Wildlife Toxicity Assessments for Chemicals of Military Concern

Introduction

Chapter 1

Characterizing Potential for Toxicity

Estimating Risks to Wildlife

Mark S. Johnson; Matthew J. McAtee

Abstract

Historical and current examples exist where valued wildlife species were harmed from exposure to chemicals that initially resulted from human events. Inadvertent and indirect exposures to pesticides, chemical residues, and waste streams have resulted in adverse effects that have affected populations of wildlife species. Many jurisdictions have regulations requiring responsible testing, characterization, and disposal of substances suspected of causing harm to wildlife species. Examples include the Toxic Substances Control Act; Resource Conservation and Recovery Act; and the Federal Insecticide, Fungicide, Rodenticide Act.

Military installations require the use of specific substances to develop and maintain weapon systems; industrial equipment; and a ready, fit, and prepared force. Additionally, military installations consist of large tracts of valuable habitat for a wide range of wildlife species. To aid in the characterization of potential risk of substances that have or may have been released in the environment, toxicity benchmarks, or Toxicity Reference Values (TRVs) are used to determine whether the risk of adverse effects from exposure to these substances in the environment is acceptable. There is no current consensus for the development of these values; however, this effort provides a method using the best techniques currently available to develop TRVs that can be used for screening purposes. This chapter provides the background, a review of methodologies, and support for the methodology used in the following chapters in the derivation and use of TRVs.

Keywords

Toxicity Reference Value

wildlife

introduction

terrestrial

models

Contents

Introduction   3

History of Wildlife Toxicity Reference Values   5

Recent Advances in Toxicity Reference Value Derivation   7

Important Research Needs   9

References   10

Introduction

Adverse effects to wildlife from unintended exposure to substances have been described in various sources and have provided the basis for many regulations that are intended to protect valued species. Additionally, many animal species, because of their greater exposure potential to substances in the environment, have been proposed as sentinels of adverse events as early warning signals for humans that may be exposed [1]. In the United States, there are regulations that protect migratory birds, threatened and endangered species, and eagles from harm, which may include adverse effects from chemical exposure in the environment. The Resource Conservation and Recovery Act and the Comprehensive Environmental Response, Compensation, and Liability Act (i.e., Superfund) were established, and require addressing adverse effects to the environment as well as to humans. Additionally, the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA) also requires an environmental evaluation of risk for applications of new pesticides to determine whether nontarget species could be at risk from exposure. These regulatory requirements have provided the initiative to develop the means to assess health risks to valued wildlife species. An important part of wildlife risk assessment for any given substance is the derivation of a toxicity benchmark.

The military utilizes and controls relatively large tracts of land, some of which are hundreds of square kilometers in size. Often land and personnel uses are highly restricted and controlled, resulting in highly valuable habitats for many wildlife species. In fact, the Nature Conservancy and the U.S. Department of Defense have been partnering for more than 30 years in managing the natural resources, long range sustainability of lands, and the more than 330 threatened or endangered species on military installations while simultaneously ensuring the success of the military mission [2]. The program known as the Readiness and Environmental Protection Initiative (REPI) has resulted in partnerships in over 24 states, protecting sensitive and valuable habitats while protecting the military’s ability to test and train at some of the nation’s most important military installations. Although these efforts are habitat-focused, clearly the value of these areas and the species that use them are of national interest.

Some activities that have occurred on military lands since before World War I have resulted in releases of substances in the environment that are, or may be, harmful to wildlife. On military properties, many of these substances represent those that are relatively military-unique, such as explosives, propellants, and pyrotechnics (collectively known as energetics). Their release often was a result of [1] earlier production and use practices before regulation that modified such practices (e.g., ammunition production during World War II), or [2] intended use where either incomplete combustion has occurred (e.g., detonation ranges), or where otherwise allowed by law (e.g., military facilities with RCRA permits). Other releases have occurred through training or use, or during war through theater operations. Characterization of risks at these locations where releases have occurred or to where substances have been transported is important to natural resource personnel, regulators, and society as in many cases, these substances are unique and data are scarce. A subset of this characterization includes those to valued wildlife species. In some cases, these releases were not of military unique substances, but of substances of military relevance. Examples include specific metals, pesticides, or organic compounds. Textbooks have been written to help provide central sources for this and other information [3,4].

