A question of clinical reliability: observed versus EuroSCORE-predicted mortality after aortic valve replacement

The study aim was to determine the clinical reliability of the EuroSCORE as a predictor of operative risk in aortic valve replacement (AVR).

Between 2000 and 2007, a total of 1497 patients underwent isolated elective AVR (no endocarditis, aortic procedure or re-do) at the authors' institution. A fitting of the deviation of expected mortality (EM) from observed mortality (OM) was performed and studied. To identify the cause of deviation of EM, a multivariate analysis of the EuroSCORE variables (using SAS JMP software) was conducted on the available data, and the results were re-evaluated.

An overestimation of EM was observed, and this was found to increase systematically with the rise in expected risk (0.3 +/- 1.0% at 5% OM versus 23.8 +/- 1.9% at 35% OM; p < 0.0001). A multivariate analysis of the EuroSCORE variables showed only age and preoperative neurological dysfunction as significant risk factors (p < 0.003 and < 0.04, respectively). All other EuroSCORE variables were statistically insignificant.

The EuroSCORE is a solid and practical concept, but is clinically unreliable as a predictor of operative risk for elective AVR; hence, it should no longer be used for this purpose in its present form. It is recommended that a statistical correction of the EuroSCORE deviation be used, and that an updated EuroSCORE or a new risk stratification tool be developed to predict operative risk for patients undergoing heart valve surgery.