TOXICOLOGICAL SCIENCES 91(1), 275–285 (2006)
doi:10.1093/toxsci/kfj125
Advance Access publication February 2, 2006
Recommended Relative Potency Factors for
2,3,4,7,8-Pentachlorodibenzofuran: The Impact of
Different Dose Metrics
Robert A. Budinsky,*,1 Dennis Paustenbach,† Donald Fontaine,* Bryce Landenberger,* and Thomas B. Starr‡
*The Dow Chemical Company, Midland, Michigan 48674; †ChemRisk, Inc., San Francisco, California 94105; and ‡TBS Associates,
Raleigh North Carolina 27615–3700
Received April 6, 2005; accepted January 18, 2006
The recent National Toxicology Program (NTP) cancer bioassays for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and
2,3,4,7,8-pentachlorodibenzofuran (4-PeCDF) permit a reevaluation of the current TEF value of 4-PeCDF. The data also allow for
the derivation of relative potency factors (RPFs) for cancer, which
are based not only on administered dose but also on potentially
more informative dose metrics, such as liver concentration, area
under the liver concentration curve, and lifetime average body
burden. Our analyses of these data indicate that chi-squared tests
of observed versus predicted liver tumor incidence for 4-PeCDF
reject the current TEF value of 0.5 value as too high. 4-PeCDF
RPFs were derived using estimation methods that either did or did
not assume parallelism of the 4-PeCDF and TCDD dose–response
curves. The resulting parallelism-based RPFs for administered
dose, liver concentration at terminal sacrifice, liver concentration
AUC, and lifetime average body burden are 0.26, 0.014, 0.021,
and 0.036, respectively. The administered dose RPF estimate is
approximately one-half the current TEF value of 0.5. However,
the use of administered dose fails to take into account pharmacokinetic differences between congeners and the generally acknowledged belief that body burden or some other measure of
cumulative dose is more appropriate for estimating the health risk
posed by persistent chemicals. The other three dose metrics do
account for these important factors, and the corresponding RPFs
are at least 10-fold lower than the current TEF for 4-PeCDF. In
summary, our analyses support an administered dose TEF no
greater than 0.25 and one in the 0.05–0.1 range for internal dose
metrics such as lifetime average liver concentration or body burden.
INTRODUCTION
2,3,4,7,8-Pentachlorodibenzofuran (4-PeCDF) is an important dioxin-like compound for a number of reasons. It has
a relatively high toxic equivalency factor (TEF) of 0.5 (Van den
Berg et al., 1998), and it contributes substantially to the toxic
1
To whom correspondence should be addressed at The Dow Chemical
Company, 1803 Building, Midland, MI 48674; Fax: 989-638-9863. E-mail:
[email protected].
equivalents (TEQ) found in many environmental and biomonitoring samples (Kreuzer et al., 1997; NHANES, 2005;
Patterson et al., 1994; Petreas et al., 2000; USEPA 2003a;
Wittsiepe et al., 2000, 2004). It also exhibits a high degree
of dose-dependent liver sequestration because of its ability
to induce as well as bind to CYP1A2 (Brewster and Birnbaum,
1987; NTP, 2003a; Van den Berg et al., 1989), and it is one of
the few dioxins for which 2-year carcinogenicity bioassay
data are available (NTP, 2003a). Because of the significant role
that 4-PeCDF can play in environmental remediation initiatives, an accurate estimate of its carcinogenic potency relative
to that of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) is of
considerable value.
Toxic equivalency factors for dioxin-like compounds were
first introduced in the late 1980s. Pluss et al. (1988) recommended a potency of 0.4 for 4-PeCDF compared to TCDD
based on clinical findings in a 13-week dietary rat study.
Relative potency factors (RPFs) ranging from 0.05 to 0.5
(in vivo endpoints) and 0.7 to 0.9 (in vitro endpoints) were
suggested for 4-PeCDF by Olson et al. (1989). Safe’s (1998)
summary of 4-PeCDF RPFs from a number of earlier in vivo
and in vitro toxicity studies ranged from 0.12 to 0.8 and 0.11
to 0.67, respectively, based on acute lethality, body weight loss,
thymic atrophy, immunotoxicity, and teratogenicity endpoints.
The current WHO TEF value of 0.5 is based on either 73 (Finley
et al., 2003) or 60 (USEPA, 2003a) RPFs, all for non-cancer
endpoints. The U.S. Environmental Protection Agency (EPA)
Draft Dioxin Reassessment identified enzyme induction, organ
and body weight changes, immune changes, developmental
toxicity, thyroid hormone changes, hepatic retinoid changes,
hepatic porphyria, and tumor promotion dose–response data
sets as supporting a TEF of 0.5 for 4-PeCDF (USEPA, 2003a).
Before completion of the recent 4-PeCDF cancer bioassay
(NTP, 2003a), the Waern et al. (1991) initiation-promotion
study came the closest to providing data that are directly relevant to cancer risk assessment. Waern et al. (1991) estimated
RPFs of 0.1, based on administered dose, and 0.007, based on
liver concentration. In contrast, the Harper et al. (1993) plaqueforming immunotoxicity study yielded a much higher range,
The Author 2006. Published by Oxford University Press on behalf of the Society of Toxicology. All rights reserved.
For Permissions, please email: [email protected]
276
BUDINSKY ET AL.
with RPFs varying from 0.58 to 4.0. Finley et al. (2003)
evaluated the World Health Organization (WHO) RPF database with a weighting scheme based on relevance to human
cancer risk assessment. They obtained a weighted mean TEF
for 4-PeCDF of 0.27. Haws et al. (2005) have recently updated
this database. The minimum RPF estimate from in vivo data
was 0.0065, with a maximum of 3.7 and a 50th percentile RPF
estimate of 0.2. Based on the 101 available in vitro and in vivo
RPFs, Haws and co-workers derived an average RPF value of
0.22. All of these estimates are for application with administered doses.
The National Toxicology Program (NTP) recently reported
findings from its cancer bioassays of TCDD, 4-PeCDF, PCB126,
and a 1:1:1 mixture (based on current WHO TEF values) of
all three compounds (first introduced by van Birgelen et al.,
1997; NTP, 2003a, 2003b, 2003c, and 2003d). The TCDD
and the TEF-adjusted doses for 4-PeCDF and PCB126 were
selected using tumor incidence rates from the original TCDD
bioassay conducted by Kociba et al. (1978). These NTP studies
provide the first opportunity (1) to evaluate the hypotheses of
parallel dose responses and dose additivity upon which the
validity of the TEF concept relies and (2) to derive RPF
sampling distributions for 4-PeCDF based on carcinogenicity.
The liver and adipose tissue concentration data that were
collected for TCDD and 4-PeCDF during the course of these
bioassays also allow for assessments of relative carcinogenic
potency based on dose metrics other than administered dose.
For example, an RPF appropriate for use with lifetime average
body burden would be consistent with the EPA dose–response
modeling for cancer and non-cancer effects and for extrapolation of estimated risks from rodents to humans.
The parallel dose–response and dose-additivity assumptions
underlying TEFs have recently been evaluated from data
gathered during the NTP bioassays (Toyoshiba et al., 2004;
Walker et al., 2005). Toyoshiba et al. (2004) used Hill model
dose–response relationships to assess the parallelism of the
congener-specific and mixture dose–response curves, dose
additivity, as well as to derive RPF estimates for 4-PeCDF
and PCB126 from data for CYP1A1 and CYP1A2 activity.
