Placebo, prozac and PLoS: significant lessons for psychopharmacology

Kirsch et al. (2008, Initial severity and antidepressant benefits: a meta-analysis of data submitted to the Food and Drug Administration. PLoS Med 5: e45), conducted a meta-analysis of data from 35 placebo controlled trials of four newer antidepressants. They concluded that while these drugs are statistically significantly superior to placebo in acute depression, the benefits are unlikely to be clinically significant. This paper has attracted much attention and debate in both academic journals and the popular media. In this critique, we argue that Kirsch et al.'s is a flawed analysis which relies upon unusual statistical techniques biased against antidepressants. We present results showing that re-analysing the same data using more appropriate methods leads to substantially different conclusions. However, we also believe that psychopharmacology has lessons to learn from the Kirsch et al. paper. We discuss issues surrounding the interpretation of clinical trials of antidepressants, including the difficulties of extrapolating from randomized controlled trials to the clinic, and the question of failed trials. We call for more research to establish the effectiveness of antidepressants in clinically relevant populations under naturalistic conditions, for example, in relapse prevention, in patients with co-morbidities, and in primary care settings.