Biological test method for toxicity tests using early life stages of rainbow trout: chapter 4

Section 4: Universal Test Procedures

• 4.1 Preparing Test Solutions
• 4.2 Beginning the Test
• 4.3 Test Conditions and Procedures
  • 4.3.1 Test Options
  • 4.3.2 Test Type and Solution Replacement
  • 4.3.3 Temperature
  • 4.3.4 Dissolved Oxygen and Aeration
  • 4.3.5 pH
  • 4.3.6 Life-stage Transition
  • 4.3.7 Fertilization Success and Thinning
  • 4.3.8 Final Phase of Embryo/Alevin/Fry (EAF) Test
  • 4.3.9 Reference Toxicant
• 4.4 Test Observations and Measurements
• 4.5 Test Endpoints and Calculations
  • 4.5.1 Biological Endpoints
  • 4.5.2 Effective and Lethal Concentrations
  • 4.5.3 Inhibiting Concentration for a Specified Percent Effect
  • 4.5.4 No-observed-effect Concentration (NOEC) and Lowest-observed-effect Concentration (LOEC)
  • 4.5.5 Student's t-test
  • 4.5.6 Tukey's Test
• 4.6 Test Validity
• 4.7 Legal Considerations

Procedures described in this Section apply to each of the toxicity tests for samples of chemical, wastewater, or receiving water described in Sections 5, 6, and 7. All aspects of the test system described in Section 3 must be incorporated into these universal procedures. The summary checklist of recommended test conditions and procedures in Table 2 includes not only universal procedures for each species, but also those for specific types of test substances.

4.1 Preparing Test Solutions

All vessels, measurement devices, stirring equipment, and fish-handling equipment must be thoroughly cleaned and rinsed in accordance with standard operating procedures. Control/dilution water should be used as the final rinse water.

For any test that is intended to estimate an EC50, LC50, ICp, or NOEC/LOEC (Section 4.5), a minimum of five concentrations plus a control solution (100% dilution water) must be prepared. For EA and EAF tests, with multiple endpoints based on both lethal and sublethal effects (Section 4.5), more concentrations (e.g., six to eight plus a control) are recommended to improve the likelihood of attaining each endpoint sought. An appropriate geometric series may be used (e.g., 100, 32, 10, 3.2, 1.0; or 100, 46, 22, 10, 4.6, 2.2, 1.0). Concentrations may be selected from other appropriate logarithmic series (see Appendix E). In instances where there is less uncertainty about the range of concentrations likely to be toxic, a geometric series in which each successive concentration is about 50% of the previous one (e.g., 100, 50, 25, 12.5, 6.3) is recommended. There is not usually a great improvement in precision from the use of steps smaller than the 50% dilution factor (i.e., concentrations closer together). If there was considerable uncertainty about the toxic levels, more concentrations should be used to obtain a wide spread, rather than using a lower factor for dilution. Volume requirements for tests will vary according to the option (E, EA, or EAF) used (see Sections 4.3.2, 5.1, 6.1, and 7.1).

Single-concentration tests could be used for regulatory purposes (e.g., pass/fail). They would normally use full-strength effluent, elutriate, leachate, or receiving water, or an arbitrary or prescribed concentration of chemical. Use of controls would follow the same rationale as multi-concentration tests. Single-concentration tests are not specifically described herein, but procedures are evident, and all items apply except for testing a single concentration and a control.

The test must be started with at least three replicates of each concentration including controls. If endpoints are to be calculated using hypothesis tests (i.e., NOEC/LOEC), a minimum of four replicates per concentration must be used.^Footnote 11 The test must start with an equal number of replicates for each concentration, including controls. If there is accidental loss of a replicate during the test, unbalanced sets of results can be analysed with less power (EC, 1998b).

For a given test, the same control/dilution water must be used for preparing the control and all test concentrations. Each test solution must be made up to an identical volume, and well mixed with a clean glass rod, Teflon™ stir bar, or other clean device made of nontoxic material.

The temperature of sample(s) or test solutions (including the control/dilution water) should be adjusted as required for each test option and each life stage (see Section 4.3.3). If necessary, the temperature of samples or test solutions may be adjusted to the test temperature by heating or chilling in a water bath, or by the use of an immersion cooler made of nontoxic material (e.g., stainless steel). Samples or test solutions must not be heated by immersion heaters, since this could alter chemical constituents and toxicity. It might be necessary to adjust the pH of the sample(s) or test solutions (see Section 4.3.5), or to provide preliminary aeration of the solutions (Section 4.3.4).

For site-specific assessments of toxic effect, "upstream" water might be used as control/dilution water. Upstream water cannot be used if it is clearly toxic according to the criteria of the test for which it was intended (see Section 4.6). In such cases, an alternate source of control/dilution water (Section 3.4) must be used.

Table 2 Checklist of Recommended Test Conditions and Procedures

4.2 Beginning the Test

Eggs must be dry-fertilized (see Appendix D for guidance) to prevent the onset of micropyle closure and water hardening before they are transferred to test solutions.^Footnote 12 Uniformity in size of the freshly fertilized eggs is important, as the egg size can affect the alevin and fry size (Beacham et al., 1985). Any eggs distinguished visually as under- or oversized should be discarded. A minimum of 120 embryos per concentration must be used for the E test; 120 to 320 embryos per concentration are recommended for an EA or EAF test (Section 4.3.1). Each treatment (concentration) including the control(s) must include a minimum of four replicate test chambers if statistics using hypothesis tests are intended; and a minimum of three replicates per treatment if point-estimation techniques (e.g., EC50, ICp) are intended (see Sections 4.1 and 4.5).

Identical numbers of embryos should be added to each test chamber. Using 40 embryos per replicate^Footnote 13, a test with three replicates (including five concentrations plus a control) requires 720 eggs. Similarly, a test which uses 80 embryos per replicate, three replicates, and five concentrations plus a control, requires 1440 eggs. The eggs must be obtained from a batch of eggs stripped from four or more females of similar size (see Section 2.2 and Appendix D).

An attempt must be made to achieve "homogeneity of the experimental units" to avoid any differences among vessels that are related to the stripping of gametes. There are two ways to achieve this. They are both valid and are suitable for the same statistical analyses of results (Hubert, 1991). In the first method, embryos from different parents or strippings which have been held separately may be combined (pooled) before exposing embryos to test solutions. In the second method, embryos from a given stripping may be divided evenly among all replicates of all concentrations, then embryos from other strippings are similarly allotted evenly to all incubation units, to make up the full number per replicate. The second method requires more care and effort in culturing and handling. It should, however, reduce the "noise" of the variation between replicates at the same concentration and avoid the chance that exists in the first method, of getting high proportions of unfertilized eggs in a particular replicate, assuming that such stripping-related variation exists.

Fertilization must be accomplished by the dry mixing of eggs and milt for a minimum of 5 minutes (Fennell et al., 1998) and a maximum of 20 minutes (Birge et al., 1985). Following this mixing, groups of freshly fertilized eggs (embryos) should be transferred as quickly as possible to test solutions (see Appendix D for guidance). This should be accomplished within the 10-minute period immediately following the 5- to 20-minute interval for fertilization, and must be completed within the 30-minute period immediately after this interval. A brief (i.e., no more than 10 seconds) rinse of each group of embryos being transferred might be necessary to wash off debris and excess milt. Either control/dilution water or an aliquot of the respective test solution to which the embryos are to be transferred should be used for this purpose. If water is used, embryos must not contact it for more than a few seconds from time of fertilization until their introduction to the test solutions. Eggs that appear abnormal in any way (e.g., opaque or milky-white in colour), or which are noticeably under- or oversized in relation to the other eggs, must not be selected for the test. Any embryos possibly damaged or injured during transfer must be discarded; they can be removed by using egg-picking tweezers or a large-bore pipette (7 to 10 mm) with rubber bulb.

