- Top of page
- TRV COMPILATION
- TRV ANALYSIS
- Supporting Information
Wildlife toxicity reference values (TRVs) are routinely used during screening level and baseline ecological risk assessments (ERAs). Risk assessment professionals often adopt TRVs from published sources to expedite risk analyses. The US Environmental Protection Agency (USEPA) developed ecological soil screening levels (Eco-SSLs) to provide a source of TRVs that would improve consistency among risk assessments. We conducted a survey and evaluated more than 50 publicly available, large-scale ERAs published in the last decade to evaluate if USEPA's goal of uniformity in the use of wildlife TRVs has been met. In addition, these ERAs were reviewed to understand current practices for wildlife TRV use and development within the risk assessment community. The use of no observed and lowest observed adverse effect levels culled from published compendia was common practice among the majority of ERAs reviewed. We found increasing use over time of TRVs established in the Eco-SSL documents; however, Eco-SSL TRV values were not used in the majority of recent ERAs and there continues to be wide variation in TRVs for commonly studied contaminants (e.g., metals, pesticides, PAHs, and PCBs). Variability in the toxicity values was driven by differences in the key studies selected, dose estimation methods, and use of uncertainty factors. These differences result in TRVs that span multiple orders of magnitude for many of the chemicals examined. This lack of consistency in TRV development leads to highly variable results in ecological risk assessments conducted throughout the United States. Integr Environ Assess Manag 2013; 9: 114–123. © 2012 SETAC
- Top of page
- TRV COMPILATION
- TRV ANALYSIS
- Supporting Information
The term toxicity reference value (TRV) is used to describe threshold values that indicate a level of exposure where the measured response is not statistically different from a control group. Alternatively, a TRV may represent an exposure dose or concentration above which ecologically relevant effects might occur to wildlife species and below which it is reasonably expected that such effects will not occur (USEPA 2005). TRVs are typically derived for defined species and life stages, or broader taxonomic groups, and are clearly linked to the assessment endpoints (generally, growth, survival, or reproduction) defined in the problem formulation of the ecological risk assessment (ERA). However, wildlife (e.g., birds, mammals, and herpetofauna) TRVs are highly variable, even for the same chemical due to differences in derivation procedures. TRVs may be derived from toxicity test data using various statistical methods, such as inferential statistical comparisons or confidence limits around point estimates from dose–response functions. Inferential statistical comparisons (e.g., analysis of variance) are regularly used to derive no observed adverse effect levels (NOAELs) or lowest observed adverse effect levels (LOAELs). TRVs may also be derived from a concentration or dose–response function where a predefined percentage (e.g., 10% or 20%) of the exposed population expresses the measured response (known as an effect dose [EDx], or when confidence intervals are applied, as a benchmark dose). More recently, TRVs have been derived from species sensitivity distributions (Posthuma et al. 2001), although this is more common for aquatic than terrestrial ecosystems.
Ecological risk assessments for terrestrial wildlife often use TRVs expressed as estimated daily oral doses, generally based on oral exposure estimates derived from concentrations of the chemicals of concern in environmental media. USEPA (2005), when developing their ecological soil screening levels (Eco-SSLs), addressed the importance of alternative exposure routes such as inhalation or dermal and stated that their analyses “support the conclusion that these pathways are generally less significant when compared to the ingestion pathways and do not warrant inclusion in the derivation of the Eco-SSLs.” More recent arguments have been made for inclusion of these nonstandard routes even if they are a minor component of total exposure (e.g., Gallegos et al. 2007; Mineau 2012).
Although TRVs are critical components of ecological risk assessments (USEPA 1997, 1998), there are few standardized, consensus-based values that are consistently adopted for wildlife. Within the United States, risk assessors often turn to established TRV compendia (Sample et al. 1996; USEPA R9 BTAG 2009; USEPA Eco-SSLs) or media screening value compilations (see review by Barron and Wharton 2005), or they derive wildlife toxicity benchmarks de novo from published literature. Wildlife TRVs are particularly challenging to develop and standardize due to limitations in the underlying toxicological database for many chemicals (Kapustka 2008). Through the publication of the Eco-SSLs, the EPA Office of Solid Waste and Emergency Response attempted to develop a set of standard values for chemicals frequently encountered at Superfund sites (As, Pb, DDT, dieldrin, low [LMW] and high [HWH] molecular PAHs, and pentachlorophenol) (USEPA 2005).
