• Tissue residue benchmarks;
  • 2,3,7,8-TCDD;
  • Toxicity equivalent quotient;
  • Species sensitivity distributions


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References

Tissue residue-based toxicity benchmarks (TRBs) have typically been developed using the results of individual studies selected from the literature. In the past, TRBs have been developed using a point estimate (e.g., LC50 value) reported in a study on a single species deemed to be most closely related to the receptor of interest. Despite attempts to maximize the protectiveness and relevance of TRBs, their relationship to specific receptors remains uncertain, and their general applicability for use in broader ecological risk assessment contexts is limited. This article proposes a novel framework that establishes benchmarks as distributions rather than single-point estimates. Benchmark distributions allow the user to select a tissue concentration that is associated with the protection of a specific percentage of organisms, rather than linked to a specific receptor. A methodology is proposed for searching, reviewing, and analyzing linked, tissue residue effect data to derive benchmark distributions. The approach is demonstrated for contaminants having a dioxin-like mechanism of toxic action and is based on residue effects data for 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD) and equivalents in early life stage fish. The calculated tissue residue benchmarks for 2,3,7,8-TCDD toxic equivalency (TEQ) derived from the resulting distribution could range from 0.057- to 0.699-ng TCDD/g lipid depending on the level of protection needed; the lower estimate is protective of 99% of fish species whereas the higher end is protective of 90% of fish species.


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References

Traditional determinations of aquatic toxicity rely on exposing test organisms to substances contained in the aqueous, sediment, or food phases of test media. However, quantifying the dose of a chemical delivered to aquatic organisms using chemical levels measured in external media has proven to be problematic. External physicochemical factors (e.g., pH, organic carbon, alkalinity, salinity, or temperature) affect the bioavailability of chemicals and, ultimately, the dose delivered to the organism (McCarty and Mackay 1993). This problem is compounded by the potential for multiple routes of exposure to the contaminant.

The internal concentration of a chemical (tissue residue) is considered to be a much better surrogate for the dose at the site of toxicological action than the external concentration because it implicitly reflects chemical bioavailability, assimilation and metabolic efficiencies, and multiple routes of exposure. Measured tissue residues also provide better evidence that the chemical has indeed been accumulated, and adverse effects in these organisms are due to accumulation of that chemical. As a result, researchers have increasingly advocated the use of tissue residues to more accurately predict dose (Friant and Henry 1985; van Hoogan and Opperhuizen 1988; McCarty et al. 1991; Landrum et al. 1992; McCarty and Mackay 1993; Barron et al. 2002; Landrum et al. 2003). Accordingly, the use of a tissue residue-based framework for identifying chemical levels associated with adverse effects to aquatic biota is attractive because it eliminates the uncertainties associated with estimating the dose from external media (Meador et al. 2002).

The impetus for this research lies in a need for toxicity benchmarks that delineate the potential for adverse biological effects resulting from bioaccumulation of chemicals in ecological risk and damage assessments, total maximum daily load (TMDL) development, dredged-sediment disposal evaluations, and remediation activities. Although analytical methods for measuring chemicals in tissues are available, better methodologies are needed for deriving tissue residue-based toxicity benchmarks (TRBs). This is particularly true for assessments of potential biological impacts to aquatic receptors.

TRBs have typically been derived by risk assessors as point estimates of levels of concern for specific ecological receptors (i.e., species). In many cases, the lowest-observed effect residue (LOER) or no-observed effect residue (NOER) from a single, published study is used to assess risks associated with measured tissue residues. Reliance on a single, published result does not consider variability in effect residue relationships associated with different test species, exposure regimes, or chemicals used in the study of uncertainty and precludes consideration of uncertainty in its application to risk assessments. Because differences in sensitivity between the selected species and the receptor of interest are poorly understood, risk assessors often seek to maximize the protectiveness of the TRB by assigning a value associated with the most sensitive toxicity endpoint of a species for which data are available and deemed to be most closely related to the receptor species. Despite this attempt to maximize the protectiveness and the relevance of the TRB, its relationship to specific receptors remains uncertain, and its general applicability for use in ecological risk assessments is limited.

In this article, a simple and straightforward methodology is proposed for searching and reviewing published literature on linked, tissue residue effect data. Specific recommendations are provided for screening, analyzing, and presenting the linked residue effect data as species sensitivity distributions (SSDs). An SSD is a statistical distribution that captures the variation in toxicological sensitivity among a given set of species to a certain chemical. The SSD is expressed as a cumulative distribution function (CDF) composed of specific effect-concentration metrics (e.g., LC50, EC50, LOEC, or NOEC values) obtained from toxicity studies on the x axis and the cumulative proportions of affected species (p) on the y axis (Posthuma et al. 2002). Benchmarks or criteria can be derived from an SSD by selecting a percentile (p) to protect 1 — p percent of species on the y axis and reading off the corresponding concentration on the x axis as dictated by the SSD. Therefore, instead of selecting a single-point benchmark from residue data reported in a single study, a risk manager can use the SSD to select a residue-based effect benchmark that corresponds to a given level of desired protectiveness (e.g., protection of 95% of species). The TRBs derived in this manner are likely more appropriate because they reflect the entire body of available linked residue effect data, and the SSD offers risk managers substantial flexibility in selecting TRBs.

In the current work, the SSD benchmark-derivation methodology is demonstrated for contaminants that have a dioxin-like mechanism of toxic action. The dataset used in the demonstration consists of fish egg and embryo residues of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) or dioxin, furan, and PCB congeners with published TCDD toxicity-equivalency factors (TEFs) referred to as dioxin-like compounds. TCDD and dioxin-like compounds were selected for this demonstration because (1) a substantial body of linked effect residue data exists for these compounds that are well suited for exploring the utility of a distribution-based approach for setting residue benchmarks, (2) they are highly lipophilic and resistant to metabolism, resulting in a propensity to accumulate in aquatic organism tissues and biomagnify to higher trophic levels of the food web (Eisler 2000), and (3) their presence in sediments of ports and harbors has made management of these sediments problematic and expensive (NRC 2001).


