meeting of the Association of British Neurologists in April vessels: a direct comparison between intravenous and intra-arterial DSA. 1992. We thank Professor Charles Warlow for his helpful Cl/t Radiol 1991;44:402-5. 6 Caplan LR, Wolpert SM. Angtography in patients with occlusive cerebro- comments. vascular disease: siews of a stroke neurologist and neuroradiologist. AmcincalsouraalofNcearoradiologv 1991-12:593-601. 7 Kretschmer G, Pratschner T, Prager M, Wenzl E, Polterauer P, Schemper M, E'uropean Carotid Surgery Trialists' Collaboration Group. MRC European ct al. Antiplatelet treatment prolongs survival after carotid bifurcation carotid surgerv trial: interim results for symptomatic patients with severe endarterectomv. Analysts of the clinical series followed by a controlled trial. (70-90()',) or tvith mild stenosis (0-299',Y) carotid stenosis. Lantcet 1991 ;337: AstniSuirg 1990;211 :317-22. 1 235-43. 8 Antiplatelet Trialists' Collaboration. Secondary prevention of vascular disease 2 North American Svmptotnatic Carotid Endarterectomv Trial Collaborators. by prolonged antiplatelet treatment. BM7 1988;296:320-31. Beneficial effect of carotid endarterectomv in svmptomatic patients with 9 Murie JA, Morris PJ. Carotid endarterectomy in Great Britain and Ireland. high grade carotid stenosis. VEngl 7,1Med 1991;325:445-53. Br7Si(rg 1986;76:867-70. 3 Hankev CJ, Warlows CP. Svmptomatic carotid ischaemic events. Safest and 10 Murie J. The place of surgery in the management of carotid artery disease. most cost effective way of selecting patients for angiography before Hospital Update 1991 July:557-61. endarterectomv. Bt4 1990;300:1485-91. 11 Dennis MS, Bamford JM, Sandercock PAG, Warlow CP. Incidence of 4 Humphrev PRD, Sandercock PAG, Slatterv J. A simple method to improve transient ischemic attacks in Oxfordshire, England. Stroke 1989;20:333-9. the accuracy of non-invasive ultrasound in selecting TIA patients for 12 Dennis M, Warlow CP. Strategy for stroke. BM7 1991;303:636-8. cerebral angiographs. 3' Veurol, NeuotsrtrgPsschiates 1990;53:966-71. 5 B3orgstein Rl, Brown MINM, W'aterston J, Butler P, Thakkar CH, Wylie IG, Swvash .M. Digital subtraction angiographv of the extracranial cerebral (Accepted I Septentiber 1992)

Quality oflife measures in health care. I: Applications and issues in assessment

Ray Fitzpatrick, Astrid Fletcher, Sheila Gore, David Jones, David Spiegelhalter, David Cox

