Evaluation of model for end-stage liver disease for prediction of mortality in decompensated chronic hepatitis B

We aimed to study the predictive ability of model for end-stage liver disease (MELD) for short-term mortality in chronic hepatitis B.All patients admitted from 1996 to 2003 because of chronic hepatitis B and its related complications were identified by electronic search of the hospital database. MELD and Child-Turcotte-Pugh (CTP) scores on initial admissions were calculated. Cox proportional hazard model was used to determine the factors associated with mortality. The area under receiver operator characteristics curve (AUC) was used to determine the predictive abilities of the two models for 3-month and 1-yr mortalities.A total of 2,073 patients was admitted because of liver-related problems and 506 patients had chronic hepatitis B-related complications. Two hundred fifty-six (51%) patients died and 16 (3%) patients underwent liver transplantation. In multivariate analysis, MELD and CTP scores were independent predictors of 3-month and 1-yr mortality. Other independent predictors of mortality included older age, hepatocellular carcinoma (HCC), lamivudine treatment, and lower serum sodium. At both 3 months and 1 yr, the AUC of the MELD score (0.65 and 0.63, respectively) was significantly lower than that of the CTP score (0.75 and 0.77, respectively) (p0.0001). The differences remained significant when only liver cirrhosis patients without HCC at presentation were analyzed, but the AUC of the two scores became comparable when patients on lamivudine were excluded.The MELD score is a valid prognostic model in decompensated chronic hepatitis B. Lamivudine treatment may affect the performance of MELD score. Other variables including those in CTP score may improve its predictive ability.