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References

Species-sensitivity distributions in the proposed methodology are based on the geometric means of the NOER and the LOER values associated with early life stage toxicity for several fish species. An SSD based on 50% lethal residues (LC50) is also presented for comparative purposes.

Data compilation

A search for data that concurrently report biological effects and whole-body residues of chemicals of interest in aquatic organisms can be accomplished by consulting the U.S. Army Corps of Engineers Environmental Residue Effects Database (ERED) (USACE 2004) and the U.S. Environmental Protection Agency (USEPA) Office of Research and Development's effect residue database (Jarvinen and Ankley 1999). These 2 databases provide information on tissue residues in various organisms exposed to individual compounds and the magnitude and nature of biological effects associated with those residues. The database investigations should be augmented by a thorough literature search covering primary scientific journals (e.g., DialogTM database) to confirm the accuracy of database citations and identify new studies published since the latest database update. Gray literature studies are not generally recommended for inclusion because of the uncertain nature and level of peer-review.

Identified studies must be screened before statistical analyses using a set of specific criteria designed to retain only those inputs that will maximize the relevance and the reliability of the derived benchmarks. The specific criteria used to screen the data in this study are discussed in the Data screening criteria section. If a benchmark is to be developed for a different substance, the list of screening criteria that were used for screening studies with TCDD and dioxin-like compounds might need to be modified to account for any chemical-specific considerations (e.g., mechanism of action, speciation, and target tissue).

Data screening criteria

Laboratory studies with single chemical exposures only—Only data that are obtained from well-designed and controlled laboratory studies with exposures to the individual chemical of interest should be used to generate an SSD. Data from studies with exposure to multiple chemicals or field-collected sediments or waters should not be used because of the potential confounding effects associated with other chemicals present in the mixture. In some cases, such as in this study, data can be pooled for different compounds that share a common toxic effect and mechanism of toxicity. In such cases, however, it is important that the relative toxicities of those compounds be accounted for in generating the SSD curve.

For this study, data for all compounds having a dioxin-like mechanism of toxicological action (aryl hydrocarbon [Ah] receptor active compounds) and published TCDD TEFs were considered. These compounds include the 17 chlorinated dioxins and furans and 12 PCB congeners having TEFs listed by the World Health Organization (van den Berg et al. 1998). Most of the studies that met the screening criteria and retained in the analysis were conducted with TCDD. Although the SSD is primarily derived using TCDD responses, it may appropriately be used to evaluate the total risk (i.e., toxic equivalent quotient [TEQ]) associated with all Ah-receptor active compounds that have published TEFs by using the following equation (van den Berg et al. 1998):

  • equation image((1))

where TEQ is the toxic equivalent of 2,3,7,8-TCDD, PCDDn is polychlorinated dibenzo-p-dioxin congener concentration, PCDFp is polychlorinated dibenzo-p-furan congener concentration, PCBq is polychlorinated biphenyl congener concentration, and TEFn,p,q is the toxic equivalency factor for appropriate individual PCDD, PCDF, and PCB congeners, respectively.

Ecologically significant endpoints—In calculating an SSD, preference should be given to those effect residue data that link residues with organism effects that can be most confidently associated with ecological consequences at the population level. These potential endpoints include significant reductions in survival, growth, or reproduction. Although the SSD approach could potentially be applied to generate benchmarks based on other physiological responses and effects (i.e., metabolic induction and avoidance behavior), the ecological relevance of those effects in terms of affecting the survival and reproduction of species are usually harder to ascertain (McCarty and Mackay 1993).

Exposure route and duration—It is important that the exposure route and duration be carefully evaluated in screening studies for inclusion before generating an SSD. Namely, the exposure duration should be sufficient to allow for the internal distribution of the chemical to reach the target organ or to equilibrate within the organism (van Wezel et al. 1995; Landrum et al. 2004). A static equilibrium is preferred, but a well-parameterized dynamic equilibrium (showing empirical evidence of a steady state) can be used to determine the approximate time to steady state. To minimize variability associated with the use of residue effect data, the exposure route should be considered (Barron et al. 2002). For ecological risk assessment, it may be preferable to generate residue-based SSDs using studies that report effects resulting from chronic exposures through dietary, waterborne, or maternal exposure routes.

Egg and embryo development was the preferred endpoint in this study; therefore, maternal TCDD uptake and transfer to eggs was deemed the most ecologically relevant exposure pathway, followed by direct absorption. Because there is evidence that TCDD effect residues in early life stage fish are independent of the route of exposure (including egg injection) (Walker et al. 1992), studies were not excluded based on this criterion. Studies using maternal, egg injection, and water exposures were used for the final SSD calculation.

Measured data only—To minimize uncertainty, only those residue and effect data that are measured using acceptable methods and directly reported in the results should be used to generate SSDs. Use of estimated or predicted tissue residue or effect data can introduce potentially large uncertainties into the SSD curve. Similarly, when data are represented graphically, concentrations should not be estimated from the axes.

Whole body versus organ residues—The SSDs can be generated using residues reported for specific tissues or using whole-body residues. In making the choice, it is important to carefully consider the mechanism of toxicological action of the contaminant and how the resulting TRB will be incorporated into an ecological risk assessment.

In the current framework, an effect residue distribution is developed for the life stage of fish that is most sensitive to the Ah-receptor mediated toxicity (egg and embryo). Therefore, only residues reported on a whole egg and embryo basis were considered further in the derivation of the effect residue distribution. Because of their lipophilicity, however, effect residues of TCDD and dioxin-like compounds in fish eggs could readily be related to maternal tissue concentrations after lipid normalization. For nonpolar organic compounds, the ratio of chemical on a lipid-normalized basis was found to be approximately 1:1 egg to adult fish. (Russell et al. 1999).

Dose-dependent response—Only data from studies with demonstrated dose-dependent response should be included in an SSD. The residues reported in candidate studies should support a dose-dependent response in which significant effects are consistently observed at higher residues. This screening ensures that the derived benchmark relies on data that indicate incremental effects and that the observed response is related to the concentration of chemical in the tissue of the organism.