Interestingly, the assumption of parallelism was unilaterally
rejected for CYP1A1 and CYP1A2 at all measured time points
(14, 31, and 53 weeks on study). The hypothesis of dose additivity was also rejected except for CYP1A1 at 14 weeks and
possibly for CYP1A2 at 53 weeks ( p ¼ 0.065). Relative potency factor estimates for 4-PeCDF from the full (unconstrained) dose–response models for CYP1A1 ranged from 0.04
to 0.19, whereas those for CYP1A2 ranged from 0.07 to 0.25.
Walker et al. (2005) investigated parallelism, dose additivity,
and the adequacy of the WHO TEF value of 0.5 as an RPF
estimate for 4-PeCDF using the carcinogenicity data from the
NTP bioassays. There was no significant evidence of nonparallelism, but significant departures from dose additivity
(presuming parallel dose responses) were obtained for two of
the four elevated tumor types: cystic keratinizing epithelioma
(CKE) of the lung ( p ¼ 0.033) and gingival squamous cell
carcinoma ( p ¼ 0.047). It was noted that the statistical power to
detect significant departures from these fundamental assumptions was low, with the quantal tumor data ranging across the
four tumor endpoints from 0.1 to 0.50.
Using similar Hill model analyses of the bioassay tumor
data, Walker et al. (2005) also estimated lower RPFs for
4-PeCDF than the current TEF value of 0.5. For cholangiocarcinoma and hepatocellular adenoma (HA) of the liver,
CKE of the lung, and gingival squamous cell carcinoma, their
‘‘optimal’’ potency estimates for 4-PeCDF were 0.16, 0.34,
0.30, and 0.26, respectively. Although only the RPFs for
cholangiocarcinoma of the liver and CKE of the lung were
significantly smaller than the WHO TEF of 0.5 (p < 0.0001),
the authors called attention to the consistency between their
lower potency estimates and those reported by Waern et al.
(1991). Further, they suggested that ‘‘the current TEF value for
4-PeCDF ought to be reevaluated for its application in
quantitative cancer risk assessments.’’
Other preliminary RPF estimates for 4-PeCDF derived specifically for internal dose metrics have been reported (Budinsky
et al., 2004; Fontaine et al., 2004). These estimates, appropriate
for use with total body burden and liver concentration, are an
order of magnitude lower than either the current WHO TEF or
the administered dose RPF estimates reported by Walker et al.
(2005) for 4-PeCDF.
In the present article we expand on the preliminary RPF
estimates reported by Budinsky et al. (2004) and Fontaine
et al. (2004). Making use of a simple statistical goodness-of-fit
test, we first determine whether or not the current TEF value
of 0.5 accurately predicts liver and lung tumor responses to
4-PeCDF exposure using the observed dose responses for
these endpoints in the TCDD bioassay. Then our analyses generate separate and distinct RPFs appropriate for use with the
following dose metrics: daily administered dose per unit of
body weight, time-averaged body burden, liver concentration,
or area under the liver concentration curve (AUC), with the
latter two dose metrics determined at terminal sacrifice. We
also derive empirical sampling distributions for the estimated
RPFs using Monte Carlo simulation.
MATERIALS AND METHODS
Tumor incidence. Tumor incidence data from the NTP bioassay studies of
TCDD and 4-PeCDF (NTP, 2003a and 2003b) were used in dose–response
modeling and in evaluating the adequacy of 0.5 as a RPF in predicting the
observed 4-PeCDF tumor response from the modeled TCDD tumor response.
The data were also used to develop the sampling distributions of 4-PeCDF
RPFs appropriate for use with different internal dose metrics. The Weibull
dose–response model (with independent background) was fit to the TCDD
data for liver tumors or combined liver or lung tumors as a function of (1)
administered dose (ng/kg per day),(2) liver concentration at terminal sacrifice
(pg/gm), (3) area under the liver concentration curve at terminal sacrifice
(AUC, pg/g 3 weeks), or (4) lifetime time average body concentration (pg/g)
using the EPA Benchmark Dose software (version 1.3.2) (USEPA, 2001).
277
RELATIVE POTENCY FACTORS FOR 4-PeCDF
Animals with multiple tumors were counted only once in incidence numerators,
and incidence denominators were adjusted for the competing risks of intercurrent mortality using the poly-3 procedure (Bailer and Portier, 1988). For
purposes of this analysis, liver tumors included both hepatocellular adenomas
and cholangiolar carcinomas; lung tumors were predominantly cystic keratinizing epitheliomas, with occasional alveolar or bronchiolar adenomas. We
chose to analyze combined liver or lung tumors in addition to liver tumors
alone because the USEPA had used this combination in developing its original
carcinogenic potency factor for TCDD of 156,000 (mg/kg/day)1 (USEPA,
1988).
Dose metrics. A summary of the specific dose metric values that were
utilized in the dose-response modeling is provided in Table 1. Liver concentrations of TCDD and 4-PeCDF at terminal sacrifice were taken directly from
Tables 13 (TR 521) and 8 (TR 525) in the Board Draft NTP Technical Reports
for these substances (NTP, 2003a, 2003b). Liver AUC at the terminal sacrifice
was generated from the liver concentrations reported at weeks 14, 31, 53, and
104 using simple linear interpolation, the trapezoidal rule, and the assumption
that the liver concentration at study onset, and other times it was reported to be
below the limit of quantification (LOQ), equaled 1/2 LOQ. The presumed
baseline AUC value of 104 * (1/2 LOQ) was then subtracted from each treatment group AUC at terminal sacrifice. To illustrate, in the 3 ng/kg/day TCDD
dose group:
Liver AUC @ 104 weeks
¼ 1=2½ð14 0Þ * ð1=2 * 50 þ 676:2Þ þ ð31 14Þ
* ð676:2 þ 598:9Þ þ ð53 31Þ * ð598:90 þ 499:38Þ
þ ð104 53Þ * ð499:38 þ 680:71Þ 104 *ð1=2 * 50Þ
¼ 55320:13 ðpg=gÞ * weeks:
To obtain lifetime average body concentrations, weekly body concentrations
during the course of the bioassays were first estimated with a simple pharmacokinetic model (Edmund Crouch, Cambridge Environmental, personal
communication). Briefly, TCDD or 4-PeCDF body burden (the product of body
concentration Cb and body weight, w) was assumed to reside entirely in liver and
adipose tissue and have a time evolution governed by the differential equation:
dðCb * wÞ=dt ¼ ðg k * f * Cb Þ * w;
ð1Þ
where g is the product of the administered dose and an oral gavage absorption
fraction (assumed equal to 0.9), k is a constant metabolic elimination rate, and
f is the fraction of the total body burden present in liver. Following Carrier et al.
