The recent increase in the clinical use of synthetic vitamin A compounds has led to concern of possible side effects. Some of these effects are known to be influenced by dietary levels of vitamin K. We therefore compared the toxic effects of 13-cis-retinoic acid (13cisRA), retinyl acetate (ROAc), and N-(4-hydroxyphenyl)retinamide (4HPR) in male Sprague-Dawley rats maintained on diets containing different levels of vitamin K. Animals were fed either an NIH-07 diet supplemented with menadione (3.1 ppm vitamin K3), an NIH-07 diet not supplemented with menadione, or an AIN-076 purified diet devoid of vitamin K. The retinoids had no effect on prothrombin times of animals fed the supplemented diet. When menadione was omitted from the diet, however, 4HPR-dosed animals had elevated prothrombin times. This effect was observed as early as Day 7 and was accompanied by one confirmed hemorrhagic death. 13cisRA-dosed animals showed no change in prothrombin times. In the high-dose ROAc group, there was a twofold increase in prothrombin times but only after prolonged dosing. In animals fed the NIH-07 diets, 13cisRA and ROAc induced multiple bone fractures at all dose levels. In contrast, 4HPR administered at the highest dose induced only one fracture in one animal. Animals fed the purified diet lost weight faster and diet sooner than those maintained on the other diets. Bone fractures were not observed in these animals because of early deaths resulting from hemorrhaging. For all retinoid-dosed groups maintained on the purified diet, changes in prothrombin times occured as early as 1 week. The order of effect was 4HPR greater than ROAc greater than 13cisRA, with increases in prothrombin times correlating with increases in hemorrhagic deaths. Hence, the degree of retinoid-induced hemorrhage, but not the incidence of bone fractures, was inversely related to vitamin K levels in the diet. 13cisRA and ROAc, but not 4HPR, caused a dose-dependent reduction in plasma osteocalcin, an effect that correlated with retinoid-induced bone effects. In contrast, serum alkaline phosphatase was elevated in animals dosed with 13cisRA or 4HPR but not in those dose with ROAc. For this enzyme, the electrophoretic pattern on agarose gel showed a decrease, compared to controls, in the major isozyme in serum of ROAc-dosed animals. Hence, plasma osteocalcin is a better predictor of retinoid-induced bone effects than serum alkaline phosphatase.