The Declaration of Helsinki is one of the major international codes of ethics for biomedical research involving human subjects (1). First promulgated in 1964 and revised several times, most recently in October 2000 (with a subsequent clarification), the Declaration represents the effort of the World Medical Association (WMA) to develop an international consensus on the ethics of medical research with human subjects (2). Article II.3 of the 1996 version of the Declaration required that every patient enrolled in a medical study be assured of receiving the best proven diagnostic and therapeutic methods. (We refer to the current version of the Declaration as "Helsinki 2000" and its immediate predecessor as "Helsinki 1996.") Placebo controls were permitted only in studies where no proven diagnostic or therapeutic method existed.
The use of placebo controls in clinical research is now a matter of intense controversy (3–7). Article II.3, for reasons we will describe later, was largely ignored, although those who hold that placebo use is unethical when there are accepted treatments for the condition under study often cited this provision of the Declaration of Helsinki in support of their position (3, 8–10). Although compliance with the Declaration of Helsinki is not required by law in the United States, it has been influential with some institutional review boards as they consider protocols that propose placing patient-subjects in a placebo control group.
The latest revision of the Declaration was motivated in part by this controversy over the use of placebos (11–14), a good deal of which has centered on the use of placebos in psychiatric research. However, the new version did not resolve key issues faced by clinician-scientists when considering the ethics of a placebo-controlled trial (14). The WMA’s Council, recognizing difficulties inherent in a literal interpretation of Helsinki 2000, issued a clarification in 2001 that substantially altered the categorical approach of Helsinki 2000 to the ethics of placebo controls. This clarification, formally approved in October 2002, represents a more reasonable approach to determining when a placebo-controlled clinical trial is ethically acceptable but does not provide substantive guidance for application. We describe the changes in the Declaration as it evolved, including Clarification 2002, and offer a framework for operationalizing the Clarification, illustrating the complexity of the issues with the example of research on schizophrenia.
The Declaration’s historical concern with placebo use in medical research is based on the moral principle that no patient should be denied the best available treatment and on the presumption that clinical ethics and clinical research ethics entail the same physician responsibilities to individual patients (14). Helsinki 1996 set forth its position on placebo controls in Article II.3:
In any medical study, every patient—including those of a control group, if any—should be assured of the best proven diagnostic and therapeutic method. This does not exclude the use of inert placebo in studies where no proven diagnostic or therapeutic method exists.
Taken at face value, Article II.3 not only limited the use of placebos but ruled out the testing of all new therapies for conditions for which even partially effective "proven" treatments existed. If all subjects were required to receive the best proven treatment, as the Article maintained, that would leave no opening for some subjects to be given a new, perhaps more efficacious, but not yet proven medication. Under this standard, except for cases in which a new medication could be tested as an "add-on" to an older drug (assuming compatibility and noncompetitive mechanisms of action), evaluation of new therapies would have been impossible. Strict application of this rule, for example, would have prevented efficacy testing of H2 receptor antagonists for the treatment of peptic ulcer, because the withholding of belladonna and its derivatives would have been considered an unethical failure to provide the "best proven therapeutic method." Similarly, the development of new and improved antihypertensive drugs would have ceased with the establishment of the effectiveness of the ganglionic blockers. This prohibition of placebo-controlled trials was not based on evidence of substantial or sustained risk. In fact, a recent meta-analysis (15) demonstrated that placebo controls in short-term clinical trials of antihypertensive agents are not associated with an increased risk for serious adverse events.
In Helsinki 2000, Article II.3 was replaced by the new Article 29:
The benefits, risks, burdens, and effectiveness of a new method should be tested against those of the best current prophylactic, diagnostic, and therapeutic methods. This does not exclude the use of placebo, or no treatment, in studies where no proven prophylactic, diagnostic, or therapeutic method exists.
