Support for Development of CatReg Software. Categorical regression is one of the newest quantitative techniques in toxicological sciences. It provides a quantitative methodology for accounting for effect severity in analyzing the dose-response of multiple toxicity endpoints. Sciences supported EPA’s National Center for Environmental Assessment in the development of the categorical regression software (CatReg) by independently verifying the validity of program results for a number of theoretical data sets. The data sets were developed using the S-Plus statistical package, the native programming language of the CatReg software. The data sets were run through CatReg and the output was evaluated against the expected regression results.

Support for Development of Benchmark Dose Software. Benchmark dose estimation is a quantitative technique for analyzing toxicity data that provides an estimation of the full dose-response curve, and can be used to develop risk estimates that are more appropriate than using simple no effect or low effect levels. Sciences assisted EPA in the development of the benchmark dose approach to dose response analysis by analyzing the performance of EPA’s Benchmark Dose Software (BMDS). This was done by coding independent S-Plus versions of the mathematical models used in BMDS (e.g. the Hill model) and then running hundreds of toxicology data sets taken from animal studies through the newly developed models and through BMDS. An analysis of the effect of data characteristics on the variations between the BMDS and the test models was conducted and made publicly available on EPA’s web site.

Meta-Analysis of Toxicological Data. Many chemicals, particularly those that have been in production for long periods of time, have multiple toxicological studies for the same endpoints. The variance in results from these studies may be the result of random variability in responses. A more accurate estimate of the dose-response can be obtained by considering all of the available data. Meta-analysis can be used to provide the most statistically accurate dose-response from multiple studies. Sciences recently conducted a meta-analysis of toxicity data for the insecticide dimethoate for the producer of the chemical. Sciences analyzed the available data and determined that the most sensitive endpoint for risk assessment was cholinesterase inhibition and that the dose-response was highly dependent on the route of exposure. This analysis was presented to EPA’s FIFRA Scientific Advisory Panel (see SAP results here) and the Panel accepted the conclusion that dimethoate is appropriately regulated by using cholinesterase inhibition as the most sensitive endpoint.