The relative sensitivity of amphibians to chemicals in the environment, including plant protection product active substances, is the subject of ongoing scientific debate. The objective of this study was to compare systematically the relative sensitivity of amphibians and fish to chemicals. Acute and chronic toxicity data were obtained from the U.S. Environmental Protection Agency (U.S. EPA) ECOTOX database and were supplemented with data from the scientific and regulatory literature. The overall outcome is that fish and amphibian toxicity data are highly correlated and that fish are more sensitive (both acute and chronic) than amphibians. In terms of acute sensitivity, amphibians were between 10- and 100-fold more sensitive than fish for only four of 55 chemicals and more than 100-fold more sensitive for only two chemicals. However, a detailed inspection of these cases showed a similar acute sensitivity of fish and amphibians. Chronic toxicity data for fish were available for 52 chemicals. Amphibians were between 10- and 100-fold more sensitive than fish for only two substances (carbaryl and dexamethasone) and greater than 100-fold more sensitive for only a single chemical (sodium perchlorate). The comparison for carbaryl was subsequently determined to be unreliable and that for sodium perchlorate is a potential artifact of the exposure medium. Only a substance such as dexamethasone, which interferes with a specific aspect of amphibian metamorphosis, might not be detected using fish tests. However, several other compounds known to influence amphibian metamorphosis were included in the analysis, and these did not affect amphibians disproportionately. These analyses suggest that additional amphibian testing is not necessary during chemical risk assessment. Environ. Toxicol. Chem. 2013;32:984–994. © 2013 SETAC
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Amphibian toxicity testing is not specifically required for the authorization of plant protection products or other chemicals in Europe. However, in recent years increasing interest has been expressed in developing a specific amphibian risk assessment approach because of perceived uncertainties that the aquatic life stages of amphibians are not protected by the current assessment schemes [1, 2]. Consequently, the latest draft of the proposed new European Union (EU) data requirements to be adopted under the new Plant Protection Products Regulation (Regulation 1107/2009) requires the risk to amphibians be addressed, although testing should not be required until an internationally validated test guideline is available. At present the only available guideline is the amphibian metamorphosis assay, Organisation for Economic Co-operation and Development (OECD) Test Guideline 231, which is a screening assay for thyroid toxicity and is therefore not designed to generate an endpoint for risk assessment. In other countries, toxicity testing (e.g., China) or specific risk assessment procedures (e.g., Canada) are required for amphibians.
The current lack of amphibian testing in most regions is attributable to (1) the assumption that the sensitivity of surrogate vertebrate species such as fish is comparable to that of aquatic life stages of amphibians, and therefore by using fish data the risks to aquatic life stages of amphibians can be assessed; (2) animal welfare concerns in some parts of the world (e.g., Europe) about the use of vertebrates for toxicity testing; and (3) an absence of standard guidelines for amphibian testing. The first assumption has been broadly confirmed in the published literature for the acute toxicity of some chemical groups. The most extensive recent review was performed by Kerby et al. . They used a species sensitivity distribution approach when analyzing the median lethal concentration (LC50) data from almost 24,000 studies retrieved from the U.S. Environmental Protection Agency (U.S. EPA)'s ECOTOX database to compare amphibian sensitivity with that of other groups of organisms. They concluded that amphibians are of low to moderate sensitivity to metals, inorganic chemicals, and plant protection product active substances (pyrethroid, carbamate, organophosphorus, or organochlorine) when compared with 13 other classes of organisms, including fishes. However, they found that amphibians were highly acutely sensitive to three phenolic chemicals.
Other authors have also found that amphibians are not usually more sensitive than fish [4-13]. However, analyses of the relative sensitivity of fish and amphibians using longer-term chronic endpoints have not been undertaken.
Despite this, several authors have implicated exposure to chemicals, particularly plant protection products, in the decline of wild amphibian populations [14-18], although the evidence for this remains contentious . Therefore, establishing whether laboratory data on amphibian sensitivity to chemicals lend any support to this hypothesis is important. Further understanding of the relative toxicity of fish and amphibians to plant protection products is crucial to help balance increased vertebrate test data needs with societal concerns and the intent of Article 62 of the new EU Plant Protection Products Regulation to reduce animal use. The objective of the present study is therefore to undertake a series of pairwise comparisons of acute and chronic toxicity data obtained from laboratory tests conducted using species of amphibians and fish, to determine whether sensitivity differs systematically between these two classes of organisms. A specific purpose of this analysis is to inform the current debate on inclusion of amphibian-specific testing under changes to the data requirements for the evaluation of plant protection products in Europe.
MATERIALS AND METHODS
Acute toxicity data for amphibians and fish were obtained from two sources. Data were principally obtained (February 2012) directly from the U.S. EPA ECOTOX database (http://cfpub.epa.gov/ecotox/), according to the following criteria. A chemical was identified for inclusion in the analysis if a pair of 96-h LC50 values was available in ECOTOX comprising any amphibian species and either rainbow trout (Oncorhynchus mykiss) or fathead minnow (Pimephales promelas). Oncorhynchus mykiss and P. promelas were selected because these are the most commonly used fish species in regulatory testing regimens worldwide and are therefore the species that would normally be used as surrogates for amphibians when assessing the adverse effects of chemicals. Oncorhynchus mykiss data were selected preferentially over P. promelas data because O. mykiss is generally considered to be more sensitive to chemicals , and for this reason O. mykiss is the only species required for acute testing under the new EU Plant Protection Products Regulation. Only tests from ECOTOX that reported measured concentrations of test chemical were included. A geometric mean value was calculated if more than one 96-h LC50 value was available for the same chemical and species. When data on several amphibian species were available, only data for the most sensitive amphibian species were selected for further analysis.
