Beyond screening: problems and prospects for risk characterization of endocrine disruptors

Over time, the NAS's original concept of "risk characterization" has broadened to include not just explication of the risk inferences themselves, but a forthright discussion of their dependence on the many less-than-certain elements of the inferential process. Ideally, such characterization communicates the span of possible conclusions that emerge from alternative reasonable interpretations of the information at hand, together with considered judgment regarding the likelihood that each interpretation provides a good representation of the actual situation. Uncertainties regarding the magnitudes of parameters in the various equations used in quantitative risk assessment are increasingly characterized using statistical tools for analysis and propagation of variance through the calculations. Of potentially greater concern, however, is "model uncertainty"-the more fundamental uncertainty regarding the soundness of our conception of what biological phenomena are operating, how they interact, and how they can be represented in quantitative models. This is particularly true for assessment of potential endocrine disruptors, since much fundamental information is lacking and unanswered questions exist about the consequences and interpretation of empirically observable effects. An approach is to define a series of alternative models, each embodying a different conception, and using expert scientific judgment to weight the individual results.