(1995), the liver fraction f was presumed to have a minimum value of 0.01 and
Michaelis-Menten dependence upon body concentration:
f ¼ 0:01 þ ðfmax 0:01Þ * Cb =ðCb þ Km Þ:
ð2Þ
To obtain estimates of the parameters fmax and Km in Equation 2, mean
adipose and liver tissue concentrations (Ca and Cl) reported for the various dose
groups at 14, 31, and 53 weeks were first combined with corresponding mean
body weights, mean liver weights, and estimated adipose tissue weights (w, wl,
and wa) to generate estimates of the body concentration Cb and liver fraction
f at each of the above-noted time points for each nonzero dose group:
Cb ¼ ðCa * wa þ Cl * wl Þ=w;
ð3Þ
f ¼ ðCl * wl Þ=ðCb * wÞ:
ð4Þ
Adipose tissue weight (wa) was not measured during the bioassays, so it was
estimated from total body weight using a relationship reported by Brown
et al. (1997):
wa =w ¼ ð1:664 þ 0:0199 * wÞ=100:
ð5Þ
The parameters fmax and Km in Equation 2 were then optimized by minimizing
the sum of squared differences between the right hand sides of Equations 2
and 4.
To obtain an estimate of the metabolic elimination parameter k, Equation 1
was integrated numerically using a 1-week time step, reported body weights
(linearly interpolated as necessary to obtain weekly values), and interatively
estimated values of k:
ðCb * wÞtþ1 ¼ g * wt þ ðCb * wÞt * ð1 k * ft Þ:
ð6Þ
The predicted body burdens were then divided by body weight to provide
predicted body concentrations for comparison with those given by Equation 3
using the data reported at 14, 31, and 53 weeks. Assuming that the prediction
errors were log-normally distributed, the parameter k was optimized by
minimizing the sum of squared differences between the logarithms of the
body concentrations predicted by Equation 6 and those given by Equation 3.
With k optimized, the predicted TCDD and 4-PeCDF body concentrations were
averaged over 104 weeks to obtain lifetime average body concentrations
suitable for use in the dose–response modeling.
Relative potency hypothesis tests, RPF estimates, and RPF sampling
distributions. The predicted numbers of tumor-bearing animals for each
4-PeCDF bioassay dose group were generated by (1) applying the current TEF
of 0.5 to the 4-PeCDF dose metrics to obtain TCDD-equivalent dose metrics;
(2) using the fitted dose–response model for TCDD to estimate the probability
of being tumor-bearing at each 4-PeCDF dose; and (3) multiplying each such
probability by the poly-3 adjusted number of animals at risk in the corresponding 4-PeCDF dose group. Under the null hypothesis that the true relative
potency (RPF) for 4-PeCDF equals 0.5, the sum of [(Observed – Predicted)2/
Predicted] over the different dose groups should be approximately distributed
as a chi-squared random variate with degrees of freedom equal to the number of
included dose groups.
Two different methods were used to estimate RPFs and their sampling
distributions from the 4-PeCDF and TCDD tumor data. The first method
employed separate, possibly nonparallel Weibull model fits to each of the
TCDD and 4-PeCDF tumor data sets. One thousand synthetic data sets for each
compound were generated using the observed poly-3-adjusted numbers of
TABLE 1
Dose Metrics for TCDD and 4-PeCDF Used in Dose–Response Modeling and Relative Potency Factor (RPF) Estimation
Administered dose
5 days/week ng/kg/day
TCDD
0
3
10
22
46
100
Liver concentration at
terminal sacrifice pg/g
Liver AUC at terminal
sacrifice (pg/g) * weeks
Body concentration
lifetime average pg/g
4-PeCDF
TCDD
4-PeCDF
TCDD
4-PeCDF
TCDD
4-PeCDF
0
6
20
44
92
200
25
681
2,213
4,364
6,413
9,325
1,615
17,615
62,457
125,110
263,557
499,565
8,176
55,320
203,100
434,792
786,705
1,507,044
36,401
1,339,162
4,589,442
10,136,372
21,445,272
41,065,441
0
44
123
255
518
1,110
0
491
1,610
3,570
7,410
16,300
278
BUDINSKY ET AL.
animals at risk and the observed binomial probabilities of response at each
dose. Weibull model parameters specific to each simulated data set were then
obtained with the EPA Benchmark Dose software (USEPA, 2001). Then TCDD
and 4-PeCDF doses were determined that gave extra risks of 5%, 10%,
25%, 50%, 75%, 90%, and 95% (referred to as EDx). Empirical sampling
distributions for each RPFx (defined as the ratio of the TCDD EDx to the
corresponding 4-PeCDF EDx) were obtained either by direct calculation of all
1 million possible ratios or from a randomly drawn subset of 1000 ratios.
The second estimation method employed fitting the TCDD and 4-PeCDF
tumor incidence data simultaneously, assuming parallel dose responses. A
single Weibull model was fit to both data sets by maximum likelihood with the
RPF for 4-PeCDF as an additional adjustable parameter (estimation routines
kindly provided by Edmund Crouch, Cambridge Environmental). As with the
first method, 1000 synthetic data sets for TCDD and 4-PeCDF tumor incidence
were generated using the poly-3 adjusted numbers of animals at risk and the
experimental binomial probabilities of response at each dose. The simultaneous
estimation procedure was then implemented with each simulated data set in
combination with each dose metric. Finally, the empirical sampling distributions for the 4-PeCDF RPF were constructed from the results, and their means,
variances, and other characteristics were determined.
TABLE 2
Observed and Predicted Numbers of Liver Tumor-Bearing
Animals in the 4-PeCDF Bioassay Using Independent Weibull
Models Fit to the TCDD Liver Tumor Data versus Administered
Dose (AD), Liver Concentration at Terminal Sacrifice (Cl), Area
under the Liver Concentration Curve at Terminal Sacrifice (AUCl),
or Lifetime Average Body Concentration (Cb), with an assumed
RPF of 0.5 for 4-PeCDF
4-PeCDF
No.
dose
observed/
group
no. at riska
0
6
20
44
92
200
Liver Tumor Incidence Data and Modeled Dose Responses
Figure 1 displays the poly-3-adjusted liver tumor incidence
for the TCDD and 4-PeCDF studies along with the independent
Weibull model fits to each data set. There is only a relatively
modest increase in liver tumor incidence (~16%) for 4-PeCDF
even at the highest dose tested, and this is consistent with
NTP’s characterization of the 4-PeCDF findings as providing
only ‘‘some evidence of carcinogenicity’’ versus the ‘‘clear
evidence of carcinogenicity’’ in the TCDD bioassay. It is also
readily apparent that 4-PeCDF is less than half as potent as
TCDD on the administered dose scale.
Tests of the Adequacy of 0.5 as an RPF for 4-PeCDF
Tables 2 and 3 provide the results of chi-squared tests of the
null hypothesis that an RPF of 0.5 for 4-PeCDF is consistent
NTP Bioassay Liver Tumor Responses
0.7
# with Tumor / # at Risk
AD
Cl
AUCl
0.
0.005
0.1
0.8
4.7
23.4
0.0001
16.4
36.7
37.9
36.1
37.2
0.00007
3.1
35.6
37.9
36.1
37.2
Cb
0.
