Fifty febrile, granulocytopenic allogeneic bone marrow transplant patients receiving prophylactic trimethoprim-sulfamethoxazole were randomized to one of two empirical antibiotic regimens to determine whether a shortened course of empirical therapy was beneficial. Of the 50 patients, 25 received empirical tobramycin and ticarcillin for only 3 days, and 25 were maintained on empirical tobramycin and ticarcillin until they were afebrile and no longer granulocytopenic. Although the incidence of bacterial infections in the two groups was not statistically significantly different, almost twice as many bacterial infections were observed in the group that received the short course of empirical therapy. Furthermore, because of the high incidence of bacterial infection and clinical concerns about occult bacterial sepsis, within 2 weeks of the randomization the overall use of parenteral antibacterial agents was similar in both groups. The incidence of invasive fungal disease and the use of amphotericin B therapy were similar in both groups. The results of this study suggests that little clinical benefit is likely to be seen in bone marrow transplant patients treated with short-course empirical tobramycin and ticarcillin, despite the administration of prophylactic trimethoprim-sulfamethoxazole, and emphasize the need for new strategies to prevent infections with gram-positive and trimethoprim-sulfamethoxazole-resistant gram-negative bacteria in these patients.