Toward evidence-based medical statistics: Re-evaluate the efficacy of antidepressants by using Bayes factors

Abstract

IntroductionStudies have been reported the similar efficacy of antidepressants (effect size around 0.3), and it is difficult for clinicians to select an antidepressant. This may partly due to the use of a p<.05 null-hypothesis significance testing (NHST) framework to evaluate “substantial evidence”. This framework only allows dichotomous conclusions and does not quantify the strength of evidence supporting efficacy. In addition, meta-analyses based on publications may offer positively biased results due to selective publications.ObjectivesDemonstrate that the Bayesian framework can provide valuable information on the strength of the evidence for drug efficacy.AimsRe-evaluate the efficacy of FDA-approved antidepressants applied to anxiety disorders and depression by means of Bayes factors.MethodsTo avoid selective publication, data of double-blind placebo-controlled trials for FDA-approved antidepressants for the treatment of anxiety disorders and depression were extracted from the FDA. Bayes factors (BFs) were calculated and compared with the results obtained under NHST framework.ResultsA large variance of evidence for the efficacy of antidepressants was found for both depression and anxiety disorders. Among trials providing “substantial evidence” according to the FDA for anxiety disorders, only 27 out of 59 dose groups obtained strong support for efficacy according to the typically used cut-off of BF≥20. For depression, all FDA-approved antidepressants had BF≥20, except for bupriopion. Moreover, it was shown that the tested antidepressants can be differentiated based on the strength of evidence and effect size.ConclusionsThe BFs quantified the comparative evidence base for the efficacy of antidepressants.Disclosure of interestThe authors have not supplied their declaration of competing interest.