Mental illness is, in spite of the recent advances in the social and neuro sciences, still a great unknown. The need for objective measurements and for valid and reliable markers of psychic phenomena has never been greater. The paucity of knowledge surrounding the causes of mental illness has inevitably led to vacuum-filling treatments which are of unpredictable or questionable efficacy. Such treatments owe their therapeutic effect to the influence of a series of nonspecific variables which we have only identified to a minor degree. They are present nevertheless, and do contribute to the improvement of the patient.

In addition to the direct therapeutic interventions that one might consider nonspecific, there are a number of other variables which form an integral part of any therapeutic strategy.

These may include environmental, financial, personality, and time variables which converge with those variables considered specific to the condition under treatment in a synergism that will hopefully lead to improvement of the individual. These nonspecific variables are grouped under the heading of "placebo," not because they represent "no treatment" but rather because of our own incomplete understanding of their total influence on the illness.

To consider the use of placebo in a well-conducted clinical trial to be "no treatment" reflects a lack of under- standing of the therapeutic process in mental disorders. The potency of the nonspecific or placebo effect should be recognized in the evaluation of any treatment because it is clear that neither biological nor psychological therapies alone are sufficient to bring about recovery. This has been demonstrated repeatedly in most medical disorders. A brief look at the Compendium of Pharmaceuticals and Specialties (Ottawa: Canadian Pharmaceutical Association), a desktop reference for practicing physicians, reveals the potency of the placebo effect in its contribution to side effects.

The potency or importance of the placebo factor is not uniform in all psychiatric conditions or even in all stages of a specific condition. There are, however, a number of situations in which the placebo effect is such that its recognition and measurement is necessary to allow for an accurate evaluation of a pharmacological effect. The drug variable is, in the context of a clinical trial, the variable to be measured and quantified. It should now be obvious that it is not quantified in isolation from the placebo effect but in light of the synergism of all variables. The failure to recognize this synergy in the evaluation of a treatment would be intellectually dishonest. Moreover, it would inevitably lead to the use of treatments of unproven value with a potential for serious side effects which have not been adequately weighed against the potential benefits.

The use of placebo as a form of reference for the treatment under study in mental disorders has been the subject of controversy in psychiatry. Many of the issues raised by opponents of placebo-controlled trials arise from a lack of understanding of mental illness. For example, the question of consent is often raised, yet being ill and being incompetent to consent are two different issues. Most patients, even if certifiable, are competent to consent to treatment. The suffering inflicted by a mental illness is difficult to appreciate by the uninitiated. Because the available treatments are not adequate for large numbers of patients, the hope of better treatments down the road is a strong motivation for many patients who volunteer for clinical studies.

The question of when it is appropriate to have a control for the placebo effect in the evaluation of a new treatment then arises. In the opinion of many, it is necessary to demonstrate the efficacy of a new treatment at the outset of its evaluation. In those conditions which are strongly influenced by nonspecific factors, a control for these through a placebo condition is important to evaluate the continuing course of the group of patients in treatment and the impact of these factors on the outcome. The early demonstration of efficacy is also essential in reducing the unnecessary exposure of many patients to an ineffective substance. This early comparison to placebo is necessary both because of the lack of ‘gold standards’ in most psychiatric treatments and because of the unpredictability of the course of most psychiatric illnesses. This is due not only to the heterogeneity of diagnostic groupings but also to the important influence of nonspecific factors, which are embodied in the so-called placebo effect. The question of equipoise is thus addressed when treatments are randomized in conditions such as depression where up to 30% of patients respond favourably to placebo conditions and as few as 50% respond to an "active" anti-depressant.

The conditions for which one would choose to have a placebo control may vary. There are no blanket statements which will provide one answer to such a complex issue. In the design of a trial, it is essential to construct safeguards that will allow easy removal from the study, but these safeguards must also apply to the active drug treatment arms.

he publication of Rothman and Michel's article, "The continuing unethical use of placebo controls," in the New England Journal of Medicine sparked an increasingly fervent controversy over the proper role for placebos in psychiatric and other clinical research¹. How should Canadian Research Ethics Boards (REBs) deal with protocols involving the use of placebo controls? Under what circum-stances may such proposals be allowed to go forward? Recent regulatory changes in Canada have a direct impact on these questions.