Although it allows the testing of new treatments, Article 29 perpetuates the proscription of placebo controls in clinical trials when even partially effective treatments exist, thus continuing many of the difficulties associated with Helsinki 1996, which we expand on later in this article. Moreover, by proscribing the use of placebo controls in clinical trials in which there is virtually no risk of serious adverse consequences, it continues to give equal moral weight to substituting placebo for insulin in diabetes research and substituting placebo for aspirin in a headache study. The category of studies in which the risk of placebo use is negligible includes research on analgesics, hypnotics, and antihistamines, where, because of variations in symptom perception and reporting, natural course of illness, cohort effects, and subjective experience of illness, placebo-controlled designs are the norm. Prohibition of placebo use greatly reduces the efficiency of conducting research of this type without any compensating enhancement of subjects’ safety. Article 29 also bars placebo controls in studies designed to test, in developing countries, inexpensive alternatives to therapies used in the developed world. (For further discussion of this topic, which is beyond the scope of this paper, see references 10–12, 16.)
Reacting to expressions of dismay over Article 29, in 2001 the WMA issued a Clarification regarding its interpretation. The WMA was unable to meet in October 2001 because of the events of September 11, but the Clarification was approved as a footnote to Article 29 at the WMA meeting in October 2002. The Clarification reads as follows:
The WMA hereby reaffirms its position that extreme care must be taken in making use of a placebo-controlled trial and that in general this methodology should only be used in the absence of existing proven therapy. However, a placebo-controlled trial may be ethically acceptable, even if proven therapy is available, under the following circumstances:
Where for compelling and scientifically sound methodological reasons its use is necessary to determine the efficacy or safety of a prophylactic, diagnostic or therapeutic method; or
Where a prophylactic, diagnostic or therapeutic method is being investigated for a minor condition and the patients who receive placebo will not be subject to any additional risk of serious or irreversible harm.
All other provisions of the Declaration of Helsinki must be adhered to, especially the need for appropriate ethical and scientific review. (WMA web site: http://www.wma.net/e/home.html)
This Clarification narrows the injunction against placebo substitution for a proven treatment. Clinical trials with a placebo control group would be ethically justified where there are good reasons for placebo use or if the condition being studied is "minor" and the additional risk is negligible. This is an important recognition of a moral distinction among circumstances in which placebos may be used. It would require a case-by-case review of the justifications for placebo use, a process sensitive to scientific merit and study-specific risk based on subject selection and risk-reduction procedures. This position, previously articulated by the leadership of both the U.S. Office of Human Research Protections (oral communication from Dr. Greg Koski, Director) and the U.S. Food and Drug Administration (5, 17), now appears to have been adopted by the drafters of the Helsinki Clarification as well. It is worthwhile to consider why this approach embodied by the Clarification represents a forward step in the regulation of research.
The conclusions of Helsinki 1996 and 2000 that placebos can never be used in place of standard therapies in a major condition were misguided. These provisions were based on the assumption that physicians, even when they are conducting research, ought to ensure the maximum benefit possible for each person in their care, consistent with the fiduciary obligation of undivided loyalty to the interests of the patient (18–20). It is generally accepted, however, that the ethics of the research setting are not identical to those of ordinary clinical care (21, 22). Whereas physicians in usual clinical settings owe a single-minded dedication to their patients’ well-being, physician-researchers are concerned as well with generating reliable scientific knowledge. This leads them to use techniques that would not be acceptable in the usual clinical context: assigning patients to treatments at random, regardless of their own preferences; working within the constraints of a clinical protocol that may specify dosage levels and exclude adjunctive treatments; not telling subjects which treatment they are receiving and remaining ignorant of this themselves (i.e., double-masked procedures); and, in some cases, using placebos rather than active medication.
How can these deviations from the usual conduct of clinical care be justified? First, studies must have the potential to yield scientifically valid and clinically useful information, and the magnitude of these benefits must be reasonable and proportional to the risks that subjects are running, including those associated with the deferral of some components of standard therapy. Second, the patient-subjects themselves, if they are competent adults, having been informed of the nature and purpose of the research study, its potential risks and benefits, and their option of receiving ordinary therapy, and being able to make a voluntary decision, must consent to these deviations from usual clinical practices. If either of these conditions is absent, departure by the physician from a single-minded focus on patients’ best interests is illegitimate.
