The Benchmark Dose Concept: Questionable Utility for Risk Assessment? T. B. Starr, ENVIRON International Corporation, 7500 Rainwater Road, Raleigh, NC 27615-3700
Estimation of benchmark doses has been proposed as an alternative and preferred methodology to the use of No Observed Adverse Effect Levels (NOAELs) in setting "safe" levels of exposure to materials that cause toxic but non-carcinogenic effects. This approach employs dose-response data to estimate the parameters of a flexible mathematical model, e.g., the Weibull model, which is then extrapolated below the observable response range to "benchmark" doses that are predicted to yield a specific level of response, such as 1% or 5%. Conservative lower bounds on the benchmark dose, e.g., the lower 95% confidence limit, have been proposed for use in regulatory decision-making. This raises several concerns. First, it must be recognized that this approach is generic, empirical, and lacking of any scientifically established mechanistic foundation. Without explicit incorporation of information regarding mechanisms of toxic action, it is nothing more than "curve-fitting." Second, the comparatively small group sizes employed in most toxicology studies impart a substantial degree of conservatism to lower bounds relative to central estimates. For example, Crump has reported that lower 95% bounds on 1% benchmark doses are, on average, about 30-fold lower than the NOAELs that could be identified using the very same studies. Is this additional conservatism warranted? Third, Crump has also noted that the small number of dose groups in most toxicology studies and their similarly small size nearly always preclude rejection of a linear lower bound. This approach will thus serve to linearize the regulation of non-cancer endpoints. Is there any justification for abandoning the traditional threshold concept of toxicology in favor of the biologically implausible "even one molecule is enough" concept for all non-cancer endpoints? These concerns merit serious consideration by the risk assessment community.