Placebo in Clinical Trials for Depression: Title and subTitle BreakComplexity and Necessity

Within the last several years, the long-simmering debate about the ethical issues surrounding placebo administration in randomized clinical trials has reached new levels of intensity. The report of the National Bioethics Advisory Commission included specific recommendations for studies involving more than minimal risk, including the use of placebo controls in clinical trials.¹ At a time of increased interest in drug discovery in medicine and the potential need for definitive randomized clinical trials, the controversy over the ethical issues places an appropriate burden on the scientific community to justify the use of placebo controls.^{2 - 3} One argument is that placebo administration is not appropriate if effective treatment for a condition exists and that assessment of efficacy can be conducted with active controls. Another view is that placebo controls may be necessary to determine the assay sensitivity of a trial and are ethical if patients provide informed consent and are not harmed.⁴

Treatment of depression, a well-recognized and worldwide public health problem,⁵ represents a special opportunity to highlight several key issues in the "placebo controversy" and to demonstrate why for the foreseeable future placebo-controlled trials will be necessary. While the treatment of depression has been marked by major successes with the introduction of new classes of antidepressants and their "uptake" by clinicians and patients,⁶ newer antidepressants will continue to be developed. These new products may target specific subtypes of depression, may provide more complete recovery and sustained remission, and may have reduced adverse effects and faster onset of action. These products will require proof of efficacy and Food and Drug Administration approval. A better understanding of the many alternative or complementary treatments for depression, such as nonprescription herbal compounds, also contributes to the need to determine when placebo controls are appropriate.

Two articles in this issue of THE JOURNAL, one from the Hypericum Depression Trial Study Group⁷ and another from Walsh et al⁸ on placebo response, focus attention on the key issues. The first report examines the efficacy of St John's wort (Hypericum perforatum) in major depressive disorder (MDD) in a 3-arm clinical trial.⁷ This study provides an opportunity to comment on the role of placebo and whether placebo efficacy rates "interfere" with the establishment of the efficacy of new treatments. The second report addresses the problems of placebo in depression trials using a 20-year perspective. Walsh and colleagues⁸ describe how placebo characteristics and clinical response levels have changed.

The current study⁷ on the use of St John's wort in the treatment of MDD is the second one within a year⁹ to conclude that St John's wort is not effective. These trials were conducted because, even though St John's wort is widely used for the treatment of major depression and depressive symptoms, its efficacy has not been clearly established, despite more than 20 randomized trials, most of which are considered to have had serious methodological flaws.^{9 - 11} Both of the recent trials were multicenter randomized, double-blind, placebo-controlled trials with a standardized extract of St John's wort. The previous 8-week trial by Shelton et al⁹ showed no significant difference between St John's wort extract and placebo on any of the depression outcome measures. Response rates in the intention-to-treat analysis were also not significantly different (26.5% for St John's wort vs 18.6% for placebo). The second study,⁷ reported in this issue, also included a large multicenter population of outpatients with MDD, but differed in that a selective serotonin reuptake inhibitor (SSRI), sertraline, was included as an active control in addition to testing St John's wort extract (hypericum) and placebo. From this 8-week clinical trial, the authors conclude that neither sertraline nor hypericum was significantly different from placebo on the 2 primary outcome measures, the Hamilton Depression Scale (HAM-D) and the Clinical Global Impression Scale (CGI-I). The overall response rates (including partial and full response) were 38.1% for hypericum, 43.1% for placebo, and 48.6% for sertraline.

As the authors point out, their study represents an excellent example of why it is critical to include both placebo and active comparators in trials of agents for which efficacy is unproven. Without a placebo control, one might prematurely conclude that hypericum has efficacy comparable to an SSRI, in this case, sertraline. However, without sertraline as the active comparator, the conclusion, as in the previously published trial, would be that St John's wort extract is not efficacious in MDD. Before drawing conclusions, the low assay sensitivity of this new trial must be acknowledged. It is likely that the study was underpowered given the assumption of a 20% drug-placebo difference in full response rates. In addition, while relatively high doses of hypericum (1500-1800 mg/d) were permitted, the maximum acute dose of sertraline (100 mg/d) almost certainly contributed to its low effect size and failure to separate out from placebo or hypericum. The overall placebo response was high compared with the previous hypericum vs placebo study,⁹ demonstrating the variability in placebo response from trial to trial. A recent review of placebo controls in depression trials¹² made a series of recommendations to enhance documentation of treatment vs placebo efficacy, including greater emphasis on effect size, rather than significance levels alone.

Walsh and colleagues⁸ have approached the placebo response issue for major depression studies in a careful review of controlled trials between 1981 and 2000 in which outpatients with MDD were randomly assigned to medication or placebo. In the 75 trials meeting this criterion, the mean (SD) proportion of patients in the placebo group who responded was 29.7% (8.3%). Since many investigations involved more than 1 active antidepressant, the greater response with an active drug in such trials was used to calculate the overall active treatment response rate (50.0% [9.0%]). The average effect size across studies for maximum proportion responding to a medication was 0.43 (0.22). The authors conclude that the response to placebo is highly variable, substantial, and has increased significantly in recent years as shown by the high positive correlation with the year of publication. While the response rate for antidepressant medication has also increased in the last 20 years, the association between response rate and year of publication was more statistically robust for placebo than for active medication. Walsh et al note that the change in placebo response rate does not appear to be directly explained by changes in study characteristics such as patient age, placebo lead-in, or minimum required HAM-D score. It is difficult to disagree with their contention that since the average rate of response has changed over time, the use of a historical standard as a valid benchmark of efficacy rates would be a poor choice.

Both of these reports, from different perspectives, point to the continuing set of problems inherent in clinical trials for major depression. Based on an extensive consensus development panel process convened in September 1999, and on summaries of presentations at that conference, including those by Lavori¹³ and Rush,¹⁴ a National Depressive and Manic-Depressive Association Consensus Statement on the use of placebo in clinical trials of mood disorders¹⁵ concluded that placebo has a definite role in mood disorder studies and that findings of equivalence between a new drug and standard treatment are not evidence of efficacy unless the new drug is also significantly more effective than placebo. Since this report expresses the consensus of an interdisciplinary group of clinical researchers, biostatisticians, bioethicists, and consumers, it is likely that the use of placebo-controlled trials in developing interventions will continue. It is also noteworthy that all the stakeholders are now much more sensitized to the necessity of seeking newer strategies for shorter durations in drug-placebo trials to minimize risks, as well as to the need for alternate research designs, when appropriate.

Taken together, the 2 reports in this issue of THE JOURNAL return full circle to the placebo response and understanding its mechanisms of action¹⁶ and highlight the perplexing complexity of the placebo and its ability to cause "mischief" in scientific inquiry. This may be nature's way of providing clues to fundamental aspects of the healing process, even as advances in medicine and the discovery of new therapies take place. It is important to learn from, rather than dismiss, the variability of the therapeutic response.