The stated purpose of deriving the Eco-SSLs was to 1) “conserve resources by limiting the need for USEPA, state, contractor, and other federal risk assessors to perform repetitious toxicity data literature searches and toxicity data evaluations,” and 2) to “increase consistency among screening risk analyses, decrease the possibility that potential risks from soil contamination to ecological receptors will be overlooked, and allow risk assessors to focus their resources on identifying key site studies needed for critical decision-making” (USEPA 2005). The objective of this article was to review agency- or state-directed ERAs that have been published before and after the release of the Eco-SSLs to determine if the goal of consistency among risk analyses has been achieved.
The development of wildlife TRVs for assessing ecological risks has clearly evolved over the past 2 decades and continues today. In this article, we explore whether wildlife TRVs are becoming more standardized and streamlined in ERAs being carried out throughout the United States. As part of this analysis we examined: 1) the frequency of use of TRVs published in USEPA Eco-SSL documents, 2) the variability in wildlife TRVs selected for use in ERAs both nationwide and within USEPA regions, and 3) the consistency of TRV derivation procedures.
Alternative (e.g., Cd; Stanton et al. 2010) or additional (e.g., TCDD; Blankenship et al. 2008) TRVs have been published for a limited number of chemicals using derivation methodologies that differ from those used by USEPA in the Eco-SSL process. For example, the USEPA Eco-SSL TRV derivation procedure uses an inferential statistical method (NOAELs and LOAELs), which historically has been standard practice in regulatory ecological risk assessment. However, alternative TRV derivation procedures (i.e., the use of a species sensitivity distribution based on a population level effect concentration, EDx or benchmark dose) have been proposed (USACHPPM 2000; Allard et al. 2010) and are now being implemented at a limited number of Superfund sites (Sample et al. 2011). In addition, ERAs frequently are unclear (or unstated) about the uncertainty factors used to derive the TRV values (extrapolation from subchronic to chronic exposure duration), and may use different assumed food ingestion rates, body weights, and interspecies scaling factors (Duke and Taggart 2000; McDonald and Wilcockson 2003; DeMott et al. 2005). Although we examine the database of ERAs for reasons why the wildlife TRVs differ from those recommended in the Eco-SSLs, it is not the purpose of this article to recommend a standardized procedure or argue for or against particular approaches; this was adequately covered previously by Allard et al. (2010).
- Top of page
- TRV COMPILATION
- TRV ANALYSIS
- Supporting Information
We conducted a survey of existing publicly available ERAs to evaluate TRV sources and derivation methods. Published ERAs were identified from national and regional USEPA websites, state environmental department websites, published journal articles, and from web-based searches. The ERAs were published between 1999 and 2011 and were identified as being pre- and postdevelopment of USEPA Eco-SSLs (i.e., before or after 2003–2008, depending on the chemical). We focused our document search on large Superfund or state hazardous waste sites that examined a large suite of inorganic and organic substances. In addition, only ERAs that were conducted in accordance with USEPA (1997, 1998) ERA guidance for Superfund were included in this analysis. This survey identified a total of 56 publicly available ERAs (Table 1) with toxicity assessment sections that documented the sources and numerical values of each TRV selected for the site-specific risk evaluation.