0.8
12.9
36.7
36.1
37.2
129.2 (5)
103.0 (5)
143.6 (5)
Chi-squared 20.9 (5)
(df)b p value 0.00084 3.1 3 1029 3.5 3 1026 1.3 3 1020
RESULTS
0.6
1/42.0
0/38.1
1/36.7
1/37.9
3/36.1
6/37.2
Number predicted with
liver tumors
a
Number at risk is the poly-3 adjusted number of animals at risk of
developing liver tumors.
b
Because predicted numbers of tumor-bearing animals in the control group
were either exactly zero or extremely small, only the five nonzero dose groups
were included in the chi-squared computations.
with liver or the combined liver and lung tumor incidences
observed in the TCDD and 4-PeCDF bioassays. For each of
these tumor endpoints, four tests of the null hypothesis were
undertaken (one for each dose metric), and in every case the
TABLE 3
Observed and Predicted Numbers of Liver or Lung
Tumor-Bearing Animals in the 4-PeCDF Bioassay Using
Independent Weibull Models Fit to the TCDD Liver or Lung
Tumor Data versus Administered Dose (AD), Liver Concentration
at Terminal Sacrifice (Cl), Area under the Liver Concentration
Curve at Terminal Sacrifice (AUCl), or Lifetime Average Body
Concentration (Cb), with an Assumed RPF of 0.5 for 4-PeCDF
TCDD
0.5
4-PeCDF
No.
dose
observed/
group
no. at riska
0.4
0.3
0.2
0
6
20
44
92
200
4-PeCDF
0.1
0
0
50
100
150
AD
0.4
0.4
0.4
0.9
4.6
26.0
Cl
0.4
17.4
36.7
37.9
36.1
38.1
Chi-squared 15.2 (6)
143.2 (6)
(df) p value 0.019 2.1 3 1028
200
Administered Dose (ng/kg/day)
FIG. 1. Separate Weibull model fits of the TCDD and 4-PeCDF liver tumor
incidence data.
1/42.0
0/38.1
1/36.7
1/37.9
3/36.1
8/38.1
Number predicted with
liver or lung tumors
a
AUCl
0.4
3.0
36.3
37.9
36.1
38.1
128.5 (6)
2.7 3 1025
Cb
0.4
0.9
13.4
37.4
36.1
38.1
103.0 (6)
5.9 3 10–20
Number at risk is the poly-3 adjusted number of animals at risk of
developing liver or lung tumors.
RELATIVE POTENCY FACTORS FOR 4-PeCDF
null hypothesis was rejected; i.e., there was no indication that
an RPF of 0.5 was consistent with the tumor data. The test
using liver tumor data and administered dose was highly
significant (p ¼ 0.00084), with the remaining tests yielding
even more highly significant differences and far smaller
p values (p < 1020), indicating that once the prominent
pharmacokinetic differences between TCDD and 4-PeCDF are
accounted for with an appropriate internal dose metric (e.g.,
AUCl), the tumor incidence rates predicted for 4-PeCDF with
an RPF of 0.5 are simply not compatible with the observed rates.
Figure 2 displays the differences between observed and
predicted numbers of liver tumor-bearing animals as a function
of the four different 4-PeCDF dose metrics. The large
discrepancies between observed and predicted incidence arise
in the dose range where there are significant treatment-related
effects for TCDD. In the lower dose range, where there are just
a few cases of liver tumors, the differences between observed
and predicted incidence are much smaller. For all dose metrics,
use of a 0.5 RPF value led to overprediction of tumor incidence
rates.
279
for administered dose, and 0.043 (0.015), 0.026 (0.013)
and 0.022 (0.020), respectively, for lifetime average body
concentration.
The dependence of the RPF estimates on the response probabilities at which they are estimated is a characteristic that
would not arise if the dose–response relationships for TCDDand 4-PeCDF-induced liver tumor incidence were parallel to
one another; i.e., parallel dose responses necessarily give the
same RPF value at all response probabilities. This incidence
RPF Sampling Distributions at Selected
Dose–Response Levels
When separate Weibull models were fitted to simulated data
sets for TCDD and 4-PeCDF, the estimated mean RPFs and
their sampling distributions varied with tumor incidence rate as
well as dose metric. Examples of the RPF distributions
obtained using administered dose and lifetime average body
concentration are shown Figure 3a and 3b, respectively. Three
sampling distributions are provided in each frame of the figure,
representing the RPFs obtained at predicted liver tumor
incidence rates of 10%, 50%, and 90%. The incidence ratedependent distributions have estimated means (standard deviations) of 0.30 (0.12), 0.19 (0.10), and 0.16 (0.15), respectively,
Assuming RPF = 0.5
40
# Observed - # Predicted
Administered Dose
30
Body Concentration
20
Liver AUC
Liver Concentration
10
0
-10
-20
-30
-40
-50
0
3
10
22
46
100
Dose, TEQ (ng/kg/day)
FIG. 2. Observed – Predicted numbers of animals with liver tumors using
four different dose metrics.
FIG. 3. a. The RPFx sampling distributions at 10%, 50%, and 90% liver
tumor incidence rates for 4-PeCDF based on administered dose. b. The RPFx
sampling distributions at 10%, 50%, and 90% liver tumor incidence rates for
4-PeCDF based on lifetime average body concentration.
280
BUDINSKY ET AL.
RPFx and RPFx/ Std Dev for Average Body Concentration
3.0
0.06
RPFx / Std Dev
0.05
Mean RPFx
imately 0.036 (0.011) for the lifetime average body concentration dose-metric.
3.5
0.04
RPF Distributions Assuming Parallel Dose–Response Curves
2.5
2.0
RPFx
0.03
1.5
0.02
1.0
0.01
0.5
0.00
RPFx / Std Dev
0.07
0.0
0
0.25
0.5
0.75
1
Estimated Fraction with Liver Tumor
FIG. 4. The estimated mean RPFx and mean RPFx divided by its estimated
standard deviation (S.D.) plotted versus fraction with liver tumor. The relative
standard deviation (S.D./RPFx) is minimized at an incidence rate of approximately 0.16.
rate dependence also motivates consideration of whether there
is an ‘‘optimal’’ tumor response rate at which the relative error
in the estimated RPFs is minimized. In Figure 4, the mean RPF
values obtained for the average body concentration dose metric
are plotted as a function of response probability at the 5%,
10%, 25%, 50%, 75%, 90%, and 95% incidence points; the
figure also displays each mean RPF estimate divided by its
estimated standard deviation, a ‘‘figure of merit’’ that is useful
in assessing the relative precision of the different responsedependent RPF estimates, with larger RPF/S.D. ratios implying
smaller relative uncertainty. The response probability at which
the estimated RPF yields the smallest relative error is approximately 16%, irrespective of which dose metric is considered
(data not shown). As indicated in Table 4, this response rate
corresponds to a mean RPF value (standard deviation) of
approximately 0.26 (0.081) for administered dose and approx-
Relative potency factor sampling distributions obtained
assuming parallel dose responses for TCDD- and 4-PeCDFinduced liver tumors with each of four dose metrics are
presented in Figure 5. Table 4 provides the corresponding
mean RPF estimates and associated standard deviations, and
Table 5 presents selected cumulative percentiles of these RPF
sampling distributions for the four dose metrics that were
considered.
The mean RPF estimates for each dose metric are 0.27
(administered dose), 0.014 (liver concentration), 0.022 (liver
AUC), and 0.036 (average body concentration). These values
are very close to the corresponding RPF16 values derived without enforcing the constraint of parallel tumor responses (see
Table 4). It is noteworthy that the standard deviations of these
RPF estimates are about 30% smaller than those obtained
without imposition of this constraint.
