Environmental Toxicology and Chemistry, Vol. 34, No. 12, pp. 2864–2869, 2015 # 2015 SETAC Printed in the USA
Short Communication IT IS TIME TO DEVELOP ECOLOGICAL THRESHOLDS OF TOXICOLOGICAL CONCERN TO ASSIST ENVIRONMENTAL HAZARD ASSESSMENT SCOTT E. BELANGER,y HANS SANDERSON,z MICHELLE R. EMBRY,*x KATIE COADY,k DICK DEZWART,# BRIANNA A. FARR,x STEVE GUTSELL,yy MARLIES HALDER,zz ROBIN STERNBERG,§§ and PETER WILSONkk yProcter & Gamble, Cincinnati, Ohio, USA zAarhus University, Roskilde, Denmark xInternational Life Sciences Institute Health and Environmental Sciences Institute, Washington, DC, USA kDow Chemical Company, Midland, Michigan, USA #National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands yyUnilever, Colworth Science Park, Sharnbrook, United Kingdom zzEuropean Commission, Joint Research Centre, Ispra, Italy xxUS Environmental Protection Agency, Washington, DC, USA kkSanofi, Bridgewater, New Jersey, USA (Submitted 9 March 2015; Returned for Revision 16 April 2015; Accepted 23 June 2015) Abstract: The threshold of toxicological concern (TTC) concept is well established for assessing human safety of food-contact substances and has been reapplied for a variety of endpoints, including carcinogenicity, teratogenicity, and reproductive toxicity. The TTC establishes an exposure level for chemicals below which no appreciable risk to human health or the environment is expected, based on a de minimis value for toxicity identified for many chemicals. Threshold of toxicological concern approaches have benefits for screening-level risk assessments, including the potential for rapid decision-making, fully utilizing existing knowledge, reasonable conservativeness for chemicals used in lower volumes (low production volume chemicals (e.g., < 1 t/yr), and reduction or elimination of unnecessary animal tests. Higher production volume chemicals (>1 t/yr) would in principle always require specific information because of the presumed higher exposure potential. The TTC approach has found particular favor in the assessment of chemicals used in cosmetics and personal care products, as well as other chemicals traditionally used in low volumes. Use of the TTC in environmental safety is just beginning, and initial attempts are being published. Key questions focus on hazard extrapolation of diverse taxa across trophic levels, importance of mode of action, and whether safe concentrations for ecosystems estimated from acute or chronic toxicity data are equally useful and in what contexts. The present study provides an overview of the theoretical basis for developing an ecological (eco)-TTC, with an initial exploration of chemical assessment and boundary conditions for use. An international collaboration under the International Life Sciences Institute Health and Environmental Sciences Institute has been established to address challenges related to developing and applying useful eco-TTC concepts. Environ Toxicol Chem 2015;34:2864–2869. # 2015 SETAC Keywords: Threshold of toxicological concern (TTC)
Risk assessment
Animal alternatives
carcinogenicity, teratogenicity, and reproductive toxicity, albeit with different databases for each endpoint, with endpointspecific TTC values as the result [3]. Thresholds for toxicological concern have benefits for screening-level risk assessments, including the potential for rapid decision-making, fully utilizing existing knowledge, reasonable conservativeness for chemicals used in lower volumes, and reduction or elimination of unnecessary animal tests. Consequently, TTCs have found particular favor in the assessment of chemicals used in cosmetics and personal care products or other chemicals traditionally used in low volumes. It is used with respect to human health screeninglevel risk assessments for indirect contact substances at low concentrations and in products in which levels may be coincident with low production volumes [4]. Approaches using TTCs have only recently been explored in environmental assessments. De Wolf et al. [5] applied the concept to address environmental thresholds of no toxicological concern (ETNC) in freshwater systems (which they termed ETNCaq) for organic chemicals. Modes of action, based on Verhaar classifications formed on chemical structure [6,7], were used to group chemicals in the European Centre for Ecotoxicology and Toxicology of Chemicals Aquatic Hazard Assessment database supplemented by additional publicly available data (Table 1). Lowest and 95th percentile hazardous concentration values (HC5) were then calculated for each group
INTRODUCTION
Risk assessment of chemicals inherently involves an assessment of toxicity, exposure, and the resulting likelihood or probability of observing an adverse response. Furthermore, it requires ethical and resource considerations regarding how much data are attainable and should be derived (e.g., via use of animal testing) versus what is considered an acceptable level of extrapolation [1]. Hence, a pragmatic optimum is sought after in tiered risk assessment methodology. A recent addition to the methodology is the concept of the threshold for toxicological concern (TTC). The TTC establishes an exposure level for chemicals below which no appreciable risk to human health or the environment is expected, based on a de minimis value for toxicity identified for many chemicals. This level can then be compared with an estimate of the likely exposure to a chemical to complete a screening-level safety assessment for a given route of exposure or environmental compartment/species of concern. The TTC concept is well established for assessing human safety of indirect food-contact substances [2] and has been reapplied for a variety of endpoints, including * Address correspondence to [email protected] Published online 25 June 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/etc.3132 2864
AChE inhibitors, surfactants, US EPA ECOTOX, US EPA OPP db
Review, multiple sources
Williams et al. [9] (CTD)
Mons et al. [16] (TTC)
Algae, invertebrates, fish, acute and chronic
Reproduction for aquatic vertebrates, crustaceans, and mollusks
MOA assigned by ASTER
5% of cumulative distribution
Cumulative Weibull and log-normal distributions per taxon, class; 1% and 5% thresholds with no AF
PNECs derived from the 5% of the cumulative distribution, EU TGD (2003) and REACH TGD (2008) for AFs and data quality assessment
Combination of probabilistic and deterministic by MOA, ACR of 10 for MOA1 and 2: typical EU AFs following calculation of 95th percentile, 50% confidence interval, used lower 95th percentile confidence limit for assessment Cumulative NOEC distributions analyzed by RIVM’s ETX2.0 software.
