Methods for Deriving Pesticide Aquatic Life Criteria - Reviews of Environmental Contamination and Toxicology - page 50

To ensure consistency in how toxicity data are used to derive criteria, the terms
“acute” and “chronic” must be defined in the methodology. Once defined, the
choice to use either acute or chronic data depends on what kind of criterion is being
calculated and what kinds of data to support it are available. Acute criteria should
be derived from acute data, and chronic criteria should be derived from chronic
data; however, when chronic toxicity data are lacking, acute data may be used to
derive chronic criteria.
As discussed in other parts of this chapter, current criteria derivation methodologies
use toxicity data that have been summarized in the form of an NOEC, LC 50 , EC 50 ,
or some other effect level (i.e., EC 5 , EC 10 , EC x , etc.). Which, then, among these
values is best to use for derivation of protective criteria? We address this question
in the following section. It comprises a discussion of toxicity data analysis methods,
which focus on the problems and challenges associated with using either NOEC or
EC x values for deriving protective criteria.
Ecotoxicity test data are usually analyzed by one of two methods: hypothesis tests
or regression analysis. Hypothesis tests are typically used to analyze results of life
cycle, partial life cycle, and early life-stage tests. In this approach, results from treat-
ment groups are compared with those from control groups to determine significant
differences in responses (Stephan and Rogers 1985). An NOEC or NOEL (no-
observed-effect level), and an LOEC or LOEL may be derived from this type of analy-
sis. Some methodologies use the geometric mean of the NOEC and the LOEC to
calculate an MATC. The other widely used method for analysis of ecotoxicity data
is regression analysis, which is most commonly applied to acute toxicity tests, but can
as easily be applied to chronic tests. In regression analysis, an equation is derived that
describes the relationship between concentrations and effects (Stephan and Rogers
1985). Thus, it is possible to make point estimates of toxicant concentrations that will
cause a given level of effect (EC x ), or to predict effects for a given level of toxicant.
Many problems with the hypothesis test approach are described in the literature.
They are summed up succinctly by Stephan and Rogers (1985) who point out seven
computational and five conceptual problems with hypothesis testing, and then
discuss why regression analysis is a better alternative. The computational points are
briefly described here; for the conceptual points and further details, the reader is
referred to Stephan and Rogers (1985):
1. Hypothesis testing can only provide quantitative information about toxicant
concentrations actually tested. The estimated effect values (i.e., NOEC and
LOEC) must be one of the tested concentrations, with the true NOEC lying
somewhere between the NOEC and the LOEC. For regulatory purposes, such
as deriving water quality criteria, a single number is needed; therefore, regula-
tors choose to use one or the other of the NOEC or LOEC, or they use an
arithmetic or geometric mean of the NOEC and the LOEC. As the authors
point out, hypothesis tests provide no basis for such interpolations. In contrast,
regression analysis determines a relationship between concentration and
effect, and so provides a means to interpolate for estimation of effects at
untested concentrations.