Care must be taken to avoid unnecessary handling of freshly fertilized eggs, or bumping or dropping them as they are transferred into the incubation units. Within the units, embryos need adequate space to ensure sufficient oxygen exchange and removal of metabolic wastes. The embryos must be distributed evenly on the bottom of each unit so that they are only one layer thick and are not clumped together or piled on top of one another. This distribution will also facilitate efficient recognition and counting of nonviable or hatching embryos.^Footnote 14

At the start of the test, the number of embryos transferred to each incubation unit needs to be counted or recounted to ensure that the required number is present and to make any necessary adjustments. During this counting procedure, the incubation unit may be raised gently to just below the surface of the test solution if this is necessary for observation. The number of embryos in the incubation unit should be adjusted as necessary by removing excess embryos using egg-picking tweezers or a large-bore pipette with rubber bulb, and by supplementing any missing embryos with the required number transferred gently and carefully from the remaining group(s) of eggs fertilized for use in the test. The appearance of all embryos in each incubation unit should also be examined at this time, and any embryos appearing atypical in size, shape, or colour should be discarded and replaced.

In addition to these procedures, there must be formal random assignment of the group of embryos in each incubation unit to particular concentrations and replicates. The test concentrations must also be in randomized positions in the test facility. Each test chamber must be clearly coded or labelled to identify the substance and concentration being tested, and the date and time of starting. Temperature, dissolved oxygen, and pH levels in the test chambers should be checked and adjusted, if required/permitted, to acceptable levels (see Sections 4.3.3, 4.3.4, and 4.3.5) before adding test organisms.

It is recommended that the conductivity of each newly prepared test solution be measured before dispensing it to the test vessels, as a pre-use check on concentration. The temperature, dissolved oxygen concentration, and pH of each newly prepared test solution should also be checked as necessary before its use.

Salmonid embryos are extremely sensitive to any disturbance or mechanical shock until they reach the "eyed" stage (see Appendix D). Therefore, throughout an E test and during the "pre-eyed" stage of an EA or EAF test, any routine maintenance procedures (e.g., renewal of test solutions in static-renewal tests) must be performed with extra care. Before embryos reach the eyed stage, any removal of obviously dead (i.e., opaque) embryos or unfertilized eggs to control fungal infection should be done very carefully (without disturbing any of the surviving embryos) using a large-bore pipette (7 to 10 mm) and rubber bulb.

4.3 Test Conditions and Procedures

4.3.1 Test Options

One or more of the following three test options may be used: an embryo (E) test for frequent or periodic monitoring; an embryo/alevin (EA) test for measuring toxic effects on multiple developmental stages; or an embryo/alevin/fry (EAF) test for definitive investigations (see Sections 5 to 7). All three options start with the onset of embryo development, and measure the development and survival of early life stages. The E test must be started with ≥ 120 embryos per concentration (e.g., three replicates of 40 embryos each, per concentration), and normally ends seven days after fertilization. However, the duration of an E test may be extended to as much as ten days after fertilization (Birge, 1996). The longer exposure might be warranted for obtaining clear results, if previous tests showed slow development of embryos. The EA test normally starts with 120 to 320 embryos per concentration^Footnote 15, and is terminated seven days after 50% hatching is seen to be achieved among the surviving embryos of the control, with no feeding of fish. The EAF test also normally starts with 120 to 320 embryos per concentration^Footnote 15, and ends 30 days after 50% of the surviving alevins in the control are seen to have exhibited swim-up behaviour (see Section 4.3.6). The swim-up fry are fed daily during all but the last of the 30 days of the EAF test. Average survival and average dry weight of surviving fry are measured at the end.

Any of these three test options may be used to evaluate samples of chemical, effluent, elutriate, leachate, or receiving water, depending on the objectives of the test. The duration must be ≥7 days for the E test, and is ~30 days for the EA test, and ~70 days for the EAF test, according to the test conditions and procedures herein.

The E test uses only one biological endpoint (nonviability of rainbow trout embryos). This test option is convenient for frequent or periodic monitoring, but in some situations, an initial comparison with or use of the more definitive EA or EAF test is recommended (see Sections 5.1, 6.1, and 7.1). Such a comparison with or use of an EA or EAF test could be appropriate for certain test substances with unusual modes of action, for programs monitoring the environmental effects of particular types of effluents, or for a particular leachate or effluent.

4.3.2 Test Type and Solution Replacement

Tests may be run in either a static-renewal or a flow-through mode. With many types of substances, static tests with 12- or 24-hour renewal of solutions, when done properly, can be as sensitive and as accurate as flow-through tests (Sprague, 1973). For some substances having high chemical or biochemical oxygen demand, volatility, or instability, use of a flow-through test with rapid replacement of test solutions might be necessary.

In static-renewal tests, solutions are changed daily or more frequently, and there are two procedures for doing that:

1. prepare new solutions in clean test chambers, and gently transfer and resuspend the incubation units containing surviving embryos or alevins in the fresh solutions; or
2. retain the organisms in the same exposure chamber while the solutions are almost completely renewed by siphoning 80%, then replacing it to the original volume.

The latter procedure should be used in static-renewal E tests, and during the first two weeks or so in the EA or EAF tests. Old solutions should be siphoned out cautiously and new solution added slowly, because embryos are very sensitive to any disturbance or mechanical shock until they have developed to the eyed stage (see Section 4.2 and Appendix D). Once the embryos have completely developed to the eyed stage, either renewal procedure may be followed.

Flow-through tests require a system that continually delivers a series of pre-mixed concentrations of the wastewater or other test substance to the test chambers, at a controlled rate. Various devices might create successive dilutions of a stock solution or test substance by means of metering pumps or proportional diluters. The flow rates of test solutions, or stock solutions and control/dilution water, should be checked daily throughout the test, and should not vary by more than 10%.

The minimum amount of test solution for each replicate is governed by one of two requirements; calculations must be done for both, and the one requiring the most new test solution must be adopted. The first requirement is based on biomass of organisms in the replicate; the amount of new test solution required each day increases in direct proportion to greater biomass. For this requirement, amounts of test solution needed will be the same in a continuous-flow test as in a static-renewal test. The second requirement is that every 24 hours, most of the old test solution in a container must be replaced with new test solution. The relative volumes required in continuous-flow and static-renewal tests will depend on the standing volume in the test container. As the standing volume becomes greater, the second requirement (for replacing it) tends to dominate, and the first requirement based on biomass tends to become less important. Accordingly, the investigator should not choose the size of container arbitrarily, but should make the calculations and decide on a suitable standing volume in the container, keeping in mind the amount of sample or test chemical that is available.

These two absolute requirements for amount of new test solution are minima. Use of minimum replacement might, in some cases, result in lower measured toxicity than would be found with a more generous supply of test solution. It is also quite possible that other items could come into play and increase the needed amounts of new solution. For example, more test solution might be required if the tests showed signs of oxygen depletion.

The biomass requirement is that there should be at least 0.5 L/g of embryo or alevin, every day (i.e., ≥0.5 L/g·d), and for swim-up fry there must be ≥0.5 L/g·d. This can be estimated for the maximum biomass expected during the test, or adjusted periodically through the longer tests. For instance, in an EA test using rainbow trout, 40 alevins of medium size (say 125 mg, as indicated in Appendix D) would represent 5 g in a replicate vessel. Therefore, at least 2.5 L of new test solution should be provided every day, in either a static-renewal or flow-through test. The flow rate to each replicate should be set to deliver that amount if it is the governing requirement.

The second requirement is to replace at least 80% of the test solution in each container every day. In static-renewal tests, a chamber would normally contain a volume of solution that equalled or exceeded the required daily supply for biomass, of 0.5 L/g·d. Every 24 hours or less, 80% of that standing volume must be renewed, according to the methods previously described. More frequent renewal of static solutions might be necessary, depending on the nature of the substance being tested.^Footnote 16 In a continuous-flow test, to achieve 80% molecular replacement of the old test solution, the daily volume of inflow to the test chamber must equal or exceed 1.6 times the volume of standing liquid in the chamber, assuming that there is complete mixing within the chamber (Sprague, 1973).

Some examples can be given. If the container is set to hold 2 L of test solution in a static-replacement mode, then every day, 80% of that volume would be replaced, i.e., 1.6 L. This is less than the 2.5 L required to satisfy biomass in the previous example, but that would not necessarily always be the case. In a continuous-flow test with 2 L in each container, the daily inflow would have to be 1.6 × 2 L = 3.2 L, more than that for a static-renewal test and more than that in the biomass example. If 1 L within the container were satisfactory for covering the embryos, then the amount could be adjusted and the required daily inflow would only be 1.6 L. A tactic of reducing the standing volume and having a relatively large continuous inflow might be desirable if volatile toxicants were present in the test substance.

