A revised model for end-stage liver disease optimizes prediction of mortality among patients awaiting liver transplantation.

Abstract

The Model for End Stage Liver Disease (MELD) was originally developed based on data from patients who underwent the transjugular intrahepatic portosystemic shunt procedure. An updated MELD based on data from patients awaiting liver transplantation should improve mortality prediction and allocation efficiency. Wait-list data from adult primary liver transplantation candidates from the Organ Procurement and Transplantation Network were divided into a model derivation set (2005-2006; n=14,214) and validation set (2007-2008; n=13,945). Cox regression analysis was used to derive and validate an optimized model that updated coefficients and upper and lower bounds for MELD components and included serum levels of sodium. Main outcomes measure was ability to predict 90-day mortality of patients on the liver transplantation wait list. Optimized MELD score updated coefficients and implemented new upper and lower bounds for creatinine (0.8 and 3.0 mg/dL, respectively) and international normalized ratio (1 and 3, respectively). Serum sodium concentrations significantly predicted mortality, even after adjusting for the updated MELD model. The final model, based on updated fit of the 4 variables (ie, bilirubin, creatinine, international normalized ratio, and sodium) had a modest yet statistically significant gain in discrimination (concordance: 0.878 vs 0.865; P<.01) in the validation dataset. Utilization of the new score could affect up to 12% of patients (based on changed score for 459 of 3981 transplants in the validation set). Modification of MELD score to update coefficients, change upper and lower bounds, and incorporate serum sodium levels improved wait-list mortality prediction and should increase efficiency of allocation of donated livers.