A prospective, randomized study was performed to compare the efficacy and safety of mezlocillin plus sisomicin (Regimen A) and cefamandole plus tobramycin (Regimen B). Sixty-one episodes of documented infections were treated in 49 adult myelosuppressed cancer patients. About two-thirds of the patients treated with either regimen responded to this antibacterial therapy. The patients failing to respond to the initial therapy were treated with additional antibiotics, either with the empiric Regimen C (cefotaxime plus amikacin plus azlocillin) or with antibiotics selected on the basis of bacteriological results (Regimen D). With these schedules, the over-all response rate was increased to about 90%. The data suggest that febrile granulocytopenic cancer patients may profit from this empiric and sequential antibiotic treatment.