A primary challenge in characterizing risk to wildlife from environmental exposure requires the interpretation and translation of toxicity information into useful risk assessment tools. Federal entities and jurisdictions collect information and do this for human health risk assessment. However, to date, there is no single updated source for toxicity values for wildlife.

History of wildlife toxicity reference values

Most evaluations of risk from exposure to man-made substances (xenobiotics) require the determination of the threshold for toxicity—to describe the point where exposure exceeds the organisms’ capability to adapt and illicit no adverse health effects. This dose-response function is the fundamental principal of toxicology—that the dose makes the poison. Ascertaining the exposure point at which adverse effects occur is the keystone in determining level of exposure that could be used in risk assessment applications. In principal, this seems a straightforward task; however, there are many confounding aspects of this exercise and determining this threshold is often a complex problem.

Estimating safe levels of exposure at or near the suspected threshold of toxicity is and has been an active area for human health risk assessment. Various methods are used. Examples include procedural cook book methods where points of departures (PODs) are determined from the dose-response relationship from a key or seminal study that are then modified with uncertainty factors that result in a toxicity benchmark. Some use a weight-of-evidence approach where professional judgment is used and described in determining the toxicity benchmark. Each endeavor to characterize the threshold for adverse responses. Examples include the U.S. Environmental Protection Agency (U.S. EPA) Integrated Risk Information System (IRIS) Reference Dose and the American Conference for Governmental Industrial Hygienist (ACGIH®) Threshold Limit Values, respectively. Toxicity benchmarks developed for wildlife and described herein are referred to as Toxicity Reference Values (TRVs).

Toxicity benchmark derivation for wildlife was initially patterned after human health risk assessment methods [5]. Here we define wildlife as wild terrestrial or riparian vertebrates that predominantly reside in terrestrial or riparian environments.

Most controlled toxicity data used for human health assessment are from animal data or terrestrial mammal data, which are largely collected from laboratory rodent animal models. To be comprehensive in the protection of humans, an incredible number of measurements or toxicity endpoint data are gleaned – from histopathology of numerous organs and organ systems to blood cell differentials, plasma clinical chemistry values, feed consumption, and body mass changes. Data are typically reviewed, and the toxicity of the most sensitive endpoint is modeled and evaluated for POD determination.

Most methods used for wildlife and humans have employed the collective use of no observed adverse effect levels (NOAELs) and/or lowest observed adverse effect levels (LOAELs) used in determining PODs or used together from many studies and species to estimate or characterize the variation in toxicity between species. NOAELs and LOAELs, or NOEC/LOECs where C represents a concentration in media, such as air, water, or soil; and the former L is typically reserved for oral exposures or doses in mg of substance/kg body mass each day or a daily oral exposure estimate, in fact, represent treatment levels from controlled studies that are statistically different in the parameter measured from those organisms that are not exposed or the controls. They are statistical artefacts of study design and have been criticized for providing little information on the dose response relationship between exposure and effect [6–12]. Additionally, care must be taken in their interpretation as statistical significance within the confines of the study and the biological relevance of the endpoint may not be the same.

The NOAEL/LOAEL approach, though used for years in human health risk assessment, has been criticized for many years. The U.S. EPA Office of Research and Development Integrated Risk Information System have now moved to the exclusive use of the benchmark dose (BMD) method in determining the POD. However, the challenge of developing chemical-specific TRVs that are applicable to many species remains.

The primary challenge in TRV derivation is endpoint selection and extrapolation to the species of interest. In the human health example, only one species (humans) is considered, though it is recognized that there is considerable range in response between humans; young and old, ones with existing conditions, genetic profiles, and so forth.i However, although there are data intended for human health concerns that can be relevant to wildlife species (e.g., rodent data can be reasonably extrapolated to other wild rodents such as voles and deer mice), the relevance of extrapolating rodent data to other species of wildlife is questionable. The variation in response that is expected from the controlled rodent data that are largely from isogenetic strains to outbred wild white-tailed deer in which the risk assessor is interested must be considered relative to the physiological differences between the species and the nature of the toxic endpoint of effect.

Early efforts in TRVs focused on NOAELs/LOAELs that were limited to endpoints expected to be most directly related to effects that would cause changes in populations [13]. Initially, they were patterned on the same measures typically measured in aquatic organisms: mortality, growth, and reproduction [5]. However, these efforts have been criticized in that other, more sensitive effects may indirectly influence population growth and sustainability. For example, exposures to substances that cause lethargy have the potential to be directly related to an increased mortality rate from predators that would reduce fitness. Therefore, the method used here takes a conservative approach and uses the most sensitive endpoint determined as adverse, not adaptive, to derivate the TRV.

