ChemRisk, 8024 Mesa Dr., #126, Austin, Texas 78731
    Exponent, Inc., 420 Lexington Ave., Suite 1740, New York, New York 10170
    ChemRisk, 10375 Richwood Ave., Suite 350, Houston, TX 77042
    U.S. Environmental Protection Agency, ORD NHEERL ETD, MD-B143-01, Research Triangle Park, North Carolina 27709
    National Institute of Environmental Health Sciences, Environmental Toxicology Program, P.O. Box 12233, Mail Drop EC-34, Research Triangle Park, North Carolina 27709
    U.S. Environmental Protection Agency, Office of Research and Development, 1200 Pennsylvania Avenue, MC 8101 R, Washington, District of Columbia 20460
    ChemRisk, 25 Jessie St., Suite 1900, San Francisco, CA 94105

    ABSTRACT

    The toxic equivalency factor (TEF) approach has been widely accepted as the most feasible method available at present for evaluating potential health risks associated with exposure to mixtures of dioxin-like compounds (DLCs). The current mammalian TEFs for the DLCs were established by the World Health Organization (WHO) following the meeting of an international expert panel in June of 1997. The TEFs recommended by WHO were determined based on a consensus of scientific judgment and were presented as point estimates. However, the relative potency estimates (REPs) underlying the TEFs were derived from a heterogeneous data set and often span several orders of magnitude. In this article, we present a refined database of mammalian REPs that we believe will facilitate better characterization of the variability and uncertainty inherent in the data. The initial step involved reviewing the REP database used by the WHO panel during its review in 1997. A set of criteria was developed to identify REPs that were determined to be the most representative measure of a biological response and of adequate quality for use in quantitative analyses. REPs were determined to be inappropriate for use in quantitative analyses if any of the established exclusion criteria were met. Comparison of data records to the established exclusion criteria resulted in the identification of a substantial number of REPs believed to be inappropriate for use in quantitative analyses. Next, studies published after 1997 were added to the database. The availability of such a refined database will improve risk assessment for this class of compounds by including additional information from new studies and facilitating the use of quantitative approaches in the further development of TEFs.

    Key Words: dioxin; polychlorinated dibenzo-p-dioxins; polychlorinated dibenzofurans; polychlorinated biphenyls; toxic equivalency factor (TEF); relative potency (REP).

    INTRODUCTION

    Polychlorinated dibenzo-p-dioxins (PCDDs), polychlorinated dibenzofurans (PCDFs), and polychlorinated biphenyls (PCBs) are persistent, bioaccumulative toxicants that are ubiquitous in the environment and have been detected in the tissues of birds, fish, and mammals, including human adipose tissue and milk (Safe, 1990). PCDDs and PCDFs have no known industrial use but occur as unwanted by-products of a number of industrial operations or combustion processes, such as chlorine bleaching of paper and pulp, production of chlorinated phenols, burning of wastes and fuels, and metal smelting (USEPA, 2003a). In contrast, PCBs have been widely used for a variety of industrial purposes, including heat transfer agents, dielectric fluids for capacitors and transformers, plasticizers, and paint additives (Safe, 1990). A subset of the PCDDs, PCDFs, and PCBs induce a similar spectrum of biochemical and toxic responses in experimental animals that is characterized by severe weight loss, thymic atrophy, hepatotoxicity, edema, fetotoxicity, teratogenicity, reproductive toxicity, immunotoxicity, and enzyme induction (Birnbaum, 1994; Birnbaum and Tuomisto, 2000; DeVito and Birnbaum, 1994; McConnell et al., 1978; Safe, 1990; Schwetz et al., 1973). These common biological effects are mediated through a common mechanism of action initiated by binding to and activating the Aryl hydrocarbon (Ah) receptor (Birnbaum, 1994; Hankinson, 1995; Martinez et al., 2003; Okey et al., 1994; Safe, 1990; Sewall and Lucier, 1995). This subset of PCDDs, PCDFs, and PCBs is commonly referred to as the "dioxin-like" compounds (DLCs) and comprises 17 laterally substituted (2,3,7,8-substituted) PCDD and PCDF congeners, and 12 non-ortho and mono-ortho chlorine-substituted PCBs.

    Assessment of the potential risk associated with exposure to the DLCs is complicated by the fact that these compounds are typically detected in the environment as part of complex mixtures of structurally related polyhalogenated aromatic hydrocarbons (Birnbaum, 1999; Safe, 1994). Because the toxicity information is incomplete for this class of chemicals, a toxic equivalency factor (TEF) methodology has been developed to assess their potential health effects. The TEF method is a relative potency scheme, with the most toxic and well-studied member of the class, 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), serving as the prototype. This method has been re-evaluated over the past 20 years in a variety of forums. One of the most recent was the review by a World Health Organization (WHO) expert panel in 1997, the outcome of which was the establishment of consensus-based TEFs for mammals, birds, and fish for the 29 DLCs (van den Berg et al., 1998). The 1997 re-evaluation by WHO was based on a review of relative potency (REP) data that had been compiled in a database by scientists at the Karolinska Institute (hereafter referred to as the REP1997 Database). In this report, we briefly review the evolution of the TEF methodology and development of the REP1997 Database. More importantly, we present a new database that is built upon the REP1997 Database. This refined database (hereafter referred to as the REP2004 Database) can provide the structure to assess the variability in the underlying data, as well as the uncertainty inherent in the TEF values assigned to individual congeners.

    Evolution of the Toxic Equivalency Factor (TEF) Approach

    The TEF methodology was first proposed by the Ontario Ministry of the Environment (OME) in 1984. In their Scientific Criteria Document, the OME concluded that environmental standards designed to protect human health from exposure to DLCs should be based on a "toxic equivalency" approach, with the most toxic and well-studied member of the class, TCDD, serving as the prototype (OME, 1984). This recommendation was based on scientific evidence indicating that the DLCs were structurally related and induced a similar spectrum of biochemical and toxic responses, which were mediated by a common mechanism of action involving binding to the Ah-receptor. The OME recommended assigning toxicity factors relative to TCDD for all of the PCDD/PCDF congeners based on their homologue group. Congeners were assigned to homologue groups based on the number of chlorine substitutions. Shortly thereafter, a TEF-like method was used to estimate potential health risks associated with exposures resulting from a PCB transformer fire in Binghamton, New York (Eadon et al., 1986). This represented the first use of a TEF approach to assess this class of compounds.

    In 1987, the U.S. Environmental Protection Agency (USEPA) formally adopted an interim TEF approach. In so doing, USEPA refined the OME approach by recommending TEFs for each specific laterally substituted congener, rather than just for each homologue group (USEPA, 1987). At that same time, an international effort was underway to harmonize the different TEF schemes that had been proposed by regulatory bodies in several countries (NATO/CCMS, 1988a,b,c). The result of this international effort was the adoption of a common set of TEFs approved by the NATO/CCMS Dioxin Information Exchange Subcommittee in April 1988. Representatives from member countries were then asked to seek formal adoption of the recommended TEFs, termed the International TEFs/89, or I-TEFs/89. In response to this request, and in keeping with their commitment to periodically review and update their interim TEF approach, the USEPA formally adopted the I-TEFs/89 as their preferred interim approach (USEPA, 1989). The key changes that were made by the USEPA in adopting the I-TEFs/89 included establishing TEFs for octachlorodibenzo-p-dioxin (OCDD) and octachlorodibenzofuran (OCDF), modifying TEF values for specific congeners based on in vivo data, and eliminating TEF values for all non-laterally substituted congeners.

    The initial TEF methodologies focused solely on PCDDs and PCDFs. However, increasing evidence indicated that the coplanar and mono-ortho coplanar PCBs were Ah-receptor agonists and induced a variety of in vitro and in vivo effects that are similar to those of TCDD. Safe first proposed TEFs for the PCBs in 1990 (Safe, 1990), and shortly thereafter, USEPA convened a workshop to discuss the establishment of TEFs for PCBs (Barnes et al., 1991). The issue of establishing TEFs for the PCBs was then evaluated on an international level as part of a joint project sponsored by the WHO European Centre for Environmental Health (WHO-ECEH) and International Programme on Chemical Safety (IPCS). In preparation for the WHO-ECEH/IPCS expert consultation held in Bilthoven, Netherlands, in December 1993, scientists with the Institute of Environmental Medicine (IEM) at the Karolinska Institute (Drs. U. G. Ahlborg, A. Hanberg, and F. Waern) compiled all relevant mammalian experimental data for the dioxin-like PCBs into a database. A specific PCB congener was determined to be suitable for inclusion in the database if the following criteria were met: (1) the congener is structurally similar to the PCDDs and PCDFs, (2) the congener binds to the Ah-receptor, (3) the congener elicits dioxin-like biochemical and toxic responses, and (4) the congener is persistent and accumulates in the food chain (Ahlborg et al., 1994). The specific studies included in this initial database were those published through 1993 that met the following established inclusion criteria: (1) at least one PCB congener was tested in the study, (2) an appropriate reference compound (TCDD, PCB77, PCB126, or PCB169) was included in the study or in a different study with the same experimental design and by the same authors, and (3) endpoints were affected by both the test congener and the reference compound (Ahlborg et al., 1994).

