background Antidepressants remain controversial, partly due to allegations that disappointing results were buried and because of their modest average efficacy.
aim To investigate bias in the antidepressant literature and the possibilities for predicting which patients with depression or anxiety do receive significant benefits from antidepressants.
method We investigated bias by comparing information from the US Food and Drug Administration with the published literature. To predict response, we used patient data from randomized trials.
results Of all studies on depression or anxiety, 50% and 72% were positive, compared to 95% and 96% of all published studies. Safety outcomes were poorly reported in published articles and unpublished studies were often ‘bundled’ into pooled-trials publications with positive conclusions. We found an association between severity and antidepressant efficacy for some, but not all, anxiety disorders; previous research has found inconsistent evidence for this association for depression. Furthermore, patients with depression that showed early improvement were more likely to attain a good response, irrespective of which symptoms improved.
conclusion These results demonstrate the severe impact of bias on the antidepressant literature. Severity and early improvement predicted a good response, but more information is needed to improve predictions. The increased accessibility of individual patient data will hopefully soon enable further progress in this area.