Timothy Rebbeck

University of Pennsylvania

PLCO (Learn more about this study)

2007-0047

Feb 15, 2008

Analysis of Case-Control Population Based Genetic Association Studies Using Propensity Scores

This is a Ph.D. thesis project for an Epidemiology student at the University of Pennsylvania. Population-based case-control genetic association studies are potentially susceptible to confounding effects introduced by population stratification. Several methods have been developed to address this issue (Devlin 1999, 2001, Pritchard 1999, 2000, Price 2006, Epstein 2007, Nievergelt 2007). These approaches are primarily focused on controlling population stratification by use of genetic markers such as substructure-informative loci. We believe that the major problem in population based case-control studies is to insure a good match between cases and controls in terms of genetic and non-genetic covariates. We propose a propensity score approach for controlling genetic and non-genetic covariates in population based genetic association studies. A propensity score approach will allow for the control of numerous covariates simultaneously by matching on a single scalar variable. This approach will allow us to account for variations that might not be captured by genetic markers alone and will be able to control for numerous genetic and non-genetic factors simultaneously. The specific aims of this project are (1) to develop a novel statistical approach to control for genetic and non-genetic factors using a propensity score approach; (2) to compare the proposed propensity score method with existing approaches in case-control studies; (3) to extend the propensity score method to study gene-gene and gene-environment interaction effects; and (4) to extend the propensity score method to haplotype case-control studies using a regression based approach. This approach will be developed under a generalized linear model framework so that it can be easily used for analyzing both case-control and cohort studies. These methods will provide a novel and important analytical strategy for obtaining less biased and more valid estimates of the effect of genetic and non-genetic factors for population-based association studies.

Specific Aim 1: To develop a novel statistical approach to control for genetic and non-genetic factors using a propensity score in population based case-control studies.

Specific Aim 2: To compare data analysis approaches using a naïve regression method, genomic control, population structure, principal components, stratification score and propensity score approach using Cancer Genetic Markers of Susceptibility (CGEMS) and prostate cancer data (PLCO).

Specific Aim 3: To extend the propensity score method to study gene-gene and gene-environment interaction effects.

Specific Aim 4: To extend the propensity score method to haplotype case-control studies using regression models to adjust for population stratification and to estimate haplotype effects simultaneously.