Obtained effect size as a function of sample size in approved antidepressants: a real-world illustration in support of better trial design.

The high failure rate of antidepressant trials has spurred exploration of the factors that affect trial sensitivity. In the current analysis, Food and Drug Administration antidepressant drug registration trial data compiled by Turner et al. is extended to include the most recently approved antidepressants. The expanded dataset is examined to further establish the likely population effect size (ES) for monoaminergic antidepressants and to demonstrate the relationship between observed ES and sample size in trials on compounds with proven efficacy. Results indicate that the overall underlying ES for antidepressants is approximately 0.30, and that the variability in observed ES across trials is related to the sample size of the trial. The current data provide a unique real-world illustration of an often underappreciated statistical truism: that small N trials are more likely to mislead than to inform, and that by aligning sample size to the population ES, risks of both erroneously high and low effects are minimized. The results in the current study make this abstract concept concrete and will help drug developers arrive at informed gate decisions with greater confidence and fewer risks, improving the odds of success for future antidepressant trials.