Soon after the plasma-derived hepatitis B vaccine became available in the US, the Centers for Disease Control and the manufacturer received over 100 reports of vaccinated groups with unexpectedly low levels of vaccine-induced antibody. To confirm previous retrospective surveys relating these failures to buttock injection and to evaluate the effect of other host factors on vaccine-induced antibody responses, we conducted a clinical trial in healthy health-care workers. Participants were randomly assigned to one of three treatment groups: 1-Ar, 1-inch needle injection in the arm; 1-Bu, 1-inch needle injection in the buttock; 2-Bu, 2-inch needle injection in the buttock. All participants were administered vaccine according to the standard vaccine dosage schedule of 20 micrograms at 0, 1 and 6 months. Antibody response rates (antibody to hepatitis B surface antigen greater than or equal to 10 sample ratio units by radioimmunoassay) and geometric mean titres of antibody two months after the third vaccine dose were 93% and 1454 mIU ml-1 for group 1-Ar, 72% and 85 mIU ml-1 for group 1-Bu, and 83% and 387 mIU ml-1 for group 2-Bu. Seroconversion rates and titres of antibody in the three groups were significantly different from each other statistically. Increasing age, increasing total skinfold thickness and cigarette smoking were independently associated with lower antibody responses in persons receiving buttock injections but not in persons receiving arm injections.(ABSTRACT TRUNCATED AT 250 WORDS)