-
PDF
- Split View
-
Views
-
Cite
Cite
Scott E. Belanger, Jane M. Rawlings, Gregory J. Carr, Use of fish embryo toxicity tests for the prediction of acute fish toxicity to chemicals, Environmental Toxicology and Chemistry, Volume 32, Issue 8, 1 August 2013, Pages 1768–1783, https://doi.org/10.1002/etc.2244
- Share Icon Share
Abstract
The fish embryo test (FET) is a potential animal alternative for the acute fish toxicity (AFT) test. A comprehensive validation program assessed 20 different chemicals to understand intra‐ and interlaboratory variability for the FET. The FET had sufficient reproducibility across a range of potencies and modes of action. In the present study, the suitability of the FET as an alternative model is reviewed by relating FET and AFT. In total, 985 FET studies and 1531 AFT studies were summarized. The authors performed FET–AFT regressions to understand potential relationships based on physical–chemical properties, species choices, duration of exposure, chemical classes, chemical functional uses, and modes of action. The FET–AFT relationships are very robust (slopes near 1.0, intercepts near 0) across 9 orders of magnitude in potency. A recommendation for the predictive regression relationship is based on 96‐h FET and AFT data: log FET median lethal concentration (LC50) = (0.989 × log fish LC50) − 0.195; n = 72 chemicals, r = 0.95, p < 0.001, LC50 in mg/L. A similar, not statistically different regression was developed for the entire data set (n = 144 chemicals, unreliable studies deleted). The FET–AFT regressions were robust for major chemical classes with suitably large data sets. Furthermore, regressions were similar to those for large groups of functional chemical categories such as pesticides, surfactants, and industrial organics. Pharmaceutical regressions (n = 8 studies only) were directionally correct. The FET–AFT relationships were not quantitatively different from acute fish–acute fish toxicity relationships with the following species: fathead minnow, rainbow trout, bluegill sunfish, Japanese medaka, and zebrafish. The FET is scientifically supportable as a rational animal alternative model for ecotoxicological testing of acute toxicity of chemicals to fish. Environ Toxicol Chem 2013;32:1768–1783. © 2013 SETAC
INTRODUCTION
The fish embryo test (FET) has emerged in the past decade as a potentially valuable tool in the assessment of the acute toxicity of chemicals and effluents to fish 4. The FET initially was developed as a 48‐h assay involving the exposure of recently fertilized embryos (<1 h) to a series of concentrations of a chemical after which lethality is measured over the course of study 1. Lethality is indicated by coagulation of the embryo, lack of somites, lack of detachment of the tail, or lack of heartbeat, any of which leads to death. The test was refined further and was eventually proposed as a German DIN 5, and an ISO method emerged for testing of wastewater effluents 6. The FET represents an advancement in fish welfare and progress toward the 3 Rs (refinement, reduction, and replacement of animal tests 7). The most commonly used species in this context is the zebrafish, Danio rerio. The FET methodology has initially focused on zebrafish with the expectation that additional fish species, such as fathead minnow (Pimephales promelas) and medaka (Oryzias latipes), will be pursued in the future. Indeed, this is now happening, but the majority of research has utilized zebrafish (P. Amcoff and A. Gourmelon, Organization for Economic Co‐Operation and Development [OECD], Paris, France, personal communication).
Additional refinements of the assay followed 2, and an OECD draft test guideline was proposed 8 that initially described the assay for durations up to 48 h, proposing the extension of exposure to 72 h, the point at which embryos hatch. During this period, discussions occurred on the international stage regarding regulatory and scientific implications for the assay as an animal alternative. Embry et al. 3 summarized an International Life Sciences Institute/Health and Environmental Sciences Institute (ISLI‐HESI) workshop held in 2008 in which the FET was explored as a model useful for the prediction of acute fish toxicity. Halder et al. 9 provided the broad regulatory context of animal alternatives as used in environmental hazard assessment. An important outcome of the international workshop was a clear consideration of the FET assay to be extended for a short period after hatch up to a total duration of 96‐h. This extension includes the embryonic developmental interval known as the eleutheroembryo stage (sensu Balon 10), the period in which the developing embryo receives nutrition solely from the yolk sac prior to exogenous feeding. Clarification through feeding experiments ensued that defined the transition of the eleutheroembryo stage of zebrafish to the larval form as occurring after 96 h 11. The extension of the FET assay through 96 h postfertilization provides a sustained period of exposure of the embryo outside the chorion.
The FET–AFT relationship is quite exceptional; however, some members of the scientific community still wondered whether the database was sufficiently robust. As additional means to provide perspective on this, comparisons were made to acute fish–acute fish toxicity relationships similar to those published in the US Environmental Protection Agency (USEPA) Web‐based interspecies correlation estimation (Web‐ICE) 19. The FET–AFT regressions were quantitatively indistinguishable from acute fish–acute fish toxicity regressions. This is an important facet of the discussion, because OECD and national regulatory agencies accept all 5 species for use in chemical hazard assessments under Mutual Acceptance of Data agreements.
The OECD FET validation program increased the diversity of compounds studied and provided substantial insight into assay variability. Intralaboratory variability (measured as coefficient of variation) was routinely <25%, and interlaboratory variability was <30% for 16 of the 20 compounds. The most toxic compounds had the highest coefficients of variation, but differences were the result of only slight changes in measured toxicity [16; Busquet et al., European Commission, Ispra, Italy, unpublished manuscript].
Validation data, literature data, and information supplied by laboratories actively involved in FET research were used to re‐evaluate the relationship between the FET and AFT. The present study describes the domain of applicability of the FET as a function of physical–chemical properties of compounds tested in the FET, their modes of action, the functional uses of the chemicals, and their chemical classes. The breadth and distribution of test types, species, durations, and other relevant considerations for toxicity testing are also summarized. Finally, FET–AFT and acute fish–acute fish toxicity relationships were developed to assess the potential of the FET to predict AFT.
MATERIALS AND METHODS
Sources of data and information
Fish embryo test data were derived from numerous sources, including: 1) studies identified and previously summarized in which the authors participated 20; 2) studies conducted under the oversight of the OECD Validation Management Group to assess the transferability and intra‐ and interlaboratory variability of the FET 16; 3) information from the peer‐reviewed literature that was identified through hand searches and via literature search engines (Google Scholar, SCOPUS/SciVerse, PubMed); 4) results made available for this summary by laboratories actively engaged in FET method development (T. Braunbeck, University of Heidelberg, Heidelberg, Germany, unpublished data; S. Scholz, Helmholtz Centre for Environmental Research, Leipzig, Germany, unpublished data; K. Schirmer, EAWAG, Dübendorf, Switzerland, unpublished data); and 5) data call‐ins made at SETAC annual meetings (SETAC Europe 21st Annual Meeting, Milan, Italy, 15–19 May, 2011; SETAC North America 32nd Annual Meeting, Boston, MA, USA, 13–17 November, 2011). Targeted emails to distribution lists of the SETAC Global Advisory Committee on Animal Alternatives in Environmental Science and the ILSI‐HESI Project Committee on Animal Alterative Needs in Environmental Risk Assessment (approximately 100 contacts) were sent seeking additional information.
The AFT data were pulled from peer‐reviewed literature, government reports, the USEPA ECOTOX database (http://cfpub.epa.gov/ecotox/), the USEPA Assessment Tools for the Evaluation of Risk (ASTER) database (http://www.epa.gov/med/Prods_Pubs/aster.htm), USEPA Ecological Structure Activity Relationships (ECOSAR) program (ver 1.1), the Pesticide Action Network database (http://www.pesticideinfo.org/Index.html), the OECD eChemPortal (http://www.echemportal.org/echemportal/substancesearch/page.action?pageID=0), and the Umwelt Bundes Amt ETOX (http://webetox.uba.de/webETOX/index.do;jsessionid=98FB01DFE86E14B210BE486882273D84?language=en). On a few occasions, specific data were made available through interested industry participants or from manufacturer's safety data sheets. Use of manufacturer's safety data sheet information, however, was limited to occasions when all other avenues for obtaining fish data were exhausted and the entries were sufficiently detailed with appropriate information.
Information was collected in spreadsheet format (Microsoft Excel 2007). Studies were entered into the database as individual line records such that contributions of independent tests on toxicities of each chemical could be assessed for contributions to variability in regression estimates. To facilitate data collection and maintenance, a data template was distributed to all contacts. Table 1 summarizes the content of the template and information utilized for developing all subsequent analyses.
Information collected from laboratories submitting fish embryo and acute fish toxicity data as well as information summarized from literature
Parameter | Comment |
CAS number | CAS number if known; otherwise sufficient detail is given for the reviewer of the information to determine the appropriate CAS number |
FET species tested | Common and scientific name |
Test (exposure) type | Static, semistatic/renewal, flow‐through |
Dilution water | Well water, reconstituted, surface water, dechlorinated domestic/city water |
Test temperature (mean, min, max) | Mean, maximum, minimum (as °C) |
Hardness (mean, min, max) | Mean, maximum, minimum (CaCO3/L) |
pH (mean, min, max) | Mean, maximum, minimum |
Exposure verification | Indicate whether exposure concentrations were measured and used to calculate the LC50 value |
Analytical samples measured | Indicate if stock solutions were measured |
Multiple values present from the same study | Indicate if LC50 values provided were from 1 test at multiple time points (e.g., 24‐, 48‐, 72‐, and 96‐h LC50 values from 1 test, each provided separately) |
Embryo interval tested | Embryo (0–72‐h exposure for zebrafish) or eleutheroembryo (after 72 h for zebrafish); other species should be identified as to interval |
Duration of exposure | Amount of time embryo or fish was exposed (h) |
Nominal LC value | LC values provided based on the endpoints known to be directly associated with mortality: coagulation, no somites, no tail detached, lack of heartbeat |
Measured LC value | LC values should be based on the following endpoints: coagulation, no somites, no tail detached, lack of heartbeat |
References | Complete a reference as possible including journals, posters, presentations, or unpublished data as appropriate |
Comments | Clarifications of information provided by the submitter if desired |
Parameter | Comment |
CAS number | CAS number if known; otherwise sufficient detail is given for the reviewer of the information to determine the appropriate CAS number |
FET species tested | Common and scientific name |
Test (exposure) type | Static, semistatic/renewal, flow‐through |
Dilution water | Well water, reconstituted, surface water, dechlorinated domestic/city water |
Test temperature (mean, min, max) | Mean, maximum, minimum (as °C) |
Hardness (mean, min, max) | Mean, maximum, minimum (CaCO3/L) |
pH (mean, min, max) | Mean, maximum, minimum |
Exposure verification | Indicate whether exposure concentrations were measured and used to calculate the LC50 value |
Analytical samples measured | Indicate if stock solutions were measured |
Multiple values present from the same study | Indicate if LC50 values provided were from 1 test at multiple time points (e.g., 24‐, 48‐, 72‐, and 96‐h LC50 values from 1 test, each provided separately) |
Embryo interval tested | Embryo (0–72‐h exposure for zebrafish) or eleutheroembryo (after 72 h for zebrafish); other species should be identified as to interval |
Duration of exposure | Amount of time embryo or fish was exposed (h) |
Nominal LC value | LC values provided based on the endpoints known to be directly associated with mortality: coagulation, no somites, no tail detached, lack of heartbeat |
Measured LC value | LC values should be based on the following endpoints: coagulation, no somites, no tail detached, lack of heartbeat |
References | Complete a reference as possible including journals, posters, presentations, or unpublished data as appropriate |
Comments | Clarifications of information provided by the submitter if desired |
CAS = Chemical Abstract Service; FET = fish embryo test; LC = lethal concentration; LC50 = median lethal concentration.
