Controlling Confounding in (Genetic) Case-Control Studies Using the Stratification Score

• Activity Number: 100
• Type: Invited
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Glen Satten*+ and Andrew Allen
• Companies: CDC and Duke University
• Keywords: Stratification score ; Confounding ; Association ; Genetics ; Propensity score ; Case-control study

Abstract

We recently introduced the stratification score, which is the probability of disease given potential confounders, as a method for control of confounding by population stratification in genetic association studies. We showed that tests using data post-stratified on the stratification score are appropriately sized and have good power to detect association in the presence of confounding. Here we consider the theoretical basis for the stratification score, and show how standardization based on the stratification score can lead to estimates of the difference in exposure (or allele frequency, in the genetic case) that would have been seen if cases and controls had the same distribution of confounding covariables. Real data and simulation studies are used as illustrations.