    Establishment of the Current Mammalian TEFs for the PCDDs, PCDFs, and Dioxin-like PCBs

    One of the outcomes of the 1993 WHO-ECEH/IPCS expert consultation on PCBs was the recommendation to expand the existing PCB database to include information on the PCDDs and PCDFs. The update and expansion of the original PCB database was again carried out by scientists with the Institute of Environmental Medicine (IEM) at the Karolinska Institute in Stockholm, Sweden (Drs. A. Hanberg, F. Waern, and P. Andersson). This expanded database, the REP1997 Database, was to serve as the basis for a subsequent WHO-ECEH/IPCS re-evaluation of the TEFs. This second WHO-ECEH/IPCS re-evaluation was conducted at the Karolinska Institute in Stockholm, Sweden, in June of 1997. During their review, the WHO-ECEH/IPCS expert panel examined data from an extensive body of in vivo and in vitro studies that had been compiled in the REP1997 Database. Consensus-based TEF values were established by the WHO expert panel based on scientific judgment, after consideration of the mammalian data in the REP1997 Database, along with previously published TEFs (van den Berg et al., 1998). Because mammalian TEFs had been established previously for PCDD/Fs (NATO/CCMS, 1988b; USEPA, 1989), as well as for the dioxin-like PCBs (Ahlborg et al., 1994), the WHO expert panel decided that the existing TEFs would remain unchanged unless there was sufficient information to warrant modification (van den Berg et al., 1998). The WHO expert panel ultimately recommended TEFs for 29 congeners, including the 17 laterally substituted PCDDs/PCDFs, 4 non-ortho PCBs, and 8 mono-ortho PCBs. The final TEFs recommended by the WHO expert panel (hereafter referred to as the WHO98 TEFs) are described as representing order-of-magnitude estimates of potency for each congener relative to the most potent member of this class of compounds, TCDD. As such, the WHO98 TEFs are assigned values, not calculated values.

    The additive TEF model has been widely accepted as the most feasible method currently available for evaluating potential health risks associated with exposures to this class of compounds (Birnbaum, 1999; Birnbaum and DeVito, 1995; NATO/CCMS, 1988b; Olson et al., 1989; van den Berg et al., 1998; Yrjanheikki, 1992). While there are clearly uncertainties associated with this approach (e.g., non-additive interactions, natural ligands for the Ah receptor, differences in species responsiveness, differences in the shape of dose-response curves), studies involving mixtures of DLCs have indicated that their toxicity can be predicted fairly well and support the use of the TEF methodology (DeCarprio et al., 1986; Silkworth et al., 1989; Suter-Hoffman and Schlatter, 1989). More recently, as part of the National Toxicology Program's (NTP's) comprehensive evaluation of the ability of the TEF approach to predict cancer risk for dioxin-like compounds, Walker and coworkers (2005) showed that the WHO98 TEFs adequately predicted the increased incidence of liver tumors induced by exposure to a mixture of TCDD, 2,3,4,7,8-pentachlorodibenzofuran (PeCDF), and PCB126. While there are innate limitations and untested assumptions in the TEF methodology, it is clearly more appropriate than the other potential alternatives such as basing the risk on TCDD alone or assuming that all chemicals are equipotent to TCDD.

    Nonetheless, because of the uncertainties and limitations that are inherent in the TEF methodology, WHO and others have clearly indicated that the approach should be thought of as an interim methodology that should be subject to periodic review as new scientific information becomes available (Birnbaum, 1999; Birnbaum and DeVito, 1995; USEPA, 1987, 1989; van den Berg et al., 1998). The need to explore alternative approaches for assessing the potential health risk associated with this class of compounds has been acknowledged by USEPA and others (USEPA, 2003b). As scientists gain a better understanding of the modes of action underlying this class of compounds, as more data become available concerning the relative potencies of these compounds, and as more sophisticated quantitative tools are developed, it may be possible to further improve the TEF methodology. In its 20-year history, this approach has evolved and each iteration has resulted in a process that is more transparent and reflects the latest science.

    The Next Iteration of Mammalian TEFs: Improving the Process

    The WHO98 TEFs are currently being used by numerous governmental agencies and other entities to regulate sources, as well as to assess and control potential risks associated with exposure to PCDD/Fs and dioxin-like PCBs in foods, consumer products, and environmental media. As has been noted by a number of investigators, for any given congener, the underlying REP values typically represent a heterogeneous data set, and the range of REPs often spans several orders of magnitude (Birnbaum et al., 2004; Finley et al., 2003; Haws et al., 2004; USEPA, 2003b; van den Berg et al., 1998). Despite this wide range of REP values for a given congener, the TEF values are presented as single point estimates. The publication by van den Berg and coworkers (1998) clearly described the qualitative criteria considered in assigning the current mammalian TEFs (i.e., in vivo studies were given greater weight than in vitro studies and/or quantitative structure activity relationship [QSAR] data; chronic studies were given greater weight than subchronic studies, which were given greater weight than subacute studies, which were given more weight than acute studies; and Ah-receptor-mediated toxic responses were given more weight than were Ah-receptor-mediated biochemical responses [e.g., enzyme induction]). However, it is difficult to determine how the expert panel actually applied these criteria during the course of their re-evaluation in 1997. It is also not possible to determine which studies received the most weight or how each specific study influenced the derivation of a TEF value for a specific congener. As a result, it is not currently possible to quantitatively characterize the uncertainty that is inherent in the WHO98 TEF values.

    As the TEF methodology has evolved over the last 20 years, so too have our approaches for assessing human health risk. In recent years, there has been a significant effort aimed at increasing transparency of the risk assessment process, as well as increasing the use of more quantitative tools such as probabilistic risk assessment methods. Application of probabilistic risk assessment techniques to the TEFs would require knowledge of the distribution of REP values underlying the current TEFs for each congener. Thus, the REP databases must be designed to allow for the evaluation of the range and distribution of REP values. Further, a standardized method of choosing a TEF from a distribution of REP values for each congener would provide greater transparency and potentially a uniform degree of conservatism in the TEF values across chemicals. This is important given the widespread use of the TEFs to assess potential health risks associated with exposures to this class of compounds or otherwise support regulation. In addition, there are examples in the literature of cases where minor changes in the TEF values have been shown to have considerable impact on dioxin risk estimates (Dyke and Stratford, 2002). In their analyses, Dyke and Stratford (2002) found that the changes in the TEFs recommended by WHO in 1998 resulted in a 3.6-fold increase in the toxic equivalent (TEQ) for PCDD/Fs in sludge samples and a 25% increase in milk.

    Given the importance of TEFs, the goal of this current project was to re-examine and update the REP1997 Database to better characterize the range of observed REPs, and in so doing, create a structure that would facilitate quantitative analyses. This, in turn, will allow for better characterization of the uncertainty inherent in the mammalian TEFs. Development of the REP2004 Database was necessary because the REP1997 Database was not intentionally designed or annotated in a way that would allow for quantitative characterization of the distribution of the REPs within the database. The analysis reported herein describes our efforts with regard to the development of the REP2004 Database. It is important to note that the REP2004 Database builds upon the REP1997 Database and, as such, should be thought of as an extension of that database, not as an alternative database. We also provide recommendations regarding possible next steps, as well as implications for risk assessment.

    Review of the Mammalian Data in the REP1997 Database

    Background.

