Pulse Brain · Growing Health Evidence Index
Tier 3 — Observational / field trialPeer-reviewed

Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator

Jack Bowden, George Davey Smith, Philip Haycock, Stephen Burgess

Genetic Epidemiology · 2016

Read source ↗ All evidence

Summary

This methodological paper addresses a critical problem in Mendelian randomization: the conventional inverse-variance weighted method produces biased estimates when some genetic instrumental variables are invalid. The authors present a novel weighted median estimator that remains consistent even when up to 50% of information derives from invalid instruments, offering improved Type 1 error rates in finite samples. The method is demonstrated through simulation and application to lipid–coronary artery disease risk, where it reveals null effects of HDL cholesterol consistent with experimental evidence, and is proposed as a complementary sensitivity analysis alongside MR-Egger regression.

UK applicability

This is a statistical methodology paper with no direct application to UK farming systems or soil health. However, the methods may be relevant to UK researchers conducting Mendelian randomization analyses of dietary or nutritional exposures and health outcomes, improving the robustness of causal inference in nutritional epidemiology.

Key measures

Type 1 error rates, consistency of causal estimates, performance compared to inverse-variance weighted and MR-Egger regression methods

Outcomes reported

The study developed and validated a weighted median estimator for Mendelian randomization that produces consistent causal estimates even when up to 50% of genetic variants are invalid instrumental variables. The method was evaluated through simulation analysis and applied to lipid–coronary artery disease associations.

Theme
Measurement & metrics
Subject
Measurement methods & nutrient profiling
Study type
Research
Study design
Methodological development with simulation analysis and empirical application
Source type
Peer-reviewed study
Status
Published
System type
Laboratory / in vitro
DOI
10.1002/gepi.21965
Catalogue ID
BFmor3gaas-czzcqc

Topic tags

Pulse AI · ask about this record

Dig deeper with Pulse AI.

Pulse AI has read the whole catalogue. Ask about this record, its theme, or how the findings apply to UK farming and policy — every answer cites the underlying studies.