Summary Many clinicians remain unsure of the relevance of Box 1: Applications of quality of life measuring quality of life to their clinical practice. measures measures have In health economics quality of life * Screening and monitoring for psychosocial become the standard means of assessing the results problems in individual patient care of health care interventions and, more contro- * versially, the means of prioritising funding; but they Population surveys of perceived health problems have many other applications. This article-the first * Medical audit of three on measuring quality of life-reviews the * Outcome measures in health services or evaluation instruments available and their application in research screening programmes, audit, health care research, * Clinical trials and clinical trials. Using the appropriate instrument * Cost-utility analyses is essential if outcome measures are to be valid and clinically meaningful. patients' judgments of quality of life differ substantially Interest in measuring quality of life in relation to and systematic assessment may improve health profes- health care has increased in recent years.' 2 The sionals' judgments.':1 Clinicians seem to find the Department ofPublic information from quality of life measures useful and Health and Primary Care, purpose is to provide more accurate assessments of University ofOxford, individuals' or populations' health and of the benefits informative but trials have found that the additional Nuffield College, Oxford and harm that may result from health care. The term information does not greatly alter clinical decisions OXI 1NF quality of life misleadingly suggests an abstract and or short term changes in health status.7 These Ray Fitzpatrick, itniversitV philosophical approach, but most approaches used in disappointing results may arise either because the lectu.rer in medical sociology medical contexts do not attempt to include more quality of life data are inappropriate to clinical decision general notions such as life satisfaction or living making or, more likely, because the information is not Royal Postgraduate standards and tend rather to concentrate on aspects of fed back to clinicians in the most useful format or at the Medical School, London personal experience that might be related to health and right time. W12 OHS Quality of life measures used for screening need to Astrid Fletcher, senior health care. Some of the commonly used synonyms for lecturer int epidemiology quality of life more accurately convey the content and be evaluated in terms of sensitivity (false negative purpose of measures-health related quality of life, results) and specificity (false positive results). Instru- Departnment of subjective health status, functional status. This is the ments whose value has been proved for screening Epidemiology and Public first of three papers intended to review measurement should not be assumed to be effective for other Health, University of issues surrounding the use of the growing number of purposes-for example, as outcome measures in trials Leicester, Leicester questionnaires and interview based instruments or in evaluation studies. A recent conference on David Jones, professor of designed to assess health related quality of life. applications of quality of life instruments in routine medical statistics patient care concluded that, in the United States at least, many practitioners have a mixture of enthusiasm Medical Research Council Alternative applications Biostatistics Unit, for their potential relevance to clinical practice and Cambridge CB2 2SR Quality of life measures can be used in many ways in unresolved doubts.' Sheila Gore, seniior statistician health care (box 1). For example, quality of life David Spiegelhalter, senior instruments have been shown to be better than conven- POPULATION APPLICATIONS statistician tional rheumatological measures as predictors of long Quality of life instruments can also be used in term outcomes in rheumatoid arthritis in terms of both surveys of the health of district or general practice Nuffield College, Oxford morbidity and mortality.' They can therefore be used populations. Such instruments can assess subjective OXI 1NF to identify patients needing particular attention. They aspects of health problems not addressed by conven- David Cox, wardeni may also be used to screen for psychosocial problems; tional epidemiological measures."' Here too, how- it is not clear how Correspondence to: to monitor patients' progress, pa'rticularly in relation to ever, yet useful quality of life Dr Fitzpatrick. the management of chronic illness; or to determine information will prove in assessing health needs." choice of treatment. In more formal studies of health service research BAl1 992;305:1074-7 Several studies have shown that clinicians' and quality of life assessments provide an important