4.3.3 Temperature

The rate of early development of rainbow trout and other species of salmonid fish depends intimately on water temperature (Peterson et al., 1977; Gordon et al., 1987; Peterson and Martin-Robichaud, 1989; Beacham and Murray, 1990), and there can be different temperatures for the optimal development and growth of each life stage and/or species. In the E test, the daily mean temperature must be 14 ± 1.0°C for rainbow trout embryos (Fennell et al., 1998); and the instantaneous temperatures of the replicate groups must not vary by more than 3°C at any time. This temperature range, although higher than the optimum for the embryos, is still within the acceptable range for successful development of trout embryos. At this temperature, development of embryos and toxic action will be modestly accelerated, allowing more definitive endpoints to be reached within the short duration of this test (Yee et al., 1996).

Throughout an EA or EAF test, daily mean temperature to which each life stage is exposed (i.e., embryos and alevins in EA test; embryos, alevins, and fry in EAF test) must be 14 ± 1°C (Fennell et al., 1998). Additionally, instantaneous temperatures for the replicate groups must not vary by more than 3°C at any time.

Sample/solution temperature must be adjusted as required to attain an acceptable value for each solution (14 ± 1°C). Samples or test solutions must not be heated by immersion heaters, since this could alter chemical constituents and toxicity. Temperature must be determined by measurements in representative test chambers (i.e., in at least the high, medium, and low concentrations plus control solutions if a multi-concentration test). For a static-renewal test, measurements must be made and recorded at the beginning and end of each 24-h (or earlier, if used) period of exposure, in both the fresh test solution and the used solution just before it is changed. For a flow-through test, measurements must be made and recorded daily. In addition, it is recommended that the temperature of at least one test solution be measured continuously throughout the test.

4.3.4 Dissolved Oxygen and Aeration

The dissolved oxygen content (DO) of the control/dilution water used for preparing test solutions should be 90 to 100% saturation before its use, and, if necessary, the water should be aerated vigorously to achieve this.

Pre-aeration (before exposure of test organisms) or aeration (during exposure) of each test solution might be required or appropriate, depending on the test substance, type, and objectives (see Sections 3.3, 4.3.2, 5.3, 6.3, and 7.3). Apparatus for exposing embryos and alevins to test solutions, with or without aeration, is described in Section 3.3.

If pre-aeration is done (see Sections 5.3, 6.3, and 7.3), each aliquot of sample or solution used for renewal should be pre-aerated^Footnote 17 for 30 minutes at a rate of 6.5 ± 1 mL/mi·L. Immediately thereafter, the dissolved oxygen content of the sample or solutions should be measured. If (and only if) the measured value in one or more solutions is <60% or >100% of air saturation, the pre-aeration of either sample or all test solutions (including the control) should be continued at the same rate (i.e., 6.5 ± 1 mL/mi·L) for an additional period not to exceed 90 minutes. This additional period of pre-aeration must be restricted to the lesser of 90 minutes and attaining 60% saturation in the highest test concentration (or 100% saturation, if supersaturation is evident).^Footnote 18 Immediately thereafter, fish must be exposed to each test solution, regardless of whether 60 to 100% saturation was achieved in the sample or all test solutions. Any pre-aeration must be reported, including the duration and rate (Section 8).

For a static-renewal test (see Section 4.3.2), each test solution including the controls should be aerated continuously throughout the test to ensure an ongoing exchange of solution across the developing embryos or alevins. The rate of aeration of each test solution must be minimal and controlled, to avoid undue stripping of volatile toxicants and/or excessive and uncontrolled detoxification of oxidizable toxic constituents. If a group of test organisms is exposed to a discrete container of solution having a volume ≥6 L, an aeration rate of 6.5 ± 1 mL/mi·L can and should be provided using conventional air-control valves and aeration apparatus (see footnote 7). If the volume of test solution is <6 L, this low rate of aeration cannot be achieved or controlled with conventional air-control valves. Accordingly, such low volumes of test solution should be aerated gently through a narrow-bore (e.g., 0.5 mm ID) aperture at a rate which does not exceed 100 bubbles per minute (EC, 1992b; USEPA, 1994). Section 3.3 describes an appropriate apparatus for aerating a static-renewal setup (see footnote 7 and Figure 3B).

A flow-through test (Section 4.3.2) can be performed with or without aeration of the test solutions, since the continuous flow of fresh solution to each test vessel provides an ongoing exchange of solution across the developing embryos or alevins. Section 3.3 describes and illustrates (Figure 3C) a suitable apparatus for conducting a flow-through test with or without aeration. The nature of the test substance (e.g., volatility, oxygen demand, stability) should be considered when deciding if a flow-through setup is appropriate and whether or not to aerate. Depending on the oxygen demand, gentle aeration of each test solution might be necessary during flow-through tests to maintain dissolved oxygen at adequate levels of 60 to 100% saturation (see Section 6.3). If aeration is used, each replicate solution (including the controls) must be aerated at a similar and controlled rate, as previously described. Alternatively or additionally, more rapid renewal of solutions might be required to maintain DO at 60 to 100% of saturation.

If the objective for certain tests (e.g., for research) is to include an appraisal of the high oxygen demand of the test substance as part of the measurement of its total effect, a flow-through setup would be used (see Sections 3.3 and 4.3.2), and no aeration of test solutions would be provided during the test.

Dissolved oxygen (DO) must be monitored and recorded throughout the test for representative solutions. In static-renewal tests, DO must be measured at the beginning and the end of each renewal interval in at least one replicate of the control(s) and the high, medium, and low concentrations. In flow-through tests, DO must be measured in each replicate at the start of the test, as well as daily thereafter in at least the control(s) and the high, medium, and low concentrations.

Oxygen in the test vessels should not fall below 60% of saturation. If it does, the investigator should be aware that the test is not measuring the toxic quality, per se, of the substance being tested. Rather, such a test would measure the total effect of the substance (e.g., effluent) including its deoxygenating influence.^Footnote 19 Initial measurements will indicate any potential problems with dissolved oxygen, and in such cases, a running check on oxygen concentrations is required. The required use of oxygen-saturated control/dilution water and daily or continuous renewal of solutions will, in most instances, keep dissolved oxygen above the levels that severely stress the developing salmonids and have a major influence on results.

4.3.5 pH

The pH must be measured in the control solutions and those of high, medium, and low concentrations at the beginning of the test, before embryos are added. The pH should also be measured in representative replicates immediately before and immediately after each renewal in static-renewal tests, and daily in flow-through tests.

Toxicity tests should normally be carried out without adjustment of pH. However, if the sample of test substance causes the pH of any solution to be outside the range 6.5 to 8.5, and the toxicity of the test substance rather than the deleterious or modifying effects of pH is being assessed^Footnote 20, the pH of the solutions or sample should be adjusted, or a second, pH-adjusted test should be conducted concurrently. For this second test, the initial pH of the sample, the stock solution (flow-through tests), or of each fresh solution before renewal (static-renewal tests) may, depending on objectives, be neutralized (adjusted to pH 7.0) or adjusted to within ± 0.5 pH units of that of the control/dilution water, before fish exposure. Another acceptable approach for this second test is to adjust the pH upwards to 6.5 to 7.0 (if sample has/causes pH <6.5), or downwards to pH 8.0 to 8.5 (if sample has/causes pH >8.5). Solutions of hydrochloric acid (HCl) or sodium hydroxide (NaOH) at strengths ≤1 N should normally be used for all pH adjustments. Some situations (e.g., effluent samples with highly-buffered pH) might require higher strengths of acid or base.

Abernethy and Westlake (1989) provide useful guidelines for adjusting pH. Aliquots of samples or test solutions receiving pH-adjustment should be allowed to equilibrate after each incremental addition of acid or base. The amount of time required for equilibration will depend on the buffering capacity of the solution/sample. For effluent samples, a period of 30 to 60 minutes is recommended for pH adjustment (Abernethy and Westlake, 1989). Once the test is initiated, the pH of each solution is monitored but not adjusted.

If the purpose of the toxicity test is to gain an understanding of the nature of the toxicants in the test substance, pH adjustment is frequently used as one of a number of techniques (e.g., oxidation, filtration, air stripping, addition of chelating agent) for characterizing and identifying sample toxicity. These "Toxicity Identification Evaluation" (TIE) techniques provide the investigator with useful methods for assessing the physical/chemical nature of the toxicant(s) and their susceptibility to detoxification (USEPA, 1991a; b).

