Summary. Non-cirrhotic patients having acute liver decompensation in flares of hepatitis B can recover spontaneously or die without liver transplantation. Criteria for identifying patients in need of liver transplantation are lacking. Fifty-one non-cirrhotic patients having acute liver decompensation in flares of hepatitis B were retrospectively reviewed. The patients were divided into three groups: group A patients (n = 18) recovered from acute liver decompensation spontaneously; group B patients (n = 22) died of acute liver failure; and group C patients (n = 11) had liver transplantation. Model of end-stage liver disease (MELD) scores were evaluated to identify the criteria for liver transplantation. The cut-off point of MELD scores for liver transplantation was evaluated by receiver operating characteristic (ROC) curve. Comparing group A and B patients, MELD score was an independent factor to predict prognosis. By analysing ROC curve, a MELD score > 30 was the most optimal cut-off point to indicate liver transplantation; however, the false positive rate was 11.1%. By weekly measurement of MELD scores, subsequent increase in MELD scores could help to avoid false positives. Moreover, a MELD score > 34 yielded 0% false positive rate and indicated the necessity of definite liver transplantation. For group C patients, ten of 11 patients were saved by liver transplantation. In conclusion, for the patients having acute liver decompensation in flares of hepatitis B, liver transplantation is definitely indicated by MELD scores > 34. Liver transplantation is also indicated if the MELD score increases in the subsequent 1–2 weeks. Liver transplantation has a good outcome if performed on time.