Regulatory Toxicology and Pharmacology 67 (2013) 278–284

Contents lists available at ScienceDirect

Regulatory Toxicology and Pharmacology

journal homepage: www.elsevier .com/locate /yr tph

A sensitivity analysis using alternative toxic equivalency factorsto estimate U.S. dietary exposures to dioxin-like compounds

0273-2300/$ - see front matter Published by Elsevier Inc.http://dx.doi.org/10.1016/j.yrtph.2013.08.007

Abbreviations: Cij, DLC average concentrations in each food category; DLCs, dioxin-like compounds; Eij, DLC intake estimate for each food category; EEnvironmental Protection Agency; fj, fat fraction estimates for each food category; HpCDF, heptachlorodibenzofuran; HxCDD, hexachlorodibenzo-p-dioxin;hexachlorodibenzofuran; PCBs, polychlorinated biphenyls; PCDDs, polychlorinated dibenzo-p-dioxins; PCDFs, polychlorinated dibenzofurans; PeCDD, pentachlorodidioxin; REP, relative estimates of potency; rj, food ingestion rate for each food category; SI, Supplementary Information; TCDD, 2,3,7,8-tetrachlorodibenzo-p-dioxin; Tequivalency factor; TEFA, alternative TEF value based on the percentile used (e.g., where ‘‘A’’ indicates the percentile—10th, 25th, 50th, 75th, 90th); TEFB, baseline TTEFL, lower TEF value; TEFU, upper TEF value; TEQ, toxicity equivalence; TEQAj, alternative daily TEQ intake rate estimates for the jth food category; TEQAT, alternadaily TEQ intake estimates; TEQL, lower TEQ value; U.S., United States; WHO, World Health Organization.⇑ Corresponding author. Address: National Center for Environmental Assessment, U.S. Environmental Protection Agency, 26 W Martin Luther King Drive (M

Cincinnati, OH 45268, USA. Fax: +1 513 487 2539.E-mail address: rice.glenn@epa.gov (G.E. Rice).

1 Present address: Department of Environmental Health Science, Indiana University Richard M. Fairbanks School of Public Health, Indiana University-Purdue UIndianapolis (IUPUI) Campus, 714 N Senate Avenue, Indianapolis 46202, USA.

Shahid Parvez a,1, Amanda M. Evans a, Matthew Lorber b, Belinda S. Hawkins c, Jeffery C. Swartout c,Linda K. Teuschler c, Glenn E. Rice c,⇑a Oak Ridge Institute for Science and Education (ORISE), MC-100-44, P.O. Box 117, Oak Ridge, TN 37831-0117, USAb National Center for Environmental Assessment, Office of Research and Development, U.S. Environmental Protection Agency, 1200 Pennsylvania Avenue, NW, Washington,DC 20004, USAc National Center for Environmental Assessment, Office of Research and Development, U.S. Environmental Protection Agency, 26 W Martin Luther King Drive, Cincinnati, OH45268, USA

a r t i c l e i n f o

Article history:Received 9 April 2013Available online 21 August 2013

Keywords:Dioxin-like compounds (DLCs)Toxic equivalency factor (TEF)Toxic equivalency (TEQ)2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD)Dietary intakeRelative estimates of potency (REP)Sensitivity analysisDose addition

a b s t r a c t

EPA recommends sensitivity analyses when applying the toxic equivalency factor (TEF) method toevaluate exposures to dioxin-like compounds (DLCs). Applying the World Health Organization’s (WHO)2005 TEF values and estimating average U.S. daily dietary intakes of 25 DLCs from eight food categories,we estimate a toxic equivalency (TEQ) intake of 23 pg/day. Among DLCs, PCB 126 (26%) and 1,2,3,7,8-PeCDD(23%) dominate TEQ intakes. Among food categories, milk (14%), other dairy (28%), beef (25%), and sea-food (18%) most influenced TEQ intakes. We develop two approaches to estimate alternative TEF values.Based on WHO’s assumption regarding TEF uncertainty, Approach1 estimates upper and lower TEFs foreach DLC by multiplying and dividing, respectively, its individual TEF by ± half a log. Based on compiledempirical ranges of relative potency estimates, Approach2 uses percentile values for individual TEFs.Total TEQ intake estimates using the lower and upper TEFs based on Approach1 were 8 and 68 pgTEQ/day, respectively. The 25th and 75th percentile TEFs from Approach2 yielded 12 and 28 pg TEQ/day, respectively. The influential DLCs and food categories remained consistent across alternative TEFs,except at the 90th percentile using Approach2. We highlight the need for developing underlying TEFprobability distributions.

Published by Elsevier Inc.

1. Introduction The World Health Organization (WHO) and the United States

Dioxin-like compounds (DLCs), including polychlorinated di-benzo-p-dioxins (PCDDs), polychlorinated dibenzofurans (PCDFs),and some polychlorinated biphenyls (PCBs), are structurally andtoxicologically related halogenated dicyclic aromatic hydrocarbons(USEPA, 2003). DLCs are persistent and lipophilic, bioaccumulatingin aquatic and terrestrial food webs. For most people, the principalpathway of exposure to DLCs is ingestion of animal based products(Lorber et al., 2009, 2010; USFDA, 2007).