4.3.6 Life-stage Transition

While salmonids go through several developmental phases during their early life stages, there are three major transitions used as benchmarks in the test. The first is the transition from recently fertilized egg to embryo, including the transition from a semipermeable to a relatively impermeable egg membrane (i.e., water hardening) and the initial period of embryo development (i.e., rapid cell division of the developing embryo). The second is the transition from embryo to alevin (i.e., successful hatching), and the third is from alevin to swim-up fry (i.e., yolk utilization to exogenous feeding).

The transition from newly fertilized egg to an embryo in its initial stages of development, before the egg membrane becomes relatively impermeable (until ~2 h post-fertilization), is a critical period when the developing embryo is highly susceptible to direct exposure to toxic solutions.^Footnote 21 Therefore, the start of the test (E, EA, and EAF) has been standardized to ensure that this period occurs during exposure to test solutions. To maximize sensitivity and comparability, the test should start as soon as possible after fertilization has taken place, and must start within the 30-minute period immediately after the complete dry-mixing of eggs and milt, for which a minimum of 5 minutes and a maximum of 20 minutes are allowed (Section 4.2).

For the transition stage from embryo to alevin (EA or EAF test only), the start of the alevin stage is defined as the time when 50% of the initial number of eggs have hatched. The observer is not likely to record a time for exactly 50% hatch; in practice, a time is adopted when the hatch is first seen to include at least one-half of the embryos, and fairly close to 50%. In the EA test, when 50% of the initial number of control eggs are first seen to have hatched, it is considered that the alevin stage has started, and the test ends after a further seven days. At the end, a complete count is made of successful alevins in each replicate, in order to deduce the number and percentage of nonviable alevins (i.e., those which were unfertilized, died as embryos, failed to hatch, or developed abnormally).^Footnote 22

The start of the swim-up fry stage is defined as the time when 50% of the surviving fish exhibit swim-up behaviour.^Footnote 23 In the EAF test, one phase of the test ends and the final phase begins, when 50% of the surviving control fish are seen to exhibit swim-up behaviour. At that time, there must be a count of the total numbers of alevins, deformed alevins, and swim-up fry in each replicate, after which the alevins are discarded.^Footnote 24 Some or all of the surviving swim-up fry in each replicate are released from the incubation unit(s) into the test chamber. The number of fry to be used, and the possibility of thinning, is discussed in Section 4.3.7.^Footnote 25 Feeding of fry is initiated (see Section 4.3.8) and continued for 29 consecutive days. Then fish are not fed for 24 hours, the exposure is ended, and mortalities, abnormalities, and average weight of fish surviving in the replicate are documented (see Section 4.4).

4.3.7 Fertilization Success and Thinning

For any E, EA, or EAF test, an early indication of fertilization success and control viability can be obtained a few days after fertilization by holding additional replicates in control/dilution water under conditions identical to the test treatments, and clearing and examining them microscopically (see footnote 28, Section 4.4) for the incidence of nonviable embryos. If the mean percentage of nonviable control embryos (including unfertilized eggs) is >30% at this time, the investigator must end the test, and restart it using another population of freshly fertilized eggs.

Successful fertilization, survival through hatching, and larval development can vary widely among various batches of gametes. Although it is desirable to have 100% fertilization and 100% control survival, such success is rarely achieved.

Thinning refers to the random removal of a number of individual test organisms from one or more replicates, to reduce crowding, maintain an acceptable loading density, and/or minimize the volumes of test solution required during each renewal (Section 4.3.2). Thinning must not be done at any time during an E or EA test, or during the embryo or alevin stages of an EAF test. It might seem desirable to start with an excess number of eggs, and select equal numbers of viable embryos when it is possible to distinguish them from apparently infertile eggs because of the possibility of poor fertilization success in all concentrations including the control. This must not be done, however, because it could compromise the validity of statistical tests. Exposure to the toxicant before thinning could influence the viability in some concentrations, creating a bias in the choice of organisms. There is only one time in the EAF test when thinning can be done if desired, i.e., when starting the final phase of this test for survival and growth of fry (see Section 4.3.8).

When preparing for an EA or EAF test, preliminary studies are recommended. Such studies should determine the maximum number of embryos that can be placed initially in each incubation unit without causing detrimental effects from crowding (such as insufficient oxygen or accumulation of metabolic waste). By distributing the embryos only one layer thick on the bottom of the incubation unit, efficient recognition and counting of viable versus nonviable embryos (including unfertilized eggs) will also be facilitated. The maximum number for an incubation unit should be determined for the embryo size, flow rate, dimensions of the incubation unit, amount of test solution provided to the test chamber, and the expected size of the alevin or fry at the end of the test. In cases where more than one incubation unit is suspended in a test chamber, embryos or alevins may be moved among the incubation units within the same test chamber to distribute them evenly. However, organisms must not be transferred from one test chamber (i.e., replicate) to another.

In the EAF test, thinning of swim-up fry may be done before starting the final 30-day exposure, i.e., at the time when 50% of the control organisms are seen to exhibit swim-up behaviour. Thinning might be done to maintain the biomass requirement and to minimize the associated sample and solution volume requirements (Section 4.3.2). Thinning might also be done to achieve better balance in numbers, e.g., to attain an identical number of individuals per replicate for the final phase of an EAF test. The extent of thinning may be independent for each replicate, and it is not required that the degree of thinning be balanced among replicates or concentrations. The number of fry within a given replicate must, however, be reduced in a random manner. Thinning cannot be done during the 30-day exposure; that would render the test invalid.^Footnote 26

There are advantages in retaining all the fry, rather than thinning, if facilities and amount of test substance allow. Other factors being equal, larger numbers of test organisms produce narrower confidence limits on endpoints, for example on the LC50 for fry. If thinning is done, ideally it should result in the same large number of fry in each replicate. Preferably, each replicate should retain ≥10 fry, but lower numbers could be used if necessary, as long as they meet the minimum requirements listed in Section 4.3.8.

4.3.8 Final Phase of EAF Test

When 50% of the surviving fish in the control of an EAF test are seen to have attained swim-up status in the EAF test, one phase of the test ends and the final 30 days of exposure commences. A count is made of the numbers of alevins and deformed alevins in each replicate, after which the alevins are terminated. Those data are used for narrative statements on the results of the EAF test (see Section 4.4). The number of individuals in each replicate that are nonviable at swim-up (see Section 4.4) is also determined and recorded at this time, as the first endpoint of the EAF test (Section 4.4).

The swim-up fry existing in all replicates should be released from the incubation unit into their test chamber, and counted. These are the fish that are used for the last phase of the test. There could be thinning of these fry if necessary or desired (Section 4.3.7). Thereafter, these groups of fish are used for subsequent observations of mortality, behaviour, and growth. At least five swim-up fry must be present in a replicate; if not, that replicate is excluded from the final (30-day) phase of the EAF test. At least two replicates must be available for a given concentration; if not, that concentration is excluded from the exposure. There must be at least two replicates of the control, each with ≥5 fry; without that, this final 30-day part of the EAF test cannot be done. Fry must not be transferred among replicates to make up the requirements. The test may proceed with unbalanced numbers of replicates and/or unbalanced numbers of fry per replicate.^Footnote 27

Feeding is initiated in each replicate and continued for 29 days, then fish are not fed during the final day, the 30-day exposure is terminated, and all final observations and measurements are made (Section 4.4).

A commercial starter feed suitable for rainbow trout swim-up fry should be used. The fry should be fed 4% of their body weight per day, with approximately equal portions of this ration offered at least four times per day. Newly hatched brine shrimp may also be used.

The bottom of each test chamber should be siphoned daily to remove any excess food or faeces that have accumulated. For static-renewal tests, this procedure can be combined with the daily siphoning and replacement of each test solution. Care should be taken during siphoning to avoid any injury to the fish. The inlet to the siphon tube should be screened to avoid drawing fish into the tube during this procedure.

