Pleiotropy-robust Mendelian randomization.

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Mendelian randomization (MR) studies are expanding due to powerful genome-wide association study (GWAS) meta-analyses and large-scale genetic surveys.
• A key assumption in MR, the exclusion restriction, is often challenged by the pleiotropic effects of genetic variants.
• Pleiotropy, where a single genetic variant influences multiple traits, can bias causal inference in MR studies.

Purpose of the Study:

• To develop and validate a novel method for Mendelian randomization that accounts for and corrects pleiotropic effects.
• To relax the strict exclusion restriction assumption in MR analyses.
• To provide unbiased causal effect estimates even when genetic variants exhibit pleiotropy.

Main Methods:

• Developed the pleiotropy-robust Mendelian randomization (PRMR) method, building on instrumental variables (IVs) estimation techniques.
• PRMR estimates the degree of pleiotropy and subsequently corrects for it.
• The method relies on specific assumptions including the existence of a subsample where genetic variants do not affect the exposure and homogeneous pleiotropic effects.

Main Results:

• Simulations demonstrated that existing MR methods yield biased estimators under realistic pleiotropy.
• PRMR produced unbiased estimators in simulations when its underlying assumptions were met.
• The method was practically applied to estimate causal effects of tobacco exposure on BMI, prostate cancer on self-reported health, and educational attainment on BMI using UK Biobank data.

Conclusions:

• PRMR enables the use of instrumental variables that violate the exclusion restriction due to pleiotropy.
• The method effectively corrects for pleiotropy in causal effect estimation.
• PRMR can also serve as a valuable sensitivity analysis tool when the degree of pleiotropy is unknown.