Abstract
OBJECTIVE: To explore the predictive role of matrix metalloproteinases (MMPs) in cardiovascular disease (CVD) outcomes. METHODS: In a cohort of 37,154 UK Biobank participants, we analyzed plasma levels of nine MMPs using baseline samples collected in the United Kingdom between 2006 and 2010, with follow-up for outcomes until April 22, 2024. Cox models estimated Hazard Ratios (HR) and 95% Confidence Intervals (CI) for CVD mortality, morbidity, and subtypes. Machine learning models were built and evaluated using Kaplan-Meier curves, receiver operating characteristic curves(ROC), and SHapley Additive exPlanations (SHAP) for feature importance. RESULTS: MMP-1, -3, -7, -8, -9, and - 12 were associated with increased risk of CVD mortality. MMP-7 (HR: 1.57, 95% CI: 1.37, 1.80) and MMP-12 (HR: 1.69, 95% CI: 1.53, 1.88) had the strongest associations. The MMP-based prediction model achieved high discrimination for CVD mortality (Area Under the Curve [AUC] = 0.89), CVD morbidity (AUC = 0.72), arrhythmia (AUC = 0.69), coronary artery disease (AUC = 0.76), cerebrovascular accident (AUC = 0.81) and heart failure (AUC = 0.81).The SHAP value identified MMP-12 as the most consistent predictor, followed by MMP-7. CONCLUSIONS: MMPs, particularly MMP-12 and MMP-7, strongly predict CVD risk. The MMP-based model shows potential for clinical risk stratification.