The USEPA employed a transparent system of literature review, quality assurance, and TRV derivation based on the evaluation of NOAELs and LOAELs [14]. The Eco Soil Screening Level (EcoSSLs) effort developed TRVs for 10 metals, based on NOAEL/LOAEL spread for growth, mortality, and reproduction, though other endpoint data were also plotted for comparison purposes. Typically, the highest NOAEL for mortality, growth, or reproduction was used above the lowest LOAEL. Acute data were not included and this effort has not been revised recently.

Recent advances in toxicity reference value derivation

Recent efforts in TRV derivation have become more sophisticated. Species sensitivity distributions (SSDs) have been used by many to consider variation in response between species [15]. However, they require discreet data; continuous data such as growth and histopathological data are difficult to plot. Additionally, the perceived variation attributed to species differences may be a reflection of differences in study designs instead [16]. Attempts to keep SSD models within the endpoints assumed to be ecologically relevant (i.e., mortality, growth, and reproduction) often ignore other endpoints of potential significance.

Many have suggested a best fit straight-line extrapolation of the dose response curve to be an improvement over the NOAEL/LOAEL approach [9,16,17]. These values are typically referred to as Effective or Ecological Dose of Concentrations at a specific level of population response (e.g., EC20); however, their predominant use has been typically relegated to soil invertebrates [14].

The application of the benchmark dose method used in human health benchmark derivation has been used as a preferred method for extrapolating laboratory data for TRVs and is the method used in this present effort [17]. As described, the benchmark dose method can be used with data of many types to include discreet (incidence), continuous, and even mixed (e.g., categorical regression). The model allows the user to define the population level of interest. It is important to note that estimates from the 1 to 5% or the 95 to 99% ranges are highly variable and confidence intervals are large as data are rarely available for the tails of the dose response distribution. The mean modeled 10% level of a sublethal effect is usually referred to as the benchmark dose (BMD) and the lower 95% confidence limit on the mean 10% response is referred to as the benchmark dose limit (BMDL). Typically, the BMDL has been used for screening purposes for ecological risk assessment purposes and for determining the point of departure for estimating noncancer values for human health risk assessment. When a less protective, more predictive value is needed, the risk assessor can move up the dose response curve and choose a higher level, or frequency of response, for a population effect or a lower level of confidence (e.g., choosing a 15 or 20% sublethal effect to a population or lower the confidence bound on the mean rather than using the lower 95% confidence bound). The 50% level of confidence is known as the BMD and is the point on the best fit modelled curve at the determined population level of effect.

Toxicity data developed for the species of interest, on the substance of interest, should be used in the derivation of the TRV. However, rarely are data available for wildlife species. Therefore, data collected on a substance within a vertebrate class (i.e., mammals, birds, reptiles, and amphibians) are used collectively to make a toxicity assessment and derive a benchmark. Typically this level of comparison can provide data that are sufficient for this purpose. Challenges remain in evaluating these data since the variation in toxicological response may be due to a host of factors. Few studies use the same methods to be able to determine if differences are due to species or methodological differences. Differences between feeding and gavage studies and vehicle choices can influence absorption, hence toxicity. There are few methods described for wildlife toxicity testing and fewer published intended for TRV derivation.

Ideally, many dose response curves could be used at one time for a given endpoint using data from many species to obtain a more accurate representation of cross-species variation. However, currently there are no models developed for this purpose. Therefore, the current method is constrained by selecting the most relevant, high quality data set to model the effect determined to be representative of the class representing the species of interest. This requires professional judgment of the entire data set and a weight of evidence determination.

Recent reviews of wildlife risk assessments suggest that the TRV can greatly influence the risk calculation more than the exposure estimation [18]. It has been suggested that the BMD method can be applied to data of species of interest, not all data within a class, to assist in deriving a predictive value for remedial decision making [16]. This and moving from the BMDL to the BMD may provide flexibility and reduce the uncertainty sufficiently to assist in making rational remedial decisions—many of which have other habitat impacts on wildlife populations. Clearly a balance between predictive adverse effects and necessary habitat alterations must be considered. Other methods are being considered. These enable the calculation of a best-fit dose response curve to a collection of curves for a specific endpoint using data for various species. This approach permits a more robust evaluation when considering species variability; however, care is exercised to ensure the variation in response is due to differences between species, and not differences in study design.

