Over the two decades, the responsiveness to both active medication and placebo increased statistically significantly, but with a significantly greater increase with placebo than with medication. Walsh and colleagues (2002) indicated that this pattern of response to placebo is "highly variable and often substantial, and has increased in recent years, as has the response to medication."
Although unable to identify causal factors for the phenomenon, they suspected there to be relevant differences between participants in recent studies who may be more likely to be recruited from the public, and those in the past who were typically referred by clinicians. "It is likely that clinically important characteristics of patients participating in treatment studies have changed as a result of these practices," Walsh and colleagues indicated.
In an accompanying editorial, Kupfer and Frank (2002a) emphasized the value of incorporating placebo control in depression treatment studies. They noted that an absence of placebo in the HDTSG trial might have led to the conclusion that hypericum is comparable to sertraline(Drug information on sertraline) in the treatment of depression, rather than to issues of study assay sensitivity. They acknowledged, however, that placebo response in clinical trials remains problematic and requires further assessment.
"Taken together, the two reports å return full circle to the placebo response and understanding its mechanism of action and highlight the perplexing complexity of the placebo and its ability to cause 'mischief' in scientific inquiry," Kupfer and Frank observed.
Schneider and Small (2002) have found that depression remission rates with active medications and placebo differed significantly between clinical sites within multisite studies, including those reviewed by Walsh et al. (2002). Among 29 studies, Schneider and Small found that remission rates at individual sites ranged from 0% to 71% with active medication and from 0% to 75% with placebo.
Private research sites showed greater rates of remission with both medication and placebo than did academic sites, with more than twice the drug-placebo effect sizes. In addition to considering individual patient characteristics to understand changes in placebo response, as Walsh and colleagues (2002) suggested, Schneider and Small (2002) recommended discerning differences between investigators and investigational sites. They noted that most depression studies utilize the interview-based HAM-D and/or unstructured CGI for investigators' assessments of improvement. Both of these can reflect interaction between interviewer and subject.
"Individual investigators, despite participating in the same clinical trial and following the same protocols, might have vastly different clinical experiences," they observed. "Subtly different methods of subject selection, perceived financial incentives, unaddressed bias, and unique characteristics of individual sites may contribute to response variability" (Schneider and Small, 2002).
In their letter on the HDTSG hypericum study, Cott and Wisner (2002) suggested that the study could serve as a role model for the difficulties in carrying out antidepressant trials in which the average failure rate is 50%. In a letter separate from their editorial, Kupfer and Frank (2002b) advocated for trial designs that emphasize clinical rather than statistical significance, albeit with adequate assay sensitivity and effect size.