United States
           Environmental Protection
           Agency
            Office of Research and
            Development
            Washington, DC 20460
EPA/600/R-92ra45
December 1991
&EPA
Dose-Response Analysis of
Ingested Benzo[a]Pyrene
(CAS No. 50-32-8)

-------

-------
                                            EPA/600/R-92/045
                                              December 1991
       DOSE-RESPONSE ANALYSIS

                   OF

       INGESTED BENZO[a]PYREIME

            {CAS No. 50-32-8)
      Human Health Assessment Group
Office of Health and Environmental Assessment
     Office of Research and Development
    U.S. Environmental Protection Agency
            Washington, DC
                                       Printed on Recycled Paper

-------
                                   DISCLAIMER
      This document has been reviewed in accordance with U.S. Environmental Protection
Agency policy and approved for publication.  Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.

-------
                                CONTENTS
List of Tables	iv
Preface		 v
Authors and Reviewers	vi
1.  INTRODUCTION			  1

2.  DATA AND IMPLICATIONS FOR DOSE-RESPONSE MODELING	  2

      2.1.  Biological and Mechanistic Data	  2
      2.2.  Empirical Observations		  3

3.  CALCULATION OF POTENCY SLOPES	  4

      3.1.  Brune et al., 1981 . .  .'..	  5
      3.2.  Neil and Rigdon, 1967		  5
      3.3.  Chouroulinkov et al.,  1967	  9

4.  EVALUATION OF DOSE-RESPONSE MODELS CONSTRUCTED ON THE BASIS OF
  THE NEIL AND RIGDON DATA .	  10

5.  DISCUSSION AND SUMMARY		. .  15

6.  OUTSTANDING ISSUES	 .	  17

7.  CONTINUING RISK ASSESSMENT WORK AND RESEARCH NEEDS .	  20


Appendix A:  Other B[a]P Data	 .  22

Appendix B:  Derivation of Dose-Response Model to Fit
           Neil and Rigdon Data	  26

References . .	  28
                                    HI

-------
                                  LIST OF TABLES
1.    Incidence rates of forestomach and total contact sites (forestomach, larynx, esophagus)
     tumors in Sprague-Dawley rats exposed to B[a]P in caffeine solution by intubation and in
     diet mix	  6

2.    Predicted vs. observed tumor incidence and data used to calculate the dose-response
     model	• • • •	  7

3.    Potency slopes per (mg/kg/day) for humans calculated from different studies  	  10

4.    Using the Weibull-type model to predict other mouse studies in additional to those in
     Tables			-	 .  12

5.    Using Clement's dose-response model to predict other mouse studies	  14
                                          IV

-------
                                     PREFACE
     This report was prepared by the Human Health Assessment Group within the Office of
Health and Environmental Assessment (OHEA) to support OHEA's preparation of a Drinking
Water Criteria Document (DWCD) on Polycyclic Aromatic Hydrocarbons (PAHs). The DWCDs
are prepared for the Office of Ground Water and  Drinking Water and the Office of Science and
Technology within the Office of Water.
     The dose-response assessment of PAHs will comprise three steps.  The first two steps
involve assessing the dose-response relationships of benzo[a]pyrene (B[a]P), one of the
PAHs, via  ingestion and inhalation. The third step involves the analysis of PAHs other than
B[a]P that  are commonly found in the environment.
     This report focuses on the dose-response analysis of ingested B[a]P as there is more
information available on B[a]P that is suitable for  quantitative assessment than is available for
the other PAHs.

-------
                           AUTHORS AND REVIEWERS
AUTHORS

Chao W. Chen
Carcinogen Assessment Statistics and Epidemiology Branch
Human Health Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency

Margaret M.L Chu
Genetic Toxicology Assessment Branch
Human Health Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
REVIEWERS

V. James Cogliano
Carcinogen Assessment Statistics and Epidemiology Branch
Human Health Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Vicki L. Dellarco
Genetic Toxicology Assessment Branch
Human Health Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
                                       vi

-------
Robert E. McGaughy
Human Health Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency

Rita S. Schoeny
Methods Evaluation and Development Staff
Office of Health and Environmental Assessment
Environmental Criteria and Assessment Office
U.S. Environmental Protection Agency
                                       VII

-------

-------
                                 1. INTRODUCTION
     This report represents the first part of the Office of Health and Environmental
Assessment's dose-response analysis of polycyclic aromatic hydrocarbons (PAHs). Because
of the tremendous amount of information bearing on approaches that can be developed to
estimate risks from exposures to PAHs, the task is divided into three steps beginning with an
analysis of benzo[a]pyrene (BAP), one of the PAHs. The steps are: (1) determining the oral
potency slope for benzo[a]pyrene (B[a]P), (2) determining the inhalation potency slope for
B[a]P, and  (3) determining the potency slopes for other PAHs using a  relative potency
approach.  The focus of this  report is on estimating the carcinogenic potency slope of B[a]P
by ingestion.  The potency slope is used to provide an upper-bound estimate for risk from
ingesting low doses of B[a]P. It should be noted that only information  relevant to analyzing
dose-response relationships by ingestion is analyzed in this report.
     The determination of potency slopes for B[a]P is chosen as the first two steps because
there is more information on  B[a]P than other PAHs that are found in environmental media.
Furthermore,  few PAHs have even minimal information for performing  dose-response analysis
of oral or inhalation routes of exposure.  However, results from skin painting experiments on
PAHs where B[a]P is used as positive controls are available.  The potential of using B[a]P as
a reference point  in deriving oral and inhalation potency slopes for other PAHs can be
developed. The validity of calculating relative potency  between two compounds based on
data from one route of exposure and then extrapolated to another route of exposure needs
verification.
     In  1987, ICF-Clement Associates (now known as  Clement International Corporation,
hereafter referred to as Clement) was contracted to prepare a quantitative risk assessment of
PAHs.  The report was peer-reviewed at an EPA-sponsored peer-review workshop on May 28-
29, 1988.  A substantial part  of the Clement report concentrated on constructing a dose-
response model using data from Neil and Rigdon (1967) for B[a]P. At  the peer-review
workshop, a number of recommendations were made by the expert review panel relevant to
the estimation of an oral potency for B[a]P. These included:
      • As the multistage model is currently used by the EPA, it should also be used to
         calculate potency, and the results should be  compared to those calculated by
         Clement.

-------
      •  Include other B[a]P studies and B[a]P data from studies using B[a]P as a positive
         control.
      •  Other tumor sites should also be considered.
      In this report, 14 studies on B[a]P-induced forestomach tumors via the oral route of
exposure are reviewed and evaluated.  In addition to the potency slope derived from the
Clement report, several other estimates are provided, using either different dose-response
models on the same data (from the Neil and Rigdon, 1967 study used by Clement) or data
from different oral studies.

           2. DATA AND IMPLICATIONS FOR DOSE-RESPONSE MODELING

      Since humans are exposed to mixtures of PAHs, information from human studies for
dose-response analysis of B[a]P alone is not available.  Tumor induction data from animal
studies on B[a]P by ingestion are also limited.  Fourteen studies are reviewed here for
applicability in the derivation of an oral potency slope. Current reviews concerning biological
and mechanistic data were consulted to guide the effort.