Important research needs

A lack of adequate toxicity data is often a limiting factor in deriving safe levels of exposure. This statement is accurate for public health concerns and for ensuring wildlife populations are protected. The latter is often confounded by the sheer number of species to be assessed. Animal data used for human health risk assessment is clearly inadequate for many reasons; the predominant one is the lack of such data for many substances used in commercial applications. New methods have been developed (and are under development) to help streamline the toxicology process and provide more useful low-level data applicable to many human exposures (e.g., Tox21) [19], but these will require further development.

Data developed for human health purposes using laboratory rodents can be useful in estimating effects on nonhuman mammals as well. Naturally, those rodent data are most applicable to other rodents and likely less useful in characterizing effects to Carnivores or Mustelids. Variation in levels of exposure that causes adverse effects can be attributed to species differences, but also differences in study design. For example, data from gavage and feeding studies can vary enormously attributed to differences in toxicokinetic differences between a single bolus and a longer-sustained daily exposure, respectively. Standardized methods in exposure and endpoint measurement are needed to compare differences between species accurately.

Still, it is impractical to test all species of concern for each potential environmental xenobiotic. Evaluations need to be conducted that compare species sensitivities to help determine which species would provide results that could best be extrapolated to others. This may be most useful where specific classes of compounds are concerned. Greater refinement, development, and use of in silico (e.g., Quantitative Structural Activity Relationship; QSAR) models may also be incorporated. Further, additional statistical tools that enable the characterization of multiple dose-response curves are needed to determine the BMD for several species simultaneously, which may evaluate all variables in the results at once. Although many terrestrial vertebrates share or show homology between biological systems that enable life, differences in how these functions are performed due to various selective pressures through history create the basis for selective toxicity at their respective targets. Finding the canary sentinel equivalent is likely to remain elusive for the vast array of environmental substances that have selective targets of toxicity.

References

[1] National Research Council (U.S.) Committee on Animals as Monitors of Environmental Hazards. Animals as sentinels of environmental health hazards. Washington (DC): National Academy Press; 1991.

[2] Nature Conservancy. The military and nature: our partnership with the department of defense. [Internet] 2014 [cited 2014 Sept 14] 2014. Available from http://www.nature.org/newsfeatures/specialfeatures/partnership-with-the-department-of-defense.xml.

[3] Kendall RJ, Smith PN. Perchlorate ecotoxicology. Pensacola, FL: SETAC Press; 2006.

[4] Sunahara GI, Lotufo G, Kuperman RG, Hawari J. Ecotoxicology of explosives. Boca Raton, FL: CRC Press; 2009.

[5] Suter GW, Sample BEOak Ridge National Laboratory, Lockheed Martin Energy Research Corp. Guide to the ORNL ecotoxicological screening benchmarks: background, development, and application. Oak Ridge, TN: Oak Ridge National Laboratory; 1998 Pub. No. 4793.

[6] Chapman PM, Cardwell RS, Chapman PF. A warning: NOECs are inappropriate for regulatory use. Environ Toxicol Chem. 1996;15(2):77–79.

[7] Warne M. St J., van Dam R. NOEC and LOEC data should no longer be generated or used. Australas J Ecotoxicol. 2008;14(1):1–5.

[8] Laskowski R. Some good reasons to ban the use of NOEC, LOEC and related concepts in ecotoxicology. Oikos. 1995;73(1):140–144.

[9] Kodell RL. Replace the NOAEL and LOAEL with BMDL01 and BMDL10. Environ Ecol Stat. 2009;16(1):3–12.

[10] Yanagawa T, Kikuchi Y. Statistical issues on the determination of the no-observed-adverse-effect levels in toxicology. Environmetrics-Chichester. 2001;12(4):319–326.

[11] Leisenring W, Ryan L. Statistical properties of the NOAEL. Reg Regul Toxicol Pharmacol. 1992 Apr;15(2):161–171.

[12] Bokkers BG, Slob W. A comparison of ratio distributions based on the NOAEL and benchmark approach for subchronic-to-chronic extrapolation. Toxicol Sci. 2005;85(2):1033–1040.

[13] U.S. Environmental Response Team. Ecological risk assessment guidance for Superfund: process for designing and conducting ecological risk assessments. Edison, NJ: U.S. Environmental Protection Agency, Environmental Response Team; 1997 Interim Final. EPA 540-R-97-006.

