Clinical Trials: active control vs placebo What is ethical?*

Transcription

1 Science and Engineering Ethics (2004) 10, Clinical Trials: active control vs placebo What is ethical?* Jacek Spławiński ß and Jerzy Kuźniar Þ ß National Institute of Public Health, Warsaw, Poland Þ District Hospital, No.2, Rzeszów, Poland Keywords: placebo, active-control, superiority, non-inferiority, informed consent, uncertainty principle ABSTRACT: The quest for effective medicines is very old. In modern times two important tools have been developed to evaluate efficacy of drugs: superiority and non-inferiority types of clinical trials. The former tests the null hypothesis of μ (the difference between a tested drug and comparator) 0 against μ > 0; the latter tests the null hypothesis of μ - Δ against, μ > - Δ, where Δ is the clinical difference from the comparator. In a superiority trial, a new drug is tested against a placebo; in a noninferiority trial, a new drug is tested against active treatment. In this paper, arguments are presented to show that a superiority trial against a placebo is scientifically sound but ethically unacceptable, whereas a non-inferiority trial against active treatment is ethically sound but scientifically not reliable. Switching from a superiority type of trial with placebo to a non-inferiority trial with an active-control following the latest revision of Declaration of Helsinki is in practice switching from the violation of the uncertainty principle to uncertainty of results. Given human and financial resources, it appears an academic question as to which is more unethical: to violate patients rights or to produce results without scientific value. All presented considerations lead to the conclusion that the use of a superiority trial of design with an active control instead of placebo will satisfy scientific needs, expectation of patients, and the ancient quest for effective medicines. In the era of Good (Clinical, Laboratory, Manufacture) Practice, the attention of those performing clinical trials is focused on the procedure, not always on its essence. However even the excellent performance of a trial which is not worth doing is fruitless. * An earlier version of this paper was presented at an international conference, Placebo: Its Action and Place in Health Research Today, held in Warsaw, Poland on April, Address for correspondence: Jacek Spławiński, National Institute of Public Health, 30/34 Chelmska Street, Warsaw, Poland; splawinski@il.waw.pl. Paper received, 17 July 2003: accepted, 28 November Opragen Publications, POB 54, Guildford GU1 2YF, UK. Science and Engineering Ethics, Volume 10, Issue 1,

2 J. Spławiński and J. Kuźniar Background Clinical trials have been developed because of the demand to use only efficacious drugs, i.e. these trials are tools used to distinguish between ineffective and effective medicines; strictly speaking between medicine and placebo, which may or may not have an effect itself. 1 There is a common belief that the problems of drug efficacy were discovered by two US senators, Kefauver and Harris, 2 in However, contrary to this belief, at least some people had recognized the problem much earlier. What was the driving force behind the statement:... if physician cures the broken leg, a patient will pay 5 shekels of silver...? 3 Was that not the recognition of the need for effective treatment if the payment was unconditionally connected with a cure? A person unfamiliar with the Babylonian period may think that there was only a weak relation between a broken leg and its cure since only five shekels were involved. But this is wrong thinking; at the same time the cost of a new boat was 2 shekels of silver and Abraham paid 400 shekels for the Machpela cave and land to bury his wife Sara. 4 Hammurabi and his advisors were apparently very serious about efficacy of a treatment and, in addition, about protection of the patients. They did not have ethical codes, GCP, etc. but they firmly stated in the law... if a physician causes the death of his patient... his hands will be amputated... (This was a clever regulation because a physician without hands was unable to treat another patient). The conclusion is that even from the very beginning of human civilization the need for efficacious treatment was recognized. In the twentieth century the golden tool for evaluation of new effective treatment was developed that of the randomized controlled trial versus placebo (superiority type of trial). Such high ranking is to acknowledge the essential role of comparison and the importance of randomization for balancing confounders. The discovery of the placebo as a comparator allowed the control of the psychological effect of receiving some treatment and also permitted blinding. Three main types of clinical trials were developed: superiority, non-inferiority, and equivalence which is a special case of non-inferiority trial. In the superiority trial, the hypothesis of μ (true treatment difference between the tested drug and the comparator) 0 is tested against μ > 0. In the non-inferiority trial, the hypothesis μ - is tested against μ > -, where is the clinically non-significant difference from the comparator. 5 Placebo trials and ethics The serious fight over the use of placebo is about ten years old, apparently without an end. In the present work, arguments were collected to show that perhaps the question, whether to use a placebo or an active control, is the wrong one. Instead it seems more appropriate to ask which design of clinical trial, superiority or non-inferiority, will deliver more credible results. The uses of placebo have many proponents and opponents (in this work the special case of using placebo as add-on therapy is not discussed). Proponents point out several 78 Science and Engineering Ethics, Volume 10, Issue 1, 2004