1074 BMJ VOILUME 305 31 OCTOBER 1992 outcome measure.'2 Increasingly it is argued that wellbeing or sense of social support. This diversity of quality of life measures should be incorporated in experience cannot be captured in a single scale. medical audit.'3 Certainly, they provide information There are two basic types of instrument, disease that is relatively economical to gather and process for specific and generic. Disease specific instruments have audit purposes. Unlike much audit data, quality of life been developed for one disease or a narrow range of focuses on outcomes and patients' concems. One ofthe diseases. Examples include the arthritis impact few studies in Britain to examine the application of measurement scales2' and the back pain disability quality of life data to regular hospital medical audit questionnaire.22 Generic instruments are intended to gave encouraging results. A demonstration project at be applicable to a wide range of health problems. the Freeman Hospital, Newcastle, was successful in Among the more commonly used are the sickness setting up a routine systemto collect outcome measures impact profile23 and the Nottingham health profile.24 including quality of life assessments in a way that was For cancer modular format instruments have recently acceptable to clinicians and managers." The feasibility been developed, which comprise a core of general of routine use of quality of life instruments has also purpose quality of life items together with more been shown in several health care settings in the United specific instruments designed for each of the main States, although these studies suggest that consider- types of cancer.2 able attention has to be given to integrating data Though some instruments are administered by gathering into clinic and surgery routines and making clinicians or interviewers, increasingly the emphasis data readily intelligible to busy clinicians.'3 '6 has been on self completed questionnaires for each and economy of use. The quantitative information pro- vided also varies. Most give scores for the different Clinical trials dimensions of quality of life which are not intended to The best understood application of quality of life be combined but others assess dimensions that may be measures is in clinical trials, where they provide summed to provide a single score. Thus the QL index, invaluable evidence of the effects of interventions. developed for use in patients with cancer, consists of Unfortunately, many trials purporting to assess impact items on five dimensions (activity, daily living, health, of treatment on quality of life do not assess the support, and outlook) which are summed to provide a construct properly or assess a single or-limited aspect of QL index total.2 Summing disparate dimensions is not what is a multidimensional construct.' Moreover, recommended because contradictory trends for dif- multidimensional end points such as quality of life ferent aspects of quality of life are missed. present particular problems of design, analysis, and interpretation.'" These problems are considered in our second paper. The most controversial use of these Requirements of measures measures is in health economics, where quality of life RELIABILITIY seems to provide a single standard means of expressing All instruments must produce the same results on the results of health care interventions in cost-utility repeated use under the same conditions. This can be analyses. In our third paper we consider this applica- examined by test-retest reliability, although practically tion, widely associated with the technique of calcu- it may be difficult to distinguish measurement error lating quality adjusted life years (QALYS). from real changes in quality of life. Reliability is often It is important to distinguish the different applica- assessed by examining internal reliability-the degree tions of quality of life measures because instruments of agreement of items addressing equivalent concepts. that have proved useful when applied in one context Inter-rater reliability also needs to be established for may be less appropriate elsewhere. A good research interview based assessments.'" tool may be impractical for clinical uses. Generally, more attention has been given to the use of quality of VALIDITY life instruments in clinical trials than to examining The validity of quality of life measures is more their value in routine clinical care, medical audit, or difficult to assess because instruments are measuring resource allocation. an inherently subjective phenomenon. An informal but essential approach is to examine face validity by asking whether instruments seem to cover the full Definitions, dimensions, instruments range of relevant topics. This process may be enhanced Although the concept of quality of life is inherently by including people with a wide range of backgrounds subjective and definitions vary, the content of the in the assessment process-for example, doctors, various instruments shows some similarity (box 2). nurses, patients, social scientists.27 In addition in depth Early measures of patients' function or general well- descriptive surveys of the relevant patient group being such as the functional scale of the American should be consulted as these provide invaluable evi- Rheumatism Association' or the Karnofsky index,2" dence of the range of patients' experiences.23 tended to use a single score. Now many instruments A more formal approach is to examine construct reflect the multidimensionality of quality of life. A validity, which is concerned with the pattern of person may be confined to a wheel chair with little relations of the quality of life instrument with other range of movement but have a strong psychological more established measures. This may require examina- tion of the extent of agreement of quality of life scores with laboratory or clinical measures of severity of Box2: DimensionsBox of quality of life ~~~~~~~~~~guishdisease'2betweenor of thepatientabilitygroupsof the instrumentconsidered totodistin-have * Physical function-for example, mobility, selfcare different health statuses.3" Exact agreement with other * Emotional function-for example, depression, measures, such as severity of disease, is not required anxiety since that would mean that quality of life scores were * Social function-for example, intimacy, social redundant. Above all, once validity has been shown for support, social contact one purpose it cannot be assumed for all possible * Roleperformance-for example, work, housework populations or applications. For example, an instru- * Pain ment validated for rheumatoid arthritis and subjective * Other symptoms-for example, fatigue, nausea, problems in the areas of pain, mobility, and fatigue disease specific symptoms gave scores for pain that were too low when applied to ______patients with severe migraine.3' This was an artefact of

BMJ VOLUME 305 31 OCTOBER 1992 1075 the instrument's approach to measurement, which emphasised pain associated with movement, a problem Box 3: Basic requirements of quality of characteristic of locomotor disorders but not of life assessments headache. * Multidimensional construct

