You are here:
ENHANCING TEST SENSITIVITY IN TOXICITY TESTING BY USING A STATISTICAL PERFORMANCE STANDARD
Citation:
Denton, D. L., F. J. Fox, F A. Fulk, AND T J. NorbergKing. ENHANCING TEST SENSITIVITY IN TOXICITY TESTING BY USING A STATISTICAL PERFORMANCE STANDARD. Presented at 22nd Annual SETAC Meeting, Baltimore, MD, November 11-15, 2001.
Description:
Previous reports have shown that within-test sensitivity can vary markedly among laboratories. Experts have advocated an empirical approach to controlling test variability based on the MSD, control means, and other test acceptability criteria. (The MSD represents the smallest difference between the control and treatment means that leads to statistical rejection of the null hypothesis of no toxicity at each concentration of the toxicity test dilution series.) We evaluated one such approach to controlling within-test variability, for commonly used test methods for chronic toxicity. We used a large dataset of reference-toxicant tests performed by laboratories routinely conducting whole effluent toxicity tests. Data underwent rigorous quality-assurance screening and verification of calculated endpoints. We report some key measures of within-test variability, and infer expected effect sizes and power. We recommend use of lower and upper bounds for the percent minimum significant difference (PMSD), defined as MSD divided by control mean, to control test sensitivity and power. We evaluate the increase in test sensitivity that might result from this practice. We conclude that test sensitivity to detect substantial toxic effects (for example, a 25% to 33% decrease in response) is occasionally insufficient at some laboratories may need to increase replication, decrease in response) is occasionally insufficient in some laboratories. To increase sensitivity, some control mean performance. We demonstrate that inadequate test sensitivity is not always signaled by control charts of EC25, LC50, and NOEC. Laboratories should therefore consider maintaining control charts for MSD or PMSD, and should consider reporting the MSD and control mean with each toxicity test. We provide a new control-chart method for PMSD.