2.1. BIOLOGICAL AND MECHANISTIC DATA
       Biological and mechanistic data that impact the dose-response analysis for determining
human risk from ingesting B[a]P include: quantitative biological data (e.g., specific DMA  • '•
adduct(s) as measures of internal dose; comparative species sensitivity data to guide cross-
species extrapolation; absorption, distribution, and elimination kinetics to determine if high
dose response is predictive of low dose response; and  data from other routes of
administration (e.g., skin painting) to derive the ingestion response.
       Reviews of B[a]P published in the literature (Cooper et al., 1983; Baird et al., 1988;
Graslund and Jernstrom, 1989; DiGiovanni, 1989; Nebert, 1988, 1989; McKay et al., 1988;
IARC, 1983; National Research Council, 1983; Santodonato et al., 1981) suggest that many
factors modulate the biological conversion  of B[a]P to its ultimate carcinogenic form.  Further,
the expression of the carcinogenic potential of B[a]P depends on factors such as species,
strain, diet, route and method of  administration, and other host-related factors such as the
age, sex, and previous and concurrent exposure of the organism to other environmental
agents.  More than one mechanism may be involved in B[a]P induced tumors.  However, the

-------
effect of dose, route of entry, and tissue or genetic factors of the organism, which determines
if one mechanism predominates or if mechanisms work jointly, is not completely understood.
       Even though the volume of information on the mechanism of carcinogenic action of
B[a]P is large, it does not seem to provide quantitative data that are directly applicable or that
can be linked to deriving oral potency estimates. Definitive conclusions can only be made
upon a more comprehensive evaluation of the data and integrating the data from other routes
of administration.  It is clear, however, that the key limitations of the data for quantitative risk
assessment are the lack of a quality bioassay for the ingestion route and the lack of
quantitative biological data linked to the bioassay. These limitations cannot be bridged by a
thorough analysis of the existing data alone.
2.2.  EMPIRICAL OBSERVATIONS
       Fourteen B[a]P ingestion studies (1 in rats and 13 in mice) were reviewed and
evaluated for quantitative risk assessment.  These studies included long- and short-term B[a]P
bioassays and using B[a]P bioassays of other compounds in positive control groups. These
data are used either for calculating potency slopes or for evaluating the validity of the model
constructed on the basis of the Neil and Rigdon (1967) data. Although its experimental
protocol is unconventional and somewhat complicates mathematical modeling, the Neil and
Rigdon (1967) study provides useful empirical information on B[a]P because of its variation in
exposure duration, dosing pattern, and large number of animals studied. It provides empirical
information that usually is not available from long-term bioassays. The main drawback of the
Neil and Rigdon study is that, for some dosed groups, only ranges of the ages when exposure
began or was terminated are reported,  and the interval of these ranges vary. To use these
data in dose-response modeling, the midpoint of the range  is used as a surrogate for the age
exposure began or was terminated.
       Although only 3 of the 14 studies are appropriate  for calculating a B[a]P potency slope,
two interesting observations can be made from these data when they are taken together:
       1. B[a]P is capable of inducing  papillomas from a single oral dose (see Appendix A:
Berenblum and Haran, 1955; Neil and Rigdon, 1967; Field and Roe, 1965; Fedorenko and
Yansheva, 1967; Roe et a!., 1970; Robinson et al., 1987).  However, to induce carcinomas,
multiple dosing  may be necessary.  It should be emphasized here that these observations

-------
should not be construed as a real biological mechanism; it is only an empirical observation of
animals exposed to B[a]P via the oral route.
       2.  In the studies cited above, the proportion of animals with tumors (papillomas and
carcinomas combined) appears to depend less on the duration of exposure (or total dose) but
more on the dose rate and duration of study, irrespective of animals being exposed to B[a]P
either from single or multiple doses (at an identical dose rate). The difference between the
single exposure and the multiple (longer) exposure is that almost all carcinomas were induced
only from multiple dose exposure; single exposure induced only papillomas.  For instance, in
the Berenblum and Haran (1955) study, 15 animals (out of 19) had carcinomas when animals
were exposed to 1.5 mg of B[a]P weekly for 20 to 30 weeks, while no  animals (out of 17) had
carcinomas when animals were exposed to a single dose (1.5 mg) of B[a]P.  However, despite
the great difference in the exposure duration and totaldose, the  proportion of animals with
tumors (papillomas and carcinomas combined) in both experiments are comparable (16/19 vs.
16/17).
       Based on these empirical observations  and the multistage theory of careinogenesis
(i.e., initiation-promotion-conversion), the available data seem to  suggest that B[a]P has
potential for initiation and conversion (from initiated cells to malignancy), without excluding the
possibility of tumor promotion, and that a continuous exposure to B[a]P may  not be required
for initiated cells to grow into detectable tumors.  However, the presence of B[a]P may be
required for progression to carcinomas.  A mathematical model that is consistent with the
empirical evidence is presented in Appendix B and is used for dose-response modeling based
on the Neil and Rigdon data.

                       3. CALCULATION OF POTENCY SLOPES
       Only data from three dietary studies (Brune et al., 1981 [Table 1]; Neil and Rigdon,
 1967 [Table 2]; Chouroulinkov et al., 1967) are considered adequate for estimating potency
 slopes. Other studies are used to evaluate the validity of the dose-response model
 constructed on the basis of the Neil and Rigdon data.
       The procedures and data for calculating slopes from each of the three studies are
 described below. In the calculations,  the body surface equivalence assumption is used for
 interspecies conversion; namely, the carcinogenic effect due to a dose expressed in units of

-------
mg per body surface area is assumed to be equivalent between animals and humans. This
assumption is recommended in EPA's Guidelines for Carcinogen Risk Assessment (U.S. EPA,
1986). In the actual calculations, a slope for humans can be derived by multiplying a factor,
(70 kg/Wa)1/3, to the animal slope, where Wa is animal body weight in kg, 70 kg is the assumed
human body weight, and the animal slope is expressed in terms of (mg/kg/day)"1.

3.1. BRUNEETAL, 1981
       The study by Brune et al. (1981) is the only 2-year ingestion study reviewed.  B[a]P
was administered to Sprague-Dawley rats either as an admixture to the diet or by gavage in
an aqueous 1.5% caffeine solution. Combined data from male and female rats are given in
Table 1.  The linearized multistage (LMS) model was used to calculate the potency slope for
B[a]P, assuming that the average body weight for the animals was 0.4 kg.  The results are
given in Table 3. The study with caffeine solution by intubation produced a larger slope
estimate.  It is not clear whether or not the larger slope estimate is due to the effects of
caffeine or intubation. As will  be discussed, there is some possibility that animals exposed to
B[a]P by intubation could  produce more tumors than animals exposed to B[ai]P in the diet.
       In the diet mix study (but not in the caffeine  solution study), more total contact-site
tumors (larynx, esophagus, and forestomach)  were  induced in males than in females.
Although it is possible to separate total tumor  incidence rates between males and females, not
enough information is provided in the paper to separate the forestomach tumor incidence
rates between males and females.  In order to derive a male-based potency slope  based on
the forestomach tumors, a factor of 1.5 is multiplied to the slopes calculated on the basis of
combined (male and female) forestomach data.  The 50% increase of slope values represents
the cancer risk increase when male data (total tumors) are used, instead of combined data.

3.2. NEIL AND RIGDON, 1967
       Since animals in the Neil and Rigdon (1967) study were exposed (partial lifetime) to
B[a]P during different periods of their lifetimes, a model that reflects this exposure pattern and
is consistent with the empirical evidence discussed previously must be developed to fit the
dose-response data and to calculate the potency slopes for B[a]P. A Weibutl-type of
dose-response model, which can incorporate partial-lifetime exposure, is used to fit the Neil
and Rigdon data (Table 2).

-------
      Table 1. Incidence rates of fbrestomach and total contact sites (forestomach, larynx,
      esophagus) tumors in Sprague-Dawley rats exposed to B[a]P in caffeine solution by
      intubation and in diet mix
Dose
(mg/kg/year)
Forestomach
tumors
Total
tumors
Median survival
Time (days)
In caffeine solution (via intubation)
0
6
18
39
3/64
11/64
25/64
14/64
6/64
13/64
26/64
14/64
102
112
113
87a
In diet mix
0
6
39
2/64
1/64
9/64
3/64"
3/64b
10/64b
129
128
131
• Data from this group were not used in the calculation because the median survival time was
much shorter than for the other groups.  No adjustment is possible because time to death
data were not available.

"The incidence of total tumors in males was 3/32, 3/32, and 8/32, respectively, for control-,
low-, and high-dose groups.  No sex-specific forestomach tumor data were available.

Source: Brune etal., 1981.


       The probability of observing tumors by age t (days) when animals are administered d

(mg/g diet) of B[a]P during the age interval (t^) is given by
                                                                        t>tr
Using the above model and data from Table 2, the parameters are estimated by the maximum

likelihood method as follows:
                            -q^O,  q2 = 2.10x10-5,   k = 3.13.