Table 1. Summary of the wildlife toxicity reference value database
|Dates of ERA publication||1999–2011|
|Total number of ERAs in database||56|
|Number of ERAs by USEPA regiona|
| Region 1||5|
| Region 2||5|
| Region 3||4|
| Region 4||4|
| Region 5||7|
| Region 6||2|
| Region 7||2|
| Region 8||9|
| Region 9||5|
| Region 10||13|
|Number of inorganic substancesb||24|
|Number of organic substancesb||34|
|Number of avian NOAEL recordsc||689|
|Number of avian LOAEL recordsc||503|
|Number of avian EDx recordsd||8|
|Number of mammalian NOAEL recordsc||776|
|Number of mammalian LOAEL recordsc||510|
|Number of mammalian EDx recordsd||10|
The documents reviewed in this analysis generally followed USEPA's process for conducting ecological risk assessment (USEPA 1997, 1998). Accordingly, the document structures included specific sections for the problem formulation, effects assessment and analysis phases of the ERA. In these sections, the TRV derivation methods were described (to varying degrees), including the original literature sources, dose extrapolation methods, and final NOAEL or LOAEL values used in the risk characterization. An electronic database (Microsoft Excel) was developed to catalogue the TRV information for each ERA. Database fields were generated to track the source of the TRVs, the original toxicity study or studies used to develop the TRV, the assumptions used to derive the TRVs (e.g., uncertainty factors, dose adjustments, interspecies adjustments, or allometric scaling), and the final oral toxicity values (NOAEL, LOAEL, or EDx) used to evaluate risks to avian or mammalian species in each ERA. Information provided within each ERA was transcribed to the appropriate fields in the database. If the ERA was deficient in any database category this was noted such that reporting practices could be assessed. This database provides a representative sampling, rather than a comprehensive account of all ERAs completed in the United States. Exploratory analyses were carried out using this database to understand current trends and practices within the ecological risk assessment community. A summary of the TRV database is provided in Table 1.
- Top of page
- TRV COMPILATION
- TRV ANALYSIS
- Supporting Information
An analysis of TRVs was performed by reviewing the text of each ERA document and compiling the TRV derivation parameters (NOAEL, LOAEL, uncertainty factors, dose-conversion assumptions) into the database. We queried the database to determine the sources of TRVs (e.g., USEPA Eco-SSLs), to investigate the variability among the TRVs, to examine consistency across geographic areas (i.e., USEPA Regions), and to identify differences in TRV derivation procedures. The results of each of these analyses are discussed in the following sections.
Sources of TRVs
Database queries were performed to determine the primary sources of wildlife TRVs used in ERAs. The ERAs used a variety of techniques to identify relevant toxicological literature including performing independent literature reviews (28 of 56 [50%]), reviewing TRV compilation sources (42 of 56 [75%]), and/or relying on previously completed USEPA-approved ERAs (6 of 56 [11%]). The wildlife TRV compendium developed by Sample et al. (1996) was the resource most frequently adopted by risk assessors in the ERAs that were evaluated (36 of 56 [64%]). This resource was used by ERAs for numerical TRVs, particularly for chemicals where no other resource was available, or as a tool for identifying relevant toxicological literature. Although 75% (42 of 56) of the ERAs in the database were completed during or after the publication of one or more of the USEPA Eco-SSL documents (between 2003 and 2008), the Eco-SSLs were not the primary source of TRVs (Figure 1). Generally, less than half (50%) of the ERAs relied on the USEPA Eco-SSL TRVs for each respective substance. This suggests that one of the goals of the USEPA Eco-SSL initiative (i.e., increased consistency among risk analyses) has yet to be realized.
Figure 1. Percent of ecological risk assessments using NOAEL TRVs from USEPA Eco-SSL documents. Only ERAs published after Eco-SSLs were established are included in the underlying data. Sample sizes [number of ERAs] ranged from 1 to 26 depending on the substance evaluated. Chemical symbols or abbreviations are shown on the horizontal axis.
Download figure to PowerPoint
Both avian and mammalian TRVs were found to be highly variable among the ERAs evaluated. Examples of this variation are depicted for metals and metalloids in Figure 2. Wildlife TRVs often spanned several orders of magnitude, even for those chemicals with established USEPA Eco-SSL TRVs (Figure 2). For example, avian NOAEL and LOAEL TRVs for As ranged from 0.691 to 20 mg × kg−1 × d−1 and 4.5 to 71 mg × kg−1 × d−1, respectively. Likewise, mammalian NOAEL and LOAEL TRVs for Cd ranged 0.0252 to 5.1 mg × kg−1 × d−1 and 0.909 to 13 mg × kg−1 × d−1, respectively. A similar breadth of variability in LOAEL TRVs for other metals and metalloids was also apparent (data not shown). The most common reasons for this variation include: selection of alternative key studies underlying the TRV, differing assumptions for dose-conversions, and use of differing uncertainty factors, all of which we discuss in more detail below. The results of this analysis indicate an evident lack of uniformity in TRVs adopted by risk assessors in ERAs published in the last decade (Figure 2).
