Given the immense variety of compounds being developed for introduction into the human environment, reliablemedium term alternatives to traditional long term rodent test protocols for carcinogen risk assessment are a highpriority. In vivo models are necessary because it has been well established that there is a lack of complete correlationbetween mutagenicity and carcinogenicity . Optimally, they should be able to detect not only complete carcinogenicor promoting potential, but also any ability to inhibit neoplasia. In order to be effective, they must take into accountthe detailed available knowledge on mechanisms of action of carcinogens and modulating agents. To allow shorteningof the time period, attention must be concentrated on preneoplaqstic lesions and other surrogate For the liver, auniquely comprehensive set of background data have already been accumulated with the Ito model, for which, has asolid scientific basis, with quantitation of glutathione S transferase positive foci as the preneoplasia-based surrogateendpoint (PSE). A very practical candidate for routine application, its predictive power, flexibility and capacity toincorporate a range of mechanism-based surrogate endpoints (MSEs) can also provide a powerful tool for attainmentof the twin goals of detecting carcinogenic agents and identifying promising chemopreventors.