Construction and evaluation of risk prediction model for major adverse cardiovascular events in peritoneal dialysis patients

Abstract

Background and objectives This study developed and validated a robust, clinically applicable risk prediction model for cardiovascular mortality in high-risk peritoneal dialysis patients. The model enables precise individualized risk stratification, supports evidence-based clinical decisions, and facilitates targeted preventive interventions to improve long-term outcomes. Methods A retrospective analysis was conducted on peritoneal dialysis patients at the Department of Nephrology, First Affiliated Hospital of Xi’an Jiaotong University, from June 2018 to May 2025. Cardiovascular disease served as the primary endpoint. Baseline characteristics and laboratory parameters were compared between groups. Univariate and multivariate regression analyses identified predictors of major adverse cardiovascular events. The predictive model was developed and validated using a 7:3 split after bootstrap resampling. Results Multivariate analysis identified age, fasting blood glucose, absolute handgrip strength, and left ventricular ejection fraction as significant predictors. ROC curves (AUC > 0.6) and Kaplan–Meier curves (P < 0.05) confirmed their diagnostic and discriminatory value for major adverse cardiovascular events. A nomogram incorporating these four variables was developed. Its AUCs for predicting 3-, 5-, and 8-year cardiovascular event incidence were 0.828 (95% CI: 0.744–0.912), 0.785 (95% CI: 0.702–0.868), and 0.836 (95% CI: 0.744–0.929), respectively—indicating strong predictive discrimination. Using the optimal cut-off, patients were stratified into low- and high-risk groups. Kaplan–Meier analysis showed significant differences in cumulative event incidence at all three time points (P < 0.001), confirming good discrimination. Calibration demonstrated close agreement between predicted and observed outcomes; decision curve analysis supported robust clinical utility. Conclusion Age, fasting blood glucose, absolute handgrip strength, and left ventricular ejection fraction are independent predictors of MACE in patients receiving PD. We have developed and internally validated a novel, interpretable nomogram-based risk prediction model that demonstrates high discriminative accuracy, calibration fidelity, and clinical utility. This tool holds substantial promise for early identification of high-risk individuals at PD initiation, thereby facilitating timely, personalized cardiovascular risk mitigation strategies.