(U.S.) Environmental Protection Agency (EPA), among others, rec-ommend using the toxic equivalence factor (TEF) method for eval-uating human health risks from exposures to DLC mixtures(USEPA, 2010; Van den Berg et al., 2006). Although the use ofwhole mixture data or data on a sufficiently similar mixture arepreferred risk assessment approaches for environmental mixtures,when data are not sufficient to apply these methods, component-based approaches, such as the TEF method are employed (USEPA,2000). The TEF method is based on dose addition (USEPA, 2000).

PA, U.S.HxCDF,

benzo-p-EF, toxic

EF value;tive total

S-A110),

niversity

S. Parvez et al. / Regulatory Toxicology and Pharmacology 67 (2013) 278–284 279

After reevaluating the additivity concept for TEFs, WHO’s 2005 ex-pert panel on the dioxin TEFs concluded that results from manyin vivo DLC mixture studies conducted between 1998 and 2005were consistent with additivity and supportive of the TEF approach(Van den Berg et al., 2006).

In the TEF method, doses of individual DLCs are scaled based ontheir toxic potency relative to the index chemical, 2,3,7,8-tetra-chlorodibenzo-p-dioxin (TCDD). These numerical scaling factorsare termed TEFs. TCDD is the index chemical for this class of chem-icals because its dose-response behavior is well studied and manystudies have compared its toxic potency to that of other DLCs. Theproduct of a DLC’s dose and its assigned TEF value is toxicologicallyequivalent to a dose of TCDD, termed the TCDD toxicity equiva-lence (TEQ) (USEPA, 2000; USEPA, 2010). The sum of these individ-ual products is the total TEQ (i.e., the TCDD dose that is the toxicequivalent of the DLC mixture dose), presented here as an oral hu-man dose and referred to hereafter as intake. The TEF for each DLCis usually presented as a point estimate. Here, we assess the impactof varying these TEF point estimates.

A WHO expert panel developed the current, consensus TEF val-ues (Van den Berg et al., 2006). In this most recent update of theTEFs, the panelists evaluated data from an empirical range of val-ues developed by Haws et al. (2006) and used expert judgmentto assign a final TEF value for each DLC. These ranges were basedon a database of in vivo and in vitro studies that compared thetoxic potency of individual DLCs to that of TCDD. From each ofthese comparative toxicity studies, Haws et al. (2006) calculatedrelative estimates of potency (REPs); the toxicity tests, test condi-tions, sex, and species used to generate these REPs varied acrossthe DLCs. REP estimates were not weighted based on study typeor study quality. The panel assigned TEF values using single pointestimates from toxicological studies, rather than specific pointswithin the REP ranges. Typically, the TEF assignment for eachDLC was between the 50th and 75th percentiles of the REP rangeand were generally closer to the 75th percentile to be health pro-tective (Van den Berg et al., 2006). The most recent TEF assign-ments (Van den Berg et al., 2006) are listed in Table 1 and arereferred to as baseline TEF (TEFB) values.

Here, we developed two approaches to address the sensitivity ofTEQ intake to alternative TEF values, as identified in EPA’s 2010 TEFdocument (USEPA, 2010). In the first approach, we considered Vanden Berg et al.’s (2006) description of the TEF as a ‘‘central valuewith a degree of uncertainty assumed to be at least ± half a log’’and implemented this by estimating upper and lower alternativeTEF values by multiplying and dividing each DLC-specific TEF by3.16 (i.e., ‘‘half a log’’), respectively (termed ‘‘Approach1’’). In thesecond approach, we estimated alternative TEF values using per-centile values from the REP data developed by Haws et al. (2006)for each DLC based on the underlying REP empirical range (termed,‘‘Approach2’’), using the observed variation as a surrogate for over-all uncertainty in the TEF values. When we used the REP empiricalrange from only the in vivo studies comparing the toxicity of DLCsto TCDD, we termed this ‘‘Approach2A’’; when we used the REPempirical range from the combined in vivo and in vitro compara-tive toxicity studies, we termed this ‘‘Approach2B’’.

In this study, we estimated the U.S. daily total TEQ dietary in-take from eight food categories to be approximately 23 pg/dayusing baseline TEF values. Previously, Lorber et al. (2010) esti-mated that food intakes from these eight categories accountedfor approximately 90% (29.8 pg/day) of the U.S. daily total TEQ die-tary intake (33.5 pg/day). There are several differences betweenour analysis and that of Lorber et al. (2010): (1) They includedinhalation, soil dermal contact, and water, soil, and vegetable oilingestion pathways that we did not consider; (2) we evaluated fiveadditional PCBs not included in Lorber et al.; and (3) we used up-dated DLC concentrations for seafood and eggs as reported by Bol-

ger and Murray (Bolger, P.M., Murray, C., III, Concentrations ofdioxin-like compounds in commercial seafood and eggs, 2011,Unpublished results.).

This manuscript focuses on EPA’s recommendation (USEPA,2010) to consider a sensitivity analysis when using the TEF meth-od. Holding constant both the individual DLC concentrations ineight food categories and the daily intake rates of these food cate-gories, we identified the potential ranges of U.S. TEQ exposuresfrom intake of these food categories and examined the influenceof alternative TEF values on the TEQ intake estimates associatedwith intakes of specific DLCs and DLC groups.

2. Materials and methods

To calculate TEQ intake estimates and conduct sensitivity anal-yses of alternative TEF (TEFA) values, we needed TEFA values foreach DLC and oral intake estimates (Eij) for each DLC in the eightfood categories. This section describes how we obtained valuesfor these factors. Additional information on TEFA values and theDLC concentrations, food ingestion rates, and fat fraction estimatesfor each food category is provided in the Supplementary Informa-tion (SI).

