Abstract
Background: Simple surgical and clinical risk scores are useful in mortality prediction. Aims: The study's aim was to validate three scores in real-world registry of percutaneous coronary intervention (PCI) for the left main coronary artery (LMCA). Methods: All data were obtained from the BIA-LM Registry. Discrimination and calibration of EuroSCORE II, ACEF, CHA(2)DS(2)-VASc, and CHA(2)DS(2)-VA were assessed with receiver operating characteristic (ROC) curves analysis and Hosmer-Lemeshow (HL) test. Results: The final cohort included 851 patients, median age was 71, and 156 patients had history of previous coronary artery bypass grafting (CABG). Median EuroSCORE II, ACEF, CHA(2)DS(2)-VASc, and CHA(2)DS(2)-VA were 3.1% (IQR 5.4%), 1.56 (IQR 0.9), 4 (IQR 2), and 4 (IQR 2), respectively. In the short- (30 days) and long-term (mean 4.1 years), there were 27 and 318 deaths. In short-term, EuroSCORE II showed the best discrimination in the overall population and subgroup with unprotected LMCA [area under the curve (AUC) 0.804, 95% CI 0.717-0.890 and AUC 0.826, 95% CI 0.737-0.913, respectively, p < 0.001 for comparisons with other models), with the best cut-off value at 7.1%. In long-term observation, EuroSCORE II and ACEF showed good predictive value (overall population: AUC 0.716, 95% CI 0.680-0.750 and AUC 0.725, 95% CI 0.690-760, respectively). In short- and long-term observation, EuroSCORE II and ACEF showed poor calibration (HL test p < 0.05) as compared to CHA(2)DS(2)-VASc (HL test p = 0.40 and 0.18). Conclusions: EuroSCORE II showed good mortality prediction in short-term observation; however, its predicted risk should be interpreted with caution due to poor calibration. ACEF and EuroSCORE II may be useful in long-term mortality prediction.