4.3.9 Reference Toxicant

The routine use of a reference toxicant or toxicants is practical and necessary to assess, under standardized conditions, the relative sensitivity of the group of embryos that are used, and the precision and reliability of data produced by the laboratory for the reference toxicants (EC, 1990a). Sensitivity of embryos to the recommended reference toxicant(s) must be evaluated at the time that each E, EA, or EAF test is performed, using a portion of the same group of freshly-fertilized eggs used to start that test. The concurrent reference toxicity test undertaken at the start of either of these tests should be an E test because of the long duration of an EA or EAF test.

Criteria used in recommending appropriate reference toxicants for this test could include:

• chemical readily available in pure form;
• stable (long) shelf life of chemical;
• highly soluble in water;
• stable in aqueous solution;
• minimal hazard posed to user;
• easily analyzed with precision;
• good dose-response curve for salmonid embryos;
• knowledge of the degree and type of any influence of pH on toxicity to test organism; and
• knowledge of the degree and type of any influence of water hardness on toxicity to rainbow trout embryos.

Reagent-grade phenol and/or zinc (prepared using zinc sulphate) are recommended for use as the reference toxicant(s) for this test. Sensitivity of rainbow trout embryos to one or both of these reference toxicants should be evaluated using E test(s), and the EC50 determined for one or both of the chemicals (see Section 4.5).

Conditions and procedures for undertaking E tests with reference toxicant(s) are to be consistent and as described elsewhere in this report.

The same procedures and conditions (e.g., static-renewal or flow-through test; same source of control/dilution water) should be used within a testing facility each time that the reference toxicity test is performed. Embryo tests with one or more reference toxicants would normally use the control/dilution water that is used at the laboratory for the definitive E tests. Alternatively, if a greater degree of standardization is desired, soft reconstituted water should be prepared (hardness 40 to 48 mg/L as CaCO₃, pH 7.2 to 7.5; see footnote 37, Section 5.4). This should be used for controls and dilutions (USEPA, 1985b; EC, 1990b).

A warning chart (EC, 1990; 1998b) must be prepared and updated for each reference toxicant used. Successive ICps are plotted on this chart and examined to determine whether the results are within ±2 SD (= warning limits) of values obtained in previous tests using the same reference toxicant and test procedure. The mean and standard deviation of available log EC50s is recalculated with each successive test until the statistic stabilizes (EC, 1990; 1998b). The warning chart should plot logarithm of EC50 on the vertical axis against date of the test (or test number) on the horizontal axis.

The logarithm of concentration (log EC50) should be used in all calculations of mean and standard deviation, and in all plotting procedures. This simply represents continued adherence to the assumption by which each EC50 was estimated on the basis of logarithms of concentrations. The warning chart may be constructed by plotting the logarithms of the mean and its limits on arithmetic paper, or by plotting arithmetic values on the logarithmic scale of semi-log paper. If it were definitely shown that the EC50s failed to fit a log-normal distribution, an arithmetic mean and limits might prove more suitable.

Each new EC50 for the reference toxicant should be compared with the established warning limits of the chart; it is considered acceptable if it falls within the warning limits.

If a particular EC50 falls outside the warning limits, the sensitivity of the embryos and the performance and precision of the test are suspect. Since this might occur 5% of the time due to chance alone, an outlying EC50 does not necessarily mean that the sensitivity or precision are in question. Rather, it provides a warning that this might be the case. A check of all pre-test and test conditions and procedures is required at this time.

Results that remained within the warning limits would not necessarily indicate that a laboratory was generating consistent results. Extremely variable data for a reference toxicant would produce wide warning limits; a new data point could be within the warning limits but still represent undesirable variation. A coefficient of variation of no more than 30% is tentatively suggested as a reasonable limit by Environment Canada (1990).

Stock solutions of phenol should be made up on the day of use. Stock solutions of zinc should either be made up just before their use, in which instance preservation is unnecessary; or acidified with nitric acid to pH <2 if stored (APHA et al., 1995). If stored, acidic zinc solutions should be held in the dark at 4 ± 2°C, and in that state they may be stored for several weeks before use. Zinc sulphate (usually ZnSO₄·7H₂O, molecular weight 4.398 times that of zinc) should be used for preparing stock solutions of zinc. The concentration of zinc should be expressed as mg Zn⁺⁺/L.

Concentrations of reference toxicant in all stock solutions should be measured chemically using appropriate methods (e.g., APHA et al., 1995). Upon preparation of the test solutions, aliquots should be taken from at least the control, low, middle, and high concentrations, and analyzed directly or stored for future analysis if the EC50 was found to be outside warning limits. If stored, sample aliquots must be held in the dark at 4 ± 2°C. Both zinc and phenol solutions should be preserved before storage, following the appropriate guidance given in APHA et al. (1995). Stored aliquots requiring chemical measurement should be analyzed promptly upon completion of the toxicity test. It is desirable to measure concentrations in the same solutions at the end of the test, after completing biological observations. Calculations of EC50 should be based on the geometric mean measured concentrations if they are appreciably (i.e., ≥20%) different from nominal ones and if the accuracy of the chemical analyses is satisfactory.

4.4 Test Observations and Measurements

In all tests, any obviously dead (i.e., opaque) embryos, alevins, or fry should be removed as soon as they are noted, and their numbers recorded. Live individuals must not be removed, whether or not they are deformed. In particular, developing embryos which are not obviously dead but appear atypical, should not be disturbed or removed for microscopic examination until the end of the test (if an E test) or at least until the eyed stage is reached (if an EA or EAF test). When removing dead individuals, extreme care should be taken not to bump or damage adjacent embryos or alevins, since they are extremely delicate and sensitive (see Section 4.2). In particular, extreme care must be taken until the eyed stage, to avoid disturbing the other embryos.

In all tests, daily tabulations should be made of any individuals removed from each replicate. In the longer tests, daily tabulations would also include the number hatched, the number exhibiting swim-up behaviour, the numbers of alevins and fry with deformities, and the number of fry showing abnormal behaviour. Abnormal behaviour includes uncoordinated swimming behaviour, atypical quiescence, atypical feeding behaviour, hyperventilation, and loss of equilibrium.

Routine measurements of test conditions must be carried out as outlined in other sections. Temperature and dissolved oxygen must be measured daily in representative test chambers (Sections 4.3.3 and 4.3.4). Dissolved oxygen must also be measured in each sample and/or test solution before the test, and pre-aeration applied if required or appropriate (Section 4.3.4). The pH must be measured in the control solutions and in representative test chambers at the beginning of the test before embryos are added, and should be measured daily thereafter (Section 4.3.5). For a multi-concentration test, measurements of temperature, dissolved oxygen, and pH must include at least the high, medium, and low concentrations plus the control solution(s). Certain measurements of conductivity are recommended in Section 4.2.

For the E test, observations of the number and percentage of nonviable embryos, including unfertilized eggs, and living but obviously deformed embryos (e.g., those with two heads), must be recorded in each replicate at the end of the test, normally seven days after fertilization. Upon completion of the exposure, the group of embryos and unfertilized eggs remaining in each incubation unit should be transferred together to a pre-labelled vial containing a fixative/clearing solution.^Footnote 28 After clearing, the contents of the vial should be transferred to a shallow container such as a weighing boat, and examined carefully under a dissecting stereo-microscope (Yee et al., 1996).

At the end of the E test, each embryo or unfertilized egg must be scored as viable or nonviable. Viable embryos appear to have developed normally to the stage typical for the controls. Those scored as nonviable would include eggs that were apparently unfertilized, embryos with marked retardation in rate of development, and obviously deformed or otherwise atypical embryos, including twins. Any unfertilized eggs, or embryos which turned opaque and were removed before the end of the test, must be included in the count of nonviable embryos. If the count indicates some missing individuals compared to the starting number, they must also be included in the nonviable category. Observations should start with the control groups, to gain familiarity with the appearance of normal, developing embryos. Yee et al. (1996) provide a series of colour photomicrographs to assist in distinguishing viable from nonviable embryos, and further information is given in Vernier (1969) and Velsen (1980).^Footnote 29

In the EA test, the number and percentage of nonviable alevins in each replicate must be determined and recorded seven days after 50% hatch is seen to be achieved in the control, marking the end of the test. All individuals are classified as viable alevins or nonviable alevins. Nonviability includes failures at any stage: non-fertilization of eggs; mortality as an embryo or alevin; failure to hatch by the end of the test; and obviously deformed or otherwise atypical embryos or alevins (e.g., two-headed individuals). If the count indicates some missing individuals compared to the number which started in the replicate, they must also be included in the nonviable alevin category.

