Biological test method for determining acute lethality of sediment to amphipods: chapter 6

Section 6: Data Analysis and Interpretation

• 6.1 Data Analysis
• 6.2 Interpretation of Results

6.1 Data Analysis

Investigators should consult Environment Canada (1998a) as well as Section 12 in USEPA (1994a) for detailed guidance regarding the appropriate statistical endpoints and their calculation.

The objective of the data analysis is to quantify contaminant effects on replicate (see Section 4.1) groups of test organisms exposed to various treatments of concern, and to determine if these effects are statistically different from those occurring in a reference or control sediment. Initially, statistical endpoints (i.e., mean ± SD, for percent survival at Day 10; see Section 4.10) are calculated for the replicate samples representing each treatment (including those representing the reference and control treatments). Each study consists of at least a control treatment (i.e., five or more groups of amphipods exposed to control sediment from the site where test organisms were collected) and one or more test treatments. A test treatment might be represented by replicate samples of dredged material from a particular depth or locale (sampling station) of interest, or replicate samples of field-collected sediment from a particular station within or adjacent to an ocean disposal site. Alternatively, a test treatment might be represented by five or more subsamples (i.e., laboratory replicates) of a single (nonreplicated) sample of sediment from a particular sampling station or site-specific depth (see Section 4.1). In each case, each test treatment is normally represented by ≥5 replicates. The same number of replicates per treatment should be used in the test wherever possible, to maximize statistical power and robustness.

Each study with samples of test sediment should, if possible, include one or more reference stations, for which ≥5 replicate samples or subsamples would be included in the toxicity test (Section 4.1). Statistical comparisons of biological data for the replicates representing each test treatment (i.e., potentially contaminated sediment from a single sampling station and depth) with that for replicate samples of reference sediment, should be applied whenever possible or appropriate (EC, 1992; EC, 1998a; 1998b). Such comparisons provide a site-specific basis for evaluating toxicity. Statistical comparisons of biological data for test sediment(s) with that for the control sediment(s) should be made if the samples of reference sediment prove unsuitable for comparison with samples from other sites (e.g., due to their toxicity or physicochemical characteristics that are atypical of test sediments), or if the control sediment(s) used are more appropriate for distinguishing contaminant effects from effects due to confounding factors such as sediment grain size. Regardless of whether or not statistical comparisons are made with reference sediment or control sediment, the experimental results obtained using control sediment must be used as a criterion for judging the validity and acceptability of the test (Section 4.6).

Samples of reference sediment must first meet the species-specific application limits for porewater salinity and grain size (see Section 2.6), if they are to be incorporated in a sediment toxicity test. Given the species-specific criteria for test validity included herein (Section 4.6), which reflect a somewhat differing ability of the four candidate test species to survive a 10-day exposure to uncontaminated sediment, the following recommendations are made for judging whether or not to compare the endpoint results for the test sediments against those for the reference sediment:

• For R. abronius or E. estuarius, the mean 10-day survival rate in the replicate samples of reference sediment must be at least 80% to be eligible and acceptable for comparison with results for test sediments.
• For E. washingtonianus, the mean 10-day survival rate in the replicate samples of reference sediment must be at least 75% to be eligible and acceptable for comparison with results for test sediments.
• For A. virginiana, the mean 10-day survival rate in the replicate samples of reference sediment must be at least 70% to be eligible and acceptable for comparison with results for test sediments.

The recommendations for judging the suitability of comparisons of toxicity data for test sediment(s) with reference sediment use a species-specific mean survival rate 10% lower than the respective species-specific criterion for a valid test which is based on the mean 10-day survival rate in control sediment (see Section 4.6). In each instance, this permits a similar and somewhat (i.e., up to 10%) lower minimum mean 10-day survival rate in reference sediment relative to that in control sediment. This allowance is provided in consideration of the possibility of some reduced survival in reference sediment due to its dissimilar physicochemical characteristics (e.g., grain size), compared to sediment that the test organisms are accustomed to (i.e., control sediment).

The statistical procedures and interpretation of the results should be appropriate to the experimental design and study intent (see USEPA/USACE, 1991; USEPA, 1994a; EC, 1998a; and EC, 1998b for further guidance). Using this reference method, pairwise comparisons of survival data for each test treatment are normally made against survival data derived for a particular reference or control sediment. Initially, all data should be tested for normality using the Shapiro-Wilk’s test, and for homogeneity of variance using Bartlett’s test or other suitable tests (USEPA, 1994a). These and other statistical procedures are included in the methods of “TOXSTAT”; a series of statistical programs on computer disk which can be purchased by contacting WEST, Inc. (2003 Central Avenue, Cheyenne, WY, USA). Instructions for use accompany the TOXSTAT programs on disk.

Survival data which pass the tests for normality and homogeneity of variance should be treated by a pairwise comparison of the results for each test treatment versus the results for the reference or control treatment (see earlier discussion). A one-tailed Student’s t-test should be used for this purpose. If a set of data cannot meet the requirements for normality and homogeneity of variance, an arcsine-square root transformation should be applied, followed by retesting for both (USEPA, 1994a). If the transformed data do not meet the assumption of normality, nonparametric statistics such as the Wilcoxon Rank Sum Test (USEPA, 1994a) or other suitable tests can be applied. If the transformed data meet the assumption of normality, Bartlett’s test or Hartley’s F test should be used to test the homogeneity of variance assumption. Failure of the homogeneity of variance assumption leads to the use of a modified one-tailed Student’s t-test, with adjusted degrees of freedom (USEPA, 1994a). Transformed data which meet the requirements for both normality and homogeneity of variance should be treated by a straightforward pairwise comparison using a one-tailed Student’s t-test.

