An empirical comparison of meta-analysis and mega-analysis of individual participant data for identifying gene-environment interactions

Meta-analysis combining results from multiple studies is a standard practice in GWAS. For genetic main effects, meta-analysis has been shown to provide comparable results as mega-analysis that jointly analyzes the pooled data from the available studies. Gene-environment interaction (GEI) studies are an important component of genetic epidemiology research since they can explain a part of the missing heritability, elucidate the biological networks underlying disease risk, and identify individuals at high risk for disease. However, it is not known whether meta- and mega-analyses of interactions also yield comparable results. In this study, we investigate whether both approaches provide comparable results for identifying interaction effects using empirical data from 4 studies: the Framingham Heart Study, GENOA, HERITAGE and HyperGEN. We performed meta-analysis of cohort-specific results and mega-analysis by analyzing the pooled data from all 4 studies. We used the standard 1 degree of freedom (df) test of main effect only, the 1 df test of the interaction effect (in the presence of main effect), and the joint 2 df test of main and interaction effects. We found that the results from meta- and mega-analyses were highly consistent for all three tests. The correlation between -log (p) values from the two analyses was 0.89 for the 1 df main effect, 0.90 for the 1 df interaction test, and 0.91 for the joint 2 df test. Although mega-analysis provided slightly better results as expected, both yielded very similar results for the most promising SNPs. Moreover, mega-analysis is not always feasible especially in very large and diverse consortia since pooling of raw data may be limited by the terms of the informed consent. Our study illustrates that meta-analysis can be an effective approach also for identifying interactions in very large consortia without losing appreciable power.