Information collected from laboratories submitting fish embryo and acute fish toxicity data as well as information summarized from literature
Parameter | Comment |
CAS number | CAS number if known; otherwise sufficient detail is given for the reviewer of the information to determine the appropriate CAS number |
FET species tested | Common and scientific name |
Test (exposure) type | Static, semistatic/renewal, flow‐through |
Dilution water | Well water, reconstituted, surface water, dechlorinated domestic/city water |
Test temperature (mean, min, max) | Mean, maximum, minimum (as °C) |
Hardness (mean, min, max) | Mean, maximum, minimum (CaCO3/L) |
pH (mean, min, max) | Mean, maximum, minimum |
Exposure verification | Indicate whether exposure concentrations were measured and used to calculate the LC50 value |
Analytical samples measured | Indicate if stock solutions were measured |
Multiple values present from the same study | Indicate if LC50 values provided were from 1 test at multiple time points (e.g., 24‐, 48‐, 72‐, and 96‐h LC50 values from 1 test, each provided separately) |
Embryo interval tested | Embryo (0–72‐h exposure for zebrafish) or eleutheroembryo (after 72 h for zebrafish); other species should be identified as to interval |
Duration of exposure | Amount of time embryo or fish was exposed (h) |
Nominal LC value | LC values provided based on the endpoints known to be directly associated with mortality: coagulation, no somites, no tail detached, lack of heartbeat |
Measured LC value | LC values should be based on the following endpoints: coagulation, no somites, no tail detached, lack of heartbeat |
References | Complete a reference as possible including journals, posters, presentations, or unpublished data as appropriate |
Comments | Clarifications of information provided by the submitter if desired |
Parameter | Comment |
CAS number | CAS number if known; otherwise sufficient detail is given for the reviewer of the information to determine the appropriate CAS number |
FET species tested | Common and scientific name |
Test (exposure) type | Static, semistatic/renewal, flow‐through |
Dilution water | Well water, reconstituted, surface water, dechlorinated domestic/city water |
Test temperature (mean, min, max) | Mean, maximum, minimum (as °C) |
Hardness (mean, min, max) | Mean, maximum, minimum (CaCO3/L) |
pH (mean, min, max) | Mean, maximum, minimum |
Exposure verification | Indicate whether exposure concentrations were measured and used to calculate the LC50 value |
Analytical samples measured | Indicate if stock solutions were measured |
Multiple values present from the same study | Indicate if LC50 values provided were from 1 test at multiple time points (e.g., 24‐, 48‐, 72‐, and 96‐h LC50 values from 1 test, each provided separately) |
Embryo interval tested | Embryo (0–72‐h exposure for zebrafish) or eleutheroembryo (after 72 h for zebrafish); other species should be identified as to interval |
Duration of exposure | Amount of time embryo or fish was exposed (h) |
Nominal LC value | LC values provided based on the endpoints known to be directly associated with mortality: coagulation, no somites, no tail detached, lack of heartbeat |
Measured LC value | LC values should be based on the following endpoints: coagulation, no somites, no tail detached, lack of heartbeat |
References | Complete a reference as possible including journals, posters, presentations, or unpublished data as appropriate |
Comments | Clarifications of information provided by the submitter if desired |
CAS = Chemical Abstract Service; FET = fish embryo test; LC = lethal concentration; LC50 = median lethal concentration.
In some cases, particularly for metals, aquatic toxicity is known to be influenced in part by water hardness. Because tests vary tremendously with respect to water hardness, all test results were normalized to a set water hardness of 100 mg/L using accepted water hardness regressions for Cd, Cu, Mn, Ni, and Zn 25. The authors of the present study recognize that water‐effect adjustments have become more mechanistic and complex with the advent of biotic ligand models for metal toxicity 26. Use of historical hardness–toxicity regressions published by US EPA represents a compromise, because most historical toxicity studies published in the gray and peer‐reviewed literature lack the specific input needs of biotic ligand models. For compounds other than metals, adjustments to toxicity values as a function of water quality were not performed because information available in publications or reports was generally lacking to perform such adjustments uniformly. This is to say that sources of variability beyond biological sensitivity of the assay are frequently encountered and normally will remain unaccounted for in comparative analyses.
Physical–chemical and quantitative structure–activity relationship information
Verification of test chemical identity for AFT and FET tests was an absolute requirement for inclusion in the database. Chemical names (including common names or trade names) were associated with a Chemical Abstract Service (CAS) number. Those CAS numbers that are particularly broad (for example, those used to describe organic mixtures) were used as infrequently as possible and only if the FET and AFT test data were obviously paired to the CAS number and common name identifier. Consultations with chemistry experts were used to align nomenclature. Molecular weight was determined following this harmonization if not provided directly in reports. Chemicals are known to exist as salts that can dissociate, such as metals and many surfactants. Molecular weight was ultimately expressed in terms of the toxic fraction of the salt.
The ECOSAR (ver 1.1) program was used to generate estimations of the octanol–water partition coeeficient (KOW) and solubility based on simplified molecular input line entry system (SMILES) notations for organic compounds. Categorization of compounds into ECOSAR chemical classes followed rules as used in ECOSAR 27 and was supplemented by ASTER (C. Russom, USEPA, Mid‐Continent Ecology Division, Duluth, Minnesota, personal communication). Functional use (e.g., industrial chemical, biocide, pesticide) designations were assigned to each chemical based on greatest apparent use. Source of information came from expert industry consultation, Internet searches, manufacturer's safety data sheets, and High Production Volume Challenge and European Union Registration, Evaluation, and Authorization of Chemicals (REACH) submission documents.
Data compilation and processing
Physical–chemical, FET and AFT information were compiled in Microsoft Excel for further manipulation, summarization, and development prior to statistical analyses. Statistical approaches and constraints placed on study inclusion described previously 4 were repeated for the present study. In brief, all toxicity data were converted to the same concentration units (µg/L). Results that could not be converted to µg/L because of lack of molecular weight information were excluded. All acute fish toxicity studies shorter than 24 h or longer than 96 h were not used. If 96‐h fish acute toxicity data were available, these were given priority, followed by 72‐h, 48‐h, and 24‐h studies. The FET investigations were almost completely summarized as 48‐, 72‐, or 96‐h toxicity values. Again, preference was given to 96‐h data when available. When LC50s were given as ranges, the geometric mean of the upper and lower bounds of the range was used if the ratio was less than 3. These values were used only if other data were unavailable for the compound (a rare event). Studies were included that employed static, semistatic (also called static renewal), and flow‐through (which collectively includes high‐frequency intermittent addition and continuous addition systems) exposure techniques with and without analytical verification applied. The influence of exposure regimes was also evaluated. Priority for AFT data was given to the previously identified 5 taxa that dominate the literature: zebrafish (D. rerio), fathead minnow (P. promelas), rainbow trout (Oncorhynchus mykiss), bluegill (Lepomis macrochirus), and medaka (O. latipes). Unlike the previous investigation 4, if FET data existed for a chemical that could not be paired with 1 of these 5 species, an alternative OECD test species was not used.
The objective was to maximize the amount of high‐quality studies with paired FET and AFT data. Ideally, the best available data per species would be used. Multiple studies per species per compound were summarized as geometric means.
Statistics
In conventional regression modeling, only the measured response is assumed to be subject to random variation. However, in these data, the measured responses in both the AFT test and the FET are results of experimental testing and are subject to random variation. The statistical treatment of this issue is called “errors‐in variables models” or “measurement error models” 28. Ordinary regression applied to data subject to measurement error underestimates the slope of the regression, proportional to the level of measurement error in the variable used as the predictor. In these data, the variability in the FET LC50, derived from chemicals for which experimental replication is present in the database, is approximately the same as the experiment‐to‐experiment variability in the AFT test LC50. Orthogonal regression 29 was used to fit the linear relationship between the 2 experimental methods, adjusted for measurement error. Orthogonal regression minimizes the sum total of squared residual distances perpendicular to the line itself, versus ordinary regression, which minimizes the sum total of squared residual distances in the vertical dimension only. Although care must be taken with the application of orthogonal regression to an errors‐in variables problem 30, it is a useful tool when it is not otherwise contraindicated. Fish embryo test and AFT data in the present study are generally chosen for the x and y axes, respectively, except for interspecies comparisons of AFT data, in which all permutations of the 5 OECD species are explored. Slope, intercept, 95% confidence intervals, and correlation coefficients were computed using the statistical package R 31. The authors have previously provided detailed explorations of the use of orthogonal regression as applied to these types of data sets 33 to augment the assessments provided in earlier reviews 4.
A variety of comparisons were considered in the orthogonal regression exercises. The FET–AFT relationships were assessed for all data, for subsets of the FET data (for example, 96‐hr only), and for FET results versus individual fish species. In addition, regressions were generated using subsets of various chemical functional classes (industrial organics, surfactants, pharmaceuticals, and pesticides). Influences of solubility and hydrophobicity (log KOW) were also probed. Finally, the relationships between different fish species in the OECD test 203 portion of the database were evaluated as a benchmark against which FET–AFT relationships could be judged. Environmental risk assessors routinely use the 5 common species nearly interchangeably for assessment of AFT as input to risk assessments. These fish–fish relationships have also been intensively explored in the USEPA WeB‐ICE (http://www.epa.gov/ceampubl/fchain/webice/index.html) program. Interspecies correlation estimation is used to predict toxicity of 1 species from a known species toxicity value 34.
Fish size is a potentially important consideration in the determination of AFT and may influence variability in AFT outcomes. To provide perspective on this variable, fish size (wet wt, dry wt, total length, fork length, standard length, age, and maturity) was evaluated for 3 of the most frequently tested compounds: diazinon (a pesticide), phenol (a neutral organic), and Cd (a heavy metal). The influence of size on AFT 96‐h LC50 values was determined for each fish species by regression and correlation analyses. Significant relationships, if they existed, were inferred at α of 0.05. In addition, adherence of fish size employed in each test was assessed in accordance with recommended test guideline recommendations, which differ by species and the test guideline being considered. Analyses were performed in JMP Pro version 10.0.1 35.
General description of the chemical domain covered in the database
Detailed descriptions of chemical coverage, chemical‐by‐chemical, is not possible in the present study due to the breadth of chemicals included. The reader is referred to the Supplemental Data containing FET and acute fish toxicity data. In broad terms, however, the following summary in Data inclusion and exclusion is provided along with comparison to Lammer et al. 4 for historical reference.
Data inclusion and exclusion
The initial data compilation for this FET–AFT comparison included the gathering of data on analytical confirmation of exposures and test chemical solubility. After data compilation, studies were evaluated in the context of their log KOW (for organic compounds) and solubility (all compounds). The ECOSAR program 27 was used to estimate log KOW. If measured values existed in the ECOSAR database, these were used. A similar exercise was conducted with respect to solubility. Compounds with a very high log KOW (e.g., >5) were critically assessed because these have very low estimated solubility (<0.5 mg/L in ion‐free water). If tests were well above the limits of solubility, allowing for some experimental error, the validity of the resulting toxicity estimate was questioned and the results were censored from this analysis.
Figure 1 shows the breadth and distribution of tests, conducted in the FET (all chemicals) and in the acute fish toxicity test for which FET data were available. The range of log KOWs was from −4.15 to 7.85 and solubility from 0.8 µg/L to >1 g/L.