6.2 Interpretation of Results

Interpretation of results is not necessarily the sole responsibility of the laboratory personnel undertaking the test; this might be a shared task which includes an environmental consultant or other qualified persons responsible for reviewing and interpreting the findings.

Environment Canada (1998b) provides useful advice for interpreting and applying the results of toxicity tests with environmental samples; and should be referred to for guidance in these respects. Initially, the investigator should examine the results and determine if they are valid. In this regard, the species-specific criteria for a valid test (see Section 4.6) must be met.

The findings of the reference toxicity test which was initiated with the same batch of organisms as those used in the sediment toxicity test (see Section 5) should be considered during the interpretive phase of the investigation. These results, when compared with historic test results derived by the testing facility using the same reference toxicant, test organism, and test procedure (i.e., by comparison against the laboratory’s warning chart for this reference toxicity test), will provide insight into the sensitivity of the test organisms as well as the laboratory’s testing precision and performance at the time that the sediment toxicity test was conducted.

All data representing the known physicochemical characteristics of each sample of test material (including that for control and reference sediment) should be reviewed and considered when interpreting the results. The analytical data determined for whole sediment and pore water (see Section 4.3) should be compared with the known tolerance limits and application limits for the species of amphipod used in the test (see Appendices D for R. abronius, E for E. washingtonianus, F for E. estuarius, and G for A. virginiana). Values which approach (but do not exceed) the known tolerance limits (e.g., for ammonia) or application limits (i.e., for sediment grain size or porewater salinity) for each species could reduce their tolerance to contaminants within the sample, and thus have influenced the test results.

Concentrations of porewater ammonia and/or hydrogen sulphide can be elevated in samples of dredged material or field-collected estuarine or marine sediment. The elevated levels might be due to organic enrichment from natural and/or anthropogenic (man-made) sources. The known tolerance limits of the species of test organism used in the study should be considered together with the measured levels of these toxic constituents, when appraising their significance in influencing the test results. Measured concentrations of ammonia should be converted to those for un-ionized ammonia (based on test conditions of pH and temperature), and the concentrations for both total ammonia and un-ionized ammonia considered with respect to reported tolerance limits for this chemical.

All physicochemical data determined for the overlying water during the sediment toxicity test (see Section 4.8) should also be reviewed and considered when interpreting the findings. If, for example, records indicate that the dissolved oxygen concentration in the overlying water within one or more test chambers fell to levels below 60% of saturation, this oxygen depression might have contributed to any toxic responses observed therein (ASTM, 1993; USEPA, 1994a). Similarly, any recorded excursions in water temperature beyond the allowable limits (Section 4.5) should be identified and assessed in conjunction with the test results and their interpretation. As with the porewater ammonia analyses, measurements of ammonia in overlying water at the start and end of the test (Section 4.8) should also be converted to the respective values for un-ionized ammonia (based on the concurrent measurements of pH and temperature for the overlying water). These values should be considered together with the known species’ tolerance to ammonia, when interpreting the test results.

Records of numbers of animals seen swimming in the overlying water, floating on its surface, or emerged from the sediment during the test (Section 4.8) should be reviewed and considered together with those indicating any initial avoidance responses during the first hour of the test (Section 4.7). Any evidence of an avoidance response to one or more samples or treatments should be appraised and interpreted in conjunction with the physicochemical results (whole sediment and pore water; Section 4.3) for the same samples.

The purpose of this sediment toxicity test is to determine whether one or more test sediments are toxic to the test organisms, using the conditions and procedures herein. Various criteria have been used by researchers and regulators to judge if samples of test sediment pass or fail a 10-day toxicity test using marine or estuarine amphipods. For instance, some investigators have ranked test sediment as toxic if 10-day survival was significantly different from that of controls, based on the results of a Student’s t-test (Schlekat et al., 1995). Others have concluded that mean 10-day survival rates in test sediment which are less than 80% (using R. abronius) and statistically different from that in reference sediment should be considered significant (Scott et al., 1990). USEPA (1994b) states that “...dredged material does not meet the benthic toxicity criteria if mortality rates for the dredged material tests exceed that of the reference sediment by more than 20% for amphipods (20% represents the minimum detectable difference of the test method)”. Interim guidelines by Environment Canada researchers and regulators have largely followed USEPA (1994b) when judging if test sediments pass or fail a 10-day test for sediment toxicity using marine or estuarine amphipods, but have also incorporated an alternate criterion based on comparison of results with control sediment in the absence of an acceptable reference sediment (Lee et al., 1995; EC, 1997).

In keeping with Environment Canada (1997), the following two-part guidance is recommended when judging if samples of test sediment pass or fail a 10-day test for sediment toxicity, using this reference method:

• Test sediment from a particular sampling station and depth is judged to have failed this sediment toxicity test if the mean 10-day survival rate for the replicate groups of test organisms exposed to this sediment is more than 20% lower than that in the reference sediment and is significantly different.
• In the absence of an acceptable reference sediment, the test sediment is judged to have failed this sediment toxicity test if the mean 10-day survival rate for the replicate groups of test organisms exposed to this sediment is more than 30% lower than that in the control sediment and is significantly different.