    The database used by the WHO expert panel to establish the current mammalian TEFs during their review in 1997, the REP1997 Database, was kindly provided to us by Dr. Frederik Waern, Karolinska Institute, Stockholm, Sweden. The database was in the form of a Microsoft Excel spreadsheet, which was created by scientists with the IEM at the Karolinska Institute. Because this database was intended to be an all-inclusive collation of pertinent information, REP values and critical study elements from not only published manuscripts, but also manuscripts in press, conference proceedings, theses, dissertations, and unpublished studies through June of 1997, were entered into the database. Studies were determined to be suitable for inclusion in the REP1997 Database when the following criteria were met: (1) at least one test congener (PCDD, PCDF, or PCB) and a reference compound (2,3,7,8-TCDD or PCB126) were included in the study, or the reference compound data (TCDD or PCB126) was from an identical experiment by the same authors; and (2) the endpoint used as the basis for the REP was Ah-receptor-mediated and was affected by both the test congener and reference compound. An effort was also made to include information about the design of each study (e.g., cell line, species, strain, duration, route, dose regimen, and chemical purity), along with the REP values. In addition, because the WHO-ECEH/IPCS re-evaluation of the TEF methodology was intended to address TEFs for wildlife as well as mammals, experimental data and associated REP values for fish and avian species were also incorporated into the REP1997 Database. In addition to incorporating REP values estimated by the authors, scientists at the IEM used several different approaches to calculate REP values from individual studies. The types of methods employed to derive the REP estimates for each study in the database included the following: (1) comparing dose-response curves or using linear interpolation of log-doses, comparing the same effect level; (2) determining ratios of median effective doses or concentrations (ED50s or EC50s), median lethal doses (LD50s), tumor promotion indices, and dissociation constants (Kds) for Ah receptor binding; or (3) estimating REP values directly from the figures presented in the publication (van den Berg et al., 1998). As indicated in the background information included in the REP1997 Database, several problems were noted with regard to the calculation of REP values: (1) different compounds resulted in different maximal effects, and as a result, calculation of REP values was based on the comparison of different effect levels; (2) in some cases, the study included only a single dose level of the test/reference compounds, and in such cases, if the response data was not within the dose-response range of the reference compound, the REP value was qualified with "<" or ">" symbols in the database; (3) in some papers, the data were presented in graphical form only; and (4) only a few of the papers cited in the database tested high dose levels, so it was questionable whether maximal effects/responses were achieved.

    While the REP1997 Database contains REP data for fish, birds, and mammals, as well as for additional PCB congeners, our review of this database was limited to the mammalian data for 28 of the 29 congeners for which WHO established TEFs in 1997. The remaining TEF established by WHO was for TCDD, the prototype to which all other congeners were compared in developing the TEFs. The REP1997 Database did not contain entries for TCDD alone, but rather, that information was included as reference compound data from the study for the test congener(s). All of the discussions, summary statistics, etc. from this point forward are based on the mammalian data for those 28 congeners.

    Findings.

    The mammalian records included in the REP1997 Database for the 28 congeners were based on data presented in 88 individual publications, a substantial number of which contained data for multiple congeners, multiple endpoints, and multiple timepoints (Table 1). In vivo data were obtained from 61 different publications, while 30 publications served as the source for the in vitro data included in the REP1997 Database (Table 1). It should be noted that some studies included in the database contained data for both in vivo and in vitro studies.

    There was wide variation in the total number of studies available for each of the congeners, with 1,2,3,7,8,9-hexachlorodibenzofuran (HxCDF) having data from just a single in vitro study, while data for PCB126 were obtained from 39 different studies (Table 1). Even for those congeners recently found by the Centers for Disease Control and Prevention to account for most of the background toxic equivalency quotient (TEQ) in blood in the U.S. population (i.e., TCDD; 1,2,3,7,8-pentachlorodibenzo-p-dioxin [PeCDD], 1,2,3,6,7,8-hexachlorodibenzo-p-dioxin [HxCDD], 2,3,4,7,8-pentachlorodibenzofuran [PeCDF], PCB118, PCB126, and PCB156 [Needham et. al., 2005]), there was great variability in the number of studies that provided REP data for the WHO 1998 expert panel to consider when establishing the current mammalian TEFs. As indicated in Table 1, the number of studies included in the REP1997 Database for each of those congeners is as follows: (1) 1,2,3,7,8-PeCDD (14 studies, including 9 in vivo); (2) 1,2,3,6,7,8-HxCDD (4 studies, all of which were in vitro); (3) 2,3,4,7,8-PeCDF (21 studies, including 15 in vivo); (4) PCB118 (18 studies, including 11 in vivo); (5) PCB126 (39 studies, including 25 in vivo); and (6) PCB156 (28 studies, including 19 in vivo).

    For the 28 congeners of interest, there were a total of 976 mammalian records in the REP1997 Database (731 in vivo records and 245 in vitro records). It is important to note that not all of the 976 records contained REP values. Of the 818 records that did include REP values, 87 were non-numeric or qualified values (e.g., expressed as a range or qualified with "<" or ">" or ""). There were 731 records that contained numeric REP values (495 in vivo and 236 in vitro REPs) (Table 2). As was observed for the number of studies in the REP1997 Database, there was a wide range in the total number of numeric REPs available for each of the 28 congeners, with 1,2,3,7,8,9-HxCDF having only a single REP value from an in vitro study and PCB126 having 107 numeric REP values, 82 of which were from in vivo studies (Table 2). Even for those congeners that account for the majority of the background TEQ in blood in the U.S. population, there was wide variation in the total number of numeric REP values, (1,2,3,6,7,8-HxCDD [9 numeric REPs, all in vitro]; PCB118 [38 numeric REPs, including 27 in vivo]; 1,2,3,7,8-PeCDD [48 numeric REPs, including 40 in vivo]; PCB156 [66 numeric REPs, including 53 in vivo]; 2,3,4,7,8-PeCDF [73 numeric REPs, including 59 in vivo]; and PCB126 [107 numeric REPs, including 82 in vivo]. Summary statistics for the REP values included in the REP1997 Database are provided in Table 3. There was much greater variability in the range of REP values for the PCBs than for the PCDDs and PCDFs. The in vivo + in vitro REPs for PCBs were typically found to span three to six orders of magnitude, while those REPs for the PCDD/Fs were spread across one to three orders of magnitude. In addition, for 1,2,3,7,8,9-HxCDD and -HxCDF, the current WHO98 TEF slightly exceeds the maximum REP value in the REP1997 Database (Table 3). For congeners that currently account for the majority of the background TEQ in blood in the U.S. population, the REP values span two to five orders of magnitude, with the PCBs again exhibiting greater variability in the range of REP estimates (i.e., in vivo and in vitro REPs for PCB118, 126, and 156 span three to five orders of magnitude, while those for the three PCDD/Fs [i.e., 1-PeCDD; 1,2,3,6,7,8-HxCDD; and 4-PeCDF] span approximately two orders of magnitude]) (Fig. 1; Table 3). These findings indicate substantial variability in the REP data underlying the current mammalian WHO98 TEFs for the dioxin-like PCBs, PCDDs, and PCDFs, even for those congeners of greatest public health concern.

    Development of the REP2004 Database

    Data retrieval and database compilation.

    The first step of our analysis involved obtaining copies of all original in vivo and in vitro mammalian studies cited in the REP1997 Database. For purposes of our review, it was assumed that the developers of the REP1997 Database had identified all of the pertinent studies for inclusion, and therefore, a literature search was not conducted to identify studies published prior to 1997. As will be discussed in the next section, however, our development of the REP2004 Database did include adding relevant studies published subsequent to the WHO-ECEH/IPCS evaluation in 1997. Before beginning our review of each original paper cited in the REP1997 Database, those study elements that were believed to be important metrics of study quality and relevance to humans were identified. The specific study elements determined to be critical were: cell line, route of administration, chemical purity, exposure duration, delay between treatment and measurement of response, measurement endpoint, strain, tissue type, number of dose levels tested, attainment of a maximal response, method of REP derivation, vehicle, animal age and sex, number of animals per treatment group, controls, and the reference compound. With the exception of three of the study elements (delay between treatment and measurement of effect, attainment of a maximal response, and controls), all critical study elements were explicitly identified in the original REP1997 Database. Delay between treatment and measurement of effect, while not explicitly identified in the REP1997 Database, could be deduced from other information provided in the database. Additionally, while the database does not include information regarding attainment of a maximal response, this factor was taken into account by the developers of the REP1997 Database in their derivation of the REPs included in the REP1997 Database (Dr. A. Hanberg, personal communication).