Distribution of the molecular weight (A), octanol–water partition coefficient (KOW; B), and solubility (C) for chemicals assessed in the fish embryo test (FET; all compounds) and acute fish toxicity test (for compounds for which FET data also existed). All species tested in the FET and all species tested in the acute fish toxicity test are included.
Chemicals tested in the FET and AFT test spanned 15 different categories of functional use (Table 2). The single largest category was industrial organic compounds, and pesticides, pharmaceuticals, surfactants, and biocides were also well represented. As can be seen in Table 2, by 2012 (the present study) the number of chemicals tested in the FET had increased by 38% and compounds with both FET and AFT test data had doubled.
Categories of functional use for chemicals tested in the fish embryo test and acute fish toxicity test
In Lammer et al. 4 | In the present study | |||
FET | OECD 203 | FET | OECD 203 | |
Biocide | 5 | 2 | 10 | 5 |
Flame retardant | 0 | 0 | 1 | 1 |
Food additive/vitamin | 3 | 1 | 4 | 2 |
Hair dye | 1 | 1 | 1 | 1 |
Industrial organic | 94 | 50 | 125 | 77 |
Inorganic | 1 | 1 | 2 | 2 |
Metal | 4 | 4 | 7 | 7 |
Natural/botanical | 1 | 0 | 4 | 1 |
Organometal | 1 | 1 | 1 | 1 |
Perfume | 0 | 0 | 1 | 1 |
Pesticide | 15 | 11 | 26 | 23 |
Petrochemical | 0 | 0 | 1 | 1 |
Pharmaceutical | 10 | 2 | 22 | 8 |
Polymer | 3 | 0 | 5 | 2 |
Surfactant | 4 | 4 | 19 | 19 |
Total | 142 | 77 | 229 | 151 |
In Lammer et al. 4 | In the present study | |||
FET | OECD 203 | FET | OECD 203 | |
Biocide | 5 | 2 | 10 | 5 |
Flame retardant | 0 | 0 | 1 | 1 |
Food additive/vitamin | 3 | 1 | 4 | 2 |
Hair dye | 1 | 1 | 1 | 1 |
Industrial organic | 94 | 50 | 125 | 77 |
Inorganic | 1 | 1 | 2 | 2 |
Metal | 4 | 4 | 7 | 7 |
Natural/botanical | 1 | 0 | 4 | 1 |
Organometal | 1 | 1 | 1 | 1 |
Perfume | 0 | 0 | 1 | 1 |
Pesticide | 15 | 11 | 26 | 23 |
Petrochemical | 0 | 0 | 1 | 1 |
Pharmaceutical | 10 | 2 | 22 | 8 |
Polymer | 3 | 0 | 5 | 2 |
Surfactant | 4 | 4 | 19 | 19 |
Total | 142 | 77 | 229 | 151 |
FET = fish embryo test; OECD 203 = Organisation for Economic Co‐Operation and Development, test 203.
Categories of functional use for chemicals tested in the fish embryo test and acute fish toxicity test
In Lammer et al. 4 | In the present study | |||
FET | OECD 203 | FET | OECD 203 | |
Biocide | 5 | 2 | 10 | 5 |
Flame retardant | 0 | 0 | 1 | 1 |
Food additive/vitamin | 3 | 1 | 4 | 2 |
Hair dye | 1 | 1 | 1 | 1 |
Industrial organic | 94 | 50 | 125 | 77 |
Inorganic | 1 | 1 | 2 | 2 |
Metal | 4 | 4 | 7 | 7 |
Natural/botanical | 1 | 0 | 4 | 1 |
Organometal | 1 | 1 | 1 | 1 |
Perfume | 0 | 0 | 1 | 1 |
Pesticide | 15 | 11 | 26 | 23 |
Petrochemical | 0 | 0 | 1 | 1 |
Pharmaceutical | 10 | 2 | 22 | 8 |
Polymer | 3 | 0 | 5 | 2 |
Surfactant | 4 | 4 | 19 | 19 |
Total | 142 | 77 | 229 | 151 |
In Lammer et al. 4 | In the present study | |||
FET | OECD 203 | FET | OECD 203 | |
Biocide | 5 | 2 | 10 | 5 |
Flame retardant | 0 | 0 | 1 | 1 |
Food additive/vitamin | 3 | 1 | 4 | 2 |
Hair dye | 1 | 1 | 1 | 1 |
Industrial organic | 94 | 50 | 125 | 77 |
Inorganic | 1 | 1 | 2 | 2 |
Metal | 4 | 4 | 7 | 7 |
Natural/botanical | 1 | 0 | 4 | 1 |
Organometal | 1 | 1 | 1 | 1 |
Perfume | 0 | 0 | 1 | 1 |
Pesticide | 15 | 11 | 26 | 23 |
Petrochemical | 0 | 0 | 1 | 1 |
Pharmaceutical | 10 | 2 | 22 | 8 |
Polymer | 3 | 0 | 5 | 2 |
Surfactant | 4 | 4 | 19 | 19 |
Total | 142 | 77 | 229 | 151 |
FET = fish embryo test; OECD 203 = Organisation for Economic Co‐Operation and Development, test 203.
The ECOSAR program was used to classify the 229 chemicals tested in the FET. Among the 229 chemicals, 207 could be classified. The remaining chemicals primarily were metals, metalloids, organometals, and complex mixtures. In total, 38 different chemical classes were identified by ECOSAR based on functional groups and structure (Table 3). Aliphatic amines, neutral organics, and phenols were the most common ECOSAR groups. Table 3 confirms that a very wide range of compounds was tested, providing the opportunity to address a broad domain of chemical applicability.
ECOSAR chemical categories for chemicals tested in the fish embryo test in this review
ECOSAR grouping | n |
Acrylates | 1 |
Aldehydes (mono) | 3 |
Aliphatic amine | 47 |
Amide | 5 |
Amide (acid) | 1 |
Analine | 13 |
Anionic surfactant | 12 |
Carbamate ester | 1 |
Carbamate ester, phenyl | 1 |
Cationic surfactants | 4 |
Epoxide, mono | 1 |
Ester | 3 |
Ester acid | 1 |
Esters (dithiophosphates) | 2 |
Esters (monothiphosphates) | 3 |
Halo alcohols | 1 |
Halopyridine | 2 |
Hydrazine | 4 |
Imidazole | 2 |
Imide | 1 |
Neutral organic | 42 |
Neutral organic (acid) | 12 |
Nonionic surfactant | 3 |
Phenol | 21 |
Phenols (acid) | 1 |
Polyphenol | 1 |
Pyrazoles/pyrroles | 3 |
Pyrethroid | 2 |
Quinone | 2 |
Quinone/hydroquinone | 1 |
Substituted urea | 3 |
Triazine | 1 |
Thiocarbamate, di | 1 |
Vinyl/allyl alcohol | 2 |
Vinyl allyl aldehyde | 1 |
Vinyl/allyl ether | 1 |
Vinyl/allyl halide | 1 |
Vinyl/allyl ketone | 1 |
ECOSAR grouping | n |
Acrylates | 1 |
Aldehydes (mono) | 3 |
Aliphatic amine | 47 |
Amide | 5 |
Amide (acid) | 1 |
Analine | 13 |
Anionic surfactant | 12 |
Carbamate ester | 1 |
Carbamate ester, phenyl | 1 |
Cationic surfactants | 4 |
Epoxide, mono | 1 |
Ester | 3 |
Ester acid | 1 |
Esters (dithiophosphates) | 2 |
Esters (monothiphosphates) | 3 |
Halo alcohols | 1 |
Halopyridine | 2 |
Hydrazine | 4 |
Imidazole | 2 |
Imide | 1 |
Neutral organic | 42 |
Neutral organic (acid) | 12 |
Nonionic surfactant | 3 |
Phenol | 21 |
Phenols (acid) | 1 |
Polyphenol | 1 |
Pyrazoles/pyrroles | 3 |
Pyrethroid | 2 |
Quinone | 2 |
Quinone/hydroquinone | 1 |
Substituted urea | 3 |
Triazine | 1 |
Thiocarbamate, di | 1 |
Vinyl/allyl alcohol | 2 |
Vinyl allyl aldehyde | 1 |
Vinyl/allyl ether | 1 |
Vinyl/allyl halide | 1 |
Vinyl/allyl ketone | 1 |
ECOSAR = Ecological Structure Activity Relationships program.
ECOSAR chemical categories for chemicals tested in the fish embryo test in this review
ECOSAR grouping | n |
Acrylates | 1 |
Aldehydes (mono) | 3 |
Aliphatic amine | 47 |
Amide | 5 |
Amide (acid) | 1 |
Analine | 13 |
Anionic surfactant | 12 |
Carbamate ester | 1 |
Carbamate ester, phenyl | 1 |
Cationic surfactants | 4 |
Epoxide, mono | 1 |
Ester | 3 |
Ester acid | 1 |
Esters (dithiophosphates) | 2 |
Esters (monothiphosphates) | 3 |
Halo alcohols | 1 |
Halopyridine | 2 |
Hydrazine | 4 |
Imidazole | 2 |
Imide | 1 |
Neutral organic | 42 |
Neutral organic (acid) | 12 |
Nonionic surfactant | 3 |
Phenol | 21 |
Phenols (acid) | 1 |
Polyphenol | 1 |
Pyrazoles/pyrroles | 3 |
Pyrethroid | 2 |
Quinone | 2 |
Quinone/hydroquinone | 1 |
Substituted urea | 3 |
Triazine | 1 |
Thiocarbamate, di | 1 |
Vinyl/allyl alcohol | 2 |
Vinyl allyl aldehyde | 1 |
Vinyl/allyl ether | 1 |
Vinyl/allyl halide | 1 |
Vinyl/allyl ketone | 1 |
ECOSAR grouping | n |
Acrylates | 1 |
Aldehydes (mono) | 3 |
Aliphatic amine | 47 |
Amide | 5 |
Amide (acid) | 1 |
Analine | 13 |
Anionic surfactant | 12 |
Carbamate ester | 1 |
Carbamate ester, phenyl | 1 |
Cationic surfactants | 4 |
Epoxide, mono | 1 |
Ester | 3 |
Ester acid | 1 |
Esters (dithiophosphates) | 2 |
Esters (monothiphosphates) | 3 |
Halo alcohols | 1 |
Halopyridine | 2 |
Hydrazine | 4 |
Imidazole | 2 |
Imide | 1 |
Neutral organic | 42 |
Neutral organic (acid) | 12 |
Nonionic surfactant | 3 |
Phenol | 21 |
Phenols (acid) | 1 |
Polyphenol | 1 |
Pyrazoles/pyrroles | 3 |
Pyrethroid | 2 |
Quinone | 2 |
Quinone/hydroquinone | 1 |
Substituted urea | 3 |
Triazine | 1 |
Thiocarbamate, di | 1 |
Vinyl/allyl alcohol | 2 |
Vinyl allyl aldehyde | 1 |
Vinyl/allyl ether | 1 |
Vinyl/allyl halide | 1 |
Vinyl/allyl ketone | 1 |
ECOSAR = Ecological Structure Activity Relationships program.
To identify possible modes of action for organic chemicals in the database, ASTER was used through a request to the USEPA National Health and Environmental Effects Research Laboratory (NHEERL), Mid‐Continent Ecology Division (MED), Office of Solid Waste and Emergency Response, and Office of Research and Development, National Center for Computational Toxicology 36. To this end, the full database of compounds, CAS numbers, and SMILES notations was forwarded in an Excel file to the USEPA. Excel output files with quantitative structure–activity relationship and toxicity data extractions, as well as confirmation of previous chemical classification by ECOSAR and ASTER mode of action, were developed and returned (C. Russom, USEPA, Mid‐Continent Ecology Division, Duluth, Minnesota, personal communication, personal communication). Modes of action for 192 of the 229 compounds in the database were assigned, the rest being compounds that were inorganic, metals, metalloids, or polymers or for which the program could not converge. Surfactants that were unclassified account for 11 of the 37 compounds that were not classified. These all would be classified as nonpolar narcotics based on their structures, and thus only 26 chemicals were ultimately not classified. In total, 17 modes of action were identified from chemicals in the database and are provided in Table 4.
