Performance of currently available risk models in a cohort of mechanically supported high-risk percutaneous coronary intervention--From the PROTECT II randomized trial

Procedural risk scores facilitate clinical decision making by using individual patient characteristics to estimate the risk of adverse events. The performance of PCI-based risk scores is not well-described among patients undergoing hemodynamically supported high risk PCI. A total of 427 patients with unprotected left main disease, last remaining vessel or three-vessel disease with severely reduced left ventricular function underwent supported high-risk PCI with an intra-aortic balloon pump (IABP, N = 211) or a left ventricular assist device (Impella 2.5, N = 216) as part of the PROTECT II trial. We examined the performance of the additive Euroscore, logistic Euroscore, STS mortality score, STS morbidity and mortality score, Mayo Clinic risk score and New York state PCI risk score on the endpoint of 90-day mortality in this unique high-risk population. Mean age was 67.2 ± 10.9 years; 65.8% of patients were in NYHA class III/IV, and mean LVEF was 24%. All-cause 90-day mortality was 10.4%. The scores were generally correlated (p < 0.0001 for all comparisons), with R(2) values ranging from 0.28 (STS morbidity/mortality and Mayo Clinic) to 0.68 (logistic Euroscore and STS mortality). However, receiver-operator curves for 90-day all-cause mortality for all risk scores demonstrated poor discriminatory performance with c-statistics of 0.542-0.616. Calibration of the risk scores was not poor, but varied according to the specific score examined. The discriminatory capacity of currently available risk models is suboptimal when applied to a cohort of mechanically supported complex high-risk PCI. A risk score designed specifically for this population could help to further refine risk assessment