Table 4 also presents the Weibull model parameter estimates
obtained with the two estimation approaches. Of particular note
is the fact the power parameter (a) for 4-PeCDF is substantially
smaller than that for TCDD for all four dose metrics, indicating
a considerably less nonlinear dose–response relationship than
that observed for TCDD, and thus providing some evidence of
non-parallel dose responses. The Weibull model parameters
obtained with the joint approach assuming parallelism strongly
resemble those obtained for TCDD alone. This may reflect the
relative weakness of the liver tumor dose response observed in
the 4-PeCDF bioassay.
DISCUSSION
When the 1997 WHO TEF evaluation was conducted,
carcinogenicity data for 4-PeCDF (and most other dioxin-like
TABLE 4
Weibull Modela Parameters from the Individual and Joint Fits of the TCDD and 4-PeCDF Liver Tumor
TCDDb
Data set
Dose metric
AD
Cl
AUCl
Cb
4-PeCDFb
Joint w/parallel dose responsec
a
b
c
a
b
c
RPF16d
S.D.
a
b
c
RPF
S.D.
2.52
5.16
3.02
2.59
0.996
0.991
0.996
0.997
0.
0.
0.
0.
1.36
1.53
1.51
1.36
0.170
0.168
0.168
0.170
0.014
0.014
0.014
0.014
0.26
0.013
0.021
0.036
0.081
0.0034
0.0059
0.011
2.38
5.10
2.87
2.44
0.945
0.945
0.947
0.945
0.010
0.016
0.011
0.011
0.27
0.014
0.022
0.037
0.054
0.0018
0.0039
0.0073
Note. Incidence, together with relative potency estimates and estimated standard deviations obtained with both procedures.
a
P(d) ¼ c þ (1c)(1exp(b * da)).
b
Doses scaled so that the d ¼ 1 for the highest dose group.
c
Doses scaled so that the d ¼ 1 for the highest TCDD dose group.
d
RPF16 is the relative potency of 4-PeCDF evaluated at the liver tumor incidence rate giving the minimum relative error in potency, namely, 0.16.
RELATIVE POTENCY FACTORS FOR 4-PeCDF
potentially more congruent with probabilistic risk assessment
needs; and (3) generation of empirical sampling distributions
for these RPFs. We propose that the cancer-based RPFs and
RPF distributions derived herein be used in conducting human
cancer risk assessments whenever 4-PeCDF is a significant
congener in the exposure mixture of concern. However, the
WHO TEFs, which were currently being re-evaluated at the
time this manuscript was prepared, may still prove useful for
other exposure circumstances and matrices.
RPF Sampling Distributions
Probability Density, Arbitrary Units
50
45
Administered
Dose
40
Liver AUC
35
30
25
Body
Concentration
Liver
Concentration
20
15
10
Adequacy of an RPF of 0.5 for 4-PeCDF
5
0
0.001
0.01
0.1
1
RPF
FIG. 5. The RPF sampling distributions obtained from simultaneous fits to
the TCDD and 4-PeCDF liver tumor data assuming dose additivity (parallel
dose–response curves).
compounds) were not available. The new NTP carcinogenicity
data permit development of RPFs and estimates of potential
human risk specifically for the cancer endpoint. Relative
potency factors and their distributions derived from these
tumor data obviate the need to rely on the TEF of 4-PeCDF,
which by definition provides only an order-of-magnitude
estimate of potency based on data for a range of biological
endpoints, most of which have limited relevance to the
carcinogenic potential of 4-PeCDF. Data from the NTP TCDD
and 4-PeCDF bioassays (NTP, 2003a, 2003b) permit (1) an
objective assessment of whether or not an RPF of 0.5 (the
current 4-PeCDF TEF value) provides an adequate description
of the 4-PeCDF carcinogenicity data in relation to that obtained
for TCDD; (2) derivation of RPFs and their distributions with
and without the assumption of parallelism, and suitable for use
with administered dose and other internal dose metrics
TABLE 5
Cumulative Percentiles of the RPF Sampling Distributions for
Different Dose Metrics Obtained with a Common Weibull Model
for TCDD- and 4-PeCDF–Induced Liver Tumors, Assuming
Parallel Dose Responses
Cumulative
percentile
1
5
10
25
50
75
90
95
99
281
Administered
dose RPF
Liver
concentration
RPF
Liver
AUC
RPF
Body
concentration
RPF
0.147
0.185
0.205
0.235
0.270
0.306
0.335
0.355
0.385
0.009
0.011
0.012
0.013
0.014
0.015
0.015
0.016
0.017
0.013
0.016
0.017
0.019
0.022
0.024
0.026
0.027
0.029
0.021
0.026
0.029
0.033
0.037
0.042
0.046
0.049
0.053
Our analyses show that an RPF of 0.5 does not adequately
characterize the carcinogenic potency of 4-PeCDF relative to
that of TCDD. Indeed, it significantly overstates this potency.
There is also some evidence of non-parallelism of the liver
tumor dose responses for TCDD and 4-PeCDF, in that the
exponents of dose (the Weibull model parameter a) in the
separately estimated models were quite different, with the dose
response for TCDD-induced liver tumors being considerably
steeper than that for 4-PeCDF. However, the low statistical
power of the bioassay designs noted by Walker et al. (2005)
precludes a rigorous test of this hypothesis.
Despite the large number of dose groups employed in these
studies, maximal tumor responses for 4-PeCDF and TCDD do
not appear to have been achieved in any of the treated groups.
Indeed, the 4-PeCDF study failed to reach a 20% liver tumor
response rate, even at the highest dose tested (200 ng/kg per
day). Although these data are strong enough to reveal the
comparatively low carcinogenic potency of 4-PeCDF relative
to that of TCDD, they fall short of providing a definitive
characterization of the shape of the dose–response curve for 4PeCDF–induced liver tumors. It is noteworthy that the more
sensitive assessments of parallelism by Toyoshiba et al. (2004)
using liver enzyme activity levels determined during the course
of these bioassays provide compelling evidence of the inappropriateness of this assumption. The practical need for
RPFs in human cancer risk assessments thus motivates the
consideration of additional approaches to identifying RPFs that
do not depend on this questionable assumption.
RPF Estimates Without the Assumption of Parallelism
In our analyses, which did not assume parallelism, the RPF
estimates depend on the response level at which they were
estimated, implying that no single RPF value can completely
characterize the potency of 4-PeCDF relative to that of TCDD.
This dependence on response level, illustrated in Figure 3a
and 3b, and Figure 4, appears monotone, decreasing with
increasing probability of being tumor-bearing, and is steepest
at the lower end of the dose–response curve where there is
minimal information about dose response for either TCDD or
4-PeCDF. However, the uncertainty of this estimate, reflected
by the spread in the RPF sampling distributions (depicted in
Figure 3a and 3b), also increases as the probability of being
282
BUDINSKY ET AL.
tumor bearing declines. The result of these two tendencies
is to create an optimum response rate at which the relative
standard error of the estimated RPF is minimized. As shown in
Figure 4 for lifetime average body concentration Cb, this
minimum occurred at a response probability of approximately
0.16 (the tumor incidence in the highest 4-PeCDF dose group),
where the central RPF estimate is approximately 0.036. The
highest tested 4-PeCDF dose thus provides the least uncertain
estimate of a ‘‘point of departure’’ suitable for low-dose risk
extrapolation.
