Impact of nonrandom selection mechanisms on the causal effect estimation for two-sample Mendelian randomization...

Area of Science:

• Epidemiology
• Statistical Genetics
• Biostatistics

Background:

• Nonrandom selection in one-sample Mendelian Randomization (MR) can lead to biased estimates and inflated type I error rates.
• In two-sample MR, differing selection mechanisms across samples can severely impact causal effect estimation.

Purpose of the Study:

• To propose sufficient conditions for causal effect invariance under different selection mechanisms in two-sample MR.
• To evaluate the impact of various selection mechanisms on causal effect estimation and the performance of pleiotropy-robust MR methods.

Main Methods:

• Simulation study considering 49 selection mechanisms based on genetic variants (G), exposures (X), and outcomes (Y).
• Comparison of eight pleiotropy-robust MR methods under different selection scenarios.
• Application to investigate the effect of obesity on HbA1c levels.

Main Results:

• Nonrandom selection in the second sample (sample II) has a greater influence on bias and type I error rates than in the first sample (sample I).
• Selections depending on X+Y, G+Y, or G+X+Y in sample II cause larger biases.
• Pleiotropy, particularly directional pleiotropy, combined with nonrandom selection, severely impacts MR method performance, violating the InSIDE assumption.

Conclusions:

• Nonrandom selection in two-sample MR significantly exacerbates bias in causal effect estimation for pleiotropy-robust methods.
• The choice of selection mechanism and the presence of pleiotropy are critical factors influencing the reliability of two-sample MR findings.
• Application results suggest nonrandom selection magnified the estimated causal effect of obesity on HbA1c levels.