Placebo-Controlled Trials

• Controlled clinical trials
• Placebos
• Ethics
• medical

IN RESPONSE:

Dr. Djulbegovic argues that the principle of equipoise must be of particular concern in placebo-controlled trials because placebo-treated patients may be disadvantaged and should not “sacrifice themselves for the benefit of others.” It seems important, as we emphasized in our papers, to distinguish between studies of treatments for serious illness and studies of symptomatic treatments. Exposure to placebo in the former requires genuine uncertainty about the outcome. Exposure to placebo in a trial of headache, anxiety, or seasonal allergy, however, cannot reasonably be said to constitute “sacrifice” of oneself. It is, at worst, the sort of choice to defer or omit therapy that people with symptomatic conditions make every day. Moreover, at least in the first trial carried out, there is, in fact, equipoise—uncertainty as to whether the drug or placebo will be superior. There may be knowledge that some other treatment is effective, but that is a different question. It should, however, be appreciated that during treatment development, studies are replicated and there are often multiple placebo-controlled trials of various doses and regimens in diverse settings and populations. These studies help define safe and effective use of the drug, but the favorable results of these trials are at least strongly suspect. Nonetheless, despite possible lack of equipoise, such trials have been conducted, have been considered ethical, and are valuable. In contrast, if in the course of drug development it becomes known that a treatment enhances survival or decreases significant morbidity, relevant equipoise no longer exists and another placebo-controlled trial cannot be conducted.

Dr. Markman asks whether placebo controls are justified where available therapy is “known to prevent reversible but highly clinically relevant morbidity,” such as “emesis caused by moderately emetogenic cancer chemotherapy.” He specifically questions the conduct of a placebo-controlled trial of ondansetron in preventing emesis after cyclophosphamide regimen–induced emesis, since dexamethasone had been shown to be effective in that population. He asks whether patients in the trial gave truly informed consent and why they should have endured emesis when an inexpensive existing treatment was available. In this setting, Dr. Markman suggests that the appropriate comparator was dexamethasone, not placebo, but does not state whether these should have been noninferiority trials or superiority trials. A superiority trial (or a trial seeking evidence of greater efficacy when ondansetron is added to dexamethasone) would have been informative and interpretable, perhaps particularly desirable given dexamethasone's low cost. An interpretable noninferiority trial, however, would have required a full evaluation of all trials (not just one or two) comparing dexamethasone with placebo. If dexamethasone had consistently been shown to be effective (superior to placebo) in decreasing emesis, an appropriately sized equivalence–noninferiority trial could have been informative. Despite the documented effectiveness of ondansetron in the setting of highly emetogenic chemotherapy, however, there are many situations, primarily postsurgical, in which ondansetron has not been consistently distinguishable from placebo as an antiemetic. In these situations, a comparison with dexamethasone would have been uninformative. To know whether ondansetron was effective in these cases, placebo-controlled trials would be necessary.

Whether patients or physicians accept a trial of a particular design depends on the trial's value and necessity. It may be that an active control design is interpretable in the initial treatment of patients receiving moderately or markedly emetogenic cancer chemotherapy. This could be shown by a thorough review of experience with placebo-controlled trials of the proposed active control. If a consistent benefit were established, most patients and physicians would not want to participate in a placebo-controlled trial of a new agent, and no such trial would be needed for regulatory approval.

• Copyright ©2004 by the American College of Physicians