The efficacy of a novel sterol-complexed preparation of amphotericin B, amphotericin B colloidal dispersion, was compared with that of deoxycholate-complexed amphotericin B in an acute murine model of systemic coccidioidomycosis. Mice (CD-1, female) were infected intravenously with 180 or 200 arthroconidia of Coccidioides immitis, and intravenous therapy was begun 3 days later. Six doses in various regimens of either preparation were given over 14 days, and deaths were tallied for an additional 35 days. All regimens that were not acutely lethal prolonged the survival of mice over that of controls (P less than 0.001). Quantitative determination of residual burdens of C. immitis in the spleen, liver, and lungs of survivors revealed that the colloidal dispersion was not as effective as the deoxycholate suspension on a milligram-per-kilogram basis. Deoxycholate suspension at 1.3 mg/kg cleared the organs in all mice, whereas colloidal dispersion at 5.0 mg/kg was the lowest dose that cleared organisms from all animals. Lower doses cleared organisms from fewer animals or cleared only selected organs. Deoxycholate suspension was more efficacious than colloidal dispersion in clearing C. immitis from the liver or lungs (P less than 0.05 to 0.01, dose and organ dependent) at identical doses. No overt toxicity was observed in mice treated with colloidal dispersion at 10 mg/kg. In contrast, deoxycholate suspension at 2.0 mg/kg was acutely toxic; 50% of the treated mice died after treatment. The two complexes were not equivalent on a milligram-per-kilogram basis; the deoxycholate suspension was three to four times more efficacious and also greater than 5- to greater than or equal to 8-fold more toxic. Thus, the therapeutic index of the colloidal dispersion complex is greater than that of the deoxycholate complex. The amount of amphotericin B per dose could also be increased when given as a colloidal dispersion to an optimally level. Amphotericin B colloidal dispersion shows promise for the therapy of disseminated coccidioidomycosis and should be tested in other animal models and in humans.