3 Clinical Trials: active control vs placebo What is ethical? advantages of placebo controlled studies. 6,7,8 The internal validation of such studies that ensures assay sensitivity is the best support for the conclusion of efficacy that does not require external validation. There is a fundamental need for evidence of assay sensitivity in trial designs other than superiority trials. 7 In other words, with placebocontrolled superiority trials, investigators are not left with the difficult question: was the new medicine really effective or was it the study that did not work? Opponents point out that placebo-controlled trials violate the uncertainty principle. As underlined by Rudawski, the patient may enter a clinical trial... when on the basis of our best judgment, an equal possibility exists that each of the compared methods of treatment will be of advantage to the patient. 9 With recent findings that placebo has no effect and even when it has it is not distinguishable from bias, 1 the researcher is left without doubts as to which arm of the trial could be beneficial to the patient: the new drug or placebo. Frankly, is a clinical trial ever undertaken by a company unless there is a good reason to suppose that an investigated drug will show some level of efficacy? Historically, placebo-controlled studies contributed to great progress in medicine, just to mention Cardiac Arrhythmia Suppression Trial (CAST) 10 or Heart and Estrogen/progestin Replacement Study (HERS) 11 results. But placebo-controlled studies, as pointed out above, violate the uncertainty principle which is a foundation of informed consent. Some authors believe that informed consent, with the explanation of all potential risks due to the lack of active treatment, supports the practice of placebocontrolled trials. 12 Passing the whole ethical burden to the patients, however, cannot be justified, since despite the best efforts to inform patients they will rarely be as well informed about their treatment options and the consequences of withholding treatment as physicians. Indeed some participants of placebo-controlled clinical trials signed their informed consent by thumbprint. 13 The conclusion is that the physicians are taking full responsibility to assure that every patient in the study should be given the best proven diagnostic and therapeutic method. 14 It seems surprising, therefore, that the most powerful agency, the US Food and Drug Administration (FDA), still considers placebo-controlled trials as the gold standard, 7 based on the argument that there are many areas in medicine where withholding accepted treatment for a short period of time (as in early escape or randomized withdrawal designs) is not leading to harm. These areas include depression 7 although it would be difficult to prove that none of the patients on placebo was harmed by withholding treatment. Moreover, it seems important to recognize, that the burden of justification as to whether a patient will be harmed or only discomforted by a placebo should not be taken solely by physicians. Researchers may be inclined to use patients as a means to a scientific end, not as the end in themselves. 14 Ethical codes have been developed to prevent such a possibility. There is another argument against placebo-controlled studies. A new drug could be better than a placebo but worse than the best available treatment. At the same time as all drugs are toxic patients would be unnecessarily exposed to medicine with a lower benefit/risk ratio than existing treatment. Such a possibility is highlighted by the Science and Engineering Ethics, Volume 10, Issue 1,

