Methods for Deriving Pesticide Aquatic Life Criteria - Reviews of Environmental Contamination and Toxicology - page 52

Hoekstra and Van Ewijk (1993) gave examples of how NOEL values (i.e, no
observable-in-this-particular-test effect level values) are often misinterpreted as
no-effect levels. They cite a study by Murray et al. (1979), in which thymus gland
weight was potentially reduced by as much as 25% at the NOEL, with the uncertainty
resulting from variability in the weight of the exposed thymus glands. A study by
Speijers et al. (1986) resulted in a NOEL that could potentially cause a 73% reduc-
tion in response, compared to control values. Mount et al. (2003), similarly noted
that tests with low variability may produce an LOEC representing responses 2-3%
different from the control, whereas, a test with high variability may produce an
LOEC representing responses 40 + % different from the control. Stephan and Rogers
(1985) found that adverse effects, ranging from 10% to 50% different from
controls, have been reported as “no statistically significant effect concentrations.”
Suter et al. (1987) found effect levels at the MATC, in fish tests, ranging from 12%,
for hatching, to 42% for fecundity. In a more recent short communication, Crane
and Newman (2000) summarized findings of studies showing that the level of effect
corresponding to reported MATCs for fish averaged 28%, with a range of 0.1-84%,
and that power analysis of hypothesis tests for standard Daphnia magna and
Ceriodaphnia dubia tests revealed that these tests are able to detect effects ranging
from 25% to 100%. Clearly, despite its name, the NOEC is not a no-effect level,
and for derivation of protective water quality criteria, it would be unacceptable to
use NOEC data corresponding to such potentially high effects.
Given the apparent agreement among toxicologists that regression analysis
provides better effect level estimates than hypothesis tests (Stephan and Rogers
1985; Bruce and Versteeg 1992; Grothe et al. 1996; Moore and Caux 1997), we
are faced with the problem of having a large, otherwise usable, historical chronic
toxicity data set in which results are reported as NOECs derived from hypothesis
tests. In some cases (i.e., if enough raw data are included in the study report), data
could be reanalyzed to determine point estimates. However, the problem of decid-
ing what effect level best represents a no-effect level, remains. The USEPA
(1991) suggests that an NOEC (for all tests and species) is approximately equiva-
lent to an IC 25 (inhibition concentration; concentration causing 25% inhibition
compared to the control), while Bruce and Versteeg (1992) chose an EC 20 as a
level of population effect that probably would not lead to adverse effects at the
community level. Bruce and Versteeg (1992) also state that the decision as to
what is a safe level should be based on biological criteria established with consid-
eration for the species, the measured endpoint, test design, compound degradability,
and the slope of the concentration-response curve. Results of a 1994 workshop
in the Netherlands indicated a preference among participants (including regulators,
industry, contract laboratories, statisticians, and risk assessors) for use of an EC 5
or EC 10 to represent a no-effect level (Van Der Hoeven et al. 1997). This was
determined via a questionnaire, with responses ranging from EC 1 to EC 25 .
Reasons given for choosing the EC 5 and the EC 10 were admittedly completely
nonscientific: the effect level should be small because an (almost) no-effect level
is intended; the effect level should not be too small because of problems with
accuracy and model dependence; and the effect level should be a round number.