Clinical trials: a summary

Journal of Internal Medicine 2004; 256: 284–287 CLINICAL TRIALS Clinical trials: a summary L. WILHELMSEN1, P. HELD1 & H. WEDEL2 From the 1Section of Preventive Cardiology, The Cardiovascular Institute, Göteborg University, Göteborg; and 2The Nordic School for Public Health, Göteborg, Sweden Abstract. Wilhelmsen L, Held P, Wedel H (Göteborg University, Göteborg; and The Nordic School for Public Health, Göteborg, Sweden). Clinical trials: a summary (Clinical Trials). J Intern Med 2004; 256: 284–287. Several well-known, experienced authors have contributed to a series of interesting articles on clinical trials in the Journal of Internal Medicine during the years 2003 and 2004. With this concluding article we would like to highlight some of the views given by these authors and add some personal views. The paper by Devereaux and Yusuf [1] mention that the number of clinical trials has increased tremendously between the late 1980s and 1998 – from 5000 to 12 000 publications per year, and that this growth relates to the acceptance that randomized trials provide the best means to assess the efficacy of an intervention. Randomized trials have been placed at the top of the therapeutic hierarchy. One important factor being that bias is eliminated due to the randomization process. These authors emphasize the importance of systematic overviews and large enough trials that are able to convince the medical profession as early as possible thereby potentially saving hundreds of thousands of lives. This was not the case with, for example, thrombolytic treatment in acute myocardial infarction, where too small trials resulted in a long delay until convincing results became available. Methodological deficiencies in clinical trials are common and can have profound effects. This is exemplified by errors at randomization including insufficient concealment of the treatment group. The importance of blindness to treatment allocation, whenever possible, is also stressed. If that is not the case, indisputable clinical end-points like death have to be used. Furthermore, complete follow-up of all patients is very important for the interpretation of the result and the primary analyses should normally be according to the intention-to-treat principle. The most reliable estimate of the treatment effect is the overall result; great caution should be exercised when interpreting subgroup result that deviates from this. In his paper Califf [2] stresses some important points for future clinical trials. In keeping with the previous authors he emphasizes the importance of large trials, and points to the need for extensive networks of academic centres as well as clinical practice, to be able to recruit many enough patients. He underlines the extremely rare occurrence of qualitative interactions in subgroups, whereas quantitative interactions regarding benefit or harm are more common and mostly relates to differences in baseline characteristics. In this context he stresses the importance of developing and implementing common data standards for comparisons across studies. 284 Ó 2004 Blackwell Publishing Ltd CLINICAL TRIALS: A SUMMARY Biomarkers have often mistakenly been positioned as surrogate end-points. In future trials he advocates the use of an array of biomarkers to screen for possible positive treatment effects. Even when such effects are detected promising therapies have to be tested in clinical trials with clinical end-points, because there is no substitute for such outcomes. Another important point is that trials have to be tailored to the relevant length of treatment in question; risks and benefits may differ with length of time. The problem of class effects is also discussed. Head-to-head comparisons of similar therapies will be very expensive, and the use of routine clinical data to compare the actions will be necessary. Trials of similar drugs may in future have to use noninferiority trials to compare benefits and risks. New and beneficial treatments seldom save money, at least not in the short run and usually not for the health provider, and regarding preventive therapies the long-term costs may in fact increase because people live longer and use more healthcare resources. Hopefully this would not deny the use of the therapy. Amongst several thoughtful comments a series of interesting examples are given in Sleight’s paper on ‘Where are clinical trials going? Society and clinical trials’ [3]. Amongst them is the early ‘randomization’ to streptomycin versus nothing when the drug supply was insufficient for treating all in need of it. He criticizes media and politicians who have allayed public anxiety thereby undermining the trust in doctors – some of whom are involved in clinical trials. Lengthy informed consent forms are questioned because patients cannot remember the content of them especially not when presented in the very early phase of a serious disease as for example an acute myocardial infarction. A short form supplemented by a fuller discussion with relatives is recommended. An important discussion relates to the partnership between industry and academia. The problem of conflicts of interest is also discussed, both economic and academic conflicts may be real and the latter ones can be very strong. There are examples where reviewers of scientific articles may have delayed or even prevented publication of rival data. Presenters of clinical trial results may give a more positive emphasis on some trial data, or a subgroup analysis, than is relevant from the overall results. 