Do GRACE (Global Registry of Acute Coronary events) risk scores still maintain their performance for predicting mortality in the era of contemporary management of acute coronary syndromes?

Background Although the GRACE risk scores (RS) are the preferred scoring system for risk stratification in acute coronary syndromes (ACS), little is known whether these RS still maintain their performance in the current era. We aimed to investigate this issue in a contemporary population with ACS. Methods The study population composed of patients enrolled in the MASCARA national registry. The GRACE RS were calculated for each patient. Discrimination and calibration were evaluated with the C statistic and the Hosmer-Lemeshow test, in the whole population and according to the type of ACS, risk strata, and whether the patient had a history of diabetes and/or chronic renal failure. We determined if left ventricular ejection fraction (LVEF) provides incremental prognostic information above that established by the RS and whether percutaneous coronary intervention (PCI) during admission affects the performance of the score for predicting 6-month mortality. Results The 5,985 patients constituted the validation cohort for the in-hospital mortality RS and 5,635 the validation cohort for the 6-month mortality RS. Overall, both GRACE RS demonstrated excellent discrimination (C > 0.80) and calibration (all P values in Hosmer-Lemeshow > .1). Although similar results were seen in all subgroups, the 6-month mortality RS performed significantly less well in patients undergoing PCI compared to those patients who did not (C = 0.73 vs 0.76, P < .004). Adding LVEF to the RS did not convey significant prognostic information. Conclusions The GRACE RS for predicting in-hospital and 6-month mortality still maintain their excellent performance in a contemporary cohort of patients with ACS. Further studies are needed to investigate the performance of the 6-month mortality GRACE score in patients undergoing in-hospital PCI. Left ventricular ejection fraction did not convey significant information over that provided by the RS. (Am Heart J 2010;160:826-834.e3.)