"Cell-kinetic multistage" (CKM) models account for clonal growth of intermediate, premalignant cell populations and thus distinguish somatic mutations and cell proliferation as separate processes that may influence observed rates of tumor formation. This paper illustrates the application of two versions of a two-stage CKM model (one assuming exponential and the other geometric proliferation of intermediate cells) for extrapolating cancer risk potentially associated with exposure to carbon tetrachloride, chloroform, and dichloromethane, three suspect human carcinogens commonly present in trace amounts in groundwater supplies used for domestic consumption. For each compound, the models were used to calculate a daily oral "virtually safe dose" (VSD) to humans associated with a cancer risk of 10(-6), extrapolated from bioassay data on increased hepatocellular tumor incidence in B6C3F1 mice. Exposure-induced bioassay tumor responses were assumed first to be due solely to "promotion" (enhanced proliferation of premalignant cells, here associated with cytotoxicity), in accordance with the majority of available data on in vivo genotoxicity for these compounds. Available data were used to model dose response for induced hepatocellular proliferation in mice for each compound. Physiologically based pharmacokinetic models were used to predict the hepatotoxic effect (metabolized) dose as a function of parent compound administered dose in mice and in humans. Resulting calculated VSDs are shown to be from three to five orders of magnitude greater than corresponding values obtained assuming each of the compounds is carcinogenic only through induced somatic mutations within the CKM framework. Key issues and uncertainties in applying CKM models to risk assessment for cancer promoters are discussed.