Because studies of the treatment of children with occult bacteremia have yielded conflicting results, we compared ceftriaxone with amoxicillin for therapy. Inclusion criteria were age 3 to 36 months, temperature > or = 39 degrees C, an acute febrile illness with no focal findings or with otitis media (6/10 centers), and culture of blood. Subjects were randomly assigned to receive either ceftriaxone, 50 mg/kg intramuscularly, or amoxicillin, 20 mg/kg/dose orally for six doses. Of 6733 patients enrolled, 195 had bacteremia and 192 were evaluable: 164 Streptococcus pneumoniae, 9 Haemophilus influenzae type b, 7 Salmonella, 2 Neisseria meningitidis, and 10 other. After treatment, three patients receiving amoxicillin had the same organism isolated from their blood (two H. influenzae type b, one Salmonella) and two from the spinal fluid (two H. influenzae type b), compared with none given ceftriaxone. Probable or definite infections occurred in three children treated with ceftriaxone and six given amoxicillin (adjusted odds ratio 0.43, 95% confidence interval 0.08 to 1.82, p = 0.31). The five children with definite bacterial infections (three meningitis, one pneumonia, one sepsis) received amoxicillin (adjusted odds ratio 0.00, 95% confidence interval 0.00 to 0.52, p = 0.02). Fever persisted less often with ceftriaxone (adjusted odds ratio 0.52, 95% confidence interval 0.28 to 0.94, p = 0.04). Although the difference in total infections was not significant, ceftriaxone eradicated bacteremia, prevented significantly more definite focal bacterial complications, and was associated with less persistent fever.