Scientific Approach 
Existing toxicity data bases, comparative toxicity literature, and methods for extrapolation among 
endpoints, life stages, and species of aquatic organisms will be reviewed to identify areas where 
improvements can be made in extrapolation based on existing data. 
Specific data gaps, particularly those addressing modes of action/structure-activity, will be 
identified and tests conducted where performance of laboratory toxicity tests would significantly 
enhance the ability to extrapolate among endpoints, life stages, and species. This research can 
best be accomplished by collaboration among the Ecology Divisions, because some of the data 
gaps will involve freshwater testing and others will involve testing of saltwater species. 
Collaboration with other governmental agencies such as the FWS, NOAA, and U.S. Geological 
Survey (USGS) is also desirable. 
Methods for acute tests will be based on standard methods such as Weber (1993) and ASTM 
(1988). Dissolved oxygen, pH, salinity, and temperature will be measured in all treatments on 
day 0 through day 4. Most of the effort with endangered species has been completed and any 
additional testing of such species will be conducted by USGS/BRD, Columbia Environmental 
Research Center, Columbia, MO through an interagency agreement jointly funded by ORD, OW, 
and Office of Prevention, Pesticides, and Toxic Substances (OPPTS). 
Statistical analyses will be performed on survival data with probit analysis or the Spearman- 
Karber program to generate as a minimum, 24-, 48-, 72- and 96-h EC/LCSOs. Three-way 
analysis of variance is performed using SAS to determine statistical differences between species 
LCSOs, among chemicals, and at each time interval. 
Interspecies correlations will be conducted using Model n least squares methodology for two 
independent variables (Mayer and Ellersieck 1986) on a combined data base from Gulf Breeze, 
FL, Columbia, MO (USGS), Office of Pesticide Programs (OPP), and AQUIRE. Slopes and 
intercepts are derived from the equation log y = a + b (log x), where x equals 48-h EC50/96-h 
LC50 values for the surrogate test species and y equals 48-h EC50/96-h LC50 values for the 
endangered or other species. Chemical groupings by mode of action will also be compared to 
taxonomic groupings of species to ascertain any differences in vulnerability. A user manual and 
software, based on a Windows platform, will be developed to use the data base as a reference 
catalog and for user interaction to derive calculated values based on their input data. 
For acute-to-chronic predictive models, both classical and non-classical time-to-event approaches 
(Crane et al. 2001, Jones 1964) will be used. The appropriate computer language(s) to combine 
all three acute-to-chronic models, linear regression, multifactor probit analysis, and accelerated 
life testing (Mayer et al. 1994, Lee et al. 1995, Sun et al. 1995), including life table subroutine, 
will be determined. The accelerated life testing model will be the population-based estimation 
procedure using an underlying life-time distribution/survival probability distribution (e.g., 
exponential, Weibull, extreme value, log-normal, gamma, or log-gamma models). This approach 
will define extra response above natural mortality due to chemical exposure. We will develop 
appropriate computer commands using compatible language that will allow for: 1) data 
input/transfer fi'om pre-established data bases, 2) selective and continuous processing of data 
102 
