Objective: This study aims to investigate their correlation and predictive utility for in-stent restenosis (ISR) in patients with acute coronary syndrome (ACS) following percutaneous coronary intervention (PCI).

Methods: We collected medical records of 668 patients who underwent PCI treatment from January 2022 to December 2022. Based on follow-up results (ISR defined as luminal narrowing ≥ 50% on angiography), all participants were divided into ISR and non-ISR groups. The XGBoost machine learning (ML) model was employed to identify the optimal predictive variables from a set of 31 variables. Discriminatory ability was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC), while calibration and performance of the prediction models were assessed using the Hosmer-Lemeshow (HL) test and calibration plots. Clinical utility of each model was evaluated using decision curve analysis (DCA).

Results: In the XGBoost importance ranking of predictive factors, LMR and RC ranked first and fourth, respectively. The AUC of the entire XGBoost ML model was 0.8098, whereas the model using traditional stepwise backward regression, comprising five predictive factors, had an AUC of 0.706. The XGBoost model showed superior predictive performance with a higher AUC, indicating better discrimination and predictive accuracy for ISR compared to traditional methods.

Conclusion: LMR and RC are identified as cost-effective and reliable biomarkers for predicting ISR risk in ACS patients following drug-eluting stent (DES) implantation. LMR and RC represent cost-effective and reliable biomarkers for predicting ISR risk in ACS patients following drug-eluting stent implantation. Enhances the accuracy and clinical utility of ISR prediction models, offering clinicians a robust tool for risk stratification and personalized patient management.