A systematic review and meta-analysis on the association between BDNF val(66)met and hippocampal volume--a genuine effect or a winners curse?

Inconsistenties have been reported with regard to an association between val(66)met, a polymorphism on the BDNF gene, and hippocampal volume. We performed a systematic review and a meta-analysis to determine the magnitude and direction of this putative association and estimated the potential influence of demographic, clinical, and methodological characteristics of studies. Tests of publication bias and time-related trends were performed and statistical power of the included studies was calculated. The literature search for MRI studies on differences in total hippocampal volume as a function of BDNF val(66)met returned 25 records that fulfilled our criteria (total N = 3,620). Meta-analysis showed that carriers of a met allele had lower hippocampal volumes relative to val/val homozygotes (d = 0.13, P = 0.02). Between-study heterogeneity in effect size estimates was substantial (Q = 54.47, P < .001) and this could not be explained by demographic, clinical, and methodological differences across studies. Funnel plot inspection and trim-and-fill estimations suggested evidence for publication bias and effect sizes decreased substantially over the years (Pearson's r = -0.54, P < .01). All included studies were underpowered. This meta-analysis shows that carriers of a met allele have lower total hippocampal volumes relative to val/val homozygotes. However, effect sizes converged closer to null with virtually each attempt at replication and were based on underpowered studies. Altogether, this may call into question whether the observed effect is a genuine biological effect of the met allele or whether it is subject to a winners curse, with large effect sizes found in a few early studies and increasingly smaller effect sizes in later studies.