NOER and LOER—Only 1 toxicity metric (e.g., LOEC and LC50) should be used for development of the SSD. If the geometric mean of NOER and LOER is to be used to generate the SSD, only studies that simultaneously report these endpoints should be included in the database. Many studies report one or the other and are, therefore, rejected based on this criterion. Although it has been suggested that NOER values can be estimated from LOER by applying a correction factor (Sample et al. 1996), it is not recommended in this proposed framework because it introduces additional uncertainty. In the current study, the geometric mean of the LOER and NOER was calculated by taking the square root of the product of the 2 values.

Distribution fitting

The SSD can be fitted to several statistical distributions including normal, lognormal, logistic, or triangular (Posthuma et al. 2002). The logistic distribution is used because de Zwart and Sterkenburg (2002) found that lognormal or log-logistic models are generally the most appropriate descriptors of SSDs. The fitting of data to a particular distribution type and model allows the calculation of benchmarks associated with percentiles between data points, as well as to extrapolate beyond the data range. The curve fitting of effect residue data to a generalized linear model (GLiM) was performed using SAS® (SAS Institute, Cary, NC, USA) (Bailer and Oris 1997). Residue data were fitted to a logistic distribution with a logit-link function and a binominal-error distribution.


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References

The application of the screening criteria to the universe of data for TCDD and dioxin-like compounds yielded 26 linked residue effect data pairs (Table 1). All data are reported as ng TCDD/g lipid (using calculated TEQ values, when appropriate). After the data set was screened using the 7 criteria, further trimming of the initial data set was conducted to eliminate records where a geometric mean of the NOER and LOER could not be calculated (e.g., missing a NOER or LOER value). In addition, when multiple records for the same species were encountered, only the record with the lowest geometric mean of the NOER and the LOER was retained for calculating the SSD. Final trimming yielded 10 data points for generating the SSD for fish exposed to TCDD and dioxin-like compounds.

Figure 1 depicts the residue data points and the fitted logistic distribution for the data set describing the geometric mean of the NOER and LOER values from Table 1. The fit of the SSD was highly significant (p < 0.0001; F = 341) and the GLiM statistics were as follows: df= 8; β0 (slope) = -1.85; β1 (intercept) = 2.21; SE β0 (SE of the slope) = 0.494; SE β1 (SE of the intercept) = 0.045; and the t (between slope and intercept) = -0.87.

For comparative purposes, Figure 2 summarizes the GliM-fitted distributions for LR50 values. The fit of the model was also highly significant (p < 0.0001; F = 345), and the fitting statistics were as follows: df = 8; β0 (slope) = -2.11; β1 (intercept) = 2.01; SE β0 (SE of the slope) = 0.552; SE β1 (SE of the intercept) = 0.416; and the t (between slope and intercept) = -0.90.

Table Table 1.. Residues of 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD) and dioxin-like compounds in fish eggs or embryos showing no observable effect residue values (NOER), lowest observable effect residue values (LOER), and their associated geometric mean (GM), and median lethal residue (LR50) valuesa
Common nameGenus speciesNOER ng/g wwLOER ng/g wwGM ng/g wwLR50 ng/g wwRoute of exposureLipid in eggs (%)GM ng/g lipidLR50 ng/g lipidUsed for SSDReference
  1. a ww = wet weight; X = GM SSD; Y = LR50 SSD.

  2. b Lipid based on Johnson et al. (1998).

  3. c Lipid based on Walker et al. (1994).

  4. d Based on toxic equivalent concentration.

  5. e Lipid based on egg lipid content of 5.4 mg/egg and mean egg weight of 82.5 mg/egg from Bellastrazzi et al. (2003).

  6. f Lipid based on Elonen et al. (1998).

Brook troutSalvelinus fontinalis0.0840.1560.1140.127Maternal6.81.681.87X, YJohnson et al. 1998
Brook troutS. fontinalis0.1350.1850.1580.200Water6.82.322.94 Walker and Peterson 1994b
Channel catfishIctalurus punctatus0.3850.8550.5740.644Water4.811.9513.42X, YElonen et al. 1998
Fathead minnowPimephales promelas0.2350.4350.3200.539Water2.413.3222.46X, YElonen et al. 1998
Japanese medakaOryzias latipes0.4550.9490.6571.110Water2.922.6638.28X, YElonen et al. 1998
Lake herringCoregonus artedii0.1750.2700.2170.902Water6.63.2913.67X, YElonen et al. 1998
Lake troutSalvelinus namaycush0.0350.042Water8.00.53YGuiney et al. 1996c
Lake troutS. namaycush0.074Water8.00.93 Walker et al. 1996c
Lake troutS. namaycush0.097Water8.01.21Walker et al.1996cd
Lake troutS. namaycush0.0230.0500.0340.058Maternal8.00.420.73XWalker et al. 1994
Lake troutS. namaycush0.0340.0400.0370.069Water8.00.610.86 Walker et al. 1994
Lake troutS. namaycush0.0440.0550.0490.080Injection8.00.611.00 Walker et al. 1994
Lake troutS. namaycush0.0340.0550.0430.065Water8.00.540.81 Walker et al. 1991c
Lake troutS. namaycush0.054Water8.00.68 Zabel et al. 1995cd
Lake troutS. namaycush0.085Water8.01.06 Zabel et al. 1995c
Lake troutS. namaycush0.0330.0440.0380.047Injection8.00.480.59 Walker and Peterson 1991c
Northern pikeEsox lucius1.1901.8001.4642.460Water4.234.8558.57X, YElonen et al. 1998
Rainbow troutOncorhyncus mykiss0.200Injection6.53.08YWalker et al. 1996e
Rainbow troutO. mykiss0.362Injection6.55.57 Walker et al. 1996de
Rainbow troutO. mykiss0.2790.439Water6.56.75 Walker et al. 1992e
Rainbow troutO. mykiss0.1940.2910.2380.421Injection6.53.666.48 Walker et al. 1992e
Rainbow troutO. mykiss0.230Injection6.53.54 Walker and Peterson 1991e
Rainbow troutO. mykiss0.1760.2240.1990.226Injection6.53.053.48XWalker and Peterson 1991de
White suckerCatastomus commersoni0.8481.2201.0171.890Water2.540.6975.60X, YElonen et al. 1998
ZebrafishDanio danio0.4242.0000.9212.610Water1.754.17153.53XElonen et al. 1998
ZebrafishD. danio0.7001.5001.0252.500Water1.760.28147.06YHenry et al. 1997f
thumbnail image

Figure Figure 1.. Logistic species sensitivity distribution for 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD) based on the geometric means of no observed effect residue and lowest observed effect residue values. The 95% confidence intervals associated with the fitted model are depicted by the outside lines.