285 The important work of safety committees is discussed. This author does not support the public release of interim data from a trial that has been advocated by some authors, because it can seriously harm the trial result. So far it has been a serious lack of published experience from safety committees, which can be of help for future safety work. (A book on such experiences is now under publication by one of the contributors to the present series of papers – DL. DeMets.) The importance of economic evaluation of a new treatment is also stressed. The translation of clinical trials into clinical practice is covered by Julian [4]. Guidelines have become popular, but the problem is that evidence of the quality of the guidelines is not always evident to the reader, who will be dependent on the expert’s opinions. Guidelines also have to be updated very often – at least annually. Trial populations are usually based on specific inclusion criteria, and there is often a problem for the clinician to know if his/her patient belongs to those who would benefit from the treatment in question. The surveys to investigate the use of a certain treatment is discussed, and criticized because of selective response of compliant therapists to these surveys. Hospitals and practices included in such surveys tend to be more from urban areas and more sophisticated. However, some surveys indicate that compliance with recommendations have improved during recent years. The reason for poor compliance to guidelines has been summarized in five ‘I’s: Ignorance, Incredulity (uncertainty), Inertia, Incompetence and Indigence. All these are commented upon by important examples. In his article of planning of trials Boissel [5] gives a comprehensive background for the development of the trial protocol and the trial conduct, starting with the important prerequisite that the hypothesis has to be fixed prior to the start of the trial. He stresses the low level of evidence and difficulties with statistical inference associated with post hoc analyses. Choice of proper control treatment is not always an easy task; traditions may vary between countries, guidelines may differ and the available evidence from previous trials may not be impressive. The problem of lower end-point rate in the trial than according to nontrial experience because of selection Ó 2004 Blackwell Publishing Ltd Journal of Internal Medicine 256: 284–287 286 L . W I L H E L M S E N et al. of patients in the trial with usually lower end-point incidence should cause action. The power of the trial has to be kept high enough. Caution is raised against the use of combined outcomes, such as a mix of death, recurrent myocardial infarction, need for angioplasty or coronary bypass surgery as interpretation may be difficult. A point is made about the difference between the target population and the studied population with an illustrative example from one trial showing marked differences between patients considered for inclusion in the trial and those who were finally included. The quality assurance in a clinical trial is often expensive but extremely important in relation to the cost of poor quality. In his paper ‘On the evaluation of statistical methods as applied to clinical trials’ Machin [6] describes some milestones that have impacted directly on the day-to-day conduct of clinical trials. One important example is the Cox proportional hazards model, which in turn needed the modern expansion in data-processing capabilities. The important papers by Peto et al. published in 1976 [7] and 1977 [8] are highlighted. Key items were the Kaplan–Meier estimate of the survival curve, the log-rank test and the stratified log-rank, and a computer program facilitated the use of these methods. Another important development was the development of tables and computer programs for sample sizes in clinical trials. This was very important in the light of the many far too small trials that were conducted and published with no possibility to decide whether the treatment was beneficial or not. Despite being published in a theoretical statistical journal the Cox proportional hazards regression model is also cited as probably the most cited paper in medical literature, and has definitely been of great importance for medicine. Special methods are available when there are competing risks that may be especially important when elderly subjects are studied. Bayesian methods are advocated for future trial analyses. Another point relates to the complexity of multilevel models. This is when in addition to a major outcome there is one or more other important outcome. An example is wound healing after burns where also repeated measurements of pain are carried out during the healing process. A second paper on ‘Statistical issues in clinical trials’ is presented by DeMets [9]. A different approach to the previous statistical paper is used. First, it is stated, ‘no clever analysis can rescue a flawed design or poorly conducted trial’. Five issues are discussed: intention to treat (ITT), surrogate outcome measures, subgroup analyses, missing data and noninferiority trials. The ITT principle means that all patients randomized into a trial are to be accounted for in the primary analysis and all primary events during follow-up are to be accounted. If these aspects are not adhered to there may easily be bias, and there are no possibilities to take this into account. Because perfect compliance is difficult and exclusion of noncompliant patients may introduce bias, it is important to keep compliance high by various means, and to increase sample size to compensate for nonadherence. Surrogate measures should be clinically relevant, sensitive to the intervention effect, be able to be ascertained in all patients and resistant to bias. These requirements are very difficult to comply with, and