-------
       Table 2. Predicted vs. observed tumoY incidence and data used to calculate
       the dose-response model3                    ;,                 :
Dose
(mg/g diet)
0.000
0.001
0.01
0.02
0.03
0.04
0.045
0.05
0.10
0.25
to
(day)
0
30
30
116
50
67
50
20
22
19
tt
(day)
300
140
140
226
160
177
160
172
132
137
- ' , t , ..-„•"
(day) ,
300
140
140
226
160
177
162
172
132
137
Observed"
0/289 (0.00)
0/25 (0.00)
0/24 (0.00)
1/23 (0.05)
0/37 (0.00)
1/40 (0.03)
4/40 (0.10)
24/34 (0.70)
19/23 (0.82)
66/73 (0.90)
Predicted
incidence rate
0.00
0.00
0.00
0.02
0.04
0.08
0.11
0.30C
0.40°
0.98
a ^ = age of mice (in days) when exposure began, t, = age of mice (in days) when exposure
was terminated, t = age of mice (in days) when the study was terminated. Data on the control
and the first nine low-dosed groups were taken from Table 1 of Neil and Rigdon (1967). Data
from other groups were not included because animals in these groups were exposed to high
doses of B[a]P and/or for only very short periods.
b Observed number of animals with tumors (incidence  rate).
0 The predicted value lies outside the 95% confidence  interval calculated from the observed
response. This crude test is used to determine whether or not the model reasonably predicts
the observed response.
This model is equivalent to the linearized-muitistage model that would be used when animals
are exposed to B|a]P for their entire lifetimes.  The derivation of the model is given in
Appendix B. The resultant dose-response model is also used to evaluate the other 12 mouse
studies (Table 4).                                     -'.••<...•
      As indicated in Table 2, the model fits the Neil and Rigdon data adequately for all dose
groups except for the two treatment groups (0.05 and 0.1 mg/g B[a]P in the diet).  There may
be some underlying biological, rather than statistical, reasons to explain the poor fit of these
two treatment groups. It is interesting to observe that animals in the three highest-dosed

-------
groups, including the two indicated above, were exposed to B[a]P at a much younger age
(about 20 weeks) than animals in other groups. The higher than expected tumor
response might be due to age-dependent physiological parameters (e.g., a difference in cell
division rates in the forestomach tissue between younger and older animals). Further
research is needed to  confirm or deny this hypothesis.
       As indicated  in Table 4, the model appears to adequately predict the tumor response in
mice in most of the studies in which the animals were exposed to B[a]P via dietary intake, but
the model does not adequately predict the tumor response in mice  exposed to B[a]P via
gavage.  The gavage studies (most of which are also single-dose experiments) tend to
produce a higher tumor incidence than predicted by the model.
       The Weibull-type model is used to calculate the potency slope at the 10% point by
setting ^=0, ^=730, and t=730 days in the model.  The 10% point corresponds to d=2.4 ppm
diet (equivalent to 0.3 mg/kg/day).
       The Neil  and Rigdon data were also used by Clement to calculate potency slopes.
The two slopes proposed in the Clement report are 2.9/(mg/kg/day) (point estimate) and
5.9/(mg/kg/day) (95%  upper-bound estimate). The linear term (slope), 2.9/(mg/kg/day),
proposed by Clement (1990a), is not used in Table 3 because it is the slope (tangent) at dose
zero.  When the dose-response curve is not linear at low doses, the tangent at dose zero
should not be used as a potency slope for low-dose extrapolation because it is a lower bound,
not an upper bound, for the dose-response curve at low doses.  In order to  obtain the 95%
upper bound, Clement found that the point estimate of the cell proliferation parameter b must
be reduced (rather than increased or held constant as one would expect) by more than 50%
from the point estimate. The value 5.9/(mg/kg/day) is not recommended for two reasons:  (1)
 Not all parameters are included in searching for an upper bound; only two dose-related
 parameters are included. If background parameters are included, the upper-bound value may
 be different. (2) It is not biologically reasonable to haye to reduce the cell growth rate by
 more than 50%  in order to get an upper bound. The upper bound for risk is supposed to be
 calculated by increasing the value of parameters starting from their maximum  likelihood
 estimates until the log-likelihood ratio statistic exceeds a given  critical value (which  depends
 on the confidence level specified).  From the approach that an  upper bound is calculated, one
 would not expect to obtain an upper bound by decreasing a parameter from the value where
 the point estimate of the tumor response is obtained.  Being unable to obtain  an  upper bound

                                           8

-------
by starting from parameter values associated with the point estimate of dose, response
suggests the peculiarity of the model. It can be demonstrated that a 50% decrease of cell-
growth rate can significantly alter the shape of the dose-response model. Note also that the
background  tumor incidence is estimated to be greater than zero, a consequence of using a
nonzero historic background rate of a different strain of mice (SWR/J).  CFW mice were used
in the Neil and Rigdon study.
       In this report, the slope is obtained by extrapolating linearly from the 10% point to the
background  on the fitted dose-response curve.  Using this gipproach, two slopes are obtained:
4.5/(mg/kg/day), based on the Weibull-type model, and 9.3/(mg/kg/day), based on the Clement
model. These two slopes are presented in Table 3. When the fitted dose-response curve is
not linear at low doses, the potency slope can be defined as a secant from
a point on the dose-response curve to the point at zero dose.  This is a reasonable way to
find an upper bound for risk at low doses. A similar concept and approach have  been
proposed by Krewski et al. (1986, 1991) and Gaylor and Kodell (1980).  These investigators
found that the potency slopes calculated from this approach are comparable  to those obtained
from the linearized multistage  model for most of the compounds they have studied.

3.3.  CHOUROULINKOV ET AL, 1967
       This study was selected for slope  calculation because it is a long-term study. Mice
(strain unknown) were given 8 mg of B[a]P in olive oil mixed in the diet over  14 months.  The
following gastric tumors were observed;             •
Dose
(mg/kg/day)
0
0.63
Incidence
rate
0/81
5/81 (0.062)
       Since there is only one dosed group, the slope is calculated by direct extrapolation
from the observed response (0.062) at 0.63 mg/kg/day to the background rate (0.00):

-------
      Table 3.  Potency slopes per (mg/kg/day) for humans calculated from different
      studies0
Study
Neil and Rigdon (1967)

Bruneetal. (1981)






Chouroulinkov et al. (1967)
Potency slope
4.5
9.3

Remarks
Weibull-type
Clement (calculated at the
10% point)

In caffeine solution via intubation
67.8b
68.7"
total tumors
forestomach tumor only
In diet mix
11.7, (17.8)°
9.5, (14.2)c
6.5
total tumors
forestomach tumor only
linearly extrapolated from
the observed response
a The potency slope is used to calculate an upper-bound estimate for risk due to ingesting low
doses of B[a]P.

b B[a]P was given to animals via intubation. There is some evidence that intubation may
increase tumor response: therefore, this value is excluded from the final recommended slopes.

c Values in parentheses represent the potency slope estimates based on male rat data alone,
by adjusting upward 50% of the slopes calculated on the basis of combined data (male and
female) data.  See discussion in the text under calculation of potency slopes.


        (0.062/0.63) x (70 kg/0.03 kg)1/3 x (24 months/14 months)3 = 6.5/(mg/kg/day),

where (24/14)3 is a factor to adjust for less-than-lifetime follow-up.


