[Meta-analysis of therapeutic trials: applications in psychiatry]

Meta-analysis is being increasingly used in therapeutic and clinical research to synthetize data from terminated clinical trials. Meta-analysis provides a powerful tool to objectively combine data in a quantitative manner, unlike the classical general review of literature, which is qualitative and subjective by definition, and thus not reproducible, and also the simple data pooling, methods which neglects the statistical heterogeneity between studies. The two most known objectives of meta-analysis are, to provide an objective decision when the trial results have produced contradictory or non-significant results, and to give a better estimation of the magnitude of the treatment effect. Precise methodological rules must be followed, but these rules are not known well enough in the scientific community. For example, particular care is necessary for the retrieval and selection of the trials: all available trials which satisfy predefined criteria of methodological validity should be taken into consideration. One of the main pitfalls in meta-analysis is the problem of unpublished trials, which represents a potential bias seen as an overestimation of treatment effect. In psychiatry, the applications of meta-analysis are numerous, especially in the domain of the treatment of depressive disorders, which will be developed in this paper. Several meta-analyses have been performed to assess the efficacy of antidepressants in major depression, especially with the "new antidepressants", which include the specific serotonine reuptake inhibitors (SSRIs). To date, these meta-analyses have shown that the main advantage of SSRIs was that they were better accepted, but without being more efficacious compared with classic antidepressants. The results of several of these meta-analyses are to be interpreted with the potential biases in mind, especially the publication bias and selection bias. Clinicians must be aware of the difficulties encountered in the choice of outcome criteria, and in deciding how to deal with patients that withdraw from treatment early (intention to treat analysis is the least biased solution for this problem.). Meta-analysis can therefore be helpful in establishing medical references in psychiatry, and to organize consensus conferences, but the limits of this method must be clearly recognized.