NMR-Derived Metabolic Profiles Associated with Incident Acute Myocardial Infarction and Pre-Diagnostic Metabolic Trajectories in the UK Biobank

Abstract Acute myocardial infarction (AMI) remains a leading cardiovascular event, yet the metabolic signatures that precede its onset and their value for long-term risk prediction remain incompletely characterized. Using NMR metabolomics data from 271 430 UK Biobank participants without baseline AMI, this study evaluated 168 absolute metabolite measures in relation to incident AMI. Cox proportional hazards models identified 60 AMI-associated metabolites after adjustment for conventional risk factors, and CatBoost modeling with SHAP interpretation and forward selection further defined eight metabolite profiles, including glycine, albumin, glycoprotein acetyls, creatinine, triglycerides in large LDL, free cholesterol in chylomicrons and extremely large VLDL (XXL-VLDL-FC), cholesteryl esters in IDL, and cholesteryl esters in HDL. Integrating these metabolite profiles with conventional risk factors modestly improved 10-year AMI discrimination in both machine learning and survival models, with a CatBoost AUC of 0.741 and a Cox C index of 0.761 in validation analyses. SHAP analyses suggested sex- and age-related heterogeneity in metabolite contributions, particularly for HDL-related and inflammatory traits. A nested case-control trajectory analysis showed persistently lower levels of glycine and albumin, and higher levels of glycoprotein acetyls, large LDL triglycerides, and creatinine during the decade before diagnosis. This study identifies NMR-derived metabolic patterns associated with AMI risk and provides a basis for further investigation of their potential contributions to metabolomics-informed cardiovascular risk assessment.