Are Placebo-Controlled Clinical Trials Ethical or Needed When Alternative Treatment Exists? FREE

The randomized clinical trial was a major methodologic breakthrough in medicine. For conditions having no effective treatment, the control regimen to which the new treatment is compared is usually either observation or administration of placebo. In some cases, untreated or placebo control groups are used even though effective treatment exists for the condition. Rothman (1) has challenged the ethics of such clinical trials; in the current issue, Temple and Ellenberg (2 - 3) address this challenge.

There is general agreement that placebo or untreated controls are not appropriate in trials of therapy for life-threatening conditions if a treatment that prolongs or preserves life is available. The disagreement centers on trials of therapy for non–life-threatening conditions, in which a delay in administration of the effective treatment is unlikely to cause permanent harm.

Critics of placebo-controlled trials or trials that include an untreated control group cite article II.3 of the Declaration of Helsinki: “In any medical study, every patient—including those of a control group, if any—should be assured of the best proven diagnostic and therapeutic methods” (4). Rothman argues that no patient should suffer unnecessary pain, even if the condition being considered is not life-threatening: “You could torture someone for 10 minutes, and he would recover, but that doesn't mean it's a great idea  … you're supposed to be acting in the best interest of the patient  …” (1). Rothman also argues that placebo-controlled trials are unnecessary and result from “a misplaced emphasis on statistical significance testing. They never know how effective any drug is because they simply ask, ‘Is it or isn't it effective? Is it significantly better than placebo?’” (1).

Temple and Ellenberg point out that, if taken literally, the requirement of the Declaration of Helsinki that all patients receive the “best proven diagnostic and therapeutic methods” would bar all clinical trials, even historically controlled ones, because when effective treatment exists the patients receiving the investigational treatment are not getting the best proven treatment. They argue that placebo-controlled trials may be ethically conducted when effective therapy exists as long as omission of such treatments would not increase risk for death or irreversible morbidity and patients are fully informed about their alternatives. Temple and Ellenberg acknowledge and describe situations in which there is ambiguity about the potential for increased risk for death or irreversible morbidity (3).

Temple and Ellenberg indicate that for many classes of drugs, medications that are widely considered “effective” cannot be demonstrated to be superior to placebo in 30% to 50% of clinical trials. “The problem may be a generally small response that varies among populations, insufficient compliance with therapy or use of concomitant medication, study samples that improve spontaneously … or that are unresponsive to the drug, or some other reason not yet recognized” (2). As a consequence, they believe that apparent equivalence of a new drug to a standard medication may not imply that the new drug is effective, because there is uncertainty about whether the standard medication is effective in that study. In such circumstances, a placebo or untreated control group is needed to demonstrate that the new medication is effective.

Temple and Ellenberg seem to be correct in their assertion that article II.3 of the Declaration of Helsinki, if taken literally, would exclude all clinical trials, including single-arm trials in which all patients receive a new treatment. Such a reading does not seem consistent with the larger body of literature on the ethics of medical experimentation. Modern writing on medical ethics identifies two key considerations: respect for patient autonomy and beneficence on the part of the physician. Respect for patient autonomy means that the patient should be informed about his or her treatment options and the pros and cons of each, and the patient's decision should then be elicited without coercion. Beneficence means that the physician should look out for the best interests of each specific patient. To evaluate these two considerations for a particular clinical trial, one must evaluate the adequacy of the informed consent process and whether there is reason to suspect compromise of beneficence in the physician–patient relationship.

Clinical trials that use placebos for indications for which effective or approved medications exist may be perfectly ethical, but close examination of whether patients are adequately informed and whether the trial is consistent with the principal of beneficence is warranted. Careful review is particularly needed for clinical trials in which the physicians entering patients have a substantial financial stake.

