OBJECTIVE To test the hypothesis that amphotericin B deoxycholate is less toxic when given by continuous infusion than by conventional rapid infusion. DESIGN Randomised, controlled, non-blinded, single centre study. SETTING University hospital providing tertiary clinical care. PATIENTS 80 mostly neutropenic patients with refractory fever and suspected or proved invasive fungal infections. INTERVENTION Patients were randomised to receive 0.97 mg/kg amphotericin B by continuous infusion over 24 hours or 0.95 mg/kg by rapid infusion over four hours. MAIN OUTCOME MEASURES Patients were evaluated for side effects related to infusion, nephrotoxicity, and mortality up to three months after treatment. Analysis was on an intention to treat basis. RESULTS Patients in the continuous infusion group had fewer side effects and significantly reduced nephrotoxicity than those in the rapid infusion group. Overall mortality was higher during treatment and after three months' follow up in the rapid infusion than in the continuous infusion group. CONCLUSION Continuous infusions of amphotericin B reduce nephrotoxicity and side effects related to infusion without increasing mortality.