2.1. Oral DLC intake estimate (Eij)

For the ith DLC (for i = 1, 2,. . ., n) and jth food category (for j = 1,2,. . ., m), oral DLC intakes (Eij, pg/day) were estimated as the prod-uct of DLC average concentrations in food (Cij, pg/g), food ingestionrate (rj, g/day), and the total fat fraction (fj) using the following Eq.(1).

Eij ¼ Cij � rj � fj ð1Þ

2.2. DLC concentrations in U.S. foods (Cij, pg/g)

Table A.1 in the SI summarizes DLC average concentrations (pg/g) in the following eight food categories: milk, other dairy (e.g.,cheeses, yogurt), beef, poultry, pork, other meats (e.g., unidentifiedmeat in casseroles), eggs, and seafood, based on U.S. estimates.From Lorber et al. (2009, 2010), we obtained DLC concentrationsfor all food categories except seafood and eggs (Bolger, P.M., Mur-ray, C., III, Concentrations of dioxin-like compounds in commercialseafood and eggs, 2011, Unpublished results.). Lorber et al. (2010)assumed that DLC concentrations reported as below the detectionlimit in individual food samples were present at concentrationsequal to one-half the detection limit. They also evaluated trendsin DLC exposures over time and limited their analyses to DLCsfor which concentration data were available over relevant timeperiods. Lorber et al. (2010) included many dioxins and furansfor which TEF values had been developed. Although concentrationdata were available for eight PCBs, they identified only three (PCB77, PCB 126, and PCB 169) that were consistently reported in allfood categories and limited their analysis to these three PCBs. Inour analysis, we included the additional concentration data forPCB 81, PCB 118, PCB 105, PCB 156, and PCB 157 that were avail-able for some food categories. We did not evaluate the varianceof any DLC concentrations in the food categories due to our focuson TEF uncertainty.

2.3. Food ingestion rate (rj, g/day)

The food ingestion rates (g/day) for each food category (rj) aremilk (175 g/day), other dairy(55 g/day), beef (49.7 g/day), pork(15.4 g/day), poultry (35 g/day), other meats (24.5 g/day), eggs(16.8 g/day), and seafood (15.5 g/day). These rates are based on

Table 1Baseline and alternative TEF values based on Approach1 and Approach2A.

Dioxin-like compounds (DLCs) Alternative toxic equivalency factors (TEFA)a

Approach1 Approach2A

TEFB TEFL TEFU Nb TEF10 TEF25 TEF50 TEF75 TEF90

Polychlorinated dibenzo-p-dioxins (PCDDs)2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD or 2,3,7,8-TCDD) 1 1 1 NA 1 1 1 1 11,2,3,7,8-Pentachlorodibenzo-p-dioxin (1,2,3,7,8-PeCDD) 1 0.32 3.2 36 0.1 0.1 0.4 0.6 0.81,2,3,4,7,8-Hexachlorodibenzo-p-dioxin (1,2,3,4,7,8-HxCDD) 0.1 0.032 0.32 15 0.03 0.03 0.06 0.09 0.11,2,3,6,7,8-Hexachlorodibenzo-p-dioxin (1,2,3,6,7,8-HxCDD) 0.1 0.032 0.32 0c 0.1 0.1 0.1 0.1 0.11,2,3,7,8,9-Hexachlorodibenzo-p-dioxin (1,2,3,7,8,9-HxCDD) 0.1 0.032 0.32 1 0.03 0.03 0.03 0.03 0.031,2,3,4,6,7,8-Heptachlorodibenzo-p-dioxin (1,2,3,4,6,7,8-HpCDD) 0.01 0.003 0.03 12 0.004 0.004 0.01 0.01 0.021,2,3,4,5,6,7,8-Octachlorodibenzo-p-dioxin (OCDD or 1,2,3,4,5,6,7,8-OCDD) 0.0003 0.000095 0.00095 1 0.0003 0.0003 0.0003 0.0003 0.0003

Polychlorinated dibenzofurans (PCDFs)2,3,7,8-Tetrachlorodibenzofuran (2,3,7,8-TCDF) 0.1 0.032 0.32 17 0.01 0.01 0.03 0.1 0.31,2,3,7,8-Pentachlorodibenzofuran (1,2,3,7,8-PeCDF) 0.03 0.009 0.09 20 0.01 0.01 0.02 0.08 0.12,3,4,7,8-Pentachlorodibenzofuran (2,3,4,7,8-PeCDF) 0.3 0.095 0.95 82 0.05 0.05 0.2 0.3 0.71,2,3,4,7,8-Hexachlorodibenzofuran (1,2,3,4,7,8-HxCDF) 0.1 0.032 0.32 6 0.03 0.03 0.05 0.07 0.11,2,3,6,7,8-Hexachlorodibenzofuran (1,2,3,6,7,8-HxCDF) 0.1 0.032 0.32 11 0.01 0.01 0.08 0.09 0.11,2,3,7,8,9-Hexachlorodibenzofuran (1,2,3,7,8,9-HxCDF) 0.1 0.032 0.32 0c 0.1 0.1 0.1 0.1 0.12,3,4,6,7,8-Hexachlorodibenzofuran (2,3,4,6,7,8-HxCDF) 0.1 0.032 0.32 3 0.02 0.02 0.02 0.06 0.081,2,3,4,6,7,8-Heptachlorodibenzofuran (2,3,4,7,8-HpCDF) 0.01 0.003 0.03 0c 0.01 0.01 0.01 0.01 0.011,2,3,4,7,8,9-Heptachlorodibenzofuran (1,2,3,4,7,8,9-HpCDF) 0.01 0.003 0.03 0c 0.01 0.01 0.01 0.01 0.011,2,3,4,6,7,8,9-Octachlorodibenzofuran (OCDF or 1,2,3,4,6,7,8,9-OCDF) 0.0003 0.000095 0.00095 6 0.00002 0.00002 0.00008 0.0006 0.001

