Twenty years ago, the American Journal of Epidemiology published David Sackett's brief description of. clinical epidemiology and its practitioners [1]. This commentary was a useful focal point for an emerging discipline. By 1983, with clinical epidemiology already thriving in many academic medical centres, Walter Holland called into question both the term, 'clinical epidemiology', and the nature of the discipline [2]. More recently, clinical epidemiology has drawn strong criticism from John Last, a noted academician whose contributions include the editorship of the Maxcy-Rosenau Textbook of Public Health. Writing in the Journal of Public Health Policy in 1988 [3], Last referred to the 'uncritical enthusiasm' for clinical epidemiology in medical schools as 'a danger to health', and staked. a claim to the term 'epidemiology' as appropriate only to the description of what classical or population epidemiologists do. Faced with such views, practitioners and proponents of clinical epidemiology can respond in three ways. They can ignore the criticism, and go on about their business. They can reaffirm their differences and resort to defensive rhetoric. Or, the critique can become an opportunity for reflection about the nature of clinical epidemiology and its relations with sister disciplines in modem medical schools. The latter course is followed here by four physicians who-despite diverse backgrounds and interests-all consider their work to be in the field of clinical epidemiology.

AS A SOCIETY, we are in conflict with ourselves about the cost of health care. 1 On one hand, we want the best care possible, regardless of cost. On the other hand, we are not willing to pay the cost of the care we want. Our conflict parallels a flaw in the medical marketplace. An essential condition for achieving an equilibrium between cost and value is that the two must be connected through decisions. When people decide what products and services (goods) they want, they must not only see the value they will receive, but they mast also be responsible for the costs. Because of a variety of features of the medical marketplace-most notably third-party coverage, third-party advice, and uncertainty about outcomes-the required connection between value and cost is severed. The result is what we see. One side of our collective mind demands more services while the other side cries that costs are too high. Resolving our conflict will require connecting value to cost. An essential step in accomplishing this will be to incorporate costs in practice policies. 1 As controversial as that thought might seem (the great majority of practice policies currently do not take costs into account except in the most rudimentary way), arriving at the conclusion is the easy part. A more difficult issue is how to implement the goal of connecting value to cost. Suppose we agree that, in principle, costs should be considered when practice policies are designed, and that an activity should be recommended and covered only if its health outcomes (benefits minus hanns) are deemed to be worth its costs. The next questions are, Who should do the deeming? What should the deemers be asked?

A model was developed for a simple clinical trial in which graders had defined probabilities of misclassifying pathologic material to disease present or absent. The authors compared Kappa between graders, and efficiency and bias in the clinical trial in the presence of misclassification. Though related to bias and efficiency, Kappa did not predict these two statistics well. These results pertain generally to evaluation of systems for encoding medical information, and the relevance of Kappa in determining whether such systems are ready for use in comparative studies. The authors conclude that, by itself, Kappa is not informative Enough to evaluate the appropriateness of a grading scheme for comparative studies. Additional, and perhaps difficult, questions must be addressed for such evaluation.

The purpose of this paper is to examine what is meant by a ralid measure of health. Guyatt, Kirshner and Jaeschke propose that health tests should be designed so as to have one of several kinds of validity: 'longitudinal construct validity' for those which are used for longitudinal research designs, and 'cross-sectional construct validity' for those which are used for cross-sectional designs. Williams and Naylor argue that this approach to test classification and validation confuses what a test purports to measure with the purpose for which it is used, and that some tests have multiple uses. A review of the meanings of validity in the psychologica test literature shows that both sets of authors use the term validity in an idiosyncratic way. Although the use of a test (evaluated by content validity) should not be conflated with whether the test actually measures a specified construct (evaluated by construct validity);' if health is actually made up of several constructs (as suggested in Hyland's interactional model) then there may be an association between types of construct and types of purpose. Evidence is reviewed that people make several, independent judgements about their health: cognitive perceptions of health problems are likely to be more sensitive to change in a longitudinal research design. whereas emotional evaluations of health provide less bias in cross-sectional designs. Thus. a classification of health measures in terms of the purpose of the test may parallel a classification in terms of what tests purport to measure.

Substitution of graphic representation for extensive lists of numerical statistical data is highly desirable by both editors and readers of medical journals, faced with an exploding abundance of contemporary medical literature. A novel graphic tool. the 'double-ring diagram', is described herein which permits visual representation of information regarding certain statistical variables used to describe the performance of a test or physical sign in the diagnosis of a disease. The diagram is relatively easy to construct on the basis of a number of primary data such as the prevalence and the true positive, true negative. false positive and false negative test results. These values are reflected in the diagram along with the values of other statistical variables derived from them. such as the sensitivity. specificity, predictive values for positive and negative test result. and accuracy. This diagram may be useful in visualizing a test's performance and facilitating visual comparison of performance of two or more tests.