Questions regarding the use of placebos must be set within the established ethical framework for research. REBs in Canada serve to protect research participants. This task is discharged by ensuring that clinical research fulfills ethical and scientific standards set out in the Tri-Council Policy Statement on Ethical Conduct for Research Involving Humans².

Ethical standards for human experimentation are typically expressed in three principles: respect for persons (the choices of autonomous persons ought to be respected; non-autonomous persons are entitled to protection), beneficence (do no harm; maximize possible benefits while minimizing risks) and justice (the benefits and harms of research ought to be distributed equitably)^3,4. The principle of respect for persons underlies requirements for informed consent and confidentiality. Beneficence demands that research participation be associated with an acceptable balance of potential benefits and risks. Justice requires that researchers neither prey upon the vulnerable nor exclude without good reason those who may benefit from research participation.

The Code of Ethical Conduct for Research Involving Humans provides REBs with clear guidance on the proper use of placebos. Article 5.4 states: "The use of placebos in clinical trials is ethically unacceptable where clearly effective therapies or inter-ventions are available." This require-ment is consistent with earlier guide-lines from the Medical Research Council of Canada: "Particular care must be taken, however, in such studies [clinical trials] with patients, to ensure that the subject's best interest is never sacrificed to that of the randomized study"⁵. It is also consonant with the principle that the medical treatment of research subjects ought not be disadvantaged by study participation, a tenet of a variety of international codes and regulations^6,7.

Article 5.4 is best understood as an entailment of clinical equipoise, a widely accepted guiding rule for the interpretation of the principle of beneficence. In order for a clinical trial to be initiated ethically, a state of clinical equipoise must exist. That is, there must be "genuine uncertainty within the expert medical community about the preferred treatment"⁸. If "clearly effective therapies or inter-ventions are available," a clinical trial in which one of the treatment arms is placebo cannot satisfy clinical equipoise (in this case placebo is, by definition, known to be inferior to standard treatment). Freedman, the author of clinical equipoise, sets out five circumstances in which a placebo control may be justified: there is no standard therapy; standard therapy is no better than placebo; standard therapy is placebo; new evidence has called into question the therapeutic index of standard therapy; and optimal therapy is not available due to cost or short supply⁹. Two cases in which placebo is unproblematic deserve to be highlighted: the study of an ‘add-on therapy’ in which all patients will receive standard treatment, and a clinical trial of a new treatment in a patient population known to be refractory to standard treatment (and for whom no effective second-line treatment exists).

When proposing a placebo-controlled trial, it is incumbent upon the investigator to prove that no "clearly effective therapies" exist for the condition in question. REBs should verify such claims by consulting with expert clinicians not involved in the study (and who have no other potential conflicts of interest) and the medical literature. Were a researcher, let's imagine, to claim that there is no clearly effective treatment for schizophrenia or depression, the overwhelming literature to the contrary ^{10, 11, 12} would force the REB to reject the protocol (indeed, a failure to do so may expose the REB and the institution to legal liability)¹³. Also, REBs must not be swayed by irrelevant claims by investigators, for example:

• "there isn't one standard treatment" (when there are several treatments known to be effective, the study should choose the best or most widely-used as the control treatment), or
• "standard treatment doesn't work for every patient" (no medical treatment does; if the issue is truly the treatment of refractory patients, then the trial's study population should be restricted to those who failed to respond to standard treatment).

REBs may encounter the assertions, "even if clearly effective therapy exists for a medical condition, good science demands the use of a placebo control" and "there are no acceptable alter-natives to this type of trial." Both claims have been examined rigo-rously and shown to be false¹⁴. (It should be noted that even if true, there are no provisions in the Code of Ethical Conduct for Research Involving Humans to allow a trial to go forward on these grounds.) The belief that comparison to placebo uniquely allows one to determine the "true biological effect" of a drug relies on a simplistic view of drug effect: true biological effect = overall response–placebo (psychological) response. If biological and psychological factors act synergistically, e.g., biological effect may potentiate psychological expectancy, then this purported advantage of placebo is nullified. Indeed, placebo-controlled trials may have distinct scientific disadvantages. As opposed to an active treatment control, the use of a placebo control makes it far more difficult to preserve blinding in a study—essential to the unbiased assessment of treatment—due to obvious disparities in efficacy or side effects between treatment arms in a clinical trial. New study designs have been developed specifically to test the equivalence of new treatments and current standard therapy¹⁴. These trial designs will be used more frequently as the allowable use of placebo controls is (appropriately) restricted.