From this perspective, the use of placebo in a particular clinical trial should not be routine but requires the demonstration that it brings substantial scientific advantages, combined with the use of adequate risk reduction/management methods. There is an extensive literature elaborating the scientific advantages of placebo use (7, 23–26). We highlight only two aspects of it here. Since the differences in response rates are likely to be greater between a new, effective medication and a placebo, as opposed to a standard effective drug, studies using placebo have greater statistical power and can recruit smaller numbers of subjects. This advantage associated with placebo use not only diminishes the cost and duration of studies but also reduces the number of subjects who will be exposed to potential adverse effects of the new medication. If placebo-controlled trials are more efficient, new therapies will be made available more quickly at less cost.
The presence of a placebo condition also may be essential for interpretation of the findings of a study, particularly when no differences are found between subjects receiving a new medication and those getting an older "proven" drug. This lack of difference could be the result of both medications being equally effective. Alternatively, both medications may have been equally ineffective in this trial because of errors in measurement or because of the selection of a subject population enriched with individuals resistant to either treatment. This appears to be a growing problem precisely in research on those disorders for which partially effective treatments exist, since those subjects who do not respond favorably to existing treatments have an incentive to enroll in trials of newer drugs. In schizophrenia studies, for example, the effect size of antipsychotic drug treatment has decreased markedly over time, and drug and placebo response (and the difference between the two) vary considerably across studies (27), precluding the use of "historical" controls. Thus, the advantages associated with placebo use include both reducing the number of subjects at risk and preventing false negative and false positive results. As we suggest later in this article, it would be wrong to say that the use of placebos does not raise concerns that require careful examination. Addressing this issue is necessary throughout clinical research on therapeutic interventions. Now, however, we turn to the use of placebos in schizophrenia research as an example of the benefits of this more nuanced approach.
Schizophrenia (28) is a chronic psychotic disorder with onset usually at the end of the second decade of life. It is marked by several pathological manifestations that are only weakly related to each other within the individual patient: positive psychotic symptoms (delusions, hallucinations, and disordered thoughts); negative symptoms (e.g., restricted affect and drive); and impairments in cognition. Therapeutic efficacy (29) has been established for several classes of antipsychotic drugs, augmented by specific psychosocial therapies. The main effects of these treatments are the reduction of positive psychotic symptoms and the delay of psychotic relapse in clinically stable patients. However, many patients are poor responders to any available medication. Moreover, efficacy for major aspects of the illness, such as cognitive impairment and primary negative symptoms, is not established. All current medications have significant side effects—ranging from tardive dyskinesia, a drug-induced movement disorder that was common with the older medications, to excessive weight gain and increased cardiac, vascular, and diabetic risk associated with many of the newer drugs. Thus, schizophrenia is one of the disorders that Article 29 of Helsinki 2000 affects most directly: there are substantial reasons to seek new medications that are more effective and safer than the partially effective treatments that are currently available.
Assuming the development of a scientifically valid research protocol and the presence of a pool of willing and competent research subjects (many persons with schizophrenia retain good decision-making capacity [30–35]), how might we assess the risks of introducing a placebo group into a schizophrenia clinical trial? First, substituting placebo for active medication has implications for some, but not all, aspects of the disease. Long-term outcomes related to functional capacity and quality of life, for example, have little relationship to positive psychotic symptoms. Rather, they are associated with the presence of primary negative symptoms and impairments in cognition; antipsychotic drugs have not been demonstrated to be efficacious for these critical aspects of the disease.
Second, using placebos does not require withholding all other aspects of treatment. Indeed, psychosocial interventions, which reduce symptoms and enhance stability and functional capacities, are often enriched in clinical trials. Adjunctive medications may address transitory symptomatic expression. The study can be designed with a low threshold for intervention with antipsychotic medication. Overall risk of trial participation is thereby diminished.
Third, persons with schizophrenia vary considerably in their responsiveness to medication. The risk of substituting placebo for active medication in poorly responsive patients—precisely the group of greatest interest in the development of new treatments—is less than it would be in good responders. Article 29 of the Declaration fails to distinguish between these groups.
