Epidemiologic input to environmental risk assessment

Epidemiological data are often rejected for risk assessment purposes because of the lack of accurate exposure data. Animal data are frequently resorted for risk assessment calculations. Such a decision should account for potential errors in animal to human risk extrapolations, which may be of 1 or 2 orders of magnitude or more. Because historical human exposure can usually be estimated with confidence that the errors would be considerably less than 1 order of magnitude, it follows that human studies should generally take precedence over animal studies, even when crude assumptions have to be made concerning human exposure levels. This paper illustrates this with cancer risk assessments for vinyl chloride and antimony, both of which relied on use of the stipulated average cohort exposure levels without use of any individual exposure data for cohort members. Linear extrapolations of risk to low exposures were used in both instances. Non-cancer risk assessments were illustrated by examining results of experimental studies of effects of ozone on ventilatory function. A simple approach, in this case involving visual inspection of study results, was again recommended. Consideration was given to concerns about synergistic effects of low-level exposure to chemical mixtures. Calculations were presented to demonstrate that estimating risks by multiplying effects of individual chemicals produces the same risk estimated by adding effects, provided that exposure levels were sufficiently low for each individual agent. It was concluded that epidemiology has much to offer to environmental health risk assessment, but that epidemiologists need to be more effective in justifying the scientific basis of the design and interpretation of epidemiological studies to laboratory scientists, who frequently reject epidemiological findings for inappropriate reasons.