[14] U.S. Environmental Protection Agency. Guidance for developing ecological soils screening levels. Washington DC: Office of Solid Waste and Emergency Response; 2005.

[15] Awkerman JA, Raimondo S, Barron MG. Development of species sensitivity distributions for wildlife using interspecies toxicity correlation models. Environ Sci Technol. 2008 May 1;42(9):3447–3452.

[16] Mayfield DB, Johnson MS, Burris JA, Fairbrother A. Furthering the derivation of predictive wildlife toxicity reference values for use in soil cleanup decisions. Integr Environ Assess Manag. 2014 Jul;10(3):358–371.

[17] U.S. Army Center for Health Promotion and Preventive Medicine (USACHPPM). Standard practice for wildlife toxicity reference values. Aberdeen Proving Ground, MD: U.S. Army Center for Health Promotion and Preventive Medicine; 2000 Technical Guide No. 254.

[18] Sample BE, Fairbrother A, Kaiser A, Law S, Adams B. Sensitivity of ecological soil-screening levels for metals to exposure model parameterization and toxicity reference values. Environ Toxicol Chem. 2014 Jun 18;doi:10.1002/etc.2675 [Epub ahead of print].

[19] National Research Council (U.S.). Committee on Toxicity Testing and Assessment of Environmental Agents. Toxicity testing in the 21st century: a vision and a strategy. Washington, DC: National Academies Press; 2007.

i It is important to note that intra- and interspecies variability is an issue for wildlife.

Part II

Methodological Derivations

Chapter 2

Methods for Derivation of Wildlife Toxicity Values for Use in Ecological Risk Assessments

Adam T. Deck; Mark S. Johnson

Abstract

To understand the potential of an environmental chemical to cause adverse health effects on wildlife, a safe level to which wildlife can be exposed needs to be determined. This level, known as a Toxicity Reference Value (TRV), is derived by compiling and analyzing controlled laboratory experiments into a TRV report. Development of a TRV begins with the collection of relevant, controlled laboratory toxicological studies that are identified based on the endpoint, the exposure duration, the effect level, and the exposure pathway. Other considerations include study design validity, the quality of the study, and whether the study results corroborate with other published data. The relevant studies and data formulate the toxicity profile. The benchmark dose approach, the no observed adverse effect level/lowest observed adverse effect level (NOAEL/LOAEL) approach, or the approximation approach is employed on toxicity profile data to derive the TRV, which is summarized into a TRV report.

Keywords

Toxicity Reference Value

wildlife

toxicity profile

derivation

Technical Guide 254

Contents

Introduction   15

Data Collection and Literature Search   16

Identifying Relevant Studies   17

Toxicity Profile Preparation   19

Derivation of Toxicity Reference Values   20

Assigning a Confidence Level   21

Creation of the Toxicity Reference Value Report   22

References   22

Introduction

Understanding the potential for chemicals in the environment to cause adverse effects to wildlife requires knowledge of exposure and toxicology. The potential for exposure can be measured directly (e.g., through a chemical analysis of crop or gastrointestinal contents [1] or modeled from environmental media (e.g., soil and/or sediment chemical concentrations). In the latter case, food web models can be used to determine oral uptake [2]. However, toxicity data are typically used to determine a safe exposure level, most often assuming exposure occurs predominantly through oral exposure routes, although inhalation and dermal pathways can also be modeled. In this paradigm, exposure is considered the numerator, and a safe level of exposure based upon toxicity data is considered the denominator. The potential for adverse effects exists when exposure estimates (numerator) exceed the safe level of exposure (denominator).

Controlled toxicity data are typically used to understand the levels at which chemicals could be considered safe. Such data are from studies conducted in a controlled laboratory environment where the only difference between groups of test animals is the concentration of the chemical to which the animals are exposed. These studies can be expensive and laboratory animal wildlife models are limited. Regardless, the data from these studies can be used to establish a dose-response curve from which a threshold for adverse events occurs and a safe level can be identified. This level is often referred to as a Toxicity Reference Value (TRV).