Modes of action identified using the US Environmental Protection Agency Assessment Tools for the Evaluation of Risk (ASTER) mode of action categorization program
Mode of action | No. of compounds classified |
Acrylate toxicity | 1 |
Alkylation/arylation based reactivity | 6 |
Carbamate mediated acetylcholinesterase inhibition | 2 |
Carbonyl reactivity | 3 |
Diester toxicity | 2 |
Ester narcosis | 3 |
Hydrazine‐based reactivity | 4 |
Neurotoxicant: cyclodiene‐type | 4 |
Nonpolar narcosis | 118 |
Nonpolar narcosis (unclassified surfactants) | 11 |
Organophosphate‐mediated acetylcholinesterase inhibition | 6 |
Polar narcosis | 33 |
Pyridinium compounds | 2 |
Quinoline reactivity | 1 |
Respiratory blocker | 1 |
Reactive | 2 |
Uncoupler of oxidative phosphorylation | 4 |
Total number of compounds | 203 |
Mode of action | No. of compounds classified |
Acrylate toxicity | 1 |
Alkylation/arylation based reactivity | 6 |
Carbamate mediated acetylcholinesterase inhibition | 2 |
Carbonyl reactivity | 3 |
Diester toxicity | 2 |
Ester narcosis | 3 |
Hydrazine‐based reactivity | 4 |
Neurotoxicant: cyclodiene‐type | 4 |
Nonpolar narcosis | 118 |
Nonpolar narcosis (unclassified surfactants) | 11 |
Organophosphate‐mediated acetylcholinesterase inhibition | 6 |
Polar narcosis | 33 |
Pyridinium compounds | 2 |
Quinoline reactivity | 1 |
Respiratory blocker | 1 |
Reactive | 2 |
Uncoupler of oxidative phosphorylation | 4 |
Total number of compounds | 203 |
Modes of action identified using the US Environmental Protection Agency Assessment Tools for the Evaluation of Risk (ASTER) mode of action categorization program
Mode of action | No. of compounds classified |
Acrylate toxicity | 1 |
Alkylation/arylation based reactivity | 6 |
Carbamate mediated acetylcholinesterase inhibition | 2 |
Carbonyl reactivity | 3 |
Diester toxicity | 2 |
Ester narcosis | 3 |
Hydrazine‐based reactivity | 4 |
Neurotoxicant: cyclodiene‐type | 4 |
Nonpolar narcosis | 118 |
Nonpolar narcosis (unclassified surfactants) | 11 |
Organophosphate‐mediated acetylcholinesterase inhibition | 6 |
Polar narcosis | 33 |
Pyridinium compounds | 2 |
Quinoline reactivity | 1 |
Respiratory blocker | 1 |
Reactive | 2 |
Uncoupler of oxidative phosphorylation | 4 |
Total number of compounds | 203 |
Mode of action | No. of compounds classified |
Acrylate toxicity | 1 |
Alkylation/arylation based reactivity | 6 |
Carbamate mediated acetylcholinesterase inhibition | 2 |
Carbonyl reactivity | 3 |
Diester toxicity | 2 |
Ester narcosis | 3 |
Hydrazine‐based reactivity | 4 |
Neurotoxicant: cyclodiene‐type | 4 |
Nonpolar narcosis | 118 |
Nonpolar narcosis (unclassified surfactants) | 11 |
Organophosphate‐mediated acetylcholinesterase inhibition | 6 |
Polar narcosis | 33 |
Pyridinium compounds | 2 |
Quinoline reactivity | 1 |
Respiratory blocker | 1 |
Reactive | 2 |
Uncoupler of oxidative phosphorylation | 4 |
Total number of compounds | 203 |
Several compounds were found to be challenging and often chosen to stretch the overall domain of applicability by researchers. Though laudable and necessary, this also resulted in a fair number of studies with questionable outcomes. To better understand the influence of such factors, the ratio of the FET or AFT LC50 values to the solubility (in pure water) was determined for the mean FET and AFT value per compound (Figure 2). Particular attention was paid to compounds with toxicity:solubility ratios above 10, which make it likely that key exposure concentrations were not completely soluble and that physical effects, in addition to toxicity, were being measured. It is important to note that testing above the limit of solubility is not confined to only high KOW compounds, although in practice, this is the key consideration. Compounds with polar or nonpolar narcotic modes of action have moderate potency, and apparently ecotoxicologists often test such compounds above their limits of solubility even at moderate KOWs. Table S1 in the Supplemental Data identifies the suite of challenging compounds or those that elicited additional evaluation. All‐trans retinol, retinoic acid, resmethrin, and tetrabromobisphenol‐A all have KOWs above 7.1, and their solubilities are 10 µg/L or less. Toxicity to solubility ratios were approximately 100 to 4000 for this group. A more environmentally reasonable approach to assess compounds such as these would be to investigate chronic, long‐term toxicity because acute toxicity is less likely to be relevant. This group of compounds was deleted from further consideration. The FET:AFT ratios for this group ranged from 0.04 to 20 000, further supporting the contention that results were questionable. Alkyl ether sulfate, dialkyl sulfosuccinate, and fatty alkyl ester sulfonate are all surfactants with moderate KOWs, and low solubilities in the range of 1.2 µg/L to 300 µg/L. The potencies of the compounds are moderate, resulting in toxicity:solubility ratios of 10 to 8000. As with the high KOW compounds, results from these toxicity tests are also questionable. The FET:AFT ratios for these compounds were in the range of 0.2 to 0.7 (i.e., both FET and AFT tests were similar and were equally flawed). In spite of the FET:AFT test ratios being similar, these are also not further considered in the subsequent analyses. Details of rationales for inclusion or exclusion of selected problematic compounds or tests are given in the Supplemental Data. It is important to note in this overall exercise of comparing FET and AFT that the OECD FET validation included several challenging compounds. Some were exceptionally toxic (e.g., methylmercury), and others were of moderate to high KOW (tetradecyl sulfate, triclosan, dibutyl maleate) or were semivolatile (6‐methyl, 5‐hepten 2‐one). In addition, several compounds were rapidly and highly biodegradable (tetradecyl sulfate, octanol).

Relationship between fish embryo test (FET) and fish acute fish toxicity tests to solubility. The horizontal solid lines indicate toxicity:solubility ratios of 1 and 10. Values above 10 are especially suspect and were candidates for deletion from further consideration. All species tested in the FET and all species tested in the acute fish toxicity test are included. KOW = octanol–water partition coefficient; LC50 = median lethal concentration.
RESULTS
Overview of FET studies
The final database used to probe the relationship between the FET and AFT consisted of 985 FET studies on 229 compounds and 1531 AFT studies on 151 compounds. FET studies in the database were dominated by zebrafish (D. rerio) at nearly 97% (Table 5). Most tests were conducted static (64.8%), followed by (34.7%) semistatic. Semistatic tests were more frequent after 2009, and semistatic tests were used nearly exclusively in the OECD validation of the zebrafish FET 16.
Comparison | AFT | FET | ||
n | % | n | % | |
Distribution of fish species tested | ||||
Zebrafish (Danio rerio) | 80 | 5.2 | 955 | 97 |
Bluegill (Lepmois macrochirus) | 392 | 25.6 | 0 | 0 |
Rainbow trout (Oncorhynchus mykiss) | 455 | 29.7 | 0 | 0 |
Medaka (Oryzias latipes) | 105 | 6.9 | 4 | 0.4 |
Fathead minnow (Pimephales promelas) | 499 | 32.6 | 24 | 2.4 |
African sharptooth catfish (Clarias gariepinus) | 0 | 0 | 2 | 0.2 |
Total | 1531 | 100 | 985 | 100 |
Test type | ||||
Flow through | 500 | 32.7 | 5 | 0.5 |
Semistatic (=static renewal) | 785 | 51.3 | 342 | 34.7 |
Static | 105 | 6.9 | 638 | 64.8 |
Not recorded | 141 | 9.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Analytical verification of exposure | ||||
Yes | 510 | 33.3 | 59 | 6.0 |
No | 651 | 42.5 | 926 | 94.0 |
Not recorded | 370 | 24.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Duration of test (hours) | ||||
24 | 8 | 0.5 | 8 | 0.8 |
48 | 115 | 7.5 | 541 | 54.9 |
72 | 7 | 0.5 | 47 | 4.8 |
91 | 3 | 0.2 | 0 | 0 |
96 | 1398 | 91.3 | 355 | 36.0 |
108 | 0 | 0 | 4 | 0.4 |
120 | 0 | 0 | 30 | 3.0 |
Total | 1531 | 100 | 985 | 100 |
Comparison | AFT | FET | ||
n | % | n | % | |
Distribution of fish species tested | ||||
Zebrafish (Danio rerio) | 80 | 5.2 | 955 | 97 |
Bluegill (Lepmois macrochirus) | 392 | 25.6 | 0 | 0 |
Rainbow trout (Oncorhynchus mykiss) | 455 | 29.7 | 0 | 0 |
Medaka (Oryzias latipes) | 105 | 6.9 | 4 | 0.4 |
Fathead minnow (Pimephales promelas) | 499 | 32.6 | 24 | 2.4 |
African sharptooth catfish (Clarias gariepinus) | 0 | 0 | 2 | 0.2 |
Total | 1531 | 100 | 985 | 100 |
Test type | ||||
Flow through | 500 | 32.7 | 5 | 0.5 |
Semistatic (=static renewal) | 785 | 51.3 | 342 | 34.7 |
Static | 105 | 6.9 | 638 | 64.8 |
Not recorded | 141 | 9.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Analytical verification of exposure | ||||
Yes | 510 | 33.3 | 59 | 6.0 |
No | 651 | 42.5 | 926 | 94.0 |
Not recorded | 370 | 24.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Duration of test (hours) | ||||
24 | 8 | 0.5 | 8 | 0.8 |
48 | 115 | 7.5 | 541 | 54.9 |
72 | 7 | 0.5 | 47 | 4.8 |
91 | 3 | 0.2 | 0 | 0 |
96 | 1398 | 91.3 | 355 | 36.0 |
108 | 0 | 0 | 4 | 0.4 |
120 | 0 | 0 | 30 | 3.0 |
Total | 1531 | 100 | 985 | 100 |
AFT = acute fish toxicity; FET = fish embryo test.
