Association of NT-proBNP/LVEF with angiographic coronary disease burden in acute myocardial infarction: a retrospective study with machine learning analyses

Background N-terminal pro-B-type natriuretic peptide (NT-proBNP) and left ventricular ejection fraction (LVEF) reflect complementary dimensions of ventricular stress and systolic function. Whether the NT-proBNP/LVEF ratio is consistently associated with angiographic coronary disease burden in patients with acute myocardial infarction (AMI), and whether its predictive contribution is robust across conventional and machine-learning approaches, remains uncertain. Methods This single-center retrospective study included 375 patients with AMI who underwent coronary angiography. The continuous Gensini score was the primary outcome, while clinically anchored anatomical outcomes, including multivessel disease and high-burden coronary anatomy, served as sensitivity outcomes. Associations between log-transformed NT-proBNP/LVEF and Gensini score were assessed using multivariable median regression. Direct discrimination of NT-proBNP/LVEF versus NT-proBNP alone was evaluated using the area under the receiver operating characteristic curve (AUC) and paired DeLong testing; the incremental information provided by adding LVEF to NT-proBNP was additionally assessed using continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Six multivariable models—logistic regression, least absolute shrinkage and selection operator (LASSO) logistic regression, random forest, support vector machine (SVM), extreme gradient boosting (XGBoost), and CatBoost—were internally validated using leakage-controlled repeated fivefold cross-validation with five repeats. Calibration, decision curve analysis, and Shapley additive explanations (SHAP)-based model interpretation were also evaluated. Results Higher NT-proBNP/LVEF was consistently associated with greater angiographic coronary disease burden. In the fully adjusted model, each doubling of NT-proBNP/LVEF was associated with a 12.20-point increase in median Gensini score (95% confidence interval [CI], 9.33–15.07; P < 0.001). Comparable associations were observed for high-burden anatomy (odds ratio [OR] per doubling, 1.69; 95% CI, 1.44–1.97) and multivessel disease (OR per doubling, 1.58; 95% CI, 1.36–1.84). For the secondary high-Gensini-burden outcome, NT-proBNP/LVEF and NT-proBNP showed comparable discrimination (AUC, 0.783 vs. 0.780; ΔAUC = 0.004; P = 0.362). Adding LVEF to NT-proBNP did not show clear incremental improvement by NRI or IDI. Among the six multivariable models, SVM yielded the numerically highest AUC (0.763; 95% CI, 0.715–0.811), but no pairwise AUC difference remained significant after correction for multiple comparisons. XGBoost and CatBoost showed highly concordant SHAP importance rankings (Spearman ρ = 0.978), with NT-proBNP/LVEF ranked first in both models. Conclusions Higher NT-proBNP/LVEF was consistently associated with greater angiographic coronary disease burden across the continuous Gensini score and clinically anchored anatomical outcomes in patients with AMI. The ratio provides a parsimonious representation of ventricular stress and systolic function and showed stable predictive importance across the two boosting algorithms examined. Its discriminative performance was comparable, rather than superior, to that of NT-proBNP alone, supporting its potential role as a readily available adjunctive marker of coronary disease burden rather than a replacement for established clinical assessment.