studied. Patients with mild pain or no Mean (mm) 49 75 pain would not. Standard deviation 17 18

This study has stayed true to this rule; trials of a Median (mm) 49 76 given intervention have been excluded if the trials 90% patients > (mm) 26 49 studied patients with mild or no initial pain. As with exclusions because of non-standard methods, there 85% patients > (mm) 30 54 have been few pharmacological trials where the rule on baseline pain has led to exclusion but for pre-emptive techniques and local anaesthetic blocks it has been a major problem. Study design and validity

How do you know what is moderate or Pain measurement is one of the oldest and most severe pain on a pain intensity VAS? studied of the subjective measures, and pain scales The usual criterion to ensure adequate sensitivity for have been used for over 40 years. Even in the early analgesic trials is to test the intervention on patients days of pain measurement there was understanding who have established pain of moderate to severe that the design of studies contributed directly to intensity. When a VAS is the only pain measure in the validity of the result obtained. Trial designs a trial we need to know what point on it represents which lack validity produce information that is, moderate pain, so that the trial can be included in a at best, difficult to use and, at worst, useless. meta-analysis with an inclusion criterion of baseline pain of at least moderate intensity. Placebo People in pain respond to placebo treatment. Some In order to answer this question, individual patient patients given placebo obtain 100% pain relief (see data from 1080 patients from RCTs of various anal- chapter 5). The effect is reproducible, and some gesics were used.8 Baseline pain had been measured work has been done to try and assess the character- using both a four-point categorical pain intensity istics of the ‘placebo responder’ by gender, race 19 Pain measurement, study design and validity

paracetamol) can be shown to have statistical difference from placebo, then the analgesic assay Categorical scale – moderate pain should be able to distinguish another analgesic of Number of patients 100 similar effectiveness. Alternatively, two different doses of a standard analgesic (such as morphine) could be used – showing the higher dose to be 75 statistically superior to the lower dose again provides confidence in the sensitivity of the assay. 50 Failure to demonstrate sensitivity in one assay invalidates the results from that particular assay. 25 However, the results could still be included in meta-analysis. 0 0 102030405060 70 80 90 100 Equivalence Baseline VAS PI score Studies of analgesics of an A versus B design are notoriously difficult to interpret (Figure 9). If there Categorical scale – severe pain is a statistical difference, then that suggests sensi- Number of patients tivity. Lack of a significant difference (Figure 9, top 100 graph) means nothing – there is no method by which to determine whether there is an analgesic 75 effect which is no different between A and B, or whether the assay lacks the sensitivity to measure a difference that is actually present. 50 This is not just a problem for pain studies.9,10 25 Designs which minimise these problems include using two doses of a standard analgesic plus placebo to establish sensitivity (Figure 9, middle and 0 bottom graphs). Simple calculations could show 0 102030405060 70 80 90 100 what dose of the new analgesic was equivalent to Baseline VAS PI score the usual dose of the standard analgesic.

Problems FIGURE 8 Frequency distributions of initial VAS pain intensity The correct design of an analgesic trial is situation- scores by initial categorical verbal pain intensity dependent. In some circumstances, very complex designs have to be used to ensure sensitivity and psychological profile. None has succeeded but and validity. women are known to respond better than men to some analgesics, getting more analgesia from the No gold standard same plasma concentration of drug. There may be circumstances in which there is no established analgesic treatment of sufficient RCT effectiveness to act as a gold standard against Because the placebo response was an established which to measure a new treatment, as is often the fact in analgesic studies, randomisation was used case in chronic pain. Clearly, the use of placebo early in studies to try to avoid any possibility of bias or no-treatment controls is of great importance, from placebo responders, and to equalise their especially when effects are to be examined over numbers in each treatment group. This was true prolonged periods of weeks or months. even in studies without placebo, since an excess of placebo responders in an active treatment arm of a However, paradoxically, it is these very circum- study might inflate the effects of an analgesic. stances in which ethical constraints act against using placebo or non-treatment controls because Sensitivity of the need to do something. In acute pain studies, Particularly for a new analgesic, a trial should prove conversely, there are few problems with using its internal sensitivity – that is that the study is an placebos, since the failure of a placebo (or any adequate analgesic assay. This can be done in several treatment) can be dealt with by prescribing 20 ways. For instance, if a known analgesic (such as additional analgesics which should work. Health Technology Assessment 1998; Vol. 2: No. 12

When there is no pain to begin with Clearly, where there is no pain it is difficult to Equivalence: danger of A vs. B negatives measure an analgesic response. Yet a number of studies seek to do this by pre-empting pain, or by Was there no difference? Was there a difference which was missed? using an intervention where there is no pain (intraoperatively, for instance) to produce analgesia when pain is to be expected. A These are difficult but not impossible circumstances in which to conduct research. Meticulous PR B attention to trial design is necessary to be able to demonstrate differences.

Time References 1. Keele KD. The pain chart. Lancet 1948;2:6–8. Protection – negative control (placebo) 2. Scott J, Huskisson EC. Graphic representation of pain. Pain 1976;2:175–84. 3. Wallenstein SL, Heidrich IG, Kaiko R, Houde RW. A Clinical evaluation of mild analgesics: the measurement of clinical pain. Br J Clin Pharmacol 1980;10:319–27S. B 4. Littman GS, Walker B.R., Schneider BE. PR Reassessment of verbal and visual analogue ratings in analgesic studies. Clin Pharmacol Therap Placebo 1985;38:16–23. 5. Sriwatanakul K, Kelvie W, Lasagna L. The quantification of pain: an analysis of words used Time to describe pain and analgesia in clinical trials. Clin Pharmacol Ther 1982;32:141–8. Protection – positive control (active) 6. Murphy DF, McDonald A, Power C, Unwin A, MacSullivan R. Measurement of pain: a comparison Standard high dose of the visual analogue with a nonvisual analogue scale. Clin J Pain 1988;3:197–9. 7. Jadad AR, McQuay HJ. The measurement of pain. A In: Pynsent P, Fairbank J, Carr A, editors. Outcome measures in orthopaedics. Oxford: Butterworth B PR Heinemann, 1993:16–29. 8. Collins SL, Moore RA, McQuay HJ. The visual analogue pain intensity scale: what is moderate Standard low dose pain in millimetres? Pain 1997;72:95–7. 9. Jones B, Jarvis P, Lewis JA, Ebbutt AF. Trials to assess Time equivalence: the importance of rigorous methods. BMJ 1996;313:36–9. 10. McQuay H, Moore A. Placebos are essential when FIGURE 9 Using placebo or active comparators to protect extent and variability of placebo response are against A vs. B negative results unknown. BMJ 1996;313:1008.

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 5 Estimating relative effectiveness

inor analgesics, such as paracetamol, questions the blindness of these trials if this result M ibuprofen and combinations with opiates like was correct, and elegantly dissects the areas where codeine or dextropropoxyphene, are used often to ‘leakage’ of blinding can occur (patient–patient, treat pain. There are few direct comparisons, one patient–doctor, patient–nurse). with another, but most trials contain a placebo, which has the potential to be the universal com- Both these observations call into question the parator. Instead of measuring relative effectiveness validity of the methods used to gather the data. through multiple comparisons of different drugs, If the methods are faulty, how reliable are the it should be possible to compare the absolute answers? Therefore, in the first part of this chapter effectiveness of analgesics against placebo. the nature of the variation in placebo responses in five randomised, double-blind, parallel-group trials Some aspects of clinical trial methods relating in postoperative pain is examined, together with to the placebo response in clinical trials of single the relationship of the variation to the analgesic doses of analgesics using classical methods are effectiveness of the active treatments. examined in this chapter. The ways in which data can be extracted from published studies for use Methods in meta-analysis are then determined. Individual patient data was used from five placebo- controlled double-blind RCTs, performed over a 10-year period by the Pain Research Group in Placebo responses in Oxford, in which the analgesic effects of various 5–9 analgesic trials drugs in postoperative pain were investigated. All were randomised, double-blind and parallel-group The placebo response is confusing.1 Two common trials of single doses of drugs given orally. Random- misconceptions are that a fixed fraction (one-third) isation was made using random number tables. of the population responds to placebo and that the Drugs were prepared outside the hospital in which extent of the placebo reaction is also a fixed frac- the studies were done. Treatment codes were not tion (again about one-third of the maximum possi- broken until the studies were finished. All drugs ble2). As Wall points out, these ideas stem from a used within a study were identical. Drugs were misreading of Beecher’s work of 40 years ago.1 given in a standardised way by the nurse observer. The methods used by the trained nurse observers In Beecher’s five acute pain studies, 139 patients to measure pain were identical. Patients were asked (31%) of 452 given placebo had 50% or more relief a standardised battery of questions in a fixed order of postoperative pain at two checked intervals.3 The at each assessment point in the studies. All patients proportion of patients who had 50% or more relief knew that a placebo was one of several possible of pain varied across the studies, ranging from 15% treatments. All patients had moderate or severe to 53%. There was neither a fixed fraction of pain within 72 hours of their operations, and all responders, nor a fixed extent of response. were aware that they could withdraw from the study at any time for any reason. Placebo responses have also been reported as varying systematically with the efficacy of the active Each study used five scales for pain; three for analgesic medicine. Evans pointed out that in seven pain intensity and two for pain relief. Of these studies the placebo response was always about 55% the five-point CATPR scale for pain relief (0 = of the active treatment, whether that was aspirin or none, 1 = slight, 2 = moderate, 3 = good, 4 = morphine: the stronger the drug, the stronger the complete) was chosen for this analysis because it placebo response.4 was closest to Beecher’s original method. For each patient the area under the curve of pain relief Randomised, double-blind trials are meant to (categorical scale) against time was calculated eliminate (or at least minimise) both selection (TOTPAR). The percentage of the maximum and observer bias; Evans’ observation suggests possible for this summary measure was then that significant observer bias occurs. Wall1 rightly calculated (% maxTOTPAR).10 23 Estimating relative effectiveness

Percentage 100 Mean y = 0.543x – 0.139, r2 = 0.765 Median y = 0.120x + 2.300, r2 = 0.278

75 Placebo (% maxTOTPAR)

50 30

20 0 <10 <20 <30 <40 <50 <60 <70 <80 <90 <100 % maxTOTPAR 10

FIGURE 10 Distribution of % maxTOTPAR scores for the 130 patients given placebo ( ) and for the 395 patients given active drugs ( ) 0 0204040 Active (% maxTOTPAR) Results In the five trials 130 patients had a placebo. Individual patients’ scores with placebo varied from 0 to 100% of the maximum possible pain relief. FIGURE 11 % maxTOTPAR placebo and active scores for five trials The distribution of these % maxTOTPAR scores is shown in Figure 10. In the five trials, 395 patients The same pattern of results was also found when were given active drugs. Individual patient scores the analysis was repeated using the results from with different active drugs varied from 0% to 97% the VAS for pain relief. of the maximum possible pain relief. The distribution of these % maxTOTPAR scores for the Comment active drugs is shown in Figure 10. The variation of the placebo response in the acute pain setting found by Beecher some 40 years ago is The mean % maxTOTPAR scores for the five confirmed by these results. Using the dichotomous placebo groups varied from 11% to 29%, and the measure of greater than 50% pain relief at 45 and mean scores for the active drugs varied from 12% 90 minutes, Beecher found that a range of 15–53% to 49%. The relationship between the mean scores of patients given placebo had better than 50% for the active drugs and the mean placebo scores relief in five acute pain studies.3 Here, using the is shown in Figure 11. Mean placebo scores were derived dichotomous measure of 50% maximum related to the mean score for the active drugs in pain relief, a range of 7–37% of patients given each trial such that the higher the active score, the placebo achieved better than 50% relief across higher the placebo score. A similar relationship the five studies (see Table 8). obtained for the best and for the worst active drug from each of the five trials. On average, the mean In analgesic trials the response of a group of placebo results were 54% of the mean active results patients to a treatment is usually described not based on a slope of 0.54; 95% confidence intervals as a dichotomous variable (like the proportion (CIs) around the slope, 0.03–1.08. of patients with at least 50% relief), but rather as a continuous variable (the mean extent of the The relationship between the median scores for response). The common description of pain active and placebo treatments is also shown in intensity difference or pain relief is thus as the Figure 11. There was little relationship between mean with standard deviations or standard the two and, on average, the median placebo score errors of the mean, as if the data were was less than 10% of the median active drug score. normally distributed. The slope to the regression line was 0.12 (95% CIs, –0.24, 0.48) and included no relationship between Patient responses were not normally distributed, placebo response and extent of the response to either for patients given placebo or for those given 24 active analgesic. an active treatment (Figure 10). The predominant Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 8 Results with placebo; mean (SD), median (interquartile range) and number of patients with > 50% of % maxTOTPAR in the five studies

Study No. Mean % Median % Number of patients % of patients with maxTOTPAR maxTOTPAR with > 50% of at least 50% of (SD) (interquartile range) % maxTOTPAR % maxTOTPAR

Porter, et al.5 21 11.9 (19.3) 3.1 (16.4) 2 10

Evans, et al.6 30 29.4 (29.1) 14.0 (53.0) 11 37

McQuay, et al.7 19 20.1 (29.1) 3.1 (27.3) 4 21

McQuay, et al.8 30 10.7 (17.8) 2.1 (8.3) 2 7

McQuay, et al.9 30 16.9 (21.2) 8.3 (25.0) 2 7

group was that getting less than 10% of maximum treatment, and the investigators knew the relief – 62% of patients given placebo and 37% study design and active treatments; it has been of those given an active treatment. In these suggested that this can modify patients’ behav- circumstances, the use of a mean as a description iour.11,12 A small number of patients may have had is not valid, and the use of a median is more opportunities to communicate with each other. sensible. Averaging results to describe them Doctors who knew the trial design obtained con- is a historical hangover. sents from patients, and this may also be a source of bias.13 The nurse observer spent most time with In describing the placebo groups, therefore, the the patients, but in standardised situations. This range of mean placebo response of 11–29% of would be the most likely source of bias, as the nurse maximum (Table 8) becomes a range of median might be able to influence a patient’s response placebo response of 2–14% and a range of the by his/her demeanour based on experience of proportion of patients with at least 50% of % other patients’ reactions. That would produce maxTOTPAR of 7–37%. Regressing median time-dependent changes in study results that placebo response against median active response have been observed before.14 from the same five trials yielded a poor correlation, with a regression line no different from the hori- Bias may still occur but its effects are slight, and zontal, which would be the expected result if there this has important consequences. It means that was no bias. The idea that there is a constant results obtained over a range of clinical conditions relationship between active analgesic and placebo and times may be combined in meta-analyses with response is therefore an artefact of using an confidence. Gøtzsche has confirmed similar magni- inappropriate statistical description. tudes of effect for NSAIDs in active and placebo- controlled studies,15 showing that the presence of It is the comparison of the mean data from placebo a placebo does not affect the active treatment – and active treatments that led to the observation4 the alternative hypothesis. that placebo is about 55% as effective as an active treatment, whatever active treatment is used. In the five trials here, comparison of the mean placebo Deriving dichotomous outcome response with the mean active treatment (Figure 11) measures from continuous data produced a regression with a slope of 0.54 – exactly in RCTs of analgesics the same result! The problem is that, in most published trial This defies logic unless there was considerable bias reports, the only value available which describes despite randomisation and the use of double-blind the magnitude of analgesic effect is the mean and methods, and would, if true, undermine the con- standard deviation of the SPID or TOTPAR. Is it fidence placed in analgesic trial results. But is possible, then, to use this to generate other, more it true? useful data with which meta-analysis can work with confidence? Meta-analytic outcomes using mean Randomisation controls for selection bias, and the values from different trials have been explored,16,17 double-blind design is there to control observer but the result is a complicated analysis which is bias. Patients knew a placebo was one possible not intuitively accessible to doctor or patient. If 25 Estimating relative effectiveness

individual patient information was available from Methods for converting mean to dichotomous every RCT of analgesics, dichotomous data could data from clinical trials of analgesics, given in be extracted for NNT calculations. The reality is single doses using classical analgesic methodology, that individual patient data is not available, so that have been determined in three stages, all of the problem is how to derive dichotomous out- which use at least 50% maxTOTPAR as a final comes from the published mean data. A full version dichotomous outcome. of these arguments is published elsewhere.18 Stage 1 Use of Oxford data from about 1500 A proposed solution patients combined with mathematical The hypothesis was examined that, in modelling using TOTPAR scales. pharmacological interventions in acute pain: Stage 2 Verification with an external data set of 3500 patients using TOTPAR scales. (i) a relationship exists between the descriptive Stage 3 Examination of the use of other scales. mean value for pain relief and a dichotomous description of the same data set For each stage, methods and results are shown (ii) knowing this relationship allows the separately and then discussed together. conversion of descriptive mean values for pain relief into dichotomous data that can be used Stage 1 methods with confidence for meta-analysis. Actual patient data Individual patient data were taken from 12 placebo- Relationships that exist between treatment group and active-controlled double-blind randomised means and some simple extractable variables from trials in which the analgesic effects of various drugs a known data set are an obvious place to start. What in postoperative pain were investigated.5–9,19–26 The is required as an extractable variable is a single trials were undertaken over a 15-year period by the value, for instance, the proportion or number of Pain Research Group, Oxford. Complete individual patients who have achieved 50% pain relief. If patient information over 4 or 6 hours was available treatment group means reliably predict the pro- for a number of pain and pain relief scales. All portion with 50% pain relief, this suggests that the drugs were administered orally, except sublingual relationship between the two variables is a product buprenorphine20 and intramuscular opioids.26 of the underlying distribution. One benefit of using the proportion of patients who have achieved at All the studies were randomised, double-blind and least 50% pain relief is that it is clinically intuitive. parallel-group. Patients were told about the study by the nurse observer the day before surgery. Informed The robustness of such a relationship can be tested consent was obtained by the doctor that evening. in various ways. The gold standard would be to Random number tables were used for random- test relationships between mean and dichotomous isation. Drugs were prepared outside the hospital in variables developed from one set of trials using which the studies were undertaken. Treatment codes data from other trials; however, this has not were not broken until the studies were completed. proved possible. All drugs within a study were identical in appearance and double-dummy methods were used when differ- In the absence of available information from real ent routes of administration were compared. Drugs trials, surrogate trials can be obtained through were given in a standardised way by the nurse simulation. Simulation methods have been used to observer. The methods used by the trained nurse generate individual patient data for large numbers observers to measure pain were identical. of trials using the underlying distribution from randomised trials of pharmacological interventions Patients were asked a standardised battery of performed in Oxford over about 15 years using questions in a fixed order at each assessment time standard methods. This approach generates in the studies. In placebo-controlled trials, all precision in defining the underlying distribution patients knew that placebo was one possible treat- of the data, and tests the assumptions made in ment. All patients had moderate or severe pain deriving the technique for converting mean pain within 72 hours postoperatively, and all were aware relief data into dichotomous data. that they could withdraw from the study at any time for any reason. At the start of the assessments the While simulation methods can give a degree of nurse observer made sure that patients had recov- confidence that the general approach has validity, it ered sufficiently from the anaesthetic and were able is testing against other, real, data sets that will allow to communicate reliably. Studies with more than 26 the method to be used in meta-analysis. one nurse observer were block randomised, with Health Technology Assessment 1998; Vol. 2: No. 12

one nurse responsible for each block. Only with relatively small patient numbers, simulations one nurse assessed any one patient. If no pain were conducted to test how robust the relationships relief was obtained from the test medication developed with actual treatments and real patients by 1 hour, or if the pain intensity subsequently might be. The simulations had three main aims, reverted to the initial value before the end of as follows. the 6-hour study, patients were given analgesia (‘escape analgesia’). 1. To generate a very large number of simulated active treatments (10,000) with a mean of Each study used five scales for pain; three for pain 30 simulated patients (standard deviation, intensity and two for pain relief. In this study, the 3 patients, minimum group size 15 patients) categorical measurement of pain relief with a in each, where the % maxTOTPAR for each five-point categorical verbal rating scale (CATPR: simulated patient was generated randomly 0 = none, 1 = slight, 2 = moderate, 3 = good, from a distribution similar to the real data. 4 = complete) was used because it has been Comparable results from real and simulated shown that with this scale placebo responses data would allow the conclusion that the are independent of active treatment efficacy.27 conversion technique was dependent only on the amalgamated distribution of For each patient, the area under the curve of % maxTOTPAR from all trials, and pain relief (categorical scale) against time was not on the underlying distribution of calculated (TOTPAR). The percentage of the % maxTOTPAR within each trial. maximum possible for this summary measure was 2. To show that, for each simulated treatment, then calculated (% maxTOTPAR), as well as the mean % maxTOTPAR could be converted numbers and proportion of each group with at to the calculated number with at least 50% least 50% maxTOTPAR (or percentage > 50% maxTOTPAR, using the techniques developed maxTOTPAR to accommodate unequal group for the 45 actual treatments, and, for these sizes). The dichotomous descriptor of at least 50% simulated treatments, to compare the calcu- maxTOTPAR was chosen because it is a simple lated number with at least 50% maxTOTPAR clinical endpoint of pain half relieved, easily with the number generated in the simulation. understood by professionals and patients. This would provide an indication of how accurate the conversion technique was The relationship between the mean % likely to be for a large data set with this maxTOTPAR and the actual number of patients underlying distribution. with at least 50% maxTOTPAR was examined by 3. To generate simulated individual patient data linear regression analysis. Using the equation to the using two different underlying distributions regression line, the calculated number of patients (normal distribution and a uniform distri- with at least 50% maxTOTPAR was then compared bution, ensuring in each case that the mean with the actual number. was similar to that for the real data), in order to test the extent to which the accuracy of the Odds ratios and their 95% CIs were calculated from conversion technique was dependent on the standard formulae incorporating a fixed-effects underlying distribution. model and NNTs, using the method of Cook and Sackett.28 Where the same treatment (placebo or Computer codes were written in Fortran and run active) had been given in different trials, data from on the Oxford University DEC Vax Cluster. Uni- individual treatment arms were combined. form random numbers in the range 0–1 (U [0,1]) were obtained using the intrinsic function ‘ran’, Simulations and these were then used to calculate both random The underlying distribution using % maxTOTPAR treatment sizes and individual patient data with for individual real patients in the actual 45 treat- the appropriate underlying distribution, as ments was approximately uniform over the range described below. 10–100% of % maxTOTPAR, with a spike in the range 0–10% of % maxTOTPAR. This was an (i) Treatment sizes were assumed to be normally amalgamation of patient data from all the treat- distributed with a mean of 30 and a standard ments and was unlikely to reflect the actual deviation of 3. These were calculated by trans- distribution within any one treatment. forming the U [0, 1] values into normal values with the required mean and standard deviation Because the possibility exists that statistical differ- using the Box–Mueller algorithm.29 If any ences in distribution could occur in treatment arms generated value of the group size was below 27 Estimating relative effectiveness

15 it was discarded and a new value generated Calculated number of patients with at least which fell within the appropriate range. 50% maxTOTPAR (ii) For generation of the ‘simulated actual’ The actual and calculated numbers of patients in distribution, the U [0, 1] value generated was each group with at least 50% maxTOTPAR are first multiplied by 140 (giving a U [0, 140] shown in Table 9. The equation to the regression distribution), but for any values greater than line was: 100 the value was discarded and a new value generated which was multiplied by 10. This Calculated number of patients > 50% maxTOTPAR process ensured that 50/140 (36%) patients = 0.93 actual + 0.93 (r2 = 0.88) were uniformly distributed in the range 0–10% maxTOTPAR while the remaining In 36 of 45 treatments, the agreement between 64% were uniformly distributed in the range actual and calculated was within two patients; in 42 10 to 100. Standard techniques were then used of 45, agreement was within three patients and in 43 to show that a distribution generated in this of 45, agreement was within four patients. The two way had a theoretical mean of 37.1 and a most aberrant results occurred in the same trial.22 standard deviation of 31.7. (iii) For generation of the ‘normal’ distribution, Simulated actual distribution – mean and the Box–Mueller algorithm was again used to proportion with at least 50% maxTOTPAR generate the appropriate values but, in this A simulated distribution, similar to that of the case, it was necessary to restrict generated actual data (‘simulated actual’ distribution) was values to the range 0–100% maxTOTPAR. used to produce 10,000 simulated treatments. This Since this restriction process altered the mean generated a regression of mean % maxTOTPAR of the underlying distribution, the appropriate against percentage > 50% maxTOTPAR which was values to be used in the simulation to give a very similar to that obtained for the actual data mean of 37 were determined by iteration. from 45 treatments: (iv) For the generation of the ‘uniform’ distribution, the value U [0, 1] was multiplied by Percentage of patients > 50% maxTOTPAR 74.0 to obtain a distribution which was = 1.34 mean % maxTOTPAR – 14.1 (r2 = 0.79) uniform on [0, 74], with a mean of 37. Simulated actual distribution – calculated Stage 1 results numbers > 50% maxTOTPAR The actual trials used in the analysis, the treatments This equation was used to obtain the calculated used, numbers in each group, mean % maxTOTPAR percentage > 50% maxTOTPAR, which was then and the numbers of patients with at least 50% regressed against the actual percentage of patients maxTOTPAR are shown in Table 9. The calculated > 50% maxTOTPAR. The equation to the regres- number of patients in each treatment group with at sion line was very similar to that obtained for the least 50% maxTOTPAR was derived from 45 actual actual data from 45 treatments: treatments, using the relationship between mean % maxTOTPAR and percentage > 50% maxTOTPAR. Calculated number of patients > 50% maxTOTPAR Mean % maxTOTPAR for each study was entered = 0.82 actual + 1.92 (r2 = 0.83) into the equation to the regression to derive the proportion with more than half relief. This From Table 10 it can be seen that, using the under- proportion was then combined with the number of lying distribution, the difference between calcu- patients to generate the actual number of patients in lated and actual number of patients with at least each group predicted to have better than 50% relief. 50% maxTOTPAR was 0–2 in 90% of the simulated Numerical values were rounded up or down to the studies and, in 99%, it was in the range 0–3. These nearest integer. results are very similar to those obtained with the actual data and, again, this suggests strongly that Actual mean and proportion with at least provided the underlying actual amalgamated 50% maxTOTPAR distribution is a reasonable reflection of the The relationship between mean % maxTOTPAR assumed ‘true’ underlying distribution of pain and proportion with at least 50% maxTOTPAR is relief, then the conversion technique is accurate shown in Figure 12. The equation to the regression and robust. line was: Normal and uniform distributions Percentage of patients > 50% maxTOTPAR In order to test the effect of different underlying 28 = 1.41 % maxTOTPAR – 14.1 (r2 = 0.89) distributions, the process of obtaining the number Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 9 Studies

Study Treatment (N) Mean % Actual number Calculated number maxTOTPAR with > 50% with > 50% maxTOTPAR maxTOTPAR

Evans, et al. 19826 paracetamol, 650 mg, + dextropropoxyphene, 65 mg (30) 46.0 18 15 placebo (30) 29.4 11 8 zomepirac, 100 mg (30) 38.4 12 12 zomepirac, 50 mg (30) 49.4 19 17

McQuay, et al. 198621 paracetamol, 500 mg (30) 31.0 8 9 ketorolac, 5 mg (30) 39.5 11 12 ketorolac, 10 mg (30) 47.0 16 15 ketorolac, 20 mg (30) 54.0 18 19 paracetamol, 1000 mg (30) 41.9 12 14

McQuay, et al. 19878 aspirin, 650 mg (30) 23.4 7 6 fluradoline, 150 mg (30) 17.3 7 3 fluradoline, 300 mg (30) 26.9 8 7 placebo (30) 10.7 2 0

Porter, et al. 19815 bicifadine, 100 mg (19) 12.0 1 1 bicifadine, 150 mg (20) 17.8 2 2 placebo (21) 11.9 2 1 codeine, 60 mg (20) 25.0 4 4

McQuay, et al. 19909 bromfenac, 5 mg (30) 25.4 6 6 bromfenac, 10 mg (30) 38.9 14 12 bromfenac, 25 mg (30) 46.3 15 15 placebo (30) 16.9 2 3 paracetamol, 1000 mg (30) 32.9 10 10

Carroll, et al. 199320 bromfenac, 10 mg (23) 58.6 16 16 bromfenac, 25 mg (21) 46.4 13 11 buprenorphine, 0.2 mg (22) 21.7 1 4 buprenorphine, 0.4 mg (24) 35.5 9 9

Bullingham, et al. 198119 paracetamol, 1000 mg (30) 51.7 17 18 paracetamol, 1000 mg + buprenorphine, 1.0 mg (30) 47.8 18 16 paracetamol, 1000 mg + buprenorphine, 1.5 mg (29) 54.9 15 18 paracetamol, 1000 mg + buprenorphine, 2.0 mg (30) 50.8 16 17

McQuay, et al. 19857 dihydrocodeine, 30 mg (18) 40.6 8 8 placebo (19) 20.1 4 3 zomepirac, 100 mg (18) 47.4 11 9

McQuay, et al. 199224 paracetamol, 1000 mg, + codeine, 16 mg, + caffeine, 60 mg (30) 39.1 10 12 ibuprofen, 400 mg + codeine, 25.6 mg (30) 54.0 21 19

McQuay, et al. 199325 dihydrocodeine, 30 mg (41) 28.7 9 11 dihydrocodeine, 60 mg (43) 32.8 13 14 ibuprofen, 400 mg (40) 60.0 31 28

continued 29 Estimating relative effectiveness

TABLE 9 contd Studies

Study Treatment (N) Mean % Actual number Calculated number maxTOTPAR with > 50% with > 50% maxTOTPAR maxTOTPAR

McQuay, et al. 198923 ibuprofen, 400 mg (23) 44.8 10 11 ibuprofen, 400 mg + codeine, 20 mg (24) 57.7 15 16

McQuay, et al. 198722 aspirin, 500 mg + paracetamol, 500 mg (47) 36.6 13 18 aspirin, 500 mg, + paracetamol, 500 mg, + codeine, 13.6 mg (48) 34.8 8 17

McQuay, et al. pethidine, 100 mg (21) 16.1 1 2 unpublished26 meptazinol, 100 mg (20) 21.8 1 3 morphine, 15 mg (22) 24.4 4 4

of patients with at least 50% maxTOTPAR was repeated for two further distributions. The results y = 1.41x – 14.1, r2 = 0.89 obtained for the ‘normal’ and ‘uniform’ distri- 80 butions (Table 10) were less accurate. Even so, these levels of agreement indicate that the conversion technique is robust, even with these gross differ- 60 ences in underlying distribution, and suggest that it will be very robust to the smaller differences likely to be encountered in practice.

40 NNTs NNTs were calculated for paracetamol, 1000 mg, zomepirac, 100 mg, bromfenac, 10 mg, bromfenac, 20 25 mg, dihydrocodeine, 30 mg, ibuprofen, 400 mg, and ibuprofen, 400 mg, plus codeine, 24.6 mg; for

Proportion with > 50% maxTOTPAR Proportion these, and for placebo, there was information from 0 at least two trials (Table 11). NNT values derived 10 20 30 40 50 60 70 from the actual and the calculated data, as well as Mean % maxTOTPAR odds ratios and CIs, were very similar or identical.

Single treatment arms from the individual reports FIGURE 12 Relationship between mean % maxTOTPAR and were combined to obtain odds ratio estimates with proportion with at least 50% maxTOTPAR 95% CIs using a fixed effects model and to derive

TABLE 10 Accuracy of the conversion in actual and simulated treatments

Difference between actual 45 actual Simulated actual Simulated normal Simulated uniform and calculated numbers treatments distribution distribution distribution (%) (%) (%) (%)

≤ 1 57.7 60.3 21.3 40.7

≤ 2 82.1 90.8 44.4 71.2

≤ 3 93.2 98.8 68.2 89.0

≤ 4 95.4 99.8 87.1 97.1

≤ 5 97.7 100 96.5 99.3 ≤ 30 6 97.7 100 99.4 99.9 Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 11 Numbers-needed-to-treat

Treatment Active > 50% NNT Odds ratio maxTOTPAR/Total (95% CI) (95% CI)

Dihydrocodeine, Actual 17/59 7.9 (3.9–∞) 2.2 (1.0–4.7) 30 mg Calculated 19/59 4.8 (3.0–13.3) 4.0 (1.8–9.0)

Paracetamol, Actual 39/90 3.7 (2.6–6.6) 3.9 (2.1–7.1) 1000 mg Calculated 42/90 2.9 (2.1–4.3) 6.2 (3.4–11.4)

Zomepirac, Actual 23/48 3.2 (2.1–6.1) 5.5 (2.5–11.7) 100 mg Calculated 21/48 3.1 (2.1–5.8) 7.3 (3.2–16.6)

Bromfenac, Actual 30/53 2.5 (1.8–3.9) 7.4 (3.6–15.1) 10 mg Calculated 28/53 2.4 (1.8–3.7) 9.8 (4.6–20.8)

Bromfenac, Actual 28/51 2.6 (1.9–4.2) 7.0 (3.4–14.6) 25 mg Calculated 26/51 2.5 (1.8–4.1) 9.4 (4.3–20.3)

Ibuprofen, Actual 41/63 2.0 (1.6–2.8) 9.3 (4.9–17.7) 400 mg Calculated 39/63 2.0 (1.6–2.7) 12.0 (6.2–23.4)

Ibuprofen, 400 mg, plus Actual 36/54 2.0 (1.6–2.7) 10.5 (5.3–20.8) codeine, 24.6 mg Calculated 35/54 1.9 (1.5–2.5) 14.6 (7.1–29.6)

NNTs for analgesic effectiveness.28 At least two within 3 hours of study drug administration, those identical treatments from different trials were needing sedatives during the observation period required. Of the 130 patients who received and those with known contraindications or placebo, 21 actually had at least 50% maxTOTPAR medical conditions which might interfere and 15 were calculated to have at least 50% with observations. maxTOTPAR. The following drugs were given as single oral doses: placebo (695 evaluable patients); codeine, 60 mg Verification from independent data (649); tramadol, 50 mg (409); tramadol, 75 mg (281); tramadol, 100 mg (468); tramadol, 150 mg Stage 2 methods (279); tramadol, 200 mg (50); aspirin, 650 mg, plus Individual patient data from 18 primary RCTs was codeine, 60 mg (305); and paracetamol, 650 mg, made available by Grünenthal GmbH, Aachen, plus propoxyphene, 100 mg (316). Germany, and Robert Wood Johnson Pharma- ceutical Research Institute, Spring House, PA, Patients were given the study drug if they had USA (RWJ). moderate or severe pain on a four-point categorical scale (0 = no pain, 1 = slight, 2 = moderate, 3 = Study protocols for postsurgical pain (including severe). Thereafter observations were made at gynaecological procedures) and pain due to the 30 minutes, and at 1, 2, 3, 4, 5 and 6 hours after extraction of impacted third molars were essentially administration. Pain intensity was measured using identical. Trials were double-blind, single-dose, the same categorical scale, together with a five- parallel-group studies; randomisation was by com- point categorical scale of pain relief (0 = no relief, puterised random-number generation, stratified 1 = a little, 2 = some, 3 = a lot, 4 = complete). Time on pretreatment pain. Criteria for patient selection of repeat medication was also recorded, together were moderate or severe pain, and that the with a global assessment of therapy (excellent, patient’s condition was appropriate for manage- very good, good, fair or poor) at the final evalu- ment with a centrally-acting analgesic or paraceta- ation. At repeat medication, pain relief scores mol combined with centrally-acting analgesics. reverted to zero and pain intensity scores to the Patients ages ranged from 18 years to 70 years. initial value; adverse event recording, but not pain Patients had to be cooperative, reliable, and evaluations, continued after repeat medication. motivated, and be able to take oral medication. Exclusion criteria included patients with mild For each patient the area under the curve of pain or no pain, those who had taken analgesic drugs relief (categorical scale) against time (TOTPAR) 31 Estimating relative effectiveness

was calculated for 6 hours after the study drug was procedures). Studies involved between 21 and given. The percentage of the maximum possible for 58 patients in each treatment (mean 40 patients). this summary measure was then calculated for each The distributions of % maxTOTPAR for all active patient (% maxTOTPAR10), mean TOTPAR was and all placebo patients in these groups are shown calculated for all patients in each treatment arm in Figure 13. and the number of patients on each treatment achieving at least 50% maxTOTPAR was noted. The relationship between actual and calculated numbers of patients with at least 50% maxTOTPAR The mean TOTPAR value was then used to in each treatment arm is shown in Figure 14, and calculate the theoretical number of patients with the equation to the regression line for this is at least 50% maxTOTPAR, using a relationship compared with 45 treatments from trials in established in clinical trials of analgesics in Oxford Oxford in Table 12, using both the relationship with 1283 patients with 45 treatments (percentage for the actual data and that from a 10,000 of patients with at least 50% maxTOTPAR = 1.41 x treatment simulation. mean % maxTOTPAR – 14.1). Actual and calculated numbers were then compared using Of the 85 treatment arms, 80 (94%) were within unweighted linear regression analysis. four patients per treatment and 74 (87%) within three (Table 13). These proportions are comparable Stage 2 results to those obtained previously for actual and simu- Individual patient information was available from lated treatments (see Table 10). Summing the over 3400 patients in 85 different treatment arms positive and negative differences between actual in nine studies involving dental surgery (mostly and calculated numbers of patients with at least third molar extraction) and nine involving general 50% maxTOTPAR gave an average difference of postoperative pain (including gynaecological 0.30 patients per treatment arm.

Postsurgical placebo Postsurgical active drug 160 450 140 400 120 350 300 100 250 80 200 Count Count 60 150 40 100 20 50 0 0 0 20406080100 0 20406080100 % maxTOTPAR % maxTOTPAR

Dental placebo Dental active drug 300 800 250 700 600 200 500 150 400

Count Count 300 100 200 50 100 0 0 0 20406080100 0 20406080100 % maxTOTPAR % maxTOTPAR

32 FIGURE 13 Distributions of % maxTOTPAR for active treatments and placebo in dental and postsurgical pain Health Technology Assessment 1998; Vol. 2: No. 12

Comparison of (actual – calculated, irrespective of sign) numbers of patients with at least 50% max- y = 0.94x + 0.33, r2 = 0.89 TOTPAR as percentages for the 45 actual treatments 40 and 10,000 simulated treatments using the simulated actual, normal and uniform distributions are shown 18 30 in Table 13. Cumulative percentages are shown at different levels of agreement and the final column adds the 85 treatment arms from the RWJ trials. 20 Combining the 85 treatments in this data set with the earlier 45 treatments18 produced a new 10 relationship for use in future conversions:

0 Proportion of patients with > 50% maxTOTPAR = 1.33 x mean % maxTOTPAR – 11.5 (r2 = 0.89)

Calculated number with > 50% maxTOTPAR Calculated number –10 0 10203040Use of pain intensity and VAS Actual number with > 50% maxTOTPAR Stage 3 methods Data for the study were from individual patient FIGURE 14 Relationship between actual and calculated data from 13 RCTs (1283 patients with 45 treatments, 18 number with at least 50% maxTOTPAR for 3400 patients Oxford data ) and 18 RCTs (3453 patients with 87 with 85 treatments treatments, RWJ data) described in stages 1 and 2.

TABLE 12 Regression equations for calculated and actual number of patients in each treatment with > 50% maxTOTPAR

Study Slope Intercept Coefficient of determination (r2)

45 treatment arms from RCTs in Oxford [1] 0.93 0.93 0.88

45 treatment arms from RCTs in Oxford [2] 0.82 1.92 0.83

85 treatment arms from RWJ RCTs (3453 patients) [3] 0.94 0.33 0.89

Results for solutions to the equation: Calculated = (Actual x slope) + intercept, using the relationship between % > 50% maxTOTPAR and mean % maxTOTPAR derived from: [1] 45 actual treatments (Oxford RCTs) [2] a 10,000 treatment arm simulation [3] 85 actual treatments (RWJ RCTs)

TABLE 13 Accuracy of the conversion in actual and simulated treatments

Difference between 45 actual Simulated Simulated Simulated 85 RWJ actual and calculated treatments actual normal uniform actual numbers (Oxford) distribution distribution distribution treatments (%) (%) (%) (%) (%)

≤ 1 57.7 60.3 21.3 40.7 50.6

≤ 2 82.1 90.8 44.4 71.2 70.6

≤ 3 93.2 98.8 68.2 89.0 87.1

≤ 4 95.4 99.8 87.1 97.1 94.1

≤ 5 97.7 100 96.5 99.3 96.6 ≤ 6 97.7 100 99.4 99.9 98.8 33 Estimating relative effectiveness

For each patient the SPID was calculated for categorical pain intensity, and the equivalent VAS A % maxSPID (4713 patients) SPID. For each individual patient the 4-hour or 6-hour SPID was divided by the maximum possible < –35 SPID; for example, a patient with a SPID of 6 and to –20 initial pain intensity of 3 would have a theoretical to –5 maximum SPID of 18, and the % maxSPID would to 10 be 33%. The area under the curve of pain relief to 25 against time was calculated for the categorical (TOTPAR) and VAS TOTPAR scales. The percent- to 40 age of the maximum possible for each summary to 55 10 measure was then calculated for each patient. to 70 Rules for calculation included that in the event of to 85 repeat medication within 6 hours, pain relief scores reverted to zero and pain intensity scores to their to 100 initial value. The mean summary measure for all 0 500 1000 1500 2000 patients in each treatment arm was calculated. The Number of patients number of patients on each treatment achieving at least 50% maxTOTPAR was noted. B % maxVAS SPID (1059 patients) The relationship between the mean % maxSPID, < –35 % maxVAS SPID and % maxVAS TOTPAR and the to –20 actual number of patients with at least 50% max- to –5 TOTPAR was examined by linear regression analysis. Using the equation to the regression line, the to 10 calculated number of patients with at least 50% to 25 maxTOTPAR was then compared with the actual to 40 number using unweighted linear regression analysis. to 55 Stage 3 results to 70 Individual patient scores for categorical pain to 85 intensity and VAS pain intensity and pain relief to 100 were asymmetrically distributed, much as was seen for TOTPAR. 0 50100 150 200 250 300 Number of patients Categorical pain intensity scale Data were available from 132 treatments with C % maxVAS TOTPAR (1082 patients) 4713 patients. Individual patient distribution < 10 of % maxSPID was asymmetric (Figure 15). Linear regression analysis performed for the Oxford and to 20 RWJ data sets separately showed similar relation- to 30 ships, so the data sets were combined for all to 40 132 treatments. to 50

Results from Oxford and RWJ data sets and the com- to 60 bined data for the regression of number of patients to 70 per treatment with at least 50% maxTOTPAR against to 80 mean %maxSPID (with 95% CI). to 90 to 100 For all 132 treatments the regression line was: 0 50100 150 200 250 300 Percentage with > 50% maxTOTPAR Number of patients = 1.36 mean % maxSPID – 2.3 (r2 = 0.85)

There was good agreement between the actual FIGURE 15 Distribution of individual patient scores for number of patients with at least 50% maxTOTPAR categorical pain intensity and VAS pain intensity and pain 34 in each treatment arm and the calculated number relief scores Health Technology Assessment 1998; Vol. 2: No. 12

using the relationship derived with % maxSPID (Figure 16): A. 40 y = 0.862x + 1.373, r2 = 0.859 Calculated number with > 50% maxTOTPAR = 0.86 actual + 1.37 (r2 = 0.86) 30 For 92% of treatments the actual and the calculated numbers with at least 50% maxTOTPAR 20 were within four patients per treatment. Agree- ment (actual – calculated) was normally distributed around zero (Figure 17). Summing the

> 50% maxTOTPAR 10 positive and negative differences between actual Calculated number with Calculated number and calculated numbers of patients with at least 50% maxTOTPAR gave an average difference 0 of –0.03 patients per treatment arm. 0 10203040 VAS pain intensity Actual number with > 50% maxTOTPAR Data were available from 40 treatments within the Oxford data set with 1059 patients. Individual B. 40 patient distribution of % maxVAS SPID was asym- 2 y = 0.897x + 1.190, r = 0.789 metric (Figure 15B). The regression line between percentage with > 50% maxTOTPAR and mean 30 % maxVAS SPID was given by:

Percentage with > 50% maxTOTPAR 20 = 1.18 mean % maxVAS SPID – 2.2 (r2 = 0.87)

There was good agreement between the

> 50% maxTOTPAR 10

Calculated number with Calculated number actual number of patients with at least 50% maxTOTPAR in each treatment arm and the calculated number using the relationship 0 derived from % maxVAS SPID (Figure 16B): 0 10203040

Actual number with > 50% maxTOTPAR Calculated number with > 50% maxTOTPAR = 0.90 actual + 1.19 (r2 = 0.79) C. 40 y = 0.891x + 1.154, r2 = 0.808 For 95% of treatments the actual and the calculated numbers with at least 50% maxTOTPAR were 30 within four patients per treatment. Summing the positive and negative differences between actual and calculated numbers of patients with at least 20 50% maxTOTPAR gave an average difference of –0.23 patients per treatment arm.

> 50% maxTOTPAR 10 VAS pain relief Calculated number with Calculated number Data were available from 40 treatments with 1082 patients. Individual patient distribution of % maxVAS 0 TOTPAR was asymmetric (Figure 15C). The regression 0 10203040 line between percentage with > 50% maxTOTPAR and mean % maxVAS TOTPAR was given by: Actual number with > 50% maxTOTPAR Percentage > 50% maxTOTPAR = 1.15 mean % maxVAS TOTPAR – 8.51 (r2 = 0.81) FIGURE 16 Correlation of actual and calculated numbers of patients with at least 50% maxTOTPAR in each treatment for There was good agreement between the actual calculations using categorical pain intensity and VAS pain intensity number of patients with at least 50% maxTOTPAR and relief scores in each treatment arm and the calculated number 35 Estimating relative effectiveness

using the relationship derived from % maxVAS in each treatment (Figure 16A), and the sum of the TOTPAR (Figure 16C): difference over all 132 treatments was –0.03 patients per treatment, with the differences distributed nor- Calculated number with > 50% maxTOTPAR mally around zero (Figure 17). This is firm evidence = 0.89 actual + 1.15 (r2 = 0.81) for the reliability of the conversion method.

For 95% of treatments the actual and the calcu- Only 40 treatments from the Oxford data set were lated numbers with at least 50% maxTOTPAR were available for calculating relationships between within four patients per treatment. Summing the patients with at least 50% maxTOTPAR and mean positive and negative differences between actual % maxVAS SPID and mean % maxVAS TOTPAR. and calculated numbers of patients with at least Despite this, the agreement between actual and 50% maxTOTPAR gave an average difference of calculated numbers with at least 50% maxTOTPAR –0.11 patients per treatment arm. was good (Figure 16B, C, and Table 15), so that over the 40 treatments the sum of actual minus Overall comments calculated was less than a quarter of a patient For SPID it was possible to use the gold standard of per treatment arm using either measure. verification by independent data sets. Regressing the percentage > 50% maxTOTPAR against mean % Although no independent verification was possible maxSPID independently for Oxford and RWJ data for VAS, the similarity of the results to those inde- sets produced very similar results (Table 14). Using pendently verified for TOTPAR18,30 and SPID the combined regression analysis, there was excel- supports the approach of using mean data from lent agreement between actual and calculated previously published reports to derive dichotomous numbers of patients with at least 50% maxTOTPAR data for meta-analysis.18

Number 25

0 –9 –8 –7 –6 –5 –4 –3 –2 –1 0 1 2 3 4 5 6 7 8 9 10 Actual minus calculated number of patients with at least 50% maxTOTPAR

FIGURE 17 Distribution of actual – calculated number of patients with at least 50% maxTOTPAR in each treatment using SPID data

TABLE 14 Summary report on SPID calculations for 132 treatments

Results from Oxford and RWJ data sets and the combined data for the regression of number of patients per treatment with > 50% maxTOTPAR against mean % maxSPID (with 95% CI).

Data set N Intercept Slope Coefficient of (95% CI) (95% CI) determination

Oxford 45 –2.3 (–8.3, 3.6) 1.44 (1.24, 1.64) 0.83

RWJ 87 –1.7 (–4.3, 1.0) 1.27 (1.16, 1.38) 0.86

36 Combined 132 –2.3 (–4.9, 0.2) 1.36 (1.26, 1.45) 0.85 Health Technology Assessment 1998; Vol. 2: No. 12

Comment From the categorical pain relief scale and its summary TOTPAR measure, dichotomous data There is an asymmetric distribution of summary (the proportion of patients achieving at least 50% values of pain relief in clinical trials of analgesics of % maxTOTPAR and the corollary, those not using standard trial methods. Using mean values to achieving 50% relief) can now be derived with describe these summary values is inappropriate and some confidence. Categorical pain relief data can may result in erroneous conclusions.27 To use also be used with confidence. information from RCTs of analgesic drugs reporting mean data, conversion to some form amenable Other data may be used as it becomes available to to meta-analysis is necessary – and, preferably, some further validate these relationships (Table 16), based dichotomous measurement. The alternative may be on a wide variety of acute pain conditions with to discard the many thousands of studies of different analgesics, including simple analgesics, analgesic interventions in the literature. NSAIDs, combinations and sublingual and intramuscular opiates. The only caution is that the Some possible methods of conversion have been validity of these relationships has been demons- subjected to the gold standard of verification by trated only in short-term single-dose studies in an independent data set. There were many acute pain models. patients, in many studies, with different clinical settings, using placebo and several different active analgesics. The result – the relationship between References the calculated and actual number of patients with at least 50% maxTOTPAR – was essentially the 1. Wall PD. The placebo and the placebo response. In: Wall PD, Melzack R, editors. Textbook of pain. 3rd ed. same as that obtained originally using the relation- Edinburgh: Churchill-Livingstone, 1994:1297–1308. ship for the actual data and from a 10,000 treatment arm simulation. Verification was also possible 2. Wall PD. The placebo effect: an unpopular topic. for SPID, but not for VAS, though there is no Pain 1992;51:1–3. obvious reason to suspect that conversions 3. Beecher HK. The powerful placebo. JAMA explored here should not be accurate. 1955;159:1602–6.

TABLE 15 Accuracy of the conversion in actual and simulated treatments

Actual minus calculated % of patients with > 50% maxTOTPAR for treatment arms: agreement to within – Data sets Scale ≤ 1 ≤ 2 ≤ 3 ≤ 4 ≤ 5 ≤ 6 45 Oxford TOTPAR 58 82 93 95 98 98 85 RWJ TOTPAR5275859496100 132 Oxford and RWJ combined SPID 45 70 87 92 95 98 40 Oxford VAS SPID 65 75 85 95 95 98 40 Oxford VAS TOTPAR 65 73 85 95 98 98 Comparisons of actual minus calculated (irrespective of sign) percentages of patients with at least 50% maxTOTPAR as cumulative percentages for the 45 Oxford treatments using TOTPAR and 85 RWJ treatments using TOTPAR.18,30 Cumulative percentages are shown at different levels of agreement.The final three rows show comparisons using SPID,VAS SPID and VAS TOTPAR as basis of calculations in Oxford and RWJ data sets.

TABLE 16 Summary of formulae to derive proportion of patients achieving at least 50% pain relief from mean data using different outcome measures

Outcome measure Formula Proportion Categorical PR 1.33 x mean % maxTOTPAR – 11.5 of patients Categorical PIachieving 1.36 mean % maxSPID – 2.3 at least VAS PR50% PR 1.15 mean % maxVAS TOTPAR – 8.51 VAS PI 1.18 mean % maxVAS SPID – 2.2 37 Estimating relative effectiveness

4. Evans FJ. The placebo response in pain reduction. 19. Bullingham RE, McQuay HJ, Moore RA, Weir L. An Adv Neurol 1974;4:289–96. oral buprenorphine and paracetamol combination 5. Porter EJB, Rolfe M, McQuay HJ, Moore RA, compared with paracetamol alone: a single dose Bullingham RES. Single dose comparison of double-blind postoperative study. Br J Clin Pharmacol bicifadine, codeine and placebo in postoperative 1981;12:863–7. pain. Curr Ther Res 1981;30:156–60. 20. Carroll D, Frankland T, Nagle C, McQuay H. Oral bromfenac 10 and 25 mg compared with sublingual 6. Evans PJ, McQuay HJ, Rolfe M, O’Sullivan G, buprenorphine 0.2 and 0.4 mg for postoperative Bullingham RE, Moore RA. Zomepirac, placebo and pain relief. Br J Anaesth 1993;71:814–17. paracetamol/dextropropoxyphene combination compared in orthopaedic postoperative pain. Br J 21. McQuay HJ, Poppleton P, Carroll D, Summerfield Anaesth 1982;54:927–33. RJ, Bullingham RE, Moore RA. Ketorolac and acetaminophen for orthopedic postoperative pain. 7. McQuay HJ, Bullingham RE, Moore RA, Carroll D, Clin Pharmacol Ther 1986;39:89–93. Evans PJ, O’Sullivan G, et al. Zomepirac, dihydrocodeine and placebo compared in postoperative 22. McQuay HJ, Carroll D, Watts PG, Juniper RP, Moore pain after day-case surgery. The relationship between RA. Does adding small doses of codeine increase the effects of single and multiple doses. pain relief after third molar surgery? Clin J Pain Br J Anaesth 1985;57:412–19. 1987;2:197–201. 8. McQuay HJ, Carroll D, Poppleton P, Summerfield 23. McQuay HJ, Carroll D, Watts PG, Juniper RP, RJ, Moore RA. Fluradoline and aspirin for ortho- Moore RA. Codeine 20 mg increases pain relief from pedic postoperative pain. Clin Pharmacol Ther ibuprofen 400 mg after third molar surgery. 1987;41:531–6. A repeat-dosing comparison of ibuprofen and an ibuprofen-codeine combination. Pain 1989;37:7–13. 9. McQuay HJ, Carroll D, Frankland T, Harvey M, [Erratum: Pain 1989;39:373]. Moore A. Bromfenac, acetaminophen, and placebo in orthopedic postoperative pain. Clin Pharmacol 24. McQuay HJ, Carroll D, Guest P, Juniper RP, Moore Ther 1990;47:760–6. RA. A multiple dose comparison of combinations of ibuprofen and codeine and paracetamol, codeine 10. Cooper SA. Single-dose analgesic studies: the upside and caffeine after third molar surgery. Anaesthesia and downside of assay sensitivity. Adv Pain Res Ther 1992;47:672–7. 1991;18:117–24. 25. McQuay HJ, Carroll D, Guest PG, Robson S, Wiffen 11. Gracely RH, Dubner R, Deeter WR, Wolskee PJ. PJ, Juniper RP. A multiple dose comparison of Clinicians’ expectations influence placebo analgesia. ibuprofen and dihydrocodeine after third molar Lancet 1985;i:43. surgery. Br J Oral Maxillofac Surg 1993;31:95–100. 12. Wall PD. Pain and the placebo response. Experi- 26. McQuay HJ, Moore RA, Poppleton P, Uppington J, mental and theoretical studies of consciousness. Bullingham RES, Lammer P. Postoperative single- CIBA Foundation Symposium 174. Chichester: Wiley, dose comparison of intramuscular meptazinol, 1993:187–216. morphine and pethidine, compared in the same 13. Bergmann J-F, Chassany O, Gandiol J, Deblois P, patients with intravenous meptazinol, morphine and Kanis JA, Segrestaa J, et al. A randomised clinical trial pethidine given by PCA. Unpublished. of the effect of informed consent on the analgesic 27. McQuay H, Carroll D, Moore A. Variation in the activity of placebo and naproxen in cancer pain. Clin placebo effect in randomised controlled trials of Trials Meta-Analysis 1994;29:41–7. analgesics: all is as blind as it seems. Pain 14. Shapiro AP, Myers T, Reiser MF, Ferris EB. 1996;64:331–5. Comparison of blood pressure response to veriloid 28. Cook RJ, Sackett DL. The number needed to treat: and to the doctor. Psychosom Med 1954;16:478–88. a clinically useful measure of treatment effect. BMJ 15. Gøtzsche PC. Meta-analysis of NSAIDs: contribution 1995;310:452–4. of drugs, doses, trial designs, and meta-analytic 29. Dudewicz EJ, Mishra SM. Modern mathematical techniques. Scand J Rheumatol 1993;22:255–60. statistics. New York: Wiley, 1988. 16. Moertel CG, Ahmann DL, Taylor WF, Schwartau 30. Moore A, McQuay H, Gavaghan D. Deriving N. Relief of pain by oral medications. JAMA dichotomous outcome measures from continuous 1974;229:55–9. data in randomised controlled trials of analgesics: 17. Jadad AR. Meta-analysis of randomised clinical trials verification from independent data. Pain in pain relief [DPhil thesis]. Oxford: University of 1997;69:127–30. Oxford, 1994. 31. Silverstein FE, Graham DY, Senior JR, Davies HW, 18. Moore A, McQuay H, Gavaghan D. Deriving Struthers BJ, Bittman RM, et al. Misoprostol reduces dichotomous outcome measures from continuous serious gastrointestinal complications in patients data in randomised controlled trials of analgesics. with rheumatoid arthritis receiving nonsteroidal anti- 38 Pain 1996;66:229–37. inflammatory drugs. Ann Intern Med 1995;123:241–9. Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 6 Combining data and interpreting results

s professionals we want to use the best • Were the question(s) and methods A treatments and, as patients, to be given them. clearly stated? Knowing that an intervention works (or does not • Were the search methods used to locate relevant work) is fundamental to clinical decision-making. studies comprehensive? • Were explicit methods used to determine which When is the evidence strong enough to justify articles to include in the review? changing practice? Some of the decisions we make • Was the methodological quality of the primary are based on individual studies, often on small studies assessed? numbers of patients, which, given the random • Were the selection and assessment of the play of chance, may lead to incorrect decisions. primary studies reproducible and free from bias? Systematic reviews identify and review all the • Were differences in individual study results relevant studies, and are more likely to give a explained adequately? reliable answer. Explicit methods and quality • Were the results of the primary studies combined standards are used to reduce bias. The results are appropriately? the closest we are likely to get to the truth in the • Were the reviewers’ conclusions supported by current state of knowledge. the data cited?

The following questions should be answered by a When systematic reviews use data from different systematic review. numbers of papers (see the paper by Vander Stichele and colleagues2 for an excellent discussion • How well does an intervention work (compared of the eligibility criteria for trials of head lice with placebo, no treatment or other interventions infection), reasons should be sought. Reviews can in current use) – or can it be forgotten? use criteria that exclude information important to • Is it safe? individual clinicians, or may include studies with • Will it work and be safe for the patients inadequate trial design. Inclusion and exclusion in my practice? criteria must be read critically to see if they make sense in the particular clinical circumstance. Clinicians live in the real world and are busy people; they need to synthesise their knowledge of Outcome measures chosen for data extraction a particular patient in their practice, their experi- should also be sensible. This is not usually a ence and expertise, and the best external evidence problem but, again, it is a part of the method that from systematic review. They can then be reason- needs to be read carefully to see if the outcome ably sure that they are doing their best. However, measure extracted appears appropriate. The the product of a systematic review, particularly a reviewer may have used all the available inform- meta-analysis, is often some sort of statistical ation, with any problems being due to the original output, which is not usually readily interpretable or trials, but outcome measures are a determinant usable in day-to-day clinical practice. A common of the clinical utility of the review. Examples, in currency to help make the best treatment decision the antibiotic treatment of Helicobacter pylori for a particular patient is needed. That common infection and peptic ulcer, would be outcome currency is, in the authors’ opinion, the NNT. measures of short-term bacterial kill rates and long-term remission. Quality control Systematic reviews of inadequate quality may be Therapeutic interventions: which study worse than none, because faulty decisions may be architectures are admissible? made with unjustified confidence. Quality control The gold standard for a systematic review of in the systematic review process, from literature therapeutic efficacy is that the eligible studies searching onwards, is vital. How the quality of a should be RCTs. If trials are not randomised, systematic review may be judged is encapsulated in estimates of treatment effect may be exaggerated the following questions1 (these are explained in by up to 40% (see Table 17).3 In a systematic review more detail in chapter 7). of TENS in postoperative pain, 17 reports on 39 Combining data and interpreting results

TABLE 17 Systematic reviews should eliminate bias Combining data: quantitative

Feature Overestimate of treatment systematic reviews effect (%) There are two parts to the question, ‘Does it work?’ Randomisation 40 – how does it compare with placebo and how does it compare with other therapies? Whichever Double-blind 17 comparison is considered, the three stages of examining a review are a L’Abbé plot (Figure 18), Duplicates 20 statistical testing (odds ratio or relative risk), and Small trials 30 a clinical significance measure such as NNT.

L’Abbé plots6 786 patients could be regarded unequivocally For therapies, a first stage is to look at a simple as RCTs in acute postoperative pain. Of these scatter plot, which can yield a surprisingly compre- 17 RCTs, 15 demonstrated no benefit of TENS hensive qualitative view of the data. Even if the over placebo. A total of 19 reports had pain review does not show the data in this way, they outcomes but were not RCTs; in 17 of these, can be extracted from information on individual TENS was considered by their authors to have trials presented in the review tables. Data extracted had a positive analgesic effect.4 When appropriate, from three different systematic reviews of treat- and particularly with subjective outcomes, the ments for painful diabetic neuropathy are shown gold standard for an efficacy systematic review in Figure 19.7–9 Each point on the graph is the is studies which are both randomised and result of a single trial, and what happens with double-blind. The therapeutic effect may the intervention in question – the experimental be exaggerated by up to 20% in trials with event rate (EER) – is plotted against the event rate deficient blinding.3 in the controls – the control event rate (CER).

Not all data can be combined Trials in which the experimental treatment in a meta-analysis: qualitative proves better than the control (EER > CER) will systematic reviews be in the upper left of the plot, between the y axis It is often not possible or sensible to combine or and the line of equality (Figure 20). All three pool data, and this results in a qualitative rather interventions presented in Figure 19 were effective than a quantitative systematic review. Combining but the figure does not indicate how effective. data is not possible if there is no quantitative If experimental is no better than control then information in the component trials of the the point will fall on the line of equality (EER = review. Combining data may not be sensible CER), and if control is better than experimental, if different clinical outcomes were used or the then the point will be in the lower right of the patients were followed-up for different lengths plot, between the x axis and the line of equality of time. Combining continuous rather than (EER < CER). dichotomous data may be difficult. Even if dichotomous data is measured and presented, if the trials are otherwise of poor quality5 it may not be sensible to combine the data. 100

Making decisions from qualitative Treatment 75 better than systematic reviews control Making decisions about whether or not a therapy works from such a qualitative systematic review 50 may look easy. In the example above, 15 of Control 17 RCTs of TENS in acute pain showed no benefit better than 25 compared with controls. The thinking clinician treatment will catch the Bayesian drift – that TENS in acute pain is not effective. The problem with such Proportion improved with treatment improved Proportion 0 simple vote-counting is that it may mislead. It 0 255075100 ignores the sample size of the constituent studies, Proportion improved with control the magnitude of the effect in the studies and the validity of their design even though they 40 were randomised. FIGURE 18 L’Abbé plot for treatment Health Technology Assessment 1998; Vol. 2: No. 12

Number given placebo EER 250 % > 50% pain relief on treatment 100

200 75

150 50

100 25

50 0 0 255075100 CER 0 % > 50% pain relief on placebo 020406080 Proportion of patients with > 50% pain relief with placebo FIGURE 19 L’Abbé plot of EER against CER (■, RCTs of anticonvulsants; ◆, antidepressants; ●, topical capsaicin in diabetic neuropathy7–9) FIGURE 20 Relationship between placebo response and trial size for pharmacological interventions in diabetic neuropathy Visual inspection gives a quick and easy indication of the level of agreement between trials. Hetero- geneity is often assumed to be due to variation in was 24–69%.11 Such variation would not be expect- the EER, the effect of the intervention. Variation ed in other circumstances, such as in the use of in CER can also be a source of heterogeneity antimicrobial agents. H. pylori eradication rates with (see Figure 19); in this case, the controls were all short-term use of ulcer healing drugs were 0–17% matched with placebo in a relatively homogenous in 11 RCTs (with 10/11 being below 10%).12 chronic condition with treatment over a period ranging from several weeks to several months. The reason for large variations in event rates with placebo may have something to do with trial design L’Abbé plots are not yet widely used. They have and population. The overwhelming reason for several benefits: large variations in placebo rates in pain studies (and probably studies in other clinical conditions) • the simple visual presentation is easy to assimilate is the relatively small group sizes in trials. Group • they lead to consideration of the reasons for such sizes are chosen to produce statistical significance wide variation in (especially) placebo responses, through power calculations – for pain studies the and about other factors in the overall package of usual size is 30–40 patients for a 30% difference care that can contribute to effectiveness between placebo and active. • the need for placebo controls is explained if ethical issues about future trials arise An individual patient can have no pain relief or • overly good or bad results for an intervention are 100% pain relief. Random selection of patients can viewed with scepticism in a single trial where the therefore produce groups with low or high placebo major influence may be how good or bad the response rates, or with a rate in between. Ongoing response was with placebo. mathematical modelling based on individual patient data shows that, while group sizes of up to Variation in control (placebo) response rates 50 patients are likely to show a statistical difference The large variation in CER (from 0% to 80%) is between 80% and 90% of the time, to generate a not unusual. Similar variation was seen in trials of close approximation to the ‘true’ clinical impact anti-emetics in postoperative vomiting10 and, in six of a therapy requires as many as 500 patients per trials of prophylactic natural surfactant for preterm group (or more than 1000 patients in a trial). infants, the CER for bronchopulmonary dysplasia This is part of the rationale of systematic reviews. 41 Combining data and interpreting results

Examples of the way group size can be a source of may be more acceptable.14 In some circumstances, variation are important in understanding how pool- for instance, in prophylaxis for nausea and ing of information in pain trials can be of help. vomiting, particular CER spreads may be One example, given in Figure 20, shows trials in determinants of trial validity.15 diabetic neuropathy in which the proportion of patients given placebo is plotted against the In most pain studies neither of these apply. number given placebo. Statistical significance A similar pattern of an inverted ‘V’ is also seen in Odds ratios topical NSAID trials, and indicates that almost all When it is legitimate and feasible to combine data, the variability in placebo responses occurs in trials the odds ratio and the relative risk are the accepted of small size. In rheumatoid arthritis, Gøtzsche13 statistical tests to show that the intervention works found a similar variability in estimates of change significantly better than the comparator. As more in erythrocyte sedimentation rate (ESR) and joint use is made of systematic reviews to compare size by sample size. therapies, clinicians need to understand these clinical epidemiological tools, which present the The lessons are that information from individual results in an unfamiliar way. trials of small size should be treated with circum- spection in pain and probably in other therapeutic The odds ratios for the trials of antidepressants in areas, and that the variation in outcomes seen in diabetic neuropathy mentioned above are shown trials of small size is probably an artefact, especially in Figure 21. Some of the component trials did not in the absence of any Bayesian drift. show statistical significance; the lower 95% CI of the odds ratio was less than 1. Conversely, other Indirect comparisons trials and the combined analysis did show statistical Indirect comparisons of the efficacy of different significance, with the lower 95% CI being greater interventions, for example, by trying to compare than 1, meaning that in 19 cases out of 20 the treatments which have each been compared ‘true’ value will be greater than 1. with placebo rather than with each other, may not be viable if the CERs are dissimilar. Post-hoc The odds ratio can give a distorted impression approaches, taking all the trials, then using when analyses are conducted on subgroups which only those which have a low or a high CER, are differ substantially in baseline risk.14 When CERs frowned on, although using particular clinical are high (certainly when they are above 50%), settings and anticipating less spread of the CER odds ratios should be interpreted with caution.

Favours control Favours treatment

Gomez-Perex et al, 1985 Kvinesdal et al, 1984 Max et al, 1987 Max et al, 1991 Max et al, 1992 Sindrup et al, 1989 Sindrup et al, 1990a Sindrup et al, 1990a Sindrup et al, 1990b Sindrup et al, 1990b Sindrup et al, 1992a Sindrup et al, 1992b Sindrup et al, 1992b Combined

0.1 1 10 100 Odds ratio

42 FIGURE 21 Odds ratios for antidepressants in diabetic neuropathy9 Health Technology Assessment 1998; Vol. 2: No. 12

Relative risk or benefit measured on continuous rather than dichotomous The fact that it is the odds ratio rather than relative scales, and even using different outcome measures. risk or benefit which is used as the test of statistical The z score output is in standard deviation units significance in systematic reviews seems to be due and is thus scale-free. to custom and practice rather than to any inherent intellectual advantage.14 Relative risk or benefit The disadvantage of effect size is that it is not may be better than an odds ratio because it is intuitive for clinicians. more robust in situations where the CER is high.16 With event rates above 10%, relative risk produces Number-needed-to-treat more conservative figures.17 The NNT concept is proving to be a very effective alternative as the measure of clinical significance In the following chapters, both relative benefit from quantitative systematic reviews. It has the and relative risk (of harm) are used, despite the crucial advantage of applicability to clinical uncertainty and disagreement between statisticians practice, and shows the effort required to achieve and reviewers. In all cases, the actual numbers a particular therapeutic target. The NNT is given are given so that, when the dust has settled, by the equation: re-calculations can be made according to the prevailing opinion. 1 NNT =

Heterogeneity (IMPact/TOTact) – (IMPcon/TOTcon) Clinicians making decisions on the basis of systematic reviews need to be confident that apples where: are not being compared with oranges. The L’Abbé IMPact = number of patients given active plot is a qualitative defence against this spectre. treatment who achieve the target

Statistical testing provides a quantitative rampart, TOTact = total number of patients given and is available as standard software.* Unfortun- active treatment ately, all of these tests lack power, so while a positive IMPcon = number of patients given a control test for heterogeneity suggests mixed fruits are treatment who achieve the target being compared, a negative test does not provide TOTcon = total number of patients given complete reassurance of no heterogeneity. control treatment

Heterogeneity will also appear to occur because of NNT is also 1 divided by the proportion obtaining variations in CERs and EERs caused by the random a particular effect with treatment minus the pro- play of chance in trials of small size. Generally trials portion obtaining the same effect with control, with fewer than ten patients per group have been when those proportions are expressed as a fraction. omitted in reviews in this report but considerable Because we have just described the absolute risk variability will occur in groups of less than reduction, so NNT is also the reciprocal of the 50 patients. absolute risk reduction.

Treatment-specific How well does an NNT is treatment-specific. It describes the intervention work? difference between active treatment and control. The threshold used to calculate NNT can vary but While odds ratios and relative risks can show NNT is likely to be relatively unchanged because a that an intervention works well compared with a change in threshold changes results for both active control, they are of limited help in showing how and control. well the intervention works – the size of the effect or its clinical significance. For example, in an individual patient data meta-analysis of postoperative pain relief, NNTs Effect size compared with placebo were calculated for para- The classical method of estimating effect size is cetamol, 650 mg, plus propoxyphene, 100 mg, at to use the standardised mean difference.18 The between 20 and 80% relief of pain (see Figure 22). advantage of this approach is that it can be used With placebo, the proportion of patients achieving to compare the efficacy of different interventions a particular level of pain relief fell quickly as the

* Cochrane Collaboration: Review Manager Software (RevMan) 1996. 43 Combining data and interpreting results

the specified range. If the odds ratio or relative % or NNT risk/benefit is not statistically significant then 80 the NNT is infinite, indicating no difference from control. An NNT with an infinite CI is then but a point estimate. It may still have clinical importance as a benchmark until further data permits finite 60 CIs but decisions must take account of this parlous state.

Disadvantages 40 The disadvantage of the NNT approach, apparent from the formula, is that it needs dichotomous data. Continuous data can be converted to dichotomous for acute pain studies so that NNTs may be 20 calculated by deriving a relationship between the two from individual patient data.20 Because of the way in which it is calculated, NNT will also be 0 sensitive to trials with high CERs. As the CER rises, 0 20406080100 the potential for treatment-specific improvement decreases, resulting in higher (and apparently less % maxTOTPAR effective) NNTs. So, as with any summary measure from a quantitative systematic review, NNT needs to be treated with caution; comparisons can only be FIGURE 22 Effect of different thresholds of pain relief on NNT made confidently if CERs are in the same range. (–■–, paracetamol plus propoxyphene; –●–, placebo; –●–, NNT) Calculating NNTs when they are target was raised. For an effective analgesic, this not provided proportion fell slowly until high relief targets Odds ratios were reached. The difference remained largely If a quantitative systematic review produces odds unaltered over a wide range of targets, thus ratios but no NNTs, these can be derived from generating stable NNTs. Table 18.

An NNT of 1 describes an event which occurs in A caveat must be added here that odds ratios every patient given the treatment but in no patient should be interpreted with caution when events in a comparator group. This could be described occur commonly, such as in treatments, and odds as the ‘perfect’ result in, say, a therapeutic trial ratios may overestimate the benefits of an effect of an antibiotic compared with placebo. For when event rates are above 50%. They are likely to therapeutic benefit, the NNT value should be as be superseded by relative risk or benefit because close as possible to 1; there are few circumstances these are more robust in situations where event in which a treatment is close to 100% effective and rates are high.14,21 the control or placebo completely ineffective, so NNTs of 2 or 3 often indicate an effective intervention. For unwanted effects, NNT becomes the Is it safe? NNH (number-needed-to-harm), which should be as large as possible. Estimating the risk of harm is a critical part of a clinical decision. Systematic reviews should report It is important to remember that the NNT is always adverse events as well as efficacy, and consider the relative to the comparator and applies to a partic- issue of rare but important adverse events. Large ular clinical outcome. The duration of treatment RCTs apart, most trials study limited numbers of necessary to achieve the target should be specified. patients. New medicines may be launched after The NNT for cure of head-lice at 2 weeks with trials on 1500 patients,22 missing any rare but permethrin 1% compared with a control was important adverse events. The rule of three is 1.1 (95% CI, 1.0–1.2).2,19 important here. If a particular serious event does not occur in 1500 patients given the Confidence intervals treatment, then we can be 95% confident The CI of the NNT is an indication that, in that the chance of it occurring is, at most, 44 19 cases out of 20, the ‘true’ value will be in 3/1500.23 Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 18 Table for estimating NNT when odds ratio or CER are known (for prophylactic interventions16)

Odds ratios

Prophylaxis Treatment

CER 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 10.0

0.05 41 46 52 59 69 83 104 139 209 43 22 15 12 9 8 7 6 3

0.1 21 24 27 31 36 43 54 73 110 23 12 9 7 6 5 4 4 2

0.2 11 13 14 17 20 24 30 40 61 14 8 5 4 4 3 3 3 2

0.3 8 9 10 12 14 18 22 30 46 11 6 5 4 3 3 3 3 2

0.4 7 8 9101215192640106 4 4 3 3 3 3 2

0.5 6 7 8 9 11 14 18 25 38 10 6 5 4 4 3 3 3 2

0.7 6 7 9101316202844138 7 6 5 5 5 5 4

0.9 12 15 18 22 27 34 46 64 101 32 21 17 16 14 14 13 13 11

Formula for prophylaxis: NNT = 1 – [CER x (1 – OR)]

(1 – CER) x CER x (1 – OR)

Formula for treatment: NNT = CER (OR – 1) + 1

CER (OR – 1) x (1 – CER) where OR = odds ratio.

Choose the column which is closest to the published odds ratio (prophylaxis left side, treatment right side) and the row which is closest to the expected CER, then read off the corresponding NNT. The table can also be used to see how different values for CER or EER for an individual patient affect the NNT at a given odds ratio.

Much the same rules apply to harm as to RCTs, rare serious harm may not be spotted. efficacy, but with some important differences – For an adverse effect systematic review, study the NNH rather than the NNT and the rules architectures of lower intrinsic quality may of admissible evidence. therefore be admissible. An extreme example is that observer blinding is superfluous if the Number-needed-to-harm outcome is death. Such rare and serious harm For minor adverse effects reported in RCTs, cannot and should not be dismissed just because NNH may be calculated in the same way as the it is reported in a case report rather than in an NNT. When there is low incidence it is likely RCT. The ‘process rules’ in this area have yet that point estimates alone will emerge (infinite to be determined. CIs). Major harm may be defined in a set of RCTs as intervention-related study withdrawal, and be calculated from those numbers. Precise estimates of Using NNTs major harm will require a much wider literature search to trawl for case reports or series. The In an ideal world you will have three numbers absence of information on adverse effects in for each intervention, an NNT for benefit and systematic reviews reduces their usefulness. an NNH for both minor and major harm.

Rules of admissible evidence This then becomes the yardstick against which The gold standard of evidence for harm, as for alternative interventions should be judged, and efficacy, is the RCT. The problem is that, in the the pivot for the clinical decision on whether or not relatively small number of patients studied in to use the intervention for an individual patient. 45 Combining data and interpreting results

References 12. Moore RA. Helicobacter pylori and peptic ulcer. A systematic review of effectiveness and an overview of 1. Oxman AD, Guyatt GH. Guidelines for reading the economic benefits of implementing that which literature reviews. Can Med Assoc J 1988;138:697–703. is known to be effective (http://www.jr2.ox.ac.uk/ 2. Vander Stichele RH, Dezeure EM, Bogaert MG. Bandolier/bandopubs/hpyl/hp0.html). Systematic review of clinical efficacy of topical Oxford: Health Technology Evaluation treatments for head lice. BMJ 1995;311:604–8. Association, 1995. 3. Schulz KF, Chalmers I, Hayes RJ, Altman DG. 13. Gøtzsche PC. Sensitivity of effect variables in Empirical evidence of bias: dimensions of rheumatoid arthritis: a meta analysis of 130 methodological quality associated with estimates placebo controlled NSAID trials. J Clin Epidemiol of treatment effects in controlled trials. JAMA 1990;43:1313–18. 1995;273:408–12. 14. Sinclair JC, Bracken MB. Clinically useful measures 4. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. of effect in binary analyses of randomized trials. Randomization is important in studies with pain J Clin Epidemiol 1994;47:881–9. outcomes: systematic review of transcutaneous 15. Tramèr M, Moore RA, McQuay HJ. A rational electrical nerve stimulation in acute postoperative approach to sensitivity analyses in meta-analysis: pain. Br J Anaesth 1996;77:798–803. factors influencing propofol’s effect on 5. Jadad AR, Moore RA, Carroll D, Jenkinson C, postoperative nausea and vomiting. Reynolds DJM, Gavaghan DJ, et al. Assessing Submitted 1997. the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 16. Sackett DL, Deeks JJ, Altman DG. Down with odds 1996;17:1–12. ratios! Evidence-Based Med 1996;1:164–6. 6. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in 17. Deeks J. What the heck’s an odds ratio? Bandolier clinical research. Ann Intern Med 1987;107:224–33. 1996;3(25):6–7. 7. Zhang WY, Li WPA. The effectiveness of topically 18. Glass GV. Primary, secondary, and meta-analysis of applied capsaicin. A meta-analysis. Eur J Clin research. Educ Res 1976;5:3–8. Pharmacol 1994;46:517–22. 19. Moore RA, McQuay H, Gray M. Bandolier – the first 8. McQuay H, Carroll D, Jadad AR, Wiffen P, Moore A. 20 issues. Oxford: Bandolier, 1995. Anticonvulsant drugs for management of pain: a systematic review. BMJ 1995;311:1047–52. 20. Moore A, McQuay H, Gavaghan D. Deriving dichotomous outcome measures from continuous 9. McQuay HJ, Tramèr M, Nye BA, Carroll D, data in randomised controlled trials of analgesics. Wiffen PJ, Moore RA. A systematic review of Pain 1996;66:229–37. antidepressants in neuropathic pain. Pain 1996;68:217–27. 21. Sackett D, Richardson WS, Rosenberg W, Haynes B. Evidence based medicine. London: Churchill 10. Tramèr M, Moore A, McQuay H. Prevention of Livingstone, 1996. vomiting after paediatric strabismus surgery: a systematic review using the numbers-needed-to-treat 22. Moore TJ. Deadly medicine. New York: Simon & method. Br J Anaesth 1995;75:556–61. Schuster, 1995. 11. Soll JC, McQueen MC. Respiratory distress 23. Eypasch E, Lefering R, Kum CK, Troidl H. syndrome. In: Sinclair JC, Bracken ME, editors. Probability of adverse events that have not yet Effective care of the newborn infant. Oxford: occurred: a statistical reminder. BMJ Oxford University Press, 1992:chapter 15, p333. 1995;311:619–20.

46 Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 7 Existing systematic reviews

s part of the evidence-gathering exercise our 1. They had to be described as systematic A aim was to find all previous systematic reviews reviews or, if not, they had to include pooled of analgesic interventions. analysis of the results of several independent primary studies. Studies in which statistical This was undertaken in two stages. In the first, a synthesis had been planned but was deemed search was undertaken for all systematic reviews to be inappropriate were also included. published before 1993/94. A total of 80 were found, 2. They had to incorporate trials in which and two judges assessed their quality using Oxman pain was an outcome measure or in which and Guyatt’s index.1 Most of the reviews looked at analgesic interventions were compared drug interventions for chronic pain conditions. Two- for outcomes other than pain within the thirds were published after 1990. Most had methodo- context of a painful condition (e.g. a study logical flaws, such as insufficient information on looking at the validity of grip strength to retrieval methods and validity assessment and design assess the effectiveness of NSAIDs in of the primary studies. Poor quality systematic re- rheumatoid arthritis). views reached significantly more positive conclusions 3. They had to be published or accepted and, when there was more than one systematic review for publication. on a particular topic, the results did not always agree. A full account of this first stage has been published.2 Search strategy A MEDLINE search (Silver Platter MEDLINE The second stage was (and is) a prospective v. 3.0, 3.1 and 3.11) was undertaken from 1966 to exercise to maintain an up-to-date database of October 1993. This MEDLINE strategy had been systematic reviews in pain relief. developed to identify the maximum possible number of randomized, double-blind studies or meta-analyses in pain research, and contained Introduction text words, ‘wild cards’ and MeSH terms.6 Forty journals were searched by hand. The register of Systematic reviews can potentially resolve conflicts systematic reviews at the UK Cochrane Centre was when reports of primary studies disagree, and in- checked for eligible studies, and lead authors of crease the likelihood of detecting small but clinically abstracts were asked for full manuscripts. The important effects.3–5 They can also be easily misused reference lists for citations of other systematic to produce misleading estimates of effectiveness. reviews were scanned.

A systematic search of the literature was used Methodological evaluation to identify the highest possible proportion of Each study was evaluated twice, using Oxman and systematic reviews assessing analgesic interventions. Guyatt’s index,1,7 with the title of the journal, the The objectives were: authors’ names, the date of publication and the source of financial support for the study obscured. (i) to produce a citation database of all A consensus score was obtained. available reviews (ii) to assess the quality of systematic reviews Statistical analysis in pain relief The chi-squared test was used to test the relation- (iii) to establish whether or not quality scores are ship between the direction of the conclusion of useful to resolve conflicts between different the systematic reviews (positive versus negative/ systematic reviews. uncertain) and the overall quality scores, and the influence of study architecture on systematic reviews which included study designs other than Methods RCTs. Prior hypotheses were that poor quality reviews and those including designs other Inclusion criteria than RCTs would be more likely to produce Reports had to meet the following criteria. positive conclusions. 47 Existing systematic reviews

Results pooling methods (Table 20). All were based on published data only (no individual patient Quality assessment of reviews data analysis), without validity checks with to 1993/94 the study investigators. A separate list of the Of the 84 reports found, 70 were included in reviews, by first author, is presented at the the quality assessment (Table 19). The exclusions end of this chapter. are specified elsewhere.2 The earliest report was from 1980, and over two-thirds appeared after Overall quality scores 1990. Reviews considered between two and 196 The median agreed overall score for the primary studies (median 28). In all, 60 reviews systematic reviews was 4 (range 1–7). Systematic reached positive conclusions, seven negative reviews of high quality were significantly less conclusions, 12 were uncertain and one did likely to produce positive results (see Table 21 not reach any conclusion. They used different and Figure 23; chi-squared 18.2, p = 0.006).

TABLE 19 Details of the systematic reviews

Setting Number (%) Outcomes Number (%)

Chronic 58 (72) Pain 63 (79)

Acute 14 (19) Adverse effects 11 (14)

Mixed 6 (7) Validity 3 (4)

Unclear 2 (2) Patient preference 1 (1)

Intervention Return to work 1 (1)

Drug 42 (54) Pulmonary function 1 (1)

Psychological 16 (20) Primary studies

Physical 10 (13) Randomised only 24 (30)

Diagnostic 3 (4) Randomised and double-blind 7 (9)

Complementary 2 (2) Double-blind only 3 (4)

Non-surgical invasive 2 (2) Combination of observational and any of the above 25 (31)

Multidisciplinary 2 (2) Observational only 4 (5)

Surgical 1 (1) Not reported 17 (21)

Preventive 1 (1)

Not specified 1 (1)

TABLE 20 Pooling methods used in the systematic reviews

Method Number (%) Method Number (%)

Standardised mean differences 26 (32) Random effects 2 (2)

Odds ratios/Mantel–Haenszel 10 (13) Kendal’s correlation 1 (1)

‘Percentage change’ comparison 15 (20) Log rank test 1 (1)

Simple addition 7 (9) Relative potency 1 (1)

Criteria-based 4 (5) Not reported 5 (6)

Weighted means 3 (4) Pooling considered inappropriate 3 (4)

Mean risk differences 2 (2) 48 Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 21 Meta-analyses: quality and conclusions Quality score Conclusion 7

Overall Positive Negative/ 6 quality score Uncertain 5 11114 21203

3612

41111 Yes – works No – doesn’t work 583 Efficacy as stated in original review 689

744 FIGURE 23 Systematic reviews: quality and estimate of efficacy2

Of 19 systematic reviews with negative or uncertain Interventions evaluated by multiple results, 16 had overall quality scores above the systematic reviews median, compared with only 20 of the 60 with There was more than one systematic review for positive results. Systematic reviews restricted to six interventions (Table 22). For acupuncture and RCTs were significantly less likely to produce NSAIDs the conclusions of the reviews were the positive conclusions (19/31) than those which same. Two reviews of acupuncture in chronic pain included other study architectures (41/49; chi- concluded that the evidence was flawed and that squared = 5.07; p = 0.024). All conclusions from acupuncture was of uncertain value.8,9 Three systematic reviews of psychological interventions reviews confirmed that the risk of gastrointestinal were positive. In only one of those reviews was complications was increased by NSAIDs.10–12 quality scored above the median. All abstracts scored below the median, and six out of eight Most systematic reviews of manipulation for chronic abstracts received the minimum possible score. back pain concluded that it was useful,13–15 as did

TABLE 22 Multiple systematic reviews on a particular intervention: quality and conclusions

Intervention Study Quality score Conclusion

Acupuncture ter Riet, et al.9 6 Uncertain Patel, et al.8 5 Uncertain

Gastrointestinal effects of NSAIDs Chalmers, et al.10 7 Positive Gabriel, et al.11 6 Positive Bollini, et al.12 5 Positive

Manipulation Koes, et al.19 6 Uncertain Shekelle, et al.13 6 Positive Ottenbacher & di Fabio14 4 Positive Anderson, et al.15 3 Positive

Second-line drugs for rheumatoid arthritis Gøtzsche, et al.20 7 Uncertain Felson, et al.17 6 Positive Capell, et al.16 4 Positive Felson, et al.18 4 Positive

Prevention of postherpetic neuralgia Schmader & Studenski23 7 Uncertain Lycka24 5 Positive Crooks, et al.26 3 Positive Naldi, et al.25 3 Negative

Laser for musculoskeletal pain Gam, et al.22 6 Negative 21 Beckerman, et al. 6 Positive 49 Existing systematic reviews

BOX 2 Using the Oxman and Guyatt scoring system for reviews

The purpose of this index is to evaluate the scientific quality (i.e. adherence to scientific principles) of research overviews (review articles) published in the medical literature. It is not intended to measure literary quality, importance, relevance, originality, or other attributes of overviews. The index is for assessing overviews of primary (‘original’) research on pragmatic questions regarding causation, diagnosis, prognosis, therapy, or prevention. A research overview is a survey of research. The same principles that apply to epidemiological surveys apply to overviews; a question must be clearly specified, a target population identified and accessed, appropriate information obtained from that population in an unbiased fashion, and conclusions derived, sometimes with the help of formal statistical analysis, as in ‘meta-analyses’. The fundamental difference between overviews and epidemiological surveys is the unit of analysis, not the scientific issues addressed by the questions in this index. Since most published overviews do not include a methods section, it is difficult to answer some of the questions in the index. Answers need to be based, as far as possible, on information provided in the overview. If the methods used are incompletely reported relative to a specific item, that item should be scored as ‘partially’. Similarly, if no information is provided regarding what was done relative to a particular question, it should be scored as ‘can’t tell’, unless there is information in the overview to suggest that the criterion either was or was not met.

Quality features 1 Were the search methods used to find evidence on the primary question(s) stated? No Partially Yes 2 Was the search for evidence reasonably comprehensive? No Can’t tell Yes 3 Were the criteria used for deciding which studies to include in the overview reported? No Partially Yes 4 Was bias in the selection of studies avoided? No Can’t tell Yes 5 Were the criteria used for assessing the validity of the included studies reported? No Partially Yes 6 Was the validity of all studies referred to in the text assessed using appropriate criteria? No Can’t tell Yes 7 Were the methods used to combine the findings of the relevant studies (to reach a conclusion) reported? No Partially Yes 8 Were the findings of the relevant studies combined appropriately relative to the primary question of the overview? No Can’t tell Yes 9 Were the conclusions reached by the author(s) supported by the data and/or analysis reported in the overview? No Partially Yes 10 How would you rate the scientific quality of this overview?

