The American Journal of Bioethics

One of the secrets to having a successful career is to write a controversial article early on. Ensuing responses generate a dialogue with interlocutors guaranteed to pad one's curriculum vitae for years to come. My own work on the issue of placebo-controlled trials is critical of the excessive reliance on placebo controls in new drug trials, because such studies often violate clinical equipoise and suffer from scientific shortcomings. While Miller and Brody's (2002) critique does offer an opportunity to continue this dialogue, it is remarkable only in its ahistoricism, misconstrual of well-accepted concepts, and failure to grasp basic issues in the methodology of randomized controlled trials (RCTs).

Miller and Brody's central thesis is that the concept of clinical equipoise conflates the ethics of clinical practice with the ethics of clinical research.

Physicians in clinical practice have a duty to promote the medical best interests of patients by offering optimal medical care. In RCTs, however, physician-investigators are not offering personalized medical therapy for individual patients. Rather, they seek to answer clinically relevant scientific questions by conducting experiments that test the safety and efficacy of treatments in groups of patients.

This attempt to separate the ethics of clinical practice from the ethics of research fails to acknowledge the historical fact that clinical equipoise emerged as a response to the question, "When may the physician offer legitimately trial enrollment to her patient?" (Weijer, Glass, and Shapiro 2000). Recognizing that the treatment arms in an RCT may be consistent with the standard of care to which physicians are held, clinical equipoise allows a trial to go forward when there is honest, professional disagreement in the community of expert clinicians as to the preferred treatment. The significance of this insight can scarcely be overstated, as it provides a moral foundation for the RCT, one in which the ethics of practice and research are inextricably intertwined.

The alternative to clinical equipoise offered by Miller and Brody is too thin to be palatable. They suggest that physician-investigators simply have a duty not to expose subjects to excessive risks and to seek informed consent. Unable to answer the historical question, they beat a hasty retreat and conclude, "as a rule it is undesirable or ethically hazardous for physician-investigators to enroll in their studies individuals with whom they have an ongoing doctor-patient relationship." Not only is this position impractical, it fails to recognize that physicians do not cease to be physicians when they act as investigators. Irrespective of any prior relationship with a patient, a physician will, and indeed ought to, be held to standards of competent medical care when administering therapeutic interventions in a research study. It may be further noted that barring "excessive risk" has the unintended effect of negating the possibility of studying risky treatments, such as novel cancer treatments or surgical interventions. Miller and Brody are not forthcoming as to the origins of their alternative or why it ought to be supported. They merely assert that it is the case.

Despite the fact that arguments over the use of placebo controls largely emerge in the context of treatments for serious diseases, placebo advocates commonly provide examples of the use of placebo controls in studies of trivial diseases. Miller and Brody continue this tradition by asking what problem there is with a placebo-controlled trial of a novel treatment of allergic rhinitis. They claim,

Notice, however, the significance of recognizing an exception in this case and in comparable clinical trials. If it is ethically justifiable to conduct a placebo-controlled trial of a new treatment for allergic rhinitis, then what counts ethically is not denial of treatment but lack of substantial risk to participants.

This is, of course, quite false. Clinical equipoise does not require, as Miller and Brody believe, that "physicians offer optimal medical care to patients." Rather it requires that competent care be offered; in other words, treatment offered must be consistent with the prevailing standard of care. If, as is the case with many minor medical conditions, no treatment is an accepted therapeutic option, then, consistent with clinical equipoise, a placebo-controlled trial may proceed. Placebo-controlled trials in the context of serious illnesses such as depression or schizophrenia are ethically egregious precisely because no competent physician would fail to offer therapy to a patient with the condition (Weijer 1999).

Clinical equipoise is not a stand-alone ethical standard. Elsewhere I have presented a detailed calculus for the ethical analysis of risk in research premised on the recognition that therapeutic and nontherapeutic procedures are offered for differing purposes (Weijer 2000a). Nontherapeutic procedures are administered solely to answer the study question at hand; therapeutic procedures are intended to also provide therapeutic benefit to the subject. Therapeutic procedures must pass the test of clinical equipoise. Nontherapeutic procedures must have risks that are

1. minimized; and

2. reasonable in relation to the knowledge to be gained. [End Page 10]

Miller and Brody raise two objections to this schema for ethical analysis. First, they say that

[i]f physician-investigators are subject to therapeutic obligation in the case of clinical trials, which makes RCTs ethical only when they conform to clinical equipoise, it is puzzling that physician-investigators can ethically perform any research procedures that pose risks but no compensating therapeutic benefits to patient volunteers.

The answer to this objection is straightforward. Only therapeutic procedures are held to the standard of clinical equipoise. Indeed, it would make no sense to speak of nontherapeutic procedures such as questionnaires, extra blood draws, or X rays, in the context of a standard of care. No care is being provided. Nontherapeutic procedures are justified by moral tests that rely on minimizing risk and ensuring commensurate gains in knowledge. The nontherapeutic procedures present in pathophysiological studies are justified in precisely the same way as nontherapeutic procedures in RCTs.

Second, Miller and Brody claim that I have erred in classifying a placebo control as a therapeutic procedure. They state that "[n]o rationale is provided for describing a placebo control as a therapeutic procedure." This is, quite naturally, false. As I explain in the paper in question (Weijer 2000a), assuming that the placebo control is a nontherapeutic procedure is an exercise in question begging. In the design of an RCT, the question at hand is to what ought the novel treatment be compared? Clinical equipoise provides guidance to the trial designer, suggesting that the morally and scientifically preferred comparison is with best proven therapy or, if none exists, placebo (Weijer 2000b). The institutional review board is similarly guided by clinical equipoise when it asks whether the comparison of novel treatment to placebo is permissible. This is consistent with the often ignored second component of clinical equipoise, which requires that a trial ought to be designed in such a way that, if successful, clinical practice will change (Weijer 2000b).

In their attempt to dismantle 30 years of reflection on the ethical analysis of risk begun by members of the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research (Weijer 2000a), Miller and Brody have left us with a calculus for risk analysis that treats all interventions as nontherapeutic interventions. As noted above, this has unfortunate consequences. The authors on a preconscious level perhaps recognize the untenability of their position when they claim "risks that are not compensated by medical benefits to participants should not exceed a tolerable threshold." Do they mean to imply that a separate moral calculus for procedures associated with medical benefits is required after all?

The scientific merits of placebo controls have been hotly debated. Miller and Brody speculate that "there remain strong methodological reasons for including placebo controls in many trials comparing new and standard treatments," but at no point do they tell us what these reasons are or argue as to why they are good reasons. They suggest that I have "fail[ed] to come to grips with the methodological limitations of active-control trials" despite the existence of a number of full-length articles written by myself and colleagues on this exact point (Weijer 1999; Freedman, Weijer, and Glass 1996). Serious authors have argued that placebo controls have methodological advantages, and these arguments have all been countered (Weijer 1999; Freedman, Weijer, and Glass 1996; Temple and Ellenberg 2000). I leave it to the reader to judge the merits of these arguments.

In short, Miller and Brody offer no novel or persuasive argument to support the claim that a more permissive stance vis-à-vis placebo controls is desirable. Each of their points has been raised by others and countered elsewhere. This inertia is perhaps the most frustrating aspect of the placebo debate. Even when an argument is definitively rejected--and I am not claiming that all raised to date have been--it is bound to raise its tired head yet again in another piece. The resistance of the position of placebo advocates to argument raises the question, "Why does argument fail?" An examination of this question would provide for a much more stimulating perspective on the placebo question than that offered by Miller and Brody.