Amphotericin B has long been the standard treatment for candidemia, but its use is complicated by its toxicity. More recently, fluconazole, a water-soluble triazole with activity against candida species and little toxicity, has become available. We conducted a multicenter randomized trial that compared amphotericin B with fluconazole as treatment for candidemia. To be eligible, patients had to have a positive blood culture for candida species, a neutrophil count > or = 500 per cubic millimeter, and no major immunodeficiency. Patients were randomly assigned to receive either amphotericin B (0.5 to 0.6 mg per kilogram of body weight per day) or fluconazole (400 mg per day), each continued for at least 14 days after the last positive blood culture. Outcomes were assessed by a group of investigators blinded to treatment assignment. Of the 237 patients enrolled, 206 met all entry criteria. The most common diagnoses were renal failure, nonhematologic cancer, and gastrointestinal disease. There was no statistically significant difference in outcome: of the 103 patients treated with amphotericin B, 81 (79 percent) were judged to have been treated successfully, as were 72 of the 103 patients treated with fluconazole (70 percent P = 0.22; 95 percent confidence interval for the difference, -5 to 23 percent). The bloodstream infection failed to clear in 12 patients in the amphotericin group and 15 in the fluconazole group; the species most commonly associated with failure was Candida albicans. There were 41 deaths in the amphotericin group and 34 deaths in the fluconazole group (P = 0.20). Intravascular catheters appeared to be the most frequent source of candidemia. There was less toxicity with fluconazole than with amphotericin B. In patients without neutropenia and without major immunodeficiency, fluconazole and amphotericin B are not significantly different in their effectiveness in treating candidemia.