Polychlorinated biphenyls (PCBs)3,30 ,4,40-Tetrachlorobiphenyl⁄ (PCB 77 or 3,30 ,4,40-TCB) 0.0001 0.000032 0.00032 16 0.000006 0.000006 0.00006 0.0001 0.023,4,40 ,5-Tetrachlorobiphenyl⁄ (PCB 81 or 3,4,40 ,5-TCB) 0.0003 0.000095 0.00095 0c 0.0003 0.0003 0.0003 0.0003 0.00032,30 ,4,40 ,5-Pentachlorobiphenyl⁄⁄ (PCB 118 or 2,30 ,4,40 ,5-PeCB) 0.00003 0.0000095 0.000095 15 0.000002 0.000002 0.00002 0.00005 0.0012,3,30 ,4,40-Pentachlorobiphenyl⁄⁄ (PCB 105 or 2,3,30 ,4,40-PeCB) 0.00003 0.0000095 0.000095 16 0.000002 0.000002 0.0004 0.0001 0.0013,30 ,4,40 ,5-Pentachlorobiphenyl⁄ (PCB 126 or 3,30 ,4,40 ,5-PeCB) 0.1 0.032 0.32 86 0.02 0.02 0.1 0.2 0.42,3,30 ,4,40 ,5-Hexachlorobiphenyl⁄⁄ (PCB 156 or 2,3,30 ,4,40 ,5-HxCB) 0.00003 0.0000095 0.000095 16 0.000005 0.000005 0.00006 0.0005 0.092,3,30 ,4,40 ,50-Hexachlorobiphenyl⁄⁄ (PCB 157 or 2,3,30 ,4,40 ,50-HxCB) 0.00003 0.0000095 0.000095 2 0.0006 0.0006 0.001 0.001 0.0023,304,40 ,5,50-Hexachlorobiphenyl⁄⁄ (PCB 169 or 3,304,40 ,5,50-HxCB) 0.03 0.009 0.09 15 0.0004 0.0004 0.02 0.2 0.6

NA = not applicable. ‘⁄’ and ‘⁄⁄’ identifies non-ortho coplanar and mono-ortho coplanar PCBs, respectively.a TEFA is the toxic equivalency factor at percentile A, where A is the 10th, 25th, 50th, 75th, or 90th percentile using Approach2A or A equals B (baseline), L (lower) or U (upper), referring to the range using Approach1.b Total number of in vivo studies in Haws et al. (2006).c No empirical range available for in vivo TEF values, hence TEFB was used for sensitivity analyses.

280S.Parvez

etal./R

egulatoryToxicology

andPharm

acology67

(2013)278–

284

S. Parvez et al. / Regulatory Toxicology and Pharmacology 67 (2013) 278–284 281

the mean adult per capita ingestion rates reported by Lorber et al.(2009), assuming a 70 kg adult. In this paper, our focus was on TEFuncertainty, so we held the ingestion rates constant while varyingthe TEFs and did not evaluate the variance in food category inges-tion rates. Fat fraction (fj).

Because DLCs partition to the fat and are reported as concentra-tions in fat in several food categories, fat fraction estimates areused to estimate concentrations in whole foods. The fat fractionestimates used for beef (0.17), pork (0.19), and poultry (0.09) wereobtained from Lorber et al. (2009, 2010). Generally, these wereweighted average estimates determined by combining informationon amounts of specific meat product consumed (e.g., hamburger,steak, and other beef products) and information on the fat contentof that food product (fat content of hamburger and steak, etc.). Thefat fraction for other meats (0.15) was estimated as the average ofthe fat fractions for beef, pork, and poultry. The fat fraction esti-mate for milk (0.018) was obtained from information in EPA’sExposure Factors Handbook (USEPA, 1997) on the percentage ofmilk types (whole, low fat, and skim) consumed by people in com-bination with data from USDA (1995) on fat composition(whole = 3.2%; low fat = 1.3%, and skim = 0.7%). The fat fractionestimate for other dairy (0.12) similarly considered amounts ofkey dairy products consumed, such as cheese, yogurt, and icecream, along with the fat fractions of these products (USEPA,1997) [Lorber et al. (2009, 2010) provides additional details]. Foreggs and seafood, the DLC concentrations were reported on awhole weight basis; thus, fat fraction estimates were not needed.

2.4. TEQ intake estimates and sensitivity analysis

In Eq. (2), alternative daily TEQ intake rate estimates for the jthfood category (TEQAj, pg/day) were calculated as the sum across nDLCs of the products of the ith DLC-specific TEFAi value and the dai-ly oral intake (Eij, pg/day) from Eq. (1).

TEQAJ ¼Xn

i¼1

Eij � TEFAi ð2Þ

In Eq. (3), the alternative total daily TEQ intake estimates(TEQAT, pg/day) were calculated by summing the TEQAj [from Eq.(2)] across the m food categories.