    Modifications to records from the original REP1997 Database.

    The relevant data concerning each of the critical study elements in each of the original in vivo and in vitro studies, as well as the REP values when presented in the publication, were compared to the data in the REP1997 Database. Those entries in the REP1997 Database that were determined to be in error or incomplete were then revised or updated. It should be noted that the REP1997 Database contained information for a number of additional study elements (e.g., class of measure, dose, result, EC50, ED50, lowest-observed-effect level [LOEL], no-observed-effect level [NOEL]); however, those study elements were not viewed as critical for assessing study quality and relevance and, therefore, were not reviewed.

    The next step of our analysis involved developing a set of criteria that could be used to identify those REP values that could not be used in quantitative analyses. Table 4 summarizes the different bases for excluding REP values in the development of the REP2004 Database. There were four primary bases for exclusion: (1) repetitive studies—REPs from the same original study presented in multiple publications; (2) repetitive endpoints—multiple REPs from a single study that used different assays to measure the same biological response (e.g., Aryl Hydrocarbon Hydroxylase [AHH] and ethoxyresorufin-o-deethylase [EROD]); (3) studies that used only a single dose level of the test and/or reference compound—incomplete dose-response data; and (4) other miscellaneous issues—data being omitted in the final peer-reviewed publication, lack of valid reference compound data, lack of a response, values derived from a mixtures study, values derived from unpublished studies that could not be obtained or verified. In cases where there were multiple REPs that represented measures of the same biological response in the same study (i.e., "repetitive endpoints"), an effort was made to identify the single most representative REP. That REP value was then retained for quantitative analysis, while all others were identified as candidates for exclusion. Examples of repetitive endpoints, as well as the specific endpoint that was determined to be the best measure of the biological response under evaluation, are outlined in Table 5. In cases where multiple REPs were calculated for the same data set using different calculation techniques (e.g., ED50 ratios and ED25 ratios based on the data presented in Leece et al., 1985), the REPs based on the ED50 ratio were retained, to be consistent with other REPs in the database.

    Efforts were made to obtain copies of all studies cited in the REP1997 Database, even when the study data were from sources other than peer-reviewed journals (e.g., theses, dissertations, "unpublished" sources, manuscripts in press, conference proceedings). A decision was made to retain records from these alternative sources if the study could be obtained and the data could be reviewed and verified. We were able to obtain the majority of studies that were from sources other than peer-reviewed journals. Some of these studies were published subsequently in peer-reviewed journals, and the original records in the REP1997 Database were updated based on those subsequent publications. However, in a few cases, we were unable to obtain the unpublished data, and the REPs associated with those studies have been identified as candidates for exclusion in the REP2004 Database.

    In addition, the data for a number of studies included in the REP1997 Database represented preliminary findings at the time of the WHO expert panel meeting in 1997 but have since been published. In several instances, the REP values and associated study characteristics were modified in the final publications. Therefore, where appropriate, we corrected and updated the REP values and associated study element information in the development of the REP2004 Database. For example, the records for DeVito et al. (1993) represented data for an interim timepoint (four weeks) in a subchronic study. Data for PCB77 and PCB157 were included in the REP1997 Database based on this study. However, the study authors later discovered a contamination problem with PCB77 and discovered that PCB157 had been mislabeled by the supplier and was actually a different congener (PCB156). As a result, data for these two congeners were excluded from the final publication (DeVito et al., 2000). The database was revised accordingly, with the REP values for PCB77 and PCB157 being identified as candidates for exclusion in the REP2004 Database.

    As a general rule, the REP values contained in the original REP1997 Database were not modified but rather were accepted as-is. Many of the REP values in the REP1997 Database were calculated from dose-response curves developed by IEM (indicated as "dose response graphs by Waern" in the REP1997 Database). We were unable to obtain the dose-response graphs developed by IEM, so the calculations and associated values could not be verified. Those instances where REP values were modified in the REP2004 Database are indicated clearly in the database, and a description of the REP calculation method is provided. As already indicated, a number of REP values in the REP1997 Database were qualified by a ">," "<," or "," which generally indicated a lack of complete dose-response data. In cases where it was determined that a numerical value could, in fact, be calculated, the qualified value was replaced with the numerical value that had been calculated. However, the majority of the qualified REPs have been identified in the REP2004 Database as being candidates for exclusion based on the criteria outlined in Table 4 as described above. Many of the qualified REPs were from studies that tested only one dose level of a test congener or reference compound. All REPs that were determined to be candidates for exclusion are clearly identified in the REP2004 Database, along with a discussion of the rationale for the exclusion.

    Our review of all of the mammalian data for the 28 congeners of interest in the REP1997 Database identified a substantial number of REP values that met one or more of the exclusion criteria and, thus, were identified as candidates for exclusion. In some cases, this resulted in the exclusion of all records for a specific publication. As such, the total number of studies retained from the REP1997 Database decreased from 88 to 53, and included 31 in vivo studies and 24 in vitro studies. In addition to the non-numeric entries, there were a number of other issues that were identified that further reduce the total number of REP values in the REP1997 Database.

    Because there may be differences of opinion regarding the exclusion of specific REP values, all records from the original REP1997 Database, as well as new REP values that met one or more of the exclusion criteria, were retained in the REP2004 Database, but the associated REP values were identified as candidates for exclusion in quantitative analyses of the REP data. The supporting rationale for excluding each specific REP value is provided in the REP2004 Database. It is believed that such an approach will provide greater transparency and will also provide flexibility for modifications should others arrive at different conclusions. It is important to note that there were also a substantial number of errors and incomplete entries identified for specific study elements described in the REP1997 Database as well (e.g., incorrect information concerning chemical purity, incorrect number of dose levels, incorrect cell culture system, etc.), and those entries were updated accordingly. While such information is important when conducting quantitative analyses of the REP data, it should not have affected the re-evaluation of the TEFs by WHO in 1997.

    Addition of new REP values.

    A comprehensive literature search was conducted to identify potentially relevant studies published since the time of the last WHO-ECEH/IPCS evaluation of the TEFs in 1997. The literature search extended from 1997 through the end of 2004. Two additional studies published early in 2005 (Vondracek et al., 2005; Walker et al., 2005) were brought to our attention and were included in the REP2004 Database, but the actual literature search did not extend beyond 2004. Each study identified as potentially relevant was then reviewed to determine whether it met the inclusion criteria established by WHO during their review in 1997. If so, then the study was considered for inclusion in the REP2004 Database, and REP values were calculated. It is important to note that, in several cases, new REP values were added to the REP2004 Database based on studies that were published prior to 1997 and were actually included in the original REP1997 Database. This occurred in cases where the pre-1997 paper was found to contain data for an additional congener, endpoint, and/or timepoint that was determined to be valid but had not been included the original REP1997 Database.

    In adding new records to the database, the goal was to be consistent with what had been done by IEM in the course of developing the original REP1997 database. For example, even though REPs from new studies that were based on single-dose-level studies or represented potentially repetitive endpoints would ultimately be identified as candidates for exclusion in the REP2004 Database, consistent with what had been done with records in the original REP1997 Database, those records were still incorporated into the REP2004 Database. Again, as already discussed, we believe that doing so provides greater transparency, as well as flexibility for others to potentially modify the database should they reach different conclusions.

    In terms of the approach used to calculate REP values from those new studies that were determined to contain valid data, various methods were employed, depending in part on the quality of the underlying dose-response data. The overall goal was to be consistent with the approach used by IEM to calculate REP values in the original REP1997 Database. Therefore, when REP values were provided by the study authors, those values were incorporated into the REP2004 Database. In cases where the study authors provided multiple REP values for the same response based on multiple methods of calculation, the single best value was identified and retained for quantitative analysis. For example, for the recent studies conducted by the NTP (Toyoshiba et al., 2004; Walker et al., 2005), multiple REPs were calculated for the same data set using different dose-response models (i.e., same-shape model, independent model, additive model). All of these REP values were added to the REP2004 Database, but only the REPs with the best model fit were retained for quantitative analysis (i.e., REPs based on the same-shape model), while the others were identified as candidates to be excluded.

    In cases where the study authors did not provide REPs in their publication, REPs were calculated by us. For purposes of describing the various types of methods used to calculate REPs, the dose-response data can be grouped into the following categories: (1) data based on a single dose level of both the test and reference compound, (2) data based on multiple dose levels of the test congener and a single dose level of the reference compound, (3) data based on a single dose level of the test compound and multiple dose levels of the reference compound, and (4) data based on multiple dose levels of both the test and reference compounds.

