The effects of pre- and postprandial levels of lipids in serum on the experimental in vivo and in vitro toxicities of amphotericin B deoxycholate (AmB-d) were studied. Normal OF1 mice were tested at baseline, after normal feeding, after 3 h of fasting, or after a sequence of feeding and fasting and vice versa. The 50% lethal dose (LD50) of AmB-d was significantly higher in fed mice than in mice which fasted or at baseline (2.38 +/- 0.12 versus 1.53 +/- 0.2 and 1.50 +/- 0.1 mg/kg of body weight, respectively; P < 0.05). When different nutritional regimens were alternated over a short period, the level of in vivo AmB-d toxicity was dictated by the last feeding regimen. Serum triglycerides, but not cholesterol in very-low-density and low-density lipoproteins, correlated significantly (P < 0.01) with the LD50 of AmB. In vitro experiments showed that the addition of human serum reduced AmB-d-induced toxicity against human erythrocytes, but serum drawn after fasting was less protective than postprandial serum. However, neither serum decreased the in vitro activity of AmB-d against Candida albicans. Circular dichroism, a method that enables the amount of free AmB to be measured, showed that both mouse and human total serum lipoproteins bound more AmB-d when serum was isolated postprandially than when it was obtained after fasting. Our results show that AmB-d toxicity is reduced by feeding-induced modifications in serum lipids. The influence of food intake on the clinical toxicity of the drug merits being investigated.