A computable metabolic–visceral adiposity phenotype and self-reported stroke across population-based surveys: cross-sectional and secondary longitudinal analyses

We evaluated the metabolic score for visceral fat (METS-VF) as a candidate computable metabolic–visceral adiposity phenotype associated with self-reported stroke across population-based datasets. Cross-sectional analyses included NHANES 2005–2010, NHANES 2011–2016, and CHARLS; a secondary longitudinal analysis used ELSA. In DAG-adjusted models, each 1-SD increase in METS-VF was associated with higher self-reported stroke prevalence across the three cross-sectional cohorts (ORs, 1.338–1.373). In ELSA, baseline METS-VF was associated with first newly reported physician-diagnosed stroke during follow-up (discrete-time OR, 1.306; 95% CI, 1.030–1.656). Head-to-head comparisons with simple and TyG-derived indices showed modest, dataset-dependent incremental information rather than consistent superiority. External machine-learning transfer from NHANES to CHARLS was limited. METS-VF therefore represents a candidate computable phenotype with reproducible stroke-related associations, but its clinical utility and prospective predictive value require adjudicated longitudinal validation.