in fact there are few trials that have used them successfully. Missing data may not be missing at random, and no statistical analysis can ever adequately adjust for missing data. Thus, all researchers must do everything possible to eliminate missing data. Noninferiority trials means that the new therapy may be easier to administer, better tolerated, less toxic, or less expensive such there is a beneficial trade-off if not much in treatment benefit is given up. There are several challenges in this context: (i) a very high trial quality is needed, (ii) there must be a very strongly effective control intervention, and (iii) the margin of difference is arbitrary, and can be subject to criticism afterwards. ‘Publication of clinical trials: accountability and accessibility’, by Tumber and Dickersin [10] is a comprehensive review with 142 useful references. Standards for reporting trials are strongly recommended as well as publication of both positive and negative trial results. Failure to publish is a form of scientific misconduct! Trials with negative results that are funded by industry are less likely to be published, but even the US Government has prevented public discussion of ‘politically controversial’ results. Authorship is also discussed, and honorary authorship is criticized and the authors cite a paper showing that 19% of peer-reviewed articles had Ó 2004 Blackwell Publishing Ltd Journal of Internal Medicine 256: 284–287 CLINICAL TRIALS: A SUMMARY honorary authors! Duplicate publication of the same article or identical data is a form of self-plagiarism and warrants concern. The problem of conflict of interest is covered with several up-to-date references. Access to trial design and results are becoming better, and the article reviews various ways to retrieve such information. There is a plea for electronic publication that will most probably increase in future. However, this may cause economic problems for the publishers, or authors if they will be charged for printing. The article finishes with a series of recommendations regarding dissemination of clinical trials, registration at inception with a unique identification number, reporting of all trials, and report of contributors and potential conflicts of interest. Solutions to paying for universal dissemination has to be explored, as well as archiving of electronic reports. Guest editors’ comments The development of methodology of clinical trials has been tremendous over the recent decades and has resulted in a large number of quality trials. This has translated into major changes in medical treatment and is the basis for the development of ‘evidencebased medicine’. The impact of the development and acceptance of clinical trial results have thus been tremendous and will hopefully continue to be a major factor in the development of medicine to benefit patients. This development has also been connected with the development of a more detailed framework of rules and guidelines, which in turn lead to more complicated trial structures with increasing costs. Some of this has been necessary and good but there are worrying aspects. New rules and regulations have made it increasingly difficult for single investigators or institutions to manage the conduct and costs of trials. There is no question that the methodological improvements are important for good quality, but the complicated framework may also divert the initiatives away from academic institutions to industry. One example is the new European Union clinical trials directive active from 1 May 2004. These rules are all individually aimed at securing quality, 287 control and safety of trials but there are aspects where the logic is difficult to see. One example is the new involvement of the local ethical committees in the assessment of serious adverse events in ongoing trials, when also this is one of the main tasks of the safety committee of the trial. Other examples exist and the sum of all well intended rules and regulations risk hampering new important and innovative research. We must make sure that bright ideas that may benefit patients are taken care of and are tested rigorously in modern clinical trials. The tools are there. Conflict of interest statement No conflict of interest was declared. References 1 2 3 4 5 6 7 8 9 10 Devereaux PJ, Yusuf S. The evolution of the randomised controlled trial and its role in evidence-based decision making. J Intern Med 2003; 254: 105–13. Califf RM. Issues facing clinical trials in the future. J Intern Med 2003; 254: 426–33. Sleight P. Where are clinical trials going? Society and clinical trials. J Intern Med 2004; 255: 151–8. Julian DG. Translation of clinical trials into clinical practice. J Intern Med 2004; 255: 309–16. Boissel JP. Planning of clinical trials. J Intern Med 2004; 255: 427–38. Machin D. On the evolution of statistical methods as applied to clinical trials. J Intern Med 2004; 255: 521–8. Peto R, Pike MC, Armitage P et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient (I) Introduction and design. Br J Cancer 1976; 34: 585–612. Peto R, Pike MC, Armitage P et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient (II) Analysis and examples. Br J Cancer 1977; 35: 1–39. DeMets DL. Statistical issues in interpreting clinical trials. J Intern Med 2004; 255: 529–37. Berni Tumber M, Dickersin K. Publication of clinical trials: accountability and accessibility. J Intern Med 2004; 256: 271–283. Correspondence: Professor Lars Wilhelmsen, Section of Preventive Cardiology, The Cardiovascular Institute, Drakegatan 6, SE412 50 Göteborg, Sweden. (fax: 46-31- 703 1890; e-mail: [email protected]). Ó 2004 Blackwell Publishing Ltd Journal of Internal Medicine 256: 284–287