For each replicate, observations are made seven days after 50% hatch is seen to be achieved in the control. There are counts for each replicate, of the number of apparently unfertilized eggs, number of dead embryos, number of live embryos, number of dead alevins, number of “living but deformed or otherwise atypical” alevins, and number of “living and apparently normal” alevins.

Narrative statements must be made for one or both of the following two categories of effect during an EA test, for which there are no formal endpoints. In each case, there must be a brief narrative statement describing apparent differences from the control, or lack of difference. Approximate numerical data on differences should be given in the statement, or in tabular form if appropriate.

• Delayed hatching: useful comparisons with the control could be: (a) approximate times to median hatch in each concentration; and/or (b) approximate percentage judged to have hatched in each concentration at the time of median hatch in the control.
• Deformed alevins: approximate percentage of deformed alevins in each concentration including control. Tally deformities throughout test and sum to obtain the total number of deformed alevins (living plus dead). Express as a percentage of the number that hatched in that concentration.

Other observations could be made and reported, if it were desired to increase the kinds of information provided by the test. These might include the proportions of nonviable embryos, and the mortality of alevins after hatching. When an E test is done in parallel with an EA test, it would provide information for the earlier stages of development.

In the EAF test, the first phase provides an endpoint based on nonviability at time of swim-up, and the second phase provides discrete endpoints on mortality and growth of fry.

For each replicate in an EAF test, the number and percentage of test organisms that are nonviable at swim-up must be determined and recorded when 50% swim-up is seen to be achieved in the control groups. Scoring as nonviable at swim-up includes failure at any stage until early swim-up: non-fertilization of eggs; mortality as an embryo, alevin, or early swim-up fry; failure to hatch; and obviously deformed or otherwise atypical embryos, alevins, or early swim-up fry. If the count indicates some missing individuals compared to the number which started in the replicate, they should also be included in the nonviable at swim-up category.

The observations are made when 50% swim-up is seen to be achieved in the control. There are counts for each replicate, of the number of apparently unfertilized eggs, number of dead embryos, number of live embryos, number of dead alevins, number of “living but deformed or otherwise atypical” alevins, number of “living and apparently normal” alevins, number of dead swim-up fry, number of “living but deformed or otherwise atypical” swim-up fry, and number of “living and apparently normal” swim-up fry. All alevins are then discarded, marking the end of the first phase of the EAF test.

The second phase of the EAF test begins at this time. This phase is a discrete 30-day exposure that includes feeding, and measures mortality and weight of fry. If thinning is to be done on the number of fry, it is performed at this time, before the 30-day exposure (see Section 4.3.7). The fry may either be held in the same open test chambers used for the first phase of the test, or transferred to other (larger) test chambers if necessary to prevent the biomass requirement (i.e., ≥0.5 L/g·d; see Section 4.3.2) from being exceeded as the fish feed and grow throughout this phase of the test.

After 29 consecutive days of feeding, the exposure continues for another day without feeding. The number of fry that died in each replicate during the 30 days is then tabulated. The total dry weight (after 24 hours at 60°C) of the group of surviving fry in each replicate must be recorded to the nearest 0.01 g. Average dry weight of surviving fry is calculated.

Observations during an EAF test must enable narrative statements on the following three categories of effect, which do not have formal endpoints. In each case, there must be a brief narrative statement describing apparent differences from the control, or lack of difference. Approximate numerical data on differences should be given in the statement, or in tabular form if appropriate.

• Deformed alevins: approximate percentage of deformed alevins in each concentration including control, as in the EA test.
• Delayed swim-up: useful comparisons with the control could be: (a) approximate times to median swim-up in each concentration; and/or (b) percentage judged to exhibit swim-up behaviour in each concentration at the time of median swim-up in the control.
• Abnormal behaviour of fry: report degree and type of abnormal behaviour.

Other observations could be made and reported, if it were desired to increase the kinds of information provided by the test. These might include the proportions of nonviable embryos, delayed hatching (as in the EA test), and mortality of alevins after hatching.

4.5 Test Endpoints and Calculations

4.5.1 Biological Endpoints

Biological endpoints to be estimated in this test depend on the option chosen (E, EA, or EAF), as indicated in the following tabulation.

• The E test assesses nonviable embryos, i.e., developmental failure in embryos which occurs sometime during the exposure period which starts immediately after fertilization. One or both of the following two endpoints are estimated for the same effect: (1) effective concentration for 25% nonviable embryos (EC25); and (2) median effective concentration for nonviable embryos (EC50).
• The EA test is based on nonviable alevins, i.e., failure to reach the alevin stage in a timely and normal manner because of deterioration at any previous stage, including failure of egg fertilization, mortality as an embryo or alevin, failure to hatch by the end of the test, and abnormal development. One or both of the following two endpoints are obtained for the same effect: (1) effective concentration for failure of 25% of individuals to develop normally to the alevin stage (EC25); and (2) median effective concentration for failure to develop normally to the alevin stage (EC50).

The EAF test has the following three endpoints. The most sensitive effect (i.e., the endpoint with the lowest concentration) is taken as the definitive indication of toxicity (Woltering, 1984; Birge and Black, 1990).

• Nonviable at swim-up includes failure to survive at this or any previous stage: failure of egg fertilization; mortality of embryos; failure to hatch; mortality of alevins, mortality of early swim-up fry; and obviously deformed or otherwise atypical embryos, alevins, or early swim-up fry. One or both of the following two endpoints are estimated for this effect: (1) effective concentration for 25% failure to develop normally to the early swim-up stage (EC25); and (2) median effective concentration for failure to develop normally to the early swim-up stage (EC50).
• Mortality of fry measures mortality within the 30-day exposure of fry in the final phase of the test. Mortality preceding that time is not included. One endpoint is estimated, the median lethal concentration for fry (LC50).
• Weight of fry measures the average dry weight of surviving fry, compared to the control, at the end of the test, after they have been exposed for 30 days in the final phase of the test. This is essentially a measurement of successful growth, except that no measurements of initial weight are made. One endpoint is estimated, the inhibiting concentration for 25% less dry weight than the controls, among surviving fry at the end of the test (IC25). An extra endpoint could be obtained by estimating the NOEC and LOEC and calculating the TOEC, if desired and if enough replicates were available.

Several narrative reports must be made on additional observations during the EA and EAF tests, as listed in the following text. These are not formal endpoints of the tests and do not require rigorous counting and statistical procedures. Nevertheless, the statements are required as part of the documentation of the test (see Section 8). In each case, there must be a brief narrative statement describing apparent differences from the control, or lack of difference. Approximate numerical data on differences should be given in the statement, or in tabular form if appropriate. Some of these observations might help explain the results for the formal endpoints, previously listed. An apparent difference from the control in any of these items is taken as an indication of toxicity, but cannot be considered definitive in the absence of formal statistical analysis.

• Delayed hatching (EA test). Useful comparisons with the control could be: (a) times to median hatch in each replicate or concentration; and/or (b) percent hatching at the time of median hatch in the control.
• Deformed alevins (EA and EAF tests). This is the number of deformed alevins throughout test (living plus dead individuals), as a percentage of the number that hatched in that concentration.
• Delayed swim-up (EAF test). Useful comparisons with the control could be: (a) times to median swim-up in each replicate or concentration; and/or (b) percent swim-up at the time of median hatch in the control.
• Abnormal behaviour of fry (EAF test). The prevalence and type(s) of abnormal behaviour are to be reported.

Other observations could be made if more information were desired, but these also would not be considered formal endpoints of the test. The other items could include proportions of nonviable embryos (EA and EAF tests), delayed hatching (EAF test), and mortality among alevins after hatching, as discrete observations within the alevin phase of development (EA and EAF tests).

4.5.2 Effective and Lethal Concentrations

For toxic effects using EC50, EC25, or LC50 (Section 4.5.1), the following steps apply when calculating the endpoint.

• Count the affected (or missing) individuals by replicates, but combine the numbers from the replicates at a given concentration.
• The EC50/LC50 cannot be estimated unless one concentration results in an effect ≥50%. The EC25 cannot be estimated unless one concentration results in an effect ≥25%.
• For EC50/EC25, use Abbott's formula on the combined numbers, to allow for a reasonable control effect (see the following text). This applies to endpoints that incorporate success of fertilization.
• Use probit analysis to calculate EC50 and EC25, or LC50, and their 95% confidence limits.
• If results do not satisfy the requirements for probit analysis, use the binomial method to calculate the EC50/LC50, and the range that would encompass the 95% confidence limits. Estimate the EC25 less formally by interpolation or by other acceptable quantal statistics.

