A reevaluation of the exclusion criteria used in antidepressant efficacy trials.

Abstract

Antidepressant efficacy trials use multiple exclusion criteria that are intended primarily to maximize differences between the drug and placebo. These exclusion criteria were implemented prior to having undergone rigorous testing, however, and it is unclear whether they accomplish their intended goal. The authors review the current state of knowledge about treatment response rates in individuals who are typically excluded from antidepressant efficacy trials. After reviewing 31 antidepressant efficacy trials published between 1994 and 1998, the authors identified 10 commonly used exclusion criteria. A computerized MEDLINE search and a manual search were used to identify studies evaluating the efficacy of antidepressant medications in populations of subjects typically excluded from antidepressant efficacy trials. Three exclusion criteria-short episode duration, mild severity of illness, and positive response during the placebo lead-in phase-are used primarily to reduce placebo response rates. Preliminary evidence suggests that, of those three criteria, only a short episode duration has been associated with less robust differences between drug and placebo response rates. The seven other exclusion criteria, including various psychiatric comorbidities, long duration of illness, medical comorbidity, and prior nonresponse to treatment, are used because individuals who meet those criteria are believed to have lower response rates to somatic therapy. Although such individuals are often found to have lower response rates to somatic therapy, they also appear to have lower response rates to placebo, and there is little evidence to suggest that drug-placebo differences are any less robust in these individuals. Many of the standard exclusion criteria currently used in antidepressant efficacy trials may not be achieving their intended goal of maximizing drug-placebo differences. The practice of excluding subjects with particular clinical profiles, which greatly reduces the generalizability of antidepressant efficacy trials, appears to lack empirical support.