One hundred and thirty-four adults and 204 children were randomized in two prospective, parallel comparative multicentre trials to receive either conventional amphotericin B 1 mg/kg/d (c-AMB), liposomal amphotericin B 1 mg/kg/d(L-AMB1) or liposomal amphotericin B 3 mg/ kg/d (L-AMB3). Patients were entered if they had a pyrexia of unknown origin (PUO) defined as temperature of 38 degrees C or more, not responding to 96 h of systemic broad-spectrum antibiotic treatment, and neutropenia (< 0.5 x 10(9)/l). The safety and toxicity of liposomal amphotericin B was compared with that of conventional amphotericin B. Efficacy of treatment was assessed, with success defined as resolution of fever for 3 consecutive days (< 38 degrees C) without the development of any new fungal infection. Clinical and laboratory parameters were collected for safety analysis. In both the paediatric and adult populations, L-AMB treated patients had a 2-6-fold decrease in the incidence (P < or = 0.01) of test-drug-related side-effects, compared to c-AMB. Severe trial-drug-related side-effects were seen in 1% of L-AMB treated patients, in contrast to 12% of patients on c-AMB (P < 0.01). Nephrotoxicity, in the patient subset not receiving concomitant nephrotoxic agents, defined as a doubling from the patients baseline serum creatinine level, was not observed in the L-AMB1 arm whereas the incidence was 3% in patients on L-AMB3 and 23% in those on c-AMB (P < 0.01). Moreover, time to develop nephrotoxicity was longer in both L-AMB arms than c-AMB (P < 0.01). Severe hypokalaemia was observed less frequently in both L-AMB arms (P < 0.01). Analysis was by intention-to-treat and included all patients randomized. Success was defined by a minimum of 3 consecutive days with fever (< 38 degrees C) continuing to study end indicated by recovery of neutrophils to 0.5 x 10(9)/l. Addition of systemic antifungal therapy or development of systemic fungal infection were failures as was persistent fever to study end. Efficacy assessments indicated success in 49% of the total group treated with c-AMB, 58% of patients responded to L-AMB1 and 64% to L-AMB3. A statistically significant difference was found between c-AMB and L-AMB3 (P = 0.03) but a Kaplan-Meier analysis of time to differvescence of fever showed there was no significant difference between the arms. It was concluded that liposomal amphotericin at either 1 or 3 mg/kg/d was significantly safer than conventional amphotericin B in children and adults. The main aim of this open-label study was to compare safety between the three trial arms. However, we provide evidence for an equivalent or possibly superior efficacy of liposomal amphotericin with regard to resolution of fever of unknown origin. Subsequent trials should compare amphotericin preparations in defined fungal infections.