MRSamePopTest: introducing a simple falsification test for the two-sample mendelian randomisation 'same population'...

Area of Science:

• Epidemiology
• Statistical Genetics
• Causal Inference

Background:

• Two-sample Mendelian randomization (MR) is widely used for causal inference in epidemiology.
• The validity of MR relies on the assumption that variant-exposure and variant-outcome associations originate from comparable populations.
• Current methods often assess population comparability by checking demographic similarity, which may be insufficient.

Purpose of the Study:

• To propose an easy-to-implement falsification test for the 'same-population' assumption in two-sample MR studies.
• To enhance the reliability and generalizability of causal effect estimates derived from MR.
• To facilitate the design of MR studies utilizing diverse populations.

Main Methods:

• Proposed a novel falsification test based on the exchangeability of effect modifiers.
• Suggested testing the homogeneity of variant-phenotype associations for phenotypes measured in both genetic association studies.
• Developed a user-friendly R package to implement the proposed test.

Main Results:

• The proposed test offers a practical approach to evaluate the 'same-population' assumption in two-sample MR.
• This method can identify potential population heterogeneity that might affect causal inference.
• The R package simplifies the application of this sensitivity analysis.

Conclusions:

• The 'same-population' assumption is critical for valid two-sample MR.
• The proposed homogeneity test provides a valuable tool for assessing this assumption.
• Increased attention to sensitivity analyses, including this test, can improve MR study design and interpretation.