Severe acute hepatitis B is a rapid deterioration of liver function, which carries a high mortality rate. The aim of this study is to evaluate the efficacy of lamivudine in patients with severe acute hepatitis B. In this study, 80 patients with severe acute hepatitis B were randomly divided into lamivudine and the control group. For the two groups, we compared HBsAg, HBeAg seroconversion rates, serum HBV DNA-negative rate, biochemical indicators, the incidence of liver failure, and mortality. The influential factors on the mortality were studied by Cox proportional hazards model. The improvement in serum TBiL, INR, and HBV DNA levels of the lamivudine group was significantly greater than that of the control group. The mortality of lamivudine group (7.5%, 3/40) was significantly lower than that of the control group (25.0%, 10/40) (p = 0.034). The incidence of liver failure (8.7%, 2/23) of patients receiving lamivudine within a week was significantly lower than that (35.3%, 6/17) of those who received it after a week (p = 0.038). In multivariate Cox proportional hazards analyses, age (p = 0.043), ratio of total to direct bilirubin (p = 0.009), treatment method (p = 0.006), and the decline of HBV DNA load during therapy (p = 0.017) were independent predictors of mortality. The HBsAg seroconversion rates (62.5%, 25/40) and HBeAg seroconversion rates (63.6%, 21/33) of the lamivudine group were significantly lower than those (85.0%, 34/40), (87.5%, 28/32) of the control group (p = 0.022, 0.026). Early treatment with lamivudine leads to a greater decrease in HBV DNA level, better clinical improvement and mortality improvement in patients with severe acute hepatitis B, but with a lower seroconversion rate. A rapid decline of HBV DNA load is a good predictor for the treatment outcome.