Comparison | AFT | FET | ||
n | % | n | % | |
Distribution of fish species tested | ||||
Zebrafish (Danio rerio) | 80 | 5.2 | 955 | 97 |
Bluegill (Lepmois macrochirus) | 392 | 25.6 | 0 | 0 |
Rainbow trout (Oncorhynchus mykiss) | 455 | 29.7 | 0 | 0 |
Medaka (Oryzias latipes) | 105 | 6.9 | 4 | 0.4 |
Fathead minnow (Pimephales promelas) | 499 | 32.6 | 24 | 2.4 |
African sharptooth catfish (Clarias gariepinus) | 0 | 0 | 2 | 0.2 |
Total | 1531 | 100 | 985 | 100 |
Test type | ||||
Flow through | 500 | 32.7 | 5 | 0.5 |
Semistatic (=static renewal) | 785 | 51.3 | 342 | 34.7 |
Static | 105 | 6.9 | 638 | 64.8 |
Not recorded | 141 | 9.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Analytical verification of exposure | ||||
Yes | 510 | 33.3 | 59 | 6.0 |
No | 651 | 42.5 | 926 | 94.0 |
Not recorded | 370 | 24.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Duration of test (hours) | ||||
24 | 8 | 0.5 | 8 | 0.8 |
48 | 115 | 7.5 | 541 | 54.9 |
72 | 7 | 0.5 | 47 | 4.8 |
91 | 3 | 0.2 | 0 | 0 |
96 | 1398 | 91.3 | 355 | 36.0 |
108 | 0 | 0 | 4 | 0.4 |
120 | 0 | 0 | 30 | 3.0 |
Total | 1531 | 100 | 985 | 100 |
Comparison | AFT | FET | ||
n | % | n | % | |
Distribution of fish species tested | ||||
Zebrafish (Danio rerio) | 80 | 5.2 | 955 | 97 |
Bluegill (Lepmois macrochirus) | 392 | 25.6 | 0 | 0 |
Rainbow trout (Oncorhynchus mykiss) | 455 | 29.7 | 0 | 0 |
Medaka (Oryzias latipes) | 105 | 6.9 | 4 | 0.4 |
Fathead minnow (Pimephales promelas) | 499 | 32.6 | 24 | 2.4 |
African sharptooth catfish (Clarias gariepinus) | 0 | 0 | 2 | 0.2 |
Total | 1531 | 100 | 985 | 100 |
Test type | ||||
Flow through | 500 | 32.7 | 5 | 0.5 |
Semistatic (=static renewal) | 785 | 51.3 | 342 | 34.7 |
Static | 105 | 6.9 | 638 | 64.8 |
Not recorded | 141 | 9.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Analytical verification of exposure | ||||
Yes | 510 | 33.3 | 59 | 6.0 |
No | 651 | 42.5 | 926 | 94.0 |
Not recorded | 370 | 24.2 | 0 | 0 |
Total | 1531 | 100 | 985 | 100 |
Duration of test (hours) | ||||
24 | 8 | 0.5 | 8 | 0.8 |
48 | 115 | 7.5 | 541 | 54.9 |
72 | 7 | 0.5 | 47 | 4.8 |
91 | 3 | 0.2 | 0 | 0 |
96 | 1398 | 91.3 | 355 | 36.0 |
108 | 0 | 0 | 4 | 0.4 |
120 | 0 | 0 | 30 | 3.0 |
Total | 1531 | 100 | 985 | 100 |
AFT = acute fish toxicity; FET = fish embryo test.
The 48‐h FET studies were most common and comprised approximately 55% of the studies included in the present study. Studies of 96‐h duration were more frequent after 2008. Accompanying FET exposures with analytical verification was not common, occurring only 6% of the time and almost exclusively from tests conducted from 2009 to 2012. Comparisons of LC50s determined using both measured exposure concentrations and nominal values in FET studies were available 16. Measured LC50s were never more than 39% lower than the nominal determination (the worst case being dibutyl maleate, a difficult test substance). Many organic substances, including the antimicrobial triclosan and the fragrance 6‐methyl, 5‐hepten 2‐one, had measured LC50s that deviated only 1% to 20% from the nominal determination.
The majority of compounds for which FET data exist were represented by a single test. In total, 128 of 229 compounds were single assays of a compound (55.6%). The chemicals involved in the OECD validation program were 20 of the 26 most intensively assayed chemicals. On average, each chemical was tested 4.3 times, and only 50 chemicals were tested more than 4 times. The FET database was dominated by zebrafish, but information was available on other species about 5% of the time. The most frequently encountered exposure regime was 48 h, static exposure with no analytical verification. However, it should also be noted that 96‐h, semistatic exposures have become dominant in the most recent literature and research and will eventually overtake the 48‐h, static component of the database.
Overview of AFT studies
Among the 1531 AFT studies, 32.6% were conducted with fathead minnow, 29.7% with rainbow trout, 25.6% with bluegill, 6.9% with medaka, and 5.2% with zebrafish (Table 5). Among the approximately 1500 acute fish toxicity studies, fathead minnow, rainbow trout, and bluegill dominated relatively equally, and zebrafish and medaka were less well represented (Table 5). Approximately one‐third of AFT studies were performed flow‐through, with about 50% of studies being semistatic. Exposure durations most frequently encountered were 96 h (91%), with shorter durations being about 9%. Data associated with Japanese medaka were frequently of 48 h duration, but not exclusively. Analytical verification of exposures was applied in 33% of the studies, and most exposures were not verified or analytical data were not explicitly mentioned.
Fish size was evaluated as a potential confounding variable in the assessment. Weight and length data for a random subset of all the 1531 studies are summarized in Table 6. Tests were conducted with fish that, on average, complied with test guideline recommendations for average length (OECD test 203 recommendations of 20 ± 10 mm for zebrafish, medaka, and fathead minnow; 50 ± 10 mm for rainbow trout) and weight (USEPA recommendations of <3 g), although exceptions existed 37. To address the implications of fish size further, 3 compounds were chosen that were among the most well studied: Cd (a metal; n = 66), phenol (a neutral organic; n = 43), and diazinon (a pesticide; n = 65). For this group of compounds, 107 of 174, or 61%, fully reported fish size information, and 81% of these studies met all OECD test 203 17 and USEPA 37 guideline requirements for the smallest fish recommended. Note that if a study was identified as having been conducted as an OECD test 203 or USEPA 850.1075 AFT but did not provide fish size information, the study was not excluded from the present study, but it did not contribute to the fish size analysis. Regressions of fish length and weight for each species and test compound resulted in 30 potential combinations (5 species, 3 compounds for length and weight). Insufficient data existed to address other size‐related measures such as age or sexual maturity. Eight species–chemical combinations had sufficient data to develop quality regressions for weight and 9 for length were possible. Fifteen of 17 regressions indicated that no relationship existed between fish size and AFT test LC50 values. Two regressions suggested that sensitivity to Cd was slightly greater with increasing fish size for fathead minnow and rainbow trout (p < 0.05). Although the regressions were statistically significant, plots of the data showed that the relationship was driven by a small number of values at the extreme ends of size tested, for which results were also more variable. In addition, in no case were significant relationships found for both length and weight for the same species–compound combination.
Comparison of fish size in a representative random subsample of all available acute fish toxicity studies
Species | n | Average wet weight (g) | Standard deviation wet weight (g) | Average length (mm) | Standard deviation length (mm) |
Zebrafish (Danio rerio) | 15 | 0.31 | 0.18 | 18.3 | 11.3 |
Bluegill (Lepomis macrochirus) | 15 | 4.55 | 12.03 | 53.5 | 31.7 |
Rainbow trout (Oncorhynchus mykiss) | 15 | 15.13 | 15.89 | 49.1 | 27.9 |
Medaka (Oryzias latipes) | 15 | 0.35 | 0.14 | 16.1 | 4.1 |
Fathead minnow (Pimephales promelas) | 15 | 0.42 | 0.23 | 19.6 | 13.8 |
Species | n | Average wet weight (g) | Standard deviation wet weight (g) | Average length (mm) | Standard deviation length (mm) |
Zebrafish (Danio rerio) | 15 | 0.31 | 0.18 | 18.3 | 11.3 |
Bluegill (Lepomis macrochirus) | 15 | 4.55 | 12.03 | 53.5 | 31.7 |
Rainbow trout (Oncorhynchus mykiss) | 15 | 15.13 | 15.89 | 49.1 | 27.9 |
Medaka (Oryzias latipes) | 15 | 0.35 | 0.14 | 16.1 | 4.1 |
Fathead minnow (Pimephales promelas) | 15 | 0.42 | 0.23 | 19.6 | 13.8 |
Comparison of fish size in a representative random subsample of all available acute fish toxicity studies
Species | n | Average wet weight (g) | Standard deviation wet weight (g) | Average length (mm) | Standard deviation length (mm) |
Zebrafish (Danio rerio) | 15 | 0.31 | 0.18 | 18.3 | 11.3 |
Bluegill (Lepomis macrochirus) | 15 | 4.55 | 12.03 | 53.5 | 31.7 |
Rainbow trout (Oncorhynchus mykiss) | 15 | 15.13 | 15.89 | 49.1 | 27.9 |
Medaka (Oryzias latipes) | 15 | 0.35 | 0.14 | 16.1 | 4.1 |
Fathead minnow (Pimephales promelas) | 15 | 0.42 | 0.23 | 19.6 | 13.8 |
Species | n | Average wet weight (g) | Standard deviation wet weight (g) | Average length (mm) | Standard deviation length (mm) |
Zebrafish (Danio rerio) | 15 | 0.31 | 0.18 | 18.3 | 11.3 |
Bluegill (Lepomis macrochirus) | 15 | 4.55 | 12.03 | 53.5 | 31.7 |
Rainbow trout (Oncorhynchus mykiss) | 15 | 15.13 | 15.89 | 49.1 | 27.9 |
Medaka (Oryzias latipes) | 15 | 0.35 | 0.14 | 16.1 | 4.1 |
Fathead minnow (Pimephales promelas) | 15 | 0.42 | 0.23 | 19.6 | 13.8 |
Test temperature was frequently reported, but not universally, just as with fish size. A random subsampling of fish studies from the >1500 conducted indicated that mean (± standard deviation) temperatures (°C) for AFT tests were as follows: zebrafish, 23.7 ± 1.9 °C (n = 7); bluegill, 18.7 ± 4.1 °C (n = 74); rainbow trout 13.1 ± 3.2 °C (n = 65); fathead minnow, 21.2 ± 4.3 °C (n = 97); medaka, 23.0 ± 2.7 °C (n = 3). Among these studies, 76.2% were conducted within temperature ranges for accepted test guidelines, and an additional 9.3% of studies were within the guidelines plus 1 °C. Because the windows of test temperatures were relatively narrow, detailed evaluations of temperature effects on LC50 results did not yield additional insight. All correlations between temperature and LC50s for frequently studied compounds were not significant.
In total, 151 compounds were tested in common between the FET and the AFT. Among these, 50 acute fish toxicity studies had only 1 entry per chemical. This resulted in an average number of repeat studies of 9.9 per chemical, with 16 chemicals having 30 or more individual test results. The AFT database was relatively balanced with respect to species composition. Fathead minnow, bluegill sunfish, and rainbow trout are the most abundant species in the database. The most frequently encountered exposure regime was 96 h, semistatic exposure, with no analytical verification.
FET–AFT test relationships
Several iterations of the FET–AFT relationship were developed, beginning with all data considered (n = 151 chemicals; Figure 3). This regression has a slope close to 1 and an intercept close to 0, with p < 0.001. The slope confidence interval surrounds unity. A measurable degree of scatter exists in both x and y planes, and for some compounds this is over a few orders of magnitude. If studies shorter than 96 h for the AFT test are eliminated, the regression changes only marginally (Figure 4). Exposure duration, from the perspective of the AFT test, appears not to be a major influence as the bulk of the AFT data is already 96 h and most of the shorter term studies are from Japanese medaka, the least commonly tested of the 5 species included in the present study. When both FET and AFT results are constrained to 96 h, the number of chemicals included was reduced by roughly 50% (Figure 5). However, the resulting only 96‐h data regression is close to the regressions presented above. The confidence intervals are completely overlapping and inclusive of the slopes from all the other regressions (Table 7). Regressions presented in Figures 4 and 5 are highly statistically significant (p < 0.001). The strength of the correlation changes the most and approaches unity. In practical terms, the regressions show a very good predictive relationship between the FET and AFT test. Based on the 96‐h FET and 96‐h AFT regression, predictions of AFT from FET results of 0.1 mg/L, 1 mg/L, and 10 mg/L would be 0.07 mg/L, 0.7 mg/L, and 6.2 mg/L, respectively.

