Use Of Lethality Data During Categorical Regression Modeling Of Acute Reference Exposures
On this page:
Categorical regression is being considered by the U.S. EPA as an additional tool for derivation of acute reference exposures (AREs) to be used for human health risk assessment for exposure to inhaled chemicals. Categorical regression is used to calculate probability-response functions for health effects data which has been classified into ordered severity categories such as no-observed-adverse-effect level (NOAEL), adverse effect level (AEL) and lethal effect level (FEL). The categorical regression model calculates the exposure concentrations and durations at which there is a 10% probability of adverse effects, the EC-T10, which can be used to derive AREs. Concerned that the use of lethality data would not be conservative, peer reviewers of the categorical regression approach recommended excluding lethality data from EC-T10 calculations. To determine the effect of removing lethality data, data sets for the acute toxicity of perchloroethylene, methyl isocyanate, phosgene, trichloroethylene, hydrogen sulfide and ethylene oxide were analyzed with and without lethality data. The effect of excluding lethality data seemed to be related to the amount of lethality data in the database. EC-T10s for databases with the largest amounts of lethality data were increased when lethality data was excluded, while EC-T10s for databases with the smallest amounts of lethality data were decrease. To confirm this observation, analyses were performed on hypothetical data sets with various amounts of lethality data. Calculations with hypothetical data sets did not confirm the observation on actual data sets, addition or removal of lethality data did not change the EC-T10. This study showed that the percentage of lethality data in a data set does not determine how the inclusion or exclusion of lethality data affects the ECT-10. Further investigations may examine the location of lethality points at low vs high concentrations and durations; and the intermingling of NOAEL, AEL and FEL data.