With the availability of new, broad-spectrum antibiotics, initial therapy with a single agent has become an alternative to classic combinations in the management of febrile, neutropenic cancer patients. The aims of this study were to compare the efficacy of ceftazidime and imipenem as empirical monotherapy of febrile episodes in neutropenic patients, and to examine the frequency with which second-line antibiotics (amikacin, vancomycin, or both) were required. A prospective clinical trial was carried out in a single centre. Eligible patients with solid tumours or lymphoma were randomised to receive monotherapy with ceftazidime or imipenem. In the event of no response, amikacin and/or vancomycin were added in 48-72 h intervals (sequentially, or according to clinical or microbiological data). Efficacy was evaluable for 111 assessable episodes. Median neutrophil count at entry was 100 cells/microliters and median duration of neutropenia was 4 days. Febrile episodes were classified as microbiologically (34%) or clinically documented (42%), and fever of unknown origin (24%). Gram-negative infections (57%) predominated over gram-positive isolates (30%). The overall success rate with monotherapy (69% versus 70%), or with modification (20% versus 23%) were equivalent for ceftazidime and imipenem (P = 0.75). The mortality in this series was 5%. Single-agent therapy with either ceftazidime or imipenem is effective for the empirical treatment of febrile episodes in neutropenic patients with solid tumours. Early addition of amikacin and/or vancomycin resolves most failures of the first step.