Comparison of the fish embryo test (FET) versus acute fish toxicity test for 151 compounds (data for all species are included). Open circles indicate individual test data; solid circles are geometric means of FET‐acute fish toxicity results for each chemical. LC50 = median lethal concentration.

Comparison of the fish embryo test (FET) versus 96‐h acute fish toxicity test for 144 compounds (data for all species are included). Open circles indicate individual test data; solid circles are geometric means of FET‐acute fish toxicity results for each chemical. LC50 = median lethal concentration.

Comparison of 96‐h fish embryo test (FET) versus 96‐h fish acute toxicity test for 72 compounds (data for all species are included). Open circles indicate individual test data; solid circles are geometric means of FET‐acute fish toxicity results for each chemical. LC50 = median lethal concentration.
Regression | Slope (95% confidence interval) | Intercept | r | n |
FET–Fish 4 | 1.070 (0.95, 1.20) | −0.350 | 0.89 | 77 |
FET–all fish (the present study) | 1.027 (0.95, 1.11) | −0.285 | 0.90 | 151 |
FET–96‐h fish (the present study) | 1.026 (0.95, 1.11) | −0.307 | 0.90 | 144 |
96‐h FET–96‐h fish (the present study) | 0.989 (0.92, 1.07) | −0.195 | 0.95 | 72 |
Zebrafish FET–zebrafish | 0.975 (0.80, 1.19) | 0.056 | 0.85 | 44 |
FET–rainbow trout | 1.033 (0.92, 1.16) | −0.617 | 0.89 | 75 |
Regression | Slope (95% confidence interval) | Intercept | r | n |
FET–Fish 4 | 1.070 (0.95, 1.20) | −0.350 | 0.89 | 77 |
FET–all fish (the present study) | 1.027 (0.95, 1.11) | −0.285 | 0.90 | 151 |
FET–96‐h fish (the present study) | 1.026 (0.95, 1.11) | −0.307 | 0.90 | 144 |
96‐h FET–96‐h fish (the present study) | 0.989 (0.92, 1.07) | −0.195 | 0.95 | 72 |
Zebrafish FET–zebrafish | 0.975 (0.80, 1.19) | 0.056 | 0.85 | 44 |
FET–rainbow trout | 1.033 (0.92, 1.16) | −0.617 | 0.89 | 75 |
Regression | Slope (95% confidence interval) | Intercept | r | n |
FET–Fish 4 | 1.070 (0.95, 1.20) | −0.350 | 0.89 | 77 |
FET–all fish (the present study) | 1.027 (0.95, 1.11) | −0.285 | 0.90 | 151 |
FET–96‐h fish (the present study) | 1.026 (0.95, 1.11) | −0.307 | 0.90 | 144 |
96‐h FET–96‐h fish (the present study) | 0.989 (0.92, 1.07) | −0.195 | 0.95 | 72 |
Zebrafish FET–zebrafish | 0.975 (0.80, 1.19) | 0.056 | 0.85 | 44 |
FET–rainbow trout | 1.033 (0.92, 1.16) | −0.617 | 0.89 | 75 |
Regression | Slope (95% confidence interval) | Intercept | r | n |
FET–Fish 4 | 1.070 (0.95, 1.20) | −0.350 | 0.89 | 77 |
FET–all fish (the present study) | 1.027 (0.95, 1.11) | −0.285 | 0.90 | 151 |
FET–96‐h fish (the present study) | 1.026 (0.95, 1.11) | −0.307 | 0.90 | 144 |
96‐h FET–96‐h fish (the present study) | 0.989 (0.92, 1.07) | −0.195 | 0.95 | 72 |
Zebrafish FET–zebrafish | 0.975 (0.80, 1.19) | 0.056 | 0.85 | 44 |
FET–rainbow trout | 1.033 (0.92, 1.16) | −0.617 | 0.89 | 75 |
When regressions included only FET and AFT for zebrafish, the domain of chemicals covered by the regression is reduced to n = 44 from 151, and the range of potency declines to 5 orders of magnitude as opposed to 9 (Figure 6A). The correlation, the regression coefficients, and the width of the slope confidence interval become somewhat worse; however, this is to be expected given that the number of tests is less than one‐third of the full database. Other possible combinations of FET with each fish species were also assessed. Rainbow trout is known to be somewhat more sensitive than many other standard OECD fish species 38 (also based on relationships in Web‐ICE). An expectation would be that the regression line runs parallel to the line of 1:1 correspondence, but the intercept more clearly departs from 0. This expectation is met in Figure 6B for the FET–rainbow trout relationship. Slope and correlation coefficients are remarkably similar to those of other FET–AFT test regressions (Figures 3,4 and 5). The intercept indicates rainbow trout as more sensitive. Coefficients of various regressions are summarized in Table 7, and all models had p < 0.001.

Representative regressions of fish embryo test (FET)–fish regressions for zebrafish only (A) and rainbow trout acute fish toxicity (B). Open circles indicate individual test data; solid circles are geometric means of FET‐acute fish toxicity results for each chemical. LC50 = median lethal concentration.
FET–AFT test comparisons as a function of KOW, chemical class, or functional use categories
In total, 41 compounds in the database have a KOW in excess of 3. When FET versus AFT values are regressed for this group of chemicals, the relationship suggests a level of departure for the less toxic compounds (low KOW); however, the sample size of FET–fish pairs is relatively small on the less toxic end of the distribution (Figure 7A). Still, the 95% confidence interval for the slope coefficient encompassed 1, and the bulk of the data lies close to the line of 1:1 correspondence (p < 0.001). See the Supplemental Data, in which chemicals were evaluated for inclusion or deletion based on KOW. The toxicity results for the highest KOW chemicals (7 and greater) for both FET and AFT might have been similar but were not considered in this analysis because tests in this KOW range were often conducted well above the limit of solubility (in other words, these results would be equally wrong but still close to the line of 1:1 correspondence). Plots of AFT LC50 versus FET LC50 versus log KOW indicate that all trends occur as expected and that there appears to be little bias toward high log KOW being poorly predicted (under‐predicted) in the FET assay. The chemical with the greatest departure from the regression line is cyclohexane, with both AFT and FET having similar departures from the line of 1:1 correspondence. Cyclohexane has a calculated KOW of 3.1 and published values ranging from 3.44 (ECOSAR, predicted and measured data). Cyclohexane was the least toxic compound in this exercise.

Fish embryo test (FET) versus acute fish toxicity for compounds with octanol–water partition coefficient (log KOW) above 3 (data for all species are included; A); and (B) overlying KOW values in the context of toxicity data. Open circles indicate individual test data; solid circles are geometric means of FET‐acute fish toxicity results for each chemical. LC50 = median lethal concentration.
The ECOSAR program was capable of assessing 207 of the 229 compounds in the database, and assignment to chemical class based on structure was consistent with that provided by ASTER. Those that were not assessed were inorganics, metals or organometals, polymers, or structures that could not be translated into SMILES for various reasons. Among these 207 compounds, most were neutral organics, aliphatic amines, and phenols (see also Tables 3 and 4). For the class of compounds categorized as neutral organic, aliphatic amine, or phenol, 27, 26, and 17 chemicals were assessed in the FET and AFT tests, respectively. Other chemical groups were insufficiently populated to provide meaningful results. Regressions for neutral organics, aliphatic amines, and phenols are summarized in Figure 8A to 8C, and all have p < 0.001. Neutral organics were the least toxic (as a group), whereas phenols were the most toxic (other, less well‐studied groups, such as epoxides or pyrethroids, may be more toxic overall but were not as completely studied as a group). Each group represented a subpopulation of potencies within the larger FET–AFT relationship of Figures 3 and 4. Slopes, intercepts, and correlation coefficients were consistent with previous regressions. They were reflective of the smaller samples and narrower ranges of potency within each group.

Fish embryo test (FET)‐acute fish toxicity comparisons for neutral organics (A), aliphatic amines (B), and phenols (C; data for all species are included). Open circles indicate individual test data; solid circles are geometric means of FET‐acute fish toxicity results for each chemical. LC50 = median lethal concentration.
The last group of FET–AFT test analyses was an assessment of major groups of chemicals based on functional use groups, including industrial organic chemicals, pesticides, pharmaceuticals, and surfactants. Industrial organics was the most prevalent overall group in the database and spanned about 5 orders of magnitude in potency, mostly above 1 mg/L (Figure 9). Pesticides were relatively less tested, but not uncommonly so. Pesticides were the most toxic group of compounds in the database. Some compounds were exceptionally variable, with some individual compounds spanning up to 5 orders of magnitude in measured toxicity to fish. The overall pesticide slope was exceptionally close to 1; however, the intercept suggests a pattern of slightly greater sensitivity for fish versus the FET. Thiram, an ectoparasticide, was the worst conforming, and azinphosmethyl, an organophosphate pesticide was the second worst conforming. AFT tests of these compounds were mostly without analytical procedures, and this may have contributed to the larger than average intercept for the FET–AFT relationship. With respect to pharmaceuticals, the database is small. Few published data are available on acute toxicity of pharmaceutical compounds overall, reflecting that the primary interest is the response of fish to long‐term, low levels of exposure. The FET–AFT relationship for pharmaceuticals is still directionally correct (positive slope, not far from 1 for the sample size, p < 0.05). Surfactants are the last relatively large functional group of chemicals, although this is a meshing of charged and uncharged chemicals. The prediction is good in the narrow range of potency covered.

Comparison of fish embryo test (FET) and acute fish toxicity for major categories of functional uses of chemicals including industrial organics (A), pesticides (B), pharmaceuticals (C), and surfactants (D; data for all species are included). Open circles indicate individual test data; solid circles are geometric means of FET‐acute fish toxicity results for each chemical. LC50 = median lethal concentration.
A summary of the various permutations of chemical groups (by functional use, by ECOSAR chemical class, and by KOW cutoff) is given in Table 8. Slopes are approximately 1, and intercepts vary (positively and negatively) depending on the group. The most robust relationships are those with samples sizes that approach 30 or greater.
Summary of regression coefficients for FET–acute fish toxicity relationships with respect to functional use, ECOSAR chemical class, and log KOW5
Chemical group | Slope | Intercept | n | r |
Industrial organics | 1.051 | −0.395 | 83 | 0.89 |
Pesticides | 1.007 | −0.717 | 18 | 0.71 |
Pharmaceuticals | 1.518 | −2.542 | 8 | 0.76 |
Surfactants | 0.902 | 0.497 | 15 | 0.85 |
Neutral organics | 1.056 | −0.507 | 27 | 0.89 |
Aliphatic amines | 1.254 | −1.422 | 26 | 0.77 |
Phenols | 1.084 | −0.434 | 17 | 0.83 |
High KOW (>3) | 0.837 | 0.240 | 41 | 0.84 |
Chemical group | Slope | Intercept | n | r |
Industrial organics | 1.051 | −0.395 | 83 | 0.89 |
Pesticides | 1.007 | −0.717 | 18 | 0.71 |
Pharmaceuticals | 1.518 | −2.542 | 8 | 0.76 |
Surfactants | 0.902 | 0.497 | 15 | 0.85 |
Neutral organics | 1.056 | −0.507 | 27 | 0.89 |
Aliphatic amines | 1.254 | −1.422 | 26 | 0.77 |
Phenols | 1.084 | −0.434 | 17 | 0.83 |
High KOW (>3) | 0.837 | 0.240 | 41 | 0.84 |
Cutoff for log KOW was > 3.