Flaws Extensive Major Minor Minimal 1234567 For Question 8, if no attempt has been made to combine findings, and no statement is made regarding the inappropriateness of combining findings, check ‘no’. If a summary (general) estimate is given anywhere in the abstract, the discussion, or the summary section of the paper, and it is not reported how that estimate was derived, check ‘no’, even if there is a statement regarding the limitations of combining the findings of the studies reviewed. If in doubt mark as ‘can’t tell’. For Question 9, if an overview is to be scored as ‘yes’, data (not just citations) must be reported that support the main conclusions relating to the primary question(s) addressed by the overview. For Question 10, overall scientific quality, the score should be based on the answers to the first nine questions. The following guidelines can be used to assist with deriving a summary score: If the ‘can’t tell’ option is used one or more times on the preceding questions, a review is likely to have minor flaws at best and it is difficult to rule out major flaws (i.e. a score of 4 or less). If Questions 2, 4, 6 or 8 are marked as ‘no’, the review is likely to have major flaws (i.e. a score of 3 or less, depending on the number and degree of the flaws). 50 Health Technology Assessment 1998; Vol. 2: No. 12

reviews of second-line drugs for rheumatoid 5. Kassirer JP. Clinical trials and meta-analysis. What do arthritis,16–18 but for both interventions one review they do for us? N Engl J Med 1992;327:273–4. questioned the validity of the findings because of the 6. Jadad AR, McQuay HJ. A high-yield strategy to high risk of bias in the primary studies.19,20 identify randomized controlled trials for systematic reviews. Online J Curr Clin Trials [serial online] Systematic reviews produced conclusions in oppo- 1993; Doc No 33 [3973 words; 39 paragraphs; site directions for lasers in musculoskeletal pain21,22 5 tables]. and for interventions to prevent post-herpetic 7. Oxman AD, Guyatt GH, Singer J, Goldsmith CH, neuralgia.23–26 Both systematic reviews evaluating Hutchison BG, Milner RA, et al. Agreement among laser treatment were given the same quality score. reviewers of review articles. J Clin Epidemiol 1991;44:91–8. 8. Patel M, Gutzwiller F, Paccaud F, Marazzi A. A meta-analysis of acupuncture for chronic pain. Comments Int J Epidemiol 1989;18:900–6. The use of systematic reviews to assess analgesic 9. ter Riet G, Kleijnen J, Knipschild P. Acupuncture interventions is increasing but most of the reviews and chronic pain: a criteria-based meta-analysis. found had methodological flaws which may J Clin Epidemiol 1990;43:1191–9. threaten their conclusions. Only eight of the 10. Chalmers TC, Berrier J, Hewitt P, Berlin J, Reitman 80 satisfied all the Oxman and Guyatt criteria D, Nagalingam R, et al. Meta-analysis of randomized (see Box 2 for explanation) and 16% were given controlled trials as a method of estimating rare complications of non-steroidal anti-inflammatory the lowest possible score. The relationship between drug therapy. Aliment Pharmacol Therap 1988;2:9–26. methodological rigour, type of primary studies included and the direction of the conclusions 11. Gabriel SE, Jaakkimainen L, Bombardier C. Risk for underscores the importance of review quality. serious gastrointestinal complications related to use of nonsteroidal anti-inflammatory drugs. A meta- Systematic reviews including only RCTs were analysis. Ann Intern Med 1991;115:787–96. less likely to produce positive conclusions. 12. Bollini P, García Rodríguez LA, Pérez Gutthann S, Walker AM. The impact of research quality and Reviewers have to work hard to reduce bias. study design on epidemiologic estimates of the The search for evidence must be comprehensive, effect of nonsteroidal anti-inflammatory drugs on decisions about which studies to include or upper gastrointestinal tract disease. Arch Intern Med exclude have to be overt, and validity criteria 1992;152:1289–95. need to be stated. Equally, readers need to be 13. Shekelle PG, Adams AH, Chassin MR, Hurwitz EL, aware of the pitfalls. Brook RH. Spinal manipulation for low-back pain. Ann Intern Med 1992;117:590–8. Several examples were found of reviews of 14. Ottenbacher K, di Fabio RP. Efficacy of spinal the same intervention producing conflicting manipulation/mobilization therapy. A meta analysis. results, despite similar quality scores. This Spine 1985;10:833–7. despite the concept that systematic reviews 15. Anderson R, Meeker WC, Wirick BE, Mootz RD, can resolve conflicting results between Kirk DH, Adams A. A meta-analysis of clinical primary studies. trials of spinal manipulation. J Manipulative Physiol Ther 1992;15:181–94. 16. Capell HA, Porter DR, Madhok R, Hunter JA. References Second line (disease modifying) treatment in rheumatoid arthritis: which drug for which patient? 1. Oxman AD, Guyatt GH. Validation of an index Ann Rheum Dis 1993;52:423–8. of the quality of review articles. J Clin Epidemiol 1991;44:1271–8. 17. Felson DT, Anderson JJ, Meenan RF. The comparative efficacy and toxicity of second-line 2. Jadad AR, McQuay HJ. Meta-analyses to evaluate drugs in rheumatoid arthritis. Arthritis Rheum analgesic interventions: a systematic qualitative 1990;33:1449–61. review of their methodology. J Clin Epidemiol 1996;49:235–43. 18. Felson DT, Anderson JJ, Meenan RF. Use of short- term efficacy/toxicity trade offs to select second-line 3. Sacks HS, Berrier J, Reitman D, Ancona-Berk VA, drugs in rheumatoid arthritis. Arthritis Rheum Chalmers TC. Meta-analyses of randomized 1992;35:1117–25. controlled trials. N Engl J Med 1987;316:450–5. 19. Koes BW, Assendelft WJ, van der Heijden GJ, Bouter 4. Gerbarg ZB, Horwitz RI. Resolving conflicting LM, Knipschild PG. Spinal manipulation and clinical trials: guidelines for meta-analysis. J Clin mobilisation for back and neck pain: a blinded Epidemiol 1988;41:503–9. review. BMJ 1991;303:1298–303. 51 Existing systematic reviews

20. Gøtzsche PC, Pødenphant J, Olesen M, Halberg P. Antczak-Bouckoms A, Tung F, Chalmers TC, Bouckoms Meta-analysis of second-line antirheumatic drugs: A, 1987. Quality assessment and meta-analysis of random- sample bias and uncertain benefit. J Clin Epidemiol ized controlled trials of ibuprofen. Pain;suppl:S51. 1992;45:587–94. Assendelft WJ, Koes BW, van der Heijden GJ, Bouter LM, 21. Beckerman H, de Bie RA, Bouter LM, de Cuyper 1992. The efficacy of chiropractic manipulation for back HJ, Oostendorp RA. The efficacy of laser therapy for pain: blinded review of relevant randomized clinical musculoskeletal and skin disorders: a criteria-based trials. J Manipulative Physiol Ther;15:487–94. meta-analysis of randomized clinical trials. Phys Ther Assendelft WJJ, Hay EM, Adshead R, Bouter LM, 1996. 1992;72:483–91. Corticosteroid injections for lateral epicondylitis: a 22. Gam AN, Thorsen H, Lonnberg F. The effect of low- systematic overview. Br J Gen Pract;46:209–16. level laser therapy on musculoskeletal pain: a meta- Assendelft WJJ, Koes BW, van der Heijden GJM, Bouter analysis. Pain 1993;52:63–6. LM, 1996. The effectiveness of chiropractic treatment of 23. Schmader K, Studenski S. Are current therapies low back pain: an update and attempt at statistical useful for the prevention of postherpetic neuralgia? pooling. J Manipulative Physiol Ther;19:499–507. A critical analysis of the literature. J Gen Intern Med Aylward M, 1975. Clinical studies on alclofenac in the 1989;4:83–9. treatment of rheumatic diseases: a drug in question. 24. Lycka BA. Postherpetic neuralgia and systemic Curr Med Res Opin;3:274–85. corticosteroid therapy. Efficacy and safety. Int J Ballantyne JC, Carr DB, Chalmers TC, Dear KB, Angelillo Dermatol 1990;29:523–7. IF, Mosteller F, 1993. Postoperative patient-controlled 25. Naldi L, Zucchi A, Brevi A, Cavalieri d’Oro L, analgesia: meta-analyses of initial randomized control Cainelli T. Corticosteroids and post-herpetic trials. J Clin Anesth;5:182–93. neuralgia. Lancet 1990;336:947. Ballantyne JC, Carr DB, Berkey CS, Chalmers TC, 26. Crooks RJ, Jones Da, Fiddian P. Zoster-associated Mosteller F, 1996. Comparative efficacy of epidural, chronic pain: an overview of clinical trials with subarachnoid and intracerebroventricular opioids in acyclovir. Scand J Infect 1991;78 suppl:62–8. patients with pain due to cance. Reg Anesth;21:542–56. Bamberg P, Caswell CM., Frame MH, Lam SK, Wong EC, Reviews in chronic pain 1992. A meta-analysis comparing the efficacy of omepra- Abenhaim L, Bergeron AM, 1992. Twenty years of zole with H2-receptor antagonists for acute treatment of randomized clinical trials of manipulative therapy for duodenal ulcer in Asian patients. J Gastroenterol back pain: a review. Clin Invest Med;15:527–35. Hepatol;7:577–85. Abram SE, 1993. Advances in chronic pain management Baron JC, Bousser MG, 1982. Essais therapeutiques since gate control [1992 Bonica lecture]. Reg controles dans la migraine. Organisation et resultats Anesth;18:66–81. [Controlled therapeutic trials in migraine: methodology and results]. Rev Neurol Paris;138:279–95. Adamson GD, Nelson HP, 1996. Medical and surgical treatment of endometriosis. Endocrinologist;6:384–91. Basinski A, Naylor CD, 1991. Aspirin and fibrinolysis in acute myocardial infarction: meta-analytic evidence for Aker PD, Gross AR, Goldsmith CH, Peloso P, 1996. synergy. J Clin Epidemiol;44:1085–96. Conservative management of mechanical neck pain: systematic overview and meta-analysis. BMJ;313:1291–6. Beckerman H, de Bie RA, Bouter LM, De Cuyper HJ, Oostendorp RA, 1992. The efficacy of laser therapy for Ali NMK, 1995. Does sympathetic ganglionic block musculoskeletal and skin disorders: a criteria-based meta- prevent postherpetic neuralgia? Reg Anesth;20:227–33. analysis of randomized clinical trials. Phys Ther;72:483–91. Allegrante JP, 1996. The role of adjunctive therapy in the Beecher HK, 1955. The powerful placebo. management of chronic nonmalignant pain. Am J JAMA;159:1602–6. Med;101:33–39S. Belcon MC, Rooney PJ, Tugwell P, 1985. Aspirin and Altman GB, Lee CA, 1996. Strontium-89 for treatment of gastrointestinal haemorrhage: a methodologic painful bone metastasis from prostate cancer. Oncol Nurs assessment. J Chronic Dis;38:101–11. Forum;23:523–7. Bernstein J, 1983. Overview of efficacy of fenbufen in Amery WK, 1988. Onset of action of various migraine rheumatoid arthritis and osteoarthritis. Am J Med;75:70–4. prophylactics. Cephalalgia;8:11–13. Bertagna C, de Gery A, Hucher M, Francois JP, Zanirato Amery WK, 1989. Prophylactic and curative treatment J, 1994. Efficacy of the combination of nilutamide plus of migraine with calcium antagonists. Drug Des orchidectomy in patients with metastatic prostatic cancer. Deliv;4:197–203. A meta-analysis of seven randomized double-blind trials (1056 patients). Br J Urol;73:396–402. Anderson R, Meeker WC, Wirick BE, Mootz RD, Kirk DH, Adams A, 1992. A meta-analysis of clinical trials of Bhatt Sanders D, 1985. Acupuncture for rheumatoid spinal manipulation. J Manipulative Physiol arthritis: an analysis of the literature. Semin Arthritis 52 Ther;15:181–94. Rheum;14:225–31. Health Technology Assessment 1998; Vol. 2: No. 12

Blanchard EB, Andrasik F, 1982. Psychological assess- Carroll D, Moore RA, Tramèr MR, McQuay HJ, 1998. ment and treatment of headache: recent developments Transcutaneous electrical nerve stimulation does not and emerging issues. J Consult Clin Psychol;50:859–79. relieve labour pain: updated systematic review. Contemp Rev Obstet Gynaecol 1997;(Sept):195–205. Blanchard EB, Andrasik F, Ahles TA, Teders ST, O’Keefe D, 1980. Migraine and tension headache: a meta-analytic Chalmers TC, Berrier J, Bewitt P, Berlin J, Reitman D, review. Behav Ther;11:613–31. Nagalingam R, et al., 1988. Meta-analysis of randomized controlled trials as a method of estimating rare Bloomfield SS, Mitchell J, Cissell G, Barden TP, 1986. complications of non-steroidal anti-inflammatory drug Analgesic sensitivity of two postpartum pain models. therapy. Aliment Pharmacol Ther;2:9–26. Pain;27:171–9. Chalmers TC, Ballantyne JC, Carr DB, Dear KBG, Bogaards MC, ter Kuile MM, 1994. Treatment of Angelillo IF, Mosteller F, 1993. Comparative effects of recurrent tension headache: a meta-analytic review. analgesic therapies upon postoperative pulmonary Clin J Pain;10:174–90. function: meta-analyses [abstract]. In: Proceeding 7th World Congress on Pain. Oxford: IASP/Elsevier;136. Bollini P, García Rodríguez LA, Pérez Gutthann S, Walker AM. The impact of research quality and study Christensen E, Juhl E, Tygstrup N, 1977. Treatment of design on epidemiologic estimates of the effect of non- duodenal ulcer. Randomized clinical trials of a decade steroidal anti-inflammatory drugs on upper gastro- (1964 to 1974). Gastroenterology;73:1170–8. intestinal tract disease. Arch Intern Med;152:1289–95. Cohen JE, Goel V, Frank JW, Bombardier C, Peloso P, Bosi Ferraz M, Atra E, 1989. Meta-análise: nova opcao em Guillemin F, 1994. Group education interventions for pesquisa clínica. Prinípos básicos e um exemplo clínico. people with low back pain. An overview of the literature. Rev Paul Med;107:5–9. Spine;19:1214–22. Collins S, Moore A, McQuay H, 1996. Paracetamol- Bosi-Ferraz M, Tugwell P, Goldsmith CH, Atra E, 1990. codeine combinations versus paracetamol alone. Actual Meta-analysis of sulfasalazine in ankylosing spondylitis. size of increase needs to be measured [letter]. J Rheumatol;17:1482–6. BMJ;313:1209. Broome ME, Lillis PP, 1989. A descriptive analysis of the Collins SL, Moore A, McQuay HJ, Wiffen PJ, 1999. Oral pediatric pain management research. Appl Nurs Res;2:74–81. ibuprofen and diclofenac in postoperative pain: a Broome ME, Lillis PP, Smith MC, 1989. Pain inter- quantitative systematic review. Eur J Pain. In press. ventions with children: a meta analysis of research. Collins SL, Edwards J, Moore A, McQuay HJ, 1998. Single Nurs Res;38:154–8. dose oral dextropropoxyphene in postoperative pain: a quantitative systematic review. Eur J Clin Brown BWJ, 1993. Meta-analysis and patient-controlled Pharmacol;54:107–12. analgesia [editorial; comment]. J Clin Anesth;5:179–81. Cooper SA, 1984. Five studies on ibuprofen for Brunarski DJ, 1984. Clinical trials of spinal manipulation: postsurgical dental pain. Am J Med;suppl:70–7. a critical appraisal and review of the literature. J Manipulative Physiol Ther;7:243–9. Cooper SA, Needle SE, Kruger GO, 1977. Comparative analgesic potency of aspirin and ibuprofen. J Oral Capell HA, Porter DR, Madhok R, Hunter JA, 1993. Surg;35:898–903. Second line (disease modifying) treatment in rheuma- Cooper SA, Gross DJ, Vogel R, Berrie P, 1983. Relative toid arthritis: which drug for which patient? Ann Rheum analgesic efficacy of ibuprofen and zomepirac Na in Dis;52:423–8. periodontal and oral surgery. Clin Pharmacol Ther;33:194. Capurso L, Koch M, 1991. La prevenzione del danno Cooper SA, Precheur H, Rosenbeck R, Engel J, 1984. The gastrico da FANS: H2 antagonisti o misoprostol? Una analgesic efficacy of ibuprofen compared to acetamino- meta-analisi degli studi clinic controllati [Prevention of phen with codeine. Clin Pharmacol Ther;35:232. NSAID-induced gastric lesions: H2 antagonists or Cooper SA, Berrie R, Cohn P, 1988. Comparison of misoprostol? A meta-analysis of controlled clinical ketoprofen, ibuprofen, and placebo in a dental surgery studies]. Clin Ter;139:179–89. pain model. Adv Ther;5:43–53. Carr D, Eisenberg E, Chalmers TC, 1993. Neurolytic Crooks RJ, Jones Da, Fiddian P, 1991. Zoster-associated celiac plexus block for cancer pain – a meta-analysis chronic pain: an overview of clinical trials with acyclovir. [abstract]. In: Proceedings 7th World Congress on Scand J Infect;78 suppl:62–8. Pain. Oxford: IASP/Elsevier;338. Cutler R, Fishbain D, Rosomoff H, Abdel-Moty E, Khalil Carroll D, Tramèr M, McQuay H, Nye B, Moore A, 1996 T, Steele-Rosomoff R, 1992. Does non-surgical pain Randomization is important in studies with pain out- treatment of chronic pain return patients to work? In: comes: systematic review of trancutaneous electrical Proceedings of American Pain Society Meeting 1992:83. nerve stimulation in acute postoperative pain. Br J Cutler RB, Fishbain DA, Rosomoff HL, Abdel-Moty E, Anaesth;77:798–803. Khalil TM, Rosomoff RS, 1994. Does nonsurgical pain Carroll D, Tramèr M, McQuay H, Nye B, Moore A, 1997. center treatment of chronic pain return patients to work? A review and meta-analysis of the literature. Transcutaneous electrical nerve stimulation in labour 53 pain: a systematic review. Br J Obstet Gynaecol;104:169–75. Spine;19:643–52. Existing systematic reviews

Dahl JB, Moiniche S, Kehlet H, 1994. Wound infiltration Eisenberg E, Carr DB, Chalmers TC, 1995. Neurolytic with local anaesthetics for postoperative pain relief. Acta celiac plexus block for treatment of cancer pain: a meta- Anaesthesiol Scand;38:7–14. analysis. Anesth Analg;80:290–5. Dahl SL, 1993. Nabumetone: a “nonacidic” nonsteroidal Ernst E, Fialka V, 1993. A review of the clinical antiinflammatory drug. Ann Pharmacother;27:456–63. effectiveness of exercise therapy for intermittent claudication. Arch Intern Med;153:2357–60. de Craen AJM, di Giulio G, LampeSchoenmaeckers AJE, Kessels AGH, Kleijnen J, 1996. Analgesic efficacy and Ernst E, Fialka V, 1993. Low-Dose-Lasertherapie: eine safety of paracetamol-codeine combinations versus kritische Prufung der klinischen Wirksamkeit [Low-dose paracetamol alone: a systematic review. BMJ;313:321–5. laser therapy: critical analysis of clinical effect]. Schweiz Devine EC, 1992. Effects of psychoeducational care for Med Wochenschr;123:949–54. adult surgical patients: a meta-analysis of 191 studies. Ernst E, Nielsen GL, Schmidt EB, 1994. Fiskeolie og Patient Educ Couns;19:129–42. reumatoid artrit [Fish oils and rheumatoid arthritis]. Devine EC, Cook TD, 1986. Clinical and cost saving Ugeskr Laeger;156:3490–5. effects of psychoeducational interventions with surgical Felson DT, Anderson JJ, Meenan RF, 1990. The patients: a meta-analysis. Res Nurs Health;9:89–105. comparative efficacy and toxicity of second-line drugs in Devine EC, Westlake SK, 1995. The effects of psycho- rheumatoid arthritis. Arthritis Rheum;33:1449–61. educational care provided to adults with cancer: meta- Felson DT, Anderson JJ, Meenan RF, 1992. Use of short- analysis of 116 studies. Oncol Nurs Forum;23:1369–81. term efficacy/toxicity tradeoffs to select second-line Dexter F, 1994. Analysis of statistical tests to compare doses drugs in rheumatoid arthritis. Arthritis Rheum;35:1117–25. of analgesics among groups. Anesthesiology; 81:610–15. Felson DT, Anderson JJ, Boers M, Bombardier C, di Fabio RP, 1992. Efficacy of manual therapy. Chernoff M, Fried B, et al., 1993. The American College Phys Ther;72:853–64. of Rheumatology preliminary core set of disease activity measures for rheumatoid arthritis clinical trials. Arthritis di Fabio RP, 1995. Efficacy of comprehensive rehabili- Rheum;36:729–40. tation programs and back school for patients with low back pain: a meta-analysis. Phys Ther;75:865–78. Felson DT, Anderson JJ, Meenan RF, 1994. The efficacy and toxicity of combination therapy in rheumatoid Dickersin K, 1989. Pharmacological control of pain arthritis. A meta-analysis. Arthritis Rheum;37:1487–91. during labour. In: Chalmers I, Enkin M, Keirse MJNC, editors. Effective care in pregnancy and childbirth. Fernandez E, Turk DC, 1989. The utility of cognitive Oxford: Oxford University Press:913–50. coping strategies for altering pain perception: a meta analysis. Pain;38:123–35. Dordain G, Aumaitre O, Eschalier A, Decamps A, 1984. La vitamine B12, une vitamine antalgique? Etude critique Ferraz MB, Atra, 1989. Metaanálise: nova opcao em de la litterature [Vitamin B12, an analgesic vitamin? pesquisa clínica. Principios básicos e um exemplo clínico Critical examination of the literature]. Acta Neurol [Meta-analysis: a new option in clinical research. Basic Belg;84:5–11. principles and a clinical example]. Rev Paul Med;107:5–9. Downie WW, 1993. Diclofenac/misoprostol. A review of Ferraz MB, Tugwell P, Goldsmith CH, Atra E, 1990. the major clinical trials evaluating its clinical efficacy and Meta-analysis of sulfasalazine in ankylosing spondylitis. upper gastrointestinal tolerability in rheumatoid arthritis J Rheumatol;17:1482–6. and osteoarthritis. Drugs;45:1–6; discussion 36–7. Finley GA, McGrath PJ, Forward SP, McNeill G, Duckro PN, Cantwell-Simmonds E, 1989. A review of Fitzgerald P, 1996. Parents’management of children’s studies evaluating biofeedback and relaxation training pain following ‘minor’ surgery. Pain;64:83–7. in the management of pediatric headache. Headache;29:428–33. Fleet RP, Dupuis G, Marchand A, Burelle D, Beitman BD, 1994. Panic disorder, chest pain and coronary artery Eandi M, della Pepa C, Rubinetto MP, 1991. Il piroxicam disease: literature review. Can J Cardiol;10:827–34. in analgesia [Piroxicam in analgesia]. Clin Ter;136:107–35. Flor H, Fydrich T, Turk DC, 1992. Efficacy of multidisciplinary pain treatment centers: a meta-analytic Edwards JE, McQuay HJ, Moore RA, 1999. Systematic review. Pain;49:221–30. review: dihydrocodeine in postoperative pain. Cochrane Library. In press. France RD, Houpt JL, Ellinwood EH, 1984. Therapeutic effects of antidepressants in chronic pain. Gen Hosp Einarson TR, McGhan WF, Bootman JL, Sabers DL, 1985. Psychiatry;6:55–63. Meta-analysis: quantitative integration of independent research results. Am J Hosp Pharmacy;42:1957–64. Fullerton T, Gengo FM, 1992. Sumatriptan: a selective 5-hydroxytryptamine receptor agonist for the acute Eisenberg E, Berkey CS, Carr DB, Mosteller F, Chalmers treatment of migraine. Ann Pharmacother;26:800–8. TC, 1994.Efficacy and safety of nonsteroidal anti- inflammatory drugs for cancer pain: a meta-analysis. Furey SA, Wksman JA, Dash BH, 1992. Nonprescription 54 J Clin Oncol;12:2756–65. ibuprofen: side effect profile. Pharmacotherapy;12:403–7. Health Technology Assessment 1998; Vol. 2: No. 12

Gabriel SE, Jaakkimainen L, Bombadier C, 1991. Risk for Gøtzsche PC, 1990. Sensitivity of effect variables in serious gastrointestinal complications related to use of rheumatoid arthritis: a meta analysis of 130 placebo nonsteroidal anti-inflammatory drugs. A meta-analysis. controlled NSAID trials. J Clin Epidemiol;43:1313–18. Ann Intern Med;115:787–96. Gøtzsche PC, 1993. Meta-analysis of NSAIDs: Gabriel SE, Jaakkimainen L, Bombadier C, 1993. The contribution of drugs, doses, trial designs, and meta- cost-effectiveness of misoprostol for nonsteroidal anti- analytic techniques. Scand J Rheumatol;22:255–60. inflammatory drug-associated adverse gastrointestinal Gøtzsche PC, Pødenphant J, Olesen M, Halberg P, 1992. events. Arthritis Rheum;36:447–59. Meta-analysis of second-line antirheumatic drugs: sample Gadsby JG, Flowerdew MW, 1997. The effectiveness of bias and uncertain benefit. J Clin Epidemiol;45:587–94. transcutaneous electrical nerve stimulation (TENS) and Grant A, 1989. The choice of suture materials and acupuncture-like transcutaneous electrical nerve techniques for repair of perineal trauma: an overview stimulation (ALTENS) in the treatment of patients with of the evidence from controlled trials. Br J Obstet chronic low back pain. In: Bombardier C, Nachemson A, Gynaecol;96:1281–9. Deyo R, de Bie R, Bouter L, Shekelle P, et al., editors. CMSG Back Module of the Cochrane Database of Grant A, Sleep J, 1989. Relief of peritoneal pain and Systematic Reviews [updated 5 December 1996]. discomfort after childbirth. In Chalmers I, Enkin M, Oxford: Cochrane Collaboration. Keirse MJNC, editors. Oxford: Oxford University Press;1347–58. Gam AN, Thorsen H, Lonnberg F, 1993. The effect of low-level laser therapy on musculoskeletal pain: a meta- Haase W, Fischer M, 1991. Statistische metaanalyse von analysis. Pain;52:63–6. multizentrischen klinischen studien mit ibuprofen im hinblick auf die kohortengrosse [Statistical meta-analysis Garzillo MJ, Garzillo TA, 1994. Does obesity cause low of multicenter clinical studies of ibuprofen with regard to back pain? J Manipulative Physiol Ther;17:601–4. cohort size]. Z Rheumatol;50:77–83. Gebhardt WA, 1994. Effectiveness of training to prevent Halpern S, Preston R, 1994. Postdural puncture job-related back pain: a meta-analysis. Br J Clin headache and spinal needle design. Meta-analyses. Psychol;33:571–4. Anesthesiology;81:1376–83. Geis GS, 1992. Overall safety of Arthrotec. Scand J Hanks GW, 1989. Controlled release morphine (MST Rheumatol;suppl 96:33–6. Contin) in advanced cancer. The European experience. Cancer;63:2378–82. Genuis ML, 1995. The use of hypnosis in helping cancer patients control anxiety, pain and emesis: a review of Harrison DL, Slack MK, 1996. Meta-analytic review recent empirical studies. Am J Clin Hypn;37:316–25. of the effect of subcutaneous sumatriptan in migraine headache. J Pharm Technol;12:109–14. Ghent WR, Eskin BA, Low DA, Hill LP, 1993. Iodine replacement in fibrocystic disease of the breast. Can J Haselkorn JK, Turner JA, Diehr PK, Ciol MA, Deyo RA, Surg;36:453–60. 1994. Meta-analysis. A useful tool for the spine researcher. Spine;19:2076–82S. Glazer S, Portenoy RK, 1991. Systemic local anesthetics in pain control. J Pain Symptom Manag;6:30–9. Hathaway D, 1986. Effect of preoperative instruction on postoperative outcomes: a meta-analysis. Nurs Good M, 1996. Effect of relaxation and music on Res;35:269–75. postoperative pain: a review. J Adv Nursing;24:905–1014. Heller CA, Ingelfinger JA, Goldman P, 1985. Goodkin K, Vrancken MAE, Feaster D, 1995. On the Nonsteroidal antiinflammatory drugs and aspirin – putative efficacy of the antidepressants in chronic, analyzing the scores. Pharmacotherapy;5:30–8. benign pain syndromes. An update. Pain Forum;4:237–47. Hersh EV, Gooper SA, Betts N, Wedell D, MacAfee K, Goodnick PJ, Sandoval R, 1993. Psychotropic treatment Quinn P, et al., 1993. Single dose and multidose analgesic of chronic fatigue syndrome and related disorders. J Clin study of ibuprofen and neclofenamate sodium after third Psychiatry;54:13–20. molar surgery. Oral Surg Oral Med Oral Pathol;76:680–7. Gøtzsche PC, 1989. Meta analysis of grip strength: most Hill AG, 1967. Review of flufenamic acid in rheumatoid common, but superfluous variable in comparative NSAID arthritis. Ann Phys Med;suppl:87–92. trials. Dan Med Bull;36:493–5. Hoffman RM, Kent DL, Deyo RA, 1991. Diagnostic Gøtzsche PC, 1989. Methodology and overt and hidden accuracy and clinical utility of thermography for lumbar bias in reports of 196 double blind trials of nonsteroidal radiculopathy. A meta-analysis. Spine;16:623–8. antiinflammatory drugs in rheumatoid arthritis. Control Holmes B, Brogden RN, Heel RC, Speight TM, Avery GS, Clin Trials;10:31–56. [erratum: Control Clin Trials 1984. Flunarizine. A review of its pharmacodynamic and 1989;50:356]. pharmacokinetic properties and therapeutic use. Gøtzsche PC, 1989. Patients’ preference in indomethacin Drugs;27:6–44. trials: an overview. Lancet;i:88–91. Holroyd KA, Penzien DB, 1986. Client variables and the Gøtzsche PC, 1989. Review of dose response studies of behavioral treatment of recurrent tension headache: NSAIDs in rheumatoid arthritis. Dan Med Bull;36:395–9. a meta analytic review. J Behav Med;9:515–36. 55 Existing systematic reviews

Holroyd KA, Penzien DB, 1989. Meta-analysis minus the Koes BW, Assendelft WJ, van der Heijden GJ, Bouter LM, analysis: a prescription for confusion. Pain;39:359–61. Knipschild PG, 1991. Spinal manipulation and mobilis- Holroyd KA, Penzien DB, 1990. Pharmacological versus ation for back and neck pain: a blinded review. non pharmacological prophylaxis of recurrent migraine BMJ;303:1298–303. headache: a meta analytic review of clinical trials. Koes BW, Bouter LM, Beckerman H, van der Heijden GJ, Pain;42:1–13. Knipschild PG, 1991. Physiotherapy exercises and back Holroyd KA, Penzien DB, Cordingley GE, 1991. pain: a blinded review. BMJ;302:1572–6. Propranolol in the management of recurrent migraine: Koes BW, van Tulder MW, van der Windt WM, Bouter a meta-analytic review. Headache;31:333–40. LM, 1994. The efficacy of back schools: a review of Holroyd KA, Athens OH, Penzien DB, Jackson MS, randomized clinical trials. J Clin Epidemiol;47:851–62. Rokicki La, Cordingley GE, 1992. Flunarizine vs Koes BW, Bouter LM, van der Heijden GJ, 1995. propranolol: a meta-analysis of clinical trials. Methodological quality of randomized clinical trials on Headache;32:256. treatment efficacy in low back pain. Spine;20:228–35. Howell C, Chalmers I, 1991. A review of prospectively Koes BW, Scholten RPM, Mens JMA, Bouter LM, 1995. controlled comparisons of epidural with non-epidural Efficacy of epidural steroid injections for low-back pain forms of pain relief during labour. Int J Obstet and sciatica: a systematic review of randomized clinical Anesth;2:1–17. trials. Pain;63:279–88. Hurwitz EL, Aker PD, Adams AH, Meeker WC, Shekelle Kubiena G, 1989. [Considerations of the placebo PG, Barr JSJ, 1996. Manipulation and mobilization of the problem in acupuncture. Reflections on usefulness, cervical spine: a systematic review of the literature. ethical justification, standardization and differentiated Spine;21:1746–60. use of placebos in acupuncture]. Wein Klin Hyman RB, Feldman HR, Harris RB, Levin RF, Malloy Wochenschr;101:362–7. GB, 1989. The effects of relaxation training on clinical Labrecque M, Dostaler LP, Rousselle R, Nguyen T, symptoms: a meta analysis. Nurs Res;38:216–20. Poirier S, 1994. Efficacy of nonsteroidal anti- Jadad AR, 1996. Meta-analysis in pain relief: a valuable inflammatory drugs in the treatment of acute renal but easily misused tool. Curr Opin Anaesth;9:426–9. colic. A meta-analysis. Arch Intern Med;154:1381–7. Jadad AR, McQuay HJ, 1996. Meta-analyses to evaluate Ladas SD, Raptis SA, 1992. Conservative treatment analgesic interventions: a systematic qualitative review of of acute pancreatitis: the use of somatostatin. their methodology. J Clin Epidemiol;49:235–43. Hepatogastoenterology;39:466–9. Jadad AR, Carroll D, Glynn CJ, McQuay HJ, 1995. Lahad A, Malter AD, Berg AO, Deyo RA, 1994. The Intravenous regional sympathetic blockade for pain relief effectiveness of four interventions for the prevention in reflex sympathetic dystrophy: a systematic review and a of low back pain. JAMA;272:1286–91. randomized, double-blind crossover study. J Pain Symptom Lancaster T, Silagy C, Gray S, 1995. Primary care Manag;10:13–20. management of acute herpes zoster: systematic review of Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds evidence from randomized controlled trials. Br J Gen DJM, Gavaghan DJ, et al., 1996. Assessing the quality of Pract;45:39–45. reports of randomized clinical trials: is blinding Laska EM, Sunshine A, Zighelbolm I, Roure C, Marrero necessary? Control Clin Trials;17:1–12. I, Wanderling J, et al., 1983. Effect of caffeine on Johanson JF, Rimm A, 1992. Optimal nonsurgical acetaminophen analgesia. Clin Pharmacol treatment of hemorrhoids: a comparative analysis of Ther;33:498–509. infrared coagulation, rubber band ligation and injection Laska EM, Sunshine A, Mueller F, Elvers WB, Siegel C, sclero-therapy. Am J Gastroenterol;87:1601–6. Rubin A, 1984. Caffeine as an analgesic adjuvant. Kaiko RF, Grandy RP, Oshlack B, Pav J, Horodniak J, JAMA;251:1711–18. Thomas G, et al., 1989. The United States experience Lataste X, Taylor P, Notter M, 1989. DHE nasal spray in with oral controlled release morphine (MS Contin the acute management of migraine attacks. tablets). Parts I and II. Review of nine dose titration Cephalalgia;9:342–3. studies and clinical pharmacology of 15 mg, 30 mg, Laurent MR, Buchanan WW, Bellamy N, 1991. Methods 60 mg, and 100 mg tablet strengths in normal subjects. of assessment used in ankylosing spondylitis clinical trials: Cancer;63:2348–54. a review. Br J Rheumatol;30:326–9. Kalso E, Tramèr M, Carroll D, McQuay H, Moore RA, Letzel H, Schoop W, 1992. Gingko-biloba-Extrakt EGb 1996. Pain relief from intra-articular morphine after knee 761 und Pentoxifyllin bei Claudicatio intermittens. surgery: a qualitative systematic review. Pain;71:642–51. Sekundaranalyse zur klinischen Wirksamkeit [Gingko Kellihan MJ, Mangino PD, 1992. Pamidronate. biloba extract EGb 761 and pentoxifylline in intermittent Ann Pharmacother;26:1262–9. claudication. Secondary analysis of the clinical Kinnison M, Powe N, Seinberg E, 1989. Results of effectiveness]. Vasa;21:403–10. randomized controlled trials of low versus high Lipton RB, Stewart WF, 1994. The epidemiology of 56 osmolality contrast media. Radiology;170:381–9. migraine. Eur Neurol;34 suppl 2:6–11. Health Technology Assessment 1998; Vol. 2: No. 12

Loosemore TM, Chalmers TC, Dormandy JA, 1994. A Moore RA, McQuay HJ, 1997. Single-patient data meta- meta-analysis of randomized placebo control trials in analysis of 3453 postoperative patients: oral tramadol Fontaine stages III and IV peripheral occlusive arterial versus placebo, codeine and combination analgesics. disease. Int Angiol;13:133–42. Pain;69:287–94. Lycka BA, 1990. Postherpetic neuralgia and systemic Moore A, McQuay H, Gavaghan D, 1996. Deriving corticosteroid therapy. Efficacy and safety. Int J dichotomous outcome measures from continuous Dermatol;29:523–7. data in randomised controlled trials of analgesics. Macarthur C, Saunders N, Feldman W, 1995. Helicobacter Pain;66:229–37. pylori, gastroduodenal disease, and recurrent abdominal Moore A, Collins S, Carroll D, McQuay H, 1997. pain in children. JAMA;273:729–34. Paracetamol with and without codeine in acute pain: Magni G, 1991. The use of antidepressants in the a quantitative systematic review. Pain;70:193–201. treatment of chronic pain. A review of the current Moore RA, Collins S, McQuay HJ, 1997. Variation in the evidence. Drugs;42:730–48. placebo effect; the impact on individual trials and conse- Mahon WA, de Gregorio M, 1985. Benzydamine: a quences for meta-analysis. In: Jensen TS, Hammond DL, critical review of clinical data. Int J Tissue React;7:229–35. Jensen TS, editors. Proceedings of the 8th World Congress on Pain. Progress in Pain Research and Major P, 1992. Bor ikke-steroide antiinflammatoriske Management, vol 2. Seattle: IASP Press:469–76. midler brukes ved akutte blotdelsskader? [Would nonsteroidal anti-inflammatory agents be used in acute soft Moore RA, Nye BA, Carroll D, Wiffen PJ, Tramèr M, tissue injuries?] Tidsskr Nor Laegeforen;112:1614–15. McQuay HJ, 1998. A systematic review of topically-applied non-steroidal anti-inflammatory drugs. BMJ;316:333–8. Malone MD, Strube MJ, 1988. Meta-analysis of non- medical treatments for chronic pain. Pain;34:231–44. Moote CA, 1996. Ibuprofen arginine in the management [erratum: Pain 1989;37:128]. of pain. A review. Clin Drug Invest;11:1–7. Malone MD, Strube MJ, Scogin FR, 1989. Reply to Morton SC, Williams MS, Keeler EB, Gambone JC, Kahn Holroyd and Penzien. Pain;39:362–3. KL, 1994. Effect of epidural analgesia for labor on the Cesarean delivery rate. Obstet Gynecol;83:1045–52. Max MB, 1995. Thirteen consecutive well-designed randomized trials show that antidepressants reduce pain Mullen PD, Laville EA, Biddle AK, Lorig K, 1987. Efficacy in diabetic neuropathy and postherpetic neuralgia. Pain of psychoeducational interventions on pain, depression, Forum;4:248–53. and disability in people with arthritis: a meta analysis. McGee J, Alexander M, 1979. Phenothiazine analgesia: J Rheumatol;14 supp:33–9. fact or fantasy? Am J Hosp Pharm;36:633–40. Murphy NG, Zurier RB, 1991. Treatment of rheumatoid McQuay HJ, 1995. Pre-emptive analgesia: a systematic arthritis. Curr Opin Rheumatol;3:441–8. review of clinical studies. Ann Med;27:249–56. Naldi L, Zucchi A, Brevi A, Cavalieri d’Oro L, Cainelli T, McQuay H, Moore RA, 1996. Epidural steroids [letter]. 1990. Corticosteroids and post-herpetic neuralgia. Anaesth Intensive Care;24:284–6. Lancet;336:947. McQuay H, Carroll D, Jadad AR, Wiffen P, Moore A, Nicholas JJ,1994. Physical modalities in rheumatological 1995. Anticonvulsant drugs for management of pain: a rehabilitation. Arch Phys Med Rehabil;75:994–1001. systematic review. BMJ;311:1047–52. NIH Technology Assessment Panel, 1996. Integration of McQuay H, Carroll D, Moore A, 1996. Variation in the behavioural and relaxation approaches into the treat- placebo effect in randomised controlled trials of ment of chronic pain and insomnia. JAMA;176:313–18. analgesics: all is as blind as it seems. Pain;64:331–5. Olesen J, 1986. Role of calcium entry blockers in the McQuay H, Tramèr M, Nye BA, Carroll D, Wiffen PJ, prophylaxis of migraine. Eur Neurol;25 suppl:72–9. Moore RA, 1996. A systematic review of antidepressants Onghena P, van Houdenhove B, 1992. Antidepressant- in neuropathic pain. Pain;68:217–27. induced analgesia in chronic non-malignant pain: a McQuay HJ, Carroll D, Moore RA, 1999. Injected meta-analysis of 39 placebo-controlled studies. morphine in postoperative pain: a quantitative systematic Pain;49:205–19. review. J Pain Symptom Manage. In press. Ottenbacher K, di Fabio RP, 1985. Efficacy of spinal McQuay HJ, Carroll D, Moore RA, 1997. Radiotherapy manipulation/mobilization therapy. A meta analysis. for painful bony metastases: a systematic review. Clin Spine;10:833–7. Oncol;42:71–9. Pace NL, 1990. Prevention of succinylcholine myalgias: Mendelson G, 1977. Acupuncture analgesia. I. Review of a meta-analysis. Anesth Analg;70:477–83. clinical studies. Aust N Z J Med;7:642–8. Pace NL, 1993. The best prophylaxis for succinylcholine Mizraji M, 1990. Clinical response to etodolac in the myalgias: extension of a previous meta-analysis [letter]. management of pain. Eur J Rheumatol Inflamm;10:35–43. Anesth Analg;77:1080–1. Montastruc JL, Senard JM, 1992. Medicaments Patel M, Gutzwiller F, Paccaud F, Marazzi A, 1989. A anticalciques et prophylaxie de la migraine. meta-analysis of acupuncture for chronic pain. Int J Pathol Biol Paris;40:381–8. Epidemiol;18:900–6. 57 Existing systematic reviews

Penzien DB, Holroyd KA, 1987. The behavioral Radack K, Wyderski RJ, 1990. Conservative management treatment of migraine headache: a meta-analytic of intermittent claudication. Ann Intern Med;113:135–46. review of the literature. Dissertation Abstracts Int;47: Rahimtoola S, 1985. A perspective on the three large No DA8629941. multicenter randomized clinical trials of coronary bypass Penzien DB, Holroyd KA, Hursey KG, Holm JE, surgery for chronic stable angina. Circulation;72:V123–35. Wittchen H, 1987. The behavioral treatment of Reeve J, Menon D, Corabian P, 1996. Transcutaneous migraine headache: a meta-analytic review of the electrical nerve stimulation (TENS): a technology literature. [Based on doctoral dissertation by assessment. Int J Technol Assess Health Care;12:299–324. Penzien DB]. Richardson PH, Williams AC, 1993. Meta-analysis of Penzien DB, Rains JC, Holroyd KA, 1992. A review of antidepressant-induced analgesia in chronic pain: alternative behavioral treatments for headache. Mis comment [letter]. Pain;52:247–9. Psychol;17:8–9. Riedemann PJ, Bersinic S, Cuddy LJ, Torrance GW, Philipp M, Fickinger M, 1993. Psychotropic drugs Tugwell PX, 1993. A study to determine the efficacy of in the management of chronic pain syndromes. safety and tenoxicam versus piroxicam, diclofenac and Pharmacopsychiatry;26:221–34. indomethacin in patients with osteoarthritis: a meta- Plantema F, 1986. Worldwide studies comparing piroxicam analysis. J Rheumatol;20:2095–103. and naproxen. Acta Obstet Gynecol Scand Suppl;138. Riethmüller-Winzen H, 1987. Flupirtine in the treatment Pope JE, Anderson JJ, Felson DT, 1993. A meta-analysis of of post-operative pain. Postgrad Med J;63:61–5. the effects of nonsteroidal inti-inflammatofy on blood pressure. Arch Intern Med;153:477–84. Robinson M, Mills RJ, Euler AR, 1991. Ranitidine prevents duodenal ulcers associated with non-steroidal Porzio F, 1993. Meta-analysis of three double-blind anti-inflammatory drug therapy. Aliment Pharmacol comparative trials with sustained-release etodolac in the Ther;5:143–50. treatment of osteoarthritis of the knee. Rheumatol Int;13:S19–24. Robinson RG, Preston DF, Schiefelbein M, Baxter KG, 1995. Strontium 89 therapy for the palliation of pain due Porzio F,1995. Sublingual piroxicam FDDF in the to osseous metastases. JAMA;274:420–4. management of pain associated with rheumatic and nonrheumatic conditions; the Italian experience. Eur J Roderick PJ, Wilkes HC, Meade TW. The gastrointestinal Rheumatol Inflamm;15:11–17. toxicity of aspirin: an overview of randomised controlled trials. Br J Clin Pharmacol;35:219–26. Post B, Philbrick J, 1988. Do corticosteroids prevent postherpetic neuralgia? A review of the evidence. Rosch W, 1989. Efficacy of cisapride in the treatment of J Am Acad Dermatol;18:605–10. epigastric pain and concomitant symptoms in non ulcer dyspepsia. Scand J Gastroenterol;24 suppl:165. Powell FC, Hanigan WC, Olivero WC, 1993. A risk/benefit analysis of spinal manipulation therapy Saag KG, Criswell LA, Sems DM, Nettleman MD, Kolluri for relief of lumbar or cervical pain [see comments]. S, 1996. Low dose corticosteroids in rheumatoid arthritis: Neurosurgery;33:73–8; discussion 78–9. a meta-analysis of their moderate-term effectiveness. Arthritis Rheum;39:1818–25. Poynard T, Valterio C, 1994. Meta-analysis of hydroxy- ethylrutosides in the treatment of chronic venous Sacks HS, Ancona-Berk VA, Berrier J, Nagalingam R, insufficiency. Vasa;23:244–50. Chalmers TC, 1988. Dipyridamole in the treatment of angina pectoris: a meta-analysis. Clin Pharmacol Pradalier A, Vincent D, 1992. Migraine et anti- Ther;43:610–15. inflammatoires non-steroidiens [Migraine and nonsteroidal anti-inflammatory agents]. Pathol Biol Sand T, 1989. Which factors affected reported headache Paris;40:397–405. incidences after lumbar myelography? A statistical Puett DW, Griffin MR, 1994. Published trials of analysis of publications in the literature. nonmedicinal and noninvasive therapies for hip and Neuroradiology;31:55–9. knee osteoarthritis. Ann Intern Med;121:133–40. Schiassi M, Bianchini C, Restelli A, Zoni G, 1989. Pustaver MR, 1994. Mechanical low back pain: etiology Tolerability profile of the antiinflammatory compound and conservative management. J Manipulative Physiol imidazole salicylate: a metanalysis of safety data in 1408 Ther;17:376–84. patients. Int J Tiss Reac;11:321–6. Quan M, 1994. Pelvic inflammatory disease: diagnosis Schmader K, Studenski S, 1989. Are current therapies and management. J Am Board Fam Pract;7:110–23. useful for the prevention of postherpetic neuralgia? A critical analysis of the literature. J Gen Intern Med;4:83–9. Radack K, Deck C, 1987. Do nonsteroidal anti inflammatory drugs interfere with blood pressure control Schuhfried O, Fialka-Moser V, 1995. Iontophoresis in the in hypertensive patients? J Gen Intern Med;2:108–12. treatment of pain [in German]. Wien Med Wochenschr;145:4–8. Radack K, Deck C, 1991. Beta-adrenergic blocker therapy does not worsen intermittent claudication in subjects Schulz KF, Grimes DA, Altman DG, Hayes RJ, 1996. with peripheral arterial disease. A meta-analysis of Blinding and exclusions after allocation in randomised randomized controlled trials. Arch Intern controlled trials: survey of published parallel group trials 58 Med;151:1769–76. in obstetrics and gynaecology. BMJ;312:742–4. Health Technology Assessment 1998; Vol. 2: No. 12

Shekelle PG, Adams AH, Chassin MR, Hurwitz EL, Brook Thomson AM, Hillier VF, 1994. A re-evaluation of the RH, 1992. Spinal manipulation for low-back pain. Ann effect of pethidine on the length of labour. J Adv Intern Med;117:590–8. Nurs;19:448–56. Siminoski K, Josse RG, Bowyer M, 1996. Calcitonin in the Trabant H, Widdra W, De LS, 1990. Efficacy and safety treatment of osteoporosis. Can Med Assoc J;155:962–5. of intranasal buserelin acetate in the treatment of endometriosis: a review of six clinical trials and comparison Simkin P, 1989.Non-pharmacological methods of pain with danazol. Prog Clin Biol Res;323:357–82. relief during labour. In: Chalmers I, Enkin M, Keirse MJNC, editors. Effective care in pregnancy and Tramèr M, Williams J, Carroll D, Wiffen PJ, McQuay HJ, childbirth. Oxford: Oxford University Press:899–911. Moore RA, 1997. Systematic review of direct comparisons of non-steroidal anti-inflammatory drugs given by differ- Simmons J, Stavinoha W, Knodel L, 1984. Update and ent routes for acute pain. Acta Anaesth Scand;42:71–9. review of chemonucleolysis. Clin Orthop;Mar:51–60. Turner JA, 1995. Efficacy of antidepressant medication Sindhu F, 1996. Are non-pharmacological nursing for chronic pain problems. Pain Forum;4:254–5. interventions for the management of pain effective? A meta-analysis. J Adv Nurs;24:1152–9. Turner JA, Denny MC, 1993. Do antidepressant medications relieve chronic low back pain? J Fam Smith MC, Holcombe JK, Stullenbarger E, 1994. A Pract;37:545–53. meta-analysis of intervention effectiveness for symptom management in oncology nursing research. Oncol Nurs Turner JA, Ersek M, Herron L, Deyo R, 1992. Surgery for Forum;21:1201–9; discussion 1209–10. lumbar spinal stenosis. Attempted meta-analysis of the literature. Spine;17:1–8. Solomon GD, 1989. Verapamil in migraine prophylaxis – a five year review. Headache;29:425–7. Turner JA, Deyo RA, Loeser JD, von Korff M, Fordyce WE, 1994. The importance of placebo effects in pain Spierings EL, Messinger HB, 1988. Flunarizine vs. treatment and research. JAMA;271:1609–14. pizotifen in migraine prophylaxis: a review of comparative studies. Cephalalgia;8 suppl:27–30. Turner JA, Loeser JD, Bell KG, 1995. Spinal cord stimulation for chronic low back pain; a systematic Stalnikowicz R, Rachmilewitz D, 1993. NSAID-induced literature synthesis. Neurosurgery;37:1088–96. gastroduodenal damage: is prevention needed? a review and meta-analysis. J Clin Gastroenterol;17:238–43. Vaiani G, Grossi E, 1990. Meta-analysis of italian clinical trials of nabumetone. Drugs;40:48–9. Stewart WF, Simon D, Shechter A, Lipton RB, 1995. Population variation in migraine prevalence: a meta- van der Heihden CJM, Beurskens AJ, Koes BW, analysis. J Clin Epidemiol;48:269–80. Assendelft WJ, de Vet HC, Bouter LM, 1995. The efficacy of traction for back and neck pain: a systematic, Stulberg B, Bauer T, Belhobek G, 1990. Making core blinded review of randomized clinical trial methods. decompression work. Clin Orthop;Dec:186–95. Phys Ther;75:93–104. Suls J, Fletcher B, 1985. The relative efficacy of avoidant van der Heijden CJM, van der Windt DAW, Kleijnen J, and nonavoidant coping strategies: a meta analysis. Koes BW, Bouter LM, 1996. Steroid injections for Health Psychol;4:249–88. shoulder disorders: a systematic review of randomized Suls J, Wan CK, 1989. Effects of sensory and procedural clinical trials. Br J Gen Pract;46:309–16. information on coping with stressful medical procedures van der Hoogen HM, Koes BW, van Eijk JT, Bouter LM, and pain: a meta-analysis. J Consult Clin Psychol;57:372–9. 1995. On the accuracy of history, physical examination, Sutters KA, Miaskowski C, 1992. The problem of pain and erythrocyte sedimentation rate in diagnosing low in children with cancer: a research review. Oncol Nurs back pain in general practice. A criteria-based review of Forum;19:465–71. the literature. Spine;20:318–27. Teasell RW, Harth M, 1996. Functional restoration: Vischer T, 1987. Efficacy and tolerability of tenoxicam: an returning patients with chronic low back pain. overview. Eur J Rheumatol Inflamm;9:51–7. Spine;21:844–7. Vischer T, 1987. Efficacy and tolerability of tenoxicam: an ter Riet G, Kleijnen J, Knipschild P, 1990. Acupuncture overview. Scand J Rheumatol Suppl;65:107–12. and chronic pain: a criteria-based meta-analysis. J Clin Volmink J, Lancaster T, Gray S, Silagy C, 1996. Epidemiol;43:1191–9. Treatments for postherpetic neuralgia – a systematic Tfelt Hansen P, 1986. Efficacy of beta blockers in review of randomized controlled trials. Fam migraine. A critical review. Cephalalgia;6 suppl:15–24. Pract;13:84–91. Tfelt-Hansen P, 1993. Sumatriptan for the treatment of Watts RW, Silagy CA, 1995. A meta-analysis on the efficacy migraine attacks – a review of controlled clinical studies. of epidural corticosteroids in the treatment of sciatica. Cephalalgia;13:238–44. Anaesth Intensive Care;23:564–9. Thomas D, Williams RA, Smith DS, 1980. The frozen White KP, Harth M, 1996. An analytical review of shoulder: a review of manipulative treatment. Rheumatol 24 controlled clinical trials for fibromyalgia syndrome Rehab;19:173–9. (FMS). Pain;64:211–19. 59 Existing systematic reviews

Wilkie DJ, Savedra MC, Holzemer WL, Tesler MD, Paul Wright JM, 1994. Review of the symptomatic treatment of SM, 1990. Use of the McGill Pain Questionnaire to diabetic neuropathy. Pharmacotherapy;14:689–97. measure pain: a meta-analysis. Nurs Res;39:36–41. Yee LY, Lopez JR, 1992. Transdermal fentanyl. Windle M, Booker L, Rayburn W, 1989. Postpartum pain Ann Pharmacother;26:1393–9. after vaginal delivery. A review of comparative analgesic trials. J Reprod Med;34:891–5. Zhang WY, Li WPA, 1994. The effectiveness of topically Wood MJ, 1996. How to measure and reduce the burden applied capsaicin. A meta-analysis. Eur J Clin of zoster-associated pain. Scand J Infect Dis;suppl:55–8. Pharmacol;46:517–22. Wood MJ, Kay R, Dworkin RH, Soong SJ, Whitley RJ, Zhang WY, Li Wan Po A, 1996. Analgesic efficacy of 1996. Oral acyclovir therapy accelerates pain resolution paracetamol and its combination with codeine and in patients with herpes zoster: a meta-analysis of placebo- caffeine in surgical pain: a meta-analysis. J Clin Pharm controlled trials. Clin Infect Dis;22:341–7. Ther;21:261–82.

60 Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 8 Paracetamol with and without codeine in acute pain

Summary which is commonly prescribed as well as being available without prescription. In England in 1995, A systematic review of RCTs to assess the analgesia paracetamol alone accounted for over 5 million obtained from single oral doses of paracetamol prescriptions for adults (16% of total non-opiate alone and in combination with codeine in analgesic prescriptions), with 4.5 million pre- postoperative pain found: scriptions of paediatric suspensions.2 In combination with codeine, paracetamol accounted • 39 trials of paracetamol against placebo with for a further 6.4 million prescriptions (20% of 4124 patients total non-opiate analgesics). Paracetamol alone • 21 trials of paracetamol plus codeine against and in combination with a variety of opioids placebo with 450 patients accounted for 93% of prescriptions in this • 12 trials of paracetamol plus codeine against the BNF classification. same dose of paracetamol with 794 patients. Policy decisions and guidelines are increasingly Pain relief information was extracted and con- being made on the basis of hard evidence. Trying verted into dichotomous information, that is, to judge the relative efficacy of analgesics against numbers of patients with at least 50% pain relief. one another is not easy because there are few such Wide variation in responses to placebo (0–72%) direct comparisons. Only five direct comparisons and active drugs (5–89%) were observed. were found of paracetamol, 1000 mg, and ibuprofen, 400 mg, in acute pain.3 In postoperative pain states paracetamol, 1000 mg, alone against placebo had an NNT of 4.6 (95% CI, Relative efficacy can also be determined indirectly, 3.9–5.4) and paracetamol, 600/650 mg, alone an from comparisons of each analgesic with placebo, NNT of 5.3 (95% CI, 4.1–7.2). Paracetamol, using a common descriptor of efficacy, and then 600/650 mg, plus codeine, 60 mg, against placebo comparing the results for various analgesic had a better NNT of 3.1 (95% CI, 2.6–3.9), with no interventions, both pharmacological and non- overlap of 95% CIs with paracetamol, 600/650 mg, pharmacological. In this review of paracetamol, its alone. In direct comparisons of paracetamol plus analgesic efficacy is compared with information codeine with paracetamol alone, the additional about other drugs, again determined by similar analgesic effect of 60 mg of codeine added to quantitative systematic reviews. paracetamol was 11 extra patients in every 100 achieving at least 50% pain relief. In indirect comparisons of each with placebo, it was 14 extra Methods patients per 100. This gave an NNT for adding codeine, 60 mg, of 7.7 (95% CI, 5.1–20). RCTs of paracetamol in postoperative pain (post dental extraction, postsurgical or postpartum The results confirm that paracetamol is an effective pain) were sought. A number of different search analgesic and that codeine, 60 mg, added to para- strategies were used to identify eligible reports in cetamol produces worthwhile additional pain relief MEDLINE (1966–May 1996), EMBASE (1980– 96), even in single oral doses. Cochrane Library (March 1996) and the Oxford Pain Relief database (1950–94).4 The words This systematic review has been published in part ‘paracetamol’, ‘acetaminophen’, and ‘trial’ by Moore and colleagues.1 were used in a free text search, both alone and in combination, and without restriction to language. Additional reports were identified from reference Introduction lists of retrieved reports, review articles (including a recent systematic review of paracetamol plus Paracetamol is an important non-opiate analgesic, codeine)5 and textbooks. 61 Paracetamol with and without codeine in acute pain

Inclusion criteria for paracetamol Relative benefits estimates were calculated with Neither pharmaceutical companies nor authors of 95% CIs using a random effects model;12 the papers were contacted for unpublished reports. random effects model was chosen because it Abstracts and review articles were not considered. produces the most conservative estimate (homo- The inclusion criteria used were: geneity was assumed when p > 0.1). NNTs and 95% CIs were calculated using the method • randomised allocation to treatment groups described by Cook and Sackett.13 A statistically which compared either paracetamol or a significant difference from control was assumed paracetamol and codeine combination with when the 95% CI of the relative benefit did placebo or a paracetamol and codeine combi- not include 1. nation with the same dose of paracetamol alone • full journal publication • established postoperative pain, with the pain Results outcome measured using a five-point pain relief scale with standard wording (none, slight, Paracetamol versus placebo moderate, good, complete) or a four-point pain The literature searches found 37 reports of intensity scale (none, mild, moderate or severe) 39 trials which fulfilled the inclusion criteria; or a VAS for pain relief or pain intensity, 2530 patients were given paracetamol and TOTPAR or SPID (at 4, 5 or 6 hours) as a 1594 patients placebo. Details of the trials derived pain relief outcome (or sufficient data are presented in Table 23. provided to allow their calculation) • postoperative oral administration Of the trials found, 21 investigated oral surgery • adult patients pain (post-dental pain, predominantly third molar • baseline pain of moderate to severe intensity extraction with bone removal), eight postsurgical (for VAS this equates to > 30 mm)6 pain (elective general, gynaecological and • double-blind design. orthopaedic surgery) and ten postpartum pain (episiotomy and post-Caesarean section). The Reports for the relief of other pain conditions were doses of paracetamol administered were 500 mg in excluded, as were those for paracetamol used in six trials, 600 mg in six trials, 650 mg in 11 trials combination with drugs other than codeine and and 1000 mg in 20 trials; for analysis purposes, data trials where the number of patients per treatment from paracetamol, 600 mg, and paracetamol, 650 group was less than ten.7 In postpartum pain, trials mg, were combined. One report on episiotomy were included if the pain investigated resulted from provided dichotomous information on the overall episiotomy or Caesarean section combined with patient global rating of pain relief.14 This report uterine cramps but trials investigating uterine was included. The proportion of patients with good cramps alone were excluded. or excellent pain relief was used.

Data extraction and analysis The variation in placebo response rates (i.e. the The numbers of patients treated, the mean proportion of patients with at least 50% pain relief) TOTPAR, SPID, VAS TOTPAR or VAS SPID, was from 0% to 72% of patients with at least 50% study duration and the dose given were taken maxTOTPAR (see Figure 24, page 69). The placebo from each report. Information on adverse events response rate ranged from 0% to 72% in post- was also extracted. For each report, the mean dental pain, from 11% to 48% in postsurgical pain, TOTPAR, SPID, VAS TOTPAR or VAS SPID and from 0% to 34% in postpartum pain. The values for active and placebo were converted to variation in response rates with all doses of para- % maxTOTPAR or % maxSPID by division into cetamol was 5–83% (Figure 24). The mean response the calculated maximum value.8 The proportion rate for paracetamol, 600/650 mg, was 42% and of patients in each treatment group who achieved for placebo 23%. at least 50% maxTOTPAR was calculated using verified equations.9–11 These proportions were Combining data across conditions, the pooled then converted into the number of patients relative benefits for all doses of paracetamol achieving at least 50% maxTOTPAR by multi- versus placebo were significant (Table 24). plying by the total number of patients in the Paracetamol, 600/650 mg, compared with treatment group. Information on the number placebo in single dose administration had an of patients with at least 50% maxTOTPAR for NNT for at least 50% pain relief of 5.3 (95% treatment and placebo was then used to calculate CI, 4.1–7.2) and at 1000 mg the NNT was 4.6 62 relative benefit and NNT. (95% CI, 3.9–5.4), with overlap between the CIs. Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 23 Trials of paracetamol versus placebo

Study Condition Design, study Outcome Dosing Analgesic Remedication Withdrawals Adverse effects Quality and number duration and measures regimen outcome results and exclusions score of patients follow-up

Beaver & Episiotomy and RCT,double-blind, single PI (4-point scale) Placebo Paracetamol signifi- Patients allowed 108 analysed. NSD between 4 McMillan, uterine cramp oral dose, 5 parallel PR (5-point scale) (n = 22); cantly superior remedication after ‘There were no groups.All mild 1980 (vaginal delivery groups. 3 hour washout 50% PR (y/n) paracetamol, to placebo 2 hours if pain drop-outs.’ and subjective. < 48 hours). prior to start and medi- Global rating 1000 mg (p < 0.05–0.01) returned to pre- Total numbers n = 108 cation given > 30 minutes (5-point scale) (n = 22). for all measures medication levels. reporting adverse Age ? before or > 2 hours NB: episiotomy except those After remedication, effects (number of after patient’s meal. pain assessed as based on change PR = 0 for all effects): placebo Evaluated in hospital ‘right now’ and in PI. further time points. 7/22 (8); para- by nurse observer at uterine cramp cetamol 11/22 (13). 0 hours then hourly pain as ‘during for 6 hours or until the last hour’. pain returned to premedication level.

Bentley & Third molar RCT,double-blind, single PI (10-point scale) Placebo Paracetamol Rescue analgesic, 120 patients 53 patients reported 3 Head, 1987 bony impacted. oral dose, four parallel PR (5-point scale) (n = 17); significantly Tylenol no 3. analysed. Exclu- one or more; 86 n = 128 groups. No information paracetamol, superior for all For patients with sions: 3 took no reported in total, Mean age, on anaesthesia except 1000 mg measures of remedication at medication; 1 majority being ‘mid-20s’ ‘no sedative or narcotic (n = 41). efficacy. < 5 hours, last PI took only part; dizziness, drowsiness, agents were used before, and PR scores 1 took none until nausea and vomiting. during or after surgery’. carried on for all day after surgery; NSD between Self-assessed at home at further time points. 1 remedicated treatment groups. 0, 1, 2, 3, 4, 5 hours then after 30 minutes; Total numbers reports posted to 1 vomited within reporting adverse investigator. 30 minutes of effects (number of taking medication; effects): paracetamol 1 did not return 21/42 (31); placebo forms. 9/19 (16).

Berry, et al., Episiotomy RCT,double-blind, single PI (5-point scale) Placebo Total PI and PR After reasonable No details. “None of the patients 3 1975 n = 225 oral dose, 3 parallel PR (5-point scale, (n = 76); scores showed para- period, rescue experienced adverse Age 15+ years. groups. 12 hour washout but non-standard) paracetamol, cetamol significantly analgesia could be drug reactions.” prior to start.Assessed Gastric discomfort 1000 mg superior to placebo. prescribed at the Number reporting by observer(s) in hospital (4-point scale) (n = 76). Global rating > good. investigator’s gastric discomfort: at 0, 0.5, 1, 2, 3, 4 hours. Global rating Placebo 18/76. discretion and patient paracetamol (slight) (5-word scale) Paracetamol 43/76. regarded as a treat- 2/76; placebo ment failure (no (moderate) 2/76. information on how data handled). Placebo 23/76; paracetamol 2/76.

Cooper & Impacted third RCT,double-blind, 2 single PI (4-point scale) Study 2: Placebo Parametric analysis Remedication 160 analysed. None serious 5 Beaver, molar oral dose studies PR (5-point scale) (n = 40); concluded that consisted of taking Exclusions: reported. Most 1976 n = 216 (paracetamol = study 2), Pain half gone? (y/n) paracetamol, paracetamol signifi- second envelope and 30 patients did not common: drowsiness, Age 16+ years 4 parallel groups. No Had the patient 600 mg (n = 40). cantly superior to evaluating it in same return forms; 12 nausea and headache. information on anaes- fallen asleep during placebo (p < 0.01). way as the first. completed forms Numbers reporting thesia. Self-assessed at the hour? (y/n) Unclear how data improperly; 4 took adverse effects home at 0, 1, 2, 3 hours was then handled. concomitant medi- (number of effects): (questionnaire) – posted 112/160 patients cation; 6 required placebo 5/40 (5); to investigator. required no medication; 4 paracetamol 5/40 (7). remedication. were randomly deleted to even out numbers on each treatment.

Cooper, Impacted third RCT,double-blind, single PI (4-point scale) Placebo (n = 38); Paracetamol showed t > 1 hour before 247 analysed. Numbers reporting 4 et al., 1980 molar oral dose, 6 parallel groups. PR (5-point scale) paracetamol, significant analgesic remedication (unclear Exclusions: 21 lost adverse events n = 298 No information on Global rating 500 mg (n = 37). efficacy for all how data handled). to follow-up; 10 (number of effects): Mean age anaesthesia. Self-assessed (5-point scale) measures. Remedication dropped out before placebo 6/38 (7); ‘early 20s’ at home at 0, 1, 2, 3, 50% relief of < 4 hours: placebo ingesting medication paracetamol 3/37 (6). 4 hours (questionnaire) – baseline pain (y/n) 26/38; paracetamol (no details); 20 majority collected 1 week Time to 25/37. ingested medication later by observer, few remedication but excluded for returned by mail (patients protocol violations telephoned if problems (no details). encountered)

continued 63 Paracetamol with and without codeine in acute pain

TABLE 23 contd Trials of paracetamol versus placebo

Cooper, Impacted third RCT,double-blind, single PI (4-point scale) Placebo All active Remedication after 200 analysed. None serious 4 et al., 1981 molar oral dose, 5 parallel groups. PR (5-point scale) (n = 37); treatments signifi- hour 1 if needed (last Exclusions: 17 did reported. Numbers n = 248 Either general or local Global rating paracetamol, cantly superior to score used for all not ingest medi- reporting adverse Mean age anaesthetic. Self-assessed (5-point scale) 650 mg placebo for all further time points). cation; 31 ingested effects (number of ‘early 20s’ at home at 0, 1, 2, 3, 50% relief of (n = 37). measures.(For mean Remedication medication but effects): placebo 4 hours (questionnaire). baseline pain (y/n) values of SPID, peak < 4 hours: placebo violated protocol 4/37 (5); paracetamol Time to PID, total PR, etc. 20/37; paracetamol (remedicated before 12/37 (15). remedication see Table 2). 2/37. 1st hour observation, constant deviation of more than 15 minutes from evaluation times, did not return questionnaire, lost to follow-up)

Cooper, Oral surgery RCT,double-blind, single PI (4-point scale) Placebo For all measures Remedication after 99 analysed. None serious 3 et al., 1986 (involving bone oral dose, 3 parallel groups, PR (5-point scale) (n = 22); paracetamol first hour if needed Exclusions: 6 reported. Over half removal) single centre and 1 surgeon. Pain half gone? (y/n) paracetamol, significantly superior (last score used for did not require those reported were n = 112 No information on Global rating 1000 mg to placebo all further time analgesia; 3 fell drowsiness; there Age 16+ years anaesthesia. Self-assessed (5-point scale) (n = 38). (p < 0.05). (see points). asleep; other 4 were two reports at home at 0, 1, 2, 3, 4, 5, Time to Table 2 for mean were for ‘various of nausea. Numbers 6 hours (diary). remedication values of SPID, protocol of events reported: TOTPAR, 50% violations’. paracetamol 12; reduction, etc.) placebo 0.

Cooper, Impacted third RCT,double-blind, single PI (4-point scale) Placebo Paracetamol Remedication after 143 analysed. None serious 4 et al., 1988 molar oral dose, 3 parallel PR (5-point scale) (n = 40); appeared clinically first hour (last or Exclusions: 11 lost reported. Numbers n = 165 groups, single centre and Pain half gone? (y/n) paracetamol, more effective than baseline score used to follow-up; 8 did of events reported: Age range 1 surgeon. Local anaes- Global rating 600 mg placebo but was for all further time not require placebo 3; 18–57 years thetic with sedative and/or (5-point scale) (n = 36). not significantly points). medication; 3 for paracetamol 8. nitrous oxide. 4-hour Time to superior. ‘various protocol washout prior to start. remedication violations’. Self-assessed at home at 0, 0.5, 1, 2, 3, 4, 5, 6 hours (diary).

Cooper, Removal of RCT,double-blind, single PI (4-point scale) Placebo Paracetamol Remedication after 184 analysed. 190 evaluated (all 5 et al., 1989 impacted teeth oral dose, 3 parallel groups. PR (5-point scale) (n = 64); significantly superior first hour (last or Exclusions: 4 slept who ingested n = 194 local anaesthetic with iv Pain half gone? (y/n) paracetamol, to placebo baseline score used through more than medication). None Age 16+ years sed., atropine and/or Global rating 1000 mg (p < 0.05–< 0.001) for all further time two observations; serious reported – nitrous oxide. 4-hour (5-point scale) (n = 59). for all measures. points).Approximately 2 lost to follow-up; drowsiness being washout prior to start. Time to (see Table 2 for 50% paracetamol and 2 did not need most common. Self-assessed at home at remedication mean values, SPID, 36% placebo com- medication; 1 had Numbers reporting 0, 0.5, 1, 2, 3, 4, 5, 6 hours TOTPAR, etc.) pleted least 4 hours inadequate baseline adverse effects (diary). before remedication. PI; 1 failed to com- (number of effects): Significantly longer plete evaluations placebo 7/64 (7); mean time to at set times. paracetamol remedication for 11/63 (13). paracetamol (p < 0.05) than for placebo.

Cooper & Removal of RCT,double-blind, single PI (4-point scale) Placebo Paracetamol was t > 1 hour before 226 analysed. All adverse effects 3 Kupperman, one or more oral dose, 6 parallel PR (5-point scale) (n = 44); only active drug not remedication, data Exclusions; 13 mild. Numbers 1991 impacted teeth groups. Local anaesthetic Global rating paracetamol, significantly superior included and baseline did not require reporting adverse n = 247 (lidocaine + epinephrine) (5-point scale) 650 mg to placebo for or last score (most medication; 3 lost effects (number of Age ‘young adult’ with i.v. diazepam and Time to (n = 37). any measure. severe) used for all to follow-up, 2 effects): placebo methohexital (nitrous remedication further time points. remedicated. with 7/44 (9); paraceta- oxide also used on slight pain before mol 6/37 (7). occasion). second hour Self-assessed at home at observation; 2 0, 0.5 hours then hourly remedicated for 6 hours (diaries). before first hour observation; 1 fell asleep for over 2 hours.

64 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 23 contd Trials of paracetamol versus placebo

Dionne, Impacted third RCT,double-blind, single PI (4-point scale) Placebo Paracetamol not t ≥ 2 hours before 124 analysed. All reported mild. 3 et al., 1994 molar oral dose, 5 parallel groups. PR (5-point scale) (n = 25); significantly different remedication, data Exclusions: 4 Numbers reporting n = 135 General and local anaes- Global rating paracetamol, to placebo for any included and baseline previously enrolled adverse effects Age 16+ years thetic. 4-hour washout (5-point scale) 650 mg measure of analgesia. used for all further in study; 3 remedi- (number of effects): prior to start. Self-assessed Time to (n = 27). time points. cated before placebo 5/25 (5); at clinic for at least first remedication 2 hours; 2 lost to paracetamol 7/27 (9). 2 hours then at home follow-up; 1 ineli- hourly for 6 hours. gible because of codeine sensitivity.

Dolci, et al., Removal of single RCT,double-blind, single PI (4-point scale) Placebo PID/PR: paracetamol Rescue analgesia 298 analysed. 4 withdrew from 4 1994 impacted third oral dose, 4 parallel groups, PR (5-point scale) (n = 76); significantly superior permitted after Exclusions: 15 lost study because of molar multicentre (11). No Global rating paracetamol, to placebo at 90 minutes, no to follow-up, 6 adverse effects. n = 336 information on anaesthesia. (5-point scale) 500 mg t = 30 minutes further evaluation remedicated before Paracetamol: Age 18+ years 24-hour washout prior to (n = 72). (p < 0.01) and at post-remedication. 1.5 hours; 3 nausea 1; swelling 1. start. Evaluations made at t = 1–4 hours Number remedicated experienced Placebo: fever 1, 0, 0.5, 1, 1.5, 2 hours in (p < 0.001). at t > 1.5 hours: adverse event and nausea and clinic then at 3 and 4 hours SPID/TOTPAR: placebo 46/76; did not complete diarrhoea 1. at home (diary). paracetamol signifi- paracetamol 15/72. assessment; 14 did Numbers reporting cantly superior at not experience > adverse effects t = 1–4 hours. moderate baseline (number of effects): Dichotomous data pain. paracetamol 7/80 (7); available for Global placebo 8/82 (12). Rating (see page 190, Table V).

Fassolt & Postoperative RCT,double-blind, single PI (5-point scale Placebo All active drugs Remedication No information None serious 2 Stocker, (‘simple surgery’ oral dose. 5 parallel groups. and VAS) (n = 28); significantly superior allowed after 2 hours given. reported – no further 1983 with 15+ surgical General anaesthetic used. PR (5-point scale) paracetamol, for all measures of (last score used for details given. techniques 4-hour washout prior to Global rating 650 mg efficacy (except all further time named) start. Evaluation at 0, (5-point scale) (n = 29). Suprofen 200 which points). Number n = 146 30 minutes then hourly for Time to showed no not remedicated at Age 18+ years 6 hours by same trained remedication significant difference 6 hours: placebo observer in hospital. in global rating). 9/28; paracetamol 25/29.

Forbes, Impacted third Double-blind, single oral PI (4-point scale) Placebo At 4 hours para- If remedication at 159 analysed. No active treatment 3 et al., 1982 molar (1 or dose, 5 parallel groups. PR (5-point scale) (n = 30); cetamol significantly < 2 hours -excluded. Exclusions: 4 lost produced more than more) No information on 50% PR? (y/n) paracetamol, superior to placebo If > 2 < 12 hours, to follow-up; 3 did placebo. None were n = 177 anaesthesia. Self-assessed Global rating 600 mg for PR, peak PR and PR = 0 and PI not take medi- serious. Age 15+ years at home at 0, 1 hours, then (5-point scale) (n = 34). 50% PR (p < 0.01). baseline or last. cation; 7 took hourly for 12 hours. Time to % patients remedi- remedication remedication cated at > 2 hours < 2 hours; 1 but < 12 hours: medicated with placebo 93%; mild pain; 3 did not paracetamol 85%. complete forms properly.

Forbes, Postoperative RCT,double-blind, single PI (4-point scale) Placebo At 6 hours para- Patients remedicated 132 analysed. None serious 5 et al., 1983 (general, oral dose, 5 parallel groups. PR (5-point scale) (n = 26); cetamol significantly on demand if There were reported. Most gynaecological General anaesthetic. Oral Pain half gone? (y/n) paracetamol, superior for all < 2 hours excluded. no exclusions. common were or orthopaedic analgesia given on request Global rating 600 mg measures; at If remedicated at drowsiness, dizziness surgery) first day able to take. Self- (5-point scale) (n = 26). 12 hours for all < 12 hours, PR = 0, and dry mouth. n = 132 assessed in hospital at 0, Time to except SPID. PI = baseline or last Numbers reporting Age 18+ years 0.5, 1, 1.5, 2 hours then remedication score for remaining adverse effects hourly for 12 hours. Single time points. (number of effects): nurse observer present for placebo 4/26 (5); first 6 hours, 2nd 6-hour paracetamol period monitored by 11/26 (11). ward staff.

Forbes, Postoperative RCT,double-blind, single PI (4-point scale) Placebo Paracetamol If remedication at 129 analysed. None serious 4 et al., 1984 (general, oral dose, 5 parallel groups. PR (5-point scale) (n = 33); significantly superior < 6 hours, PR = 0, Exclusions: 2 recorded. Sedation gynaecological General anaesthetic.Trial Pain half gone? (y/n) paracetamol, for all measures of PI = baseline or last remedicated accounted for 65–70%. or orthopaedic drug given on request first Acceptability 650 mg efficacy at t = for all remaining < 2 hours; 1 Numbers reporting surgery) day after surgery. Assessed (5-point scale, (n = 31). 1–4 hours. But only time points. received interfering adverse effects n = 132 in hospital by single nurse each hour) marginally significant medication. (number of effects): Age 18+ years observer at 0, 15 and Time to for peak PI difference placebo 8/33 (9); 30 minutes, then hourly remedication and peak PR. paracetamol 9/33 (10). for 6 hours.

continued 65 Paracetamol with and without codeine in acute pain

TABLE 23 contd Trials of paracetamol versus placebo

Forbes Impacted third RCT,double-blind, single PI (4-point scale) Placebo Paracetamol Patients could 148 analysed. NSD between 5 et al., 1984 molar (1 or oral dose, 4 parallel groups. PR (5-point scale) (n = 36); significantly superior remedicate after Exclusions: 1 did treatments, none more) General/local anaesthetic Pain half gone? (y/n) paracetamol, for all measures of 2 hours but were not return results; serious. Most n = 191 (unclear). Self-assessed at Global rating 650 mg total and peak asked to complete 1 lost to follow-up; frequent was Age 15+ years home at 0, 1 hours, then (5-point scale) (n = 39). analgesia. next evaluation 26 did not require drowsiness. hourly for 6 hours. Time to before doing so (PI medication, 8 did Numbers reporting Returned 5 days later for remedication last or baseline score; not follow instruc- adverse effects review and debriefing. PR = 0 for all further tions; 7 remedi- (number of effects): time points). cated < 2 hours. placebo 2/40 (2); % remedicating by paracetamol 1/43 (2). hour 6: paracetamol 74%; placebo 97%.

Forbes, Impacted third RCT,double-blind, single PI (4-point scale) Placebo For 12 hours: Patients could 88 analysed. None serious et al., 1989 molar (1 or oral dose, 4 parallel groups PR (5-point scale) (n = 23); paracetamol not remedicate after Exclusions: 9 did reported. more) at 2 centres. General and Global rating paracetamol, significantly superior 2 hours but were not take medi- Numbers reporting n = 107 local anaesthetic (unclear). (5-point scale) 600 mg for any measure. asked to complete cation; 2 remedi- adverse effects Age 15+ years Patients self-assessed at Time to (n = 22). For 4 hours: para- next evaluation cated before 2 hour (number of effects): home at 0, 1, 2 hours then remedication cetamol significantly before doing so (PI point; 1 remedi- placebo 2/26 (2); hourly for 12 hours or superior for SPID, last or baseline score; cated with slight paracetamol 3/26 (3). until remedication (diary). TOTPAR and PR = 0 for all further pain; 4 did not com- NB: includes those Returned 5 days later for hours of 50% time points). plete evaluation; reported post- review and debriefing. relief. % remedicating by 1 took only part of remedication. hour 12: paracetamol the medication; 2 95%; placebo 91%. remedicated despite having some relief from the study medication.

Forbes, Impacted third RCT,double-blind, single PI (4-point scale) Placebo All active medi- Remedication 206 analysed. None serious 5 et al., 1990 molar (1 or then multiple oral dose, PR (5-point scale) (n = 34); cations significantly allowed after 2 hours Exclusions; 3 lost reported. more) 6 parallel groups. General/ Pain half gone? (y/n) paracetamol, superior for all but asked to to follow-up; 1 lost Numbers reporting n = 269 local anaesthetic (unclear). Global rating 600 mg measures of total complete next report card; 22 did adverse effects Age 15+ years Self-assessed at home at 0, (5-point scale) (n = 36). and peak analgesia. evaluation (PI last or not require (number of effects): 1 hours, then hourly for Time to baseline score; medication; 8 placebo 0/38; 6 hours. Returned 5 days remedication PR = 0 used for all remedicated despite paracetamol 5/41 (5). later for review and remaining time having relief from debriefing. points). study medication; % remedicated by 6 remedicated with hour 6: placebo 33%; only slight pain; paracetamol 29%. 13 remedicated < 2 hours; 7 failed to follow instructions; 3 did not complete forms.

Honig & Postoperative RCT,single oral dose, 4 PI (4-point scale) Placebo TOTPAR and global If remedicated, last No details given. None severe except 3 Murray, (elective surgery, parallel groups. No PR (5-point scale) (n = 30); rating: paracetamol score used for all 1 severe dry mouth. 1984 viz. abdominal, information on anaesthesia, 50% PR? (y/n) paracetamol, significantly superior further time points. Reported in all orthopaedic, 4-hour washout prior to Global rating 600 mg to placebo Number of patients groups; primarily rectal, thoracic, start. Interviewed in (5-point scale) (n = 28). (p < 0.05). not remedicated in central nervous vascular) hospital by nurse observer 6 hours: placebo system and gastro- n = 116 at 0, 0.5, 1 hours then 14/30; paracetamol intestinal effects. Age range hourly for 6 hours. 16/28. 19–87 years

Jain, Postoperative RCT,double-blind, single PI (4-point scale) Placebo Paracetamol only Remedicated at 122 analysed. None serious 4 et al., 1986 (general, oral dose, 4 parallel groups. PR (5-point scale) (n = 32); significantly superior < 2 hours, data Exclusions: 2 had reported. NSD gynaecological General anaesthetic.Trial paracetamol, for maximum excluded. If > 2 hours, improperly blinded between groups. or orthopaedic drug given within 72 hours 650 mg relief (p < 0.05). last measurement drugs; 2 remedi- Numbers reporting surgery) of surgery on request for (n = 30). used for all remaining cated < 2 hours adverse effects n = 128 analgesia. 4-hour washout time points. and 2 received (number of effects): Age range prior to start.Assessed in Total numbers interfering placebo 6/32 (8) 18–70 years hospital by nurse observer remedicated: placebo medication. paracetamol 9/30 (9). at 0, 15 and 30 minutes, 11/32; paracetamol then hourly for 6 hours. 10/30; Nalbu 9/34; Combo 5/32.

66 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 23 contd Trials of paracetamol versus placebo

Kiersch, Impacted third RCT,double-blind, single PI (4-point scale) Placebo Paracetamol “Patients were asked 226 analysed. None serious 4 et al., 1994 molar (3 or 4) oral dose, 3 parallel groups. PR (5-point scale) (n = 30); significantly superior to allow 2 hours... Exclusions: 1 reported. n = 232 Local anaesthetic, 48-hour Pain half gone? (y/n) paracetamol, to placebo for most before taking experienced nausea Numbers reporting Age 14+ years washout prior to start. Global rating 1000 mg efficacy measures in alternate medication.” and vomiting so adverse effects Self-assessed in clinic for (5-point scale) (n = 30). first 6 hours. Time to remedi- did not ingest (number of effects): first 2 hours, then at home Time to cation (median): treatment; 2 did not placebo 13/45 (18); at 0, 20, 30, and 40 minutes, remedication placebo 2.0 hours; require analgesia; paracetamol 31/92 1 hour then hourly for VAS PI (100 mm) paracetamol 1 failed to follow (35). 12 hours. 3.1 hours. instructions; 2 vomited within 10 minutes of taking trial drug.

McQuay, Postoperative RCT,double-blind, single PI (4-point scale, Placebo Paracetamol If remedicated after 150 analysed. None serious 4 et al., 1988 (elective oral dose, 5 parallel groups. VAS and 8-word (n = 30); significantly superior 1 hour, PI scored at Exclusions: 2 reported; NSD orthopaedic General anaesthetic.Trial verbal rating) paracetamol, to placebo for all baseline and PR = 0. discharged before between groups. surgery) drug given 1/2 days after PR (5-point scale 1000 mg integrated measures For patients. end; 3 received Numbers reporting n = 158 surgery. 3-hour washout and VAS) (n = 30). NB: of efficacy. remedicated at drugs prohibited by adverse effects Age range prior to start. Assessed in Pain half gone? (y/n) these are < 6 hours: all active protocol; 1 vomited (number of effects): 18–70 years hospital by nurse observer Global rating numbers after treatments intact medication placebo 6/30 (8); at 0, 0.5, 1 and 1.5 hours (5-point scale, both exclusions. significantly superior within 15 minutes; paracetamol 6/30 (10). then hourly for 6 hours. patient and observer) (p < 0.01) to placebo. 2 pain assessments Time to inadequately remedication completed. Vital signs

Mehlisch, Third molar RCT,double-blind, single PI (4-point scale) Placebo Paracetamol If remedicated before 399 analysed. None serious 3 et al., 1995 (at least 1 oral dose, 3 parallel groups. PR (5-point scale) (n = 40); significantly superior 1 hour, data excluded Exclusions:1 failed reported. embedded) Local anaesthetic, 12-hour Global rating paracetamol, to placebo for all from analysis.Value to complete diary. Numbers reporting n = 240 washout prior to start. (5-point scale) 1000 mg measures of efficacy. of 0 was assigned for adverse effects Age 15+ years Self-assessed at 0, 15 and (n = 101). PI and PR at all time (number of effects): 45 minutes, 1 hour and points after remedi- placebo 4/40 (?); 90 minutes, then hourly cation.% patients paracetamol for 6 hours. remedicating: placebo 17/101 (?). 88%; paracetamol 52%.

Mehlisch & Oral surgery RCT,double-blind, single PI (4-point scale) Placebo Paracetamol Remedicated if 162 analysed. NSD in numbers 4 Frakes, (involving bone oral dose, 3 parallel groups. PR (5-point scale) (n = 55); significantly superior required after 1 hour Exclusions: 9 failed for paracetamol 1984 removal) No information on Pain half gone? (y/n) paracetamol, (p < 0.05) to and patient to comply with and placebo – n = 174 anaesthesia except no Global rating 1000 mg placebo for all considered treatment protocol; 3 lost to no other Age 16+ years long-acting i.m. or i.v. (5-point scale) (n = 58). measures, and to failure (no details on follow-up. details given. anaesthesia used. 4-hour Time to aspirin for maximum how data was washout period prior to remedication PID (p < 0.05), handled). Full dichoto- start. Assessed in clinic by maximum PR mous data on times of nurse observer at (p < 0.03) medication in Table IV. 30 minutes then hourly and global rating Number of patients for 6 hours. (p < 0.02). not remedicated before t > 6 hours: paracetamol 13/58; placebo 3/55.

Schachtel, Episiotomy (post RCT,double-blind, single PI (4-point scale) Placebo Paracetamol If remedicated after 111 analysed. None reported. 4 et al., 1989 uncomplicated oral dose, 3 parallel groups. PR (5-point scale) (n = 38); significantly superior 1 hour considered Exclusions: 4 Placebo 0/37; delivery) 4-hour washout prior to Global rating paracetamol, (p < 0.05) to placebo treatment failure; remedicated but paracetamol 0/37. n = 115 start. Assessed (where and (5-point scale) 1000 mg for TOTPAR, global last/baseline PI and did not record Age range by whom not clear) at 0, (n = 37). rating and number PR = 0 scored for at what time. 16–37 years 0.5, 1 hour, then hourly of remedications. remaining time points. for 4 hours. Remedicated < 6 hours: paracetamol 13/37; placebo 22/38.

Sunshine, Impacted RCT,double-blind, single PI (4-point scale) Placebo Paracetamol If remedicated at 182 analysed. None serious 5 et al., 1986 third molar oral dose, 6 parallel groups. PR (5-point scale) (n = 30); significantly superior < 1 hour, excluded. No exclusions. reported; NSD n = 182 Local anaesthetic Global rating paracetamol, (p < 0.05) to If at > 1 hour, last PI between groups. Age 16+ years (lidocaine/epinephrine). (4-point scale) 650 mg placebo for: PID or baseline and Numbers reporting 4-hour washout prior to Overall (n = 30). at 1–3 hours, PR = 0 used. Full adverse effects: start. Evaluations at 0, 0.5, improvement SPID at 4 hours, dichotomous data placebo 1/30; 1, 2, 3 hours in clinic by (7-point scale) PR at 2 hours and in Table III. paracetamol 1/30. single observer. At 4, 5, Time to time to peak effect Number of patients 6 hours self-assessed. remedication (see Table II). remedicated: placebo 1 week later met with 13/30; paracetamol observer to review forms. 14/30. continued 67 Paracetamol with and without codeine in acute pain

TABLE 23 contd Trials of paracetamol versus placebo

Sunshine, Episiotomy RCT,double-blind, single PI (4-point scale) Placebo Paracetamol If remedicated at 200 analysed. Only 2 patients 4 et al., 1989 (multiparous oral dose, 3 parallel groups. PR (5-point scale) (n = 50); significantly superior < 2 hours excluded. No exclusions. reported adverse in-patients) Only included patients with Global rating paracetamol, (p < 0.05) to If remedicated at efffect; they were n = 200 severe pain. 4-hour (4-point scale) 650 mg placebo for all > 2 hours, PI last or in neither Age 18+ years washout prior to start. Overall improvement (n = 75). measures of efficacy. baseline and PR = 0 paracetamol nor Evaluations in hospital by (7-point scale) used for remaining placebo groups. nurse-observer at 0, 0.5, Time to time points. Those reported 1 hours, then hourly for remedication No remedications were mild dizziness, 6 hours (if asleep, woken). at < 2 hours. sleepiness and Interviews in patients’ first Remedications at sweating. language (Spanish). > 2 hours: placebo 8/50; paracetamol 2/75.

Sunshine, Postoperative RCT,double-blind, single PI (4-point scale) Placebo Paracetamol not If remedicated at All 240 enrolled No details for single 4 et al., 1993 (Caesarean oral dose then multi-dose, PR (5-point scale) (n = 48); significantly superior < 1 hour after first were analysed. dose phase. section) 5 parallel groups. Only Global rating paracetamol, (p < 0.05) to dose, dropped and n = 240 included patients with (5-point scale) 650 mg placebo for replaced. If Age 18+ years severe pain, 4-hour Time to (n = 48). any measure. remedicated > 1 hour washout prior to start. meaningful relief after first dose, Evaluation by same nurse- eligible for repeat observer at 0, 0.5, 1 hours, dose phase. then hourly for 8 hours. Number of patients Interviews in patients’ first remedicated language (Spanish). < 8 hours: placebo 35/48; paracetamol 42/48.

Winter, Oral surgery RCT,double-blind, single PI (4-point scale) Placebo Paracetamol Remedication 164 analysed. None serious 4 et al., 1983 (various oral dose, 4 parallel groups. PR (5-point scale) (n = 41); significantly superior allowed after 2 hours Exclusions: 3 reported. procedures) 4-hour washout period 50% PR? (y/n) paracetamol, to placebo for all or if pain returned to protocol violations; Numbers reporting n = 168 prior to start. General Global rating 1000 mg measures of pre-medication levels. 1 did not receive adverse effects: Age range and/or local anaesthetic. (5-point scale) (n = 41). analgesic efficacy. 2 patients study medication. placebo 1/41 16–75 years Self-assessed at 0, 0.5, 1, 2, Both produced remedicated (severe headache); 3 and 4 hours. significant analgesia ≥ 2 hours (1 placebo paracetamol 0/41. as early as and 1 paracetamol). t = 0.5 hours.

Young, Postoperative RCT,double-blind, single PI (4-point scale) Study 1: Paracetamol only “Any concomitant 119 analysed. None serious 4 et al., 1979 (various elective oral dose, 4 parallel groups. PR (5-point scale) Placebo significantly superior or additional Exclusion: 1 reported. procedures) General anaesthetic 2 Global ratings (n = 29); to placebo medication given was received analgesia Numbers reporting n = 120 (halothane/nitrous (5-point scale, both paracetamol, (p < 0.05) at duly noted.” No within 2 hours adverse effects: Age range oxide/oxygen). 4-hour patient and observer 650 mg t = 2 hours details of this data of study. placebo 1/30 12–83 years washout prior to start. opinions) (n = 30). PR score. given or how it was (sedation); Evaluated in hospital by handled. paracetamol single observer at 0, 0.5, 1, 3/30 (nausea). 2, 3 and 4 hours.

Drug-related study withdrawals rarely occurred. paracetamol, 650 mg, plus codeine, 60 mg, in One study15 had three withdrawals, one patient on five trials, and paracetamol, 1000 mg, plus codeine, placebo and two on paracetamol. Another study16 60 mg, in two trials. One report on episiotomy had two patients who withdrew on paracetamol. provided dichotomous information on the These studies reported a variable incidence of overall patient global rating of pain relief.17 adverse events which were mild and transient, This report was included. The proportion of with no difference in incidence between patients with good or excellent pain relief was paracetamol and placebo. used. Pain relief was measured over 4–6 hours in 19 of the reports; one had observations for Paracetamol plus codeine just 3 hours. versus placebo A total of 20 reports of 20 trials were found The variation in placebo response rate was which fulfilled the inclusion criteria; 721 patients from 0% to 72% of patients with at least 50% were given paracetamol and 664 placebo. Details maxTOTPAR. The variation in response rate of the trials are given in Table 25. The doses for all doses of paracetamol plus codeine was administered were paracetamol, 300 mg, plus 20% to 83% (see Figure 25, page 73). The mean codeine, 30 mg, in five trials; paracetamol, response rate for paracetamol, 600/650 mg, plus 68 600 mg, plus codeine, 60 mg, in eight trials; codeine, 60 mg, was 51% and for placebo, 21%. Health Technology Assessment 1998; Vol. 2: No. 12

Paracetamol plus codeine versus paracetamol alone At least 50% PR with paracetamol A total of 12 reports of 12 trials were found which 100 fulfilled the inclusion criteria; 395 patients were given paracetamol and 399 placebo. Four reports were identified which fulfilled the inclusion criteria 75 except that they had pain outcomes other than five- point categorical pain relief scores. Details of the trials are given in Table 27 (see page 75). Ten trials 50 were in oral surgery and three in postsurgical pain. Doses were paracetamol, 600 mg, plus codeine, 60 mg, in seven trials, paracetamol, 650 mg, plus codeine, 60 mg, in four trials, and paracetamol, 25 1000 mg, plus codeine, 60 mg, in two trials. Pain relief was measured over 4–6 hours in 12 reports; one measured pain relief over 3 hours. 0 0 25 50 75 100 The variation in response rates for paracetamol alone was from 5% to 89% of patients with at At least 50% PR with placebo least 50% maxTOTPAR. The variation in response rates to all doses of paracetamol plus codeine was 24–83%. FIGURE 24 Single dose studies of paracetamol against placebo (◆, 500 mg; ●, 600/650 mg; , 1000 mg; +, 1500 mg) Only one of the reports had a lower 95% CI of the relative benefit that did not include 1 (Figure 26, see page 74). The combined relative benefit (fixed Doses of paracetamol were given first. Studies effects model) for this homogeneous data set was were predominantly in oral surgery (14 trials for 1.25 (95% CI, 1.09– 1.43). Combining data across paracetamol plus codeine against placebo and conditions, the NNT for addition of codeine, ten trials for paracetamol plus codeine against 60 mg, to all doses of paracetamol in single dose paracetamol alone). administration for at least 50% pain relief was 7.7 (95% CI, 5.1–17). Combining data across conditions, paracetamol, 300 mg, plus codeine, 30 mg, compared with There were no serious adverse events which placebo in single dose administration had an necessitated patient withdrawal from any study. NNT for at least 50% pain relief of 5.3 (95% CI, 3.8–8.0), paracetamol, 600/650 mg, plus codeine, 60 mg, an NNT of 3.1 (95% CI, 2.6–3.8) and Comment paracetamol, 1000 mg, plus codeine, 60 mg, an NNT of 1.9 (95% CI, 1.5–2.6), although Paracetamol, 1000 mg, alone had an overall NNT in only two trials (see Table 26, page 74). of 4.6 for at least 50% pain relief compared with placebo in single dose administration. This means There were no serious adverse events which that one in every five patients with pain of moder- necessitated patient withdrawal from ate to severe intensity will get at least 50% pain any study. relief; they would not have done had they been

TABLE 24 Summary risk ratios and NNTs for trials of paracetamol against placebo

Number of trials Paracetamol > 50% max TOTPAR > 50% max TOTPAR RB NNT dose (mg) on paracetamol on placebo (95% CI) (95% CI)

6 500 194/353 109/296 1.4 (1.1–1.9) 5.6 (3.9–9.5)

17 600/650 243/594 125/573 1.7 (1.3–2.2) 5.3 (4.1–7.2)

20 1000 620/1376 207/907 2.3 (1.7–2.9) 4.6 (3.9–5.4)

3 1500 133/207 63/141 1.4 (1.2–1.9) 5.0 (3.3–11) 69 Paracetamol with and without codeine in acute pain

TABLE 25 Studies of paracetamol plus codeine versus placebo

Bentley & Third molar RCT,double-blind, single PI (10-point scale) Placebo Combination Rescue analgesic: 120 analysed. 53 had 1 or more; 3 Head, 1987 bony impacted oral dose, 4 parallel groups. PR (5-point scale) (n = 17); significantly superior Tylenol no. 3. Exclusions: 3 did 86 reported in total, n = 128 No information on paracetamol, to placebo for all Patients who not take medication; majority being Mean age: anaesthesia except ‘no 1000 mg, + measures of efficacy, remedicated 1 took only a dizziness, drowsiness, mid-20s sedative or narcotic agents codeine, but not significantly < 5 hours: last PI and portion; 1 took no nausea and vomiting. were used before, during 60 mg, (n = 41) different from PR scores carried on medication until No significant or after surgery’. Self- paracetamol for for all further day after surgery; difference between assessed at home at 0, 1, 2, any measure. time points. 1 remedicated after treatment groups. 3, 4, 5 hours; reports 30 minutes; 1 Numbers reporting posted to investigator. vomited within adverse effects 30 minutes of taking (number of effects): medication; 1 placebo 9/19 (16); patient did not paracetamol + return forms. codeine 15/42 (24).

Cooper & Impacted RCT,double-blind, 2 single PI (4-point scale) Study 2: Parametric analysis Remedication 160 analysed. None serious 5 Beaver, third molar oral dose studies (second PR (5-point scale) Placebo concluded para- consisted of taking Exclusions: 30 reported. Most 1976 n = 216 is paracetamol), 4 parallel Pain half gone? (y/n) (n = 40); cetamol significantly second envelope and patients did not common were Age: 16+ years groups. No information on Had the patient paracetamol, superior to placebo evaluating it in same return forms, 12 drowsiness, nausea anaesthesia. Self-assessed at fallen asleep during 600 mg, + (p < 0.01) and non- way as first; unclear filled in the forms and headache. home at 0, 1, 2, 3 hours the hour? (y/n) codeine, parametric factorial how data was then incorrectly, 4 took Numbers reporting (questionnaire posted to 60 mg (n = 40) analysis showed handled. 112/160 concomitant medi- adverse effects investigator). codeine had effect patients required cation, 6 required (number of effects): for hour 1 (p < 0.01) remedication. no medication and placebo 5/40 (5); and peak PI 4 were randomly paracetamol + (p < 0.05). NSD deleted to even out codeine 11/40 (11). between combi- number of patients nation and its on each treatment. constituents

Cooper, Impacted RCT,double-blind, single PI (4-point scale) Placebo All active treatments Remedication after 200 analysed. None serious 4 et al.., 1981 third molar oral dose, 5 parallel groups. PR (5-point scale) (n = 37); significantly superior first hour if needed Exclusions: 17 did reported. Numbers n = 248 General or local anaesthetic. Global rating paracetamol, to placebo for all (last score used for not ingest medi- reporting adverse Mean age: Self-assessed at home at (5-point scale) 650 mg, + measures. Combi- all further time cation, 31 ingested effects (number of early 20s 0, 1, 2, 3, 4 hours 50% relief of codeine, nation slightly more points). Remedication medication but effects): placebo (questionnaire). baseline pain? (y/n) 60 mg, effective than para- < 4 hours: placebo violated protocol 4/37 (5); paracetamol Time to (n = 42) cetamol alone but 20/37; paracetamol (remedication before + codeine 10/42 (10). remedication difference not 2/37. hour 1 observations; significant. constant deviation of more than 15 minutes from evaluation times, not returning questionnaire and lost to follow-up).

Cooper, Impacted RCT,double-blind, single oral PI (4-point scale) Placebo Combination Remedication after 143 analysed. None serious 4 et al.., 1988 third molar dose, 3 parallel groups, single PR (5-point scale) (n = 40); significantly superior first hour (last or Exclusions: 11 lost reported. Number n = 165 centre and 1 surgeon. Local Pain half gone? (y/n) paracetamol, to placebo for every baseline score used to follow-up, 8 did of effects reported: Age range: anaesthetic with sedation Global rating 600 mg, + measure and to for all further time not require placebo 3; 18—57 years and/or nitrous oxide. 4-hour (5-point scale) codeine, paracetamol for points). medication and 3 for paracetamol + washout prior to start. Self- Time to 60 mg (n = 31) TOTPAR (p < 0.05). ‘various protocol codeine 4. assessed at home at 0, 0.5, 1, remedication violations’. 2, 3, 4, 5, 6 hours (diary).

Cooper & Removal of RCT,double-blind, single oral PI (4-point scale) Placebo Combination t > 1 hour before 226 analysed. All mild. Numbers 3 Kupperman, one or more dose, 6 parallel groups. Local PR (5-point scale) (n = 44); significantly superior remedication; data Exclusions; 13 did reporting adverse 1991 impacted teeth anaesthetic (lidocaine + Global rating paracetamol, to placebo for most included and baseline not require medi- effects (number n = 247 epinephrine) with i.v. (5-point scale) 650 mg, + measures and to or last score (most cation, 3 lost to of effects): placebo Age:‘young diazepam and methohexital Time to codeine, paracetamol for severe) used for all follow-up, 2 7/44 (9); paracetamol adults’ – on occasion nitrous oxide remedication 60 mg (n = 39) TOTPAR and global further time points. remedication. with + codeine 8/39 (11). was also used. Self-assessed rating. slight pain before at home at 0, 0.5 hours then hour 2 observations, hourly for 6 hours (diaries). 2 remedications. Before first hour observations, 1 fell asleep for over 2 hours.

70 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 25 contd Studies of paracetamol plus codeine versus placebo

Desjardins, Oral surgery RCT,double-blind, single oral PI (4-point scale) Placebo Combination only If t ≥ 1 hour before 123 analysed. None serious 4 et al., 1986 n = 137 dose, 3 parallel groups, local PR (5-point scale) (n = 41); significantly superior remedication, data Exclusions:14 did reported. Numbers Age: 18+ years anaesthetic. 4-hour washout 50% PR (y/n) paracetamol, to placebo for global included and last not medicate, lost reporting adverse prior to start. Self-assessed at Global rating 300 mg, + rating and total score for PI and PR to follow-up, or effects (number home at 0, 0.5 hours, then (5-point scale) codeine, anxiety (p < 0.05). used for all further provided uninter- of effects): placebo hourly for 6 hours. Anxiety (4-point 30 mg (n = 39) time points. pretable results. 4/41 (4); scale) paracetamol + Relaxation (4-point codeine 2/39 (3). scale)

Dionne, Impacted RCT,double-blind, single oral PI (4-point scale) Placebo Neither paracetamol t ≥ 2 hours before 124 analysed. All mild. Numbers 3 et al., 1994 third molar dose, 5 parallel groups. PR (5-point scale) (n = 25); nor combination remedication; data Exclusions;4 reporting adverse n = 135 General and local anaesthetic. Global rating paracetamol, significantly different included and baseline previously enrolled effects (number Age: 16+ years 4-hour washout prior to start. (5-point scale) 650 mg, + from placebo for any used for all further in the study, 3 of effects): placebo Self-assessed at clinic for at Time to remedication codeine, measure of analgesia. time points. remedicated before 5/25 (5); paracetamol least the first 2 hours, then 60 mg (n = 24) t = 2 hours, 2 lost to + codeine 9/24 (10). at home hourly for 6 hours. follow up, 1 ineligible because of codeine sensitivity.

Forbes, Impacted Double-blind, single oral dose, PI (4-point scale) Placebo Combination signifi- Excluded if remedi- 159 analysed. No active treatment 3 et al., 1982 third molar 5 parallel groups. No PR (5-point scale) (n = 30); cantly superior to cated at < 2 hours. Exclusions: 4 lost to produced more than (1 or more) information on anaesthesia. 50% PR (y/n) paracetamol, placebo for all If > 2 hours but follow-up, 3 did not placebo. None were n = 177 Self-assessed at home at 0, Global rating 600 mg, + measures. < 12 hours, PR = 0 take medication, 7 serious. Age: 15+ years 1 hours, then hourly for (5-point scale) codeine, and PI baseline or took remedication 12 hours. Time to remedication 60 mg (n = 31) last. Patients remedi- < 2 hours, 1 took catedat > 2 hours but medication with mild < 12 hours: placebo pain and 3 did not 93%; paracetamol complete the forms 85%. properly.

Forbes, Postoperative RCT,double-blind, single PI (4-point scale) Placebo For 6-hour data, Patients remedicated 132 analysed.There None serious 5 et al., 1983 (general, oral dose, 5 parallel groups. PR (5-point scale) (n = 26); combination signifi- on demand at were no exclusions. reported. Most gynaecological General anaesthetic, given on Pain half gone? (y/n) paracetamol, cantly superior to < 2 hours excluded. common were or orthopaedic request first day able to take Global rating 600 mg, + placebo for all If remedication drowsiness, dizziness surgery) oral analgesia. Self-assessed (5-point scale) placebo, measures; for < 12 hours, PR = 0, and dry mouth. n = 132 in hospital at 0, 0.5, 1, 1.5, Time to remedication 60 mg (n = 26) 12-hour data, all PI = baseline or last Numbers reporting Age: 18+ years 2 hours, then hourly for except SPID. score for remaining adverse effects 12 hours. Single nurse time points. (number of effects): observer present for first placebo 4/26 (5); 6 hours, next 6 hours moni- paracetamol + tored by ward staff. codeine 11/26 (16).

Forbes, Impacted RCT,double-blind, single oral PI (4-point scale) Placebo Combination signifi- If remedicated at 122 analysed. None serious 5 et al., 1986 third molar dose, 3 parallel groups, local PR (5-point scale) (n = 38); cantly superior to < 2 hours, excluded. Exclusions:1 did not reported. Numbers (1 or more) anaesthetic. Self-assessed at 50% PR (y/n) paracetamol, placebo for all If > 2 hours but return the form, reporting adverse n = 146 home at 0, 1 hours, then Global rating 300 mg, + measures of efficacy. < 6 hours, PR = 0 6 did not need effects (number Age: 15+ years hourly for 6 hours (diary); (5-point scale) placebo, and PI baseline or analgesia and 17 of effects): placebo returned 5 days later for Anxiety (4-point 30 mg (n = 43) last (which ever had invalid data. 9/46 (11); review and debriefing. scale) was greater). paracetamol + Relaxation (4-point codeine 6/46 (8). scale)

Forbes, Impacted RCT,double-blind, single oral PI (4-point scale) Placebo At 4 hours: combi- Patients could 88 analysed. None serious 5 et al., 1989 third molar dose, 4 parallel groups at PR (5-point scale) (n = 23); nation significantly remedicate after Exclusions: 9 did not reported. Numbers (1 or more) 2 centres. General and local Global rating paracetamol, superior to placebo 2 hours but were take medication, 2 reporting adverse n = 107 anaesthetic (unclear). Self- (5-point scale) 600 mg, + paracetamol for asked to complete remedicated before effects (number Age: 15+ years assessed at home at 0, 1, Time to remedication codeine, TOTPAR (p < 0.05). next evaluation the 2-hour point, 1 of effects): placebo 2 hours, then hourly for 60 mg (n = 17) before doing so (PI remedicated with 2/26 (2); paracetamol 12 hours or until remedi- last or baseline score, slight pain, 4 did not + codeine 1/17 (1). cation (diary); returned 5 days PR = 0 for all further complete their NB: includes those later for review and debriefing. time points). Patients evaluation, 1 took reported post- remedicating by hour only part of the remedication. 12: paracetamol 95%; medication and 2 placebo 91%. remedicated despite getting some relief from study medication.

continued 71 Paracetamol with and without codeine in acute pain

TABLE 25 contd Studies of paracetamol plus codeine versus placebo

Forbes, Impacted RCT,double-blind, single oral PI (4-point scale) Placebo Combination signifi- If remedication 128 analysed. None serious 5 et al., third molar dose, 4 parallel groups, local PR (5-point scale) (n = 32); cantly superior to < 2 hours, excluded. Exclusions:1 failed to reported. Numbers 1990a (1 or more) anaesthetic. Self-assessed at Pain half gone? (y/n) paracetamol, placebo for all If > 2 hours but return form, 19 did reporting adverse n = 162 home at 0, 1 hours, then Global rating 600 mg, + measures of efficacy. < 6 hours, PR = 0 not require analgesia effects (number of Age: 15+ years hourly for 6 hours (diary); (5-point scale) codeine, and PI baseline or and 14 had invalid effects): placebo returned 5 days later for Time to remedication 60 mg (n = 27) last which ever data. 5/34 (6); paracetamol review and debriefing. was greater. + codeine 9/31 (12). Number remedicating before t = 6 hours: placebo 27; combination 23.

Forbes, Impacted RCT,double-blind, single then PI (4-point scale) Placebo All active medi- Remedication allowed 206 analysed. None serious 5 et al., 1990b third molar multiple oral dose, 6 parallel PR (5-point scale) (n = 34); cations significantly after 2 hours but Exclusions; 3 lost reported. Numbers (1 or more) groups. General/local anaes- Pain half gone? (y/n) paracetamol, superior for all asked to complete to follow-up, 1 lost reporting adverse n = 269 thetic (unclear). Self-assessed Global rating 600 mg, + measures of total next evaluation (PI report card, 22 did effects (number of Age: 15+ years at home at 0, 1 hours, then (5-point scale) codeine, and peak analgesia. last or baseline score, not need medication. effects): placebo hourly for 6 hours; returned Time to remedication 60 mg (n = 38) PR = 0, used for all 8 remedicated 0/38; paracetamol 5 days later for review and remaining time despite having relief + codeine 8/40 (9). debriefing. points). Patients from study mediation, remedicating. by 6 remedicated with hour 6: placebo 33%; only slight pain, 13 paracetamol 29%. remedicated at < 2 hours, 7 failed to follow instructions, 3 did not complete forms.

