CARCINOGENIC RISK ASSESSMENTS
OF
POLYCHLORINATED BIPHENYLS (PCBs)
Executive Summary
This document presents the HERD risk assessment for
carcinogenicity of PCBs and summarizes the results of four
previous PCS risk assessments for cancer conducted by FDA, OTA,
and CAG/SPA, and OTS. (Unfortunately, no consolidated
assessment can be developed from these sources, because of the
different units and different techniques used. However, the
results from these various risk assessments are consistent,
though not directly comparable).
In light of these earlier assessments we chose studies by
NCI (three positive dose levels) and Kimbrough (one pusitive
dose level), respectively, from which to extrapolate
carcinogenic risks at low exposures. The NCI study has been
criticized by other authors as having too few animals in each
sex-dose group; nowever, the study was used because no
consistent differences or trends in responses by sex were found
-------
and with pooled data for males and females it allows the best
characterization of dose-response, unlike the Kimbrough study
with only one dose level. Cur results are within a factor of
three of the FDA risk assessment of carcinogenicity of PCS,
which used both Kimbrough's data and the NCI data.
Estimates of human exposure come from the Exposure
Evaluation Division of OTS. Other relevant features of the
risk assessment are : 1) The extrapolation distance between the
lowest experimental dose (1.25 rag/kg/day) in the NCI study and
many of the exposure estimates can be very large. The larger
the distance the less confidence one might have in the
assessment. 2) The dose-response data for total malignanciess
nre also linear, this corresponds well with the "linearized"
upper 95% confidence limits from' the GAG model.
For regulatory purposes and for the purposes of comparing
risks of PCBs with those for other chemicals, it is suggested
that the number from the 95% upper confidence bound found in
the multistage model in that table be used. This model is
routinely used by CAG to set air and water quality criteria and
standards. The. estimates of exposure for which carcinogenic
risks are calculated are based on conservative assumptions.
Few, if any, individuals will be exposed to these relatively
large amounts, of PCBs. The carcinogenic risks calculated for
-------
typical members of exposed groups are believed to be no higher
and probably much lower than those presented in this document.
Previous Quantitative cancer Risk Assessment
This memorandum summarizes and pulls together for the
first time the information available from previous quantitative
risk assessments of PCBs, and updates and fills in gaps in that
area. It concludes with on an HERD risk assessment of PCBs.
The first risk assessment to be summarized is that
conducted by the EPA Carcinogen Assessment Group (CAG) for the
Office of Water Regulations and Standards, Criteria and
Standards Division (EPA, 1980) . This risk assessment was used
to set an ambient water quality criteria level as shown in
Table 1. (CAG Table page C-84). In Table 1 are the estimated-
concentrations of PCBs (or virtually safe doses) corresponding
to lifetime cancer risk levels of 10"~7, 10"^, and 10~5. These
concentrations are, as CAG said, "exceedingly low." The cancer
risk levels stand for additional incidence of cancer in an
exposed population. CAG used the "linearized" multistage model
( Crump, 1931) which is linear at low doses or exposures so that
the additional lifetime risk is directly proportional to PCB
intake. CAG's risk assessment considered only the
bioaccumulation of PCBs in fish and shellfish, plus exposure
-------
through drinking water. Exposures from other food sources,
air, or occupational exposures were not included in the
criteria level derived by GAG using Crump's model.
For experimental data, CAG chose the study by Kimbrough et
al. (1975) of Sherman rats, although CAG indicated that rats
appear to be much less sensitive to the acute and subacute
effects of PCBs than man or non-human primates. In 184 rats
treated with 100 ppm PCS in their diet, Kimbrough observed 26
hepatocellular carcinomas compared to 1/173 hepatocullar
carcinomas in controls. CAG considered all of the mouse
studies of PCBs unsuitable for a quantitative risk assessment
of cancer because none of them involved feeding for most or all
of an animal lifetime. They also criticized the NCI study of
PCBs (NCI, 1978), the only other long-term major study of PCB
exposure that suggested a carcinogenic effect, as using too few
experimental animals to establish a basis for assessming
carcinogenic risk.
