Genistein (GEN) has recently generated considerable interest as a potential agent for the prevention and treatment of cancer. The present investigation was undertaken to determine if the concentrations of drug shown to inhibit the growth of human tumor cell lines by 50% in vitro (IC50 = 2-27 microg/ml) can be achieved and sustained systemically in mice. We found that GEN plasma levels decreased biexponetially from 64 microg/ml to 0.55 microg/ml during the initial 40 min after i.v. injection of a 52 mg/kg dose. Mean half-lives of the two initial disposition phases were 2.5 plus minus 0.4 min and 7.1 plus minus 1.1 min in mice treated with doses of 9-52 mg/kg. Plasma profiles of i.v. GEN exhibited a prominent secondary peak near 78 min followed by a terminal decay phase with a 39.5 plus minus 16.8 min half-life. Although these features are suggestive of enterohepatic cycling, the mean apparent total plasma clearance of GEN (66.5 plus minus 7.3 ml/min/kg) was nevertheless similar to hepatic blood flow. The systemic availability of GEN from a 180 mg/kg p.o. dose, which afforded 1.1 microg/ml peak plasma concentration, was only 12%. Thus, bolus i.v. and p.o. administration of GEN failed to either achieve or adequately sustain plasma levels of the drug within the target range established by in vitro antitumor studies. Plasma levels resulting from i.p. injection of a 185 mg/kg dose were 5-times greater on average than achieved by the p.o. route. While the plasma concentration exceeded the IC50 values for the majority of human cancer cell lines responsive to GEN for only a short period of time, drug levels remained above 2 microg/ml, the IC50 of the most sensitive cell lines, for 4 h. Extrapolation from the single dose study suggests that repetitive i.p. injection of at least 200 mg/kg GEN every 8 h will afford continuous systemic exposure to potentially cytostatic concentrations of the drug against these cell lines. This information should facilitate efforts to assess the effectiveness of GEN in appropriate in vivo tumor models.