    In cases where the study authors provided benchmark values in the publication (e.g., ED50s, EC50s, LD50s, benchmark dose–lower confidence limit [BMDLs]), REPs were calculated based on the ratio of ED50s or EC50s (e.g., EC50_TCDD or PCB126/EC50_Test Congener) or similar values such as LD50s, NOELs, LOELs, or BMDLs. For example, in a study by Fattore et al. (2004), the authors provided BMDL, NOEL, and LOEL values. Ratios of these values (e.g., BMDL TCDD or PCB126/BMDLTest Congener) were then used to calculate REPs. In this example, while all of the REPs (e.g., BMDL ratio, NOEL ratio, LOEL ratio) were included in the REP2004 Database, the ratio based on BMDLs was determined to be most informative and was identified as the single most representative REP value to be used in quantitative analyses, while the others were identified as candidates for exclusion. In other cases, when ED50/EC50 values were presented along with other benchmark values, the REP based on the ED50/EC50 ratio was determined to be most informative and was retained for quantitative analyses, while all others were identified as candidates for exclusion.

    In cases where the study authors tested only a single dose level of either the test and/or reference compound, EC50 or ED50 ratios could not be calculated. In keeping with the approach employed by IEM, REPs were calculated based on the ratios of doses associated with equivalent responses. When equivalent magnitudes of responses were not reported in the single-dose-level studies, linear extrapolation techniques were employed to estimate the dose levels at equivalent response levels within the experimental dose range. Given the significant uncertainty in estimating REPs from single-dose-level studies, all such REP values, although included in the database for purposes of transparency, were ultimately identified as candidates for exclusion.

    In cases where the study authors tested multiple dose levels of both the test and reference compounds, attempts were made to estimate EC50 or ED50 values using Probit or linear/log-linear interpolation analyses. In cases where ED50 or EC50 values could be determined, REPs were calculated based on the ratio of ED50s or EC50s (e.g., EC50_TCDD or PCB126/EC50_Test Congener). In some cases, Probit analyses could not be used because of an insufficient number of dose levels (e.g., only two dose levels), a lack of a dose-related response, or a lack of goodness-of-fit in the Probit model. In such cases, REPs were calculated based on linear or log-linear interpolation analyses. For a number of studies, the dose-response data were only presented graphically in the form of figures. In such cases, the data were first evaluated by drawing a straight line on the graph to estimate numeric information. The aforementioned techniques were then employed to derive REP values. In addition, in keeping with the inclusion criteria established by WHO in 1997, in some cases, it was necessary to obtain reference compound data from another study with the same experimental design and conducted by the same authors. Finally, consistent with the approach used by IEM in developing the original REP1997 Database, a factor of 10 was applied to derive REP values in all cases when PCB126 served as the reference compound, regardless of the specific REP calculation method employed (i.e., the REP value was multiplied by 0.1, because PCB126 has been determined to be one-tenth as potent as 2,3,7,8-TCDD in mammals).

    Findings Related to the REP2004 Database

    As indicated in Tables 1 and 2, our refinement of the REP1997 Database resulted in the identification of a substantial number of mammalian studies and associated REP values that met one or more of our established exclusion criteria. However, this was offset to some extent by the addition of a number of new REP values from studies published since 1997. Even with the addition of 43 new studies (24 in vivo + 19 in vitro), the overall number of studies in the REP2004 Database decreased from 88 in the original REP1997 Database to 83 in the REP2004 Database (Table 1). The total number of in vivo studies in the REP2004 Database (48) also decreased relative to the total number in the original REP1997 Database (61), even with the addition of 24 new in vivo studies (Table 1). In contrast, the total number of in vitro studies in the REP2004 Database (37) increased relative to the original REP1997Database (30) (Table 1). A similar pattern was observed for the total number of REP values (Table 2). Close to 50% of all of the REP values in the REP1997 Database met one or more of the established exclusion criteria and were identified as candidates for exclusion (Table 2). Of the 818 REP values in the REP1997 Database, approximately 10% (87 of 818) were non-numeric, with PCBs accounting for the majority (96%) of non-numeric entries. The substantial decrease in the total number of REP values was offset to some extent by the addition of 253 new REP values, 167 of which were from in vivo studies. This brought the total number of REP values in the REP2004 Database to 634, 60% of which were from in vivo studies. As was observed for the original REP1997 Database, there was significant variability in the number of REP values among the various congeners, with 1,2,3,7,8,9-HxCDD, 1,2,3,6,7,8-HxCDD, OCDD, 1,2,3,7,8,9-HxCDF, 1,2,3,4,6,7,8-HpCDF, 1,2,3,4,7,8,9-HpCDF, OCDF, PCB114, PCB123, PCB157, PCB167, and PCB189 all having fewer than 10 retained REP values, while 2,3,4,7,8-PeCDF and PCB126 each having close to 100 retained REP values (Table 2). For those congeners that account for most of the background TEQ in blood in the human population, with the exception of 1,2,3,6,7,8-HxCDD (which had only five retained REP values in the REP2004 Database), all were fairly data rich, having anywhere from 25 to 115 retained REP values.

    It is important to note that, while we have identified a substantial number of REP values in the original REP1997 Database that met one or more of our exclusion criteria, and thus, have been identified as candidates for exclusion when using the database for quantitative purposes, this does not invalidate the review conducted by the WHO expert panel in 1997. In fact, expert panel members were aware of a number of the issues that have been outlined in this publication and those issues were considered qualitatively in establishing the WHO98 mammalian TEFs. At the time of the 1997 WHO re-evaluation of the TEFs, the REP1997 Database represented the most comprehensive collection of data on the relative potencies of the PCDDs, PCDFs, and dioxin-like PCBs that was available and was appropriate for the type of evaluation that was being conducted by the WHO expert panel. However, as the level of detail necessary to use the REP database quantitatively exceeds that needed when using the information qualitatively, it was necessary to update and refine the REP1997 Database.

    Summary statistics (i.e., percentiles) describing the distribution of REP values for each congener having at least 10 retained REP values are provided in Table 6. There are 16 congeners in the REP2004 Database that had 10 or more REP values (Table 6). As can be seen in Table 6, the WHO98 mammalian TEFs for PCDD/Fs range from the lower bound to the upper bound of the distribution, while the WHO98 TEFs for the PCBs range from the lower bound to the mid-point. As such, the WHO98 TEFs for the PCDD/Fs are generally more conservative than the WHO98 TEFs for the PCBs. A similar pattern was observed for those congeners that have been shown to account for the majority of the background TEQ in blood in the U.S. population (Fig. 1 and Table 6). The WHO98 TEFs for 1,2,3,7,8-PeCDD and 2,3,4,7,8-PeCDF are generally consistent with the upper bound of the distribution, while the WHO98 TEFs for PCB118, PCB126, and PCB156 are generally consistent with the mid-point. These findings are consistent with what has been reported previously by Finley et al. (2003) and by Haws et al. (2004). In addition, the range of REP values (in vivo + in vitro combined) for each PCDD/F congener was typically spread over two to three orders of magnitude, while the range for PCBs spanned three to five orders of magnitude. While there did appear to be some correlation between the number of retained REP values and the variability observed (i.e., more studies resulted in greater variability) for those congeners with more REP values, the differences in the number of REP values do not account for all of the variability. For example, 2,3,4,7,8-PeCDF and PCB126 had similar numbers of REP values—99 and 115, respectively—yet PCB126 exhibited a much broader range of REP values (i.e., three orders of magnitude for 4-PeCDF and five orders of magnitude for PCB126). Comparing the REP2004 Database to the REP1997 Database, we found little change in the range of REPs (in vivo + in vitro combined) for the four congeners that had the greatest number of REP values in the REP2004 Database (Fig. 1 and Table 6). For 2,3,4,7,8-PeCDF and PCB126, the range of associated REP values was broader than in the original REP1997 Database, but the 25th to 75th percentiles ranges were similar, suggesting that although some of the REP values that were added represented the extremes of the distribution, the majority of the REP values were clustered. Box plots depicting the range of REP values for other congeners are provided in the Appendix of this publication (Fig. A-1 through A-6).