  4.  EVALUATION OF DOSE-RESPONSE MODELS CONSTRUCTED ON THE BASIS OF

                           THE NEIL AND RIGDON DATA
       Data from 12 mouse studies are used to evaluate the two dose-response models (the
Weibull-type and the Clement models) calculated on the basis of the Neil and Rigdon data.
                                         10

-------
       As shown in Table 4, the Weibull-type model appears to adequately predict the tumor
response in most of the dietary studies, but is inadequate for most of the intubation studies.
In those animals exposed to B[a]P by intubation, B[a]P induced greater tumor incidence than
predicted by the model.  There are two possible explanations for the discrepancy observed
between these two types of studies (dietary vs. gavage):  (1) B[a]P by gavage intake induces
more tumor responses than dietary intake (i.e., the gavage effect), and (2) the effect of B[a]P
continues  after termination of exposure (i.e., the B[a]P effect is longer than that implied by the
term ^ in the model). Note that it is impossible to separate the two possible effects (i.e., the
gavage effect and the long-lasting effect of B[a]P) because all of the gavage studies (except
for those by Brune et al. [1981] and Robinson et al. [1987]) are also single dose studies.
Considering the empirical evidence observed previously that a single dose of B[a]P is
sufficient to induce tumors, one is tempted to conclude that the second reason explains all the
discrepancy observed between the two sets of data.  However, the second reason  cannot
totally explain the discrepancy because there are cases in the gavage studies showing
comparable to or less than the predicted values.  It is possible that both effects are operating.
Pharmacokinetic information on B[a]P during and after exposure will be useful in answering
the questions about the effect of B[a]P after exposure is terminated.
       For the Clement model, two approaches are used  to calculate predicted values.
Despite the fact that it is not reasonable to substitute a different parameter (Approach II in
Table 5) after the model has been derived, we have followed the Clement model using two
approaches (Clement, 1990a; [Table 8]) to obtain predicted values in Table 5.  The first
approach  assumes that the growth rate for initiated cells during the nonexposed period  is the
same as the background growth rate. The second approach assumes that the growth rate for
initiated cells during the nonexposed period is the same as that during the exposed period.
As shown  in Table 5, the Clement model (i.e., under Approach I) has predicted  values
equal to 0 for almost  all of the cases where ^ is less than t (i.e., when animals were not
sacrificed  immediately after exposure was terminated). There seems to be no clear pattern
for Approach II.
                                          11

-------
Table 4.  Using the Weibull-type model to predict other mouse studies in
addition to those in Table 3.
Study
Dose
(mg/g
diet)
V
(days)
ti
(days)
t
(days)
Observed
forestomach
tumors
Predicted
Dietary Studies: ; ;
Trioloetal. (1977)

Wattenberg (1972)


Wattenberg (1974)
Chouroulinkov et al.
(1967)
0.2
0.3
0.1
0.3
1.0
0.3
0.0032
63
63
70
70
63
63
70
147
147
196
287
147
105
495
147
147
196
287
147
203
495
6/9 (0.66)
9/19 (1.00)
11/20 (0.55)
13/19 (0.68)
12/12 (1.00)
8/20 (0.40)
5/81 (0.06)
0.59
0.86
0.55
0.99a
1.00
;1.00a
0.04
Gavage Studies: ,
Berenblum and
Haran (1955)
Single dose
Ei-Bayoumy (1985)
Biancifiori et al.
(1967)
Field and Roe (1965)
Single dose


Fedorenko and
Yansheva (1967)




0.054
0.375
0.14 -
0.023

0.00313
0.0313
0.05
3.6x1 Q-5
3.6x1 0"4
3.6x1 Q-3
3.6x1 0'2
3.6x1 0'1
92
92
63
56

63
63
63
76
76
76
76
76
302
93
91
161

64
64
64
146
146
146
146
146
302
302
175
301

569
569
569
578
578
578
578
578
16/19 (0.84)
16/17 (0.94)
17/20, (0.85)
5/25 (0.20)

5/35 (0.14)
17/109
(0.16)
8/17 (0.47)
0/16 (0.00)
2/26 (0.08)
5/24 (0.20)
23/30 (0.77)
23/27 (0.85)
0.68
0.60a
0.47a
0.24

0.00a
0.01a
0.09a
0.00
0.00
0.03a
0.94a
1.00a
                                    12

-------
Study


Robinson et al.
(1987)
Roe et al. (1970)
Benjamin etal.
(1988)
Dose
(mg/g
diet)
0.027

0.0125
0.11

to
(days)

63

70
71

V
(days)

119

71
97

t
(days)

274

548
211

Observed
forestomach
tumoirs
24/36 (0.67)

21/61 (0.34)
42/44 (0.95)

Predicted


0.1 6a

.0.0.1 a
0.47a

a The predicted incidence rate Hes outside the 95% confidence interval constructed from the
observed tumor response.
                                         13

-------
Table 5. Using Clement's dose-response models to predict other mouse
studies
Study
Dose
(mg/g
diet)
to
(days)
V
(days)
t
(days)
Observed
(days)
Predicted8
1 1"
Dietary Studies:
Trioloetal. (1977)
Wattenberg (1972)


Wattenberg (1974)
Chouroulinkov et
al. (1967)
0.2
0.3
0.1
0.3
1.0
0.3
0.0032
63
63
70
70
63
63
70
147
147
196
287
147
105
495
147
147
196
287
147
203
495
6/9 (0.66)
9/19 (1.00)
11/20 (0.55)
13/19 (0.68)
12/12 (1.00)
8/20 (0.40)
5/81 (0.06)
0.98"
1.00
1 .00b
1.00b
1.00
0.13b
0.03
Gavage Studies:
Berenblum and
Haran (1955)
Single dose
Ei-Bayoumy
(1985)
Biancifiori et al.
(1967)
Field and Roe
(1965)
Single dose


Fedorenko and
Yansheva(1967)



0.054
0.375
0.14
0.023

0.00313
0.0313
0.05
3.6x1 0'5
3.6x1 Q-4
3.6x1 0'3
3.6x1 0'2
92
92
63
56

63
63
63
76
76
76
76
302
93
91
161

64
64
64
146
146
146
146
302
302
175
301

569
569
569
578
578
578
578
16/19 (0.84)
16/17 (0.94)
17/20 (0.85)
5/25 (0.20)

5/35 (0.14)
17/109
(0.16)
8/17 (0.47)
0/16 (0.00)
2/26 (0.08)
5/24 (0.20)
23/30 (0.77)
1.00
0.00b
0.00b
0.01b

0.00b
0.00b
0.00b
0.00
0.00
o.oob
0.12"

NA
NA
NA
NA
NA
1.00b
NA

NA
1.00
0.1 2b
0.14

0.01 b
1.00b
1.00b
0.00
0.00
0.04b
1.00b
                                 14

-------
Study

Robinson et al.
(1987)
Roe etal. (1970)
Benjamin et al.
(1988)
Dose
(mg/g
diet)
3.6x1 0'1
0.027
0.0125
0.11
t0
(days)
76
63
70
71
t!
(days)
146
119
71
97
t
(days)
578
274
548
211
Observed
(days)
23/27 (0.85)
24/36 (0.67)
21/61 (0.34)
42/44 (0.95)
Pred
1
1.00b
0.00"
o.oob
o.oob
icted"
II
1.00b
0.07"°
0.98b
0.46"
a I and II are different approaches used to calculate predicted values under two different
assumptions:  (1) the growth rate for initiated cells during the nonexposed period is the same
as the background rate, and (2) the growth rate during the nonexposed period is the same as
that during the exposed period.
b The predicted incidence rate lies outside the 95% confidence interval constructed from the
observed tumor response.
c The data  (ages at which exposure began and was terminated)  used for this prediction differ
from that used in Table 8 of the Clement (1990a) report.  It is not clear why the difference
exists. When  age last observed is the same as age last exposed, there is no issue of cell
growth rate during the nonexposure period.
NA = not applicable.
                           5. DISCUSSION AND SUMMARY

       Information from 14 oral studies was integrated to arrive at a quantitative risk
assessment for B[a]P. Although most of the studies are not suitable for potency slope
calculation, when taken together, they do provide usable information about the dose-response
relationship of B[a]P. An important observation is that dose rate and the time elapsed since
exposure began are more relevant to tumor (papillomas in the forestomach) occurrence than
the total dose given to animals.  However, available bioassay data indicate that multiple
dosings are required to induce carcinomas in the forestomach.
       Seven potency slopes are provided in Table 3.  The two largest slope estimates  come
from the Brune et al. (1981) study in  which  animals were given B[a]P in a caffeine solution via
intubation.  Since there is some evidence that intubation may increase tumor response,  these
two slope estimates are excluded from further consideration.  The remaining five slope
                                          15

