A total of 318 children were prospectively randomized in group 1 with two 5-μg doses of recombinant vaccine given at 0 and 1 month; in group 2 with three 5-μg doses of recombinant vaccine given at 0, 1, and 6 months; or in group 3 with three doses of plasma-derived vaccine given at 0, 1, and 6 months. Eleven subjects with a hepatitis B surface antigen antibody (anti-HBs) titer of less than 10 mIU/mL at 12 months were given an extra dose of vaccine and were excluded from analysis. No booster doses were given to any other subjects. All children were followed up yearly for the level of anti-HBs titers and for the detection of hepatitis B infection. At the 12th year of follow-up, there were significantly fewer subjects with anti-HBs of 10 mIU/mL or above in group 1 (60.4%) when compared with group 2 (81.4%; P = .0287) and group 3 (79.0%; P = .0381). The geometric mean titers (GMTs) of subjects of group 1 were significantly lower than those of group 2 and group 3 throughout the 12 years of follow-up. A total of 65 subjects had one or more episodes of anamnestic response. No subject became positive for hepatitis B surface antigen (HBsAg); 2 became positive for hepatitis B core antigen antibody (anti-HBc). In conclusion, the long-term protective immunity was better with three doses of hepatitis B vaccine (either the recombinant or plasma-derived) than with two doses. However, protection from hepatitis B infection could be equally achieved by either two doses or three doses of the vaccine. Booster doses were not necessary, probably because of effective anamnestic response.