In a prospective randomized trial, 318 children aged between 3 mo and 11 yr who were negative for all hepatitis B markers were randomized to receive two 5-μg doses of hepatitis B recombinant DNA yeast vaccine at 0 and 1 mo (group 1), three 5-μg doses of hepatitis B recombinant DNA yeast vaccine at 0, 1 and 6 mo (group 2) or three 10-μg doses of plasma-derived hepatitis B vaccine (group 3). The HBs antibody response rate at 8 mo was between 93% and 99%; it was still 75% to 87% at 5 yr in all three groups. Geometric mean titers at 1 yr were 83, 1,085 and 858 mIU/ml in groups 1, 2 and 3, respectively. These values had decreased after 5 yr to 47, 131 and 250 mIU/ml. Subjects in group 1 showed a significantly less proportional drop in geometric mean titer at the fifth year than did subjects in group 2 (p = 0.05) or group 3 (p = 0.015). None of the children developed HBc antibody, even after 5 yr of follow-up. We noted 42 episodes of significantly increased HBs antibody titers, probably due to anamnestic response, even when the titers had dropped to low levels. The mean age at which anamnestic response occurred was 8.7 yr. We conclude that (a) the recombinant vaccine and plasma-derived vaccine are comparable in safety and immunogenicity; (b) two doses of vaccine was as effective in protecting hepatitis B infection as three doses, despite lower HBs antibody titers; (c) anamnestic responses occurred most frequently around 4 yr after a child began attending school; and (d) a booster dose was not necessary at 5 yr, probably because of ability of healthy child vaccinees to mount good anamnestic responses. (HEPATOLOGY 1993;18:763-767).