Download figure to PowerPoint

Selection of tissue residue benchmarks

Several options for selecting TRBs from the SSDs are presented in Table 2. The benchmark ultimately selected depends on risk management considerations, levels of protection required, and degree of conservatism desired. For example, if a TRB is to be selected to be protective of 95% of fish species (i.e., 5% affected level), the benchmark would be 0.321 ng TCDD/g lipid (based on the geometric means of NOERs and LOERs). Some uncertainty exists about the reported mean value, however, as evidenced by the upper and lower confidence limits (Table 2 and Figure 1). Therefore, a risk management decision must be made as to the conservatism that is desired in applying this TRB.


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References

The SSDs obtained using the described methodology yield TRBs that are intended for broad application in ecological risk assessments and offer specific levels of protection against adverse toxicological effects associated with exposure to a chemical of interest for a range of fish species. Results from the described methodology provide risk managers with the distinct advantage of allowing residue effect data from multiple studies and species to be considered in selecting a tissue residue-based evaluative benchmark, rather than relying on data from a single study. The range of fish species protected, or the level of the protection afforded, depends on the magnitude of the percentile chosen from the ranked effects residue distributions for use as the TRB.

thumbnail image

Figure Figure 2.. Logistic species sensitivity distribution for 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD) based on lethal residue associated with 50% mortality (LR50) for various fish species. The 95% confidence intervals associated with the fitted model are depicted by the outside lines.

Download figure to PowerPoint

Table Table 2.. Tissue-residue benchmarks for the protection of fish exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD) and dioxin-like compounds at selected species protection level. Benchmark options include the predicted mean values as well as the lower (LCL) and upper (UCL) 95% confidence limitsa
 Benchmark value (ng TCDD/g lipid)
  1. a NOER = no observable effects residue; LOER = lowest observable effects residue; LR50 = median lethal residue.

Species protection levelLCLMeanUCL
Geometric mean of NOER and LOER

The SSD distribution presented in this paper for use in assessing the risk of TCDD and dioxin-like compounds incorporates residue effect data for lake trout and other salmonids that are substantially lower than residues associated with effects in the other fish species. Inclusion of the salmonid species in the curve results in TRBs that are likely to be conservative for many other fish species. It is recognized that species for which residue effect data does not exist may be the focus of risk assessments and that there is a potential for these unstudied species to have sensitivities exceeding those of the salmonids. Consequently, it is difficult to make assumptions about the relative toxicity of the species for which observations are available to those that could be of interest in field applications in deriving and presenting the SSD and resulting TRBs for assessing risk of TCDD and dioxin-like compounds.

In the past, many jurisdictions have derived SSD-based benchmark values (e.g., U.S. water quality criteria) (Sample et al. 1996) using the 95th percentile as the desired level of protection. Other applications, such as U.S. National Oceanic and Atmospheric Administration's sediment screening values (Long et al. 1995), have adopted the 90th percentile. The ultimate decision regarding the appropriate protection level used to identify an appropriate benchmark is left to risk managers and policy makers. Because benchmarks derived from SSD-based effect residue distributions are based on statistical representations of available data for multiple species, they should be cautiously applied within species-specific risk assessments (e.g., risk to endangered species). In some cases, it may be more appropriate to develop the distribution using data from studies that have reported residue effect concentrations for the target species or from closely related phylogeny.

Although SSDs can be derived for classes of chemicals sharing a common mechanism of toxic action, they cannot account for nonadditive interactions with other chemicals that are likely to occur in field mixtures (Eisler 2000). As a result, adverse responses in the field might be observed at lower or higher residues of the chemical than those reported in the single chemical exposures conducted in laboratories. This limitation in application is attendant to any TRBs that are based on the actions of single chemicals or chemical classes; this analysis does not seek to reconcile this difficulty in the application here. To address this issue, other lines of evidence based on site-specific information such as ecological surveys, in situ experiments, and bioaccumulation studies should be considered in the application of TRB values. This information may identify populations that have greater sensitivity to contaminants or lesser ability to metabolize these contaminants. Differences in metabolism are assumed to be considered in the development of the TRB; however, the presence of additional contaminants in field situations may alter the metabolic ability of an organism through up regulation or down regulation of metabolic enzymes (Rice and Roszell 1998). The use of in situ experiments or bioaccumulation studies may provide insight regarding the potential interactions of contaminants. Because very little is understood regarding these mechanisms of interaction, additional levels of protection may be necessary.

The SSD methodology described in this paper allows identification of TRBs for use as numerical guidelines in interpreting risks associated with chemicals measured in field-collected or laboratory-exposed organisms. TRBs developed in this manner can also be related to other media using bioconcentration factors, using biota-sediment accumulation factors (BSAFs), or applying more sophisticated food web and exposure models, as appropriate. For example, in a screening-level risk assessment, tissue concentrations in fish can be predicted from sediments using BSAFs (Wong et al. 2004). The resulting tissue concentrations can then be compared with a specific TRB derived from the SSD. In a more detailed assessment, such as that conducted by von Stackelberg et al. (2002) in which PCBs were modeled in a food web using the Gobas model (Gobas 1993), TRBs can be considered using probabilistic considerations of exposure.

