In a prospective randomized study, 100 episodes of fever (greater than 38 degrees C) and granulocytopenia (less than 1000/microliters) in cancer patients were empirically treated with ceftazidime (2 g every 8 h) plus teicoplanin (400 mg every 8 h on day 1; 400 mg every day thereafter) or ceftazidime (2 g every 8 h) plus amikacin (500 mg every 8 h). Bacteraemia, clinically documented infection and possible infection were documented in seven, 11 and 19 patients treated with ceftazidime plus teicoplanin and in 11, four and 17 patients treated with ceftazidime plus amikacin. Overall, the response rate was similar in the two groups of patients as was the need for treatment modifications and the rate of death. For documented Gram-positive bacteraemia, the response rate was 2/5 patients treated with ceftazidime plus teicoplanin and 2/7 with ceftazidime plus amikacin; for documented Gram-negative bacteraemia, the response rate was 1/2 and 3/4 patients respectively. No breakthrough bacteraemia was observed. Tolerance was excellent, although renal toxicity (elevation of serum creatinine) was observed in three patients treated with ceftazidime plus teicoplanin and in none allocated to ceftazidime plus amikacin.