Background:
Mendelian randomisation (MR) has rapidly become a go-to method for causal inference in population-scale studies. The growing availability of East Asian biobank data is accelerating this trend—but methodological rigour has not always kept pace. Recent critiques highlight a broader credibility crisis in MR research, especially when instruments derived from European populations are used without appropriate validation. Whether these concerns extend to the rising body of East Asian MR studies remains under-examined.
Methods:
We first conducted a systematic review of MR studies involving East Asian (EAS) populations published between 2014 and 2023. An AI-assisted workflow was used to extract structured study characteristics, including design type, ancestry alignment, instrument selection, and sensitivity analyses. We applied a purpose-built scoring rubric incorporating STROBE-MR and cross-ancestry methodological guidance, covering 23 core MR quality items (Part A) and 15 ancestry-related items (Part B). Two-sample EAS MR studies were classified into: (i) ancestry-matched (EAS–EAS), (ii) European (EUR) instruments validated in EAS (EUR-val-EAS), and (iii) unvalidated European instruments applied in EAS (EUR–EAS).
Results:
From 567 screened articles, 253 studies met inclusion criteria. Two-sample MR designs surged after 2019, driven largely by studies from China and Japan. Among the 194 eligible two-sample studies, reporting of instrument strength, linkage disequilibrium clumping, and harmonisation procedures were often incomplete. Rubric scores revealed a clear quality gradient: EAS–EAS studies scored highest (median Part A=20, Part B=NA), EUR-val-EAS intermediate (Part A=20; Part B=4), and EUR–EAS lowest (Part A=10; Part B=7).
Conclusions:
As MR continues to expand into global populations, ensuring methodological robustness—particularly in cross-ancestry contexts—is vital. Our review highlights both progress and persistent gaps, underscoring the need for tools that support ancestry-aware MR practices and more credible causal inference in diverse settings.