RPF Estimates Assuming Parallelism
The central RPF estimates we obtained with the assumption
of parallel dose–response curves were nearly identical to those
obtained previously with the independent model approach for
the response probability of 0.16 (see Table 4). However, it
is worth noting that the estimated standard deviations of the
RPFs obtained with the parallel dose–response assumption are
approximately 33% smaller than those obtained without it at
the response probability of 0.16 (shown in Table 4). The greater
variance of the non-parallel RPF estimates is a reflection of the
additional freedom that non-parallel dose–response curves
have to diverge from each another. In fact, the assumption of
parallelism leads to estimates of RPF sampling variability that
are biased low when dose responses are not parallel. Allowance
for non-parallelism, at least for 4-PeCDF, reveals additional
uncertainty that should be accounted for when using TEFs (or
RPFs) in risk assessment.
Recently, Walker et al. (2005) published results from
independent and simultaneous fits of the new NTP bioassay
data against administered dose for TCDD, 4-PeCDF, PCB126,
and the presumably equipotent mixture of the three chemicals
with saturable (Hill type) dose–response models. For bile duct
tumors (cholangiocarcinomas), liver adenomas, and lung
tumors (cystic keratinizing epitheliomas), they obtained
endpoint-specific RPFs (standard errors) for 4-PeCDF of 0.03
(0.10), 0.19 (0.25), and 0.36 (not determined), respectively,
without assuming parallel dose responses, and RPF50s of 0.16
(0.04), 0.34 (0.08), and 0.26 (0.12), respectively, with the
assumption of parallel dose responses. These values are quite
close to our results for liver tumors (combined cholangiocarcinomas and hepatocellular adenomas), and the pattern of larger
variability of RPF estimates obtained without assuming
parallelism is also consistent with our results.
RPF Sampling Distributions
The development of empirical RPF sampling distributions
allows for objective tests of a hypothesis that implicitly
underlies the TEF concept, namely, that RPFs for different
endpoints are actually just estimates of a common underlying
population parameter, namely, the ‘‘true’’ TEF. They also
permit the development of TEF estimates using objective
weighting— e.g., inverse variance weighting—of endpoint-
specific RPF estimates. In our view, this potential for objectivity is a marked improvement over previous approaches to
the development of TEF values.
Relative potency factor sampling distributions such as those
we have developed also facilitate the exploration of probabilistic risk assessment approaches that accommodate uncertainty and variability in toxicity parameters as well as exposure
inputs (USEPA, 2004). It is noteworthy that EPA recently
concluded that sufficient data exist to conduct Monte Carlo
analyses for 4-PeCDF: ‘‘For some chemicals, such as PCB 126,
PeCDD and 4-PeCDF, there are sufficient data to apply these
methods.’’ (USEPA, 2003a, Part II, Chapter 9).
RPFs for Different Dose Metrics
It is important to recognize that RPFs derived with
administered dose do not account for important pharmacokinetic differences between congeners such as the more pronounced sequestration in the liver of 4-PeCDF relative to that
of TCDD. Others have suggested that TEFs should account for
congener-specific differences in pharmacokinetics (DeVito and
Birnbaum, 1995; Neubert et al., 1992). DeVito et al. (1997)
recommended that TEF values be developed on both an intake
and a tissue concentration equivalent basis. In addition, the use
of internal dose metrics has been endorsed by the USEPA for
risk assessment purposes:
. . .the scientific principle that the internal dose to the tissue of interest is
the ultimate determinant of toxicity. . . .In its 1994 report, the NRC states
that: . . .the target-site dose is the ultimate determinant of risk . . .. Thus,
the manner or route of systemic exposure is secondary (except as it
affects internal doses, e.g., by first-pass metabolism) to the internal dose.
Others in government also reflect this fundamental principle. (USEPA,
2004).
Certainly, TEFs developed with a body concentration dose
metric would be the most compatible with the USEPA’s current
use of lifetime average body burden when assessing the
potential human risks posed by dioxin-like compounds. The
RPF values we obtained assuming parallel dose–response
curves range from about 0.27 using administered dose to about
one order of magnitude smaller using liver concentration
(0.014), area under the liver concentration curve (0.022), or
lifetime average body concentration (0.036). The corresponding RPF16 estimates we obtained without the parallel dose–
response assumption are very close to these values. If potential
human cancer risk is the primary concern, it would seem that an
appropriate RPF for 4-PeCDF would be either 0.27 or 0.036,
depending on whether administered dose or lifetime average
body burden is the preferred dose metric.
Lipid concentration is another internal dose metric that
merits discussion. At low exposure levels, it should be tightly
correlated with both administered dose and liver concentration,
but when exposure is sufficiently high to cause CYP1A2
induction and substantial sequestration of TCDD or 4-PeCDF
in the liver, these two dose metrics should diverge from one
283
RELATIVE POTENCY FACTORS FOR 4-PeCDF
another. This could be important for non-hepatic lesions such
as gingival tumors, lung, pancreatic and uterine tumors, where
the target organ ligand concentrations are presumably better
represented by those in fat. Even in the liver, the most appropriate dose metric is unclear because the CYP1A2-bound
fraction may be biologically inactive. It is worth noting in this
regard that there is some evidence of hepatic sequestration in
humans, particularly for more highly chlorinated congeners
(Grassman et al., 2000; Iida et al., 1999; Kitamura et al., 2001).
Sequestration is thus relevant for estimating the potential
hazards of dioxin-like compounds to humans.
Given the fact of hepatic sequestration, the appropriate dose
metric for estimating relative potency depends, at least in part,
on whether CYP1A2-bound dioxins are inactive or active.
Studies conducted in CYP1A2 knockout mice have reported
mixed results concerning this issue. Some report enhanced
sensitivity to dioxin-induced immunotoxicity in CYP1A2
(/) mice (Smialowicz et al., 2004), others report no
difference in dioxin-induced enzyme induction or hepatic
oxidative stress between CYP1A2 (–/–) mice and CYP1A2
(þ/þ) mice (Liang et al., 1997; Slezak et al., 1999), and still
others report that hepatotoxicity is dependent on CYP1A2
expression (Smith et al., 2001).
Interestingly, enzyme activity data collected during the
NTP bioassays suggest that the hepatic-sequestered fraction
may not be inactive. In 4-PeCDF-treated animals receiving
100 ng TEQ/kg per day, CYP1A1 activity at 53 weeks was
approximately double (3674 ± 401 pmol/min per milligram)
that in TCDD-treated animals (1871 ± 109 pmol/min per
milligram) receiving the same TEQ dose, despite comparably
induced levels of CYP1A2 and a nearly 30-fold greater hepatic
sequestration of 4-PeCDF than TCDD (Toyoshiba et al., 2004).
If the CYP1A2 bound fraction were biologically inactive, one
would expect much less, rather than more, CYP1A1 activity in
the 4-PeCDF-exposed animals.
An apparent limitation of RPFs based on internal dose
metrics relative to those derived using administered doses
relates to the uncertainty with which internal doses are estimated. The tissue concentration data that we used to estimate
internal doses came from satellite groups of approximately 10
animals that may or may not be representative of the main
study groups in the NTP bioassays, and we just assumed representativeness. There is also measurement error associated with
the resulting internal dose estimates, and we have not taken this
explicitly into account. One would expect that the internal dose
metric RPF sampling distributions would widen, possibly substantially, if measurement errors were properly accounted for,
compared to those we derived based solely on the binomial
sampling variability of tumor incidence. However, it is worth
noting in this regard that some distribution widening should also
occur in the administered dose RPF sampling distributions,
provided one properly accounted for the substantial uncertainty
associated with the assumption that administered dose is
a perfect linear surrogate for the unknown causal dose metric.