Techniques applied for purposes of extrapolation
Not eco-TTC per se Distribution of data by test species Implications for SSDs and extrapolation
stances, metals, organometals (and others)
Distributions suggest exclusion of bioaccumulative sub-
included
Surfactant distributions with embedded SAR effects
result in a lower eco-TTC
Questioned appropriateness of standard tests Results comparable to [5] Cationic chemicals should be analyzed separately as they
quality of data
Endocrine MOA only Unknown need, by regulators, of amount, type, and
Water solubility not addressed Questions re: db quality
Additional notes
a See reference for complete information on database descriptions. ETNC ¼ ecological threshold of no concern; TTC ¼ toxicological threshold of concern; CTD ¼ chemical toxicity distribution; SSD ¼ species sensitivity distribution; db ¼ database; FHM ¼ fathead minnow; ECETOC AHA ¼ European Centre for Ecotoxicology and Toxicology of Chemicals Aquatic Hazard Assessment; USEPA ¼ US Environmental Protection Agency; EU ¼ European Union; MOA ¼ mode of action; ACR ¼ acute to chronic ratio; AF ¼ application factor; US FDA ¼ US Food and Drug Administration; EDKB ¼ Endocrine Disruptor Knowledge Base; UK RE ¼ UK Risk Evaluation Reports; NOEC ¼ no-observed-effect concentration; PNEC ¼ predicted no-effect concentration; TGD ¼ technical guidance document; REACH ¼ European Commission’s Registration, Evaluation, Authorisation, and Restriction of Chemicals; AChE: acetylcholinesterase; SAR ¼ structure activity relationships; ASTER ¼ Assessment Tools for the Evaluation of Risk.
US EPA Web-ICE Dutch eToxBase AQUIRE
Human health focus for drinking water; compounds covered were genotoxic or carcinogenic Acute toxicity
Internal data: polymers, surfactants, fragrances, inorganics, organometallics, organics
Gutsell et al. [10] (eco-TTC)
Hendricks et al. [17] (SSD)
Acute toxicity to invertebrates, fish
Endocrine estrogenic MOA USEPA ECOTOX US FDA EDKB UK RE
Gross et al. [8] (TTC)
Divided into Verhaar MOA groups; assessment driven by most sensitive group
ECETOC AHA db US EPA FHM db EU Existing substances db Utrecht guppy db
De Wolf et al. [5] (ETNC)
Endpoint coverage
Chemical coverage as described by the authorsa
Reference and term applied to use
Table 1. Overview of research to derive ecological thresholds of toxicological concern
Development of ecological thresholds of toxicological concern Environ Toxicol Chem 34, 2015 2865
2866
Environ Toxicol Chem 34, 2015
of chemicals. The ETNC values for Verhaar modes of action 1 to 3 were approximately 0.1 mg/L; however, further stratification of the data indicated that mode of action 1 and mode of action 2 alone would be higher. Gross et al. [8] developed a case study on the application of TTCs to endocrine-active substances with an estrogenic mode of action. Cumulative rank and frequency distribution analysis (similar to species sensitivity distribution determinations) were applied to several datasets. The largest (n ¼ 22) and least variable dataset on which to base a TTC for estrogen agonists (excluding steroidal estrogens) for all endpoints and the most sensitive taxon resulted in an HC5 of 13.79 ng/L. The authors proposed application of an assessment factor of 5 to the HC5, which produced a threshold value of 2.8 ng/L [8]. No individual no-observed-effect concentration in the nonsteroidal estrogen dataset fell below this value. A slightly different take on this concept was developed by Williams et al. [9] and termed chemical toxicity distributions. The authors applied the approach to look at sensitivity distributions of particular test organisms to selected categories of compounds. The intent was to ascertain a lower value on which to build a read-across case in the context of the European Commission’s Registration, Evaluation, Authorisation, and Restriction of Chemicals (REACH). The authors explored the development of chemical toxicity distributions for acetylcholinesterase inhibitors and several classes of surfactants as representatives of specifically acting and general industrial (nonspecifically acting) modes of action. The approach used differed from that of De Wolf et al. [5] in that distributions were for specific test species as opposed to the generally utilized algae, invertebrate, and fish. More recently, Gutsell et al. [10] presented a TTC-like approach (also termed ecological [eco]TTC) for consumer product chemicals. The method included an exploration of the impact of functional chemical class on the eco-TTC. Data from algal, invertebrate, and fish acute and chronic studies were folded into the analysis. Eco-TTC values varied by approximately 2 orders of magnitude, depending on chemical class, and were mostly comparable with the determinations from De Wolf et al. [5]. An exception to this was the cationic chemical class, which gave rise to a significantly lower eco-TTC value. Gutsell et al. [10] importantly concluded that it is possible to conduct a screeninglevel environmental risk assessment of chemicals that may have poorly defined class assignments in terms of mode of action. A comprehensive list of known applications of TTC or TTC-like assessments for environmental or ecological endpoints is given in Table 1. DEVELOPING ECO-TTCS
We propose that now is a suitable time to develop a critical assessment of the potential scope, utility of application, and limitations of eco-TTCs, with an eye toward potential regulatory application. Data availability from international chemical management programs provide unprecedented insight into ecotoxicity; regulatory needs continue to advance toward assessing the universe of chemicals, including those of very low volumes; and greater computing power allows probing of sophisticated and complex chemical information data. Various statistical and methodological approaches have been undertaken in the early forays of eco-TTC development by environmental risk scientists. Data availability and quality verification are always considered to be key aspects, but additional methodological needs exist. The process of deriving TTCs (ecological or human) is fairly well established. Given a
S.E. Belanger et al.
