Hidden assumptions in environmental research.

A major challenge in both the design of epidemiologic studies of exposure to toxic agents and the interpretation of the findings is to recognize and to accommodate the complex biologic assumptions that underlie population-based research. Environmental health research is grounded on a scientific foundation that blends toxicology with epidemiology. Toxicology without human population studies often leads to findings uninterpretable for purposes of risk assessment. Epidemiologic studies are often less helpful than they could be because the variation in response is seldom addressed, the effects under study are often small in magnitude and easily confounded, and exposure estimates are crude. The net bias is to reduce the risk estimate, perhaps substantially. Biological variation has usually been treated as "noise" in the system but it is a more basic problem that lends itself to investigation and quantification. Improved modeling of the biological determinants of response is needed if results are to be generalizable and interpretable in mechanistic terms.