Re-Analyzing a Randomized Controlled Trial of Combination Antidepressant Treatment with Mirtazapine: Confidence Intervals Suggest Substantial Uncertainty

This article subjects a randomized controlled trial published in the American Journal of Psychiatry to a methodological and statistical critique, including a re-analysis of the effect size statistics presented. The published trial tested the use of combination antidepressants (mirtazapine co-prescribed with either buproprion, venlafaxine, or fluxoetine) at treatment initiation as compared to fluoxetine monotherapy. The authors report that combination therapy was effective, with a number-needed-to-treat statistic (NNT) of 3-5, a strong effect size. Close scrutiny of the methodology and clinical trial registration shows that 4 of 6 pre-registered outcomes were statistically non-significant, 1 outcome was not reported, and 1 unregistered outcome was published. The well-critiqued Hamilton Depression Inventory was the only positive outcome measure. Calculating confidence intervals for the reported NNT demonstrates substantial uncertainty (95% CI for NNT = 2.3 – 18.0). In an era of evidence-based psychiatric practice, there is insufficient evidence to recommend combination therapy at initiation of treatment.