coherent, quality dataset, each chemical can be categorized with respect to mode of action, chemical functional use, or chemical category. Each of these classifications can be a complex decision because various options to accomplish the classification are available. A hazard decision associated with each chemical and test result is required. This could be based on a predicted no-effect concentration (PNEC) determination for the chemical or simply based on results for a particular test species or trophic level (e.g., algae, invertebrates, or fish). The hazard determinations for chemicals with a given mode of action or chemical category can then be viewed as a statistical distribution (probability density function) of values (Figure 1A). The statistical distribution can be assumed or identified on the basis of the empirical values present, and a small percentile (most often 5%) can be estimated from it. The probability density function is most often converted into a cumulative distribution that looks like an S-shaped regression along with an upper and lower confidence interval that informs the user of the quality and noise in the data (Figure 1B). A common interpretation for ecoTTCs, similar to that already in use for human safety, is that the 5th percentile is an appropriately conservative marker to which exposure concentrations for a chemical matching the same mode of action or chemical category can be compared. In this sense, eco-TTCs will markedly differ in their use and interpretation from chemical-specific environmental hazard assessments or environmental quality standards; eco-TTCs are more generic starting points from which chemical-specific evaluations may begin. The hazard data used to derive the ecoTTC can vary depending on the goals of the assessment, as indicated in the Introduction. Some eco-TTCs focus on distributions developed for individual species [8,9], whereas others focus on PNECs derived from all available taxa into which appropriate, but varying, application factors were embedded prior to the computation of the eco-TTC [5,10]. A need exists to evaluate and understand patterns of sensitivity for the broad groups of tested organisms, such as photosynthetic microbes (cyanobacteria and algae), invertebrates, and fish in concert with the breadth of information available supporting various uses of application factors [11,12]. A number of possible approaches can be considered; for example, eco-TTCs based on ecosystem PNECs could be established via application of regionally appropriate extrapolation factors applied to datasets of different levels of complexity or completeness. In this regard, one may accept that an appropriate level of uncertainty is already embedded in the assessment. Different regions (e.g., North America, Europe, Asia) having different levels of risk tolerance for datasets of the same level of complexity would have different eco-TTCs as a result. Another aspect of evaluating the potential for eco-TTCs in an environmental risk assessment context would be to categorize chemicals based on a variety of tools such as the mode of action, functional groups present in the molecule, and functional chemical category based on intended use (e.g., surfactant). The development of eco-TTCs is likely initially most useful for compounds that can be extrapolated from established patterns of acute toxicity or that have wellestablished modes of action. Not all chemicals, classes, or categories may be suitable to be covered by an eco-TTC approach; however, this will require technical discussions and research investments to fully explore. Costs associated with developing eco-TTCs in a given chemical domain will probably vary. We believe, for example, that low-volume chemical substances of unknown or variable composition, complex reaction products, and biological materials (UVCBs) would be a
Development of ecological thresholds of toxicological concern
Environ Toxicol Chem 34, 2015
2867
Figure 1. Generic example of an ecological threshold of toxicological concern (eco-TTC) based on a probability density function (A) that is then converted to the cumulative probability distribution (B). Key statistical indicators include the calculated 5th centile and 95% lower confidence limit of the 5th percentile in panel B. The distribution was constructed from a theoretical data set of 19 different chemicals from which the individual chemical-specific predicted no-effect concentrations were derived.
group for which relevant eco-TTCs could be applied based on acute toxicity and at a relatively low cost. It is likely that the potential utility would be for assessments of hazard associated with ingredients in formulated products by private-sector organizations and perhaps in certain regulatory situations. An interesting group would be industrial narcotic compounds, which represent the majority of chemicals in commerce. Deeper investigations to confirm the importance of modes of action, test species influences, and specific functional groups would be challenging and important and would need a moderate research investment. On the other end of the scale from UVCBs would be non-narcotic compounds requiring substantially more mode of action insight that are also supported by primarily chronic toxicity data. We envision a growth in eco-TTCs that could be developed. Efforts to develop eco-TTCs can be envisioned as an everexpanding scope of chemical and species coverage with increasing costs and breadth of research questions requiring answers as coverage expands. Several key challenges for ecoTTC developers are also envisioned (Figure 2). The universe of ecotoxicological data, particularly of standard test guidelinetype studies, will need to be assessed. Data collection, chemical
categorization, and study quality assessment will be a large effort. A rigorous, transparent architecture to organize, probe, and harmonize as much as possible and share information will be exceedingly useful, as with any metadata analysis. Principles of chemistry and ecotoxicology will be applied, as will categorization of chemical uses, which relies heavily on industry knowledge and product applications. For example, different types of mode of action assignments will be used [6,7,13,14] to explore robustness of assignments, and datasets that vary in completeness with respect to included taxa (algae, invertebrates, and fish) leading to different levels of uncertainty in PNEC derivation should also be explored. Eco-TTCs will differ markedly in their use and interpretation from chemicalspecific environmental hazard assessments or environmental quality standards, as eco-TTCs are more generic starting points from which chemical-specific evaluations may begin. A new tool for early tier hazard assessment and identification will result and will have unique benefits compared with existing measures. For example, eco-TTCs can be developed as an alternative to quantitative structure-activity relationships (QSARs) or to assess complex mixtures that have otherwise been problematic for other technical reasons. Eco-TTCs could also be informed
2868
Environ Toxicol Chem 34, 2015
S.E. Belanger et al.
Figure 2. Proposed ecological threshold of toxicological concern (eco-TTC) development process with key challenges identified. HC5 ¼ 5% hazardous concentration; PNEC ¼ predicted no-effect concentration; QSAR ¼ quantitative structure–activity relationship.
