Accuracy of chronic aquatic toxicity estimates determined from acute toxicity data and two time-response models.

Traditionally, chronic toxicity in aquatic organisms and wildlife has been determined from either toxicity test data, acute to chronic ratios, or application of safety factors. A more recent alternative approach has been to estimate chronic toxicity by modeling the time course of mortality as determined in standard acute toxicity tests, but these approaches have received limited validation. The accuracy of chronic toxicity estimates from two time-response models, linear regression analysis (LRA) and accelerated life testing (ALT), was investigated using a dataset of more than 150 matched species pairs of standard acute toxicity test data and measured chronic no-observed-effect concentrations (NOECs). Chronic survival was more accurately modeled by both ALT (accuracy, 69%) and LRA (accuracy, 76%) than was reproduction, growth, or the most sensitive endpoint (accuracy, 50-60%). In general, LRA estimates of chronic toxicity were less conservative than ALT estimates, with 66 to 79% of LRA estimates being greater than the measured NOEC. Acute datasets with early mortality produced estimates of chronic survival that were more accurate (ALT, 92%; LRA, 89%) compared to all datasets but were less conservative (84% of ALT estimates were overestimated vs 93% of LRA estimates). Acute datasets with late mortality resulted in poor ALT and LRA estimates of chronic toxicity for all endpoints. Additional survival time measurements did not improve the accuracy of ALT or LRA estimates of chronic toxicity over the standard four acute measurement times (24, 48, 72, and 96 h). The time course of mortality should be considered when applying time-response models to estimate chronic aquatic toxicity, with greater accuracy likely for chronic survival than for growth or reproduction.