As determined by in vitro tests, anhydroretinol, a metabolic product of retinol, was bound specifically by serum retinol-binding protein and by cellular retinol-binding protein but not by cellular retinoic acid-binding protein or the nuclear receptors, RARs and RXRs. For rats dosed with the mammary carcinogen, N-methyl-N-nitrosourea (45 mg/kg body weight) and given diets containing either the retinoid vehicle, anhydroretinol (67, 134, 268, or 536 mg/kg of diet), or retinyl acetate (328 mg/kg of diet), there were, over a 90-day observation period, no significant differences in body weights. The compound did not accumulate in liver tissue or cause an increase in hepatic levels of retinyl palmitate (potential problems observed with other retinoids). The numbers of mammary cancers were as follows: no retinoid, 4.5/rat; retinyl acetate, 2.1/rat; and increasing doses of anhydroretinol, 2.9, 3.3, 3.0, and 1.7/rat, respectively. Thus, anhydroretinol, at non-toxic levels, was effective as a preventive agent in this experimental model of breast cancer.