The resulting dataset of chemicals contained relatively few plant protection product active substances, so amphibian and fish data from Aldrich  were reanalyzed and used to supplement the ECOTOX-derived dataset described previously. Aldrich  reports acute amphibian data identified from the scientific literature over a preceding 15-year period (http://webofknowledge.com/WOS), or via either the PAN Pesticides database (http://www.pesticideinfo.org/) or the U.S. EPA ECOTOX database. Aldrich  extracted accompanying toxicity data for fish from documentation relevant to the EU plant protection product reevaluation procedure (http://dar.efsa.europa.eu/dar-web/provision). The Aldrich  analysis did not differentiate between amphibian studies that reported LC50 values based on measured concentrations or those based on nominal concentrations. When the Aldrich  study reported a range of LC50 values, or when LC50 values were available for different amphibian life stages (embryos, tadpoles, or adults), the lowest LC50 value was selected for the comparative analysis presented here. If plant protection products were not registered for use in the EU, Aldrich  occasionally reported amphibian toxicity data without matching fish toxicity data. Where possible, we filled these data gaps using acute toxicity data reported in U.S. re-registration eligibility decision documents (http://www.epa.gov/oppsrrd1/reregistration/status.htm) to remain consistent with Aldrich's original approach. If LC50 data for the same chemical were reported in both the ECOTOX and US-supplemented Aldrich  datasets, the lowest amphibian or fish data reported in either dataset were selected for the comparison here and are subsequently referred to as the combined ECOTOX/Supplemented Aldrich dataset in the results and figures.
Pairwise comparisons of the LC50 values from amphibian and fish studies were then undertaken for each chemical. Ratios (fish LC50/amphibian LC50) of greater than 1 indicate that amphibians are relatively more sensitive than fish. Ratios greater than 100 indicate that the standard EU acute toxicity assessment factor of 100 routinely applied to the results of acute fish tests during risk assessment would not be sufficient to cover the sensitivity of aquatic life stages of amphibians.
Chronic laboratory toxicity data for amphibians were primarily obtained (February 2012) from the U.S. EPA ECOTOX (http://cfpub.epa.gov/ecotox/) database. These searches were supplemented with data from the U.S. EPA Office of Pesticide Programs (OPP) Pesticide Ecotoxicity databases (http://www.ipmcenters.org/Ecotox/index.cfm) and with data from the scientific literature (http://webofknowledge.com/WOS) to encompass data published since the latest update of the ECOTOX database (September 2011–January 2012). This search used terms based on the amphibian genera recorded in the ECOTOX database. No specification of the identity or type of chemicals was made during any of the database or literature searches. Additional studies of relevance that were not present in the ECOTOX or OPP databases were also included, as well as data from Gross et al. , which included the results of phases 1 to 3 of the OECD validation program [22-24] for the Amphibian Metamorphosis Assay (OECD TG 231). Data were retained for analysis if they were from studies of at least a 10-d duration, employed either static-renewal or flow-through aqueous exposure study designs, and reported apical endpoints of potential population relevance (i.e., they were related to survival, growth, development [including metamorphosis], or reproduction).
Data from studies using either dietary or intraperitoneal exposure, or based on only a single exposure concentration, were not included in the analysis. Studies based on nominal rather than measured concentrations were not excluded from the analysis, because this would have considerably reduced the number of chemicals with amphibian data available for comparison with fish data. Therefore, in some instances these data may not be appropriate for direct regulatory purposes. For example, the amphibian atrazine data in Rohr et al. , although providing the lowest available no observed effect concentration (NOEC), would not necessarily be suitable for regulatory purposes after detailed review in light of a weight of evidence evaluation of all the data . However, this conservative approach is suitable for the general data comparisons made here.
Amphibian data were subsequently evaluated by endpoint and species, and the lowest long-term population-relevant NOEC, a key NOEC, was identified for each chemical for subsequent comparison with fish data. Key NOEC data were preferentially selected from studies that reported bounded NOEC values (i.e., those accompanied by a corresponding lowest observed effect concentration [LOEC]) and where the spacing factor between test concentrations was less than 100. Studies with unbounded NOEC values (i.e., where the highest tested concentration is reported as the NOEC) or those using a spacing factor of at least 100 were considered to be potentially unreliable, because in these cases, the true NOEC may be considerably higher than the reported value. However, when no bounded NOEC data for a chemical were available from a study employing a low spacing factor, such potentially unreliable amphibian NOECs were retained in the analysis, but clearly identified, as they represent a worst-case comparison of relative amphibian and fish sensitivity.
