A Linear Mixed Model With Measurement Error Correction (LMM-MEC): A Method for Summary-Data-Based Multivariable...

Area of Science:

• Genetics and Epidemiology
• Statistical Genetics
• Causal Inference

Background:

• Multivariable Mendelian randomization (MVMR) methods assess causal effects of multiple risk factors on disease.
• Existing MVMR methods face challenges in accounting for variances in risk factor summary statistics.
• Accurate causal inference requires robust methods that handle uncertainty in genetic and phenotypic data.

Purpose of the Study:

• To propose a novel linear mixed model with measurement error correction (LMM-MEC) for MVMR.
• To address the challenge of accounting for variances in summary statistics for both disease outcomes and risk factors.
• To evaluate the performance of LMM-MEC compared to existing MVMR methods under various conditions, including pleiotropy and linkage disequilibrium.

Main Methods:

• Developed a linear mixed model (LMM) to account for variance in disease summary statistics (fixed- or random-effects).
• Relaxed the NOME assumption and incorporated estimation error from risk factor summary statistics using regression calibration.
• Validated the LMM-MEC method through simulation studies and an application to cholesterol biomarkers and longevity.

Main Results:

• LMM-MEC demonstrated comparable performance to existing MVMR methods under no or balanced pleiotropy.
• The method showed improved coverage rates and power under directional pleiotropy compared to some existing methods.
• In an application study, LMM-MEC identified a causal association between high LDL-c levels and a lower likelihood of longevity, using 739 genetic variants with low linkage disequilibrium.

Conclusions:

• The proposed LMM-MEC method effectively accounts for variances in summary statistics in MVMR analyses.
• LMM-MEC offers improved performance, particularly under directional pleiotropy and specific linkage disequilibrium scenarios.
• The study highlights a causal link between elevated LDL-c and reduced longevity, demonstrating the utility of LMM-MEC in genetic epidemiology.