Meta-Analysis of Published and Unpublished Studies

META-ANALYSIS OF PUBLISHED AND UNPUBLISHED STUDIES

In addition to individual published trial analysis, there has been a meta-analysis of registration studies from the United Kingdom (Whittington et al., 2004). The authors divided the studies into peer-reviewed published and unpublished studies residing with the Committee on Safety of Medicines (CMS). Efficacy measures included remission, response to treatment and depression symptom scores. Safety measures included serious adverse events (AEs), suicide-related behaviours and discontinuation of treatment because of AEs. The results were startling: published stud-ies reported efficacy, but the effect was lost when assessed along with results from unpublished studies. Overall, the authors found limited empirical support for the use of SSRIs to treat major depressive disorder (MDD) in youth: fluoxetine use was supported but one paroxetine and two sertraline trials had equivo-cal or weak positive risk–benefit profiles. However, in these two cases, the addition of unpublished data shifted the results to an unfavourable risk–benefit profile. Data from citalopram and venlafaxine showed unfavourable risk–benefit profiles.

The meta-analysis is consistent with UK recommen-dations for a contraindication on the use of all SSRIs except fluoxetine to treat childhood depression. By contrast, the FDA has been silent on the efficacy issue focusing its attention entirely on the question of SSRI safety with respect to treatment-emergent suicidality.

From a research methodology perspective, unpub-lished trial data are a source of publication bias and render meta-analytic approaches virtually meaning-less. Demands for an end to tolerate unpublished studies based on the so-called proprietary rights have been advocated within US child psychiatry (Zito et al., 2004). In addition, leading academic medi-cal journals have announced new rules restricting publication unless trials have been registered and all relevant data are available for review (DeAngelis et al., 2005). The journal policy strengthens the goal of registration of all trials in a government-sponsored database (http://www.clinicaltrials.gov). Thus far, database compliance has been increasing but is in need of substantial improvement especially regard-ing completion of the field for recording the primary outcome measurement (Zarin, Tse and Ide, 2005).