A Cautionary Note on Testing for Association of Low-Frequency Variants in Case-Control Studies

• Activity Number: 357
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Guan Xing*+ and Chao Xing
• Companies: Bristol-Myers Squibb and The University of Texas Southwestern Medical Center
• Keywords: Wald test ; case-control study, ; low-frequency variation

Abstract

Wald test is commonly employed test in genetic association studies because of its computational simplicity. Hauck and Donner (1977) observed an anomalous behavior of Wald test that in a binary logit model, as the distance between the parameter estimate and the null value increases, the test statistic decreases to zero. The human genetic mapping enters an era of whole-genome sequencing aiming at directly identifying disease predisposing variants, the frequency of which is constrained to be low under the pressure of selection. In a sample of case-control study, it is likely that this variant appears only scarcely in cases, but not at all in controls. Thus the anomalous behavior of Wald test raises concerns on using it. In this study, we compared the behavior of four tests: likelihood ratio test, Wald test, score test, and exact test in such a genetic scenario with and without considering multiple covariates. We further compared the methods in a real data of whole-genome nonsynonymous variants screening. To conclude, the Wald test should be used with caution in testing for association of low-frequency variants in case-control studies.