TEQAT ¼Xm

j¼1

TEQAj ð3Þ

Eqs. (2) and (3) were used in the sensitivity analysis by applyingTEFA values to estimate TEQ intakes for (1) individual food catego-ries, (2) all food categories together, (3) DLC classes (i.e., dioxins,furans, and PCBs), and (4) individual DLCs. In Eqs. (2) and (3), theTEFA value being used was held constant throughout the summa-tions (e.g., the 25th percentile TEFA values from Approach2A wereused for all DLCs in those calculations).

2.5. Alternative TEF values

Table 1 summarizes the TEFA values for the DLCs used in thisanalysis. For Approach1, we estimated TEFA values by multiplyingand dividing the baseline TEF value for each DLC (TEFB) by 3.16(i.e., one half a log), denoted as TEFU (upper) and TEFL (lower) val-ues, respectively, while keeping the TEF for TCDD constant at 1. ForApproach2A, we based TEFA values on the percentile estimates(10th, 25th, 50th, 75th, and 90th) of the REP empirical ranges,based only on in vivo studies in the database for each DLC (Hawset al., 2006). Designated TEFA, here, subscript ‘A’ indicates the per-centile used (e.g., TEF10 is an alternative TEF value based on the10th percentile). For five DLCs [1,2,3,6,7,8-hexachlorodibenzo-p-

dioxin (1,2,3,6,7,8-HxCDD); 1,2,3,7,8,9-hexachlorodibenzofuran(1,2,3,7,8,9-HxCDF); 1,2,3,4,6,7,8-heptachlorodibenzofuran (1,2,3,4,6,7,8-HpCDF); 1,2,3,4,7,8,9-HpCDF; and PCB 81], there were noin vivo studies in the REP database, so the TEFB values derived fromin vitro studies were used as a placeholder for these DLCs through-out the sensitivity analyses. These in vitro assays were admitted tothe REP database because the measured responses were mediatedthrough the aryl hydrocarbon receptor activation, the mechanismof action in the in vivo DLC studies (Haws et al., 2006). Hawset al. (2006) also developed an REP empirical range based on com-bined in vivo and in vitro data (Table A.2 in SI); alternative TEQ in-take estimates based on this empirical range (i.e., Approach2B) arepresented in the SI (Table A.3 in SI).

3. Results

Using the oral DLC intake estimates (Table A.4 in SI) and TEFB

values (Table 1), we estimated the U.S. daily average TEQ dietaryintake to be approximately 23 pg/day based on DLC concentrationsin the eight food categories and average daily consumption rates(Table 2). Fig. 1(a) shows that PCB 126 (26%); 1,2,3,7,8-penta-chlorodibenzo-p-dioxin (1,2,3,7,8-PeCDD) (23%); and 1,2,3,6,7,8-HxCDD (14%) account for 63% of this estimated total TEQ intake.The inset pie chart [Fig. 1(b)] shows that consumption of foodsfrom the following food categories account for 85% of this TEQ in-take estimate: other dairy (28%), beef (25%), seafood (18%), andmilk (14%).

Using Approach1, Table 2 shows the total TEQ intake estimatesrange from 8 to 68 pg/day between the TEFL and TEFU values,respectively. Using Approach2A, the total TEQ intake ranges from8 to 106 pg/day at the TEF10 and TEF90 percentile values, respec-tively, and from 12 to 28 pg/day at the TEF25 and TEF75 percentilevalues, respectively.

The individual food category-specific total TEQ daily intake esti-mates using the TEFB values are always between those generatedusing TEF50 and TEF75 values using Approach2A (Fig. 2). This pat-tern is consistent with the conservative perspective WHO usedwhen developing the TEFB values. Using Approach2A, the TEQ in-take estimates gradually increase in all food categories betweenthe TEF10 and TEF75 percentile values, as expected; however, be-tween TEF75 and TEF90, TEQ intake estimates increase substantiallyfrom 28 to 106 pg/day. Generally, for all TEFA values, using eitherApproach1 or Approach2A, other dairy, beef, and seafood contrib-ute the most to the total daily TEQ intake estimates (Table 2).The only notable exception is the TEQ intake from milk exceedsthose from beef and seafood using TEF90 values due to the contri-bution of PCB 156 from milk (Table 2).

Using Approach1, the contribution of dioxins (53%, 58%, and51%), furans (18%, 16%, and 19%), and PCBs (29%, 26%, and 30%)to the average daily TEQ intake estimate are comparable relativeto each other when comparing the TEFB, TEFL, and TEFU values,respectively (Table 2). Using Approach2A, these relative contribu-tions change substantially across the percentiles. Using TEF10,TEF25, and TEF50 values, the dioxins are the primary contributorsto total TEQ intake estimates. In contrast, the percentage contribu-tion from the PCBs increases between TEF10 and TEF90; at TEF75 andTEF90, the PCBs are the primary contributors to the total TEQ intake(Table 2).

Regarding the individual DLCs, the same three (PCB 126;1,2,3,7,8-PeCDD; and 1,2,3,6,7,8-HxCDD) that account for most ofthe total TEQ intake using TEFB values are also the primary contrib-utors to total TEQ intake based on Approach2A for TEF25, TEF50, andTEF75 (Table 2). Recall that for 1,2,3,6,7,8-HxCDD, which had noin vivo data, we used the TEFB value throughout the analysis. UsingTEF90 values, the percentage contribution to total TEQ intake frommilk consumption replaces beef consumption among the top three

Table 2The percentage contributions to total TEQ daily intake at alternative TEF values based on Approach1 and Approach2A for different DLC groups, individual DLCs (Top 3), andindividual food categories (Top 3).

