Chemicals can increase carcinogenic risk by either directly damaging DNA or increasing cell replication or they can do both. These effects have different implications for a biologically-based extrapolation from rodent bioassays to humans. 2-Acetylaminofluorene (2-AAF) administered at low doses to mice for a lifetime has a different dose-response for the liver (approximately linear) compared to the urinary bladder (apparent no effect dose of 45 ppm with a sigmoidal dose response at 60-150 ppm), which can be explained if carcinogen metabolism, DNA adduct formation and cell proliferation effects are considered. In contrast to 2-AAF and other genotoxic chemicals, chemicals which form calculi in the urine do not generally damage DNA directly but increase cell proliferation dramatically by eroding the bladder surface, leading to regenerative hyperplasia. This occurs only at doses at which calculi form; lower doses do not produce calculi and, therefore, do not increase cell proliferation or cause tumors. Extrapolation to humans from the rodent bioassay should be dependent on dose requirements for formation of calculi rather than any type of statistical extrapolation to lower doses. Saccharin and other sodium salts administered at high doses to rats also produce bladder cancer by increasing cell proliferation. These salts do not affect mice, hamsters, guinea pig or monkeys. Based on dose and mechanistic considerations, saccharin and these other sodium salts are unlikely to be human carcinogens. Extrapolation to possible human cancer risk requires biological determinations rather than simply using statistical formulations.