by the emerging science of critical body burdens or chemical activity approaches; however, data gaps may be problematic for this to be realized in the short term. We expect environmental risk assessors in industry and the public sector (i.e., regulators) to be the direct beneficiaries of developing a consensus approach and putative eco-TTCs for application to previously untested compounds. In addition to being helpful as a screening hazard assessment approach, ecoTTCs could also be helpful in weight-of-evidence assessments for existing compounds or chemical categories that may be controversial or less well understood, including those for which QSARs appear to be presently unattainable; or, when the prediction is questionable, the eco-TTC can be used as a corroborating or refuting line of evidence. If an appropriate and high-quality QSAR is available for a chemical and species of interest, however, more realistic PNECs may be derived and will likely be used instead of an eco-TTC. Identification of compounds that are at the extremes of the PNEC or chemical– species distribution function may be exceptionally helpful in the evaluation of hypotheses of chemical–biological interactions, modes of action, and the like. Private-sector risk assessors may gain an important tool for screening-level hazard assessments for guiding newly developed chemical technologies prior to commercialization in addition to devising inputs to risk assessment for regulatory use. The public will benefit greatly
from approaches such as these by improving smart use of existing information as well as negating or reducing the need for additional animal testing. In human safety, improved animal welfare through the reduction, replacement, or refinement of vertebrate tests associated with the use and acceptance of TTC approaches is profound [15]. CALL FOR AN INTERNATIONAL COLLABORATION
The International Life Sciences Institute Health and Environmental Sciences Institute has established a working group in the Animal Alternatives in Environmental Risk Assessment Technical Committee. The committee is composed of environmental risk assessment scientists representing private industry, government, and academic interests. The group is in the process of collecting data from diverse public-sector (and in some cases private-sector) sources to develop eco-TTCs. The committee intends to clarify boundary conditions of application, probe strengths and weaknesses of various approaches, and provide an overall process that can lead to regulatory use and acceptance if methods prove scientifically sound. We contend that the time is ripe for such an approach, as international chemical management programs proliferate, interest in chemicals of emerging concern gain public prominence, and attention moves toward chemicals that are low volume and subsequently
Development of ecological thresholds of toxicological concern
have fewer data on their environmental effects. Key to this process is engagement of a diverse array of scientists engaged in different industry sectors, regulators and academic scientists. Data availability for an even wider diversity of chemicals and ecotoxicity results, as well as healthy debate about the strengths and limitations of eco-TTCs, would be welcome. Interested scientists can participate directly by joining these existing research efforts, making ecotoxicological data available for others to use and/or incorporate into the assessments, or following the development of these programs and deriving ecoTTCs and sharing them independently. The intent is to expand aquatic hazard inputs to the eco-TTC assessments beyond chemicals that are traditionally employed in research and to widen the array and depths of modes of action that are included. An intentional outcome, in addition to eco-TTC data assessment, will be to increase the likelihood for regulatory acceptance where use can be technically supported. Rigorous assessments of data quality and public transparency of the data, process, and results will be important attributes for regulatory acceptance. Acknowledgment—D. de Zwart received funding from the SOLUTIONS project via the European Union’s Seventh Framework Programme for research, technological development, and demonstration under grant agreement no. 603437. The authors thank the collaborators on this Health and Environmental Sciences Institute project for valuable input and discussions: A. Beasley, R. Hummel, A. Kienzler, and J. Suski. Disclaimer—The views, conclusions, and recommendations expressed in the present article are those of the authors and do not necessarily represent views or policies of the European Commission. This article was prepared by R. Sternberg as part of her official duties for the US Environmental Protection Agency. It has not been formally reviewed by the Agency and does not necessarily reflect the Agency’s views. Data Availability—No actual data have been analyzed in this manuscript; however, all references are publicly available. REFERENCES 1. Solomon KR, Brock TCM, Zwart Dd, Dyer SD, Posthuma L, Richards SM, Sanderson H, Sibley PK, van den Brink PJ. 2008. Extrapolation Practice for Ecotoxicological Effect Characterization of Chemicals. SETAC & CRC, Pensacola, FL, USA. 2. US Food and Drug Administration Center for Food Safety and Applied Nutrition. 1993. Toxicological principles for the safety assessment of direct food additives and color additives used in food, Redbook II. Fed Reg 58:16536. 3. Kroes R, Renwick AG, Cheeseman M, Kleiner J, Mangelsdorf I, Piersma A, Schilter B, Schlatter J, Van Schothorst F, Vos JG, W€ urtzen G. 2004. Structure-based thresholds of toxicological concern (TTC): Guidance for application to substances present at low levels in the diet. Toxicol Sci 42:65–83.
Environ Toxicol Chem 34, 2015
2869
4. European Food Safety Authority Scientific Committee. 2012. Scientific opinion on exploring options for providing advice about possible human health risks based on the concept of threshold of toxicological concern (TTC). EFSA J 10:2750–2853. 5. De Wolf W, Siebel-Sauer A, Lecloux A, Koch V, Holt M, Feijtel T, Comber M, Boeije G. 2005. Mode of action and aquatic exposure thresholds of no concern. Environ Toxicol Chem 24:479–485. 6. Verhaar HJM, Ramos EU, Hermens JLM. 1996. Classifying environmental pollutants. 2: Separation of class 1 (baseline toxicity) and class 2 (‘polar narcosis’) type compounds based on chemical descriptors. J Chemometr 10:149–162. 7. Verhaar HJM, Van Leeuwen CJ, Hermens JLM. 1992. Classifying environmental pollutants. 1: Structure-activity relationships for prediction of aquatic toxicity. Chemosphere 25:471–491. 8. Gross M, Daginnus K, Deviller G, de Wolf W, Dungey S, Galli C, Gourmelon A, Jacobs M, Matthiessen P, Micheletti C, Nestmann E, Pavan M, Paya-Perez A, Ratte HT, Safford B, Sokull-Kl€ uttgen B, Stock F, Stolzenberg HC, Wheeler J, Willuhn M, Worth A, Comenges JMZ, Crane M. 2010. Thresholds of toxicological concern for endocrine active substances in the aquatic environment. Integr Environ Assess Manag 6:2–11. 9. Williams ES, Berninger JP, Brooks BW. 2011. Application of chemical toxicity distributions to ecotoxicology data requirements under REACH. Environ Toxicol Chem 30:1943–1954. 10. Gutsell S, Hodges G, Marshall S, Roberts J. 2015. Ecotoxicological thresholds—Practical aplications in an industrial inventory. Environ Toxicol Chem 34:935–942. 11. Nabholz JV, Miller P, Zeeman M. 1993. Environmental risk assessment of new chemicals under the Toxic Substances Control Act (TSCA) Section Five. In Landis WG, Hughes JS, Lewis MA, eds, Environmental Toxicology and Risk Assessment. ASTM Special Technical Publication 1179. ASTM International, Philadelphia, PA, USA, pp 40–55. 12. European Chemicals Agency. Guidance on information requirements and chemical safety assessment, Chapter R.10: Characterization of dose [concnetration]-response for the environment. In Guidance on Information Requirements and Chemical Safety Assessment. Helsinki, Finland. 13. Russom CL, Bradbury SP, Broderius SJ, Hammermeister DE, Drummond RA. 1997. Predicting modes of toxic action from chemical structure: Acute toxicity in the fathead minnow (Pimephales promelas). 16:948–967. 14. Computer Sciences Corporation. 2012. ASTER User Guide—ASsessment Tools for the Evaluation of Risk, Ver 2.0. US Environmental Protection Agency, Duluth, MN, USA. 15. Hauge-Nilsen K, Keller D. 2014. Feasibility study: Refinement of the TTC concept by additional rules based on in silico and experimental data. Arch Toxicol 89:25–32. 16. Mons MN, Heringa MB, van Genderen J, Puijker LM, Brand W, Van Leeuwen CJ, Stoks P, van der Hoek JP, van der Kooij D. 2013. Use of the threshold of toxicological concern (TTC) approach for deriving target values for drinking water contaminants. Water Res 47:1666– 1678. 17. Hendriks AJ, Awkerman JA, de Zwart D, Huijbregts MAJ. 2013. Sensitivity of species to chemicals: Dose-response characteristics for various test types (LC50, LR50 and LD50) and modes of action. Ecotoxicol Environ Saf 97:10–16.