Figure 2. Variability in wildlife TRVs selected for metals/metalloids in ERAs conducted from 1999 to 2011. Solid circles represent TRVs from individual ERAs. Solid squares represent the NOAEL TRV from USEPA Eco-SSLs. The chemical symbol for each metal/metalloid is shown on the vertical axis.
Download figure to PowerPoint
Few USEPA Eco-SSLs are available for organic substances, particularly persistent and bioaccumulative chemicals that are often drivers of risk at contaminated sites (e.g., Hg, PCBs, and TCDD). The ERAs we evaluated consistently (36 of 56 [64%]) relied on TRVs published by Sample et al. (1996) or independently derived TRVs based on published toxicity studies. A summary of the variability of TRVs for some selected persistent and bioaccumulative substances is presented in Table 2. The TRV variability was due to many factors (e.g., dose extrapolation assumptions or choice of primary source) and these are further discussed in the following sections. In many cases, TRVs for these organic substances vary by several orders of magnitude. As mentioned previously, TRVs were often compiled from secondary sources (42 of 56 [75%]) and less than half (18 of 42 [43%]) of the ERAs provided detailed information on the derivation of the TRVs (e.g., dose–extrapolation methods). Thus, in a majority of the ERAs it was not documented whether the original interpretations of the toxicity literature were confirmed. It is possible that previous interpretations may differ from those based on current knowledge of chemical-specific toxicology. Therefore, in a majority of ERAs it appears that limited critical reviews are being performed such that errors in the original interpretation could be overlooked.
Table 2. Toxicity reference value (mg × kg−1 × d−1) distributions for persistent and bioaccumulative substances
|Substance||Eco-SSL (NOAEL) TRVa||NOAEL TRVs from databaseb||LOAEL TRVs from databaseb|
| Hg||na||0.45c (21, 0.019–3.25)||0.9c (17, 0.18–0.91)|
| Methylmercury||na||0.0064c (32, 0.0051–2.6)||0.064c (28, 0.02–0.0.9)|
| DDT and metabolites||0.227||0.0028c (21, 0.0028–1.0)||0.028c (19, 0.028–2.62)|
| PAHs (HMW)||na||0.14d (15, 0.004–81.4)||1.4d (12, 0.04–395)|
| PAHs (LWM)||na||40e (13, 3.4–1000)||400e (6, 10.2–400)|
| PCBs||na||0.18c (30, 0.09–4.1)||1.8c (26, 0.58–41)|
| TCDD||na||1.4 × 10−5c (12, 7.0 × 10−6–4.4 × 10−5)||1.4 × 10−4c (9, 1.0 × 10−4–2.5 × 10−2)|
| Hg||na||1.0c (25, 0.027–13.2)||56f (14, 0.25–56)|
| Methylmercury||na||0.032c (29, 0.0017–0.25)||0.16c (23, 0.0084–4.0)|
| DDT and Metabolites||0.147||0.8c (16, 0.04–19.1)||4.0c (13, 0.274–191)|
| PAHs (HMW)||0.615||1.0c (20, 0.39–10)||10c (13, 3.07–38.4)|
| PAHs (LWM)||65.6||65.6 (16, 1.0–350)||114g (8, 32.8–700)|
| PCBs||na||0.14c (27, 0.0034–0.45)||0.69c (22, 0.034–3.45)|
| TCDD||na||1.0 × 10−6c (12, 8.0 × 10−8–3.6 × 10−6)||1.0 × 10−5c (9, 2.2 × 10−6–3.6 × 10−5)|
Two examples are provided demonstrating the continued use of toxicity studies that fail to meet Eco-SSL guidelines for inclusion in TRV derivation; these are for TCDD and high molecular weight (HMW) PAHs. The most frequently used avian toxicity values for TCDD are the NOAEL (1.4 × 10−5 mg × kg−1 × d−1) and LOAEL (1.4 × 10−4 mg × kg−1 × d−1) derived by USEPA (1995) and Sample et al. (1996) from a study by Nosek et al. (1992) (Table 2). These TRVs were used by all 12 ERAs in this database that examined risks to avian species from exposure to TCDD. Nosek et al. (1992) examined reproductive endpoints in ring-necked pheasants dosed weekly via intraperitoneal injection for 10 weeks. Both USEPA (1995) and Sample et al. (1996) assumed that intraperitoneal exposure was similar to oral exposure and, given the lack of other appropriate oral studies, determined that this study was appropriate for use in risk analyses. The TRVs