The Data-Generating Process for Measures of Relative Toxicity: Implications for Risk Assessment. K. P. Brand, L. Rhomberg, and J. S. Evans, Harvard SPH, Boston, MA
Measures of relative toxicity observed between different exposure scenarios (e.g., subchronic versus chronic), different endpoints (e.g., adult versus developmental), or different species are often used in the specification of extrapolations used to estimate a toxicity of interest, for example chronic human toxicity, from a surrogate bioassay outcome such as a sub-chronic toxicity obtained in a rat bioassay. Ratios of NOAELs and ratios of Benchmark doses represent two such measures of relative toxicity. Summary statistics (central tendency and spread) of the former have widely served as a reference in the specification or "groundtruthing" of noncancer adjustment factors (aka, safety factors or uncertainty factors), while the latter has been used to explore the relative plausibility of interspecies scaling rules. How informative (accurate and precise) are NOAEL-ratio or BMD-ratio summary statistics? The answer to this question, which requires a careful consideration of the role of artifact (experimental design and other exogenous factors in influencing the observed bioassay outcomes), is essential if appropriate interpretations of bioassay evidence are to be made. We have developed an approach to meet these needs. By simulating calibrations for any summary statistic of interest, we can identify artifacts and characterize ensuing uncertainties. Our analysis reveals several artifacts and demonstrates that summary statistics of NOAEL ratios can not be taken at face value. BMD ratios are similarly confounded. The calibration approach is demonstrated, and implications for regulatory toxicology discussed.
Funding from HCRA gratefully acknowledged.
Go to . . .