Optimally, toxicity data would be available for chemicals of environmental concern and the wildlife species of interest; this is rarely the case. Typically, toxicity data collected for another purpose are used to develop a TRV. These data are from studies that use different designs, a factor that also contributes to the variability in results. Although TRVs are designed to apply to classes of vertebrates (e.g., mammals, birds, reptiles, and amphibians), chemical-specific data are often scarce, and TRV developers use whatever data are available to determine the threshold for adverse effects that would be useful for an entire class of vertebrates. An essential element of this process is to collect all potentially relevant information and to document the procedure. Once obtained, this information is documented in the form of a chemical-specific toxicity profile. Pertinent studies are gleaned and highlighted, professional judgment is used to ascertain the toxicity endpoints of relevance, and TRVs are then derived logically and transparently. Herein, we describe the procedure used to develop the TRVs provided in this book.

Data collection and literature search

The first step in creating the toxicity profile is to search the available scientific literature and collect data that will be used to develop the toxicity profile. Background information on manufacturing, environmental fate and transport, field studies, and any other data that provide context for the need to develop TRVs for wildlife are also helpful. When searching for study data, it is recommended that all available and pertinent databases be searched to ensure the toxicity profile represents a complete and comprehensive summary of all relevant material and data. The literature search should be conducted based on the vertebrate class (e.g., mammals, birds, and reptiles) and cross-referenced with the specific chemical that is being detailed. It is also important to document the parameters and results of the search. Important information includes databases and dates searched, keywords used, number of database accessions, and the abstracts and papers retrieved. Primary sources of information are preferred.

Identifying relevant studies

After the literature search has been completed and the pertinent studies gathered, they are analyzed to determine their applicability for use in developing the TRV. Seven criteria are considered in determining the relevance of the studies: endpoint, exposure duration, effect level, exposure pathway, validity of the study design, quality of the study, and corroboration of study results with other data.

While human risk assessments are concerned with protecting the health of individual human lives, ecological risk assessments are concerned with sustaining populations of species [2]. Because of this, a study must have endpoints that can conceivably affect the sustainment of the population in order to be included as relevant. For example, studies that find endpoints of uncertain toxicological significance (e.g., enzyme induction, adaptive responses, changes in clinical chemistry parameters within the normal ranges, etc.) are often not included in TRV development, but they may be used for corroboration of the relevant adverse effect. Studies that clearly identify adverse effects such as those that result in mortality; impaired reproduction, development, or growth; changes in behavior related to reproduction, feeding, and predator avoidance; and resistance to disease are considered because they are most likely to affect the sustainment of the population [2]. Additionally, endpoints that are adverse on an individual level are assumed to conceivably result in an adverse population level impact through changes in somatic repair, which could result in higher energetic demands in a food-limited environment in which reproduction could also be indirectly affected.

The exposure duration, defined as the length of time to which the study subjects were exposed to the test chemical, must be indicated for each study. For the purposes of developing a TRV, chronic exposures are the most relevant because representative species in each of the vertebrate classes could be resident and, as such, could conceivably be exposed for sustained periods. Data for chronic exposures are few, however, and there are certain circumstances under which shorter exposures are more relevant to field populations. Lengthy, sustained exposure periods typically result in lower exposure levels that often result in toxicity at much lower daily exposure rates than acute, short-term exposures. Thus, a conservative approach is used in preferring studies of chronic duration. For the purposes of developing TRVs, chronic exposures are those equal to or greater than 10% of the lifespan of the test species. While chronic exposures are preferred, subchronic exposures (less than 10% of the lifespan of the test species but greater than 14 days), subacute exposures (repetitive exposure of up to 14 days), and acute exposures (a single exposure event) may be relevant in providing information on mechanism, etiology, or species differences [3].

Studies are considered relevant when both no adverse effects and sublethal adverse effects are reported. Statistical comparisons are used to define effect levels that include the no observed adverse effect levels (NOAELs), the lowest observed adverse effect levels (LOAELs), and the doses, which produce health effects in 50% of the exposed population (ED50). Studies with no and low adverse effect are desirable because they capture the threshold for adverse effects, which is useful for developing TRVs for screening purposes.

For each study, the exposure pathway (i.e., the route by which a chemical is introduced into a receptor) should be considered along with the exposure pathway for the receptor, and the two should match. The exposure pathway of a wild animal that eats contaminated wild vegetation is more similar to that of a study animal that eats contaminated food than to a study animal that has been dosed by gavage. However, the characterization of dose in feeding studies can be problematic. In these studies, exposure is quantified by evenly coating the feed material with the test chemical and weighing the feed before and after its exposure to the test animal. The ability of animals to select feed particles that are less toxic than others has occurred [4], resulting in an increase in the chemical concentration of uneaten feed over time. Moreover, animals typically spill feed or change the mass of the test feed by urinating or defecating in feed troughs, adding variability to the exposure estimate [5].