derived from this study have been used as the basis for evaluating risks to TCDD in ERAs for almost 20 years. If the Nosek et al. (1992) data were reviewed according to USEPA's wildlife TRV derivation procedure for the Eco-SSLs it would be rejected based on the route of exposure (intraperitoneal) (USEPA 2005). Furthermore, a number of limitations for this study (e.g., the exposure route does not account for bioavailability or metabolism) have been discussed in the literature (most recently by Fredricks et al. ), suggesting this study is inappropriate (or at least overly conservative) for risk assessment purposes. In another case, 8 of 16 (50%) of the ERAs we reviewed (published between 2002 to 2011) that examined avian risks to HMW PAHs adopted NOAEL or LOAEL TRVs derived from Hough et al. (1993) (see Table 2). In this study, benzo[a]pyrene (a surrogate for HMW PAHs) was applied to pigeons via weekly intramuscular injection for 5 months and indices for reproduction were measured. Six of the ERAs (completed during 2007 or after) used the study by Hough et al. (1993) even though it was rejected by USEPA for use in its Eco-SSL for PAHs (USEPA 2007). Only one of these ERAs provided a discussion regarding the uncertainty associated with this study and/or why it was appropriate for the risk analyses. Thus, some ERAs are using substandard TRVs in the absence of more robust data without appropriately documenting the inherent uncertainties. The TRV variability illustrated in Figure 2 and Table 2 is examined further in the following sections, which highlight the many challenges facing the wildlife risk assessment community, particularly for substances with limited data sets (e.g., persistent and bioaccumulative chemicals).
Geographic TRV consistency
We examined the consistency of TRVs throughout the United States by stratifying the data according to the 10 USEPA geographic regions. A representative example showing the variability in regional selection of TRVs is presented in Figure 3. In this case, avian and mammalian NOAEL-based TRVs are plotted by geographic region. In many cases, the same approximate TRVs have been selected (Figure 3). However, there is a notable lack of consistency within and across geographic regions, as selected TRVs differ by several orders of magnitude, even following the publication of the final lead (Pb) Eco-SSL document in March 2005. For ERAs completed when the Eco-SSL TRV was available, the avian Eco-SSL TRV was used in 8 of 25 (32%) ERAs and the mammalian Eco-SSL TRV was used in 8 of 22 (36%) ERAs (see Figure 1). The lowest avian NOAEL TRVs (and sources) were 0.014 mg × kg−1 × d−1 (USEPA R9 BTAG 2009) or 0.025 mg × kg−1 × d−1 (USEPA 1999) and the lowest mammalian NOAEL TRV was 0.0375 mg × kg−1 × d−1 (USEPA 1999). This trend (i.e., wide variability in TRVs) is consistent for several of the chemicals we evaluated in the database (see Figure 2 and Table 2). These data suggest that TRVs are still being developed in an inconsistent fashion on both a regional and national scale and that USEPA's goal of national consistency has not yet been achieved.
Figure 3. Case example of intra- and inter-USEPA regional variability for Pb wildlife TRVs from ERAs conducted from 1999 to 2011.
Download figure to PowerPoint
Wildlife TRV derivation procedures
Toxicity reference values reported in ERAs in this database were derived (with limited exceptions) from acute, subchronic, or chronic oral exposure studies using typical laboratory or domesticated species (e.g., rodents, chickens, or game-farm mallards) with growth, survival, or reproduction as the common toxicological endpoints. The variance among TRVs observed in Figure 2 is due to several methodological inconsistencies among the derivation procedures as described below.