Forbes, Impacted RCT,double-blind, single oral PI (4-point scale) Placebo Combination signifi- If remedication at 232 analysed. None serious 5 et al., 1994 third molar dose, 3 parallel groups, local PR (5-point scale) (n = 45); cantly superior to < 2 hours, excluded. Exclusions:1 lost to reported. Numbers (1 or more) anaesthetic. Patients self- Pain half gone? (y/n) paracetamol, placebo for all If > 2 hours but follow-up, 32 did not reporting adverse n = 324 assessed at home at 0, 0.5, Global rating 300 mg, + measures of mean < 6 hours, PR = 0 require analgesia and effects (number of Age: 15+ years 1 hours, then hourly for (5-point scale) placebo, and peak analgesia. and PI baseline or 51 had invalid data. effects): placebo 6 hours (diary); returned Time to remedication 30 mg (n = 93) last, whichever 10/65 (12); 5 days later for review and was greater. paracetamol + debriefing. codeine 18/107 (21).

Heidrich, Orthopaedic RCT,double-blind, single oral PI (4-point scale) Placebo Orthogonal analyses No information No information on No differences 2 et al.., 1985 surgery dose, 3 parallel groups. No PR (5-point scale) (n = 40); of variance showed on remedication. withdrawals or between treatments n = 120 information on anaesthesia, Global rating paracetamol, combination gave exclusions. in terms of side- Age range: 3-hour washout prior to start. (5-point scale) 300 mg, + greater relief from effects. No patient 18–65 years Interviewed in hospital by VAS PI and PR placebo, pain than placebo. withdrew because nurse observer at 0, 0.5, McGill questionnaire 30 mg, (n = 40) of adverse events. 1 hours, then hourly for Mood questionnaire 6 hours. and VAS

Honig & Postoperative RCT,double-blind, single oral PI (4-point scale) Placebo Combination signifi- If remedicated, last No details given. None severe except 3 Murray, (elective dose, 4 parallel groups. No PR (5-point scale) (n = 30); cantly superior to score used for all 1 (severe dry 1984 surgery: information on anaesthesia, 50% PR (y/n) paracetamol, placebo for most further time points. mouth). Primarily abdominal, 4-hour washout prior to start. Global rating 600 mg, + measures of efficacy. Number who did not central nervous orthopaedic, Interviewed in hospital by (5-point scale) codeine, remedicate in 6 hours: system and rectal, thoracic, nurse observer at 0, 0.5, 60 mg (n = 30) placebo 14/30; gastrointestinal vascular) 1 hours, then hourly for paracetamol 16/28. effects in all groups. n = 116 6 hours. Age range: 19–87 years

Petti, 1985 Orthopaedic RCT,double-blind, single oral PI (4-point scale) Placebo Does not say Remedication 129 analysed. None serious 2 or general dose, 4 parallel groups. No PR (5-point scale) (n = 32); anything directly allowed after 2 hours. Exclusions: 12 reported. Only 1 surgery information on anaesthesia, Global rating paracetamol, about combination. If remedicated, scores excluded for reported (dry n = 141 4-hour washout prior to start. (5-point scale) 300 mg, + of PR = 0 and PI = 2 protocol violations. mouth) in Age range: Interviewed in hospital by Severity of adverse placebo, 30 mg allocated and patient paracetamol + 18–80 years observer at 0, 0.5, 1 hours, effects (5-point scale) (n = 31) excluded from further codeine group. then hourly for 6 hours. NB: evaluations. all patients had a baseline PI of 2 (moderate).

continued 72 Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 25 contd Studies of paracetamol plus codeine versus placebo

Sunshine, Impacted RCT,double-blind, single oral PI (4-point scale) Placebo Combination signifi- If at < 1 hour 182 analysed. None serious 5 et al., 1986 third molar dose, 6 parallel groups. Local PR (5-point scale) (n = 30); cantly superior to excluded. If > 1 hour, No exclusions. reported; NSD n = 182 anaesthetic (lidocaine/ Global rating paracetamol, placebo for most last PI or baseline between groups. Age: 16+ years epinephrine); 4-hour washout (4-point scale) 650 mg, + measures of efficacy and PR = 0 used Numbers reporting prior to start. Evaluations at Overall codeine, (p < 0.05). (Full dichotomous adverse effects: 0, 0.5, 1, 2, and 3 hours in improvement 60 mg (n = 31) data:Table III. placebo 1/30; clinic by single observer. Self- (7-point scale) Number remedicated: paracetamol + assessed at 4, 5 and 6 hours. Time to remedication placebo 13/30; codeine 3/31. Met with observer 1 week paracetamol 14/30. later and reviewed forms.

Turek & Elective RCT,double-blind, single oral PI (4-point scale) Placebo Combination signifi- If remedicated at 160 analysed. None serious 3 Baird, 1988 surgery dose, 4 parallel groups. No PR (5-point scale) (n = 41); cantly superior to < 1 hours, excluded. Exclusions:1 from reported. Numbers n = 161 information on anaesthesia, Global rating paracetamol, placebo for most If > 1 hours but placebo for taking reporting adverse Age: 18+ years 3 hour washout prior to start. (5-point scale) 650 mg, + measures of efficacy. < 6 hours, PR = 0 concomitant effects (number of Interviewed in hospital by Subjective assessment codeine, and PI = baseline medication. effects): placebo nurse observer at 0, 0.5, of improvement 60 mg (n = 39) allocated for all 4/41 (6); paracetamol 1 hours then hourly for (7-point scale) further time points. + codeine 11/39 (20). 6 hours.

was no overlap between the 95% CI of the NNT for paracetamol, 600/650 mg, plus codeine, At least 50% PR with paracetamol plus codeine 60 mg, in 13 trials (95% CI, 2.6–3.9) and that 100 of paracetamol, 600/650 mg, alone in 17 trials (95% CI, 4.1–7.2). This indicates that addition of codeine, 60 mg, provides a substantial increase 75 in analgesia in single dose administration. This is demonstrated clearly in Figure 27 (see page 77), where wide CIs accompany point estimates of the NNT in trials with small numbers of patients. 50 Despite this, paracetamol combined with codeine, 60 mg, is clearly a powerful analgesic.

25 The extra analgesic effect of adding codeine, 60 mg, to paracetamol can be estimated in two ways. Since both paracetamol alone and paracetamol plus codeine were compared with placebo, 0 then any increased response rate (proportion of 0 25 50 75 100 patients with at least 50% pain relief) may be At least 50% PR with placebo ascribed to the addition of codeine. For paracetamol, 600/650 mg, alone against placebo, the difference between active (42%) and control FIGURE 25 Single dose studies of paracetamol plus codeine (23%) response rates was 19%. For paracetamol, against placebo (◆, 300 mg + 60 mg; ●, 600/650 mg + 60 mg; , 1000 mg + 60 mg) 600/650 mg, plus codeine, 60 mg, the difference between active (51%) and control (21%) response rates was 33%. Thus the extra 11% response was given placebo. The equivalent NNT at 600/650 mg due to the addition of codeine, 60 mg. was 5.3, indicating lower efficacy, although the dose response was not significant. There were also direct comparisons of paracetamol (all doses) plus codeine, 60 mg, with the same dose Paracetamol, 600/650 mg, plus codeine, 60 mg, of paracetamol alone. Here again, any increased compared with placebo in single dose adminis- response rate can be ascribed to the addition of tration had an NNT of 3.1 for at least 50% pain codeine. For paracetamol plus codeine, 60 mg, relief, meaning that one in every three patients versus the same dose of paracetamol the difference with pain of moderate to severe intensity will get between active (55%) and control (41%) response at least 50% pain relief; they would not have done rates was 14%. This agreement between direct so had they been given placebo (Table 26). There and indirect measures helps to justify the 73 Paracetamol with and without codeine in acute pain

TABLE 26 Summary risk ratios and NNTs for trials of paracetamol and codeine against placebo and paracetamol alone

Number Drug dose > 50% > 50% > 50% RB NNT of trials (mg) maxTOTPAR maxTOTPAR maxTOTPAR (95% CI) (95% CI) paracetamol on paracetamol on placebo on paracetamol +codeine + codeine alone

Paracetamol + codeine versus placebo 5 300 + 30 69/246 17/196 3.0 5.3 (1.8–5.0) (3.8–8.0) 13 600/650 + 60 200/398 80/418 2.6 3.1 (1.7–4.0) (2.6–3.9) 2 1000 + 60 48/77 5/50 6.2 1.9 (0.8–47) (1.5–2.6)

Paracetamol + codeine versus same dose of paracetamol alone 10 600/650 + 60 165/309 129/313 1.3 8.3 (1.0–1.6) (5.0–23) 2 1000 + 60 57/86 44/86 1.3 6.7 (1.1–1.5) (3.4–174) 12 All doses + 60 222/395 173/399 1.2 7.7 (1.1–1.5) (5.1–17)

meta-analytical methods. The extra 14% response for codeine, 60 mg, corresponds to an NNT for at Cooper et al., 1981 least 50% pain relief in single dose administration of 7.7 (95% CI, 5.1–17) (Table 26). This means that for every eight patients given paracetamol, 600/650 Forbes et al., 1982 mg, plus codeine, 60 mg, one extra will achieve at least 50% pain relief who would not have Forbes et al., 1983 done had they received paracetamol, 600/ 650 mg, alone. Honig & Murray, 1984

The variation in placebo and active response rates Gertzbein et al., 1986 was large, but this degree of variation is common in pain studies,18 as well as in studies with more Sunshine et al., 1986 objective outcomes like postoperative vomiting,19 and in the response of infants to pulmonary Bentley & surfactant.20 The variability in both the placebo Head, 1987 and active response rates (see Figures 24 and 25) Cooper et al., 1988 underpins the use of standard methods in pain research, where sensitivity of the model is demons- Forbes et al., 1989 trated by separation of standard analgesic from placebo. This variability also emphasises both the Forbes et al., 1990 need to include placebo groups in analgesic trials, and the need to understand better those factors Cooper & that contribute to the variability in placebo Kupperman, 1991 responses in pain. Dionne et al., 1994 The power of the systematic review method is demonstrated here in several ways. The analgesic Combined data effect of paracetamol at two doses has been determined with confidence from all the available pub- 0.1 1 10 100 lished data. The rather slight effects of codeine RB added to paracetamol (which are difficult to measure in single trials with limited numbers FIGURE 26 Relative benefit for individual trials of codeine, of patients) has been confirmed in direct and 60 mg, added to paracetamol, compared with the same dose 74 indirect comparison. of paracetamol alone Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 27 Studies of paracetamol plus codeine versus paracetamol

Bentley & Third molar RCT,double-blind, single PI (10-point scale) Paracetamol, Both paracetamol Rescue analgesic: 120 analysed. 53 had one or 3 Head, 1987 bony impacted oral dose, 4 parallel groups. PR (5-point scale) 1000 mg, + and combination Tylenol no. 3. For Exclusions: 3 did more, 86 reported n = 128 No information on anaes- codeine, 60 mg significantly superior patients who not take medication, in total; majority Mean age: thesiaexcept ‘no sedative or (n = 41); to placebo for all remedicated at 1 took only a were dizziness, mid-20s narcotic agents were used paracetamol, measures of efficacy; < 5 hours, last PI portion, 1 took no drowsiness, nausea before, during or after 1000 mg combination not and PR scores medication until day and vomiting. NSD surgery’. Self-assessed at (n = 41). significantly different carried on for all after surgery, 1 between treatment home at 0, 1, 2, 3, 4, 5 hours from paracetamol further time remedicated after groups. then reports posted to for any measure. points. 30 minutes, 1 Numbers reporting investigator. vomited within adverse effects 30 minutes of taking (number of effects): medication and 1 paracetamol + did not return codeine 15/42 (24); forms. paracetamol 21/42 (31).

Cooper & Impacted RCT,double-blind, 2 single PI (4-point scale) Study 2: Parametric analysis Remedication 160 analysed. None serious 5 Beaver, third molar oral dose studies (study 2 PR (5-point scale) paracetamol, concluded para- consisted of taking Exclusions: 30 did reported. Most 1976 n = 216 is paracetamol), 4 parallel Pain half gone? (y/n) 600 mg, + cetamol significantly second envelope not return forms, common were Age: groups. No information on Had patient fallen codeine, 60 mg, superior to placebo and evaluating it in 12 filled forms in drowsiness, nausea 16+ years anaesthesia. Self-assessed asleep during the (n = 40); (p < 0.01); non- same way as the improperly, 4 took and headache. at home at 0, 1, 2, 3 hours hour? (y/n) paracetamol, parametric factorial first; unclear how concomitant Numbers reporting (questionnaire) then 600 mg analysis showed data were then medication, 6 adverse effects posted to investigator. (n = 40). codeine had effects handled. 112/160 required no (number of effects): for first hour patients required medication, 4 were paracetamol + (p < 0.01) and peak remedication. randomly deleted to codeine 11/40 (11); PID (p < 0.05). No even out numbers paracetamol 5/40 (7). significant difference of patients on each found between treatment. combination and its constituents.

Cooper, Impacted RCT,double-blind, single PI (4-point scale) Paracetamol, All active treatments Remedicated after 200 analysed. None serious 4 et al., 1981 third molar oral dose, 5 parallel groups. PR (5-point scale) 650 mg, + significantly superior first hour if needed Exclusions: 17 did reported. Numbers n = 248 General/local anaesthesia. Global rating codeine, 60 mg to placebo for all (last score used for not ingest medi- reporting adverse Mean age: Self-assessed at home at (5-point scale) (n = 42); measures. Combi- all further time cation, 31 ingested effects (number early 20s 0, 1, 2, 3, 4 hours 50% relief of baseline paracetamol, nation was slightly points). medication but of effects): (questionnaire). pain? (y/n) 650 mg more effective than Remedication violated protocol paracetamol + Time to remedication (n = 37). paracetamol alone < 4 hours: placebo (remedicated before codeine 10/42 (10); but difference was 20/37; paracetamol first hour observ- paracetamol not significant. 2/37. ation, constant 12/37 (15). deviation of more than 15 minutes from evaluation times, not returning questionnaire and lost to follow-up).

Cooper, Impacted RCT,double-blind, single PI (4-point scale) Paracetamol, Combination signifi- Remedicated after 143 analysed. None serious 4 et al., 1988 third molar oral dose, 3 parallel groups, PR (5-point scale) 600 mg, + cantly superior to first hour (last or Exclusions: 11 lost reported. Numbers n = 165 single centre and 1 surgeon. Pain half gone? (y/n) codeine, 60 mg placebo for every baseline score used to follow-up, 8 did reporting adverse Age range: Local anaesthetic with Global rating (n = 31); measure and to for all further time not require effects: paracetamol 18–57 years sedative and/or nitrous (5-point scale) paracetamol, paracetamol for points). medication, 3 for + codeine 4; oxide. 4-hour washout Time to remedication 600 mg TOTPAR. Para- ‘various protocol paracetamol 8. prior to start. Self-assessed (n = 36). cetamol appeared violations’. at home at 0, 0.5, 1, 2, 3, 4, clinically more effec- 5, 6 hours (diary). tive than placebo but not significant. (p < 0.05).

Cooper & Removal of RCT,double-blind, single PI (4-point scale) Paracetamol, Paracetamol was t > 1 hour before 226 analysed. All reported mild. 3 Kupperman, one or more oral dose, 6 parallel groups. PR (5-point scale) 650 mg, + only active drug not remedication; data Exclusions; 13 did Numbers reporting 1991 impacted Local anaesthetic (lidocaine Global rating codeine, 60 mg significantly superior included and not require medi- adverse effects teeth + epinephrine) with i.v. (5-point scale) (n = 39); to placebo for any baseline or last cation, 3 lost to (number of effects): n = 247 diazepam and methohexital Time to remedication paracetamol, measure. Combi- score (most severe) follow-up, 2 remedi- paracetamol + Age:‘young (nitrous oxide also used on 650 mg nation significantly used for all further cated with slight codeine 8/39 (11); adults’ occasion). Self-assessed at (n = 37). superior to placebo time points. pain before hour 1 paracetamol 6/37 (7). home at 0, 0.5 hours then for most measures observations, 2 re- hourly for 6 hours (diary). and to paracetamol medicated before for TOTPAR and hour 1 observations, global rating. 1 fell asleep for over 2 hours.

continued 75 Paracetamol with and without codeine in acute pain

TABLE 27 contd Studies of paracetamol plus codeine versus paracetamol

Dionne, Impacted RCT,double-blind, single PI (4-point scale) Paracetamol, Neither paracetamol t ≥ 2 hours before 124 analysed. All reported mild. 3 et al., 1994 third molar oral dose, 5 parallel groups. PR (5-point scale) 650 mg, + nor combination remedicating; data Exclusions: 4 Numbers reporting n = 135 General/local anaesthesia. Global rating codeine, 60 mg were significantly included and previously enrolled adverse effects Age: 16+ years 4-hour washout prior to (5-point scale) (n = 24); different from baseline used for all in study, 3 remedi- (number of effects): start. Self-assessed at clinic Time to remedication paracetamol, placebo for any further time points. cated before paracetamol + for at least first 2 hours 650 mg measure of analgesia. t = 2 hours, 2 lost codeine 9/24 (10); then at home hourly for (n = 27). to follow-up, 1 paracetamol 7/27 (9). 6 hours. ineligible because of codeine sensitivity.

Forbes, Impacted Double-blind, single oral PI (4-point scale) Paracetamol, At 4 hours para- If remedicated 159 analysed. None of active 3 et al., third molar dose, 5 parallel groups. No PR (5-point scale) 600 mg, + cetamol significantly < 2 hours, excluded. Exclusions: 4 lost treatments 1982 (1 or more) information on anaesthesia. 50% PR? (y/n) codeine, 60 mg superior to placebo If > 2 < 12 hours, to follow-up, 3 did produced more n = 177 Self-assessed at home at Global rating (n = 31); for PR, peak PR and PR = 0 and PI not take medication, than placebo. None Age: 0, 1 hours then hourly for (5-point scale) paracetamol, 50% PR (p < 0.01). baseline or last. 7 took remedication were serious. 15+ years 12 hours. Time to remedication 600 mg Combination signifi- % patients remedi- at < 2 hours, 1 (n = 34). cantly superior to cated at > 2 < 12 medicated with mild placebo for all hours: placebo 93%; pain, 3 did complet- measures. paracetamol 85%. ed forms improperly.

Forbes, Postoperative RCT,double-blind, single PI (4-point scale) Paracetamol, For 6-hour data Patients remedi- 132 analysed; None serious 5 et al.., (general, oral dose, 5 parallel groups. PR (5-point scale) 600 mg, + both paracetamol cated on demand no exclusions. reported. Most 1983 gynaecological General anaesthetic, given Pain half gone? (y/n) placebo, 60 mg, and combination but if at < 2 hours, common were or orthopaedic on request on first day able Global rating (n = 26); significantly superior excluded. If remedi- drowsiness, dizziness surgery) to take oral analgesia. Self- (5-point scale) paracetamol, to placebo for all cated < 12 hours, and dry mouth. n = 132 assessed in hospital at 0, Time to remedication 600 mg measures and for PR = 0, and PI = Numbers reporting Age: 18+ years 0.5, 1, 1.5, 2 hours then (n = 26). 12-hour data, for all baseline or last adverse effects hourly for 12 hours. Single except SPID. score for remaining (number of effects): nurse observer present for time points. paracetamol + first 6 hours, next 6 hours codeine 11/26 (16); monitored by ward staff. paracetamol 11/26 (11).

Forbes, Impacted RCT,double-blind, single PI (4-point scale) Paracetamol, At 4 hours Remedication 88 analysed. None serious 5 et al., third molar oral dose, 4 parallel groups PR (5-point scale) 600 mg, + paracetamol and allowed after Exclusions: 9 did reported. Numbers 1989 (1 or more) at 2 centres. General/local Global rating codeine, 60 mg combination signifi- 2 hours but asked not take medication, reporting adverse n = 107 anaesthesia (unclear). Self- (5-point scale) (n = 17); cantly superior to complete next 2 remedicated effects (number of Age: 15+ years assessed at home at 0, 1, Time to remedication paracetamol, to placebo. evaluation before before 2 hours, 1 effects): paracetamol 2 hours, then hourly for 600 mg Combination also doing so (PI last or remedicated with + codeine 1/17 (1); 12 hours or until remedi- (n = 22). significantly superior baseline score, PR slight pain, 4 did not paracetamol 3/26 (3). cation (diary). Diary to paracetamol for = 0 for all further complete evaluation, NB: includes those returned 5 days later for TOTPAR (p < 0.05). time points). 1 took only part of reported post- review and debriefing. % remedicating medication, 2 remedication. by hour 12: remedicated despite paracetamol 95%; of having some relief placebo 91%. from study medication.

Forbes, Impacted RCT,double-blind, single PI (4-point scale) Paracetamol, All active medi- Remedication 206 analysed. None serious 5 et al., third molar then multiple oral dose, PR (5-point scale) 600 mg, + cations significantly allowed after Exclusions; 3 lost reported. Numbers 1990 (1 or more) 6 parallel groups. General/ Pain half gone? (y/n) codeine, 60 mg superior for all 2 hours but asked to follow-up, 1 lost reporting adverse n = 269 local anaesthesia (unclear). Global rating (n = 38; measures of total to complete next report card, 22 did effects (number Age: 15+ years Self-assessed at home at 0, (5-point scale) paracetamol, and peak analgesia. evaluation (PI last not require medi- of effects): 1 hours then hourly for Time to remedication 600 mg or baseline score, cation, 8 remedi- paracetamol + 6 hours. Returned 5 days (n = 36). PR = 0 for all re- cated despite having codeine 8/40 (9); later for review and maining time points). relief from study paracetamol 5/41 (5). debriefing. % remedicated by medication, 6 re- hour 6: placebo 33%; medicated with only paracetamol 29%. slight pain, 13 remedicated at < 2 hours, 7 failed to follow instructions, 3 did not complete forms.

Gertzbein, Postoperative RCT,single oral dose, 3 PI (5-point scale) Paracetamol, Combination seen Remedication 113 analysed. None serious 4 et al., (elective surgery parallel groups. No PR (5-point scale) 1000 mg, + to give higher allowed after 1 hour Exclusions: 1 refused reported. Numbers 1986 – orthopaedic information on anaesthesia, 50% PR? (y/n) codeine, 60 mg efficacy results than (PI = last score and to comply with reporting adverse or general) First oral analgesic given. Global rating (n = 45); paracetamol alone PR = 0 for all re- instructions, 1 effects (number of n = 116 Interviewed in hospital by (4-point scale): patient paracetamol, but was not signifi- maining time points). vomited within effects): paracetamol Age range: nurse-observer at 0, 0.5, and observer VAS PI 1000 mg cantly superior for Mean time to re- 1 hour of taking + codeine 13/47 (13); 16–65 years 1 hours, then hourly for (‘No pain’ to ‘Worst (n = 45). any measure. medication: combi- study medication, paracetamol 13/46 (15). 5 hours. pain I can imagine’) nation 230 minutes; 1 took concomitant paracetamol analgesia. 214 minutes.

76 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 27 contd Studies of paracetamol plus codeine versus paracetamol

Honig & Postoperative RCT,single oral dose, 4 PI (4-point scale) Paracetamol, For TOTPAR and If remedicated, last No details given. None severe except 3 Murray, (elective surgery parallel groups. No PR (5-point scale) 600 mg, + global rating both score used for all 1 severe dry mouth. 1984 – abdominal, information on anaesthesia, 50% PR? (y/n) codeine, 60 mg paracetamol and further time points. Reported in all orthopaedic, 4-hour washout prior to Global rating (n = 30); combination signifi- Number not groups; primarily rectal, thoracic start. Interviewed in (5-point scale) paracetamol, cantly superior to remedicated at central nervous and vascular) hospital by nurse observer 600 mg placebo (p < 0.05). 6 hours: placebo system and gastro- n = 116 at 0, 0.5, 1 hours then (n = 28). Combination also 14/30; paracetamol intestinal effects. Age range: hourly for 6 hours. significantly superior 16/28. 19–87 years for SPID and number of patients remedicating before 6 hours.

Sunshine, Impacted RCT,double-blind, single PI (4-point scale) Paracetamol, Both paracetamol Excluded if at 182 analysed; None serious 5 et al., third molar oral dose, 6 parallel groups. PR (5-point scale) 650 mg, + and combination < 1 hour. If at no exclusions. reported; NSD 1986 n = 182 Local anaesthetic (lidocaine/ Global rating codeine, 60 mg significantly superior > 1 hour, last PI or between groups. Age: 16+ years epinephrine). 4-hour (4-point scale) (n = 31); (p < 0.05) to baseline and PR = 0 Numbers reporting washout prior to start. Overall improvement paracetamol, placebo for: PID at used. Full adverse effects: Evaluations at 0, 0.5, 1, 2 (7-point scale) 650 mg 1–3 hours, SPID at dichotomous data paracetamol + and 3 hours in clinic by Time to remedication (n = 30). 4 hours, PR at given in Table III. codeine 3/31; single observer. At 4, 5 2 hours and time to Number remedi- paracetamol 1/30. and 6 hours self-assessed. peak effect. Combi- cating: placebo 1 week later met with nation also signifi- 13/30; paracetamol observer and reviewed cantly superior 14/30. forms. for TOTPAR.

Paracetamol, 500 mg 353

Paracetamol, 600/650 mg 594

Paracetamol, 1000 mg 1376

Paracetamol, 1500 mg 207

Paracetamol, 600/650 mg, + codeine, 60 mg 398

Paracetamol, 1000 mg, + codeine, 60 mg 77

12345678910 NNT (95% CI)

FIGURE 27 Relative effectiveness of paracetamol doses and paracetamol/codeine combinations

References 4. Jadad AR, Carroll D, Moore A, McQuay H. Developing a database of published reports of 1. Moore A, Collins S, Carroll D, McQuay H. randomised clinical trials in pain research. Pain Paracetamol with and without codeine in acute 1996;66:239–46. pain: a quantitative systematic review. Pain 1997;70:193–201. 5. de Craen AJM, di Giulio G, LampeSchoenmaeckers AJE, Kessels AGH, Kleijnen J. Analgesic efficacy and 2. Government Statistical Service. Prescription cost safety of paracetamol–codeine combinations versus analysis for England 1995. London: Department of paracetamol alone: a systematic review. BMJ Health, 1996. 1996;313:321–5. 3. Jadad AR. Meta-analysis of randomised clinical trials 6. Collins SL, Moore RA, McQuay HJ. The visual in pain relief [DPhil thesis]. Oxford: University of analogue pain intensity scale: what is moderate Oxford, 1994. pain in millimetres? Pain 1997;72:95–7. 77 Paracetamol with and without codeine in acute pain

7. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in 21. Moore RA, McQuay HJ. Single-patient data meta- clinical research. Ann Intern Med 1987;107:224–33. analysis of 3453 postoperative patients: oral tramadol versus placebo, codeine and combination 8. Cooper SA. Single-dose analgesic studies: the upside analgesics. Pain 1997;69:287–94. and downside of assay sensitivity. Adv Pain Res Ther 1991;18:117–24. Studies included 9. Moore A, McQuay H, Gavaghan D. Deriving Beaver WT, McMillan D, 1980. Methodological dichotomous outcome measures from continuous considerations in the evaluation of analgesic combi- data in randomised controlled trials of analgesics. nations: acetaminophen (paracetamol) and hydro- Pain 1996;66:229–37. codone in postpartum pain. Br J Clin Pharmacol;10 10. Moore A, McQuay H, Gavaghan D. Deriving suppl:215–23S. dichotomous outcome measures from continuous Bentley KC, Head TW, 1987. The additive analgesic data in randomised controlled trials of analgesics: efficacy of acetaminophen, 1000 mg, and codeine, verification from independent data. Pain 60 mg, in dental pain. Clin Pharmacol Ther;42:634–40. 1997;69:127–30. Berry FN, Miller JM, Levin HM, Bare WW, Hopkinson 11. Moore A, Moore O, McQuay H, Gavaghan D. JH, Feldman AJ, 1975. Relief of severe pain with Deriving dichotomous outcome measures from acetaminophen in a new dose formulation versus continuous data in randomised controlled trials of propoxyphene hydrochloride 65 mg and placebo: a analgesics: use of pain intensity and visual analogue comparative double blind study. Curr Ther Res;17:361–8. scales. Pain 1997;69:311–15. Bjune K, Stubhaug A, Dodgson MS, Breivik H, 1996. 12. DerSimonian R, Laird N. Meta-analysis of clinical Additive analgesic effect of codeine and paracetamol trials. Control Clin Trials 1986;7:177–88. can be detected in strong, but not moderate, pain after Caesarean section. Baseline pain-intensity is a deter- 13. Cook RJ, Sackett DL. The number needed to treat: minant of assay-sensitivity in a postoperative analgesic a clinically useful measure of treatment effect. trial [see comments]. Acta Anaesthesiol Scand;40:399–407. BMJ 1995;310:452–4. Cooper SA, Kupperman A, 1991. The analgesic efficacy 14. Berry FN, Miller JM, Levin HM, Bare WW, of flurbiprofen compared to acetaminophen with Hopkinson JH, Feldman AJ. Relief of severe pain codeine. J Clin Dent;2:70–4. with acetaminophen in a new dose formulation Cooper SA, Precheur H, Rauch D, Rosenheck A, Ladov versus propoxyphene hydrochloride 65 mg and M, Engel J, 1980. Evaluation of oxycodone and placebo: a comparative double blind study. acetaminophen in treatment of postoperative dental Curr Ther Res 1975;17:361–8. pain. Oral Surg Oral Med Oral Pathol;50:496–501. 15. Dolci G, Ripari M, Pacifici L, Umile A. Evaluation Cooper SA, Breen JF, Giuliani RL, 1981. The relative of piroxicam-beta-cyclodextrin, piroxicam, para- efficacy of indoprofen compared with opioid analgesic cetamol and placebo in post-operative oral surgery combinations. J Oral Surg;39:21–5. pain. Int J Clin Pharmacol Res 1994;14:185–91. Cooper SA, Erlichman MC, Mardirossian G, 1986. 16. Kiersch TA, Halladay SC, Hormel PC. A single-dose, Double blind comparison of an acetaminophen codeine double-blind comparison of naproxen sodium, caffeine combination in oral surgery pain. Anesth acetaminophen, and placebo in postoperative Prog;33:139–42. dental pain. Clin Ther 1994;16:394–404. Cooper SA, Firestein A, Cohn P, 1988. Double blind 17. Turek M, Baird W. Double blind parallel com- comparison of meclofenamate sodium with acetamino- parison of ketoprofen (Orudis), acetaminophen phen, acetaminophen with codeine and placebo for plus codeine, and placebo in postoperative pain. relief of postsurgical dental pain. J Clin Dent;1:31–4. J Clin Pharmacol 1988;28:S23–8. Cooper SA, Schachtel BP, Goldman E, Gelb S, Cohn P, 1989. Ibuprofen and acetaminophen in the relief of 18. McQuay HJ, Tramèr M, Nye BA, Carroll D, Wiffen acute pain: a randomized, double blind, placebo PJ, Moore RA. A systematic review of antidepressants controlled study. J Clin Pharmacol;29:1026–30. in neuropathic pain. Pain 1996;68:217–27. Desjardins PJ, Cooper SA, Finizio T, 1986. Efficacy of low 19. Tramèr M, Moore A, McQuay H. Prevention of dose combination analgesics: acetaminophen/codeine, vomiting after paediatric strabismus surgery: a aspirin/butalbital/caffeine/codeine, and placebo in oral systematic review using the numbers-needed-to-treat surgery pain. Anesth Prog;33:143–6. method. Br J Anaesth 1995;75:556–61. Dionne RA, Snyder J, Hargreaves KM, 1994. Analgesic 20. Soll JC, McQueen MC. Respiratory distress efficacy of flurbiprofen in comparison with acetamino- syndrome. In: Sinclair JC, Bracken ME, editors. phen, acetaminophen plus codeine, and placebo after Effective care of the newborn infant. Oxford: impacted third molar removal. J Oral Maxillofac 78 Oxford University Press, 1992: chapter 15, p. 333. Surg;52:919–24; discussion 25–6. Health Technology Assessment 1998; Vol. 2: No. 12

Dolci G, Ripari M, Pacifici L, Umile A, 1994. Evaluation Jain AK, Ryan JR, McMahon FG, Smith G, 1986. of piroxicam-beta-cyclodextrin, piroxicam, paracetamol Comparison of oral nalbuphine, acetaminophen, and and placebo in post-operative oral surgery pain. Int J Clin their combination in postoperative pain. Clin Pharmacol Pharmacol Res;14:185–91. Ther;39:295–9. Fassolt A, Stocker H, 1983. Behandlung des post- Kiersch TA, Halladay SC, Hormel PC, 1994. A single- operativen Wundschmerzes mit Suprofen [Treatment dose, double-blind comparison of naproxen sodium, of postoperative wound pain with suprofen]. acetaminophen, and placebo in postoperative dental Arzneimittelforschung;33:1327–30. pain. Clin Ther;16:394–404. Forbes JA, Beaver WT, White EH, White RW, Neilson GB, Laska EM, Sunshine A, Zighelboim I, Roure C, Shackleford RW, 1982. Diflunisal: a new oral analgesic Marrero I, Wanderling J, et al., 1983. Effect of caffeine with an unusually long duration of action. on acetaminophen analgesia. Clin Pharmacol JAMA;248:2139–42. Ther;33:498–509. Forbes JA, Kolodny AL, Beaver WT, Shackleford RW, Lehnert S, Reuther J, Wahl G, Barthel K, 1990. Scarlett VR, 1983. A 12 hour evaluation of the analgesic Wirksamkeit von Paracetamol (Tylenol) und efficacy of diflunisal, acetaminophen, and acetamino- Acetylsalizylsaure (Aspirin) bei postoperativen phen codeine combination, and placebo in postoperative Schmerzen [The efficacy of paracetamol (Tylenol) and pain. Pharmacotherapy;3:47–54S. acetyl salicylic acid (Aspirin) in treating postoperative pain]. Dtsch Zahnarztl Z;45:23–6. Forbes JA, Barkaszi BA, Ragland RN, Hankle JJ, 1984. Analgesic effect of acetaminophen, phenyltoloxamine McQuay HJ, Carroll D, Frankland T, Harvey M, Moore and their combination in postoperative oral surgery pain. A, 1990. Bromfenac, acetaminophen, and placebo in Pharmacotherapy;4:221–6. orthopedic postoperative pain. Clin Pharmacol Ther;47:760–6. Forbes JA, Kolodny AL, Chachich BM, Beaver WT, 1984. Nalbuphine, acetaminophen, and their combination in Mehlisch DR, Frakes LA, 1984. A controlled comparative postoperative pain. Clin Pharmacol Ther;35:843–51. evaluation of acetaminophen and aspirin in the treatment of postoperative pain. Clin Ther;7:89–97. Forbes JA, Jones KF, Smith WK, Gongloff CM, 1986. Analgesic effect of an aspirin codeine butalbital Mehlisch DR, Sollecito WA, Helfrick JF, Leibold DG, caffeine combination and an acetaminophen codeine Markowitz R, Schow CEJ, et al., 1990. Multicenter clinical combination in postoperative oral surgery pain. trial of ibuprofen and acetaminophen in the treatment Pharmacotherapy;6:240–7. of postoperative dental pain. J Am Dent Assoc;121:257–63. Forbes JA, Butterworth GA, Burchfield WH, Yorio CC, Mehlisch DR, Jasper RD, Brown P, Korn SH, McCarroll K, Selinger LR, Rosenmertz SK, et al., 1989. Evaluation Murakami AA, 1995. Comparative study of ibuprofen of flurbiprofen, acetaminophen, an acetaminophen- lysine and acetaminophen in patients with postoperative codeine combination, and placebo in postoperative dental pain. Clin Ther;17:852–60. oral surgery pain. Pharmacotherapy;9:322–30. Pande AC, Pyke RE, Greiner M, Cooper SA, Benjamin R, Forbes JA, Butterworth GA, Burchfield WH, Beaver Pierce MW, 1996. Analgesic efficacy of the kappa- WT, 1990. Evaluation of ketorolac, aspirin, and an receptor agonist, enadoline, in dental surgery pain. acetaminophen–codeine combination in postoperative Clin Neuropharmacol;19:92–7. oral surgery pain. Pharmacotherapy;10:77–93S. Petti A, 1985. Postoperative pain relief with pentazocine Forbes JA, Kehm CJ, Grodin CD, Beaver WT, 1990. Evalu- and acetaminophen: comparison with other analgesic ation of ketorolac, ibuprofen, acetaminophen, and an combinations and placebo. Clin Ther;8:126–33. acetaminophen–codeine combination in postoperative Rubin A, Winter LJ, 1984. A double blind randomized oral surgery pain. Pharmacotherapy;10:94–105S. study of an aspirin/caffeine combination versus Forbes JA, Bates JA, Edquist IA, Burchfield WH, Smith acetaminophen/aspirin combination versus acetamino- FG, Schwartz MK, et al., 1994. Evaluation of two opioid- phen versus placebo in patients with moderate to severe acetaminophen combinations and placebo in post- post partum pain. J Int Med Res;12:338–45. operative oral surgery pain. Pharmacotherapy;14:139–46. Schachtel B, Thoden W, Baybutt R, 1989. Ibuprofen and Gertzbein SD, Tile M, McMurty RY, Kellam JF, Hunter acetaminophen in the relief of postpartum episiotomy GA, Keith RG, et al., 1986. Analysis of the analgesic pain. J Clin Pharmacol;29:550–3. efficacy of acetaminophen 1000 mg, codeine phosphate Seymour RA, Kelly PJ, Hawkesford JE, 1996. The efficacy 60 mg, and the combination of acetaminophen 1000 mg of ketoprofen and paracetamol (acetaminophen) in and codeine phosphate 60 mg in the relief of post- postoperative pain after third molar surgery. Br J Clin operative pain. Pharmacotherapy;6:104–7. Pharmacol;41:581–5. Heidrich G, Slavic Svircev V, Kaiko RF, 1985. Efficacy and Stubhaug A, Grimstad J, Breivik H, 1995. Lack of quality of ibuprofen and acetaminophen plus codeine analgesic effect of 50 and 100 mg oral tramadol after analgesia. Pain;22:385–97. orthopaedic surgery: a randomized, double-blind, Honig S, Murray KA, 1984. An appraisal of codeine as an placebo and standard active drug comparison. 79 analgesic: single dose analysis. J Clin Pharmacol;24:96–102. Pain;62:111–18. Paracetamol with and without codeine in acute pain

Sunshine A, Marrero I, Olson N, McCormick N, Laska Turek M, Baird W, 1988. Double blind parallel com- EM, 1986. Comparative study of flurbiprofen, zomepirac parison of ketoprofen (Orudis), acetaminophen plus sodium, acetaminophen plus codeine, and acetamino- codeine, and placebo in postoperative pain. J Clin phen for the relief of postsurgical dental pain. Pharmacol;28:S23–8. Am J Med;80:50–4. Winnem B, Samstad B, Breivik H, 1981. Paracetamol, tiaramide and placebo for pain relief after orthopedic Sunshine A, Zighelboim I, De Castro A, Sorrentino JV, surgery. Acta Anaesthesiol Scand;25:209–14. Smith DS, Bartizek RD, et al., 1989. Augmentation of acetaminophen analgesia by the antihistamine Winter LJ, Appleby F, Ciccone PE, Pigeon JG, 1983. A phenyltoloxamine. J Clin Pharmacol;29:660–4. comparative study of an acetaminophen analgesic combination and aspirin in the treatment of postoperative oral surgery pain. Curr Ther Res;33:115–22. Sunshine A, Olson NZ, Zighelboim I, de Castro A, 1993. Ketoprofen, acetaminophen plus oxycodone, and Young RE, Quigley JJ, Archambault WAJ, Gordon LL. acetaminophen in the relief of postoperative pain. Butorphanol/acetaminophen double blind study in Clin Pharmacol Ther;54:546–55. postoperative pain. J Med 1979;10:239–56.

80 Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 9 Oral ibuprofen and diclofenac in postoperative pain

Summary scriptions (31% of total NSAID prescriptions) and diclofenac for nearly 6 million prescriptions The aim of this review was to compare ibuprofen (34%), although it is not known how many were and diclofenac in postoperative pain. Studies were for acute pain conditions.2 identified by an extensive literature search, with additional reports being identified from the refer- With an increasing amount of surgery being ence lists of reports, review articles and textbooks. performed as day cases, it is important to know which drug should be recommended for post- The studies of interest were randomised, control- operative pain relief. The relative efficacy of the led, single dose comparisons of ibuprofen or diclo- two drugs was compared to allow a balanced fenac against placebo. Summed pain relief or SPID decision to be made, based on efficacy, safety over 4–6 hours was extracted, and converted into and cost. dichotomous information yielding the numbers of patients with at least 50% pain relief. This was then used to calculate the relative benefit and the NNT Methods for one patient to achieve at least 50% pain relief. Single dose, RCTs of ibuprofen and diclofenac in In all, 34 reports compared ibuprofen and placebo postoperative pain (post-dental extraction, post- (3591 patients), six compared diclofenac with surgical or postpartum pain) were sought. Different placebo (840 patients), and there were two direct search strategies were used to identify eligible comparisons of diclofenac, 50 mg, and ibuprofen, reports from MEDLINE (1966–December 1996), 400 mg, (130 patients). In postoperative pain, EMBASE (1980–January 1997), the Cochrane ibuprofen, 200 mg, had an NNT of 3.3 (95% CI, Library (August 1996), Biological Abstracts 2.8–4.0) compared with placebo, ibuprofen, (1985–96) and the Oxford Pain Relief Database 400 mg, had an NNT of 2.7 (95% CI, 2.5–3.0) (1950–94).3 A search was undertaken for each drug and ibuprofen, 600 mg, had an NNT of 2.4 (95% using the terms, ‘clinical trial’, ‘trial’, ‘study’, CI, 1.9–3.3). Diclofenac, 50 mg, had an NNT of ‘random*’, ‘double blind’, ‘analgesi*’ and ‘pain*’, 2.3 (95% CI, 2.0–2.7) compared with placebo in together with ‘ibuprofen’, ‘Brufen’, ‘propionic established postoperative pain; diclofenac, acid’ and ‘isobutylphenyl propionic acid’ for the 100 mg, had an NNT of 1.8 (95% CI, 1.5–2.1). ibuprofen search and, for the diclofenac search, ‘diclofenac’, and 76 brand names.4 Each was a When diclofenac, 50 mg, was compared directly with broad free text search, including various combi- ibuprofen, 400 mg, there was no significant differ- nations of the words, and without restriction to ence between the two treatments. Ibuprofen showed language. Additional reports were identified from a clear dose response with relative efficacy similar to the reference lists of retrieved reports, review that for diclofenac. Both drugs work well. Choosing articles and textbooks. between them is an issue of dose, safety and cost. Included reports This chapter of the review has been published in Neither pharmaceutical companies nor authors of full by Collins and colleagues.1 papers were contacted for unpublished reports. Abstracts and review articles were not considered. The inclusion criteria used were: Introduction • randomised allocation to treatment groups Ibuprofen and diclofenac are two of the most which included ibuprofen or diclofenac widely used NSAIDs, with ibuprofen commonly and placebo available without prescription. In England in 1996, • full journal publication ibuprofen accounted for nearly 5.5 million pre- • established postoperative pain with the pain 81 Oral ibuprofen and diclofenac in postoperative pain

outcome measured using a five-point pain NNT and 95% CIs were calculated using Cook relief scale with standard wording (none, slight, and Sackett’s method.12 A statistically significant moderate, good, complete) or a four-point pain difference from control was assumed when the intensity scale (none, mild, moderate or severe) 95% CI of the relative benefit did not include 1. or a VAS for pain relief or pain intensity, TOTPAR or SPID (at 4, 5 or 6 hours) as a The number of patients experiencing at least 50% derived pain relief outcome (or sufficient data pain relief with placebo (the CER) can vary greatly provided to allow their calculation) with the relatively small sample sizes used typically • postoperative oral administration for ibuprofen in analgesic trials affecting the apparent efficacy of and postoperative oral, rectal, intravenous or an analgesic.13 To allow for this variation, the rela- intramuscular administration for diclofenac tive benefit and NNT for each dose of ibuprofen • adult patients and diclofenac were also calculated using a fixed • baseline pain of moderate to severe intensity CER of 19%. This value was obtained from data (for VAS this equates to > 30 mm5) for 4378 patients given placebo, pooled from • double-blind design. 124 single dose analgesic trials meeting identical inclusion criteria included in this and similar Excluded reports systematic reviews (843/4378 patients experienced Reports were excluded of: at least 50% pain relief).14–17

• ibuprofen or diclofenac used for the relief of other pain conditions Results • controlled release formulations • ibuprofen or diclofenac used in combination Ibuprofen versus placebo with other drugs There were 34 reports of 35 trials that fulfilled our • trials which reported data from a crossover inclusion criteria; a total of 2214 patients were design as a single data set given ibuprofen and 1377 placebo. Details of the • trials in which the number of patients per studies are presented in Table 28 with their refer- treatment group was less than ten6 ences being listed at the end of this chapter. The • trials which included pain relief data collected author of one report was contacted for information after additional analgesia was given. on the number of patients in each treatment arm.18

Data extraction and analysis Oral surgery pain was investigated in 25 trials From each report the numbers of patients treated (predominantly third molar extraction with bone was taken, together with the mean TOTPAR, SPID, removal), in five trials postpartum pain was investi- VAS TOTPAR or VAS SPID, study duration and the gated (predominantly episiotomy and Caesarean doses given. Information on adverse events was also section), and in a further four trials postoperative extracted. For each report, the mean TOTPAR, pain was studied (one tonsillectomy, one inguinal SPID, VAS TOTPAR or VAS SPID values for active hernia, one orthopaedic surgery and one general and placebo were converted to % maxTOTPAR or surgery). The doses of ibuprofen prescribed were % maxSPID by division into the calculated maxi- 50 mg in one trial, 100 mg in two, 200 mg in eight, mum value.7 The proportion of patients in each 400 mg in 30, 600 mg in three, and 800 mg treatment group who achieved at least 50% in one. maxTOTPAR was calculated using verified equations.8–10 The proportions were then con- The CER (the proportion of patients given verted into the number of patients achieving at placebo experiencing at least 50% pain relief) least 50% maxTOTPAR by multiplying by the ranged from 0% to 67% (median 12%) (Figure 28). total number of patients in the treatment group. For the single trial of ibuprofen, 50 mg, the EER Information on the number of patients with (the proportion of patients given ibuprofen experi- at least 50% maxTOTPAR for active drug and encing at least 50% pain relief) was 28%. For ibu- placebo was then used to calculate the relative profen, 100 mg, two trials gave EER values of 27% benefit and NNT. and 8%. For ibuprofen, 200 mg, the EER varied between 6% and 57% (median 39%), and for ibu- Relative benefit estimates were calculated with profen, 400 mg, it varied between 13% and 100% 95% CIs using a random effects model;11 the (median 60%). For the single trial of ibuprofen, random effects model was chosen because it 800 mg, the EER was 100%. The 100 mg and produces the most conservative estimate 200 mg data sets were homogeneous but the 82 (homogeneity was assumed when p > 0.1). 400 mg and 600 mg data sets were not. Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 28 Studies of ibuprofen versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Ahlstrom, Third molar RCT,double-blind, single VAS PI:‘no Ibuprofen, Ibuprofen significantly Patients allowed to 97 analysed. None serious reported and et al., 1993 extraction oral dose, parallel groups. pain at all’ to 400 mg superior to placebo by remedicate after Exclusions: 30 for no patient withdrew as a n = 127 4-hour washout before ‘agonising pain’ (n = 32); 40 minutes (p = 0.01); this 1 hour. After various protocol result. Numbers reporting Age range: start. Evaluated at 0, 20, Global rating placebo continued for 6 hours. remedication, PI = violations. adverse effects (number of 18–40 years 40 minutes 1 hour then by patient (n = 30). TOTPI & SPID: ibuprofen last score carried effects): ibuprofen 3/32 (?); hourly intervals for significantly superior to forward for all placebo 2/30 (?). 6 hours. Medication taken placebo (p < 0.0001). further time when baseline PI at least 6 hour SPID: ibuprofen points. moderate (> 30 mm). 188 mm; placebo 32 mm.

Arnold, General surgery RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen not significantly After remedication No information on Difference in occurrence not et al., 1990 (including oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for PR = 0 and PI = any exclusions. significant between groups. gynaecological Assessed by single nurse- PR (5-point scale), (n = 15); either SPID or TOTPAR. baseline score for No patient withdrew from and orthopaedic) observer at 0, 0.5, 1 hour viz, standard time placebo 6-hour TOTPAR: ibuprofen all further time either ibuprofen or placebo n = 59 then hourly intervals for to meaningful (n = 14). 4.2; placebo 1.5. points. group as a result of adverse Age range: 6 hours. Medication taken relief events. 22–70 years when baseline PI moderate Global rating to severe. (5-point scale) by patient

Bakshi, Third molar RCT,double-blind, single VAS PI:‘no pain’ Ibuprofen, Ibuprofen significantly Patients allowed to 245 analysed. None serious reported and et al., 1994 extraction oral dose, parallel groups. to ‘pain could 400 mg superior to placebo for remedicate after Exclusions: 9 did not no patient withdrew as a n = 257 Local anaesthetic. Self- not be worse’ (n = 80); TOTPAR and both global 1 hour. If they experience severe result. Numbers reporting Age: adults up assessed at 0, 20 and PR (5-point placebo ratings (p < 0.01). 6-hour remedicated pain, 2 remedicated adverse effects (number of to 65 years 40 minutes, 1, 1.5, and scale): none, (n = 82). TOTPAR: ibuprofen 14.9; earlier, data before 1 hour, 1 effects): ibuprofen 6/80 (?); 2 hours, then hourly for poor, moderate, placebo 8.85. excluded from completed diary placebo 5/82 (?) 6 hours. Medication taken sufficient, total efficacy analysis. incorrectly. when baseline PI was at Global rating After remedication least severe. (5-point scale) by PR = 0 and PI = patient and by last score for all observer further time points.

Cooper, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen at both doses Patients allowed to 192 analysed. None serious reported and et al., 1977 extraction oral dose, parallel groups. 4-point scale) 400 mg significantly superior to remedicate after Exclusions: 17 no patient withdrew as a n = 245 Local anaesthetic. Self- PR (standard (n = 40); placebo for all measures 2 hours. If they provided uninter- result. No individual data Age: ? assessed at home at 5-point scale) ibuprofen, of efficacy (p < 0.05). remedicated earlier pretable data, 12 provided but NSD in 0 hours then hourly for 50% PR? (y/n) 200 mg 4-hour TOTPAR: ibuprofen, data excluded from took confounding occurrence between groups. 4 hours. Medication taken Global rating (n = 38); 400 mg, 7.32; ibuprofen, efficacy analysis. medication, 10 lost when baseline PI moderate (5-point scale) placebo 200 mg, 6.27; placebo 3.32. After remedication to follow-up, 9 did to severe. by patient (n = 40). PR = 0 and PI = not need medication, baseline score for 5 fell asleep. all further time points.

Cooper, Third molar RCT,double-blind, single PI (standard Ibuprofen, All active treatments Patients allowed 249 analysed. None serious reported; no et al., 1982 extraction oral dose, parallel groups. 4-point scale) 400 mg significantly superior to to remedicate after Exclusions: 30 lost patient withdrew as a result. n = 316 Mostly local anaesthetic. PR (standard (n = 38); placebo for SPID and 1 hour. If they to follow-up, 15 did Numbers reporting adverse Age range: Self-assessed at home at 5-point scale) placebo TOTPAR (no p-value remedicated earlier not require effects (number of effects): 16–65 years 0 hours then hourly for 50% PR? (y/n) (n = 46); given). 4-hour TOTPAR: data excluded from medication, 11 ibuprofen 11/38 (12); placebo 4 hours. Medication taken Global rating ibuprofen, ibuprofen 8.39; placebo efficacy analysis. remedicated before 5/46 (6); ibuprofen + codeine when baseline PI was (5-point scale) 400 mg, + 2.65; ibuprofen + codeine After remedication 1 hour, 6 missed 18/41 (20); codeine moderate to severe. by patient codeine, 9.39; codeine 4.12. PR = 0 and PI = more than 1 evalu- 11/41 (11). 60 mg baseline score for ation, 3 medicated (n = 41); all further time with slight pain, 1 codeine, points. did not take all 60 mg medication, 1 medi- (n = 41). cated over 24 hours after surgery.

Cooper, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 161 analysed. None serious reported; no et al., 1988 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 20 did patient withdrew as a result. n = 201 Local anaesthetic + PR (standard (n = 37); all measures of efficacy 1 hour. If they not require Numbers reporting adverse Age:Adult sedative. Self-assessed at 5-point scale) placebo (p < 0.01). 6-hour remedicated earlier medication, 13 lost effects (number of effects): home at 0 hours, then 50% PR? (y/n) (n = 43). TOTPAR: ibuprofen data excluded from to follow-up, 7 for ibuprofen 10/40 (14); placebo hourly for 6 hours. Medi- Global rating 11.32; placebo 4.67. efficacy analysis. various protocol 7/45 (7). cation taken when baseline (5-point scale) After remedication violations. PI was moderate to severe. by patient PR = 0 and PI = baseline score for all further time points.

continued 83 Oral ibuprofen and diclofenac in postoperative pain

TABLE 28 contd Studies of ibuprofen versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Cooper, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 184 analysed. None serious reported; no et al., 1989 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 2 lost to patient withdrew as a result. n = 194 Local anaesthetic. Self- PR (standard (n = 61); all measures of efficacy 1 hour. If they follow-up, 2 did not Numbers reporting adverse Age: 16+ years assessed at home at 0, 0.5, 5-point scale) placebo (p < 0.001). 6-hour remedicated earlier require medication, effects (number of effects): 1 hours then hourly for 50% PR? (y/n) (n = 64). TOTPAR: ibuprofen data excluded from 4 missed more than ibuprofen 5/63 (6); placebo 6 hours. Medication taken Global rating 11.32; placebo 4.67. efficacy analysis. 1 evaluation, 1 had 7/64 (7). when baseline PI moderate (5-point scale) No details on how insufficient baseline to severe. by patient data were handled pain, 1 failed to for remedication. complete diary at appropriate time.

Forbes, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 109 analysed. None serious reported; no et al., 1984 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 21 patient withdrew as a result. n = 136 General anaesthetic. Self- PR (standard (n = 28); all measures of efficacy 2 hours. If they did not require Numbers reporting adverse Age: 15+ years assessed at home at 5-point scale) placebo (p < 0.01). 6-hour remedicated earlier medication, 2 took effects (number of effects): 0 hours, then hourly for 50% PR? (y/n) (n = 28). TOTPAR: ibuprofen data excluded from rescue medication ibuprofen 5/28 (6); placebo 12 hours. Medication Global rating 15.79; placebo 3.79. efficacy analysis. instead of trial 3/28 (3). taken when baseline PI (5-point scale) 4-hour TOTPAR: After remedication medication, 2 moderate to severe. by patient ibuprofen 10.75; PR = 0 and PI = remedicated despite Follow-up 5 days post- placebo 2.79. baseline or last having some relief, surgery with research nurse. score (whichever 2 remedicated was greater) for all before 2 hours. further time points.

Forbes, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 206 analysed. None serious reported; no et al., 1990 extraction then multiple oral dose, 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 3 lost to patient withdrew as a result. n = 269 parallel groups. General/ PR (standard (n = 32); all measures of analgesia 2 hours. If they follow-up, 1 lost Numbers reporting adverse Age: 15+ years local anaesthetic (unclear). 5-point scale) placebo (p < 0.05 at least). 6-hour remedicated earlier report card, 22 did effects (number of effects): Self-assessed at home at 50% PR? (y/n) (n = 34). TOTPAR: ibuprofen data excluded from not require medi- ibuprofen 8/43 (9); placebo 0, 1 hours then hourly for Global rating 10.47; placebo 1.88. efficacy analysis. cation, 8 remedi- 0/38 (0). 6 hours. Medication taken (5-point scale) After remedication cated despite having when baseline PI moderate by patient PR = 0 and PI = relief from study to severe. Follow-up 5 days baseline or last medication, 6 remedi- postsurgery. score (whichever cated with only slight was greater) for pain, 13 remedicated all further time at < 2 hours, 7 failed points. to follow instructions, 3 did not complete forms.

Forbes, Third molar RCT,double-blind, single PI (standard Ibuprofen, All ibuprofen treatments Patients allowed to 298 analysed. None serious reported; no et al., 1991 extraction oral dose, parallel groups. 4-point scale) 50 mg significantly superior to remedicate after Exclusions: 33 did patient withdrew as a result. n = 395 Local anaesthetic. 4 hour PR (standard (n = 57); placebo for all measures 2 hours. If they not require Numbers reporting adverse Age: 15+ years caffeine washout prior to 5-point scale) ibuprofen, (p < 0.05 at least). 8-hour remedicated earlier medication, 14 effects (number of effects): start. Self-assessed at 0, 0.5, 50% PR? (y/n) 100 mg TOTPAR: ibuprofen, data excluded from remedicated at ibuprofen, 50 mg, 10/63 1 hours then hourly for Global rating (n = 49); 50 mg, 8.82; ibuprofen, efficacy analysis. < 2 hours, 1 ate (15); ibuprofen, 100 mg, 5/62 8 hours. Medication taken (5-point scale) ibuprofen, 100 mg, 8.46; ibuprofen, After remedication caffeine-containing (6); ibuprofen, 200 mg, 6/60 when baseline PI moderate by patient 200 mg 200 mg, 10.00; placebo PR = 0 and PI = food, 2 medicated (6); placebo 8/61 (8); 100 mg to severe. Follow-up (n = 48); 2.58; 100 mg combination baseline or last for headache, 1 caffeine combination 12/58 5 days postsurgery. placebo 11.29; 200 mg combination score (whichever rated only one side (15); 200 mg caffeine Multicentre (2 sites). (n = 51); 15.58 (6-hour TOTPAR was greater) for of mouth, 1 form combination 8/58 (9). ibuprofen, calculated from mean all further time completed by 100 mg, + hourly scores). points. relative, 3 lacked caffeine, consistency, 22 100 mg evaluated at (n = 49); incorrect time, 3 ibuprofen, incomplete forms. 200 mg, + caffeine, 100 mg (n = 44).

Forbes, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 241 analysed. None serious reported; no et al., 1991 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 7 lost to patient withdrew as a result. n = 288 Local anaesthetic. Self- PR (standard (n = 37); all measures of efficacy 2 hours. If they follow-up, 12 did not Numbers reporting adverse Age: 15+ years assessed at home at 0, 5-point scale) placebo (p < 0.05 at least). 8-hour remedicated earlier require medication, effects (number of effects): 1 hours then hourly for 50% PR? (y/n) (n = 39). TOTPAR: ibuprofen 14.30; data excluded from 4 remedicated with ibuprofen 7/43 (8); placebo 8 hours. Medication taken Global rating placebo 2.59. 6-hour efficacy analysis. some relief, 1 re- 3/47 (3). when baseline PI moderate (5-point scale) TOTPAR (calculated from After remedication medicated with slight to severe. Follow-up by patient mean hourly scores): ibu- PR = 0 and PI = pain, 19 remedicated 5 days postsurgery. profen 10.97; placebo 2.49. baseline or last before 2 hours, 2 score (whichever lacked consistency, 1 was greater) for did not complete all further time form, 1 took only points. part of medication. 84 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 28 contd Studies of ibuprofen versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Forbes, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 280 analysed. None serious reported; no et al., 1992 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions; 3 did not patient withdrew as a result. n = 338 Local anaesthetic. Self- PR (standard (n = 38); all measures of efficacy 2 hours. If they return form, 14 did Numbers reporting adverse Age: 15+ years assessed at home at 0 and 5-point scale) placebo (p < 0.01). 8-hour remedicated earlier not require effects (number of effects): 1 hour, then hourly for 50% PR? (y/n) (n = 38). TOTPAR: ibuprofen 14.82; data was excluded medication, 4 re- ibuprofen 4/45 (8); placebo 8 hours. Medication taken Global rating placebo 2.34. 6-hour from efficacy medicated despite 2/46 (5). when baseline PI moderate (5-point scale) TOTPAR (calculated from analysis.After some relief, 6 to severe. Follow-up 5 days by patient mean hourly scores): ibu- remedication PR remedicated with postsurgery. profen 11.79; placebo 2.06. = 0 and PI = slight pain, 18 baseline or last remedicated before score (whichever 2 hours, 2 lacked was greater) consistency, 2 did for all further not complete form, time points. 2 took only part of medication, 5 took back-up medication.

Frame, Third molar RCT,double-blind, single PI (non-standard Ibuprofen, At 2 and 3 hours Patients allowed to 123 analysed. None serious reported; no et al., 1989 extraction oral dose, parallel groups. 9-point scale) 400 mg ibuprofen significantly remedicate after Exclusions: 9 did not patient withdrew as a result. n = 148 Local anaesthetic. Self- PR (standard (n = 42); superior to placebo. 2 hours. No take medication, 7 Numbers reporting adverse Age: 16+ years assessed at home at 0, 0.5, 5-point scale) placebo 5-hour TOTPAR information lost to follow-up, 1 effects (number of effects): and 1 hours, then hourly 50% PR? (y/n) (n = 38). (calculated from graph): provided on how was asleep so did ibuprofen 2/42 (2); placebo for 5 hours. Medication ibuprofen 12.85; data were handled not complete form, 1/38 (3). taken when baseline PI placebo 7.95. for patients who 1 had complications at least moderate. remedicated. so did not complete form, 7 had slight pain.

Fricke, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 201 analysed. None serious reported; 1 et al., 1993 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 1 took patient in ibuprofen group n = 207 72-hour washout prior to PR (standard (n = 81); all measures after 2 hours.After medication twice, withdrew as a result of Age: 15+ years start. Local anaesthetic. 5-point scale) placebo 30 minutes. 6-hour remedication PR 5 had insufficient vomiting which investigators Self-assessed at home at 0, 50% PR? (y/n) (n = 39). TOTPAR: ibuprofen 10.9; = 0 and PI = pain. did not attribute to 20, 30, 40, and 60 minutes, Global rating placebo 2.9. baseline or last medication. Numbers then hourly for 12 hours. (5-point scale) score (whichever reporting adverse effects Medication taken when by patient was greater) for all (number of effects): baseline PI moderate. VAS PI:‘no pain’ further time points. ibuprofen 8/81 (13); placebo Review 1/2 days after to ‘worst pain 1/39 (1). the trial. imaginable’

Gay, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 194 analysed. None serious reported; no et al., 1996 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 2 patient withdrew as a result. n = 206 12-hour washout prior to PR (standard (n = 41); all summary measures of 1 hour. If they remedicated before Numbers reporting adverse Age range: start. Local anaesthetic. 5-point scale) placebo analgesia (p < 0.05). remedicated 1 hour, 10 failed to effects (number of effects): 18–60 years Self-assessed at ‘regular 50% PR? (y/n) (n = 39). 6-hour TOTPAR: earlier data was complete assess ibuprofen 3/41 (3); placebo intervals’ for 6 hours. Global rating ibuprofen 13.6; excluded from ment within 4/41 (7). Medication taken when (5-point scale) placebo 5.2. efficacy analysis. 15 minutes of baseline PI moderate by patient After remedication scheduled time. to severe. VAS PI:‘no pain’ PR = 0 and PI = to ‘worst pain baseline or last imaginable’ score (whichever was greater) for all further time points.

Heidrich, Orthopaedic RCT,double-blind, single PI (4-point scale) Ibuprofen, Orthogonal analyses of No information No information “There were no differences et al., 1985 surgery oral dose, parallel groups. PR (5-point scale) 400 mg variance showed ibuprofen given on patients given on any among treatments in terms n = 120 4-hour washout prior to VAS PI:‘no relief’ (n = 40); produced greater PR than who remedicated. exclusions. of side effects. No patient Age range: start. Assessed by trained to ‘complete placebo placebo. 6-hour VAS withdrew because of adverse 18–65 years nurse observer at 0, 0.5 relief’ (n = 40). TOTPAR: ibuprofen 234; events.” and 1 hour, then hourly for VAS PI:‘no pain’ placebo 104. 6 hours. Medication taken to ‘worst pain when baseline PI moderate imaginable’ to severe. McGill pain questionnaire

continued 85 Oral ibuprofen and diclofenac in postoperative pain

TABLE 28 contd Studies of ibuprofen versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Hersh, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen at both doses Patients allowed to 254 analysed. None serious reported; no et al., 1993 extraction oral dose, parallel groups. 4-point scale) 400 mg significantly superior to remedicate after No exclusions. patient withdrew as a result. n = 254 4-hour washout prior to PR (standard (n = 49); placebo for all measures 1 hour. If they Numbers reporting adverse Age: 16+ years start. Local anaesthetic. 5-point scale) ibuprofen, of analgesia (p < 0.05). remedicated earlier effects (number of effects): Self-assessed at 0, 0.5, and Global rating 200 mg 6-hour TOTPAR (calcu- data excluded from ibuprofen, 400 mg, 6/49 (7); 1 hour then hourly for (5-point scale) (n = 51); lated from the graph): efficacy analysis. ibuprofen, 200 mg, 4/51 (4); 8 hours (for the first by patient placebo ibuprofen, 400 mg, 9.1; After remedication placebo 9/51 (9). 2 hours in clinic). Medi- (n = 51). ibuprofen, 200 mg, 6.68; PR = 0 and PI = cation taken when baseline placebo 1.18. baseline or last PI moderate to severe. score (whichever was greater) for all further time points.

Hersh, Third molar RCT,double-blind, pre- PI (standard Placebo then Ibuprofen significantly Patients allowed to 81 analysed. No information given. et al., 1993 extraction surgery placebo then single 4-point scale) ibuprofen, superior to placebo for remedicate after Exclusions: 19 lost n = 114 oral dose, parallel groups. PR (standard 400 mg all summary measures of 1 hour.After to follow-up, 11 Age: not specified Local anaesthetic. Self- 5-point scale) (n = 12); analgesia. 6-hour remedication PR did not require assessed at 0, 0.5, and 50% PR? (y/n) placebo then TOTPAR: ibuprofen = 0 and PI = medication, 3 1 hours, then hourly for Global rating placebo 15.67; placebo 9.00. baseline or last excluded for various 6 hours. Medication taken (5-point scale) (n = 16). score (whichever protocol violations. when baseline pain was of by patient was greater) for moderate to severe all further time intensity. points.

Jain, et al., Third molar RCT,double-blind, single PI (4-point scale) Ibuprofen, All ibuprofen doses Patients allowed to 227 analysed. None serious reported; no 1986 extraction oral dose, parallel groups. – standard word- 400 mg significantly superior to remedicate after Exclusions: 10 patient withdrew as a result. n = 260 Self-assessed at home at ing but scale 1–4 (n = 49); placebo (p < 0.001). 1 hour. If they remedicated before Numbers reporting adverse Age range: 0 and 1 hours, then hourly PR (non-standard ibuprofen, 6-hour SPID: ibuprofen, remedicated earlier 1 hour, 19 did not effects (number of effects): 18–65 years for 6 hours. Medication 5-point scale) 200 mg 400 mg, 3.0; ibuprofen, data excluded from take medication or ibuprofen, 400 mg, 10/? (12); taken when baseline PI Global rating (n = 47); 200 mg, 2.26; ibuprofen, efficacy analysis. were lost to ibuprofen, 200 mg, 6/? (8); moderate to severe. (5-point scale) ibuprofen, 100 mg, 1.54; placebo After remedication follow-up, 2 had ibuprofen, 100 mg, 13/? (15); by patient 100 mg –1.73. PR = 0 and PI = mild baseline pain, placebo 12/? (15). VAS PI:‘no pain’ (n = 39); last score for all 1 missed > 2 evalu- to ‘worst ever placebo further time ations, 1 used pain’ (n = 47). points. confounding drugs.

Jain, et al., Episiotomy RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 147 analysed. None serious reported; no 1988 n = 161 oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 11 patient withdrew as a result. Age: 18+ years 4-hour washout prior to PR (standard (n = 49); most summary measures 2 hours. If they remedicated before Numbers reporting adverse start. Assessed by trained 5-point scale) placebo of analgesia (p < 0.01). remedicated earlier 2 hours, 2 received effects (number of effects): nurse observer at 0, 0.5, Time to (n = 48). 6-hour TOTPAR: data excluded from confounding agents, ibuprofen 2/49 (2); placebo and 1 hour, then hourly meaningful relief ibuprofen 14.4; efficacy analysis. 1 was aged under 1/48 (1). for 6 hours. Medication Global rating placebo 8.61. After remedication 18 years. taken when baseline pain (5-point scale) PR = 0 and PI = was of moderate to severe by patient baseline or last intensity. Overall improve- score (whichever ment (7-point was greater) scale) by patient for all further time points.

Kiersch, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 203 analysed. None serious reported; no et al., 1993 extraction oral dose, parallel groups. 4-point scale) 200 mg superior to placebo for remedicate after Exclusions: 2 for patient withdrew as a result. n = 205 72-hour washout prior to PR (standard (n = 81); all summary measures 2 hours. No protocol violations. Numbers reporting adverse Age: 15+ years start. Self-assessed at home 5-point scale) placebo of analgesia (p < 0.001). information given effects (number of effects): at 0, 20, 30, 40, 60 minutes, 50% PR? (y/n) (n = 42). 6-hour TOTPAR: on how data were ibuprofen 16/81 (20); placebo then hourly for 12 hours. Global rating ibuprofen 10.3; placebo then handled. 5/43 (5). Medication taken when (5-point scale) 3.7. baseline PI at least by patient moderate. Review VAS PI:‘no pain’ 1/2 days after the trial. to ‘worst pain imaginable’

86 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 28 contd Studies of ibuprofen versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Laska, Third molar RCT,double-blind, single PI (standard Ibuprofen, All three doses of Patients allowed to 195 analysed. No withdrawals as a result. et al., 1986 extraction oral dose, parallel groups. 4-point scale) 400 mg ibuprofen significantly remedicate after Exclusions: 4 Numbers reporting adverse n = 200 4-hour washout prior to PR (standard (n = ?); superior to placebo for 1 hour. If they remedicated before effects (number of effects): Age: 16+ years start. Self-assessed at 0, 0.5, 5-point scale) ibuprofen, % SPID. 6-hour SPID remedicated earlier 1 hour, 1 vomited ibuprofen, 400 mg, 1/? (1); and 1 hour then hourly Global rating 600 mg (calculated from the PI data excluded from within 5 minutes placebo 3/? (3). for 6 hours. Medication by patient (n = ?); difference graph): efficacy analysis. of taking study taken when baseline PI Blood serum ibuprofen, ibuprofen, 400 mg, 13.4; After remedication medication. moderate to severe. levels 800 mg ibuprofen, 600 mg, 14.1; PR = 0 and PI = (n = ?); ibuprofen, 800 mg, 13.9; baseline or last placebo placebo 5.3. score (whichever (n = ?). was greater) for Assumed all further time equal distri- points. bution of 40 patients per group.

Lavenziana, Postoperative RCT,double-blind, single VAS PI:‘no pain’ Ibuprofen Patients with 61–80 mm Patients allowed to 124 analysed. None reported. et al., 1996 (inguinal hernia) oral dose, parallel groups. to ‘unbearable arginine baseline pain significantly remedicate after Exclusions: 1 patient n = 125 6-hour washout prior to pain’ soluble, superior to placebo 1 hour. Patients for insufficient pain. Age range: start.Assessed in hospital Global rating 400 mg (p < 0.05). Patients with asked to wait until 18–75 years at 0, 15, 30, 45, 60, (5-point scale) (n = 42); > 81 mm baseline pain, pain returned to 90 minutes, 2 hours then by patient placebo NSD. baseline intensity hourly for up to 6 hours. (n = 41). VAS SPID from graph: before remedicating Medication taken when ibuprofen 250; placebo but no information baseline PI moderate 215. given on how data to severe. were then handled.

McQuay, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen at both doses Patients allowed to 161 analysed. None serious reported; no et al., 1996 extraction oral dose, parallel groups. 4-point scale) 400 mg significantly superior to remedicate after Exclusions: 15 no patient withdrew as a result. n = 218 12-hour washout prior to PR (standard (n = 30); placebo for all measures 45 minutes If they pain, 10 concurrent Numbers reporting adverse Age range: start. Local anaesthetic. 5-point scale) ibuprofen, of analgesia. 6-hour remedicated earlier illness, 7 analgesics effects (number of effects): 16–53 years Self-assessed for 6 hours Global rating 200 mg TOTPAR (calculated from data excluded from within 48 hours, 4 ibuprofen, 400 mg, 2/30 (3); (time points not identified). (5-point scale) (n = 31); the graph): ibuprofen, efficacy analysis. withdrew before ibuprofen, 200 mg, 4/31 (4); Medication taken if baseline by patient placebo 400 mg, 9.1; ibuprofen, After remedication study began, 4 did placebo 1/11 (1). PI moderate to severe VAS PI:‘no pain’ (n = 11). 200 mg, 6.68; placebo PR = 0 and PI = not attend, 3 within 2 hours of surgery. – ‘worst pain 1.18. baseline for all previous NSAID imaginable’ further time allergy, 1 possible VAS PR:‘no points. pregnancy, 1 migraine relief’ to ‘com- after surgery, 1 plete relief’ surgery cancelled, 3 Random 8-word remedicated before scale 45 minutes. Mood (VAS) Stopwatch to meaningful relief

Mehlisch, Various oral RCT,double-blind, single PI (4-point scale): Ibuprofen, Ibuprofen significantly Patients allowed to 697 analysed. None serious reported; no et al., 1990 surgery oral dose, parallel groups. standard wording 400 mg superior for most remedicate.After Exclusions: 4 lost to patient withdrew as a result. procedures 6-hour washout prior to scale 1–4 (n = 306); summary measures of remedication PR follow-up, 4 entered Numbers reporting adverse n = 706 start. Self-assessed at 0, PR (non-standard placebo efficacy. 6-hour SPID: = 0 and PI = in trial twice (only effects (number of effects): Age range: 0.5, and 1 hour, then hourly 4-point scale) (n = 85). ibuprofen 5.84; placebo baseline score for first entry analysed ibuprofen 31/310 (?); placebo 18–64 years for up to 6 hours. 0.99. all further time for efficacy but both 12/85 (?). Medication taken when points. included in safety baseline PI moderate analysis), 1 failed to to severe. meet inclusion criteria.

Mehlisch, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 239 analysed. None serious reported; no et al., 1995 extraction oral dose, parallel groups. 4-point scale) 400 mg superior to placebo for remedicate after Exclusion: 1 patient patient withdrew as a result. n = 205 12-hour washout prior to PR (standard (n = 98); all measures of analgesia 1 hour (but had only 1 molar Numbers reporting adverse Age: 15+ years start. Local anaesthetic. 5-point scale) placebo (p < 0.05). 6-hour encouraged to removed and failed effects (number of effects): Self-assessed at 0, 15, 30, Global rating (n = 40). TOTPAR: ibuprofen wait for 4 hours). to complete diary. ibuprofen 12/98 (?); placebo 45, 60, 90 minutes, 2 hours, (5-point scale) 14.39; placebo 2.62. If remedicated 4/40 (?). then hourly for 6 hours. by patient earlier data Medication taken when excluded from baseline PI moderate efficacy analysis. to severe. After remedication PR and PID = 0 for all further time points.

continued 87 Oral ibuprofen and diclofenac in postoperative pain

TABLE 28 contd Studies of ibuprofen versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Nelson, Third molar RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 180 analysed. None serious reported; no et al., 1994 extraction oral dose, parallel groups. 4-point scale) 200 mg superior to placebo for remedicate after Exclusions: 2 patient withdrew as a result. n = 183 12-hour washout prior to PR (standard (n = 75); PR and PI differences 1 hour. If they remedicated before Numbers reporting adverse Age: 15+ years start. Local anaesthetic. 5-point scale) placebo from 30 minutes remedicated earlier 1 hour, 1 did not effects (number of effects): Self-assessed at 0, 15, 30, 50% PR? (y/n) (n = 40). onwards. 6-hour data excluded from record baseline ibuprofen 16/77 (?); placebo 45, 60, 90 minutes, 2 hours, Global rating TOTPAR: ibuprofen efficacy analysis. pain intensity. 11/41 (?). then hourly for 6 hours. (5-point scale) 12.31; placebo 5.56. After remedication Medication taken when by patient PR = 0 and PI = 0 baseline PI moderate for all further to severe. time points.

Pagnoni, Caesarean RCT,double-blind, single VAS PI:‘no Ibuprofen Sum of PID and mean Patients allowed to 92 analysed. None reported. et al., 1996 section oral dose, parallel groups. pain’ to ‘unbear- arginine AUC showed ibuprofen remedicate after No exclusions. n = 92 6-hour washout prior to ablepain’ soluble, significantly superior to 1 hour. If they Age: 18+ years start. General anaesthetic. Global rating 400 mg placebo (p < 0.001); remedicated earlier Assessed in hospital at (5-point scale) (n = 30); mean peak PI difference data excluded from 0, 15, 30, 45, 60, 90 minutes, by patient placebo value was also signifi- efficacy analysis. 2 hours, then hourly for (n = 32). cantlysuperior to After remedication 6 hours. Medication taken placebo (p < 0.05). PI = last recorded when baseline pain was 6-hour VAS SPID from value for all further > 55 mm. graph: ibuprofen time points. 181 mm; placebo 65 mm.

Parker, Tonsillectomy RCT,double-blind, single PI (non-standard Ibuprofen Ibuprofen significantly No information 110 analysed. No details given of those et al., 1986 n = 139 oral dose then multiple 9-point scale) syrup, superior to placebo at given on patients No information occurring during single dose. Age range: doses, parallel groups. PR (standard 600 mg 30 minutes and 1 hour. who remedicated. given on For multiple doses, 1 patient 16–66 years General anaesthetic. 5-point scale) (n = 44); 4-hour TOTPAR: 29 exclusions. in placebo group withdrew as Assessed in hospital at 50% PR? (y/n) placebo ibuprofen 10.92; a result. Numbers reported 0, 0.5, 1 hours, then hourly Observers (n = 33). placebo 9.37. were similar for both groups. for 4 hours. Medication noted their taken when baseline PI impression of moderate to severe. patients’ progress

Schachtel, Episiotomy RCT,double-blind, single PI (standard Ibuprofen, Ibuprofen significantly Patients allowed to 111 analysed. None reported. et al., 1989 n = 115 oral dose then multiple 4-point scale) 400 mg superior to placebo for remedicate after Exclusions: 4 Age range: doses, parallel groups. PR (standard (n = 36); all measures of analgesia 1 hour. If they re- remedicated 16–37 years 4-hour washout prior to 5-point scale) placebo (p < 0.05) at least. medicated earlier before 1 hour. start.Assessed in hospital Global rating (n = 38). 4-hour TOTPAR: data excluded from at 0, 0.5, 1 hours, then (5-point scale) ibuprofen 10.4; efficacy analysis. hourly for 4 hours. by patient placebo 5.5. After remedication Medication taken when PR = 0 and PI = baseline PI moderate last or baseline to severe. (which ever was greater) for all further time points.

Seymour, Third molar RCT,double-blind, single VAS PI: ‘no Study 1: ibu- Study 1: both forms of Patients allowed to 187 analysed. None serious reported; no et al., 1991 extraction oral dose, parallel groups. pain’ to ‘unbear- profen (tablets), ibuprofen significantly remedicate.After Claimed to have patient withdrew as a result. n = 205 General anaesthetic. able pain’ 400 mg (n = 31); superior to placebo; no remedication PI = enrolled only 180? Only 1 patient from the Age: Adult Assessed in hospital by Global rating ibuprofen significant difference last score for all placebo group (Study 1) same observer at 0, 10, (5-point scale) (liquid in gelatin between 2 active groups. further time points. reported an adverse event. 20, 30, 45, 60, 90 minutes, by patient capsules), Study 2: soluble ibu- 2 hours, then hourly for 400 mg (n = 32); profen significantly 6 hours. Medication taken placebo (n = 32). superior to placebo when baseline pain was Study 2: from 20 minutes; tablets > 30 mm. ibuprofen from 30 minutes. 6-hour (tablets), VAS SPID: gelatin 400 mg 233 mm; tablets (n = 30); 243 mm; placebo ibuprofen 120 mm; soluble 228 mm; (soluble), tablets 214 mm; placebo 400 mg 86 mm. (n = 32); placebo (n = 30).

88 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 28 contd Studies of ibuprofen versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Seymour, Third molar RCT,double-blind, single PI (VAS):‘no Ibuprofen All ibuprofen treatments Patients allowed to 199 analysed. Reported by 4 patients: et al., 1996 extraction oral dose, parallel groups. pain’ to ‘unbear- (tablets), 600 mg except 200 mg resulted remedicate.After Exclusions: 4 for 3 took ibuprofen (dose not n = 148 General anaesthetic. able pain’ (n = 17); ibu- in significantly less pain remedication PI = ‘unwanted effects’, specified) and 1 placebo. Age: Adult Assessed in hospital by Global rating profen (soluble), than placebo for all last score for all 25 failed to reach nurse observer at 0, 10, (5-point scale) 600 mg (n = 17); efficacy measures further time sufficient baseline PI. 20, 30, 45, 60, 75, 90,120, by patient ibuprofen (p < 0.05). 6-hour VAS points. 150 minutes, 3 hours, (tablets), 400 mg SPID: ibuprofen, 600 mg then hourly for 6 hours. (n = 15); ibuprofen tablets, 230; ibuprofen, Medication taken when (soluble), 400 mg 600 mg soluble, 148; baseline pain was (n = 16); ibuprofen ibuprofen, 400 mg tablets, > 30 mm. (tablets), 200 mg 258; ibuprofen, 400 mg (n = 18); ibuprofen soluble, 238; ibuprofen, (soluble), 200 mg 200 mg tablets, 140; (n = 17); placebo ibuprofen, 200 mg soluble, (n = 19). 198; placebo 44.

Sunshine, Episiotomy RCT,double-blind, single PI (standard Ibuprofen, 400 mg Ibuprofen significantly Patients allowed to 120 analysed. None reported. et al., 1983 n = 115 oral dose, parallel groups. 4-point scale) (n = 30); placebo superior to placebo for remedicate after No exclusions. Age: 18+ years 4-hour washout prior to PR (non- (n = 30). all measures of analgesia 1 hour. If they start. Assessed in hospital standard from 1 hour onwards. remedicated earlier by same observer at 0, 5-point scale) 4-hour SPID: ibuprofen data excluded from 0.5, 1 hours, then hourly (percentages 6.47; placebo 1.12. efficacy analysis. for 4 hours. Medication not descriptive After remedication taken when baseline PI wording) PR = 0 and PI = moderate to severe. Global rating last for all further of medication time points. (4-point scale) by patient Global rating of personal improvement (7-point scale) by patient

Sunshine, Episiotomy, RCT,double-blind, single PI (standard Ibuprofen, 400 mg All active treatments Patients allowed to 195 analysed. None reported in either et al., 1987 Caesarean oral dose, parallel groups. 4-point scale) (n = 38); placebo significantly superior to remedicate after Exclusions: 1 non- placebo or ibuprofen groups. section or 4-hour washout prior to PR (non- (n = 40). placebo for TOTPAR and 1 hour. If they compliant with gynaecological start. Assessed in hospital standard all except codeine for remedicated earlier washout period, 4 surgery by same observer at 0, 5-point scale) SPID. 4-hour SPID: data excluded from did not complete n = 200 0.5, 1 hours, then hourly (percentages ibuprofen 8.1; placebo efficacy analysis. evaluations. Age: not for 4 hours. Medication not descriptive 5.2. After remedication specified taken when baseline PI wording) PR = 0 and PI = moderate to severe. Global rating last for all further of medication time points. (4-point scale) by patient Global rating of personal improvement (7-point scale) by patient

Each point in Figure 28 represents one trial with the NNT for a single dose of ibuprofen, 400 mg, proportion of patients achieving at least 50% pain standard formulation tablets (1356 patients) relief on the study drug plotted on the y axis, and compared with placebo was 2.8 (95% CI, 2.5–3.1) the proportion of patients achieving the same end and for ibuprofen, 400 mg, soluble formulations point with placebo on the x axis. (250 patients) the NNT was 2.5 (95% CI, 2.1–3.1). All formulations were therefore pooled for the In one trial,19 a syrup formulation of ibuprofen overall analysis. was used, in two trials20,21 soluble ibuprofen and liquid in gelatin capsules were used, in two trials22,23 The single data set for ibuprofen, 50 mg, showed ibuprofen lysine was used, and in two trials24,25 no significant difference from placebo. The pooled soluble ibuprofen arginine was used. When the relative benefits for ibuprofen, 100 mg, 200 mg, results from these more readily absorbed formu- 400 mg and 600 mg, were significantly different lations were pooled and compared with those for from placebo, as was the single data set for ibupro- the standard tablet formulation, no differences fen, 800 mg (Table 29). At a dose of 50 mg, ibupro- were found in the relative benefit or NNTs. The fen had an NNT of 3.6 (95% CI, 2.5–6.1) for at least 89 Oral ibuprofen and diclofenac in postoperative pain

Number of patients At least 50% PR with ibuprofen (%) given active treatment 100

Ibuprofen, 100 mg 88 80 Ibuprofen, 200 mg 406

60 Ibuprofen, 400 mg 1606

Ibuprofen, 600 mg 114 40 Ibuprofen, 800 mg 39

20 12345678910 NNT (95% CIs) 0 0 25 50 75 100 FIGURE 29 Dose response for ibuprofen trials At least 50% PR with placebo (%)

At least 50% PR with diclofenac (%) FIGURE 28 Ibuprofen against placebo in postoperative pain 100 (◆, ibuprofen, 50 mg; ●, ibuprofen, 100 mg; ▲, ibuprofen, 200 mg; ■, ibuprofen, 400 mg; ▼, ibuprofen, 600 mg) 80 50% pain relief over 4–6 hours compared with placebo in pain of moderate to severe intensity; at 100 mg, the NNT was 5.6 (95% CI, 3.8–9.9), at 60 200 mg, the NNT was 3.3 (95% CI, 2.8–4.0), at 400 mg, the NNT was 2.7 (2.5–3.0), at 600 mg, the 40 NNT was 2.4 (1.9–3.3) and, at 800 mg, the NNT was 1.6 (1.3–2.2). The dose response for ibuprofen is presented in Figure 29. 20

When a fixed CER of 19% was applied, there was a clear dose response with no overlap in CIs 0 except for the 600 mg and 800 mg doses (Table 29; 0204060 80 100 see also Figure 32). At least 50% PR with placebo (%) Drug-related study withdrawals occurred rarely. In one study,26 one withdrawal on ibuprofen was for FIGURE 30 Diclofenac against placebo in postoperative pain vomiting which the authors did not attribute to the (◆, diclofenac, 25 mg; ●, diclofenac, 50 mg; , diclofenac, 100 mg)

TABLE 29 Summary of relative benefit and NNTs for trials of ibuprofen versus placebo

Number Dose of Number of Number of RB – NNT NNT with of trials ibuprofen patients with patients with random (95% CI) 19% CER (mg) > 50% PR: > 50% PR: effects model (95% CI) ibuprofen placebo (95% CI)

1 50 16/57 0/51 144 (0.3– > 1000) 3.6 (2.5–6.1) 12.5 (4.1–∞)

2 100 16/88 0/98 72 (16–318) 5.6 (3.8–9.9) –100 (10–∞)

8 200 151/406 22/320 3.5 (2.3–5.3) 3.3 (2.8–4.0) 5.6 (4.1–8.5)

30 400 858/1606 214/1292 3.3 (2.5–4.3) 2.7 (2.5–3.0) 2.9 (2.7–3.2)

3 600 90/114 40/108 2.5 (1.2–5.5) 2.4 (1.9–3.3) 1.7 (1.4–2)

90 1 800 39/39 14/37 2.6 (1.8–4) 1.6 (1.3–2.2) 1.2 (1.1–1.5) Health Technology Assessment 1998; Vol. 2: No. 12

medication. In one study,21 there were three with- an NNT of 2.6 (95% CI, 1.9–4.5) for at least 50% drawals on ibuprofen and one on placebo for vomit- pain relief over 4–6 hours compared with placebo ing soon after ingestion of the study drug. In another in pain of moderate to severe intensity. The NNT study,19 one patient withdrew on placebo. The studies at 50 mg was 2.3 (95% CI, 2.0–2.7) and at 100 mg reported a variable incidence of minor adverse events 1.8 (95% CI, 1.5– 2.1), with overlapping CIs. which were all mild and transient, with no difference The dose response for diclofenac is shown in in incidence between ibuprofen and placebo. Figure 31.

Diclofenac versus placebo With a fixed value of CER of 19%, the NNT for Although the search identified nearly 2000 trials, diclofenac, 25 mg, was 3.9 (95% CI, 2.3–12), for the majority were in chronic pain or the drug was 50 mg 2.3 (95% CI, 2.0–2.7) and for 100 mg 2.2 administered before the patient experienced pain. (95% CI, 1.8–2.8), with overlapping CIs (Table 29; Over 500 reports were found of trials involving Figure 32). rectal, intravenous and intramuscular diclofenac. Predominantly, the reports were not in established Drug-related study withdrawals rarely occurred. postoperative pain, were not placebo-controlled or In one study,33 there was one withdrawal on diclo- did not use standard pain outcome measures. Only fenac, 100 mg, for nausea and vomiting. The six trials fulfilled the inclusion criteria; all were for studies reported a variable incidence of minor oral diclofenac (528 patients were given diclofenac adverse events none of which were serious and and 312 placebo). Five reports identified by the there was no difference in incidence between search could not be obtained despite attempts to diclofenac and placebo. contact the authors, ordering through the British Library and help from the librarians at Novartis Diclofenac versus ibuprofen and Knoll Pharmaceuticals.27–31 Details of the six There were two direct comparisons of diclofenac, included studies are presented in Table 30, with 50 mg, and ibuprofen, 400 mg.35,36 Both trials were their references being given at the end of this in dental pain (third molar removal); 118 patients chapter. The authors of one study were contacted received diclofenac and 112 patients ibuprofen. for information regarding the number of There was no significant difference between patients in each treatment arm;32 however, they diclofenac, 50 mg, and ibuprofen, 400 mg were unable to provide this information and (relative benefit 1.0; 95% CI, 0.9–1.2). so an equal split of 50 patients per group was assumed. Comment Five trials investigated oral surgery pain (third molar extraction with bone removal) and one A single dose of ibuprofen, 400 mg, had an NNT pain following gynaecological surgery. The doses of 2.7 for at least 50% pain relief compared with of diclofenac prescribed were 25 mg in one trial, placebo. This means that one from every three 50 mg in six trials and 100 mg in three trials. In patients with pain of moderate to severe intensity three trials the immediate release diclofenac will experience at least 50% pain relief with potassium formulation was used,32–34 and in two ibuprofen which they would not have had with trials dispersible diclofenac was used.35,36 In one placebo. The equivalent NNT for a single dose trial, both the immediate release and enteric- of ibuprofen, 600 mg, was 2.4 and for ibuprofen, coated formulations were used.37 To ensure 200 mg, 3.3, showing a dose response although comparability, only data from the immediate the CIs overlapped (Table 29). When a fixed CER release formulation were included. was used to smooth out the CER variations of individual trials, the CIs for the NNTs did not CER values ranged from 8% to 38% (median 10%) overlap, supporting the dose–response finding (Figure 30). The EER for the single trial of diclo- (Figure 32). Moreover, the use of a fixed (popu- fenac, 25 mg, was 46%. The EER for diclofenac, lation) CER had little effect on the NNT in circum- 50 mg, varied between 53% and 75% (median stances where there were either large numbers of 58%) and for diclofenac, 100 mg, between 56% patients or where there were large effects (Tables 29 and 72% (median 67%). The 100 mg data set was and 31). Only in small trials with limited analgesic homogeneous but the 50 mg data set was not. efficacy (low doses) did the use of the fixed CER alter the NNT significantly. The pooled relative benefits for all doses of diclofenac versus placebo were significant A single dose of diclofenac, 50 mg, had an NNT (Table 31). At a dose of 25 mg, diclofenac had of 2.3 for at least 50% pain relief compared with 91 Oral ibuprofen and diclofenac in postoperative pain

TABLE 30 Studies of diclofenac versus placebo

Study Condition Design, study Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Ahlstrom, Third molar RCT,double-blind, single VAS PI:‘no Diclofenac Diclofenac significantly Patients allowed to 97 analysed. None serious reported; no et al., 1993 extraction oral dose, parallel groups. pain at all’ to (drinkable), superior to placebo by remedicate after Exclusions: 30 for patient withdrew as a result. n = 127 4-hour washout prior to ‘agonising pain’ 50 mg 40 minutes (p = 0.001); 1 hour. After various protocol Numbers reporting adverse Age range: start. Evaluated at 0, 20, Global rating (n = 35); this continued for 6 hours. remedication violations. effects (number of effects): 18 – 40 years 40 minutes, 1 hour then by patient placebo TOTPI and SPID: diclo- PI = last score was diclofenac 6/35 (?); placebo hourly for 6 hours. Medi- (n = 30). fenac significantly super- carried forward 2/30 (?). cation taken when baseline ior to placebo (p < 0.0001). for all further pain was at least moderate 6-hour SPID: diclofenac time points. intensity (> 30 mm). 173 mm; placebo 32 mm.

Bakshi, Third molar RCT,double-blind, single PI (standard Diclofenac Diclofenac potassium Patients allowed to 151 analysed. None serious reported; no et al., 1992 extraction oral dose, parallel groups. 4-point scale) potassium significantly superior to remedicate after Exclusions: 26 patient withdrew as a result. n = 180 4-hour washout prior to PR (non-standard (sugar-coated), placebo for SPID, 1 hour. If they re- did not require Numbers reporting adverse Age: adult start. Self-assessed at 0, 15, 4-point scale) 50 mg (n = 51); TOTPAR, MAXPID and medicated earlier medication, 3 lost effects (number of effects): 30 minutes, 1 hour, then 50% PR? (y/n) diclofenac MAXPAR (p < 0.001). data excluded from to follow-up. diclofenac potassium 3/51 hourly for 6 hours. Medi- Global rating sodium Diclofenac sodium efficacy analysis. (5); diclofenac sodium 1/54 cation taken when baseline (4-point scale) (enteric- significantly superior to After remedication (1); placebo 3/46 (3). pain was of at least by patient coated), placebo for SPID PR = 0 and PI = moderate intensity. 50 mg (n = 54); (p = 0.023) and MAXPID last score or placebo (p = 0.018). 6-hour SPID: baseline (whichever (n = 46). diclofenac potassium 7.8; was greater) for diclofenac sodium 5.3; all further placebo 2.6. time points.

Bakshi, Third molar RCT,double-blind, single VAS PI: ‘no Diclofenac Diclofenac significantly Patients allowed to 245 analysed. None serious reported; no et al., 1994 extraction oral dose, parallel groups. pain’ to ‘pain (dispersible), superior to placebo for remedicate after Exclusions: 9 did patient withdrew as a result. n = 257 Local anaesthetic. could not 50 mg TOTPAR and both global 1 hour. If they not experience Numbers reporting adverse Age range: Self-assessed at 0, 20, be worse’ (n = 83); ratings (p < 0.01). remedicated earlier severe pain, 2 effects (number of effects): adults up to 40 minutes, 1, 1.5, PR (5-point placebo 6-hour TOTPAR: data excluded from remedicated diclofenac 4/83 (?); placebo 65 years 2 hours, then hourly for scale): none, (n = 82). diclofenac 15.45; placebo efficacy analysis. before 1 hour, 1 5/82 (?). 6 hours. Medication taken poor, moderate, 8.85. After remedication completed diary when baseline pain was sufficient, total PR = 0 and PI = incorrectly. of at least severe intensity. Global rating last score for all (non-standard further time points. 5-point scale) by patient and by observer

Hebertson, Gynaecological RCT,double-blind, single PI (standard Diclofenac, Both diclofenac doses Patients allowed to 209 analysed for All were gastrointestinal et al., 1994 surgery oral dose, parallel groups. 4-point scale) 50 mg significantly superior to remedicate after at least 1 efficacy except one in placebo group n = 217 4-hour washout prior to PR (standard (n = 52); placebo for PR at each 1 hour. If they analysis. 194 (not defined); 1 withdrew Age: 16+ years start.Assessed by observer 5-point scale) diclofenac, time point from 1 hour remedicated earlier analysed for from diclofenac, 100 mg, for at 0, 0.5, 1 hour, then Global rating 100 mg onwards.6-hour TOTPAR data excluded from 8-hour SPID and nausea and vomiting. hourly for 8 hours. Medi- (5-point scale) (n = 52); from graph: diclofenac, efficacy analysis. TOTPAR. No Numbers reporting adverse cation taken when baseline by patient placebo 50 mg, 12.1; diclofenac, After remedication information given effects (number of effects): pain was of moderate to (n = 52). 100 mg, 12.2; placebo patients discon- on any exclusions. diclofenac, 50 mg, 3/54 (?); severe intensity and groups 3.7. tinued in study; no diclofenac, 100 mg, 2/55 (?); stratified by baseline PI. information given on placebo 2/54 (?). how their data were then handled.

Mehlisch, Third molar RCT,double-blind, single PI (standard Diclofenac, Both diclofenac groups Patients allowed to There were no None serious reported; no et al., 1994 extraction oral dose, parallel groups. 4-point scale) 50 mg significantly superior to remedicate after exclusions. patient withdrew as a result n = 208 Local anaesthetic. 4-hour PR (standard (n = 53); placebo for all time 2 hours.After of adverse events. Numbers Age range: washout prior to start. Self- 5-point scale) diclofenac, points with regard to PR. remedication PR reporting adverse effects 16–70 years assessed at 0, 0.5, 1 hour, Global rating 100 mg 6-hour TOTPAR from = 0 and PI = last (number of effects): then hourly for up to 8 hours. (5-point scale) (n = 52); graph: diclofenac, 50 mg, score for all further diclofenac, 50 mg, 2/53 (?); Medication taken when base- by patient placebo 11.6; diclofenac, 100 mg, time points. diclofenac, 100 mg, 2/52 (?); line pain was of moderate (n = 52). 14.3; placebo 3.3. placebo 2/52 (?). to severe intensity.

