Résumé : Empirical therapy with cefoperazone was compared with cefoperazone plus amikacin in granulocytopenic and nongranulocytopenic febrile patients. In nonneutropenic patients the overall response rate to cefoperazone was 88%; 10 of 12 gram-negative bacteremic patients were cured. Cefoperazone plus amikacin resulted in an 88% overall response rate and cured 14 of 15 patients with bacteremia. In neutropenic patients the overall response rate was 77% with cefoperazone alone and 73% with cefoperazone plus amikacin; the cure rates for gram-negative bacteremias were 8 of 11 and 6 of 12 patients, respectively. Our findings support the concept of single-drug empirical therapy with cefoperazone in febrile cancer patients, whether granulocytopenic or not, especially when gram-negative bacteremias are predominantly caused by Escherichia coli or Klebsiella species. The issue of Pseudomonas spp. and other more resistant pathogens needs further assessment with a larger number of patients.