MR-BOIL: Causal inference in one-sample Mendelian randomization for binary outcome with integrated likelihood method.

Area of Science:

• Statistical genetics
• Epidemiology
• Causal inference

Background:

• Mendelian randomization (MR) uses genetic variants as instrumental variables to infer causality.
• Existing MR methods struggle with binary outcomes due to confounding factors and logistic model limitations.
• Accurate causal inference is crucial in understanding disease etiology and developing interventions.

Purpose of the Study:

• To develop a novel method, MR-BOIL, for robust causal inference in one-sample Mendelian randomization with binary outcomes.
• To address the bias introduced by unobserved confounders in previous approaches.
• To improve the reliability and statistical power of causal effect estimation.

Main Methods:

• Proposed an integrated likelihood method (MR-BOIL) treating confounders as latent variables.
• Utilized an expectation-maximization algorithm for parameter estimation under joint normality assumption of confounders.
• Validated the method through extensive simulations and application to real-world data.

Main Results:

• MR-BOIL provides asymptotically unbiased causal effect estimates for binary outcomes.
• The method enhances statistical power without increasing the type I error rate.
• Application to Atherosclerosis Risk in Communications Study data demonstrated superior reliability in identifying causal relationships compared to existing methods.

Conclusions:

• MR-BOIL offers a reliable and statistically powerful approach for causal inference in one-sample Mendelian randomization with binary outcomes.
• The method effectively handles confounding factors, improving the accuracy of genetic epidemiology research.
• Available R code facilitates the application of MR-BOIL in future studies.