Abstract
The shape of the dose-cancer incidence curve observed in an animal bioassay for carcinogenicity of a chemical is the result of a superposition of various mechanisms contributing to the process of carcinogenesis. For genotoxic carcinogens, for instance, a sublinear (convex; up-bent) shape could be the result of a saturation of DNA repair or of a higher rate of cell turnover associated with high-dose cytotoxicity and regenerative hyperplasia. In human cancer epidemiology, the situation is more complex. The human population is very heterogeneous with respect to both genetic and life-style factors that modulate the process of tumor formation. Therefore, individuals are expected to show widely variable susceptibility to carcinogenic factors, and the dose-response curve is in fact a reflection of the tolerance distribution. Each modulating factor divides the population up into sub populations of different susceptibility so that nonlinearities that could be present in a homogeneous population are flattened out. A linear extrapolation of a human cancer risk to low dose might therefore be appropriate under certain conditions even if the dose-response curve in animals has a strongly sigmoidal shape. For cancer prevention, the investigation of susceptibility factors is expected to be of great value. The elimination of concurrent risk factors in high-risk subpopulations or individuals might be more effective than a minor general reduction of a tolerable exposure level.