the relationships between variables and the health effect.
Consider the situation of a new chemical that is proposed for use or that will be a byproduct of some new operation. Direct evidence that the chemical causes adverse effects in humans is lacking. Some important considerations include the potential for adverse health effects in humans exposed to the chemical and, if adverse effects do result from exposure, the magnitude of the effects after exposures of different severities. It is clear that predictions are required. However, the basis for the predictions cannot be the previous human experience; there is no previous experience. It might be possible to extrapolate, but the model from which the extrapolation is made will of necessity be a nonhuman model.
In the scenario just described, the need for extrapolations from experimental systems to humans is apparent. Many other scenarios, both clinical and "population-based," will require the prediction of human responses from data obtained in nonhuman test systems, especially in light of the thousands of chemicals that are produced, used, and released into the environment as byproducts of our way of life. Indeed, it is the desire to be predictive that drives the need to develop and apply good experimental systems. Such systems have at least four advantages: they allow predictions of human health effects and the magnitude of those effects before human exposure occurs or before adverse effects are manifested in exposed populations; they can be altered to clarify aspects of the process leading from exposure to adverse health effects when similar experimentation in humans would be unethical; they can be designed to eliminate many factors that confound the determination of cause-effect relationships in epidemiologic studies; and they can suggest directions for epidemiologic investigation by providing the hypotheses that epidemiologic studies might be able to test.
Previous chapters have focused on markers of susceptibility, exposure, and effect (particularly early effect) and their value in clinical situations; the sooner a disease state or precursor of a disease state can be identified, the greater the chance of successful therapy or treatment. This chapter focuses on the prediction of effects, not in an individual patient but rather in a (hypothetical) population of humans potentially exposed to a supposed toxicant. In this context, one is concerned about maintaining the health of the population by predicting whether an activity or an exposure is likely to produce harmful consequences in that population—often without previous observations of humans exposed at the magnitudes of interest. The objective is to learn how to tie chemical exposure under various scenarios to the dose or amount of the chemical that reaches the body, to the amount that is absorbed and distributed to target tissues, and ultimately to the effect.
How does one make such predictions? The discipline of risk assessment addresses that question. Human-health risk assessment is a complex,