It is clear from the results of this modeling effort that the wide confidence intervals associated with calculated TRBs make probabilistic consideration of sensitivities (i.e., quantitative uncertainty in the TRB value) problematic in risk assessment applications. To account for this, a level of protection should be preselected in accordance with risk management objectives, and a mean value associated with that level used as the TRB in the assessment.

Uncertainties associated with species sensitivity distribution development

Toxicological metric used in SSD modeling—To assess risks to ecological receptors, assessors must select an appropriate metric (or metrics) of toxic response (e.g., NOER, LOER, or LR50) for use in deriving a TRB that would serve as a threshold for inferring risk due to a chemical. NOERs and LOERs tend to be most commonly applied in ecological risk assessments (Russell et al. 1999). However, NOERs employed as acceptable levels are often criticized for potentially underestimating the “true” threshold, and LOERs employed as lower limits of potential concern are criticized for overestimating the “true” threshold (Chapman et al. 1996; Crane and Newman 2000). In any case, both NOER and LOER values share the limitation that they reflect the experimental design employed to test a specific hypothesis and, therefore, may not be readily reproducible across studies, even across studies using the same species (Landis and Yu 1995). For these reasons, use of the geometric mean of the LOER and NOER for relating species sensitivities in generating SSDs is proposed solely as a compromise. The appropriateness of the degree of ecological protection actually conferred by selecting one metric over another for use as in generating SSDs (and TRBs) remains ill defined. The final selection of a metric may ultimately be based solely on the risk managers' perception of the relative advantages or be dictated by stakeholder preferences.

TRBs derived from the LR50SSD for TCDD and dioxin-like compounds did not differ significantly from those based on the geometric means of NOER and LOER values. The similarity in the TRB values can be explained by the steep dose-response associated with dioxin-like toxicity. It is probable that SSD-based TRBs could differ significantly for contaminants with less-steep dose-response relationships. The confidence intervals associated with mean TRB levels for TCDD and dioxin-like compounds were narrower using the geometric means of NOER and LOER than using LR50 values and, therefore, were deemed preferable for use in risk assessment.

Consideration of exposure route and duration—The appropriateness of the use of a whole body (or organ) tissue residue measured at a given exposure time point to estimate a threshold for adverse effects relies on the relationship between tissue concentration and the concentration at the locus of toxic action being constant. Chemicals that partition between various environmental compartments (i.e., sediment, water, and biota) and between tissues of exposed organisms (e.g., lipid pools) to approach a steady state in the different compartments over time are best suited for evaluation using an SSD approach. Because high log octanol-water coefficient (Kow) (hydrophobic), organic chemicals (such as TCDD) are known to partition in this manner, the SSD approach outlined in this article is particularly well suited to evaluating accumulations of those compounds. Chemicals that do not partition in this manner such as cationic metals may cause toxicity that is not reflected in measured whole-body residues (Rainbow 2002).

The critical body (or organ) residue for adverse effects should, theoretically, be largely independent of the exposure route (absorption, ingestion, and injection) and rate of chemical uptake (Mackay et al. 1992; Kane-Driscoll and Landrum 1997). However, tissue residues are a function of the rates of chemical uptake (flux) and loss (metabolism and elimination), and tissue residue concentrations associated with adverse effects are not entirely independent of exposure route and duration. Effective tissue residues obtained from short-term exposures are generally expected to be higher than concentrations associated with effects occurring at longer exposures (Lee et al. 2002; Landrum et al. 2004).

The extent of partitioning is critically important in evaluating the relative utility of toxicological data considered for deriving SSDs. Equilibrium between uptake and elimination may take several weeks depending on the chemical, the size of the organism, its behavior, and the exposure route (Boese et al. 1997). If an exposure pathway leads to concentrations of the chemical at the locus of toxic action and results in an effect before steady state is approached in the organism, measured whole-body residue levels would underestimate the actual effective residue at the site of toxic action. For example, if disruption of gill function is the mode of toxic action of a chemical and expression of this disruption is dependent on a critical residue within the gill tissue, then absorption pathways could lead to expression of toxicity due to accumulated residues at the gill before dynamic equilibrium is attained throughout the body. Toxicity for that same chemical when exposure occurs via ingestion would be dependent on attaining equilibrium between the gill and the remainder of the body and would better approximate the threshold residue at the site of toxic action. Local intoxication of the gills and incomplete internal distribution of PCBs was offered by van Wezel et al. (1995) as a possible explanation for lower lethal body burdens measured in fathead minnows that died following ⩽20 h of exposure to aqueous Aroclor mixtures than in minnows surviving the same exposures for 4 to 10 d.

In general, data from chronic studies are preferred for use in SSDs because they minimize the influence of incomplete distribution of chemicals within organisms and maximize the relevance of the resulting TRBs for interpreting risks associated with field-measured residues. In addition, the variation in residue effect data as a result of different exposure pathways should be considered when selecting studies used to support the SSD. In this study, residue data with broadly consistent exposure regimes (i.e., eggs and embryos exposed for 48 h) were available in the literature. Furthermore, TCDD effect residues associated with fish early life stage toxicity have been shown to not differ between exposure pathways (maternal, aqueous, or injected). Thus, time and route of exposure was a relatively small source of uncertainty. However, in other efforts where exposure regimes differ, these parameters could introduce a large degree of uncertainty to the resultant SSD. In most cases, using residue effect data associated with a similar or common exposure to generate SSDs is preferable to combining data obtained from studies using different exposure pathways.

Field application of TCDD toxic equivalency SSD

It is difficult to validate environmental quality benchmarks that are intended for broad application (such as the non-species-specific TRBs that are introduced in this article). To demonstrate the application of the methodology and TRB developed in this study, the TRBs derived for TCDD equivalents were used to evaluate the potential for effects to 6 species of commercially and recreationally important fish in and around New York-New Jersey harbor, USA, a system known to be contaminated with dioxins, furans, and PCBs. The TRB values derived were converted from egg concentrations to whole-fish concentrations using the relationship described by Russell et al. (1999) who indicate that the concentration of chlorinated organic compounds in eggs, on a lipid-normalized basis, can be estimated from adult fish concentrations using a 1:1 ratio.