Typically, test animals exposed to toxic compounds in feed resist feeding and frequently show a marked reduction in feed consumption and body mass. This often changes with continued exposure, during which feed consumption is resumed, and body mass increases to normal levels. However, the wide fluctuations in body mass contribute to the variation in the dose estimate because dose is calculated on a body mass basis: milligrams/chemical/kilogram body mass/day (mg chemical/kg body mass/day). This variation is rarely captured in dose estimates. Still, when these factors are best controlled for, feeding studies best emulate environmental conditions in many cases where exposures occur throughout the day, and that metabolism, disposition, and excretion may be able to prevent toxic systemic levels from occurring. Although quantification of exposure remains a challenge, compounds administered in the test animals’ drinking water are less prone to some of these drawbacks.

Gavage studies, that is, those in which a precise amount of chemical is directly administered in the gut of each individual test animal specific to its body weight and treatment, can provide conservative, less variable responses. Such studies are conservative in that each animal is given one daily dose each day, a bolus that can result in a greater probability of eliciting adverse effects by overwhelming the ability of the liver to metabolize the test compound, or overwhelming the ability for excretion via the urine or feces. Clearly, there are drawbacks and benefits associated with each.

The validity and quality of the study needs to meet minimum requirements for inclusion. The study must be statistically and biologically relevant. Thus, the chemical being investigated should occur in such a form as to best replicate the bioavailability of the test compound under natural conditions, and the dose of the chemical should be derived with limited inherent variability. Sufficient information must be provided to allow the study to be repeated, and the study must corroborate results from similar studies [3].

After each study has been considered and evaluated as relevant, the final step prior to writing the toxicity profile is to determine whether the collection of relative studies provides adequate data with which to derive the TRV. Typically, studies that contain data from three species within a class, data for species from at least two orders, and a minimum of two chronic LOAELs and one chronic NOAEL are required [3]. However, for many substances, these criteria were not satisfied. In such cases, weight-of-evidence techniques have been employed to use the best science in developing a value that would be relevant for screening purposes. Where insufficient data exist, uncertainty factors spanning an order of magnitude have been applied. Lastly, each value has been qualified with a level of confidence based on the abundance and concordance of the data and the biological systems that are expected to be perturbed to cause an adverse effect.

Toxicity profile preparation

With the relevant studies in hand, the toxicity profile can be prepared. In addition to providing a summary of the review of the literature on the toxicology of a given compound, the toxicity profile includes the logic on the relevance of the studies in deriving a TRV. Specifics include the dates on which the search was conducted, which sources were reviewed, and the search strategy used. The summarized toxicity data are organized by taxonomical class, route of exposure, and exposure duration, respectively. When sufficient, the quantitative data are depicted in a scatter diagram and a table indicating the most relevant data, to include the study endpoint, the study species, and the chemical’s effect levels.

Derivation of toxicity reference values

Data from the completed toxicity profile are used to develop two TRVs for each class of vertebrate and for each exposure pathway: low-value (useful for screening using conservative criteria) and high-value (uses mean values or the lowest dose at which adverse effects occur). By creating two TRVs, risk assessors have a range of toxicity reference values upon which risk estimates in an ecological risk assessment can be developed.

The methods recommended for deriving TRVs include the benchmark dose approach, the NOAEL/LOAEL approach, and the approximation approach [3]. Where possible, the benchmark dose was the preferred method; it fits a dose-response curve to the most relevant data and endpoint and considers the variation in the measurement of the endpoint. This variation is either captured in the low-value TRV, based on the lower 95% confidence interval on a 10% effect level (i.e., a dose about which there is 95% confidence that 10% or fewer animals are expected to develop an adverse effect) or captured using the 95% confidence level of one standard deviation. The latter is used when normal ranges of values do not exist for a particular test species. The high-value TRV is the dose that either corresponds to a 10% population effect level or represents the threshold for adverse effects that is on the best fit dose-response curve. The benchmark dose is preferred because it uses all the data from a specific endpoint from a study that was considered relevant and plots the best fit dose-response function that can be the means of departure for the derivation of other values (e.g., for remedial decision making) [6]. However, there are cases where dose-responses are multimodal (e.g., essential nutrients, hormetic responses, etc.), where a dose-response curve cannot be mathematically fitted with a level of confidence; in such cases, the NOAEL/LOAEL and/or the approximation approach can be