Interesting questions, therefore, remain about what the most
appropriate dose metric is for estimating, evaluating, and using
RPFs and TEFs. Although these questions are certainly of
scientific interest, they may be academic when it comes to
practical risk assessment applications, which typically rely on
estimated intake or extrapolated body burdens. We offer
several recommendations regarding the practical application
of RPFs in risk assessment.
First, the RPFs for carcinogenicity of 4-PeCDF should not
be combined with RPFs obtained from enzyme induction or
other non-cancer endpoints, such as immune function changes.
It is noteworthy that draft EPA guidance on the use of TEFs in
ecological risk assessments already permits the use of endpoint-specific RPFs in place of generic TEFs. Specifically, the
guidance states:
Selection of an RPF based on a few data points, or even a single RPF
value, is appropriate if the RPF data are of high quality and the overall
species, endpoint, and dose specificity is greater than for the comparable
TEF. (USEPA, 2003b)
Second, cancer risk assessments should consider the full
range of 4-PeCDF RPFs that can be generated from the
available NTP (2003a, 2003b) bioassay data.
Finally, we encourage the use of probabilistic risk assessment approaches that can accommodate both RPF point
estimates and their sampling distributions. Such approaches
provide a great deal more transparency regarding variability
and uncertainty than the current practice of using upper bound
point estimate TEFs (Finley et al., 2003).
ACKNOWLEDGMENTS
D.P. and T.B.S. provide consulting services to The Dow Chemical Company
and they received financial support for conducting this research. D.P. is also
serving as an expert witness to The Dow Chemical Company on dioxin-related
matters. The authors appreciate the assistance of Dr. Edmund Crouch who
provided the body burden estimates and the maximum likelihood estimation
routines.
REFERENCES
Bailer, A. J., and Portier, C. J. (1988). Effects of treatment-induced mortality
and tumor-induced mortality on tests for carcinogenicity in small samples.
Biometrics 44, 417–431.
Brewster, D. W., and Birnbaum, L. S. (1987). Disposition and excretion of
2,3,4,7,8-pentachlorodibenzofuran in the rat. Toxicol. Appl. Pharmacol. 90,
243–252.
Brown, R. P, Delp, M. D., Lindstedt, S. L., Rhomberg, L. R., and Beliles, R. P.
(1997). Physiological parameters values for PBPK models. Toxicol. Ind.
Health 13, 407–484.
Budinsky, R., Starr, T. B., Paustenbach, D., Landenberger, B., and Fontaine, D.
(2004). A preliminary evaluation of the toxic equivalency factor (TEF) for
2,3,4,7,8-PCDF (4-PCDF) using data from the recent NTP dioxin bioassays.
Organohalogen Cmpd. 66, 3433–3438.
Carrier, G., Brunet R. C., and Brodeur, J. (1995). Modeling of the
toxicokinetics of polychlorinated dibenzo-p-dioxins and dibenzofurans in
284
BUDINSKY ET AL.
mammalians, including humans. II. Kinetics of absorption and disposition of
PCDDs. Toxicol. Appl. Pharmacol. 131, 267–276.
DeVito, M. J., and Birnbaum, L. S. (1995). The importance of pharmacokinetics in determining the relative potency of 2,3,7,8-tetrachlorodibenzo-pdioxin and 2,3,7,8-tetrachlorodiebenzofuran. Fundam. Appl. Toxicol. 24,
145–148.
DeVito, M. J., Diliberto, J. J., Ross, D. G., Menache, M. G., and Birnbaum,
L. S. (1997). Dose-response relationships for polyhalogenated dioxins and
dibenzofurans following subchronic treatment in mice. Toxicol. Appl.
Pharmacol. 147, 267–280.
Finley, B. L., Connor, K. T., and Scott, P. K. (2003). The use of toxic
equivalency factor distributions in probabilistic risk assessment for dioxins,
furans, and PCBs. J. Toxicol. Environ. Health 66, 533–550.
Fontaine, D. D., Paustenbach, D., Landenberger, B. D., Budinsky, R. A., and
Starr, T. B. (2004). An evaluation of the toxic equivalency factor (TEF) for
2,3,4,7,8-PCDF (4-PCDF) using data from the recent NTP dioxin bioassays.
Paper W8.7. Society for Risk Analysis Annual Meeting, 2004.
Grassman, J. A., Needham, L. L., Masten, S. A., Patterson, D., Portier, C. J.,
Lucier, G. W., and Walker, N. J. (2000). Evidence of hepatic sequestration of
dioxins in humans? An examination of tissue levels and CYP1A2 expression.
Organohalogen Cmpds. 48, 87–90.
Harper, N., Connor, K., and Safe, S. (1993). Immunotoxic potencies of polychlorinated biphenyl (PCB), dibenzofuran (PCDF) and dibenzo-p-dioxin
(PCDD) congeners in C57BL/6 and DBA/2 mice. Toxicology 80, 217–227.
Haws, L. C., Su, S. H., Harris, M., DeVito, M. J., Walker, N. J., Farland, W. H.,
Finley, B., and Birnbaum, L. S. (2005). Development of a refined database of
mammalian relative potency estimates for dioxin-like compounds. Toxicol.
Sci. 89, 4–30.
Iida, T., Hirakawa, H., Matsueda, T., Nagayama, J., and Nagata, T. (1999).
Polychlorinated dibenzo-p-dioxins and related compounds: Correlations of
levels in human tissues and in blood. Chemosphere 38, 2767–2774.
Kociba, R. J., Keyes, D. G., Beyer, J. E., Carreon, R. M., Wade, C. E., Dittenber,
D. A., Kalnins, R. P., Frauson, L. E., Park, C. N., Barnard, S. D., et al.
(1978). Results of a two-year chronic toxicity and oncogenicity study of
2,3,7,8-tetrachlorodibenzo-p-dioxin in rats. Toxicol. Appl. Pharmacol. 46,
279–303.
Kitamura, K., Nagao, M., Yamada, T., Sunaga, M., Hata, J., and Watanabe, S.
(2001). Dioxins in bile in relation to those in the human liver and blood. J.
Toxicol. Sci. 26, 327–336.
NTP (2003c). Toxicology and carcinogenesis studies of 3,3#,4,4#,5-pentachlorobiphenyl (PCB 126) in female Harlan Sprague-Dawley rats. NTP TR
520, NIH Publication No 04–4454.
NTP (2003d). Toxicology and carcinogenesis studies of a mixture of 2,3,7,8tetrachlorodibenzo-p-dioxin (TCDD), 2,3,4,7,8-pentachlorodibenzofuran
(PeCDF) and 3,3#,4,4#,5-pentachlorobiphenyl (PCB 126) in female Harlan
Sprague-Dawley rats. NTP TR 526, NIH Publication No 04–4462.
Olson, J. R., Bellin, J. S., and Barnes. D. C. (1989). Reexamination of data
used for establishing toxicity equivalence factors (TEFS) for chlorinated
dibenzo-p-dioxins and dibenzofurans (CDDS and CDFS) Chemosphere 18,
371–381.