Chronic fish toxicity data for pairing with amphibian data were identified from the ECOTOX and OPP databases, and EU or U.S. regulatory dossiers, as well as further searches of the scientific literature. Fish data were retained if they were from static-renewal or flow-through laboratory aqueous exposures of at least 21 d and reported apical endpoints. Studies that reported potentially unreliable NOECs (e.g., limit tests, unbounded NOECs, or NOECs derived from studies using a factor of greater than 100 between concentrations) were not selected for comparison with amphibian data, unless they had been accepted for regulatory purposes in a previous review.
Pairwise comparisons of key NOECs from amphibian and fish studies were undertaken for each chemical, and the ratio of amphibian to fish sensitivity was calculated. Ratios of greater than one (fish NOEC/amphibian NOEC) indicate that amphibians are relatively more sensitive than fish. Factors greater than 10 indicate that the standard EU assessment factor of 10 routinely applied to the results of chronic fish tests during risk assessment would not be sufficient to cover the sensitivity of aquatic life stages of amphibians. Where spacing factors between amphibian and fish studies were of unequal size, an additional ratio was calculated of the respective maximum acceptable toxicant concentration (MATC), which is calculated as the geometric mean of the NOEC and LOEC value, as this measure is less sensitive to differences in spacing factors, and thus may provide a more suitable basis for comparison in these instances.
Statistical correlation between amphibian and fish LC50 and NOEC values was investigated using Spearman's correlation (Analyse-it Software Limited, version 2.26).
Ninety-six-hour LC50 comparisons of amphibian and fish sensitivity were made for 55 chemicals (eight inorganic chemicals and 47 organic chemicals, of which 32 were plant protection product active substances; Supplemental Data, Table S1). A very highly significant positive correlation was found between amphibian and fish LC50 values (Spearman's rank correlation rs = 0.81 [95% confidence interval, 0.72–1.00], p < 0.0001). The median sensitivity ratio was 0.52, implying that overall fish were acutely more sensitive than amphibians. Amphibians were more sensitive than fish in 16 of 55 cases, despite the data selection criteria favoring a higher sensitivity for amphibians. In 10 of these cases the sensitivity of amphibians was less than 10-fold that of fish.
Amphibians were between 10-fold and 100-fold more sensitive than fish on only four occasions (2,4-dichlorophenoxyacetic acid [2,4-D], aluminum chloride, malathion, and pentachlorophenol). On only two further occasions (dimethoate and p-nonylphenol), the difference was greater than 100-fold (Fig. 1). These two cases deviate from the overall high correlation between fish and amphibian toxicity and potentially indicate instances in which the standard EU assessment factor of 100 applied to acute fish LC50 values during risk assessment may not sufficiently cover the sensitivity of amphibians. Therefore, these instances are discussed in detail later in the present study.
Searches for apical amphibian data from the ECOTOX and OPP databases and the other sources returned data for 68 chemicals. Of these, corresponding chronic data for fish were available for 52 chemicals (10 inorganic chemicals and 42 organic chemicals, of which 20 were plant protection product active substances). The amphibian data comprised studies involving 14 amphibian species (predominantly studies using species from the genera Rana, Bufo, and Xenopus). The median exposure duration for chronic amphibian and fish studies was very similar at 42 and 37.5 d, respectively. Most of the fish NOEC values are based on measured exposure concentrations, whereas most of the amphibian data are based on nominal concentrations, making the latter less reliable.
Table 1 [27-98] summarizes chronic NOECs that were available for both amphibians and fish. A highly significant positive correlation was found between the amphibian and fish NOECs (Spearman's rs = 0.58 [95% confidence interval 0.51–1.00], p < 0.0001). The median sensitivity ratio was 0.31, implying that overall fish were chronically more sensitive than amphibians. Amphibians were more sensitive than fish in 11 of 52 cases, which is a remarkably low number considering the fact that the data selection criteria favored a higher sensitivity for amphibians. In five of these cases, the sensitivity of amphibians was less than 10-fold that of fish.
