Risk stratification algorithm for clinical outcomes in anemic patients undergoing percutaneous coronary intervention

To explore the crosstalk between baseline or visit hemoglobin and major adverse cardiovascular and cerebral events (MACCE) in percutaneous coronary intervention (PCI) patients and to construct risk stratification models to predict MACCE amongst these patients.

We conducted a retrospective cohort in patients undergoing PCI procedures at Beijing Friendship Hospital between January 2013 and December 2020. Multivariate Cox proportional hazards models were employed for data analyses. The composite MACCE was the primary endpoint and we used machine learning algorithms to evaluate risk factors associated with MACCE. Model performance was measured using Brier scores and receiver-operating characteristic curves. The association between risk factors and MACCE probability was examined using partial dependency plots.

8,298 PCI-treated patients were enrolled in the study. 1,919 of these patients had anemia. During a four-year median follow-up period, 1,636 patients (19.71%) had MACCE. The visit hemoglobin and hemoglobin change was associated with higher risk of MACCE respectively (visit hemoglobin: hazard ratio [HR]: 0.98; 95% confidence interval [CI]: 0.98-0.99; p < 0.001; hemoglobin change: HR: 0.99; 95%CI: 0.98-0.99; p < 0.001). Gradient Boosting (GB) was the BPM, with a mean C-statistic value of 0.78 (95% CI: 0.76-0.80) for predicting MACCE (Brier score: 0.26). The best indicator for MACCE was a low estimated glomerular filtration rate [eGFR] (71 mL/min/1.73m²) at admission, followed by a high serum HbA1c (6.6%) level. A simple risk tree successfully classified patients (17-40.5%) with increased risks of MACCE. The high- vs. low-risk HR for MACCE was 2.04 (95% CI: 1.48-2.82).

Visit hemoglobin and long-term hemoglobin changes were more predictive of MACCE risk than baseline hemoglobin levels. Our findings indicated that increasing hemoglobin levels might improve the long-term prognosis of anemia patients. We established a new risk stratification model for MACCE, which may more efficiently prioritize targeted screening for at-risk anemic patients undergoing PCI.