Short Communication IT IS TIME TO DEVELOP ECOLOGICAL THRESHOLDS OF TOXICOLOGICAL CONCERN TO ASSIST ENVIRONMENTAL HAZARD ASSESSMENT SCOTT E. BELANGER,y HANS SANDERSON,z MICHELLE R. EMBRY,*x KATIE COADY,k DICK DEZWART,# BRIANNA A. FARR,x STEVE GUTSELL,yy MARLIES HALDER,zz ROBIN STERNBERG,§§ and PETER WILSONkk yProcter & Gamble, Cincinnati, Ohio, USA zAarhus University, Roskilde, Denmark xInternational Life Sciences Institute Health and Environmental Sciences Institute, Washington, DC, USA kDow Chemical Company, Midland, Michigan, USA #National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands yyUnilever, Colworth Science Park, Sharnbrook, United Kingdom zzEuropean Commission, Joint Research Centre, Ispra, Italy xxUS Environmental Protection Agency, Washington, DC, USA kkSanofi, Bridgewater, New Jersey, USA (Submitted 9 March 2015; Returned for Revision 16 April 2015; Accepted 23 June 2015) Abstract: The threshold of toxicological concern (TTC) concept is well established for assessing human safety of food-contact substances and has been reapplied for a variety of endpoints, including carcinogenicity, teratogenicity, and reproductive toxicity. The TTC establishes an exposure level for chemicals below which no appreciable risk to human health or the environment is expected, based on a de minimis value for toxicity identified for many chemicals. Threshold of toxicological concern approaches have benefits for screening-level risk assessments, including the potential for rapid decision-making, fully utilizing existing knowledge, reasonable conservativeness for chemicals used in lower volumes (low production volume chemicals (e.g., < 1 t/yr), and reduction or elimination of unnecessary animal tests. Higher production volume chemicals (>1 t/yr) would in principle always require specific information because of the presumed higher exposure potential. The TTC approach has found particular favor in the assessment of chemicals used in cosmetics and personal care products, as well as other chemicals traditionally used in low volumes. Use of the TTC in environmental safety is just beginning, and initial attempts are being published. Key questions focus on hazard extrapolation of diverse taxa across trophic levels, importance of mode of action, and whether safe concentrations for ecosystems estimated from acute or chronic toxicity data are equally useful and in what contexts. The present study provides an overview of the theoretical basis for developing an ecological (eco)-TTC, with an initial exploration of chemical assessment and boundary conditions for use. An international collaboration under the International Life Sciences Institute Health and Environmental Sciences Institute has been established to address challenges related to developing and applying useful eco-TTC concepts. Environ Toxicol Chem 2015;34:2864–2869. # 2015 SETAC Keywords: Threshold of toxicological concern (TTC)
Risk assessment
Animal alternatives
carcinogenicity, teratogenicity, and reproductive toxicity, albeit with different databases for each endpoint, with endpointspecific TTC values as the result [3]. Thresholds for toxicological concern have benefits for screening-level risk assessments, including the potential for rapid decision-making, fully utilizing existing knowledge, reasonable conservativeness for chemicals used in lower volumes, and reduction or elimination of unnecessary animal tests. Consequently, TTCs have found particular favor in the assessment of chemicals used in cosmetics and personal care products or other chemicals traditionally used in low volumes. It is used with respect to human health screeninglevel risk assessments for indirect contact substances at low concentrations and in products in which levels may be coincident with low production volumes [4]. Approaches using TTCs have only recently been explored in environmental assessments. De Wolf et al. [5] applied the concept to address environmental thresholds of no toxicological concern (ETNC) in freshwater systems (which they termed ETNCaq) for organic chemicals. Modes of action, based on Verhaar classifications formed on chemical structure [6,7], were used to group chemicals in the European Centre for Ecotoxicology and Toxicology of Chemicals Aquatic Hazard Assessment database supplemented by additional publicly available data (Table 1). Lowest and 95th percentile hazardous concentration values (HC5) were then calculated for each group
INTRODUCTION
Risk assessment of chemicals inherently involves an assessment of toxicity, exposure, and the resulting likelihood or probability of observing an adverse response. Furthermore, it requires ethical and resource considerations regarding how much data are attainable and should be derived (e.g., via use of animal testing) versus what is considered an acceptable level of extrapolation [1]. Hence, a pragmatic optimum is sought after in tiered risk assessment methodology. A recent addition to the methodology is the concept of the threshold for toxicological concern (TTC). The TTC establishes an exposure level for chemicals below which no appreciable risk to human health or the environment is expected, based on a de minimis value for toxicity identified for many chemicals. This level can then be compared with an estimate of the likely exposure to a chemical to complete a screening-level safety assessment for a given route of exposure or environmental compartment/species of concern. The TTC concept is well established for assessing human safety of indirect food-contact substances [2] and has been reapplied for a variety of endpoints, including * Address correspondence to [email protected] Published online 25 June 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/etc.3132 2864
AChE inhibitors, surfactants, US EPA ECOTOX, US EPA OPP db
Review, multiple sources
Williams et al. [9] (CTD)
Mons et al. [16] (TTC)
Algae, invertebrates, fish, acute and chronic
Reproduction for aquatic vertebrates, crustaceans, and mollusks
MOA assigned by ASTER
5% of cumulative distribution
Cumulative Weibull and log-normal distributions per taxon, class; 1% and 5% thresholds with no AF
PNECs derived from the 5% of the cumulative distribution, EU TGD (2003) and REACH TGD (2008) for AFs and data quality assessment
Combination of probabilistic and deterministic by MOA, ACR of 10 for MOA1 and 2: typical EU AFs following calculation of 95th percentile, 50% confidence interval, used lower 95th percentile confidence limit for assessment Cumulative NOEC distributions analyzed by RIVM’s ETX2.0 software.