|NOEC (µg/L)||LOEC (µg/L)||Species||Exposure duration in days (type)||Endpoint/study type||Reference||NOEC (µg/L)||LOEC (µg/L)||Species||Exposure duration in days (type)||Endpoint/study type||Reference|
|17β-Estradiol||0.015||0.2||Xenopus laevis||73 (r, m)||Development||||0.003||0.009||Oryzias latipes||167 (f, m)||Growth||||0.19|
|2,4 Dichloroaniline||320||1,000||X. laevis||100 (r, n)||Development||||320||1,000||O. latipes||40 (r, n)||Survival||||1.0|
|3-Benzylidene camphor (3-BC)||50||-||X. laevis||35 (r, m)||Development||||33||74||Pimephales promelas||21 (r, m)||Reproduction||||0.66|
|4-Tert-octylphenol||206.33||515.82||Rana sylvatica||14 (r, n)||Growth||||14||65||Zoarces viviparus||35 (f, m)||Survival||||0.07|
|Acetochlor||2.7||-||Rana catesbeiana||52 (r, n)||Development||||130||270||Oncorhynchus mykiss||60 (f, m)||Early life stage||EPA-OPP||48.1|
|Acrylaldehyde||23||-||Bufo arenarum||42 (r, m)||Development||||14||35||P. promelas||32 (f, m)||Survival||||0.61|
|Alpha-cypermethrin||0.1||1||R. arvalis||60 (r, n)||Growth||||0.03||0.10||P. promelas||34 (f, m)||Survival||||0.3|
|Ammonium chloride||210||348||Rana pipiens||21 (f, m)||Growth||||6||12||Ictalurus punctatus||30 (f, m)||Growth||||0.03|
|Ammonium nitrate||3,300||6,600||Pseudacris regilla||10 (r, m)||Growth||||3||6||O. mykiss||21 (r, n)||Growth||||0.0009|
|Ammonium perchlorate||59||14,000||X. laevis||70 (r, m)||Development||||1,000||10,000||P. promelas||28 (r, n)||Growth||||16.9|
|Atrazine||40||400||Ambystoma barbouri||56 (r, n)||Growth||||8.4||82.4||Salmo salar||396 (f, m)||Growth||||0.21|
|Azinphos-methyl||30||110||Ambystoma maculatum||10 (f, m)||Growth||||0.17||0.34||Cyprinodon variegatus||28 (f, m)||Growth||||0.01|
|Azoxystrobin||10||-||Rana temporaria||42 (r, n)||Growth||||147||193||P. promelas||33 (f, m)||Growth||EPA-OPP||14.7|
|Benzophenone-2||1,500||3,000||X. laevis||21 (m)||Growth||||540||990||P. promelas||32 (f, m)||Growth||||0.36|
|Bisphenol A||228.12||2,282.9||X. laevis||36 (r, n)||Development||||2.4||5||Salmo trutta||61 (f, n)||Reproduction||||0.011|
|Cadmium||1.09||7.59||R. pipiens||42 (r, m)||Survival||||0.6||1.3||O. mykiss||53 (f, m)||Survival||||0.55|
|Carbaryl||5||50||A. barbouri||53 (r, n)||Survival||||210||680||P. promelas||270 (r, m)||Full life-cycle||EPA-OPP||42|
|Chlorothalonil||16.4||82||Osteopilus septentrionalis||10 (r, m)||Survival||||0.96||6.2||Salmo salar||21 (r, m)||Growth||||0.06|
|Chlorpyrifos||10||100||Rana sphenocephala||12 (r, n)||Growth||||0.12||2.6||P. promelas||200 (f, m)||Full life-cycle||EPA-OPP||0.01|
|Copper||5||25||Rana pipiens||55 (r, n)||Growth||||1.6||10.6||Tilapia guineensis||28 (r, n)||Growth||||0.32|
|Dexamethasone||3.9||39||X. laevis||21 (r, n)||Growth||||42.7||424||P. promelas||29 (f, m)||Growth||||10.9|
|Diazinon||1000||2000||Bufo americanus||16 (r, n)||Growth||||0.916||1.82||P. promelas||34 (f, m)||Early life stage||EPA-OPP||0.001|
|Dichlorvos||150,000||3.75 x 106||Rana hexadactyla||98 (r, n)||Growth||||5.2||10.1||O. mykiss||61 (f, m)||Early life stage||EPA-OPP||0.00003|
|Dieldrin||0.8||1.8||X. laevis||24 (f, m)||Survival||||0.55||0.95||O. mykiss||90 (f, m)||Growth||||0.69|
|Dimethoate||1,000||3,200||X. laevis||100 (r, n)||Survival||||430||840||O. mykiss||96 (f, m)||Early life stage||EPA-OPP||0.43|
|Dinitro o cresol||320||1,000||X. laevis||100 (r, n)||Development||||100||320||O. latipes||40 (r, n)||Survival||||0.31|
|Diuron||7,600||14,500||Rana aurora||14 (r, m)||Growth||||26.4||61.8||P. promelas||60||Early-life stage||EPA-OPP||0.003|
|Endosulfan||1||10||A. barbouri||50 (f, m)||Growth||||0.2||0.4||P. promelas||280 (f, m)||Full life-cycle||EPA-OPP||0.2|
|Ethinylestradiol||0.002||-||X. tropicalis||49 (f, m)||Development||||0.0002||0.0008||P. promelas||301 (f, m)||Growth||||0.09|
|Ethylenethiourea (ETU)||5,000||10,000||X. laevis||28 (r, n)||Development||||32,000a||100,000||O. mykiss||60 (r, n)||Growth||||6.4|
|Flutamide||210||-||Xenopus tropicalis||210 (f, m)||Growth||||62.7||651||P. promelas||21 (m)||Reproduction||||0.30|
|Glufosinate (as SL formulation)||1,000||-||Spea multiplicata||30 (r, n)||Growth||||92.5||185||O. mykiss||21 (f, m)||Juvenile growth||EFSA - DAR||0.09|
|Glyphosate (as RoundupTM)||3,960b||19,840b||R. sylvatica||16 (r, n)||Survival||||2,400||7,000||O. mykiss||21 (f, m)||Survival||||0.61|
|Malathion||1,000||5,000||B. americanus||16 (r, n)||Survival||||21||44||O. mykiss||97 (f, m)||Early life stage||EPA-OPP||0.02|