    Summary statistics are provided in Table 7 for those congeners that have fewer than 10 retained REP values. As is evident in Table 7, the range of REP values was much narrower than was observed for those congeners described in Table 6, due primarily to the limited number of REP values (<10) available for each congener. In terms of the impact of our refinement on the overall spread of REP values, there was very little change in the variability of REPs for most of the congeners in Table 7. With respect to where the WHO98 TEFs fall within the range of REP values for those congeners in Table 7, there did not appear to be any consistent pattern, with 1,2,3,7,8,9-HxCDD, 1,2,3,6,7,8-HxCDD, PCB123, PCB157, and PCB189 being above the median (50th percentile) value and WHO98 TEF for all of the other congeners being at or below the median value.

    Given that the WHO expert panel emphasized the in vivo REP values during the course of the re-evaluation in 1997, it is also important to look at the in vivo and in vitro REPs separately. Data for the in vivo REP values are presented in Table 8 and Figure 2, while data for the in vitro REPs are presented in Table 9 and Figure 3. Box plots depicting the range of REP values for additional congeners are provided in the Appendix to this publication. For the in vivo REP values, development of the REP2004 Database generally resulted in a broader range of REP values, while the medians remained fairly constant between the two databases (Table 8). A similar pattern was generally observed for those congeners that account for most of the background TEQ in blood in the U.S. population (Fig. 2), the only exceptions being the decrease in the range of in vivo REP values for PCB118 and the decrease in the median value for PCB156. In both the REP1997 and REP2004 Databases, the WHO98 TEF typically exceeded the median in vivo REP value for both the PCDD/Fs and PCBs (Table 8). For the in vitro REP values, refinement and updating of the REP database did not result in any consistent pattern in terms of the impact on the range of REPs (Table 9). The range of in vitro REP values was generally comparable between the two databases, in some cases broader, and in other cases narrower (Table 9). Similar to what was observed for the in vivo REP values, there was very little change in the median values between the two databases, with the exception of PCB114 and PCB118 (Table 9). Similar to what was observed for the in vivo REPs, there was no consistent pattern for those congeners that account for most of the background TEQ in blood in the U.S. population (Fig. 3). The range for the two PCDD congeners decreased, while the range for 2,3,4,7,8-PeCDF and PCB126 remained fairly constant and the range for PCB118 and PCB156 increased. In terms of where the WHO98 TEFs fell within the range of REP values for each congener, the current mammalian TEFs were at or above the median value for at least half of the PCDD/F and PCB congeners (Table 9).

    In addition to comparing the differences in the ranges of in vivo and in vitro REP values between the REP1997 and REP2004 Databases, it is also important to determine whether the ranges of in vivo vs. in vitro REP values are different, because this would suggest that the wide range of REP values observed for some congeners in those analyses where the in vivo + in vitro REPs have been combined is due primarily to the in vitro REP values. This is a reasonable assumption, because one would expect there to be greater uncertainty associated with the in vitro REP values, and therefore, in combining the in vivo + in vitro REP values, the spread of REP values could potentially be exaggerated. However, comparison of the range of in vivo REPs in the REP2004 Database (Table 8) to the corresponding range of in vitro REPs (Table 9) indicates that, for most congeners, the ranges are actually very similar. As an example, evaluation of those specific congeners that account for most of the background TEQ in blood in the U.S. population, indicates the following:

    1,2,3,7,8-PeCDD: in vivo range = 0.044 to 1.5; in vitro range = 0.095 to 1.1

    1,2,3,6,7,8-HxCDD: cannot compare due to a lack of in vivo REP values

    2,3,4,7,8-PeCDF: in vivo range = 0.0065 to 3.7; in vitro range = 0.1 to 3

    PCB126: in vivo range = 0.000067 to 0.86; in vitro range = 0.00069 to 0.77

    PCB118: in vivo range = 0.00000042 to 0.0023; in vitro range = 0.000002 to 0.075

    PCB156: in vivo range = 0.0000021 to 0.42; in vitro range = 0.000026 to 0.51.

    As is evident from this comparison, the range of in vitro REP values is within the range of in vivo REPs. In fact, with the exception of PCB118, the range of in vitro REP values is actually less than the range of in vivo REPs. Even for PCB118, the difference between the range of in vivo and in vitro REPs is within an order of magnitude.

    An additional observation when comparing the range of in vivo and in vitro REPs for each congener is that the classes of compounds exhibit the following hierarchy with regard to the spread of REP values (from greatest to least): PCBs > PCDFs > PCDDs. The spread of in vivo and in vitro REP values is generally within an order of magnitude for PCDDs. For PCDFs, the in vivo and in vitro REP values are generally spread across 1–2 orders of magnitude. For PCBs, the in vivo REPs are generally spread across 1–5 orders of magnitude, while the in vitro REPs are generally spread across 1–4 orders of magnitude. Finally, analyses were conducted to determine whether the distributions of in vivo and in vitro REPs were statistically different. Such analyses were carried out for those congeners that had at least 10 in vivo REPs and 10 in vitro REPs. The first step involved determining whether the REP data were normally distributed using the Shapiro Wilk Test. Our analyses indicated that the data were not normally distributed, so the distributions of REP values were compared using the Mann Whitney U Test. This analysis indicated that distributions for TCDF, PCB169, PCB105, PCB118, and PCB156 were similar, while those for 2,3,4,7,8-PeCDF, PCB77, and PCB126 were different (at p < 0.05). The next step of the analysis involved comparing the median values (i.e., 95% confidence intervals around the medians) for these congeners. This analysis indicated that the median values for the in vivo REPs were similar to those for the in vitro REPs for TCDF, PCB126, PCB169, PCB105, PCB118, and PCB156. Only 2,3,4,7,8-PeCDF and PCB77 had in vivo and in vitro median REP values that were statistically different.

    These findings indicate that the wide ranges observed for some congeners cannot be explained solely by the fact that the in vivo and in vitro REPs have been combined for discussion purposes in this paper. It is also important to note that we are currently in the process of developing and applying weighting factors to the REP2004 Database to differentiate between studies of different quality and relevance (e.g., there should be less uncertainty in extrapolating from in vivo studies than from in vitro studies, and therefore, in vivo studies should be given more weight than in vitro studies). Preliminary analyses by Finley and coworkers (2003) indicated that the application of such weighting factors did not have much of an impact on the distributions. Nonetheless, it is important to answer this question with regard to the REP2004 Database. This work is currently underway.

    Finally, it has been suggested that the variability in REP values for any given congener may be due in part to the fact that they are often based on many different endpoints (USEPA, 2003b). To examine this issue, the retained REP values in the REP2004 Database were sorted based on the following classes of measure: biochemical changes, toxicity, carcinogenicity. Examples of specific endpoints within each of these categories are as follows:

    Biochemical changes (e.g., enzyme induction, cytochrome P450 1A1 [CYP1A1] messenger RNA [mRNA], liver lipid, phosphenolpyruvate carboxykinase [PEPCK], porphyrin, vitamin A, and thyroxine [T4]).

    Toxicity (e.g., hyperplasia, body-weight changes, cleft palate, fixed ratio [FR] reinforcement rate, immunoglobulin-M [IGM], plaque-forming cells [PFCs]/spleen, alanine aminotransferase [ALAT], aspartate aminotransferase [ASAT]).

    Carcinogenicity (e.g., cholangiocarcinoma, hepatocellular carcinoma, promotion index, volume fraction, percent S phase, number of cells, cell size, cyclin A, intercellular communication).

    Data for the in vivo REP values are presented in Table 10, while data for the in vitro REP values are presented in Table 11. Essentially equal numbers of in vivo REP values were based on biochemical (181 out of 383) and toxicity endpoints (188 out of 383) as indicated in Table 10. There were very few in vivo REPs based on carcinogenicity endpoints, with carcinogenicity endpoints accounting for only 4% of all in vivo REPs. PCB126 and 2,3,4,7,8-PeCDF had the greatest number of in vivo REPs based on carcinogenicity endpoints, due primarily to the recent NTP studies (Walker et al., 2005). Within classes, the range of in vivo REPs for the PCDD/Fs was generally narrower than the range for PCBs (Table 10). A similar pattern was observed for the in vitro REPs (Table 11). Across classes, the ranges of both in vivo and in vitro REPs based on toxicity endpoints were generally narrower than for those based on biochemical endpoints (Tables 10 and 11). This observation for the in vitro REPs is difficult to interpret, however, given the limited number of in vitro REP values that are based on toxicity endpoints. As indicated in Table 11, the majority of in vitro REP values (81%) are based on a biochemical endpoint, with receptor binding accounting for 10% of the in vitro REP values and toxicity endpoints accounting for only 5% of the values.