-------
estimates are remarkably similar. If a single value must be adopted, we recommend that
some form of mean slope be used.  One approach would be to take the geometric mean of
the four forestomach-based slope estimates (4.5, 9.3, 14.2, and 6.5; geometric mean = 7.9)
calculated on the basis of forestomach tumors. The potency slope based on forestomach
tumors is recommended  because it is the common tumor site observed in all 14 studies
reviewed in this report.
       In this report, an empirical model is used to fit the Neil and Rigdon data.  This model is
selected to reflect empirical evidence, rather than biological information. The linearized
multistage model, which  is also a curve-fitting procedure, is used to calculate potency slope
on the basis of the Brune et al. (1981) study.  These empirical/curve-fitting procedures are
used because there is no sufficient biological information on B[a]P to construct a biologically
based dose-response  (BBDR) model.
       The Clement model was developed by adapting a simplified version of the two-stage
carcinogenesis model  (also known as the MVK model) proposed by Moolgavkar and Vehzon
(1979) and Moolgavkar and Knudson (1981).  The Clement model  invokes an additional
assumption that the relative initiation rate (B[a]P-induced divided by spontaneous rate)  is
equal to that of the second transition (from an initiated cell to a malignant cell) rate. A
simplified MVK model  can be obtained by assuming that the second transition rate in the
exact MVK model is negligible (this implies that the rate is not dose-related). The Clement
model, by assuming identical relative transition rates, thus implies that the initiation rate is also
independent of dose.  Clearly, this implication is not reasonable because  B[a]P is known to
have initiation potential in many tissues.  In general, the use of the simplified MVK model is
not recommended for  animal data with high tumor response.  Chen and Moini (1990)
discussed in detail the problems associated with the use of the simplified MVK model.  The
Clement model is more appropriately described as a curve-fitting procedure.
                                          16

-------
                              6. OUTSTANDING ISSUES
       1.  The selection of tumor site and mouse-to-human dose conversion.
       Forestomach tumors were used to model the dose-response of B[a]P by ingestion for
several reasons: the background incidence for forestomach tumors in mice is low (a rare
occurrence), forestomach tumors are found in several studies in which different strains of mice
were used, and the studies contained more quantitative information for modeling.
       Because forestomach tumors in mice are produced as a result of contact at the site of
administration, we are concerned about the appropriateness of using this tumor site for
extrapolation to human ingestion risks. We are concerned that projecting human oral risk
from mouse forestomach tumor data is analogous to using dermal data.  Adriaenssens et al.
(1983) studied dose-response relationships for binding of B[a]P metabolites to DMA and
protein in lung, liver, and forestomach of control and butylated hydroxyanisole-treated female
A/HeJ mice. Over an oral dose range of 0.012 to 7.5 mg/mouse, no indication of saturation of
DNA B[a]P metabolite binding was observed for forestomach even though it was indicated for
lung and liver. Linear dose response for adduct formation was also observed by Pereira et al.
(1979) for mouse skin following topical application which spans the range of B[a]P doses of
0.01 to 300 fig/mouse. These  observations support the concern of possible similarity between
the reaction of forestomach and skin of mice to B[a]P. Is the use of a body surface area
scaling factor for extrapolating from mice (site of contact tumors) to humans (site unknown)
appropriate?  What is the alternative that can be explored?
       One possible alternative is to use  data from other routes of exposure.  However,
Nebert's review (1989) and earlier studies (Nebert and Jensen, 1979; Legraverend et al.,
1983) on lethality, marrow toxicity, and leukemia in Ahb/Ahb and Ahd/Ahd mice from oral
exposure to B[a]P indicated the potential for a first-pass elimination kinetics for metabolism.
Even though much more information on B[a]P can be obtained from dermal and intratracheal
studies, this information suggests that using data from inhalation or dermal exposure to derive
an oral potency estimate may be inappropriate. Gould other tumor sites be used?
       2.  Use of other tumor sites for dose-response analysis.
       In addition to forestomach tumor in mice, ingestion of B[a]P also induces tumors at
distal sites ( e.g., lung and mammary gland). Qualitatively the observation of tumors other
than forestomach strengthens the evidence that ingestion of B[a]P could present a
                                          17

-------
carcinogenic hazard for humans.  However, the cjuahtitatiye information available is
inadequate for meaningful dose-response analysis. For example, the distal tumors observed
in the Neil and Rigdon studies (1967) do not show a dose response.  Individual animal data
                                       • . .  •'--,:'• i' '•   ...
are not available for the forestomach to correct for competing risks.  Many of the other studies
are single-dose experiments frequently with high tumor incidence (approaching 100%), or the
site is associated with high background incidence  (e.g., one study showed mammary tumors
in the control group of female LEW/Mai rats to be  30%).  For these reasons no potency slope
is calculated on the basis of these tumor sites. However, it should not be interpreted that
other tumor sites are not biologically significant. On the contrary, if quantitative information is
available, it may provide insight for risk estimation. For example, Nebert and Jensen's (1979)
study of B[a]P-initiated leukemia in mice associated with allelic difference  at the  Ah locus may
contain information for such an analysis.  However, their study, as reported,.lacked the
necessary detailed information to proceed at this point.
      3. Selection of measures for dose in modeling.
       Based on a number of review articles (DiGiovanni, 1989; Graslund and Jernstrom,
1989; Thakker et al., 1988; Cooper et al.,  1983; Conney, 1982; Baird and Preuss-Schwartz,
1988), it is apparent that the metabolism of B[a]P  is well studied, and specific DMA adduct
levels have been determined in various rodent and human tissues.  B[a]P metabolites and
DNA adduct formation in the forestomach  have been studied (Adriaenssens et al., 1983;
loannou et al., 1982) in mice.  However, the strains used are different from the studies with
quantitative data for dose-response analysis.  Can other comparative, quantitative-specific
DNA adduct data be used as a measure of internal dose for modeling?
       The information reviewed thus far seems to indicate that most of the data provide
qualitative  mechanistic insight but no quantitative  measure for dose-response analysis. A
large number of studies on PAHs have used dermal assay systems. As stated  by DiGiovanni
(1989),  qualitative metabolic events associated with tumor initiation seem to be similar
between human and mouse epidermal cells.  But  it is not clear how to relate a quantitative
parameter, such as a specific DNA adduct level, to tumor formation in the two species.
       4.  Range of human susceptibility to B[a]P carcinogenesis.        •
       Calabrese (1988) compared data on induced AHH activities and DNA binding of
hydrocarbons between various tissues in humans and rodents.  In general, humans have  a
                                          18

-------
wider range of individual variation on these parameters, but the values seem to fall within the
range observed in rodents of different sensitivity.
       The results of Autrup's (1990) analysis of data from cultured human tissues and cells
versus rodent tissue/cell culture seem to indicate no qualitative difference in the metabolism of
humans compared to that of animals, but the binding of B[a]P to DNA is higher in humans.
       Nebert's (1989) review of the Ah locus and genetic differences in toxicity and cancer
highlighted the importance of genetic determinants of PAH toxicity and carcinogenicity.  The
Ah locus, localized to mouse chromosome 12, encodes a cytosolic receptor which regulates
cytochrome P1-450 induction that mediates hydrocarbon metabolism. In Nebert and Jensen's
experiment (1979), mice that were homozygous for the high-affinity Ah receptor (with alleles
Ahb/Ahb or an Ah  responsive phenotype), when exposed to B[a]P orally, developed no
leukemia. Mice that were homozygous for the low-affinity Ah receptor (with alleles Ahd/Ahd or
an Ah nonresponsive phenotype), when similarly exposed, developed leukemia. An estimated
oral dose of 120 mg/kg-day for 10 days is lethal to the (Ah"/Ahd) homozygote (100%).  Time to
death was shortened as dose increased. The responsive strains were unaffected for up to 6
months of exposure. Nonresponsive mice also developed marrow toxicity from oral B[a]P.
Nebert suggest that these data, taken together, are consistent with "first-pass elimination"
kinetics.  Other experiments indicated that responsive mice develop site-of-contact tumors
when exposed dermally or orally, while nonresponsive mice  develop thymoma/leukemia.
What are the contributions of the Ah locus-mediated pharmacokinetic factors and target organ
susceptibility in the expression of carcinogenicity?
       What is the range of human susceptibility to B[a]P carcinogenicity and other toxicities
when ingested?  Humans are polymorphic with respect to the Ah locus.  Some human studies
(Kellermann et al., 1978) have implicated the involvement of the  Ah locus and human lung
cancers.  Which animal models can be used to bound the range of human susceptibilities?
       5.  Duration of exposure and concurrent exposure to  dietary PAHs and other factors.
       Thus far, this analysis raises questions about the pattern  of exposure on the effect of
B[a]P-induced tumor progression: the effect of terminating exposure, the effect of caffeine
intake, and the effect of loading dose (e.g., of B[a]P intake by gavage or intubation). These
patterns can potentially be  mimicked by humans through differences in lifestyle.
       Dietary ingestion constitutes the major route of human exposure to PAHs.  Estimates
by Santodonato et al. (1981) suggest that dietary exposure could be more than 100 times
                                          19

-------
higher than drinking water as a source of human exposure.  What is the site of carcinogenic
action of PAHs on humans when ingested? Humans are exposed to PAHs as a group of
mixtures with variable composition.  Skin painting studies have indicated interactions. A study
by Robinson et al. (1987) in A/J mice of particulates of a coal tar paint used in potable water
systems suggested the potential for interaction via the oral route.  Should we set standards
using data from individual PAHs one at a time?