|Molinate||187.3||9,365||Bombina orientalis||13 (r, n)||Survival||||390||830||O. mykiss||60 (f, m)||Early life stage||EPA-OPP||2.08|
|Nonylphenol||25||50||X. laevis||14 (r, n)||Development||||0.65||8.1||P. promelas||21 (m)||Reproduction||||0.03|
|Pentachlorophenol||5||10||X. laevis||14 (r, n)||Development||||15||39||Etheostoma fonticola||30 (f, m)||Growth||||3.0|
|Perfluorooctane sulfonate, potassium salt||1,000||3,000||R. pipiens||112 (f, m)||Development||||50||250||Danio rerio||70 (r, n)||Growth||||0.05|
|Piperonyl butoxide||10,000||-||X. laevis||10 (r, n)||Survival||||230||480||P. promelas||35 (f, m)||Early life stage||EPA-OPP||0.02|
|p-Nitro toluene||3,200||10,000||X. laevis||100 (r, n)||Development||||1,000||3,200||O. latipes||40 (r, n)||Survival||||0.31|
|p-Nonylphenol||100||-||R. pipiens||124 (r, n)||Growth||||10||100||D. rerio||240 (r, n)||Reproduction||||0.1|
|p-Octylphenol||206.3||-||R. pipiens||77 (r, n)||Growth||||12||35||D. rerio||78 (f, m)||Growth||||0.06|
|Potassium dichromate||1,000||32,000||X. laevis||100 (r, n)||Survival||||4,580||49,490||Lepomis macrochirus||182 (r, m)||Growth||||4.58|
|Propoxur||1,000,000||-||R. hexadactyla||98 (r, n)||Development||||1,400||3,020||O. mykiss||28||-||IUCLID||0.001|
|Propylthiouracil (PTU)||2,500||5,000||X. laevis||21 (m)||Growth||||2,500||5,000||D. rerio||35 (f, n)||Growth||||1.0|
|Sodium bromide||32,000||100,000||X. laevis||100 (r, n)||Survival||||100,000||320,000||Poecilia reticulata||28 (r, n)||Survival||||3.13|
|Sodium nitrate||3,965||-||Hyla chrysoscelis||42 (r, m)||Development||||1,600||6,250||Salvelinus namaycush||120 (r, m)||Growth||||0.40|
|Sodium perchlorate||33.49||70.42||X. laevis||66 (r, m)||Development||||7,310||99,250||Gambusia holbrooki||56 (r, m)||Reproduction||||218|
|Tetrabromobisphenol A (TBBPA)||250||500||X. laevis||21 (n)||Development||||160||310||P. promelas||35||Survival||||0.64|
|Tetrapropylene benzene sulphonate||3,200||10,000||X. laevis||100 (r, n)||Survival||||3,200||10,000||Oryzias latipes||40 (r, n)||Survival||||1.0|
|Triclosan||29.6||-||X. laevis||32 (f, m)||Growth||||15.1||40||O. mykiss||35 (f, n)||Survival||||0.51|
|Zinc||2,000||-||Bufo fergusonii||22 (n)||Development||||26||51||Jordanella floridae||30 (f, m)||Growth||||0.01|
Amphibians were between 10-fold and 100-fold more sensitive than fish in a total of five cases (azoxystrobin, acetochlor, ammonium perchlorate, carbaryl, and dexamethasone) and greater than 100-fold more sensitive than fish on a single occasion (sodium perchlorate). However, when potentially unreliable NOECs were excluded from the analysis (i.e., those for azoxystrobin, acetochlor, and ammonium perchlorate; see below), amphibians were between 10-fold and 100-fold more sensitive than fish on only two occasions (carbaryl and dexamethasone) and greater than 100-fold more sensitive than fish on only a single occasion (sodium perchlorate). The influence of including unreliable NOECs on the outcome of the analysis is shown in Figure 2.
The analysis presented here shows that fish and amphibian toxicity data are highly correlated and that fish are generally more sensitive than amphibians (acute and chronic). Relatively few instances were seen of amphibians showing marked higher sensitivity than fish (i.e., ≥10-fold), after either acute or chronic exposure. These cases are discussed in detail in the following sections.
The criteria applied to the acute ECOTOX dataset for measured concentrations to be reported potentially resulted in fewer acute comparisons than may have been expected for an analysis of this type. However, the reduced uncertainty associated with this refined acute dataset is likely to outweigh the potential benefits of a larger dataset, which may include unreliable studies. Because Aldrich  did not require studies to be based on measured concentrations, some amphibian data based on nominal concentrations were included.
In this analysis, amphibians were between 10-fold and 100-fold more sensitive than fish in comparisons for aluminum chloride (25-fold), 2,4 dichloroaniline (12-fold), malathion (34-fold), and pentachlorophenol-sodium salt (12-fold). In all of these cases, application of the standard EU assessment factor of 100 to the regulatory fish LC50 value would have covered the amphibian LC50 value, and these results are therefore not discussed further.
