be predicted with certainty. Nonstochastic events such as enzyme inhibition following exposure to a certain toxicant are more predictable. Nonstochastic effects may not be seen if lifetime exposures are below a certain threshold concentration, whereas stochastic effects such as carcinogenesis may have no threshold. Different approaches to risk assessment are usually used for these two categories of response.

Information on health risks may be obtained directly from epidemiological and other studies of humans or indirectly through toxicological experiments conducted in animal models. Although results of laboratory studies are applicable to humans only indirectly, they can be used to predict potential health hazards in advance of actual human exposure and thus continue to be used widely to identify substances with potential toxicity.

Three major problems are inherent in the translation of results of animal tests to humans:

• Laboratory tests are conducted at relatively high doses to induce measurable rates of response in a small sample of animals. These results must often extrapolated to lower doses that correspond to anticipated human exposure levels.

• Interspecies differences must be considered when extrapolating between the animal model and humans.

• It may be necessary to extrapolate from a route of exposure chosen for experimental practicality to a different but more likely route of human exposure.

Classical methods in toxicological risk assessment are applicable with nonstochastic toxic effects. For many kinds of agents and end points, toxicity is manifest only after the depletion of a physiological reserve. In addition, the biological repair capacity of many tissues can accommodate a certain degree of damage by reversible toxic processes (Aldridge, 1986; Klaassen, 1986). Above this threshold, however, the compensatory mechanisms that maintain normal biological function may be overwhelmed, leading to organ dysfunction. The objective of classical toxicological risk assessment, which has focused on nonstochastic end points, has been to establish a threshold dose below which adverse health effects were expected to be rare or absent.

Historically, the threshold concept was introduced by Lehman and Fitzhugh (1954), who proposed that an acceptable daily intake (ADI) could be calculated for chemical contaminants in human food. This concept was