Diethanolamine (DEA) is a chemical used widely in a number of industries and is present in many consumer products. Studies by the National Toxicology Program (NTP) have indicated that lifetime dermal exposure to DEA increased the incidence and multiplicity of liver tumors in mice, but not in rats. In addition, DEA was not carcinogenic when tested in the Tg.Ac transgenic mouse model. Short-term genotoxicity tests have yielded negative results. In view of these apparent inconsistencies, we have critically evaluated the NTP studies and other data relevant to assessing the carcinogenic potential of DEA. The available data indicate that DEA induces mouse liver tumors by a non-genotoxic mode of action that involves its ability to cause choline deficiency. The following experimental evidence supports this hypothesis. DEA decreased the hepatic choline metabolites and S-adenosylmethionine levels in mice, similar to those observed in choline-deficient mice. In contrast, DEA had no effect in the rat, a species in which it was not carcinogenic at a maximum tolerated dose level. In addition, a consistent dose-effect relationship had been established between choline deficiency and carcinogenic activity since all DEA dosages that induced tumors in the NTP studies were also shown to cause choline deficiency. DEA decreased phosphatidylcholine synthesis by blocking the cellular uptake of choline in vitro, but these events did not occur in the presence of excess choline. Finally, DEA induced transformation in the Syrian hamster embryo cells, increased S-phase DNA synthesis in mouse hepatocytes, and decreased gap junctional intracellular communication in primary cultured mouse and rat hepatocytes, but all these events were prevented with choline supplementation. Since choline is an essential nutrient in mammals, this mode of action is qualitatively applicable to humans. However, there are marked species differences in susceptibility to choline deficiency, with rats and mice being far more susceptible than other mammalian species including humans. These differences are attributed to quantitative differences in the enzyme kinetics controlling choline metabolism. The fact that DEA was carcinogenic in mice but not in rats also has important implications for human risk assessment. DEA has been shown to be less readily absorbed across rat and human skin than mouse skin. Since a no observed effect level for DEA-induced choline deficiency in mice has been established to be 10 mg/kg/d, this indicates that there is a critical level of DEA that must be attained in order to affect choline homeostasis. The lack of a carcinogenic response in rats suggests that exposure to DEA did not reach this critical level. Since rodents are far more sensitive to choline deficiency than humans, it can be concluded that the hepatocarcinogenic effect of DEA in mice is not predictive of similar susceptibility in humans.