TEF empirical range TEQT (pg/day) TEQ contribution (%)a

PCDDs 65b PCDFs 145b PCBs 166b Top 3 DLCs Top 3 food categories

TEFL 8 58 16 26 PCB 126 (24) Other dairy (27)1,2,3,7,8-PeCDD (21) Beef (25)TCDD (16) Seafood (19)

TEF10 8 71 13 16 1,2,3,6,7,8-HxCDD (38) Other dairy (28)1,2,3,7,8-PeCDD (16) Beef (28)PCB 126 (15) Seafood (15)

TEF25 12 56 12 32 PCB 126 (30) Other dairy (28)1,2,3,6,7,8-HxCDD (26) Beef (25)1,2,3,7,8-PeCDD (11) Seafood (17)

TEF50 17 47 14 38 PCB 126 (35) Other dairy (29)1,2,3,6,7,8-HxCDD (18) Beef (24)1,2,3,7,8-PeCDD (12) Seafood (19)

TEFB 23 53 18 29 PCB 126 (26) Other dairy (28)1,2,3,7,8-PeCDD (23) Beef (25)1,2,3,6,7,8-HxCDD (14) Seafood (18)

TEF75 28 33 13 54 PCB 126 (42) Other dairy (29)1,2,3,6,7,8-HxCDD (11) Beef (21)1,2,3,7,8-PeCDD (11) Seafood (21)

TEFU 68 51 19 30 PCB 126 (28) Other dairy (29)1,2,3,7,8-PeCDD (24) Beef (25)1,2,3,6,7,8-HxCDD (14) Seafood (18)

TEF90 106 11 6 83 PCB 156 (44) Other dairy (47)PCB 126 (22) Milk (22)PCB 169 (7) Seafood (12)

a Percentage TEQ contribution of a DLC group = [(TEQ of a DLC group using TEFA values) � TEQAT] � 100; Percentage TEQ contribution of a DLC = [(TEQ of a DLC using TEFA

value) � TEQAT] � 100; Percentage TEQ contribution by food category = [(TEQAj for the jth food category) � TEQAT] � 100; see Eqs. (2) and (3).b Total number of studies for each DLC group.

282 S. Parvez et al. / Regulatory Toxicology and Pharmacology 67 (2013) 278–284

food categories. Additionally, at TEF90, other dairy and milk con-sumption accounts for approximately 50 and 24 pg/day, respec-tively, together comprising 70% of the 106 pg TEQ/day totalintakes from all eight food categories. PCB 156 alone contributes44% of this total. At baseline, PCB 156 contributes less than 1% toTEQ intake from milk and other dairy. PCB 156 has TEF values of0.00003, 0.0005, and 0.09 at the baseline, TEF75, and TEF90 values,respectively, spanning a range of 3 orders of magnitude, which sig-nificantly raises the total TEQ intake estimate at TEF90 (Tables 1and 2).

A comparison of the total TEQ intake estimates using TEFL fromApproach1 and TEF10 from Approach2A shows these total TEQ esti-mates to be the same (TEQ = 8 pg/day); however, the contributionsof individual DLCs to these totals differ. For example, PCB 126(24%) is the primary contributor to the total TEQ intake estimatebased on TEQL, and 1,2,3,6,7,8-HxCDD (38%) is the main contribu-tor at TEF10 (Table 2). In contrast, a comparison between total TEQintake estimates using TEFU from Approach1 (TEQ = 68 pg/day) andTEF90 from Approach2A (TEQ = 106 pg/day) differ markedly; theindividual DLC contributions also differ. PCB 126 (28%); 1,2,3,7,8-PeCDD (24%); and 1,2,3,6,7,8-HxCDD (14%) remain the primarycontributors using TEFU, but at TEF90, PCB 156 (44%), PCB 126(22%), and PCB 169 (7%), are the top three contributors to totalTEQ intake. Finally, the total TEQ intake estimates based on TEF25

and TEF75 values are 12 and 28 pg/day, respectively; these arecomparable to the TEFL but not the TEFU total TEQ intake estimates(8 and 68 pg/day, respectively).

The SI compares Approaches 2A and 2B (Fig. A.2, Table A.3, andTables A.5–A.8 in SI). Total daily TEQ intake estimates using Ap-proach2B are comparable to the estimates based on Approach2Aat the TEF10, TEF25, and TEF50 percentiles, although the percentagecontributions from the various DLC classes differ. The estimates atthe TEF75 and TEF90 percentiles differ substantially, as the TEQ in-take estimates from Approach2B are 50% and 100% larger, respec-tively, than those estimated using Approach2A.