Nelson, Third molar RCT,double-blind, single PI (standard Diclofenac, All diclofenac groups Patients allowed to 252 analysed for None serious reported; no et al., 1994 extraction oral dose, parallel groups. 4-point scale) 25 mg (n = ?); significantly superior to remedicate at at least 1 efficacy patient withdrew as a result n = 255 Local anaesthetic. 4-hour PR (standard diclofenac, placebo for PR from 1 hour, no analysis. No of adverse events. Numbers Age range: washout prior to start. 5-point scale) 50 mg (n = ?); 1 hour onwards. 6-hour information given exclusions given reporting adverse effects 16–70 years Self-assessed at 0, 0.5, diclofenac, TOTPAR from graph: as to how their data other than (number of effects): 1 hour then hourly for 100 mg (n = ?); diclofenac, 25 mg, 10.5; were then handled. remedicators. diclofenac, 25 mg, 5 (?); up to 8 hours. Medication placebo (n = ?). diclofenac, 50 mg, 12.2; diclofenac, 50 mg, 4 (?); taken when baseline pain NB: assumed diclofenac, 100 mg, 14.8; diclofenac, 100 mg, 4 (?); was of moderate to equal distri- placebo 3.6. placebo 6 (?). severe intensity. bution of 51 per group (including 92 aspirin group). Health Technology Assessment 1998; Vol. 2: No. 12

placebo. The equivalent NNT for diclofenac, 100 mg, was 1.8 and for diclofenac, 25 mg, 2.6, indicating a dose response although the CIs Number of patients overlapped (Table 31). With a fixed CER of 19%, given active treatment the CIs for the two higher doses overlapped completely (Table 31; Figure 32). Diclofenac, 25 mg 50 Diclofenac is widely regarded as a more effective Diclofenac, 50 mg 324 NSAID than ibuprofen. These results do not support this conclusion. When diclofenac, 50 mg, Diclofenac, 100 mg 154 and ibuprofen, 400 mg, were compared directly 12345678910 there was no significant difference between them. NNT (95% CIs) When compared with placebo, diclofenac, 50 mg, and ibuprofen, 600 mg, had very similar NNTs with complete overlap of CIs. Single trials of NSAIDs have often reported flat dose–response curves, FIGURE 31 Dose response for diclofenac trials typified by that for diclofenac presented in

TABLE 31 Summary of relative benefit and NNTs for trials of diclofenac versus placebo

Number Dose of Number of Number of RB – NNT NNT of trials diclofenac patients with patients with random effects (95% CI) with 19% CER (mg) > 50% PR: > 50% PR: model (95% CI) diclofenac placebo (95% CI)

1 25 23/50 4/50 5.8 (2.1–15.4) 2.6 (1.9–4.5) 3.9 (2.3–12.1)

6 50 203/324 57/312 4.3 (2.4–7.8) 2.3 (2–2.7) 2.3 (2–2.7)

3 100 100/154 13/154 7.2 (5.5–9.4) 1.8 (1.5–2.1) 2.2 (1.8–2.8)

NNT 1 39

114 2 324 154 207

1606 3

50 1376 4

616 5

406 6 10 100 1000 10,000 Oral dose (mg)

FIGURE 32 Dose response for diclofenac (◆), ibuprofen (●) and paracetamol (■) using a fixed 19% placebo response rate. Numbers indicate patients given active treatments 93 Oral ibuprofen and diclofenac in postoperative pain

Figure 32. With the advantage of the much larger 11. DerSimonian R, Laird N. Meta-analysis of clinical numbers of patients in this meta-analysis, the ‘true’ trials. Control Clin Trials 1986;7:177–88. dose response for ibuprofen is shown (Figure 32), 12. Cook RJ, Sackett DL. The number needed to treat: with the interesting finding that higher doses a clinically useful measure of treatment effect. (> 1 g) of paracetamol also follow traditional BMJ 1995;310:452–4. dose–response curve contours. 13. Moore RA, Collins S, McQuay HJ. Variation in the placebo effect; the impact on individual trials and The issue of the relative efficacy of the two drugs consequences for meta-analysis. In: Jensen TS, therefore comes down to dose (Figures 29, 31 and Hammond DL, Jensen TS, editors. Proceedings 8th 32) and safety.38 Ibuprofen, 400 mg, is one-sixth of World Congress on Pain. Progress in pain research the maximum daily dose. Diclofenac, 50 mg, is one- and management, vol. 2. Seattle: IASP Press, 1997. third of the maximum daily dose. This may explain 14. Moore A, Collins S, Carroll D, McQuay H. prescriber confusion. Paracetamol with and without codeine in acute pain: a quantitative systematic review. Pain 1997;70:193–201. References 15. Moore RA, McQuay HJ. Single-patient data meta- analysis of 3453 postoperative patients: oral 1. Collins SL, Moore A, McQuay HJ, Wiffen PJ. Oral tramadol versus placebo, codeine and combination ibuprofen and diclofenac in postoperative pain: a analgesics. Pain 1997;69:287–94. quantitative systematic review. Eur J Pain. In press. 16. Collins SL, Edwards J, Moore A, McQuay HJ. 2. Government Statistical Service. Prescription cost Single dose oral dextropropoxyphene in post- analysis for England 1996. London: Department operative pain: a quantitative systematic review. of Health, 1997. Eur J Clin Pharmacol 1998;54:107–12.

3. Jadad AR, Carroll D, Moore A, McQuay HJ. 17. Edwards JE, McQuay HJ, Moore RA. Systematic Developing a database of published reports of review: dihydrocodeine in postoperative pain. randomised clinical trials in pain research. Cochrane Library. In press. Pain 1996;66:239–46. 18. Laska EM, Sunshine A, Marrero I, Olson N, Siegel C, McCormick N. The correlation between blood 4. Reynolds JEF. Martindale: the extra pharmacopoeia. levels of ibuprofen and clinical analgesic response. 30th edition. London: Pharmaceutical Press, Clin Pharmacol Ther 1986;40:1–7. 1993:292–314. 19. Parker D, Gibbin K, Noyelle R. Syrup formulations 5. Collins SL, Moore RA, McQuay HJ. The visual for post tonsillectomy analgesia: a double blind analogue pain intensity scale: what is moderate study comparing ibuprofen, aspirin and placebo. pain in millimetres? Pain 1997;72:95–7. J Laryngol Otol 1986;100:1055–60. 20. Seymour RA, Hawkesford JE, Weldon M, Brewster 6. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in D. An evaluation of different ibuprofen prepara- clinical research. Ann Intern Med 1987;107:224–33. tions in the control of postoperative pain after third 7. Cooper SA. Single-dose analgesic studies: the upside molar surgery. Br J Clin Pharmacol 1991;31:83–7. and downside of assay sensitivity. In: Max MB, 21. Seymour R, WardBooth P, Kelly P. Evaluation of Portenoy RK, Laska EM, editors. The design of different doses of soluble ibuprofen and ibuprofen analgesic clinical trials. Adv Pain Res Ther tablets in postoperative dental pain. Br J Oral 1991;18:117–24. Maxillofac Surg 1996;34:110–14.

8. Moore A, McQuay H, Gavaghan D. Deriving 22. Mehlisch DR, Jasper RD, Brown P, Korn SH, dichotomous outcome measures from continuous McCarroll K, Murakami AA. Comparative study of data in randomised controlled trials of analgesics. ibuprofen lysine and acetaminophen in patients Pain 1996;66:229–37. with postoperative dental pain. Clin Ther 1995;17:852–60. 9. Moore A, McQuay H, Gavaghan D. Deriving 23. Nelson SL, Brahim JS, Korn SH, Greene SS, dichotomous outcome measures from continuous Suchower LJ. Comparison of single-dose ibuprofen data in randomised controlled trials of analgesics: lysine, acetylsalicylic acid, and placebo for verification from independent data. Pain moderate-to-severe postoperative dental pain. 1997;69:127–30. Clin Ther 1994;16:458–65. 10. Moore A, Moore O, McQuay H, Gavaghan D. 24. Laveneziana D, Riva A, Bonazzi M, Cipolla M, Deriving dichotomous outcome measures from Migliavacca S. Comparative efficacy of oral ibu- continuous data in randomised controlled trials of profen arginine and intramuscular ketorolac in analgesics: use of pain intensity and visual analogue patients with postoperative pain. Clin Drug Invest 94 scales. Pain 1997;69:311–15. 1996;11:8–14. Health Technology Assessment 1998; Vol. 2: No. 12

25. Pagnoni B, Ravanelli A, Degradi L, Rossi R, Tiengo 37. Bakshi R, Jacobs L, Lehnert S, Picha B, Reuther J. M. Clinical efficacy of ibuprofen arginine in the A double-blind, placebo-controlled trial comparing management of postoperative pain associated with the analgesic efficacy of two formulations of diclo- suction termination of pregnancy. A double-blind fenac in postoperative dental pain. Curr Ther Res placebo-controlled study. Clin Drug Invest Clin Exp 1992;52:435–42. 1996;11:27–32. 38. Henry D, Lim LL, Rodriguez LAG, Gutthann SP, 26. Fricke J, Halladay S, Francisco C. Efficacy and safety Carson JL, Griffin M, et al. Variability in risk of of naproxen sodium and ibuprofen for pain relief gastrointestinal complications with individual non- after oral surgery. Curr Therapeut Res Clin Exp steroidal anti-inflammatory drugs: results of a 1993;54:619–27. collaborative meta-analysis. BMJ 1996;312:1563–6. 27. Carlos D. Comparative study of the efficacy of diclofenac Na, meperidine HCl and nalbuphine Studies included HCl in post-operative analgesia. Philip J Intern Med Ahlström U, Bakshi R, Nilsson P, Wåhlander L, 1993. The 1984;22:51–5. analgesic efficacy of diclofenac dispersible and ibuprofen in postoperative pain after dental extraction. Eur J Clin 28. Frezza R, Bolognesi P, Bernardi F. Confronto Pharmacol;44:587–8. sull’azione di tre antinfiammatori non steroidei nel controllo de dolore odontoiatrico post- Arnold JD, 1990. Ketoprofen, ibuprofen, and placebo in chirurgico. Efficaaci i FANS contro il dolore the relief of postoperative pain. Adv Ther;7:264–75. [Comparison of the action of 3 non-steroidal anti- Bakshi R, Jacobs LD, Lehnert S, Picha B, Reuther J, 1992. inflammatory agents in the control of post-operative A double-blind, placebo-controlled trial comparing the pain. Effectiveness of NSAID against pain]. analgesic efficacy of two formulations of diclofenac in Attual Dent 1985;1:40–2. postoperative dental pain. Curr Ther Res;52:435–42. 29. Iqbal K, Biswas G, Mondo S, Afzalunnessa B. Bakshi R, Frenkel G, Dietlein G, Meurer Witt, B, Assessment of post operative analgesia: a com- Schneider B, Sinterhauf U, 1994. A placebo-controlled parative study of pethidine and diclofenac sodium. comparative evaluation of diclofenac dispersible versus J Bangladesh Coll Physicians Surg 1986;4:1–7. ibuprofen in postoperative pain after third molar 30. Joubert JJ. An assesment of the efficacy and surgery. J Clin Pharmacol;34:225–30. tolerability of Voltaren in the treatment of inflam- Cooper SA, Needle SE, Kruger GO, 1977. Comparative mation after extraction of teeth. J Dent Assoc S Africa analgesic potency of aspirin and ibuprofen, J Oral 1977;32:581–3. Surg;35:898–903. 31. Vigneron JR, Thys R. Etude de l’action anti- Cooper SA, Engel J, Ladov M, Precheur H, Rosenheck A, inflammatoire et antalgique du diclofenac en Rauch D, 1982. Analgesic efficacy of an ibuprofen traumatologie et chirurgie orthopedique [Study codeine combination. Pharmacotherapy;2:162–7. of the anti-inflammatory and analgesic actions of diclofenac in traumatology and orthopedic Cooper SA, Berrie R, Cohn P, 1988. Comparison of surgery]. Rev Med Liege 1977;32:10–14. ketoprofen, ibuprofen, and placebo in a dental surgery pain model. Adv Ther;5:43–53. 32. Nelson S, Brahim J. An evaluation of the analgesic efficacy of diclofenac potassium, aspirin, and Cooper SA, Schachtel BP, Goldman E, Gelb S, Cohn P, placebo in postoperative dental pain. Today’s Ther 1989. Ibuprofen and acetaminophen in the relief of Trends 1994;12:3–14. acute pain: a randomized, double blind, placebo controlled study. J Clin Pharmacol;29:1026–30. 33. Hebertson R, Storey N. The comparative efficacy of diclofenac potassium, aspirin, and placebo in the Forbes JA, Barkaszi BA, Ragland RN, Hankle JJ, treatment of patients with pain following gyneco- 1984. Analgesic effect of fendosal, ibuprofen and logic surgery. Today’s Ther Trends 1994;12:33–45. aspirin in postoperative oral surgery pain. Pharmacotherapy;4:385–91. 34. Mehlisch D, Brown P. Single-dose therapy with diclofenac potassium, aspirin, or placebo following Forbes JA, Kehm CJ, Grodin CD, Beaver WT, 1990. dental impaction surgery. Today’s Ther Trends Evaluation of ketorolac, ibuprofen, acetaminophen, and 1994;12:15–31. an acetaminophen codeine combination in postoperative oral surgery pain. Pharmacotherapy;10:94–105S. 35. Ahlström U, Bakshi R, Nilsson P, Wåhlander L. The analgesic efficacy of diclofenac dispersible Forbes JA, Beaver WT, Jones KF, Kehm CJ, Smith WK, and ibuprofen in postoperative pain after dental Gongloff CM, et al., 1991. Effect of caffeine on ibuprofen extraction. Eur J Clin Pharmacol 1993;44:587–8. analgesia in postoperative oral surgery pain. Clin Pharmacol Ther;49:674–84. 36. Bakshi R, Frenkel G, Dietlein G, Meurer Witt B, Schneider B, Sinterhauf U. A placebo-controlled Forbes JA, Edquist IA, Smith FG, Schwartz MK, Beaver comparative evaluation of diclofenac dispersible WT, 1991. Evaluation of bromfenac, aspirin, and versus ibuprofen in postoperative pain after third ibuprofen in postoperative oral surgery pain. molar surgery. J Clin Pharmacol 1994;34:225–30. Pharmacotherapy;11:64–70. 95 Oral ibuprofen and diclofenac in postoperative pain

Forbes JA, Beaver WT, Jones KF, Edquist IA, Gongloff McQuay HJ, Angell K, Carroll D, Moore RA, Juniper RP, CM, Smith WK, et al., 1992 Analgesic efficacy of 1996. Ibuprofen compared with ibuprofen plus caffeine bromfenac, ibuprofen, and aspirin in postoperative after third molar surgery. Pain;66:247–51. oral surgery pain. Clin Pharmacol Ther;51:343–52. Mehlisch DR, Brown P, 1994. Single-dose therapy with Frame JW, Evans CR, Flaum GR, Langford R, Rout PG, diclofenac potassium, aspirin, or placebo following 1989. A comparison of ibuprofen and dihydrocodeine in dental impaction surgery. Today’s Ther Trends;12:15–31. relieving pain following wisdom teeth removal. Br Dent Mehlisch DR, Sollecito WA, Helfrick JF, Leibold DG, J;166:121–4. Markowitz R, Schow CE, et al., 1990. Multicenter clinical Fricke JR, Halladay SC, Francisco CA, 1993. Efficacy and trial of ibuprofen and acetaminophen in the treatment safety of naproxen sodium and ibuprofen for pain relief of postoperative dental pain. J Am Dent Assoc;121:257–63. after oral surgery. Curr Ther Res;54:619–27. Mehlisch DR, Jasper RD, Brown P, Korn SH, McCarroll K, Gay C, Planas E, Donado M, Martinez JM, Artigas R, Murakami AA, 1995. Comparative study of ibuprofen Torres F, et al., 1996. Analgesic efficacy of low doses of lysine and acetaminophen in patients with postoperative dexketoprofen in the dental pain model: a randomised, dental pain. Clin Ther;17:852–60. double-blind, placebo-controlled study. Clin Drug Nelson S, Brahim J, 1994. An evaluation of the anal- Invest;11:320–30. gesic efficacy of diclofenac potassium, aspirin, and Hebertson RM, Storey N, 1994. The comparative efficacy placebo in postoperative dental pain. Today’s Ther of diclofenac potassium, aspirin, and placebo in the Trends;12:3–14. treatment of patients with pain following gynecologic Nelson SL, Brahim JS, Korn SH, Greene SS, Suchower surgery. Today’s Ther Trends;12:33–45. LJ, 1994. Comparison of single-dose ibuprofen lysine, Heidrich G, Slavic Svircev V, Kaiko RF, 1985. Efficacy and acetylsalicylic acid, and placebo for moderate-to-severe quality of ibuprofen and acetaminophen plus codeine postoperative dental pain. Clin Ther;16:458–65. analgesia. Pain;22:385–97. Pagnoni B, Ravanelli A, Degradi L, Rossi R, Tiengo M, Hersh EV, Cooper S, Betts N, Wedell D, MacAfee K, 1996. Clinical efficacy of ibuprofen arginine in the Quinn P, et al., 1993. Single dose and multidose analgesic management of postoperative pain associated with study of ibuprofen and meclofenamate sodium after suction termination of pregnancy. A double-blind third molar surgery. Oral Surg Oral Med Oral placebo-controlled study. Clin Drug Invest;11:27–32. Pathol;76:680–7. Parker DA, Gibbin KP, Noyelle RM, 1986. Syrup Hersh EV, Ochs H, Quinn P, MacAfee K, Cooper SA, formulations for post tonsillectomy analgesia: a double 1993. Narcotic receptor blockade and its effect on the blind study comparing ibuprofen, aspirin and placebo. analgesic response to placebo and ibuprofen after oral J Laryngol Otol;100:1055–60. surgery. Oral Surg Oral Med Oral Pathol;75:539–46. Schachtel BP, Thoden WR, Baybutt RI, 1989. Ibuprofen Jain AK, Ryan JR, McMahon FG, Kuebel JO, Walters PJ, and acetaminophen in the relief of postpartum Noveck C, 1986. Analgesic efficacy of low dose ibuprofen episiotomy pain. J Clin Pharmacol;29:550–3. in dental extraction pain. Pharmacotherapy;6:318–22. Seymour RA, Hawkesford JE, Weldon M, Brewster D, Jain AK, Mcmahon FG, Ryan JR, Narcisse C, 1988. A 1991. An evaluation of different ibuprofen preparations double-blind study of ibuprofen 200 mg in combination in the control of postoperative pain after third molar with caffeine 100 mg, ibuprofen 400 mg, and placebo in surgery. Br J Clin Pharmacol;31:83–7. episiotomy pain. Curr Ther Res;43:762–79. Seymour RA, WardBooth P, Kelly PJ, 1996. Evaluation Kiersch TA, Halladay SC, Koschik M, 1993. A double- of different doses of soluble ibuprofen and ibuprofen blind, randomized study of naproxen sodium, ibuprofen, tablets in postoperative dental pain. Br J Oral Maxil and placebo in postoperative dental pain. Clin Surg;34:110–14. Ther;15:845–54. Sunshine A, Olson NZ, Laska EM, Zighelboim I, de Laska EM, Sunshine A, Marrero I, Olson N, Siegel C, Castro A, de Sarrazin C, 1983. Ibuprofen, zomepirac, McCormick N, 1986. The correlation between blood aspirin, and placebo in the relief of postepisiotomy pain. levels of ibuprofen and clinical analgesic response. Clin Pharmacol Ther;34:254–8. Clin Pharmacol Ther;40:1–7. Sunshine A, Roure C, Olson N, Laska EM, Zorrilla C, Laveneziana D, Riva A, Bonazzi M, Cipolla M, Rivera J, 1987. Analgesic efficacy of two ibuprofen Migliavacca S, 1996. Comparative efficacy of oral ibu- codeine combinations for the treatment of post- profen arginine and intramuscular ketorolac in patients episiotomy and postoperative pain. Clin Pharmacol with postoperative pain. Clin Drug Invest;11:8–14. Ther;42:374–80.

96 Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 10 Comparison of the analgesic efficacy of NSAIDs given by different routes in acute and chronic pain

Summary Introduction The aim of this review was to test the evidence Oral NSAIDs are an important component for a difference in analgesic efficacy and adverse of simple ‘low-technology’ pharmacological effects of NSAIDs administered by different control of both acute and chronic pain. Oral routes. The relevant published RCTs were NSAIDs can be surprisingly effective in patients comparisons of the same drug given by different with moderate to severe postoperative pain. routes. Presence of internal sensitivity was sought Compared with placebo, oral ibuprofen, 400 mg, as a validity criterion. Analgesic and adverse will result in one in every three patients getting effect outcomes were summarised and at least 50% relief of pain over 6 hours.2 This is a synthesised qualitatively. high standard of effectiveness and one against which more complicated methods of delivering In 26 trials (2225 patients analysed) eight different adequate analgesia have to be judged. Invasive NSAIDs were tested in 58 comparisons. In 15 trials procedures like continuous extradural opiate (58%) different routes for the same drug were infusion, or PCA, carry recognised risks3 and compared. Drugs were administered by intraven- may not be available or appropriate for the ous, intramuscular, intrawound, rectal and oral majority of patients with acute or chronic routes in postoperative pain (14 trials), renal colic pain. (4), acute musculoskeletal pain (1), dysmenorrhoea (1) and rheumatoid arthritis (6). Five of While oral NSAIDs can be effective, the advent the 15 direct comparisons were invalid because of rectal and injectable formulations has led they reported no difference between routes but to a fashion for using these routes. This is without evidence of internal sensitivity. reflected in the US Agency for Health Care and Policy Research Acute Pain Guidelines, in In all three direct comparisons in renal colic, which the options for postoperative pain include intravenously administered NSAID had a faster systemic administration of NSAIDs with no onset of action than intramuscular or rectal. In one mention of the oral route.4 There are clinical direct comparison in dysmenorrhoea, orally admin- circumstances in which use of the oral route is istered NSAID was better than rectal. In the five not possible, such as patients who cannot swallow, direct comparisons in postoperative pain the results who are unconscious, nauseated or who have an were inconsistent. In one direct comparison in ileus. If NSAIDs are indicated for such patients, rheumatoid arthritis intramuscular NSAID was then rectal or injectable formulations are the better than oral. Injected and rectal administration only options. However, in the much commoner had some specific adverse effects. circumstance of preoperative premedication of conscious day-case patients who can swallow, In renal colic there is evidence that NSAIDs act or in other acute and chronic pain conditions, quickest when given intravenously. This may be are there any reasons for using rectal or injected clinically relevant. In all other pain conditions formulations rather than oral? The reasons there is a lack of evidence of any difference be- could be greater efficacy, or similar efficacy tween routes. In pain conditions other than renal with fewer adverse effects. However, the evi- colic, there is, therefore, a strong argument in dence for any such advantage over oral use favour of giving NSAIDs orally rather than is unclear and, in this review, the existing by injection. evidence from published reports of direct comparisons of NSAIDs given by different This chapter of the review has been published in routes in both acute and chronic pain full by Tramèr and colleagues.1 is assessed. 97 Comparison of the analgesic efficacy of NSAIDs given by different routes in acute and chronic pain

Methods focus was on trials which compared the same drug given by different routes. Only such direct com- Full reports of published RCTs of direct comparisons were regarded as relevant to this review. parisons of NSAIDs administered by different Comparisons of different drugs across routes were routes and tested in acute or chronic pain with regarded as irrelevant and were not analysed. pain outcomes were sought. A number of different Comparisons between NSAID and non-NSAID search strategies5 were used to identify eligible controls (opioid, placebo) were not considered. reports in MEDLINE (Knowledge Server®, Silver Platter, 1966–96), EMBASE (1986–96) and the Each trial was checked for specific design details Oxford Pain Relief Database (1950–93). The terms with potential impact on trial validity. These details ‘NSAID’, ‘non-steroidal anti-inflammatory’, and were first, whether or not the design included individual drug names were used in conjunction internal sensitivity measures, either a negative with ‘postoperative pain’, ‘renal colic’, ‘*colic’, control (placebo or no treatment) or at least two ‘intravenous’, ‘intramuscular’, and ‘rectal’ in dose levels of an active drug. There was a prior searching, including combinations and without agreement that trials which reported equivalence restriction in language. Additional reports were (i.e. no difference) between routes but which had identified from the reference lists of retrieved no index of internal sensitivity would be regarded reports, review articles and textbooks, by hand- as invalid and not considered for data synthesis. searching locally available anaesthesia journals, Second, the extent to which blinding was protected and by contacting pharmaceutical companies with by using a double-dummy design was checked. licensed parenteral or rectal NSAID preparations. Finally, baseline pain intensity was recorded in trials in which pain was treated (chronic pain settings, Excluded reports for instance), and pain intensity without analgesic Abstracts, letters, review articles and use of topical intervention was recorded in prophylaxis trials formulations (skin, mucous membranes, eye) or (such as postoperative setting). intra-articular use were not considered. Unpublished reports were not sought. Reports in which the Information on the clinical setting, inclusion numbers of patients per treatment group were less criteria, number of patients, study design, and than ten were excluded. Authors were not contacted. drugs, route and doses used was extracted from the reports, together with information on analgesic Included reports measurements and results, and adverse effects. Each report which could possibly meet the Analgesic efficacy was estimated by extracting data inclusion criteria was read by at least two authors of significant difference (p < 0.05, as reported in independently and scored for inclusion and the original trials) between NSAID arms. Relevant methodological quality using a validated 3-item, outcomes were pain intensity at rest or on move- 5-point scale.6 Authors met to agree scores. Reports ment, and additional analgesic consumption. which were described as ‘randomised’ were given Quantitative analysis of combined data was pro- 1 point, and a further point if the method of ran- posed. There was a prior hypothesis that there was domisation was described and adequate (such as no clinically relevant difference between routes of a table of random numbers). There was a pre hoc administration with NSAIDs and, specifically, that agreement that trials without concealment of treat- the oral route would be no different from the ment allocation (allocation according to patient’s other routes of administration. date of birth, for instance) would be excluded from further analysis because of the well-documented risk of overestimation of treatment effects in such Results trials.7,8 One point was given when the trial was described as ‘double-blind’. When the method A total of 26 RCTs (2225 patients analysed), of double-blinding was described and adequate published between 1970 and 1996, were considered (double-dummy method, for instance), a further eligible for the review (Table 32). point was given. Finally, reports which described the numbers of and reasons for withdrawals were Of the 26 trials, 14 were in postoperative given 1 point. Thus, the maximum score for an pain (1268 patients),9–22 four in renal colic included RCT was 5 and the minimum score was 1. (647 patients),23–26 one in acute musculoskeletal pain (77 patients),27 one in dysmenorrhoea Data extraction and analysis (32 patients),28 and six in rheumatoid arthritis These trials compared drug efficacy across different (201 patients).29–34 Different doses of eight different 98 routes of administration. Therefore, the primary NSAIDs (diclofenac, ibuprofen, indomethacin, Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 32 RCTs comparing analgesic efficacy of NSAIDs administered by different routes

Study Quality Regimen: Number Setting Pain outcomes Internal sensitivity Double- Overall Adverse score drug, dose, of patients dummy efficacy effects (1–5) route (no. (> better than; of patients) < worse than; = similar) Index Given

[Bold: relevant [Bold: fulfilled trials, i.e. same validity criteria drug across route] in relevant trials]

Postoperative

Parke, et al., 5 Ketorolac, 113 Major Time to onset of analgesia: Placebo Yes Yes i.v. = i.m. No serious 19959 30 mg i.v., + orthopaedic NSD. adverse events. saline i.m. (38) (moderate Time to first subsequent Ketorolac, or severe analgesic: ketorolac i.v. = 30 mg i.m., + pain) i.m. > placebo. saline i.v. (38) Number achieving 1-point Saline i.m. + decrease of pain scale within i.v. (37) 30 minutes: ketorolac i.m. = placebo; i.v. > placebo. Patients’ rating: ketorolac i.v.> placebo.

Hyrkas, 2 Diclofenac, 151 Third molar PR 0–8 hours Placebo Yes Yes p.o. + p.o. = Diclofenac p.o. et al., 199210 50 mg p.o. (VAS mean): diclofenac p.o. p.o. + i.m. 17/51; rapid + 100 mg and i.m. significantly better diclofenac i.m. p.o. retard than placebo. 16/51; placebo premedication Rescue analgesics: 16/49. (20 mins) (51) significantly less needed in Diclofenac, diclofenac groups. 50 mg i.m. + 100 mg p.o. retard premedication (20 mins) (51) Placebo (49)

Ben-David, 2 Ketorolac, 90 Inguinal hernia Dipyrone at 90 minutes: No Yes No i.m. = i.v. = Stated as none. et al., 199611 30 mg i.m. repair (local i.m. 2/14; i.w. 3/14; i.v. treatment i.w. > p.o. induction (14) anaesthesia) 1/14; p.o. 7/14 (significant); control Ketorolac, 30 mg control 10/14. i.w. induction (14) Buprenorphine at Ketorolac 30 mg 90 minutes: i.m., i.w., i.v., and i.v. induction (14) p.o. 0/14; control 5/14. Ketorolac, 30 mg VAS PI supine/sitting p.o. premedi- (90 minutes): i.m., i.w., and cation (1 hour) i.v. significantly better than (14) p.o. and control. No treatment (14)

Campbell, 1 Diclofenac, 160 Third molar Analgesic needs ‘nil’: Placebo Yes No i.v. > i.m. Bleeding time et al., 1 mg/kg i.v. diclofenac i.v. 14/40; intra-operational 1990a12 induction (40) diclofenac i.m. 12/40; (30 minutes post- Diclofenac, fentanyl 3/40; saline 5/40. injection): 1 mg/kg i.m. VAS PI (mean) at significantly induction (40) 30 minutes postoperatively increased with Fentanyl, 1 µg/kg i.v. (60 minutes postinjection): i.m. diclofenac. induction (40) diclofenac i.v. significantly No other adverse Saline i.m. induction (40) better than all other groups. effects reported.

Jakobsson, 2 Diclofenac, 75 mg 200 Minor No pain (discharge): Placebo Yes No i.m. > p.o. Emesis: no et al., i.m. premedication gynaecology diclofenac i.m. 43/50 and difference. 199613 (10–20 mins) (50) p.o. 34/50; ketorolac i.m. Anxiety: Diclofenac, 50 mg 44/50; placebo 34/50 (i.m. significantly less p.o. premedication NSAID vs. p.o. or placebo in i.m. groups. (10–20 mins) (50) significant difference). No other adverse Ketorolac, 30 mg i.m. No analgesics (discharge): effects reported. premedication diclofenac i.m. 37/50 and (10–20 mins) (50) p.o. 27/50; ketorolac i.m. Saline, i.m. premedication 38/50; placebo 27/50. (10–20 mins) (50)

continued 99 Comparison of the analgesic efficacy of NSAIDs given by different routes in acute and chronic pain

TABLE 32 contd RCTs comparing analgesic efficacy of NSAIDs administered by different routes

[Bold: relevant [Bold: fulfilled trials, i.e. same validity criteria drug across route] in relevant trials]

Postoperative contd

Moore, 1 Diclofenac, 75 mg 32 Thoracotomy PI (VAS mean): i.m. vs. p.r., No No No i.m. = p.r. i.m. 1/16; p.r. 5/16 et al., i.m. postoperative NSD. (2 diarrhoea). 199314 + 75 mg/12 hours: Analgesic consumption total 300 mg/ (mean mg papaveretum): 48 hours (16) NSD. Diclofenac, 100 mg p.r. postoperative + 100 mg/12 hours: total 500 mg/ 48 hours (16)

Dennis, 3 Diclofenac, 100 mg 40 Knee Pain after 24 hours (nil or No No Yes i.v. = p.r. No adverse effects et al., p.r. premedication arthroscopy mild): ketorolac i.v. 11/20; reported. 199515 (1 hour) (20) diclofenac p.r. 14/20. Ketorolac, 10 mg i.v. Activity restriction (none induction (20) or mild): ketorolac i.v. 12/20; diclofenac p.r. 16/20.

Lysak, 4 Ketorolac, 60 mg 84 Gynaecology PI (mild at discharge): No No Yes i.m. = p.o. Ketorolac i.m. et al., i.m. premedication laparoscopy ketorolac i.m.90%; 7/29; piroxicam 199416 (30 minutes) (29) piroxicam p.o. 97%; p.o. 18/28; fentanyl Piroxicam, 40 mg fentanyl 63%. 8/27. p.o. premedication Morphine required in (90 minutes) (28) postanaesthetic care unit: Fentanyl, 100 µg i.v. ketorolac i.m. 16/29; induction + 2 x piroxicam p.o. 25/28; 25 µg i.v. intra- fentanyl 20/27. operative (27)

Morrison, 3 Paracetamol, 650 mg 60 Strabismus VAS PI at 5 hours: ketorolac No Yes No i.v. > p.o. Nausea: ketorolac et al., 199417 p.o. postoperative (adult) i.v. significantly better than i.v. 0/20, para- (30 minutes) (20) paracetamol and cetamol 2/20, Ibuprofen, 600 mg ibuprofen p.o. ibuprofen 2/20. p.o. postoperative No additional analgesia at No other adverse (30 minutes) (20) 5 hours: paracetamol 4/20; effects mentioned. Ketorolac, 60 mg i.v. ibuprofen p.o. 0/20; end of surgery (20) ketorolac i.v. 13/20.

Morley- 3 Ketorolac, 30 mg i.m. 87 Gynaecology VAS PI: significantly lower Placebo Yes Yes p.r. = .i.m Ketorolac i.m. 7/31; Forster, induction (31) or breast with NSAID at 15 and indomethacin p.r. et al., 199318 Indomethacin, 100 mg 90 minutes but not at 9/31; placebo 8/25. p.r. induction (31) 60 minutes (p.r. = i.m.). Placebo i.m. and No additional analgesics: p.r. (25) both NSAIDs significantly better than placebo.

Murrell, 2 Indomethacin, 100 mg 137 Gynaecology Analgesic use up to Placebo No Yes p.r. = i.m. No difference in et al., 199619 p.r. + saline i.m. laparoscopy 180 minutes (fentanyl, frequency of induction (38) paracetamol/codeine): NSD. complaints of pain Ketorolac, 30 mg i.m. VAS PI: no difference at 30 at injection site. + placebo p.r. and 60 minutes. induction (51) Post hoc significant differ- Placebo p.r. and i.m. ence between ketorolac induction (48) and placebo at 180 minutes.

Roelofse, 1 Tenoxicam, 20 mg i.v. 25 Third molar VAS PI: significantly lower No Yes No (i.m. + p.o.) > Discomfort due et al., 199320 intra-operative + 20 mg with diclofenac at 1, 2 and (i.v. + p.o.) to i.m. injection: p.o. postoperative (12) 3 hours. 13/13 with Diclofenac, 75 mg i.m. diclofenac. intra-operative + 50 mg p.o. postoperative (13)

100 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 32 contd RCTs comparing analgesic efficacy of NSAIDs administered by different routes

[Bold: relevant [Bold: fulfilled trials, i.e. same validity criteria drug across route] in relevant trials]

Postoperative contd

Rusy, et al., 2 Ketorolac, 1 mg/kg i.v. 50 Tonsillectomy ‘Objective pain score’ (blood No No No i.v. = p.r. Extra homeostatic 199521 induction (25) (children) pressure, crying, agitation, measurements: Paracetamol, 35 mg/kg movement, verbal report): significantly more p.r. induction (25) ketorolac > paracetamol at with ketorolac 2 hours. No difference at (not related 30 minutes, 1 and 3 hours. to route). Additional analgesics up to 3 hours: morphine and codeine, NSD; paracetamol: significantly less with ketorolac.

Wakeling, 4 Diclofenac, 100 mg p.r. 39 Third molar Median time to rescue No No Yes p.o. = p.r. Vomiting: piroxicam et al., 199622 premedication (1 hour) analgesic: diclofenac 3/20; diclofenac + placebo p.o. (19) 350 minutes; piroxicam 0/19. Piroxicam, 40 mg p.o. 305 minutes (p > 0.05). Nausea: piroxicam premedication (1 hour) No analgesic after 18 hours: 2/20; diclofenac + placebo p.r. (20) diclofenac 2/19; piroxicam 6/20. 1/19. VAS PI similar at any time.

Renal colic

Muriel- 4 Dipyrone, 1 g i.m. 302 Renal colic Proportion of patients with Dose Yes Yes i.v. > i.m.; i.v.: vomiting x 1 Villoria, + placebo i.v. (71) (VAS PI > 50) > 50% improvement: response i.v. > i.m.; (diclofenac). et al., 199523 Dipyrone, 1 g i.v. + significant differences i.v. > i.m. i.m.: drowsiness placebo i.m. (30) dipyrone, 2 g i.v. >1 g i.v. at x 3 (1 dipyrone, Dipyrone, 2 g i.v. + 10 minutes; diclofenac, 75 mg 1 g; 1 dipyrone, placebo i.m. (76) i.v. > i.m. at 20 minutes; 2 g; 1 diclofenac); Dipyrone, 2 g i.m. + dipyrone, 2 g i.v. > 2 g i.m. at drowsiness x placebo i.v. (71) 10 and 20 minutes; dipyrone, 1 (diclofenac). Diclofenac, 75 mg 1 g i.v. > 1 g i.m. at 20 minutes. i.m. + placebo i.v. (32) Diclofenac, 75 mg i.v. + placebo i.m. (22)

Nelson, 1 Indomethacin, 84 Renal colic VAS PI at 10 minutes: i.v. No Yes Yes i.v. > p.r. 30/84 non-drug- et al., 198824 100 mg p.r. (47) significantly lower than p.r.; related with- Indomethacin, at 30 minutes, NSD. drawals (non- 50 mg i.v. (37) Supplementary analgesics: retention of p.r. 16/47; i.v. 8/37. suppositories and others). Drug-related: p.r. 8/47; i.v. 18/37.

Nissen, 3 Indomethacin, 116 Renal colic VAS PI at 10 and 20 minutes: No Ye s No i.v. > p.r. i.v. 44/53; p.r. 29/63 et al., 100 mg p.r. (63) i.v. significantly lower than (p = 0.03). 199025 Indomethacin, p.r.; at 30 minutes, NSD. 50 mg i.v. (53) Supplementary analgesics: p.r. 17/63; i.v. 5/53 (p = 0.03).

El-Sherif, 2 Indomethacin, 50 mg 145 Renal colic PR (‘complete’) after first No Yes No i.v. > i.m. Indomethacin et al., 199026 i.v. (44) dose at 30 minutes: indome- i.v. 5/44; diclofenac Diclofenac, 50 mg thacin i.v. 37/44; diclofenac i.m. i.m. 3/44; avafortan i.m. (47) 31/47; avafortan 45/54. 0/54. Avafortan (dipyrone + NSAID i.v. significantly better antispasmodic) i.v. (54) than i.m.

continued 101 Comparison of the analgesic efficacy of NSAIDs given by different routes in acute and chronic pain

TABLE 32 contd RCTs comparing analgesic efficacy of NSAIDs administered by different routes

[Bold: relevant [Bold: fulfilled trials, i.e. same validity criteria drug across route] in relevant trials]

Acute musculoskeletal pain

Turturro, 5 Ketorolac, 60 mg 77 Acute musculo- VAS PI 0–120 minutes: NSD No No Yes i.m. = p.o. Dyspepsia: et al, 199527 i.m. + placebo p.o. (40) skeletal pain between ketorolac i.m. and ketorolac 1/40; Ibuprofen, 800 mg p.o. (treatment) ibuprofen p.o. ibuprofen 2/37. + placebo i.m. (37) Sedation: ketorolac 1/40; ibuprofen 0/37. Dry mouth: ketorolac 0/40; ibuprofen 1/37.

Dysmenorrhoea

Ylikorkala, 4 Crossover (n = 32) 32 Primary No difference in number of No Yes Yes p.o. > p.r. p.r. irritation et al., 198028 Naproxen, 500 mg dysmenorrhoea additional analgesics taken. after naproxen p.o. 6-hourly + Spasmodic PR (score): suppositories: placebo p.r. significantly better with p.o. 2/32. Naproxen, 500 mg All other symptoms (score): p.r. 6-hourly + NSD. Patients’ overall placebo p.o. assessment: NSD.

Rheumatoid arthritis

Dougados, 3 Ketoprofen, 2 x 40 Rheumatoid Decrease in VAS PI, patients’ No Yes Yes i.m. > p.o. Active p.o. (6/20): et al., 199229 50 mg p.o. + placebo arthritis global judgement, maximum 3 nausea, i.m. (20) (VAS PI > 40) decrease in VAS PI after 1 vomiting, Ketoprofen, 100 mg treatment: p.o. better than 1 indigestion, i.m. + placebo i.m. (NSD). 1 diarrhoea, p.o. (20) Delay until lowest pain 1 headache, intensity: i.m. shorter than 2 vertigo, p.o. (p < 0.05). 1 dyspnea. Active i.m. (5/20): 1 headache, 2 somnolence, 1 pruritus, 1 vertigo.

Baber, 4 Crossover (n = 13) 13 Rheumatoid Articular index, grip strength, No No Yes p.o. = p.r. p.o.: 17/13, et al., Placebo p.r. + placebo arthritis pain score (analogue 0–9), pr: 11/13. 198030 p.o. for 1 week, then morning stiffness, digital joint Diarrhoea: 4/13 p.r. Indomethacin, size, rescue analgesics Indigestion: 5/13 100 mg p.r. + placebo (paracetamol) and patients’ p.o., 2/13 p.r. p.o. for 1 week preference: NSD between Indomethacin, p.o. and p.r. 100 mg p.o. + placebo p.r. for 1 week

Hansen, 3 Crossover (n = 12) 12 Rheumatoid Morning stiffness and pain No No Yes p.o. = p.r. Endoscopy: equal et al., 198431 Placebo p.r. + placebo arthritis (PI on 50 mm VAS): amounts gastric p.o. for 2 weeks, then no difference between mucosal damage. Indomethacin, p.r. and p.o. p.r. discomfort 100–150 mg p.o. + with suppositories: placebo p.r. for 8 p.o. active + p.r. 2 weeks placebo; 7 p.r. active Indomethacin, 100– + p.o. placebo; 7 150 mg p.r. + placebo p.o. and p.r. placebo. p.o. for 2 weeks

Huskisson, 3 Crossover (n = 20) 20 Rheumatoid Patients’ preference: p.o. No No Yes p.o. = p.r. Nausea, anorexia, et al., 197032 Indomethacin, arthritis significantly better than p.r. epigastric dis- 100 mg p.r. + placebo No difference in pain, morning comfort: p.o. 2/20; p.o. for days 1 and 3 stiffness, duration of stiffness. p.r. 2/20. Indomethacin, 100 mg p.o. + placebo p.r. for days 2 and 4

102 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 32 contd RCTs comparing analgesic efficacy of NSAIDs administered by different routes

[Bold: relevant [Bold: fulfilled trials, i.e. same validity criteria drug across route] in relevant trials]

Rheumatoid arthritis contd

Iversen, 4 Crossover (n = 22) 22 Rheumatoid Day and night pain (4-point No No Yes p.o. = p.r. p.o. (5): headache, et al., 198133 Indomethacin SR, arthritis scale), morning stiffness dizziness,‘loose 75 mg p.o. + placebo (minutes): NSD. bowel motions’, p.r. for 2 weeks Conventional grip strength: stomach pain. Indomethacin, p.r. significantly better than p.r. (8): headache, 100 mg p.r. + placebo p.o. (6 mm). ‘loose bowel p.r. for 2 weeks motions’.

Uddenfeldt, 4 Crossover (n = 94) 94 Rheumatoid Morning stiffness (duration): No No Yes p.o. = p.r p.r. irritation: et al., Ketoprofen CR, 200 mg arthritis p.o. significantly better than p.r. indomethacin 7/94. 199334 p.o. + placebo p.r. for Awakenings during night: p.r. Gastro-intestinal 3 weeks significantly better than p.o. problems: keto- Indomethacin, 100 mg NSD in pain on awakening profen 40/94; p.r. + placebo p.o. for (VAS), articular index, patients’ indomethacin 3 weeks and doctors’ assessment 28/94. (7-step scale), rescue analgesics. ketoprofen, ketorolac, naproxen, piroxicam, sensitivity (i.e. which were valid), five were in tenoxicam), given by intravenous, intramuscular, postoperative pain,9–13 three were in renal colic,23–25 intrawound, rectal and oral routes, were tested in one was in dysmenorrhoea,28 and one was in 58 single-dose or multiple-dose comparisons. rheumatoid arthritis.29 Six of them used a double-dummy design.9,10,23,24,28,29 The median quality score of all trials was 3 (range 1–5). Quantitative analysis was not considered Postoperative pain appropriate because of the variety of clinical Of 14 trials in postoperative pain, five were valid settings, drugs, doses, routes and pain outcomes direct comparisons. They compared diclofenac or reported. Instead, any statistically significant differ- ketorolac across routes. ence between treatments was extracted from the original reports and documented in table format as In one trial, diclofenac, 1 mg/kg injected had been done previously for other qualitative intravenously at induction of anaesthesia, led to systematic reviews.35,36 A ‘vote counting’ procedure significantly lower pain intensity scores 30 minutes was then agreed, giving positive or negative votes if after surgery than the same dose given intramuscu- there was evidence of presence or absence, respec- larly at induction.12 In two other trials no difference tively, of a significant difference between routes. was found between ketorolac, 30 mg, given either intravenously or intramuscularly at induction.9,11 In In all, 15 trials (58% of all analysed trials) were one of these trials, inguinal hernia repair was per- relevant to this review because the same drug formed under local anaesthesia with very low pain was compared by different routes of administra- scores during the postoperative observation period tion.9–14,23–25,28–-33 In nine of them (35% of all whether or not an NSAID or no treatment was trials), the same drug was compared at the given.11 In the same trial, both intramuscular and same dose.9,11,12,23,28–32 intravenous ketorolac, 30 mg, at induction led to significantly lower pain scores and less rescue In five of the 15 relevant trials equivalence was analgesics at 90 minutes after surgery than the same reported between routes but there was no index dose taken orally but 1 hour before surgery.11 Group of internal sensitivity.14,30–33 These trials were not, sizes in this trial were small (i.e. 14 patients per therefore, analysed further. group) and it was not of double-dummy design.

Of the ten relevant trials in which a significant In another trial with larger groups (50 patients per difference between routes was reported, or equi- group) but, again, not of double-dummy design, valence was reported but with an index of internal less pain and rescue analgesics at discharge were 103 Comparison of the analgesic efficacy of NSAIDs given by different routes in acute and chronic pain

reported with diclofenac, 75 mg, intramuscularly Other pain conditions compared with the same drug given orally but at a No direct comparisons from other pain conditions lower dose (50 mg).13 In yet another trial, diclo- were found. fenac, 150 mg, taken orally was compared with diclofenac, 50 mg, intramuscularly plus 100 mg Adverse effects orally.10 The drugs were given as premedication Commonly reported adverse effects independent using a double-dummy design, and group sizes of the route of administration were nausea, vomit- were large (50 patients per group). No difference ing, dizziness, drowsiness, sedation, anxiety, dys- was found between the two forms of administration. pepsia, indigestion, and dry mouth (Table 32). Two studies reported bleeding time changes.12,21 In Renal colic 12 patients with rheumatoid arthritis treated for Of four trials in renal colic, there were three valid 2 weeks with indomethacin, 100–150 mg orally and direct comparisons.23–25 Dipyrone, diclofenac, and rectally, respectively, in a study of crossover design, indomethacin given by different routes were com- endoscopically diagnosed gastric mucosal damage pared. Two were of double-dummy design.23,24 In was independent of the route of administration.31 one, baseline pain before treatment was started (at least 50 mm on a 100 mm VAS) was defined.23 Adverse effects related to the route of adminis- Group sizes in these three trials were between tration were most often reported for intramuscular 22 and 76 patients. and rectal regimens (Table 32). Discomfort at the site of injection19,20 was the most frequent complaint In one trial, pain relief was tested with dipyrone, relating to intramuscular injections. After rectal 1 g or 2 g, and diclofenac, 75 mg, given intra- administration, diarrhoea,14,30 rectal irritation,28,31,34 muscularly compared with intravenously.23 At and non-retention of suppositories24 were reported. 10 and 20 minutes after administration, the proportion of patients with at least 50% improvement was significantly in favour of the intravenous route Comment with each drug and dose. Many doctors use injected or rectal NSAIDs when In the two other trials, intravenous indomethacin, the oral route could be used. This is despite advice 50 mg, was compared with the same drug given to use the least invasive route possible, with the rectally but at double the dose.24,25 Despite the statement that no study has specifically compared intravenous dose being only half the rectal dose, the analgesic efficacy of alternative routes of the significant improvement (less pain intensity, fewer same drug.37 Reasons for preferring injected or rescue analgesics) was reported for the intravenous rectal formulations when the oral route could be compared with the rectal route. Again these differ- used might be greater efficacy or faster onset of pain ences were apparent only at 10 or 20 minutes relief. The safety argument would be that these effi- post-administration. cacy benefits were achieved at no greater (or accept- ably greater) level of adverse effects. Patients may Dysmenorrhoea prefer oral to rectal dosing.38 There are also legal Only one trial in dysmenorrhoea was found.28 ramifications, because of the obligation for consent This crossover trial in 38 patients compared oral if drugs are given rectally while a patient is asleep.39 with rectal naproxen, both 500 mg at 6-hourly intervals, using a double-dummy design. Relief Using evidence from systematically searched of spasmodic pain was significantly better with published reports of RCTs with direct comparisons, the oral route. we wanted to compare the benefits and risks of NSAIDs given by different routes in acute and Chronic pain chronic pain. Systematic reviews are powerful instru- Six trials in rheumatoid arthritis were found. ments to gain more insight in treatment efficacy and Five of them were direct comparisons but only harm. Ideally, dichotomous outcomes would be one small trial using a double-dummy design with extracted from original reports and combined using 20 patients per group was valid.29 Patients with biostatistical methods. In some circumstances, dich- defined baseline pain (at least 40 mm on a 100 mm otomous data may be extracted from measurements scale) receiving ketoprofen, 100 mg, intramuscu- which were originally not binary outcomes.40 larly reported a significantly shorter delay until the lowest pain intensity score was achieved than However, such quantitative analysis was not patients receiving the same dose of the same possible because of the variety of clinical settings, 104 drug orally. drugs, doses, routes and pain outcomes reported. Health Technology Assessment 1998; Vol. 2: No. 12

Systematic reviewers then have to rely on blinding of the other trials must be questioned. In statistically significant results as reported in trials with deficient blinding, the therapeutic effect the original reports and apply a vote-counting may be exaggerated.7 While all trials in rheumatoid procedure.35 It is obvious that such a qualitative arthritis used a double-dummy technique, this was approach is vulnerable to bias. Vote counting take true for only half of the surgical trials. An extreme no account of the size of the trial or of the size of example was the comparison of an oral drug given any difference in effect. In such analyses, pre-set 1 hour before surgery with the same drug given by validity criteria become especially important in injection at induction.11 This trial did not use a order to minimise the risk of bias.36 double-dummy method and reported better analgesic efficacy for the parenteral route These trials highlighted different methodological compared with the oral. problems affecting the validity of the trials. Finally, a pre hoc defined pain intensity, sufficient to First, if the null hypothesis was that there was no provide measurable change after study treatment, difference between the routes, comparisons of was reported in only a minority of trials measuring the same drug given by different routes might be pain relief.9,23,29 Very low pain intensity scores expected. It might be desirable to concentrate independent of the treatment were reported in only on comparisons of the same drugs at the some trials where pain was meant to be prevented, same dose. However, only about half of these trials such as in a surgical setting. If there is no pain, compared the same NSAID across routes, thus analgesia cannot be measured. A pain trial without addressing what was regarded as the clinically an adequate baseline pain intensity is not a relevant question. Only one-third of all trials would valid assay.36 have satisfied stricter validity rules (i.e. direct comparison of the same drug at the same dose). Applying our rules of validity to these 26 trials revealed that only 15 (58%) of all systematically Second, the classical approach to design of searched trials were relevant to this review (i.e. analgesic trials is to build in an index of internal investigated the same drug given by different sensitivity, either by using a placebo (or no treat- routes). Five of them had to be excluded because ment) control, or by including a high and a low their results could not be interpreted (i.e. they dose of a standard analgesic to establish a dose– reported equivalence but had no index of internal response relationship. What such designs seek to sensitivity). This meant that only ten trials (38% of achieve is a defence against equivalence of treat- all trials) could be analysed, and only six of them ments. Lack of internal sensitivity is a key issue in were of double-dummy design. In renal colic, there equivalence trials.41 This has been shown in syste- was evidence from three valid direct comparisons matic reviews of analgesic trials.36 Without such that the intravenous route acted significantly faster controls, equivalence in a comparison of two or than the rectal or intramuscular route. Although more drugs may mean that the methods of measur- this difference was only evident during the first ing pain or its relief failed in that study, rather than 10 to 20 minutes, the faster onset of action is likely that this was a true negative result of no difference to be clinically relevant in this specific setting. In between the analgesic effect of the drugs. Power the only trial in dysmenorrhoea, one outcome calculations cannot be a defence against methodo- measurement indicated that oral NSAID may be logical failure. Only a positive result (significant better than rectal. In one trial in rheumatoid difference) despite the lack of negative controls is arthritis, one isolated endpoint suggested that an adequate vindication of such methods. intramuscular ketoprofen may be superior to oral; however, the clinical utility of this is unclear. Eight (31%) of the 26 trials has a method built in Finally, in the surgical setting, results were far that ensured internal sensitivity in the form of a from being conclusive. placebo control,9,10,12,13,18,19 a no-treatment control,11 or two dose levels of the same drug given by the Reporting of adverse effects was generally poor same route.23 Five direct comparisons reported and mostly not related to route of administration. equivalence between routes but had no index of Rectal and intramuscular routes were most likely to internal sensitivity. These trials were, therefore, have specific local adverse effects. These have to be invalid and were excluded from further analysis. taken into account when the advantages of one route over the other are discussed. Third, although all these trials were, by definition, comparisons of NSAIDs given by different routes, With the exception of the renal colic setting, only 17 (65%) were of double-dummy design. The these trials constitute a lack of evidence for any 105 Comparison of the analgesic efficacy of NSAIDs given by different routes in acute and chronic pain

difference rather than evidence of lack of 7. Schulz KF, Chalmers I, Hayes RJ, Altman DG. difference between NSAIDs given by different Empirical evidence of bias: dimensions of methodo- routes. This is not just semantics; if there is ade- logical quality associated with estimates of treatment effects in controlled trials. JAMA 1995;273:408–12. quate evidence of a lack of difference then practice should change, reverting to the safest and simplest 8. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. option, the oral route. If there is a lack of evidence Randomization is important in studies with pain (rather than evidence of a lack of difference), then outcomes: systematic review of transcutaneous a research agenda is set, to determine whether or electrical nerve stimulation in acute postoperative not there is any clinical advantage of one route pain. Br J Anaesth 1996;77:798–803. over another. It could be argued that patients 9. Parke TJ, Millett S, Old S, Goodwin APL, Rice ASC. should again receive the safest and simplest option Ketorolac for early postoperative analgesia. J Clin unless they agree to participate in a randomised Anesth 1995;7:465–9. comparison of different routes of administration. The research agenda should be to design simple 10. Hyrkas T, Ylipaavalniemi P, Oikarinen VJ, Hampf G. comparisons of the same drug at the same dose Postoperative pain prevention by a single-dose across route, with validity and, ideally, with stand- formulation of diclofenac producing a steady ardised outcome measures in the various studies plasma concentration. J Oral Maxillofac Surg to make combined quantitative analysis possible. 1992;50:124–7. 11. Ben-David B, Baune-Goldstein U, Goldik Z, Gaitini It does not seem right that over 2200 patients have L. Is preoperative ketorolac a useful adjunct to already participated in trials over the past 26 years regional anesthesia for inguinal herniorrhaphy? and, yet, for the majority of clinical settings we still Acta Anaesthesiol Scand 1996;40:358–63. cannot answer the simple question, ‘Is it better to 12. Campbell WI, Kendrick R, Patterson C. Intravenous give NSAIDs by injection or suppository than to diclofenac sodium. Does its administration before take them orally?’ operation suppress postoperative pain? Anaesthesia 1990;45:763–6.

References 13. Jakobsson J, Rane K, Davidson S. Intramuscular NSAIDs reduce post-operative pain after minor 1. Tramèr M, Williams J, Carroll D, Wiffen PJ, outpatient anaesthesia. Eur J Anaesth 1996;13:67–71. McQuay HJ, Moore RA. Systematic review of direct comparisons of non-steroidal anti-inflammatory 14. Moore AP, Thorpe JA, Mulley BA, Cruise M. Which drugs given by different routes for acute pain. diclofenac form is best? Post-thoracotamy pain: Acta Anaesth Scand 1997;42:71–9. intramuscular v. rectal administration of diclofenac. Hosp Pharmacy Pract 1993;September:431–2,435. 2. Collins SL, Moore A, McQuay HJ, Wiffen PJ. Oral ibuprofen and diclofenac in postoperative pain: a 15. Dennis AR, Leeson-Payne CG, Hobbs GJ. A com- quantitative systematic review. Eur J Pain. In press. parison of diclofenac with ketorolac for pain relief after knee arthroscopy. Anaesthesia 1995;50:904–6. 3. Bates DW, Cullen DJ, Laird N, Petersen LA, Small SD, Servi D, et al. Incidence of adverse drug events 16. Lysak SZ, Anderson PT, Carithers RA, DeVane and potential adverse drug events. JAMA GG, Smith ML, Bates GW. Postoperative effects of 1995;274:29–34. fentanyl, ketorolac, and piroxicam as analgesics for outpatient laparoscopic procedures. Obstet Gynecol 4. Acute Pain Management Guideline Panel. Acute 1994;83:270–5. Pain Management: operative or medical procedures and trauma. Clinical Practice Guideline No. 1. 17. Morrison NA, Repka MX. Ketorolac versus Agency for Health Care Policy and Research, acetaminophen or ibuprofen in controlling US Department of Health and Human Services. postoperative pain in patients with strabismus. AHCPR Publication No 92-0032. Rockville, MD: Ophthalmology 1994;101:915–18. Public Health Service, 1992:24–5. 18. Morley-Forster P, Newton PT, Cook M. Ketorolac 5. Jadad AR, McQuay HJ. A high-yield strategy to and indomethacin are equally efficacious for the identify randomized controlled trials for systematic relief of minor postoperative pain. Can J Anaesth reviews. Online J Curr Clin Trials [serial online] 1993;40:1126–30. 1993;Doc No 33:3973 words;39 paragraphs;5 tables. 19. Murrell GC, Leake T, Hughes PJ. A comparison of 6. Jadad AR, Moore RA, Carroll D, Jenkinson C, the efficacy of ketorolac and indomethacin for Reynolds DJM, Gavaghan DJ, et al. Assessing the postoperative analgesia following laparoscopic quality of reports of randomized clinical trials: is surgery in day patients. Anaesth Intens Care 106 blinding necessary? Control Clin Trials 1996;17:1–12. 1996;24:237–40. Health Technology Assessment 1998; Vol. 2: No. 12

20. Roelofse JA, van der Bijl P, Joubert JJD. An open 30. Baber N, Sibeon R, Laws E, Halliday L, Orme M, comparative study of the analgesic effects of Littler T. Indomethacin in rheumatoid arthritis: tenoxicam and diclofenac sodium after third molar comparison of oral and rectal dosing. Br J Clin surgery. Anesth Pain Control Dentistry 1993;2:217–22. Pharmacol 1980;10:387–92. 21. Rusy LM, Houck CS, Sullivan LJ, Ohlms LA, Jones 31. Hansen TM, Matzen P, Madsen P. Endoscopic DT, McGill TJ, et al. A double-blind evaluation of evaluation of the effect of indomethacin capsules ketorolac tromethamine versus acetaminophen in and suppositories on the gastric mucosa in pediatric tonsillectomy: analgesia and bleeding. rheumatic patients. J Rheumatol 1984;11:484–7. Anesth Analg 1995;80:226–9. 32. Huskisson EC, Taylor RT, Burston D, Chuter PJ, 22. Wakeling HG, Barry PC, Butler PJ. Postoperative Hart FD. Evening indomethacin in the treatment of analgesia in dental day case surgery. A comparison rheumatoid arthritis. Ann Rheum Dis 1970;29:393–6. between Feldene ‘Melt’ (piroxicam) and diclofenac suppositories. Anaesthesia 1996;51:784–6. 33. Iversen O, Fowles M, Vlieg M, Smidt N, Wigley R. Sustained release indomethacin: a double blind 23. Muriel-Villoria X, Muriel-Villoria C, Zungri-Telo E, comparison with indomethacin suppositories. et al. Comparison of the onset and duration of the N Z Med J 1981;93:261–2. analgesic effect of dipyrone, 1 or 2 g, by the intramuscular or intravenous route, in acute renal 34. Uddenfeldt P, Leden I, Rubin B. A double-blind colic. Eur J Clin Pharmacol 1995;48:103–7. comparison of oral ketoprofen ‘controlled release’ and indomethacin suppository in the treatment of 24. Nelson CE, Nylander C, Olsson AM, Olsson R, rheumatoid arthritis with special regard to morning Pettersson BA, Wallstrom I. Rectal v. intravenous stiffness and pain on awakening. Curr Med Res Opin administration of indomethacin in the treatment of 1993;13:127–32. renal colic. Acta Chir Scand 1988;154:253–5. 35. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. 25. Nissen I, Birke H, Olsen J, Wurtz E, Lorentzen K, Transcutaneous electrical nerve stimulation in Salomon H, et al. Treatment of ureteric colic. labour pain: a systematic review. Br J Obstet Intravenous versus rectal administration of Gynaecol 1997;104:169–75. indomethacin. Br J Urol 1990;65:576–9. 36. Kalso E, Tramèr M, Carroll D, McQuay H, Moore 26. El-Sherif A, Foda R, Norlen L, Yahia H. Treatment RA. Pain relief from intra-articular morphine after of renal colic by prostaglandin synthetase inhibitors knee surgery: a qualitative systematic review. Pain and avafortan (analgesic antispasmodic). Br J Urol 1997;71:642–51. 1990;66:602–5. 27. Turturro MA, Paris PM, Seaberg DC. Intramuscular 37. Cashman J, McAnulty G. Nonsteroidal anti- ketorolac versus oral ibuprofen in acute musculo- inflammatory drugs in perisurgical pain manage- skeletal pain. Ann Emerg Med 1995;26:117–20. ment. Mechanisms of action and rationale for optimum use. Drugs 1995;49:51–70. 28. Ylikorkala O, Puolakka J, Kauppila A. Comparison between naproxen tablets and suppositories in 38. Vyvyan HAL, Hanafiah Z. Patients’ attitudes to primary dysmenorrhea. Prostaglandins rectal drug administration. Anaesthesia 1980;20:463–8. 1995;50:983–4. 29. Dougados M, Listrat V, Duchesne L, Amor B. 39. Mitchell J. A fundamental problem of consent. Efficacite comparee du ketoprofene en fonction de BMJ 1995;310:43–8. sa voie d’administration (intra-musculaire ou per 40. McQuay HJ, Moore RA. Using numerical results os). Une etude en double aveugle contre placebo au from systematic reviews in clinical practice. cours de la polyarthrite rhumatoide [Comparative Ann Intern Med 1997;126:712–20. efficacy of ketoprofen related to the route of administration (intramuscular or per os). A double- 41. Jones B, Jarvis P, Lewis JA, Ebbutt AF. Trials to assess blind study versus placebo in rheumatoid arthritis]. equivalence: the importance of rigorous methods. Rev Rhum Mal Osteoartic 1992;59:769–73. BMJ 1996;313:36–9.

107

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 11 Topically-applied NSAIDs

Summary rubefacients.2,3 This systematic review was undertaken to examine the evidence that topical In this chapter the effectiveness and safety of NSAIDs are effective and safe, and to determine topical NSAIDs in acute (soft tissue trauma, strains whether there is evidence of differences between and sprains) and chronic pain conditions (osteo- topical preparations. arthritis, tendinitis) are assessed. In all, 86 RCTs involving 10,160 patients were found. Measures approximating at least 50% pain relief, local and Methods systemic adverse effects were extracted. Analysis was undertaken at 1 week for acute and 2 weeks for Reports were sought of RCTs of topical noNSAIDs chronic conditions using relative benefit and NNT. in which pain was an outcome. Reports were included which compared topical NSAID(s) with In acute pain conditions, placebo-controlled trials placebo, with another topical NSAID, or with an had a relative benefit of 1.7 (95% CI, 1.5–1.9) and oral NSAID. A number of different search strategies an NNT of 3.9 (95% CI, 3.4–4.4). Analysing by drug in MEDLINE (1966–September 1996), EMBASE (at least three trials), ketoprofen (NNT 2.6), fel- (1981–September 1996) and the Oxford Pain binac (3.0), ibuprofen (3.5) and piroxicam (4.2) Relief Database (1950–94)4 were used to locate had significant efficacy. Benzydamine and indo- reports, using individual drug name (generic and methacin were not distinguished from placebo. proprietary), together with the words ‘administration’, ‘topical’, ‘gel’, ‘ointment’, ‘aerosol’, ‘cream’, In chronic pain conditions, placebo-controlled and combinations of these, without restriction on trials had a relative benefit of 2.0 (95% CI, 1.5–2.7) language. Additional reports were identified from and an NNT of 3.1 (95% CI, 2.7–3.8). Small trials the reference lists of retrieved reports and review (< 40 treated patients) exaggerated the effective- articles. Librarians and medical directors of 12 ness of topical NSAIDs in acute conditions only pharmaceutical companies in the UK identified as (by 24%). There was no relationship between marketing topical NSAIDs were asked for reports of trial quality and treatment effect. RCTs of their products, including any unpublished reports. Abstracts were not sought. Authors were In both acute and chronic pain, local and systemic not contacted. adverse events and drug-related study withdrawal had a low incidence and were no different from RCTs of NSAIDs with pain as an outcome in acute placebo. Topical NSAIDs are effective in relieving conditions (strains, sprains, sports injuries) or pain in acute and chronic conditions. chronic conditions (arthritis, rheumatism) were included. Those in vaginitis, oral or buccal conditions, thrombophlebitis or experimental pain Introduction settings were not.

Some topical NSAIDs are available without Reports were screened by two members of the team prescription and are widely advertised for acute to eliminate those without pain outcomes, which and chronic painful conditions. In the UK, some were definitely not randomised, or were abstracts 20–24 million (predominantly oral) NSAID pre- or reviews. Each report which could possibly be scriptions are written each year, 5% of total NHS described as an RCT was read independently by prescriptions, with many more available without each of the authors and scored using a 3-item, 1–5 prescription. The attributable risk of going to score, quality scale.5 Consensus was then achieved. hospital with gastrointestinal problems is between The maximum score for an included RCT was 1.3% and 1.6% annually for regular users of oral 5 and the minimum 1. NSAIDs.1 This raises the question whether using oral NSAIDs is worse than the disease for some Information about treatment(s) and control(s), patients.2 Despite licensed status there is scepticism condition studied, number of patients randomised that topical NSAIDs have any action other than as and analysed, study design, observation periods, 109 Topically-applied NSAIDs

outcome measures used for pain or global Results evaluation, analgesic outcome results, local skin irritation, systemic adverse effects and study The literature searches found 86 reports withdrawal because of adverse effects was taken (10,160 patients) which fulfilled the inclusion from each report by authors meeting to concur. criteria, 76 of which had dichotomous pain outcomes, including three unpublished reports A clinically relevant outcome was defined as at least with 1695 patients from a pharmaceutical company. 50% pain relief. Only information that was avail- The number of reports, patients, and the distri- able in dichotomous form was used for analysis. bution of quality scores divided by acute or A hierarchy of measures was used for extraction chronic, both placebo and active drug controlled, which approximated, in this order of preference: is shown in Table 33. Over 75% of placebo- controlled trials had quality scores of 3 or more. (i) patient global judgement (excellent/good) Conversely, 60% of active drug controlled trials (ii) pain on movement (no pain/slight pain) had scores of 2 or less. Full details of trial design, (iii) spontaneous pain or pain at rest (no outcome measures, and results are presented in pain/slight pain) Tables 34–37. (iv) physician global judgement (excellent/good) if defined against a stated scale. TABLE 33 Number of reports, patients and the distribution of quality scores The denominator was taken as the number of patients randomised, that is an intention-to-treat Trials Number Number Quality score analysis. For acute conditions the effectiveness of patients (1–5) measure nearest to 1 week after start of treatment 12345 was taken and, for chronic conditions, 2 weeks. Prior hypotheses were that topical NSAIDs were Acute pain no better than placebo and that there were no placebo-controlled 37 3556 1 6 10 13 7 differences between them. Acute pain active drug controlled 24 4171 4 11 4 5 0 The scatter of success rates with topical NSAIDs against success rate with placebo6 was used as a Chronic pain graphical means of exploring the consistency of placebo-controlled 13 1161 0 3 5 5 0 efficacy and the homogeneity of the data. On such Chronic pain active plots a scatter lying predominantly between the line drug controlled 12 1272 2 5 3 2 0 of equality and the axis of the active intervention (topical non-steroidal) would suggest consistent efficacy with the intervention and relative Acute conditions homogeneity. In all, 37 reports of 40 placebo-controlled trials of topical NSAIDs were found (see Tables 34 and 35). Relative risk or benefit with 95% CI was calculated The mean size of group treated with topical drug for pain data using a random effects model7 was 47 patients (median 32). Studies were con- because the results were heterogeneous. Hetero- ducted in recent soft tissue injury, sprains, strains geneity was assumed when p > 0.1. This was per- or trauma. Dichotomous pain outcomes were formed by pooling all data, by pooling data for an available for 1747 patients on active drug treatment individual drug where there were at least three and 1492 on placebo. An additional 24 reports trials and, for sensitivity analysis, by quality score of 24 trials compared different topical NSAIDs and treatment group size. A fixed effect model8 or formulations or route of administration in was used for the (homogeneous) adverse effect 4171 patients. In three studies, a topical NSAID data. A statistically significant improvement over was compared with oral; one such study also had control was assumed when the lower limit of the a placebo control. 95% CI of the relative benefit was > 1. NNTs and 95% CIs were calculated for effect data.9,10 The Relative benefit and 95% CIs are shown for each NNT indicated how many patients with acute or placebo controlled trial in Figure 33. Of the 37 chronic pain have to be treated with topical comparisons, 27 showed statistical superiority for NSAIDs for one of them to achieve at least 50% topical NSAIDs over placebo. The scatter of the pain relief who would not have done with placebo. proportion of patients with at least 50% pain A significant difference between NNTs was relief with topical NSAID or placebo is shown in 110 assumed when CIs did not overlap. Figure 34. Of the 37 comparisons, 36 were in the Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 34 Placebo-controlled trials in acute painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Drug-related Quality study design regimen measures results withdrawals and score and follow-up adverse effects

Airaksinen, Ketoprofen, Acute soft tissue n = 56 5 g twice 1.VAS PI on rest 1. Overall significant 5/29 ketoprofen; Withdrawals: 0/29 2 et al., 1993 2.5% gel; placebo injuries, < 1 week parallel group daily and movement reduction in pain at 4/27 placebo. ketoprofen; 0/27 gel 0, 3, 7 days 2. Patient and rest with ketoprofen placebo.Adverse investigator (NSD for placebo). effects: 1/29 global rating 2. Significant difference ketoprofen; (p < 0.05) in number 0/29 placebo. of patients improved (patient global): 24/29 ketoprofen; 14/27 placebo.

Åkermark, Indomethacin, 1% Repetitive sports n = 70 Indomethacin 1. Improvement 1. Indomethacin spray 4/23 indomethacin Withdrawals: 1/23 5 et al., 1990 spray; indomethacin, injuries parallel group spray, 0.5– VAS showed significant improve- spray; 0/23 indomethacin spray; oral; placebo, double-dummy 1.5 ml three 2. Physician global ment (p < 0.05) days 3 and 7. indomethacin oral; 1/23 indomethacin spray/oral 3, 7, 14 days to five times 5-point scale 2. Marked improvement 0/24 placebo. oral; 0/24 placebo. daily; 3. Pain on move- or symptom free at 1 week: Adverse effects: indomethacin ment, palpation, 10/22 indomethacin spray; 4/23 indomethacin 3 x 25 mg activity, 4-point 5/23 indomethacin oral; spray; 10/23 tablets scale 3/24 placebo. indomethacin oral; 3. Marked improvement or 0/24 placebo. symptom free at 2 weeks: 16/22 indomethacin spray.

Aoki, et al., Piroxicam, 0.5% gel; Acute orthopaedic n = 252 1 g 3–4 times Multiple outcomes: 1. Significant difference in 1/84 piroxicam; Withdrawals and 4 1984 indomethacin, 1% trauma multicentre, daily 1. Overall overall improvement 2/84 indomethacin; adverse effects: 0/84 gel; placebo, gel parallel group improvement (p < 0.05), piroxicam best. 2/84 placebo. piroxicam; 0/84 0, 3, 7 days patient 2. Improvement day 7: indomethacin; 0/84 2. PI – movement, piroxicam significantly better placebo. spontaneous than placebo (p < 0.01). 3. Patients better or much better: 56/84 piroxicam; 41/84 indomethacin; 33/84 placebo.

Auclair, Niflumic acid, 2.5% Achilles heel n = 243 5 g gel three 1. Pain (VAS) on 1. Significantly more pain 5/123 niflumic Adverse effect 3 et al., 1989 gel; placebo, gel tendinitis of recent parallel group times daily palpation reduction than placebo. acid; 6/116 withdrawal: 1/123 origin 7, 21 days 2. PI 2. Pain on dorsiflexion placebo. niflumic acid; 0/116 at dorsifexion disappeared or improved: placebo. 3. Global patient 75/117 niflumic acid; 69/110 placebo. 3. Global very good/good : 69/117 niflumic acid; 54/109 placebo.

Baracchi, Ibuprofen, 10% Acute soft tissue n = 40 Twice daily 1. Categorical 1. Ibuprofen significantly No report of Well tolerated. 4 et al., 1982 cream; placebo, trauma parallel group spontaneous pain, better than placebo (for local effects. cream 3, 5, 7, 10, 12, pain on movement spontaneous pain, p < 0.001). 14 days and pressure 2. Global (good or excellent 2. Investigator response): 17/20 ibuprofen; global 3/20 placebo.

Campbell & Ibuprofen, 5% Acute ankle sprain, n = 100 4 inches four VAS on rest and 1. Ibuprofen better than Not reported. Withdrawals: 0/50 4 Dunn, 1994 cream; placebo, < 24 hours parallel group times daily movement placebo on days 2 and 3. ibuprofen; 0/50 cream 2-week diaries 2. Improved walking ability placebo.Adverse at day 7: 21/50 ibuprofen; effects: 1/50 19/50 placebo. ibuprofen; 0/50 placebo. 55/100 returned diaries.

Candela, Ketoprofen, gel; Traumatic sport n = 30 Twice daily CAT scales, pain 1. Ketoprofen better Not reported. Not reported. 1 et al., 1986 placebo, gel injuries parallel group on pressure, on than placebo. 5, 10, 15 days movement, 2. Better/much better at functional limitation day 10: 10/15 ketoprofen; 2/15 placebo.

Chaterjee, Benzydamine, 3% Soft tissue injuries n = 51 Three times VRS pain – 1. Benzydamine better than Not reported. Withdrawals and 4 1977 cream; placebo, parallel group daily spontaneous, placebo at 6 days spontane- adverse effects: 0/25 cream 6 days pressure, movement ous pain, pressure, movement. benzydamine; 0/25 2. None or slight pain on placebo. movement on day 6: 21/25 benzydamine; 12/25 placebo.

continued 111 Topically-applied NSAIDs

TABLE 34 contd Placebo-controlled trials in acute painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Drug-related Quality study design regimen measures results withdrawals and score and follow-up adverse effects

Diebshlag, Diclofenac, gel; Ankle sprains n = 20 Ad libitum 1.Ankle joint Reduced swelling and less Not reported. Withdrawals and 3 1986 placebo, gel crossover volume measurement pain with diclofenac. adverse effects: 0/20 2 x 1 week 2.VAS PI diclofenac; 0/20 placebo.

Diebshlag, Ketorolac, 2% gel; Acute ankle sprain n = 37 3 g three times 1.Ankle joint 1. Ketorolac better than 1/13 ketorolac; Withdrawals and 4 et al., 1990 etofenamate, 5% gel; parallel group daily volume measurement placebo and etofenemate. 1/12 etofenamate; adverse effects: placebo, gel 2, 3, 4, 8, 14 days 2.VAS PI 2. Improved by day 3: 0/12 0/12 placebo. 0/12 placebo; 0/13 placebo; 11/13 ketorolac; ketorolac; 0/12 6/12 etofenamate. etofenamate.

Diebshlag & Salicyclic acid, Ankle sprain, n = 80 10–15 cm 1.Ankle joint Salicylic acid better than Not reported. Withdrawals and 5 Knocker, ointment; placebo, < 48 hours parallel group ointment volume measurement placebo for all measures. adverse effects: 1987 ointment 2, 3, 4, 8, 15 days twice daily 2.VAS pain on rest 0/40 salicylic acid; and movement 0/40 placebo.

Dreiser, Ibuprofen, 5% cream; Acute tendinitis, n = 64 4 cm cream VAS pain on rest, 1. Ibuprofen better than 0/32 ibuprofen; Withdrawals and 3 1988 placebo, cream < 1 month parallel group three times pressure, movement placebo (p < 0.01). 0/32 placebo. adverse effects: 7 days daily 2. Global improvement: 0/32 ibuprofen; 26/32 ibuprofen; 13/28 0/32 placebo. placebo.

Dreiser, Ketoprofen, 5% gel; Simple sprains n = 60 5 cm twice daily VAS PI on rest, Global improvement: 0/30 ketoprofen; Withdrawals and 5 1989 placebo, gel parallel group movement, patient 18/30 ketoprofen; 1/30 placebo. adverse effects: 7 days global 5/30 placebo. 0/30 ketoprofen; 0/30 placebo.

Dreiser, Niflumic acid, 2.5% Uncomplicated n = 60 5 g gel three VAS PI, investigator Patient global (improved 3/30 niflumic Withdrawals and 5 et al., 1990 gel; placebo, gel ankle sprains parallel group times daily PI, patient and or healed): 23/30 niflumic acid; 1/30 placebo. adverse effects: 7 days investigator global acid; 10/30 placebo. 0/30 niflumic acid; 0/30 placebo.

Dreiser, Furbiprofen, patch Ankle joint pain n = 131 Two patches VAS spontaneous Mean VAS significantly lower Not reported. Withdrawals: 4 et al., 1994 40 mg; placebo, after post-traumatic parallel group daily pain by patient with furbiprofen. Pain better 0/64 furbiprofen; patch strain 7 days than moderate at day 7: 0/66 placebo. 53/64 furbiprofen; 52/66 placebo.

Fantato & de Benzydamine, 3% Oedema and n = 52 Three times 1. 4-point verbal 1. Benzydamine better Not reported. Not reported. 5 Gregorio, cream; placebo, post-traumatic pain parallel group daily rating than placebo. 1971 cream 6 days 2. Investigator global 2. ≥ 50% fall in symptom with CAT scale score: 22/26 benzydamine; 14/26 placebo.

Frahm, Salicylic acid, cream Acute knee or n = 156 10 cm cream VAS PI on movement VAS PI significantly less on 0/78 salicylic acid; Withdrawals: 0/78 5 1993 2%; placebo, cream ankle sprains parallel group twice daily and at rest day 9 for salicylic acid 0/78 placebo. salicylic acid; 0/78 9 days cream. placebo.Adverse effects: 0/78 salicylic acid; 1/78 placebo.

Fujimaki, Piroxicam, 0.5% gel; Musculoskeletal pain n = 271 1 g three times 1. 4-point verbal 1. Both active treatments 1/92 piroxicam; Withdrawals: 0/92 3 et al., 1985 indomethacin, 1% gel; multicentre, daily rating scale pain on better than placebo in 6/90 indomethacin; piroxicam; 1/90 placebo, gel parallel group rest and movement producing marked 2/89 placebo. indomethacin; 0/89 7 days 2. patient global improvement. placebo.Adverse 2. Overall improvement effects: 0/92 better or much better: piroxicam; 2/90 44/92 piroxicam; 44/90 indomethacin; indomethacin; 40/89 placebo. 0/89 placebo.

Haig, 1986 Benzydamine, 3% Acute soft tissue n = 43 Six times daily 4-point verbal rating 1. Pain on movement 0/21 benzydamine; Withdrawals and 4 cream; placebo, injuries parallel group scale – spontanous improved on day 6: NSD – 0/22 placebo. adverse effects: cream 2, 4, 6 days pain and pain on 18/21 benzydamine; 0/21 benzydamine; movement 13/22 placebo. 0/22 placebo. 2. Spontanous pain improved on day 6: significantly different – 20/21 benzydamine; 14/22 placebo.

Julien, 1989 Ketoprofen, 0.5% gel; Tendinitis n = 60 5 cm twice 1. Patient VAS PI on Overall patient assessment 1/30 ketoprofen; Withdrawals and 4 placebo, gel parallel group daily rest and movement (recovery, improvement) on 0/30 placebo. adverse effects: 7 days 2. Overall patient day 7: 25/30 ketoprofen; 0/30 ketoprofen; assessment 13/30 placebo. 0/30 placebo.

112 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 34 contd Placebo-controlled trials in acute painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Drug-related Quality study design regimen measures results withdrawals and score and follow-up adverse effects

Kockelbergh, Ketoprofen, 2.5% Acute soft tissue n = 74 7.5 g gel twice 1. 4-point verbal 1. Ketoprofen better than 1/38 ketoprofen; Withdrawals and 2 et al., 1985 gel; placebo, gel injuries, < 1 week parallel group daily rating VAS PI placebo in producing 1/36 placebo. adverse effects: baseline and 2. Global improved symptoms. 0/38 ketoprofen; 1 week 2. Global, patient, good: 0/36 placebo. 30/38 ketoprofen; 22/36 placebo.

Kockelbergh, Ketoprofen, 2.5% Acute low back pain, n = 40 15 g gel with 1.VAS pain and 1. Significantly more 4/20 ketoprofen; Withdrawals and 3 et al., 1985 gel; placebo, gel < 10 days parallel group physiotherapy 5-point verbal rating patients with moderate/ 1/20 placebo. adverse effects: 2 weeks and ultrasound; 2. 3-point patient severe pain at end in 1/20 ketoprofen; 10 sessions global rating placebo group. 0/20 placebo. 2. Global rating good: 13/20 ketoprofen; 9/20 placebo.

Lester, 1983 Salicylic acid, 2% Sprained ankle n = 42 Not stated 1.Ankle movement Pain relieved by 7 days: 0/20 salicylic acid; Not reported. 3 cream; placebo, parallel group 2. Swelling 18/20 salicylic acid; 2/22 placebo. cream 7 days 3. Pain 13/22 placebo. 4. Return to normal activity

Linde, et al., Benzydamine, 5% Sprained ankle n = 100 Three times 1. Swelling Significant reduction in Not reported. Not reported. 2 1985 cream; placebo, parallel group daily 2. Pain on walking swelling with benzydamine; cream 8 days 3. Fit for work NSD for pain. Free of walking pain on day 8: 35/50 benzydamine; 40/50 placebo.

McLatchie, Felbinac, 3% gel; Acute soft tissue n = 231 3 cm gel three 1.VAS PI rest, Good/very good treatment 3/118 felbinac; Withdrawals and 4 et al., 1989 placebo, gel injury parallel group times daily movement, night pain response (physician 2/113 placebo. adverse effects: baseline and 2. Investigator global assessment): 85/118 felbinac; 0/118 felbinac; 7 days 46/113 placebo. 0/113 placebo.

Morris, Felbinac, 3% gel; Acute soft tissue n = 100 1 cm gel three 1. Multiple global 1. Felbinac better than 0/50 felbinac; Withdrawals and 4 et al., 1991 placebo, gel sports injuries multicentre, times daily rating placebo. 0/50 placebo. adverse effects: parallel group 2.VAS pain 2. Patients with good/very 0/50 felbinac; baseline and good results (patient 0/50 placebo. 7 days global): 23/50 felbinac; 13/50 placebo.

Noret, Ketoprofen, 2.5% Minor sports n = 98 7.5 g gel twice 1.VAS PI 1. Ketoprofen better than 1/51 ketoprofen; Withdrawals and 3 et al., 1987 gel; placebo, gel injuries multicentre, daily 2. 4-point pain on placebo on many indices. 0/47 placebo. adverse effects: parallel group pressure 2. Global patient good or 1/51 ketoprofen; 1, 3, 8 days 3. Global better: 39/51 ketoprofen; 0/47 placebo. 9/47 placebo.

Parrini, Ketoprofen, 15% Soft tissue injuries n = 169 2 g three times 1. Categorical scale 1. Ketoprofen better 0/83 ketoprofen; Withdrawals and 4 et al., 1992 foam; placebo, foam parallel group daily for spontaneous than placebo for pain on 0/86 placebo. adverse effects: 7 days pain, on movement pressure, movement 0/83 ketoprofen; 2. Categorical scale and at rest. 0/86 placebo. global physician 2. Global physician excellent/good: 67/83 ketoprofen; 38/86 placebo.

Ramesh, Ibuprofen, 5% Soft tissue trauma n = 80 5–10 cm three 1. Pain on rest, Pain on movement on 1/40 ibuprofen; Withdrawals and 4 et al., 1983 cream; placebo, parallel group to four times pressure and move- day 7 none/slight: 28/40 1/40 placebo. adverse effects: cream 0, 3, 7, 10 days daily ment; 4-point scale ibuprofen; 16/40 placebo. 0/40 ibuprofen; 2. Investigator global 0/40 placebo. 3-point scale

Russell, Piroxicam, 0.5% gel; Soft tissue injuries n = 214 1 g four times 1.VAS pain on rest 1. Piroxicam better than 4/102 piroxicam; Withdrawals: 5 1991 placebo, gel parallel group daily and movement placebo at reducing pain 10/102 placebo. 1/102 piroxicam; 7 days up to 2. Global 4-point by day 8. 8/102 placebo. 21 days scale 2. Better joint mobility Adverse effects: 3. Daily pain charts with piroxicam. 4/102 piroxicam; 3. Global assessment good/ 7/102 placebo. excellent: 79/100 piroxicam; 45/100 placebo.

continued 113 Topically-applied NSAIDs

TABLE 34 contd Placebo-controlled trials in acute painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Drug-related Quality study design regimen measures results withdrawals and score and follow-up adverse effects

Sanguinetti, Biphenyl acetic acid, Soft tissue traumas n = 82 Three times Various scales Global patient good/very 0/42 biphenyl Adverse effects: 4 1989 3% gel; placebo, gel parallel group daily good: 34/42 biphenyl acetic acetic acid; 0/42 biphenyl acetic 7 days acid; 11/40 placebo. 0/40 placebo. acid; 0/40 placebo.

Sinneger & Fentiazac, 5% cream; Sport microtrauma n = 20 Twice or three Pain at rest, pressure, TOTPAR achieved 0/10 fentiazac; Withdrawals and 2 Blanchard, placebo, cream parallel group times daily movement by within 10 days: 7/10 0/10 placebo. adverse effects: 1981 10 days physician fentiazac; 1/10 placebo, 0/10 fentiazac; 0/10 placebo.

Taboada, Piroxicam, gel; Acute musculo- n = 40 Dose of drug Patient global Excellent or good: Not reported. Not reported. 2 1992 placebo, gel skeletal pain parallel group and duration 16/20 piroxicam; 5–10 not stated; gels 6/20 placebo. applications used with ultrasound and infrared treatment

Thorling, Naproxen, 10% gel; Sport injuries n = 120 2–6 times 1. Physician scoring Global patient, good/very 1/60 naproxen; Withdrawals and 3 et al., 1990 placebo, gel parallel group daily of pain at rest, good on day 7: 38/60 0/60 placebo. adverse effects: 7 days movement, swelling naproxen; 27/60 placebo. 0/60 naproxen; 2. Patient and 0/60 placebo. physician global

Vecchiet & Meclofenamic acid, Soft tissue injuries n = 60 4 g twice daily 1. Categorical scale 1. Meclofenamic acid 0/30 meclofenamic Withdrawals and 3 Colozzi, 5% gel; placebo, gel parallel group for spontaneous pain, better than placebo. acid; 0/30 placebo. adverse effects: 1989 5, 10 days pain on movement 2. Global patient: 30/30 0/30 meclofenamic 2. Patient and meclofenamic acid; acid; 0/30 placebo. physician global 19/30 placebo.

Wanet, Diethylamine Traumatic n = 56 Three times Pain on rest and Global assessment at end Not reported. Not reported. 3 et al., 1979 salicylate; placebo, rheumatological parallel group daily movement of treatment, good/very gel injuries 15 days good: 20/32 salicylate; 9/24 placebo.

Zerbi, Ketoprofen, foam; Painful traumatic n = 154 Twice daily 1. Pain at rest, under 1. Both active formulations 2/46 ketoprofen Withdrawals: 0/46 2 et al., 1992 ketoprofen, gel; injuries parallel group application, pressure, movement significantly better than foam; 0/49 ketoprofen foam; placebo, foam 7 days equivalent to 2. Global evaluation placebo. ketoprofen gel; 0/49 ketoprofen gel; 200 mg each 2. Global patient (positive 0/42 placebo. 1/42 placebo. No time result): 33/46 foam; 35/49 adverse effects gel; 13/42 placebo. withdrawals.

segment favouring treatment over placebo. The and benzydamine were no better than placebo three trials which did not have dichotomous out- (Table 38). comes also reported statistical benefit for topical NSAID over placebo. The percentage of patients achieving at least 50% pain relief with active treatment or placebo in all Pooled relative benefit for all 37 comparisons was studies in all trials (placebo and active drug con- 1.7 (95% CI, 1.5–1.9) and the NNT was 3.9 (95% trolled) in acute conditions is shown in Figure 35 CI, 3.4–4.3) (Table 38). Pooling data only from trials (lower panel). The range with placebo was from with a quality score of at least 3 produced the same 0% to 80%. With topical NSAID it was from 30% results. Sensitivity analysis by treatment group size to 100%. There was no significant difference in showed that trials with a group size of at least 40 the (low) frequency of local or systemic adverse treated patients produced higher (worse) estimates effects, or drug-related withdrawal (Table 38). for NNT of 4.8 (95% CI, 4.0–5.7) than all trials together. Trials with fewer than 40 treated patients Chronic conditions produced a significantly lower (better) NNT of The 13 placebo-controlled trials (see Tables 36 and 2.6 (95% CI, 2.3–3.1) than either larger trials 37) were predominantly in single joint arthritis and or all trials. rheumatological disorders, with dichotomous outcomes from 547 patients on active drug treatment Pooling data for each drug studied in three or and 550 on placebo in 12 trials. In 12 other trials more trials showed ketoprofen, felbinac, ibuprofen different NSAIDs were compared in 1272 patients. and piroxicam to be statistically superior to placebo In two of these trials, topical NSAIDs were 114 with NNTs ranging from 2.6 to 4.2. Indomethacin compared with oral. Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 35 Active drug controlled trials in acute painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Withdrawals Quality study design regimen measures results and adverse score and follow-up effects

Arioli, Piroxicam, 1% cream; Acute musculo- n = 75 1 g piroxicam 1. Categorical and 1. Piroxicam better than 0/38 piroxicam; 0/38 piroxicam; 2 et al., 1990 diclofenac, 1% gel skeletal disorders parallel group, cream, 4 g VAS scales for pain on diclofenac for some 0/37 diclofenac. 0/37 diclofenac. open design diclofenac gel, movement, at rest, etc. measures. 3, 7, 14 days four times daily 2. Patient global 2. Patient global 7 days (better/much better): 34/38 piroxicam; 27/37 diclofenac.

Baixauli, Naproxen, 10% gel; Acute soft tissue n = 30 5 cm naproxen, 1. Patient and 1. Cured or improved 0/15 naproxen; Withdrawals and 3 et al., 1990 ketoprofen, 10% gel trauma < 24 hours parallel group 3–5 cm investigator global day 3: 10/15 naproxen; 0/15 ketoprofen. adverse effects: 3, 7 days ketoprofen, rating 5-point 12/14 ketoprofen. 0/15 naproxen; twice daily 2. Improved 3-point 2. Cured or improved 0/15 ketoprofen. day 7: 15/15 naproxen; 13/15 ketoprofen. 3. Patient global (good/ very good): 13/15 naproxen; 9/15 ketoprofen.

Bouchier- Diclofenac, 1% gel; Acute soft tissue n = 386 4 g gel three VAS pain on rest 1. Diclofenac better than Not reported. 0/191 diclofenac; 1 Hayes, et al., felbinac, 3% gel injuries multicentre, times daily pressure and felbinac for some measures. 2/195 felbinac. (probably 1990 parallel group movement 2. ≥ 50% improvement in not 3, 7 days pain on movement on day 7: double- 110/191 diclofenac; 100/195 blind) felbinac.

Butron, Naproxen, 10% gel; Sprains and n = 64 As required 1. VAS pain on rest, Patient global good/ 3/34 naproxen; 0/34 naproxen; 2 et al., 1994 diclofenac, 1% gel contusions parallel group movement excellent: 30/34 naproxen; 4/30 diclofenac. 0/30 diclofenac. 4 days 2. Patient and 28/30 diclofenac. physician global

Commandre, Niflumic acid, 2.5% Acute sprains n = 100 15 gm of each 1. Patient VAS 1. Niflumic acid significantly 3/51 niflumic acid; 0/51 niflumic acid; 2 et al., 1993 gel; piroxicam, 0.5% or tendinitis parallel group daily 2. Investigator better than piroxicam 4/49 piroxicam. 0/49 piroxicam. gel 7, 14 days categorical days 8 and 15. Adverse effects 3. Patient global 2. Patient global day 8: withdrawals: 41/51 niflumic acid; 1/51 niflumic acid; 23/49 piroxicam. 1/49 piroxicam.

Curioni, Ibuproxam; Soft tissue injuries n = 60 Twice daily 1. Pain – spontaneous, Some differences between 2/20 ibuproxam; No information. 4 et al., 1985 ketoprofen; parallel group application on palpation, groups. 3/20 ketoprofen; etofenamate 10 days movement 1/20 etofenamate. 2. Patient global

Diebschlag, Indomethacin, 1% Acute ankle sprain n = 42 Three times Swelling, pain No difference in swelling 0/19 (A); 1/22 (B). No differences. 3 et al., 1992 gel (A); indomethacin, 1 parallel group daily or pain between two % gel (B) (different 2 weeks preparations. vehicles) Patient global, excellent or good: 19/19 (A); 21/22 (B).

Gallachi, Diclofenac, 1% gel; Painful inflammatory n = 50 2 g four times 1. Spontaneous pain 1. NSD. 0/25 both groups. No data. 2 et al., 1990 diclofenac, 1.16% gel symptoms parallel group daily 2. Pain on pressure 2. NSD. 7, 14 days 3. Patient global 3. 19/25 both groups good/excellent.

Governali Ketoprofen, 5% gel; Soft tissue injuries n = 30 2–3 g of gel or 1. Pain – spontaneous, 1. Gel significantly better 0/15 ketoprofen Withdrawals and 2 & Casalini, ketoprofen, 1% parallel group cream three movement, pressure than cream. gel; 0/15 adverse effects: 1995 cream 7, 14 days times daily 2. Patient global 2. On day 7, excellent/ ketoprofen cream. 0/15 ketoprofen good: 14/15 ketoprofen gel; 0/15 gel; 9/15 ketoprofen cream. ketoprofen cream.

Gualdi, et al., Flunoxaprophene, Soft tissue injuries n = 60 3–5 cm of gel, 1. PI NSD between groups. 1/30 flunox- No information. 2 1987 gel; ketoprofen, gel parallel group twice daily 2. Function aprophene; 1, 4, 7, 10 days 3. Patient global 3/30 ketoprofen.

Hallmeier & Etofenamate, 10% Sports injuries n = 60 Not given 1. Oedema 0/30 etofenamate; Withdrawals: 2 Michelbach, gel plus dressing; parallel group 2. Erythema 2/30 heparin. 0/30 etofenamate; 1986 heparin/ dexpanthenol/ 4 days 3. Movement 0/30 heparin. dimethyl-sulphoxide 4. Patient global 4. Patient global: 26/30 ‘success’ etofenamate; 10/30 heparin.