Crump conducted the PCB risk assessment (conversation with
Qiao Chen, November 18, 1982) but documentation of some details
is incomplete. He combined the Kimbrough animal data for
hepatocellular carcinoma and liver neoplastic nodules, the two
tumor types observed in the Kimbrough study. At control the
proportion responding was 1/173, but at the high dose the
proportion responding was 170/184. Crump assumed that the
-------
average weight of a test rat was 400 grams resulting in a
conversion from 100 ppm to 4.42 mg/kg/day. ( CAG normally
assumes rats weigh 350 grams).
The formula for calculating VSD's for humans is then
VSD = 70
q2 x (2 + CT. OOb^kg x 46,000)
where 2 is a daily consumption of 2 liters of drinking water,
6. 5g is an assumed daily consumption of fish and shellfish,
46,000 is the bioconcentration factor (because of the size of
this factor all of risk comes from comsumption of fish and
shellfish, not water), 10"^ is the presumed acceptable lifetime
risk, and 70kg is.the weight of a human. The above formula
appears on p.353 of the Friday, November 28, 1980, Federal
Register EPA Water Quality Criteria Document; Availability
(L'.S. EPA 1980), and q2 is the potency calculated from Crump's
model using 1/173 at 0 mg/kg/day and 170/184 at 4.42
n;g/kg/day. Thus, q2 for rats = .69 and qi. for humans is
achieved by multiplying by a species conversion of 5.85. This
gives 4.01. Thus. VSD = 5.8 x 10~7 mg/1 or 0.58ng/l. This is
close to .79ng/l, the value derived by CAG for 10~5 risj< in
(Table 1). The small difference between CAGs risk assessment
-------
and our reconstruction of it is unexplained, perhaps Crump made
an adjustment for the number of days in the experiment (730)
versus the numer of days the animals were exposed (645) .
CAG noted several drawbacks of the Kinbrough study. Since
it had o.nly one dose level it could not provide any evidence of
the shape of the -dose-response curve and it tested one species,
one sex, and only one commercial mixture of PCBs. CAG added
that Kimbrough's experimental design was good in that the
dosing was for a large proportion of the liftapan of the
animals, the food route was appropriate, and the dose over time
was uniform. Also, there was good documentation of the intake
dose, a sufficient number of test animals were used for
statistical tests of increases in tumors, and a thorough
description of the pathology was provided. Finally, CAG noted
that an "acceptable no carcinogenic level" could be established
with greater certainty .if better quantitative data on
c^arcinogenicity had been available. They also felt that
studies with larger numbers of animals designed to measure
relatively small effects were needed.
The second risk assessment to be summarized is that
conducted by the PCB Risk Assessment Work Force of the Food and
Drug Administration ( Cordell et al., 1982). This group used
both the Kimbrough (1975) data and the NCI (1978) data to
estimate carcinogenic risk to humans from consumption of fish
-------
contaminated with PCBs. FDA used linear extrapolation from
high doses in the animal studies to the lower exposures of
consumers of PCB contaminated fish and water. Linear
extrapolation was defended by the authors on the grounds that
one cannot deduce or directly observe the shape of the low end
of the dose response curve with precision, and among the
available methods, linear extrapolation from high to low doses
appears to be consistent with what is known about the
biological mechanisms of carcinogenesis as well as least likely
to underestimate cancer risk. They also point out that linear
extrapolation as used by Qrump is the limiting case (upper
confidence limit on risic) on his multistage model at low
doses. Hence, "linearized" multistage model (Crump, 1981).
"Instead of using the entire Crump's model, however, the authors
placed 99%" upper confidence bounds on the animal response..data
to control, fdr.the effect of sample size so. that they could
make comparisons between experiments. Use of upper bounds on
risk along with linear extrapolation involved an additional
degree of conservatism in their risk assessment.
An additional element of uncertainty was brought out by
the FDA that was not considered in the CAG assessment. That
is, there is an absence of toxicity data on the particular set
of PCBs that occur as residues in fish. Due to environmental
transformation, the chemical composition of PCB residues found
in fish differs from that of industrial PCB products (e.g.,
-------
ArocLor 1254 and 1260) , although a typical PC3 residue in fish
resembles the Aroclor 1254 mixture more closely that it does
the other Aroclors. For this reason it was uncertain that the
available tqxicity data from Kimbrough and MCI represented the
toxicity of the PC3 mixture ingested by humans who consume
fish.
