Robust Semiparametric Methods for Analysis of Multivariate Secondary Phenotypes in Case-Control Genome-Wide Association Studies

• Activity Number: 532
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Chuanhua Julia Xing* and Elizabeth Schifano and Xihong Lin
• Companies: XPrecision and University of Connecticut and Harvard T.H. Chan School of Public Health
• Keywords: Multiple secondary phenotype analysis ; the scaled marginal model ; genome-wide association studies ; case-control design ; robustness analysis ; inverse weighted probability

Abstract

Joint analysis of multiple phenotypes in genome-wide association studies (GWASs) is important for detecting pleiotropic effects. For secondary phenotypes in case-control studies, the over-sampling of the cases relative to the primary (disease) phenotype can induce bias in genetic association analysis with secondary phenotypes. To account for this ascertainment bias, we propose using weighted estimating equations for fitting the scaled mean model. In contrast to earlier work by Schifano et al. (2013) where the weight is the inverse of fixed, known sampling fractions/probabilities of the cases and controls, the proposed weight is a population-based measure, defined as the inverse of the conditional probability of case-control status. The proposed method is robust to departures from normality of multiple continuous phenotypes and the misspecification of within-individual correlation across the multiple phenotypes. We show through simulation that the proposed method out-performs other methods and other weighting strategies. We illustrate our approach using a lung cancer case-control study.