Hallmeier, Etofenamate, 10% gel; Sprains and n = 60 2–4 times daily Patient global Patient global very good/ 0/30 etofenamate; Withdrawals and 1 1988 diclofenac, 1% gel contusions parallel group, good: 27/30 etofenamate; 0/30 diclofenac. adverse effects: single blind 13/30 diclofenac. 0/30 etofenamate; 7 days 0/30 diclofenac.

continued 115 Topically-applied NSAIDs

TABLE 35 contd Active drug controlled trials in acute painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Withdrawals Quality study design regimen measures results and adverse score and follow-up effects Hosie, 1993 Felbinac, 3% foam; Acute lower back n = 287 2 g gel three 1. Pain 5-point scale 1. No difference between 1/140 felbinac 25/140 felbinac 4 ibuprofen, 400 mg injury multicentre, times daily; 2. Investigator global groups in symptom severity. foam; 3/147 foam; 19/147 tablets parallel group, 1 tablet three 2. Both showed significant ibuprofen oral. ibuprofen oral. double-dummy times daily improvement. 7, 14 days 3. No/mild pain on movement at 14 days: 99/140 felbinac foam; 109/147 ibuprofen oral. Kroll, et al., Piroxicam, 0.5% gel; Sprains and n = 173 1 g piroxicam, 1. Patient score of Patient global excellent/ 4/84 piroxicam; Withdrawal: 2 1989 diclofenac, 1.16% gel tendinitis parallel group, 2–4 g diclofenac, pain on movement good: 63/84 piroxicam; 3/89 diclofenac. 2/84 piroxicam; open four times daily (21-point VAS) 62/89 diclofenac. 1/89 diclofenac. to 14 days 2. Patient global Adverse effects: 2/84 piroxicam; 0/89 diclofenac. Montagna, Meclofenamic acid, Painful musculo- n = 40 Prescribed 1. Pain – spontaneous 1. No statistical difference No data. No data. 1 et al., 1990 5% gel; naproxen, skeletal disorders parallel group amounts twice and on movement between groups. 10% gel 4, 8, 15 days daily 2. Patient global 2. Excellent/good on day 8: 13/20 meclofenamic acid; 10/20 naproxen. Oakland, Felbinac + placebo Acute injuries n = 220 1–2 g gel 1. Pain at rest NSD. 3/147 felbinac; 0/147 felbinac; 3 1993 ultrasound; placebo of lateral ankle parallel group two to three 2. Investigator global 3/73 placebo. 2/73 placebo. gel + ultasound; ligaments days 3, 5, 7 times daily felbinac + ultrasound Picchio, Ibuprofen, 10% gel; Acute sports n = 40 Three times 5-point pain for pain 1. Ibuprofen significantly Not reported. 0/20 ibuprofen; 3 et al., 1981 ketoprofen, 1% gel injuries parallel group daily at rest, on movement, better and faster than 0/20 ketoprofen. 4, 8, 12, 16 days spontaneous ketoprofen. 2. No pain on movement at 12 days: 16/20 ibuprofen; 10/20 ketoprofen. Pineda, Felbinac, 3% gel; Acute soft tissue n = 172 Felbinac, 1. Multiple, 10-point 1. Complete recovery at 5/86 felbinac; 1/86 felbinac; 1 et al., 1983 piroxicam gel, (?0.5%) injuries multicentre, 180 mg/day, pain on rest, 7 days: felbinac better than 1/86 piroxicam. 0/86 piroxicam. parallel group piroxicam, movement and piroxicam (p = 0.008). 3, 7 days 18 mg/day, night pain 2. Good/very good global: three times 2. Global 5-point 68/86 felbinac; 65/86 daily piroxicam. Rosemeyer, Diclofenac, 1% gel; Distortion of n = 91 10 cm diclofenac, 1. Pain at rest 1 and 2. NSD at any time 4/44 diclofenac; Adverse effects 4 1991 piroxicam, 0.5% gel ankle joint parallel group 3 cm piroxicam, 2. Pain on pressure 3. Patient global excellent/ 5/47 piroxicam. withdrawal: 0/44 3, 7, 10, 14 days four times daily 3. Patient global good: 35/44 diclofenac; diclofenac; 1/47 40/47 piroxicam. piroxicam. RPR1 Ketoprofen, gel; Acute soft tissue n = 1575 Ketoprofen, Patients’ global Greatly improved: 396/1048 Not reported. Not reported. 2 piroxicam, gel; injury parallel group 4–5 g, piroxicam assessment of injury ketoprofen; 69/263 diclofenac, gel 5 days 1 g, diclofenac, piroxicam; 80/264 2–4 g, three times diclofenac. daily for 5 days Selligra & Naproxen, 10% gel; Aoft tissue injuries n = 100 2–6 times daily Patients’ global Good/very good: 31/49 1/49 naproxen; Withdrawals: 2 Inglis, 1990 flufenamic acid, 3% gel parallel group, naproxen; 28/51 flufenamic 0/51 flufenamic 1/49 naproxen; single-blind acid. acid. 0/51 flufenamic 7 days acid. Adverse effects: 0/49 naproxen; 0/51 flufenamic acid. Sugioka, Piroxicam, 0.5% gel; Non-traumatic n = 366 1 g three or Multiple pain 4-point. Piroxicam better than 1/183 piroxicam; Withdrawals: 4 et al., 1984 indomethacin, 1% gel disease of muscle multicentre, four times daily 7-point symptom indomethacin. 12/183 indo- 1/183 piroxicam; or tendon parallel group improvement Patient self-assessment methacin. 12/183 indomethacin. 1, 2 weeks better/much better: Adverse effects: 85/183 piroxicam; 6/183 piroxicam; 55/183 indomethacin. 26/183 indomethacin. Tonutti, Ketoprofen, 5% gel; Soft tissue trauma n = 30 2–3 grams gel 1. Pain, spontaneous, 1. Comparable efficacy. 0/15 ketoprofen; 0/15 ketoprofen; 4 1994 etofenamate, 5% gel parallel group three times on movement, 2. Day 7 good/excellent: 0/15 etofenamate. 0/15 etofenamate. 7 days daily for up to pressure 10/15 ketoprofen; 11/15 3 weeks 2. Patient global etofenamate. Vander- Etofenemate, 10% Strains and sprains n = 60 5 cm gel, Categorical scales for 1. Day 7 no/slight pain: 1/30 etofenemate 0/30 etofenemate 2 straeten & gel; naproxen, of lower limbs parallel group 1 tablet, three spontaneous pain and 13/30 etofenemate gel; gel; 0/30 naproxen gel; 6/30 naproxen Scheumans, 275 mg tablets within 3 days 7, 17 days times daily pain on palpation 15/30 naproxen oral. oral. oral. Withdrawals: 1990 2. Clinical global good/ 1/30 etofenemate excellent improvement: gel; 2/30 naproxen 12/30 etofenemate gel; oral. 116 13/30 naproxen oral. Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 36 Placebo controlled trials in chronic painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Withdrawals Quality study design regimen measures results and adverse score and follow-up effects Algozzini, Trolamine salicylate, Osteoarthritis n = 26 3.5 g cream, 1. 4-point PI 1. NSD. 0/26 salicylate; Withdrawals and 4 et al., 1982 10% cream; placebo, of knee crossover four times daily 2. Numerical rating 2. Patient preference: 0/26 placebo. adverse effects: cream 1 week 0–10 8/26 salicylate; 6/26 placebo; 0/26 salicylate; 11/26 no preference. 0/26 placebo, 3. PR from diaries: 9/26 salicylate; 6/26 placebo. Bolten, 1991 Felbinac, 3% gel; Acute extra-articular n = 281 1 g, 3 times 1. Categorical and 1. Felbinac significantly 2/142 felbinac; Withdrawals and 3 placebo, gel rheumatic disorders parallel group daily VAS on rest and better than placebo. 4/139 placebo. adverse effects: 0, 7, 14 days movement 2. Global estimation of 0/142 felbinac; 2.VAS good/very good responses 0/139 placebo. 3. Investigator (p < 0.001): 67/142 felbinac; global rating 39/139 placebo. Camus, Diethylamine Rheumatic disorders n = 20 Three times 4-point verbal rating 1. Salicylate better than 0/10 salicylate; Withdrawals and 2 1975 salicylate, cream; parallel group daily placebo in giving relief 0/10 placebo. adverse effects: placebo, cream 10 days over 10 days. 0/10 salicylate; 2. Pain reduced: 8/10 0/10 placebo. salicylate; 3/10 placebo. Dreiser & Diclofenac, plasters; Osteoarthritis n = 155 Applied twice 1.VAS 1. Diclofenac better than 1/78 diclofenac; Withdrawals: 0/78 4 Tisne- placebo, plasters of knee parallel group daily (each 2. Global rating placebo from day 4 3/77 placebo. diclofenac; 0/77 Camus, 4, 7, 15 days plaster contain- 5-point 2. Global rating excellent/ placebo. Adverse 1993 ed 180 mg diclo- good: 55/78 diclofenac; effects: 0/78 diclofenac derivative) 21/77 placebo. fenac; 1/77 placebo. El-Hadidi & Diclofenac, gel; Painful rheumatic n = 120 Three times Physician judgement Diclofenac significantly 2/60 diclofenac; Withdrawals: 3 El-Garf, ultrasound coupling conditions parallel group per week plus VAS PI by patient better than regular coupling 1/60 regular. 1/60 diclofenac; 1991 gel 4 weeks at rest and on gel on all measures. 0/60 regular. movement Complete PR on passive Adverse effects: Patient global movement at 2 weeks: 26/60 0/60 diclofenac; diclofenac; 18/60 regular. 0/60 regular. Fotiades & Flufenamate, 3%, plus Cervical, lumbar and n = 100 Three or four Point-scoring system Scoring very good/good: 0/48 active drug; Withdrawals and 3 Bach, 1976 salicylate, 2%, gel; shoulder pain and parallel group times daily for including pain at rest, 43/48 active drug; 0/52 placebo. adverse effects: placebo, gel gonarthroses up to 20 days 6–20 days on pressure, pain 26/52 placebo. 0/48 active drug; relief, muscle spasm 0/52 placebo. and movement Galiazzi & Diclofenac, plaster Rheumatological n = 60 Applied twice 1. Multiple 4-point 1. Diclofenac better than 0/30 diclofenac; 0/30 diclofenac; 3 Marcolongo, (slow-release); disorders parallel group daily (each verbal rating and VAS placebo in reducing pain. 0/30 placebo. 0/30 placebo. 1993 placebo, plaster 3, 5, 7, 14 days plaster con- 2. Investigator 2. Assessment of good/ tained 180 mg global scale excellent response: 26/30 diclofenac diclofenac; 2/30 placebo. derivative) Ginsberg & Indomethacin, 4% Tendinitis n = 30 2–4 sprays 1.VAS 1. Indomethacin better 2/30 indomethacin; 0/30 indomethacin; 2 Famaey, spray; placebo, spray crossover 3–5 times daily 2. 4-point verbal than placebo on various 0/30 placebo. 0/30 placebo. 1991 2 x 2 weeks lightly massaged rating pain indices. into skin 2. Subjective improvement: 26/30 indomethacin; 18/30 placebo. Gui, et al., Ibuprofen, cream; Osteoarthritis n = 40 Application Spontaneous pain and 1. Improved spontaneous 0/19 ibuprofen; 0/19 ibuprofen; 3 1982 placebo, cream parallel group twice daily pain on pressure and pain: 17/19 ibuprofen; 0/20 placebo. 0/20 placebo. 21 days movement 9/20 placebo. 2. Improved pain on movement: 14/19 ibuprofen; 7/20 placebo. Hohmeister, Flufenamate, 3%, Cervical and lumbar n = 100 Three times Symptom improve- Complete PR at 8/49 active gel; Withdrawals: 4 1983 plus salicylate, 2%, back pain parallel group daily ment, complete PR 21 days: 38/49 active gel; 0/51 placebo. 0/49 active gel; gel; placebo, gel 7, 14, 21 days 3/51 placebo. 0/51 placebo. Mattara, Flurbiprofen, 40 mg Scapulo-humoral n = 80 Twice daily VAS PI for extension, Day 14 no pain or slight 4/40 flurbiprofen; Withdrawals: 4 et al., 1994 patch; placebo, patch periarthritis parallel group flexion and abduction pain: 14/40 flurbiprofen; 1/40 placebo. 0/40 flurbiprofen; 14 days 13/40 placebo. 0/40 placebo. Adverse effects: 5/40 flurbiprofen; 2/40 placebo. Rose, et al., Piroxicam, 0.5% gel; Gonarthrosis n = 30 1 g gel, four 1. Pain on movement 1. No pain 7/15 piroxicam; 1/15 piroxicam; 0/15 piroxicam; 2 1991 placebo, gel parallel group times daily 2. Pain at rest 2/15 placebo. 1/15 placebo. 0/15 placebo. 14 days 3. Patient global 2. Excellent/good: 8/15 piroxicam; 5/15 placebo. Roth, 1995 Diclofenac, gel; Osteoarthritis n = 119 Four times Overall pain NSD. 12/59 diclofenac; Adverse effects 4 placebo, gel breakthrough pain parallel group daily for 26/60 placebo. withdrawals: 3/59 14 days 2 weeks diclofenac; 4/60 placebo. 117 Topically-applied NSAIDs

TABLE 37 Active controlled trials in chronic painful conditions: trial design, outcome measures and results

Study Drug(s) Condition Numbers, Dosing Outcome Analgesic outcome Skin irritation Withdrawals Quality study design regimen measures results and adverse score and follow-up effects

Ammer, Diclofenac, gel; Soft tissue n = 227 2–4 days per 1. Pain at rest and 1. NSD. Adverse effects 2 1991 indomethacin, 1% gel rheumatism with parallel group week on movement 2. Good/excellent: 76/89 withdrawal: pain of medium 14 days 2. General efficacy diclofenac; 62/84 1/89 diclofenac; intensity indomethacin. 0/84 indomethacin.

Balthazar- Diclofenac, gel; Rheumatological n = 50 Twice daily 1. Symptom intensity Improved at 14 days: 0/25 diclofenac; Withdrawals and 4 Letawe, indomethacin, gel disorders parallel group 3-point scale 15/25 diclofenac; 0/25 indomethacin. adverse effects: 1987 7, 14 days 2. Investigator global 17/25 indomethacin. 0/25 diclofenac; 0/25 indomethacin.

Browning & Normal oral NSAID; Mild to moderate n = 191 Three to four 1. Patients’ 1. Significant reduction in 1/106 piroxicam. Withdrawal: 2 Johson, half normal oral osteoarthritis parallel group, times daily, assessment of pain mean score for tenderness 1/106 piroxicam. 1994 plus piroxicam open study piroxicam and stiffness and restiction of active Adverse effects: 14, 28 days 2. Patients’ overall movement for topical 1/106 piroxicam; assessment of NSAID. 1/85 oral alone. efficacy day 2. Patients’ overall assessment of efficacy day, excellent/good: 54/85 oral alone; 71/106 oral plus topical NSAID.

Dickson, Piroxicam, 0.5% gel; Chronic osteo- n = 235 1 g gel three 1. Pain 9-point scale NSD between treatments; 3/117 piroxicam; Adverse effects: 4 1991 oral ibuprofen arthritis of knee multicentre, times daily; 2.Analgesic patient global rating good/ 4/118 ibuprofen. 30/117 piroxicam; parallel group, 400 mg consumption better: 68/117 piroxicam; 27/118 ibuprofen. double-dummy ibuprofen three 3. Global 4-point 65/118 ibuprofen. Withdrawals: 4 weeks times daily scale 9/117 piroxicam; 7/118 ibuprofen.

Geller, 1980 Diethylamine Chronic disorders n = 50 Not recorded 1. Pain at rest and 1. Diethylamine salicylate Two local effects Not reported. 2 salicylate, 10% gel; crossover in movement, better than etofenamate but drug respons- etofenamate, 5% gel 7 days; 4-day 4-point scale on all scores. ible not given. washout 2. Global 5-point 2. After first phase, good/ scale patient very good results patient global: 24/25 salicylate; 8/25 etofenamate.

Giacovazzo, Diclofenac, gel; Osteoarthritis n = 40 Diclofenac, VAS PI No difference between two 0/20 diclofenac; 0/20 diclofenac; 1 1992 felbinac, gel (biphenyl parallel group 160 mg/day, treatments. Improvement in 0/20 felbinac. 0/20 felbinac. acetic acid) 1 week felbinac, pain scores: 14/20 diclo- 90 mg/day, fenac; 14/20 felbinac. three times daily

Golden, Triethylamine salicy- Rheumatic pain n = 40 Application of Daily diaries, Good/excellent results: 1/20 salicylate 2/20 salicylate 3 1978 late, 10% cream; oral parallel group, cream four categorical scales 13/20 salicylate cream; cream; 1/20 oral cream; 6/20 oral aspirin, 325 mg tablet double-dummy times daily; two 10/20 oral aspirin. aspirin. aspirin. 7 days tablets four times daily

Matucci- Ketoprofen, 2.5% Soft tissue n = 36 Twice daily VAS PI and Ketoprofen better than 0/18 ketoprofen; 0/18 ketoprofen; 2 Cerinic & gel; etofenemate, rheumatic disorders parallel group tenderness etofenemate for pain on 0/18 etofenemate. 0/18 etofenemate. Casini, 1988 5% gel 3, 7 days active and passive movement.

Reginster, Indomethacin, 1% Rheumatoid arthritis n = 20 Three times % improvement on Both improved significantly 2/20; 2/20. Withdrawals and 2 et al., 1990 gel; indomethacin, crossover daily, 100 mg swelling and pain at from baseline. adverse effects: 4% spray 14 days daily total rest on flexion 0/20; 0/20.

Ritchie, Flurbiprofen, patch; Soft tissue n = 131 Flurbiprofen, Pain, tenderness Statistically more PR with Adverse effects 3 1996 piroxicam, 0.5% gel rheumatism of crossover at 40 mg patch, flurbiprofen. withdrawals: shoulder or elbow 4 days twice daily; 3 cm 1/133 flubiprofen; 4, 8, 14 days piroxicam gel 3/133 piroxicam. four times daily

Rosenthal & DHEP,1% plaster; Periarticular, n = 190 Plaster twice Spontaneous pain, Patient global, good/ 2/96 plaster; Withdrawals and 3 Bahous, diclofenac, 1% gel tendinous parallel group daily; gel four pain on pressure, excellent: 79/96 plaster; 3/94 gel. adverse effects: 1993 inflammations 14 days times daily patient global 39/94 gel. 0/96 plaster; 0/94 gel.

Vitali, 1980 Ketoprofen, 1, 2.5 Orthopaedic n = 62 5–15 cm twice Spontaneous pain, 1. 2.5% gel was most useful. Not reported. Not reported. 3 and 5% gel parallel group daily for palpation, movement 2. Spontaneous pain 14 days 6–13 days better/much better: 13/20 1% gel; 18/20 2.5% gel; 16/20 118 5% gel. Health Technology Assessment 1998; Vol. 2: No. 12

Diethyl salicylate Successful outcome with topical Salicylic acid non-steroidal (%) Benzydamine 100 Benzydamine Benzydamine Benzydamine Etofenamate 75 Felbinac Felbinac Felbinac Fentiazac 50 Flurbiprofen Ibuprofen Ibuprofen Ibuprofen Ibuprofen 25 Indomethacin Indomethacin Indomethacin Ketoprofen 0 Ketoprofen Ketoprofen 0 25 50 75 100 Ketoprofen Ketoprofen Successful outcome with topical Ketoprofen placebo (%) Ketoprofen Ketoprofen Ketoprofen FIGURE 34 Placebo-controlled trials of topical NSAIDs in acute Ketorolac (◆) and chronic (●) conditions Meclofenamic acid Naproxen Niflumic acid Niflumic acid single topical NSAID was tested in as many as Piroxicam three placebo-controlled studies and combined Piroxicam estimates could not therefore be calculated for Piroxicam Piroxicam any single drug. Combined The percentage of patients achieving at least 50% 0.1 1 10 100 pain relief with active drug treatment or placebo RB for successful outcome in all studies in all trials (placebo and active drug controlled) in chronic conditions is shown in Figure 35 (upper panel). The range with placebo FIGURE 33 Placebo-controlled trials of topical NSAIDs: one was from 5% to 60%. With topical NSAIDs it was week outcome in acute painful conditions from 30% to 95%. There was no significant difference in the (low) frequency of local or systemic Relative benefit and 95% CIs for each drug adverse effects, or drug-related withdrawal compared with placebo are shown in Figure 36. (Table 38). Seven of the 12 studies showed statistical superiority for topical NSAIDs over placebo. The scatter Comparison with oral NSAIDs of the proportion of patients with at least 50% pain Five studies compared topical with oral NSAIDs, relief with topical NSAID or placebo is shown in three in acute11,12,13 and two in chronic condi- Figure 34. All 12 comparisons were in the segment tions.14,15 None showed statistical benefit of oral favouring treatment over placebo. The one trial over topical NSAIDs. which did not have dichotomous outcomes also reported statistical benefit for topical NSAIDs over placebo. Comments

The pooled relative benefit for all 12 comparisons Topical NSAIDs were significantly more effective was 2.0 (95% CI, 1.5–2.7) and the NNT was 3.1 than placebo. This is not just due to rubbing. (95% CI, 2.7–3.8) (Table 38). Sensitivity analysis Placebo preparations were also rubbed on affected by quality score or treatment group size produced parts. The significant difference was therefore no significant change in these estimates. No additional to any effect of rubbing. Topical pre- 119 Topically-applied NSAIDs

TABLE 38 Combined results and sensitivity analysis for topical NSAIDs in acute and chronic painful conditions

Drug Trials Patients Average CER EER RR NNT number of treated patients

Acute painful conditions

Combined efficacy data 37 3239 47 39 71 1.7 (1.5–1.9) 3.9 (3.4–4.4)

Local adverse effects 3 2.6 1.2 (0.8–1.7)

Systemic adverse effects 0.7 0.8 1.0 (0.6–1.8)

Withdrawal due to adverse effects 0.4 0.6 0.8 (0.4–1.4)

Trials with quality score 3–5 only 30 2834 52 38 72 1.7 (1.5–1.9) 3.9 (3.4–4.4)

Trials with treatment groups of < 40 patients 20 933 24 35 76 1.9 (1.6–2.2) 2.6 (2.3–3.1)

Trials with treatment groups of 40–80 patients 8 810 51 44 66 1.6 (1.1–2.2) 5.0 (3.7–7.4)

Trials with treatment groups of > 80 patients 7 1496 123 41 67 1.6 (1.3–1.9) 4.6 (3.7–5.9)

Ketoprofen 9 724 43 36 74 2.0 (1.5–2.6) 2.6 (2.3–3.2)

Felbinac 3 413 70 32 66 2.0 (1.5–2.7) 3.0 (2.4–4.1)

Ibuprofen 4 284 36 34 70 1.9 (1.2–3.0) 3.5 (2.5–5.6)

Piroxicam 4 589 74 39 69 1.6 (1.2–2.2) 4.2 (3.1–6.1)

Benzydamine 4 245 31 62 84 1.4 (0.9–2.0) 6.7 (3.8–23)

Indomethacin 3 394 66 32 47 1.3 (0.9–1.8) 10 (5.0–∞)

Chronic painful conditions

Combined efficacy data 12 1097 30 65 2.0 (1.5–2.7) 3.1 (2.7–3.8)

Local adverse effects 5.3 5.9 0.9 (0.4–1.7)

Systemic adverse effects 1.3 1.1 1.1 (0.5–2.3)

Withdrawal due to adverse effects 0.7 0.7 1.0 (0.4–2.4)

Trials with quality score 3–5 only 9 987 55 27 62 2.2 (1.5–3.1) 3.1 (2.6–3.8)

Trials with treatment groups of < 40 patients 6 261 22 31 69 2.2 (1.5–3.1) 2.6 (2.0–3.6)

Trials with treatment groups of > 40 patients 6 836 70 29 61 2.0 (1.7–2.4) 3.3 (2.8–4.3)

Response is either the proportion of patients with successful outcome or percentage of patients with an adverse effect. An infinite NNT CI indicates that there may be no benefit from the treatment compared with placebo.

parations produced NNTs in the range 3–5 (Table studies in both acute and chronic conditions 38). At least one patient in about three using a had quality scores of 3 or more on a scale of 1–5 topical NSAID will achieve at least 50% pain relief in over 75% of reports (see Table 33). This is who would not have done had they used placebo. important, since trials of lower methodological quality (2 or less using the same validated scale While this result may surprise some, it is not as here) have been shown to have a more 120 because the trials were poor. Placebo-controlled favourable outcome.16 Health Technology Assessment 1998; Vol. 2: No. 12

Etofenamate (1) Chronic Flurbiprofen (1) Ibuprofen (1) Flufenamate plus salicylate (2) Ketoprofen (2) Felbinac (2) Piroxicam (3) Indomethacin (3) Diclofenac plaster (3) Salicylate (4) Diclofenac gel (5) Placebo (12)

Ketorolac (1) Acute Flurbiprofen (1) Flufenamic acid (1) Fentiazac (1) Meclofenamic acid (2) Saliscylate (2) Niflumic acid (3) Felbinac (4) Benzydamine (4) Etofenamate (5) Naproxen (4) Indomethacin (4) Ibuprofen (5) Felbinac (5) Diclofenac (7) Piroxicam (11) Ketoprofen (11) Placebo (35)

0 20406080100

Patients with successful outcome (%)

FIGURE 35 Successful outcome for treatment arms from placebo- and active controlled trials in chronic and acute painful conditions – drug and number of trials

It was judged sensible to pool only data for withdrawal from the study were also rare (0.5%). individual drugs when there were at least three Adverse effects were no more common than RCTs. In acute conditions there was enough with placebo. information to make comparisons (Table 38). The average response for placebo was similar for Topical NSAIDs are less associated with the gastro- individual drugs apart from benzydamine. Keto- intestinal adverse effects seen with the same drugs profen, felbinac, ibuprofen and piroxicam were taken orally.17 The low incidence of systemic all statistically superior to placebo, in contrast to adverse effects for topical NSAIDs probably results indomethacin and benzydamine which were not. from the much lower plasma concentrations from CIs for the NNT for ketoprofen did not overlap similar doses applied topically to those adminis- with those of benzydamine or indomethacin. tered orally.13,18 Topical application of ibuprofen There is no clear message as to which of ketopro- resulted in significant tissue concentrations in deep fen, felbinac, ibuprofen or piroxicam was best, tissue compartments, more than enough to inhibit or indeed whether there was any difference in inflammatory enzymes.18,19 efficacy. They all work. These positive results for topical NSAIDs could, it Local skin reactions were rare (3.6%) and systemic may be argued, be skewed by publication restricted effects were rarer (less than 0.5%). Local or syste- to positive findings. It is next to impossible to rebut mic adverse effects of sufficient severity to cause this argument. Strenuous efforts were made to 121 Topically-applied NSAIDs

patients, who may benefit from using topical rather than oral NSAIDs. We need to compare the pain relief and mobility, harm and cost for these alter- Trolamine salicylate natives. The few studies identified which compared Diethylamine salicylate oral with topical NSAIDs were of inadequate design and power to answer these important Diclofenac plaster questions. In the meantime, the message is that topical NSAIDs are effective and safe. Diclofenac plaster

Diclofenac gel References Felbinac 1. Wynne HA, Campbell M. Pharmacoeconomics of Flufenamate + salicylate nonsteroidal anti-inflammatory drugs (NSAIDs). PharmacoEconomics 1993;3:107–23. Flufenamate + salicylate 2. Bateman DN, Kennedy JG. Non-steroidal anti- Flurbiprofen inflammatory drugs and elderly patients. The medicine may be worse than the disease. BMJ Ibuprofen 1995;310:817–18. Indomethacin 3. Rational use of NSAIDs for musculoskeletal disorders. Drug Therapeut Bull 1994;32:91–5. Piroxicam 4. Jadad AR, Carroll D, Moore A, McQuay H. Combined Developing a database of published reports of randomised clinical trials in pain research. 0.1 1 10 100 Pain 1996;66:239–46. RB for successful outcome 5. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJM, Gavaghan DJ, et al. Assessing the quality of reports of randomized clinical trials: is FIGURE 36 Placebo-controlled trials of topical NSAIDs: 2-week blinding necessary? Control Clin Trials 1996;17:1–12. outcome in chronic painful conditions 6. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in clinical research. Ann Intern Med 1987;107:224–33. unearth unpublished data. Ironically, one pharma- 7. DerSimonian R, Laird N. Meta-analysis of clinical ceutical company withheld results they claimed to be trials. Control Clin Trials 1986;7:177–88. positive and favourable to their product. Rosenthal’s 8. Yusuf S, Peto R, Lewis J, Collins R, Sleight P. Beta- 20 file drawer argument says there would need to be blockade during and after myocardial infarction: an many negative results (more than 692 for acute, 37 overview of the randomized trials. Prog Cardiovasc for chronic) to overturn these positive results. Dis 1985;27:335–71. 9. Laupacis A, Sackett DL, Roberts RS. An assessment More important is the empirical evidence that of clinically useful measures of the consequences of small trials (arbitrarily set at fewer than 40 patients treatment. N Engl J Med 1988;318:1728–33. per group as being between the mean and median sizes of 47 and 32 patients per treated group) 10. Cook RJ, Sackett DL. The number needed to treat: produced exaggerated estimates of clinical efficacy a clinically useful measure of treatment effect. BMJ 1995;310:452–4. by 24% (4.8 minus 3.9/3.9, Table 38) with CIs which did not overlap. By contrast, trial quality made no 11. Åkermark C, Forsskåhl B. Topical indomethacin in difference despite evidence to the contrary from overuse injuries in athletes. A randomized double other settings.16 Size of treatment group may be blind study comparing Elmetacin® with oral an important issue for credibility of estimates of indomethacin and placebo. Int J Sports Med clinical efficacy in treatments, just like random- 1990;11:393–6. isation21,22 and double-blinding.21 Just as it may 12. Hosie GAC. The topical NSAID, felbinac, versus oral be hazardous to change practice on the basis of ibuprofen: a comparison of efficacy in the treatment a single small trial, similarly beware meta-analysis of acute lower back injury. Br J Clin Res 1993;4:5–17. 23 restricted to multiple small trials. 13. Vanderstraeten G, Schuermans P. Study on the effect of etofenamate 10% cream in comparison The important research agenda is to identify those with an oral NSAID in strains and sprains due to 122 patients with chronic disease, particularly elderly sports injuries. Acta Belg Med Phys 1990;13:139–41. Health Technology Assessment 1998; Vol. 2: No. 12

14. Browning RC, Johson K. Reducing the dose of oral Aoki T, et al., 1984. Comparative study of piroxicam gel NSAIDs by use of feldene gel: an open study in with indomethacin gel and piroxicam placebo gel in the elderly patients with osteoarthritis. Adv Ther treatment of pain from orthopedic trauma. Jpn Pharmacol 1994;11:198–206. Ther;12:101–17. 15. Golden EL. A double-blind comparison of orally Arioli G, Scaramelli M, Pillosu W, 1990. La terapia topica ingested aspirin and a topically applied salicylate delle affezioni acte musculoscheletriche. Risultati di uno cream in the relief of rheumatic pain. Curr Ther Res studio comparativo sul piroxicam crema 1% versus 1978;24:524–9. diclofenac emulgel 1% [Topical therapy of acute skeletal muscle diseases. Results of a comparative study on 16. Khan KS, Daya S, Jadad AR. The importance of piroxicam cream 1% versus diclofenac emulgel 1%]. quality of primary studies in producing unbiased Clin Ter;134:363–9. systematic reviews. Arch Intern Med 1996;156:661–6. Auclair J, Georges M, Grapton X, Gryp L, d’Hooghe M, 17. Evans JMM, McMahon A, McGilchrist M, White G, Meisser RG, et al., 1989. A double-blind controlled multi- Murray F, McDevitt D, et al. Topical non-steroidal center study of percutaneous niflumic acid gel and anti-inflammatory drugs and admission to hospital placebo in the treatment of achilles heel tendinitis. for upper gastrointestinal bleeding and perforation: Curr Ther Res;46:782–8. a record linkage case-control study. BMJ 1995;311:22–6. Baixauli F, Inglés F, Alcántara P, Navarrete R, Puchol E, Vidal F, 1990. Percutaneous treatment of acute soft tissue 18. Berner G, Engels B, Vögtle-Junkert U. Percutaneous lesions with naproxen gel and ketoprofen gel. J Int Med ibuprofen therapy with Trauma-Dolgit gel: bioequi- Res;18:372–8. valence studies. Drugs Exp Clin Res 1989;15:559–64. Balthazar-Letawe D, 1987. Voltaren Emulgel en practique 19. Treffel P, Gabard B. Feasibility of measuring the rheumatologique. Essai comparatif avec Indocid gel biavailability of topical ibuprofen in commercial [Voltaren Emulgel in rheumatological practice. A formulations using drug content in epidermis and comparative trial with Indocid gel]. Acta Belg Med a methyl nicotinate skin inflammation assay. Skin Phys;10:109–10. Pharmacol 1993;6:268–75. Baracchi G, Messina Denaro S, Piscini S, 1982. 20. Rosenthal R. The “file drawer problem” and Esperienza sull’impiego topico dell’acido isobutil-fenil- tolerance for null results. Psychol Bull proprioico (Ibuprofen) nell flogosi traumatica. 1979;86:638–41. Confronto a doppia cecità con placebo [Experience of the topical use of isobutylfenylproprionic acid 21. Schulz KF, Chalmers I, Hayes RJ, Altman DG. (Ibuprofen) in traumatic inflammation. A double-blind Empirical evidence of bias: dimensions of methodo- comparison with placebo]. Gazz Med Ital;141:691–4. logical quality associated with estimates of treatment effects in controlled trials. JAMA 1995;273:408–12. Bolten W, 1991. Felbinac-Gel zur Behandlung lokalisierter extraartikulärer rheumatischer Besch- 22. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. werden – eine multizentrische, placebokontrollierte, Randomization is important in studies with pain randomisierte Studie [Felbinac gel for treatment of outcomes: systematic review of transcutaneous localized extra-articular rheumatic diseases – a multi- electrical nerve stimulation in acute postoperative centre, placebo controlled, randomized study]. pain. Br J Anaesth 1996;77:798–803. Z Rheumatol;50:10–13. 23. Egger M, Smith GD. Misleading meta-analysis. Bouchier-Hayes TA, Rotman H, Darekar BS, 1990. Lessons from “an effective, safe, simple” Comparison of the efficacy and tolerability of diclofenac intervention that wasn’t. BMJ 1995;310:752–4. gel (Voltarol Emulgel) and felbinac gel (Traxam) in the treatment of soft tissue injuries. Br J Clin Pract;44:319–20. Studies included Boutrón F, Galicia A, Zamora G, Martinez-Zurita F, 1994. Airaksinen O, Venãlãinen J, Pietilãinen T, 1993. Single-blind study to evaluate the efficacy and safety of Ketoprofen 2.5% gel versus placebo gel in the treatment naproxen gel compared with diclophenac emulgel in the of acute soft tissue injuries. Int J Clin Pharmacol Ther treatment of soft tissue injuries. Proc West Pharmacol Toxicol;31:561–3. Soc;37:153–6. Åkermark C, Forsskåhl B, 1990. Topical indomethacin in Browning RC, Johson K, 1994. Reducing the dose of oral overuse injuries in athletes. A randomized double blind NSAIDs by use of feldene gel: an open study in elderly study comparing Elmetacin® with oral indomethacin and patients with osteoarthritis. Adv Ther;11:198–206. placebo. Int J Sports Med;11:393–6. Campbell J, Dunn T, 1994. Evaluation of topical Algozzine GJ, Stein GH, Doering PL, Araujo OE, Akin ibuprofen cream in the treatment of acute ankle sprains. KC, 1982. Trolamine salicylate cream in osteoarthritis of J Accid Emerg Med;11:178–82. the knee. JAMA;247:1311–13. Camus J, 1975. Action de la Myrtécaïne associée au Ammer K, 1991. Perkutane antirheumatische therapie Salicylate de Diéthylamine, en traitement local, dans beim weichteilrheumatismus. Rheumatologie;6:412–21. diverses affections rheumatismales. Rheumatologie;27:61–6. 123 Topically-applied NSAIDs

Candela V, Bagarone A, 1986. Studio in cieco semplice Fantato S, de Gregorio M, 1971. Clinical evaluation of sull’attivita terapeutica del “Fastum Gel” nelle lesioni topical benzydamine in traumatology. traumatiche da sport. Med Sport;39:57–63. Arzneimittelforschung;21:1530–5. Chatterjee DS, 1977. A double-blind clinical study with Fotiades P, Bach GL, 1976. Wirkung einer benzydamine 3% cream on soft tissue injuries in an flufenaminsäurehaltigen salbe bei verschiedenen occupational health centre. J Int Med Res;5:450–8. rheumatischen erkrankungen. Fortschr Med;94:1036–8. Commandre F, Zakarian H, Corriol-Rohou S, 1993. Frahm E, Elsasser U, Kämmereit A, 1993. Topical Comparison of the analgesic and anti-inflammatory treatment of acute sprains. Br J Clin Pract;47:321–2. effects of topical niflumic acid gel versus piroxicam gel in the treatment of musculoskeletal disorders. Fujimaki E, et al., 1985. Clinical evaluation of piroxicam Curr Ther Res;53:113–21. gel versus indomethacin gel and placebo in the treatment of muscle pain: a double-blind, multicenter study. Curioni GB, Di Domenica F, Daolio P, Spignoli G, 1985. Jpn Pharmacol Ther;12:119–37. Valutazione dell’efficacia terapeutica di ibudros gel in traumatologia. Clin Europa;24:456–60. Galeazzi M, Marcolongo R, 1993. A placebo-controlled study of the efficacy and tolerability of a nonsteroidal Dickson DJ, 1991. A double-blind evaluation of topical anti-inflammatory drug, DHEP plaster, in inflammatory piroxicam gel with oral ibuprofen in osteoarthritis of the peri- and extra-articular rheumatological diseases. Drugs knee. Curr Ther Res;49:199–207. Exp Clin Res;19:107–15. Diebschlag W, 1986. Diclofenac bei stumpf-traumatischen Sprunggelenkschwellungen. Volumetrische Verlaufs- Gallacchi G, Mautone G, Lualdi P, 1990. Painful bestimmung im plazebokontrollierten Doppelblind- inflammatory conditions. Topical treatment with diclo- versuch [Diclofenac in blunt traumatic ankle joint fenac hydroxyethylpyrrolidine. Clin Trials J;27:58–64. swelling. Volumetric monitoring in a placebo controlled Geller O, 1980. Vergleich eines Salizylat/Heparin-Gels double-blind trial]. Fortschr Med;104:437–40. mit einem Mono-Substanzpräparat. Ergebnisse einer Diebschlag W, Nocker W, 1987. Einfluß einer topischen Doppelblind-cross-over-Studie [Comparison of a sali- Behandlung auf den Krankheitsverlauf bei cylate heparin gel with a monosubstance preparation. Sprunggelenks-Distorsionen [The effect of topical Results of a double-blind crossover study]. Münch Med treatment in the treatment of disease in ankle sprains]. Wschr;122:1231–2. Arzneimittelforschung;37:1076–81. Giacovazzo M, 1992. Clinical evaluation of a new NSAID Diebschlag W, Nocker W, Bullingham R, 1990. A double- applied topically (BPAA Gel) vs. diclofenac emulgel in blind study of the efficacy of topical ketorolac trometha- elderly osteoarthritic patients. Drugs Exp Clin mine gel in the treatment of ankle sprain, in comparison Res;18:201–3. to placebo and etofenamate. J Clin Pharmacol;30:82–9. Ginsberg F, Famaey JP, 1991. Double-blind, randomized Diebschlag W, Nocker W, Lehmacher W, 1992. Treatment crossover study of the percutaneous efficacy and toler- of acute sprains of the ankle joint. A comparison of the ability of a topical indomethacin spray versus placebo in effectiveness and tolerance of two gel preparations the treatment of tendinitis. J Int Med Res;19:131–6. containing indomethacin. Fortschr Med;110:64–72. Golden EL, 1978. A double-blind comparison of orally Dreiser RL, 1988. Clinical trial of efficacy and tolerability ingested aspirin and a topically applied salicylate cream of topical ibuprofen in the treatment of tendinitis. J Int in the relief of rheumatic pain. Curr Ther Res;24:524–9. Méd;119:15–31. Governali E, Casalini D, 1995. Ricerda clinica controllata Dreiser RL, 1989. Clinical trial – Fastum gel FG-6. tra ketoprofene gel 5% e ketoprofene crema 1% in Unpublished. pazienti con postumi di lesioni traumatiche. Dreiser RL, Tisne-Camus M, 1993. DHEP plasters as a Riabilitazione;28:61–9. topical treatment of knee osteoarthritis – a double-blind Gualdi A, Bonollo L, Martini A, Forgione A, 1987. placebo-controlled study. Drugs Exptl Clin Res;19:117–23. Antinflammatori no steroidi per uso topico in Dreiser RE, Charlot J, Lopez A, Ditisheim A, 1990. traumatologia: studio clinico con flunoxaprofene Clinical evaluation of niflumic acid gel in the treatment e chetoprofene. Riforma Med;102:401–4. of uncomplicated ankle sprains. Curr Med Res;12:93–9. Gui L, Pellacci F, Ghirardini G, 1982. Impiego Dreiser RL, Roche R, de-Sahbe R, Thomas F, Leuteneger dell’ibuprofen crema in pazienti ambulatoriali di E, 1994. Flurbiprofen local action transcutaneous: clin- interesse ortopedico. Confronto in doppia cecità con ical evaluation in the treatment of acute ankle sprains. placebo. Clin Ter;101:363–9. Eur J Rheum Inflamm;14:9–13. Haig G, 1986. Portable thermogram technique for El-Hadidi T, El-Garf A, 1991. Double-blind study com- topically applied benzydamine cream in acute soft-tissue paring the use of Voltaren Emulgel versus regular gel injuries. Int J Tissue React;8:145–7. during ultrasonic sessions in the treatment of localized traumatic and rheumatic painful conditions. J Int Med Hallmeier B, 1988. Efficacy and tolerance of etofenamate 124 Res;19:219–27. and diclofenac in acute sports injuries. Rheuma;8:183–6. Health Technology Assessment 1998; Vol. 2: No. 12

Hallmeier B, Michelbach B, 1986. Etofenamat unter tape- Picchio AA, Volta S, Longoni A, 1981. Studio clinico verbänden. Med Welt;37:1344–8. controllato sull’impiego dell’ibuprofen per uso topico in traumatologica sportiva. Med Sport;34:403–6. Hohmeister R, 1983. Die behandlung von weichteil- rheumatischen erkrankungen mit mobilisin gel. Fortschr Pineda MD, Capulong OM, de Guzman DL, Te CV, 1993. Med;101:1586–8. The topical NSAIDs, felbinac and piroxicam: a compari- Hosie GAC, 1993. The topical NSAID, felbinac, versus son of efficacy and safety in the treatment of acute soft oral ibuprofen: a comparison of efficacy in the treatment tissue trauma. Br J Clin Res;4:63–72. of acute lower back injury. Br J Clin Res;4:5–17. Ramesh N, Steuber U, 1983. Dolgit® cream in accident- Julien D, 1989. Clinical trial – Fastum gel FG-8. and sports-related injuries in medical practice. Results of Unpublished. a double blind study. Therapiewoche;33:4563–70. Kockelbergh M, Verspeelt P, Caloine R, Dermaux F, 1985. Reginster JY, Crommen J, Renson M, Franchimont P, Traitement anti-inflammatoire local par un gel de 1990. Percutaneous administration of indomethacin in kétoprofène: données cliniques récentes [Local anti rheumatoid arthritis. Curr Ther Res;47:548–53. inflammatory treatment with a ketoprofen gel: current clinical findings]. Acta Belg Med Phys;8:205–13. Ritchie LD, 1996. A clinical evaluation of flurbiprofen Kroll MP, Wiseman RL, Guttadauria M, 1989. A clinical LAT and piroxicam gel: a multicentre study in general eveluation of piroxicam gel: an open comparative trial practice. Clin Rheumatol;15:243–7. with diclofenac gel in the treatment of acute musculoskeletal disorders. Clin Ther;11:382–91. Rose W, Manz G, Lemmel E, 1991. Behandlung der akitivierten gonarthrose mit topisch appliziertum Lester AA, 1983. Management of sprained ankles. piroxicam-gel. Münch Med Wschr;133:562–6. Practitioner;225:935–6. Rosemeyer B, 1991. Behandlung von distorsionen des Linde F, Hvass I, Jürgensen U, Madsen F, 1985. Ankel- sprunggelenkes. Med Welt;42:166–70. forstuvninger behandlet med benzydamin 5% creme. Ugeskr Laeger;148:12–13. Rosenthal M, Bahous I, 1993. A controlled clinical study Mattara L, Trotta F, Biasi D, Cervetti R, 1994. Evaluation on the new topical dosage form of DHEP plasters in of the efficacy and tolerability of a new locally acting patients suffering from localized inflammatory diseases. preparation of flurbiprofen in scapulohumeral Drugs Exp Clin Res;19:99–105. periarthritis. Eur J Rheum Inflamm;14:15–20. Russell AL, 1991. Piroxicam 0.5% topical gel compared Matucci-Cerinic M, Casini A, 1988. Ketoprofen vs. to placebo in the treatment of acute soft tissue injuries: etofenamate in a controlled double-blind study: evidence a double-blind study comparing efficacy and safety. Clin of topical effectiveness in soft tissue rheumatic pain. Invest Med;14:35–43. Int J Clin Pharmacol Res;8:157–60. Sanguinetti C, 1989. Trattemento con BPAA gel dei McLatchie GR, McDonald M, Lawrence GF, Rogmans D, traumi dei tessuti moli. Clin Ter;130:255–8. Lisai P, Hibberd M, 1989. Soft tissue trauma: a randomised controlled trial of the topical application of felbinac, Seligra A, Inglés F, 1990. A comparative study of a new NSAID. Br J Clin Pract;43:277–80. naproxen gel and flufenamic acid gel in the treatment Montagna CG, Turroni L, Martinelli D, Orlandini MC, of soft tissue injuries. Curr Med Res;12:249–54. 1990. Single-blind comparative study of meclofenamic acid gel versus naproxen gel in acute musculoskeletal Sinniger M, Blanchard P, 1981. Controlled clinical trial disorders. Curr Ther Res;47:933–9. with fentiazec cream in sport microtraumatology. J Int Med Res;9:300–2. Morris WD, Scott HV, Peters WA, Ketelbey JW, 1991. Felbinac topical gel for acute soft tissue sports injuries. Sugioka Y, et al., 1984. Multicenter clinical evaluation of N Z J Sports Med;19:45–7. piroxicam gel vs. indomethacin gel in the treatment of non-traumatic diseases of tendon or muscle. Jpn Noret A, Roty V, Allington N, Hauters P, Zuinen C, Poels Pharmacol Ther;12:139–53. R, 1987. Ketoprofen gel as topical treatment for sport injuries. Acta Ther;13:367–78. Taboada A, 1992. Experencia controlada con gel de Oakland C, 1993. A comparison of the efficacy of the piroxicam asociado a ultrasonidos en afecciones agudas topical NSAID felbinac and ultrasound in the treatment del aparato locomotor. Prens Méd Argent;79:630–2. of acute ankle injuries. Br J Clin Res;4:89–96. Thorling J, Linden B, Berg R, Sandahl A, 1990. A double- Parrini M, Cabitza P, Arrigo A, Vanasia M, 1992. Efficacia blind comparison of naproxen gel and placebo in the e tollerabilità del ketoprofene sale di lisina schiuma treatment of soft tissue injuries. Curr Med Res;12:242–8. per uso topico nel trattamento di alcune patologie traumatiche dell’apparato locomotore [Efficacy and Tonutti A, 1993. Utilizzazione del ketoprfene gel 5% tolerability of ketoprofen lysine salt foam for topical use (Oridis gel) nella pratica traumatologica: studio in in the treatment of traumatic pathologies of the doppio cieco controllato verso etofenamato. Ort locomotor apparatus]. Clin Ter;141:199–204. Traum;XIV:119–25. 125 Topically-applied NSAIDs

Vanderstraeten G, Schuermans P, 1990. Study on the Wanet G, 1979. Etude clinique contrôlée d’un effect of etofenamate 10% cream in comparison with an topique associant la nopoxamine au salicylate oral NSAID in strains and sprains due to sports injuries. de diéthylamine (Algésal suractivé) en médecine Acta Belg Med Phys;13:139–41. physique et réhabilitation [Controlled clinical study of a topic associating nopoxamine with diethylamine Vecchiet L, Colozzi A, 1991. Effects of meclofenamic salicylate (Algésal suractivé) in physical medicine acid in the treatment of lesions deriving from minor and rehabilitation)]. J Belge Méd Phys traumatology. Clin J Pain;7:S54–9. Réhab;2:119–26. Vitali G, 1980. Sperimentazione clinica controllata sull’ impiego topico dell’acido 2-(3-benzoil-fenil) proprionico Zerbi E, Pace A, Demarchi F, Bassi F, Arrigo A, a tre diverse concentrazioni [Controlled clinical Garagiola U, 1992. Ketoprofen lysine salt in a new experiment with the topical use of 2-(3-benzoyl-phenyl) foam formulation for the topical treatment of traumatic proprionic acid at 3 different concentrations]. Clin injuries: a controlled, between-patient, clinical trial. Ter;94:257–73. Curr Ther Res;51:823–9.

126 Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 12 Injected morphine in postoperative pain

Summary Methods The pain relief after injected morphine compared Single-dose, randomised, placebo-controlled with placebo is examined in this chapter, in trials of injectable (intramuscular, subcutaneous patients with moderate or severe pain after surgery, and intravenous) morphine in acute postoperative and the efficacy of injected morphine is related pain were sought. A number of different search to that of oral analgesics. A literature search of strategies were used to identify eligible reports various databases was for randomised, single-dose, in MEDLINE (1966–97), EMBASE (1980–97), placebo-controlled trials. the Cochrane Library (1997 issue 2) and the Oxford Pain Relief Database (1950–94).2 The Pain relief or pain intensity difference over last electronic search was conducted in March 4–6 hours and adverse effects were extracted. 1997. The words ‘morphine’, ‘diamorphine’, The number of patients with at least 50% pain ‘heroin’ were used to identify relevant reports, relief was derived and then used to calculate the using a combination of free text words and relative benefit and the NNT for one patient to MeSH terms, and without restriction to language. achieve at least 50% pain relief for 4–6 hours. Additional reports were identified from reference lists of retrieved reports, review articles, and In 15 trials intramuscular morphine, 10 mg specialist textbooks. (486 patients), was compared with placebo (460 patients); the NNT was 2.9 (95% CI, 2.6–3.6). Included and excluded reports One in three patients with moderate or severe Inclusion criteria were full journal publication postoperative pain achieved at least 50% pain of RCTs which included single-dose treatment relief; they would not have done had they been groups of injected (intravenous, intramuscular given placebo. Minor adverse effects were more or subcutaneous) morphine and placebo, acute common with morphine (34%) than with placebo postoperative pain, blinded design, baseline (23%) (relative risk 1.49 (95% CI, 1.09–2.04)) but pain of moderate to severe intensity, adult drug-related study withdrawal was rare and not patients, and assessments of pain intensity or different from placebo. Intramuscular morphine, pain relief over 4–6 hours with results for 10 mg, gives analgesia equivalent to oral NSAID, in TOTPAR, SPID, VAS TOTPAR, or VAS SPID), keeping with historic results from single trials. For or with data from which these could be calcu- patients who can swallow, oral NSAID may be the lated. Review articles, letters or abstracts were best choice. not included.

This chapter of the review has been published in Reports were screened to eliminate those without full by McQuay and colleagues.1 pain outcomes, those which were definitely not randomised, or were abstracts or reviews. Each report which could possibly be described as an Introduction RCT was read independently by each of the authors and scored using a three-item, 1–5 score, Perhaps understandably, we all tend to believe that quality scale.3 Consensus was then achieved. The injecting drugs provides better pain relief than maximum score for an included study was 5 and taking the same drug by mouth. Indeed, it took the minimum 1. generations for doctors to be persuaded that oral morphine was effective in cancer pain – they all Data extraction and analysis wanted to inject. This chapter focuses on the post- Data extracted from the reports were the pain operative pain relief produced by injection of setting, study treatment groups, numbers of morphine, using the same methods as for the oral patients treated, study duration, the route and drugs. The aim was to achieve an estimate of the dose of morphine, and mean or derived analgesic efficacy of injected morphine which could TOTPAR, SPID, VAS TOTPAR or VAS SPID be compared with the estimates for the oral drugs. or any dichotomous global pain relief outcome. 127 Injected morphine in postoperative pain

Information on minor and major adverse events, Nine reports which appeared to fulfill inclusion as defined by the authors of the original reports, criteria were omitted. Three studies31–33 had pain was also extracted. relief or intensity information for 1 hour or less. Two reports34,35 appeared to duplicate previously For each report with mean TOTPAR, SPID, VAS published information and four36–39 used TOTPAR or VAS SPID values for morphine and non-standard assessments which could not placebo, the data was converted to percentage be used. of maximum by division into the calculated maximum value.4 The proportion of patients in each Only for 10 mg doses of intramuscular morphine treatment group who achieved at least 50% max- was data available from at least three trials, which TOTPAR was calculated using verified equations.5–7 could be pooled for meta-analysis. In 15 com- These proportions were then converted into the parisons, 486 patients were given intramuscular number of patients achieving at least 50% max- morphine, 10 mg, and 460 placebo (Table 40). The TOTPAR by multiplying by the total number of size of the active treatment group in these trials patients in the treatment group.5 varied between 9 patients and 51 patients (mean 33, median 30). Information on the number of patients with > 50% maxTOTPAR for morphine and placebo The placebo response rate (i.e. the proportion of was used to calculate relative risk (or benefit) patients given placebo experiencing at least 50% and NNT by pooling data when available from pain relief) varied from 0% to 47% (mean 15%), at least three comparisons between morphine and the response rate for intramuscular morphine, and placebo with a particular dose and route of 10 mg, was 7–93% (mean 46%; Figure 37). Of the administration. Relative risk or benefit estimates 15 comparisons between intramuscular morphine, were calculated with their 95% CIs using a random 10 mg, and placebo, eight showed it to be statistic- effects model8 for analgesic data which were not ally superior to placebo and had a lower CI of the homogenous (p < 0.1) and a fixed effects model9 relative benefit greater than 1 (Table 40). The for adverse effect data which were homogenous pooled relative benefit was 2.8 (95% CI, 2.0–3.8). (p > 0.1). Homogeneity of the analgesic results was also explored graphically.10 The NNT11 was The pooled NNT for intramuscular morphine, calculated with a 95% CI.12 A statistically signifi- 10 mg, compared with placebo was 2.9 (95% CI, cant difference from control was assumed when 2.6–3.6). Omitting a trial which included acute the 95% CIs of the relative risk/benefit did not non-surgical pain18 did not affect this result. The include 1. Statistical difference between NNTs pooled NNT without this study was 3.1 (95% CI, was assumed when CIs did not overlap. 2.7–3.8). The NNT for trials in which fewer than the median number of patients were given morphine (that is, fewer than 32 patients treated) Results was 2.9 (95% CI, 2.3–4.1), the same as for larger trials (with 32 patients or more) – NNT of 3.0 In all, 18 reports of 20 trials fulfilled the inclusion (95% CI, 2.5–3.8). criteria; 696 patients were given morphine and 563 placebo. No trials of subcutaneous morphine Minor adverse effects occurred in 34% of patients or of diamorphine by any route of administration given intramuscular morphine compared with 23% met the inclusion criteria. Morphine was given by of patients given placebo. This was a significantly intramuscular injection in all studies except one,13 increased rate with a relative risk of 1.49 (95% CI, in which it was given intravenously. Morphine doses 1.09–2.04). Major adverse effects (drug-related were 5 mg,14,15 8 mg,13,16 10 mg,14,15,17,18–29 12.5 mg30 study withdrawal) were rare (overall 1.2%) and and 20 mg.28 Details of these studies are presented did not differ between morphine and placebo in Table 39. (Table 40).

Two studies18,30 included a mixed population of patients with postoperative and other acute pain. Comment Trials otherwise investigated pain relief predominantly after orthopaedic and gynaecological Morphine is the archetypal analgesic for use in surgery. Pain outcomes were over 6 hours except moderate or severe pain. It is also the ‘gold stand- for two studies in which they were over 4 hours.22,30 ard’ against which other injected analgesics are Quality scores were 2 for two reports, 3 for six, tested. It was surprising, therefore, that rigorous 128 4 for nine and 5 for one. searching revealed so few placebo-controlled trials Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 39 Injected morphine in postoperative pain: patients, methods, outcomes and results of included studies

Study Condition Design, Outcome Treatment Analgesic outcome Withdrawals Adverse Comment Quality and number study duration measures groups results (morphine and adverse events score of patients and follow-up vs. placebo) effects

Beaver & General and RCT,double-blind, Standard 1. i.m. morphine, Morphine superior to Withdrawals Sedative adverse Morphine better 3 Feise, gynaecological single dose, parallel 4-point PI 8 mg, n = 24 placebo PID not reported. effects: morphine than placebo. 197616 surgery group. Assessments by Standard 2. Placebo, n = 24 (see Figure 1). 19/24, placebo n = 96 single nurse observer, 5-point PR 3. i.m. hydroxyzine, 3/24. Age: not given hourly assessments 9-item tension/ 100 mg, n = 24 up to 6 hour. anxiety question- 4. i.m. hydroxyzine Moderate to severe naire + morphine, baseline pain. n = 24

Brown, Various surgical RCT,double-blind, PI (5-point) none, 1. i.m. morphine, Significant difference Patients remaining No study with- NSAID better 4 et al., 198420 procedures single dose, parallel slight, moderate, 10 mg, n = 30. between morphine and in study at drawals reported. than morphine. n = 90 group, double-dummy. severe, very severe 2. Placebo, n = 30 placebo for most 6 hours: placebo NSD between Age range: Assessed by single PR (5-point) none, 3. p.o. naproxen, outcomes. 5 (17%), morphine treatments. 18–68 years nurse observer at 0, poor, fair, good, 550 mg, n = 30 1. % patients with 13 (43%), p = 0.01. Patients reporting 0.5, 1 hour then hourly very good > 50% PR: placebo Early termination one or more: intervals for 6 hour. 50% PR 37%, morphine 77%, due to inadequate placebo 14 (47%), Medication taken when at 6 hours p = 0.002. relief: morphine morphine 22 (73%), baseline pain was at Time to next 2. Mean hours in study 16 (20%), placebo naproxen 13 (43%). least moderate. analgesic (SD): placebo 3.2 (1.5), 25 (83%), p = 0.01. morphine 4.7 (1.4) 3. 6-hour mean SPID: placebo 2.1, morphine 5.5 4. 6-hour mean TOTPAR values not given. NB: 5-point SPID.

Brown, Various surgical Study 3 only. RCT, PI (5-point) 1. i.m. morphine 1. 6-hour mean SPID Completing 6-hour No study with- NSAID better than 4 et al., procedures double-blind, single 1 = none, 2 = slight, 10 mg, n = 30 (SEM): morphine 5.2 study: morphine drawals reported. morphine. Wrong 199127 n = 150 dose, double-dummy, 3 = moderate, 4 = 2. Placebo, n = 30 (0.6), placebo 0.9 (1.0), 43%, placebo 30%. NSD in % patients calculations used for Age range: parallel group. Assessed severe, 5 = very severe 3. p.o. anirolac, p < 0.003. reporting: morphine Figure 2 TOTPAR. 18–66 years by single nurse observer PR (5-point) 600 mg, n = 30 2. 6 hour TOTPAR mean 41%, placebo 23%. at 0, 0.5, 1 hour then 5 = none, 4 = poor, 4. p.o. anirolac, (SEM): morphine 17.7 hourly intervals for 3 = fair, 2 = good, 125 mg, n = 30 (1.0), placebo 24.1 (1.1), 6 hours. Medication 1 = very good 5. p.o. anirolac, p < 0.003. NB: wrong taken when baseline Pain half gone at 5 mg, n = 30 calculation, 50% pain was at least end of study, or at TOTPAR not reported. moderate. time of withdrawal

Campos Various acute RCT,single dose, parallel Standard 4-point PI 1. i.m. morphine, 1. SPID: morphine Patients completing: Patients reporting: Morphine better 3 & Solis, and medical group, double-blind. 50% PR 10 mg, n = 30 12.32, placebo 6.48. morphine 28/30, morphine 66, than placebo. 198018 patients Assessments by single 2. Placebo, n = 30 2. 50% PR: morphine placebo 23/40. placebo 20. (2 part studies) nurse observer at 0, 3. i.m. nefopam, 76%, placebo 49.6%. Reporting no relief n = 120 30 minutes, 1, 2, 3, 4, 5, 20 mg, n = 30 at 2 hours: (each study) 6 hours. Baseline pain at 4. i.m. diphen- morphine 1/30, Age: adult least moderate. hydramine, 20 mg, placebo 6/30. n = 30 Dropped out for other reasons: morphine 1, placebo 1.

Davie, et al., Various day- RCT,single dose, parallel VAS PI 10 cm 1. i.v. morphine, Morphine gave signifi- Withdrawals at Patients reporting: Morphine better 4 198213 surgery group, double-blind. 8 mg, n = 30 cantly less pain than 2 hours: morphine morphine 4, than placebo. procedures Assessments by single 2. Placebo, n = 30 placebo at all assess- 3/30, placebo 20/30, placebo 3, n = 90 observer at 0, 30 minutes, 3. p.o. fenoprofen, ment times. No VAS fenoprofen 8/30. fenoprofen 2. Age range: 1, 2, 3, 4, 5, 6 hours. Base- 200 mg, n = 30 SPID, but data can be 23–69 years line pain at least moderate. calculated from table in text.

de Andrade, Orthopaedic RCT,double-blind, PI (5-point) 1. i.m. morphine, 1. 6-hour mean SPID 105 remaining in 5 patients withdrew: NSAID comparable 4 et al., surgery (hip single dose, parallel 0 = none, 1 = mild, 10 mg, n = 51 (SD): morphine,10 mg, study at 6 hours; morphine 2; to morphine. 199414 and knee group, double-dummy? 2 = moderate, 2. i.m. morphine, 7.2 (3.9); morphine, morphine, 10 mg, placebo 1. Difference in sample replacement) Assessed by patients 3 = severe, 5 mg, n = 50 5 mg, 6.7 (4.6 ); placebo 33; morphine, 5 mg, Morphine, 10 mg, size of active vs. n = 176 at 0, 0.5, 1 hour then 4 = very severe 3. Placebo, n = 25 2.4 (3.8); p < 0.01. 32; placebo 7; significantly more placebo controls. Age: adult hourly intervals for VAS PI 100 mm 0–99 4. p.o. ketorolac, 2. 6-hour mean p < 0.01. than placebo, 6 hours. Medication PR (5-point) 0 = none, 10 mg, n = 50 TOTPAR (SD): p = 0.01 (Table 4). taken when baseline 1 = a little, 2 = some, morphine, 10 mg, Patients reporting: pain was at least 3 = a lot, 4 = complete 14.0 (6.0); morphine, morphine, 10 mg, moderate. Patient and investi- 5 mg, 12.4 (6.7); placebo 26/51; morphine, gator global evaluation 5.0 (7.3); p < 0.01. 5 mg, 17/50; placebo end of study 5/25.

continued 129 Injected morphine in postoperative pain

TABLE 39 contd Injected morphine in postoperative pain: patients, methods, outcomes and results of included studies

de Lia, Gynaeco- RCT,double-blind, single Standard 4-point PI 1. Morphine, 1. 6-hour mean SPID: Cumulative drop- Study withdrawals NSAID comparable 4 et al., 198624 logical surgery dose, parallel group, Investigator rating end 10 mg, n = 30 morphine 5.17, placebo out rate shown in not stated. Patients to morphine. n = 92 double-dummy. Assessed of treatment: poor/fair 2. Placebo, n = 30 1.80, p < 0.01. Table II.Patients reporting adverse Age: not stated by patients at 0, 1 hour = no effect, good/ 3. p.o. flurbi- 2. Pain more than half dropped-out at effects: morphine then hourly intervals for excellent = effective profen, 50 mg, gone at 6 hours: 6 hours: morphine 4, placebo 4. 6 hours. Medication PR derived score n = 30 morphine 6 (20%), 21 (70%), placebo taken when baseline half gone placebo 0 (0%). 66 (74%). pain was at least Need for additional 3. 6-hour mean moderate analgesia TOTPAR: morphine 16.98, placebo 10.80, p < 0.01.

Fragen, Orthopaedic RCT,double-blind, single Standard 4-point PI 1. i.m. morphine, 1. 6-hour mean SPID: Cumulative drop- Study withdrawals Morphine better 3 et al., and major dose, parallel group, PR (5-point): 10 mg, n = 36 morphine 5.0, out rate shown not reported. than synthetic 198319 gynaecological double-dummy. Assessed 1 = worse, 0 = none, 2. Placebo, n = 35 placebo 1.6. in Figure 2. Number of patients mixed agonist surgery by single observer at 1 = a little, 2 = 3. i.m. ciramadol, 2. 6-hour mean reporting not given. antagonist control. n = 139 0, 15, 30 minutes, 1 hour moderate, 3 = a lot, 30 mg, n = 34 TOTPAR: morphine Nausea: morphine Age range: then hourly intervals for 4 = complete 4. i.m. ciramadol, 10.1, placebo 3.63. 5, placebo 1. 18–65 years 6 hours. Medication VAS PI 10 cm 60 mg, n = 34 Figure 1 shows % Vomiting: morphine taken when baseline Patient rating end of patients with > 50% 4, placebo 1. pain was at least treatment (4-point): PR at each time point. moderate. Injections 1 = poor, 2 = fair, into deltoid muscle. 3 = good, 4 = excellent 50% PR (derived)

Gravenstein, Postoperative RCT,double-blind, single Standard 4-point PI 1. i.m. morphine, 1. PI scores mean Not given. No study with- High dose of 3 et al., 198421 wound pain dose, parallel group. PR (6-point): 1 = 10 mg, n = 40 shown in Table 3. drawals reported. dezocine better n = 160 Assessed by more than worse, 0 = none, 2. Placebo, n = 40 2. PR mean scores NSD reported than morphine. Age range: 1 observer at 0, 15, 1 = slight, 3. i.m. dezocine, shown in Figure 2. between groups. 19–70 years 30 minutes, 1 hour then 2 = moderate, 10 mg, n = 40 3. Patients with hourly intervals for 3 = substantial, 4. i.m. dezocine, moderate to complete 6 hours. Medication 4 = complete 15 mg, n = 40 PR by time shown taken when baseline VAS PI 100 mm in Table 3.Patients pain was at least Patient rating of rating treatment as moderate. treatment (4-point): good or excellent: poor, fair, good, morphine 15/36, excellent placebo 6/24. Investigator rating of treatment (satisfactory, unsatisfactory)

Kaiko, Acute RCT,double-blind, single Standard 4-point PI 1. i.m. morphine, 1.VAS TOTPAR: placebo 9 of 17 post- Patients reporting: Morphine better 3 et al., 198717 postoperative dose, crossover design. PR (5-point) 10 mg, n = 9 90, morphine 135, operative patients placebo 2/12, than placebo. pain Assessed by nurse VAS PI 2. Placebo, n = 9 cocaine 70, cocaine + completed morphine 6/13, n = 9/17 observer at 0, 30 mins, VAS PR 3. Oral cocaine, morphine 161. crossover. cocaine 4/16. (completed 1 hour then hourly VAS mood 10 mg, n = 9 2.VAS SPID: placebo morphine + crossover each intervals for 6 hours. 4. Oral cocaine, 58, morphine 99, cocaine 7/13. treatment) Medication taken when 10 mg, + cocaine 35, cocaine + Age range: baseline pain was at morphine, n = 9 morphine 108. 22–65 years least moderate.

Kantor, Postoperative RCT,double-blind, Standard 4-point PI 1. i.m. morphine, 1. Derive SPID data for Not reported. Total side-effects Morphine better 4 et al., and acute 4 doses of same drug Patients with 12.5 mg, n = 50 dose 1, day 1 from per dose: morphine than placebo. 198130 traumatic pain given over 2 days. ≥ 50% PR 2. Placebo, n = 49 Figure 1. (n = 14/50) 34; n = 250 Nurse observers, 3. p.o. codeine, 2. SPID for day 1: placebo (n = 4/49) Age range: assessments at 0, 90 mg, n = 50 morphine 4.90, placebo 9; pentazocine 21–75 years 30 minutes, 1 hour then 4. p.o. penta- 1.20, pentazocine 3.43, (n = 12/50) ?; hourly intervals for zocine, 75 mg, codeine 3.63, codeine (n = 11/50) 4 hours. Medication n = 50 oxycodone 4.35. 25; oxycodone taken when baseline pain 5. p.o. oxycodone (n = 2/49) 3. was at least moderate. compound (oxycodone hydrochloride, 4.5 mg, oxycodone terepthalate, 0.30 mg, aspirin, 224 mg, phena- cetin, 160 mg, caffeine, 32 mg), n = 49

130 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 39 contd Injected morphine in postoperative pain: patients, methods, outcomes and results of included studies

Lippmann, Major abdo- RCT,double-blind, single PI (4-point): 0 = none, 1. i.m. morphine, Morphine superior to See Figure 6 for % Patients with none: High dose of mixed 2 et al., minal and ortho- dose, parallel group. 1 = slight, 2 = 10 mg, n = 30 placebo for most remedicating by morphine 10/30, agonist antagonist 198926 paedic surgery Assessments at 0, moderate, 3 = severe 2. i.m. placebo, outcomes, as were time. placebo 25/30. better than n = 151 30 minutes, 1 hour then PR (scale not n = 30 other active treatments morphine; only Age range: hourly intervals for described) 3. i.m. tonazocine, 1. SPID see Figure 4. extractable data 18–65 years 6 hours. Medication Additional analgesia 2 mg, n = 29 2.TOTPAR see Figure 3. from physician rating taken when baseline Treatment effective, 4. i.m. tonazocine, of effectiveness. pain was at least partially effective, 4 mg, n = 30 moderate. ineffective (derived 5. i.m. tonazocine, score) 2 mg, n = 31

Morrison, Major obstetric RCT,double-blind, single Standard 4-point PI 1. i.m. morphine, 1. SPID mean 6-hour: Dropped-out at 1 patient reported No differences 3 et al., 198625 and gynaeco- dose, parallel group. PR: 1 = unchanged, 10 mg, n = 51 morphine 10.7, placebo 6 hours: morphine adverse effect with between NSAID logical surgery Assessments at 0, worse, 2 ≤ half gone, 2. Placebo, n = 55 5.07 (see Figure 2). 15, placebo 34. morphine; no and morphine. n = 181 30 minutes, 1 hour then 3 = half gone, 4 ≥ 3. p.o. flurbi- 2.TOTPAR mean others. Zomepirac treat- Age range: hourly intervals for half gone, 5 = profen, 50 mg, 6-hour: morphine ments incomplete 19–65 years 6 hours. Medication completely gone n = 53 23.91, placebo 14.54 as license withdrawn taken when baseline Investigator global: 4. p.o. zomepirac, (see Figure 2) during study. Non- pain was at least 1 = no effect, 2 = poor, 100 mg, n = 22 3. Patients requesting standard PR scale, moderate. 3 = fair, 4 = good, additional analgesia: no extractable data. 5 = excellent morphine 12/47, Need for supple- placebo 34/50. mentary analgesia

Nørholt, Third molar RCT,double-blind, single PI (5-point): 0 = gone, 1. i.m. morphine, 1. 4-hour mean SPID Median time to Patients reporting: No differences 5 et al., 199628 extraction dose, double-dummy, 1 = slight, 2 = 10 mg, n = 37 (SD): morphine, 10 mg, remedication, morphine, 10 mg, between NSAID n = 252 parallel group. Assess- moderate, 3 = severe, 2. i.m. morphine, 1.9 (2.7); morphine, minutes (range): 32/37; morphine, and morphine. Age range: ments at 0, 15, 30, 4 = unbearable 20 mg, n = 37 20 mg, 3.9 (3.0); morphine, 10 mg, 20 mg, 37/37; Extractable data – 18–40 years 45 minutes, 1 hour then PI (11-point) 0–10 3. i.m. placebo, placebo –5 (2.5). 185 (65–540); placebo 22/37. global rating? hourly intervals for PR (5-point): 0 = none n = 37 2. 4-hour mean morphine, 20 mg, 8 hours. Medication to 4 = complete 4. i.m. lornoxicam, TOTPAR (SD): 540 (100–540); taken when baseline Patient rating of 4 mg, n = 33 morphine, 10 mg, 5.1 placebo 80 pain was at least treatment 4 & 8 hours: 5. i.m. lornoxicam, (3.8); morphine, 20 mg, (30–540). moderate. 1–5 8 mg, n = 38 8.8 (3.6); placebo 1.2 Onset of relief 6. i.m. lornoxicam, (2.3). 8-hour values Duration of relief 16 mg, n = 38 available. Remedication time 7. i.m. lornoxicam, 20 mg, n = 37

Pande, et al., Obstetric and RCT,double-blind, single PI 0–3 1. i.m. morphine, 1. 6-hour mean SPID Study stopped early None reported No difference 4 199629 gynaecological dose, parallel group. 4 and 8 hour SPID 10 mg, n = 14 values not given, only due to neuro- for morphine between enadoline surgery Assessments at 0, 15, PR 0–4 2. i.m. placebo, levels of significance. psychiatric effects or placebo. and morphine. 2 studies in n = 53 30 minutes,1 hour then Pain reduced by half n = 12 2. 6-hour mean from enadoline. Problems with this report; Age range: hourly intervals for Patient rating of 3. i.m. enadoline, TOTPAR (SEM): Patients completing methods in study 2. Study 2 26–61 years 6 hours. Medication treatment at end 15 mg, n = 14 morphine 2.7 (1.1), study: morphine 1, only used. taken when baseline of study (kappa agonist) placebo 0.9 (1.2). placebo 0. pain was at least 4. i.m. enadoline, moderate. 25 mg, n = 13

Pandit, Orthopaedic, RCT,double-blind, single PI (3-point): mild, 1. i.m. morphine, Morphine superior to Not stated. Patients reporting: Dezocine (mixed 4 et al., 198522 gynaecological dose, parallel group. moderate, severe 10 mg, n = 39 placebo for some but morphine 10%, agonist antagonist) and general Single observer, % PR: 0 = none, 2. i.m. placebo, not all outcomes. placebo 8%. better than surgery assessments at 0, 15, 1 ≤ 50%, 2 = 50%, n = 38 1. 4-hour mean morphine on some n = 190 30 minutes,1 hour then 3 ≥ 50%, 4 = 100% 3. i.m. dezocine, TOTPAR: morphine outcomes. Age range: hourly intervals for Pain reduced by half 5 mg, n = 38 5.8, placebo 3.1. 26–61 years 4 hours. Medication 4. i.m. dezocine, 2. 50% relief at 4 hours: taken when baseline 10 mg, n = 37 morphine 36.8%, pain was at least 5. i.m. dezocine, placebo 11.4%. moderate. 15 mg, n = 38 3. Difference in proportion of patients with adequate relief at 2 and 4 hours only given for morphine vs. placebo.

continued 131 Injected morphine in postoperative pain

TABLE 39 contd Injected morphine in postoperative pain: patients, methods, outcomes and results of included studies

Powell, Orthopaedic, RCT,double-blind, Standard 4-point PI 1. i.m. morphine, Morphine superior % remedicated at Patients reporting: Little difference 4 198523 gynaecological multiple dose, parallel PR: –1 = worse, 10 mg, n = 40 to placebo for all 6 hours: morphine morphine 6 (15%), between ciramadol and general group, more than one 0 = none, 1 = a little, 2. i.m. placebo, outcomes, as were 48.7, placebo 90. placebo 7 (18%). and morphine. surgery observer, assessments 2 = moderate, 3 = n = 40 other active treatments. Extractable data: n = 160 at 0, 15, 30 minutes, a lot, 4 = complete 3. i.m. ciramadol, 1.TOTPAR, see TOTPAR from Age range: 1 hour then hourly Patient rating end of 30 mg, n = 40 figures in text. figures in text. 18–65 years intervals for 6 hours. treatment: 1 = poor, 4. i.m. ciramadol, 2. SPID, see figures. Medication taken when 2 = fair, 3 = good, 60 mg, n = 40 3.VAS SPID, see figures. baseline pain was at 4 = excellent 4. Patient rating of good least moderate. or excellent: morphine 48.7%, placebo 26%. Global: morphine 85%, placebo 26% (see text).

van den Orthopaedic, RCT,double-blind, Standard 4-point PI 1. i.m. morphine, Morphine superior to Mean drop-out Incidence: Ciramadol better 2 Abeele & gynaecological single, parallel group, PR: –1 = worse, 5 mg, n = 20 placebo for most rate, see text. morphine, 10 mg, than morphine on Camu, and general medical observer, 0 = none, 1 = a little, 2. i.m. morphine, outcomes. 0/2; morphine, 5 mg, some outcomes. 198515 surgery assessments at 0, 15, 2 = moderate, 3 = 10 mg, n = 20 1. 6-hour TOTPAR: 4/20; placebo 4/20. n = 100 30 minutes,1 hour then a lot, 4 = complete 3. i.m. placebo, morphine, 10 mg, 13.6; Age range: hourly intervals for VAS PI 10 cm n = 20 morphine, 5 mg, 9.5; 18–65 years 6 hours. Medication Sedation: 0–3 4. i.m. ciramadol, placebo 1.1. taken when baseline Patient and investigator 30 mg, n = 40 2. 6-hour SPID: pain was at least rating of treatment: 5. i.m. ciramadol, morphine,10 mg, 9.3; moderate. poor, fair, good, 60 mg, n = 40 morphine, 5 mg, 6.3; excellent placebo 1.3. 3. 6-hour VAS SPID: morphine, 10 mg, 262.4; morphine, 5 mg, 132.0; placebo 37.9. Other data available.

TABLE 40 Analgesia and adverse effects of intramuscular morphine, 10 mg

Trial (date order) At least 50% PR At least 50% PR RB or RR NNT with morphine with placebo (95% CI) (95% CI)

Campos, et al., 198018 28/30 14/30 2.0 (1.4–3.0) 2.1 (1.5–3.7) van den Abeele & Camu, 198315 15/20 2/20 7.5 (2.0–28.6) 1.5 (1.1–2.4) Fragen, et al., 198319 17/36 4/35 4.1 (1.5–11.1) 2.8 (1.8–6.1) Brown, et al., 198420 23/30 11/29 2.0 (1.2–3.4) 2.6 (1.6–6.5) Gravenstein, 198421 10/40 0/40 101 (0.2–> 250) 4.0 (2.6–8.8) Pandit, et al., 198522 8/39 2/38 3.9 (0.9–17.2) 6.7 (3.4–138) Powell, 198523 11/39 0/40 114 (0.2–> 250) 3.6 (2.4–7.1) de Lia, et al., 198624 15/30 5/30 3.0 (1.3–7.2) 3.0 (1.8–9.1) Morrison, et al., 198625 47/51 25/55 2.0 (1.5–2.7) 2.1 (1.6–3.2) Kaiko, et al., 198717 2/9 1/9 2.0 (0.2–18.8) 9.1 (2.2–∞) Lippmann, et al., 198926 4/30 0/30 41 (0.1–> 250) 7.7 (3.9–85) Brown, et al., 199127 17/30 6/30 2.8 (1.3–6.2) 2.7 (1.7–7.2) de Andrade, et al., 199414 34/51 4/25 4.2 (1.7–10.5) 2.0 (1.4–3.2) Nørholt, et al., 199628 9/37 0/37 91 (0.2–> 250) 4.2 (2.6–9.5) Pande, et al., 199629 1/14 0/12 9.4 (0.0–> 250) 14.3 (4.9–∞)

Combined analgesic data 241/486 74/460 2.8 (2.0–3.8) 2.9 (2.6–3.6)

Trials with < 32 treated patients 101/293 40/198 2.2 (1.8–2.8) 2.9 (2.3–4.1) Trials with > 32 treated patients 136/293 35/270 4.0 (1.6–9.8) 3.0 (2.5–3.8)

Minor adverse effects 108/320 68/295 1.49 (1.09–2.04) 9.1 (5.6–27.7)

Major adverse effects 2/334 6/304 0.31 (0.07–1.38) 132 Health Technology Assessment 1998; Vol. 2: No. 12

analgesics, the NNT of 2.9 (95% CI, 2.6–3.8) for At least 50% PR with 10 mg intramuscular morphine, 10 mg, can be compared intramuscular morphine (%) with those obtained for oral tramadol, 100 mg, 100 (4.8 (95% CI, 3.4–8.2)),41 for oral paracetamol, 1000 mg, (4.6 (95% CI, 3.9–5.4)), for paracetamol, 600/650 mg, plus codeine, 60 mg (3.1 (95% CI, 80 2.6–3.8)),42 and for ibuprofen, 400 mg, (2.7 (95% CI, 2.5–3.0)). The equivalence of the NNTs for oral NSAIDs and intramuscular morphine, 10 mg, is 60 supported by the repeated failure to separate them in analgesic trials.43,44 A crucial issue here is dose. 40 Clearly with opioids, there should be dose titration against effect. The NNT value of 2.9 is for 10 mg of intramuscular morphine; giving 20 mg improved 20 the NNT value.28

Rank ordering of analgesics in this way is 0 potentially less accurate than taking the relative 02040 60 80 100 efficacy of the individual drugs from within one At least 50% PR with placebo (%) very large trial with a single randomisation. In the absence of such head-to-head’ comparisons, the authors consider that this indirect ranking, the FIGURE 37 Relationship of the proportion of patients achieving relative efficacy of the drugs against placebo, is at least 50% pain relief with intramuscular morphine, 10 mg, to helpful in making clinical decisions. The trials used the proportion obtaining at least 50% pain relief with placebo to produce NNTs for analgesics compared with in 15 trials placebo are all single-dose, postoperative, randomised and double-blind. The patients must have had in which morphine was given by intravenous, moderate to severe pain before being treated and intramuscular or subcutaneous injection, and in standard measures of pain were required. These which single-dose analgesic efficacy was tested using uniform quality standards and patient selection standard, validated methods. No subcutaneous criteria allow a credible indirect ranking of efficacy studies were found and only one intravenous study, to be made. Internal validity is demonstrated in and only for intramuscular morphine, 10 mg, was the ranking by the dose–response relationships there sufficient information (494 treated patients) obtained for analgesics, with better analgesia for it to be pooled for meta-analysis. No studies of (lower NNT) obtained with higher doses diamorphine were found which met the criteria. (Figure 56). External validity will come when there are direct (head-to-head) comparisons which A single intramuscular dose of morphine, 10 mg, confirm the rank order in the indirect table. had an NNT of 2.9 for at least 50% pain relief compared with placebo. This means that one in At first sight the fact that the analgesia from every three patients with pain of moderate to severe intramuscular morphine, 10 mg, is no better than intensity will experience at least 50% pain relief the analgesia from a therapeutic dose of oral with morphine which they would not have had with NSAID is surprising.45 Injected drugs are generally placebo. Sensitivity analysis found that size of trial thought of as more ‘powerful’ than oral drugs. did not make a difference (Table 40). Sensitivity In reality, there is a considerable body of direct analysis was not performed for quality of trials, evidence that confirms the indirect ranking. since all but two reports had quality scores of 3 or For many years investigators have been unable more. Overestimation of the effect of treatment has to distinguish the analgesia resulting from intra- been shown in trials with quality scores of 2 or less muscular morphine, 10 mg, and oral NSAID using the same validated quality scale as here.40 (where comparisons were within the same trial and, hence, randomised). The NNT for morphine can be compared with those of other analgesics from similar meta-analyses This is a clinically useful observation for patients in which the efficacy of analgesics was compared who can swallow and who have no contraindication with placebo in patients with moderate or severe to NSAID. Oral NSAID appears to be the best anal- postoperative pain. While there is as yet no com- gesic choice. There is no advantage to giving that parable information available for other injected dose of NSAID by a suppository or injection.46 If 133 Injected morphine in postoperative pain

the patient can swallow but speedy analgesia is 14. de Andrade JR, Maslanka MA, Maneatis T, Bynum L, required, then intravenous rather than intra- Burchmore M. The use of ketorolac in the manage- muscular analgesia seems more logical. If the patient ment of postoperative pain. Orthopedics 1994;17:157–66. cannot swallow, then intramuscular morphine, 10 mg, gives analgesia equivalent to oral NSAID, and 15. van den Abeele G, Camu F. Comparative evaluation doubling the dose does indeed increase the anal- of ciramadol (WY 15.705), morphine and placebo gesia.28 There is not, as yet, a ranking of injected for treatment of postoperative pain. Acta Anaesthesiol NSAID compared with injected opioid. Belg 1985;36:97–110. 16. Beaver WT, Feise G. Comparison of the analgesic effects of morphine, hydroxyzine, and their References combination in patients with postoperative pain. 1. McQuay HJ, Carroll D, Moore RA. Injected Adv Pain Res Ther 1976;1:553–7. morphine in postoperative pain; a quantitative 17. Kaiko R, Kanner R, Foley K, Wallenstein S, Canel A, systematic review. J Pain Symptom Manage. In press. Rogers A, et al. Cocaine and morphine interaction 2. Jadad AR, Carroll D, Moore A, McQuay H. in acute and chronic cancer pain. Pain Developing a database of published reports of 1987;31:35–45. randomised clinical trials in pain research. Pain 18. Campos VM, Solis EL. The analgesic and hypo- 1996;66:239–46. thermic effects of nefopam, morphine, aspirin, 3. Jadad AR, Moore RA, Carroll D, Jenkinson C, diphenhydramine, and placebo. J Clin Pharmacol Reynolds DJM, Gavaghan DJ, et al. Assessing the 1980;20:42–9. quality of reports of randomized clinical trials: is 19. Fragen RJ, Kouzmanoff C, Caldwell NJ. blinding necessary? Control Clin Trials 1996;17:1–12. Intramuscularly administered ciramadol for 4. Cooper SA. Single-dose analgesic studies: the upside management of postoperative pain: a comparative and downside of assay sensitivity. Adv Pain Res Ther study. J Clin Pharmacol 1983;23:219–26. 1991;18:117–24. 20. Brown CR, Sevelius H, Wild V. A comparison of 5. Moore A, McQuay H, Gavaghan D. Deriving single doses of naproxen sodium, morphine sulfate, dichotomous outcome measures from continuous and placebo in patients with postoperative pain. data in randomised controlled trials of analgesics. Curr Ther Res 1984;35:511–18. Pain 1996;66:229–37. 21. Gravenstein JS. Dezocine for postoperative wound 6. Moore A, Moore O, McQuay H, Gavaghan D. pain. Int J Clin Pharmacol Ther Toxicol 1984;22:502–5. Deriving dichotomous outcome measures from continuous data in randomised controlled trials of 22. Pandit SK, Kothary SP, Pandit UA, Kunz NR. Double analgesics: use of pain intensity and visual analogue blind placebo controlled comparison of dezocine scales. Pain 1997;69:311–15. and morphine for post operative pain relief. Can 7. Moore A, McQuay H, Gavaghan D. Deriving Anaesth Soc J 1985;32:583–91. dichotomous outcome measures from continuous 23. Powell W. A double blind comparison of multiple data in randomised controlled trials of analgesics: intramuscular doses of ciramadol, morphine, and verification from independent data. Pain placebo for the treatment of postoperative pain. 1997;69:127–30. Anesth Analg 1985;64:1101–7. 8. DerSimonian R, Laird N. Meta-analysis of clinical 24. de Lia JE, Rodman KC, Jolles CJ. Comparative trials. Control Clin Trials 1986;7:177–88. efficacy of oral flurbiprofen, intramuscular 9. Gardner MJ, Altman DG. Confidence intervals morphine sulfate, and placebo in the treatment rather than p values: estimation rather than of gynecologic postoperative pain. Am J Med hypothesis testing. BMJ 1986;292:746–50. 1986;80:60–4. 10. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in 25. Morrison J, Harris J, Sherrill J, Heilman C, Bucovaz clinical research. Ann Intern Med 1987;107:224–33. E, Wiser W. Comparative study of flurbiprofen and 11. Laupacis A, Sackett DL, Roberts RS. An assessment morphine for postsurgical gynecologic pain. Am J of clinically useful measures of the consequences of Med 1986;80:55–9. treatment. N Engl J Med 1988;318:1728–33. 26. Lippmann M, Mok M, Farinacci J, Lee J. Tonazocine 12. Cook RJ, Sackett DL. The number needed to treat: mesylate in postoperative pain patients: a double a clinically useful measure of treatment effect. blind placebo controlled analgesic study. J Clin BMJ 1995;310:452–4. Pharmacol 1989;29:373–8. 13. Davie IT, Slawson KB, Burt RA. A double blind 27. Brown CR, Schwartz KE, Wild VM, Koshiver JE. comparison of parenteral morphine, placebo, and Comparison of anirolac with morphine and placebo oral fenoprofen in management of postoperative for postoperative pain. Curr Ther Res 134 pain. Anesth Analg 1982;61:1002–5. 1991;50:379–85. Health Technology Assessment 1998; Vol. 2: No. 12

28. Norholt S, SindetPedersen S, Larsen U, Bang U, 36. Cohen RI, Edwards W T, Kezer EA, Ferrari DA, Ingerslev J, Nielsen O, et al. Pain control after Liland AE, Smith ER. Serial intravenous doses of dental surgery: a double-blind, randomised dezocine, morphine, and nalbuphine in the trial of lornoxicam versus morphine. Pain management of postoperative pain for outpatients. 1996;67:335–43. Anesth Analg 1993;77:533–9.

29. Pande AC, Pyke RE, Greiner M, Wideman GL, 37. Lasagna L, Mosteller F, von Felsinger JM, Beecher Benjamin R, Pierce MW. Analgesic efficacy of HK. A study of the placebo response. Am J Med enadoline versus placebo or morphine in 1954:770–9. postsurgical pain. Clin Neuropharmacol 38. Levine J, Gordon N, Smith R, Fields H. Analgesic 1996;19:451–6. responses to morphine and placebo in individuals with post-operative pain. Pain 1981;14:379–88. 30. Kantor TG, Hopper M, Laska E. Adverse effects of commonly ordered oral narcotics. J Clin Pharmacol 39. Foley WL, Edwards RC, Jacobs L3. Patient- 1981;21:1–8. controlled analgesia: a comparison of dosing regimens for acute postsurgical pain. J Oral 31. Rice A, Lloyd J, Miller C, Bullingham RE, O’Sullivan Maxillofac Surg 1994;52:155–9; discussion 159–60. G. A double-blind study of the speed of onset of analgesia following intramuscular administration of 40. Khan KS, Daya S, Jadad AR. The importance of ketorolac tromethamine in comparison to intra- quality of primary studies in producing unbiased muscular morphine and placebo. Anaesthesia systematic reviews. Arch Intern Med 1996;156:661–6. 1991;46:541–4. 41. Moore RA, McQuay HJ. Single-patient data meta- analysis of 3453 postoperative patients: oral 32. Whitehead EM, O’Sullivan GM, Lloyd J, Bullingham tramadol versus placebo, codeine and combination RES. A new method for rate of analgesic onset: two analgesics. Pain 1997;69:287–94. doses of intravenous morphine compared with placebo. Clin Pharmacol Ther 1992;52:197–204. 42. Moore A, Collins S, Carroll D, McQuay H. Paracetamol with and without codeine in acute 33. Verborgh C, Camu F. Post surgical pain relief with pain: a quantitative systematic review. Pain zero order intravenous infusions of meptazinol and 1997;70:193–201. morphine: a double blind placebo controlled evaluation of their effects on ventilation. Eur J Clin 43. Mansfield M, Firth F, Glynn C, Kinsella J. A Pharmacol 1990;38:437–42. comparison of ibuprofen arginine with morphine sulphate for pain relief after orthopaedic surgery. 34. Maslanka MA, de Andrade JR, Maneatis T, Bynum L, Eur J Anaesthesiol 1996;13:492–7. DiGiorgio E. Comparison of oral ketorolac, intramuscular morphine, and placebo for treatment of 44. Schachtel BP, Thoden WR, Baybutt RI. Ibuprofen pain after orthopedic surgery. South Med J and acetaminophen in the relief of postpartum 1994;87:506–13. episiotomy pain. J Clin Pharmacol 1989;29:550–3. 45. Collins SL, Moore A, McQuay HJ, Wiffen PJ. Oral 35. Brown CR, Moodie JE, Dickie G, Wild VM, Smith ibuprofen and diclofenac in postoperative pain: a BA, Clarke PJ, et al. Analgesic efficacy and safety of quantitative systematic review. Eur J Pain. In press. single dose oral and intramuscular ketorolac tromethamine for postoperative pain. 46. McQuay HJ, Justins D, Moore RA. Treating acute Pharmacotherapy 1990;10:59–70S. pain in hospital. BMJ 1997;314:1531–5.

135

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 13 Dihydrocodeine in postoperative pain

Summary as an antitussive. In 1995, nearly one-tenth of all analgesic prescriptions (opiate, non-opiate and The aim of this review was to determine the NSAID) issued in the UK were for dihydrocodeine.4 analgesic efficacy and adverse effects of oral and The proportion of dihydrocodeine used for the injectable dihydrocodeine from single dose studies treatment of postoperative pain is not known. in moderate to severe postoperative pain. Publish- ed studies were identified by searching electronic databases and checking reference lists of retrieved Methods reports. Summed pain relief and pain intensity data were extracted and converted to dichotomous A search was undertaken for RCTs of dihydro- information yielding the number of patients with codeine in postoperative pain which covered at least 50% pain relief. This was used to calculate MEDLINE (1966–February 1997), EMBASE the relative benefit and NNT for one patient to (1980–97), the Cochrane Library (January 1997), achieve at least 50% pain relief. Biological Abstracts (1985–97), and the Oxford Pain Relief Database (1950–94).5 The terms In three reports (194 patients) oral dihydro- ‘dihydrocodeine’, ‘random*’, ‘clinical trial’, codeine was compared with placebo and in one ‘trial’, analgesi*’, ‘pain’ and 36 brand names and (120 patients) dihydrocodeine, 30 mg or 60 mg, preparations6 were used in a broad free-text search was compared with ibuprofen, 400 mg. For a single without restriction to language. Additional reports dose of dihydrocodeine, 30 mg, in moderate to were identified from reference lists of retrieved severe postoperative pain, the NNT for at least 50% articles. Unpublished data were not sought. pain relief was 9.7 (95% CI, 4.5–∞) when compared with placebo over a period of 4–6 hours. Pooled Included reports data showed no significant difference in adverse The inclusion criteria used were: effect incidence for dihydrocodeine, 30 mg, compared with placebo. • full journal publication • postoperative pain The 95% CIs of the NNT included no benefit of • postoperative administration dihydrocodeine, 30 mg, over placebo. A statistical • adult patients superiority for ibuprofen, 400 mg, over dihydro- • baseline pain of moderate to severe intensity codeine, 30 mg or 60 mg, was shown. • double-blind design • random allocation to treatment groups which This chapter of the review has been published in included dihydrocodeine and placebo. full by Edwards and colleagues.1 Pain outcomes used were TOTPAR or SPID over 4–6 hours or sufficient data provided to allow Introduction their calculation. Pain measures allowed for the calculation of TOTPAR were a standard five-point Opioids are extensively used in the management pain relief scale (none, slight, moderate, good, of pain and are believed capable of relieving severe complete), and for SPID a standard four-point pain pain more effectively than NSAIDs.2 The aim of intensity scale (none, mild, moderate, severe). this quantitative systematic review was to assess the efficacy and safety of a single dose of oral dihydro- Data extraction and analysis codeine in the management of postoperative pain Extracted from each study were: of moderate to severe intensity. • the number of patients treated Dihydrocodeine is a synthetic opioid analgesic • the mean TOTPAR or SPID developed in the early 1900s. Its structure and • study duration pharmacokinetics are similar to that of codeine3 and • the dose of dihydrocodeine it is used for the treatment of postoperative pain or • information on adverse effects. 137 Dihydrocodeine in postoperative pain

Mean TOTPAR or SPID values were converted and NNH with 95% CIs were calculated.11 The to % maxTOTPAR or % maxSPID by division into 95% CI of the NNT indicates no benefit of one the calculated maximum value.7 The referenced treatment over the other when the upper limit equations were used to estimate the proportion of includes infinity. patients achieving at least 50% maxTOTPAR.8,9 This was then converted to the number of patients achieving at least 50% maxTOTPAR by multiplying Results by the total number of patients in the treatment group. The number of patients with at least 50% A total of 48 published reports of dihydrocodeine maxTOTPAR was then used to calculate estimates in postoperative pain were identified, of which two of relative benefit and NNT. could not be obtained from the British Library. Of the retrieved reports, 18 studies were not random- Estimates of relative benefit and risk, with 95% CIs, ised and were excluded, leaving 28 randomised were calculated using a random effects model.10 studies. Of these, two included other pain con- Homogeneity was assumed when p > 0.1. A statistic- ditions, five had no extractable pain outcome data, ally significant benefit of active treatment over seven were not double-blind, four used dihydro- control was assumed when the CI did not include 1. codeine as a rescue analgesic only, and six did A statistically significant benefit of control over not specify baseline pain of moderate to severe active treatment was assumed when the upper limit intensity. These reports were also excluded. Details of the 95% CI of the relative benefit was < 1. NNT of the included studies are given in Table 41.

TABLE 41 Dihydrocodeine in postoperative pain: summary of included trials

Study Condition Study design, Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Dihydrocodeine versus placebo

Frame, Impacted third RCT,double-blind, single PI (9-point scale), Dihydrocodeine, Dihydrocodeine 30 mg Remedication 18 withdrew: No serious adverse effects et al., 198912 molar removal oral dose, parallel groups. non-standard 30 mg (n = 49); not significantly different allowed at 2 hours. 9 insufficient pain, reported; no patients n = 148 Assessed at 0.5, 1 hour PR (5-point scale) placebo (n = 50) to placebo. If remedicated 7 did not return withdrew as result. Age: adult and then hourly for standard 4-hour TOTPAR: patients withdrawn assessment forms, Dihydrocodeine, 30 mg: 5 hours. Medication taken dihydrocodeine, 30 mg, and PR set to zero 1 did not complete 1/49 with 1 adverse effect. when pain of moderate to 0.5; placebo 0.3 for all further time assessment forms, Placebo: 1/50 with 3 adverse severe intensity. points. 1 postoperative events. complications.

Galasko, Orthopaedic RCT,double-blind, PI (5-point scale), Dihydrocodeine, Dihydrocodeine not Multiple dose study; 9 withdrew because No patients experienced et al., surgery multiple oral dose, parallel non-standard 30 mg (n = 30); significantly different second dose given of inadequate adverse effects in single dose 198913 n = 89 groups.Assessed at 0.5, PR (5-point scale), placebo (n = 28) to placebo. as required. If analgesia after first analysis. Age range: 1 hour and then hourly standard Mean TOTPAR at 6 hours: remedicated, dose. Dihydro- 18–80 years for 6 hours. Medication VAS, 100 mm dihydrocodeine 11.3, patients excluded codeine, 30 mg taken when pain of moder- (‘no pain’ to placebo 11.1 from analysis. (n = 3); placebo ate to severe intensity. ‘worst pain I have (n = 6). ever felt’)

McQuay, Minor day-case RCT,double-blind, PI (4-point scale), Dihydrocodeine, 4-hour SPID and Allowed after Single dose analysis: Dihydrocodeine, 30 mg: et al., 198514 surgery multiple oral dose, parallel standard 30 mg (n = 18); TOTPAR presented. 1 hour. If remedi- all adverse effects 6/18 with 6 adverse events. (general) groups.Assessed at 0.5, PR (5-point scale), placebo (n = 19) TOTPAR: dihydrocodeine cated patients initial mild, no patients Placebo: 3/19 with 3 n = 54 1 hour and then hourly standard significantly better than PI and PR scores withdrew as result. adverse events. Age: adult for 4 hours. Medication VAS, 100 mm placebo (p < 0.05); used for all further NSD between taken when pain of moder- dihydrocodeine, 30 mg, time points. dihydrocodeine ate to severe intensity. 6.5; placebo 3.2 and placebo.

Dihydrocodeine versus ibuprofen

McQuay, Lower third RCT,double-blind, PI (4-point scale), Dihydrocodeine, TOTPAR at 6 hours: If remedicated at 3 patients withdrew. Single-dose adverse effects et al., 199315 molar removal multiple oral dose, cross- standard 30 mg (n = 40); dihydrocodeine, 30 mg, 6 hours, initial PI data were not presented. n = 68 over design. Self-assessed PR (5-point scale), dihydrocodeine, 3.3; dihydrocodeine, score and PR score Age: adult at 0.5, 1 hour and then standard 60 mg (n = 40); 60 mg, 4.7; ibuprofen, of zero used for all hourly for 6 hours. Medi- Global rating placebo (n = 40) 400 mg, 10.0 further time points. cation allowed when pain (5-point scale), Ibuprofen significantly of moderate to severe standard better than dihydro- intensity. codeine, 30 mg or 60 mg, (p < 0.01) 138 Health Technology Assessment 1998; Vol. 2: No. 12

Four studies met the inclusion criteria: three were placebo-controlled and one used ibuprofen, 400 mg, as an active control. All four studies exam- At least 50% PR with i.m. ined the effects of oral dihydrocodeine. Three morphine, 10 mg (%) 60 trials12–14 compared dihydrocodeine, 30 mg, with placebo and one15 compared dihydrocodeine, 30 mg or 60 mg, with ibuprofen, 400 mg. 50

Oral dihydrocodeine versus placebo 40 No reports comparing dihydrocodeine, 60 mg, with placebo met our inclusion criteria. Three 30 reports compared dihydrocodeine tartrate, 30 mg, (91 patients) with placebo (85 patients). One trial investigated dental pain,12 one ortho- 20 paedic pain,13 and one pain following minor day-case surgery.14 10

The proportion of patients experiencing at 0 least 50% pain relief with dihydrocodeine varied 0 102030405060 between 14% and 50%, with a mean value of 35%. The proportion of patients experiencing at least At least 50% PR with control (%) 50% pain relief with placebo varied between 5% and 50%, with a mean of 23% (Figure 38). The data sets were homogeneous (p = 0.12). Dihydro- FIGURE 38 Trials of oral dihydrocodeine in postoperative pain codeine, 30 mg, was not significantly different from ( , dihydrocodeine, 30 mg, vs. placebo; ●, dihydrocodeine, 30 mg, vs. ibuprofen, 400 mg; ▲, dihydrocodeine, 60 mg, vs. ibuprofen, 400 mg) placebo, relative benefit 1.7 (95% CI, 0.7–4.0) (Table 42). For a single dose of dihydrocodeine, 30 mg, compared with placebo the NNT was 60 mg (40 patients), was compared with ibuprofen, 9.7 (95% CI, 4.5–∞) for at least 50% pain relief 400 mg (40 patients), in dental pain. over a period of 4–6 hours in postoperative pain of moderate to severe intensity. The proportion of patients experiencing at least 50% pain relief with dihydrocodeine, 30 mg, was Adverse effects 8%, with dihydrocodeine, 60 mg, it was 15% and Details of adverse effects are given in Table 43. The with ibuprofen, 400 mg, (active control) it was incidence of adverse effects with dihydrocodeine 45% (Figure 38). A statistical superiority of ibu- was not significantly different from placebo. All profen, 400 mg, over dihydrocodeine, 30 mg, and adverse effects were mild and transient in nature dihydrocodeine, 60 mg, was shown, with relative and no patients withdrew as a result. benefit values of 0.2 (95% CI, 0.1–0.5) and 0.3 (95% CI, 0.2–0.8), respectively. Oral dihydrocodeine vs ibuprofen In one study,15 the efficacy and safety of either Ibuprofen, 400 mg, was significantly better than dihydrocodeine tartrate, 30 mg (40 patients) or dihydrocodeine, 30 mg, or dihydrocodeine, 60 mg,

TABLE 42 Summary of relative and NNT for trials of dihydrocodeine against placebo and ibuprofen, 400 mg

Number Dose of Number of Number of RB NNT of trials dihydrocodeine patients with patients with (95% CI) (95% CI) > 50% PR: > 50% PR: placebo dihydrocodeine or ibuprofen, 400 mg

Versus placebo 3 30 mg 29/97 19/97 1.7 (0.7, 4.0) 9.7 (4.5, ∞)

Versus ibuprofen, 400 mg 1 30 mg 3/40 18/40 0.2 (0.1, 0.5) –2.7 (–1.8, –5) 1 60 mg 6/40 18/40 0.3 (0.2, 0.8) –3.3 (–2.1, –9)

Negative NNTs in the comparison with ibuprofen mean that ibuprofen is better than dihydrocodeine. 139 Dihydrocodeine in postoperative pain

(Figure 38). When compared with ibuprofen, dihydrocodeine, 30 mg, although the NNT 95% 400 mg, the NNT for a single dose of dihydro- CIs for many of these overlap. The 95% CIs for ibu- codeine, 30 mg, was –2.7 (95% CI, –1.8, –5) for at profen, 200 mg (2.8–4.0) and 400 mg (2.5–3.0), least 50% pain relief over a period of 4–6 hours in and diclofenac, 50 mg (2.0–2.7) do not overlap postoperative pain of moderate to severe intensity with those those for dihydrocodeine, 30 mg, (Table 42). Similarly, for a single dose of dihydro- indicating greater analgesic efficacy. codeine, 60 mg, the NNT was –3.3 (95% CI, –2.1, –9) for at least 50% pain relief over a period of This rank order of relative efficacy against placebo 4–6 hours. is supported by a head-to-head comparison with ibuprofen. The analgesic efficacy of a single dose Adverse effects of oral dihydrocodeine, 30 mg or 60 mg, was signifi- No single dose adverse effect data were presented.15 cantly inferior to ibuprofen, 400 mg. For a single dose of dihydrocodeine, 30 mg, compared with ibuprofen, 400 mg, the NNT was –2.7 (95% CI, Comment –1.8, –5) for at least 50% pain relief over a period of 4–6 hours in postoperative pain of moderate to Dihydrocodeine is the second most commonly severe intensity. This means that for every three prescribed opioid in England, with 1.5 million patients with moderate to severe postoperative pain prescriptions issued for dihydrocodeine tartrate treated with ibuprofen, 400 mg, one will experi- tablets alone in 1995. This increased to 1.6 million ence at least 50% pain relief who would not have in 1996. No papers were found which investigated done if given dihydrocodeine, 30 mg. injected dihydrocodeine in the evaluation of postoperative pain with standard analgesic Similarly, for a single dose of dihydrocodeine, measurement methods. 60 mg, compared with ibuprofen, 400 mg, the NNT was –3.3 (95% CI, –2.1, –9) over a period For a single dose of oral dihydrocodeine tartrate, of 4–6 hours. So, one in every three patients with 30 mg, compared with placebo the NNT was 9.7 moderate to severe postoperative pain treated with (95% CI, 4.5–∞) for at least 50% pain relief over ibuprofen, 400 mg, would experience at least 50% a period of 4–6 hours in postoperative pain of pain relief who would not have done if given moderate to severe intensity. This means that one dihydrocodeine, 60 mg. in every ten patients with moderate to severe postoperative pain would experience at least 50% pain Nausea, vomiting, headache, dizziness, drowsiness relief with dihydrocodeine, 30 mg, who would not and confusion were the most commonly reported have done so with placebo. However, the estimate adverse effects for a single dose of oral dihydro- of relative benefit showed no significant difference codeine, 30 mg, when compared with placebo. between dihydrocodeine, 30 mg, and placebo. The incidence of adverse effects was not significantly different for dihydrocodeine, 30 mg, than A rank order of single dose analgesic efficacy in for placebo (Table 43). postoperative pain of moderate to severe intensity has been established by comparing orally adminis- Our results suggest dihydrocodeine to be less tered analgesics from methodologically similar effective than other analgesics when administered studies in other chapters in this review. A number as a single oral dose. Few of the retrieved reports of analgesics demonstrated greater efficacy than for investigating oral dihydrocodeine met the criteria

TABLE 43 Summary of adverse effects of oral dihydrocodeine and placebo

Number Adverse effect Number of patients Number of patients RR NNH of trials with adverse effects: with adverse effects: (95% CI) (95% CI) dihydrocodeine placebo

3 Nausea or vomiting 7/97 0/97 25 (0.7–907) N/A

3 Headache 3/97 0/97 1.05 (0.3–4.4) N/A

3 Dizziness, drowsiness 5/97 1/97 4.2 (0.6–28)N/A or confusion

N/A: Not calculated because NSD from placebo was shown for RR. 140 Health Technology Assessment 1998; Vol. 2: No. 12

for inclusion in this quantitative systematic 8. Moore A, McQuay H, Gavaghan D. Deriving review. This resulted in very little patient data dichotomous outcome measures from continuous being available for analysis, particularly for data in randomised controlled trials of analgesics: verification from independent data. Pain dihydrocodeine, 60 mg, which is often the 1997;69:127–30. preferred dose. Administering dihydrocodeine in multiple doses may improve its analgesic 9. Moore A, Moore O, McQuay H, Gavaghan D. efficacy but may also increase the incidence Deriving dichotomous outcome measures from of adverse effects. continuous data in randomised controlled trials of analgesics: use of pain intensity and visual analogue scales. Pain 1997;69:311–15. References 10. DerSimonian R, Laird N. Meta-analysis of clinical trials. Control Clin Trials 1986;7:177–88. 1. Edwards JE, McQuay HJ, Moore RA. Systematic review: dihydrocodeine in postoperative pain. 11. Cook RJ, Sackett DL. The number needed to treat: Cochrane Library. In press. a clinically useful measure of treatment effect. BMJ 1995;310:452–4. 2. Alexander JI, Hill RG. Postoperative pain. London: Blackwell Scientific Publications, 1987. 12. Frame J, Evans C, Flaum G, Langford R, Rout P. A comparison of ibuprofen and dihydrocodeine in 3. Rowell FJ, Seymour RA, Rawlins MD. Pharmaco- relieving pain following wisdom teeth removal. Br kinetics of intravenous and oral dihydrocodeine Dent J 1989;166:121–4. and its metabolites. Eur J Clin Pharmacol 1983;25:419–24. 13. Galasko C, Russel S, Lloyd J. Double-blind investigation of the efficacy of multiple oral doses of 4. Government Statistical Service. Prescription cost ketorolac tromethamine compared with dihydro- analysis for England 1995. London: Department of codeine and placebo. Curr Ther Res 1989;45:844–52. Health, 1996. 14. McQuay HJ, Bullingham RE, Moore RA, Carroll D, 5. Jadad AR, Carroll D, Moore A, McQuay H. Develop- Evans PJ, O’Sullivan G, et al. Zomepirac, dihydro- ing a database of published reports of randomised codeine and placebo compared in postoperative clinical trials in pain research. Pain 1996;66:239–46. pain after day-case surgery. The relationship 6. Reynolds JEF. Martindale: the extra pharmacopoeia. between the effects of single and multiple doses. 30th edition. London: Pharmaceutical Press, Br J Anaesth 1985;57:412–19. 1993:292–314. 15. McQuay HJ, Carroll D, Guest PG, Robson S, Wiffen 7. Cooper SA. Single-dose analgesic studies: the upside PJ, Juniper RP. A multiple dose comparison of and downside of assay sensitivity. Adv Pain Res Ther ibuprofen and dihydrocodeine after third molar 1991;18:117–24. surgery. Br J Oral Maxillofac Surg 1993;31:95–100.