Selection of key studies
The method used by USEPA (2005) to select key studies for its Eco-SSLs is to review a number of published toxicological studies with appropriate quality, identify dose levels for each of several critical endpoints (e.g., growth, survival, or reproduction) and select a conservative NOAEL. Although USEPA (2005) methods follow an established protocol, the ERAs reviewed herein typically selected a TRV value from an established resource or identified a key study based on a literature review. ERAs either presented 1) no discussion of current toxicity literature and only reported the TRVs used in the risk assessment (24 of 56 [43%]), 2) a limited description of only the study supporting the TRV (14 of 56 [25%]), or 3) a rigorous evaluation of the toxicity literature provided in a toxicity (effects) profile (18 of 56 [32%]) for the chemicals of concern. Thus, document reviewers are not consistently provided the information necessary to verify the scientific judgments used in the ERA, and their implications for the risk characterization.
The ERAs in this database often selected TRVs based on different judgments of what constitutes the most relevant toxicological study, leading to a variety of TRV sources for the same chemical. For example, avian NOAEL TRVs for Cd in the database ranged from 0.08 to 7.3 mg × kg−1 × d−1 and were derived from multiple toxicity studies (White and Finley 1978; Leach et al. 1979; Cain et al. 1983), as well as the USEPA Eco-SSL TRV of 1.47 mg × kg−1 × d−1. Furthermore, Stanton et al. (2010) recently reviewed the USEPA Eco-SSL value for Cd and recommended an alternative NOAEL TRV of 0.7 mg × kg−1 × d−1 (derived from Mayack et al.  using kidney histology as the critical endpoint). In another example, several studies have been selected as the point of departure for deriving mammalian TRVs for TCDD (Murray et al. 1979; Heaton et al. 1995; Tillitt et al. 1996; Hochstein et al. 1998, 2001), providing a range of TRV values (see Table 2).
Blankenship et al. (2008) examined mammalian TCDD TRVs from many studies in a comprehensive review and noted a number of deficiencies that may lead to inappropriate conclusions about relevant adverse ecological effects if these TRVs are exceeded. These deficiencies included use of inappropriate species extrapolations, presence of cocontaminants in each of the studies on which the TRVs are based, different TCDD congener mixes between study exposures and site exposures, uncertainty in the toxicity equivalency factors used to adjust for potency differences among congeners, use of short-term studies on less sensitive live stages, and using nonecologically relevant endpoints as the basis for the TRV. This ERA review demonstrates the importance of critically evaluating and documenting the underlying toxicity studies used to derive TRVs; it was not always apparent in the published ERAs we reviewed whether this analysis had been completed.
In some cases, risk assessors examine dose–response relationships and adopt TRVs based on point estimates other than a NOAEL or LOAEL (e.g., ED10). Dose–response metrics avoid the limitations in the NOAEL/LOAEL approach (see review by Allard et al. ). However, a dose–response assessment was explored or conducted in only a small proportion of the ERAs (3 of 56 [5%]) in our database and only a few TRVs were estimated by this method (see Table 1). These dose–response assessments were typically applied in the detailed or baseline risk analyses to further refine ecological risks. For example, a dose–response analysis of dietary Pb in 3 ground-feeding songbird species recently was used to aid in the determination of preliminary remediation goals for the Coeur d'Alene River basin in Idaho and Washington (Sample et al. 2011). Although the calculated ED20 (9.9 mg × kg−1 × d−1) based on a multiparameter logistic model using data from a chicken reproduction study was higher than the reported LOAEL (3.52 mg × kg−1 × d−1) or NOAEL (0.35 mg × kg−1 × d−1), the resulting calculated soil preliminary remediation goal was comparable to clean up values previously derived for waterfowl and humans, and therefore considered appropriately protective of ground-feeding songbirds.
