Leveraging Surrogate Phenotypes to Improve Inference on a Partially Missing Target Phenotype

• Activity Number: 40 - Recent Advances in Statistical Methods for Genome-Wide Association Studies
• Type: Contributed
• Date/Time: Sunday, July 29, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Zachary McCaw* and Xihong Lin
• Companies: Harvard School of Public Health and Harvard University
• Keywords: GWAS; Association testing; Missing data

Abstract

We consider Genome Wide Association Studies (GWAS) in which the phenotype of primary interest is missing at random for a subset of subjects, however surrogates of the target phenotype are available for all subjects. Motivation for this work derives from the study of obstructive sleep apnea (OSA), where the gold standard diagnostic, the apnea-hypopnea index (AHI), is difficult to ascertain, however surrogate outcomes, such as sleep duration and snoring, are plentiful. We propose a genetic association test that jointly models the target and surrogate outcomes. Our framework accommodates phenotype-specific regressions, and allows for both continuous and binary surrogates. The resulting score test is no less efficient than complete case analysis when the surrogate outcome is uncorrelated with the target, but is more powerful when the surrogate is associated with the target. We demonstrate application of our method to genetic association testing of AHI in the Sleep Genetics Epidemiology Consortium (ISGEC), utilizing surrogate phenotypes from the UK Biobank (UKB).