Commentary on “Deciphering the Risk of Developing Lung Cancer after the Diagnosis of Gastric Cancer with Genetic Evidence: a European and East Asian Populations-Based Mendelian Randomization Analysis” Authors: Fei Ying, Xuan Zhao, Tao Wang Published: 2026 We read with great interest the recent article by Chen and colleagues investigating the bidirectional causal relationship between gastric cancer (GC) and lung cancer (LC) using a two-sample summary-level Mendelian randomization (MR) framework [1]. The authors’ rigorous analytical approach represents an important contribution to understanding shared genetic susceptibility between these malignancies. We wish to offer several suggestions that, while not directly implementable within the published two-sample MR framework, may inform the design of complementary future investigations. First, the current study evaluates GC and LC as unified phenotypes, yet both encompass histological and molecular subtypes with potentially distinct genetic architectures. As a future research direction contingent upon the development of subtype-resolved genome-wide association study (GWAS), stratified MR analyses—reported in accordance with the STROBE-MR guidelines—could reveal heterogeneous causal effects [2]. For GC, comparing intestinal-type versus diffuse-type, or molecular subtypes such as microsatellite instability–high versus microsatellite stable cases, might distinguish inflammation-driven carcinogenesis from epithelial-mesenchymal transition pathways. For LC, MR stratified by driver mutation status—such as EGFR (epidermal growth factor receptor)-mutant LC (prevalent among East Asian never-smokers) versus KRAS-mutant LC—could clarify pathway-specific relationships. We emphasize, however, that the current feasibility of such analyses is constrained by the limited availability and statistical power of subtype- and mutation-defined GWAS summary statistics, and that genome-wide significant, strongly relevant instruments (F-statistic > 10) for these strata remain scarce. Realization of this agenda will require coordinated international consortium efforts to expand subtype-specific GWAS sample sizes. Second, we note the authors’ transparent acknowledgment that fewer covariates were available in the East Asian versus European cohorts. In the MR setting, this limitation is distinct from conventional regression-based covariate adjustment; it pertains instead to the feasibility of multivariable MR (MVMR) extensions, which require valid genetic instruments for each additional exposure satisfying the relevance, independence, and exclusion restriction assumptions, with formal assessment of conditional instrument strength and pleiotropy [3]. A promising future direction would be the development of well-powered GWAS of candidate confounding exposures—such as Helicobacter pylori seropositivity, dietary patterns, and smoking behavior—in East Asian populations, leveraging resources such as Biobank Japan, the China Kadoorie Biobank, and the Taiwan Biobank, to generate the instrument-ready summary statistics necessary for MVMR analyses [4]. Federated, privacy-preserving analytic frameworks may further facilitate such multi-biobank integration [5]. Third, we wish to clarify a point from our initial reading of the manuscript: characterization of how risk evolves after GC diagnosis is not amenable to MR, which by design estimates the effects of lifelong germline genetic exposure. Post-diagnosis surveillance-window characterization is better addressed through prospective longitudinal cohort studies with validated diagnosis dates and follow-up, rather than through any extension of the MR framework. In summary, while comprehensive implementation of these proposals would require the maturation of subtype-specific GWAS resources, MVMR-ready instruments for confounding exposures, and linked longitudinal biobank infrastructure, even incremental progress along these lines would meaningfully advance the precision and clinical translatability of future genetic-epidemiological studies examining shared cancer susceptibility. We thank the authors for their valuable contribution, which provides an important foundation for these future investigations. genome-wide association study. Footnotes Conflicts of Interest Conflict of interest relevant to this article was not reported. References