    Implications for Risk Assessment

    Quantitative uncertainty analysis is now an essential element in the risk assessment process. Such analyses, which include probabilistic techniques, provide a means of incorporating information concerning uncertainty and variability into the final risk estimates. Therefore, we suggest that some form of a refined REP database that can better support quantitative uncertainty analyses ultimately be developed for use in risk assessment applications. A refined database could be used to facilitate quantitative analyses, such as the development of REP distributions, as a supplement to and/or in addition to the derivation of the "point-estimate" TEFs. For example, REP distributions could be used to establish a consistent percentile point-estimate TEF to represent the "central tendency" and "plausible upper bound" for each congener (e.g., the 50th and 90th percentiles of the distribution, respectively). The use of REP distributions in a probabilistic analysis of risk would theoretically allow for a more informed discussion of the uncertainty in the risk estimates.

    The analysis presented here provides one possible methodology for developing such a database; certainly, other approaches might be equally valid. In addition, it is important to be aware of the heterogeneous nature of the REP data, particularly with respect to data quality and relevance. Development of a quantitative, transparent, and reproducible weighting scheme for individual REP values would likely increase consistency in the derivation of the TEF values, facilitate characterization of uncertainty, and could also be used to evaluate new REP data as they become available.

    Clearly, additional steps can be taken to refine the database even further. The wide variety of methods used to calculate REP values, ranging from the sophisticated models used by Walker and coworkers (2005) to crude graphical techniques, as well as the host of problems associated with the underlying dose response data (e.g., non-parallel dose-response curves, differences in maximal response among congeners within a study, incomplete dose-response data due to insufficient dose levels), likely contribute to the substantial variation in REP values. Clearly, this is an area that warrants further evaluation. It should be noted, however, that the method employed to calculate REP values is limited by the type of data available. Development of a more detailed database that includes the individual data sets would allow comparative meta-analyses across studies.

    As indicated in our discussions concerning development of the REP2004 Database, REP values presented in both the REP1997 and REP2004 Database were, for the most part, taken as-is. Not all REP values were verified, nor were the underlying dose-response data evaluated to determine whether the dose response curves were parallel. Such analyses may be warranted as the database is refined further in the future. Additionally, while we chose to characterize the distributions of REP values for each congener having at least 10 REP values based on percentiles, a logical next step would be to determine whether the data fit a specific distributional shape, and then to develop probability density functions that could be used in probabilistic analyses such as Monte Carlo analyses.

    The TEF approach has been clearly identified as "interim" and is subject to periodic review as new scientific information becomes available. Therefore, the conclusions and recommendations outlined in this current review should be considered an extension of earlier efforts by WHO and others. As we gain a better understanding of the modes of action underlying this class of compounds, as more data become available concerning the relative potencies of these compounds, and as we develop more sophisticated quantitative tools, we will be able to further improve the TEF methodology. With each subsequent review of the TEFs, the scientific rigor, clarity, and transparency of the TEF approach have been improved. The recommendations set forth within this manuscript should serve to improve the process even further.

    SUPPLEMENTARY DATA

    Supplementary data include a complete copy of the REP2004 Database, which consists of two PDF files (REP2004–2 VIVO.PDF and REP2004–2 VITRO.PDF) that contain all of the relevant mammalian in vivo and in vitro data. The data elements shown in the PDF files are those elements that were reviewed by the authors of the REP2004 Database as described in this publication. The data are sorted by congener and are grouped further based on whether they represent records in the original REP1997 Database or are new records, as well as based on whether they should be retained or excluded in quantitative analyses. Within each group of records, they are sorted in order from lowest to highest REP values. The REP2004 Database can be obtained on the publisher's website (www.toxsci.oupjournals.org).

    APPENDIX

    FIG. A-1. Distributions of REP values in the REP1997 Database (in vivo + in vivo combined).

    FIG. A-2. Distributions of REP values in the REP2004 Database (in vivo + in vivo combined).

    FIG. A-3. Distributions of in vivo REP values in the REP1997 Database.

    FIG. A-4. Distributions of in vivo REP values in the REP2004 Database.

    FIG. A-5. Distributions of in vitro REP values in the REP1997 Database.

    FIG. A-6. Distributions of in vitro REP values in the REP2004 Database.

    NOTES

    The contents of this paper reflect the opinions and views of the authors and do not represent the official views or policies of NIEHS, NIH, or USEPA. The mention of trade names and commercial products does not constitute endorsement or use recommendation.

    ACKNOWLEDGMENTS

    We extend our thanks to Dr. Annette Santamaria and Amy Bradley for their help with many aspects of this project, as well as to Drs. Richard Peterson and Martin van den Berg for their valuable input and peer review of the REP2004 Database. We also thank Janet Diliberto and Drs. Annika Hanberg and Helen Hakansson for their review of this manuscript. Finally, we extend our thanks to the Institute of Environmental Medicine at the Karolinska Institute for making the REP1997 Database available to us and acknowledge Drs. U. G. Ahlborg, A. Hanberg, F. Waern, and P. Andersson for their efforts in developing this database. Their work was made possible by funding from the WHO and Swedish government agencies, and we acknowledge these organizations for their contributions to this important work. Development of the REP2004 Database was funded in part by Tierra Solutions, Inc. Additional funding was provided by Exponent, Inc. This research was also supported in part by the International Research Program of the NIH, National Institute of Environmental Health Sciences.

    REFERENCES

    Ahlborg, U. G., Becking, G. C., Birnbaum, L. S., Brouwer, A., Derks, H. J. G. M., Feeley, M., Golog, G., Hanberg, S., Larsen, J. C., Liem, A. K. D., et al. (1994). Toxic equivalency factors for dioxin-like PCBs: Report on a WHO-ECEH and IPCS consultation, Dec. 1993. Chemosphere 28, 1049–1067.

    Barnes, D., Alford-Stevens, A., Birnbaum, L. et al. (1991). Toxicity equivalency factors for PCBs Qual. Assur. 1, 70–81.

    Birnbaum, L. S. (1994). Mechanism of dioxin toxicity: Implications for risk assessment. Environ. Health Perspect. 102(Suppl. 9), 157–167.

    Birnbaum, L. S. (1999). TEFs: A practical approach to a real-world problem. Human Ecol. Risk Assess. 5, 13–24.

    Birnbaum, L. S., and DeVito, M. J. (1995). Use of toxic equivalency factors for risk assessment for dioxins and related compounds. Toxicology 105, 391–401.

    Birnbaum, L. S., and Tuomisto, J. (2000). Non-carcinogenic effects of TCDD in animals. Food Additives and Contaminants 17, 275–288.

    Birnbaum, L. S., Emond, C., and DeVito, M. J. (2004). Application of a PBPK model to aid in understanding the relative potencies (REPs) of dioxin-like chemicals. The Toxicologist 78, 362.

    DeCarprio, A., McMartin, D., O'Keefe, P., Rej, R., Silkworth, J. B., and Kaminsky, L. S. (1986). Subchronic oral toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the guinea pig: Comparison with a PCB-containing transformer fluid pyrolysate. Fundam. Appl. Toxicol. 6, 454–463.

    DeVito, M. J., and Birnbaum, L. S. (1994). Toxicology of the dioixins and related chemicals. In Dioxins and Health (A. Schecter, Ed.), pp. 139–162. Elsevier, New York.

    DeVito, M. J., Maier, W. E., Diliberto, J. J., and Birnbaum, L. S. (1993). Comparative ability of various PCBs, PCDFs, and TCDD to induce cytochrome P450 1A1 and 1A2 activity following 4 weeks of treatment. Fundam. Appl. Toxicol. 20, 125–130.

    DeVito, M. J., Menache, M. G., Diliberto, J. J., Ross, D. G., and Birnbaum, L. S. (2000). Dose-response relationships for polyhalogenated dioxins and dibenzofurans following subchronic treatment in mice. 1. CYP1A1 and CYP1A2 enzyme activity in liver, lung, and skin. Toxicol. Appl. Pharmacol. 147, 267–280.

    Dyke, P. H., and Stratford, J. (2002). Changes in the TEF schemes can have significant impacts on regulation and management of PCDD/F and PCB. Chemosphere 47, 103–116.

