Synergistic impact of insulin resistance and hepatic fibrosis on cardiovascular events in patients with coronary artery disease: A machine learning subtyping and interaction analysis.

BACKGROUND AND AIMS: Insulin resistance (IR) and hepatic fibrosis are significant yet underexplored synergistic risk factors for cardiovascular events in coronary artery disease (CAD). We investigated the interaction between the triglyceride-glucose (TyG) index and liver fibrosis scores (FIB-4, BARD) for risk prediction. METHODS AND RESULTS: Within a prospective cohort of 14,660 CAD patients, we performed a data-driven phenotypic stratification using K-means clustering-an unsupervised machine learning algorithm-to deconvolute heterogeneous metabolic profiles. Subsequently, Cox regression models were employed to evaluate associations of these phenotypes, along with continuous and tertiled TyG, FIB-4, and BARD scores, with incident cardiovascular events (a composite of cardiovascular mortality, nonfatal myocardial infarction, or stroke). The analysis identified three mechanistically distinct metabolic subtypes. During a median follow-up of 3 years, 463 cardiovascular events occurred (overall event rate: 3.16 %). The "metabolic-fibrosis mixed" subtype exhibited the highest risk (adjusted HR = 1.71, 95 %CI:1.34-2.19), with an event rate of 4.40 % (183/4159) compared to 2.43 % (116/4782) in the low-risk subtype. Both the TyG index (HR = 1.52, 95 %CI:1.40-1.65) and BARD score were independent predictors. A significant multiplicative interaction existed between TyG and BARD (P = 0.041). Additive interaction analysis confirmed synergy, with a relative excess risk (RERI) of 1.42 when both biomarkers were elevated. Risk escalated nonlinearly once TyG exceeded 9.0, potentiated by fibrosis. CONCLUSION: IR and hepatic fibrosis synergistically increase cardiovascular risk in CAD patients. Combining TyG and BARD scores with metabolic subtyping enhances risk stratification, potentially guiding targeted interventions.