SENSITIVITY OF CHANGE 0 Reliability Measures of quality of life that can distinguish * Validity between patients at a point in time are not necessarily * Sensitivity to change as sensitive to changes in patients over time when * Appropriatenesstoquestionoruse repeated. However, sensitivity to change, sometimes 0 Practical utility referred to as responsiveness, is a crucial requirement for most applications, especially in clinical trials, evaluation research, or cost-utility analyses. There are effects of treatment on quality of life cannot be several reasons why instruments may be insensitive to predicted and investigators use wide ranging quality of change in quality of life. One reason is that larger more life measures to uncover unexpected problems.4" A generic instruments may include several items not "scatter-gun" approach clearly has problems, in relevant to the particular disease or treatment group.32 particular the large volume of data generated, burdens A second related factor is that instruments may include on patients, and the risk that results will prove items that assess areas that are relatively static or not a significant by chance because of the number of vari- feasible target of the health care intervention-for ables tested. example, patterns of social relationships.3 A third Established instruments can not be assumed to be problem is that quality of life measures may be subject the most appropriate. One of the instruments most to ceiling or floor effects. For patients with very poor often used to assess quality of life in rheumatoid quality of life who obtain minimum scores before arthritis-the arthritis impact measurement scales2- treatment there may be no scope to register any further does not assess fatigue, a dimension that patients deterioration.34 Finally, some quality of life instru- report as one of the most distressing consequences of ments still contain too few broad categories to be the disease.4' sensitive to subtle but important changes in patients.35 The importance of different dimensions of quality of It is not surprising therefore that when patients life varies among individuals and the instrument complete several quality of life instruments a different should reflect patients' priorities and preferences. In impression of quality of life changes over time may be one study of women with metastatic breast cancer most obtained with different measures.6 37 women regarded issues such as self care, mobility, and The absence of a standard against which to assess the family relationships as of greater concern than side measurement properties of a quality of life instrument effects of treatment.42 One approach to improving the is a particular problem when examining instruments' appropriateness of quality of life measures is to use sensitivity to change. One approach is to examine the instruments that let patients select the dimensions of associations between quality of life change scores and most concem. This baseline of quality of life scores, other changes in health status.3' The alternative is to which will vary from patient to patient, can then be examine the sensitivity and specificity of quality of life used in assessing changes over time.43 There is some change scores against an external criterion such as the evidence that this approach may produce greater view of the clinician or the patient that a significant sensitivity to change over time than conventional change has occurred.3' However, one of the most standardised measures.44 important areas for further development is in making quantitative change scores for quality of life more PRACTICALITY clinically meaningful. At present quality of life measures are most practical APPROPRIATENESS for use in clinical trials and formal evaluation studies, To ensure that the quality of life measure used is the where they are used alongside other information about most appropriate, the health problem and likely range patients, treatments, and outcomes to address fairly of impacts of the treatment being investigated need to precise questions. Even in these contexts greater effort be carefully considered. Sometimes the possible side is needed to make quality of life data clinically meaningful."' For regular use in clinical care or medical audit the more detailed and comprehensive measures of quality of life are both impractical to administer and process and hard for health professionals to interpret and incorporate into decision making. Instruments that are to be used routinely need to be briefer and simpler to use.44 Several instruments such as the Dartmouth Coop charts45 and medical outcome study short form general health survey46 have been developed -b k t nwith such practical concers in mind. Brevity may mean that potentially important information about patients' experiences is missed and the validity and responsiveness of shorter instruments need to be

_c onsiderthis issue found that patients liked completing _ _ ~~~~~~~~~suchquestionnaires and thought that the information ~~~~~~~~~~~~~o was important for their doctor to know.47 __° ~ ~~~~~~~~~~~~~~~~~~~In clinical trials many scientific questions cannot be answered properly without adequate measurement of H quality of life. It is disappointing, therefore that, even > in this best understood of applications, many trials Quality of lfe is multifactorial. Being in a wheelchair doesn 'tpreclude a satisfying life either omit quality of life measures altogether or use

