Mendelian randomization studies for a continuous exposure under case-control sampling.

Area of Science:

• Epidemiology
• Genetic Epidemiology
• Biostatistics

Background:

• Mendelian randomization (MR) uses genetic variants as instrumental variables to infer causal relationships between an exposure and a disease.
• Case-control sampling is a common design in genetic epidemiology but can introduce bias in MR analyses.
• The two-stage instrumental variables (2SIV) method is a standard approach for MR.

Purpose of the Study:

• To evaluate the impact of case-control sampling on Mendelian randomization analyses.
• To compare different estimators for the first-stage association and their effect on causal effect estimates in MR.

Main Methods:

• Theoretical development and simulation studies were employed.
• Comparison of naïve, inverse probability weighted (IPW), and maximum likelihood (ML) estimators for the first-stage association.
• Evaluation of 2SIV estimates of causal effect and comparison with structural mean models (SMM) using double-logistic regression.

Main Results:

• The naïve estimator is biased under the alternative but unbiased under the null hypothesis.
• The maximum likelihood estimator provides smaller variance and mean squared error compared to other estimators.
• Structural mean models yield the smallest bias but often have larger variance and convergence issues.

Conclusions:

• Case-control sampling necessitates careful adjustment in MR analyses.
• Maximum likelihood estimation is recommended for improved precision in MR under case-control sampling.
• Structural mean models offer low bias but require careful implementation due to potential instability.