A published data set of dioxin, furan, and lipid concentrations in reproductive-sized fish collected from the harbor (n = 43) (Skinner et al. 1997) was used to calculate lipid-normalized concentrations for each fish. Sampling and analytical methods are described in the original report. Briefly, standard fillets (skin on and scales off) were taken from fish sampled from various areas of the estuary. The 2,3,7,8-substituted dioxin and furan concentrations were determined by USEPA Method 8290. The method for lipid analysis was not reported. Unfortunately, agonist PCB congeners were not quantified in the harbor study and, therefore, could not be considered in this exercise. Following conversion to molar concentrations, the TEFs reported for fish in van den Berg et al. (1998) were used to calculate the total TEQ for the accumulated dioxin and furan mixture in each specimen using Equation 1. The calculated TEQs are summarized in Table 3 and compared with TRB90, TRB95, and TRB99 (tissue residue benchmarks associated with the protection of 90, 95, and 99% of species, respectively) levels in Figure 3.

If TRB95 were to be employed as a benchmark for fish in the system, then the analysis would suggest that early life stage mortality due to dioxin-like toxicity is not likely for most of the harbor fish examined. This is because accumulations for all species, except the larger-sized bluefish (>56 cm) and striped bass (>76 cm), were below the TRB95. Moreover, white perch, tautog, American eel, and smaller bluefish (30–56 cm) were at or below the TRB99. The upper limit of the interquartile range (i.e., 75th percentile) of residues in larger bluefish exceeded the TRB95 and the upper limits of the distribution exceeded the TRB90.

Closer examination of the data revealed that the 2 individual bluefish and the single striped bass that exceeded the TRB95 level were collected from areas of the inner harbor where chemical levels (including the Ah-receptor agonists) are elevated with respect to the rest of the estuary. Had the data been available to include the additional contribution of the unmeasured agonist PCBs, it is possible that the range of TEQs measured in additional striped bass taken from the inner harbor may also have exceeded the TRB95. Because bluefish and striped bass are upper trophic level predators, they tend to accumulate biomagnifying compounds and are most likely to exceed levels of concern. The application illustrates the level of potential risk of dioxin-like toxicity to early life stages of bluefish and striped bass relative to the other species sampled. It does not equate to a probability of effect; to assess that probability, specific data on the sensitivity of early life stage of these species to dioxin like toxicity are required.

Although this application does not specifically validate the TCDD SSD, the SSD approach appears to yield a useful TRB that relies on a statistically robust treatment of laboratory data and appears to be selective in flagging potential concerns that are both ecologically and geographically relevant. Furthermore, this application illustrates how the SSD can be applied to screen field data sets in which (1) the sensitivities of individual species are unknown, (2) general protection of fish is the management goal, rather than protection of a specific targeted species, and (3) lipid-normalized data for individual toxicologically important compounds (in this case, Ah-receptor agonists) are available for only specific tissues assuming that residue levels are near equilibrium within the fish.

Table Table 3.. Calculated toxi-equivalent quotient values and lipid content values for individual reproductive-sized fish collected from New York-New Jersey harbor (Skinner et al. 1997). Toxic equivalents were calculated on a molar basis (TEQmol) rather than wet-weight concentration
SpeciesLipid weight%, wetTEQ ng/g, wetTEQmol ng/g, wetTEQmol ng/glipid, wetSpeciesLipid weight%, wetTEQ ng/g, wetTEQmol ng/g, wetTEQmol ng/glipid, wet
American eel7.90.002670.002260.0287Striped bass6.40.01220.007560.118
 17.60.001040.000560.00321(61–76 cm, TL)2.50.009250.007710.309
 22.20.01820.01660.0748 3.20.01160.009850.308
 15.30.01310.01220.0796 4.90.004110.00240.0491
 9.80.002830.002250.023 4.30.006350.004980.116
 4.80.000740.000540.0112 3.90.003240.002320.0595
Bluefish2.50.001640.001270.0507 3.10.002160.001250.0404
(30–56 cm, TL)8.30.004470.003110.0375 20.002280.001520.0758
 10.30.007050.005040.0489Striped bass2.40.007560.006420.267
 14.20.002870.002120.0149(>76 cm, TL)4.80.01180.01050.218
 11.40.004880.003850.0338 2.20.008590.007330.333
Bluefish7.40.06480.06210.839 3.60.004650.003570.099
(>56 cm, TL)90.03820.03610.401Tautog0.70.000470.000320.0451
 5.30.009020.007360.139 2.80.002070.001510.0538
 5.30.00490.003830.0723 1.40.001290.001080.0772
 13.50.005290.00360.0267 2.30.001970.001670.0727
 2.50.002590.001830.0733Winter flounder0.50.000860.000680.137
White perch4.10.005330.003620.0884 0.60.001370.000990.165
 3.40.003430.002570.0757 1.20.002860.002410.201
 1.90.001360.000910.0478 0.60.001530.001270.211
 2.70.002860.002180.0809 0.30.000830.000530.178
thumbnail image

Figure Figure 3.. Box and whisker plot of lipid-normalized dioxin and furan residues measured in 6 species of reproductively sized fish collected from New York- New Jersey harbor estuary, expressed as toxicity equivalent values (Skinner et al. 1997). Lower and upper lines of box denote the 25th and 75th percentiles, respectively. Filled circles indicate toxic equivalent values measured in individual fish. The tissue residue benchmarks associated with protection of 90, 95, and 99% of fish species are shown as horizontal lines.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References

The tissue-residue benchmark for TCDD TEQ for fish (based on the geometric means of NOERs and LOER values) could range from 0.057 to 0.699 ng TCDD TEQ/g lipid, depending on the level of protection needed (99 and 90% of fish species protected, respectively). However, the ultimate decision regarding the appropriate protection level of a TRB is left to risk managers and policy makers. Risk management decisions (i.e., appropriate levels of protection) are site-specific and entail balancing of ecological goals and economic, sociopolitical, and technical considerations. The proposed framework provides risk managers with the distinct advantage of allowing data from multiple studies using different species to be considered in selecting a TRB, rather than relying on data from a single study or species.