Techniques applied for purposes of extrapolation
Not eco-TTC per se Distribution of data by test species Implications for SSDs and extrapolation
stances, metals, organometals (and others)
Distributions suggest exclusion of bioaccumulative sub-
included
Surfactant distributions with embedded SAR effects
result in a lower eco-TTC
Questioned appropriateness of standard tests Results comparable to [5] Cationic chemicals should be analyzed separately as they
quality of data
Endocrine MOA only Unknown need, by regulators, of amount, type, and
Water solubility not addressed Questions re: db quality
Additional notes
a See reference for complete information on database descriptions. ETNC ¼ ecological threshold of no concern; TTC ¼ toxicological threshold of concern; CTD ¼ chemical toxicity distribution; SSD ¼ species sensitivity distribution; db ¼ database; FHM ¼ fathead minnow; ECETOC AHA ¼ European Centre for Ecotoxicology and Toxicology of Chemicals Aquatic Hazard Assessment; USEPA ¼ US Environmental Protection Agency; EU ¼ European Union; MOA ¼ mode of action; ACR ¼ acute to chronic ratio; AF ¼ application factor; US FDA ¼ US Food and Drug Administration; EDKB ¼ Endocrine Disruptor Knowledge Base; UK RE ¼ UK Risk Evaluation Reports; NOEC ¼ no-observed-effect concentration; PNEC ¼ predicted no-effect concentration; TGD ¼ technical guidance document; REACH ¼ European Commission’s Registration, Evaluation, Authorisation, and Restriction of Chemicals; AChE: acetylcholinesterase; SAR ¼ structure activity relationships; ASTER ¼ Assessment Tools for the Evaluation of Risk.
US EPA Web-ICE Dutch eToxBase AQUIRE
Human health focus for drinking water; compounds covered were genotoxic or carcinogenic Acute toxicity
Internal data: polymers, surfactants, fragrances, inorganics, organometallics, organics
Gutsell et al. [10] (eco-TTC)
Hendricks et al. [17] (SSD)
Acute toxicity to invertebrates, fish
Endocrine estrogenic MOA USEPA ECOTOX US FDA EDKB UK RE
Gross et al. [8] (TTC)
Divided into Verhaar MOA groups; assessment driven by most sensitive group
ECETOC AHA db US EPA FHM db EU Existing substances db Utrecht guppy db
De Wolf et al. [5] (ETNC)
Endpoint coverage
Chemical coverage as described by the authorsa
Reference and term applied to use
Table 1. Overview of research to derive ecological thresholds of toxicological concern
Development of ecological thresholds of toxicological concern Environ Toxicol Chem 34, 2015 2865
2866
Environ Toxicol Chem 34, 2015
of chemicals. The ETNC values for Verhaar modes of action 1 to 3 were approximately 0.1 mg/L; however, further stratification of the data indicated that mode of action 1 and mode of action 2 alone would be higher. Gross et al. [8] developed a case study on the application of TTCs to endocrine-active substances with an estrogenic mode of action. Cumulative rank and frequency distribution analysis (similar to species sensitivity distribution determinations) were applied to several datasets. The largest (n ¼ 22) and least variable dataset on which to base a TTC for estrogen agonists (excluding steroidal estrogens) for all endpoints and the most sensitive taxon resulted in an HC5 of 13.79 ng/L. The authors proposed application of an assessment factor of 5 to the HC5, which produced a threshold value of 2.8 ng/L [8]. No individual no-observed-effect concentration in the nonsteroidal estrogen dataset fell below this value. A slightly different take on this concept was developed by Williams et al. [9] and termed chemical toxicity distributions. The authors applied the approach to look at sensitivity distributions of particular test organisms to selected categories of compounds. The intent was to ascertain a lower value on which to build a read-across case in the context of the European Commission’s Registration, Evaluation, Authorisation, and Restriction of Chemicals (REACH). The authors explored the development of chemical toxicity distributions for acetylcholinesterase inhibitors and several classes of surfactants as representatives of specifically acting and general industrial (nonspecifically acting) modes of action. The approach used differed from that of De Wolf et al. [5] in that distributions were for specific test species as opposed to the generally utilized algae, invertebrate, and fish. More recently, Gutsell et al. [10] presented a TTC-like approach (also termed ecological [eco]TTC) for consumer product chemicals. The method included an exploration of the impact of functional chemical class on the eco-TTC. Data from algal, invertebrate, and fish acute and chronic studies were folded into the analysis. Eco-TTC values varied by approximately 2 orders of magnitude, depending on chemical class, and were mostly comparable with the determinations from De Wolf et al. [5]. An exception to this was the cationic chemical class, which gave rise to a significantly lower eco-TTC value. Gutsell et al. [10] importantly concluded that it is possible to conduct a screeninglevel environmental risk assessment of chemicals that may have poorly defined class assignments in terms of mode of action. A comprehensive list of known applications of TTC or TTC-like assessments for environmental or ecological endpoints is given in Table 1. DEVELOPING ECO-TTCS
We propose that now is a suitable time to develop a critical assessment of the potential scope, utility of application, and limitations of eco-TTCs, with an eye toward potential regulatory application. Data availability from international chemical management programs provide unprecedented insight into ecotoxicity; regulatory needs continue to advance toward assessing the universe of chemicals, including those of very low volumes; and greater computing power allows probing of sophisticated and complex chemical information data. Various statistical and methodological approaches have been undertaken in the early forays of eco-TTC development by environmental risk scientists. Data availability and quality verification are always considered to be key aspects, but additional methodological needs exist. The process of deriving TTCs (ecological or human) is fairly well established. Given a
S.E. Belanger et al.
