Cancer and non-cancer risk assessment: not so different if you consider mechanisms

Default risk assessment procedures use threshold models for non-carcinogens and a non-threshold model for carcinogens. This a priori distinction reflects the fact that the default procedures do not consider mechanisms of action of specific chemicals. When mechanisms are considered, the distinction is not necessary. Starting with the premise that the goal of risk assessment is to identify actual risk for specific chemicals, three major, generic components of the overall mechanism translating exposure into a response of regulatory interest are identified. These are the specific mechanisms linking (1) exposure with dose to target tissue, (2) target tissue dose with short-term responses such as cytolethality or mutation, and (3) short-term responses with ensuing long-term responses such as cancer or cirrhosis. (Short-term responses may be regulatory end points of interest, or they may be intermediate steps on the way to longer-term sequelae). On-going research on formaldehyde and chloroform is described to illustrate how these three components of the overall mechanism can be examined experimentally and used in specific models. The impact of mechanism-based risk assessment on uncertainty is also considered. Uncertainty is a function of the extent to which the model used for risk assessment misspecifies the actual mechanism of action for the chemical in question. There is a trade-off between (a) mechanism-based models that may reduce uncertainty but are expensive and time-consuming to develop and (b) default models that are not chemical-specific but can be used with minimal data sets. Experience with mechanism-based risk assessment may allow modification of default procedures to minimize this trade-off. A future default procedure for carcinogen risk assessment might allow specification of mode of action. For example, while DNA reactive-carcinogens would still be assumed to have linear low-dose risk, carcinogens acting through purely cytotoxic mechanisms might be assumed to have sharply non-linear or even threshold dose-response curves.