The Rise and Risk of MR
The exponential increase in MR publications—from 3,545 in 2023 to over 6,600 in 2024—is a double-edged sword. While the proliferation indicates growing interest and recognition, it also reflects a decline in study quality. A significant portion of recent literature relies heavily on simplified two-sample MR designs and summary-level data, often without the robust framework needed to draw meaningful conclusions. Many of these studies lack careful evaluation of MR assumptions, making them susceptible to misleading or overstated findings. The editorial warns that this trend could erode the field’s credibility, overwhelm journal editors and reviewers, and ultimately diminish MR’s role in evidence-based medicine.
Understanding MR and Its Assumptions
MR infers causality by leveraging the random assortment of genes to mimic the randomisation in clinical trials. To be valid, MR must satisfy three critical assumptions: (1) Relevance: The genetic instruments must strongly correlate with the exposure of interest. (2) Independence: There should be no confounding factors that influence both the instrument and the outcome. (3) Exclusion Restriction: The instrument affects the outcome only through the exposure, not through alternative pathways (i.e., no horizontal pleiotropy). However, assumptions (2) and (3) are rarely testable and often violated in practice, particularly when using summary data or unvetted genetic instruments.
Common Pitfalls in MR Studies
(1) Weak or Non-Genetic Exposures: Some studies have attempted to evaluate exposures that lack clear genetic underpinnings—like air pollution or dietary habits (e.g., noodle consumption). Such exposures are primarily environmental and not suitable for MR unless a gene–environment interaction is biologically justified.
(2) Inadequate Instrument Selection: Instrument choice is central to MR validity. Many studies rely solely on statistical significance from genome-wide association studies (GWAS) without confirming biological plausibility. This can lead to horizontal pleiotropy, where genetic variants influence outcomes via multiple pathways. The authors argue for a balanced strategy that combines statistical strength with biological insight, supported by SNP pruning strategies to reduce linkage disequilibrium bias.
(3) Misuse of Two-Sample MR: The accessibility of large-scale GWAS has led to a surge in two-sample MR studies. While efficient, these studies often neglect confounding, pleiotropy, and sensitivity analyses. The authors caution that replication without rigour creates an illusion of robustness and contributes to publication bloat.
Best Practices for Conducting Robust MR Analyses
The editorial outlines essential criteria for valid MR research: (1) Ensure Genetic Relevance: Exposures must be influenced by well-characterised genetic variants. For instance, HDL-C is genetically regulated by HMGCR variants, making it a valid MR exposure. In contrast, exposures like air pollution require alternative strategies, such as studying metabolite interactions under genetic control. (2) Careful Instrument Selection: Researchers should combine biological plausibility with statistical thresholds. Preferably, instruments should be close to the causal gene and pruned using stringent thresholds (e.g., clump r² ≤ 0.01). Using mechanistic knowledge can reduce pleiotropic bias but may limit the number of usable instruments. (3) Use Multiple Statistical Models: No single model suffices for all MR analyses. Standard models like inverse variance-weighted (IVW) assume all instruments are valid, while more robust alternatives (e.g., MR-Egger, MR-CAUSE, MR-APSS) accommodate violations and reduce type I errors. Employing diverse models strengthens the credibility of findings. (4) Integrate Complementary Evidence: MR findings should be validated using external datasets, colocalisation analyses, or real-world observational studies. When functional experiments are unavailable, in silico methods can provide corroborative evidence. This multi-method approach ensures that MR results are not interpreted in isolation. (5) Cautious Interpretation: Researchers should describe their results as “genetically predicted associations” rather than definitive causal relationships. Comparisons with randomised trials and exploration of mediating pathways can contextualise findings and highlight translational relevance. Furthermore, the effect sizes of genetic associations should be contrasted with those of modifiable risk factors such as medications or behaviours.
Recommendations for the Scientific Community
For Readers:
Look for adherence to STROBE-MR guidelines. Scrutinise genetic plausibility and IV assumptions. Be wary of exaggerated causal claims lacking sensitivity analyses. Cross-reference MR findings with other epidemiological data or RCTs. For Editors and Reviewers:
Reject submissions that analyse weakly genetic or environmental exposures without justification. Demand detailed instrument selection processes and bias assessments. Discourage redundant publications and require robust methodological justification for novel approaches. Encourage interdisciplinary collaborations to bridge genetics, clinical practice, and biological mechanism. Reclaiming MR's Potential
Despite the current challenges, MR remains a transformative method in modern medicine. Its ability to draw causal inferences from observational data is unparalleled when correctly applied. The editorial underscores that methodological rigour, biological reasoning, and cross-validation with complementary evidence can restore confidence in MR research. As the field matures, its future will hinge not on volume, but on quality. To preserve its promise, MR must evolve from a tool of convenience to one of precision.
See the article:
Chen L, Guillot A, Schneider CV. Attention to the misuse of Mendelian randomisation in medical research. eGastroenterology 2025;3:e100187. doi:10.1136/egastro-2025-100187
About eGastroenterology
eGastroenterology is a new, open-access, and open peer-reviewed BMJ Journal, which focuses on basic, clinical, translational, and evidence-based medicine research in all areas of gastroenterology (including hepatology, pancreatology, esophagology, and gastrointestinal surgery). eGastroenterology is now indexed by PubMed, Scopus, CAS, DOAJ, Dimensions, OpenAlex, ROAD, and COPE, with more to come!
For more information, please visit: egastroenterology.bmj.com and follow us on Twitter (@eGastro_BMJ).
Sign-up to Email Alerts for eGastroenterology: https://emails.bmj.com/k/Bmj/jausu/egastroenterology
END