FET = fish embryo test; ECOSAR = Ecological Structure Activity Relationships program; KOW = octanol–water partition coefficient.
Summary of regression coefficients for FET–acute fish toxicity relationships with respect to functional use, ECOSAR chemical class, and log KOW5
Chemical group | Slope | Intercept | n | r |
Industrial organics | 1.051 | −0.395 | 83 | 0.89 |
Pesticides | 1.007 | −0.717 | 18 | 0.71 |
Pharmaceuticals | 1.518 | −2.542 | 8 | 0.76 |
Surfactants | 0.902 | 0.497 | 15 | 0.85 |
Neutral organics | 1.056 | −0.507 | 27 | 0.89 |
Aliphatic amines | 1.254 | −1.422 | 26 | 0.77 |
Phenols | 1.084 | −0.434 | 17 | 0.83 |
High KOW (>3) | 0.837 | 0.240 | 41 | 0.84 |
Chemical group | Slope | Intercept | n | r |
Industrial organics | 1.051 | −0.395 | 83 | 0.89 |
Pesticides | 1.007 | −0.717 | 18 | 0.71 |
Pharmaceuticals | 1.518 | −2.542 | 8 | 0.76 |
Surfactants | 0.902 | 0.497 | 15 | 0.85 |
Neutral organics | 1.056 | −0.507 | 27 | 0.89 |
Aliphatic amines | 1.254 | −1.422 | 26 | 0.77 |
Phenols | 1.084 | −0.434 | 17 | 0.83 |
High KOW (>3) | 0.837 | 0.240 | 41 | 0.84 |
Cutoff for log KOW was > 3.
FET = fish embryo test; ECOSAR = Ecological Structure Activity Relationships program; KOW = octanol–water partition coefficient.
Fish–fish toxicity relationships
The importance of acute fish–acute fish toxicity relationships were evaluated as a comparative benchmark for FET–AFT relationships. Fish–fish relationships represent accepted levels of prediction at the international level because each species is commonly used and is an option for use in hazard assessment. All possible combinations of fish–fish toxicity are given in Figure 10, with regression coefficients provided in Table 9. Overall, relationships appear most robust for rainbow trout versus any other species and fathead minnow versus any other species. The most variable is medaka versus any other species, which has a smaller sample size compared with other species (Table 9). As a means to further evaluate fish–fish relationships from the FET–AFT database, additional fish–fish regressions were obtained from the USEPA Web‐ICE program (http://www.epa.gov/ceampubl/fchain/webice/index.html; Table 9, entries marked as “Web‐ICE”). Regression coefficients for same‐species pairs were very similar, and overall trends are concordant.

All possible acute fish‐acute fish toxicity relationships. Plots are for rainbow trout (Oncorhynchus mykiss; A) versus each additional species, fathead minnow (Pimephales promelas; B) versus each additional species, bluegill (Lepomis macrochirus; C) versus each additional species, and medaka (Oryzias latipes; D). LC50 = median lethal concentration.
Fish 1 (x) | Fish 2 (y) | Slope | Intercept | n | r | Source |
Bluegill | Medaka | 0.883 | 0.552 | 23 | 0.899 | FET−fish db |
Rainbow trout | Fathead minnow | 0.888 | 0.669 | 79 | 0.909 | Web‐ICE |
Rainbow trout | Medaka | 0.927 | 0.719 | 32 | 0.961 | FET−fish db |
Fathead minnow | Rainbow trout | 0.929 | −0.100 | 79 | 0.909 | Web‐ICE |
Fathead minnow | Bluegill | 0.933 | 0.022 | 66 | 0.889 | Web‐ICE |
Bluegill | Zebrafish | 0.942 | 0.362 | 22 | 0.880 | FET−fish db |
Rainbow trout | Bluegill | 0.949 | 0.262 | 307 | 0.943 | Web‐ICE |
Bluegill | Fathead minnow | 0.973 | 0.340 | 56 | 0.940 | FET−fish db |
Bluegill | Rainbow trout | 0.982 | −0.127 | 48 | 0.918 | FET−fish db |
FET (96 h) | All Fish (96 h) | 0.989 | −0.195 | 72 | 0.95 | FET−fish db |
Medaka | Fathead minnow | 0.996 | 0.133 | 37 | 0.934 | FET−fish db |
Fathead minnow | Rainbow trout | 0.998 | −0.396 | 57 | 0.929 | FET−fish db |
Rainbow trout | Fathead minnow | 1.002 | 0.397 | 57 | 0.929 | FET−fish db |
Fathead minnow | Medaka | 1.004 | −0.134 | 37 | 0.934 | FET−fish db |
Rainbow trout | Bluegill | 1.019 | 0.130 | 48 | 0.918 | FET−fish db |
Fathead minnow | Bluegill | 1.028 | −0.350 | 56 | 0.940 | FET−fish db |
Fathead minnow | Medaka | 1.040 | −0.161 | 5 | 0.961 | Web‐ICE |
Fathead minnow | Zebrafish | 1.048 | −0.337 | 26 | 0.834 | FET−fish db |
Medaka | Rainbow trout | 1.079 | −0.775 | 32 | 0.961 | FET−fish db |
Rainbow trout | Zebrafish | 1.110 | 0.001 | 25 | 0.858 | FET−fish db |
Medaka | Bluegill | 1.132 | −0.625 | 23 | 0.899 | FET−fish db |
Medaka | Zebrafish | 1.593 | −2.225 | 16 | 0.719 | FET−fish db |
Fish 1 (x) | Fish 2 (y) | Slope | Intercept | n | r | Source |
Bluegill | Medaka | 0.883 | 0.552 | 23 | 0.899 | FET−fish db |
Rainbow trout | Fathead minnow | 0.888 | 0.669 | 79 | 0.909 | Web‐ICE |
Rainbow trout | Medaka | 0.927 | 0.719 | 32 | 0.961 | FET−fish db |
Fathead minnow | Rainbow trout | 0.929 | −0.100 | 79 | 0.909 | Web‐ICE |
Fathead minnow | Bluegill | 0.933 | 0.022 | 66 | 0.889 | Web‐ICE |
Bluegill | Zebrafish | 0.942 | 0.362 | 22 | 0.880 | FET−fish db |
Rainbow trout | Bluegill | 0.949 | 0.262 | 307 | 0.943 | Web‐ICE |
Bluegill | Fathead minnow | 0.973 | 0.340 | 56 | 0.940 | FET−fish db |
Bluegill | Rainbow trout | 0.982 | −0.127 | 48 | 0.918 | FET−fish db |
FET (96 h) | All Fish (96 h) | 0.989 | −0.195 | 72 | 0.95 | FET−fish db |
Medaka | Fathead minnow | 0.996 | 0.133 | 37 | 0.934 | FET−fish db |
Fathead minnow | Rainbow trout | 0.998 | −0.396 | 57 | 0.929 | FET−fish db |
Rainbow trout | Fathead minnow | 1.002 | 0.397 | 57 | 0.929 | FET−fish db |
Fathead minnow | Medaka | 1.004 | −0.134 | 37 | 0.934 | FET−fish db |
Rainbow trout | Bluegill | 1.019 | 0.130 | 48 | 0.918 | FET−fish db |
Fathead minnow | Bluegill | 1.028 | −0.350 | 56 | 0.940 | FET−fish db |
Fathead minnow | Medaka | 1.040 | −0.161 | 5 | 0.961 | Web‐ICE |
Fathead minnow | Zebrafish | 1.048 | −0.337 | 26 | 0.834 | FET−fish db |
Medaka | Rainbow trout | 1.079 | −0.775 | 32 | 0.961 | FET−fish db |
Rainbow trout | Zebrafish | 1.110 | 0.001 | 25 | 0.858 | FET−fish db |
Medaka | Bluegill | 1.132 | −0.625 | 23 | 0.899 | FET−fish db |
Medaka | Zebrafish | 1.593 | −2.225 | 16 | 0.719 | FET−fish db |
Web‐ICE = US Environmental Protection Agency Web‐based interspecies correlation estimation; FET = fish embryo test; db = database.
Fish 1 (x) | Fish 2 (y) | Slope | Intercept | n | r | Source |
Bluegill | Medaka | 0.883 | 0.552 | 23 | 0.899 | FET−fish db |
Rainbow trout | Fathead minnow | 0.888 | 0.669 | 79 | 0.909 | Web‐ICE |
Rainbow trout | Medaka | 0.927 | 0.719 | 32 | 0.961 | FET−fish db |
Fathead minnow | Rainbow trout | 0.929 | −0.100 | 79 | 0.909 | Web‐ICE |
Fathead minnow | Bluegill | 0.933 | 0.022 | 66 | 0.889 | Web‐ICE |
Bluegill | Zebrafish | 0.942 | 0.362 | 22 | 0.880 | FET−fish db |
Rainbow trout | Bluegill | 0.949 | 0.262 | 307 | 0.943 | Web‐ICE |
Bluegill | Fathead minnow | 0.973 | 0.340 | 56 | 0.940 | FET−fish db |
Bluegill | Rainbow trout | 0.982 | −0.127 | 48 | 0.918 | FET−fish db |
FET (96 h) | All Fish (96 h) | 0.989 | −0.195 | 72 | 0.95 | FET−fish db |
Medaka | Fathead minnow | 0.996 | 0.133 | 37 | 0.934 | FET−fish db |
Fathead minnow | Rainbow trout | 0.998 | −0.396 | 57 | 0.929 | FET−fish db |
Rainbow trout | Fathead minnow | 1.002 | 0.397 | 57 | 0.929 | FET−fish db |
Fathead minnow | Medaka | 1.004 | −0.134 | 37 | 0.934 | FET−fish db |
Rainbow trout | Bluegill | 1.019 | 0.130 | 48 | 0.918 | FET−fish db |
Fathead minnow | Bluegill | 1.028 | −0.350 | 56 | 0.940 | FET−fish db |
Fathead minnow | Medaka | 1.040 | −0.161 | 5 | 0.961 | Web‐ICE |
Fathead minnow | Zebrafish | 1.048 | −0.337 | 26 | 0.834 | FET−fish db |
Medaka | Rainbow trout | 1.079 | −0.775 | 32 | 0.961 | FET−fish db |
Rainbow trout | Zebrafish | 1.110 | 0.001 | 25 | 0.858 | FET−fish db |
Medaka | Bluegill | 1.132 | −0.625 | 23 | 0.899 | FET−fish db |
Medaka | Zebrafish | 1.593 | −2.225 | 16 | 0.719 | FET−fish db |
Fish 1 (x) | Fish 2 (y) | Slope | Intercept | n | r | Source |
Bluegill | Medaka | 0.883 | 0.552 | 23 | 0.899 | FET−fish db |
Rainbow trout | Fathead minnow | 0.888 | 0.669 | 79 | 0.909 | Web‐ICE |
Rainbow trout | Medaka | 0.927 | 0.719 | 32 | 0.961 | FET−fish db |
Fathead minnow | Rainbow trout | 0.929 | −0.100 | 79 | 0.909 | Web‐ICE |
Fathead minnow | Bluegill | 0.933 | 0.022 | 66 | 0.889 | Web‐ICE |
Bluegill | Zebrafish | 0.942 | 0.362 | 22 | 0.880 | FET−fish db |
Rainbow trout | Bluegill | 0.949 | 0.262 | 307 | 0.943 | Web‐ICE |
Bluegill | Fathead minnow | 0.973 | 0.340 | 56 | 0.940 | FET−fish db |
Bluegill | Rainbow trout | 0.982 | −0.127 | 48 | 0.918 | FET−fish db |
FET (96 h) | All Fish (96 h) | 0.989 | −0.195 | 72 | 0.95 | FET−fish db |
Medaka | Fathead minnow | 0.996 | 0.133 | 37 | 0.934 | FET−fish db |
Fathead minnow | Rainbow trout | 0.998 | −0.396 | 57 | 0.929 | FET−fish db |
Rainbow trout | Fathead minnow | 1.002 | 0.397 | 57 | 0.929 | FET−fish db |
Fathead minnow | Medaka | 1.004 | −0.134 | 37 | 0.934 | FET−fish db |
Rainbow trout | Bluegill | 1.019 | 0.130 | 48 | 0.918 | FET−fish db |
Fathead minnow | Bluegill | 1.028 | −0.350 | 56 | 0.940 | FET−fish db |
Fathead minnow | Medaka | 1.040 | −0.161 | 5 | 0.961 | Web‐ICE |
Fathead minnow | Zebrafish | 1.048 | −0.337 | 26 | 0.834 | FET−fish db |
Medaka | Rainbow trout | 1.079 | −0.775 | 32 | 0.961 | FET−fish db |
Rainbow trout | Zebrafish | 1.110 | 0.001 | 25 | 0.858 | FET−fish db |
Medaka | Bluegill | 1.132 | −0.625 | 23 | 0.899 | FET−fish db |
Medaka | Zebrafish | 1.593 | −2.225 | 16 | 0.719 | FET−fish db |
Web‐ICE = US Environmental Protection Agency Web‐based interspecies correlation estimation; FET = fish embryo test; db = database.
