Gram-positive bacteria have become the predominant infecting organisms in granulocytopenic cancer patients. Empiric antibiotic regimens used in febrile neutropenic patients often include an extended-spectrum cephalosporin, but the response to therapy in gram-positive coccal bacteremia has been unsatisfactory. Thus, new antibiotics with better activity against gram-positive bacteria should be tested. The objective of this prospective randomized controlled study was to evaluate and compare the efficacy and tolerance of piperacillintazobactam plus amikacin with that of ceftazidime plus amikacin, the standard regimen of the International Antimicrobial Therapy Cooperative Group of the European Organization for Research and Treatment of Cancer, in the empiric treatment of febrile granulocytopenic cancer patients. A total of 858 episodes were eligible for this study, and 706 episodes were assessable for efficacy. The antibiotic treatment was successful in 210 (61%) of 342 episodes in the piperacillin-tazobactam-amikacin group compared with 196 (54%) of 364 episodes treated with ceftazidime plus amikacin (P = 0.05). The time to defervescence was significantly shorter (P = 0.01) and the time to failure was significantly longer (P = 0.02) in the piperacillin-tazobactam-amikacin group. A significant difference in response to bacteremic infections between the two patient groups was found: piperacillin-tazobactam plus amikacin was successful in 40 of 80 episodes (50%), and ceftazidime plus amikacin was successful in 35 of 101 episodes (35%) (P = 0.05). A multivariate analysis showed that the probability of failure was significantly greater with ceftazidime plus amikacin than with piperacillin-tazobactam plus amikacin (P = 0.02). This trial suggests that piperacillin-tazobactam plus amikacin is more effective than ceftazidime plus amikacin for the empiric treatment of fever and bacteremia in granulocytopenic cancer patients. Although cutaneous reaction was more frequently associated with piperacillin-tazobactam plus amikacin than with ceftazidime-amikacin, this unwanted effect was relatively mild and its incidence was comparable to that of other penicillin compounds.