Informed conditioning on clinical covariates increases power in case-control association studies

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Genetic association studies often collect clinical covariate data (e.g., BMI, smoking status, age).
• These covariates can modify genetic risk, impacting the interpretation of genetic associations.
• Existing methods for handling covariates in ascertained case-control studies have limitations in maximizing statistical power.

Purpose of the Study:

• To develop and validate a novel statistical method for genetic association studies that effectively incorporates clinical covariates.
• To enhance statistical power in case-control studies, particularly under non-random ascertainment of phenotype and covariates.
• To improve the identification of genetic variants associated with complex diseases.

Main Methods:

• Developed an informed conditioning approach based on the liability threshold model.
• Incorporated external epidemiological information to inform model parameters.
• Accounted for disease prevalence, phenotype ascertainment, and covariate ascertainment.
• Compared the new method against standard case-control tests, gene-covariate interaction tests, and prior covariate handling methods.

Main Results:

• The informed conditioning approach significantly increases statistical power while maintaining a controlled false-positive rate.
• Outperformed standard methods, especially in case-control-covariate ascertained studies.
• Empirically validated in large datasets (89,726 samples) across seven common diseases (type 2 diabetes, prostate cancer, lung cancer, breast cancer, rheumatoid arthritis, age-related macular degeneration, end-stage kidney disease).
• Demonstrated superior performance over logistic regression for 115 of 157 known variants (P < 1x10^-9), with a median 16% increase in chi-squared test statistics.

Conclusions:

• The informed conditioning method offers a substantial improvement in power for genetic association studies by leveraging clinical covariate data.
• This approach is robust to non-random ascertainment and provides a more accurate estimation of genetic effects.
• Application of this method to existing and future studies holds promise for discovering novel disease-associated genetic loci.