In a multicenter, randomized clinical trial, the efficacy of ciprofloxacin plus azlocillin was compared with that of a standard regimen of ceftazidime plus amikacin for the initial empiric treatment of fever in neutropenic cancer patients. In addition, the efficacy of early conversion from intravenous therapy to orally administered ciprofloxacin was compared with that of continued ceftazidime plus amikacin. Seventy-one oncology patients with 79 episodes of fever and neutropenia were randomly assigned to receive initial empiric antibiotic therapy with either intravenously administered ciprofloxacin and azlocillin followed by orally administered ciprofloxacin (regimen 1, 25 episodes); ceftazidime and amikacin (regimen 2, 30 episodes); or ceftazidime and amikacin followed by oral ciprofloxacin (regimen 3, 24 episodes). Microbiologically documented infections were the cause of fever in 10 (40 percent), seven (23 percent), and nine (38 percent) episodes in regimens 1, 2, and 3, respectively, including six, five, and four episodes of bacteremia. Patient survival was 90 to 92 percent in each regimen; however, some modification of antimicrobial therapy occurred in 65, 44, and 41 percent of surviving patients in regimens 1, 2, and 3, respectively. The rate of clearance of initial bacteremia was 67 percent (four of six) in regimen 1, 100 percent (five of five) in regimen 2 and 50 percent (two of four) in regimen 3. Patients in regimens 1 and 3 were able to convert to orally administered ciprofloxacin in 32 (65 percent) of 49 episodes after a mean of six days of intravenous therapy. Superinfections occurred in 24, 10, and 12 percent of patients receiving regimens 1, 2, and 3, respectively, and occurred similarly for patients receiving orally administered ciprofloxacin, 12 percent (four of 32), and intravenous therapy, 17 percent (eight of 47). Parenteral ciprofloxacin was generally well tolerated. One (4 percent) of 25 patients receiving regimen 1 experienced oto- or nephrotoxicity, compared with eight (15 percent) of 54 patients receiving regimens 1, 2, and 3 (p = 0.15), including three patients who required premature termination of aminoglycoside therapy. Our data suggest that the combination of ciprofloxacin and azlocillin is an effective alternative to ceftazidime and amikacin for the initial empiric therapy of febrile neutropenic patients, is generally well tolerated, and avoids the oto- and nephrotoxicity associated with aminoglycoside use. In addition, a majority of patients could change to orally administered ciprofloxacin alone after six days of parenteral therapy.