1076 BMJ VOLUME 305 31 OCTOBER 1992 inappropriate assessments.4t This may stem partly 23 Bergner M, Bobbitt R, Carter W, Gilson B. The sickness impact profile: from erroneous concerns about the lack of reproduci- development and final revision of' a health status measure. Med Care 198 1;19:787-805. bility of interview or questionnaire based data com- 24 Hunt S, McEwsen J, McKenna S. Measunrng health stattos. London: Croom pared with hard conventional measures.4 In addition Helm, 1986. 25 Aaronson N, Bullinger M, Ahmedzai S. A modular approach to quality of life data from quality of life measures are unfamiliar and assessment in cancer clinical trials. Recent Rcsults Cancer Res 1988;111: lack the intuitive meaning of more established clinical 23 1-49. 26 Spitzer W, Dobson A, Hall J, Chesterman E, Levi J, Shepherd R, et al. or laboratory measures. Providing that careful atten- Measuring the quality of life of cancer patients: a concise QL-index for use tion is given to six basic issues (box 3), it is feasible to by phvsicians. ] Chrons Dis 1981;34:585-97. assess health related quality of life. Outside clinical 27 Lomas J, Pickard L, Mohide A. IPatient versus clinician item generation for qualitv of life measures. Mled Care 1987;25:764-9. trials fundamental concerns about quality of life as 28 Anderson R, Bury M. Liting trith chronic illness: the expeietnce of pati'enits aid applied to health care have emerged regarding ethical theirfanniiies. Lsondon: Unwin Hyman, 1988. 29 Guvatt G, Berman L, Townsend M, Pugsey SO, Chambers LW. A measure of and political ramifications of the use of the concept in quality of life for clinical trials in chronic lung disease. Thoirax 1987;42: relation to resource allocation. We will consider the 773-8. 30 Stewart A, Greenfield S, Hays R, Wells K, Rogers 'H, Berrv SD, et al. issues behind such concerns in a later paper. Functional status and well-being of patients with chrottic conditions. JAMA 1989:262:907-13. I Spilker B, M1olinck F, Johnston K, Simpson RL Jr, Tilson HH. Quality of life 31 Jenkinson C, Fitzpatrick R. Measurement of health status in patients with bibliography and indexes. Wfed Care 1990;28(suppl):DS 1-77. chronic illness: comparison of the Nottingham health profile and the general 2 Wilkin D, Hallam I, Doggett M. Measures of need and outconte j nrparv care. health questionnaire. Fatin Pract 1990;7:121-4. Oxford: Oxford Unixersitv Press, 1992. 32 Patrick D, Devo R. Generic and disease-specific measures in assessing health 3 Wolfe F, Cathev M. The assessment and prediction of functional disability in status and quality of life. Med Care 1 989;27:S2 17-32. rheumatoid arthn'tis. J Rhesoxatol 199 1;18:1298-306. 33 Fitzpatrick R, Ziebland S, Jenkinson C, Mowat A, Mowat A. 'The social 4 Leigh P, 'ries J. \ortality predictors among 263 patients with rheumatoid dimension of health status measures in rheumatoid arthritis. Ist Disalil Stuld arthritis. 7Rhreuartol 1991;18:1307-12. 199 113:34-7. 5 Slevin M, I'lant H, Lvnch D, Drinkwater J, Gregory WM. Who should 34 Bindman A, Keane D, Lurie N. Measuring health changcs among severely ill measurc quality of life, the doctor or the patient? Bt J Cancer 1988;57: patients. Med Care 1990;28:1142-52. 10(9-12. 35 Deyo R, Inui T. Toward clinical applications of health status measures: 6 IPearlman R, Uhlmann R. I atient and physician perceptions of patient quality sensitivity of scales to clinically important changes. Health Serz, Res of life across chronic diseases. Y Geroitool 1988;43:25-30. 984;19:275-89. 7 Rubenstein L, Calkins D, Young R, Cleary PD, Fink A, Kosecoff J, et al. 36 'I'andon P, Stander H, Schwarz R. Analysis of quality of life data from a Improving patient function: a randomised trial of functional disability randomized, placebo-controlled heart failure trial. fln Epidentioll989;42: screening. Attn IniteriMed 1989;1 11:836-42. 955-62. 8 Kazis I., Callahan L, Meenan R, IPincus T. Health status reports in the care of 37 Fitzpatrick R, Ziebland S, Jenkinson C, Mowat A, Mowat A. Importance of patients with rheumatoid arthritis. 7Clin E.pidenio(l 199(0;43:1243-53. sensitivity to change as a criterion for selecting health status measures. 9 Lohr K. Applications of health status assessment measures in clinical practice. Quzality! in Health Care 1992;1:89-93. Med Care 1992;30(suppl):MS 1-14. 38 Fitzpatrick R, Bury M, Frank A, Donnelly T. Problems in the assessment of 10 Hunt S, McEwen J, McKenna S. Measuring health status: a new tool for outcome in a back pain clinic. Int Di'sabilSttud 1987;9:161-5. clinicians and epidemiologists. ]7 R Coll Getz P3ract 1985;35:185-8. 39 Fitzpatrick R, Newman S, Lamb R, Shipley M. A comparison sf nteasures of 11 Frankel S. 'I'he epidemiology of indications. J Epideitiol Contnnilrv Health health status in rheumatoid arthritis. Br]Rheuniatol 1989;28:20)1-6. 