Common Methods for Performing Mendelian Randomization.

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Mendelian randomization (MR) is a powerful framework for causal inference.
• It leverages genetic variants as instrumental variables to assess causal relationships between exposures and outcomes.
• Recent advances in genome-wide association studies (GWAS) provide rich datasets for MR.

Purpose of the Study:

• To summarize statistical methods for conducting Mendelian randomization analyses.
• To describe sources and types of bias that can arise when MR assumptions are violated.
• To discuss methods for sensitivity analyses and assumption relaxation in MR.

Main Methods:

• Review of statistical methods for Mendelian randomization.
• Explanation of bias in causal estimates due to violated MR assumptions.
• Summary of mathematical formulas for quantifying bias.
• Discussion of sensitivity analysis techniques.

Main Results:

• Large-scale GWAS data enable powerful MR studies.
• Violation of MR assumptions can lead to biased causal estimates.
• Methods exist to estimate the magnitude and direction of potential bias.
• Techniques for sensitivity analyses can help assess the robustness of findings.

Conclusions:

• MR is a valuable tool for causal inference when assumptions are met.
• Understanding and addressing potential biases is crucial for valid MR studies.
• Future research directions include non-linear effects and invalid instrument detection.