Secondary Trait Analysis for Case-Control Association Studies in the Presence of Covariates

• Activity Number: 514
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Godwin Yung*+ and Xihong Lin
• Companies: Harvard and Harvard School of Public Health
• Keywords: secondary trait ; case-control ; association study ; model misspecification ; secondary analysis

Abstract

Case-control genome-wide association studies often collect from their subjects extensive information on secondary traits. Reusing the data and studying the association between genes and secondary traits provide an attractive and cost effective approach that can lead to discovery of new genetic associations. For that purpose, a number of approaches have been considered, including standard methods that ignore ascertainment or stratify on case-control status, and more complex methods based on weighted or semi-parametric likelihoods. However, justification for a number of these approaches rely on the assumption of no covariates or a logistic penetrance model for the primary disease. Of course, neither are necessarily true in practice. In this paper we carry out an extensive investigation of robustness for standard methods in the presence of covariates and nuisance model misspecification. We show that in practical scenarios, standard methods have proper type I error rates when the disease is not associated with the SNP. However, contrary to popular belief and with exception of the controls-only analysis, they may be inappropriate when the disease is rare.