         7. CONTINUING RISK ASSESSMENT WORK AND RESEARCH NEEDS

       This quantitative risk assessment of B[a]P focuses on the use of available
route-specific information to estimate potential human carcinogenic risks.  Because of the
multiple mechanisms in which B[a]P can induce cancers, and because of the genetically
determined detoxification/toxification pathways that influence the  ultimate site and magnitude
of carcinogenic action, this analysis has brought to light the need for a more comprehensive
analysis.
       Since other PAHs found in drinking water have a more limited data base than B[a]P,
the Agency needs to develop other approaches (e.g., a relative potency approach) to assess
the risks of PAHs other than B[a]P. Some preliminary work has  been attempted (Chu and
Chen, 1984; Clement, 1990a, 1990b; Richard and Woo, 1990), however, validation of any
relative potency approach is needed before adaptation. This opinion was also expressed by
the panelists at the 1988 EPA-sponsored workshop on  PAHs. It became apparent during the
literature search for preparing this report that much information is available on
3-methylcholanthrene (3-MC) and dimethylbenzanthracene (DMBA).  Even though these two
compounds are not included in the  Drinking Water Criteria Document for Polycyclic Aromatic
Hydrocarbons, evaluations of these compounds can serve to validate relative potency
approaches.
        This is not to underscore the fact that a key limitation of the data is the lack of quality
bioassay and biological/toxicological data which can be used for dose-response analysis.
 Research that focuses on providing such data may be needed.  The Electric Power Research
 Institute is in the process of conducting bioassays by ingestion of manufactured gas plant
 residues using B[a]P as a positive control focusing on providing  data for dose-response
 analysis. When the results of their study become available (most likely in about 5 years) it will

                                          20

-------
provide additional dose-response information, which coupled to a comprehensive analysis of
the large volume of mechanistic information, will help to reduce the uncertainties of
extrapolating from animals to humans.
       From the viewpoint of dose-response  modeling for B[a]P, a carcinogenesis bioassay
(e.g., one that includes the study of the number and size of papiHomas and carcinomas over
time with serial sacrifice) can be useful for constructing a dose-response model that reflects
the dynamics of tumor formation and progression.  Such a bioassay, when coupled with
quantitative pharmacokinetic and biological information, would be useful for improving low-
dose extrapolation. Furthermore, comparative quantitative metabolic and pharmacokinetic
data on B[a]P between test species and humans by ingestion would be particularly useful.
       jf  B[a]P were present in drinking water at 0.03 iig/L, the highest detection limit value
reported in the National Organic Monitoring Survey (U.S. EPA, 1989), daily human exposure
to B[a]P from water would be several orders of magnitude  lower than levels used in
bioassays.  To probe the dose-response characteristics closer to levels of human exposure,
molecular parameters linkable to tumor formation are needed; the use of specific DNA
adducts has the potential. Using the enzyme-linked immunosorbent assay (ELISA) technique,
increased specific DNA adduct levels have been identified  in peripheral white blood cells of
human volunteers ingesting charcoal-broiled beef (Rothman et a!., 1990).
       Autrup's (1990) commentary on carcinogen metabolism in cultured human tissues and
cells suggested a weak association between the carcinogenic potency of PAHs in
experimental animals and the level of PAH-binding to human bronchial DNA. The major
B[a]P-DNA adducts were the same in all cultured tissues and cells from  hurnans and
experimental animals, and the same metabolic profile was  seen when rodent and human
tissues were incubated with  B[a]P.  The effects of diet on levels of PAH-DNA adducts were
measured (Kwei and  Bjeldanes, 1990; O'Neill et al., 1990a, b). The composition of the diet
seems to modulate the type  and level of specific adducts formed. For in vivo PAH-DNA
adduct data to be useful, factors, such as  diet as a modulating factor, has to be the same for
carcinogenesis studies.
                                         21

-------
                                    APPENDIX A
                                 OTHER B[a]P DATA

      The following data are extracted from B[a]P bioassay studies and non-B[a]P bioassay
studies (positive control groups).
1. Trioloetal., 1977.
      Nine-week-old Ha/ICR mice were fed B[a]P for 12 weeks and sacrificed.
B[a]P
(mg/g diet)
0
0.2
0.3
Estimated
dose
(mg/kg/day)
0.0
27.3
41.0
Proportion of mice
with forestomach
tumors (%)
0/9(0%)
6/9(66.6%)
9/9(100%)
2. Wattenberg, 1972.
       Ten-week-old Ha/ICR mice were fed B[a]P in the diet.
B[a]P
(mg/g diet)
0.1
0.3
1.0
Estimated
dose
(mg/kg/day)
14.3
41.0
Weeks in
study
18
20
12.0
Proportion of mice
with forestomach
tumors (%)
1 1/20 (55%)
13/19 (68%)
12/12 (100%)
 3.  Ei-Bayoumy, 1985.
       Nine-week-old CD-1 mice were given 1 mg of B[a]P by gavage two times/week for 4
 weeks and then followed for 12 weeks.  Proportion of mice with forestomach tumors: 17/20
 (0.85%).

 4.  Biancifiori et al., 1967.
       A positive control in a study of the action of oestrone in intact and ovariectomized
 BALB/c/CB/SE mice. B[a]P or DMBA dissolved in almond oil at a concentration of 0.5%: 0.1

                                          22

-------
ml_ (0.5 mg) was administered by stomach tube twice weekly for 15 weeks when mice
reached 8 weeks of age.
    Results in the B[aJP group: 5 out of 25 mice had squamous carcinomas of the
forestomach. Tumors occurred 28 to 65 weeks post-treatment. Mean age was about 43
weeks.

5. Berenblum and  Haran, 1955.
       Age study began: about 3 months old.
       Study 1: 0.3 mL of 0.5% of B[a]P in polyethylene glycol-400 was given by stomach
tube to C3H male mice weekly for 30 weeks.
                                 Papillomas
                                Carcinomas
 1/19
15/19
       Study 2: Single dose was given to male mice as in Study 1 above.
                                Papillomas
                                Carcinomas
16/17
 0/17
6. Field and Roe, 1965.
      The following data are taken from positive control groups in a study that investigated
the effect of citrus oils on forestomach tumors. A single dose of B[a]P was given by stomach
tube to Albino mice.  Only mice that survived 60 days were included in the following table.
Animals were killed at 569 days.
                                    Forestomach*
               ,     Dose (|ig)                  (No. of tumors/animal)
                      12.5
                      50.0
                     200.0
                    all papillomas.
     5/17
     17/109
     8/17
                                        23

-------
7. Wattenberg, 1974.
       This is a positive control group in a bioassay studying the effect of sulfur-containing
compounds on B[a]P-induced neoplasms of the forestomach in Ha/ICR mice.
       Nine-week-old mice were given 0.3 mg/g of B[a]P in the diet for 6 weeks and were
sacrificed after 20 weeks in the study. Eight out of 20 animals had forestomach tumors.

