Page 789 - The Toxicology of Fishes

Ecological Risk Assessment 769

120
NOEC = 3.6
100 EC 20 = 6.3
Response (% of Control) 60 *
LOEC = 7.2
80

40
*
20
*
0
1 10 100
Exposure Concentration (µg/L)
A
120
100 NOEC = 5.2
Response (% of Control) 60 * EC 20 = 6.3
80

LOEC = 11.5
40

20
*
0 *
1 10 100
Exposure Concentration (µg/L)
B
FIGURE 18.6 Hypothesis testing vs. point estimation. Results of toxicity tests are generally analyzed in one of two ways:
(1) hypothesis testing or (2) point estimation. Hypothesis testing involves statistical testing to determine which treatment
means are significantly different from the control treatment or other pertinent reference. Results of this analysis are typically
given as the no-observable-effect concentration (NOEC), which is the highest concentration not producing a statistically
significant reduction in performance, and the lowest-observable-effect concentration (LOEC), which is the lowest concen-
tration that does cause a statistically significant reduction. Point estimation involves regression analysis of the concentration
response curve. The resulting regression is then used to estimate the concentration associated with a particular level of effect;
for example, the LC 50 is often used to express the results of acute toxicity test; it refers to the concentration estimated to
cause 50% mortality of the test population at the specified time period (e.g., 96-hour LC 50 ). For sublethal effects (e.g., growth,
reproduction), it is common to express point estimates as “effect concentration percent ” values, such as EC 20 (the concentration
estimated to cause a 20% decrease in performance). Although point estimation has been commonly used for acute test results,
hypothesis testing has been widely used historically for chronic test endpoints. There has been a growing movement, however,
toward replacing (or at least supplementing) hypothesis testing with point estimation for both acute and chronic data. The
reasons are several. For one, despite its name, the NOEC is not necessarily a no-effect concentration; it simply means that
no statistically significant reduction was found. The power of a test to detect a reduction varies with the study design and
inherent variability of the measure, which is often high for endpoints such as reproduction. This also means that the degree
of adverse effect at the NOEC or LOEC varies across studies and endpoints, compromising comparability among tests. In
fact, the NOEC and LOEC can vary just as a function of test concentrations. Parts A and B are the theoretical results from
two tests showing the same concentration response curve and having equal variability but with different placement of test
concentrations. Although the NOEC and LOEC vary between the tests, the EC 20 values are identical, as they should be
because the underlying response was identical. Moreover, the actual ecological impact of toxic effects on a species will be
driven by the magnitude of the effect, not by its statistical significance; thus, point estimates provide a better means of
comparing results among tests. In addition, regression analyses can be used to develop confidence intervals for point estimates,
which allows for quantitative evaluation of uncertainty in risk analyses. For more information on point estimation and its
comparison to hypothesis testing, consult Stephan and Rogers (1985), Suter (1996), and Crane and Newman (2000).