Abstract: Single nucleotide polymorphisms (SNPs) are the most common type of sequence variations in human genome. They may predispose individual subjects to disease phenotypes or be related to disease progression or treatment efficacy. For genetic association studies conducted in the setting of randomized clinical trials, it is of interest to assess both genetic effects and gene-treatment interactions. These effects are usually quantified by regression parameters in an appropriate regression model, such as logistic regression model if the response is binary or Cox proportional hazards (CPH) regression model if the response is right censored survival time. Standard regression analysis can be applied to estimate parameters of interest. In this work, we apply semiparametric theory to derive a class of consistent and asymptotical normal estimators for genetic main effects and gene-treatment interaction effects, taking into account the fact that genetic variants and baseline auxiliary covariates are independent of treatment assignment due to randomization. We consider both binary and survival response variables. Our method effectively exploits the baseline auxiliary covariates that correlate with the response variable to improve the efficiency of the estimators compared with standard regression analysis.
