Application of a Monte Carlo Based Approach for Modeling Noncarcinogenic Dose Response. Jeff Swartout, Linda Knauf-Teuschler, Paul S. Price, Carol A. Gillis, and Russell E. Keenan, U.S. Environmental Protection Agency, National Center for Environmental Assessment, Cincinnati, OH; and McLaren/Hart-ChemRisk, Portland, ME
Probabilistic Reference Doses (RfDs) and dose-response curves are estimated for six chemicals using methods developed under the auspices of the Cooperative Research and Development Agreement (CRADA) between the U.S. EPA and ChemRisk, Inc. The technical details of the methodology are presented in a companion presentation (Price et al., 1995). The approach requires the estimation of a dose level that results in no effects in the test animals (ED0a) and a dose level that causes the critical effect in 50% of the test animals (ED50a). The No Observed Adverse Effect Level (NOAEL) is used as an estimate of ED0a. The ED50a is estimated using the Benchmark Dose (BMD) procedure of Crump (1984). Uncertainty factors for differences in laboratory animal and human susceptibilities (UFA) and for differences in human sensitivity (UFH) are used to scale the dose levels to humans. Distributions are assigned to each of the inputs of the noncancer risk model of Price et al. (1995) and Monte Carlo analyses are conducted. The distributions for ED0a and ED50a are based on BMD confidence intervals estimated from toxicology studies for each chemical. The uncertainty factor distributions for UFA and UFH are hypothetical but consistent with the RfD concept. A description of UFA and UFH based on empirical data is necessary to characterize the risk more accurately. Results of the Monte Carlo procedure are presented as distributions of human risk corresponding to exposures above the RfD for the six chemicals. Median risk is zero for all exposures at 10 x RfD or less for all simulations as a result of the uncertainty factor distribution assumptions. The upper 95% confidence interval on the risk at 10 x RfD ranges from 0.8 to 0.23 depending on the steepness of the dose-response curve for the chemical. Median exposure levels associated with a 0.05 risk range from 15 x RfD to 40 x RfD for the chemicals studied. The risk levels are highly dependent on the estimate of ED0a. An unbiased method for the estimation of ED0a and its distribution has not yet been found. The NOAEL is limited in its usefulness as an estimator for the ED0a because it is limited to tested exposure levels and is somewhat insensitive to the number of animals in the test group.