Receiver operating characteristic (ROC) curves have been frequently used to compare probability models applied to medical problems. Though the curves are a measure of the discriminatory power of a model. they do not reflect the model's accuracy. A supplementary accuracy curve is derived which will be coincident with the ROC curve if the model is reliable. will be above the ROC curve if the model's probabilities are too high or below if they are too low. A clinical example of this new graphical presentation is given.

The evaluation of diagnostic tests attempts to obtain one or more statistical parameters which can indicate the intrinsic diagnostic utility of a test. Sensitivity. specificity and predictive value are not appropriate for this use. The likelihood ratio has been proposed as a useful measure when using a test to diagnose one of two disease states (e.g. disease present or absent). In this paper, we generalize the likelihood ratio concept to a situation in which the goal is to diagnose one of several non-overlapping disease states. A formula is derived to determine the post-test probability of a specific disease state. The post-test odds are shown to be related to the pre-test odds of a disease and to the usual likelihood ratios derived from considering the diagnosis between the target diagnosis and each alternate in turn. Hence, likelihood ratios derived from comparing pairs of diseases can be used to determine test utility in a multiple disease diagnostic situation.

Slightly elevated urinary albumin excretion rate (microalhuminuria) is a marker of early diabectic nephropathy, but it is unclear if the established definition of microalbuminuria ($20-200{\mu}g/min$) is correct for children and adolescents. We investigated th.: albumin excretion rate, albumin/creatinine ratio and urinary albumin concentration in 150 healthy schoolchildren and adolescents to (a) obtain a reterence value for albumin excretion rate, (b) relate albumin excretion to pubertal stages and (c) evaluate albumin/creatinine ratio and morning albumin concentration as screening methods for elevated albumin excretion rate. Albumin concentration was measured by immunoturbidimetry in timed overnight urine samples. The albumin excretion showed a skewed distribution (geometric mean $3.2{\mu}g/min$, 95 percentile ($15.1{\mu}g/min$). In girls. a peak in the albumin excretion rate was found at the pubertal stage 4 (Tanner) and in boys at stage 5. Albumin/creatinine ratio of 2.5 mg/mmol as a screening level for elevated albumin excretion ($15{\mu}g/min$) showed a high positive (0.88) and negative (0.99) predictive value.

Single serologic tests may occasionally influence clinicians in making diagnoses. The antistreptolysin O (ASO) test is a frequently used tool for detecting recent Streptococcus pyogenes infection and is helpful in the diagnosis of diseases like rheumatic fever. Using data from a 1989 prospective study of 600 healthy male military recruits, in which 43% experienced S. pyogenes upper respiratory tract infection (2-dilution rise in ASO), this report compared two methods of interpreting a single ASO titer. Using the 'upper limit of normal' (80 percentile) method, recruits with an ASO titer of greater than 400 showed evidence of recent S. pyogenes infection. This method had a sensitivity and specificity of only 65.9 and 81.9% respectively. In contrast to the 'yes-no'. dichotomy of the 'upper limit of normal' method. the likelihood ratio method statistics were ASO value specific, more consistent with clinical judgment, and better emphasized the caution clinicians must use in interpreting a single ASO test.

B-mode ultrasound is being used to assess carotid atherosclerosis in epidemiological studies and clinical trials. Recently the interpretation of measurements made from ultrasound images has been questioned. This study examines the anatomical correlates of B-mode ultrasound of carotid arteries in vitro and in situ in cadavers. Twenty-seven segments of human carotid artery were collected at autopsy. pressure perfusion fixed in buffered 2.5% gluteraldehyde and 4% paraformaldehyde and imaged using an ATL UM-8 (10 MHz single crystal mechanical probe). Each artery was then frozen, sectioned and stained with van Gieson or elastin van Gieson. The thickness of the intima. media and adventitia were measured 'to an accuracy of 0.01 mm from histological sections using a calibrated eye graticule on a light microscope. Shrinkage artifact induced by histological preparation was determined to be 7.8%. Digitised ultra sound images of the artery wall were analysed off-line. The distance from the leading edge of the first interface ($LE_{1}$) to the leading edge of the second interface ($LE_2$) was measured using a dedicated programme. $LE_{1}$-$LE_{2}$ measurements were correlated against histological measurements corrected for shrinkage. Mean values for the far wall were: ultra sound $LE_{1}$-$LE_{2}$ (0.97 mm, S.D. 0.26), total wall thickness (1.05 mm, S.D. 0.37), adventitia (0.35 mm, S.D. 0.16), media (0.61 mm, S.D. 0.18). intima (0.09 mm, S.D. 0.13). Ultrasound measurements corresponded best with total wall thickness, rather than elastin or the intima-media complex. Excision of part of the intima plus media or removal of the adventitia resulted in a corresponding decrease in the $LE_{1}$-$LE_{2}$ distance of the B-mode image. Furthermore. increased wall thickness due to intimal atherosclerotic thickening correlated well with $LE_{1}$-$LE_{2}$ distance of the B-mode images. B-mode images obtained from the carotid arteries in situ in four cadavers also corresponded best with total wall thickness measured from histological sections and not with the thickness of the intima plus media. In conclusion, the $LE_{1}$-$LE_{2}$ distance measured on B-mode images of the carotid artery best represents total wall thickness of intima plus media plus adventitia and not intima plus media alone.