Finally, safeguards may be built into clinical trials to diminish the risks that subjects face. Careful monitoring of subjects allows rapid initiation of active medication at the first signs of symptom exacerbation or, in an acute therapy design, if symptom reduction and patient management progress too slowly. A clinical plan that combines placebo with close monitoring and rapid medication intervention, referred to as a "clinical out" procedure, may have greater risks than concurrent medication; it is substantially safer, however, than what many critics assume when they equate placebo use to the withholding of "all" treatment. We view placebo substitution as a treatment modification associated with risks; these risks should be minimized with enhanced clinical care and the determination as to whether the safety level achieved is acceptable.
There is extensive evidence that the actual experience of subjects in placebo-controlled trials in schizophrenia is not associated with sustained adverse consequences. Even though much of the data come from an era when subjects might have been maintained without medication until substantial symptom exacerbation, follow-up studies have found no differences in clinical course or social outcome between patient-subjects randomly assigned to placebo or to continuous medication (36–38). Moreover, rates of suicide do not appear to be higher among subjects with schizophrenia who received placebo in clinical trials (39–41). The observed risks of placebo use are thus limited to the period during which subjects are participating in the protocol and do not include long-term disability or progression of the underlying pathology.
Use of placebos in schizophrenia research, therefore, presents a range of risk, including—we believe—circumstances in which subjects might reasonably conclude that they would be willing to endure such risks for the sake of advancing knowledge about their illness (42, 43). Schizophrenia, moreover, is one of those conditions for which placebo use will sometimes present distinct advantages in clinical trials. Some of these advantages are the same that would accrue with any disorder, including smaller sample sizes. Variable rates of response in schizophrenia subjects assigned to placebo underscore the importance of having a nonactive control as a comparison condition. The presence of partially effective treatments means that subjects are likely to be drawn disproportionately from treatment-resistant populations, where an ineffective investigational drug may not be differentiated from a standard treatment that is ineffective in the study cohort.
Although the Helsinki Clarification takes a more reasonable position with regard to placebo controls, it will require considerable work before all of its implications are clear for investigators, institutional review boards, and funding and regulatory agencies. As this process is beginning, we believe the following approach to be consistent with the intent of the Clarification.
Placebo controls in clinical trials can be categorized in four distinct groups, each of which differs from the others in ways that are relevant to their ethical justification and practical aspects of oversight: 1) evaluation of therapies for conditions for which there are no existing therapies; 2) evaluation of therapies intended to provide symptomatic relief of conditions that, if untreated, do not have serious adverse consequences and for which there are existing therapies; 3) evaluation of therapies for conditions that could, if untreated, result in serious adverse consequences and for which there are existing therapies; 4) evaluation for use in "low resource" countries of inexpensive alternatives to expensive therapies regarded as the best proven therapeutic methods in wealthy countries. (Although we are addressing only the use of placebos in studies evaluating therapeutic agents, similar arguments could be applied to their use in studies on diagnostic or preventive interventions.)
Placebo use in category 1 would have been permitted even under Helsinki 2000, including, we believe, patients who refuse standard treatment for their condition (e.g., some patients with multiple sclerosis who reject the benefit-risk ratio of current treatments [44, 45]). Category 2 involves placebo use with the "minor" disorders addressed by the Clarification and represents the least controversial challenge to the categorical rejection by Helsinki 2000 of placebo substitution for a standard treatment. Category 4 raises important issues but is beyond the scope of this discussion. Our attention for the remainder of this article is directed toward placebo studies in category 3—which would include clinical trials in schizophrenia—disallowed in Helsinki 2000 but apparently permissible in at least some circumstances under Clarification 2002.
When determining whether any specific research plan meets the standard of Clarification 2002, we propose that investigators and review boards consider the following criteria:
1. There must be a plausible hypothesis that the experimental treatment will have either greater efficacy or reduced side effects compared with existing treatment, or will be effective for aspects of the disease for which standard treatment is not proven efficacious. Hypothesized differences should involve clinically significant benefits, not clinically trivial statistical differences.
