ALTHOUGH their primary concerns were the safety and ethics of placebo use, Khan et al¹ also used the Food and Drug Administration files on second-generation antidepressants to assess "symptom reduction among patients assigned to antidepressants and placebo." In this sample of approximately 19 000 patients, placebo and drug had comparable morbidity and mortality and the authors report that "symptom reduction was 40.7% with investigational . . . 41.7% with active comparators . . . and 30.9% with placebo. . . . "

We believe that these data can appropriately assess morbidity but do not adequately assess efficacy. The only data available were pretreatment and posttreatment mean total Hamilton Depression Rating scores, most without SEs. Although constrained by database inadequacies, effect size estimates were calculated and Khan et al report that "nearly half the studies had less than medium effect size (<0.5)."