4. Discussion

Our estimated U.S. total daily TEQ intake estimate (23 pg/day)from 8 food categories is approximately 7 pg/day lower than thatpredicted by Lorber et al. (2010) using the same food categories.Relative to their previous estimate, this decrease is due primarilyto our use of updated DLC concentration estimates in seafood (Bol-ger, P.M., Murray, C., III, Concentrations of dioxin-like compoundsin commercial seafood and eggs, 2011, Unpublished results.),which were lower than those used by Lorber et al. Our inclusionof five more PCBs (PCB 81, PCB 105, PCB 118, PCB 156 and PCB157) than Lorber et al. (2010) included in their analysis did not sig-nificantly change the TEQ intake estimates, because the intakerates and TEF values for these PCBs are relatively small. Our exclu-sion of other exposure routes that were included in Lorber et al.(2009) would be unlikely to substantially increase the TEQ intakeestimates for the general population; Lorber et al. (2009) estimatedthese intakes to total 3.7 pg TEQ/day. If we had assumed that non-detected congeners were not present, our TEQ estimate wouldhave been minimally reduced.

Sensitivity analyses based on Approach1 and Approaches 2Aand 2B provide a general characterization of the uncertainty inthe TEQ estimates across alternative TEF values. Approach1 relieson an overall assessment of confidence in TEF values based onthe informed opinions of expert panelists familiar with the data-base of relevant toxicology studies. The expert panelists’ assess-ment of uncertainty reflected their evaluation of the quality ofstudies across the entire set of DLC studies, and our implementa-tion resulted in a uniform range of uncertainty for all TEFs.

Approaches 2A and 2B are based on empirical data and encom-pass all studies for each DLC that met the study authors’ inclusioncriteria (Haws et al., 2006). These approaches assume that the ob-served variation is an appropriate surrogate for uncertainty in aTEF value. There are many inconsistencies across studies for indi-

Fig. 1. (a) Percent Contribution of Individual Dioxin-like Compounds to the estimated U.S. mean daily

average 2,3,7,8--tetrachlorodibenzo-p-dioxin toxicity equivalence (TEQ) dietary intake. (b) Contributions of

the Individual Food Categories to the estimated U.S. mean daily average 2,3,7,8--tetrachlorodibenzo-p-dioxin

toxicity equivalence (TEQ) dietary intake.

PCBs (29%)(Polychlorinated biphenyls)

PCDFs (18%)(Polychlorinated dibenzofurans)

PCDDs (53%)(Polychlorinated dibenzo-p-

dioxins)

pg/day

Fig. 1. (a) Percent contribution of individual dioxin-like compounds to the estimated U.S. mean daily average 2,3,7,8–tetrachlorodibenzo-p-dioxin toxicity equivalence (TEQ)dietary intake. (b) Contributions of the Individual Food Categories to the estimated U.S. mean daily average 2,3,7,8-tetrachlorodibenzo-p-dioxin toxicity equivalence (TEQ)dietary intake.

0

1

10

Pork Poultry Other meat

Eggs Milk Beef Seafood Other dairy

TEQ

dai

ly in

take

[Log

(10)

pg/

day]

TEFL TEF10 TEF25 TEF50 TEFB TEF75 TEFU TEF90

Fig. 2. Total TEQ daily intake estimates by food category using Alternative Valuesfor the Toxicity Equivalency Factors (TEF) based on Approach1 and Approach2A. Forthe individual dioxin-like compounds, TEFL and TEFU were estimated usingApproach1. Approach1 estimates upper and lower TEFs for each dioxin-likecompound by multiplying and dividing, respectively, its individual TEF by ± half alog. Approach2A estimates alternative TEF values based on compiled empiricalranges of relative potency estimates for each dioxin-like compound.

S. Parvez et al. / Regulatory Toxicology and Pharmacology 67 (2013) 278–284 283

vidual DLCs, and the percentiles identified through Approaches 2Aand 2B are not directly comparable across DLCs. In addition, thereis no weighting factor or other method applied to distinguishamong REPs derived from studies of different types or quality lev-els. Haws et al. (2006) indicate that they are developing weightingfactors for the REP values; if these lead to the development of con-sensus TEF distributions for individual DLCs, these could signifi-cantly improve the conduct of sensitivity analyses around TEQestimates.

We have limited confidence in the use of extreme values fromthe empirical ranges in sensitivity analyses. Previously, EPA (USEPA,2010) cautioned that extreme total TEQ intake estimates are sumsof multiple extreme values, representing unlikely combinations ofthe DLCs’ TEF values and likely yielding overestimates of the higheralternative TEF values and underestimates of the lower alternativeTEF values. Specifically, for the TEF90 values we used, the relativecontributions of the DLCs and of the food categories to total TEQ in-take result in a much different pattern than for other alternativeand baseline TEF values. This suggests that the choice of the moreextreme TEF values in the empirical range may be less reasonableas alternatives, and their resulting TEQ intakes and influence pat-terns require close examination. Recall that the DLCs and food cat-egories contributing the most to the total TEQ intake estimates areconsistent across all the alternative TEFs, with the exception of thelower and upper extremes (i.e., TEFL and TEF90). Further, none ofthese approaches should be interpreted as yielding confidence lim-its for the total TEQ intake.

On the other hand, the interquartile REP values reported byHaws et al. (2006) may yield appropriate alternative values foruse in sensitivity analyses, and they may provide crude estimates

284 S. Parvez et al. / Regulatory Toxicology and Pharmacology 67 (2013) 278–284

of plausible TEQ intake ranges. The SI compares Approaches 2A and2B. In general, in vivo studies are preferred over in vitro studies forderiving TEF values because they account for DLC toxicokinetics(Haws et al., 2011; Van den Berg et al., 2006). However, there isan informational trade-off in making this choice because, for manyDLCs, there are relatively few in vivo studies.