The only two comparisons of acute LC50 values that suggested greater than 100-fold sensitivity for amphibians were those for p-nonylphenol (2,111-fold) and dimethoate (7,300-fold). The acute amphibian data for p-nonylphenol were taken from a paper by Bridges et al. , in which southern leopard frog (Rana sphenocephala) tadpoles were exposed to measured concentrations of p-nonylphenol for 96 h. They reported a 96-h LC50 of 0.34 µg/L, which is much lower than the corresponding fish geometric mean value from ECOTOX of 718 µg/L (560–920 µg/L), which is based on two 96-h LC50 values reported by Ernst et al. . However, along with data on amphibian sensitivity to p-nonylphenol, Bridges et al.  also cite a rainbow trout 96-h LC50 value for p-nonylphenol of 0.19 µg/L and a fathead minnow 96-h LC50 value for p-nonylphenol of 0.27 µg/L from U.S. EPA . They also cite a boreal toad (Bufo boreas) tadpole 96-h LC50 for p-nonylphenol of 0.12 µg/L . These additional data suggest that there is in fact little difference in the acute toxicity of p–nonylphenol to amphibians and fish. This is further supported by the chronic data comparison for p–nonylphenol, which shows that fish are at least 10 times more sensitive than amphibians (see Table 1).
In our analysis and in the analysis by Aldrich , the acute amphibian LC50 for dimethoate was taken from a paper by Khangarot et al. , in which Rana hexadactyla tadpoles were exposed to formulated dimethoate (Rogor 30 EC). The reported 96-h LC50 is 7.82 ppb Rogor 30 EC (nominal concentration), which would correspond to 2.4 µg/L dimethoate active substance, although Khangarot et al.  are not clear on the expression of endpoints in units of formulation or active substance. However, other data sources suggest that amphibians and fish are similarly sensitive to dimethoate. For example, the Environment Agency of England and Wales  collated acute and chronic data for dimethoate, and the value reported by Khangarot et al.  is again identified as the most sensitive amphibian result (however, the study is considered not reliable, because of missing analytical values and insufficient description of the test methodology). For fish, the Environment Agency report also identified a low 96-h LC50 value for mullet exposed to dimethoate of 2.3 µg/L from a paper by Aboul-Eta and Khalil , which was also considered not reliable, because of missing chemical analysis. Furthermore, the chronic data pair for dimethoate shows a higher sensitivity of fish (see Table 1). Therefore, the weight of evidence suggests that fish and amphibians are similarly sensitive, rather than amphibians being significantly more sensitive than fish.
After additional scrutiny of the two instances of greater than 100-fold amphibian sensitivity initially identified in this analysis, in general, amphibians are of comparable acute sensitivity to fish in laboratory toxicity studies and, on average, slightly less sensitive than fish (Fig. 1). This finding was conspicuous despite the stringent data selection criteria applied in the present study, which allowed any amphibian species to be included in the analysis, whereas fish data were limited to only two species (O. mykiss and P. promelas) which potentially excluded more sensitive fish species, as demonstrated in the case of dimethoate.
Other authors have found that amphibians are, in general, either equally or less sensitive than fish to acute chemical exposures [3-13, 29]. For example, Hoke and Ankley  found that LC50 values from the frog embryo teratogenesis assay-Xenopus (FETAX) were not the most sensitive result when compared with other acute toxicity endpoints from traditional aquatic test species (including fish) for Cd, Cu, Se, Hg, Zn, ammonia, aniline, pentachlorophenol, atrazine, malathion, and parathion. The only exception was aluminum, for which FETAX was more sensitive, but only by a factor of approximately 2. Bridges et al.  report comparative 96-h LC50 results for southern leopard frog tadpoles (Rana sphenocephala), boreal toad tadpoles (Bufo boreas), bluegill sunfish (Lepomis macrochirus), fathead minnow, and rainbow trout for five chemicals (p-nonylphenol, carbaryl, Cu, pentachlorophenol, and permethrin). The most sensitive tadpole and fish results were within a factor of 2 for all five chemicals.
Therefore, the overall picture from this analysis and other reviews is that fish are generally more acutely sensitive than amphibians, and the standard EU risk assessment factor of 100 accounts for potential species sensitivity differences between fish and amphibians.
Further details of the studies with potentially unreliable NOECs that suggest that amphibians are more sensitive than fish are provided in the following sections.
The azoxystrobin amphibian NOEC of 10 µg/L was taken from a study by Johannsson et al.  in which common frogs (Rana temporaria) were exposed to 1 or 10 µg/L azoxystrobin. No effects were observed at the highest test concentration of 10 µg/L. This unbounded NOEC cannot be regarded as a true NOEC and is therefore an unreliable basis for comparison with the fish NOEC of 147 µg/L taken from the OPP database. Equally, the amphibian NOEC for acetochlor, which was reported by Helbing et al. , is based on exposure of Rana catesbeiana tadpoles to either 1 or 10 nM acetochlor. At 10 nM (equivalent to 2.7 µg/L, based on a molecular weight of 269.77 g/mol), no effects on apical endpoints (forelimb emergence, mouth development, or tail regression) were seen. This again is an unbounded NOEC and an unreliable basis for comparison with the fish NOEC of 130 µg/L taken from the OPP database. In both of these instances, the true amphibian NOEC will likely be higher than the reported NOEC initially used for the comparisons, and this therefore does not provide evidence of greater amphibian sensitivity to these chemicals, despite the ratios calculated in this analysis.