Comments on each of those steps follow.

Replicates: the affected (or missing) individuals are counted by replicates, but then the numbers at a given concentration are combined. The procedure uses three replicates for convenience in handling and achieving desired loading during the test, and as insurance in case of accidental loss or other problem in one test chamber. The best use of the ensuing data is to combine the replicates to obtain larger numbers of individuals in a single analysis, which provides narrower confidence limits.

Restrictions on data: it is not valid to estimate the endpoints by extrapolation from low levels of effect. The EC50/LC50 cannot be estimated unless at least one concentration results in an effect ≥50%. Similarly, the EC25 cannot be estimated unless one concentration results in an effect ≥25%.

Abbott's formula (see Finney, 1971, or EC, 1998b): unless indicated otherwise in Environment Canada (1998b), this formula should be used when calculating ECx in the E and EA tests, and for nonviability at swim-up in the EAF test. The formula corrects the effect in each test concentration for the percent effect in the controls, helping to adjust for the variable and gamete-dependent differences in fertilization from test to test (Yee et al., 1996). The formula is applied after the data from replicates have been combined.

A limit of 30% failure of fertilization must be met for validity of each of these tests (Section 4.6). The same value applies to the control results in the E test for nonviable embryos, and Abbott's formula should be used for any effect up to 30% in the control. The longer EA test for nonviable alevins allows up to 35% effect in the controls at the end of the test, before it is considered invalid (Section 4.6), and Abbott's formula should be used to correct for the effect. The still longer EAF test for viability as swim-up fry allows up to 40% effect in the control before it is considered invalid (Section 4.6), and Abbott's formula should be used to correct for the control effect.

Abbott's formula must not be used to correct for control mortality in the second phase of the EAF test with swim-up fry unless advised otherwise in Environment Canada (1998b). If control mortality exceeds 20% during that 30-day phase of the test, the test is invalid (Section 4.6). There would be little advantage in using Abbott's formula for corrections up to 20%, because it would not greatly influence the value calculated for the LC50.^Footnote 30

Probit analysis: the choice of statistical procedures is the same for each analysis to determine ECx or LC50. General instructions on statistical approaches are provided here; further advice is found in Environment Canada (1998b).

Provided that a suitable range of test concentrations was selected, and partial effects occurred at two concentrations, probit analysis can be used. If the effect in at least one concentration does not attain 50% after use of Abbott's correction for control effect, the EC50 or LC50 cannot be estimated. Similarly, EC25 cannot be estimated unless at least one concentration achieved 25% effect. If there is no effect at a certain concentration, that information is used, being an effect of zero percent. However, if successive concentrations yield a series of 0% effects, only one such value should be used in estimating the EC50 or LC50, and that should be the highest concentration of the series, i.e., the zero-effect that is "closest to the middle" of the distribution of data. Similarly, if there were a series of successive complete effects (e.g., 100% unhatched embryos at the high concentrations in the test), only one value of 100% effect would be used, again the one "closest to the middle", i.e., the 100% effect at the lowest of those concentrations. Using additional values of 0% and/or 100% effect would likely distort the estimate of EC50 or LC50.

TOXSTAT^TM (West and Gulley, 1996) or other commercial software packages can be used for standard probit analysis. They estimate EC50/LC50 and 95% confidence limits. The programs will also estimate EC25 and its confidence limits, or any other selected ECx.

A statistical program in BASIC language, adopted from Stephan (1977) and available from Environment Canada (see Appendix B), is simple to use for calculating the EC50 or LC50 with 95% confidence limits by the probit method. The program also estimates EC50/LC50 by the binomial method. The EC50 with 95% confidence limits is also estimated by the method of moving averages, but this has no advantage over probit analysis.

The EC25 should be calculated in addition to the EC50, to provide a somewhat more sensitive endpoint. Some monitoring programs, regulations, or experiments might require calculation of another endpoint such as the EC20. Commercial computer programs including TOXSTAT can be used to calculate the EC25 and its 95% confidence limits. The investigator should be aware, however, that precision decreases progressively when determining such "lesser-effect" values, and confidence limits become correspondingly wider. Estimates below EC20 are not recommended (EC, 1998b).

The binomial method must be used to estimate EC50/LC50 if the data do not provide at least two partial effects (i.e., between 0% and 100% response). The program of Stephan (1977), previously mentioned, is the only computer program known to be available at present, that provides the binomial method. The program accompanies that with conservative (wide) outer limits for the EC50/LC50, within which the true confidence limits would lie. Unfortunately, the Stephan program does not estimate the EC25.

A simple equivalent of the binomial estimate of EC50/LC50 can be done by hand calculation for those cases in which one concentration produces 0% effect and the next higher concentration produces 100% effect. The geometric mean of the two concentrations is a best estimate of the EC50. The two concentrations almost always represent conservative estimates of the confidence limits, but that is not invariably the case and some caution should be expressed in offering them as probable limits. (The usefulness of the Stephan computer program is that it calculates probabilities and selects concentrations that will definitely have a wider span than the true confidence limits.) Hand calculation of the geometric mean can be done as the mean of the logarithms of the concentrations, converted back to an arithmetic value. The geometric mean can also be calculated as the square root of the product of the two concentrations that produce zero and complete effects.

The EC25 can also be calculated by hand, in those cases for which probit analysis is not valid. Calculate it using probits of the observed proportions, to interpolate to the expected probit for 25% effect. Use logarithms of concentration for the calculations. Convert the logarithm obtained for EC25 to an arithmetic value. Alternatively, estimate the EC25 graphically on logarithmic-probability paper. Log-probit paper can be purchased at some university or technical bookstores, or copied from the figure in Environment Canada (1998b). Percentages can be converted to probits from tables in Finney (1971), Newman (1995), or some handbooks of statistics.

No confidence limits on the EC25 are provided by the current binomial/hand calculation methods.

4.5.3 Inhibiting Concentration for a Specified Percent Effect

The ICp, and in particular the IC25 is recommended as a point-estimate of the concentration causing a certain degree of effect on quantitative (graded) biological functions, such as weight of swim-up fry attained in the EAF test (Section 4.5.1). The percentage “p” is selected by the investigator, but is customarily 25% (or 20%) lower performance than in the control (EC, 1998b). IC25 is a formal endpoint which must be calculated in the EAF test, for average dry weight of fry after 30 days of exposure with feeding. The 95% confidence limits must also be calculated and reported for each ICp, to allow statistical comparisons with other such values.

An analysis to determine the IC25 for attained dry weight of fry should begin with a hand plot of percent lower weight compared to the control, against the logarithm of test concentration. The purpose of the hand plot is to check for reasonable results from later mathematical computations. The percent lower weight is calculated for a given test replicate from the average dry weight of fry surviving in that replicate, in relation to the overall average weight attained in the control replicates. The percent "deficit" for each test replicate should be plotted separately. The approximate IC25 should be read from an eye-fitted line. Any major disparity between the approximate graphic IC25 and the subsequent computer-derived IC25 must be resolved. The graph would also show whether a positive and logical relationship was obtained between concentration and effect, a desirable feature of a valid test (EC, 1998b).

At present, the standard computerized method for estimating the ICp with 95% confidence limits is based on smoothing and interpolation, using the program ICPIN (Norberg-King, 1993; USEPA, 1994; EC, 1998b). This modification of BOOTSTRP (Norberg-King, 1988) is included in the latest version of TOXSTAT^TM (West and Gulley, 1996). ICPIN first smooths the data as necessary, then estimates the ICp by simple interpolation, and obtains the confidence limits by a “bootstrap” method of many random resamplings from the actual observations (USEPA, 1994, Appendix M; or EC, 1998b). To use this program, Canadian investigators must either (a) enter concentrations as logarithms, or (b) if a logarithmic transformation is offered in a software package, make sure that it is actually retained for analysis. At time of writing, ICPIN appears to be the only method routinely used for obtaining an ICp with confidence limits, but linear or general-purpose regression would provide better estimates (EC, 1998b).^Footnote 31 Investigators should be alert for improved methods which might become available as computerized packages for environmental toxicology.

