Model of End-Stage Liver Disease Score and Derived Variants Lack Prognostic Ability after Liver Transplantation.

BACKGROUND The model of end-stage liver disease (MELD) score is currently used for donor liver allocation in many regions. The objective of this retrospective study was to assess the MELD score and its diverse variants as prognostic models for mortality after liver transplantation. MATERIAL AND METHODS An analysis of 454 consecutive adult liver transplants since the introduction of MELD-based donor liver allocation was conducted. Eight different MELD score variants were investigated. Receiver operating characteristic (ROC) curve analysis was performed to calculate the sensitivity, specificity, and overall model correctness of the investigated scores as a predictive model. The Brier score was used for the prediction of model accuracy and calculated as described before. Study endpoints were 90-day mortality and long-term patient mortality. RESULTS A 90-day mortality of 15.4% (n=69) and long-term mortality of 25% (n=112) were observed. All investigated models fail to reach relevant areas under the ROC curve greater than 0.700 for the prediction of mortality after liver transplantation. All calculated Brier scores were greater than 0.25, indicating a significant lack of model discrimination and calibration of the investigated scores. CONCLUSIONS A prognostic model for the prediction of outcome after transplantation still needs to be identified and should allow weighing urgency against utility in liver transplantation.