BACKGROUND AND AIM OF THE STUDY The aim was to evaluate the early and long-term results of redo-aortic valve replacement (AVR) for dysfunctioning bioprostheses, according to the type of bioprosthesis failure. The performance of the EuroSCORE in predicting operative mortality was addressed. METHODS A retrospective analysis was performed of 164 patients who had undergone redo-AVR at the authors' institution. All data were collected prospectively. The type of bioprosthesis failure was categorized as either structural valve dysfunction (SVD), non-structural vale dysfunction (NSVD), or infective endocarditis (IE). The logistic EuroSCORE I was obtained, and a formal analysis of its predictive performance was conducted. RESULTS The mean logistic EuroSCORE was 26.3 +/- 20.6%, while the observed operative mortality was 10.6%. The EuroSCORE overestimated mortality in all subgroups of indication; operative mortality was higher among the NSVD (14.7%) and IE (13.9%) groups, and lower among SVD patients (8.5%). The performance of the EuroSCORE in predicting mortality was better in SVD patients, and worse in NSVD patients (areas under the ROC curve of 0.857 and 0.751, respectively). Only patient-related factors independently predicted the operative mortality (logistic regression). The SVD patients displayed a significantly better long-term survival compared to NSVD and IE patients. Patient-related factors were seen to be the determinants of long-term survival. CONCLUSION Redo-AVR can be performed with good early and long-term results, despite a high predicted risk. The logistic EuroSCORE I globally overestimates the risk, and its performance varies considerably according to the indication for redo-AVR. Therefore, it should not be used to determine the application of valve-in-valve procedures. The type of indication for redo-AVR (SVD, NSVD, IE) defines three different profiles of patients having distinctive early and long-term behaviors.