4 J. Spławiński and J. Kuźniar Committee for Proprietary Medicinal Products (CPMP) a request for a head-to-head trial before registration of the drug (alefacept) that was approved in US on the basis of placebo-controlled trial. 15 Potential harm to patients and the logic of an ethical approach make the choice of placebo-controlled study, when effective treatment exists, unacceptable. The alternative is straightforward: an active-control study. In the next section, arguments will be presented that in spite of using a control with proven efficacy, non-inferiority design is not scientifically reliable and in some cases may lead to the inferiority of a new treatment. Active control trials and ethics In active-control trials, the investigated drug is compared to the best available treatment (the add-on type of the trial is not discussed here) and therefore the uncertainty principle is not violated. In the non-inferiority type of trial with an active control, investigators are testing the null hypothesis that a new drug is worse than the active control (standard) and when they can reject the null hypothesis they accept the alternative, that the new drug is not worse than the active control. 5 There are, from an ethical point of view some questionable points. In the superiority type of trial with two arms, the null hypothesis that there is no difference between a placebo and a new drug is tested. In line with this approach, informed consent could be rationally explained to patients. In the case of non-inferiority trials, the patient is informed that the tested new drug could have the same or higher efficacy than the control, but to prove it scientists start with the assumption that it is worse. From the patient s point of view such a statement is usually taken as a request to sacrifice him/herself for the good of others. It is difficult to explain that the alternative hypothesis (i.e. that a tested drug is not worse than the control) cannot be assessed directly but only accepted if the null hypothesis is rejected on the basis of a performed clinical trial. The second important disadvantage of an active-control, non-inferiority type of trial is the issue of incentive. In any clinical trial of a new drug, the sponsors are looking forward to demonstrating the efficacy of the tested new substance. Without that demonstration, the project may be dropped, company shares go down, and a number of people lose their jobs. Therefore, in the superiority trial, all involved parties have a powerful incentive to control the precision of taking measurements, to control concomitant therapy that may affect the outcome, to control the administration of right treatments to appropriate patients, and, very importantly, to control the compliance of the patients. After all, when patients in both arms will not take the medicines, the null hypothesis will not be rejected and the superiority over placebo will not be established. The condition of the trial favors type II error, i.e. non-rejection of the null hypothesis when in fact it should be rejected. Such conditions may slow down progress but guard society against the introduction of ineffective drugs. a. The Committee for Proprietary Medicinal Products (CPMP) evaluates medicines for humans within the European Agency for the Evaluation of Medicinal Products (EMEA). 78 Science and Engineering Ethics, Volume 10, Issue 1, 2004

5 Clinical Trials: active control vs placebo What is ethical? In contrast, in the active-control, non-inferiority type of trial, the opposite might be true. 7 It has already been suggested that sloppy conduct may increase the rate of false positive conclusions in non-inferiority trials. 16 In addition, it seems that deviations from inclusion criteria, from the schedule, administration of wrong treatment and patient non-adherence to treatment are factors that make a conclusion of no difference more likely. In extreme situations when non-compliance is very likely, the hypothesis that a tested drug is inferior to the standard could easily be rejected and the alternative, of no difference between tested compounds, accepted. Since such a conclusion is in line with the financial interest of the parties involved in the trial, incentives to control the precision of the trial may be low. When poor execution of the trial will favor results expected by the sponsor, the chances that the trial will be poorly performed are increased. Non-inferiority trap The non-inferiority type of trial also has disadvantages scientifically. It is well known that these trials suffer from not having internal validity 7 and external validity is required. There is a necessity to use historic results, i.e. the standard that was proven efficacious in the past. Also the assay sensitivity could be in doubt. Probably, the intention to treat analysis should not be used and instead per protocol analysis seems more appropriate. It appears that standard methodological criteria for evaluation of placebo-superiority trials are inadequate for the interpretation of active control, noninferiority trials and no difference between a new drug and an active control does not in itself demonstrates that the new treatment is effective. In other words, switching from the superiority type of trial with placebo to the non-inferiority type with an active control following the latest revision of Declaration of Helsinki is in practice switching from the violation of the uncertainty principle to uncertainty of results. The first approach is scientifically sound but ethically unacceptable, the second approach is ethically sound but scientifically weak. The aim of a non-inferiority trial is to evaluate a point estimate and confidence interval of a new intervention in comparison with the best available standard in order to show that a new treatment is no less effective than the standard. Two sides 95% confidence intervals should lie entirely above the value of minus delta, which defines the acceptable difference with the standard. Delta should be chosen in advance and defined in the protocol. Choosing its value is difficult and has consequences for the entire trial it can determine the success or failure of the trial. When delta is chosen for a superiority trial, it is to show what clinical difference between the tested drug and placebo is desirable. Choosing delta for a non-inferiority trial is looking for the opposite: what difference is clinically irrelevant. When death is the studied variable, what difference in the death rate is meaningless? Obviously, choosing delta, is another disadvantage of the non-inferiority trial. Theoretically, the black scenario of the non-inferiority trial could be depicted as follows. Let us suppose that the lower bound of confidence interval of a tested drug lies above the border of minus delta so the drug is non-inferior to the standard and, at the Science and Engineering Ethics, Volume 10, Issue 1,

