A study was carried out to compare the use of prophylactic imipenem administered at the onset of profound neutropenia (immediate) with therapeutic imipenem administered at the onset of neutropenic fever (delayed) in cancer patients treated with high-dose chemotherapy. A total of 65 patients who were scheduled to receive two cycles of high-dose cyclophosphamide, etoposide, cisplatin (CEP) chemotherapy were randomized to receive imipenem either at presentation of neutropenia (immediate imipenem arm, prophylactic arm) or at commencement of neutropenic fever (delayed imipenem arm, therapeutic arm). Treatment was crossed over when the second CEP chemotherapy cycle was received. Of the 65 patients, 41 received the two planned cycles and 24 received only the first. Compared with the delayed imipenem arm, the immediate imipenem arm was associated with lower fever incidence (86.3% vs. 100%, p=0.0142) and Gram-negative bacteria infection [4/51 (7.8%) vs. 14/55 (25.5%), OR=0.24, p =0.031]. There were fewer episodes of pneumonia (2% vs. 12.7%), septic shock (0% vs. 3.6%) and deaths from infection (0% vs. 3.6%), but these differences did not reach statistical significance. With regard to delayed imipenem, for every seven patients with immediate imipenem, one episode of febrile neutropenia was avoided; for every six patients administered immediate imipenem, one case of Gram-negative infection was avoided; and for every nine patients administered immediate imipenem, one episode of pneumonia was avoided. There were no differences in the incidence of Gram-positive infections nor in the length of hospitalization between the two treatment arms. In conclusion, compared to its conventional delayed use, immediate imipenem significantly reduces the frequency of febrile neutropenia and Gram-negative infections in patients with high-dose chemotherapy.