In this prospective multicenter trial, treatment strategies for 1573 patients with neutropenia < 1000/microliters and fever > or = 38.5 degrees C after cytotoxic chemotherapy were compared. Patients with unexplained fever were randomized to a three-phase sequential study for different established drug regimens. If an infection could be defined microbiologically or clinically, treatment modifications were determined. In phase I, treatment for all patients consisted of acylaminopenicillin (PEN) plus aminoglycoside (AMG); or third-generation cephalosporin (CEPH) plus AMG; or PEN plus CEPH. In 800 patients with unexplained fever the response rates were: PEN/AMG (n = 258): 74.4%, CEPH/AMG (n = 252): 73.4%; PEN/CEPH (n = 290): 70.0%. Total response rate was 72.5%. In phase II, patients not responding after 3 days received PEN/CEPH/vancomycin (n = 70) or PEN/CEPH/AMG (n = 74). The respective response rates were 52.9% and 55.4%, total 54.2%. If fever did not resolve, the patients received either PEN/CEPH (n = 40) or imipenem/cilastatin (n = 59) both in combination with amphotericin-B/5-flucytosin/rifampin. The response rates were 62.5% and 79.7%, respectively (p = 0.07), total 72.7%. No significant differences between the treatment modalities compared were found. Analyzing all three phases together, 91.3% of patients with unexplained fever were cured. The response rate was also analyzed according to patients with gram-positive bacteremia (n = 183), response rate = 82.5%; gram-negative organisms (n = 145) 78.6%; fungemia (n = 51) 43.1% (p < 0.001); lung infiltrates (n = 269) 61.3% (p < 0.001); clinically documented infections (n = 198) 84.4%; and clinically and microbiologically documented infections (n = 84) 82.1%. If infections were diagnosed after at least 5 febrile days, more lung infiltrates and fungal infections occurred (p < 0.001). Leukocytes rising above 500/mu during the infection predicted better response rates (p < 0.001): in unexplained fever 97.8% vs 86.5% and lower death rates 1.5% vs 8.5%. In documented infections the response rates were then 89.9% vs 62.3% and the death rates 7.0% vs 20.5%. Therapy of neutropenic fever and infections must be adapted according to risk factors and should include early empiric antifungal therapy. The therapeutic and prophylactic use of hematopoietic growth factors to overcome neutropenia should be evaluated.