Abstract
BACKGROUND: Cardiovascular disease (CVD) remains the leading global cause of death. While metabolic dysregulation is central to CVD pathogenesis, the extent to which distinct clinical subtypes exhibit unique or shared metabolic signatures remains unclear. OBJECTIVES: The purpose of this study was to systematically characterize metabolomic patterns across a broad spectrum of CVD subtypes and to delineate both shared and subtype-specific metabolic features. METHODS: We analyzed nuclear magnetic resonance-based metabolomic data (325 metabolites) from 244,567 UK Biobank participants, including 27,950 with prevalent CVDs classified into 87 phenotypes (37 classes, 50 subclasses) using International Classification of Diseases-10th Revision codes. Logistic regression, random forest, and XGBoost models assessed cross-sectional metabolite-disease associations. SHapley Additive exPlanations analysis identified key discriminative features. Internal geographic validation used Scotland and Wales cohorts. RESULTS: Metabolomic profiles demonstrated substantial heterogeneity across CVD subtypes. We identified 21 metabolites consistently associated with multiple conditions, including Intermediate-Density Lipoprotein cholesteryl esters, linoleic acid percentage, and small very low_density lipoprotein particles, reflecting shared alterations in lipoprotein metabolism and inflammatory pathways. Cross-sectional discrimination models achieved moderate-to-high performance for prevalent disease status (eg, chronic ischemic heart disease area under the curve = 0.876). Disease similarity clustering revealed reproducible organizational structures: ischemic entities formed tight clusters, hypertensive-renal diseases showed graded patterns, while rheumatic and pulmonary conditions remained distinct. These patterns were confirmed in geographic validation cohorts. CONCLUSIONS: This comprehensive metabolomic atlas reveals both shared and subtype-specific metabolic alterations across prevalent CVD. The identified metabolite set provides a hypothesis-generating framework for understanding cardiovascular metabolic heterogeneity. However, the cross-sectional design and inclusion of treated patients preclude causal or predictive inference, requiring validation in prospective, treatment-naive cohorts.