Inferring genetic causal effects on survival data with associated endo-phenotypes

Age-at-onset phenotypes are important traits in genetic association analyses. Often, intermediate phenotypes that are related to the age-at-onset phenotype are also associated with the marker loci that are associated with the age-at-onset phenotype. In order to understand the genetic etiology of the observed associations, statistical methodology is needed to distinguish between a direct genetic effect on the age-at-onset phenotype and an indirect effect induced by the genetic association with the endo-phenotype that is correlated with the age-at-onset phenotype. In this communication, we introduce a new statistical approach to detect causal genetic effects on survival data in the presence of genetic associations with secondary phenotypes that might influence survival as well and thereby induce seemingly causal relationships. Derived using causal inference methodology, the proposed method is based on standard statistical methodology and can be implemented straight-forwardly, using standard software. Using simulation studies, the theoretical properties of the approach are verified and the power is assessed under realistic scenarios. The practical relevance of the approach is illustrated by an application to survival after cardiac surgery, where genetic components of myocardial infarctions are determined to not influence post-surgery hospital duration except through the MI-pathway.