Robustness of meta-analyses in finding gene × environment interactions

Area of Science:

• Genetics
• Statistical genetics
• Epidemiology

Background:

• Meta-analyses are crucial for synthesizing genetic study evidence.
• Genome-wide association studies (GWAS) focus on genetic main effects, with growing interest in gene × environment interactions.
• Standard analysis of gene × environment interactions includes confounders as covariates, but may fail if covariate × environment interactions exist.

Purpose of the Study:

• To evaluate the robustness of meta-regression and joint meta-analysis to covariate × environment confounders.
• To compare the performance of these two meta-analysis methods in the context of gene × environment interaction studies.

Main Methods:

• Simulation studies were conducted to assess the impact of covariate × environment confounders.
• Two common meta-analysis methods, meta-regression and joint meta-analysis, were evaluated.
• The analysis focused on testing gene × environment interactions and joint effects.

Main Results:

• Meta-regression demonstrated robustness against covariate × environment confounders.
• Joint meta-analysis exhibited susceptibility to confounding, leading to inflated type I error rates.
• Non-significant covariate × environment interactions at the study level can significantly increase consortium-level type I errors.

Conclusions:

• Meta-regression offers a robust approach for analyzing gene × environment interactions in the presence of covariate × environment confounders.
• Joint meta-analysis requires inclusion of covariate × environment terms at the study level to control type I errors.
• These findings are critical for large-scale genome-wide gene × environment interaction studies.