The amphibian NOEC of 59 µg/L, for effects on development of X. laevis exposed to ammonium perchlorate, is based on a paper by Goleman et al. . This study employed a very large gap between exposure concentrations with a corresponding LOEC of 14,000 µg/L. The large spacing factor between the NOEC and LOEC means that this result is insufficient for a reliable comparison of amphibian and fish sensitivity to ammonium perchlorate, as the true NOEC could span more than two orders of magnitude between the reported NOEC and LOEC values. Furthermore, this study was conducted in FETAX medium, which may have contributed to an artificially high sensitivity (see later discussion on sodium perchlorate). The corresponding fish NOEC, based on a study by Crane et al. , is reported as 1,000 µg/L, which is between the NOEC and LOEC for amphibians as reported by Goleman et al. . In this case, the maximum acceptable toxicant concentration ratio is 3.48 (compared with the NOEC ratio of 16.9), indicating that the observed sensitivity difference is at least partly determined by experimental design and may in fact be much smaller.
After removing the above data, amphibian NOECs were more than 10 times more sensitive than corresponding fish NOECs for three substances: carbaryl (42-fold), dexamethasone (11-fold), and sodium perchlorate (218-fold).
The amphibian NOEC for carbaryl is based on results from a paper by Rohr et al. . Streamside salamander (Ambystoma barbouri) embryos were exposed to carbaryl at 0.5, 5, or 50 µg/L. Survival was reduced at 50 µg/L by approximately 20%, which is why an NOEC for survival of 5 µg/L is reported. The corresponding fish NOEC of 210 µg/L stems from a fathead minnow early life-stage test taken from the OPP database. The maximum acceptable toxicant concentration ratio is 10.6 (compared with the NOEC ratio of 42), indicating that the observed sensitivity difference is at least partly determined by experimental design and may in fact be much smaller. The U.S. EPA Office of Prevention, Pesticides, and Toxic Substances reviewed the relative acute and chronic toxicity of carbaryl to amphibian and fish species, including the studies already discussed, as part of a Pesticide Effects Determination of the risks of carbaryl use to the federally listed endangered Barton Springs salamander (Eurycea sosorum ). This assessment concluded that none of the amphibian toxicity data available in the open literature was sufficiently robust to use for quantitative risk assessment. They also report that the available evidence suggests that amphibians are less sensitive to carbaryl than the most sensitive fish species for which data were available (i.e., Atlantic salmon, Salmo salar, with an NOEC of 6.8 µg/L).
The amphibian NOEC for the corticosteroid drug dexamethasone is derived from a study by Lorenz et al. , who exposed X. laevis tadpoles (stage 51) to 10, 100, or 500 nM dexamethasone for 21 d. They reported effects on hindlimb growth at 100 nM, leading to identification of a NOEC of 10 nM from their study, which is 3.9 µg/L (based on a molecular weight of 392.5 g/mol). The corresponding fish value is a growth NOEC of 42.7 µg/L (the LOEC is 424 µg/L) from a 29-d fish early life-stage test with fathead minnow . The NOEC ratio is 11, with a similar maximum acceptable toxicant concentration ratio, as both studies employed a (relatively large) spacing factor of 10. Both of these studies seem to have been performed and reported adequately, so amphibians may be more sensitive than fish to dexamethasone. Lorenz et al.  suggest that the effects that they measured in Xenopus are likely to have resulted from complex mechanisms involved in the modulatory actions of corticosteroids on amphibian metamorphosis and thyroid hormones, via prolactin synthesis. This biochemical pathway does not occur in fish species, which could explain the apparent difference in sensitivity.
The amphibian NOEC for sodium perchlorate is derived from a study by Brausch et al. , who investigated the relative sensitivity of X. laevis to perchlorate (as sodium perchlorate) in natural stream water and synthetic (FETAX) test media. Embryos (Nieuwkoop-Faber Stage 11) of X. laevis were exposed in the laboratory to measured concentrations of perchlorate in either FETAX medium (33.5 and 70.4 µg/L) or natural stream water (29.3 and 98.8 µg/L) for 66 d. The NOEC and LOEC values of 33.5 µg/L and 70.4 µg/L were derived for developmental effects on metamorphosis and hindlimb length, respectively, in FETAX media. However, no statistically significant effects were observed in X. laevis exposed to perchlorate in natural stream water. In a separate experiment, Brausch et al.  report that exposure of juvenile (5-d-old) S. multiplicata to perchlorate in natural stream water at measured concentrations of 50.1, 107.1, and 1,038 µg/L for 42 d resulted in no statistically significant effects on either metamorphosis or survival. The authors conclude that natural surface water mitigates the anti-metamorphic effect of perchlorate in X. laevis and is likely to contribute to the lack of effects observed in S. multiplicata. The precise mechanism by which perchlorate effects are modified by test media has not been established, but it may be that natural surface waters contain sufficient iodide or some other water quality characteristic that mitigates the anti-metamorphic effects in amphibians. Similar contrasting observations on perchlorate toxicity (high in FETAX medium, low or absent in natural water) have been reported by Carr et al. . Goleman and Carr  report a NOEC for sodium perchlorate for X. laevis exposed in FETAX medium of 22.6 µg/L, which is similar to the NOEC in FETAX medium reported by Brausch et al. . The Goleman and Carr study was not selected as a key NOEC for the present study, because the spacing factor between test concentrations was greater than 100. However, it and other studies  provide supporting evidence for the NOEC reported by Brausch et al.  in FETAX medium. Perchlorate is a potent inhibitor of the uptake of iodide into the thyroid, reducing the formation of thyroid hormones that control amphibian metamorphosis [91, 108]. The specific mechanism of perchlorate toxicity may explain the relative sensitivity of amphibians to this substance, although the environmental relevance of laboratory studies on perchlorate using FETAX medium is uncertain. The observed high sensitivity of Xenopus to perchlorate in FETAX media could be a potential artifact of the medium. Garber  also reported on deficiencies of the FETAX medium, which hinders tadpole metamorphosis. Because fish are usually tested in natural well water or dechlorinated tap water, such artifacts would not occur under chronic testing conditions for fish. Finally, after 40 weeks of exposure to measured concentrations of perchlorate of up to 1,500 µg/L in natural water, X. tropicalis did not show any effects on apical endpoints. Hence the NOEC was at least 1,500 µg/L . Thus, in comparable media no difference is found between fish and amphibians in their chronic sensitivity to perchlorate.
