In a prospective randomized study, 120 febrile, granulocytopenic patients received as initial therapy ceftazidime with or without teicoplanin. At the onset of fever, patients had no obvious infectious focus. For 103 assessable episodes, initial bacteremias were detected in 18 of 51 patients (35%) given ceftazidime and 20 of 52 patients (38%) given the combination; 13 and 17 bacteremias caused by gram-positive bacteria occurred in these groups, respectively. There was no difference in terms of the final response (25 of 51 patients [49%] treated with ceftazidime alone versus 33 of 52 patients [63%] given the combination), and the morbidity was comparable for both treatment groups. The duration of fever and of total antibiotic therapy were similar in both groups. Initial therapy was modified in 26 patients (51%) treated with ceftazidime, with 20 surviving the infection, and in 19 patients (37%) treated with the combination, with 15 surviving. Persistent fever was the main reason for changing treatment, and no patient died of a gram-positive infection. Subsequent infective events occurred in 16 patients (31%) given ceftazidime and in 25 patients (48%) given the combination. Lung infiltrates developed in 12 and 13 patients, respectively, but more new infections occurred in the combination group. Allergic skin reactions were also more frequent in this group. Thus, while teicoplanin provides simple, reliable, and safe treatment of patients with presumed gram-positive infections, it is not useful when given empirically to this patient population, and treatment may result in more infective complications and adverse events.