Using genetic path analysis to control for pleiotropy in a Mendelian randomization study

Abstract

BackgroundWhen a randomized experimental study is not possible, Mendelian randomization studies use genetic variants or polygenic scores as instrumental variables to control for gene-environment correlation while estimating the association between an exposure and outcome. Polygenic scores have become increasingly potent predictors of their respective phenotypes, satisfying the relevance criteria of an instrumental variable. Evidence for pleiotropy, however, casts doubt on whether the exclusion criteria of an instrumental variable is likely to hold for polygenic scores of complex phenotypes, and a number of methods have been developed to adjust for pleiotropy in Mendelian randomization studies.\n\nMethodUsing multiple polygenic scores and path analysis we implement an extension of genetic instrumental variable regression, genetic path analysis, and use it to test whether educational attainment is associated with two health-related outcomes in adulthood, body mass index and smoking initiation, while estimating and controlling for both gene-environment correlations and pleiotropy.\n\nResultsGenetic path analysis provides compelling evidence for a complex set of gene-environment transactions that undergird the relations between educational attainment and health-related outcomes in adulthood. Importantly, results are consistent with education having a protective effect on body mass index and smoking initiation, even after controlling for gene-environment correlations and pleiotropy.\n\nConclusionsThe proposed method is capable of addressing the exclusion criteria for a sound instrumental variable and, consequently, has the potential to help advance Mendelian randomization studies of complex phenotypes.