    Eadon, G., Kaminsky, L. S., Silkworth, J., et al. (1986). Calculation of 2,3,7,8-TCDD equivalent concentrations of complex environmental contaminant mixtures. Environ. Health Perspect. 70, 221–227.

    Fattore, E., Chu, I., Salomon, S., Fanelli, R., Falk-Filippson, A., and Hakansson, H. (2004). Dose-response assessment using benchmark dose approach of changes in hepatic EROD activity for individual polychlorinated biphenyl congeners. Organohalogen Compounds 66, 3402–3407.

    Finley, B. L., Connor, K. T., and Scott, P. K. (2003). The use of toxic equivalency factor distributions in probabilistic risk assessments for dioxins, furans, and PCBs. J. Toxicol. Environ. Health Part A 66, 533–550.

    Hankinson, O. (1995). The aryl hydrocarbon receptor complex. Ann Rev. Pharmacol. Toxicol. 35, 307–340.

    Haws, L., Harris, M., Su, S., Walker, N., Birnbaum, L., DeVito, M., Farland, W., Connor, K., Santamaria, A., and Finley, B. (2004). A preliminary approach to characterizing variability and uncertainty in the mammalian PCDD/F and PCB TEFs. Organohalogen Compounds 66, 3439–3445.

    Leece, B., Denomme, M. A., Towner, R., Li, S. M. A., and Safe, S. (1985). Polychlorinated biphenyls: Correlation between in vivo and in vitro quantitative structure activity relationships (QSARs). J. Toxicol. Environ Health. 16, 379–388.

    Martinez, J. M., DeVito M. J., Birnbaum L. S. and Walker N. J. (2003). The toxicology of dioxins and dioxin-like compounds. In Dioxins and Health (T. Gasiewicz and A Schecter, Eds.), 2nd ed., pp. 139–158. John Wiley & Sons, Hoboken, NJ.

    McConnell, E. E., Moore, J. A., Haseman, J. K., and Harris, M. W. (1978). Comparative toxicity of chlorinated dibenzo-p-dioxins in mice and guinea pigs. Toxicol. Appl. Pharmacol. 44, 335–356.

    NATO/CCMS (North Atlantic Treaty Organization, Committee on the Challenges of Modern Society). (1988a). International toxicity equivalency factor (I-TEF) method of risk assessment for complex mixtures of dioxins and related compounds. Report No. 176.

    NATO/CCMS (North Atlantic Treaty Organization, Committee on the Challenges of Modern Society). (1988b).Scientific basis for the development of international toxicity equivalency factor (I-TEF) method of risk assessment for complex mixtures of dioxins and related compounds. Report No. 178.

    NATO/CCMS (North Atlantic Treaty Organization, Committee on the Challenges of Modern Society). (1988c). Inventory of regulations/statutes concerning dioxins and related compounds. Report No. 169.

    Needham, L., Barr, D., Caudill, S. A., Pirkle, J. L., Turner, W., Osterloh, J., Jones, R., et al. (2005). Concentrations of environmental chemicals associated with neurodevelopmental effects in U.S. population. Neurotoxicology Online 11 January 2005 (doi:10.1016/j.neuro.2004.09.05).

    Okey, A. B., Riddick, D. S., and Harper, P. A. (1994). The Ah receptor: Mediator of the toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and related compounds. Toxicol. Lett. 70, 1–22.

    Olson, J. R., Bellin, J. S., and Barnes, D. G. (1989). Reexamination of data used for establishing toxicity equivalency factors (TEFs) for chlorinated dibenzo-p-dioxins and dibenzofurans (CDDs and CDFs). Chemosphere 18, 371–381.

    OME (Ontario Ministry of the Environment) (1984). Scientific criteria document for standard development. No. 4–84. Polychlorinated Dibenzo-p-dioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs).

    Safe, S. (1990). Polychlorinated biphenyls (PCBs), dibenzo-p-dioxins (PCDDs), dibenzofurans (PCDFs) and related compounds: Environmental and mechanistic considerations which support the development of toxic equivalency factors (TEFs). Crit. Rev. Toxicol. 21, 51–88.

    Safe, S., (1994). Polychlorinated biphenyls (PCBs): Environmental impact, biochemical and toxic responses, and implications for risk assessment. Crit. Rev. Toxicol. 24, 87–149.

    Schwetz, B. A., Norris, J. M., Sparschu, G. L., Rowe, V. K., Gehring, P. J., Emerson, J. L., and Gerbig, C. G. (1973). Toxicology of chlorinated dibenzo-p-dioxins. Environ. Health Perspect. 5, 87–99.

    Sewall, C. H., and Lucier, G. W. (1995). Receptor-mediated events and the evaluation of the Environmental Protection Agency (EPA) of dioxin risks. Mutat. Res. 333, 111–122.

    Silkworth, J. B., Cutler, D. S., Antrim, L., Huston, D., Tumasonis, C., and Kaminsky, L. S. (1989). Teratology of 2,3,7,8-tetrachlorodibenzo-p-dioxin in a complex environmental mixture from Love Canal. Fundam. Appl. Toxicol. 13, 1–15.

    Suter-Hofmann, J., and Schlatter, C. H. (1989). Subchronic relay toxicity study with a mixture of polychlorinated dioxins (PCDDs) and polychlorinated furans (PCDFs). Chemosphere 18, 277–282.

    Toyoshiba, H., Walker, N. J., Bailer, A. J., and Portier, C. J. (2004). Evaluation of toxic equivalency factors for induction of cytochromes P450 CYP1A1 and CYP1A2 enzyme activity by dioxin-like compounds. Toxicol. Appl. Pharmacol. 194, 156–168.

    USEPA (U.S. Environmental Protection Agency). (1987). Interim procedures for estimating risks associated with exposures to mixtures of chlorinated dibenzo-p-dioxins and dibenzofurans (CDDs and CDFs). EPA/625/3-87/012.

    USEPA (U.S. Environmental Protection Agency). (1989). Interim procedures for estimating risks associated with exposures to mixtures of chlorinated dibenzo-p-dioxins and -dibenzofurans (CDDs and CDFs) and 1989 update. EPA/625/3-89/016.

    USEPA (U.S. Environmental Protection Agency). (2003a). Exposure and human health reassessment of 2,3,7,8-tetrachlorodibnzo-p-dioxin (TCDD) and related compounds. Part I: Estimating exposure to dioxin-like compounds. Volume 1: Sources of dioxin-like compounds in the United States. EPA/600/P-00/001Cb. December 2003. NAS Review Draft.

    USEPA (U.S. Environmental Protection Agency). (2003b). Exposure and human health reassessment of 2,3,7,8-tetrachlorodibnzo-p-dioxin (TCDD) and related compounds. Part II Health assessment for 2,3,7,8-tetrachlordibenzo-p-dioxin (TCDD) and related compounds. Chapter 9: Toxic Equivalency Factors (TEF) for dioxin and related compounds. NCEA-I-0836. December 2003. NAS Review Draft.

    van den Berg, M., Birnbaum, L., Bosveld, A. T. C., Brunstrom, B., Cook, P., Feeley, M., Giesy, J. P., Hanberg, A., Hasegawa, R., Kennedy, S., et al. (1998). Toxic equivalency factors (TEFs) for PCBs, PCDDs, PCDFs for humans and wildlife. Environ. Health Perspect. 106, 775.

    Vondracek, J., Machala, M., Bryja, V., Chramastova, K., Kremar, R., Dietrich, C., Hampl, A., and Kozubik, A. (2005). Aryl hydrocarbon receptor-activating polychlorinated biphehyls and their hydroxylated metabolites induce cell proliferation in contact-inhibited rat liver epithelial cells. Toxicol. Sci. 83, 53–63.

    Walker, N. J., Crockett, P. W., Nyska, A., Brix, A. E., Jokinen, M. P., Sells, D. M., Hailey, J. R., Easterling, M., Haseman, J. K., Yin, M., et al. (2005). Dose-additive carcinogenicity of a defined mixture of "dioxin-like compounds." Environ. Health Perspect. 113, 43–48.

    Yrjanheikki, E. J. (1992). Review of the models of TEFs in assessing health risks of PCDDs and PCDFs. Toxic Sub. J. 12, 283–288.

查询更多dioxin相关信息在本站>>

《毒物学科学杂志》2006年1月第89卷第1期