This analysis highlights important toxicological researchneeded for several DLCs. Perhaps the most critical is the undertak-ing of additional in vivo toxicological studies comparing the poten-cies of 1,2,3,6,7,8-HxCDD and TCDD. Intakes of 1,2,3,6,7,8-HxCDDaccount for 14% of the total TEQ intake estimate (Fig. 1). Only fivein vitro studies and no in vivo studies for this DLC met the inclu-sion criteria of Haws et al. (2006), suggesting increased uncertaintyin its contribution to the TEQ intake estimate that could be de-creased with additional comparative in vivo studies. While for sev-eral PCBs there are also relatively few studies that met theirinclusion criteria (e.g., there are no in vivo studies and 12 in vitrostudies for PCB 81), conducting additional comparative toxicologystudies is not likely warranted from the perspective of our analysis,as they would not be expected to dramatically change the TEQ in-take estimates. PCB 81 contributes less than 1% to the total TEQ in-take using baseline TEF values (Fig. 1). Importantly, relative to thecontributions of the other DLCs, our analysis may underestimatethe potential contribution of TCDD to total TEQ intake estimates.Recall that the WHO panelists assigned TEFs for individual DLCsthat were generally in the range of the 50th to the 75th percentilesof the empirical data range to be health protective (Van den Berget al., 2006). As this biases the estimates of the other DLCs high rel-ative to TCDD, the TCDD contribution is likely underestimated.

Finally, this analysis highlights the importance of developingprobability distributions for the TEFs. Continued efforts are neededto increase quality data on selected DLCs and to develop new ap-proaches for deriving representative TEF values for the DLCs. Hawset al. (2011) proposed weighting criteria for improved determina-tion of TEFs, and Van den Berg et al. (2006) suggested the use oflowest-observed-adverse-effect levels to estimate REPs, if dose-re-sponse relationships could not be confidently predicted. Such ef-forts likely would improve the scientific basis of the TEFestimates and could be sources of meaningful distributions of theTEFs for the DLCs. Given the resources needed to develop ‘‘com-plete’’ data sets for each DLC, it is likely that, in the near term, suchdistributions will need to be informed by expert judgment, perhapsthrough a consensus panel. Our analyses systematically varied allTEF values to estimate TEQ intake rate ranges. Other approachessuch as varying TEF values only for several selected DLCs could alsobe employed to examine the sensitivity of TEQ intake rate esti-mates to such changes.

Disclaimer

The views expressed in this article are those of the authors anddo not necessarily reflect the views or policies of the EPA.

Funding source

The authors declare no competing financial interest.

Conflict of interest

The authors declare that there are no conflicts of interest.

Acknowledgments

We thank P. Michael Bolger, Ph.D., D.A.B.T. and Clarence MurrayIII, Ph.D. for providing concentration data on dioxin-like com-pounds in U.S. seafood and eggs. We thank Drs. Richard C. Hertz-berg, Cynthia V. Rider, Harlal Choudhury, and Michael J. DeVitofor their helpful comments on this manuscript.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.yrtph.2013.08.007.

References

Haws, L.C., Su, S.H., Harris, M., Devito, M.J., Walker, N.J., Farland, W.H., Finley, B.,Birnbaum, L.S., 2006. Development of a refined database of mammalian relativepotency estimates for dioxin-like compounds. Toxicol. Sci. 89, 4–30.

Haws, L. C., DeVito, M. J., Walker, N. J., Harris, M. A., Tachovsky, J. A., Brinbaum, L. S.,Farland, W. H., Wikoff, D. S., 2011. Development of a consensus-basedweighting framework for evaluating estimates of relative potency for dioxin-like compounds that includes consideration of data from human cells. In:Dioxin 2011. Brussels, Belgium.

Lorber, M., Patterson, D., Huwe, J., Kahn, H., 2009. Evaluation of backgroundexposures of Americans to dioxin-like compounds in the 1990s and the 2000s.Chemosphere 77, 640–651.

Lorber, M., Huwe, J., Rawn, D.F.K., 2010. An update to estimates of intake of dioxin-like compounds for the general population of the United States. OrganohalogenCompd. 72, 154–155.

USDA, 1995. Food and Nutrient Intakes by Individuals in the United States, 1 day,1989–1991. Agricultural Research Service, Washington, DC.

USEPA, 1997. Exposure Factors Handbook. Office of Research and Development,Washington, DC.

USEPA, 2000. Supplementary guidance for conducting health risk assessment ofchemical mixtures. Risk Assessment Forum, Washington, DC.

USEPA, 2003. Chapter 9. Toxic equivalency factors (TEF) for dioxin and relatedcompounds. In: Exposure and human health of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and related compounds. Part II: health assessment for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and related compounds. Office of Researchand Development, National Center for Environmental Assessment, Washington,DC.

USEPA, 2010. Recommended Toxicity Equivalence Factors (TEFs) for Human RiskAssessments of 2,3,7,8-tetrachlorodibenzo-p-dioxin and Dioxin-likeCompounds. Risk Assessment Forum, Washington, DC.

USFDA, 2007. Dioxin Analysis Results/Exposure Estimates. U.S. Department ofHealth and Human Services, Silver Spring, MD.

Van den Berg, M., Birnbaum, L., Denison, M., Devito, M.J., Farland, W.H., Feeley, M.,Fiedler, H., et al., 2006. The 2005 World Health Organization re-evaluation ofhuman and mammalian toxic equivalency factors for dioxins and dioxin-likecompounds. Toxicol. Sci. 93, 223–241.