Get access

Revised approach to toxicity test acceptability criteria using a statistical performance assessment

Authors

  • Glen B. Thursby,

    Corresponding author
    1. U.S. Environmental Protection Agency, Atlantic Ecology Division, 27 Tarzwell Drive, Narragansett, Rhode Island 02882
    • U.S. Environmental Protection Agency, Atlantic Ecology Division, 27 Tarzwell Drive, Narragansett, Rhode Island 02882
    Search for more papers by this author
  • James Heltshe,

    1. Department of Computer Science and Statistics, University of Rhode Island, Kingston, Rhode Island 02881, USA
    Search for more papers by this author
  • K. John Scott

    1. Science Applications International Corporation, Environmental Testing Center, 165 Dean Knauss Drive, Narragansett, Rhode Island 02882, USA
    Search for more papers by this author

  • Contribution 1732 of the U.S. Environmental Protection Agency, Narragansett, Rhode Island. Although the research described in this article has been funded in part by the U.S. Environmental Protection Agency (EPA), it has not been subjected to EPA review. Therefore, it does not necessarily reflect the views of the EPA.

Abstract

Current acceptability requirements for toxicity tests are often more restrictive than necessary. They focus primarily on response in a control and generally ignore what a test was “designed” to detect as a significant difference from that control. An approach is presented that takes into account the performance of an entire test and the magnitude of the deviation from the current acceptability requirements. The procedure is based on analyzing the past statistical performance of a test method (i.e., what kind of difference from the control was the test designed to detect). It takes into account traditional control acceptance criteria, but adds a requirement for selecting a difference from the control desired to be detected as statistically significant (a threshold value). Choice of statistical procedure is not relevant to the approach. The proposed method allows a sliding scale of acceptance. The greater the deviation of mean control response below current requirements, the less likely a test is to be accepted. An example is presented using data from a 10-d sediment test using the marine amphipod Ampelisca abdita. Use of the proposed acceptability criterion will reduce the frequency of required retesting without sacrificing defensibility of data. Using the old acceptability criterion, 19% of the samples in the amphipod data set would require retesting. The proposed criterion reduces the potential percentage of retests to 9%.

Get access to the full text of this article

Ancillary