A prospective, randomized trial was initiated to evaluate the efficacy of two antibiotic regimens, differing in the agent included with activity against gram-positive bacteria, for the empirical treatment of febrile episodes in neutropenic patients with hematologic malignancies (group 1, piperacillin plus amikacin; group 2, piperacillin plus amikacin plus teicoplanin). After 72 hours of therapy, patients in group 1 who were still febrile were administered teicoplanin and those in group 2 were administered amphotericin B. A total of 158 evaluable episodes were observed within 8 months. The success rate was 50.6% in group 1 and 60% in group 2. The response rate among patients who did not respond to the original regimen increased to 86.7% with the addition of teicoplanin (group 1) and to 90% with the addition of amphotericin B (group 2). There were 86 unexplained febrile episodes and 56 documented episodes of bacteremia (34 caused by gram-positive organisms). Our results indicate that teicoplanin is safe, well tolerated, and effective for the treatment of documented episodes of gram-positive bacteremia and as an empirical agent. The inclusion of teicoplanin in the initial empirical regimen appears unnecessary if a combination of antibiotics active against gram-positive organisms is used, unless infections are due to oxacillin-resistant staphylococci.