Several of the chronic studies with amphibians that are evaluated here have been conducted to study the effects of chemicals on the hypothalamic–pituitary–thyroid axis; that is, they specifically investigate the influence on amphibian development of an endocrine mode of action. To characterize such effects, the amphibian metamorphosis assay with X. laevis (Amphibian Metamorphosis Assay, OECD test guideline 231) was developed. For other endocrine modes of action, such as (anti-)estrogenic or (anti-)androgenic mechanisms, regulatory concerns for aquatic vertebrates are usually addressed through testing with fish. Nevertheless, the present analysis of chronic data shows that even for chemicals that are known to influence the hypothalamic–pituitary–thyroid axis in amphibians, such as ethylenethiourea and propylthiouracil, fish apical endpoints are comparably sensitive (the only possible exception being the drug dexamethasone). Hence, fish data are appropriate to cover the chronic risks of such substances to amphibians. The potential to study specific thyroid-mediated effects in fish, including an assessment of thyroid histology and sensitivity considerations, was explored for propylthiouracil by Schmidt and Braunbeck , sodium perchlorate by Mukhi and Patino , and ammonium perchlorate by Crane et al. . This demonstrates the potential to use fish higher-tier toxicity tests to address the risk of substances active on the hypothalamic–pituitary–thyroid endocrine axis.
Kerby et al.  found that amphibians were highly acutely sensitive to the three phenolic chemicals that they analyzed (triclosan and two unidentified others). In our analysis of chronic data, amphibians were not more sensitive than fish to the phenolic substances 17-β–estradiol, 4-tert-octylphenol, 17-α-ethinylestradiol, bisphenol A, p-nonylphenol, nonylphenol, and triclosan. Amphibians were slightly more sensitive to pentachlorophenol. However, in this instance a standard assessment factor of 10 would be sufficient to cover the observed difference in sensitivity. Our analysis therefore does not suggest a need for additional amphibian testing for phenolic substances.
Therefore, the overall picture from this analysis and other reviews is that amphibians are not generally more chronically sensitive than fish and the standard EU risk assessment factor of 10 accounts for potential species sensitivity differences between fish and amphibians.
This analysis of acute and chronic data demonstrates that in most cases fish are more sensitive than amphibians to chemicals and thus are appropriate representative species to cover the sensitivity of aquatic vertebrates in current risk assessment procedures. Nearly all of the apparent exceptions identified in this analysis appear to be artifacts of either study design or data selection. Thus, careful checks on the quality and reliability of data are required before use, to avoid drawing erroneous conclusions. Only those substances that specifically interfere with biochemical pathways involved in amphibian metamorphosis may not be detected when using fish as surrogates (the corticosteroid drug dexamethasone may be an example), although several other compounds known to influence amphibian metamorphosis through thyroid-mediated toxicity included in the analysis were not highlighted, suggesting that substances that interfere with conserved developmental mechanisms affect amphibians and fish similarly. In common with most other comparative studies, no evidence suggests that amphibians are routinely more sensitive than fish to chemicals. Finally, other taxonomic groups, such as primary producers and invertebrates, may drive an aquatic risk assessment, especially for substances with herbicidal (e.g., acetochlor) and insecticidal (e.g., carbaryl) modes of action, respectively. These analyses and other reviews support the notion that additional aquatic amphibian testing is not necessary and that data generated using the standard aquatic test species, including acute and chronic fish data during chemical risk assessment, will satisfy requirements for both robust aquatic risk assessment and minimization of vertebrate animal testing.
Table S1. (47 KB DOC).
This study was funded by Syngenta Ltd and BASF SE, and some of the substances reviewed here are products of these companies. This paper benefitted from the constructive comments of J. Brausch, A. Hosmer, P. Dohmen, C. Bögi, and two anonymous referees.