FDA's upper confidence limits, on risk were thus computed
from the NCI Aroclor 1254 Fisher 344 rat data, for both sexes
combined, on total malignancies and liver carcinomas plus
adenomas. These data are shown in Table 2 (FDA Table 5).
However, estimated risks from the one sex Sherman rat data on
Aroclor 1260 from the Kimbrough study were also computed.
Because the relative susceptibilities of humans and
experimental animals to the carcinogenic effects of PCBs were
not known, FDA noted .that the risk estimates could obviously
severely over-estimate or under-estimate human risks. The FDA
risk estimates from the NCI and Kimbrough data, however,
demonstrated remarkable agreement showing that the various rat
strains used reacted similarly to PCS carcinogenic insult.
These risks are presented as rates per 100,000 in Table 3 (FDA
Table 6) for various tolerances or "acceptable: d'aily intakes"
of PCBs and percentiles of fish-eating populations. Exposures
in ng/1 or ppb are not directly given. Therefore it is
difficult to compare the result of this risk assessment with
that of CAG's (already discussed) or that of OTA's (to be
-------
discussed). Table 4 (FDA Table 7) was derived by assuming that
the carcinogenic risk is evenly distributed over a 70-year
period and by multiplication of the size of the fish-eating
population at risk times the risk estimates shown in Table 2.
FDA also noted that our Table 3 (from their Table 6) may
underestimate- risks to neonates who receive PCBs in mother's
milk, but who also continue to consume food contaminated with
PCBs after childhood and throughout their lives.
The third risk assessment uo be summarized was prepared
for the U.S. Congressional Office of Technology Assessment by
Kenneth Crump, who was involved in the first risk assessment
discussed here ( Crump and Masterman, 1979). Crump selected
three data sets to use in his computation of virtually safe
doses by extrapolation of the results of- animal experiments to
low dose levels. These', data sets are shown in Table 5 (OTA
Table 11). Crump's rationale for selection of data sets as the
basis for risk extrapolation was as follows.' First, Crump
agreed with CAG in not using the NCI 1978 study was not used
because it involved relatively few animals (only 24 per dose
group, whereas the usual NCI bioassay uses 50 (NCI, 1978)).
Second, the Kimbrough study was used because it provided the
most convincing evidence of the carcinogenicity of PCBs
(Kimbrough et al., 1975). The study involved a relatively
x
large number of animals and the increased incidence of
hepatocellular carcinomas in the treated group was highly
-------
significant. This study was used with two endpoints,
hepatocellular carcinomas and liver neoplastic nodules. Third,
Crump used the Industrial Biotest study (Industrial Biotest
Laboratories, 1971 from an unpublished report. Reasons for
this choice are not clear (Some of the data from this testing
.company have been found to be fradulent). The slides from this
"study were examined at least .twice, with markedly different
results. Even so,.the rediagnosis of the liver pathology had
indicated a significant tumorigenic effect. There was also
unusually high mortality and interim sacrifices further reduced
the numbers of animals on test.
He then selected several high-to-low dose extrapolation
procedures: the probit moilel (Mantel, et al. , 1975), the one-
stage or one-hit model (Crump et al.., 1977) and the multistage
model (Crump, 1981). The gamma multihit model (Rai and Van
Ryzin, 1978) was also considered but could not be used : it
requires data with at least two positive experimental doses and
thus could not be applied to either of the Kimbrough and
datasets, and when the computer algorithm for computing the
estimates failed to converge for the third data set (Industrial
Biotest) .
Table 6 (OTA Table 12) presents the computed maximum
likelihood estimates in ppb of the lifetime dose required to
produce additional risks over background of 10"^, 10"^, and
10
-------
10"5. Recall that GAG considered risk levels of 10~7, 10~6,
and 10"5 for its virtually safe doses (Table 1). Lower 95%
confidence bounds for each of these virtually safe doses were
also computed from various models. Likewise, as mentioned
earlier, the gamma multihit model was applied only to the third
data set since this model could not be applied to data
containing only one experimental dose." In the case of one
experimental dose, however, the one-hit model and multistage
model produced identical results.