2. There must be sufficient uncertainty regarding efficacy to justify a clinical trial (14, 16).
3. The scientific design must be sound, with good reason to anticipate that members of the expert clinical community will find the results convincing.
4. Compelling reasons must exist for placebo use. These may include the following: standard medications cannot be presumed to be effective with the population from which the sample is drawn; the condition being studied is susceptible to substantial fluctuation in severity or to spontaneous remission; the measurement techniques being used in the study are unavoidably imprecise; substantial reduction of risk of exposure to experimental interventions is possible with placebo use; or substantial benefits are likely as a result of more rapidly determining whether the experimental intervention is effective.
5. Subject selection must minimize the likelihood of serious adverse consequences, and subjects must be capable of providing informed, voluntary consent, or have consent provided by a suitable surrogate.
6. The advantages likely to accrue from placebo use outweigh the risks to subjects. This risk-benefit analysis requires a complex judgment that takes into account the efficacy and adverse effects of existing treatments; the likelihood of response by the particular study cohort to existing treatments; the probability (and intensity and duration) for this cohort of adverse consequences from placebo substitution; whether the number of subjects receiving placebo and the duration of its use have been minimized; the availability of other forms of treatment during the study and their potential to mitigate adverse consequences; the procedures in place for close monitoring of subjects and prompt restoration of active treatment; and the likelihood that adverse consequences can be terminated or reversed by reinstatement of the withheld treatment.
Categorical rejection of placebo controls when an effective treatment exists ignores the complexity of the issues involved in designing ethical clinical trials. Adoption of this position in the iterations of the Declaration of Helsinki through 2000 has limited its usefulness and influence. The Helsinki Clarification 2002 provides an important advance, which we interpret as requiring a case-by-case evaluation of research proposals in order to render a valid judgment on a complex issue in research ethics. Our argument here has relied on the clause that allows placebo use "where for compelling and scientifically sound methodological reasons its use is necessary to determine the efficacy or safety of a prophylactic, diagnostic or therapeutic method." The clause regarding placebo use "where a prophylactic, diagnostic or therapeutic method is being investigated for a minor condition and the patients who receive placebo will not be subject to any additional risk of serious or irreversible harm" raises questions of its own that we have not addressed here but are worthy of further exploration.
We have underscored the complex issues involved in determining the reasonableness of the decision to use placebo in a clinical trial. Drug therapy research in schizophrenia was used by way of illustration, but the fact that the issue is common throughout clinical research on therapeutic interventions was emphasized by recent attention to the question of sham surgery (46–48). The simplistic approach taken by the framers of Helsinki 2000 failed to take into account the value to affected populations of identifying more effective and less noxious forms of treatment and the utility of placebos in accomplishing these goals. The 2000 version of the Declaration failed to acknowledge the different methods that can reduce risk and enhance clinical care during placebo-controlled trials. In rejecting the call of Helsinki 2000 for categorical exclusion of placebo use in favor of case-by-case consideration of scientific merit, actual risk, risk-reduction methods, and the interest of patients in advancing knowledge on their own diseases, the 2002 Clarification to the Declaration of Helsinki represents a more reasonable basis for ethical justification and legal regulation of placebo use in clinical research.
Received March 20, 2002; revision received Aug. 17, 2002; accepted Aug. 22, 2002. From the Maryland Psychiatric Research Center, University of Maryland School of Medicine; the University of Massachusetts Medical School, Worcester; and Yale University School of Medicine, New Haven, Conn. Address reprint requests to Dr. Carpenter, Maryland Psychiatric Research Center, P.O. Box 21247, Baltimore, MD 21228; email@example.com (e-mail). Supported by NIMH grants MH-58898 and MH-40279 and the Distinguished Investigator Award, National Alliance for Research on Schizophrenia and Depression (Dr. Carpenter) and by NIMH/National Institute on Drug Abuse grant MH/DA 56826 and NIMH grant MH-62294 (Dr. Levine).