199 1;45:257-9. 40 Croog S, Levine S, Testa M, Brown B, Bulpitt CJ, Jenkins D, et al. The effects 12 Ware J, Brook R, Rogers W, Keeler EB, Davies AR, Sherboume CD, et al. of antihypertensive therapy on quality of life. Net Enlgl lfed 1986;314: Comparison of health outcomes at a health maintenance organisation with 1657-63. those of fee for sertvice care. Lancet 1986;848:1017-22. 41 Fitzpatrick R, Ziebland S, Jenkinson C, Mowat A, Mowat A. A genertc health 13 Frater A, Costain D. Any better? Outcome measures in medical audit. BA.1J status instrument in the assessment of rheumatoid arthritis. BrDv Rhietplnatol 1992;304:5 19-20. 1 992331 :87-90. 14 Bardsley NM, Coles J. Practical experiences in auditing patient outcomes. 42 Sutherland H, L.ockwood G, Boyd N. Ratings of the importance of quality of Qlnalitxs int Healti Care 1992;1:124-30. life variables: therapeutic implications for patients with metastatic breast 15 Wasson J, Keller A, Rubenstein 1, Mavs R, Nelson E, Johnson D. Benefits cancer. ] Cliii Epide?ninls 1990;43:661-6. and obstacles of health status assessment in ambulatory settings: the 43 Guvatt G, Walter S, Norman G. Measuring change over timne: assessing the clinician's point of view. Med Care 1 992;30(suppl):MS42-9. usefulness of evaluative instruments.]7 Chron Dis 1987;40:171-8. 16 Lanskv D, Butler J, WValler F. Using health status measures in the hospital 44 Tugsell P, Bombardier C, Buchanan W, Goldsmith C, Grace E, Bennett KJ, setting: from acute care to "outcomes management." Aled Care 1992;30 ei al. Methotrexate in rheumatoid arthritis: impact on quality of life assessed (suppl):MS57-73. by traditional standard item and individualised patient preference health 17 Aaronson N. Quality of life assessment in clinical trials: methodologic issues. status questionnaires. Arch Intern Med 1990,150:59-62. Controlled Clii 7irials 1989;10: 195-208S. 45 Nelson EC, Landgraf J, Hays R, Wasson JH, Kirk JWX'. T'he functional status 18 Cox D, Fitzpatrick R, Fletcher A, Gore S, Spiegelhalter D, Jones D. Quality of of patients: how can it be measured in physicians' offices? Med Care life assessment: can we keep it simple? ]oxirnial of the Ros'al Statistical Societs' 1990;28: 1111-26. ScritsA. 1992;155:353-92. 46 Stewart A, Hays R, Ware J. 'I'he MOS short form general health survey: 19 Steinbrocker O, Traeger C, Battman R. 'Fherapeutic criteria in rheumatoid reliability and validity in a patient population. Med Care 1988;26:724-35. arthritis.JAMA 1949;140:659-62. 47 Nelson E, Berwick D. Tlhe measurement of health status in clinical practice. 20 Kamofsky D, Burchenial J. The clinical evaluation of chemotherapeutic agents Med Care 1989;27:S77-90. in cancer. In: Macleod C, ed. Etaluation of chtenrothlerapeutic agelzts. 1949; 48 Schumacher M, Olschewski M, Schulgen G. Assessment of quality of life in New York: Columbia University Press, 1949:191-205. clinical trials. Stat Med 1991;10:1915-30. 21 Meenan R, Gertman P, Mason J, Dunaif R. The arthritis impact measurement 49 Feinstein A. Clinical biostatistics. XLI. Hard science, soft data, and the scales: further investigation of a health status instrument. Arth Rheut challenges of choosing clinical variables in research. Clin Phansiacol Ther 1982;25: 1048-53. 1 977;22:485-98. 22 Roland M, Morris R. A study of the natural historv of back pain. 1. Development of a reliable and sensitive measure of disability in low back pain. Sptne 1983;8:141-4. (Accepted 9 Septeniber 1992)

ANY QUESTIONS

Should a woman taking the combined contraceptive pill who patients with already deranged liver function it would be contracts hepatitis A and has abnormal results of liverfunction advisable to avoid the combined oral contraceptive. In tests by advised to stop taking the pill? such cases the liver's metabolic and excretory function will probably also be impaired, and even with a low dose oral Yes. The latest datasheets contraindicate oral contra- contraceptive the pharmacological effects may be ceptives in women with abnormal liver function test unpredictable and greater than those in a woman with results or acute or severe chronic liver disease. Ingestion normal liver function.-P B TERRY, consultant in obstetrics of contraceptive steroids alters hepatocellular function, and gynaecology, Aberdeen and observable effects include changes in the composition 1 Association of the British Pharmaceutical Industry. The ABP1 datasheet of bile, reduced volume of biliary secretion, a rise in conmpenidium, 1991-92. London: ABPI, 1992. cholesterol concentration, and a fall in the bile acid level. 2 Neinstein IS, Katz B. Contraceptive use in the chronically ill adolescent These changes are reversible and dose related.2 Thus in female: part 1. JAdolesc Health Care 1986;7:123-33.

BMJ VOLUME 305 31 OCTOBER 1992 1077