Efficient design for Mendelian randomization studies: subsample and 2-sample instrumental variable estimators.

Area of Science:

• Epidemiology
• Biostatistics
• Genetic Epidemiology

Background:

• Mendelian randomization (MR) estimates causal relationships using genetic instrumental variables (IVs).
• Traditional MR requires complete data on IVs, exposure, and outcomes.
• High costs or lack of biospecimens can hinder complete exposure data collection.

Purpose of the Study:

• To assess the statistical power and bias of MR when exposure data is available for a subset of participants.
• To evaluate the cost-efficiency of partial exposure data collection in MR studies.
• To determine the impact of subset size and IV strength on MR analysis.

Main Methods:

• Utilized simulated datasets to evaluate Mendelian randomization (MR) under partial exposure data scenarios.
• Assessed statistical power and bias in relation to subset size and instrumental variable (IV) strength.
• Examined various confidence interval calculation methods for MR with subset data.

Main Results:

• Collecting exposure data from a subset of participants is a cost-efficient MR strategy.
• Partial data collection often has negligible effects on statistical power compared to complete data.
• Maximum power depends on IV strength and approaches that of traditional IV estimators.
• Weak IVs can introduce bias towards the null (small subset) or confounded association (large subset).

Conclusions:

• Partial exposure data collection enhances the cost-efficiency and feasibility of Mendelian randomization (MR) studies.
• The findings support using subsets for exposure data in MR, especially when full data is prohibitive.
• Careful consideration of subset size and IV strength is crucial for accurate MR inference.