Genome-wide gene-environment interactions on quantitative traits using family data.

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Gene-environment interactions (GxE) are crucial for personalized interventions.
• Genome-wide GxE studies necessitate large sample sizes, often from consortia.
• Family studies can contribute but require methods to handle within-family correlations.

Purpose of the Study:

• To evaluate statistical methods for analyzing GxE with binary exposures and quantitative outcomes in family studies.
• To compare the performance of generalized estimating equations (GEE) and linear mixed-effects models.

Main Methods:

• Simulated cross-sectional and longitudinal data with family structures.
• Analysis using generalized estimating equations (GEE) and linear mixed-effects models.
• Assessment of type I error rates under various model specifications and exposure prevalences.

Main Results:

• Both GEE and mixed models perform well with sufficient exposure prevalence and correct specification.
• Mixed models exhibit inflated type I error rates when models are misspecified.
• GEE with robust variance estimates are less sensitive to misspecification but may need adjustments for infrequent exposures.

Conclusions:

• GEE methods offer a more robust approach for GxE analysis in family studies, especially when model misspecification is a concern.
• Careful consideration of model assumptions and exposure frequency is necessary for accurate GxE inference in family-based consortia.