coherent, quality dataset, each chemical can be categorized with respect to mode of action, chemical functional use, or chemical category. Each of these classifications can be a complex decision because various options to accomplish the classification are available. A hazard decision associated with each chemical and test result is required. This could be based on a predicted no-effect concentration (PNEC) determination for the chemical or simply based on results for a particular test species or trophic level (e.g., algae, invertebrates, or fish). The hazard determinations for chemicals with a given mode of action or chemical category can then be viewed as a statistical distribution (probability density function) of values (Figure 1A). The statistical distribution can be assumed or identified on the basis of the empirical values present, and a small percentile (most often 5%) can be estimated from it. The probability density function is most often converted into a cumulative distribution that looks like an S-shaped regression along with an upper and lower confidence interval that informs the user of the quality and noise in the data (Figure 1B). A common interpretation for ecoTTCs, similar to that already in use for human safety, is that the 5th percentile is an appropriately conservative marker to which exposure concentrations for a chemical matching the same mode of action or chemical category can be compared. In this sense, eco-TTCs will markedly differ in their use and interpretation from chemical-specific environmental hazard assessments or environmental quality standards; eco-TTCs are more generic starting points from which chemical-specific evaluations may begin. The hazard data used to derive the ecoTTC can vary depending on the goals of the assessment, as indicated in the Introduction. Some eco-TTCs focus on distributions developed for individual species [8,9], whereas others focus on PNECs derived from all available taxa into which appropriate, but varying, application factors were embedded prior to the computation of the eco-TTC [5,10]. A need exists to evaluate and understand patterns of sensitivity for the broad groups of tested organisms, such as photosynthetic microbes (cyanobacteria and algae), invertebrates, and fish in concert with the breadth of information available supporting various uses of application factors [11,12]. A number of possible approaches can be considered; for example, eco-TTCs based on ecosystem PNECs could be established via application of regionally appropriate extrapolation factors applied to datasets of different levels of complexity or completeness. In this regard, one may accept that an appropriate level of uncertainty is already embedded in the assessment. Different regions (e.g., North America, Europe, Asia) having different levels of risk tolerance for datasets of the same level of complexity would have different eco-TTCs as a result. Another aspect of evaluating the potential for eco-TTCs in an environmental risk assessment context would be to categorize chemicals based on a variety of tools such as the mode of action, functional groups present in the molecule, and functional chemical category based on intended use (e.g., surfactant). The development of eco-TTCs is likely initially most useful for compounds that can be extrapolated from established patterns of acute toxicity or that have wellestablished modes of action. Not all chemicals, classes, or categories may be suitable to be covered by an eco-TTC approach; however, this will require technical discussions and research investments to fully explore. Costs associated with developing eco-TTCs in a given chemical domain will probably vary. We believe, for example, that low-volume chemical substances of unknown or variable composition, complex reaction products, and biological materials (UVCBs) would be a
Development of ecological thresholds of toxicological concern
Environ Toxicol Chem 34, 2015
2867
Figure 1. Generic example of an ecological threshold of toxicological concern (eco-TTC) based on a probability density function (A) that is then converted to the cumulative probability distribution (B). Key statistical indicators include the calculated 5th centile and 95% lower confidence limit of the 5th percentile in panel B. The distribution was constructed from a theoretical data set of 19 different chemicals from which the individual chemical-specific predicted no-effect concentrations were derived.
group for which relevant eco-TTCs could be applied based on acute toxicity and at a relatively low cost. It is likely that the potential utility would be for assessments of hazard associated with ingredients in formulated products by private-sector organizations and perhaps in certain regulatory situations. An interesting group would be industrial narcotic compounds, which represent the majority of chemicals in commerce. Deeper investigations to confirm the importance of modes of action, test species influences, and specific functional groups would be challenging and important and would need a moderate research investment. On the other end of the scale from UVCBs would be non-narcotic compounds requiring substantially more mode of action insight that are also supported by primarily chronic toxicity data. We envision a growth in eco-TTCs that could be developed. Efforts to develop eco-TTCs can be envisioned as an everexpanding scope of chemical and species coverage with increasing costs and breadth of research questions requiring answers as coverage expands. Several key challenges for ecoTTC developers are also envisioned (Figure 2). The universe of ecotoxicological data, particularly of standard test guidelinetype studies, will need to be assessed. Data collection, chemical
categorization, and study quality assessment will be a large effort. A rigorous, transparent architecture to organize, probe, and harmonize as much as possible and share information will be exceedingly useful, as with any metadata analysis. Principles of chemistry and ecotoxicology will be applied, as will categorization of chemical uses, which relies heavily on industry knowledge and product applications. For example, different types of mode of action assignments will be used [6,7,13,14] to explore robustness of assignments, and datasets that vary in completeness with respect to included taxa (algae, invertebrates, and fish) leading to different levels of uncertainty in PNEC derivation should also be explored. Eco-TTCs will differ markedly in their use and interpretation from chemicalspecific environmental hazard assessments or environmental quality standards, as eco-TTCs are more generic starting points from which chemical-specific evaluations may begin. A new tool for early tier hazard assessment and identification will result and will have unique benefits compared with existing measures. For example, eco-TTCs can be developed as an alternative to quantitative structure-activity relationships (QSARs) or to assess complex mixtures that have otherwise been problematic for other technical reasons. Eco-TTCs could also be informed
2868
Environ Toxicol Chem 34, 2015
S.E. Belanger et al.
Figure 2. Proposed ecological threshold of toxicological concern (eco-TTC) development process with key challenges identified. HC5 ¼ 5% hazardous concentration; PNEC ¼ predicted no-effect concentration; QSAR ¼ quantitative structure–activity relationship.
by the emerging science of critical body burdens or chemical activity approaches; however, data gaps may be problematic for this to be realized in the short term. We expect environmental risk assessors in industry and the public sector (i.e., regulators) to be the direct beneficiaries of developing a consensus approach and putative eco-TTCs for application to previously untested compounds. In addition to being helpful as a screening hazard assessment approach, ecoTTCs could also be helpful in weight-of-evidence assessments for existing compounds or chemical categories that may be controversial or less well understood, including those for which QSARs appear to be presently unattainable; or, when the prediction is questionable, the eco-TTC can be used as a corroborating or refuting line of evidence. If an appropriate and high-quality QSAR is available for a chemical and species of interest, however, more realistic PNECs may be derived and will likely be used instead of an eco-TTC. Identification of compounds that are at the extremes of the PNEC or chemical– species distribution function may be exceptionally helpful in the evaluation of hypotheses of chemical–biological interactions, modes of action, and the like. Private-sector risk assessors may gain an important tool for screening-level hazard assessments for guiding newly developed chemical technologies prior to commercialization in addition to devising inputs to risk assessment for regulatory use. The public will benefit greatly
