Powerful three-sample genome-wide design and robust statistical inference in summary-data Mendelian randomization.

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Mendelian randomization (MR) is a popular method for estimating causal effects of risk exposures.
• Increasing sample sizes in Genome-Wide Association Studies (GWAS) allow for the use of weaker genetic instruments.
• Weak instruments can introduce challenges in MR analyses.

Purpose of the Study:

• To develop and validate a novel statistical approach for MR that effectively utilizes numerous weak genetic instruments.
• To improve the efficiency and robustness of MR analyses in the context of large-scale GWAS data.
• To re-evaluate the causal relationships between body mass index (BMI), blood lipids, and cardiovascular disease (CVD) outcomes.

Main Methods:

• A three-sample, genome-wide MR design incorporating approximately 1000 independent genetic instruments.
• An empirical partially Bayes statistical analysis weighting instruments by their strength.
• Robustness checks for balanced and sparse horizontal pleiotropy.

Main Results:

• The proposed method yields substantially shorter confidence intervals (CIs) for causal effect estimates.
• A significant causal effect of BMI on ischemic stroke (OR 1.19, 95% CI: 1.07-1.32) was identified.
• The causal effect of high-density lipoprotein cholesterol (HDL-C) on coronary artery disease (CAD) was estimated (OR 0.78, 95% CI: 0.73-0.84), but became non-significant using only strong instruments.

Conclusions:

• Genome-wide designs significantly enhance the statistical power of MR studies.
• While robust methods mitigate pleiotropy, they do not fully resolve it.
• The heterogeneous relationship between HDL-C and CAD suggests further investigation is warranted.