We performed a randomized clinical trial in granulocytopenic patients with carcinoma or leukemia. Patients with persistent fever for more than 2 days despite antibiotic therapy were randomized to antibiotic plus fluconazole therapy group (FLCZ group) or antibiotic therapy only group (antibiotic group) by the envelope method. It was possible to evaluate clinical efficacies in 62 patients (37 patients in FLCZ group and 25 patients in antibiotics group). In patients whose neutrophil counts were less than 100/microliters on the initial day of therapy, clinical efficacy rates were 72.0% (18/25) in FLCZ group and 57.1% (8/14) in antibiotics group. In patients whose neutrophil counts continued to be less than 100/microliters during therapy, clinical efficacy rates were 64.3% (9/14) and 50.0% (3/6), respectively. Further, in patients whose neutrophil counts continued to be less than 500/microliters during therapy, they were 76.9% (20/26) and 53.3% (8/15), respectively. No severe side effects nor severe case of abnormal change in laboratory test values due to fluconazole were observed in this trial. These data suggest that empiric antifungal therapy with fluconazole is effective for fungal infections in granulocytopenic patients with carcinoma and leukemia.