Background: Expansion of the donor pool by the use of grafts with extended donor criteria reduces waiting list mortality with an increased risk for graft and patient survival after liver transplantation. This study investigates the ability of the Donor-Risk-Index (DRI), the Extended-Criteria-Donor-Score (ECD-score) and the D-MELD-score to predict early outcome after liver transplantation.
Material/Methods: 291 consecutive adult liver transplants (01.01.2007–31.12.2010) were analysed in a single centre study with ongoing data collection. Primary study endpoints were 30-day mortality, 3-month mortality, 3-month patient and graft survival and the necessity of acute retransplantation within 30 days. For the primary study endpoints ROC-curve analysis was performed to calculate the sensitivity, specificity, and overall model correctness of the Donor-Risk-Index (DRI), Extended-Criteria-Donor-Score (ECD-score) and the D-MELD-Score as predictive models. Cut-off values were selected with the best Youden index.
Results: ROC-curve analysis showed areas under the curve (AUROCs) <0.7 for the DRI, the ECD-Score and the D-MELD-Score as models for the prediction of 30-day mortality, 3-month mortality, 3-month patient survival, 3-month graft survival as well as the necessity of acute retransplantation within 30 days after transplantation with unacceptable low levels of overall model correctness (<62%) and specificity (<56%).
Conclusions: The DRI, the ECD-Score and the D-MELD-Score all fail to predict short-term outcome after liver transplantation with acceptable overall model correctness in a current European transplant setting.

Keywords: Liver Transplantation, Prognostic models, receiver operating characteristics