141

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 14 Dextropropoxyphene in postoperative pain

Summary Co-proxamol® and Distalgesic®. In 1996, there were 10 million prescriptions in England for co- The aim of this review was to determine the proxamol alone, which represents one-fifth of all analgesic efficacy and adverse effects of oral analgesics prescribed (opiate, non-opiate and dextropropoxyphene alone and in combination NSAIDs); however, it is not clear how much was with paracetamol for moderate to severe post- used for postoperative pain.2 operative pain. Published reports were identified from a variety of electronic databases and Patient surveys have shown that postoperative pain additional studies were identified from the is often not managed well3 and there is a growing reference lists of retrieved reports. need to assess the efficacy and safety of commonly used analgesics as newer treatments become avail- Summed pain intensity and pain relief data were able. Judging relative analgesic efficacy is difficult extracted and converted into dichotomous inform- as clinical trials use a variety of comparators. It ation to yield the number of patients with at least can, however, be determined indirectly by 50% pain relief. This was used to calculate the comparing analgesics with placebo in similar relative benefit and NNT for one patient to achieve clinical circumstances to produce a common at least 50% pain relief. In six studies (440 patients) analgesic descriptor such as the NNT for at least dextropropoxyphene was compared with placebo 50% pain relief. This quantitative systematic review and in five (963 patients) dextropropoxyphene of the analgesic efficacy of dextropropoxyphene plus paracetamol, 650 mg, was compared has been produced using this method, both with with placebo. and without paracetamol, allowing comparison with other analgesics. For a single dose of dextropropoxyphene, 65 mg, in postoperative pain, the NNT for at least 50% pain relief was 7.7 (95% CI, 4.6–∞) when compared Methods with placebo over 4–6 hours. For the equivalent dose of dextropropoxyphene in combination with A search was undertaken of MEDLINE paracetamol, 650 mg, the NNT was 4.4 (95% CI, (1966–November 1996), EMBASE (1980–96), the 3.5–5.6) when compared with placebo. Pooled data Cochrane Library (November 1996), Biological showed increased incidence of central nervous Abstracts (1985–96) and the Oxford Pain Relief system adverse effects for dextropropoxyphene Database (1950–94)4 for RCTs of dextropropoxy- plus paracetamol compared with placebo. phene, and its combinations in postoperative pain. The terms ‘dextropropoxyphene’, ‘d-propoxy- Dextropropoxyphene, 65 mg, plus paracetamol, phene’, ‘propoxyphene’, ‘random*’, ‘clinical trial’, 650 mg, has a similar analgesic efficacy to tramadol, ‘trial’, ‘study’, analgesi*’, ‘pain’ and 41 brand 100 mg, but with a lower incidence of adverse names (including Distalgesic and Co-proxamol)5 effects. Ibuprofen, 400 mg, has a lower (better) were used in a broad free text search without NNT than both dextropropoxyphene, 65 mg, plus restriction to language. Additional reports were paracetamol, 650 mg, and tramadol, 100 mg. identified from reference lists of retrieved articles and reviews. Unpublished data were not sought. The review has been published in full by Collins and colleagues.1 Included reports The inclusion criteria used were:

Introduction • full journal publication • postoperative pain Dextropropoxyphene is an opioid analgesic which • postoperative oral administration has been widely available since the 1950s. It is • adult patients commonly used both alone, and in combination • baseline pain of moderate to severe intensity with paracetamol under such brand names as • double-blind design 143 Dextropropoxyphene in postoperative pain

• random allocation to treatment groups which Results included dextropropoxyphene and placebo or a combination of dextropropoxyphene plus A total of 130 published articles were identified. paracetamol and placebo. Two could not be obtained and attempts to contact the authors were unsuccessful. Five citations Pain outcomes used were TOTPAR or SPID over obtained from reference lists of retrieved reports 4–6 hours or sufficient data provided to allow their could not be traced by the British Library. Of the calculation. Pain measures allowed for the calcu- 123 retrieved reports, 33 were not RCTs, 24 were lation of TOTPAR or SPID were a standard five- not postoperative pain models or included other point pain relief scale (none, slight, moderate, pain conditions, 21 were not placebo-controlled, good, complete) or a standard four-point pain and in five dextropropoxyphene was used as a intensity scale (none, mild, moderate, severe). rescue analgesic only.

Data extraction and analysis Of the 40 RCTs that were placebo-controlled, The following were extracted from each study: patients did not have baseline pain of at least moderate severity in ten studies, in 16 there were • the number of patients treated no pain outcomes which were compatible with our • the mean TOTPAR or mean SPID inclusion criteria, and two studies were not double- • study duration blind. The data from one study was duplicated and • dose of dextropropoxyphene and paracetamol, therefore one of the duplicates11 was excluded. where appropriate This left 11 reports which met the inclusion criteria • information on adverse effects. and were included in the analysis. Details of the individual studies are presented in Table 44. Mean TOTPAR and mean SPID values were converted to % maxTOTPAR or % maxSPID Dextropropoxyphene versus placebo by division into the calculated maximum value.6 Six studies compared dextropropoxyphene The referenced equations were used to estimate hydrochloride, 65 mg (214 patients), with placebo the proportion of patients achieving at least (226 patients), and one trial also compared a dose 50% maxTOTPAR.7,8 The proportions were of 130 mg (25 patients) with placebo (25 patients). converted to the number of patients achieving at least 50% maxTOTPAR by multiplying by the In two trials12,13 postpartum pain (episiotomy) was total number of patients in the treatment group. investigated, and there were single studies of pain The number of patients with at least 50% following peridontal surgery,14 post-urogenital maxTOTPAR was then used to calculate surgery,15 post-gynaecological surgery,16 and after relative benefit and NNT. various surgical interventions.17

Relative benefit and relative risk estimates with The placebo response rate (the proportion of 95% CIs were calculated using the random effects patients experiencing at least 50% pain relief with model which provides a more conservative estimate placebo) varied between 4% and 76%. The dextro- of relative benefit than the fixed effects model.9 propoxyphene response rate (the proportion of Homogeneity was assumed when p > 0.1. A statist- patients experiencing at least 50% pain relief with ically significant benefit of active treatment over dextropropoxyphene) varied between 19% and placebo was assumed when the lower limit of the 84% (Figure 39). Data were homogenous (p = 0.13). 95% CI of the relative benefit was > 1. A statistically Dextropropoxyphene, 65 mg, was not significantly significant benefit of placebo over active treatment different from placebo, relative benefit 1.4 (95% was assumed when the upper limit of the 95% CI CI, 0.97–2.0) (Table 45). of the relative benefit was < 1. The NNT and NNH with their 95% CIs were calculated.10 The For a single dose of dextropropoxyphene, 65 mg, 95% CI includes no benefit of one treatment the NNT was 7.7 (95% CI, 4.6–∞) for at least 50% over the other when the upper limit is pain relief over a period of 4–6 hours compared with represented as infinity. placebo for pain of moderate to severe intensity.

Dextropropoxyphene is available as either the Pooled relative benefit estimates were calculated hydrochloride or napsylate salt. Equivalent molar using the random effects model.9 doses are 65 mg of dextropropoxyphene hydrochloride and 100 mg of dextropropoxyphene One trial17 used a dose of 130 mg of dextropro- 144 napsylate. poxyphene (25 patients). The relative benefit Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 44 Dextropropoxyphene in postoperative pain: details of included studies

Study Condition Study design, Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Dextropropoxyphene plus paracetamol

Cooper, Dental surgery RCT,double-blind, single PI (4-point scale) Dextropropoxy- Combination of dextro- Allowed at 200 analysed, 48 None serious and no et al., 198119 n = 248 oral dose, parallel groups, standard phene napsylate, propoxyphene with > 1 hour; if excluded: 31 violated patients withdrew as result. Age: adult general or local anaes- PR (5-point scale) 100 mg, + para- paracetamol significantly remedicated before protocol, 17 did not Dextropropoxyphene + thetic Self-assessed at standard cetamol, 650 mg better than placebo for patient withdrawn take medication. paracetamol, 5/42 with home at 0, 1 hour then Global evaluation (n = 42); placebo SPID and TOTPAR from study. If 5 adverse effects; placebo, hourly for 4 hours. by patient (n = 37) (p < 0.001). remedicated, PR 4/37 with 5. Medication given when (5-point scale) 4-hour TOTPAR: recorded as 0, and pain of moderate to at 4 hours dextropropoxyphene last PI score prior severe intensity + paracetamol 8.31, to remedication placebo 3.38. taken for all further time points.

Cooper, Dental surgery RCT,double-blind, single PI (4-point scale) Dextropropoxy- Combination of dextro- Did not state when 179 analysed. None serious and no 198018 n = 179 oral dose, parallel groups, standard phene napsylate, propoxyphene with remedication No withdrawals patients withdrew as a result. Age: adult mostly local anaesthetic. PR (5-point scale) 100 mg, + para- paracetamol significantly allowed. If remedi- reported. Dextropropoxyphene + Self-assessed at 0, 1 hour standard cetamol, 650 mg better than placebo for cated, last PR and paracetamol, 10/40 with then hourly for 4 hours. Global evaluation (n = 40); placebo SPID and TOTPAR PI score before 13 adverse events. Medication given when by patient (n = 48) (p < 0.05). remedication used Placebo: 13/48 with 17 pain of moderate to (5-point scale) 4-hour TOTPAR: for all further adverse events. severe intensity. at 4 hours dextropropoxyphene + time points. paracetamol 5.65, placebo 4.17.

Evans, et al., Minor RCT,double-blind, single PI (4-point scale) Dextropropoxy- Dextropropoxyphene + If remedicated 120 analysed. No None serious and no 198220 orthopaedic oral dose, parallel groups, standard phene hydro- paracetamol significantly before 4 hours, last withdrawals were patients withdrew as a surgery general anaesthetic. PR (5-point scale) chloride, 65 mg, better than placebo PI and PR score reported. result. Dextropropoxyphene n = 120 Assessed by same nurse standard + paracetamol, (p < 0.05) for TOTPAR. prior to remedi- + paracetamol: 16/30 with Age: adult observer at 0, 0.5, 1 hour 650 mg (n = 30); 4-hour TOTPAR: cation used for all 16 adverse events. then hourly for 4 hours. placebo (n = 30) dextropropoxyphene + further time points. Placebo: 13/30 with Medication given when paracetamol 7.37, placebo 13 adverse events. pain of moderate to 4.70. severe intensity.

Honig & Postoperative, RCT,double-blind, single PI (4-point scale) Dextropropoxy- Combination of dextro- If remedicated 196 analysed. Authors did not give details Murray, primarily oral dose, parallel groups. standard phene napsylate, propoxyphene with para within 6 hours No withdrawals of adverse events but 198121 orthopaedic Assessed by nurse PR (5-point scale) 100 mg, + paracetamol significantly better patient’s overall reported. reported NSD between n = 196 observer at 0, 0.5, 1 hour non-standard cetamol, 650 mg than placebo (p < 0.05) rating of drug taken active drug and placebo Age range: then hourly for 6 hours. Global evaluation (n = 50); placebo for SPID and TOTPAR. at time of groups. 19–74 years Medication given when by patient at (n = 48) 6-hour TOTPAR: remedication. pain of moderate to 5 hours (5-point) dextropropoxyphene + severe intensity. paracetamol 8.04, placebo 5.49.

Moore & Dental + Individual patient data Number of Dextropropoxy- Number of patients with None reported. None reported. None serious and no McQuay, postoperative from 18 double-blind, patients with at phene napsylate, at least 50% maxTOTPAR: withdrawals. 19977 pain RCTs. Study duration least 50% 100 mg, + para- dextropropoxyphene Dextropropoxyphene + n = 638 8 hours. Single oral dose, maxTOTPAR cetamol, 650 mg napsylate, 100 mg, + paracetamol, 88/316 adverse Age: adult parallel groups. Medi- (n = 316); placebo paracetamol, 650 mg, events; placebo, 66/322 cation was given when (n = 322) 112/316; placebo, 41/322. adverse events. Significantly pain of moderate to higher incidence with active severe intensity. treatment than placebo for: dizziness, relative risk 2.0 (1.1–4.0); drowsiness/ somnolence, 2.16 (1.5–3.2).

Dextropropoxyphene alone

Berry, Postpartum RCT,double-blind, single PI (4-point scale) Dextropropoxy- Dextropropoxyphene Patients allowed to 225 analysed; No adverse effects were et al., 197512 pain oral dose, parallel groups. standard phene hydro- significantly better than remedicate after no details given. reported with either active (episiotomy) Assessed by observer in PR (5-point) chloride, 65 mg placebo (p < 0.01). reasonable amount treatment or placebo. n = 225 hospital at 0.5, 1 hour, non-standard (n = 730; placebo Global evaluation (good of time. If remedi- Age range: then hourly for 4 hours. Global rating (n = 76) or excellent PR): dextro- cated, patients were 15–39 years Medication taken when (5-point scale) propoxyphene 26/73, regarded as a pain of moderate to by patient placebo 18/76. treatment failure. severe intensity.

continued 145 Dextropropoxyphene in postoperative pain

TABLE 44 contd Dextropropoxyphene in postoperative pain: details of included studies

Study Condition Study design, Outcome Dosing Analgesic outcome Remedication Withdrawals Adverse effects and number duration and measures regimen results and exclusions of patients follow-up

Dextropropoxyphene alone contd

Bloomfield Postpartum RCT,double-blind, single PI (4-point scale) Dextropropoxy- Dextropropoxyphene If remedicated 100 analysed. No serious adverse effects et al., 198013 pain oral dose, parallel groups. standard phene hydro- not significantly better patients withdrawn 6 withdrew: no pain were reported and no (episiotomy) Assessed, in hospital, by PR not measured chloride, 65 mg than placebo at 10% from study. Subse- relief or patients patients withdrew as a result. n = 100 same nurse observer at (n = 25); placebo probability level. quent PR readings remedicated. Dextropropoxyphene, Age: adult 0, 0.5, 1 hour then hourly (n = 25) SPID at 6 hours: set to pre-treatment 6/25 with 12 adverse events; for 6 hours. Medication dextropropoxyphene score. placebo, 9/25 with 9 adverse taken when pain of 9.32, placebo 8.12. events. moderate to severe intensity.

Cooper, Peridontal RCT,double-blind, single PI (4-point scale) Dextropropoxy- Dextropropoxyphene Allowed after 1 hour. 212 analysed, 91 No serious adverse effects et al., surgery oral dose, parallel groups, standard phene hydro significantly better than Last score prior to excluded: 48 did not were reported and no 198614 n = 301 local anaesthetic. Self- PR (5-point scale) chloride, 65 mg placebo (p < 0.1). remedication was medicate, 17 missed patients withdrew as a result. Age: adult assessed at 0, 0.5, 1 hour standard (n = 50); placebo TOTPAR at 6 hours: used for duration readings, 9 lost to Dextropropoxyphene, 10/50 then hourly for 6 hours. Global evaluation (n = 56) dextropropoxyphene of study. follow-up, 4 remedi- with 10 adverse events; Medication taken when by patient at 7.7, placebo 5.2. cated at < 1 hour, placebo, 5/56 with 5 adverse pain of moderate to 6 hours (5-point) 3 remedicated with events. severe intensity. slight pain, 4 uninter- pretable data, 2 took other medication, 2 did not receive study medicine, 1 lost form.

Coutinho, Urogenital RCT,double-blind, single PI (4-point scale) Dextropropoxy- Dextropropoxyphene Allowed at 4 hours No exclusions No serious adverse effects et al., surgery oral dose, parallel groups, standard phene hydro- not significantly better if no PR. If or withdrawals. were reported and no 197615 n = 15 local anaesthetic.Assessed PR (5-point scale) chloride, 65 mg than placebo (p not remedicated before patients withdrew as a result. Age: adult by observer at 0, 0.5, non-standard (n = 15); placebo given). 4 hours, patients Dextropropoxyphene, 1/15 1 hour then hourly for (n = 15) Mean SPID at 5 hours: withdrawn from study. with 1 adverse event; 5 hours. Medication dextropropoxyphene placebo, 0/15. taken when pain of 4.5, placebo 3.3. moderate to severe intensity.

Trop, et al., Postoperative RCT,double-blind, single PI (4-point scale) Dextropropoxy- Dextropropoxyphene, Did not state mini- 78 patients analysed. Authors reported significant 197917 pain, various oral dose, parallel groups, standard phene hydro 130 mg, significantly mum time allowed for 47 excluded due to difference from placebo for procedures local anaesthetic.Assessed PR (5-point scale) chloride, 65 mg better than placebo remedication. Last PR protocol violation. central nervous system n = 125 by observer at 0, 0.5, standard (n = 25); dextro- (p < 0.01). score before remedi- adverse events (p = 0.05). Age range: 1 hour then hourly for propoxyphene SPID and TOTPAR given cation used for all None serious and no 18–73 years 5 hours. Medication taken hydrochloride, at 6 hours. further time points. withdrawals. when pain of moderate to 130 mg (n = 25); TOTPAR: dextropro- Dextropropoxyphene severe intensity. placebo (n = 25) poxyphene, 65 mg, 8.54; hydrochloride, 65 mg, 19/25 dextropropoxyphene, with 27 adverse events; 130 mg, 9.03; placebo, dextropropoxyphene 2.68. hydrochloride, 130 mg, 23/25 with 34 adverse events; placebo, 10/25 with 12 adverse events.

van Staden, Gynaeco- RCT,double-blind, PI (4-point scale) Dextropropoxy- Dextropropoxyphene Allowed after 1 hour 80 patients analysed, No serious adverse effects 197116 logical surgery crossover design, general standard phene hydro- not significantly better if no PR. PR scored 11 excluded: 6 were reported and no n = 91 anaesthetic. Self-assessed PR measured as chloride, 65 mg than placebo (p not as zero for all violated protocol, patients withdrew as a result. Age: adult at 1 hour then hourly for PI difference (n = 26); placebo given). subsequent time 2 vomited, 3 had Dextropropoxyphene, 4/26 8 hours. Medication given (n = 29) SPID at 4 hours: dextro- points. insufficient pain. with 4 adverse events; when pain of moderate to propoxyphene hydro- placebo, 1/29. severe intensity. chloride, 65 mg, 1.64; placebo, 1.57.

estimate for dextropropoxyphene, 130 mg, and all were reported as transient and of compared with placebo was 10 (95% CI, 1.4–73). mild to moderate severity. One study reported The NNT was 2.8 (95% CI, 1.8–6.5) for at least 50% no adverse effects with either placebo or active pain relief over a period of 5 hours compared with treatment.12 placebo for pain of moderate to severe intensity. In one study, the authors reported that dextropro- Adverse effects poxyphene, both 65 mg and 130 mg, have a signifi- Details of adverse effects are presented in Table 46. cantly higher incidence of grogginess, sleepiness, 146 No patients withdrew as a result of adverse effects and lightheadedness than placebo (p = 0.05).17 Health Technology Assessment 1998; Vol. 2: No. 12

However, pooled data from the four trials reporting either drowsiness, sleepiness or somnolence13–15,17 At least 50% PR with dextropropoxyphene showed no significant difference in incidence alone or in combination with paracetamol (%) between dextropropoxyphene, 65 mg, (18/115) 100 and placebo (15/121), with a relative risk of 1.3 (95% CI, 0.7–24). No other trial reported lightheadedness or grogginess in the 80 dextropropoxyphene group.

Dextropropoxyphene plus paracetamol 60 versus placebo In four reports dextropropoxyphene napsylate, 100 mg, plus paracetamol, 650 mg, was compared 40 with placebo, and in one dextropropoxyphene hydrochloride, 65 mg, plus paracetamol, 650 mg. A total of 478 patients received dextropropoxy- 20 phene plus paracetamol, and 485 patients received placebo. 0 0 20406080100 One report7 was a meta-analysis of individual At least 50% PR with placebo (%) patient data from 18 studies with dichotomous information (the number of patients achieving at least 50% maxTOTPAR); eight reports FIGURE 39 Trials of oral dextropropoxyphene in postoperative investigated dextropropoxyphene napsylate, pain ( , dextropropoxyphene HCI, 65 mg; ●, dextropropoxyphene 100 mg, plus paracetamol, 650 mg. Only one (napsylate, 100 mg, or HCI, 65 mg) + paracetamol, 650 mg; of the studies had been published and the ▲, dextropropoxyphene HCI, 130 mg;) duplicate publication was excluded.11 CI, 1.9–3.1) (Table 45). For a single dose of In two reports18,19 pain following dental surgery dextropropoxyphene (hydrochloride, 65 mg, (impacted third molar) was studied, in two or napsylate, 100 mg) plus paracetamol, others20,21 pain post orthopaedic surgery, and 650 mg, the NNT was 4.4 (95% CI, 3.5–5.6) in one report7 pain following both dental and for at least 50% pain relief over 4–6 hours general surgery (abdominal, orthopaedic and compared with placebo for pain of moderate gynaecological) was studied. to severe intensity.

The placebo response rate varied between 6% Adverse effects and 27%. The dextropropoxyphene plus para- Details of adverse effects are given in Table 46. cetamol response rate varied between 25% and No patients withdrew as a result of adverse effects 57% (Figure 39). The trial results were homo- and all were reported as transient and of mild to genous (p = 0.35). Dextropropoxyphene moderate severity. In one trial,21 details of adverse (hydrochloride, 65 mg, or napsylate, 100 mg) effects were not given but it was reported that plus paracetamol, 650 mg, was significantly there was no significant difference between active superior to placebo, relative benefit 2.4 (95% and placebo groups. The individual patient meta-

TABLE 45 Summary of relative benefit and NNT for trials of dextropropoxyphene and dextropropoxyphene plus paracetamol against placebo

Number Dose of Number of patients Number of patients RB NNT of trials dextropropoxyphene with > 50% PR: with > 50% PR: (95% CI) (95% CI) dextropropoxyphene placebo

Dextropropoxyphene alone against placebo 6 65 mg 85/214 60/226 1.4 (0.97–2.0) 7.7 (4.6–∞) 1 130 mg 10/25 1/25 10.0 (1.4–73) 2.8 (1.8–6.5)

Dextropropoxyphene plus paracetamol against placebo 6 65 mg hydrochloride or 184/478 74/485 2.4 (1.9–3.1) 4.4 (3.5–5.6) 100 mg napsylate 147 Dextropropoxyphene in postoperative pain

TABLE 46 Summary of adverse effects for trials of dextropropoxyphene and dextropropoxyphene plus paracetamol against placebo

Number Adverse events Number of patients Number of patients RR NNH of trials with adverse events: with adverse events: (95% CI) (95% CI) drug placebo

Dextropropoxyphene 2 Nausea 3/75 2/81 1.6 (0.3–9.4) N/A 3 Drowsiness/sleepiness/ 18/115 15/121 1.3 (0.7–2.4) N/A somnolence 2 Headache 5/75 3/81 1.6 (0.5–4.9) N/A

Dextropropoxyphene plus paracetamol 3 Nausea 12/405 33/799 0.7 (0.4–1.4) N/A 1 Vomiting 2/323 6/714 1.4 (0.3–6.7) N/A 4 Dizziness 17/435 16/829 2.2 (1.1–4.3) 50 (24–∞) 3 Drowsiness/somnolence 57/405 55/799 2.2 (2.0–2.4) 14 (9.1–30) 4 Headache 14/435 51/829 0.5 (0.4–0.6) –33 (–170, –19)

Negative NNTs indicate that fewer headaches occur with dextroropoxyphene plus paracetamol than with placebo.

analysis7 pooled data on adverse effects from all Comment 18 placebo groups; 714 patients received placebo. For a single dose of dextropropoxyphene, 65 mg, The incidence of drowsiness or somnolence was the NNT was 7.7 (95% CI, 4.6–∞) for at least 50% reported in three studies.7,18,19 The pooled data pain relief compared with placebo. This means indicated a significantly higher incidence in the that one in every eight patients with pain of dextropropoxyphene combination group (57/405) moderate to severe intensity would experience at than in the placebo group (55/799), with a relative least 50% pain relief with dextropropoxyphene risk of 2.2 (95% CI, 2.0–2.4) and an NNH of 14 hydrochloride, 65 mg, who would not have done (95% CI, 9.1–30). so with placebo. The 95% CI included no benefit. The equivalent NNT for a single dose of dextropro- Dizziness was reported in four trials.7,18–20 Pooled poxyphene (65 mg, hydrochloride or 100 mg, data indicated a significantly higher incidence of napsylate) plus paracetamol, 650 mg, was 4.4 dizziness with dextropropoxyphene plus para- (95% CI, 3.5–5.6), indicating higher efficacy. The cetamol (17/435) than with placebo (16/829), 95% CIs of dextropropoxyphene alone and the with a relative risk of 2.2 (95% CI, 1.1–4.3) and combination with paracetamol overlapped. The an NNH of 50 (95% CI, 24–∞). dextropropoxyphene/paracetamol combination had an NNT similar to that of both paracetamol, The incidence of headache was reported in four 1000 mg, and ibuprofen, 200 mg. Both ibuprofen, trials.18–21 The pooled data showed dextropropoxy- 400 mg, and diclofenac, 50 mg, had NNTs with phene plus paracetamol (14/435) to have a signifi- 95% CIs lower (better) than that of the combi- cantly lower incidence of headache than placebo nation and which did not overlap with it. (51/829), with a relative risk of 0.5 (95% CI, 0.4–0.6) and an NNH of –33 (95% CI, –170, –19). For a single dose of dextropropoxyphene, 130 mg, the NNT was 2.8 (95% CI, 1.8–6.5). This appears Three trials considered the incidence of to show a dose response for dextropropoxyphene. nausea.7,18,19 Pooled data showed no significant However, given the overlapping CIs and the very difference with dextropropoxyphene plus para- small number of patients in the dextropropoxy- cetamol (12/405) than with placebo (33/799), phene, 130 mg, trial,50 this conclusion is relative risk 0.7 (95% CI, 0.4–1.4). not robust.

Incidence of vomiting was reported in one study.7 A single dose of dextropropoxyphene, 65 mg, plus For dextropropoxyphene plus paracetamol (2/323) paracetamol, 650 mg, showed a significantly higher it was not significantly different from placebo incidence of central nervous system adverse effects 148 (6/714), relative risk 1.4 (95% CI, 0.3–6.7). (somnolence, dizziness) than placebo (Table 46). Health Technology Assessment 1998; Vol. 2: No. 12

These adverse effects have also been shown for 10. Cook RJ, Sackett DL. The number needed to treat: tramadol, 100 mg, with lower (worse) NNHs for a clinically useful measure of treatment effect. both dizziness and somnolence.7 Tramadol, 100 BMJ 1995;310:452–4. mg, also showed a significantly higher incidence of 11. Sunshine A, Olson NZ, Zighelboim I, DeCastro nausea and vomiting than placebo. These adverse A, Minn FL. Analgesic oral efficacy of tramadol effects were reported with dextropropoxyphene, hydrochloride in postoperative pain. Clin 65 mg, plus paracetamol, 650 mg, but the inci- Pharmacol Ther 1992;51:740–6. dence was not significantly different from placebo. 12. Berry FN, Miller JM, Levin HM, Bare WW, Hopkinson JH III, Feldman AJ. Relief of severe pain The combination of dextropropoxyphene, 65 mg, with acetaminophen in a new dose formulation with paracetamol, 650 mg, showed similar efficacy versus propoxyphene hydrochloride 65 mg and to tramadol, 100 mg, for single dose studies in placebo: a comparative double blind study. Curr postoperative pain with a lower incidence of Ther Res 1975;17:361–8. adverse effects. 13. Bloomfield SS, Barden TP, Mitchell J. Nefopam and propoxyphene in episiotomy pain. Pharmacol References Ther 1980;27:502–7. 1. Collins SL, Edwards J, Moore A, McQuay HJ. Oral 14. Cooper SA, Wagenberg B, Zissu J, Kruger GO, dextropropoxyphene in postoperative pain: a Reynolds DC, Gallegos LT, et al. The analgesic quantitative systematic review. Eur J Clin Pharmacol efficacy of suprofen in peridontal and oral surgical 1998;54:107–12. pain. Pharmacotherapy 1986;6:267–76.

2. Government Statistical Service. Prescription cost 15. Coutinho A, Bonelli J, Nanci de Carvelho P. analysis for England 1995. London: Department of A double-blind study of the analgesic effects of Health, 1996. fenbufen, codeine, aspirin, propoxyphene and 3. Bruster S, Jarman B, Bosanquet N, Weston D, Erens placebo. Curr Ther Res 1976;19:58–65. R, Delbanco TL. National survey of hospital 16. van Staden MJ. The use of gilfanan in postoperative patients. BMJ 1994;309:1542–6. pain. S Afr Med J 1971;45:1235–7. 4. Jadad AR, Carroll D, Moore A, McQuay H. Developing a database of published reports of 17. Trop D, Kenny L, Grad BR. Comparison of nefopam randomised clinical trials in pain research. Pain hydrochloride and propoxyphene hydrochloride in 1996;66:239–46. the treatment of postoperative pain. Can Anaesth Soc J 1979;26:296–304. 5. Reynolds JEF. Martindale: the extra pharmacopoeia. 30th edition. London: Pharmaceutical Press, 18. Cooper SA. Double-blind comparison of zomepirac 1993:292–314. sodium, propoxyphene/acetaminophen, and placebo in the treatment of oral surgical pain. 6. Cooper SA. Single-dose analgesic studies: the upside Curr Ther Res 1980;28:630–8. and downside of assay sensitivity. Adv Pain Res Ther 1991;18:117–24. 19. Cooper SA, Breen JF, Giuliani RL. The relative efficacy of indoprofen compared with opioid- 7. Moore RA, McQuay HJ. Single-patient data meta- analgesic combinations. J Oral Surg 1981;39:21–5. analysis of 3453 postoperative patients: oral tramadol versus placebo, codeine and combination 20. Evans PJ, McQuay HJ, Rolfe M, O’Sullivan G, analgesics. Pain 1997;69:287–94. Bullingham RE, Moore RA. Zomepirac, placebo 8. Moore A, Moore O, McQuay H, Gavaghan D. and paracetamol/dextropropoxyphene Deriving dichotomous outcome measures from combination compared in orthopaedic continuous data in randomised controlled trials of postoperative pain. Br J Anaesth 1982;54:927–33. analgesics: use of pain intensity and visual analogue 21. Honig S, Murray KA. Surgical pain: zomepirac scales. Pain 1997;69:311–15. sodium, propoxyphene/acetaminophen 9. DerSimonian R, Laird N. Meta-analysis of clinical combination, and placebo. J Clin Pharmacol trials. Control Clin Trials 1986;7:177–88. 1981;21:443–8.

149

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 15 Oral tramadol, codeine and combination analgesics in postoperative pain

Summary lations allow comparisons to be made with standard analgesics. The analgesic effectiveness and safety of oral tramadol was compared with standard analgesics This chapter of the review has been published in using a meta-analysis of individual patient data full by Moore and McQuay.1 from RCTs in patients with moderate or severe pain after surgery or dental extraction. Calculation of % maxTOTPAR from individual patient data, Introduction and the use of at least 50% maxTOTPAR defined clinically acceptable pain relief. NNT for one Study of analgesics still poses problems, some patient to have at least 50% maxTOTPAR com- 40 years after Beecher first described methods pared with placebo was used to examine the of measuring pain and pain relief.2,3 These can effectiveness of different single oral doses of be of different sorts, starting with the obvious but tramadol and comparator drugs. important: the many possible comparisons of drugs, doses, routes of administration and pain A total of 18 randomised, double-blind, parallel condition which makes meaningful comparison group single-dose trials with 3453 patients using difficult. Many controlled trials have been per- categorical pain relief scales allowed the calculation formed and many published; some 10,000 RCTs of % maxTOTPAR. The use of at least 50% (over 4000 in pharmacological interventions in maxTOTPAR was a sensitive measure to acute pain) have been identified.4 discriminate between analgesics. Quantitative systematic reviews pool data from a Tramadol and comparator drugs gave significantly number of trials; while individual trials may have more analgesia than placebo. In postsurgical pain, relatively small numbers of patients receiving a tramadol, 50, 100 and 150 mg, had NNTs for at particular treatment, meta-analysis allows the result least 50% maxTOTPAR of 7.3 (95% CI, 4.6–18), to be confirmed using data from many patients in 4.8 (3.4–8.2) and 2.4 (2.0–3.1), respectively, many trials, thereby increasing the power to deter- comparable with aspirin, 650 mg, plus codeine, mine the ‘true’ result. It can therefore provide a 60 mg (NNT 3.6 (95% CI, 2.5–6.3)) and para- higher quality of evidence on which to base deci- cetamol, 650 mg, plus propoxyphene, 100 mg sions by prescribers, policy-makers and patients. (NNT 4.0 (3.0–5.7)). With the same dose of drug, postsurgical patients had more pain relief than Choice of RCTs for systematic reviews is essential. those having dental surgery. Tramadol showed a Randomisation (and concealment) of treatment dose response for analgesia in patients with both allocation limits selection bias, and blinding of postsurgical and dental pain. treatments controls observer bias. Inadequacies of randomisation or blinding exaggerate estimates of With the same dose of drug, postsurgical pain treatment effect.5 In RCTs in pain relief, standard patients had fewer adverse events than those having methods of measuring pain relief and trial conduct dental surgery. Adverse events – headache, nausea, appear to be effectively blinded.6 vomiting, dizziness, somnolence – for tramadol, 50 mg and 100 mg, had a similar incidence to Results of systematic reviews have to be easily comparator drugs. There was a dose response with understood to be useful and used. The elegant tramadol, tending towards higher incidences at NNT approach7 involves defining a clinical end- higher doses. point, and comparing the rate of that event in a treatment group with the rate in a comparator Single patient meta-analysis using more than half group; NNT calculations require dichotomous pain relief provides a sensitive description of the data. NNTs derived for particular benefits or harm analgesic properties of a drug, and NNT calcu- can provide a useful starting point for simple verbal 151 Oral tramadol, codeine and combination analgesics in postoperative pain

and numerical results accessible to any doctor by reference to the in-house data from Grünenthal or patient.8 GmbH, from Searle (UK) Ltd, and by searching MEDLINE (1960–95) and the Oxford Pain Relief Meta-analysis using individual patient data Database (1950–95)4 using tramadol as a sometimes produces lower estimates of effect of free-text term. treatment than does meta-analysis using group descriptions,9 although the generality of this has There was a prior hypothesis that analgesic been challenged.10 Using individual patient drugs may produce different analgesic responses information may not always be possible but where in painful dental procedures (such as third molar possible it is preferred11 because it is claimed to extractions) than postsurgical procedures (such as have the least bias of any meta-analytical method. abdominal, orthopaedic or gynaecological operations). The prior intention, therefore, was to Studies of pain relief may present an additional analyse these conditions separately. Included complication. The classic design of single dose oral reports were scored for inclusion and methodo- medication with both placebo and active controls logical quality using a 3-item scale.19 to demonstrate analgesic sensitivity3 is explanatory.12 Such trials provide evidence that a drug is Protocols for the RWJ studies of postsurgical an analgesic rather than information about the pain and of pain due to the extraction of impacted best way to use the compound in practice. In this third molars were essentially identical. Trials context, NNT methods are useful indicators of were of double-blind, single-dose, parallel-group relative efficacy.13 design; randomisation was by computerised random-number generation, stratified on pre- Tramadol has been used in many European treatment pain intensity. Criteria for patient countries since the late 1970s, in many different selection were moderate or severe pain and that pain conditions. Since most studies with tramadol the patient’s condition was appropriate for manage- in Europe had not been conducted according to ment with a centrally acting analgesic and para- US regulatory requirements, a completely new cetamol. The age range was from 18 to 70 years. programme of clinical studies for registration in Patients had to be cooperative, reliable and the USA of an oral tramadol formulation took motivated, and be able to take oral medication. place in the late 1980s and early 1990s. A total of Exclusion criteria included patients with mild 18 single-dose studies were conducted, nine in or no pain, those who had taken analgesic drugs dental pain models and nine in postsurgical within 3 hours of study drug administration, pain, and the results of the studies have those needing sedatives during the observation been summarised.14 period and those with known contraindications or medical conditions which might interfere We performed a single-patient data meta-analysis with observations. of these 18 studies, and any others (published or unpublished) which could be found and which had Drugs were given as single oral doses: placebo categorical pain relief scales, allowing the calcu- (695 evaluable patients); codeine, 60 mg (649); lation of the percentage of maximum pain relief tramadol, 50 mg (409); tramadol, 75 mg (281); obtained by individual patients. Combining data tramadol, 100 mg (468); tramadol, 150 mg (279); from many studies will help shed additional light tramadol, 200 mg (50); aspirin, 650 mg, plus on the debate15,16 about the conflicting results of codeine, 60 mg (305); paracetamol, 650 mg, tramadol in postoperative pain.17,18 plus propoxyphene, 100 mg (316).

Patients were given the study drug if they had Methods moderate or severe pain on a four-point categorical scale (0 = no pain, 1 = slight, 2 = moderate, Primary trials 3 = severe). Thereafter observations were made Individual patient data from 18 primary trials at 30 minutes, and 1, 2, 3, 4, 5 and 6 hours after were made available by Grünenthal GmbH, administration. Pain intensity was measured using Aachen, Germany and Robert Wood Johnson the same categorical scale, together with a five- Pharmaceutical Research Institute, Spring House, point CATPR (0 = no relief, 1 = a little, 2 = some, Pennsylvania, USA. One of these studies had been 3 = a lot, 4 = complete). Time of remedication was published.17 Other studies which used single also recorded, as well as a global assessment of doses of oral tramadol with categorical pain relief therapy (excellent, very good, good, fair or poor) 152 scoring in acute painful conditions were sought at the final evaluation. Health Technology Assessment 1998; Vol. 2: No. 12

Adverse experiences volunteered by the patient difference between tramadol, 50 and 100 mg, and after non-directive questioning were recorded placebo, used a pain intensity scoring system rather regardless of any rescue medication used. than pain relief, and therefore had to be excluded from this analysis. Another study22 using several Calculations dose levels of oral tramadol after dental surgery was For each patient the area under the curve of multiple dose and used only pain intensity scoring. pain relief (categorical scale) against time was It showed significant differences between all calculated (TOTPAR) for 6 hours after the study tramadol doses and placebo, but again could not drug was given. If patients remedicated, pain relief be used. No other relevant studies were identified. scores reverted to zero and pain intensity scores to the initial value; adverse event recording but not Analgesic efficacy pain evaluations continued after remedication. The relative benefits and NNTs for each drug The percentage of the maximum possible for tested are shown in Table 48, for dental and post- this summary measure was then calculated operative pain both separately and combined. (% maxTOTPAR).20 The number of patients The proportions of patients achieving at least on each treatment who achieved more than 50% maxTOTPAR are shown in Figure 40. There 50% maxTOTPAR was determined. was a clear dose response for tramadol (p < 0.0001, Kruskal–Wallis test). Relative benefit (which indicates how much more likely is an individual given a particular treatment Dental pain to have a specific outcome than someone not given Among the dental studies, all treatments showed the treatment) and its 95% CIs were calculated for significantly greater pain relief (greater proportion individual trials using a fixed-effects model,21 and of patients with at least 50% of % maxTOTPAR) NNT using the method of Cook and Sackett.7 The than with placebo (relative risk lower CI > 1) same method was used to calculate the NNH for except for codeine, 60 mg. There was a clear adverse effects. Relative risk and NNT are given with dose response for tramadol, with higher odds 95% CIs in text and tables. Significance testing for ratios and lower NNT values with the higher doses. dose response of tramadol was performed using the Tramadol, 100 mg and 150 mg, produced NNT Kruskal–Wallis test unstratified for type of surgery. values of 4.6 (95% CI, 3.6–6.4) and 4.1 (95% CI, 2.9–7.3) respectively, lower than aspirin/codeine (NNT 6.3 (95% CI, 4.5–9.8)) and paracetamol/ Results propoxyphene (NNT 5.3 (95% CI, 3.4–11.4)). Search results Postsurgical pain Individual patient data for 3453 patients from All treatments showed statistically significantly 18 studies was supplied. It consisted of pain superior analgesia to placebo. There was a clear intensity and pain relief scores from start of study dose response for tramadol; tramadol, 100 mg, to 8 hours post dose and aggregate adverse effect had an NNT of 4.8 (95% CI, 3.4–8.2) and tramadol, information. Studies with their codes, drug treat- 150 mg, an NNT of 2.4 (95% CI, 2.0–3.1). This was ments and numbers of patients, are presented in lower than aspirin/codeine and paracetamol/ Table 47. Data on pain measurements and adverse propoxyphene combinations with NNT values effects for these single-dose parallel-group double- of 3.5 (95% CI, 2.5–6.3) and 4.0 (95% CI, blind studies was provided. Of the nine post- 3.0–5.7), respectively. surgical pain studies, two (TR and TV, see Table 47) followed Caesarean section, and one (TX) was Dental and postsurgical pain models compared conducted with outpatients. Of the nine dental With the exception of tramadol, 100 mg, NNTs pain studies, three (TI, TI2 and TO) were con- were lower in postsurgical pain than in dental pain. ducted with outpatients. Tramadol, 200 mg, was When the numbers of patients with more than given in only one study in dental pain and these 50% pain relief were compared for each treatment data were excluded. Study reports were of high between postsurgical and dental pain (Figure 40), methodological quality, scoring the maximum some treatments produced significantly more pain of 5 points on a validated scale.19 relief in postsurgical pain. This was the case for codeine, 60 mg (relative benefit 2.4 (95% CI, Literature searches through MEDLINE found two 1.6–3.5)), tramadol, 75 mg (2.4 (95% CI, 1.5–3.8)) relevant studies of oral tramadol in postoperative and 150 mg (1.9 (95% CI, 1.4–2.6)), aspirin plus pain.17,18 The first17 formed part of the data set sup- codeine (1.6 (95% CI, 1.04–2.3)) and paracetamol plied. The other,18 which did not show a significant plus propoxyphene (1.6 (95% CI, 1.1–2.4)). 153 Oral tramadol, codeine and combination analgesics in postoperative pain

TABLE 47 Oral tramadol, codeine and combination analgesics in postoperative pain: trials, treatments and patient numbers

Postsurgical Dental

Trial Drug Patients Trial Drug Patients

TA Tramadol,50 mg 52 TE Tramadol,50 mg 28 Tramadol,100 mg 58 Tramadol,100 mg 30 Codeine, 60 mg 26 ASA 650 & C60 28 Placebo 28 Codeine, 60 mg 28 Placebo 29 TC Tramadol,50 mg 40 Tramadol,100 mg 39 TE2 Tramadol,50 mg 23 ASA 650 & C 60 40 Tramadol,100 mg 21 Codeine, 60 mg 39 ASA 650 & C60 22 Placebo 40 Codeine, 60 mg 24 Placebo 21 TR Tramadol,75 mg 40 Tramadol,150 mg 40 TF Tramadol,50 mg 47 APAP 650 & P 100 41 Tramadol,100 mg 49 Placebo 40 ASA 650 & C60 45 Codeine, 60 mg 47 TV Tramadol,50 mg 31 Placebo 49 Tramadol,100 mg 31 ASA 650 & C 60 30 TG Tramadol,50 mg 50 Codeine, 60 mg 29 Tramadol,100 mg 49 Placebo 28 ASA 650 & C60 41 Codeine, 60 mg 33 TW Tramadol,50 mg 40 Placebo 27 Tramadol,100 mg 40 APAP 650 & P100 39 TH Tramadol,50 mg 47 Codeine, 60 mg 39 Tramadol,100 mg 51 Placebo 40 ASA 650 & C60 51 Codeine, 60 mg 48 TW2 Tramadol,75 mg 41 Placebo 50 Tramadol,150 mg 40 APAP 650 & P100 39 TI Tramadol,50 mg 51 Codeine, 60 mg 41 Tramadol,100 mg 50 Placebo 40 ASA 650 & C60 48 Codeine, 60 mg 50 TX Tramadol,75 mg 35 Placebo 49 Tramadol,150 mg 36 APAP 650 & P100 37 TI2 Tramadol,75 mg 45 Codeine, 60 mg 33 Tramadol,150 mg 45 Placebo 36 APAP 650 & P100 42 Codeine, 60 mg 47 TY Tramadol,75 mg 31 Placebo 44 Tramadol,150 mg 28 APAP 650 & P100 31 TO Tramadol,100 mg 50 Codeine, 60 mg 30 Tramadol,200 mg 50 Placebo 30 Codeine, 60 mg 47 Placebo 5 TZA Tramadol,75 mg 39 Tramadol,150 mg 40 TQ Tramadol,75 mg 50 APAP 650 & P100 38 Tramadol,150 mg 50 Codeine, 60 mg 38 APAP 650 & P100 49 Placebo 41 Codeine, 60 mg 50 Placebo 51

ASA 650 & C60 = aspirin, 650 mg, plus codeine, 60 mg 154 APAP 650 & P100 = paracetamol, 650 mg, plus propoxyphene, 100 mg Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 48 Analgesic effectiveness of oral tramadol, codeine and combination analgesics in postoperative pain

Improved on active Improved on control RB (95% CI) NNT (95% CI)

Dental Codeine, 60 mg 36/374 28/373 1.3 (0.8–2.1) 47.2 (16.3–∞) Tramadol,50 mg 41/246 13/225 2.9 (1.6–5.2) 9.2 (6.1–18.8) Tramadol,75 mg 16/95 6/95 2.7 (1.1–6.5) 9.5 (5.1–64.5) Tramadol,100 mg 89/300 22/278 3.8 (2.4–5.8) 4.6 (3.6–6.4) Tramadol,150 mg 29/95 6/95 4.8 (2.1–11.1) 4.1 (2.9–7.3) Paracetamol, 650 mg, and propoxyphene, 100 mg 23/91 6/95 4.0 (1.7–9.4) 5.3 (3.4–11.4) Aspirin, 650 mg, and codeine, 60 mg 52/235 13/225 3.8 (2.2–6.8) 6.3 (4.5–9.8)

Postsurgical Codeine, 60 mg 63/275 35/283 1.9 (1.3–2.7) 9.5 (6–23.4) Tramadol,50 mg 38/163 13/136 2.4 (1.4–4.4) 7.3 ( 4.6–17.9) Tramadol,75 mg 74/186 31/187 2.4 (1.7–3.5) 4.3 (3.1–7) Tramadol,100 mg 51/168 13/136 3.2 (1.8–5.6) 4.8 (3.4–8.2) Tramadol,150 mg 106/184 31/187 3.5 (2.5–4.9) 2.4 (2–3.1) Paracetamol, 650 mg, and propoxyphene, 100 mg 91/225 34/227 2.7 (1.9–3.8) 3.9 (3–5.7) Aspirin, 650 mg, and codeine, 60 mg 24/70 4/68 5.8 (2.1–15.9) 3.5 (2.5–6.3)

Combined Codeine, 60 mg 99/649 63/656 1.6 (1.2–2.1) 16.7 (11–48) Tramadol,50 mg 79/409 26/361 2.7 (1.8–4.1) 8.3 (6.0–13) Tramadol,75 mg 90/281 37/282 2.4 (1.7–3.5) 5.3 (3.9–8.2) Tramadol,100 mg 140/468 35/414 3.5 (2.5–5.0) 4.8 (3.8–6.1) Tramadol,150 mg 135/279 37/282 3.7 (2.7–5.1) 2.9 (2.4–3.6) Paracetamol, 650 mg, and propoxyphene, 100 mg 114/316 40/322 2.9 (2.1–4.0) 4.2 (3.3–5.8) Aspirin, 650 mg, and codeine, 60 mg 76/305 17/293 4.3 (2.6–7.1) 5.3 (4.1–7.4)

Placebo Codeine, 60 mg Tramadol, 50 mg Tramadol, 75 mg Tramadol, 100 mg Tramadol, 150 mg Aspirin, 650 mg, and codeine, 60 mg Paracetamol, 650 mg, and propoxyphene, 100 mg

0 102030405060 Patients with at least 50% maxTOTPAR (%)

FIGURE 40 Proportion of patients with at least 50% maxTOTPAR ( , dental; , postoperative)

This difference was not due to a greater proportion less severe initial pain in the dental group, relative of postsurgical patients with more moderate than risk 0.81 (95% CI, 0.77–0.85)). Initial pain intensity severe initial pain intensity. The ratio of moderate stratification produced no consistent or signif- to severe initial (baseline) pain intensity was icant differences in the proportion of patients 931:663 (1.40:1) in postsurgical pain compared with at least 50% maxTOTPAR, or in NNTs with 1294:505 (2.56:1) in dental pain (significantly (data not shown). For postsurgical pain, 44 of 155 Oral tramadol, codeine and combination analgesics in postoperative pain

323 patients (13.6%) given placebo had at rose rapidly and were not significantly different from least 50% maxTOTPAR, compared with 28 of placebo by > 60% maxTOTPAR. 373 patients (7.5%) for dental pain patients given placebo (relative benefit 1.8 (95% CI, 1.2–2.8). Adverse events The incidence of the more common adverse events Combined data reported is shown in Figure 42 for dental and NNTs for dental and postsurgical patients postsurgical pain. Headache, vomiting, nausea, combined are also shown in Table 48. Few data sets dizziness and somnolence were the most commonly have sufficient information to allow calculation of reported adverse events, though predominantly of analgesic efficacy for dental and postsurgical pain mild intensity. models separately. These numbers are those used for comparisons with other analgesic drugs from For dental but not postsurgical pain, the adverse published reports without individual meta-analysis. event incidence was generally sufficiently high to achieve a statistical difference from placebo for Choice of half pain relief vomiting, nausea, dizziness and somnolence, but In order to test the effect of choices other than not headache. For tramadol there was a distinct half pain relief, NNTs were calculated using dichot- dose response in dental pain, with higher doses omous data for 20–80% maxTOTPAR (combined producing greater incidence of adverse events; this data). The results are shown in Figure 41. trend was not present in postsurgical pain. NNHs can be calculated for adverse effects in dental NNTs for an effective drug, tramadol, 150 mg, were patients because their incidence was sufficiently essentially the same, at about 2–3, over a wide range high. The clear dose response is shown in of decision points. Those for a slightly less effective Figure 43. analgesic (aspirin, 650 mg, plus codeine, 60 mg) rose slightly with pain relief cut-off values above 50% maxTOTPAR but for codeine, 60 mg, NNT values Comment which started at about 10 for > 20% maxTOTPAR Tramadol is an effective analgesic in postoperative pain. All doses of tramadol were statistically superior to placebo in both postsurgical and dental pain, NNT 60 and there was a significant dose response. Single oral doses of tramadol, 75 mg to 150 mg, had analgesic efficacy equivalent to combinations of paracetamol plus propoxyphene and aspirin 50 plus codeine. Internal sensitivity was demonstrated by two comparator analgesics being statistically superior to placebo, and by the dose response 40 for tramadol. The study methodology, randomised, double-blind trials, avoided known sources of major bias. 30 Search strategies identified a total of 20 randomised trials of oral tramadol in postoperative acute 20 pain with standardised measurements of pain intensity and pain relief. Only two of these had been published in full17,18 with one other in press.22 10 Results of the other trials had been published in summary form only.14 Meta-analytic tools so far developed have concentrated on patients achieving 0 02040 60 80 100 at least 50% pain relief as a single dichotomous measure of clinical effectiveness,13 so the two % maxTOTPAR studies which used pain intensity and not pain relief scales could not be included. Of the two 18 FIGURE 41 Effect of % maxTOTPAR on NNT(––––, paraceta- studies excluded, one could not distinguish oral mol 650 mg/propoxyphene 100 mg; ...... , aspirin 650 mg/codeine tramadol from placebo after orthopaedic surgery; 60 mg; – – –, codeine 60 mg; ––––, tramadol 150 mg; the other22 found oral tramadol effective in dento- 156 - - -, tramadol 100 mg; - . - . - ., tramadol 50 mg) alveolar surgical pain. Health Technology Assessment 1998; Vol. 2: No. 12

Because the only two studies published in full17,18 came to contrary conclusions about the efficacy of oral tramadol in postoperative pain, a controv- Placebo ersy has arisen about its analgesic properties.15,16 Codeine, 60 mg Examining all the available information (published Tramadol, 50 mg Tramadol, 75 mg and unpublished) has demonstrated clear anal- Tramadol, 100 mg gesic efficacy; three standard analgesics were Tramadol, 150 mg distinguished from placebo, as were tramadol, Aspirin, 650 mg, and codeine, 60 mg 50, 100 and 150 mg. Larger doses of tramadol Paracetamol, 650 mg, and propoxyphene, 100 mg produced more analgesia. This was performed 0 5 10 15 20 25 in a single patient meta-analysis, a method which Headache (%) is claimed to be more conservative than 9 aggregating mean data. Placebo Codeine, 60 mg Sunshine,14 Cooper20 and others have pointed out Tramadol, 50 mg the variability that can occur in clinical trials of Tramadol, 75 mg analgesics even in standard settings. This variability Tramadol, 100 mg may have a number of causes. One is almost certain Tramadol, 150 mg Aspirin, 650 mg, and codeine, 60 mg to be caused by the random play of chance in Paracetamol, 650 mg, and clinical trials where group sizes are of the order of propoxyphene, 100 mg 30 patients, although there may also be systematic 0 5 10 15 20 25 causes. These may be apparent only in the syste- Nausea (%) matic examination of large numbers of clinical trials with common endpoints. Placebo Codeine, 60 mg Tramadol, 50 mg The other consideration is that analgesic trial Tramadol, 75 mg 12 designs are explanatory. They are designed to Tramadol, 100 mg demonstrate that a particular compound is an Tramadol, 150 mg analgesic in single doses in acute pain. They cannot Aspirin, 650 mg, and codeine, 60 mg in themselves determine the value of the inter- Paracetamol, 650 mg and propoxyphene, 100 mg vention in clinical practice, although meta-analysis 0 5 10 15 20 25 of such trials may be helpful in determining Dizziness (%) relative efficacy.13 Placebo This unique opportunity to use the individual Codeine, 60 mg patient data from 18 trials conducted to a common Tramadol, 50 mg Tramadol, 75 mg protocol allowed several questions to be addressed. Tramadol, 100 mg The first, the original purpose of the studies, was to Tramadol, 150 mg compare the efficacy and adverse effects of the Aspirin, 650 mg, and codeine, 60 mg novel analgesic with placebo and standard oral Paracetamol, 650 mg and propoxyphene, 100 mg analgesics. The data also allowed the confirmation 0 5 10 15 20 25 of the usefulness of at least 50% pain relief as an Vomiting (%) indicator of efficacy, comparison of the various analgesics in patients with either moderate or Placebo severe baseline pain and comparison of dental Codeine, 60 mg with postsurgical pain. Tramadol, 50 mg Tramadol, 75 mg As a clinical outcome, 50% relief of pain has Tramadol, 100 mg historical provenance over 40 years,2 and is more Tramadol, 150 mg readily clinically interpretable than the summary Aspirin, 650 mg, and codeine, 60 mg Paracetamol, 650 mg, and TOTPAR measure. For comparisons of analgesics, propoxyphene, 100 mg the question arises whether the 50% relief is a 0 5 10 15 20 25 better cut-off than 20% or 80% relief. The best Somnolence (%) performing analgesic, tramadol, 150 mg, had NNT values of 2–3 across a range of cut-offs, from 30% to FIGURE 42 Incidence of adverse events with oral tramadol – 60% maxTOTPAR (Figure 43). The least analgesic headache, nausea, vomiting, dizziness and somnolence ( , dental; drug, codeine, 60 mg, showed a rapid rise in NNT , postoperative) 157 Oral tramadol, codeine and combination analgesics in postoperative pain

NNH for nausea NNH for vomiting 1 1 6 6 11 11 16 16 21 21 26 26 100 100 Oral tramadol (mg) Oral tramadol (mg)

NNH for dizziness NNH for somnolence 1 1 6 6 11 11 16 16 21 21 26 26 100 100 Oral tramadol (mg) Oral tramadol (mg)

FIGURE 43 NNH for adverse events in dental patients

beyond 50% maxTOTPAR, and the aspirin– preted; patients with less than 50% maxTOTPAR codeine combination showed a gradual rise in NNT can also obtain useful pain relief; conversely those with higher cut-offs. This provides some empirical with at least 50% maxTOTPAR may have near support for the use of 50% as the cut-off. Not only maximal pain relief. The reality, though, is that does it have a clinically useful resonance but it also multiple dosing is the norm in pain management, provides sensible discrimination between the best where adverse effects may drive practice as much and worst analgesics. as analgesia.

That the NNT (the reciprocal of the absolute It has been suggested that differences might be risk reduction, or risk difference) should be seen between analgesics when tested on pain of relatively unaffected by choice of cut-off is not initial moderate as opposed to severe intensity. unexpected. With placebo the proportion of This was not supported by these data. Stratification patients achieving a particular level of pain relief by initial pain intensity revealed no consistent or falls quickly as % maxTOTPAR increases. For significant differences in the proportion of effective analgesics, this proportion falls slowly patients with at least 50% maxTOTPAR. until high % maxTOTPAR levels are reached. The difference will remain largely unaltered In these trials the analgesics were more effective over a wide range of % maxTOTPAR – in postsurgical pain than in dental pain (Table 48), generating stable NNTs. producing lower NNTs despite there being significantly more patients with severe pain intensity at The imposition of an arbitrary dichotomous baseline in postsurgical pain. In postsurgical trials, outcome, at least 50% maxTOTPAR, on continuous significantly more patients given placebo (14%) data – a spectrum of response between no pain had at least 50% pain relief than was the case with relief and complete pain relief – is justified because dental models (8%). These average figures for nine it allows analgesics to be compared across many trials in each group are lower than those found 158 different trials. However, it should not be overinter- by Cooper.20 Health Technology Assessment 1998; Vol. 2: No. 12

Cooper’s figures were derived from study mean 3. Houde R, Wallenstein S, Beaver W. Clinical TOTPAR. McQuay and colleagues6 have pointed measurement of pain. In: de Stevens G, editor. out that means are inadequate descriptors of Analgetics. New York: Academic Press, 1965; 75–122. asymmetrically distributed pain measurements, 4. Jadad AR, Carroll D, Moore A, McQuay H. making comparison between estimates derived by Developing a database of published reports of single patient meta-analysis difficult. What Cooper’s randomised clinical trials in pain research. data did show was the great between-trial variability Pain 1996;66:239–46. of placebo and active responses. This variability is not limited to acute pain studies, and is seen also 5. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias: dimensions of methodo- in chronic pain studies,23 as well as in studies with logical quality associated with estimates of treatment more objective outcomes like postoperative effects in controlled trials. JAMA 1995;273:408–12. vomiting,24 and in the response of infants to pulmonary surfactant.25 6. McQuay H, Carroll D, Moore A. Variation in the placebo effect in randomised controlled trials of Despite analysing results on nearly 3500 patients, analgesics: all is as blind as it seems. Pain there were only 18 trials, nine each in dental and 1996;64:331–5. postsurgical pain. In order to make definitive 7. Cook RJ, Sackett DL. The number needed to treat: statements about differences between pain models a clinically useful measure of treatment effect. information from many more trials would need to BMJ 1995;310:452–4. be available. Cooper’s 1991 analysis20 had information from as many as 63 studies. The differences 8. Chalmers I. Applying overviews and meta-analyses at in analgesic efficacy and adverse events seen in the bedside. J Clin Epidemiol 1995;48:67–70. this study support the view that dental and post- 9. Stewart LA, Parmar MKB. Meta-analysis of the operative pain should be considered separately in literature or of individual patient data: is there a meta-analytical comparisons of analgesic efficacy, difference? Lancet 1993;341:418–22. at least when opioid analgesics or combinations with opioids are used. 10. Harvey I, Peters TJ, Toth B. Meta-analysis [letter]. Lancet 1993;341:964.

Analyses of other analgesics in postsurgical and 11. Oxman AD, Clarke MJ, Stewart LA. From science to dental pain models are needed to allow compari- practice. Meta-analyses using individual patient data sons of relative effectiveness to be made. This will are needed. JAMA 1995;274:845–6. not be easy, partly because few studies report data in ways which allow meta-analysis of the published 12. Schwartz D, Lellouch J. Explanatory and pragmatic attitudes in therapeutic trials. J Chronic Dis reports, and partly because many patients are 1967;20:637–48. needed to obtain estimates with narrow CIs. Single patient meta-analysis is the most useful 13. Moore A, McQuay H, Gavaghan D. Deriving method of generating comparative information. dichotomous outcome measures from continuous However, it will involve much cooperation data in randomised controlled trials of analgesics. between clinical investigators and sponsoring Pain 1996;66:229–37. pharmaceutical companies. 14. Sunshine A. New clinical experience with tramadol. Drugs 1994;47 suppl 1:8–18. Authors of reports of trials of analgesics can aid future meta-analysis by including dichotomous 15. Vielvoye-Kerkmeer APE. Comments on Stubhaug, outcomes as part of their analysis and report. Grimstad and Breivik: Pain 1995;62:111–18 [letter]. This can easily be done as an addition to, not Pain 1996;64:401–2. to the exclusion of, classical pain measures 6 16. Stubhaug A, Grimstad J, Breivik H. Reply to and analysis. Vielvoye-Kerkmeer [letter]. Pain 1996;64:402.

17. Sunshine A, Olson NZ, Zighelboim I, DeCastro A, References Minn FL. Analgesic oral efficacy of tramadol hydrochloride in postoperative pain. Clin Pharmacol 1. Moore RA, McQuay HJ. Single-patient data meta- Ther 1992;51:740–6. analysis of 3453 postoperative patients: oral tramadol versus placebo, codeine and combination 18. Stubhaug A, Grimstad J, Breivik H. Lack of analgesic analgesics. Pain 1997;69:287–94. effect of 50 and 100 mg oral tramadol after orthopaedic surgery: a randomized, double-blind, 2. Denton J, Beecher H. New analgesics. placebo and standard active drug comparison. JAMA 1949;141:1051–7. Pain 1995;62:111–18. 159 Oral tramadol, codeine and combination analgesics in postoperative pain

19. Jadad AR, Moore RA, Carroll D, Jenkinson C, 23. McQuay HJ, Tramèr M, Nye BA, Carroll D, Reynolds DJM, Gavaghan DJ, et al. Assessing the Wiffen PJ, Moore RA. A systematic review of quality of reports of randomized clinical trials: is antidepressants in neuropathic pain. Pain blinding necessary? Control Clin Trials 1996;17:1–12. 1996;68:217–27.

20. Cooper SA. Single-dose analgesic studies: the upside 24. Tramèr M, Moore A, McQuay H. Prevention of and downside of assay sensitivity. Adv Pain Res Ther vomiting after paediatric strabismus surgery: a 1991;18:117–24. systematic review using the numbers-needed-to- 21. Gardner MJ, Altman DG. Statistics with confidence. treat method. Br J Anaesth 1995;75:556–61. London: British Medical Journal, 1989. 25. Soll JC, McQueen MC. Respiratory distress 22. Collins M, Young I, Sweeney P, Fenn GC, Stratford syndrome. In: Sinclair JC, Bracken ME, editors. ME, Wilson A, et al. The effect of tramadol on Effective care of the newborn infant. Oxford: dento-alveolar surgical pain. Br J Oral Maxillofac Oxford University Press, 1992;chapter 15, Surg; in press. p. 333.

160 Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 16 Pain relief from intra-articular morphine after knee surgery

Summary studies and reviews have been published on this subject, consensus on whether intra-articular Reduction of postoperative pain by injecting opioids offer clinically relevant pain relief is opioid into the knee joint is believed to support the still lacking. hypothesis of peripheral opioid receptor activation in inflammation. This systematic review of RCTs The issue of the sensitivity of analgesic measure- was designed to examine the evidence for this. ment is particularly important. Over 40 years ago, Main outcomes were pain intensity and the use of Beecher3 and Houde4 described methods for supplementary analgesics. Efficacy of intra-articular measuring analgesic drugs which were sensitive and bupivacaine against placebo was used as an index of reproducible. Sensitive analgesic assays depended internal sensitivity. Evidence of efficacy was sought upon patients experiencing pain of moderate or in both early (0–6 hours after intra-articular severe intensity before test drug administration. injection) and late (6–24 hours) periods. The aim of this systematic review, using the In all, 36 RCTs in knee surgery were found. Six had evidence from all RCTs, was to investigate the both a local anaesthetic control and placebo; four evidence for an analgesic effect of intra-articular showed internal sensitivity and had at least one morphine and to examine those features of trial outcome showing efficacy of intra-articular methodology which influence judgement of morphine against placebo. Six studies compared experimental or clinical effectiveness. intra-articular morphine with intravenous or intramuscular morphine or with intra-articular saline without a bupivacaine control. Four of the Methods six studies showed greater efficacy for intra- articular morphine. There was no dose response RCTs of intra-articular opioids were sought evident. No quantitative analysis of pooled data systematically. A number of different search was undertaken. strategies in both MEDLINE (1966–May 1996), EMBASE and the Oxford Pain Relief Database Intra-articular morphine may have some effect (1950–94) were used, without language restriction. in reducing postoperative pain intensity and Search terms used included ‘intra-articular’, consumption of analgesics. These studies had ‘opiates’, ‘opioids’ and ‘morphine’ and ‘random*’.5 significant problems in design, data collection, Additional reports were identified from the refer- statistical analysis and reporting. Trials of better ence lists of retrieved reports and from review methodological quality are needed for a conclusive articles. Unpublished reports, abstracts and reviews answer that intra-articular morphine is analgesic, were not considered. Authors were not contacted and that any analgesia produced is clinically useful. for original data.

This chapter of the review has been published in Reports considered full by Kalso and colleagues.1 Reports were considered if they were randomised comparisons of intra-articular morphine with placebo (saline), or different doses of intra- Introduction articular morphine, or comparisons of intra- articular morphine with systemic (intravenous Intra-articular morphine has been used as a clinical or intramuscular) morphine. Reports of direct test of the hypothesis that peripheral opioid comparisons of intra-articular morphine and receptors are activated in inflammation.2 The local anaesthetic agents6,7 were not considered. judgement that exogenous opioids can provide Reports of pethidine8 were not considered effective postoperative analgesia has been taken as because of potential confounding due to its confirmation of the hypothesis.2 Even though many local anaesthetic properties. 161 Pain relief from intra-articular morphine after knee surgery

Each report which could possibly be described as an • duplicate publications11,12 RCT was read independently by each of the authors • the influence of tourniquet time on the efficacy and scored using a three-item quality scale.9 The of intra-articular morphine as the only outcome13 scale takes into account proper randomisation, • number of patients per group less than ten14,15 double-blinding and reporting of drop-outs and • double-dummy technique not used for withdrawals. Consensus was then achieved. Inform- intramuscular administration16 (study 2) ation was taken from each report on treatments and • control group not blinded17 controls, types of surgery and anaesthesia, number • controls were intra-articular bupivacaine and of patients enrolled and analysed, study design, unblinded lumbar plexus block only18 observation periods, outcome measures used for • spinal anaesthesia19 (study 1)20 pain intensity and consumption of supplementary • epidural anaesthesia21 analgesics and adverse effects. • operative intra-articular local anaesthetic11,20,22,23 (study 2)24 Validity and inclusion criteria • non-standardised anaesthesia (general Pre-hoc validity criteria were number of patients per anaesthesia, spinal or epidural anaesthesia)25 treatment group ≥ 10,10 standardised methods of • inadequate standardisation of the timing of measuring pain intensity, and general anaesthesia. pain measurements.26 Spinal or epidural anaesthetics were not accepted, nor were infiltrations of local anaesthetic into the Details of the included studies are shown in joints, because it was judged that low pain scores in Table 49. In all these trials morphine, 0.5–5 mg, the immediate postoperative period could render was used as the intra-articular opioid. Controls studies insensitive. used were bupivacaine, 0.25–0.5%, as the only intra-articular local anaesthetic, intra-articular Two periods, early (up to 6 hours from the intra- saline or intravenous or intramuscular morphine, articular injection) and late (6–24 hours) were 1–2 mg. No quantitative analysis of pooled data defined for the evaluation of effectiveness. was performed because results were presented as means, which are inadequate descriptors of Effectiveness was defined as a significant difference assymmetrically distributed data.27 (as reported in the original trials) between the active and the control in pain intensity (early and Morphine versus saline with bupivacaine late) or total consumption of rescue analgesics. as an index of internal sensitivity Six studies compared intra-articular morphine with There was a pre-hoc agreement that an adequate both bupivacaine and saline (Table 49). One28 was description of internal sensitivity was a requirement only analysed for an early effect (fewer than ten for the demonstration of an analgesic action of evaluable patients in the late period). intra-articular morphine. Such sensitivity would be derived (not necessarily exclusively) from a statis- In two studies,16,39 intra-articular bupivacaine could tically significant difference between a known not be differentiated from intra-articular saline and analgesic (intra-articular local anaesthetic) and the sensitivity of the analgesic assay was not proven. placebo, from intra-articular morphine being There was no difference between intra-articular different from placebo, or from a dose response morphine and saline in either. for intra-articular morphine. Four trials28–31 showed significantly lower VAS pain Quantitative analysis of morphine against placebo intensity scores with intra-articular bupivacaine com- was planned. pared with intra-articular saline during the early period (0–6 hours) and so had internal sensitivity.

Results All four sensitive studies reported early outcomes. Three of the four studies showed significantly lower A total of 33 RCTs were found in 31 reports, early pain intensity scores after intra-articular studying nearly 1500 patients (about 900 of whom morphine compared with intra-articular saline28,29,31 received morphine). All were in knee surgery. Two (Figure 44A). reports were in Danish, one in German and the rest in English. In the late period, from 6 hours onwards, intra- articular morphine produced significantly lower The reasons for exclusion of studies were pain intensity scores compared with placebo in all 162 as follows: three evaluable sensitive studies29–31 (Figure 44B). Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 49 Pain relief from intra-articular morphine after knee surgery: included studies

Study Included or reason Quality Drugs and routes Results: compared with placebo (saline) for exclusion (1–5) (number of patients)

VAS PI: VAS PI: VAS PI: VAS PI: Analgesic early bupivacaine late bupivacaine early morphine late morphine consumption: 24-hour total

Included studies with active control (bupivacaine) and placebo

Björnsson, Included 2 i.a. morphine, 1 mg, in normal No difference No difference at 8, No difference No difference at No difference. et al., 199416 saline, 20 ml (21) between bupivacaine 24 or 48 hours. between morphine 8, 24 or 48 hours. study 1 i.a. normal saline, 20 ml, (19) and normal saline and normal saline i.a. 0.25% bupivacaine, 20 ml, (19) at 0.5, 1, 1.5 or at 0.5, 1, 1.5 or i.a. 0.25% bupivacaine, 20 ml, + 2 hours. 2 hours. morphine, 1 mg (19)

Haynes, et al., Included in early, 3 i.a. normal saline, 40 ml (10) Bupivacaine better n < 10 Morphine better n < 10 n < 10 199428 excluded from late i.a. morphine, 1 mg, in 39 ml than normal saline: than normal saline: (inadequate number of normal saline (10) at 2 hours, p = 0.01; at 2 hours, p = 0.01; patients per group) i.a. 0.25% bupivacaine, 40 ml, + 1 at 4 hours, p < 0.05; at 4 hours p < 0.05; in 200,000 adrenaline (10) at 6 hours, p < 0.05 at 6 hours, p < 0.05. i.a. 0.25% bupivacaine, 40 ml, + 1 in 200,000 adrenaline + morphine, 1 mg (10)

Joshi, et al., Included 2 i.a. morphine, 5 mg, in 25 ml (10) Bupivacaine better NSD (8 or 24 hours). Morphine better Morphine better Significance is not 1993a29 i.a. 0.25% bupivacaine, 25 ml (10) than normal saline than normal saline than normal saline mentioned. i.a. morphine, 5 mg, in 0.25% at 1, 2, 4 hours. No at 1, 2, 4 hours; at 8, 24 hours; bupivacaine, 25 ml (10) p-values given. statistics as for statistics as for i.a. normal saline, 25 ml (10) bupivacaine. bupivacaine.

Karlsson, et al., Included 4 i.a. morphine, 1 mg, in 20 ml (10) Bupivacaine better NSD (24 or NSD (2, 4, 6 hours). Morphine better Morphine 199530 i.a. 0.375% bupivacaine, 20 ml, (10) than normal saline 48 hours). than normal saline significantly better i.a. morphine, 1 mg, in 0.375% at 2, 4, 6 hour; no at 24, 48 hours; than normal saline bupivacaine, 20 ml (10) actual p-values given. statistics as for (0–24 hours and i.a. normal saline, 20 ml (10) bupivacaine. 24–48 hours). No p-value given.

McSwiney, et al., Included 2 i.a. morphine, 5 mg, in 0.25% Bupivacaine better Bupivacaine better Morphine better Morphine better Morphine and 199331 bupivacaine, 12.5 ml, + normal than normal saline than normal saline than normal saline than normal saline bupivacaine saline, 12.5 ml (10) at: 0.5, 1, 1.5, 2, at 8, 12 hours; no at 0.5, 1, 1.5, 2, at 8, 12, 24 hours; significantly i.a. 0.25% bupivacaine, 25 ml (10) 4 hours; no SEM/SD SEM/SD bars; no 4 hours; no SEM/SD no SEM/SD bars; better (p < 0.05) i.a. morphine, 5 mg, in normal bars; no p-values. p-values. bars; no p-values. no p-values. than normal saline. saline, 25 ml (10) i.a. normal saline, 25 ml (10)

Included studies with placebo but no local anaesthetic as active control

Joshi, et al., Included 2 i.a. morphine, 5 mg, in 25 ml i.a. morphine better i.a. morphine i.a. morphine 199232 normal saline (10) than i.a. normal better than i.a. better than i.a. i.a. normal saline, 25 ml (10) saline at 0, 0.5, 1, normal saline at normal saline; 1.5, 2 and 4 hours; 8 and 12 hours; p < 0.05. p < 0.05 p < 0.05.

Joshi, et al., Included 2 i.a. morphine, 5 mg, in 25 ml NSD (1, 2 or NSD (8 or i.a. morphine 199340 normal saline (10) 4 hours). 24 hours). better than i.a. i.a. normal saline, 25 ml (10) normal saline, p < 0.01.

Lyons, et al., Pethidine arm 3 i.a. morphine, 5 mg, in 25 ml i.a. morphine i.a. morphine 199541 not considered normal saline (20) better than i.a. better than i.a. i.a. normal saline, 25 ml (20) normal saline at normal saline at i.a. pethidine, 50 mg, in 25 mg 0, 0.5, 1, 2 and 8, 12 and 24 hours; normal saline (20) 4 hours; p < 0.01. p < 0.01.

Included studies; cross-route morphine

Dierking, et al., Included 4 i.a. morphine, 2 mg, in normal NSD (1, 2, 4, Not evaluated. Not evaluated. 199433 saline, 40 ml, + i.m. normal saline, 6 hours). 1 ml (18) i.a. normal saline, 40 ml, + i.m. morphine, 2 mg (15)

continued 163 Pain relief from intra-articular morphine after knee surgery

TABLE 49 contd Pain relief from intra-articular morphine after knee surgery: included studies

Study Included or reason Quality Drugs and routes Results: compared with placebo (saline) for exclusion (1–5) (number of patients)

VAS PI: VAS PI: VAS PI: VAS PI: Analgesic early bupivacaine late bupivacaine early morphine late morphine consumption: 24-hour total

Included studies; cross-route morphine contd

Hege-Scheuing, Included 3 i.a. morphine, 1mg, in 10 ml + i.v. NSD (1, 2, 3, 4, NSD (8 or No difference. et al., 199534 normal saline, 10 ml (29) 6 hours). 24 hours). i.v. morphine, 1 mg, in 10 ml + i.a. normal saline, 10 ml (30)

Stein, et al., Included (low dose 2 i.a. morphine, 1 mg, in normal i.a. morphine better NSD (24 hours). i.a. morphine 199135 morphine excluded), saline, 40 ml, + i.v. normal saline, than i.v. morphine significantly inadequate number 1 ml (18) at 3, 4, 6 hours; better than i.v. of patients i.a. normal saline, 40 ml, + i.v. p < 0.05. morphine. morphine, 1 mg (15) i.a. morphine, 0.5 mg, in normal saline, 40 ml, + i.v. normal saline, 1 ml (10) i.a. morphine, 1 mg, + naloxone, 0.1 mg, in normal saline, 40 ml, + i.v. normal saline, 1 ml (9)

Included studies: morphine dose response

Allen, et al., Included 5 i.a. 0.25% bupivacaine, 30 ml (30) 199336 i.a. morphine, 1 mg, in normal saline, 30 ml (30) i.a. morphine, 2 mg, in normal saline, 30 ml (30) i.a. morphine, 1 mg, in 0.25% bupivacaine, 30 ml (30)

Heine, et al., Included 2 i.a. 0.5% bupivacaine, 20 ml (11) 199437 i.a. morphine, 1 mg, in 0.5% bupivacaine, 20 ml (10) i.a. morphine, 3 mg, in 0.5% bupivacaine, 20 ml (10)

Juelsgaard, Included 3 i.a. morphine, 2 mg, in 5 ml (25) et al., 199323 i.a. morphine, 4 mg in 5 ml (25)

Laurent, et al., Included 4 i.a. morphine, 5 mg, in 0.25% 199438 bupivacaine, 40 ml (20) i.a. morphine, 2 mg, in 0.25% bupivacaine, 40 ml (20) i.a. 0.25% bupivacaine, 40 ml (18)

Included studies: not analysed

Khoury, et al., Included (early), 3 i.a. morphine, 1 mg, in 20 ml (11) 19926 excluded (late) i.a. 0.25% bupivacaine, 20 ml (11) inadequate number i.a. morphine, 1 mg, + 0.25% of patients per group bupivacaine, 20 ml (11)

VanNess, et al., Included 1 i.a. 0.25% bupivacaine, 30 ml, + 1 19947 in 200 k adrenaline (41) i.a. morphine, 2 mg (2 ml), + normal saline, 28 ml (40)

Total consumption of supplementary analgesics In the early period, morphine VAS pain intensity over 24 hours was significantly lower after intra- scores were significantly lower in two of the three articular morphine compared with saline in the studies which reported early outcomes32,41 two sensitive studies in which it was analysed.30,31 (Figure 44C).

Morphine versus saline, no active In the late period, the same two studies32,41 (bupivacaine) control indicated that intra-articular morphine produced Three studies compared only intra-articular significantly lower pain intensity scores compared 164 morphine with intra-articular saline.32,40,41 with saline. Two of the three studies32,40 had a Health Technology Assessment 1998; Vol. 2: No. 12

Studies with active (local anaesthetic) control

A: Early (4-hour) PI B: Late (24-hour) PI 50 50

equality equality 40 40

30 30 Björnsson et al., 1994

Karlsson et al., 1995 20 Aasbø et al., 1996 20 Haynes et al., 1994 Aasbø et al., 1996 10 McSwiney et al.,1993 10 VAS PI with morphine (mm) VAS McSwiney VAS PI with morphine (mm) VAS et al., 1993 Joshi et al., Joshi et al., 1993a 0 1993a 0 01020304050 01020304050

VAS PI with control – saline (mm)

Studies without active (local anaesthetic) control

C: Early (4-hour) PI D: Late (24-hour) PI 50 50

equality equality 40 40

30 30 Joshi et al., 1993b Joshi et al., 1993b

20 Aasbø et al., 1996 20 Dierking Joshi et al., 1992 et al., 1994 Stein et al., 1991 Lyons et al., 1995 10 10 VAS PI with morphine (mm) VAS

VAS PI with morphine (mm) VAS Joshi et al., Lyons et al., 1995 1992 Stein et al., 1991 Hege-Scheuing et al., 1995 Hege-Scheuing et al., 1995 0 0 01020304050 01020304050

VAS PI with control – saline or systemic morphine (mm)

FIGURE 44 L’Abbé plots for intra-articular morphine trials significantly lower total consumption of analgesics intensity scores, although in one study33 there were over 24 hours after morphine. no evaluations beyond 6 hours (Figure 44D). Lower total consumption of analgesics over 24 hours was Morphine versus systemic morphine control found in only one study.35 Three studies compared intra-articular with intravenous or intramuscular morphine.33–35 Combination of morphine plus bupivacaine versus saline In the early period, one study35 showed greater All four sensitive studies which compared intra- efficacy for intra-articular morphine compared articular morphine with both saline and bupi- with intravenous morphine, 1 mg (Figure 44C). vacaine also included a group with a combination of intra-articular morphine plus bupivacaine. All In the late period, no study indicated that intra- the studies which were sensitive to bupivacaine articular morphine had statistically lower pain alone and showed a positive effect for morphine 165 Pain relief from intra-articular morphine after knee surgery

also showed a significant effect for the combination early and late, and frequently below 25% of compared with placebo, both early and late.28–31 maximum (Figure 44). If there is no pain, reduction The two studies which were insensitive for bupi- in pain intensity cannot be measured. The reduced vacaine and morphine showed no efficacy for the sensitivity of analgesic studies with low pain combination.16,39 intensity has been evaluated.43

Dose response For this reason a hierarchy of evidence was chosen. Two studies addressed the question of a dose The highest rank was when active (bupivacaine) response with intra-articular morphine alone.35,36 In control was used as well as placebo, and analgesic one study,35 intra-articular morphine, 1 mg, could efficacy of intra-articular morphine was interpreted be differentiated from control but not 0.5 mg; the only when intra-articular bupivacaine was more researchers could not differentiate 0.5 mg from effective than placebo (i.e. established internal 1 mg of morphine. The other study36 demonstrated sensitivity). Intra-articular bupivacaine is known to a reversed dose response between 1 and 2 mg. provide reliable analgesia of predictable duration Neither study had evidence of internal sensitivity. following knee surgery,44,45 and it was therefore a valid active control. In two studies different doses of morphine in combination with a standard dose of bupivacaine Outcome measures were compared.37,38 No dose response was detected The special feature of these studies was that between either 1 and 3 mg or 2 and 5 mg necessarily the intervention was made before of morphine. the patient had pain, analogous to pre-emptive studies.46 The VAS pain intensity levels were low Adverse effects for several reasons. Diagnostic arthroscopies were No adverse effects that could have been attributed included in the primary studies, and opioids and to the intra-articular treatment were reported. NSAIDs were given both pre- and perioperatively. Diagnostic arthroscopies may not cause enough postoperative pain to be sufficiently sensitive for Comment an analgesic assay. Figure 44 indicates that studies which were sensitive generally had VAS pain These reports of the use of intra-articular intensity levels above 30% of the maximum possible morphine emphasise the importance of in the control group in the early period (and most considering potential bias and issues of validity in had high values in the late period also). We clinical studies before interpreting results. excluded studies which used spinal or epidural anaesthesia or infiltration of the knee joint with Bias high doses of local anaesthetic because these It is now well recognised that studies which are measures further reduce postoperative pain and, either not randomised or randomised without hence, sensitivity. concealment of treatment allocation, or which are not adequately blinded, result in an overestimation VAS pain intensity was usually measured at rest; of the effect of treatment.42 Method and conceal- sensitivity might have been increased by assessing ment of randomisation, double-blinding and it on movement. Arthroscopic surgery is usually withdrawals and dropouts were inadequately performed as day-case surgery. Sensitivity might described in all these studies. The method have been increased by following the patients of randomisation was explicit in three studies.30,36,38 in hospital for a longer period. Patients were In many studies, it was unclear who was blinded. instructed in the use of VAS PI before anaesthesia in only a minority of studies. Most patients were Design and validity sent home with a questionnaire within 2–6 hours of Classic analgesic trial design includes both active the end of surgery. Few studies mentioned whether and placebo controls. The reason is to ensure that the assessment was performed by a trained, or even if no difference is found between test analgesic and the same, observer. All these issues should be placebo, the correct interpretation of a negative addressed in study design. Sensitivity of the result can be made if the standard (active control) analgesic assay is crucial. analgesic gives a significant difference from placebo. This is particularly important when pain is Consumption of supplementary analgesics within of only mild to moderate intensity. The mean pain the first 24 hours after surgery was the second intensities after placebo were, with one exception, commonest outcome measure but usually was not 166 less than 50% of the maximum possible,30 both standardised. Other indicators of pain and pain Health Technology Assessment 1998; Vol. 2: No. 12

relief, such as time to first analgesic, time to weight early and late. These results again provide some bearing, time to discharge, were also used, but evidence for an analgesic effect of morphine in the in only a minority of studies. VAS pain intensity and knee joint, while raising the issue of whether this is the total consumption of supplementary analgesics a systemic as opposed to a local effect. were therefore used as primary outcome measures in the analysis. Dose–response studies No dose response was detectable in any study, over Early and late periods a dose range of 0.5–5 mg. The minimum dose Analysis by early and late periods was used for tested (0.5 mg) did not show analgesic efficacy35 several reasons. For the first 2–6 hours patients whereas a dose of 1 mg did.28,30,35 No greater effect were still in hospital where VAS pain intensity was found using morphine doses of 2 mg compared measurements were made by researchers or with 1 mg.36 In combination with local anaesthetic, (trained) nurses at predetermined intervals. morphine doses of 3 mg compared with 1 mg37 and Secondly, the effect of intra-articular bupivacaine, 5 mg compared with 2 mg38 showed no increased the index of internal trial sensitivity, should have efficacy. None of these studies had proven internal been most pronounced over this time. Thirdly, any sensitivity. Failure to demonstrate dose response systemic effect of morphine should have been more may therefore have been due to lack of sensitivity obvious during this period rather than later. The in the methods. late period was considered to be important as several studies suggested a prolonged effect of intra- However, the lowest effective dose of morphine articular morphine. Most studies provided inform- used, 1 mg, would, in a 20 ml injection, be ation on VAS pain intensity values at 24 hours and equivalent to a concentration of about 200 µmol consumption of supplementary analgesics was per litre (50 µg/ml). Typical blood or tissue levels reported as a total amount taken over 24 hours. after systemic injections of analgesic doses of morphine are found at concentrations of nanomols No biological reason for suspecting a late rather per litre, at least 1000 times lower.47 The very high than an early effect was apparent in the original concentrations of morphine in the knee joint study on intra-articular morphine.35 This indicated would be expected to saturate any opioid receptors that intra-articular morphine, 1 mg, provided present. If morphine is acting on local opioid significantly better analgesia after knee surgery receptors, then the minimal effective dose may well than the same dose given intravenously at 3, 4 or be much less than 1 mg. Failure to demonstrate a 6 hours. No difference was found between VAS dose response might then be because the doses pain intensity values at 24 hours, although the total tested were at the top end of the dose–response consumption of supplementary analgesics during curve. Late efficacy might be a consequence of the 24-hour period was significantly less after residual high morphine concentrations. intra-articular morphine. Is intra-articular morphine effective? Studies with both active and Taken together, these results render some support placebo controls for the hypothesis that intra-articular morphine Only six studies included groups receiving saline, provides pain relief after knee surgery.2 Using a bupivacaine and morphine. Four of the studies were simple ‘vote-counting’ approach on Figure 44, the judged sensitive as defined by significant analgesic points from the majority of the trials fall in the effect of bupivacaine compared with saline. All four lower right quadrant, indicating greater efficacy studies demonstrated significant analgesic effect of with morphine than control. Convincing evidence intra-articular morphine compared with placebo at for an early effect is lacking. There was more both early and late times (Figure 44). This provides consistent evidence for a prolonged analgesic some evidence for a prolonged biological effect of effect, mostly a single estimate of pain intensity at morphine in the knee joint. The two negative 24 hours or consumption of analgesic. These are studies16,39 failed the sensitivity test. weak measures.

Studies with no active control The problem is that this evidence rests on four Three studies of morphine against saline showed trials which fulfilled the sensitivity requirements an analgesic effect, two in the early period and all but which had only ten patients per treatment three in the late period. Comparisons of intra- group, and two others which were methodologically articular with intravenous or intramuscular weak but did distinguish morphine from saline. morphine were less compelling; only one of the Against these studies stands the failure to demons- three studies35 showed a significant effect, both trate a dose response for intra-articular morphine. 167 Pain relief from intra-articular morphine after knee surgery

Overall, the evidence is not compelling. The 14. Boden BP, Fassler S, Cooper S, Marchetto PA, Moyer lessons for future studies are obvious, but the RA. Analgesic effect of intra-articular morphine, current agenda is one of research rather than bupivacaine, and morphine/bupivacaine after arthroscopic knee surgery. Arthroscopy clinical utility. 1994;10:104–7. 15. Joshi GP, McCarroll SM, Brady OH, Hurson BJ, References Walsh G. Intra-articular morphine for pain relief after anterior cruciate ligament repair. Br J Anaesth 1. Kalso E, Tramèr M, Carroll D, McQuay H, Moore 1993;70:97–8. RA. Pain relief from intra-articular morphine after knee surgery: a qualitative systematic review. Pain 16. Björnsson A, Gupta A, Vegfors M, Lennmarken C, 1997;71:642–51. Sjöberg F. Intra-articular morphine for postoperative analgesia following knee arthroscopy. 2. Stein C. The control of pain in peripheral tissue by Reg Anesth 1994;19:104–8. opioids. N Engl J Med 1995;332:1685–90. 17. Chan ST. Intra-articular morphine and bupivacaine 3. Beecher HK. The powerful placebo. JAMA for pain relief after therapeutic arthroscopic knee 1955;159:1602–6. surgery. Singapore Med J 1995;36:35–7. 4. Houde R. On appraising pain and analgesic drugs. 18. De Andrés J, Bellver J, Barrera L, Febre E, Bolinches In: Keele C, Smith R, editors. The assessment of R. A comparative study of analgesia after knee pain in man and animals. Edinburgh: Livingstone, surgery with intra-articular bupivacaine, intra- 1962;202–12. articular morphine, and lumbar plexus block. Anesth Analg 1993;77:727–30. 5. Jadad AR, McQuay HJ. A high-yield strategy to identify randomized controlled trials for systematic 19. Ho ST, Wang JJ, Liaw WJ, Wong CS, Cherng CH. reviews. Online J Curr Clin Trials [serial online] 1993; Analgesic effect of intra-articular morphine after Doc No 33:3973 words; 39 paragraphs. 5 tables. arthroscopic knee surgery in Chinese patients. Acta Anaesthesiol Sin 1995;33:79–84. 6. Khoury GF, Chen ACN, Garland DE, Stein C. Inta- 20. Niemi L, Pitkänen M, Tuominen M, Björkenheim articular morphine, bupivacaine, and JM, Rosenberg PH. Intra-articular morphine for morphine/bupivacaine for pain control after knee pain relief after knee arthroscopy performed under videoarthroscopy. Anesthesiology 1992;77:263–6. regional anaesthesia. Acta Anaesthesiol Scand 7. VanNess SA, Gittins ME. Comparison of intra- 1994;38:402–5. articular morphine and bupivacaine following knee 21. Raja SN, Dickstein RE, Johnson CA. Comparison of arthroscopy. Orthop Rev 1994;23:743–7. postoperative analgesic effects of intra-articular 8. Ekblom A, Westman L, Söderlund A, Valentin A, bupivacaine and morphine following arthroscopic Eriksson E. Is intra-articular pethidine an alternative knee surgery. Anesthesiology 1992;77:1143–7. to local anaesthetics in arthroscopy? A double-blind 22. Jaureguito JW, Wilcox JF, Cohn SJ, Thisted RA, study comparing prilocaine with pethidine. Knee Reider B. A comparison of intra-articular morphine Surg Sports Traumatol Arthrosc 1993;1:189–94. and bupivacaine for pain control after outpatient 9. Jadad AR, Moore RA, Carroll D, Jenkinson C, knee arthroscopy. A prospective, randomized, Reynolds DJM, Gavaghan DJ, et al. Assessing the double-blinded study. Am J Sports Med quality of reports of randomized clinical trials: is 1995;23:350–3. blinding necessary? Control Clin Trials 1996;17:1–12. 23. Juelsgaard P, Dalsgaard J, Felsby S, Frokjaer J. Analgetisk effekt af to doser intraartikulaer morfin 10. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in efter ambulant knaeartroskopi. Et randomiseret, clinical research. Ann Intern Med 1987;107:224–33. prospektivt, dobbeltblindt studie. Ugeskr Laeger 11. Dalsgaard J, Felsby S, Juelsgaard P, Frokjaer J. 1993;155:4169–72. Analgesi af lavdosis intraartikulaer morfin efter 24. Reuben SS, Connelly NR. Postarthroscopic meniscus ambulant knaeartroskopi [Analgesic effect of low- repair analgesia with intra-articular ketorolac or dose intra-articular morphine after ambulatory knee morphine. Anesth Analg 1996;82:1036–9. arthroscopy]. Ugeskr Laeger 1993;155:4166–9. 25. Heard SO, Edwards WT, Ferrari D, Hanna D, Wong 12. Lehrberger K, Stein C, Hassan A, Yassouridis A. PD, Liland A, et al. Analgesic effect of intra-articular Opioids as novel intra-articular agents for analgesia bupivacaine or morphine after arthroscopic knee following arthroscopic knee surgery. Knee Surg Sports surgery: a randomized, prospective, double-blind Traumatol Arthrosc 1994;2:174–5. study. Anesth Analg 1992;74:822–6. 13. Klinken C. Effects of tourniquet time in knee 26. Ruwe PA, Klein I, Shields CL. The effect of intra- arthroscopy patients receiving intra-articular articular injection of morphine and bupivacaine on morphine combined with bupivacaine. CRNA postarthroscopic pain control. Am J Sports Med 168 1995;6:37–42. 1995;23:59–64. Health Technology Assessment 1998; Vol. 2: No. 12

27. McQuay H, Carroll D, Moore A. Variation in the 38. Laurent SC, Nolan JP, Pozo JL, Jones CJ. Addition of placebo effect in randomised controlled trials of morphine to intra-articular bupivacaine does not analgesics: all is as blind as it seems. Pain improve analgesia after day-case arthroscopy. Br J 1996;64:331–5. Anaesth 1994;72:170–3. 28. Haynes TK, Appadurai IR, Power I, Rosen M, Grant A. Intra-articular morphine and bupivacaine 39. Aasbø V, Raeder JC, Grøgaard B, Røise O. No analgesia after arthroscopic knee surgery. additional analgesic effect of intra-articular Anaesthesia 1994;49:54–6. morphine or bupivacaine compared with placebo after elective knee arthroscopy. Acta Anaesthesiol 29. Joshi GP, McCarroll SM, O’Brien TM, Lenane P. Scand 1996;40:585–8. Intra-articular analgesia following knee arthroscopy. Anesth Analg 1993;76:333–6. 40. Joshi GP, McCarroll SM, McSwiney M, O’Rourke P, 30. Karlsson J, Rydgren B, Eriksson B, Jarvholm U, Hurson BJ. Effects of intra-articular morphine on Lundin O, Sward L, et al. Postoperative analgesic analgesic requirements after anterior cruciate effects of intra-articular bupivacaine and morphine ligament repair. Reg Anesth 1993;18:254–7. after arthroscopic cruciate ligament surgery. Knee Surg Sports Traumatol Arthrosc 1995;3:55–9. 41. Lyons B, Lohan D, Flynn CG, Joshi GP, O’Brien TM, McCarroll M. Intra-articular analgesia for arthro- 31. McSwiney MM, Joshi GP, Kenny P, McCarroll SM. scopic meniscectomy. Br J Anaesth 1995;75:552–5. Analgesia following arthroscopic knee surgery. A controlled study of intra-articular morphine, 42. Schulz KF, Chalmers I, Hayes RJ, Altman DG. bupivacaine or both combined. Anaesth Intensive Empirical evidence of bias: dimensions of Care 1993;21:201–3. methodological quality associated with estimates 32. Joshi GP, McCarroll SM, Cooney CM, Blunnie WP, of treatment effects in controlled trials. JAMA O’Brien TM, Lawrence AJ. Intra-articular morphine 1995;273:408–12. for pain relief after knee arthroscopy. J Bone Joint Surg Br 1992;74:749–51. 43. Stubhaug A, Grimstad J, Breivik H. Lack of analgesic effect of 50 and 100 mg oral tramadol after ortho- 33. Dierking GW, Østergaard MD, Dissing CK, paedic surgery: a randomized, double-blind, Kristensen JE, Dahl JB. Analgesic effect of intra- placebo and standard active drug comparison. Pain articular morphine after arthroscopic 1995;62:111–18. meniscectomy. Anaesthesia 1994;49:627–9.