Some common-sense limitations should be applied to estimates of the IC25. It should not be derived from an extrapolation. To estimate the IC25, there should be at least one concentration causing more than 25% lower performance than the control, and at least one concentration causing less than 25% lower performance (but still lower than the control, i.e., not 0% effect).^Footnote 32 Variability is greater near the extremes of the relationship, and in particular, observed impairments of 0% and 100% would add little information for an accurate estimation of ICp.

Calculation of the ICp assumes a reduction in performance compared to the control. In some cases there could be a stimulatory effect at low concentrations (e.g., increased growth), but with an inhibitory effect at higher concentrations. Stimulation cannot be assumed to be a strictly positive or beneficial effect, any more than inhibition can always be assumed to represent a strictly negative effect. What is being measured is a difference from the norm (i.e., the control). Current thinking is divided on whether to consider stimulatory effects at low concentrations (hormesis) as a sublethal effect when calculating the ICp, whether to regard it as some kind of parallel "control" performance, or whether to combine it with the control performance (as is automatically done in the smoothing of the ICPIN program. The latter option is not recommended for growth of fry in the EAF test. It is suggested here, that if a stimulatory effect occurs, the test results should be reported in two ways. First, the stimulation should be treated as a deleterious deviation, and a narrative statement should be made on the degree of stimulation and the concentration(s) associated with it. Second, when entering data into the program for calculation of the IC25, the concentrations showing a stimulatory effect should be ignored by not entering them. That way, the control performance will not be changed upwards in the calculations.

4.5.4 NOEC and LOEC

The hypothesis-testing approach can be used, if desired, by estimating the no-observed-effect concentration (NOEC) and lowest-observed-effect concentration (LOEC). They can be derived statistically from the same quantitative (graded) data used for estimating the IC25 for weight of fry (see Section 4.5.3). If NOEC is used, the Minimum Significant Difference must also be calculated and reported (see the following text).

Using NOEC/LOEC as an endpoint has certain limitations. The NOEC is not a "no-effect" concentration, but rather, it is a "no-statistically-significant-difference" concentration. The concentration that became designated as the NOEC might depend largely on sample size, number of replicates, and variability within replicates. A laboratory that had high variation, or that used few replicates, could obtain a higher NOEC than a laboratory with lower variation and more replicates.

NOEC and LOEC could be determined for the average dry weight of surviving individuals in each replicate, following the final 30 days of exposure. If there were complete mortality in a replicate, that replicate would be excluded, leading to an unbalanced analysis. Similarly, if there were complete mortality in all replicates of a given concentration, that concentration would be excluded from the analysis.

The statistical procedures to be followed are given in TOXSTAT^TM.^Footnote 33 The methods start with a check of normality and homogeneity of data, and provide suitable tests of significance for particular types of distribution. TOXSTAT also provides appropriate tests in cases where the numbers of replicates are unequal because of accidental loss or other cause.

If the data are normally distributed or can be made so by suitable transformation, an analysis of variance is carried out. Usually, differences of each concentration from the control will be ascertained by Williams’ test, which is available in TOXSTAT and is designed to be sensitive to the association between the degree of effect and the ordering of concentrations by magnitude. This test (Williams, 1971; 1972) is recommended as a more powerful tool than Dunnett's test, which ignores the ordering of test concentrations by magnitude (Masters et al., 1991). If there are unequal numbers of replicates, the Bonferroni t-test is substituted for Williams’ test. All of these are multiple-comparison tests, which provide estimates of the Minimum Significant Difference, the magnitude of the difference in averages that would have to exist between the control and a test concentration before a significant effect could be concluded for that concentration (discussed in USEPA, 1989; 1994; and EC, 1998b).

If a set of data cannot meet the requirements for normality or homogeneity, and cannot be transformed to do so, there are nonparametric tests provided in TOXSTAT that may be substituted (Steel's many-one rank test, or the Wilcoxon rank sum test in the case of unequal replicates). Those nonparametric options may be used, and are powerful tools for data that are not distributed normally. The nonparametric tests are less powerful than parametric tests, however, when used on normally distributed data, and in that situation they might fail to detect real differences in effect, i.e., an underestimate of sublethal toxicity might result. It should also be remembered that four replicates are required to make use of the nonparametric methods.

A geometric mean of the NOEC and LOEC can be calculated for the convenience of having one number rather than two (the threshold-observed-effect concentration, or TOEC). Such a value may be used and reported, recognizing that it represents an arbitrary estimate of a threshold for a statistically detected effect that might lie anywhere in the range bounded by the LOEC and NOEC. The calculated value of the TOEC is governed by whatever concentrations the investigator happened to select for the test. No confidence limits can be estimated for the TOEC, and that is also the case for NOEC and LOEC, although they indicate the outer limits of the estimate.

The meaning of "threshold" in TOEC is in the dictionary sense, a point at which an effect begins to be observed. Undetected effects might be present at lower concentrations. The geometric mean of NOEC and LOEC is often called the chronic value in the United States, but that term would be somewhat misleading here. The E, EA, and EAF test options herein represent less than 10% of the anticipated life span of rainbow trout, and therefore should not be classified as chronic.

4.5.5 Student's t-test

In a single-concentration test, Student's t-test is normally the appropriate method of comparing data from the test concentration with those of the control. The procedure for a t-test can be taken from any statistics textbook. An effect of the test substance is accepted if the effect measured in a standard endpoint is significantly different than the same statistic for the control (i.e., percent nonviable embryos, nonviable alevins, nonviable individuals at swimup, mortality of fry, and average weight of fry). The test could also be applied to those effects recommended for additional observations and associated narrative statements, if the effect was firmly and numerically documented (i.e., delayed hatching, deformed alevins, mortality of alevins, delayed swim-up, or abnormal behaviour of fry). Requirements for homogeneity of variance and normality must be satisfied (USEPA, 1994, Appendix B; EC, 1998b) before using the standard t-test. If the data do not satisfy the requirements, a nonparametric test could be selected with advice from a statistician; no particular test appears to have become standard practice as yet.

4.5.6 Tukey's Test

In some cases, the treatments in a test might not represent various concentrations of a single sample of wastewater or chemical, but rather a set of different samples, such as full-strength effluents from different industries, or samples of surface waters from different places. It might be desired to test not only whether each sample is different from the control, but also whether the samples are different from each other. That can be done using Tukey's test (one option in the statistical program TOXSTAT^TM; West and Gulley, 1996) . Such sets of tests should report the results of each sample tested, as the percent effect for the endpoint(s) selected, expressed as a percentage of the control(s), and should determine (using Tukey's test) whether that number was significantly different from the corresponding value for the control(s).

4.6 Test Validity

Assuming that all recommended procedures and conditions were followed^Footnote 34, the validity of the test must be based on each of the following: stability of temperature; maintenance of DO levels; the incidence in the control of nonviable embryos (E test), nonviable alevins (EA test), or nonviable individuals at swim-up (EAF test); the incidence of control mortality among fry (EAF test); and variation in control weight (EAF test).

For all tests, a failure rate greater than 30% for fertilization invalidates the test (Yee et al., 1996). Direct measurement of the fertilization rate might not be available since it is not required, but the limit is implicit in the following validity requirements, with some adjustments for the longer tests.

For an E test to be valid, the average percentage of nonviable control embryos must be ≤30%. Unfertilized eggs are included in the count of nonviable control embryos, and in fact, the criterion is the same rate as allowed for fertilization failure.

For an EA test, the average percentage of nonviable alevins in the control at the end of the test must not be greater than 35%.

For an EAF test, the average percentage of nonviable control individuals at the time of 50% swim-up of survivors, must not be greater than 40%.

For a valid EAF test, there is an additional requirement that mortality of control fry during the final 30-day period of exposure must not be >20%.

4.7 Legal Considerations

Care must be taken to ensure that samples collected and tested with a view to prosecution will be admissible in court. For this purpose, legal samples must be: representative of the substance being sampled; uncontaminated by foreign substances; identifiable as to date, time, and location of origin; clearly documented as to the chain of custody; and analyzed as soon as possible after collection. Persons responsible for conducting the test and reporting the findings must maintain continuity of evidence for court proceedings (McCaffrey, 1979), and ensure the integrity of the results.