Bayesian meta-analyses of the tolerability of selective serotonin reuptake inhibitors and tricyclic antidepressants for treating patients with depression

Abstract

Background. This study aimed to compare, via meta-analysis, the tolerability of selective serotonin reuptake inhibitors (SSRIs) with tricyclic antidepressants (TCAs) for treating depression. A Bayesian model was developed and the results were compared with the classical models. Methods. The outcome of interest was the combined odds ratio of premature withdrawal from treatments due to drug-related side-effects. Two separate meta-analyses - one for primary-care and one for the general setting - were conducted. Unlike the classical approach, the proposed Bayesian model allows researchers to combine their expert opinions with published data. The iterative Markov Chain Monte Carlo method (written in Stata 7.0) was applied for generating the posterior distributions for Bayesian analyses. Results. The classical models showed that significantly fewer patients receiving SSRIs withdrew prematurely due to drug-related side effects. The Bayesian models based on SSRI-favoured priors gave similar results, although the interpretation was philosophically different. Moreover, the likelihoods for the analyses were rather weak, so one must interpret the results with extra care. Conclusion. The Bayesian analyses with SSRI-favoured priors revealed that SSRIs might be better tolerated than TCAs in the general setting. However, as the posteriors were strongly influenced by the priors, more studies need to be conducted before any conclusion can be made. The Bayesian models provided more insights to the problem and the nature of data selected for meta-analyses.