Methods for Deriving Pesticide Aquatic Life Criteria - Reviews of Environmental Contamination and Toxicology - page 53

Participants felt that the effect level should depend on ecological consequences,
but that would require species-dependent values when, politically, a single effect
value for all species is preferable.
Other, novel ways of analyzing toxicity data have been proposed. These include the
use of parametric threshold models to derive a parametric no-effect concentration
(parNEC; Van Der Hoeven et al. 1997; Bedaux and Kooijman 1993; Cox 1987), models
based on dynamic energy budget (DEB) theory (Kooijman 1993; Kooijman et al.
1996; Kooijman and Bedaux 1996a, b; Péry et al. 2002), the use of life table evaluation
techniques (Daniels and Allan 1981; Gentile et al. 1982), case-based reasoning models
(Van Den Brink et al. 2002), and the use of a double bootstrap procedure to estimate
demographic toxicity (e.g., toxicant effect on population growth rate; Grist et al. 2003).
These models are not well developed, and the results they produce have not been
thoroughly compared to existing data analysis methods.
A sound approach, then, may be the one proposed by participants in the 1994
workshop in the Netherlands (Van Der Hoeven et al. 1997). There was overwhelming
support for replacing the NOEC with a more appropriate measure. However, they
recognized the need for a transition period and concluded that NOEC data may be
used as a summary statistic in ecotoxicity testing, if the following are reported: (a)
the minimum significant difference; (b) the actual observed difference from control;
(c) the statistical test used; and (d) the test concentrations. Of the alternative NOEC
replacements considered at the workshop, there was no preference for either the
EC x or parNEC approach, because both have merit, and further research is needed
before a choice can be made. However, according to workshop participants, if the
EC x approach is used, then the x value should be 5% or 10%.
Statistical regression methods are commonly used and widely accepted for
analysis of acute toxicity data. For analysis of chronic data, hypothesis tests have
been more widely used, but have fallen out of favor, primarily because of their
dependence on experimental design and unrestrained Type II error rates. Regression
methods are currently preferred for analysis of chronic data. The problem with
regression methods is that they yield EC 5 , EC 10 , or other EC x values, and science
has not yet decided which of those values best represents a true no-effect level.
Policy decisions are needed to decide what kind of chronic data are acceptable for
use in criteria derivations.
The USEPA methodology (1985) points out that it would be ideal if aquatic no-effect
concentrations could be determined by adding various concentrations of a chemical
of concern to several clean water bodies, and then determine the highest concentra-
tion that causes no effect. Because such an approach is not an option, we must
instead rely on smaller-scale toxicity studies, ranging from single- and multispecies
laboratory tests to multispecies field or semi-field (microcosm or mesocosm) tests.
As models of environmental exposure, the order of preference for obtaining needed
results is field tests, followed by mesocosm/microcosm tests, multispecies laboratory
tests, and single-species laboratory tests. However, the most abundant, reliable, and
easily interpretable toxicity data are from single-species laboratory tests. All of the
other types of studies are criticized for lack of standardization, lack of replication,
and difficulty of interpretation.