34. Hege-Scheuing G, Michaelsen K, Bühler A, 44. Chirwa SS, MacLeod BA, Day B. Intra-articular Kustermann J, Seeling W. Analgesie durch bupivacaine (Marcaine) after arthroscopic intraartikuläres Morphin nach Kniegelenks- meniscectomy: a randomized double blind arthroskopien? Anaesthesist 1995;44:351–8. controlled study [see comments]. Arthroscopy 35. Stein C, Comisel K, Haimerl E, Yassouridis A, 1989;5:33–5. Lehrberger K, Herz A, et al. Analgesic effect of intra- articular morphine after arthroscopic surgery. N 45. Kaeding CC, Hill JA, Katz J, Benson L. Bupivacaine Engl J Med 1991;325:1123–6. use after knee arthroscopy: pharmacokinetics and pain control study. Arthroscopy 1990;6:33–9. 36. Allen GC, St. Amand MA, Lui ACP, Johnson DH, Lindsay MP. Postarthroscopy analgesia with intra- 46. McQuay HJ. Pre-emptive analgesia: a systematic articular bupivacaine/morphine. Anesthesiology review of clinical studies. Ann Med 1995;27:249–56. 1993;79:475–80. 37. Heine MF, Tillet ED, Tsueda K, Loyd GE, Schroeder 47. Moore RA, Hand CW, McQuay HJ. Opioid JA, Vogel RL, et al. Intra-articular morphine after metabolism and excretion. In: Budd K, editor. arthroscopic knee operation. Br J Anaesth Update in Opioids. Clinical anaesthesiology. 1994;73:413–15. London: Baillère Tindall, 1987:829–58.

169

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 17 Analgesic efficacy of peripheral opioids

Summary This chapter of the review has been published in full by Picard and colleagues.1 Anaesthetists, using basic scientific concepts of peripheral opioid activity, try to improve regional anaesthesia and postoperative analgesia by inject- Introduction ing opioids, with or without local anaesthetic, close to nerve trunks or nerve endings. The aim of this For over 10 years anaesthetists have been trying to systematic review was to test the evidence that improve efficacy of regional anaesthesia and post- peripherally applied opioids (all except intraoperative analgesia by injecting opioids close to the articular) have an analgesic effect outside the nerve trunks or the nerve endings. The biological knee joint. basis for this approach is the presence of opioid receptors and their endogenous ligands in the A systematic search, 1966–96, was carried out for peripheral nervous system, and their effect on published reports of RCTs which compared the modulation of inflammatory pain.2 efficacy of peripheral opioids with placebo, local anaesthetic, or systemic opioids in acute pain. There are several distinct clinical approaches Reports of pethidine or intra-articular opioids to this topic. First, do opioids, when injected in were not included. Data on intraoperative efficacy combination with local anaesthetics, improve the (onset, quality, duration of sensory block), and quality and duration of a sensory block? This could postoperative efficacy (pain intensity, analgesic lead to improved surgical conditions. Second, does consumption) were extracted. Statistical signifi- this method allow the dose of the local anaesthetic cance as indicated in the original reports and to be reduced? This would minimise the risk of clinical relevance of differences between opioids systemic toxicity of local anaesthetics. Third, do and controls were taken into account to estimate opioids, when applied alone in peripheral sites, qualitatively overall efficacy. decrease postoperative pain intensity and analgesic requirements? This is a purer test of the biological Data for 952 patients in 26 trials were analysed. question of whether opioids have analgesic The opioids used were morphine (16 trials), effects peripherally. fentanyl (8), alfentanil (1), buprenorphine (1), and butorphanol (1). Two from four experimental The aim of this systematic review was to test pain trials reported a statistically significant the evidence that peripheral opioids (all except difference in favour of the opioid. In 22 clinical intra-articular) improve the quality of either trials, efficacy of opioid injections into the brachial intraoperative regional anaesthesia or plexus (10 trials), Bier’s block (4), perineural (3), postoperative analgesia and other sites (5) was tested.

In five of ten clinical trials measuring intraoperative Methods efficacy, statistically significant efficacy with opioids compared with control was reported; none were Full published reports of RCTs of peripheral judged clinically relevant. In five of 17 clinical trials opioids were sought systematically. A number measuring postoperative efficacy, a significant of different search strategies in MEDLINE difference in favour of the opioid was reported; (1966–September 1996), EMBASE (1981–96) none were judged clinical relevant. Trials of lower and the Oxford Pain Relief Database (1950–94) quality were more likely to report increased efficacy were used, without language restriction. Additional with opioids. Adverse events related to route of reports were identified from the reference lists of administration were not reported. retrieved reports and from review articles. Un- published reports and abstracts were not con- These trials provide no evidence for clinically sidered. Authors were not contacted for original relevant peripheral analgesic efficacy of opioids data. Reports were included if they were random- in acute pain. ised comparisons of peripheral opioids with 171 Analgesic efficacy of peripheral opioids

either local anaesthetics, placebo (saline), TABLE 50 Analgesic efficacy of peripheral opioids: excluded no treatment, or an opioid given by a different trials route, or comparisons of different doses of peripheral opioids. Reports of analgesic efficacy Excluded trial Reason for exclusion of intra-articular opioids are considered in Acalovschi & Cristea, 19958 Pethidine chapter 16.3 Armstrong, et al., 19935 Pethidine Inclusion criteria Davidas, et al., 19929 Pethidine Each report which could possibly be described as an RCT was read independently by each of the El Bakry, et al., 198910 Pethidine authors and scored using a three item, 1–5 score, 11 quality scale.4 The scale takes into account proper Oldroyd, et al., 1994 Pethidine randomisation, double-blinding and reporting of Gobeaux & Landais, 198812 Pethidine, not random withdrawals and drop-outs. Consensus was then achieved. Information on doses and routes of Arendt-Nielsen, et al., 199013 Not random administration of opioids and controls, types 14 of surgery and anaesthesia, number of patients Kepplinger, et al., 1995 Not random enrolled and analysed, study design, observation Moore, et al., 199415: Study 2 Number of patients periods, outcome measures, and adverse effects per group < 10 was taken from each report. Pere, 199316 Number of patients Validity criteria per group < 10 Validity criteria for included studies were number 17 ≥ Wajima, et al., 1995b Number of patients of patients per treatment group 10, any opioid per group < 10 except pethidine, which has shown local anaesthetic properties,5 any peripheral site of injection Welte, et al., 199218 Number of patients except intra-articular,3 and standardised methods per group < 10 of measuring sensory block and pain intensity. Tenant, et al., 199319 Number of patients per group < 10 Intraoperative efficacy was estimated by comparing onset and quality (loss of pinprick and Ben-Ameur, et al., 199320 No pain outcomes touch sensation), and duration of a sensory block with opioid compared with control. Arendt-Nielsen, et al., No opioid evaluated 199121: Study 2 Postoperative efficacy was estimated by comparing pain intensity, delay until first analgesic, and total Bullingham, et al., 198422 Not analysable analgesic consumption with opioid compared 23 with control. Pain intensity measurement was Mays, et al., 1987 Chronic pain analysed when reported as a VAS or verbal rating scale (VRS). Data from 26 RCTs, published in 25 reports, were Data showing any statistically significant difference analysed. In all, 952 patients, 485 of whom received (p < 0.05) between opioid and control, as indicated an opioid, were studied (Table 51). in the original report, were extracted. The authors then met to achieve consensus (vote-counting The average size of trial was 15 patients per group procedure) whether such a statistically significant (range 10–32). The median quality score was 2 difference was of clinical relevance. Finally, the (range 1–4). Three reports31,32,34 covered four trials decision (i.e. the vote counting) reached on (8%) which included a treatment arm with an clinical relevance was compared with the original analgesic method of proven efficacy and, therefore, authors’ conclusion of efficacy. had an index of internal sensitivity. Eight trials (16%) used a double-dummy design.15,32,34,36,40,42,44

Results Efficacy of peripheral opioids was tested in experimental pain trials in healthy volunteers and in a In all, 45 trials were considered for analysis, of wide variety of surgical settings with intravenous which 17 were subsequently excluded (Table 50). regional anaesthesia (Bier’s block), intrapleural, Two further reports were not considered because intraperitoneal, incisional and dental injections, 172 no copies were available in the UK.6,7 perineural blocks (femoral, ankle block, intercostal) Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 51 Analgesic efficacy of peripheral opioids: analysed RCTs

Study Treatments (number of patients) Setting Efficacy intraoperatively Efficacy postoperatively Adverse effects (‘anaesthesia’) (< less (‘analgesia’) (< less effective, p < 0.05; effective, p < 0.05; > more effective, > more effective, p < 0.05; = no difference) p < 0.05; = no difference)

Experimental

Arendt-Nielsen, 1. Morphine, 4 mg, 10 ml (10) Left and right ulnar nerve Pain and sensory thresholds N/A None et al., 199121 2. Saline, 10 ml (10) block and brain potentials: morphine Study 1 Laser stimulation > saline at 15 minutes only

Armstrong, et al., 1. Prilocaine 0.5%, 40 ml + fentanyl, Bier’s block Onset, speed of recovery and N/A Nausea: fentanyl 7; 19935 100 µg, 2 ml (15) Needle and temperature quality of sensory block: saline 1 2. Prilocaine 0.5%, 40 ml + saline, 2 ml (15) stimulation fentanyl = saline

Arthur, et al., 199224 1. Lignocaine 0.5%, 100 mg, 40 ml + saline, 2 ml (10) Bier’s block Sensory and motor block (quality N/A Nausea: lignocaine + 2. Lignocaine 0.5%, 100 mg, 40 ml + fentanyl, Needle, temperature and onset): fentanyl < lignocaine fentanyl 2; fentanyl 1 100 µg, 2 ml (10) Grip strength = lignocaine + fentanyl 3. Fentanyl, 100 µg, + saline, 40 ml (10)

Moniche, et al., 1. Morphine, 2 mg, 5 ml (12) Drugs injected s.c. in N/A Heat pain threshold: morphine Erythema at the site 199325 2. Saline, 5 ml (12) the injury > saline (30–330 minutes) of injection: Burn injury (49°C on Pressure pain threshold: morphine 5 calf bilaterally) morphine > saline at 30 minutes

Bier’s block

Abdulla & Fadhil, 1. Lignocaine, 100 mg, 40 ml (15) Upper limb surgery VRS: (4) > (1); (4) > (3); no N/A None 199226 2. Lignocaine, 100 mg, + fentanyl, 50 µg (15) significant result for (2). 3. Lignocaine, 100 mg, + pancuronium, 0.5 mg (15) Neuro muscular block: 4. Lignocaine, 100 mg, + fentanyl, 50 µg, + (3) and (4) pancuronium, 0.5 mg (15) > (1) or (2)

Ericyes, et al., 1. Prilocaine 1%, 30 ml, + saline, 10 ml (10) Upper limb surgery Onset and recovery of sensory N/A None 199527 2. Prilocaine 1%, 30 ml, + morphine, 6 mg, 10 ml (10) block: morphine > saline

Gupta, et al., 1. Prilocaine 0.5%, 3 mg/kg, + saline, 5 ml (20) Upper limb surgery N/A VAS PI, total analgesic Mild localised urti- 199328 2. Prilocaine 0.5%, 3 mg/kg, + morphine, 1 mg/5 ml (17) consumption: morphine = saline caria: morphine 1

Pitkanen, et al., 1. Prilocaine 0.5%, 40 ml, + saline, 4 ml, (12) Minor surgery of upper Loss of pinprick at 15 minutes: N/A Nausea and 199229 2. Prilocaine 0.5%, 40 ml, + fentanyl, 100 µg, 4 ml (13) extremity fentanyl, 200 mg, > fentanyl, dizziness: saline 1; 3. Prilocaine 0.5%, 40 ml, + fentanyl, 200 µg, 4 ml (12) 100 mg, = saline fentanyl, 100 mg, 7; Time to develop analgesia: fentanyl, fentanyl, 200 mg, 6 200 mg, = fentanyl, 100 mg, = saline

Other peripheral sites of injection: all drugs injected postoperatively

Aykac, et al., 199530 1. Morphine, 20 mg, intrapleural, 20 ml (14) Thoracotomy (lobectomy N/A VRS: intrapleural > i.v. Confusion: 2. Morphine, 20 mg, i.v. (ml not applicable) (14) for lung cancer). Morphine Blood levels: i.v. > intrapleural intrapleural 4; i.v. 4 injected after pleural Urinary retention: closure. intrapleural 3; i.v. 6 Respiratory depression: i.v. 1

Moore, et al., 1. Morphine, 30 ng local, 0.3 ml + oral placebo (10) Bilateral third molar surgery; N/A VAS PI, analgesic consumption: None 199415 2. Placebo local, 0.3 ml + morphine, locally applied morphine. morphine = saline Study 1 50 ng p.o. (10)

Rosenstock, et al., 1. Morphine, 5 mg, 6 ml, incisionally (10) Inguinal herniotomy; N/A VAS PI (rest, movement), Not reported 199631 2. Morphine, 5 mg, 1 ml i.v. (10) incisional morphine analgesic consumption: 3. Morphine, 5 mg, 6 ml s.c. (10) postoperatively. morphine incision = i.v. = 4. Saline, 6 ml, incisionally (10) s.c. = saline

Schulte-Steinberg, 1. Morphine, 1 mg, i.p. + saline i.v. (18) Laparoscopic N/A VAS PI, VRS, McGill, analgesic Not reported et al., 199532 2. Saline i.p. + morphine, 1 mg, i.v. (17) cholecystectomy; i.p. consumption: morphine i.p. Study 1 3. Bupivacaine 0.25% i.p. + saline i.v. (15) injection at the end = morphine i.v. = bupivacaine i.p. NB:All drugs diluted in 20 ml. of surgery.

Schulte-Steinberg, 1. Morphine, 1.5 mg, intrapleural + saline i.v. (20) Laparoscopic N/A VAS PI, VRS, McGill, analgesic Not reported et al., 199532 2. Saline intrapleural + morphine, 1.5 mg, i.v. (20) cholecystectomy; consumption: morphine Study 2 3. Bupivacaine 0.25%, intrapleural + saline i.v. (20) intrapleural injection at the intrapleural = morphine i.v. NB: All drugs diluted in 30 ml. end of surgery. < bupivacaine intrapleural

continued 173 Analgesic efficacy of peripheral opioids

TABLE 51 contd Analgesic efficacy of peripheral opioids: analysed RCTs

Perineural: drugs injected pre- or postoperatively

Bullingham, et al., 1. Morphine 0.02% one side + saline other side (10) Ankle nerve block (nerves); N/A VAS PI and PR: morphine Not reported 198333 2. Morphine 0.04% one side + morphine 0.02% bilateral foot minor surgery 0.02% = morphine 0.01% = saline other side (10) NB:Volume 15–20 ml per injection

Dahl, et al., 198834 1. Morphine, 4 mg, epidural, 10 ml + saline femoral, Femoral block and epidural N/A VAS PI: epidural > femoral Nausea, vomiting: 10 ml (10) catheter after knee surgery. Morphine consumption: epidural > femoral 2. Saline epidural, 10 ml + morphine, 4 mg, femoral, Treatment reversed for the epidural = femoral 10 ml (10) next 24 hours.

Sternlo & 1. Bupivacaine 0.5%, 20 ml (24) Intercostal block (4 ml per N/A VAS PI, delay for analgesic, None Hagerdal35 2. Bupivacaine 0.5%, 20 ml, + morphine, 4 mg (26) 5 ribs); biliary surgery analgesic consumption: NB: ml N/A bupivacaine + morphine = bupivacaine

Brachial plexus

Bourke & 1. Lignocaine 1.5%, 0.55 ml/kg, + morphine, Hand and forearm surgery; N/A VAS, recovery of sensory and Mild nausea: axillary Furman36 0.1 mg/kg (ml N/A) + saline i.v., 0.1 ml/kg (20) axillary block motor block: i.v. = axillary block block 1; i.v. 2 2. Lignocaine 1.5%, 0.55 ml/kg, + morphine, Analgesic consumption: axillary 0.1 mg/kg i.v. (ml N/A) + saline, 0.1 ml/kg (20) block > i.v.

Fletcher, et al., 1. Lignocaine 1.5%, 38 ml, + fentanyl, 100 µg, 2 ml (26) Orthopaedic surgery; Onset and duration of block: N/A None 199437 2. Lignocaine 1.5%, 38 ml, + saline, 2 ml (25) axillary block fentanyl = saline Success rate of each nerve block: fentanyl = saline

Flory, et al., 199538 1. Bupivacaine 0.5%, 40 ml, + morphine, 5 mg, Shoulder surgery; N/A VAS PI, delay until first and total Nausea, vomiting: 5 ml (20) interscalene block dose of analgesic: morphine saline 5; morphine 2. Bupivacaine 0.5%, 40 ml, +saline, 5 ml (20) = saline 10 Pruritus: saline 3; morphine 0 Urine retention: saline 1; morphine 1

Gobeaux & 1. Lignocaine 1.5%, 30 ml (12) Upper limb surgery; Onset and intensity of block: N/A None Landais12 2. Lignocaine 1.5%, 30 ml, + fentanyl, 100 µg (12) axillary block fentanyl > no treatment between NB: ml N/A 5 and 10 minutes

Gormley, et al., 1. Lignocaine 1.5%, 7 mg/kg, + alfentail, 10 µg /kg, Upper limb surgery; Duration of sensory and motor VAS: alfentanil > saline (h 3) None 199639 10 ml (28) axillary block block: alfentanil > saline Delay for first analgesic, block 2. Lignocaine 1.5%, 7 mg/kg, + saline 10 ml (32) (10–40 minutes) recovery: alfentanil = saline

Kardash, et al., 1. Mepivacaine 1.5%, 30 ml, + fentanyl, 75 µg, 1.5 ml, Upper limb surgery; Onset, duration of sensory and VAS PI axillary block > i.m. None 199540 + saline i.m., 1.5.ml (10) supraclavicular block motor block: axillary block = i.m. (0–1 hours) 2. Mepivacaine 1.5%, 30 ml,+ fentanyl, 75 µg i.m., 1.5 ml, + saline, 1.5.ml (10)

Morros Vinoles, 1. Mepivacaine 1%, 40 ml Upper limb surgery; Onset and quality of block: Delay until first analgesia: None et al., 199141 2. Mepivacaine 1%, 40 ml, + fentanyl, 100 µg axillary block (1) = (2) = (3) (1) = (2) = (3) (ml N/A) 3. Mepivacaine 1%, 40 ml, + fentanyl, 100 µg s.c.

Racz, et al., 199142 1. Bupivacaine 0.5% + lignocaine 1%, 40 ml, + Arm and forearm minor Onset and quality of sensory and VAS PI, delay until first analgesic: None morphine, 5 mg, 1 ml (19) surgery; axillary block motor block: axillary block = i.m. axillary block = i.m. 2. Bupivacaine 0.5% + lignocaine 1%, 40 ml, + morphine, 5 mg i.m., 1 ml (21)

Viel, et al., 198943 1. Bupivacaine 0.5%, 40 ml, + morphine, 50 µg/kg (20) Upper limb surgery; Sensory and motor block: Quality, duration of analgesia: Pruritus: morphine 1 2. Bupivacaine 0.5%, 40 ml, + buprenorphine, supraclavicular block buprenorphine = morphine buprenorphine > morphine Nausea, vomiting: 3 µg/kg (20) buprenorphine 1

Wajima, et al., 1. Butorphanol, 83 µg/h + saline i.v. (12) Upper extremity surgery; N/A VAS PI at 9, 12, 18, 20 hours: Nausea: i.v. 6; 199544 2. Butorphanol, 83 µg/h i.v. + saline (10) axillary block axillary block > i.v. axillary 4 NB: all perfusion 50 ml/72 hours) (postoperatively continuous Supplementary analgesia: i.v. = Vomiting: i.v. 2; infusion) axillary block axillary 2 Drowsiness: i.v. 1; 174 axillary 2 Health Technology Assessment 1998; Vol. 2: No. 12

and brachial plexus sheath injections (axillary, Other peripheral sites (5 trials) supraclavicular and interscalene approaches). The All five trials used morphine. Four of them could opioids used were morphine (16 trials), fentanyl (8), not demonstrate any difference in the post- alfentanil (1), buprenorphine (1) and butorphanol operative period between morphine and control (1). Intraoperative efficacy assessments were when applied into a tooth socket,15 into a surgical performed in ten clinical trials. Postoperative wound,31 or by intraperitoneal or intrapleural block efficacy was evaluated in 17 clinical trials. (study I and II).32 The fifth trial reported a statistically significant improvement with morphine, Experimental pain trials (4 trials) 20 mg, given intrapleurally compared with the In one trial, morphine was applied perineurally;21 same drug and dose given intravenously;30 verbal sensory and pain thresholds were significantly in- pain rating scores were lower for 20 hours in the creased compared with saline but for not longer than intrapleural group. Morphine plasma levels were 15 minutes. One trial used morphine subcutaneously lower in the intrapleural group. This was con- at the site of injury and reported higher heat and sidered to be of clinical relevance by these authors. pain thresholds compared with saline.25 Two other Analgesic consumption was not reported. We trials failed to demonstrate any benefit from adding considered the outcome to be of little clinical fentanyl to a local anaesthetic in a Bier’s block.24,45 relevance because of the unconventional (high) dose of morphine used. The experimental nature of these reports makes it difficult to judge clinical relevance. Therefore they Perineural (3 trials) were not taken into account in estimating overall None of these trials reported any significant efficacy of peripheral opioids. difference between the opioid and control.33–35

Bier’s block (4 trials) Brachial plexus (ten trials) Fentanyl was used in two trials.26,29 Abdulla and Opioids were given by interscalene (one trial), Fadhil could not demonstrate any significant supraclavicular (2) or axillary (7) approaches to difference between the combination of fentanyl the brachial plexus sheath. plus local anaesthetic and local anaesthetic alone,26 but nevertheless concluded that the method was of In three trials, morphine was combined with a local clinical relevance. We disagreed with these authors anaesthetic and applied either by axillary36,42 or because they did not comment on the comparison interscalene route.38 Comparators were systemic of interest to us (i.e. opioid versus no treatment) morphine or axillary saline. No intraoperative or but rather based their conclusion on the com- postoperative improvement could be demonstrated parison between opioid plus curare versus no with peripheral morphine in two of the three treatment or curare alone. trials.38,42 The third trial (axillary route) reported similar pain scores in the groups but a significantly Pitkånen and colleagues reported a significantly lower postoperative analgesic consumption improved quality of the sensory block after (number of tablets of oxycodone, 5 mg, plus 15 minutes with fentanyl, 200 µg, compared with acetaminophen, 500 mg) with the opioid; the either saline or fentanyl, 100 µg.29 No measure- authors concluded that this difference was clinic- ments were taken after 15 minutes. Nausea and ally important.36 The median number of tablets dizziness were more frequent with fentanyl. These was two with axillary morphine and four with authors concluded that their finding was not systemic morphine. We did not consider this clinically relevant. difference to be of clinical importance in this acute setting. Morphine was used in two trials.27,28 In one there was no significant difference between morphine and In four trials, fentanyl was combined with a local saline, and the authors concluded that morphine anaesthetic and compared with a local anaesthetic was of no value in Bier’s block.28 In the other trial, alone or with another route of injection.37,40,41,46 In both onset of and recovery from anaesthesia and two a significant improvement with fentanyl was analgesia were significantly better with morphine reported.40,46 Gobeaux and colleagues concluded compared with local anaesthetic alone.27 These that a faster speed of onset of the sensory block authors concluded that the differences of 1 minute with the opioid was clinically relevant.46 However, and 2 minutes, respectively, were clinically relevant. this difference was only 5 minutes. Kardash and We disagreed, because the difference between the colleagues reported a lower VAS for pain intensity two groups was of very short duration only and for the first postoperative hour with fentanyl but therefore of no practical importance. did not consider this to be clinically important.40 175 Analgesic efficacy of peripheral opioids

Alfentanil added to a local anaesthetic led to a efficacy. Quality scores for these trials were 2 or significant improvement compared with the local below.21,25–27,30,36,39,43,44,46 In the 16 remaining trials anaesthetic plus placebo;39 duration of sensory and (two experimental and 14 clinical) conclusions motor block after surgery was 40 minutes longer were negative. Seven, including one with two with the opioid. This was considered to be clinically studies,32 had a quality score of 3 or 4.28,37,38,42,45 relevant by these authors, although there was no difference between the two groups in the delay Adverse effects until the first analgesic rescue medication. No adverse effects attributable to the route of administration were reported. Butorphanol perfusion into the plexus sheath led to significantly lower VAS scores for pain intensity up to hour 24 postoperatively compared with the Comment same butorphanol perfusion given intravenously.44 There was no difference in postoperative analgesic The aim of this systematic review was to test the requirements. The authors concluded that this evidence for an analgesic action of peripheral difference was clinically relevant. However, average opioids and the clinical relevance of such action. In VAS scores were very low, irrespective of the route all, 26 RCTs with data from more than 950 patients of administration (i.e. axillary route 6–7% of the were analysed. These trials described a variety of maximum on a VAS for pain intensity; intravenous surgical procedures and experimental designs. Five route 17–33%). different opioids with several different doses were administered with ten different regional anaes- Buprenorphine, 3 µg/kg, was compared with thetic techniques. Trials were not consistent in morphine, 50 µg/kg, in one trial; both opioids either analysing or reporting quality of surgical were added to the same local anaesthetic before blocks, postoperative analgesia, or observation supraclavicular injection.43 A placebo group was periods. Estimation of efficacy based on data of lacking in this trial. Duration and quality of such methodological and clinical heterogeneity postoperative analgesia were significantly better was, therefore, difficult. Quantitative analysis was with buprenorphine; ‘good’ pain relief, as judged impossible. Unfortunately the different procedures by the patients, lasted for 35 hours with bupre- or blocks operate, or may operate, in different norphine and 18 hours with morphine. Authors ways, so that a negative result from one procedure concluded that buprenorphine is efficacious and does not preclude a positive result with another. long-acting as an analgesic when injected into the brachial plexus sheath. However, they did not take It was decided to judge the studies in two ways: into account equi-analgesic dosing. those which had any result which was statistically in favour of a peripheral action of opioid, and those There was a relationship between quality scores where the result was of a sufficient magnitude or of the reports and original authors’ conclusions importance to be clinically relevant. Because most on efficacy of peripheral opioids (Figure 45). of the studies had a number of different measure- Authors of ten trials (two experimental and ments at different times, the possibility that some eight clinical) reported positive estimates of statistical differences could occur by chance was high. Use of the conventional level of statistical significance in clinical and scientific studies of 5 YES NO the 95th percentile implies that, if 20 different measurements are made, one will show significance 4 just by chance. So, in 26 different studies with a 3 large number of outcome measurements, some statistically significant differences with opioid 2 would be expected. Judging clinical relevance

Quality score may be easier or more difficult. Most practising 1 clinicians would claim they could tell when a 0 result was going to benefit their patients. Codifying Efficacy as stated in original report what that entails is not easy. Reaching a consensus on clinical efficacy of peripheral opioids was influenced by whether all the measures in a study FIGURE 45 Efficacy of peripheral opioids (relationship between showed statistically significant differences, or quality of trials and overall examination of efficacy as stated in whether the magnitude of any difference shown original reports). ●, ●, clinical trials; ◆, ◆, experimental pain trials. 176 in a trial was sufficiently large to make change in Health Technology Assessment 1998; Vol. 2: No. 12

practice a reasonable consideration. The As well as the issue of validity, there is the issue of authors are conscious that others might reach methodological quality. There are other examples different conclusions. where trials with low scores (two or below on a scale of 1–5) on the validated quality scale4 used in this Of 26 trials, 14 were unequivocally negative. The review, have overestimated the effectiveness of remaining 12 trials21,25–27,29,30,36,39,40,43,44,46 reported at treatment.47 In the present review, none of the ten least one statistically significant result in favour of trials which claimed efficacy of peripheral opioids the peripheral opioid. Of these 12 positive trials, had a score above two (Figure 45). Seven of the two were in experimental pain;21,25 their results may 14 unequivocally negative reports had scores of not be directly applicable to clinical practice. Of three or four. This means that the trials of highest the remaining ten positive trials, authors of two29,40 methodological quality in this data set could not did not regard their findings as being clinically show any difference between peripheral opioid relevant. This means that results from only eight and control. out of the 26 trials (31%) were judged by their original authors as clinically relevant. This subgroup analysis by trial quality emphasises that, in these clinical models, peripheral opioids We could not, for different reasons, support the have no efficacy. The question is then why good conclusions of any of these positive reports. An quality trials showed some efficacy of morphine in isolated significant outcome in favour of the the knee joint but no efficacy of different opioids in opioid, such as a longer duration of a sensory block peripheral sites outside the knee joint. This may be which was not correlated with a delay until the because the knee joint model better reflects the patient needed a first analgesic, was not judged inflammatory process which is thought to be of clinically relevant by us.39 Differences of doubtful importance in sensitising peripheral opioid clinical importance were reported, such as the receptors.48 It may also be related to inadequately shortening of the onset of a surgical block by a few low doses of opioids used in these trials. Doses of minutes27,46 or a minimal difference in the average morphine between 0.5 mg and 5 mg tested in the analgesic consumption.36 A significant difference confined space of the knee joint produced very high between two opioids was shown but without a local concentrations.3 Similar doses of morphine placebo control,43 so that clinical relevance of injected into the peritoneal or pleural cavity, or into this greater efficacy of buprenorphine relative to an isolated limb would produce much lower local morphine remains questionable. Furthermore, in concentrations than in the knee joint. this trial the two opioids were compared in non- equi-analgesic doses, and a systemic analgesic effect This qualitative analysis of pain trials highlights the of buprenorphine with its long duration of action importance of critical appraisal of the literature cannot be ruled out. Other drawbacks in studies and some of the difficulty encountered in doing it. with positive findings were the very low pain Authors of original reports tended to overinterpret intensity scores irrespective of the treatment,44 their findings and to confuse statistical significance the unconventional dose of opioids used,30 or with clinical relevance. Inattentive or uncritical the comparison of treatment arms which were readers may be misled into a false perception of of no interest to this review.26 Such trials cannot treatment efficacy. Some 30 years ago, Schwartz be regarded as valid assays for evaluation of and Lellouch distinguished between explanatory analgesic efficacy. studies, designed to prove a hypothesis, and pragmatic studies, designed to tell us whether instiga- Do these trials represent evidence of a lack of ting a change was of benefit.49 The distinction is efficacy of peripheral opioids, or rather a lack still important and the clinical use of peripheral of evidence of their efficacy? In the systematic opioids requires much more evidence than exists review of the relevant published literature on the at present. analgesic efficacy of intra-articular morphine, only a minority of the analysed data could be regarded as valid; yet this limited amount of data provided References some evidence for its analgesic efficacy.3 Validity in those trials was assumed when baseline pain was 1. Picard PR, Tramèr MR, McQuay HJ, Moore RA. Analgesic efficacy of peripheral opioids (all except sufficiently high to allow measurement of pain intra-articular): a qualitative systematic review of relief, when an index of internal sensitivity was randomised controlled trials. Pain 1997;72: 309–18. given, and when blinding was adequate.3 Most of the trials in the present review did not meet 2. Stein C. The control of pain in peripheral tissue by these criteria. opioids. N Engl J Med 1995;332:1685–90. 177 Analgesic efficacy of peripheral opioids

3. Kalso E, Tramèr M, Carroll D, McQuay H, Moore 17. Wajima Z, Shitara T, Nakajima Y, Kim C, Kobayashi RA. Pain relief from intra-articular morphine after N, Kadotani H, et al. Comparison of continuous knee surgery: a qualitative systematic review. Pain brachial plexus infusion of butorphanol, mepiva- 1997;71:642–51. caine and mepivacaine-butorphanol mixtures for postoperative analgesia. Br J Anaesth 1995;75:548–51. 4. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJM, Gavaghan DJ, et al. Assessing the 18. Welte M, Haimerl E, Groh J, Briegel J, Sunder- quality of reports of randomized clinical trials: is Plassmann L, Herz A, et al. Effect of intrapleural blinding necessary? Control Clin Trials 1996;17:1–12. morphine on postoperative pain and pulmonary function after thoracotomy. Br J Anaesth 5. Armstrong PJ, Morton CPJ, Nimmo AF. Pethidine 1992;69:637–9. has a local anaesthetic action on peripheral nerves in vivo. Anaesthesia 1993;48:382–6. 19. Tenant F, Moll D, de Paulo V. Topical morphine for peripheral pain. Lancet 1993;342:1047–8. 6. Mocavero G. Analgesia selettiva con morfina perinervosa. Incontri Anest Rianim Sci Affini 20. Ben-Ameur M, Ecoffey C, Kuhlman G, Mazoit X, 1981;16:1–3. Gobeaux D. Réponse ventilatoire au dioxyde de carbone après bloc axillaire avec fentanyl et 7. Vinoles CM, Cuenca P, Monsegur JC, Valloba FC. lidocaïne. Ann Fr Anesth 1993;12:22–6. Adicion de fentanilo a la mepivacaina en el bloqueo axilar del plexo braquial. Rev Esp Anestiol Reanim 21. Arendt-Nielsen L, Bjerring P, Berg Dahl J. A 1991;38:87–9. quantitative double-blind evaluation of the antinociceptive effects of perineurally administered 8. Acalovschi I, Cristea T. Intravenous regional morphine compared with lidocaine. Acta Anesth anesthesia with meperidine. Anesth Analg Scand 1991;35:24–9. 1995;81:539–43. 22. Bullingham RE, McQuay HJ, Moore RA. Studies 9. Davidas JL, Blond JL, Rochette A, Manchon M, on the peripheral action of opioids in postoperative Degoute CS, Bansillon V. Action analgésique locale pain in man. Acta Anaesthesiol Belg 1984;35 suppl: par effet directe sur les troncs nerveux de la 285–90. péthidine. Thérapie 1992;47:485–7. 23. Mays KT, Lipman JJ, Schnapp M. Local analgesia 10. El Bakry MS, El-Shafei SB, Seyam EM, El-Kobbia without anesthesia using peripheral, perineural NM, Ebrahim UH. Use of pethidine as an morphine injections. Anesth Analg 1987;66:417–20. intravenous regional anesthetic. Middle East J Anesth 1989;10:189–94. 24. Arthur JM, Heavner JE, Tanmian MB, Rosenberg PH. Fentanyl and lidocaine versus lidocaine for Bier 11. Oldroyd GJ, Tham EJ, Power I. An investigation of block. Reg Anesth 1992;17:223–7. the local anaesthetic effects of pethidine in volunteers. Anaesthesia 1994;49:503–6. 25. Moniche S, Dahl JB, Kehlet H. Peripheral antinociceptive effects of morphine after burn 12. Gobeaux D, Landais A. Utilisation de deux injury. Acta Anaesth Scand 1993;37:710–12. morphiniques dans les blocs du plexus brachial. Cah Anesthesiol 1988;36:437–40. 26. Abdulla WY, Fadhil NM. A new approach to intravenous regional anesthesia. Anesth Analg 13. Arendt-Nielsen L, Oberg B, Bjerring P. Laser- 1992;75:597–601. induced pain for quantitative comparison of intravenous regional anesthesia using saline, 27. Erciyes N, Akturk G, Solak M, Dohman D. morphine, lidocaine, or prilocaine. Reg Anesth Morphine/prilocaine combination for intravenous 1990;15:186–93. regional anesthesia. Acta Anaesth Scand 1995;39:845–6. 14. Kepplinger B, Schmid H, Rettensteiner G, Derfler C, Papst H, Erhart H, et al. Wirkungsvergleich 28. Gupta A, Bjornsson A, Sjoberg F, Bengtsson M. Lack zwischen intramuskulär und periradikuär of peripheral analgesic effect of low dose morphine verabreichtem Tramadol. Neuropsychiatrie during intravenous regional anesthesia. Reg Anesth 1995;9:196–200. 1993;18:250–3.

15. Moore UJ, Seymour A, Gilroy J, Rawlins MD. 29. Pitkånen MT, Rosenberg PH, Pere PJ, Tuominen The efficacy of locally applied morphine in post- MK, Seppala TA. Fentanyl–prilocaine mixture for operative pain after bilateral third molar surgery. intravenous regional anaesthesia in patients Br J Clin Pharmacol 1994;37:227–30. undergoing surgery. Anaesthesia 1992;47:395–8.

16. Pere P. The effect of continuous interscalene 30. Aykac B, Erolcay H, Dikmen Y, Oz H, Yillar O. brachial plexus block with 0.125% bupivacaine plus Comparison of intrapleural versus intravenous fentanyl on diaphragmatic motility and ventilatory morphine for post thoracotomy pain management. 178 function. Reg Anesth 1993;18:93–7. Cardiothor Vasc Anesth 1995;9:538–40. Health Technology Assessment 1998; Vol. 2: No. 12

31. Rosenstock C, Andersen G, Antonsen K, Rasmussen 41. Morros Vinoles C, Perez Cuenca MD, Castillo H, Lund C. Analgesic effect of incisional morphine Monsegur J, Cedo Valloba F. Adicion de fentanilo a following inguinal herniotomy under spinal la mepivacaina en el bloqueo axilar del plexo anesthesia. Reg Anesth 1996;21:93–8. braquial. Rev Esp Anestesiol Reanim 1991;38:87–9.

32. Schulte-Steinberg H, Weninger E, Jokisch D, 42. Racz H, Gunning K, Della Santa D, Forster A. Hofstetter B, Misera A, Lange V, et al. Intra- Evaluation of the effect of perineuronal morphine peritoneal versus interpleural morphine or on the quality of postoperative analgesia after bupivacaine for pain after laparoscopic cholecys- axillary plexus block: a randomised double-blind tectomy. Anesthesiology 1995;82:634–40. study. Anesth Analg 1991;72:769–72. 33. Bullingham R, O’Sullivan G, McQuay H, Poppleton 43. Viel EJ, Eledjam JJ, de La Coussaye JE, d’Athis F. P, Rolfe M, Evans P, et al. Perineural injection of Brachial plexus block with opioids for postoperative morphine fails to relieve postoperative pain in pain relief: comparison between buprenorphine humans. Anesth Analg 1983;62:164–7. and morphine. Reg Anesth 1989;14:274–8. 34. Dahl JB, Daugaard JJ, Kristoffersen E, Johannsen HV, Dahl JA. Perineuronal morphine: a comparison 44. Wajima Z, Nakajima Y, Kim C, Kobayashi N, with epidural morphine. Anaesthesia 1988;43:463–5. Kadotami H, Adachi H, et al. I.v. compared with brachial plexus infusion of butorphanol 35. Sternlo IE, Hagerdal M. Perineuronal morphine in for postoperative analgesia. Br J Anaesth intercostal block. Anaesthesia 1992;47:613–15. 1995;74:392–5.

36. Bourke DL, Furman WR. Improved postoperative 45. Armstrong PJ, Power I, Wildsmith JAW. Addition analgesia with morphine added to axillary block of fentanyl to prilocaine for intravenous regional solution. J Clin Anesth 1993;5:114–17. anesthesia. Anaesthesia 1991;46:278–80. 37. Fletcher D, Kuhlman G, Samii K. Addition of 46. Gobeaux D, Landais A, Bexon G, Cazaban J, fentanyl to 1.5% lidocaine does not increase the Levron JC. Adjonction de fentanyl à la lidocaéne success of axillary plexus block. Reg Anesth adrénalinée pour le blocage du plexus brachial. 1994;19:183–8. Cah Anesthesiol 1987;35:195–9. 38. Flory N, Van Gessel E, Donald F, Hoffmeyer P, Gamulin Z. Does the addition of morphine to 47. Khan KS, Daya S, Jadad AR. The importance of brachial plexus block improve analgesia after quality of primary studies in producing unbiased shoulder surgery? Br J Anaesth 1995;75:23–6. systematic reviews. Arch Intern Med 1996;156:661–6. 39. Gormley WP, Murray JM, Fee JPH, Bower S. Effect 48. Nagasaka H, Awad H, Yaksh T. Peripheral and of the addition of alfentanil to lignocaine during spinal actions of opioids in the blockade of the axillary brachial plexus anaesthesia. Br J Anaesth autonomic response evoked by compression 1996;76:802–5. of the inflamed knee joint. Anesthesiology 1996;85:808–16. 40. Kardash K, Schools A, Concepcion M. Effects of brachial plexus fentanyl on supraclavicular block: 49. Schwartz D, Lellouch J. Explanatory and pragmatic a randomized, double-blind study. Reg Anesth attitudes in therapeutic trials. J Chronic Dis 1995;20:311–15. 1967;20:637–48.

179

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 18 Pre-emptive analgesia: a systematic review of clinical studies to 1994

Summary stimuli and the methods which could pre-empt these changes, to clinical management of Basic science evidence suggests that an analgesic postoperative pain. Since that editorial was intervention made before surgery will produce a published, the issues have become much more better outcome than the same intervention made focused. after surgery. The evidence from RCTs which tested this hypothesis in patients is reviewed The central question is whether an intervention in this chapter. made before pain starts has greater analgesic effect than the same intervention (same dose, same Four studies with paracetamol or NSAIDs did not route) made after the pain. The aim is to define show any pre-emptive effect. Of seven studies with the questions that need to be asked and, by local anaesthetic, six did not show a pre-emptive reviewing the clinical evidence systematically, effect. In four studies with opioids, there was weak to see whether or not definitive answers exist. evidence of a pre-emptive effect in three. This is a very active area of clinical research, so that any conclusions may be overtaken by There are few perfect RCTs and, unfortunately, new evidence. this applies in the pre-emptive analgesia field. Many of the studies which did not show a pre- The concept is a simple one. The effect of the pre- emptive effect lacked power. The opioid studies emptive analgesia is to prevent or reduce the which did show a pre-emptive effect had other development of any ‘memory’ of the pain stimulus technical weaknesses. in the nervous system. Preventing or reducing the pain memory should lower any subsequent anal- One way to combat lack of power would be to gesic needs.3 The scientific interest in this pheno- combine data (meta-analysis). This is very difficult menon is in the underlying mechanism. The in this field because of the outcome measures clinical interest is in the potential for improving which investigators are using. postoperative pain management.

The concept, and the explanation, are very Introduction attractive. Management of postoperative pain has rightly been criticised many times over the Pre-emptive analgesia is analgesia given before last 30 years. Despite the advent of increasingly the painful stimulus begins. The reason for giving high-technology approaches, it is doubtful that analgesia before the painful stimulus is to prevent most patients are any better served. If pre- or reduce subsequent pain. The concept that pre- emptive analgesia worked then these patients’ emptive analgesia might provide better pain pain might be reduced. Unfortunately there control came from basic science studies. Initial are difficulties with the details, and difficulties observations were that noxious stimuli induced in interpreting conflicting evidence from changes in neural function,1 such as hyperexcit- clinical studies. ability, in the spinal cord. Later studies suggested that analgesia given before the nociceptive stimulus We have been slow in distinguishing that pre- began was more effective than the same dose given emptive treatment with one kind of analgesic after the stimulus. intervention, for instance opioids, may not give the same answer as pre-emptive treatment with another, The editorial by Wall2 focused clinicians’ attention such as NSAIDs. We have also been slow to disting- on pre-emptive analgesia and linked fundamental uish between two very different outcomes, the work to clinical studies. He related the findings in outcome of a pre-emptive treatment on nociceptive fundamental studies, the ways in which the central pain and the outcome of a pre-emptive treatment nervous system changed following nociceptive on neuropathic pain. 181 Pre-emptive analgesia: a systematic review of clinical studies to 1994

Problems with the fundamental evidence autotomy after peripheral nerve section, longer- Timing is one critical problem. If pre-emptive term ‘pre-emptive’ effects have been reported with treatment reduces the memory of the subsequent the use of local anaesthetic. The speed with which noxious stimulus, how long does this effect last? autotomy developed in response to nerve section, Evidence of any pre-emptive effect is of great and the severity of the autotomy, was altered by academic interest but a very short-lived effect, less applying local anaesthetic to the nerve fibre before than 2 hours, for example, might be of little clinical the operation.8 Pre-emptive use of local anaesthetic relevance, particularly if the pre-emptive treatment delayed the onset of autotomy (42 versus 23 days) carried any risk of increased morbidity. Conversely, and reduced its severity (15 versus 41%). Similarly, pre-emptive treatments which lasted for 10 hours 50 µg of intrathecal morphine reduced autotomy with minimal increase in morbidity would be of following unilateral sciatic nerve section.9 These immense clinical importance. Extrapolating from studies are perhaps more analogous to chronic brief effects demonstrated in various animal models rather than acute pain. The effect of such nerve to clinical pain is not easy. injury is believed to be analogous to neuropathic as opposed to nociceptive pain. A second problem is whether any pre-emptive effect is an effect on acute postoperative pain (nociceptive From basic science then comes the idea that the pain), or on the development of long-term sequelae same dose of an analgesic given by the same route such as phantom limb pain (neuropathic pain), or may be more effective if given before surgery on both. Different pre-emptive interventions might rather than after. Neither of these models operates be required to tackle these two different problems. on a time-scale which is a totally convincing analogy Positive or negative evidence of an effect of a of the clinical operative and postoperative states. particular intervention on nociceptive pain might The formalin model is perhaps too brief and the not apply to neuropathic pain, and vice versa. autotomy model too long. The formalin model involves inflammatory change, the autotomy model One animal model in which pre-emptive analgesic nerve damage. Clinical procedures may involve effects have been shown is the formalin test. Sub- both inflammatory response and nerve damage. cutaneous injection of formalin into the paw gives Could pre-emptive analgesia alter outcome in all rise to two ‘peaks’ of nociceptive input. Interven- pain contexts, or is it limited, operating in, for tions may be made at various times relative to the example, somatic but not visceral pain, and are injection of formalin, and the relative efficacy of the underlying mechanisms the same? Clinical the same intervention made before the formalin demonstration of pre-emptive analgesia might fail injection may be compared with the same injection if the wrong setting was chosen. The secondary made after the formalin injection. With opioids, issue is which is the pertinent outcome? For clinical intrathecal injection of the enkephalin DAMGO postoperative pain the outcome is measured over before the formalin produced 70% greater hours extending to days. The prevention of chronic inhibition of the C-fibre response than the same pain development requires outcome measurement dose injected intrathecally after the formalin.4,5 over weeks, months and, perhaps, years. With peripheral infiltrations of local anaesthetic, one before and one after the formalin injection, the behavioural response to the formalin was The clinical evidence abolished.6 Infiltration with local anaesthetic 25 minutes after the formalin made the hindpaw The aim of this study is to provide a systematic anaesthetic but did not abolish the behavioural review of the evidence that an intervention given response. Intrathecal injection of local anaesthetic before the pain starts has greater effect than the before the formalin abolished the behavioural same intervention (same dose, same route) given response; the same intrathecal dose 5 minutes after the pain. The review is performed separately after the formalin had no effect.6 NSAIDs injected for each of three classes of intervention, NSAIDs, systemically or intrathecally before the formalin local anaesthetics and opioids. injection produce a reduction in the behavioural response;6,7 it is not clear whether the same dose The inclusion criterion for the review was RCTs of NSAID given after the formalin is less effective. which addressed the question of pre-emptive treatment versus the same treatment given after The ‘end’ of the second peak of the formalin the pain had begun (Figure 46). Randomised model occurs within an hour. This is very brief studies reduce the chance of selection bias; studies when compared with clinical pain. In another which are not randomised have no such protection. 182 animal model, however, the development of Ideally, the studies should be double-blind, and Health Technology Assessment 1998; Vol. 2: No. 12

Comparisons of pre-emptive treatment with pre-emptive plus post-treatment Pre Post These studies (all were NSAID studies) were Analgesia Surgery outcome designed to compare an NSAID given before measures surgery with the same NSAID given both before and after surgery. It is not possible from these studies to answer the question of whether an analgesic intervention made before surgery is more effective, less effective or as effective as Active Placebo Any the same intervention made after surgery. Placebo Active difference? Studies included NSAID and paracetamol FIGURE 46 Design of study required to show Three RCTs with NSAIDs and one with paracetamol pre-emptive effect met the inclusion criterion. All were in oral surgery patients. Flath and colleagues11 studied four groups also double-dummy if different routes are to be of 30 patients each having endodontic surgery compared in the treatment and control groups. (Table 53). One of these four groups had pre- Studies were excluded from this review if they operative flurbiprofen and postoperative placebo. were not RCTs and if they were RCTs which did A second group had preoperative placebo and not compare pre-emptive with post treatment postoperative flurbiprofen. The preoperative dose (Figure 46). was given 30 minutes before surgery and the postoperative dose 3 hours after. On the outcome Studies were identified by a MEDLINE search and measures of categorical scale of pain intensity and by hand searching. The MEDLINE search (Silver VAS pain intensity, there was no evidence of a pre- Platter MEDLINE v. 3.0 and 3.1) covered 1966–May emptive effect. The study had adequate sensitivity to 1993. The strategy was designed to identify the detect a difference because one of the groups had maximum number of randomised and/or double- preoperative and postoperative placebo, and the blind reports by using a combination of text words, pain scores in that group were significantly higher ‘wild cards’ and MeSH terms as described previ- than those in the groups who had flurbiprofen. ously.10 Medical journals were searched by hand. They were selected from a list of the 50 journals Sisk and colleagues14 compared diflunisal, 1 g, with with the highest number of reports in MEDLINE, placebo in 20 patients having third molar extrac- and nine specialist journals which were not tions (Table 53). The design was crossover, as shown included in that list or which were not indexed. in Figure 46. Over 8 hours there was no significant The search process included volumes published difference between preoperative and postoperative between 1950 and 1994. The studies included dosing, using categorical and VAS pain intensity (and excluded) are shown in Table 52.11–40 scores. Sisk and Grover17 used a similar design to investigate naproxen, 550 mg, in third molar Excluded studies extraction (Table 53). Again there was no significant Comparisons of pre-emptive treatment with no difference between pre- and postoperative dosing, treatment (whether or not randomised) using categorical and VAS pain intensity scores. Several of the excluded papers (Table 52) are often quoted as showing evidence of a pre-emptive effect. Gustafsson and colleagues20 also used a two- They were, however, designed to show that an occasion crossover design in third molar extrac- analgesic intervention made before surgery was tion, comparing paracetamol, 1 g, with placebo more effective than no intervention at all, and did (Table 52). Using a VAS pain intensity scale and not ask whether an analgesic intervention made time to first analgesic as outcome measures there before surgery is more, less or as effective than the was no significant difference between preoperative same intervention made after surgery. Nonetheless, and postoperative dosing. reviewers have reached the conclusion that these studies produce evidence of a pre-emptive effect. These four studies provide a consistent answer to While a positive result in such studies suggests a the question. No measurable difference was found worthwhile clinical benefit, it is not evidence for or between the same dose given preoperatively and against a pre-emptive effect. Such evidence requires postoperatively. All four studies necessarily used the control of the same intervention made after local anaesthetic; none used opioids. The balance surgery (Figure 46). of the evidence is therefore that, at normal 183 Pre-emptive analgesia: a systematic review of clinical studies to 1994

therapeutic oral doses of NSAID, no pre-emptive had significantly longer time until remedication, effect was demonstrable. clear evidence of a pre-emptive effect.

Local anaesthetic Turner and Chalkiadis25 compared infiltration Studies of pre-emptive effect with local anaesthetics after induction (29 patients) with infiltration may be divided into trials of epidural (spinal), after surgery (32 patients) and with no infiltration nerve block and infiltration (Table 54). (29 patients) in appendicectomy (Table 54). The outcome measures were VAS pain intensity scores Epidural and PCA consumption. They found no significant Dahl and colleagues12 used a parallel group design difference between the groups. on 32 colonic surgery patients (Table 54). Epidural bolus and infusion of a local anaesthetic and opioid The study which did show a pre-emptive effect combination were given 40 minutes before surgery is the least subject to criticism,24 but it is still for the preoperative group, and after surgery for balanced by six negative studies, one of which the postoperative group (some 2 hours after the is an infiltration study of similar design. The preoperative group). There was no dummy paediatric studies21,23 both had the problem injection. On categorical and VAS pain of pain scoring in children but neither showed intensity scales there was no evidence of any pre-emptive effect. Importantly neither study a pre-emptive effect. involved the use of opioid; both studies therefore sought but did not find a pre-emptive effect of local Pryle and colleagues15 used a similar study design anaesthetic alone. It is difficult to understand why in 36 abdominal hysterectomy patients (Table 54). one infiltration study should have produced a Local anaesthetic with adrenaline was given as a positive result24 when the other did not.25 lumbar epidural bolus either 40 minutes before incision or after surgery (75 minutes after the In all the negative studies the local anaesthetic preoperative group). On the outcome measures of intervention worked well, whether given before or categorical and VAS pain intensity scales, time to after surgery. The power of these studies (is the lack first use of intravenous morphine via PCA and of difference a true result?) thus becomes a major amount of intravenous morphine via PCA, there issue. Also, it is not known whether these studies was no demonstrable pre-emptive effect. were sufficiently sensitive to measure an effect if there was one. The one study which did use a no- Rice and colleagues21 compared caudal blocks treatment control25 did not measure any significant pre- and postsurgery (Table 54) in 40 children difference between infiltration (pre or post) and no having outpatient surgery (mean operation time infiltration. This presents us with a set of negative 30 minutes). An objective pain score did not show studies, without internal sensitivity checks, none of any pre-emptive effect. Gunter and colleagues23 which is of adequate size to be totally convincing. used a similar design in 24 boys having hypospadias repair (Table 54). The caudal block before surgery Opioids did reduce operating time and blood loss signifi- Four opioid studies (Table 55) conform to the cantly compared with the same block after surgery, design required to answer the pre-emptive but there was no significant difference in the pain question. Katz and colleagues13 looked at spinal outcomes of time to first analgesic or on overall opioid, while Richmond and colleagues,16 analgesic consumption. Amanor-Boadu and colleagues19 and Wilson and colleagues22 investigated the intravenous route. Nerve block Dierking and colleagues18 compared inguinal field In the study by Katz and colleagues,13 30 thoracic block pre- and postoperatively in 32 patients having patients were randomised to lumbar epidural herniorrhaphy (Table 54). Using categorical and fentanyl infusion, either pre- or intraoperatively VAS pain intensity scales there was no evidence (Table 55). The infusion in the post-incision group of a pre-emptive effect. was started 15 minutes after incision. Using VAS pain intensity and PCA intravenous morphine as Infiltration outcome measures, they found significantly lower Ejlersen and colleagues24 investigated pre- VAS pain intensity scores at 6 hours in the pre- and postoperative wound infiltration in 37 operative group (with no significant difference in herniorrhaphy patients (Table 54). Using time to the PCA morphine consumption). From 12 hours remedication as the outcome measure, the patients to 24 hours, they found significantly lower PCA 184 who had the infiltration 5 minutes before incision morphine consumption in the preoperative Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 52 Pre-emptive analgesia: studies included or excluded from the review

NSAID Local anaesthetic Opioid

Studies included Flath, et al., 198711 Dahl, et al., 199212 Katz, et al., 199213 Sisk, et al., 198914 Pryle, et al., 199315 Richmond, et al., 199316 Sisk & Grover, 199017 Dierking, et al., 199218 Amanor-Boadu, et al., 199319 Gustafsson, et al., 198220 Rice, et al., 199021 Wilson, et al., 199422 Gunter, et al., 199023 Ejlersen, et al., 199224 Turner & Chalkiadis,199425

NSAID + local anaesthetic + opioid: Kavanagh, et al., 199426

Studies excluded Pre-emptive treatment with no postoperative comparison whether or not randomised Hutchison, et al., 199027 Tverskoy, et al., 199028 McQuay, et al., 198829 McGlew, et al., 199130 Jebeles, et al., 199131 Koskinen, et al., 199132 Smith & Brook, 199033 Bugedo, et al., 199034 Campbell, et al., 199035 Bach, et al., 198836 Narchi, et al., 199137

RCT but pre-emptive treatment plus postoperative treatment compared with postoperative only Hill, et al., 198738 Dupuis, et al., 198839 Murphy & Medley, 199340

TABLE 53 Pre-emptive analgesia: NSAID and paracetamol studies reviewed

Study Design Number Procedure Treatments Outcome Timing Timing Outcome of measures pre- post- (pre- vs.post- patients operative operative operative)

Flath, et al., Parallel 120 Endodontic Flurbiprofen, categorical/ 15 minutes 3 hours NSD 198711 100 mg, vs. placebo VAS PI

Sisk, et al., Cross- 20 Third molar Diflunisal, 1 g, vs. categorical/ 30 minutes 30 minutes NSD 198914 over placebo VAS PI

Sisk & Cross- 36 Third molar Naproxen, 550 mg, categorical/ 30 minutes 30 minutes NSD Grover, over vs. placebo VAS PI 199017

Gustafsson, Cross- 50 Third molar Paracetamol, 1 g, VAS PI 45 minutes 35 minutes NSD et al., 198220 over vs. placebo TFA group with no significant difference in the VAS In a study of similar design, Amanor-Boadu and pain intensity score. colleagues19 looked at the effect of morphine, 5 mg intravenous, given at induction or at closure to Richmond and colleagues16 randomised 76 total 41 body surface surgery patients (Table 55). Out- abdominal hysterectomy patients to morphine, come measures were the time to first analgesic and 10 mg intramuscular, as premedication, morphine, the categorical and VAS pain intensity scores at that 10 mg intravenous, at induction or morphine, time. The categorical pain intensity scores were 10 mg intravenous, at closure (Table 55). Analgesic significantly lower at the time of remedication in outcome measures were VAS pain intensity scores the group given morphine at induction compared and PCA intravenous morphine consumption. with the group given morphine at closure. They found significantly lower PCA intravenous morphine consumption in the group who had Wilson and colleagues22 randomised 40 total received intravenous morphine at induction abdominal hysterectomy patients to alfentanil, compared with the group who received the 40 µg/kg intravenous, at induction or after skin same dose by the same route at closure. incision (Table 55). Analgesic outcome measures 185 Pre-emptive analgesia: a systematic review of clinical studies to 1994

TABLE 54 Pre-emptive analgesia: local anaesthetic studies reviewed

Study Design No. of Procedure Treatments Outcome Timing Timing Outcome patients measures pre- post- (pre- vs. postoperative operative operative)

Epidural

Dahl, et al., Parallel 32 Major Epidural VAS and bolus + bolus + NSD 199212 colonic (T9–T12): categorical infusion infusion after surgery bupivacaine PI at rest, 40 minutes surgery and morphine cough and preoperative (still asleep) sitting up

Pryle, et al., Parallel 36 Abdominal Epidural VAS and 40 minutes after surgery NSD 199315 hysterectomy (lumbar): categorical pre-incision (still asleep) bupivacaine + PI, PCA adrenaline morphine, i.v.

Rice, et al., Parallel 40 Hernia, Caudal: Paediatric after induction after surgery NSD 199021 orchidopexy bupivacaine objective score,TFA

Gunter, et al., Parallel 24 Distal Caudal: TFA, after induction after surgery NSD 199023 hypospadias bupivacaine analgesic needs

Nerve block

Dierking, Parallel 32 Herniorr- Inguinal field VAS and 15 minutes after closure NSD et al., 199218 rhaphy block: lignocaine categorical preoperative PI at rest, cough and moving

Infiltration

Ejlersen, et al., Parallel 37 Hernior- Infiltration: TFA, 5 minutes pre- before closure Significantly 199224 rhaphy lignocaine analgesic incision (19)(18) delayed needs remedication time in preoperative group

Turner & Parallel 90 Appendicec- Infiltration: VAS PI lying 3 minutes pre- at closure (32) NSD on VAS Chalkiadis, tomy (29 lignocaine and sitting, incision (29) PI or PCA 199425 controls with PCA no infiltration)

were VAS pain intensity scores and PCA intra- 24 hours.16 In the morphine, 5 mg intravenous, venous morphine consumption. They found no study only one of the two outcomes showed a significant difference in PCA morphine consump- significant effect. These three studies showing a tion but significantly higher pain scores at rest in weak positive pre-emptive effect with opioid are the pre-emptive group. balanced by a negative effect.22 The negative study is difficult to interpret because the difference in Three studies suggest that opioids may have a pain score at the same PCA consumption may pre-emptive effect. Unfortunately none of them mean a failure of study sensitivity. is perfect. In the epidural study, there was a significant pre-emptive effect at only one of the six Using pre-emptive NSAID, local anaesthetic and VAS pain intensity measurement points. In the opioid together morphine, 10 mg intravenous, study, PCA Kavanagh and colleagues26 compared a premedi- 186 consumption was reversed in the subsequent cation of intramuscular morphine, 0.15 mg/kg, Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 55 Pre-emptive analgesia: opioid studies reviewed

Study Design No. of Procedure Treatments Outcome Timing Timing Outcome patients measures pre- post- (pre- vs. postoperative operative operative)

Epidural

Katz, et al., Parallel 30 Thoracotomy Lumbar VAS PI and 30-minute 30-minute Significant 199213 epidural: PCA infusion started infusion started difference fentanyl morphine, i.v. 55 minutes 15 minutes (VAS PI lower pre-incision after incision at 6 hours and PCA morphine lower at 12–24 hours in pre-emptive group)

Intravenous

Richmond, Parallel 76 Total Morphine, VAS PI and i.m. pre- i.v. at closure Significant et al., 199316 abdominal 10 mg, i.v. or PCA medication (16) (21) difference hysterectomy i.m. morphine, i.v. or i.v. at (PCA mor- induction (23) phine lower to 24 hours in i.v. pre-emptive group at equivalent VAS PI)

Amanor- Parallel 41 Body surface Morphine, VAS and At induction At closure (20) Significant Boadu, et al., 5 mg, i.v. categorical (21) difference 199319 PI at TFA (categorical PI lower at TFA in pre-emptive group)

Wilson, et al., Parallel 40 Total Alfentanil, VAS PI and i.v. at induction i.v. 1 minute Significant 199422 abdominal 40 µg/kg,i.v. PCA (20) after incision difference hysterectomy morphine, i.v. (20) (VAS PI (rest) higher in pre- emptive group at same PCA) perphenazine, 0.03 mg/kg, and rectal (three drug classes) interventions to demonstrate a indomethacin, 100 mg, coupled with intercostal pre-emptive effect. local anaesthetic with intramuscular midazolam premedication, 0.05 mg/kg, and saline intercostal blocks. Thirty thoracotomy patients were random- Comment ised to pre-emptive or control and compared using VAS pain intensity and PCA morphine consump- The evidence for pre-emptive effects should be tion. There were no significant differences in VAS answered separately for each of the three drug pain intensity scores. PCA consumption was classes reviewed, because the answers may be significantly lower at 6 hours in the pre-emptive different. For NSAIDs and paracetamol there are group, there was no significant difference at four good studies, all with no evidence of a pre- 12 hours and, at 24 hours and 48 hours, the emptive effect. If there is a pre-emptive effect of morphine consumption in the pre-emptive group NSAID it is unlikely to be seen with conventional was significantly greater than control. This study dosing. In the case of local anaesthetics, one did not compare the same intervention made infiltration study showed a pre-emptive effect, before and after but has been included because another of similar design did not. Five other it is the clearest example of an RCT using multiple studies, spinal and nerve block, did not show any 187 Pre-emptive analgesia: a systematic review of clinical studies to 1994

effect. These negative local anaesthetic studies References have been criticised41 because they were 1. Woolf CJ. Evidence for a central component of post- ‘contaminated’ by opioid but, in two of the injury pain hypersensitivity. Nature 1983;306:686–88. studies,21,13 patients received no opioid. Perhaps a stronger criticism of these negative studies is 2. Wall PD. The prevention of postoperative pain. that they lacked power. Pain 1988;33:289–90. 3. McQuay HJ, Dickenson AH. Implications of nervous The evidence with opioids is inconclusive. The system plasticity for pain management. Anaesthesia evidence from fundamental studies for a pre- 1990;45:101–2. 4,9 emptive effect of opioid is stronger than its 4. Dickenson AH, Sullivan AF. Subcutaneous formalin- equivalents for local anaesthetic and NSAID. induced activity of dorsal horn neurones in the rat: Inevitably what is now needed are studies of differential response to an intrathecal opiate adequate design and size to establish whether administered pre or post formalin. Pain or not there is indeed a measurable pre-emptive 1987;30:349–60. effect of opioid in man. If the intravenous route 5. Dickenson AH. Recent advances in the physiology can be used to answer the question, studies are and pharmacology of pain: plasticity and its easier to perform than if the effect was only found implications for clinical analgesia. J Psychopharmacol with spinal routes. The caveat must be that the 1991;5:342–51. dose used in the (intrathecal) basic studies was 6. Coderre TJ, Vaccarino AL, Melzack R. Central large (up to 50 µg) and that the effect was nervous system plasticity in the tonic pain response demonstrated via the intrathecal route. This to subcutaneous formalin injection. Brain Res makes the human findings with relatively small 1990;535:155–8. intravenous doses all the more remarkable. Evidence of a pre-emptive effect with opioid 7. Malmberg AB, Yaksh TL. Hyperalgesia mediated by would offer great potential benefit to patients spinal glutamate of substance P receptor blocked by spinal cyclooxygenase inhibition. Science with postoperative pain. It would also be important 1992;257:1276–9. to know if such a pre-emptive effect applied to neuropathic pain.9,36 8. Selzer Z, Beilin BZ, Ginzburg R, Paran Y, Shimko T. The role of injury discharge in the induction of One important methodological issue is that neuropathic pain behavior in rats. Pain 1991;46:327–36. increasingly investigators are using two postoperative outcomes, a VAS for pain intensity 9. Puke MJC, Wiesenfeld-Hallin Z. The differential and PCA consumption. The assumption is made effects of morphine and the alpha2-adrenoreceptor that patients will use PCA to achieve similar levels agonists clonidine and dexmedetomidine on the of VAS pain intensity. If the VAS pain intensity prevention and treatment of experimental values for pre-emptive and control are not neuropathic pain. Anesth Analg 1993;77:104–9. significantly different but the PCA consumption 10. Jadad AR, McQuay HJ. A high-yield strategy to does show a significant difference, that is a valid identify randomized controlled trials for systematic result. The corollary is not valid. There may also reviews. Online J Curr Clin Trials [serial online] 1993; be advantage (at least to the systematic reviewer) Doc No 33:3973 words; 39 paragraphs; 5 tables. in using pain relief rather than pain intensity 11. Flath RK, Hicks ML, Dionne RA, Pelleu GB. Pain outcomes. Combining data across studies is much suppression after pulpectomy with preoperative more valid for pain relief than pain intensity.42 flurbiprofen. J Endodont 1987;13:339–47. Ideally such combination of data would increase 12. Dahl JB, Hansen BL, Hjortsø NC, Erichsen CJ, power and help to answer the clinical pre- Møiniche S, Kehlet H. Influence of timing on the emptive question. effect of continuous extradural analgesia with bupivacaine and morphine after major abdominal One final point which gets forgotten is that surgery. Br J Anaesth 1992;69:4–12. acute tolerance is well known with opioids. Two 16,26 13. Katz J, Kavanagh BP, Sandler AN, Nierenberg H, of the pre-emptive studies showed that pre- Boylan JF, Friedlander M, et al. Preemptive emptive treatment led to significant increase in analgesia: clinical evidence of neuroplasticity postoperative analgesic consumption. It may be contributing to postoperative pain. Anesthesiology that any pre-emptive effect of opioids would be 1992;77:439–46. counteracted by induction of acute tolerance. 14. Sisk AL, Mosley RO, Martin RP. Comparison of This, however, is not in accord with the basic preoperative and postoperative diflunisal for science demonstration of a pre-emptive suppression of postoperative pain. J Oral Maxillofac 188 opioid effect. Surg 1989;47:464–8. Health Technology Assessment 1998; Vol. 2: No. 12

15. Pryle BJ, Vanner RG, Enriquez N, Reynolds F. 29. McQuay HJ, Carroll D, Moore RA. Postoperative Can pre-emptive lumbar epidural blockade reduce orthopaedic pain – the effect of opiate pre- postoperative pain following lower abdominal medication and local anaesthetic blocks. Pain surgery? Anaesthesia 1993;48:120–3. 1988;33:291–5.

16. Richmond CE, Bromley LM, Woolf CJ. Preoperative 30. McGlew IC, Angliss DB, Gee GJ, Rutherford A, morphine pre-empts postoperative pain. Lancet Wood AT. A comparison of rectal indomethacin 1993;342:73–5. with placebo for pain relief following spinal surgery. Anaesth Intensive Care 1991;19:40–5. 17. Sisk AL, Grover BJ. A comparison of preoperative and postoperative naproxen sodium for suppression 31. Jebeles JA, Reilly JS, Gutierrez JF, Bradley EL, Kissin of postoperative pain. J Oral Maxillofac Surg I. The effect of pre-incisional infiltration of tonsils 1990;48:674–8. with bupivacaine on the pain following tonsillectomy under general anesthesia. Pain 1991;47:305–8. 18. Dierking GW, Dahl JB, Kanstrup J, Dahl A, Kehlet H. Effect of pre- vs postoperative inguinal field 32. Koskinen R, Tigerstedt I, Tammisto T. The effect of block on postoperative pain after herniorrhaphy. peroperative alfentanil on the need for immediate Br J Anaesth 1992;68:344–8. postoeprative pain relief. Acta Anaesthesiol Scand 1991;35 suppl 96:O21. 19. Amanor-Boadu SD, Jadad AR, Glynn CJ, Jack TM, McQuay HJ. Influence of pre and postoperative 33. Smith AC, Brook IM. Inhibition of tissue morphine administration on pain after body surface prostaglandin synthesis during third molar surgery: surgery: a double-blind randomised controlled use of preoperative fenbufen. Br J Oral Maxillofac study. In: Proceedings 7th World Congress on Surg 1990;44:251–3. Pain, 1993. Oxford: IASP/Elsevier;538–9. 34. Bugedo GJ, Cárcamo CR, Mertens RA, Dagnino JA, 20. Gustafsson I, Nystrom E, Quiding H. Effect of Muñoz HR. Preoperative percutaneous ilioinguinal preoperative paracetamol on pain after oral surgery. and iliohypogastric nerve block with 0.5% bupi- Eur J Clin Pharmacol 1983;24:63–5. vacaine for post-herniorrhaphy pain management 21. Rice LJ, Pudimat MA, Hannalah RS. Timing of in adults. Reg Anesth 1990;15:130–3. caudal block placement in relation to surgery does 35. Campbell WI, Kendrick R, Patterson C. Intravenous not affect duration of postoperative analgesia in diclofenac sodium. Does its administration before paediatric ambulatory patients. Can J Anaesth operation suppress postoperative pain? Anaesthesia 1990;37:429–31. 1990;45:763–6. 22. Wilson RJT, Leith S, Jackson IJB, Hunter D. Pre- 36. Bach S, Noreng MF, Tjellden NU. Phantom limb emptive analgesia from intravenous administration pain in amputees during the first 12 months of opioids. Anaesthesia 1994;49:591–3. following limb amputation, after preoperative 23. Gunter JB, Forestner JE, Manley CB. Caudal lumbar epidural blockade. Pain 1988;33:297–301. epidural anesthesia reduces blood loss during hypospadias repair. J Urol 1990;144:517–19. 37. Narchi P, Benhamou D, Fernandez H. Intra- peritoneal local anaesthetics for shoulder pain after 24. Ejlersen E, Andersen HB, Eliasen K, Mogensen T. day-case laparoscopy. Lancet 1991;338:291–5. A comparison between preincisional and post- incisional lidocaine infiltration and postoperative 38. Hill CM, Carroll MJ, Giles AD, Pickvance N. pain. Anesth Analg 1992;74:495–8. Ibuprofen given pre and post operatively for the relief of pain. Int J Oral Maxillofac Surg 25. Turner GA, Chalkiadis G. Comparison of pre- 1987;16:420–4. operative with postoperative lignocaine infiltration on postoperative analgesia requirements. Br J 39. Dupuis R, Lemay H, Bushnelle MC, Duncan GH. Anaesth 1994;72:541–3. Preoperative flurbiprofen in oral surgery: a method of choice in controlling postoperative pain. 26. Kavanagh BP, Katz J, Sandler AN, Nierenberg H, Pharmacotherapy 1988;8:193–200. Roger S, Boylan JF, et al. Multimodal analgesia before thoracic surgery does not reduce 40. Murphy DF, Medley C. Preoperative indomethacin postoperative pain. Br J Anaesth 1994;73:184–9. for pain relief after thoracotomy: comparison with postoperative indomethacin. Br J Anaesth 27. Hutchison GL, Crofts SL, Gray IG. Preoperative 1993;b:298–300. piroxicam for postoperative analgesia in dental surgery. Br J Anaesth 1990;65:500–3. 41. Katz J. Preop analgesia for acute pain. Lancet 1993;342:65–6. 28. Tverskoy M, Cozacov C, Ayache M, Bradley EL, Kissin I. Postoperative pain after inguinal 42. Jadad-Bechara AR. Meta-analysis of randomised herniorrhaphy with different types of anesthesia. clinical trials in pain relief [DPhil thesis]. Oxford: Anesth Analg 1990;70:29–35. University of Oxford, 1994. 189

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 19 TENS in acute postoperative pain

Summary outcome was sought in studies of TENS in acute postoperative pain. The aim was to examine the evidence for the importance of randomisation of TENS in acute postoperative pain. Controlled trials were sought; Methods randomisation and analgesic and adverse effect outcomes were summarised. A total of 46 reports A number of different search strategies were used were identified by searching strategies, of which to identify controlled trials for TENS in acute 17 with 786 patients could be regarded unequivoc- postoperative pain in both MEDLINE (1966–95: ally as RCTs in acute postoperative pain. No meta- Knowledge Server v. 3.25: January 1996) and the analysis was possible. In 15 of the 17 RCTs, TENS Oxford Pain Relief Database (1950–92).13 The was judged to have no benefit over placebo. Of the terms ‘TENS’ and ‘transcutaneous electrical nerve 29 excluded trials, 19 had pain outcomes but were stimulation’ were used in searching, including not RCTs; in 17 of these, their authors concluded combinations of these words. Additional reports that TENS had a positive analgesic effect. No were identified from the reference lists of retrieved adverse effects were reported. Non-randomised reports, review articles and textbooks. trials overestimate treatment effects. Inclusion criteria were: This chapter was published in full in 1996 by Carroll and colleagues.1 • full journal publication • TENS • postoperative pain with pain outcomes. Introduction Reports of TENS for the relief of other acute pain TENS was originally developed as a way of conditions, such as labour pain, acute infections controlling pain through the ‘gate’ theory.2 There and procedures, or those in which there were is conflicting professional opinion about the use less than ten patients per treatment group were of TENS in acute postoperative pain. The recom- excluded. Abstracts and review articles were not mendations of the Agency for Health Care Policy considered. Unpublished reports were not sought. and Research3 for acute pain management state that Neither authors of reports nor manufacturers of TENS is “effective in reducing pain and improving TENS equipment were contacted. physical function”, while an earlier report from the UK College of Anaesthetists’ working party on pain Two types of control predominated: open studies after surgery4 says that “TENS is not effective as the compared TENS with conventional postoperative sole treatment of moderate or severe pain after analgesia (intramuscular opiate) or with disabled surgery”. Some textbooks recommend or strongly TENS instruments (sham TENS). Some studies recommend TENS for postoperative pain,5–9 used blinded observers. While there was no prior although one at least is uncertain.10 TENS is of hypothesis that TENS could not be blinded doubtful benefit in labour pain11 and no systematic adequately, it was determined that, despite the review of its use in chronic pain could be found. considerable efforts documented in some reports, adequate blinding was impossible in practice. Quality of methods used in clinical trials has been shown to be a key determinant of the eventual Each report which could possibly meet the results. Schulz and colleagues12 have demonstrated inclusion criteria was read by each author that trials which are not randomised or are inade- independently and scored for inclusion and quality quately randomised exaggerate the estimate of using a 3-item scale.14 Included reports had one treatment effect by up to 40%. Studies which are point for randomisation, a further point if this had not fully blinded can exaggerate the estimate of been done correctly, and a third if the number and treatment effect by up to 17%. Evidence of the reasons for withdrawals were given. Authors met to effect of randomisation in trials with pain as an agree that studies were randomised, or whether 191 TENS in acute postoperative pain

the description of the method of randomisation overall effectiveness. Reports which had pain was adequate.12 measures but which were not randomised or were inadequately randomised were examined Information about the surgery, numbers of for positive or negative analgesic effectiveness patients, study design and duration of treatment of TENS using the judgement of their authors. was extracted from randomised reports. The type of TENS equipment, its settings and the method and frequency of its use and placement of elec- Results trodes was also extracted. Control group design and the use of TENS in these controls was similarly Of the 46 reports that were considered, three did noted. Pain outcomes, overall findings and con- not have pain outcomes, three had fewer than clusions were noted for each report, together with ten patients per group, three had methodological any adverse effect information. problems and one reported on pain during rather than after a procedure. These were not A judgement was then made as to whether the considered further. overall conclusion of randomised reports was positive or negative for the analgesic effectiveness A total of 19 reports were either not RCTs or of TENS. Post-hoc sub-group analysis in the original the method of randomisation was inappropriate reports was not considered in the judgement of (Table 56).15–33 Of the 19 reports with pain measures

TABLE 56 Non-randomised reports of TENS in acute postoperative pain

Study Pain condition or Description Authors’ judgement about operation type analgesic effectiveness

Ali, et al., 198115 Upper abdominal Not RCT Positive

Bussey & Jackson, 198116 Cholecystectomy, Retrospective study not RCT Positive hernia repair

Cooperman, et al., 197717 Upper abdominal Inadequate randomisation Positive

Cornell, et al., 198418 Foot Not RCT: matched case control Positive

Hollinger, 198619 Caesarian section Not RCT Positive

Hymes, et al., 197420 General Not RCT Positive

Issenman, et al., 198521 Spinal fusion Not RCT Positive

Lanham, et al., 198422 Foot Retrospective, not RCT Positive

Merrill, 198723 Urological Not RCT Positive

Merrill, 198824 Urological Not RCT Positive

Merrill, 198825 Urological Not RCT Positive

Neary, 198126 Abdominal, thoracic Not RCT Positive

Reuss, et al., 198827 Cholecystectomy Not RCT Negative

Rooney, et al., 198328 Thoracotomy Inadequate randomisation method Positive

Schomberg & Carter- Laparotomy Retrospective, not RCT Positive Baker, 198329

Schuster & Infante, 198030 Low back Not RCT Positive

Solomon, et al., 198031 Lumbar, hip, Retrospective, not RCT Positive gynaecological

Stabile & Mallory, 197832 Knee and hip joint Not RCT Positive

33 192 Strayhorn, 1983 Gastric bypass Not RCT Negative Health Technology Assessment 1998; Vol. 2: No. 12

excluded because they were either not randomised Pike43 studied 40 patients after total hip replace- or inadequately randomised, 17 were judged by ment. The study had as its main outcome measure their authors to have positive analgesic results for the number of pethidine (meperidine) injections TENS in acute postoperative pain. in the first 2 postoperative days and a retrospective global rating. Patients with active TENS had Of the 17 randomised studies with pain outcomes significantly fewer pethidine injections on the first found, 15 were judged to show no analgesic benefit postoperative day as well as higher scores on global of TENS in acute postoperative pain (Table 57). rating of treatment. VanderArk and McGrath47 recruited 100 patients having abdominal and TABLE 57 Analgesic benefit of TENS in postoperative pain thoracic surgery in 2 months and, although there was more success with active TENS used for 20 Analgesic result minutes three times a day, maximal relief was “almost invariably associated with the first stimu- Positive Negative lation”. Generally there were no obvious differ- Randomised 2 15 ences between the use of TENS in these two positive studies and the 15 which showed no benefit. Inadequate or not randomised 17 2 Adverse events No report described systematic recording of Randomised studies adverse events nor were any reported. The randomised studies had information from 786 patients (Table 58).27,34–49 TENS was used after various operative procedures including cardio- Comment thoracic, major orthopaedic and gastrointestinal surgery. Ten different TENS machines were used The gold standard in clinical trials is adequate ran- with different control settings and durations of domisation.12 For nearly 20 years, non-randomised treatment; in six studies, individual titration of studies have been shown to yield larger estimates of settings was reported. TENS was compared with treatment effects than studies using random allo- sham TENS without batteries, with batteries cation.50 The degree of the exaggeration of treat- reversed or with sub-threshold stimulation in ment effect when randomisation is inappropriate can 14 studies; in the other three, TENS plus intra- be as much as 40%.12 These findings underpin the muscular opiate was compared with intramuscular inclusion criteria chosen in systematic reviews. opiate alone. Quality scores were generally 1 or 2 out of a maximum of 3. The most common out- For TENS in acute postoperative pain, 17 of come measures reported were analgesic consump- 19 reports with pain outcomes which were either tion and a variety of pain score measurements. not randomised or inappropriately randomised Information was not presented in formats which claimed TENS to be effective, compared with allowed extraction for meta-analysis (Table 58). two of 17 RCTs (Table 57).

TENS versus sham TENS The possibility of bias exists. The method of ran- Of the 17 included RCTs, 14 compared TENS domisation was described in only two reports.17,28 with sham TENS. Not one found any difference. The method described was inadequate in both, One of the 1449 reported no significant difference one using a nurse to randomise patients17 and the between TENS and sham TENS for analgesic con- other using alternate allocation.28 Reports which sumption but did report a statistically significant said only that they were randomised may also difference for pain intensity in favour of the active have used an inadequate method. TENS; the published results, however, used a one- tailed statistical test which was judged inappropriate. That these data represent the lowest common denominator of information, essentially vote TENS versus opiate control counting rather than a more sophisticated analysis, Of the 17 included RCTs, seven compared opiate reflects the nature of the analgesic scoring methods plus TENS with opiate alone, four of which also that predominated in the original reports. Pain included sham TENS. Of the seven studies, five scoring using analogue or categorical scales was failed to detect any differences in analgesic con- reported as a mean (an unreliable statistic51); sumption or pain measurements between TENS alternatively, mean analgesic consumption or time and non-TENS controls. Two reports were judged to first analgesic was used. None of these allowed by their authors and by us to be positive.43,47 data extraction for further statistical analysis or 193 TENS in acute postoperative pain

TABLE 58 Randomised studies of TENS in acute postoperative pain

Study Operation Study design Number of TENS details TENS control TENS control Pain Results for Judgement Score type and duration patients setting outcomes pain outcomes of treatment periods

Conn, et al., Appendicectomy Parallel group: 42 Dow Corning Fixed rate Sham TENS VAS PI at NSD between sham and Negative 1 198634 TENS 15; sham Wright, single (tingling sensation (not turned on) 48 hours; active TENS for pain and TENS 13; stand channel, electrodes preoperatively). analgesic drug consumption; ard postoperative (either side of consumption significant difference for analgesia 14. wound). (24, 48 hours). PI control vs.TENS and 48 hours sham TENS (p < 0.01).

Cushieri, et al., Abdominal Parallel group: 106 Codman, dual Fixed rate (tingling Sham TENS VAS PI: average NSD between TENS and Negative 2 198535 sham TENS 53; channel, 2 sensation pre- (batteries pain twice daily, sham TENS. TENS 53. electrodes (either operatively), reversed) morphine 72 hours side of wound). rectangular wave consumption. form, pulse width (170 ms), pulse rate 80/second, output 15 milliamps.

Davies, Caesarean section Parallel group 35 Stim-Tec EPC Mini, Amplitude Sham TENS VAS PI: hourly, No overall difference in Negative 2 198236 (female): general Model 6011, dual individually titrated, (no batteries) time to first analgesic consumption anaesthetic + channel (1 channel wave fixed analgesic, or pain. TENS 10; only used), 2 (during surgery). analgesic epidural + TENS electrodes (either consumption. 11; general anaes- end of wound). thetic + sham TENS 8; epidural + sham TENS 6. 24 hours

Forster, Coronary artery Parallel group 45 Nuwave, Staodyn. Individually titrated Sham TENS Pain (0–10) on NSD TENS vs.sham Negative2 et al., 199437 bypass (males):TENS 15; 1 pair electrodes (tingling sensation). (no current) cough and rest; TENS. sham TENS 15; (T1–T5), 1 pair narcotic postoperative (either side of consumption. analgesia 15. wound). 72 hours

Galloway, Cholecystectomy Parallel group: 40 3M Tenzcare, dual Individually titrated. Sham TENS VAS PI; NSD TENS vs.sham Negative 1 et al., 198438 TENS 14; remote channel, site of (remote non- categorical PI TENS for pain or analgesic TENS 12; post- electrodes not segmental) (4-point scale) consumption. Pain but not operative described. at 24, 48 hours; analgesic consumption analgesia 14. analgesic significantly worse in 48 hours consumption. control group (p < 0.05) immediately after surgery.

Gilbert, et al., Herniorrhaphy Parallel group 40 Dow Corning, Individually titrated: Sham TENS VAS PI twice NSD TENS vs.sham Negative 1 198639 (males):TENS Wright Care, pulse duration (no current) daily; analgesic TENS. 20; sham TENS dual channel, 2 180 µs, frequency consumption. 20. electrodes (either 70 Hz, amplitude 72 hours side of wound). 7.5.

Lim, et al., Abdominal Parallel group: 34 Neuromed 3722, Individually titrated Sham TENS VAS PI (2, 4, 6, NSD TENS vs.sham Negative 2 198340 TENS 15; sham 2 electrodes (tingling sensation). (batteries 24, 48 hours); TENS. TENS 15. (either side of reversed) analgesic 48 hours wound). consumption.

McCallum, Laminectomy Parallel group: 20 Dow Corning, Individually titrated, Sham TENS PCA morphine NSD TENS vs.sham Negative 2 et al., 198841 TENS 10; sham Wright Care, 180 µs pulse width, (no current) consumption, TENS. TENS 10. dual channel, 4 frequency 70 Hz. 24 hours. 24 hours electrodes (at each end and on either side of wound).

Navarathnam, Cardiac Parallel group: 31 3M Tenz care Individually titrated, Sham TENS 5-point NSD TENS and sham Negative 2 et al., 198442 TENS 14; sham Model 6240, dual pulse rate 5, width (no batteries) categorical PI; TENS. TENS 17. channel, 2 pairs control 3, analgesic 72 hours electrodes (either amplitude. consumption. side of the wound and mid-thoracic region).

194 continued Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 58 contd Randomised studies of TENS in acute postoperative pain

Study Operation Study design Number of TENS details TENS control TENS control Pain Results for Judgement Score type and duration patients setting outcomes pain outcomes of treatment periods

Pike, 197843 Total hip Parallel group: 40 EPC TimeTech Individually titrated, Global Significantly less Positive 1 replacement TENS 20; opiate clinical stimulator, continual assessment; pethidine consumed in control 20. dual channel, 2 pairs stimulation. analgesic TENS group on day 1 24 hours electrodes (1 pair consumption. (p < 0.001). paravertebrally (L2–S2), between trochanter and coccyx, 1 pair above iliac crest, head of fibula).

Reuss, et al., Cholecystectomy Parallel group: 64 EPC, electrodes Pulse rate Daily dose of NSD Negative 1 198827 TENS 30; opiate placed within 2 cm 50/second, pulse pethidine for control 34. of the wound. width 170 ms, 3 postoperative amplitude 0–50. days.

Smedley, Inguinal hernia Parallel group 62 3M Tenzcare dual Individually titrated Sham TENS VAS PI, 6,12, 24, NSD TENS vs.sham Negative 2 et al., 198844 repair (males):TENS channel, 2 pairs of (tingling sensation), (controls 36, 48 hours; TENS. 34; sham TENS electrodes (over 70 Hz rectangular turned off) opiate 28. first lumbar verte- pulse, amplitude. consumption. 48 hours brae and on either side of wound).

Stubbing & Thoracotomy Parallel group: 40 Dow Corning, Individually titrated, 5-point PI, 6, NSD TENS vs.sham Negative 1 Jellicoe, TENS + i.m. Wright Care 2 fixed pulse rate 24, 48 hours; TENS. 198845 omnopon 20; i.m. channel, 2 70/second, analgesic omnopon 20. electrodes (either rectangular wave- consumption; 48 hours side of incision). form, pulse width time to oral 180 µs. analgesics; length of hospital stay.

Taylor, et al., Abdominal Parallel group: 77 MedGen, Fixed pulse width Sham TENS Daily 10-point NSD TENS vs.sham Negative 1 198346 TENS 30; sham electrode, 80 ms, frequency (no current) PI. TENS or control. TENS 22; i.m. placement not 40 Hz, amplitude narcotics 25. described. individually titrated. 72 hours

VanderArk Abdominal and Parallel group: 100 Neuromed Model Frequency Sham TENS Pain; analgesic Significant difference Positive 1 & McGrath, thoracic TENS 61; sham 3700, Meditronic, 100–150/second, (no batteries) consumption; reported. 2/39 partial 197547 TENS 39. electrode site output 20–35, duration relief or complete relief 24 hours post- individually chosen. pulse duration of relief. sham TENS vs.34 /61 surgery until 250–400 ms. with active TENS. discharge,TDS Analgesic consumption x 20 minutes. not reported.

Walker, et al., Total knee Phase 2 – 48 Strodynamics, Amplitude setting Sham TENS Analgesic NSD TENS vs.sham Negative 1 199148 replacement parallel group: continuous. No individually titrated, (sub-threshold consumption; TENS or control. TENS 18; sham other information pulse duration stimulation) length of TENS 18; post- given, electrode 100 µs at hospital stay. operative placement not 70/second. analgesia 12. described. 72 hours

Warfield, Thoracotomy Parallel group: 24 3M Tenzcare 6240, Continuous stimu- Sham TENS PI 0–10; NSD TENS vs.sham Negative 2 et al., 198549 TENS 12; sham 2 electrodes placed lation, amplitude 7, (no current) analgesic TENS. Positive result TENS 12. on either side pulse rate 3, pulse consumption. reported with one-tailed 48 hours of incision. width 5. test of statistical significance. comparison between reports. While more rigorous this may be less of an issue with pharmacological pain scoring might have been used, there is no interventions.51 Blinding of procedures is much evidence that all of the reports suffered a systematic more difficult than blinding of drug studies. Most failure in analgesic measurement. of the TENS studies did make attempts at blinding, for instance by removing batteries from the TENS Inadequacy of blinding in clinical trials of anal-gesic apparatus (sham TENS) or by using staff with no interventions continues to be of concern,52 although knowledge of the study or allocation to conduct the 195 TENS in acute postoperative pain

patient assessments. Lack of blinding has been 8. Chaney ES. The management of postoperative pain. estimated to exaggerate the estimate of treatment In: Wells P, Frampton V, Bowsher D, editors. Pain: effect of trials by some 17%.12 Adequate blinding of management and control in physiotherapy. Oxford: Heinemann, 1988:253–68. TENS for both carers and patients is particularly difficult.53 None of the reports was judged to have 9. Park G, Fulton B. The management of acute pain. been blinded and this lowered the quality scores Oxford: Oxford University Press, 1991. given to the 17 randomised studies. The fact that 10. Justins DM, Richardson PH. Clinical management only two of the reports showed any positive effect of acute pain. Br Med Bull 1991;47:561–83. of TENS in acute postoperative pain is all the more striking because of this potential overestimation 11. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. of treatment effect due to lack of blinding. Transcutaneous electrical nerve stimulation in labour pain: a systematic review. Br J Obstet Gynaecol The clear message from the studies considered in 1997;104:169–75. this systematic review is that adequate random- 12. Schulz KF, Chalmers I, Hayes RJ, Altman DG. isation is an important quality standard in studies Empirical evidence of bias: dimensions of methodo- with pain outcomes. Including non-randomised logical quality associated with estimates of treatment studies in reviews may give the wrong answer. The effects in controlled trials. JAMA 1995;273:408–12. Agency for Health Care Policy and Research 13. Jadad AR, McQuay HJ. A high-yield strategy to guidelines on acute pain management included identify randomized controlled trials for systematic non-randomised reports, and this may explain their reviews. Online J Curr Clin Trials [serial online] 1993; more positive attitude towards TENS.3 Doc No 33:3973 words; 39 paragraphs; 5 tables. 14. Jadad AR, Moore RA, Carroll D, Jenkinson C, References Reynolds DJM, Gavaghan DJ, et al. Assessing the quality of reports of randomized clinical trials: is 1. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. blinding necessary? Control Clin Trials 1996;17:1–12. Randomization is important in studies with pain 15. Ali J, Yaffe CS, Serrette C. The effect of transcutane- outcomes: systematic review of transcutaneous ous electric nerve stimulation on postoperative pain electrical nerve stimulation in acute postoperative and pulmonary function. Surgery 1981;89:507–12. pain. Br J Anaesth 1996;77:798–803. 16. Bussey JG, Jackson A. TENS for postsurgical 2. Melzack R, Wall PD. Pain mechanisms: a new theory. analgesia. Contemp Surg 1981;18:35–41. Science 1965;150:971–8. 17. Cooperman AM, Hall B, Mikalacki K, Hardy R, 3. Acute Pain Management Guideline Panel. Acute Sadar E. Use of transcutaneous electrical stimu- pain management: operative or medical procedures lation in the control of postoperative pain. Am J and trauma. Clinical Practice Guideline No. 1. Surg 1977;133:185–7. Agency for Health Care Policy and Research, US 18. Cornell PE, Lopez AL, Malofsky H. Pain reduction Department of Health and Human Services. with transcutaneous electrical nerve stimulation AHCPR Publication No. 92-0032. Rockville, MD: after foot surgery. J Foot Surg 1984;23:326–33. Public Health Service, 1992:24–5. 19. Hollinger JL. Transcutaneous electrical nerve 4. Royal College of Surgeons of England and the stimulation after cesarean birth. Phys Ther College of Anaesthetists. Report of the working 1986;66:36–8. party on pain after surgery. London: Royal College of Surgeons, 1990. 20. Hymes AC, Raab DE, Yonehiro EG, Nelson GD, Printy AL. Electrical surface stimulation for control 5. Charman RA. Physiotherapy for the relief of pain. of acute postoperative pain and prevention of ileus. In: Carroll D, Bowsher D, editors. Pain management Surg Forum 1974;24:447–9. and nursing care. Oxford: Butterworth-Heinemann, 21. Issenman J, Nolan MF, Rowley J, Hobby R. 1993:146–65. Transcutaneous electrical nerve stimulation for pain control after spinal fusion with Harrington rods. 6. Woolf CJ, Thomson JW. Stimulation-induced Phys Ther 1985;65:1517. analgesia: transcutaneous electrical nerve stimulation (TENS) and vibration. In: Wall PD, 22. Lanham RH, Powell S, Hendrix BE. Efficacy of Melzack R, editors. Textbook of pain. 3rd edition. hypothermia and transcutaneous electrical nerve Edinburgh: Churchill Livingstone, 1994:1191–217. stimulation in podiatric surgery. J Foot Surg 1984;23:152–8. 7. McCaffery M, Beebe A. Pain: a clinical manual for nursing practice. St. Louis, Missouri: CV Mosby 23. Merrill DC. Electroanalgesia in urologic surgery. 196 Company, 1989. Urology 1987;29:494–7. Health Technology Assessment 1998; Vol. 2: No. 12

24. Merrill DC. FasTENS – a disposable transcutaneous 40. Lim AT, Edis G, Kranz H, Mendelson G, Selwood T, electrical nerve stimulator designed specifically for Scott DF. Postoperative pain control: contribution of use in postoperative patients. Urology 1988;31:78–9. psychological factors and transcutaneous electrical stimulation. Pain 1983;17:179–88. 25. Merrill DC. Clinical evaluation of FasTENS, an inexpensive, disposable transcutaneous electrical 41. McCallum MI, Glynn CJ, Moore RA, Lammer P, nerve stimulator designed specifically for post- Phillips AM. Transcutaneous electrical nerve operative electroanalgesia. Urology 1989;33:27–30. stimulation in the management of acute 26. Neary JM. Transcutaneous electrical nerve postoperative pain. Br J Anaesth 1988;61:308–12. stimulation for the relief of post-incisional surgical 42. Navarathnam RG, Wang IY, Thomas D, Klineberg pain. Am Assoc Nurs J 1981;49:151–5. PL. Evaluation of the transcutaneous electrical 27. Reuss R, Cronen P, Abplanalp L. Transcutaneous nerve stimulator for postoperative analgesia electrical nerve stimulation for pain control after following cardiac surgery. Anaesth Intensive Care cholecystectomy: lack of expected benefits. South 1984;12:345–50. Med J 1988;81:1361–3. 43. Pike PM. Transcutaneous electrical stimulation. Its 28. Rooney SM, Jain S, Goldiner PL. Effect of trans- use in the management of postoperative pain. cutaneous nerve stimulation on postoperative pain Anaesthesia 1978;33:165–71. after thoracotomy. Anesth Analg 1983;62:1010–12. 44. Smedley F, Taube M, Wastell C. Transcutaneous 29. Schomberg FL, Carter-Baker SA. Transcutaneous electrical nerve stimulation for pain relief following electrical nerve stimulation for postlaparotomy inguinal hernia repair: a controlled trial. Eur Surg pain. Phys Ther 1983;63:188–93. Res 1988;20:233–7.

30. Schuster GD, Infante MC. Pain relief after low back 45. Stubbing JF, Jellicoe JA. Transcutaneous electrical surgery: the efficacy of transcutaneous electrical nerve stimulation after thoracotomy. Pain relief and nerve stimulation. Pain 1980;8:299–302. peak expiratory flow rate: a trial of transcutaneous 31. Solomon R, Viernstein MC, Long D. Reduction in electrical nerve stimulation. Anaesthesia postoperative pain and narcotic use with trans- 1988;43:296–8. cutaneous electrical nerve stimulation. Surgery 46. Taylor AG, West BA, Simon B, Skelton J, Rowlingson 1980;87:142–6. JC. How effective is TENS for acute pain? Am J Nurs 32. Stabile ML, Mallory TH. The management of 1983;83:1171–4. postoperative pain in total joint replacement. 47. VanderArk GD, McGrath KA. Transcutaneous Orthop Rev 1978;7:121–3. electrical stimulation in treatment of postoperative 33. Strayhorn G. Transcutaneous electrical nerve pain. Am J Surg 1975;130:130–8. stimulation and postoperative use of narcotic analgesics. J Natl Med Assoc 1983;75:811–16. 48. Walker RH, Morris BA, Angulo DL, Schneider J, Colwell CWJ. Postoperative use of continuous 34. Conn IG, Marshall AH, Yadav SN, Daly JC, Jaffer M. passive motion, transcutaneous electrical nerve Transcutaneous electrical nerve stimulation follow- stimulation, and continuous cooling pad following ing appendicectomy: the placebo effect. Ann R Coll total knee arthroplasty. J Arthroplasty 1991;6:151–6. Surg Engl 1986;68:191–2. 49. Warfield CA, Stein JM, Frank HA. The effect of 35. Cuschieri RJ, Morran CG, McArdle CS. Transcutane- transcutaneous electrical nerve stimulation on pain ous electrical stimulation for postoperative pain. after thoracotomy. Ann Thorac Surg 1985;39:462–5. Ann R Coll Surg Engl 1985;67:127–9. 50. Chalmers TC, Matta RJ, Smith JJ, Kunzler AM. 36. Davies JR. Ineffective transcutaneous nerve Evidence favoring the use of anticoagulants in the stimulation following epidural anaesthesia. hospital phase of acute myocardial infarction. Anaesthesia 1982;37:453–7. N Engl J Med 1977;297:1091–6. 37. Forster EL, Kramer JF, Lucy SD, Scudds RA, Novick RJ. Effect of TENS on pain, medications, and 51. McQuay H, Carroll D, Moore A. Variation in the pulmonary function following coronary artery placebo effect in randomised controlled trials of bypass graft surgery. Chest 1994;106:1343–8. analgesics: all is as blind as it seems. Pain 1996;64:331–5. 38. Galloway DJ, Boyle P, Burns HJ, Davidson PM, George WD. A clinical assessment of electro- 52. Wall PD. The placebo effect: an unpopular topic. analgesia following abdominal operations. Pain 1992;51:1–3. Surg Gynecol Obstet 1984;159:453–6. 53. Deyo RA, Walsh NE, Schoenfeld LS, Ramamurthy S. 39. Gilbert JM, Gledhill T, Law N, George C. Controlled Can trials of physical treatments be blinded? The trial of transcutaneous electrical nerve stimulation example of transcutaneous electrical nerve stimu- (TENS) for postoperative pain relief following lation for chronic pain. Am J Phys Med Rehabil 197 inguinal herniorrhaphy. Br J Surg 1986;73:749–51. 1990;69:6–10.