DISCUSSION
The FET has been developed over the past 20+ yr as a potential replacement for the AFT test. Schulte and Nagel 1 initially focused on a 48‐h assay with zebrafish, which encompasses much of the development prior to hatch. By the early 2000s, the FET had matured to a 72‐h assay with a substantial database that encouraged its use internationally, especially as animal welfare concerns have continued to evolve 2. In the most recent iteration, the FET has become well established as a 96‐h assay entirely within the period of embryonic development 11. Embry et al. 3 discussed the results from an ILSI‐HESI international workshop on FETs as an animal alternative, where the 96‐h assay was endorsed. The 96‐h FET assay is consistent with the definition of nonprotected life stages of fish in Directive 201/63/EU of the European Parlaiment and the Council of 22 September 2010 on the protection of laboratory animals used for scientific purposes 39 and increases the sensitivity of embryos to chemicals by providing a period of exposure posthatch without the chorion.
In the present study, the current FET–AFT relationship is seen to be exceptionally robust. Regressions of all available data (Figure 3), 96‐hr FET and 96‐h AFT only (Figure 5), and various subsets of data separated by compound physical–chemical properties (Figures 7 and 8) and functional use (Figure 9) all show slopes near 1 and intercepts approaching zero. Some noise, particularly in the AFT axis, is not possible to eliminate and likely is associated with allowed variability in factors typically associated with conducting fish toxicity tests 40. The AFT has experienced significant evolution over time, as has the FET. Factors such as differences in‐test system exposure, type and volume of the exposure system, use of analytical confirmation of the test compound, organism loading, water quality, and so forth, in addition to biological variability, contribute to variability in the AFT. The FET may seem less variable, but this may be more a function of FET tests being conducted in fewer laboratories at this point, the assay having been very recently developed with higher standardization for test conditions and principle reliance on zebrafish as the test organism. With respect to toxicity results, this is evident in the FET–AFT test plots, in which the span of toxicity for some chemicals reported in the literature may be as large as 4 orders of magnitude. A review of contributors to acute toxicity variability in a constrained data set used for Interspecies Correlation Estimation model development found that the common factors explored in test variability analyses resulted in a difference in toxicity of a factor of no more than 2, all other things being equal 40. Inclusion in the analysis was restricted primarily to USEPA‐generated data and had greater uniformity than the acute toxicity data in the present study (for example, comparisons of Raimondo et al. 40 did not include static studies). With respect to the FET, the validation studies performed for OECD indicate that coefficients of variation between laboratories were as high as 50% (for exquisitely toxic compounds with LC50s < 100 µg/L) and as low as 4% (LC50s > 100 mg/L) 16. These coefficients of variation are well within the range of those typical of other aquatic toxicity tests 44. Limited interlaboratory data are available for the AFT 45. In an interlaboratory evaluation of the fathead minnow acute toxicity test involving 13 laboratories, the interlaboratory coefficient of variation for potassium chloride was 19.7% and the intralaboratory coefficient of variation was 7.6%, which are remarkably similar to data for the zebrafish FET using sodium chloride (interlaboratory coefficient of variation of 18.9%, intralaboratory range of coefficients of variation from 6.3–19.5%) 16. These well‐controlled interlaboratory studies represent, however, the “best case” situation as evidenced by the regression plots of Figures 3 and 4 that indicate less variability in the x axis (FET) than in the y axis (AFT). The number of repeat tests for chemicals in the FET assay, however, is somewhat more limited than it is for the AFT test, which has a longer history of use. Overall, there appears to be no reason to expect the FET to be more or less variable than the AFT test.
It is surprising to many that noticeable gaps in chemical coverage for the AFT test exist and that there are many chemicals tested in the FET for which AFT data are not yet available. This is not to say that fish are infrequently tested, because the total universe of chemicals evaluated for acute fish toxicity is clearly quite large; however, noticeable gaps remain, as has been pointed out by numerous researchers 46 and has prompted much of the interest in systematic review through internal chemicals management such as in Canada via the Categorization of the Domestic Substances List and in Europe via the REACH legislation. Several reasons contribute to the difference in the availability of fish data versus FET data in the present study. First, scientists interested in understanding the limitation of the FET have gone to great lengths to probe the edges of what chemicals can be evaluated with the assay. Conducting tests with very challenging chemicals from a physical–chemistry perspective, assays of newly identified and emerging chemicals of concern, and an intentional effort to increase breadth of chemicals tested contribute to this phenomenon. Indeed, this is extremely important to do if the intent is to use the FET as a replacement for the AFT test, which is one of the most common early tier assays used globally in environmental hazard assessment for agrochemicals, pharmaceuticals, biocides, and industrial chemicals, as well as for the purpose of classification and labeling. The breadth of chemical categories and modes of action, though not completely exhaustive, is certainly representative of the universe of chemicals. Occasionally, a chemical might be encountered for which prediction of AFT may be problematic. For instance, the example of allyl alcohol being poorly metabolized by fish embryos, resulting in a disparity between embryo and juvenile to adult fish toxicity measurements, has been well described 48. In a recent study, Knöbel et al. 47 found zebrafish embryos to suffer no mortality to permethrin and 4‐decylanaline, compounds that were previously unstudied in the FET assay. Outliers in the zebrafish FET of Knöbel et al., identified as outliers based on nominal concentrations, were moved close to the line of unity in their regression analysis if based on measured concentrations. However, these FET assays were all 48 h, before hatching, and numerous studies have now shown that sensitivity can change markedly after the embryo hatches and transitions to the eleutheroembryo 16. Compounds such as the cationic polymers Merquat and Luviquat, the insecticide malathion, and the fungicide prochloraz had little or no measurable toxicity at 48 h but were clearly toxic to embryos by 96 h. Knöbel et al. 47 still determined that a least‐squares type II regression (not orthogonal as in the present study) was similar to the regressions in Table 9 in spite of being on a much smaller range of compounds.
Fish embryo–acute fish toxicity test relationships are similar when evaluated across a range of considerations, including the species considered, durations of test exposures, hydrophobicity, chemical categories as classified by ECOSAR, and functional groupings of chemicals by intended use. Although the database is dominated by narcotically acting chemicals, this representation is similar to their representation in commerce. Subgroups of chemicals, defined by either functional groupings or functional use, have similar FET–AFT relationships, as can be seen in Figures 8 and 9. It should be expected that any relationship will progressively improve with time as FET data accrue. The most significant departures from 1:1 relationships occur when the subcategory becomes smaller.
Acute fish–acute fish toxicity regressions are quantitatively similar to FET–AFT test regressions as initially reported 4 but for a much smaller database. The strength of any of these relationships is directly related to the number of data points and range of toxicity being considered. In essence, FET–AFT regressions cannot be distinguished from fish–fish regressions. In the USEPA Web‐ICE database for fish–fish relationships, similar types of regressions are documented for sample sizes in the range of FET–AFT regressions presented here when pairs of data are greater than 10 chemicals (see also Dyer et al. 49). In Table 9, FET information is indistinguishable from all internationally accepted fish species for use in acute toxicity tests of chemicals.
With the addition of approximately 50% more chemicals and several hundred more studies, relative to previous studies 47, it is clear that additional data will have marginal, if any, impact on the regressions. For the FET–AFT test regression in particular, the correlation coefficient increased from 0.89 to 0.90, the slope declined slightly from 1.070 to 1.027, and the intercept declined from −0.350 to −0.285 when comparing regressions from 2009 and the current model.
In summary, the FET based primarily on results from zebrafish embryos predicts AFT well. The size of the database encompasses a broad range of chemical classes, modes of action, and functional use categories. In relevant statistical aspects, the FET behaves like the AFT test. If the FET is regarded as an alternative to the AFT, the FET will provide nearly equivalent predictions of hazard while improving overall animal welfare. Incorporation of assays with additional FET species that is already underway, primarily fathead minnow and medaka, along with greater numbers of less well‐studied compounds, is likely to continue improving the FET methodology and approach.
SUPPLEMENTAL DATA
Tables S1 to S3.
Acute Fish Toxicity Data Literature. (938 KB PDF)
Acknowledgment
The authors are indebted to numerous collaborators who kindly provided data and input for this article. We especially thank the OECD Validation Management Group (T. Braunbeck, F. Busquet, M. Halder, S. Walter‐Rhode, A. Gourmelon, and A. Lillicrap), in which the authors also participated. Additional previously unpublished data were provided by T. Braunbeck and R. Strecker (University of Heidelberg), S. Scholz (Helmholtz Centre, Leipzig), and K. Schirmer and M. Knöbel (EAWAG, Switzerland) for the present study. Finally, the authors gratefully thank C. Russom and associated staff at the USEPA Mid‐Continent Ecology Division Laboratory, Duluth, Minnesota, for performing mode‐of‐action analyses using USEPA's ASTER program.
Disclaimer
The contents of this research have been fully shared with the OECD Environmental Directorate, laboratories participating in the fish embryo test validation, and the OECD Fish Embryo Test Validation Management Group. The research is intended to be broadly applicable for use in many contexts, including those in which the authors participate, such as chemical‐specific hazard assessment. No claim is laid to the work, other than Freedom to Practice, and the authors do not have any financial stake in the outcome of the research and its review. The contents of this paper represents a review of research conducted to aid development of alternatives to testing of protected laboratory animals and did not involve additional de novo testing of either embryos or juvenile or adult fish. Relevant research outcomes that did involve the de novo testing of embryos, juveniles, or adult fish will be summarized in a separate paper.
REFERENCES
International Standards Organization. Water quality‐determination of the acute toxicity of wastewater to zebrafish eggs (Danio rerio). ISO 1508 8: 2007(E). Geneva, Switzerland.