8. Fedorenkoetal.,  1967
       B[a]P was introduced directly (by syringe probe)  into the stomach of animals (2- to 3-
month-old white mice) weekly for 10 weeks in 0.2  mL B[a]P solution in triethyleneglycol. The
experiment lasted for 19 months.
       Tumors in the antrum of the stomach (the denominator is the number of animals at risk
at the occurrence of the first tumor in the study) are given below.
Dose (mg) for
1 0 times
0.001
0.01
0.1
1.0
10.0
Carcinomas
0/16
0/26
0/24
11/30
16/27
Papillomas
0/16
2/26
5/24
12/30
7/27
 9.  Robinson etal., 1987.
       This is a positive control in a coal tar paint study. Animals (8-week-old A/J mice) were
 treated with B[a]P for a total of 6 mg in 8 weeks and sacrificed at 9 months of age.
       Dose (mg/mouse)
        total dose (mg)
            Forestomach
Papillomas  carcinomas    Total
Lung
0
6
0/36
24/36
0/36
22/36
33/36
10/36
22/36
                                          24

-------
10. Roeetal., 1970.
       Fifty jig of B[a]P in 0.2 mL of PEG-400 was given by intragastric instillation to female
Swiss Albino mice (9 to 14 weeks of age). Tumor incidence for animals examined at autopsy
at 18 months is given below.
Dose (ng)
(single)
0
50
Forestomach
Pap Car
2/65
20/61
0/65
1/61
Lung
15/65
18/61
Hepatomas Maiignant
lymphomas
5/65
9/61
3/65
0/61
11. Benjamin etal., 1988.
       These are two positive controls in a study of reduction of B[a]P-induced forestomach
neoplasms in ICR mice given nitrite and dietary soy sauce. Mice were given eight doses of
B[a]P (1.5 mg, twice weekly for 4 weeks) starting at age 71 days and killed at 211 days of
age.
       Exp. 1:  23/24 (96% or 5.2+-0.7 neoplasms/mouse).
       Exp. 2:  19/20 (95% or 5.8+-1.1 neoplasms/mouse).
                                         25

-------
                                    APPENDIX B
      DERIVATION OF DOSE-RESPONSE MODEL TO FIT NEIL AND RIGDON DATA

       An empirical model used to fit data from Neil and Rigdon (1967) is presented below.
This model is selected to reflect empirical evidence rather than biological information. There
is no sufficient biological information to construct a biologically based dose-response model.

Derivation of Empirical Dose-Response Model:
       Assume that the hazard rate for tumor occurrence is given by

                                  h(t) = h0(t) + Mt),

where h0(t) = atk"1 is the background rate of tumor occurrence and the dose-induced hazard
rate is
                                        t
                               h,(l)=   /  (t-sf2 D(s)ds,
                                       0
where
    D(s)-f(d)
        = a,d + a2d2 + ... + amdm for some positive integer m
            if to < s < t,, and
    D(s) = 0 otherwise: d is dose given to animal at time s.

The dose-related hazard function h^t) implies that the earlier B[a]P exposure contributes more
to the hazard rate than the later exposure. This assumption is consistent with the observation
that a single dose is sufficient to induce tumors if the follow-up period is long enough.
                                         26

-------
         The cumulative hazard by time T for background is given by
                            Hn(t) = /  at Mdt
        The cumulative hazard by time T>^ for the B[a]P-induced hazard is given by
            T
  H.fT) =   .h
           0
            ^       -• tj"          T

        = J'h1(t)dt+ /  hi(t)dtt'+ I  h^tjdt
           0 ".to     .      t,
                                      T  t
        = f(d)       (t-s)k-2dsdt + f(d)       (t-s)k-2ds dt
                 to                  t, to
        Therefore, the probability of getting cancer by time t given that animals are exposed
to B[a]P only during the time interval [t, t  has the form
                    P(d,t) = 1 - expHdo + q,d + .:. + qmdm)[(t-gk - (t-
A quadratic model with m = 2 is adequate to fit the Neil and Rigdon data.
                                           27

-------
                                    REFERENCES
Adriaenssens, P.I.; White, C.M.; Anderson M.W. (1983) Dose-response relationships for the
binding of benzo[a]pyrene metabolites to DNA and protein in lung, liver, and forestomach of
control and butylated hydroxyanisole-treated mice. Cancer Res. 43:3712-3719.

Autrup, H. (1990) Carcinogen metabolism in cultured human tissues and cells. Carcinogenesis
11:707-712.

Baird, W.M.; Pruess-Schwartz, D. (1988) Polycyclic aromatic hydrocarbon-DNA adducts and
their analysis: a powerful technique for characterization of pathways of metabolic activation of
hydrocarbons to ultimate carcinogenic metabolites. In: Yang, S.K.; Silverman, B.D., eds.
Polycyclic aromatic hydrocarbon carcinogenesis: structure-activity relationships, Vol. 2. Boca
Ratan, FL: CRC Press, Inc., pp. 141-206.

Benjamin, H.; Storkson, J.; Tallsa, P.; Pariza, M. (1988) Reduction of benzo[a]pyrene-induced
forestomach neoplasms in mice given nitrite and dietary soy sauce. Food Chern. Toxicol.
26(8):671-678.

Berenblum, I.; Haran, N. (1955) The influence of croton oil and of polyethylene glycoi-400 on
carcinogenesis in the forestomach of the mouse. Cancer Res. 15:510-516.

Biancifiori, C.; Caschera, F.; Giornelli-Santilli, F.; Bucciarelli,  E. (1967) The action of oestrone
and four chemical carcinoges in intact and  ovariectomized BALB/c/Cb/Se mice. Br. J. Cancer
21:452-459.

Brune, H.; Deutsch-Wenzel, R.; Habs, M.; Ivankovic, S.; Schmahl, D. (1981) Investigation of
the tumorigenic response to benzo[a]pyrene in aqueous caffeine solution applied orally to
Sprague-Dawley rats.  J. Cancer Res. Clin. Oncol. 102:153-157.

Calabrese, E.J. (1988) Comparative biology of test species.  Environ. Health Perspect. 77:55-
62.

Chen, C.; Moini, A.  (1990) Cancer dose-response models incorporating clonal expansion. In:
Moolgavkar,  S., ed. Scientific issues in quantitative cancer risk assessment. Boston, MA:
Birkhauser, pp 153-175.

Chouroulinkov, I.; Gentil, A.; Guerin, M. (1967) Study of the  carcinogenic activity of 9,10-
dimethyl-benzanthracene and of 3,4-benzopyrene given orally. Bull. Cancer 54(1):67-68.

Chu, M.M.L.; Chen,  C.W. (1984) Evaluation and estimation of potential carcinogenic risks of
polynuclear aromatic hydrocarbons.  Presented at the Symposium on Polynuclear Aromatic
Hydrocarbons in the Workplace. 1984 International Chemical Congress of Pacific Basin
Societies. Available from: NTIS, Springfield, VA, PB89-22/329.
                                           28

-------
Clement International Corporation (1990a) Ingestion dose-response model for benzo[a]pyrene.
Prepared by Clement Internation Corporation, Fairfax, VA, for the Office of Health and
Environmental Assessment, U.S. Environmental Protection Agency, Washington, DC, under
EPA contract no. 68-02-4601.

Clement International Corporation (1990b) Development of relative potency estimates for
PAHs and hydrocarbon combustion product fractions compared to benzo[a]pyrene and their
use in carcinogenic risk assessments. Prepared by Clement International Corporation,
Fairfax, VA, for the Office of Health and Environmental Assessment, U.S. Environmental
Protection Agency, Washington, DC, under EPA contract no. 68-02-4601.

Conney, A.H. (1982) Induction of microsomal enzymes by foreign chemicals and
carcinogenesis by polycyclic aromatic hydrocarbons. G.H.A. Clowes Memorial Lecture. Cancer
Res. 42:4875-4917.

Cooper, C.S.; Grover, P.L.; Sims, P. (1983) The metabolism and activation of benzo[a]pyrene.
Prog. Drug Metab. 7:295-396.

DiGiovanni, J. (1989) Metabolism of polycyclic aromatic hydrocarbons and phorbol esters by
mouse skin: relevance to mechanism of action and trans-species/strain carcinogenesis. Prog.
Clin. Biol. Res. 298:167-199.

Ei-Bayoumy, K. (1985) Effects of organoselenium compounds on induction of mouse
forestomach tumors by benzo[a]pyrene. Cancer Res. 45:3631-3635.

Fedorenko, Z.; Yansheva, N. (1967) Experimental reproduction of tumors of the antral part of
the stomach in mice by administration of various dose of 3,4-benzpyrene. Hygiene Sanitation
32(5):168-173.