Patterson, D. G., Todd, G. D., Turner, W. E., Maggio, V., Alexander, L. R., and
Needham, L. L. (1994). Levels of non-ortho-substituted (coplanar), monoand di-ortho-substituted polychlorinated biphenyls, dibenzo-p-dioxins, and
dibenzofurans in human serum and adipose tissue. Environ. Health Perspect.
103(Suppl. 1), 195–204.
Petreas, M., She, J., Winkler, J., Visita, P., McKinney, M., Reynolds, P.,
Smith, D., Gilliss, D., Hurley, S., Jeffrey, S., et al. (2000). Body burdens
of organohalogens in California populations. Organohalogen Cmpds. 48,
17–20.
Pluss, N., Poiger, H., Hohbach, C., Suter, M., and Schlatter, C. (1988).
Subchronic toxicity of 2,3,4,7,8-pentachlorodibenzofuran (PeCDF) in rats.
Chemosphere 17, 1099–1110.
Safe, S. H. (1998). Development validation and problems with the toxic
equivalency factor approach for risk assessment of dioxins and related
compounds. J. Animal Sci. 76, 134–141.
Slezak, B. P., Diliberto, J. J., and Birnbaum, L. S. (1999). 2,3,7,8-Tetrachlorodibenzo-p-dioxin-mediated oxidative stress in CYP1A2 knockout
(CYP1A2–/–) mice. Biochem. Biophys. Res. Commun. 264, 376–379.
Smialowicz, R. J., Burgin, D. E., Williams, W. C., Diliberto, J. J., Setzer, R. W.,
and Birnbaum, L. S. (2004). CYP1A2 is not required for 2,3,7,8tetrachloordibenzo-p-dioxin-induced immunosuppression Toxicology 197,
15–22.
Smith, A. G., Clothier, B., Carthew, P., Childs, N. I., Sinclar, P. R., Nebert, D. W.,
and Dalton, T. P. (2001). Protection of the Cyp 1a2 (–/–) null mouse
against urophorphyia and hepatic injury following exposure to 2,3,7,8tetrachlorodibenzo-p-dioxin. Toxicol. Appl. Pharmacol. 173, 89–98.
Toyoshiba, H., Walker, N. J., Bailer, J. A., and Portier, C. J. (2004). Evaluation
of toxic equivalency factors for induction of cytochrome P450 CYP1A1 and
CYP1A2 enzyme activity by dioxin-like compounds. Toxicol. Appl.
Pharmacol. 194, 156–168.
Kreuzer, P. E., Csanady, G. A., Baur, C., Kessler, W., Papke, O., Greim, H., and
Filser, J. G. (1997). 2,3,7,8-Tetrachlorodbenzo-p-dioxin (TCDD) and congeners in infants. A toxicokinetic model of human lifetime body burden by
TCDD with special emphasis on its uptake by nutrition. Arch. Toxicol. 71,
383–400.
USEPA. (1988). A cancer risk-specific dose estimate for 2,3,7,8-TCDD.
Appendices A Through F. EPA/600/6–88/007-AB.
Liang, H. C. L., McKinnon, R. A., and Nebert, D. (1997). Sensitivity of
CYP1A1 mRNA inducibility by dioxin is the same in CYP1a2 (þ/þ)
wild-type and Cyp1A2 (/) null mutant mice. Biochem. Pharmacol. 54,
1127–1131.
USEPA. (2003a). Exposure and Human Health Reassessment of 2,3,7,8Tetrachlorodibenzo-p-Dioxin (TCDD) and Related Compounds. Washington, DC, National Academy Sciences (NAS) Review Draft, December 2003.
Neubert, D., Golor, G., and Neubert, R. (1992). TCDD-toxcity equivalencies
for PCDD/PCDF congeners: Prerequisites and limitations. Chemosphere 25,
65–70.
USEPA. (2003b). Framework for Application of the Toxicity Equivalence
Methodology for Polychlorinated Dioxins, Furans, and Biphenyls in
Ecological Risk Assessment. EPA/630/P-03/002A, June 2003.
NHANES (2005). National Center for Health Statistics. Third National Report
on Human Exposure to Environmental Chemicals. U. S. Department of
Health and Human Services, Centers for Disease Control and Prevention,
National Center for Environmental Health, NCEH Pub. No. 05–0579.
USEPA. (2004). An Examination of EPA Risk Assessment Principles and
Practices. EPA/100/B-04/001, March 2004.
NTP (2003a). Toxicology and carcinogenesis studies of 2,3,4,7,8-pentachlorodibenzofuran (PeCDF) in female Harlan Sprague-Dawley rats. NTP TR
525, NIH Publication No 04–4461.
NTP (2003b). Toxicology and carcinogenesis studies of 2,3,7,8-tetrachlorodibenzo-p-dioxin in female Harlan Sprague-Dawley rats. NTP TR 521, NIH
Publication No 04–4455.
USEPA. (2001). Benchmark Dose Software. Version 1.3.2 downloaded from
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid¼20167, March 2004.
van Birgelen, A. P. J. M., DeVito, M. J., Orzech, D., Walker, N., Birnbaum,
L. S., Bucher, J., and Lucier, G. (1997). Design of 2-year bioassays with
dioxin-like compounds in female Sprague-Dawley rats. Organohalogen
Cmpds. 34, 154–159.
Van den Berg, M., Bouwman, C., and Seinen, W. (1989). Hepatic retention of
PCDDs and PCDFs in C57BL/6 and DBA/2 mice. Chemosphere 19, 795–782.
Van den Berg, M., Birnbaum, L., Bosveld, A. T., Brunstrom, B., Cook, P.,
Feeley, M., Giesy, J. P., Hanberg, A., Hasegawa, R., Kennedy, S. W., et al.
RELATIVE POTENCY FACTORS FOR 4-PeCDF
285
(1998). Toxic equivalency factors (TEFs) for PCBs, PCDDs, PCDFs for
humans and wildlife. Environ. Health Perspect. 106, 775–792.
Dose-additive carcinogenicity of a defined mixture of dioxin-like compounds. Environ. Health Perspect. 113, 43–48.
Waern, F., Flodstrom, S., Busk, L., Kornevi, T., Nordgren, I., and Ahlborg,
U. G. (1991). Relative liver tumour promoting activity and toxicity of some
polychlorinated dibenzo-p-dioxin- and dibenzofuran-congeners in female
Sprague-Dawley rats. Pharmacol. Toxicol. 69, 450–458.
Wittsiepe, J., Furst, P., Schrey, P., Lemm, F., Kraft, M., Eberwein, G., Winneke, G.,
and Wilhelm, M. (2004). PCDD/F and dioxin-like PCB in human blood and
milk from German mothers. Organohalogen Cmpds. 66, 2865–2872.
Walker, N. J., Crockett, P., Nyska, A., Brix, A., Kokinen, M. P., Sells, D. M.,
Hailey, J. R., Easterling, M., Haseman, J. K., Yin, M., et al. (2005).
Wittsiepe, J., Schrey, P., Ewers, U., Selenka, F., and Wilhelm, M. (2000).
Decrease of PCDD/F levels in human blood from Germany over the past ten
years (1989–1998). Chemosphere 40, 103–110.