Crump believed that the Kimbrough study provided the most
convincing evidence of the carcinpgenicity of PCBs, he used
these data to calculate risks for estimated PCS exposures. The
fact that only one dose level of PCB was tested in the
Kimbrough study meant that the .one-hit model was the -only one
that could be utilized. "Crump then, converted PCB exposure
levels from the the FDA Total Diet Study (Johnson and Marshe,
1977) to equivalent exposures in rats. The FDA Total Diet
Study exposure levels come from a market basket of food
(representing the basic 2-week diet of a 16-to-19 year old
male) which was collected in each of several geographic
areas. The various foods were prepared in the manner in which
they could normally be eaten and were then analyzed for the
presence of various substances, including PCBs. The estimates
of dose from this study were felt to be tenuous, however.
Exposures to PCBs in Michigan sport fish came from a 2-year
11
-------
(1973-1974) study made under an FDA contract (Humphrey, 1977)
of persons who regularly consumed PCB-contaminated Lake
Michigan sport fish and randomly selected persons who did not
consume such fish (Humphrey, 1977).
Crump's then fitted a one-st.jp model to the animal data to
estimate the extra risk of hepatocellular carcinomas at this
dose. Crump did not mention allowing for any difference
between the composition of PCBs in food to which humans are
exposed being and the composition of Aroclor 1260 used in the
experiment. But, neither CAG nor FDA considered the difference
between the types of PCBs in fish and Aroclor 1254 and 1260.
Crump did comment that PCB levels in human adipose tissue are
likely to.be much greater than corresponding levels in rats for
a given level of dietary exposure, but this also was not used
in the risk estimates. ..The estimates are given in Table 7 (OTA
Table 13). Upper confidence bounds on these risks are not
listed, but could as the author said that they could be
obtained by multiplying all corresponding risks by 1.5.
From the estimates, in Table 7 Crump computed that
nationwide exposure at the dietary level detected in the 1976
Total Diet Study (3.3 ug/day to 8.7 ug/day) »ould cause 4.1 to
11.1 extra hepatocellular carcinomas in the U.S. Crump also
used the one-hit model with estimates of PCB levels in Lake
Michigan fish, to find an average excess of 0.22 hepatocellular
12
-------
carcinomas among residents of Michigan Cbunty exposed to PCBs
in Lake Michigan. Crump did not indicate hepatocellular
carcinomas in rats might translate into cancer at other sites
in man. Crump went on to estimate the risk of hepatocellular
carcinoma in breast-fed infants and concluded that the risk
could be at much higher than for the U.S. population in
general. .
Comparison of results from the FDA, OTA, and CAG risk
assessment is difficult due to the different units in which
risk is expressed in each assessment and the paucity of
explanation in each risk assessment of some crucial assumptions
that may have been made in each risk assessment. Table 8
presents the extra lifetime risks of cancer associated with
consumption of PCBs in food calculated by FDA and OTA in their
separate risk assessments. CAG1s risk assessment is not
compared here as they presented their data only in terms of
virtually safe dose (Table 1). Differences are readily seen.
These are primarily due to the different assumptions made in
the amount of PCBs that would be ingested and in the size of
the exposed population. FDA's risk assessment applies to the
15% of the total U.S. population expected to consume the fish
species known to be the most highly contaminated with PCBs.
The OTA calculations are based on FDA's Total Diet Study (3.3
and 8.7 ug/day) and on estimates of the PCB intake of people
/
who consumed more than 24 Ibs. of Lake Michigan sport fish per
year.
13
-------
CAG and OTA both used the Kimbrough experimental data.
However, CAG used conversions for exposure concentrations due
to bioaccumulation levels in aquatic organisms which OTA did
not use. All three risk assessments are in general poorly
documented and justified. It is not always clear whether
estimated virtually safe doses or increases in risk app:.y to
for animals or humans. None of the three published risk
assessments explicitly refers to species-to-species conversions
(or any other conversions) applied to either the exposure
estimates or the risks from the model. Though we learned from
personal communication that CAG did use species-to-species
conversions.
F~)A used linear extrapolation while OTA (Crump) used the
one-hit model, but this difference, is not great because the
one-hit model is nearly linear at"low doses. Another
difference is that FDA used the NCI study on Aroclor 1254 and
while OTA used the Kimbrough study on Aroclor 1260. Not only
were the number of dose levels different, but the responses and
tumor types were different. Finally, CAG, FDA, and OTA
probably used different exposure assumptions about the
bioaccumulation or biocohcentration levels in aquatic
organisms. It is at least known that CAG used a
bioconcentration factor of 46,000.