A slightly modified version of the Quality-of-Life after Myocardial Infarction (QLMI) questionnaire developed by Oldridge and colleagues was applied in a self-administered mode to patients with suspected acute myocardial infarction (AMI) in a randomized controlled trial of secondary prevention. Acceptability of the questionnaire was good, with 93% of responders answering all items. Factor analysis suggested three quality-of-life (QL) dimensions which we called 'emotional', 'physical' and 'social'. These differed somewhat from the dimensions proposed by Oldtidge and colleagues. However, a sensitivity analysis showed relative invariance of results to weighting schemes. Scores on our three dimensions were responsive to differences between the treatment groups, and demonstrated construct validity based on associations between the measured QL and variables expected to affect QL. We conclude that the QLMI questionnaire has good potential as an instrument for assessing QL in post-AMI patients and that it can be successfully self-administered.

To an increasing extent, case-control studies are being undertaken to determine if use of early detection procedures is associated with reduced mortality from cancer. The authors recommend that in such studies the analysis focus on screening activity in cases that occurs during an interval prior to diagnosis in which the cancer is believed to be detectable and still curable and to a corresponding time period in controls. This approach places a heavy burden on the investigator to estimate accurately the period during which the tumor ought to be detectable using the test in question and to sort out reliably tests done in response to signs or symptoms of the cancer from screening tests per se. Nonetheless, the authors feel that it offers the greatest ability to discern a true benefit of screening, while minimizing the numerous potential biases that can be present in this type of study.

The Medical Research Council has for some years encouraged collaborative clinical trials in leukaemia and other cancers, reporting the results in the medical literature. One unreported result which deserves such publication is the development of the expertise to design and analyse such trials. This report was prepared by a group of British and American statisticians, but it is intended for people without any statistical expertise. Part!, which appears in this issue, discusses the design of such trials; Part II, which will appear separately in the January 1977 issue of the Journal, gives full instructions for the statistical analysis of such trials by means of life tables and the logrank test, including a worked example, and discusses the interpretation of trial results, including brief reports of particular trials. Both parts of this report are relevant to all clinical trials which study time to death, and would be equally relevant to clinical trials which study time to other particular classes of untoward event: first stroke, perhaps, or first relapse, metastasis, disease recurrence, thrombosis, transplant rejection, or death from a particular cause. Part I, in this issue, collects together ideas that have mostly already appeared in the medical literature, but Part II, next month, is the first simple account yet published for non-statistical physicians of how to analyse efficiently data from clinical trials of survival duration. Such trials include the majority of all clinical trials of cancer therapy; in cancer trials, however, it may be preferable to use these statistical methods to study time to local recurrence of tumour, or to study time to detectable metastatic spread, in addition to studying total survival. Solid tumours can be staged at diagnosis; if this, or any other available information in some other disease is an important determinant of outcome, it can be used to make the overall logrank test for the whole heterogeneous trial population more sensitive, and more intuitively satisfactory, for it will then only be necessary to compare like with like, and not, by chance, Stage I with Stage III.

This paper proposes a new method for planning randomized clinical trials. This method is especially suited to comparison of a best standard or control treatment with an experimental treatment. Patients are allocated into two groups by a random or chance mechanism. Patients In the first group receive standard treatment; those in the second group are asked if they will accept the experimental therapy; if they decline. they receive the best standard treatment. In the analyses of results. all those in the second group. regardless of treatment. are compared with those in the first group. Any loss of statistical efficiency can be overcome by Increased numbers. This experimental plan is indeed a randomized clinical trial and has the advantage' that, before providing consent, a patient will know whether an experimental treatment is to be used.

