Mendelian randomization with Egger pleiotropy correction and weakly informative Bayesian priors.

Area of Science:

• Biostatistics
• Genetic Epidemiology
• Statistical Genetics

Background:

• The MR-Egger (MRE) estimator corrects for pleiotropy in instrumental variable (IV) analysis but has limited statistical power.
• Conventional IV estimators lack the ability to correct for directional pleiotropy.
• There is a need for more powerful methods to address pleiotropic effects in genetic studies.

Purpose of the Study:

• To introduce a Bayesian implementation of the MR-Egger estimator (BMRE).
• To evaluate the utility of weakly informative priors on the intercept term for enhancing the power of IV slope estimates.
• To compare the performance of BMRE against existing IV estimators under various simulation scenarios.

Main Methods:

• A simulation study was conducted to compare different IV estimators.
• Scenarios included variations in causal effects, pleiotropy, proportion of pleiotropic instruments, and 'Instrument Strength Independent of Direct Effect' (InSIDE) assumption violations.
• A method for determining prior distributions for pleiotropy was developed using empirical plasma urate data.

Main Results:

• Applying a weakly informative prior on the intercept term improved the power of the slope estimate in BMRE.
• Type 1 error rates were maintained near the nominal 0.05 level.
• Under the InSIDE assumption, pleiotropy did not affect estimator performance; however, InSIDE violations biased all estimators, with BMRE being more affected than MRE.

Conclusions:

• The BMRE estimator provides increased power at the expense of greater susceptibility to InSIDE assumption violations.
• BMRE represents a trade-off between MRE and conventional IV estimators.
• BMRE is a valuable tool for accounting for observed pleiotropy in IV analyses.