A methodology for using physiologically based pharmacokinetic (PBPK) models to derive predicted safe concentrations of noncarcinogens in drinking water for humans based on experimentally determined no observed adverse effect levels (NOAELs) in animals is presented and applied to the case of 1,1,1-trichloroethane (methyl chloroform, MC). For each toxic endpoint and lowest corresponding NOAEL identified for MC, we considered a set of toxicologically plausible options regarding the presumed toxic agent and the metric for effective dose to target tissue. A four-compartment PBPK model for rodents was used to estimate corresponding effective doses to the animals used to obtain the experimental NOAELs. A five-compartment PBPK model was then applied, in conjunction with a multiroute (inhalation, ingestion and dermal) human-exposure scenario, to calculate alternative concentrations of MC in drinking water predicted to result in corresponding effective doses to the same target tissues in humans. In the case of MC, the PBPK approach to interspecies and interroute extrapolation of toxicity data resulted in lower drinking water concentrations predicted to be nontoxic to humans than corresponding concentrations obtained using a traditional method for determining safe levels.