Field E.; Roe, F. (1965) Tumor promotion in the forestomach epithelium of mice by oral
administration of citrus oils. J. Natl. Cancer Inst. 35(5):775-784.

Gaylor, D.;  Kodell, R. (1980) Linear extrapolation algorithm for low dose risk assessment of
toxic substances. J. Environ. Pathol. Toxicol. 4:305-312.

Grasiund, A.; Jernstrom, B. (1989) DNA-carcinogen interaction: covalent DNA-adducts of
benzo[a]pyrene 7,8-dihydrodiol 9,10-epoxides studied by biochemical and biophysical
techniques. Q. Rev. Biophys. 22:1-37.

loannou, Y.M.; Wilson, A.G.E.; Anderson, M.W. (1982)  Effect of butylated hydroxyanisole,  a-
Angelica Lactone, and 6-naphthoflavone on benzo[a]pyrene: DNA adduct formation in vivo in
the forestomach, lung, and liver of mice.  Cancer Res. 42:1199-1204.

International Agency for Research on Cancer (IARC) (1983) Monographs on the Evaluation of
the Carcinogenic Risk of Chemicals for Humans.  Polynuclear Aromatic Compounds. Part  I.
Chemical, Environmental, and Experimental Data, Vol. 32. Lyon, France: IARC.
                                         29

-------
Kellermann, G.; Luyten-Kellermann, M.; Jett, J.;- Moses, H.; Fontana, R. (1978) Aryl
hydrocarbon hydroxylase in man and lung cancer.  In: Human genetic variation in response to
medical and environmental agents: pharmacogenetics and ecogenetics.  Hum. Genet. Suppl.
1:161-168.

Krewski, D.; Murdoch, D.; Dewanji, A. (1986) Statistical modeling and extrapolation of
carcinogenesis data. In: Moolgavkar, S.; Prentice, R., eds. Modern statistical methods in
chronic disease epidemiology. New York, NY: Wiley-lnterscience,  pp.259-282.

Krewski, D.; Gaylor, D.; Szyszkowicz, M. (1991) A model-free approach to low dose
extrapolation. Environ. Health Perspect.  90:279-285.

Kwei, G.Y.; Bjeldanes, L.F.  (1990) Stimulation of binding of benzo[a]pyrene metabolites to
DNA by diet-induced peroxidative stress. Food Chem. Toxicol. 28(7) :491-495.

Legraverend, C.; Harrison, D.E.; Ruscetti, F.W.; Nebert, D.W. (1983) Bone marrow toxicity
induced by oral benzo[a]pyrene: protection resides at the level of the intestine and liver.
Toxicol. Appl. Pharmacol. 70:390-401.

McKay, S.; Hulbert, P.B.; Grover, P.L. (1988) Mechanisms of metabolic activation involving
epoxides: the possible role  of phenolic hydroxyl groups. In: Yang, S.K.; Silverman, B.D., eds.
Polycyclic aromatic hydrocarbon carcinogenesis: structure-activity relationships, Vol. 2. Boca
Raton, FL: CRC Press, Inc., pp. 1-109.

Moolgavkar, S.; Venzon, D. (1979) Two event model for carcinogenesis: incidence curves for
childhood and adult tumors. Math. Biosci. 47:55-77.

Moolgavkar, S.; Knudson, A. (1981) Mutation and cancer: a model for human carcinogenesis.
J. Natl. Cancer Inst. 66:1037-1052.

National Research Council  (NRC) (1983) Risk Assessment in the Federal Government:
Managing the process. Washington,  DC. Ndational Academy Press.

Nebert, D.W. (1988) The 1986 Bernard  B. Brodie Award lecture: the genetic regulation of drug
metabolizing enzymes. Drug Metab.  Dispos. 16:1-8.

Nebert, D.W. (1989) The Ah locus: genetic differences in toxicity: cancer, mutation and birth
defects. Crit. Rev. Toxicol. 20:153-170.

Nebert, D.W.; Jensen, N.M. (1979) Benzo[a]pyrene-initiated leukemia in mice: association with
allelic differences at the Ah locus. Biochem. Pharmacol. 27:149-151.

Neil J.; Rigdon, H. (1967) Gastric tumors in  mice fed benzo[a]pyrene: a quantitative study.
Texas Reports on Biology and Medicine 25(4): 553-557.
                                          30

-------
O'Neill, I.K.; Bingham, S.; Povey, A.C.; Brouet, I.; Bereziat, J.-C. (1990a) Modulating effects in
human diets of dietary fibre and beef, and of time and dose on the reactive microcapsule
trapping of benzo[a]pyrene metabolites in the rat gastrointestinal tract. Carinogenesis
11(4):599-607.

O'Neill, I.K.; Povey, A.C.; Bingham, S.; Cardis, E. (1990b) Systematic modulation by human
diet levels of dietary fibre and beef on metabolism and disposition of benzo[a]pyrene in the
gastrointestinal tract of Fischer F344 rats.

Pereira, M.A.;  Burns, F.J.; Albert, R.E. (1979) Dose response for benzofajpyrene adducts in
mouse epidermal DNA.  Cancer Res. 39:2556-2559.

Richard, A.M.; Woo, Y. (1990) A CASE-SAR analysis of polycyclic aromatic hydrocarbon
carcinogenicity. Mutat. Res. 242:285-303.

Robinson, M.;  Laurie, R.; Bull, R.; Stober, J. (1987)  Carcinogenic effects in A/J mice of
particulates of coal tar paint used in potable water systems. Cancer Lett. 34:49-54.

Roe, F.; Levy, L.; Carter, R. (1970) Feeding studies on sodium cyclamate, saccharin and
sucrose for carcinogenic and tumor-promoting activity. Food Cosmet. Toxicol. 8:135-145.

Rothman, N.; Poirier, M.C.; Baser, M.E.; Hansen, J.A.; Gentile, C.; Bowman, E.D.; Strickland,
P.T. (1990) Short communication: formation of polycyclic aromatic hydrocarbon-DNA adducts
in peripheral white blood cells during consumption of charcoal-broiled beef. Carcinogenesis
11:1241-1243.                                        -      .'       '     .

Santodonato, J.; Howard, P.; Basu, D. (1981) Environmental sampling,  levels, sources, fates
and human exposure. J. Environ. Pathol. Toxicol. 5:165.

Thakker, D.R.; Levin, W.; Wood, A.W.; Conney, A.M.; Yagi, H.; Jerina, D.M. (1988)
Stereoselective biotransformation of polycyclic aromatic hydrocarbons to ultimate carcinogens.
In: Wainer, I.W.;  Drayer, D.E., eds. Drug stereochemistry: Analytical Methods and
Pharmacology. New York, NY:  Marcell Deker, Inc., pp. 271-296.

Triolo, A.; Aponte, G.; Herr, D. (1977) Induction of aryl hydrocarbon hydroxylase and
forestomach tumors by benzo[a]pyrene. Cancer Res. 37:3018-3021.

U.S. Environmental  Protection Agency (EPA) (1986) Guidelines for Carcinogen Risk
Assessment. Federal Register 51:33992-34003.

Wattenberg, L. (1972) Inhibition of carcinogenic and toxic effects of polycyclic hydrocarbons
by phenolic antioxidants and ethoxyquin. J. Natl. Cancer Inst. 48(5):1425-1430.

Wattenberg, L. (1974) Inhibition of carcinogenic and toxic effects of polycyclic hydrocarbons
by several sulfur-containing compounds. J.  Natl. Cancer Inst. 52(5):1583-1587.
                                          31
                                                  •&U.S. GOVERNMENT PRINTING OFFICE: 1993 - 750-002/60135

-------

-------

-------
ts
Official Bus
Penalty fo
00
O o m c:
5' § < » '
  —• CD

  18
  CD


  CO
  CD
o^s.
X<. £.

32-*
(O   O
O> 3 en
CO CD o


   if
   33 >
   CD CO
   CO CD
   CD 3
   SB O
CO
5T

en

         Q. =5
         Q -,
         o ffl
         3^
         o°
         3 w
         $ o
         e x
          D
           -o
           m
           3)
      -a
      §
      i?
           p •
           9
           CO
           CJ1

-------