Identification of effect modifiers using a stratified Mendelian randomization algorithmic framework.

Area of Science:

• Genetics
• Epidemiology
• Statistical genetics

Background:

• Mendelian randomization (MR) uses genetic data to infer causal relationships.
• Stratified MR methods like DRMR and residual stratification MR identify nonlinearities but have limited application for effect modifier detection.
• Identifying effect modifiers is crucial for personalized medicine and understanding differential risk factor impacts.

Purpose of the Study:

• To develop and validate a stratified MR algorithm for identifying effect modifiers of causal relationships.
• To adapt existing stratified MR techniques for broader application in detecting effect modification.
• To apply the algorithm to large-scale biobank data to uncover novel effect modifiers.

Main Methods:

• Developed a stratified MR algorithm adapting doubly-ranked MR (DRMR) and residual stratification MR.
• Validated the algorithm through simulations, assessing robustness to nonlinearity and collider bias with binary and continuous outcomes.
• Applied the algorithm to 1,715 exposure-stratifying variable-outcome combinations in the UK Biobank.

Main Results:

• The algorithm demonstrated robustness in simulations for detecting nonlinear relationships and handling collider bias.
• Two statistically significant effect modifiers were identified in the UK Biobank data.
• The causal effect of body mass index on type 2 diabetes mellitus was found to be attenuated by age.
• The causal effect of LDL cholesterol on coronary artery disease was exacerbated by increased serum urate levels.

Conclusions:

• Introduced a novel stratified MR tool for detecting effect modifiers in causal relationships.
• Identified age and serum urate as significant effect modifiers for cardiometabolic disease risk factors.
• Findings have clinical implications for personalized risk assessment and targeted interventions in cardiometabolic diseases.