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References

Acknowledgement—The authors would like to thank Tala Henry and Phil Cook for their technical advice in the development of technical concepts and review of an early draft of the manuscript. Funding for this work was provided, in part, by USEPA, Region 2, the U.S. Army Corps of Engineers, Dredging Operations Technical Support (program manager, Doug Clarke), and the U.S. Army Corps of Engineers, New York district.

Disclaimer—The work presented in this article has been internally reviewed by USEPA; its publication, however, does not signify that the contents reflect the views of the Agency. Permission has been granted by the Chief of Engineers to publish this material.


  1. Top of page
  2. Abstract
  8. Acknowledgements
  9. References
  • Bailer AJ, Oris JT. 1997. Estimating inhibition concentrations for different response scales using generalized linear models. Environ Toxicol Chem 16: 15541559.
  • Barron MG, Hansen JA, Lipton J. 2002. Association between contaminant tissue residues and effects in aquatic organisms. Rev Environ Contam Toxicol 173: 137.
  • Bellastrazzi R, Simona R, Tulli F, Bracelli A. 2003. The effect of dietary coconut oil on reproductive traits and egg fatty acid composition in rainbow trout (Onchorhynchis mykiss). Aquac Int 11: 289299.
  • Boese BL, Lee H II, Echols S. 1997. Evaluation of a first-order model for the prediction of the bioaccumulation of PCBs and DDT from sediment into the marine deposit-feeding clam Macoma nasuta. Environ Toxicol Chem 16: 15451553.
  • Chapman PM, Chapman PF, Caldwell RS. 1996. A warning: NOECs are inappropriate for regulatory use. Environ Toxicol Chem 15: 7779.
  • Crane M, Newman MC. 2000. What level of effect is a no observed effect. Environ Toxicol Chem 19: 516519.
  • de Zwart D, Sterkenburg A. 2002. Toxicity-based assessment of water quality. In: PosthumaL, SuterIIIGW, TraasTP, editors. Species Sensitivity Distributions in Ecotoxicology. New York (NY), USA: Lewis.
  • Eisler RE. 2000. Handbook of Chemical Risk Assessment. Health Hazards to Humans Plants and Animals. Vol 2. Organics. New York (NY), USA: Lewis.
  • Elonen GE, Spehar RL, Holcombe GW, Johnson RD, Fernandez JD, Erickson RJ, Tietge JE, Cook PM. 1998. Comparative toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin to seven freshwater fish species during early life stage development. Environ Toxicol Chem 17: 472483.
  • Friant SL, Henry L. 1985. Relationship between toxicity of certain organic compounds and their concentrations in tissues of aquatic organisms: A perspective. Chemosphere 14: 18971907.
  • Gobas FAPC. 1993. A model for predicting the bioaccumulation of hydrophobic organic chemicals in aquatic food-webs: Application to Lake Ontario. Ecol Model 993: 117.
  • Guiney PD, Cook PM, Casselman JM, Fizsimmons JD, Simonin HA, Zabel EW, Peterson RE. 1996. Assessment of 2,3,7,8-tetrachlorodibenzo-p-dioxin induced sac fry mortality in lake trout (Salvelinus namaycush) from different regions of the Great Lakes. Can J Fish Aquat Sci 53: 20802092.
  • Henry TR, Spitsbergen JM, Hornung MW, Abnet CC, Peterson RE. 1997. Early life stage toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in zebrafish (Danio rerio). Toxicol Appl Pharmacol 142: 5668.
  • Jarvinen AW, Ankley GT. 1999. Linkage of effects to tissue residues: Development of a comprehensive database for aquatic organisms exposed to inorganic and organic chemicals. Pensacola (FL), USA: SETAC technical publication series. 357 p.
  • Johnson RD, Tiege JE, Jensen KM, Fernandez JD, Linnum AL, Lothenbach DB, Holcombe GW, Cook PM, Christ SA, Lattier DL, Gordon DA. 1998. Toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin to early life stage brook trout (Salvelinus fontinalis) following parental dietary exposure. Environ Toxicol Chem 17: 24082421.
  • Kane-Driscoll S, Landrum PF. 1997. A comparison of equilibrium partitioning and critical body residue approaches for predicting toxicity of sediment-associated fluoranthene to freshwater amphipods. Environ Toxicol Chem 16: 21792186.
  • Landis WG, Yu M-H. 1995. An introduction to environmental toxicology: Impacts of chemicals on ecological systems. Boca Raton (FL), USA: Lewis.
  • Landrum PF, Lee H II, Lydy MJ. 1992. Toxicokinetics in aquatic systems: Model comparison and use in hazard assessment. Environ Toxicol Chem 11: 17091725.
  • Landrum PF, Lotufo GR, Gossiaux DC, Gedeon ML, Lee JH. 2003. Bioaccumulation and critical body residue of PAHs in the amphipod, Diporeia spp: Additional evidence to support toxicity additivity for PAH mixtures. Chemosphere 51: 481489.
  • Landrum PF, Steevens JA, Gossiaux DC, McElroy M, Robinson S, Begnoche L, Chernyak S, Hickey J. 2004. Time-dependent lethal body residues for the toxicity of pentachlorobenzene to Hyalella azteca. Environ Toxicol Chem 23: 13351343.
  • Lee JH, Landrum PF, Koh CH. 2002. Toxicokinetics and time-dependent PAH toxicity in the amphipod Hyalella azteca. Environ Sci Technol 36: 31243130.
  • Long ER, MacDonald DD, Smith SL, Calder FD. 1995. Incidence of adverse biological effects within ranges of chemical concentrations in marine and estuarine sediments. Environ Manag 19: 8197.
