Analyzing genetic association studies with an extended propensity score approach

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Confounding bias is a challenge in observational and genetic association studies.
• Existing propensity score methods are not optimized for genetic data.
• Accurate genetic effect estimation requires robust bias correction.

Purpose of the Study:

• To extend propensity score methodology for genetic association studies.
• To develop a method that simultaneously adjusts for genetic and non-genetic confounders.
• To provide a tool for obtaining less biased estimates of genotype-disease associations.

Main Methods:

• Proposed an extension of a previous multilevel propensity score method.
• Developed the extended genetic propensity score (eGPS) approach.
• Validated the method using simulation studies and a real-world genetic association study.

Main Results:

• The eGPS method demonstrated effective bias adjustment across various simulation scenarios.
• Estimates from eGPS showed minimal bias (mean=0.018, SE=0.01).
• The method preserved key statistical properties including coverage, Type I error, and power.

Conclusions:

• The extended genetic propensity score (eGPS) offers a novel analytical strategy.
• eGPS provides a broadly applicable approach for reducing bias in genetic association studies.
• This method enhances the validity of estimates for genetic associations.