Another alternative method used in 2 ERAs (4%) was the development of a maximum acceptable toxicant concentration (or dose) (MATC). In addition to NOAEL and/or LOAEL TRVs, these ERAs calculated the geometric mean of the NOAEL and LOAEL to estimate a MATC. The MATC (that is routinely used in aquatic, soil invertebrate, and plant risk assessments) suffers from the same limitations as NOAELs/LOAELs (Allard et al. 2010; Kapustka 2008; USEPA 2005). In addition, 3 of 56 ERAs (5%) estimated LOAEL TRVs by calculating the geometric mean of LOAELs for growth and reproduction endpoints, as published in the Eco-SSL documents (similar to the Eco-SSL process for estimating NOAEL TRVs). This process, however, is uncertain due to the incorporation of multiple endpoints, exposure durations, and dose-levels. The Eco-SSL protocol for NOAEL TRV selection includes a verification that the geometric mean NOAEL is below the lowest bounded LOAEL (USEPA 2005). Verification of the appropriateness and reliability of a geometric mean based LOAEL TRV was not incorporated into the 3 ERAs, nor did they include a detailed uncertainty section to describe their potential shortcomings.
Identification of the NOAELs and LOAELs associated with the critical adverse effect can be difficult due to limited information provided in the original journal article. This can lead to alternative interpretations of the critical concentration or dose. The current convention in TRV development (such as the Eco-SSL protocol) is to convert a dietary-based concentration (ppm or mg/kg-food) to a dose-based concentration (mg/kg-body-weight per day [mg × kg−1 × d−1]) for use in exposure modeling and risk assessment. Ostensibly, this is done to standardize the threshold for all species and to accommodate simultaneous multiple routes of exposure. However, implicit in this approach is the assumption that all species react similarly to the ingested chemical, and the only differences in response among species is due to body size (e.g., metabolic rate and relative ingestion rate). Several authors have shown that this is not true (Mineau et al. 1996; Sample and Arenal 1998) and consensus is emerging that standardization through a dose metric most likely inserts inaccuracies into the threshold derivation process (Allard et al. 2010). Furthermore, the conversion from a reported threshold concentration in food to a dose metric necessitates that assumptions are made about body weights and feeding rates for the test organism in the critical study as these are rarely provided in the publication. The choice of these parameters often varies among ERAs, which can lead to alternative derived values for dose-based TRVs with potentially wide uncertainty bounds.
All ERAs in our database used some form of dose conversion to derive TRVs; however, the underlying assumptions differed even when the same critical study was used. For example, 7 ERAs (published between 2001 and 2009) selected a study by Ambrose et al. (1976) to develop a NOAEL TRV for mammals exposed to Ni. Based on different interpretations of the critical toxicological endpoint and dose conversion procedures, a range of NOAELs were determined from the same data set that span nearly an order of magnitude (8.4 to 60 mg × kg−1 × d−1). In another case, 8 ERAs (published between 2003 and 2009) developed an avian NOAEL TRV based on Laskey and Edens (1985) and the resulting TRVs ranged from 78 to 977 mg × kg−1 × d−1. These examples illustrate the inconsistency in dose-conversions that were apparent for many chemicals in our database.
McDonald and Wilcockson (2003) provided a similar example of the effect of differing dose-conversion assumptions on the derivation of a LOAEL TRV for a short-tailed shrew from a study conducted in rats. Alternative assumptions were tested for body weights (0.25, 0.35, 0.40, 0.45, and 0.51 kg), food consumption rate equations, concentration metrics (nominal vs measured), and allometric scaling (extrapolation) factors (0.75 or 0.94). This analysis resulted in LOAEL TRVs ranging from 87 to 222 mg × kg−1 × d−1 (McDonald and Wilcockson 2003). We observed the same pattern in the ERAs in our database, as differing assumptions were adopted, leading to variable TRVs (see Figures 2 and 3).
In addition, approximately 25% (14 of 56) of the ERAs we reviewed used an allometric scaling approach for adjusting TRVs for differing ecological receptors. It is notable that USEPA (2005) did not recommend allometric scaling, when establishing the methods for deriving their Eco-SSLs due to the added uncertainty in this approach. Six of the ERAs using allometric adjustments were published after the Eco-SSL guidance document (USEPA 2005) was available, thus some risk assessors are continuing to apply this approach. Furthermore, this uncertainty may be compounded nationwide due to an inconsistent set of unified dose conversion assumptions. Application of the dose-based TRV to site-specific receptors also adds uncertainty, as the same suite of concentration-to-dose conversion values need to be estimated for the target species of concern (i.e., body weight and food ingestion rate). Thus, dose-based TRVs introduce multiple layers of uncertainty, resulting in widely variable TRVs even if based on the same critical study.