6 J. Spławiński and J. Kuźniar same time, the upper bound of the confidence interval is below zero (the difference between the tested drug and the control), and the drug obviously is inferior to the standard (Fig. 1). When the above scenario is repeated several times, each time with the point estimate for the new drug below zero, the proliferation of drugs that are less effective than active controls is inevitable. In these circumstances, within the frame of perfect procedure, the fruitless results are obtained the scenario quoted at the beginning. Figure 1. Black scenario of the non-inferiority trial. The new drug (T) tested against the standard (S). The point estimate for the new drug and confidence intervals (CI) are shown. The lower bound of CI is > - Δ (where Δ is accepted in advance, allowed difference from standard), the upper bound is < 0 (zero difference between tested and standard). H 0 an H A denote null and alternative hypotheses, respectively. H 0: H A : T S - Δ T S > - Δ - Δ 0 Active control superiority trial? There are many disadvantages connected with an active-control trial based on noninferiority design as compared to the placebo-superiority trial. It seems that difficulties and consequences for the individual and the society are not created by the use of an active control instead of placebo, but by the design of the trial. If there is a proven treatment available, an ethical investigator would not hesitate to decline to use a placebo. However, such an investigator driven by the same principles could be right to think that a clinical trial which does not provide a scientifically meaningful outcome is unethical. 14 In other words, with given human and financial resources, it is unethical to produce results without scientific value. Furthermore, investigators are more likely to look forward to finding new treatment rather better than not worse than the proven one. It seems therefore that the use of a superiority type of design with an active control, perhaps with less stringent statistics (p < 0.1?), instead of placebo will satisfy the expectations of the investigator and the patient. Such an approach will most probably assure progress in effective treatments and the ancient quest for efficacy of medicines would not be abandoned. 78 Science and Engineering Ethics, Volume 10, Issue 1, 2004

7 Clinical Trials: active control vs placebo What is ethical? REFERENCES 1. Hrobjartsson, A. & Gøtzsche, P.C. (2001) Is placebo powerless? The New England Journal of Medicine 344: Berkowitz, B.A. & Katzung, B.G. (2001) Basic & Clinical evaluation of new drugs, in: Katzung, B.G. ed. Basic & Clinical Pharmacology. Lange Medical Books/McGraw-Hill, New York: Szumowski, W. Historia Medycyny. Gebethner i Wolf, Kraków, Holy Bible Genesis XXIII Aras, G (2001) Superiority, noninferiority, equivalence, and bioequivalence revisited. Drug Information Journal 35: Emanuel, E.J. & Miller, F.G. (2001) The ethics of placebo-controlled trials a middle ground. The New England Journal of Medicine 345: Temple, R. & Ellenberg, S.S. (2000) Placebo controlled trials and active-control trials in the evaluation of new treatments. Part 1: Ethical and scientific issues. Annals of Internal Medicine 133: Rothman, K.J. & Michelis, K.B. (1994) The continuing unethical use of placebo controls. The New England Journal of Medicine 331: Rudowski, W. (1980) World Health Organization Biomedical Research Guidelines and the Conduct of Clinical Trials. Journal of Medical Ethics: The CAST Investigators (1989) Preliminary report: effect of encainide and flecainide on mortality in randomized trial of arrhythmia suppression after myocardial infarction. The New England Journal of Medicine 321: Hulley, S., Grady, D., Bush, T., Furberg, C., Riggs, B. & Hulley, S. (1989) Randomized trial of estrogen plus progestin for secondary prevention of coronary artery disease in women. Journal of American Medical Association 280: Ling, Y.-L., Chern, H.-D., Chu, M.-L. (2002) Placebo controlled clinical trials are as ethical as active controlled trials even if effective therapies exist. Drug Information Journal 36: Green, B.M., Taylor, H.R., Cupp, E.W., Murphy, R.P., White, A.T., Aziz, M.A., Schulz-Key, H., D Anna, S., Newland, H.S., Goldschmidt, L.P., Auer, C., Hansos, A.P., Freeman, S.V., Reber, E.W., Williams, N.P. (1985) Comparison of ivermectin and diethylcarbamazine in treatment of onchocerciasis. The New England Journal of Medicine 313: Mason, J.K. & McCall Smith, R.A. (1994) Law and Medical Ethics. Butterworths, London, Dublin, Edinburgh. 15. Scrip (2003) No. 2827: Koch, A. & Rohmel, J. (2002) The impact of sloppy study to conduct on noninferiority studies. Drug Information Journal 36: 3-6. Science and Engineering Ethics, Volume 10, Issue 1,