If an “effective” treatment exists, the scientific motivation for performing a placebo-controlled trial is strongest when the effectiveness of the available treatment is modest and inconsistent and when the new treatment is not expected to be more effective than the available treatment. If the available treatment is highly and consistently effective, a scientifically convincing therapeutic equivalence trial is feasible and a placebo-controlled trial is not needed to establish the effectiveness of the new drug. If the available treatment is modestly and inconsistently effective but the new treatment is expected to be even more effective, a placebo-controlled trial is not needed.

One must question how important it is to develop new drugs that are not more effective than an available treatment that is only moderately and inconsistently effective. Temple and Ellenberg claim that it is important to have available several drugs with different toxicity profiles. To the extent that the drugs are of limited and inconsistent effectiveness, however, it may be more worthwhile to attempt to develop more effective medications or medications that are effective for patients who are refractory to or intolerant of the standard drugs. Nevertheless, it may frequently be ethical to conduct placebo-controlled clinical trials in such situations because patients and physicians may believe that it is in the patients' interest to avoid or delay use of marginally effective standard medications. In fact, one may question whether marginally effective drugs have been adequately evaluated and whether their limits of effectiveness should have been better characterized before approval.

Serious problems arise in interpreting active-control clinical trials in which the effectiveness of the active control is limited and inconsistent. I previously introduced a Bayesian method for the design and analysis of active control clinical trials (5), and my quantitative conclusions are much in line with the qualitative concerns expressed by Temple and Ellenberg. My study demonstrated that to evaluate the effectiveness of the new drug, one requires a prior distribution of the degree of effectiveness of the active control relative to placebo and that prior distribution should reflect both average effectiveness and variability of effectiveness among studies. I recommended that the prior distribution be based on a random-effects meta-analysis of randomized clinical trials comparing the active control to placebo. If that body of data does not exist or if it does not indicate a substantial and consistent superiority of active control to placebo, an active-control clinical trial is neither adequate nor appropriate for establishing the effectiveness of the new drug. The usual statistical methods for planning and analyzing active control trials are often inadequate and may encourage misleading publications because they ignore the degree of uncertainty in the estimate of effectiveness of active control relative to placebo.

Rothman's claims that concerns about the active control design are based on an inappropriate fixation on statistical significance testing overlook the real problems of interpreting many active-control trials. Nevertheless, placebo-controlled clinical trials proposed in situations in which effective treatments exist should receive careful scrutiny with regard to 1) whether the effectiveness of the active control is sufficient that interpretable active-control trials could be conducted and 2) whether the trial can be conducted in a manner consistent with principles of patient autonomy and physician beneficence.

Richard Simon, DSc

National Cancer Institute; Bethesda, MD 20892

Requests for Single Reprints: Richard Simon, DSc, National Cancer Institute, 6130 Executive Boulevard, Room 739, Rockville, MD 20892.

Correction: Ethics of Placebo-Controlled Trials. In the editorial by Simon published in the 19 September issue (1), the source of two quotations attributed to Dr. Kenneth Rothman was identified incorrectly. These quotations were “You could torture someone for 10 minutes, and he would recover, but that doesn't mean it's a great idea  … you're supposed to be acting in the best interest of the patient”, and “[placebo-controlled trials result from] a misplaced emphasis on statistical significance testing. They never know how effective any drug is because they simply ask, ‘Is it or isn't it effective? Is it significantly better than placebo? ’”. Both quotations were actually taken from the Current Clinical Issues article by Tom Reynolds that also appeared in the 19 September issue of the journal (2), rather than from the article by Rothman and Michels (3) cited in the Simon editorial.

References

1. Simon R. Are placebo-controlled trials ethical or needed when alternative treatment exists? [Editorial] Ann Intern Med. 2000; 133:474-5.

2. Reynolds T. The ethics of placebo-controlled trials. Ann Intern Med. 2000; 133:491-2.

3. Rothman KJ, Michels KB. The continuing unethical use of placebo controls. N Engl J Med. 1994; 331:394-8.