  • Mackay D, Puig H, McCarty LS. 1992. An equation describing the time course and variability in uptake and toxicity of narcotic chemicals in fish. Environ Toxicol Chem 11: 941951.
  • McCarty LS, Mackay D, Smith AD, Ozburn GW, Dixon DG. 1991. Interpreting aquatic toxicity QSARs: The significance of toxicant body residues at the pharmacologic endpoint. Sci Total Environ 109/110: 515525.
  • McCarty LS, Mackay D. 1993. Enhancing ecotoxicological modeling and assessment: Body residues and modes of toxic action. Environ Sci Technol 27: 17191728.
  • Meador JP, Collier TK, Stein JE. 2002. Use of tissue and sediment-based threshold concentrations of polychlorinated biphenyls (PCBs) to protect juvenile salmonids listed under the US Endangered Species Act. Aquatic Conservation: Marine and Freshwater Ecosystems 12: 493516.
  • [NRC] National Research Council. 2001. A risk management strategy for PCB-contaminated sediments, board on environmental studies and toxicology, division on life and earth studies. Washington DC: National Academy.
  • PosthumaL, SuterGW, TraasTP, editors. 2002. Species sensitivity distributions in ecotoxicology. New York (NY), USA: Lewis.
  • Rainbow PS. 2002. Trace metal concentrations in aquatic invertebrates: Why and so what? Environ Pollut 120: 497507.
  • Rice CD, Roszell LE. 1998. Tributyltin modulates 3,3′,4,4′,5-pentachlorobiphenyl (PCB-126)-induced hepatic CYP1A activity in channel catfish, Ictalurus punctatus. J Toxicol Environ Health 55: 197212.
  • Russell RW, Gobas FAPC, Haffner GD. 1999. Maternal transfer and in ovo exposure of organochlorines in oviparous organisms: A model and field verification. Environ Sci Technol 33: 416420.
  • Sample BE, Opresko DM, Suter GW II 1996. Toxicological benchmarks for wildlife. ES/ER/TM-86/R3. Washington DC: U.S. Department of Energy.
  • Skinner LC, Prince R, Waldman J, Newell AJ, Shastay J Jr. 1997. Chemical residues in fish, bivalves, crustaceans, and a cephalopod from the New York-New Jersey harbor estuary: Dioxins and furans. Final report. New York State Department of Environmental Conservation.
  • [USACE] U.S. Army Corps of Engineers. 2002. ERED (Environmental Residue Effects Database). Accessed May 2004.
  • van den Berg M, Birnbaum L, Bosveld AT, Brunström B, Cook PM, Feeley M, Giesy J, Hanberg A, Hasegawa R, Kennedy SW, Kubiak T, Larsen JC, van Leeuwen FX, Liem AK, Nolt C, Peterson RE, Poeliinger L, Safe S, Schrenk D, Tillitt D, Tysiind M, Younes M, Waern F, Zacharewski T. 1998. Toxic equivalency factors (TEFs) for PCBs, PCDDs, and PCDFs for humans and wildlife. Environ Health Perspect 106: 775792.
  • van Hoogen G, Opperhuizen A. 1988. Toxicokinetics of chlorobenzenes in fish. Environ Toxicol Chem 7: 213219.
  • van Wezel AP, de Vries DAM, Kostense S, Sijm DTHM. 1995. Intraspecies variation in lethal body burdens of narcotic compounds. Aquat Toxicol 33: 325342.
  • von Stackelberg K, Burmistrov D, Linkov I, Cura J, Bridges TS. 2002. The use of spatial modeling in an aquatic food web to estimate exposure and risk. Sci Total Environ 288: 97110.
  • Walker KW, Cook PM, Butterworth BC, Zabel EW, Peterson RE. 1996. Potency of a complex mixture of polychlorinated dibenzo-p-dioxin, dibenzofuran, and biphenyl congeners compared to 2,3,7,8-tetrachlorodibenzo-p-dioxin in causing fish early life stage mortality. Fundamental Applied Toxicology 30: 178186.
  • Walker MK, Cook PM, Batterman AR, Butterworth BC, Berini C, Libal JJ, Hufnagle, Peterson RE. 1994. Translocation of 2,3,7,8-tetrachlorodibenzo-p-dioxin from adult female lake trout (Salvelinus namaycush) to oocytes: Effects on early life stage development and sac fry survival. Can J Fish Aquat Sci 51: 14101419.
  • Walker MK, Hufnagle LC, Clayton MK, Peterson RE. 1992. An egg injection method for assessing early life stage mortality of polychlorinated dibenzo-p-dioxins, dibenzofurans, and biphenyls in rainbow trout (Oncorhynchus mykiss). Aquat Toxicol (Amst) 22: 1538.
  • Walker MK, Peterson RE. 1991. Potencies of polychlorinated dibenzo-p-dioxin, dibenzofuran, and biphenyl congeners, relative to 2,3,7,8-tetrachlorodibenzo-p-dioxin, for producing early life stage mortality in rainbow trout (Oncorhynchus mykiss). Aquat Toxicol (Amst) 21: 219238.
  • Walker MK, Peterson RE. 1994. Toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin to brook trout (Salvelinus fontinalis) during early development. Environ Toxicol Chem 13 (5): 817820.
  • Walker MK, Spitsbergen JM, Olson JR, Peterson RE. 1991. 2,3,7,8-tetrachlorodibenzo-p-dioxin toxicity during early life stage development of lake trout (Salvelinus namaycush). Can J Fish Aquat Sci 48: 875883.
  • Wong CS, Mabury SA, Whittle DM, Backus SM, Teixeira C, DeVault DS, Bronte CR, Muir DCG. 2004. Organochlorine compounds in Lake Superior: Chiral polychlorinated biphenyls and biotransformation in the aquatic food web. Environ Sci Technol 38: 8492.
  • Zabel EW, Cook PM, Peterson RE. 1995. Toxic equivalency factors of polychlorinated dibenzo-p-dioxin, dibenzofuran and biphenyl congeners based on early life stage mortality in rainbow trout (Oncorhynchus mykiss). Aquat Toxicol (Amst) 31: 315328.