MetsObesity: a novel classification system for predicting 15-year cardiovascular risk in the UK Biobank population

• A novel five-class system (MetsObesity) was developed to stratify cardiovascular risk in individuals with obesity and/or metabolic abnormalities. • The classification integrates anthropometric, metabolic, and additional markers (CRP, liver enzymes, urate) using machine learning techniques. • MetsObesity outperformed BMI, waist circumference, body fat percentage, trunk fat percentage, and metabolic syndrome definitions in predicting major adverse cardiovascular events (MACE). • The model demonstrated strong discrimination, calibration, and clinical utility across a 15-year follow-up in a large UK Biobank cohort (n = 346,001). • MetsObesity provides a scalable, interpretable tool for risk-based prevention and may inform precision cardiometabolic care globally. Obesity phenotyping often fails to capture the heterogeneity in cardiovascular risk. Current approaches rely heavily on BMI or conventional metabolic syndrome criteria, which underestimate cardiometabolic risk in a large proportion of population. We aimed to develop and validate a novel, clinically scalable classification system, MetsObesity, to predict long-term cardiovascular outcomes in people with obesity and/or other metabolic abnormalities. Using data from 346,001 participants in the UK Biobank (median follow-up: 15 years), we applied random forests and recursive feature elimination algorithms to identify key predictors of cardiovascular events across metabolic, anthropometric, and additional abnormalities domains. The SHapley Additive exPlanations (SHAP) model was employed to explore the interpretability of variables selection. These variables informed a five-tier classification system (MetsObesity Classes 1–5). Cox proportional hazard models were used to assess associations with major adverse cardiovascular events (MACE). Model performance was evaluated using discrimination (C-index), calibration, and decision curve analysis. In the validation cohort (n = 173,000), the incidence of MACE was 1.4%, 3.3%, 5.0%, 6.0%, and 8.5% across the five classes, respectively. MetsObesity significantly outperformed BMI, waist circumference, body fat percentage, and existing metabolic syndrome definitions in predicting incident MACE. Participants in Classes 4 and 5 had 2–4-fold higher hazard ratio than traditional obesity classifications. The model demonstrated robust calibration and clinical net benefit across subgroups. Lack of external validation is the key limitation of this study. The MetsObesity offers a novel, interpretable, and implementable framework for cardiovascular risk stratification in people with obesity and metabolic abnormalities. Integration of routinely available clinical variables allows scalable application across healthcare settings. This classification has potential to inform risk-based prevention strategies and redefine cardiometabolic screening in global health systems.