We conducted a prospective, randomized trial in 132 patients undergoing bone marrow transplantation comparing cefoperazone in combination with sulbactam (S), N = 66, vs. cefoperazone plus mezlocillin (CM), N = 66, as empiric antibiotic therapy for fever and neutropenia. Overall duration of neutropenia was 3-55 (median, 13) days. Forty-one patients had positive initial cultures (S = 22 and CM = 19). Twelve of these 41 patients responded to initial study antibacterial agent treatment (S = 6 and CM = 6). Twenty-nine of 41 patients were withdrawn from study because of clinical deterioration, continued fever, or persistently positive cultures (S = 16 and CM = 13). Of the 90 patients who had culture-negative fever (S = 44 and CM = 46), 44 subjects responded with or without the addition of amphotericin B (S = 21 and CM = 23). Thirty-seven of 90 patients were withdrawn from study due to continued fever or clinical deterioration (S = 17 and CM = 20). Nine patients were withdrawn as a result of rash or diarrhea (S = 6 and CM = 3). We conclude that in patients undergoing bone marrow transplantation, there was no difference in efficacy between cefoperazone/sulbactam and the combination of cefoperazone plus mezlocillin in the empiric treatment of the febrile neutropenic patient. Since the majority of initial infections were due to gram positive bacteria, consideration should be given to broadening initial empiric antibacterial agent therapy with drugs that possess potent activity against these organisms.