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 20 TENS in labour pain

Summary for the relief of other pain conditions or those in which there were fewer than ten patients per TENS is used widely for relief of pain in labour. treatment group were excluded.3 However, two previous systematic reviews have questioned its effectiveness in this context. Each report which could possibly meet the Reports were sought by searching MEDLINE, inclusion criteria was read by each author inde- EMBASE, CINAHL and the Oxford Pain Relief pendently and scored for inclusion and quality Database. Outcomes included pain and adverse using a 3-item scale4 which examined random- effect measures. isation, blinding and withdrawal, and drop-outs. An included report could have a maximum score Ten RCTs were found; of the 877 women of 5 and a minimum of 1. Where the method of involved, 436 received active TENS and 441 acted treatment allocation was unconcealed (alternate as controls (sham TENS or no treatment). There allocation, for instance) the report was excluded. were no significant differences reported for A pre-hoc judgement was made that it would be prospective primary pain outcomes in any of the difficult to blind TENS and thus quality scores ten studies. Three studies reported significant were unlikely to exceed 3. differences between active and sham TENS for secondary pain outcomes. The use of additional Information about inclusion criteria for women in analgesic interventions was not different with labour, stage of labour, cervical dilatation, number active or sham TENS (relative risk 0.88 (095% of women, study design and timing and duration CI, 72–1.07)). of treatment was extracted from the reports, together with information on other analgesic The findings suggest that TENS has no significant interventions and preferences for future child- effect on pain in labour. Women in labour should birth. The type of TENS equipment, its settings be offered more effective interventions for the and the method and frequency of its use and relief of pain. positioning of electrodes was also extracted. Control group design and the use of TENS in these controls was similarly noted, including Methods the methods used to disable TENS devices (e.g. sham TENS with no battery). A number of different search strategies were used to identify eligible reports of RCTs of TENS in The effectiveness of TENS was judged by whether labour pain in MEDLINE (1966–97), EMBASE or not a statistically significant difference between (1980–97), CINAHL (1982–97), the Cochrane TENS and the control group (sham TENS or no Library (issue 2, 1997) and the Oxford Pain Relief treatment) was reported in the original report for Database (1950–95).1 The date of last search was at least one of the outcome measures used. Out- April 1997. The terms ‘TENS’, ‘transcutaneous comes were judged by us as being either primary or electrical nerve stimulation’, ‘labour’ and ‘child- secondary. Primary outcomes were defined as any birth’ were used in searching, including in combi- prospective assessment of pain intensity or relief nations, and there was no language restriction. made at the time of labour and when TENS was Additional reports were identified from the refer- in use. Secondary outcomes were defined as any ence lists of retrieved reports, review articles2 and retrospective assessment of pain or pain relief or textbooks. Manufacturers of TENS equipment were any other measure, or judgement made after not contacted. Abstracts and review articles were delivery, or after TENS had been discontinued. not considered. Unpublished reports were Secondary outcomes included the use of any not sought. additional pain interventions, the timing of such interventions and any retrospective global Inclusion criteria were full journal publication, evaluation of the study treatments. A judgement TENS, labour pain with pain outcomes and was then made as to whether the overall conclusion randomised treatment allocation. Reports of TENS of the report was positive or negative for the anal- 199 TENS in labour pain

gesic effectiveness of TENS on primary and care compared with 21 additional interventions secondary outcomes separately. Post-hoc sub-group in 16 women receiving TENS. Nesheim14 recorded analysis in the original reports was not considered that 35 women with TENS needed 49 analgesic in the judgement of overall effectiveness. Any interventions compared with 63 interventions in information on adverse effects was summarised. 35 women with sham TENS. In one study,13 all analgesic interventions were reported, other than Relative risk or benefit was calculated with the epidurals which were discouraged in this study. 95% CI using a random effect model5 for analgesic data which were not homogeneous (p < 0.1). A Figures for the number of women who received statistically significant difference from control was any other analgesic intervention were given in assumed when the 95% CI of relative risk did not five studies.11–13,15,16 The results of this secondary include 1. An NNT was calculated with 95% CIs6 for outcome for the comparisons of active TENS with any comparison which showed significance with sham TENS, together with the number of women relative risk. in the comparison, are presented in Figure 47. Overall, of 292 women, 227 (78%) given active TENS had an additional analgesic intervention Results compared with 239 of 280 women (85%) having sham TENS. There was no difference between Two additional reports were found which were active and sham TENS in the three largest studies not included in a previous review.7 Ten reports (Figure 47). The combined result of all five studies involving 877 women were included; 436 women had a relative risk of 0.88 (95% CI, 0.72–1.07; received active TENS and 441 acted as controls. Figure 48). The lack of any statistical difference The methodological details of the study designs, made the calculation of an NNT irrelevant. instructions to women before and during labour, TENS details and settings, control conditions and None of the studies were judged to have a positive methods of blinding are presented in Table 59. One result for the primary outcome measures, which study8 used cranial TENS; others used TENS with were prospective measures of pain intensity or dorsal or suprapubic stimulation. Nine different relief. For the secondary outcomes of additional TENS devices were used in the ten studies, analgesics taken or time to next analgesic, three predominantly with individual titration. studies8,11,15 were judged to have a positive result. The three positive studies included a study of Three studies used conventional analgesic cranial TENS in 20 women,16 in which only the administration (no TENS) as the control group.8–10 single outcome of other analgesics taken was used, In seven studies disabled TENS instruments (sham and a comparison of TENS plus epidural compared TENS) were used as a control group.11–17 In one with epidural alone.17 study both sham TENS and a no TENS control were used.13 In only one study16 was a sufficiently There were no reports of adverse events in any of determined attempt at blinding made to merit any the ten studies. inclusion points for blinding. This study had an inclusion quality score of 4; seven studies scored 2 and two scored 1. Comment

Pain outcomes and results for the ten studies are None of the ten studies included in this review presented in Table 60. There was no consistency reported any significant difference between the in the method of measuring pain intensity or relief. active TENS treatments and controls for any of In some studies suprapubic and back pain was the primary pain outcome measures used. This measured separately, and in others pain was strengthens the findings of previous negative measured at different stages of labour or at reviews on TENS in labour.2,18 The weak evidence different degrees of cervical dilatation. No study from secondary outcome measures that the need recorded any difference in pain intensity or relief for additional analgesics may be diminished18 was scores between TENS and control during labour. negated by an additional large trial,13 in which no difference was found between active TENS and Additional analgesic interventions were recorded sham TENS. Trial size is likely to be important in eight of the ten reports.9,11–17 In two studies, the when assessing even primary outcomes but for total number of interventions was noted. Bundsen9 weak secondary pain outcomes, such as additional recorded 17 additional analgesic interventions analgesic requirements, the effect of trial size 200 in 11 women receiving usual obstetric analgesic shown in Figure 47 was dramatic. The two trials Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 59 Randomised studies of TENS in labour pain

Study Early Study design Pre- Intra-study TENS TENS Electrode Control Blinding Quality criteria and duration study instructions machine setting details group(s) score of treatment instructions periods

Bunsden, Induced labour Parallel group: Standard Women in control Custom-built Individually Two electrodes: Individually titrated Open, no attempt 2 et al., 19829 only (amniotomy 1.TENS (16); information group allowed stimulator titrated by one one supra-pubic, conventional to blind.TENS ± oxytocin), who 2. control group about study conventional author: two one low back. obstetric analgesia given by one of did not desire (11). Dorsal and and available obstetric analgesia frequencies (high (nitrous oxide, authors over painful specific alternative suprapubic stimu- methods of PR. as required. frequency 50 Hz, epidural, pethidine, areas (low back or pain intervention, lation at two dif- Women pulse train), until diazepam, pudendal supra-pubic) ran- all attended ferent frequencies; requested to paraesthesia to block, paracervical domly; if first method antenatal clinic TENS from time try TENS before painful areas. block). not effective or pre-delivery. of first contraction receiving other caused discomfort to parturition. pain inter- then other site used ventions. after 15–30 minutes.

Champagne, Primipara and Parallel group: Not described. Not described. Anesthelec High frequency, Three electrodes: Sham cranio TENS; Described as 2 et al., 198411 multipara requiring 1. Limoge cranio MPO2 166 kHz, 1.2 ms, two post (no current). double-blind, light analgesia during TENS (10); 2. cranial 20%; low mastoid, one hidden on both labour and Sham cranio TENS Limoge frequency, 83 Hz, between machines, indepen- delivery. (10). Began when TENS 4 ms, 33%. eyebrows. dent person set up analgesia machine. Patient and requested. observer blind.

Chia, et al., Surgical induction Partial crossover: TENS unit Could use burst Spembly Individually Two pairs, one Entonox. Open; no attempt 2 199010 or early labour, 1.TENS (10); 2. explained and mode if needed (Obstetric titrated, fixed either side of to blind treatments. primigravida who Entonox (10). demonstrated during painful Pulsar) dual pulse rate of midline 5 cm had not previously Study began when before painful contractions. channel 200 ms, apart; one pair experienced patients requested contractions, amplitude at T11,T10; 1 pair other forms of analgesia in early midwife taught 48 mA upper sacral analgesia in labour. stage of labour; patients vertebrae. complete once breathing second stage of technique for labour reached Entonox. or other form of analgesia requested.

Harrison, Primigravida and Parallel group: 1. Patients Titrate controls 3M Tenzcare, Individually Two pairs: one pair Sham TENS Described as 1 et al., 198612 third labour, who Parity 0,TENS assured that as contractions dual channel titrated, pulse T10–L1 derma- (no current). double-blind.TENS did not desire (49); 2. Parity 0, they could use increased, they width 60–80 µs, tome, 5 cm away machine with red specific alternative sham TENS (51); TENS until were not made repetition rate from either side of light, no current. pain intervention. 3. Parity 3 TENS confident; that aware of any 80–100/second. spine; one pair Neither patient or (27); 4. Parity 3 other forms of sensations at site posterior rami attending midwife sham TENS (23). analgesia would of electrodes. S2–L1. aware which treat- From admission be available; ments were allocated, to labour ward. nature of TENS third party changed explained. numbers in attempt to maintain blinding.

Lee, et al., Primigravida and Parallel group: Patients given Low-frequency Eastleigh Biphasic Two pairs silicone Sham TENS Neither patients 1 199013 second uncompli- 1.TENS (38); verbal explan- TENS was com- Obstetric rectangular wave, electrodes, 4 x (no current); nor obstetric staff cated labour; age 2. sham TENS (33); ation on ration- menced during device frequency range 12 cm;T10–L1 control (no knew which devices range 18–35 years. 3. No treatment ale for TENS. first stage of 2–200 Hz bursts/ and S2–4 spinal TENS device). were active or control (34). Additional anal- labour, high fre- second. Amplitude level. inactive. gesia given as quency TENS used individually necessary, but during contractions titrated by epidurals not and second stage. patients. encouraged. Patients not made aware of sensations at site of stimulation.

Nesheim, Expected birth Parallel group: 1. Aim was to try Two sets of Travisens, Individually Two pairs: both Sham TENS Red light, not blind 2 198114 following normal TENS (35); 2. sham TENS to instructions: 1. Dan-Sjo titrated and pairs T10–L1. (no current) to observer as labour and normal TENS (35). Began determine its Explanation of Elektronik decreased until patients given pregnancy; cervical when dilation effectiveness in active treatment to comfortable, different dilation < 4 cm. < 4 cm. labour pain, no patient and partner, pulse 0–40 mA, instructions. risks and alter- encouraged to use frequency 100 Hz, native methods when contraction pulse duration would be began, to use more 0.25 ms, frequency available if PR intensely as pain 50–150 Hz. inadequate. increased, free to stop at any time. 2. Sham – to expect no sensation other than warmth.

continued 201 TENS in labour pain

TABLE 59 contd Randomised studies of TENS in labour pain

Steptoe & Normal vaginal Parallel group: 1. Identical short Same to both Elpha 500 Individually Two pairs carbon Sham TENS + Red light, same 2 Bo, 198415 delivery, primi- TENS (13); 2. verbal and groups to titrate titrated over rubber:T10–T12, standard obstetric instruction to both gravida, > 3 cm sham TENS (12). written up to level of 30 minutes, S2–S3. analgesia. groups. dilation. Stimulation for information to comfort or 0–60 mA, pulse 30 minutes. both groups. adequate PR width 0.2 ms, during first frequency 1–4 Hz, 30 minutes. 100 Hz.

Thomas, Early labour in Parallel group: 1. No ante-partum Both groups 3M, dual Individually 2 pairs electrodes: Sham TENS Good attempt to 4 et al., 198816 primigravida and TENS (132); 2. instruction advised to increase channel titrated. one pair para- (no current). blind study, both multigravida, sham TENS (148). given on TENS. TENS settings as vertebrally on active and sham normal or induced TENS applied Standard needed during either side of machines had delivery, < 7 cm when discomfort protocol given contractions; spinous process, flashing light,TENS cervical dilation. reported. to both groups patients free to T10–L1, one pair, applied by staff not by instructor use other analgesia S2–4. associated with who only if required.TENS trial, labour advised on turned off for two managed by non- TENS use. contractions each study staff in hour and differ- normal way. ences in pain Instructions not assessed. given by assessor.

van der First stage of Parallel group: 1. Use of TENS Low-frequency Agar GK Individually 2 pairs electrodes, Sham TENS; No description of 2 Ploeg, et al., labour, primiparae TENS (46); 2. explained to TENS used until (Klinerva titrated. 50 x 100 mm. (identical placebo sham TENS other 199617 multiparae, when Sham TENS (48). expectant contractions when Holland) 1 cm lateral spine, device). than identical analgesia TENS used until parents by high-frequency L1–L3 and L4–S1. device. requested. cervix fully dilated. attending TENS used physician. (range 1–6). Patients super- Titrated by partner vised until com- (mostly). PCA petent in the pethidine and use of TENS. promethazine escape analgesia.

Wattrisse, Primigravida, Parallel group: 1. Not described. Not described. Anesthelec High frequency, 3 electrodes: 2 Epidural 0.25% None. 2 et al., 19938 gestation of at epidural (60); 2. MPO2 166 kHz, 1 ms, posterior mastoid, bupivacaine as least 38 weeks, epidural + TENS 20%; low 1 between required, first bolus < 3 cm cervical (60).TENS applied frequency, 83 Hz, eyebrows. with fentanyl 100 µg. dilation, expected with epidural to 4 ms, 33%. normal delivery end of labour. with extradural analgesia.

in which TENS was seen to reduce additional measure but one commonly used in these studies. analgesic interventions11,15 had only 20 and The implications of these results for current 25 patients in the comparisons compared with practice is that women who are offered TENS 527 patients in the comparisons in the other three are at risk of having their pain inadequately trials (Figure 47). In only one study of moderate controlled and may experience delays in size8 had a secondary outcome been judged by us receiving effective interventions. to be positive for cranial TENS. This emphasises how individual small studies may mislead because This review was restricted to RCTs, unlike that by of the random play of chance. Reeve and colleagues.2 RCTs represent the gold standard in clinical trials of efficacy.19 For nearly The choice of outcome measure is an important 20 years, non-randomised studies have been known determinant of how studies are to be judged. If to yield larger estimates of treatment effects than the objective of TENS is to alleviate pain, then it studies using random allocation.20 The size of the is fair that judgements of its effectiveness are overestimation of treatment effect when random- based on prospective subjective measures of pain isation is inappropriate can be as much as 40%. In intensity or relief (primary outcomes), and that postoperative pain, non-randomised trials of TENS these assessments are done at appropriate time were more likely to show a positive result than points. Retrospective measures of pain are randomised trials, with 15 from 17 randomised notoriously unreliable. Subsequent need for other trials being negative and 17 from 19 non- 202 analgesic interventions is a secondary outcome randomised trials positive. Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 60 Results of randomised studies of TENS in labour pain

Study Pain outcomes (primary outcomes Results for pain outcomes Withdrawals and Adverse effects Significant difference Overall in bold; secondary outcomes drop-outs for at least one judgement in italics) primary/secondary outcome

Bunsden, et al., 1. 5-point PI (hourly) Back pain severe: ≤ 5 cm dilated, 1 in each group excluded due No specific effect No/no Negative 19829 2. Use of any other pain-relieving TENS 3/15, control 5/9; > 5 cm to subsequent Caesarean of TENS on foetal interventions dilated,TENS 1/7, control 5/6. section. 1 in each group heart rate. 3. Duration of labour Suprapubic pain severe: ≤ 5 cm received epidural because of 4. Questionnaire on day after delivery; dilated,TENS 10/15, control 7/9; special problems which were abdominal and back pain assessed > 5 cm dilated,TENS 7/8, control 5/6. not described. independently. Stage 2: pudendal block 13/15 TENS, 7/9 control analgesic.

Champagne, 1. Additional pain relieving interventions. 5/10 required additional analgesic None reported. Not described. N/A /yes Positive et al., 198411 intervention active stimulation, 10/10 secondary control group.

Chia, et al., 199010 1. 3-point PI pre-escape analgesia NSD. No relief 11% TENS,50% 1 woman in Entonox group Not described. No/no Negative 2. 3-point PR Entonox, but contractions significantly delivered without further 3.Time to next analgesic. higher in Entonox group. Additional analgesia, and was excluded. analgesia not described.

Harrison, et al., Research midwife assessments: NSD between TENS and sham TENS Not described. Not described. No/no Negative 198612 1. 5-point PI (hourly) for pain or for those requiring extra 2. Baseline pain threshold (Mosanto gun) analgesia (12% TENS,14% sham TENS). 3. 4-point PR Pain score > 50% at 1 hour:TENS 4. Site of pain. 63/64, sham TENS 55/59. Additional analgesia needed: 57/76 TENS, 58/74 controls.

Lee, et al., 199013 Every 30 minutes: NSD between TENS and sham TENS. Not described. None reported. No/no Negative 1. VAS PI 0–10 every 30 minutes Use of additional analgesic 2. Strength of uterine contractions interventions (excluding epidurals): (weak, moderate, strong). 40/62 TENS, 22/35 sham TENS, 28/37 Retrospective questionnaire at 24 hours control. NSD between three postpartum (1. did patient find TENS treatments for 30-minute pain scores. helpful or not? (0–3); 2. future use?).

Nesheim, 198114 1. Overall 5-point PR after childbirth NSD between TENS and sham TENS. Not described. Not described. N/A /no Negative 2. Escape analgesia. Pain free: 1/35 TENS,0/35 sham TENS. Good relief: 4/35 TENS 5/35 sham TENS. 63 analgesic interventions in TENS, 49 in controls.

Steptoe & Bo, 1. VAS PI 0–10 at baseline and No difference in pain measurements. 1/13 excluded in TENS group 0/12 TENS;0/13 No/yes Positive 198415 30 minutes after TENS Additional analgesia: 5/12 TENS, due to failed battery in sham TENS. secondary 2. Other analgesic interventions 13/13 control group. device; 1/13 in control group 3.Time of contractions. had Caesarean section but included in analysis.

Thomas, et al., 1.VAS PI (hourly) for abdominal and NSD between TENS and sham TENS 52/148 control group Not described. No/no Negative 198816 back pain at < 7 cm dilated, 7–10 cm dilated, or requested to withdraw 2. Use of other analgesic interventions during stage 2. No difference in use compared with 54/132 in 3. Postpartum overall assessment by patient. of other methods of PR TENS group. Only 96 patients (Entonox, pethidine or epidural). could continue with VAS Postpartum assessment of excellent/ beyond 7 cm and 16 into good relief by 29/132 active group. second stage of labour.

van der Ploeg, 1. VAS PI (bad = no; good = yes) NSD between TENS and sham TENS. 2 refused to take part and None reported. No/no Negative et al., 199617 2. Other analgesic interventions Mean number of requests for other received standard analgesia. 3. Patient’s impression of medication analgesia: 18.2 TENS,26.2 sham TENS. No other details given. during labour Number of times analgesia 4.Would patient choose TENS in future administered: 5.9 ± 2.32 TENS, deliveries? 6.5 ± 1.77 sham TENS.Amount of 5. Patient’s impression of effect of TENS pethidine administered (mg): on pain. 60.8 ± 21.6 TENS.

Wattrisse, et al., 1. VAS for global pain quality (patient) Quality of analgesia during dilatation 7 withdrawals, 1 Caesarean Not described. N/A /yes Positive 19938 during labour at 2 hours post delivery and at delivery not different between section, 2 technical problem secondary 2. Duration of analgesia after first active and control groups. Duration with epidural, 2 electrodes epidural bolus of first epidural local anaesthetic bolus fell off, 1 had other treatment. 3.Time between epidural boli. increased in TENS group by mean of NB: not all described. 22 minutes (p < 0.01).Time between boli significantly prolonged in TENS. 203 TENS in labour pain

Patients having additional analgesics with active TENS (%) 100 Champagne, et al., 280 1984 (10)11

150 Steptoe & Bo, 75 1984 (13)15 97 Harrison, et al., 1986 (76)12 50 20 25 Thomas, et al., 1988 (132)16 25 Lee, et al., 1990 (97)13

Combined 0 0255075 100 Patients having additional analgesics 0.1 1 10 with sham TENS (%) RR

FIGURE 47 Additional analgesic requirements (numbers of FIGURE 48 Relative risk for additional analgesic intervention patients in each trial is given next to the circle, whose diameter is proportional to the number) The overall methodological quality of the trials References reported was low, reflecting the fact that it is diffi- 1. Jadad AR, McQuay HJ. A high-yield strategy to cult, if not impossible, to blind studies of TENS.21 identify randomized controlled trials for systematic Inadequate blinding may be an important source of reviews. Online J Curr Clin Trials [serial online] observer bias and may contribute to overestimation 1993 Feb 27;Doc No 33:3973 words; 39 paragraphs; of treatment effects. Four of the studies considered 5 tables. here made no attempt at blinding and, of the seven 2. Reeve J, Menon D, Corabian P. Transcutaneous that used sham TENS, only one16 described the electrical nerve stimulation (TENS): a technology method of blinding in sufficient detail to indicate assessment. Int J Technol Assess Health Care that blinding may have been adequate. 1996;12:299–324.

The 1994 Maternity Service Charter22 tells 3. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in clinical research. Ann Intern Med 1987;107:224–33. women that “...you have the right to be given an explanation of any treatment proposed, including 4. Jadad AR, Moore RA, Carroll D, Jenkinson C, the benefits and risks and of any alternatives before Reynolds DJM, Gavaghan DJ, et al. Assessing the you decide whether you will agree with the treat- quality of reports of randomized clinical trials: is ment”. Those involved in the provision of maternity blinding necessary? Control Clin Trials 1996;17:1–12. services therefore need to be aware of current 5. DerSimonian R, Laird N. Meta-analysis of clinical research findings concerning effective inter- trials. Control Clin Trials 1986;7:177–88. ventions for the relief of pain, so they can apply 6. Cook RJ, Sackett DL. The number needed to these findings in their clinical practice and provide treat: a clinically useful measure of treatment women with accurate information so they can be effect. BMJ 1995;310:452–4. involved in decisions concerning their care. 7. Carroll D, Tramèr M, McQuay H, Nye B, Moore A. Transcutaneous electrical nerve stimulation in On the basis of these findings the continued use labour pain: a systematic review. Br J Obstet Gynaecol of TENS in childbirth needs to be carefully recon- 1997;104:169–75. sidered. The continued use of TENS in labour pain 8. Wattrisse G, Leroy B, Dufossez F, Bui Huu Tai R. has considerable implications both for maternity Electrostimulation cerebrale transcutanee: étude services and the women who use TENS, in terms of comparative des effets de son association a receiving prompt and effective pain relief during l’anesthesie peridurale par bupivacaine-fentanyl au childbirth. Instead of TENS, women should be cours de l’analgesie obstetricale. Cah Anesthesiol 204 given the option of more effective interventions. 1993;41:489–95. Health Technology Assessment 1998; Vol. 2: No. 12

9. Bundsen P, Ericson K, Peterson LE, Thiringer K. 16. Thomas IL, Tyle V, Webster J, Neilson A. An Pain relief in labor by transcutaneous electrical evaluation of transcutaneous electrical nerve nerve stimulation. Testing of a modified stimulation stimulation for pain relief in labour. Aust N Z J technique and evaluation of the neurological and Obstet Gynaecol 1988;28:182–9. biochemical condition of the newborn infant. Acta Obstet Gynecol Scand 1982;61:129–36. 17. van der Ploeg JM, Vervest HAM, Liem AL, van Leeuwen S. Transcutaneous nerve stimulation 10. Chia YT, Arulkumaran S, Chua S, Ratnam SS. (TENS) during the first stage of labour: a Effectiveness of transcutaneous electric nerve randomized clinical trial. Pain 1996;68:75–8. stimulator for pain relief in labour. Asia Oceania J Obstet Gynaecol 1990;16:145–51. 18. Carroll D, Moore RA, Tramèr MR, McQuay HJ. 11. Champagne C, Papiernik E, Thierry JP, Noviant Y. Transcutaneous electrical nerve stimulation does Electrostimulation cerebrale transcutanee par les not relieve labour pain – updated systematic review. courants de Limoge au cours de l’accouchement. Curr Rev Obstet Gynaecol 1997;(Sept):195–205. Ann Fr Anesth Reanim 1984;3:405–13. 19. Schulz KF, Chalmers I, Hayes RJ, Altman DG. 12. Harrison RF, Woods T, Shore M, Mathews G, Empirical evidence of bias: dimensions of Unwin A. Pain relief in labour using transcutaneous methodological quality associated with estimates electrical nerve stimulation (TENS). A TENS/TENS of treatment effects in controlled trials. JAMA placebo controlled study in two parity groups. Br J 1995;273:408–12. Obstet Gynaecol 1986;93:739–46. 13. Lee EWC, Chung IWY, Lee JYL, Lam PWY, Chin 20. Chalmers TC, Matta RJ, Smith J Jr, Kunzler AM. RKH. The role of transcutaneous electrical nerve Evidence favoring the use of anticoagulants in the stimulation in management of labour in obstetric hospital phase of acute myocardial infarction. patients. Asia Oceania J Obstet Gynaecol N Engl J Med 1977;297:1091–6. 1990;16:247–54. 21. Deyo RA, Walsh NE, Schoenfeld LS, Ramamurthy S. 14. Nesheim BI. The use of transcutaneous nerve Can trials of physical treatments be blinded? The stimulation for pain relief during labor. A con- example of transcutaneous electrical nerve trolled clinical study. Acta Obstet Gynecol Scand stimulation for chronic pain. Am J Phys Med 1981;60:13–16. Rehabil 1990;69:6–10. 15. Steptoe P, Bo JO. Transkutan nervestimulations smertelindrende effekt ved fødsler. Ugeskr Laeger 22. The patient’s charter: maternity services. London: 1984;146:3186–9. Department of Health, 1994.

205

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 21 Efficacy and harm of anti-emetic interventions in the surgical setting

he project described in this chapter had There is a validity issue which has been addressed T the aim of establishing a league table of the by limiting the range of included trials to those relative prophylactic efficacy and the likelihood of which duplicate usual clinical experience. Two harm from anti-emetic interventions which are narrow bands of CER were therefore defined: currently used to treat or prevent PONV. 20–60% CER for early outcomes (0–6 hours after surgery) and 40–80% CER for late outcomes (0–48 hours). Only trials with CERs within these bands Data source were analysed. Validity criteria defined lower and upper boundaries. Trials with very low CERs do not Systematic reviews of the literature, using data allow a valid assay of anti-emetic efficacy; PONV extraction from relevant reports, critical appraisal cannot be prevented if nobody is going to vomit of data and meta-analytical combination of data with without prophylaxis. Trials with very high CERs do biostatistical methods, were chosen as the main not represent daily clinical practice but enable even instrument for this study. Thus the ‘essence’ of this marginally active interventions to show statistically work comes from ten different systematic reviews, of significant efficacy. Trials with very low or very high which eight were designed to investigate the efficacy CERs were, therefore, regarded as invalid. The and potential for harm of anti-emetic interventions. arbitrary limits of the CER banding were applied At least 860 reports were screened, from which to all analysed interventions. 215 RCTs with data from 31,801 patients were analysed. The two remaining systematic reviews were The league table of anti-emetic efficacy designed to investigate harm from interventions. Five prophylactic interventions were analysed For this purpose, after screening about 1200 reports, within the CER banding, each with data from 300 reports of different study architectures with systematically searched RCTs: propofol induction data from 1,432,817 patients were analysed. (reworking of data from Tramèr and colleagues2), propofol maintenance,1 omitting nitrous oxide,3 a total intravenous anaesthetic with propofol,1 and Efficacy of anti-emetic ondansetron.4 Propofol was compared with non- interventions propofol anaesthetics, which were regarded as ‘no treatment’ controls. Omitting nitrous oxide was A model was proposed to compare anti-emetic compared with using it (‘no treatment’ control). interventions indirectly, that is, without the need In ondansetron trials, comparators were placebos for direct comparisons,1 which is formally equi- or no treatment. Results are presented in valent to that used to generate the league table for Figures 49–52. analgesic interventions. There is one difference, however. When placebo responses for analgesics Interpreting the league table of efficacy vary, they tend to vary around a population placebo Four criteria need to be taken into account. event rate of about 19%, that is, about 19% of patients with moderate or severe acute pain given Endpoint a placebo will experience at least 50% pain relief. The endpoint is prevention of nausea or vomiting, For PONV, the spread is much greater, with per- within 6 hours of surgery (early efficacy) or within centages of patients vomiting without prophylactic 48 hours of surgery (late efficacy). Thus for each interventions in studies ranging from close to zero intervention, anti-nausea and anti-vomiting efficacy to over 80%, and there seems to be no central can be interpreted separately, as well as short-term figure or population response. This is the case and long-term efficacy. even when operation, anaesthetic and patient population are highly standardised, as in the case Short-term efficacy has an economic impact mainly of paediatric strabismus surgery. Clearly PONV is in day-case surgery where patients are meant to be different from postoperative pain in this respect. discharged within hours of the procedure; they 207 Efficacy and harm of anti-emetic interventions in the surgical setting

Ondansetron, 8 mg i.v.

Omitting nitrous oxide

Total i.v. anaesthetic

Propofol induction

Propofol maintenance

Ondansetron, 4 mg i.v.

12345102030 Log number NNT (95% CI)

FIGURE 49 Prevention of early nausea (0–6 hours): CER 20–60%

Propofol induction

Ondansetron, 1 mg i.v.

Ondansetron, 8 mg i.v.

Ondansetron, 4 mg i.v.

Propofol maintenance

Omitting nitrous oxide

Total i.v. anaesthetic 12345102030 Log number NNT (95% CI)

FIGURE 50 Prevention of early vomiting (0–6 hours): CER 20–60%

have to be free of PONV in order to fulfil 20% of treated patients will profit from the discharge criteria. Long-term efficacy is a better prophylaxis. The lower boundary of the CER indicator of the drug’s anti-emetic efficacy and banding for early outcomes was set at 20%.1 Thus, patients’ comfort. It indicates if the patient interventions which had no scope to show an NNT will remain PONV-free at home (or on the of at least 5 for efficacy were not considered in journey home). the model.

NNT point estimate Confidence interval Worthwhile anti-emetic efficacy in the surgical The upper boundary of the CI for the NNT setting was arbitrarily defined as an NNT to prevent places the treatment in the least favourable light. nausea or vomiting compared with placebo (or If this upper limit lies within what would be 208 no treatment) of ≤ 5.2 This means that at least considered to be the minimal clinically relevant Health Technology Assessment 1998; Vol. 2: No. 12

Total i.v. anaesthetic No data

Omitting nitrous oxide No effect (NNT ∞)

Propofol induction No effect (NNT 35)

Ondansetron, 1 mg i.v.

Ondansetron, 4 mg i.v.

Ondansetron, 8 mg i.v.

Propofol maintenance

12345102030 Log number NNT (95% CI)

FIGURE 51 Prevention of early nausea (0–6 hours): CER 20–60%

Ondansetron, 1 mg i.v.

Propofol induction

Propofol maintenance

Ondansetron, 4 mg i.v.

Omitting nitrous oxide

Total i.v. anaesthetic

Ondansetron, 8 mg i.v. 12345102030 Log number NNT (95% CI)

FIGURE 52 Prevention of early vomiting (0–6 hours): CER 20–60% effect (for instance, an NNT of 5 to prevent Results PONV),2 the result indicates a definitely useful The best anti-emetic prophylaxis has the treatment. Lack of overlap between CIs can be lowest NNT (but not above 5), the largest used as a simple test of the statistical difference symbol, the narrowest CI (the upper limit between NNTs of two interventions. below 5), and the most consistent efficacy (that is, anti- both nausea and vomiting, and both Size (area) of the symbol short- and long-term). No anti-emetic intervention Areas of symbols were plotted that were which has been tested meets all these criteria. proportional to the number of analysed patients. Intravenous ondansetron, 8 mg, comes closest. The larger the number of analysed patients (that is, Propofol maintenance looks promising but the the larger the symbol), the greater the confidence long-term effect is based on a limited number in the point estimate. of patients only; the symbol is small and the 209 Efficacy and harm of anti-emetic interventions in the surgical setting

CI large. Omitting nitrous oxide looks promising Interpreting the league table of harm too but only for anti-vomiting efficacy. Endpoint A specific endpoint and its risk have to be An intervention which indicates good efficacy interpreted in their proper context. For instance, (i.e. NNT < 5) but which is based on a limited elevated liver enzymes with ondansetron might be number of patients and, therefore, with a small perceived as trivial. The biological basis for this symbol and/or a wide CI, may lead to a research adverse drug reaction, however, is not known. agenda. The question then is: ‘Are further trials There may be an argument for not giving needed?’ and, if the answer is affirmative, ‘What ondansetron to patients with pre-existing abnormal trials are needed?’ liver function tests or underlying liver disease. It is unknown if ondansetron should be avoided when other potentially hepatotoxic drugs, such as Potential for harm of anti-emetic halothane or paracetamol, are to be used. interventions NNH point estimate Interpretation of any intervention’s clinical The NNH indicates in how many patients the usefulness must take into account both efficacy and adverse drug reaction will occur which would harm. An objective measurement of severity does not have happened had the patient not received not exist; acceptability and interpretation of harm the drug. Interpretation of this result has to take are likely to be multifactorial. Unlike the efficacy into account other factors, such as severity of the league table, the league table on intervention- reaction, severity of the underlying disease, and related harm is not based on truncated data sets. availability of alternative treatments and their All comparisons are between the active inter- potential for harm. ventions and placebo or no treatment or in, the case of propofol, between active (propofol) and Statistical significance another, propofol-free anaesthetic (control). In contrast to the league table of anti-emetic efficacy, symbol sizes are not plotted proportional The league table of harm to the quantity of analysed data (i.e. the symbol The league table of harm presented in Figure 53 area is fixed) and no CIs are shown. The reason is is based on analyses of four systematic reviews: that some NNHs are based on a limited number of droperidol,5 omitting nitrous oxide,3 propofol,6 patients who had the adverse drug reaction. Yet and ondansetron.4 these NNHs may be clinically relevant. Plotting

Propofol: asystole during surgery 660

Droperidol, 75 µg/kg: extrapyramidal reactions in children 123

Omitting nitrous oxide: intraoperative awareness 46

Ondansetron: headache 36

Ondansetron: elevated liver enzymes 31

Ondansetron: constipation 23

Propofol: bradycardia at induction/during surgery 11.3

Droperidol, 75 µg/kg, in children: drowsiness, restlessness 5.3

Propofol: oculo-cardiac reflex in children (with prophylactic anticholingerics) 4.1 1 10 100 1000 Log NNH

210 FIGURE 53 Harm with anti-emetic interventions. All data are from RCTs. Significance = statistical significance, assumed when p < 0.05 Health Technology Assessment 1998; Vol. 2: No. 12

symbol size in relation to the quantity of analysed able to wait and see if a patient is going to vomit or data would detract from such potentially important become nauseated before starting treatment. Also results and tend to overinterpret more trivial (but there is a widespread belief that it may be more better documented) reactions. difficult to treat established PONV than to prevent it,8 although there is no substantial evidence to A graphical distinction was made between support this view. Using a decision-analysis treat- significant and non-significant findings. Statistical ment model, it has been suggested that prophy- significance was arbitrarily set at a value of p < 0.05. lactic anti-emetic therapy was more cost-effective This value may be unnecessarily conservative and compared with treatment of established symptoms narrower CIs could be chosen. Black symbols for operations associated with a high incidence represent adverse drug reactions which happened of emesis.9 statistically significantly (p < 0.05) more often with the intervention. Accordingly, white symbols indi- The pivotal evidence to resolve the debate was the cate absence of statistical significance (p > 0.05). relative effectiveness of treatment and prophylaxis of PONV. This comparison was possible with Conclusions ondansetron. The finding of a dose–response Each intervention introduces a certain risk relationship with ondansetron in prevention of of adverse drug reactions. The league table of PONV4 contrasted with the analysis of the efficacy intervention-related harm is an important contri- of ondansetron in the treatment of established butor to a rational risk–benefit assessment. It will PONV.10 For this, no dose response between 1 mg help both physicians and patients to take decisions and 8 mg could be established; the NNT to prevent about the use (prophylactically or therapeutically) further PONV in a nauseated or vomiting patient of anti-emetic interventions. with the lowest dose tested, 1 mg, compared with placebo was between 4 and 5, and higher doses were no more effective. Thus, 1 mg is as efficacious Prophylaxis versus treatment for treatment of established PONV as an eight-fold higher dose (i.e. 8 mg) is for the prevention of One of the main conclusions of the first systematic PONV.4 This challenges the usefulness of pro- review of efficacy and harm of anti-emetic inter- phylactic ondansetron when risk–benefit and ventions in the surgical setting5 was that it would cost–benefit arguments are considered. perhaps be better to wait and see which patient vomits and then treat. This conclusion was based Two scenarios on the somewhat unexpected result that even in Two scenarios may be described that illustrate the paediatric strabismus surgery, a clinical setting with relationship between prophylaxis and treatment a high risk for PONV, only 25% of the children based on these numbers. For simplicity, several actually profited from the best prophylaxis. The assumptions have to be made: subsequently calculated NNTs of all the other prophylactic interventions (propofol induction and (i) one ondansetron, 2 mg, ampoule is set at £6.75 maintenance, omitting nitrous oxide, ondansetron) and a 4 mg ampoule at £13.5011 did not prove to be more efficacious. (ii) one ondansetron, 2 mg, ampoule (the smallest commercially available) per patient will be The justification of prophylactic postoperative used to treat established PONV; two 4 mg antiemetics was queried 35 years ago by Adriani ampoules per patient will be used for and colleagues.7 They noted that no more than prophylaxis 25% of patients in the recovery room vomited in (iii) ondansetron’s prophylactic anti-emetic the immediate postanaesthesia period, that most efficacy is independent of CER, although of this vomiting was short-lived and subsided spon- its prophylactic efficacy is likely to be lower taneously without the use of antiemetics. Similar (i.e. higher NNT) with low CERs average PONV incidence has subsequently been (iv) the success rate with ondansetron in patients reported repeatedly in large RCTs, in case series with established PONV is 40%,10 which means and in systematic reviews of RCTs, although it that 40% of all vomiting or nauseated patients may be higher in specific clinical settings, such receiving ondansetron will not vomit or be as paediatric strabismus surgery. If the incidence nauseated further without prophylaxis is only 25% and treatment (v) NNTs for the optimal ondansetron doses are is effective, then arguably prophylaxis may be set at 5 for both prevention and treatment unnecessary on grounds of adverse effects and cost. (instead of 5–6 for prevention and 4–5 The humanitarian argument is that it is unaccept- for treatment) 211 Efficacy and harm of anti-emetic interventions in the surgical setting

(vi) NNH is set at 30 for both prevention and Cost is maintained at £27,000 in the prophylaxis treatment, and there is no dose response group but decreases to £2025 (300 x £6.75) in the (i.e. the risk of an adverse drug reaction is treatment group, a benefit of £24,925, or almost similar for both 2 mg and 8 mg); the adverse £25 per patient. In the prophylaxis group, 100 drug reaction could be a headache or patients are failures, compared with 180 in the elevated liver enzymes. treatment group who continue to be nauseated or to vomit. More than three times as many patients Any calculation based on these assumptions will in the prophylaxis group have an adverse drug yield an overoptimistic result for prevention and reaction compared with patients who are treated a conservative result for treatment of PONV for established PONV. with ondansetron. These risk–benefit–cost calculations are the strongest In the first scenario, for 1000 ‘high risk’ patients argument against prophylactic use, and in favour of undergoing surgery the CER is 60% (see Box 3). therapeutic use of anti-emetic interventions in the surgical setting. The main arguments are that costs In this situation, prophylaxis costs (1000 x (2 x with treatment are dramatically lower and about half £13.50)) are £27,000, whereas treatment costs as many patients will have an adverse drug reaction. (600 x £6.75) are £4050. The benefit is £22,950, or The number of failures is comparable in the high-risk £22.95 per patient. In the prophylaxis group, 400 scenario but in favour of prophylaxis in the low-risk patients are failures (i.e. they are nauseated or they scenario. The efficacy of ondansetron or other anti- vomit despite prophylaxis) compared with 360 in the emetics in patients who already have received ondan- treatment group (i.e. they continue to be nauseated setron and in whom the prophylaxis or treatment or to vomit because the treatment failed). Almost failed is unknown. The ‘price’ patients in the treat- twice as many patients receiving the drug prophylac- ment group have to pay, is that they need to vomit or tically will have an adverse drug reaction. feel nauseated before they receive treatment.

In the second scenario, a ‘low risk’ situation, The cost-effectiveness arguments for prophylaxis 1000 patients undergo surgery and the CER is versus treatment are investigated more fully in the 30% (see Box 3). next chapter.

BOX 3 Risk–benefit–cost analysis of anti-emetic prophylaxis versus treatment in a clinical setting with high and low CERs High CER 1000 patients undergoing surgery: CER = 60% Prevention in 1000 patients: Treatment of 600 patients with one episode of PONV: ondansetron, 8 mg i.v. (NNT ± 5) ondansetron, 1 mg i.v. (NNT ± 5) 1000 x 8000 mg x 13 600 x 1 mg = 600 mg 400 no PONV anyway + 400 no PONV anyway + 200 successful preventions 240 successful treatments 400 failures 360 failures Adverse drug reaction in 33 Adverse drug reaction in 20 (NNH 30) (NNH 30) Low CER 1000 patients undergoing surgery: CER = 30% Prevention in 1000 patients: Treatment of 300 patients with one episode of PONV: ondansetron, 8 mg i.v. (NNT ± 5) ondansetron, 1 mg i.v. (NNT ± 5) 1000 x 8 mg = 8000 mg x 27 300 x 1 mg = 300 mg 700 no PONV anyway + 200 successful preventions 700 no PONV anyway + 120 successful treatments 100 failures 180 failures Adverse drug reaction in 33 Adverse drug reaction in 10 (NNH 30) (NNH 30) 212 Health Technology Assessment 1998; Vol. 2: No. 12

Research agenda 3. Tramèr M, Moore A, McQuay H. Omitting nitrous oxide in general anaesthesia: meta-analysis of Further systematic reviews on the efficacy and harm intraoperative awareness and postoperative emesis in randomized controlled trials. Br J Anaesth of the other anti-emetic drugs which are currently 1996;76:186–93. used in anaesthesia and surgery are needed. Drugs of interest include droperidol, metoclopramide 4. Tramèr MR, Reynolds DJM, Moore RA, McQuay HJ. Efficacy, dose-response, and safety of ondanse- and hyoscine (scopoderm). New 5-HT3 receptor antagonists are increasingly used (granisetron, tron in prevention of postoperative nausea and tropisetron, dolasetron). Dose–response relation- vomiting: a quantitative systematic review of ships, optimal doses and adverse effect profiles of randomised placebo-controlled trials. Anesthesiology 1997;87:1277–89. these drugs have to be established. Opioid-induced nausea and vomiting, for instance, related to 5. Tramèr M, Moore A, McQuay H. Prevention of PCA or spinal opioid administration, remains a vomiting after paediatric strabismus surgery: a particular problem. The proposed banding model systematic review using the numbers-needed-to-treat may be used for indirect comparisons of relative method. Br J Anaesth 1995;75:556–61. anti-emetic efficacy. The risk–benefit–cost model 6. Tramèr MR, Moore RA, McQuay HJ. Propofol and could be used for further process analysis as new bradycardia: causation, frequency and severity. Br J data on prevention and treatment with other drugs Anaesth 1997;78:642–51. becomes available. Systematic reviews are likely to 7. Adriani J, Summers FW, Antony SO. Is the inform our future clinical management. However, prophylactic use of antiemetics in surgical patients such projects are time-consuming, expensive, and justified? JAMA 1961;175:666–71. dependent on efficacious team-work. Interest groups will need to be formed. 8. Zahl K. The role of ondansetron and other antiemetics in ambulatory surgery. J Clin Anesth 1993;5 suppl 1:52–56S. References 9. Watcha M, Smith I. Cost-effectiveness analysis of antiemetic therapy for ambulatory surgery. J Clin 1. Tramèr M, Moore A, McQuay H. Meta-analytic Anesth 1994;6:370–7. comparison of prophylactic antiemetic efficacy for postoperative nausea and vomiting: propofol 10. Tramèr MR, Moore RA, Reynolds DJM, McQuay HJ. anaesthesia vs omitting nitrous oxide vs a total i.v. A quantitative systematic review of ondansetron in anaesthesia with propofol. Br J Anaesth treatment of established postoperative nausea and 1997;78:256–9. vomiting. BMJ 1997;314:1088–92. 2. Tramèr M, Moore A, McQuay H. Propofol 11. Drugs used in nausea and vertigo (chapter 4.6). anaesthesia and postoperative nausea and vomiting: British National Formulary, no. 32, 1996. London: quantitative systematic review of randomized British Medical Association and Royal controlled studies. Br J Anaesth 1997;78:247–55. Pharmaceutical Society of Great Britain, 1996.

213

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 22 Cost-effectiveness of ondansetron: prophylaxis compared with treatment in PONV

Summary because of the high success rate with the lowest dose tested (1 mg) in established PONV and PONV is a frequent complication of anaesthesia the disappointing anti-nausea effect of prophy- and surgery. There is no general agreement as lactic ondansetron even at an eight-fold to whether prophylaxis of PONV is better than higher dose. treatment of established PONV. This lack of agreement may be due to the great variation in the proportion of patients who experience PONV. Introduction Given this equipoise, choice of prophylaxis or treatment could be determined by substantial The aim was to assess the relative cost-effectiveness cost-effectiveness difference between the of strategies for treating PONV with ondansetron. two approaches. The intention was to establish cost-effectiveness relationships based on the strongest evidence Cost-effectiveness was calculated for ondansetron currently available. Data from two quantitative as prophylaxis or as treatment of PONV using systematic reviews of published valid RCTs of information from systematic reviews and published ondansetron in the surgical setting were used.1,2 meta-analyses. Modelling, which was based on a cohort of 1000 patients, examined CERs (vomiting in the absence of prophylaxis) between Methods 10% and 90%, and different doses. Anti-emetic efficacy was assumed to be constant across CERs. Decision tree, endpoints, and estimates In a sensitivity analysis, cost-effectiveness of recom- of efficacy mended doses (4 mg for both treatment and The two strategies to deal with PONV, treatment prophylaxis) was compared with cost-effectiveness versus prophylaxis, were displayed graphically as a of the most effective doses as demonstrated by decision tree (Figure 54). meta-analysis (1 mg for treatment and 8 mg for prophylaxis). Treatment arm Some patients will have no PONV symptoms at any For all CERs fewer patients will suffer any PONV time (T1) (see Figure 54). Success with treatment symptom (nausea and or vomiting/retching) at (T2) was defined as a nauseated or vomiting any time after surgery with prophylaxis compared patient who had no further episode of nausea or with treatment. However, with both effective vomiting after one dose of ondansetron. These treatment doses, 1 mg and 4 mg, fewer milligrams patients vomited or felt nauseated once before are required for each patient who suffers at most they received ondansetron. A treatment failure one episode of PONV, compared with 4 mg or (T3) was a vomiting or nauseated patient who 8 mg for prophylaxis. For the endpoint of continued to vomit or to feel nauseated despite maintaining a patient PONV-free throughout, treatment with ondansetron. treatment with 1 mg was still the most cost- effective, followed by treatment with 4 mg Anti-emetic efficacy of 1 mg of ondansetron for but only at a CER below 80%. treatment of established PONV was shown not to be different from 4 mg or 8 mg.1 However, the Fewer patients will experience any PONV optimal dose recommended by the manufacturer symptoms with prophylaxis compared with was 4 mg. For the purpose of this study, therefore, treatment. But prophylaxis was not much more both 1 mg and 4 mg doses were chosen for effective than treatment, and treatment of sensitivity cost-effectiveness analyses of treatment. established PONV with effective doses (i.e. 1 mg Success rate was set (Table 61) at 40% for the or 4 mg) is more cost-effective than prophylaxis 1 mg dose and at 45% for the 4 mg dose.1 This with effective doses (i.e. 4 mg or 8 mg). This is meant that 40% and 45% of nauseated or 215 Cost-effectiveness of ondansetron: prophylaxis compared with treatment in PONV

No PONV symptoms at any time T1 No PONV Successful treatment T2 PONV T Failure of treatment T3 PONV strategy No PONV symptoms anyway P1 Successful prophylaxis P P2

Failure PONV P3

FIGURE 54 Strategies for dealing with PONV (T, treatment; P, prophylaxis)

TABLE 61 Ondansetron: total estimates of efficacy and harm

Definition Value Reference

Success rate with 1 mg treatment 40% Tramèr, et al., 19971

Success rate with 4 mg treatment 45% Tramèr, et al., 19971

Success rate with 8 mg treatment 44% Tramèr, et al., 19971

NNT of 1 mg to prevent nausea 21 Tramèr, et al., 19972

NNT of 1 mg to prevent vomiting 15 Tramèr, et al., 19972

NNT of 4 mg to prevent nausea 16 Tramèr, et al., 19972

NNT of 4 mg to prevent vomiting 6.4 Tramèr, et al., 19972

NNT of 8 mg to prevent nausea 6.4 Tramèr, et al., 19972

NNT of 8 mg to prevent vomiting 5 Tramèr, et al., 19972

NNH for headache with any dose 36 Tramèr, et al., 19972

NNH for elevated liver enzymes with any dose 31 Tramèr, et al., 19972

Success rate = percentage of vomiting or nauseated patients who are treated with the respective dose of ondansetron and who do not continue vomiting or being nauseated. Success rate with placebo in these trials was 20%.

vomiting patients will not experience any Prophylaxis arm further PONV symptoms (nausea or retching or Some patients who receive prophylactic anti- vomiting) after administration of ondansetron, emetics would not have vomited anyway (P1) (see 1 mg and 4 mg, respectively. It was assumed that Figure 54). Success with PONV prophylaxis (P2) was the success rate with either dose was independent represented by a patient who never experienced of both clinical setting and patient. The placebo any PONV symptoms because he or she had response of about 20% found in the original received prophylactic ondansetron (Figure 54). A systematic review1 was not taken into account in failure in the prophylaxis arm was represented by a 216 the analysis. patient who received prophylactic ondansetron but Health Technology Assessment 1998; Vol. 2: No. 12

nevertheless vomited or felt nauseated after surgery desired endpoint (a PONV-free patient, for instance) (P3). For prophylaxis, the appropriate estimate of divided by the number of patients involved. efficacy was the NNT. This indicated how many patients would have to be treated prophylactically Incremental cost-effectiveness ratios were calcu- with ondansetron in order to prevent PONV in one lated to indicate how much it costs to produce the patient who would have vomited or been nauseated additional effects.3 Starting from the least effective had placebo been given. Thus the placebo response strategy, the difference between the costs of pairs was taken into account. of strategies was divided by the difference in their effectiveness. The optimal prophylactic dose as recommended by the manufacturer is 4 mg. Meta-analysis, however, Drug acquisition costs differ widely because showed that 8 mg was the most effective prophy- hospitals have different purchasing strategies. The lactic dose.2 Sensitivity analyses to compare cost- main cost parameter was the number of milligrams effectiveness were therefore undertaken with both of ondansetron required rather than the actual the 4 mg and 8 mg doses. The effect of ondan- price. Thus, milligram outcomes as reported in this setron on vomiting was consistently more pro- paper may be multiplied by the price per milligram nounced than its effect on nausea.2 However, as to generate the actual drug costs of each strategy. prophylaxis of nausea may be regarded as being BNF costs (1997) for ondansetron were £6.75, as important as prophylaxis of vomiting, the NNT £13.50 and £27.00 for 2, 4 and 8 mg ampoules, for prevention of nausea was chosen as the appro- giving an approximate price of £3.40 per milligram. priate estimate of efficacy for prophylactic doses (Table 53): 16 for a 4 mg dose and 6.4 for 8 mg.2 It For the same reason, wide differences between was assumed that ondansetron’s prophylactic anti- and within hospitals, expenses involved in adminis- emetic efficacy was independent of the CER. tration of the drug, cleaning, extra staff time and materials used, costs for rescue anti-emetic Estimate of harm medication, and costs for unscheduled admission The NNH was regarded as the appropriate due to prolonged PONV were not considered. The estimate of the likelihood for drug-related adverse sum of these supplementary costs may be regarded effects (Table 53). The NNH was assumed to be as a hospital-specific constant which may then be 30 for both prophylaxis and treatment with added to the reported costs. Costs which may be ondansetron.2 The assumption was made that relevant would be staff time for administering a there was no dose– response (i.e. the additional treatment to a patient with PONV (say 3 minutes risk for an adverse drug reaction was similar for per patient at £15 per hour) and materials (mainly 1 mg treatment and 8 mg prophylaxis). The drugs, so say £1 on the basis that older and cheaper adverse drug reaction could be a headache or drugs than those in the ondansetron class would elevated liver enzymes. be used). Any major costs would derive from re- admission or non-discharge costs assumed to Modelling be that of an in-patient day cost (£200). The total cost of arriving at each of the endpoints (P1–P3; T1–T3) (Figure 54) was calculated for both Endpoints strategies, treatment and prophylaxis, based on a Two endpoints were considered to be particularly cohort of 1000 patients. This was calculated for important: the number of patients who suffered no increasing CERs, from 10% to 90%, and different more than a single episode of PONV, and the num- doses of ondansetron (1 mg and 4 mg for treatment, ber of patients who suffered no PONV symptoms. 4 mg and 8 mg for prophylaxis). Cost-effectiveness ratios were calculated and displayed graphically. In the treatment arm, the number of patients who suffered no more than one PONV episode were Relevant ratios were: patients who did not vomit or feel nauseated at all (i.e. who did not need any treatment) or, when (i) cost per patient having symptoms, responded promptly to treat- (ii) cost per patient who experienced no more ment (Figure 54: T1 + T2). This more pragmatic than one episode of PONV approach would thus accept that a patient vomits (iii) cost per patient who never suffered any once or feels nauseated briefly before an effective PONV symptoms. treatment is administered. The second, stricter endpoint (i.e. PONV-free patients) was the number Such ratios represent the average cost-effectiveness of patients in the treatment arm who needed no ratios, which are simply the costs of generating the treatment (T1). 217 Cost-effectiveness of ondansetron: prophylaxis compared with treatment in PONV

In the prophylaxis arm both endpoints were the compared: a low-risk setting (CER, 30%) and a number of patients who never experienced any high-risk setting (CER, 60%). symptoms of PONV, either because they would have had none anyway or because prophylaxis was successful (Figure 54: P1 + P2). Results Subgroup analyses Modelling: graphical display Subgroup analyses were performed to test the Number of patients who are PONV-free impact of different risks of PONV and of different For each CER value between 10% and 90%, doses of ondansetron on cost-effectiveness. It was prophylaxis yielded more PONV-free patients assumed that the CER (i.e. what happens without than treatment (Figure 55A). This relationship was anti-emetic prophylaxis) would accurately reflect linear because it was assumed that ondansetron’s the true underlying risk in a study population. anti-emetic efficacy was independent of CER. Hence, two arbitrarily defined clinical settings were With a CER of 30% (low-risk setting), 700 of

A: Patients who are PONV-free B: Milligrams required per patient Number of patients Dosage (mg) 1000 10

8 750

6 500 4

250 2

0 0 0 20 40 60 80 100 0 20 40 60 80 100 CER (%) CER (%)

C: Milligrams required per patient who does not experience more than D: Milligrams required per patient one PONV episode who is PONV-free at any time Dosage (mg) Dosage (mg)

10 10

1 1

0.1 0.1 0 20 40 60 80 100 0 20 40 60 80 100 CER (%) CER (%)

218 FIGURE 55 PONV strategy – treatment versus prophylaxis (Treatment: , 1 mg; , 4 mg. Prophylaxis: ●, 4 mg; , 8 mg) Health Technology Assessment 1998; Vol. 2: No. 12

1000 patients will be completely PONV-free with (Table 63A). In the high-risk setting it was 13 times treatment doses of both 1 mg and 4 mg, compared more (Table 63B). with 763 patients (+ 9%) with 4 mg prophylaxis and 856 patients (+ 22%) with 8 mg prophylaxis. Incremental cost-effectiveness analysis With a CER of 60% (high-risk setting), 400 of No more than one episode of PONV 1000 patients will be completely PONV-free with With a low CER value (i.e. 30%), prophylaxis with treatment doses of both 1 mg and 4 mg, compared 8 mg yielded the highest number of patients who with 463 patients (+ 16%) with 4 mg prophylaxis experienced no more than one episode of PONV, and 556 patients (+ 39%) with 8 mg prophylaxis. while prophylaxis with 4 mg yielded the lowest number (Table 64A). The crude calculation of the Cost per patient number of milligrams of ondansetron required to The cost per patient, expressed as milligrams obtain one patient who experienced no more than required per patient, was stable across all CERs with one episode of PONV showed that the best result prophylaxis. With treatment, the cost per patient was treatment with 1 mg, at 0.4 mg/patient. was related directly to CER value (Figure 55B). Fewer milligrams per patient were required for Changing from the least successful strategy, each CER with 1 mg treatment than with 4 mg prophylaxis with a dose of 4 mg, to treatment with treatment. The highest cost per patient was with a dose of 1 mg resulted in a higher success rate the most effective prophylactic dose, 8 mg. and, because many fewer milligrams were required, resulted in a sparing effect: almost 65 mg were Cost per patient who experiences at most one saved for each additional success. Only a slight episode of PONV improvement would be achieved by switching from Costs (milligrams required per patient) increased 1 mg to 4 mg treatment; 60 mg would be required with increasing CERs with all strategies but were to generate one additional patient who had no lowest with 1 mg and 4 mg treatment, respectively more than one episode of PONV. Finally, another (Figure 55C). For this endpoint, treatment with 324 mg would be required to gain one additional both doses remained more cost-effective than patient who profited when changing the strategy prophylaxis for all CERs tested. from treatment with 4 mg to prophylaxis with 8 mg.

Cost per patient who is PONV-free at any time With a high CER value (i.e. 60%), both treatment Fewest milligrams per PONV-free patient were doses yielded more patients who had no more than required with 1 mg treatment (Figure 55D). one episode of PONV than both prophylaxis doses Treatment with 4 mg doses also demonstrated (Table 64A). The crude calculation of milligrams of a lower cost per PONV-free patients than both ondansetron required to obtain one patient who prophylaxis doses but only below a CER of 80%. experienced no more than one episode of PONV showed that the best result was treatment with Subgroup analyses 1 mg, at 0.9 mg/patient. Treatment with 4 mg versus prophylaxis with 4 mg Changing from prophylaxis at 4 mg to prophylaxis In both the low-risk (CER = 30%) and high-risk at 8 mg would result in an additional 93 patients setting (CER = 60%), 63 extra PONV-free patients who would profit, but this would require 43 mg (6% of all patients) were gained with 4 mg per additional patient. Switching then to 1 mg prophylaxis compared with 4 mg treatment treatment increases the number of patients who (Table 62A and B). However, in the low-risk setting, have no more than one episode of PONV and, at 3.3 times as many milligrams were required with the same time, would save 88 mg for each addi- prophylaxis than with treatment to achieve this tional case. Another 60 mg would be required per (Table 62A). In the high-risk setting it was 1.7 times case when changing to 4 mg treatment; the gain more (Table 62B). would be a further 30 patients who experienced no more than one episode of PONV. Treatment with 1 mg versus prophylaxis with 8 mg When 1 mg for treatment and 8 mg for prophylaxis Patients who are PONV-free were used, 156 extra PONV-free patients (16% of With both low and high CERs, treatment with 1 mg all patients) were gained with prophylaxis com- or 4 mg yielded fewer patients who were PONV-free pared with treatment in both low- and high-risk at any time compared with prophylaxis with 4 mg settings (Table 63A and B). However, for the low-risk or 8 mg (Table 64B). The crude calculation of setting 27 times as many milligrams were required milligrams of ondansetron required to obtain with prophylaxis than with treatment to achieve this one patient who experienced no more than one 219 Cost-effectiveness of ondansetron: prophylaxis compared with treatment in PONV

TABLE 62 Treatment with ondansetron, 4 mg, versus prophylaxis with ondansetron, 4 mg

Patients Decision CER mg NNT No. of Total mg mg spent mg spent per mg spent per mg spent tree patients spent per success non-failure never PONV per patient

A: CER = 30%

Treatment of established PONV with ondansetron,4 mg No PONV (no need for treatment) T1 700 0 Successful treatment (45%) T2 135 540 Failure with treatment T3 165 660 All successes T2 135 540 8.9 All non-failures T1 + T2 835 1200 1.4 PONV-free T1 700 0 1.7 All patients T1 + T2 + T3 30% 4 1000 1200 1.2

Prevention of PONV with ondansetron, 4 mg No PONV anyway P1 700 2800 No PONV because of prophylaxis (NNT 16) P2 63 250 PONV despite prophylaxis P3 238 950 All successes P2 63 250 16.0 All non-failures = PONV-free P1 + P2 763 3050 5.2 All patients P1 + P2 + P3 30% 4 16.0 1000 4000 4.0

Difference in mg spent (factor x) 3 x

Incremental cost-effectiveness ratio Difference in non-failures –73 1850 Difference in PONV-free 63 3050 Difference in costs between treatment and prophylaxis 2800 mg to generate an additional non-failure –26 mg mg to generate an additional PONV-free 49 mg

B CER = 60%

Treatment of established PONV with ondansetron, 4 mg No PONV (no need for treatment) T1 400 0 Successful treatment (45%) T2 270 1080 Failure with treatment T3 330 1320 All successes T2 270 1080 8.9 All non-failures T1 + T2 670 2400 3.6 PONV-free T1 400 0 6.0 All patients T1 + T2 + T3 60% 4 1000 2400 2.4

Prevention of PONV with ondansetron, 4 mg No PONV anyway P1 400 1600 No PONV because of prophylaxis (NNT 16) P2 63 250 PONV despite prophylaxis P3 538 2150 All successes P2 63 250 16.0 All non-failures = PONV-free P1 + P2 463 1850 8.6 All patients P1 + P2 + P3 60% 4 16.0 1000 4000 4.0

Difference in mg spent (factor x) 2 x

Incremental cost-effectiveness ratio Difference in non-failures –208 –550 Difference in PONV-free 63 1850 Difference in costs between treatment and prophylaxis 1600 mg to generate an additional non-failure 3 mg mg to generate an additional PONV-free 30 mg

episode of PONV showed that the lowest result PONV-free patients, would require an additional was treatment with 1 mg, at 0.4 mg/patient at a 43 mg for each patients who benefited. 30% CER and 1.5 mg/patient at a 60% CER. Drug-related adverse effects However, switching from 1 mg or 4 mg treatment With 4 mg or 8 mg prophylaxis, 33 extra patients to prophylaxis with 4 mg would require 44 mg for from 1000 would have an adverse drug reaction each additional PONV-free patient when the CER who would not have had this reaction had they was 30%, and 25 mg when the CER was 60%. A received placebo. Because all patients received the further change to 8 mg prophylaxis, in order to drug with prophylaxis, this outcome was indepen- 220 achieve the highest possible number of absolutely dent of the CER. With treatment (1 mg or 4 mg), Health Technology Assessment 1998; Vol. 2: No. 12

TABLE 63 Treatment with ondansetron, 1 mg, versus prophylaxis with ondansetron, 8 mg

Patients Decision CER mg NNT No. of Total mg mg spent mg spent per mg spent per mg spent tree patients spent per success non-failure never PONV per patient

A CER = 30%

Treatment of established PONV with ondansetron, 1 mg No PONV (no need for treatment) T1 700 0 Successful treatment (40%) T2 120 120 Failure with treatment T3 180 180 All successes T2 120 120 2.5 All non-failures T1 + T2 820 300 0.4 PONV-free T1 700 0 0.4 All patients T1 + T2 + T3 30% 1 1000 300 0.3

Prevention of PONV with ondansetron, 8 mg No PONV anyway P1 700 5600 No PONV because of prophylaxis (NNT 6.4) P2 156 1250 PONV despite prophylaxis P3 144 1150 All successes P2 156 1250 6.4 All non-failures = PONV-free P1 + P2 856 6850 9.3 All patients P1 + P2 + P3 30% 8 6.4 1000 8000 8.0

Difference in mg spent (factor x) 27 x

Incremental cost effectiveness ratio Difference in non-failures 36 6550 Difference in PONV-free 156 6850 Difference in costs between treatment and prophylaxis 7700 mg to generate an additional non-failure 181 mg mg to generate an additional PONV-free 42 mg

B Control event rate 60%

Treatment of established PONV with ondansetron, 1 mg No PONV (no need for treatment) T1 400 0 Successful treatment (40%) T2 240 240 Failure with treatment T3 360 360 All successes T2 240 240 2.5 All non-failures T1 + T2 640 600 0.9 PONV-free T1 400 0 1.5 All patients T1 + T2 + T3 60% 1 1000 600 0.6

Prevention of PONV with ondansetron, 8 mg No PONV anyway P1 400 3200 No PONV because of prophylaxis (NNT 6.4) P2 156 1250 PONV despite prophylaxis P3 444 3550 All successes P2 156 1250 6.4 All non-failures = PONV-free P1 + P2 556 4450 14.4 All patients P1 + P2 + P3 60% 8 6.4 1000 8000 8.0

Difference in mg spent (factor x) 13 x

Incremental cost-effectiveness ratio Difference in non-failures –84 3850 Difference in PONV-free 156 4450 Difference in costs between treatment and prophylaxis 7400 mg to generate an additional non-failure –46 mg mg to generate an additional PONV-free 25 mg

ten extra patients experienced an adverse drug patients who are likely to experience PONV. The reaction when the CER was 30%, and 20 had such systematic review suggests that ondansetron, 1 mg, a reaction when the CER was 60%. is the most effective treatment and ondansetron, 8 mg, is the most effective prophylaxis. In practice, 4 mg is the recommended dose for both treatment Comment and prophylaxis.

These results show that treatment of established Using only the point estimates of effectiveness from PONV with ondansetron, 1 mg, is the most cost- systematic reviews, there was little difference in the effective option, irrespective of the proportion of number of patients who were either PONV-free or 221 Cost-effectiveness of ondansetron: prophylaxis compared with treatment in PONV

TABLE 64 Incremental costs to generate one additional patient who either experiences no more than one episode of PONV or is completely PONV-free

A Not more than one episode of PONV

Strategy Number with Additional Total mg Additional ‘Cost’ of generating Comment Total cost/ ≤ 1 episode patients spared spent mg required one additional patient patients PONV ≤ 1 episode with ≤ 1 episode PONV spared PONV (= x drug cost)

CER = 30% Prevention, 4 mg 763 N/A 4000 N/A 5 5.2 Treatment, 1 mg 820 57 300 –3700 –65 Sparing effect 0.4 Treatment, 4 mg 835 15 1200 900 60 1.4 Prevention, 8 mg 856 21 8000 6800 324 9.3

CER = 60% Prevention, 4 mg 463 N/A 4000 N/A 9 8.6 Prevention, 8 mg 556 93 8000 4000 43 14.4 Treatment, 1 mg 640 84 600 –7400 –88 Sparing effect 0.9 Treatment, 4 mg 670 30 2400 1800 60 3.6

B Completely PONV-free

Strategy Number Additional Total mg Additional ‘Cost’ of generating Comment Total cost/ patients PONV-free required mg spent one additional patients PONV-free patients PONV-free patientspared (= x drug mg cost)

CER = 30% Treatment, 1 mg 700 N/A 300 N/A 0.4 0.4 Treatment, 4 mg 700 0 1200 900 > 900 No benefit 1.7 Prevention, 4 mg 763 63 4000 2800 44.4 5.2 Prevention, 8 mg 856 93 8000 4000 43.0 9.3

CER = 60% Treatment, 1 mg 400 N/A 600 N/A 1.5 1.5 Treatment, 4 mg 400 0 2400 1800 > 1800 No benefit 6.0 Prevention, 4 mg 463 63 4000 1600 25.4 8.6 Prevention, 8 mg 556 93 8000 4000 43.0 14.4

who experienced no more than one episode of expensive but there is no reason to expect that this PONV in any of these strategies. Because of this, would be rarer with prophylactic ondansetron than the incremental analysis was highly unfavourable to with treatment. Adverse events consequent on the prophylaxis because prophylaxis will consume very much greater use of ondansetron might incur much higher quantities of anti-emetic but few addi- greater costs – either through treatment of head- tional patients will have a beneficial outcome. Even ache and an unplanned overnight stay because of this may overstate the case, because point estimates headache, or through the unnecessary investigation have CIs about them. It might be considered legiti- of raised liver enzymes. Again, there is no evidence mate to say that the number of additional patients that these are major concerns. apparently obtaining a benefit with prophylaxis was within the uncertainties of our estimates, so that no Can these results be extrapolated to other anti- additional patients would be likely to benefit. If this emetics? No direct comparisons have been found were the case, prophylaxis exposes patients to much to indicate that any of the commonly prescribed higher doses of drug for no benefit. anti-emetics are significantly better or worse than ondansetron (unpublished observations). Most In money terms, how much more expensive other anti-emetics have a much lower cost per dose prophylaxis will be depends on the unit cost of than ondansetron, and so monetary differences drug bought by hospital pharmacies. In practice between prophylaxis and treatment may not be this may be very much lower than the advertised great. But if other anti-emetics do not differ in price to GPs. However, hospitals in which prophy- efficacy, but are cheaper, then that does point to laxis with ondansetron was the norm would enjoy other possible savings, albeit recognising the need significant savings if a change were made to also to examine their safety. treatment of established PONV. What is clear is that the great variability in results What might change this view? Unplanned overnight shown in individual trials, both in respect to PONV 222 stays in day-case surgery because of PONV are clearly rates without treatment and in the apparent effect Health Technology Assessment 1998; Vol. 2: No. 12

of treatment,1,2 makes economic evaluation point- 2. Tramèr MR, Reynolds DJ, Moore RA, McQuay less outside of evidence systematically gathered. HJ. Efficacy, dose-response and safety of ondansetron in prevention of postoperative nausea and vomiting: a quantitative systematic review of randomised placebo-controlled trials. References Anesthesiology 1997;87:1277–89. 1. Tramèr MR, Moore RA, Reynolds DJM, McQuay HJ. A quantitative systematic review of ondansetron in 3. Karlsson G, Johannesson M. The decision rules treatment of established postoperative nausea and of cost-effectiveness analysis. PharmacoEconomics vomiting. BMJ 1997;314:1088–92. 1996;9:113–20.

223

Health Technology Assessment 1998; Vol. 2: No. 12

Chapter 23 Conclusion

Interventions ment led to faster recovery would increase the pressure to improve current practice, which is In treating acute pain, as in other areas of often less than ideal. medicine, tradition and ill-informed prejudice sometimes hold sway over evidence and common Non-opioids: paracetamol, combinations sense. This study concentrated on gathering evi- and NSAIDs dence for the treatments which are simple. In this Effective relief can be achieved with oral non- chapter that evidence is drawn together in a wider opioids and NSAIDs. These drugs are appropriate frame, including interventions for which there for many post-surgical and post-traumatic pains, are no systematic reviews. Wherever possible our especially when patients go home on the day of the recommendations are based on randomised trials. operation. The evolving league table for analgesic efficacy compiled from randomised trials after all Effective pain management is fundamental to kinds of surgery is shown in Figure 56. Analgesic quality care and, while we consider that good pain efficacy is expressed as the NNT, the number of control speeds recovery, there is still no compelling patients who need to receive the active drug for evidence that this is so, although advantage can be one to achieve at least 50% relief of pain compared shown with proxy measures like mobility or cough- with placebo over a 6-hour treatment period. The ing. However, evidence that good pain manage- most effective drugs have low NNT values of about

Codeine, 60 mg

Dihydrocodeine, 30 mg

Tramadol, 50 mg

Dextropropoxyphene, 65 mg

Paracetamol, 500 mg

Aspirin, 650 mg, + codeine, 60 mg

Tramadol, 100 mg

Paracetamol, 1000 mg

Paracetamol, 650 mg, + dextropropoxyphene, 65 mg

Ibuprofen, 200 mg

Morphine, 10 mg i.m.

Ibuprofen, 400 mg

Diclofenac, 25 mg

Diclofenac, 50 mg

1 10 100 NNT (95% CI)

FIGURE 56 Oral analgesic NNT league table 225 Conclusion

2, meaning that for every two patients who receive combination is also effective. Initial prescription the drug one patient will get at least 50% relief of oral NSAIDs can be supplemented with para- because of the treatment (the other patient may cetamol. As pain wanes then the prescription obtain relief but it does not reach the 50% level). should be paracetamol-based, supplemented if necessary by NSAIDs. For paracetamol, 1 g, the NNT is nearly 5. Combination of paracetamol, 650 mg, with There is an old adage that if a patient can swallow dextropropoxyphene, 65 mg, improves the NNT then it is best to take drugs by mouth. There is no value slightly. Ibuprofen is better with an NNT evidence that NSAIDs given rectally or by injection of 3, as is diclofenac at about 2.5. perform better (or faster) than the same drug at the same dose given by mouth (see chapter 11). These NNT comparisons are against placebo; These other routes become appropriate when the best NNT value of 2 means that while 50 from patients cannot swallow. Topical NSAIDs are 100 patients will get at least 50% relief because effective in acute musculoskeletal injuries of the treatment, another 20% will have a placebo ibuprofen has an NNT of 3 for at least 50% response which gives them at least 50% relief. relief at one week compared with placebo Hence, with diclofenac, 70 from 100 patients (see chapter 12). will have effective pain relief. Adverse effect data on NSAIDs from long-term This alternative way of looking at the effect of the dosing, where gastric bleeding is the main worry, various analgesics is shown in Figure 57. The range rates ibuprofen the safest.1 In acute pain, the is from about 25% of patients getting at least 50% main concerns are renal and coagulation pain relief with codeine, 60 mg (largely because of problems. Acute renal failure can be precipitated the effect of placebo), to over 70%, at the high end in patients with pre-existing heart or kidney of the 95% CI, with oral NSAID. For comparison, disease, those on loop diuretics or those who with morphine, 10 mg i.m., about 53% of patients have lost more than 10% of blood volume. get more than 50% pain relief. Because the effect NSAIDs cause significant lengthening (~30%) of placebo is added in, the comparisons between of bleeding time, usually still within the normal analgesics are not as stark as with NNT. range. This can last for days with aspirin, hours with non-aspirin-based NSAIDs. Whether or not The clear message is that, of the oral analgesics, NSAIDs cause significant increase in blood loss NSAIDs perform best and paracetamol alone or in remains contentious.

NNT Codeine, 60 mg 16.7 Dihydrocodeine, 30 mg 9.7 Tramadol, 50 mg 8.3 Dextropropoxyphene, 65 mg 7.7 Paracetamol, 500 mg 5.6 Aspirin, 650 mg, + codeine, 60 mg 5.3 Tramadol, 100 mg 4.8 Paracetamol, 1000 mg 4.6 Paracetamol, 650 mg, + dextropropoxyphene, 65 mg 4.4 Ibuprofen, 200 mg 3.3 Morphine, 10 mg i.m. 2.9 Ibuprofen, 400 mg 2.7 Diclofenac, 25 mg 2.6 Diclofenac, 50 mg 2.3 25 50 75 100 Patients with at least 50% PR (%) (95% CIs)

226 FIGURE 57 Oral analgesic league table – percentage with at least 50% pain relief Health Technology Assessment 1998; Vol. 2: No. 12

Other drugs of administration changed to achieve faster As yet there is no evidence from systematic reviews control. Delayed release formulations, oral or for a number of niche analgesic interventions. transdermal, should not be used in acute pain, These include: inhaled nitrous oxide, which can because delayed onset and offset are dangerous provide fast-onset, fast-offset analgesia for obstetrics in this context. and wound dressings; corticosteroids to reduce pain and swelling after head and neck surgery, and There is no compelling evidence that one opioid when swelling causes pain in cancer; ketamine in is better than another, but there is good evidence emergency analgesia and anaesthesia; and that pethidine has a specific disadvantage4 and clonidine. no specific advantage. Given in multiple doses, the metabolite norpethidine can accumulate and Opioids act as a central nervous system irritant, ultimately For severe acute pain, opioids are the first-line causing convulsions, especially in renal dysfunction. treatment and, to date, only one systematic review Pethidine should not be used when multiple has been found – that for injected morphine (see injections are needed. The long-held view that chapter 13). Intermittent opioid injection can pro- pethidine is better than other opioids when vide effective relief of acute pain.2 Unfortunately, dealing with colicky pain is no longer tenable.5 adequate doses are withheld because of traditions, misconceptions, ignorance and fear. Doctors and Morphine (and its relatives diamorphine nurses fear addiction and respiratory depression. and codeine) has an active rather than a toxic Addiction is not a problem with opioid use in acute metabolite, morphine-6-glucuronide. In renal pain. Over 11,000 patients were followed-up a year dysfunction, this metabolite can accumulate and after opioids were given for acute pain, and just result in greater effect from a given dose, because four were considered addicts.3 it is more active than morphine. If dose is being titrated against effect, as it should be, this will not Irrespective of the route, opioids used for matter. Less morphine will be needed. Accumu- people who are not in pain, or in doses larger lation can be a problem with unconscious intensive than necessary to control the pain, can slow or care patients on fixed dose schedules when renal indeed stop breathing. The key principle is to function is compromised. titrate the dose against the desired effect – pain relief – and minimise unwanted effects (Figure 58). Opioid adverse effects include nausea and If the patient is still complaining of pain and you vomiting, constipation, sedation, pruritus, urinary are sure that the drug has all been delivered and retention and respiratory depression. There is absorbed, then it is safe to give another, usually no good evidence that the incidence is different smaller, dose (5 minutes after intravenous, 1 hour with different opioids at the same level of anal- after intramuscular or subcutaneous, 90 minutes gesia. There is good evidence that the risk of after oral). If the second dose is also ineffective, adverse events is increased when high-technology then the process should be repeated or the route approaches are used for drug administration.6

Principle for safe and effective opiate use – titrate to effect Serious adverse If the patient is asking for more opioid than it usually events signals inadequate pain control: – too little drug Pain free – trust the patient – too long between doses – too little attention paid to the patient – too much reliance on rigid (inadequate) prescriptions Pain intensity Dose of opiate

Time

FIGURE 58 Titrate opioids to effect 227 Conclusion

There are strong arguments, based on minimising adverse effects. This advantage is more obvious risk, for using one opioid only, so that everyone is when it is possible to give further doses via a familiar with dosage, effects and problems. Our catheter, extending the duration of analgesia. first choice opioid is morphine. Whichever drug Details are given in Table 65. is chosen, simple changes in the way opioids are used, good staff education and implemention of There is a necessary distinction between blocks an algorithm for intermittent opioid dosing, can undertaken to permit surgery, and blocks under- have a powerful impact on pain relief and taken together with a general anaesthetic to patient satisfaction.2 provide postoperative pain relief. There is clear evidence that blocks can indeed provide good relief Nurse-administered intermittent opioid injection in the initial postoperative period8 and no evidence requires good staffing levels to minimise delay to suggest that patients with blocks then experience between need and injection. Staffing shortages, ‘rebound’, and need more postoperative pain ward distractions and controlled drug regulations relief. The risk of neurological damage is the major all increase the delay. PCA overcomes these drawback9 and, ideally, blocks should not be logistical problems. The patient presses a button undertaken on anaesthetised patients. and receives a pre-set dose of opioid, from a syringe driver connected to an intravenous or Epidural analgesia subcutaneous cannula. This delivers the opioid Epidural infusion via a catheter can offer to the same opioid receptors as an intermittent continuous relief after trauma or surgery for lower injection but allows the patient to circumvent limb, spine, abdominal or chest. The current delays. Not surprisingly, there is little difference in optimal infusate is an opioid/local anaesthetic outcome between efficient intermittent injection mixture. Opioids and local anaesthetics have and PCA.7 Good risk management with PCA a synergistic effect, so that lower doses of each should emphasise the same drug, protocols are required for equivalent analgesia with fewer and equipment throughout a hospital. adverse effects.10

Novel routes of opioid administration, intended The risks are those of an epidural (dural to improve analgesia and reduce adverse effects, puncture, infection, haematoma, nerve damage), include intra-articular (see chapter 17), nasal, those of the local anaesthetic (hypotension, active transdermal and inhalational. These may motor block, toxicity), and those of the opioid prove to have advantage over conventional (nausea, sedation, urinary retention, respiratory routes, different kinetic profiles or greater depression, pruritus). Wrong doses do get convenience, but their place in mainstream administered,6 so increased surveillance is care is unproven. mandatory. The risk of persistent neurological sequelae after an epidural is about 1 in 5000.11 Regional analgesia Debate continues as to whether patients with The perceived advantage of regional analgesia with epidural infusions can be nursed on general local anaesthetic is that it can deliver complete pain wards. These techniques are only appropriate relief by interrupting pain transmission from for major trauma or surgery when the potential a localised area, so avoiding generalised drug benefits outweigh the risks.

TABLE 65 Regional analgesia summary

Indications Advantages Problems

Low technology Topical Surface wounds Simple Short duration Wound infiltration Most wounds Simple Short duration Peripheral nerve blocks Limb surgery/trauma Catheter possible ?? Plexus blocks Limb surgery Catheter possible Nerve damage, motor block

High technology Epidural (including caudal) Major surgery Catheter possible; Adverse effects (thoracoabdominal, reduced thrombo- surveillance lower limb) embolism Intrathecal Major surgery Long duration relief (thoracoabdominal, possible from single lower limb) injection low-dose opioid 228 Health Technology Assessment 1998; Vol. 2: No. 12

Other techniques BOX 4 How to achieve successful pain management While experts can obtain good results with specialised procedures, such as paravertebral or interpleural Factors to consider when choosing therapy injections, the evidence that in less skilled hands Co-existing illness these are better than standard methods (should-do Staff availability Equipment available rather than can-do evidence) is often lacking. Risks and unwanted effects of the various options Appropriateness of the chosen intervention for that pain TENS and acupuncture Evidence of efficacy for the chosen intervention TENS is not effective for postoperative pain (see Cost chapter 20) and is of limited value for labour pain Steps to successful management (see chapter 21). Systematic reviews of acupuncture Regular assessment of pain and adverse effects are confined to chronic pain. Protocols for monitoring and treating pain Protocols for monitoring and treating adverse effects Psychological methods Titration of doses at short intervals until pain relieved There is evidence that psychological approaches Consideration of more than one approach are beneficial.12 Cognitive behavioural methods Appropriate back-up by identified personnel Continuing in-service training and education can reduce pain and distress in patients with burns. Preparation before surgery can reduce Predictable problems postoperative analgesic consumption. Patient Problems Babies and infants Communication; drug handling Elderly Co-existing illness; drug handling Clinical settings and Respiratory disease Respiratory depression; recommendations (see Box 4) NSAIDs and asthma Renal failure Drug handling; NSAIDs General Head injury or impaired Assessment; dose titration The tenets of good management of acute pain are consciousness that, with good staff (and patient) education in place, Drug addiction or Dose titration; weaning; already taking opioids respiratory depression after appropriate drug doses are given when needed by the nerve block which stops pain appropriate route and delivery method. Schemes Sickle cell disease Assessment; varying analgesic needs have to be flexible enough to respond to individual patient need and different clinical settings. A general Key points for improving acute pain management strategy is presented in Figure 59. 1 Opt for safety and simplicity. 2 Measure and record pain regularly – be proactive. 3 Choose evidence-based interventions. There is controversy about the optimal timing of 4 Individualise treatment and allow patient to control analgesia. initial analgesia. The idea is that analgesia is more 5 Choose appropriate drug, route and mode of delivery. effective when given before pain begins than when 6 Provide education for staff and patients. given after. Most randomised trials comparing the

Low tech Intermittent opioid bolus injection

High tech • Epidural infusion • Patient-controlled analgesia to deliver opioid

Pain Oral NSAID ± intensity paracetamol

Oral paracetamol ± NSAID

Time (pain decreases or goes away)

FIGURE 59 General strategy 229 Conclusion

same intervention given before or after pain starts 2. Gould TH, Crosby DL, Harmer M, Lloyd SM, have not shown clinical advantage of so-called pre- Lunn JN, Rees GAD, et al. Policy for controlling emptive analgesia.13 Whether poorly-controlled pain after surgery: effect of sequential changes in management. BMJ 1992;305:1187–93. acute pain generates chronic pain is also controversial. 3. Porter J, Jick H. Addiction rate in patients treated with narcotics. N Engl J Med 1980;302:123. Problem pains and patients 4. Szeto HH, Inturrisi CE, Houde R, Saal S, Cheigh J, Standard interventions and protocols will cope Reidenberg M. Accumulation of norperidine, an with most acute pain problems but some patients, active metabolite of meperidine, in patients with particularly in hospital, will require special manage- renal failure or cancer. Ann Intern Med ment. Expertise can be developed in specific units 1977;86:738–41. but, if not available, the advice of your acute pain 5. Nagle CJ, McQuay HJ. Opiate receptors; their role service should be sought. In particular do not let in effect and side-effect. Curr Anaesth Crit Care pain go untreated in children. 1990;1:247–52. 6. Bates DW, Cullen DJ, Laird N, Petersen LA, Small Conclusion SD, Servi D, et al. Incidence of adverse drug events and potential adverse drug events. JAMA The key to successful pain management is 1995;274:29–34. education, not new drugs or high-technology 7. Ballantyne JC, Carr DB, Chalmers TC, Dear KB, delivery systems. Existing tools can do the job if Angelillo IF, Mosteller F. Postoperative patient- doctors and nurses are educated, both to dispel the controlled analgesia: meta-analyses of initial random- myths and misconceptions and to take responsi- ized control trials. J Clin Anesth 1993;5:182–93. bility for providing good pain control. It is much 8. McQuay HJ, Carroll D, Moore RA. Postoperative easier to dispel myths when you have the evidence. orthopaedic pain – the effect of opiate pre- For many years patients were not given adequate medication and local anaesthetic blocks. Pain analgesia for abdominal pain in case it masked the 1988;33:291–5. signs necessary for diagnosis. This was wrong.14 9. Bridenbaugh PO. Complications of local anesthetic neural blockade. In: Cousins MJ, Bridenbaugh PO, Pain relief should not be seen as someone else’s editors. Neural blockade. 2nd edition. Philadelphia: responsibility; nor should it be just dismissed, Lippincott, 1988:695–717. because ‘in the end the pain and the patient go away’. Freedom from pain is important to patients. 10. McQuay H. Epidural analgesics. In: Wall P, Melzack In 1846, the first anaesthetic provided pain-free R, editors. Textbook of pain. 3rd edition. London: surgery. Some 150 years later patients should Churchill Livingstone, 1994:1025–34. not have to endure unrelieved pain anywhere 11. Kane RE. Neurologic deficits following epidural or in hospital. spinal anesthesia. Anesth Analg 1981;60:150–61. 12. Justins DM, Richardson PH. Clinical management References of acute pain. Br Med Bull 1991;47:561–83. 13. McQuay HJ. Pre-emptive analgesia: a systematic 1. Henry D, Lim LL, Rodriguez LAG, Gutthann SP, review of clinical studies. Ann Med 1995;27:249–56. Carson JL, Griffin M, et al. Variability in risk of gastrointestinal complications with individual non- 14. Attard AR, Corlett MJ, Kidner NJ, Leslie AP, Fraser steroidal anti-inflammatory drugs: results of a IA. Safety of early pain relief for acute abdominal collaborative meta-analysis. BMJ 1996;312:1563–6. pain. BMJ 1992;305:554–6.

230 Health Technology Assessment 1998; Vol. 2: No. 12

Acknowledgements

he authors owe a huge debt to our colleagues Oxford or who have come here to work. They T in Oxford and around the world who have include Muir Gray, Iain Chalmers, David Sackett, contributed to this work. Alejandro Jadad, Dawn Doug Altman and Jon Deeks. Many other people Carroll and our volunteers broke the back of hand- have acted as sounding-boards, such as Douglas searching for randomised trials of pain-relieving Justins, Peter Evans, Ed Charlton and Tony interventions. The chapters on finding trials, on Dickenson. Others have helped and supported, appraising their quality and on existing systematic including John Williams, Dick Waite, Ann reviews borrow heavily from the publications which Southwell and Aprille Cornell. Our libraries in and came from Alejandro’s DPhil thesis. The Oxford around Oxford have been magnificent, especially contributions of Dawn Carroll, Sally Collins, David Claire Abbott here in the Churchill Hospital, as Gavaghan, Jayne Edwards, John Reynolds and Phil have many libraries in academia and industry who Wiffen are gratefully acknowledged. Without their have provided the raw material for the work. endeavour and their patience much of this would not have happened. Martin Tramèr (Oxford and Financial support for the work has been provided Geneva), Eija Kalso (Helsinki), Pascale Picard from many different sources for different aspects (Clermont-Ferrand), Clara Faura (Alicante) and of the work, including: Pain Research Funds, Göran Leijon (Linköping) collaborated on partic- European Union Biomed 2, NHS Research ular reviews. Martin applied the same obsession he and Development HTA Programme, and NHSE used on systematic reviews of anti-emetics to reviews Anglia and Oxford Research and Development of analgesics, and his work on anti-emetics is sum- Directorate, as well as pharmaceutical companies. marised in later chapters, together with an economic analysis which includes much thought and Our thanks are also due to the referees for their work by Ceri Phillips, our health economics corre- perseverance in reading the report and for the spondent from Swansea. Our clinical colleagues quality of their comments. Chris Glynn and Tim Jack have been helpful, enthusiastic and protective sounding-boards. Finally, it is a pleasure to acknowledge our old mentor, John Lloyd. Not only did John start the The authors’ thinking over the past 6 years or so Oxford Pain Relief Unit but he set us and many has been informed by discussion and argument others on a quest for better treatments and with some very special people who were either in greater understanding.

231

Health Technology Assessment 1998; Vol. 2: No. 12

Health Technology Assessment panel membership This report was identified as a priority by the Pharmaceutical Panel. Acute Sector Panel Chair: Professor John Farndon, University of Bristol †

Professor Senga Bond, Professor Richard Ellis, Mr Ian Hammond, Professor Jon Nicholl, University of Newcastle- St James’s University Hospital, Bedford & Shires Health & University of Sheffield † † upon-Tyne Leeds Care NHS Trust Professor John Norman, Professor Ian Cameron, Mr Leonard Fenwick, Professor Adrian Harris, University of Southampton Southeast Thames Regional Freeman Group of Hospitals, Churchill Hospital, Oxford Dr John Pounsford, Health Authority Newcastle-upon-Tyne † Professor Robert Hawkins, Frenchay Hospital, Bristol † † Professor David Field, University of Bristol Professor Gordon Stirrat, Ms Lynne Clemence, † Mid-Kent Health Care Trust † Leicester Royal Infirmary Dr Gwyneth Lewis, St Michael’s Hospital, Bristol † Ms Grace Gibbs, Department of Health Professor Michael Sheppard, Professor Francis Creed, West Middlesex University Dr Chris McCall, Queen Elizabeth Hospital, University of Manchester † Hospital NHS Trust † General Practitioner, Dorset † Birmingham † Professor Cam Donaldson, Dr Neville Goodman, Professor Alan McGregor, Dr William Tarnow-Mordi, University of Aberdeen Southmead Hospital St Thomas’s Hospital, London University of Dundee † Mr John Dunning, Services Trust, Bristol Mrs Wilma MacPherson, Professor Kenneth Taylor, Papworth Hospital, Professor Mark P Haggard, St Thomas’s & Guy’s Hospitals, Hammersmith Hospital, Cambridge † MRC † London London

Diagnostics and Imaging Panel Chair: Professor Mike Smith, University of Leeds †

Professor Michael Maisey, Professor MA Ferguson-Smith, Professor Chris Price, Mr Stephen Thornton, Guy’s & St Thomas’s Hospitals, University of Cambridge † London Hospital Medical Cambridge & Huntingdon * † London Dr Mansel Hacney, School Health Commission Professor Andrew Adam, University of Manchester Dr Ian Reynolds, Dr Gillian Vivian, † UMDS, London Professor Sean Hilton, Nottingham Health Authority Royal Cornwall Hospitals Trust † Dr Pat Cooke, St George’s Hospital Professor Colin Roberts, Dr Jo Walsworth-Bell, RDRD, Trent Regional Medical School, London University of Wales College South Staffordshire Health Authority Mr John Hutton, of Medicine Health Authority † Ms Julia Davison, MEDTAP International Inc., Miss Annette Sergeant, Dr Greg Warner, St Bartholomew’s Hospital, London Chase Farm Hospital, General Practitioner, † London Professor Donald Jeffries, Enfield Hampshire † Professor Adrian Dixon, St Bartholomew’s Hospital, Professor John Stuart, University of Cambridge † London † University of Birmingham Mr Steve Ebdon-Jackson, Dr Andrew Moore, Dr Ala Szczepura, Department of Health † Editor, Bandolier † University of Warwick †

Methodology Panel Chair: Professor Martin Buxton, Brunel University †

Professor Anthony Culyer, Dr Rory Collins, Mr John Henderson, Dr Maurice Slevin, University of York * University of Oxford Department of Health † St Bartholomew’s Hospital, Dr Doug Altman, Institute of Professor George Davey-Smith, Mr Philip Hewitson, Leeds FHSA London Health Sciences, Oxford † University of Bristol Professor Richard Lilford, Dr David Spiegelhalter, Regional Director, R&D, Professor Michael Baum, Dr Vikki Entwistle, Institute of Public Health, West Midlands † † Royal Marsden Hospital University of Aberdeen † Cambridge Mr Nick Mays, King’s Fund, Professor Charles Warlow, Professor Nick Black, Professor Ray Fitzpatrick, London † University of Oxford † Western General Hospital, London School of Hygiene Professor Ian Russell, † † Edinburgh & Tropical Medicine Professor Stephen Frankel, University of York † Professor Ann Bowling, University of Bristol Professor David Sackett, University College London Dr Stephen Harrison, Centre for Evidence Based * Previous Chair Medical School † University of Leeds Medicine, Oxford † † Current members 235 continued Health Technology Assessment panel membership

continued Pharmaceutical Panel Chair: Professor Tom Walley, University of Liverpool †

Professor Michael Rawlins, Mr Barrie Dowdeswell, Dr Keith Jones, Mr Simon Robbins, University of Newcastle- Royal Victoria Infirmary, Medicines Control Agency Camden & Islington upon-Tyne* Newcastle-upon-Tyne Health Authority, Professor Trevor Jones, † † London Dr Colin Bradley, Dr Tim Elliott, ABPI, London University of Birmingham Department of Health † Dr Frances Rotblat, Ms Sally Knight, † Lister Hospital, Stevenage † Medicines Control Agency Professor Alasdair Dr Desmond Fitzgerald, Breckenridge, RDRD, Dr Andrew Mortimore, Mrs Katrina Simister, Mere, Bucklow Hill, Cheshire † Northwest Regional Southampton & SW Hants Liverpool Health Authority Health Authority Dr Felicity Gabbay, Health Authority † † Dr Ross Taylor, Transcrip Ltd University of Aberdeen † Ms Christine Clark, Mr Nigel Offen, Essex Rivers † Hope Hospital, Salford † Dr Alistair Gray, Healthcare, Colchester Dr Tim van Zwanenberg, Health Economics Research Dr John Posnett, Northern Regional Mrs Julie Dent, † Unit, University of Oxford University of York Health Authority Ealing, Hammersmith & Hounslow Health Authority, Professor Keith Gull, Mrs Marianne Rigge, Dr Kent Woods, RDRD, London University of Manchester The College of Health, London † Trent RO, Sheffield †

Population Screening Panel Chair: Professor Sir John Grimley Evans, Radcliffe Infirmary, Oxford †

Dr Sheila Adam, Dr Tom Fahey, Professor Alexander Markham, Dr Sarah Stewart-Brown, Department of Health* University of Bristol † St James’s University Hospital, University of Oxford † † Ms Stella Burnside, Mrs Gillian Fletcher, Leeds Ms Polly Toynbee, Altnagelvin Hospitals Trust, National Childbirth Trust † Professor Theresa Marteau, Journalist † Londonderry † UMDS, London Professor George Freeman, Professor Nick Wald, Dr Carol Dezateux, Institute of Charing Cross & Westminster Dr Ann McPherson, University of London † Child Health, London † Medical School, London General Practitioner, † Professor Ciaran Woodman, Dr Anne Dixon Brown, Dr Mike Gill, Brent & Harrow Oxford † Centre for Cancer NHS Executive, Health Authority Professor Catherine Peckham, Epidemiology, Manchester Anglia & Oxford † Dr JA Muir Gray, RDRD, Institute of Child Health, † Professor Dian Donnai, Anglia & Oxford RO London St Mary’s Hospital, Dr Anne Ludbrook, Dr Connie Smith, Manchester † University of Aberdeen † Parkside NHS Trust, London

Primary and Community Care Panel Chair: Dr John Tripp, Royal Devon & Exeter Healthcare NHS Trust †

Professor Angela Coulter, Professor Shah Ebrahim, Mr Edward Jones, Dr Fiona Moss, King’s Fund, London * Royal Free Hospital, London Rochdale FHSA Thames Postgraduate Medical and Dental Education † Professor Martin Roland, Mr Andrew Farmer, Professor Roger Jones, University of Manchester * Institute of Health Sciences, UMDS, London Professor Dianne Newham, Oxford † King’s College London Dr Simon Allison, Mr Lionel Joyce, Professor Gillian Parker, University of Nottingham Ms Cathy Gritzner, Chief Executive, Newcastle City † The King’s Fund † Health NHS Trust University of Leicester Mr Kevin Barton, Dr Robert Peveler, East London & City Professor Andrew Haines, Professor Martin Knapp, University of Southampton † Health Authority † RDRD, North Thames London School of Economics Regional Health Authority & Political Science Dr Mary Renfrew, Professor John Bond, University of Oxford University of Newcastle- Dr Nicholas Hicks, Dr Phillip Leech, † † † Ms Hilary Scott, upon-Tyne Oxfordshire Health Authority Department of Health Tower Hamlets Healthcare Ms Judith Brodie, Professor Richard Hobbs, Professor Karen Luker, NHS Trust, London † Age Concern, London † University of Birmingham † University of Liverpool Dr Nicky Cullum, Professor Allen Hutchinson, Professor David Mant, * Previous Chair University of York † University of Sheffield † NHS Executive South & West † † Current members

236 National Coordinating Centre for Health Technology Assessment, Advisory Group Chair: Professor John Gabbay, Wessex Institute for Health Research & Development †

Professor Mike Drummond, Professor James Raftery, Professor Andrew Stevens, Centre for Health Economics, Health Economics Unit, Department of Public Health University of York † University of Birmingham † & Epidemiology, † Ms Lynn Kerridge, Dr Paul Roderick, University of Birmingham Wessex Institute for Health Research Wessex Institute for Health Research & Development † & Development Dr Ruairidh Milne, Professor Ian Russell, Wessex Institute for Health Research Department of Health, Sciences & Clinical & Development † Evaluation, University of York † Ms Kay Pattison, Dr Ken Stein, Research & Development Directorate, Wessex Institute for Health Research NHS Executive † & Development † † Current members Health TechnologyAssessment 98 o.2 o 12 No. 2: Vol. 1998;

Copies of this report can be obtained from: