You are here:
Incorporating Aquatic Interspecies Toxicity Estimates into Large Databases: Model Evaluations and Data Gains
Citation:
Bejarano, A., Sandy Raimondo, AND M. Barron. Incorporating Aquatic Interspecies Toxicity Estimates into Large Databases: Model Evaluations and Data Gains. SETAC North America 37th Annual Meeting, Orlando, FL, November 06 - 10, 2016.
Impact/Purpose:
This abstract summarizes how the Chemical Aquatic Fate and Effects (CAFÉ) database has been expanded using EPA's Web-ICE tool. The work is important because it will provide hazard chemical responders and managers and expanded chemical toxicity database.
Description:
The Chemical Aquatic Fate and Effects (CAFE) database, developed by NOAA’s Emergency Response Division (ERD), is a centralized data repository that allows for unrestricted access to fate and effects data. While this database was originally designed to help support decisions in the event of spill incidents in aquatic environments, this data repository could help address a diverse range of research and response needs. One of CAFE’s features is that it contains aquatic toxicity data for several thousand chemicals allowing for the visualization of data in the form of probabilistic distributions of toxicity data across taxa or Species Sensitivity Distributions (SSDs). While CAFE currently has enough data to generate over 2,000 chemical-specific SSDs, 70% of compounds have only limited data (<5 species), including hazardous substances that may have a higher risk of spill events. Data gaps may be filled by Interspecies Correlation Estimation (ICE) models, which are log-linear regressions between a surrogate species and a predicted taxa. ICE models were integrated into CAFE, using a multistep process to ensure data quality and identify the best available models. Evaluations of over 3,500 ICE models included the influence on SSD estimates from models that: 1) did not meet reliability requirements (i.e., Mean Square Errors <0.95, slope >0.6, adjusted coefficients of determination [adj-R2] >0.6); and 2) were originally developed based on broad and specific mode of toxic action (MOA), or from combined MOA models. Preliminary evaluations of toxicity data for a variety of chemicals and surrogate species suggest that in most cases (79%) the exclusion of those less reliable models does not lead to substantial changes in SSD estimates, which were generally within a 3-fold of values obtained with models meeting the three reliability requirements. Although MOA appears to have a greater influence on SSD estimates than model reliability, MOA is uncertain for many chemicals and thus limiting their use. The most significant contribution of this work is that the incorporation of ICE models into CAFE would increase the number of SSDs by at least 30%, and would allow for the development of SSDs for at least 67% of 4,800 total chemicals for which aquatic toxicity data are limited. Most data gains are for standard exposure durations (24, 48, 72 and 96 hours), which comprise 85% of data generated using ICE models.