Selecting a more realistic uncertainty factor: Reducing compounding effects of multiple uncertainties

Available toxicology datasets provide a unique opportunity to validate some of the currently used Uncertainty Factors in the development of acceptable exposure levels for noncancer effects. Toxicity studies from two separate sources, the FAO/WHO database on pesticides (1978–1987) and the Monsanto database (through 1988) were chosen to evaluate three of the five currently used Uncertainty Factors. Interspecies differences in NOELs between the three mammalian species evaluated are equal to or less than a factor of 10 for both the FAO/WHO data and the Monsanto data in greater than 90% of the cases evaluated. Median values for the comparison of interspecies NOELs were 3.0 or less for all comparisons except the comparison between the mouse and rat for the Monsanto dataset where the median value was 7.5. Analyses of the Monsanto toxicity database show that the reprotoxicity NOELs were always equivalent to or higher than the chronic or subchronic NOELs for the same material. Therefore, even without conduct of a specific study to address reproductive effects, reasonable protection from adverse reproductive effects can be afforded by use of either subchronic or chronic study NOELs without application of UFD. The median ratio of subchronic NOELs and chronic NOELs was 4, and for a majority of the studies the difference between the NOELs was within one order of magnitude. Our analysis aids in validating the assumption that the upperbound for individual uncertainties maybe accounted for by use of 10‐fold uncertainty factors. However, the current U.S. Environmental Protection Agency (USEPA) reference doses/concentrations may be overly conservative because upperbounds of each of the uncertainty factors are used and each of the uncertainty factors are considered to be independent variables. Because uncertainties are probably not independent variables, the influence of compounding upperbounds when multiple uncertainty factors are used is generally only considered when four or more areas of uncertainty are outstanding. When multiple uncertainties exist, we recommend upperbound estimates only be used for the first two Uncertainty Factors, and median values be used to account for the remaining uncertainties.