A brief review is presented of the strengths and weaknesses of historical controls, data banks, and randomized trials in the evaluation of clinical treatment. Use of pre randomized versus postrandomized informed consent is discussed. Recommendations are made for the development of an appropriate clinical research strategy.

Interim analyses of randomized trials enable investigators to make more efficient use of limited research resources and to satisfy ethical requirements that a regimen be discontinued as soon as it has been established to have an inferior efficacy/toxicity profile. Unfortunately. the integrity and credibility of these trials can be compromised if inappropriate procedures are used in monitoring interim data. 'In this paper we discuss how group sequential designs provide useful guidelines that enable one to satisfy the valid objectives of interim monitoring while avoiding undesirable consequences, and we consider how flexible one can be in the way such designs are implemented. We also provide motivation for the role of data-monitoring committees in preserving study integrity and credibility in either government- or industry-sponsored trials. In our view. these committees should have multidisciplinary representation and membership limited to individuals free of apparent significant conflict of interest, and ideally should be the only individuals to whom the data analysis center provides interim results on relative efficacy of treatment regimens. Finally. we discuss some important practical issues such as estimation following group sequential testing, anal ysis of secondary outcomes after using a group sequential design applied to a primary outcome, early stopping of negative trials. and the role of administrative analyses.

At the conclusion of a double-blinded, randomized clinical trial of propranolol hydrochloride, but before un blinding, the patients and clinic personnel were asked to guess the treatment group assignment of each patient. While 79.9% of the patients receiving propranolol correctly Identified their treatment group assignment, 57.2% of the patients receiving placebo incorrectly guessed that they were aiso in the propranolol group. No specific mechanism was identified to explain why more patients receiving propranolol were better able to guess' their group assignment. Clinic physicians correctly identified the group assignment of 69.9% of the patients receiving propranolol and 68.8% of the patients receiving placebo. Clinic coordinators correctly identified the group assignment of 67.1% of the patients receiving propranolol and 70.6% of the patients receiving placebo. For clinic personnel, heart rate level and heart rate change seem to be the mechanisms employed to Identify their patients' treatment assignment.

Over the last decade, considerable interest has focused on sample size estimation in the design of clinical trials. The resulting literature is scattered over many textbooks and journals. This paper presents these methods in a single review and comments on their application in practice.

Phase I of the Trials of Hypertension Prevention was a multicenter, randomized, controlled trial designed to determine the efficacy of seven nonpharmacologic interventions in reducing blood pressure among persons with high-normal diastolic blood pressure. the initial goal for recruitment was to enroll 2,100 participants over a nine-month period. The yield from the first screening visit to randomization was 13% overall, with clinic-specific yields ranging from 4.5% to 31.7%. After five months of recruitment, approximately 60% of the goal for that point in the recruitment timetable had been randomized. Clinical centers falling short of their goals at that time altered their recruitment strategies and intensified their efforts, and centers that had exceeded their goals recruited additional participants. As a result, 2,182 participants, or 104\% of the goal for recruitment, were randomized over a 13-month period. Those clinics using a cohort, or wave, type of enrollment were most successful in achieving their recruitment goals within the prescribed timetable.

To promote the inclusion of quality of life (QOL) end-points In clinical research on cancer, the National Cancer Institute (USA) sponsored a workshop on QOL assessment In cancer clinical trials In July, 1990. Experts In clinical trials and QOL research formed four working groups to Identify current areas of cancer treatment In which QOL end-points are most Important; to discuss methodologic problems In QOL assessment; to address common problems In Implementing clinical studies with QOL end-points; and to consider statistical Issues In design, Implementation, and data analysis. Recommendations made by the working groups are summarized in this paper.

Compared to other variables being considered in therapeutics, patient compliance has long been given minor attention although it affects every aspect of medical care; Limited methodology of compliance measurement, in particular, has hampered major progress in research, and pre-conceptions have been reiterated. However, there is a recent surge in interest derived from new data revealed by reliable methods, i.g. continuous medication (compliance) monitoring. The visualization of dynamics in drug regimen compliance over time offers unique opportunities, both to scientific drug evaluation and therapeutics in medical practice. New perspectives related to the descriptive and explanatory side of the problem are outlined by giving examples from various therapeutic fields.

Multicenter controlled clinical trials require It synthesis of the scientific method and the precepts of modern management. The management tasks associated with these studies are akin to those found in other kinds of complex corporate endeavors. It is recommended that clinical investigators become more knowledgeable about management concepts and methods and management specialists be given a major role in the planning and conduct of large-scale clinical trials.