Revision of MELD to Include Serum Albumin Improves Prediction of Mortality on the Liver Transplant Waiting List

Background: Allocation of donor livers for transplantation in most regions is based on the Model for End-Stage Liver Disease (MELD) or MELD-sodium (MELDNa). Our objective was to assess revisions to MELD and MELDNa that include serum albumin for predicting waiting list mortality. Methods: Adults registered for liver transplantation in the United States (2002-2007) were identified from the United Network for Organ Sharing (UNOS) database. Cox regression was used to determine the association between serum albumin and 3-month mortality, and to derive revised MELD and MELDNa scores incorporating albumin ('MELD-albumin' and '5-variable MELD [5vMELD]'). Results: Among 40,393 patients, 9% died and 24% underwent transplantation within 3 months of listing. For serum albumin concentrations between 1.0 and 4.0 g/dL, a linear, inverse relationship was observed between albumin and 3-month mortality (adjusted hazard ratio per 1 g/dL reduction in albumin: 1.44; 95% CI 1.35-1.54). The c-statistics for 3-month mortality of MELD-albumin and MELD were 0.913 and 0.896, respectively (P < 0.001); 5vMELD was superior to MELDNa (c-statistics 0.922 vs. 0.912, P < 0.001). The potential benefit of 5vMELD was greatest in patients with low MELD (<15). Among low MELD patients who died, 27% would have gained >= 10 points with 5vMELD over MELD versus only 4-7% among low MELD survivors and high MELD (>= 15) candidates (P < 0.0005). Conclusion: Modification of MELD and MELDNa to include serum albumin is associated with improved prediction of waiting list mortality. If validated and shown to be associated with reduced mortality, adoption of 5vMELD as the basis for liver allograft allocation may improve outcomes on the liver transplant waiting list.