This work highlights several pertinent issues for interpreting reports of replication and for conducting future meta-analyses of genetic association studies. First, it was not possible to conduct a traditional meta-analysis using the standards of randomized clinical trials because few of the studies included sufficient descriptive data to conduct a standard meta-analysis. This explains why a previous report included only 5 studies that dichotomized the exposure.50 To supplement our analysis of published data, we also requested original data from many of the authors in order to classify the data in the same way as those of the original study.10 Second, as highlighted in earlier reviews of this topic45,50 and a recent critical review of life stressors across these studies,51 the samples, study designs, measures, and analyses were highly divergent across studies, thereby limiting the comparability of the studies and their evidence regarding replication. For example, several studies only found significant effects for 1 genotype (SS vs SL plus LL or SS plus SL vs LL) rather than the dose effect reported by Caspi et al21,22,24,26,35,37 or found an interaction effect in
