Feed

Background: Coronary in-stent restenosis (ISR) remains a major limitation of percutaneous coronary intervention (PCI) with drug-eluting stents (DES), adversely affecting long-term outcomes. Most available prediction models rely on invasive procedural variables and have been developed predominantly in high-income populations, limiting their generalizability. This study aimed to identify independent predictors of coronary ISR and to develop and internally validate a clinically applicable risk stratification model based on routinely available clinical, laboratory, and procedural factors in a cohort of patients from Kazakhstan. Methods: In this retrospective case-control study, 910 patients with coronary artery disease (CAD) who underwent follow-up coronary angiography after PCI between January 2018 and July 2025 were included. The study comprised 455 patients with angiographically confirmed coronary in-stent restenosis and 455 patients without restenosis selected using a consecutive sampling approach. Clinical characteristics, laboratory parameters, echocardiographic findings, and angiographic data were analyzed. Independent predictors were identified using multivariable binary logistic regression. Model discrimination was assessed using receiver operating characteristic (ROC) curve analysis, and internal validation was performed using bootstrap resampling. Results: The mean age was 62.9 ± 8.9 years, and 75.2% of patients were male. Restenosis was independently associated with prior myocardial infarction (MI) (OR 2.20; 95% CI 1.65-2.80), type 2 diabetes mellitus (T2DM) (OR 2.60; 95% CI 1.93-3.47), and smoking (OR 1.40; 95% CI 1.01-1.89). Patients with restenosis demonstrated a less favorable inflammatory and metabolic profile, including higher NLR, MHR, atherogenic index, and TyG index (all p < 0.05). LVEF was significantly lower, while multivessel disease and the number of implanted stents was higher (p < 0.001). A risk stratification model incorporating T2DM, the number of implanted stents, MPV, neutrophil count, HDL-C, LVEF demonstrated good discrimination (AUC 0.828) and 74.4% accuracy. Conclusions: The proposed model demonstrated good discrimination and satisfactory internal validity with limited optimism after internal bootstrap validation. It may serve as a useful tool for patient risk stratification after PCI. External validation in independent cohorts is required before widespread clinical implementation.