Research Product

Ecological risks to wildlife are typically assessed using toxicity data for only a few standard test species and with limited understanding of differences in species sensitivity to contaminants. lnterspecies Correlation Estimation (ICE) models were developed as log-linear least square regression models of the relationship between the acute toxicity (LD5O; mg/kg bodyweight) of a range of chemicals tested in two species. A total of 558 models were developed for 49 wildlife species and 951 chemicals. The standard wildlife test species Japanese quail (Coturnix) and mallard, and other avian species (redwinged blackbird, rock dove, ring-necked pheasant) were determined to be good surrogates for many species within the database. For mammals, Norway rat and house mouse provided good surrogacy. Cross-validation of all models predicted toxicity values within 5-fold and 10-fold of the actual values with 85% and 95% certainty, respectively. Taxonomic distance was inversely related to cross-validation prediction success (chi­square = 299.1, d.f. = 12, p <0.0001), with uncertainty increasing with larger taxonomic distance. Models built with two species within the same genus, family, or order predicted within 5-fold of the actual value with 90% confidence. Models were built with seven chemical model of action (MOA)/class or subgroups for species pairs with degrees of freedom > 50 in the original model. Average model error was not reduced by developing ICE models within most MOA/chemical class or subgroups; however improved models for carbamates and direct acting organophosphorouS acetylcholeneSterase inhibiting compounds indicate that toxicity estimates may improve if MONchemiCal älass specific models are built with robust datasets. ICE models can be used when toxicity data for a specific chemical are available for a selected surrogate species, and there is an existing model between the species pair of interest. ICE models for wildlife species provide a tool for estimating contaminant sensitivity with known levels of uncertainty for a diversity of wildlife species. The models described here, in addition to family-level ICE models, will be available in 2007 as a predictive modeling tool through the EPA Center for Exposure Assessment Modeling website (http://www.epa.gov/ceampubl/fchain/index.htm).