Credible Mendelian Randomization Studies in the Presence of Selection Bias Using Control Exposures

Area of Science:

• Epidemiology
• Genetic Epidemiology
• Biostatistics

Background:

• Selection bias is a significant limitation in Mendelian randomization (MR) studies.
• Existing methods for assessing selection bias in MR are limited.
• Understanding and mitigating selection bias is crucial for accurate causal inference.

Purpose of the Study:

• To propose and validate a novel method using control exposures to detect selection bias in MR studies.
• To assess the potential effect of higher transferrin on stroke using MR and control exposures.
• To provide a framework for selecting appropriate control exposures in MR analyses.

Main Methods:

• Simulation studies to illustrate the data-generating process of selection bias in MR.
• Conceptualization and application of control exposures (transferrin saturation, iron status) to validate MR estimates.
• Investigation of MR-Pleiotropy RESidual Sum and Outliers (MR-PRESSO) and other statistical methods.

Main Results:

• The application showed inconsistent effects of genetically predicted transferrin and transferrin saturation on stroke, suggesting potential selection bias.
• Genetically predicted iron status showed expected associations with stroke and longevity, indicating no systematic selection bias in that context.
• The proposed control exposure method successfully identified potential biases in the primary MR analysis.

Conclusions:

• Routine use of control exposures is a valuable tool for validating MR estimates and detecting selection bias.
• Appropriate control exposures include antagonists, decoys, or exposures with similar biological activity and shared bias sources.
• An additional validated control exposure with a known outcome association can further explore systematic selection bias.