Uncertainty factors have been used to address variability in species sensitivity, to account for differences between short- and long-term exposures, or laboratory to field extrapolations (Chapman et al. 1998). In the ERAs we reviewed, 61% (34 of 56) applied uncertainty factors in the derivation of TRVs and only 1 of 56 ERAs included an analysis of species sensitivity as part of its TRV selection process. A variety of factors were used to account for any one source of uncertainty. For example, 231 TRVs in the database were developed using an uncertainty factor of either 2 (<1%), 3 (2%), 5 (25%), or 10 (73%) to extrapolate from a LOAEL to a NOAEL. This inconsistency in the use of uncertainty factors has been pointed out previously (Duke and Taggart 2000; McDonald and Wilcockson 2003; Allard et al. 2010), and continues to be pervasive in ERAs conducted in the past decade.
- Top of page
- TRV COMPILATION
- TRV ANALYSIS
- Supporting Information
Despite the publication of USEPA's Eco-SSLs that were intended to standardize the derivation and use of screening level wildlife TRVs in Agency-directed ERAs, the TRVs have not yet become standardized. Our analysis of TRVs used in EPA-directed ERAs over the past decade highlights this variability in wildlife toxicity thresholds used in the United States, and suggests that there is a need for improved consistency in the practice of ERA. There are substantial differences, regionally and nationally among ERAs, primarily from inconsistent derivation methods. The end result is TRVs spanning an order of magnitude or more for many chemicals. Our review shows that the primary reasons for differing values are the choice of seminal toxicity studies on which to base the TRV, the methods used to convert the test concentration into a dose, and the use of differing uncertainty factors.
We recognize that every ERA is subject to a range of technical challenges and that selection of alternative TRVs (i.e., other than USEPA Eco-SSLs) is a common and useful practice under certain circumstances, particularly for higher tier (beyond screening) risk assessments. However, as evident in this review, TRVs adopted in recent ERAs are inconsistent and widely variable (as shown in Figures 2 and 3) and generally lack justification for selection of a particular TRV. Furthermore, this analysis suggests that adoption of USEPA Eco-SSL TRVs has yet to become prevalent in ERAs, even in screening-level assessments directed by the Agency. Under the current approach, a location could be determined to be highly risky to wildlife and not pass the screen, whereas another location with the same wildlife species and chemical concentrations of concern could be determined to pose no unacceptable risk. This erodes confidence in the risk assessment process and is a disservice to risk managers who make decisions based on the belief that the risk assessment provides meaningful results.
The risk assessment community would benefit from additional direction on the identification of scientifically appropriate TRVs and derivation procedures, particularly for substances with dated or no preexisting consensus-based TRVs (i.e., persistent and bioaccumulative chemicals). Given the number of uncertainties in the NOAEL and/or LOAEL approach, we support recent suggestions to adopt a dose–response approach for determining toxicity thresholds whenever possible (Chapman et al. 1996; Crane and Newman 2000; Allard et al. 2010; Landis and Chapman 2011). In view of the many complexities inherent in dose–response assessment, detailed guidelines should be developed to demonstrate appropriate TRV derivation using these methods. If uncertainty factors are to be used, a consistent set of guidelines should be followed, such as has been done for European risk assessments (ECB 2003). However, it is our opinion that other methods of incorporating uncertainty should be used instead of arbitrary uncertainty factors, such as development of species sensitivity distributions and error bounds on point estimates (Allard et al. 2010). If, in fact, data are so limited that even uncertainty estimates cannot be provided, the risk manager may choose to apply a margin of safety to the risk determination, but it is our opinion that this is a qualitative decision and not a scientifically based result of the risk assessment. Furthermore, a central clearinghouse of wildlife TRVs (similar to the USEPA Integrated Risk Information System [IRIS] used for human health risk assessment) would aid ecological risk assessors in selecting peer-reviewed, consensus-based wildlife TRVs with technical background documents describing their derivation. Although a website is already provided for the USEPA Eco-SSLs, it could be expanded to include recommendations on TRVs for other common chemicals. We strongly recommend that a renewed effort to standardize wildlife TRVs be initiated to reduce the variability in ecological risk assessments conducted throughout the United States.