Conditioning on parental mating types can reduce necessary assumptions for Mendelian randomization.

Area of Science:

• Epidemiology
• Genetic Epidemiology
• Statistical Genetics

Background:

• Mendelian randomization (MR) is widely used in epidemiology.
• Bias can arise in MR when confounding variables correlate with instrumental variables (genetic variants).
• Assortative mating and population admixture can create such correlations, violating MR assumptions.

Purpose of the Study:

• To propose and evaluate a novel method for reducing bias in Mendelian randomization.
• To address bias caused by correlations between instrumental variables and confounders due to assortative mating or population stratification.

Main Methods:

• Simulated datasets with assortative mating and population admixture.
• Analyzed data using conventional MR, MR with parental genotype adjustment, and MR with parental mating type conditioning.
• Assessed type I error rates and estimation bias.

Main Results:

• Conventional MR showed inflated type I error rates and biased estimates under assortative mating and population admixture.
• Adjusting for parental genotypes offered only partial bias reduction.
• Conditioning on parental mating types effectively eliminated type I error inflation and bias.

Conclusions:

• Conditioning on parental mating types is a superior strategy for Mendelian randomization in the presence of specific confounding.
• This method reduces the assumption burden and potential bias in MR studies.
• It is particularly useful when confounding arises from assortative mating or population stratification.