from approaches such as these by improving smart use of existing information as well as negating or reducing the need for additional animal testing. In human safety, improved animal welfare through the reduction, replacement, or refinement of vertebrate tests associated with the use and acceptance of TTC approaches is profound [15]. CALL FOR AN INTERNATIONAL COLLABORATION
The International Life Sciences Institute Health and Environmental Sciences Institute has established a working group in the Animal Alternatives in Environmental Risk Assessment Technical Committee. The committee is composed of environmental risk assessment scientists representing private industry, government, and academic interests. The group is in the process of collecting data from diverse public-sector (and in some cases private-sector) sources to develop eco-TTCs. The committee intends to clarify boundary conditions of application, probe strengths and weaknesses of various approaches, and provide an overall process that can lead to regulatory use and acceptance if methods prove scientifically sound. We contend that the time is ripe for such an approach, as international chemical management programs proliferate, interest in chemicals of emerging concern gain public prominence, and attention moves toward chemicals that are low volume and subsequently
Development of ecological thresholds of toxicological concern
have fewer data on their environmental effects. Key to this process is engagement of a diverse array of scientists engaged in different industry sectors, regulators and academic scientists. Data availability for an even wider diversity of chemicals and ecotoxicity results, as well as healthy debate about the strengths and limitations of eco-TTCs, would be welcome. Interested scientists can participate directly by joining these existing research efforts, making ecotoxicological data available for others to use and/or incorporate into the assessments, or following the development of these programs and deriving ecoTTCs and sharing them independently. The intent is to expand aquatic hazard inputs to the eco-TTC assessments beyond chemicals that are traditionally employed in research and to widen the array and depths of modes of action that are included. An intentional outcome, in addition to eco-TTC data assessment, will be to increase the likelihood for regulatory acceptance where use can be technically supported. Rigorous assessments of data quality and public transparency of the data, process, and results will be important attributes for regulatory acceptance. Acknowledgment—D. de Zwart received funding from the SOLUTIONS project via the European Union’s Seventh Framework Programme for research, technological development, and demonstration under grant agreement no. 603437. The authors thank the collaborators on this Health and Environmental Sciences Institute project for valuable input and discussions: A. Beasley, R. Hummel, A. Kienzler, and J. Suski. Disclaimer—The views, conclusions, and recommendations expressed in the present article are those of the authors and do not necessarily represent views or policies of the European Commission. This article was prepared by R. Sternberg as part of her official duties for the US Environmental Protection Agency. It has not been formally reviewed by the Agency and does not necessarily reflect the Agency’s views. Data Availability—No actual data have been analyzed in this manuscript; however, all references are publicly available. REFERENCES 1. Solomon KR, Brock TCM, Zwart Dd, Dyer SD, Posthuma L, Richards SM, Sanderson H, Sibley PK, van den Brink PJ. 2008. Extrapolation Practice for Ecotoxicological Effect Characterization of Chemicals. SETAC & CRC, Pensacola, FL, USA. 2. US Food and Drug Administration Center for Food Safety and Applied Nutrition. 1993. Toxicological principles for the safety assessment of direct food additives and color additives used in food, Redbook II. Fed Reg 58:16536. 3. Kroes R, Renwick AG, Cheeseman M, Kleiner J, Mangelsdorf I, Piersma A, Schilter B, Schlatter J, Van Schothorst F, Vos JG, W€ urtzen G. 2004. Structure-based thresholds of toxicological concern (TTC): Guidance for application to substances present at low levels in the diet. Toxicol Sci 42:65–83.
Environ Toxicol Chem 34, 2015
2869
4. European Food Safety Authority Scientific Committee. 2012. Scientific opinion on exploring options for providing advice about possible human health risks based on the concept of threshold of toxicological concern (TTC). EFSA J 10:2750–2853. 5. De Wolf W, Siebel-Sauer A, Lecloux A, Koch V, Holt M, Feijtel T, Comber M, Boeije G. 2005. Mode of action and aquatic exposure thresholds of no concern. Environ Toxicol Chem 24:479–485. 6. Verhaar HJM, Ramos EU, Hermens JLM. 1996. Classifying environmental pollutants. 2: Separation of class 1 (baseline toxicity) and class 2 (‘polar narcosis’) type compounds based on chemical descriptors. J Chemometr 10:149–162. 7. Verhaar HJM, Van Leeuwen CJ, Hermens JLM. 1992. Classifying environmental pollutants. 1: Structure-activity relationships for prediction of aquatic toxicity. Chemosphere 25:471–491. 8. Gross M, Daginnus K, Deviller G, de Wolf W, Dungey S, Galli C, Gourmelon A, Jacobs M, Matthiessen P, Micheletti C, Nestmann E, Pavan M, Paya-Perez A, Ratte HT, Safford B, Sokull-Kl€ uttgen B, Stock F, Stolzenberg HC, Wheeler J, Willuhn M, Worth A, Comenges JMZ, Crane M. 2010. Thresholds of toxicological concern for endocrine active substances in the aquatic environment. Integr Environ Assess Manag 6:2–11. 9. Williams ES, Berninger JP, Brooks BW. 2011. Application of chemical toxicity distributions to ecotoxicology data requirements under REACH. Environ Toxicol Chem 30:1943–1954. 10. Gutsell S, Hodges G, Marshall S, Roberts J. 2015. Ecotoxicological thresholds—Practical aplications in an industrial inventory. Environ Toxicol Chem 34:935–942. 11. Nabholz JV, Miller P, Zeeman M. 1993. Environmental risk assessment of new chemicals under the Toxic Substances Control Act (TSCA) Section Five. In Landis WG, Hughes JS, Lewis MA, eds, Environmental Toxicology and Risk Assessment. ASTM Special Technical Publication 1179. ASTM International, Philadelphia, PA, USA, pp 40–55. 12. European Chemicals Agency. Guidance on information requirements and chemical safety assessment, Chapter R.10: Characterization of dose [concnetration]-response for the environment. In Guidance on Information Requirements and Chemical Safety Assessment. Helsinki, Finland. 13. Russom CL, Bradbury SP, Broderius SJ, Hammermeister DE, Drummond RA. 1997. Predicting modes of toxic action from chemical structure: Acute toxicity in the fathead minnow (Pimephales promelas). 16:948–967. 14. Computer Sciences Corporation. 2012. ASTER User Guide—ASsessment Tools for the Evaluation of Risk, Ver 2.0. US Environmental Protection Agency, Duluth, MN, USA. 15. Hauge-Nilsen K, Keller D. 2014. Feasibility study: Refinement of the TTC concept by additional rules based on in silico and experimental data. Arch Toxicol 89:25–32. 16. Mons MN, Heringa MB, van Genderen J, Puijker LM, Brand W, Van Leeuwen CJ, Stoks P, van der Hoek JP, van der Kooij D. 2013. Use of the threshold of toxicological concern (TTC) approach for deriving target values for drinking water contaminants. Water Res 47:1666– 1678. 17. Hendriks AJ, Awkerman JA, de Zwart D, Huijbregts MAJ. 2013. Sensitivity of species to chemicals: Dose-response characteristics for various test types (LC50, LR50 and LD50) and modes of action. Ecotoxicol Environ Saf 97:10–16.