14
-------
HERD Risk Assessment
We now updated and expanded the quantitative risk
assessments conducted by CAG, FDA, and OTA. Oncologists from
the HERD Oncology Branch reviewed the NCI report "Bioassay of
Aroclor 1254 for Possible Carcinogenicity" and extracted the
detailed tumor data presented in' TabLe 9. Categories for
leukemias, malignant lymphomas, liver carcinomas, stomach
adenocarcinoma and gastrointestinal tract malignancies, and all
malignant tumors combined were established. The data are
presented for males, females, and sexes combined.
The categories of stomach adenocarcinoma and
gastrointestinal tract malignancies ha/e an additional combined
"*
subcategory, labelled "Morgan, et al.". Morgan et al., 1981
reviewed the specimen's of the stomachs of the Fischer 344 rats
in the MCI study of PCBs and found an incidence of '
adenocarcinoma of the glandular stomach of 0/47, 1/48, 3/48 and
2/48. The "Morgan, et al." subcategory for stomach
adenocarcinoma includes the tumors in the combined subcategory
because Morgan included but did not review the adenocarcinomas
of the glandular stomach already noted by NCI scientists. The
Morgan, et al." s-ubcategory for gastrointestinal malignancies
is also a summation of the combined subcategory alone and the
Morgan, et al. data.
15
-------
The tumor categories in Table 9 correspond with the
categories in Table 2 (used by FDA for animal data in risk
extrapolation to humans) except that FDA combined liver
adenomas with carcinomas, and leukemias with malignant
lymphomas labelled hematopoietic neoplasms by FDA). Liver
adenomas are frequently defined to be benign. (Personal
conversation with scientists in the Oncology Branch). Note
also that a summation of all the tumor categories in Table 9
beginning with leukemias would not be expected to add to'the
category of any malignancy. This is because any animal may
have had more than one type of tumor. Thus, perhaps any
malignancy would best be called "at least one malignancy."
Also, not all the tumor types observed in the NCI study are
listed in Table 9, only those that were malignant.
In parentheses after some of the tumor incidences in Table
9 are p-values derived using the Fisher Exact Test (Cox,
1970) . In the cases where more than one Fisher Exact Test p-
value has been calculated in a given row, a simultaneous
comparison using the same control group has taken place and the
Bonferroni inequality was used to maintain a significance of
p = .05. This occurs for the following reason. When results
for a number of treated groups are compared simultaneously with
those for a control group, a correction is usually employed to
ensure the overall significance level that one has chosen. • The
Bonferroni inequality requires that for k treated groups the p-
16
-------
value for such simultaneous comparisons be less than or equal
to p-value/k. In the case of Table 9 a p-value or a-level of
0.05 was chosen. The only dose levels or categories which were
statistically significant at a = 0.05 were the combined data by
sex at 100 ppm in feed (or 5.00 mg/kg/day) for any malignancy.
NCI's 'report on their PCS; bioassay ( NCE, 1978) states on
pp. 25-26,
"It is concluded that under the conditions
of this bioassay, Aroclor 1254 was not
carcinogenic in Fischer 344 rats; however
a high incidence of hepatocellular
pro-1 iterative lesions, [hyperplastic
nodules, adenomas, and carcinomas]
in both male and female rats was related
to treatment. In addition, the carcinomas
of the gastrointestinal tract may be
associated with tre'atment in both males
and females [none of these lesions were
found in control animals in this study."]
NCI did not establish a category of "any malignancy' ;
therefore they did not conduct any test of statistical
significance for such a category. However, as shown in Table 9
the p-value for the statistical significance of the 21/48
17
-------
animals with any malignancy at 5 mg/kg/day compared with 9/48
animals in control was highly significant at 0.0074. The data
for female rats at 1.25 mg/kg/day were also significant; it is
not clear whether this is a statistical aberration, as none of
the higher doses was significant.
Leukemias and malignant lymphomas are an. important tumor •
category in this study. 66.1 percent (41/62) of the animals
with any malignancy had one of these tumor types. However,
this category is difficult to study because of the high
background incidence of these malignancies in Fisher 344
rats. Historically (up to 1979, at least), 6.5 percent of the
male rats and 5.4 percent of the females have had spontaneous
occurrences of leukemias or lymphomas (Gart et al., 1979) .
Similar rates in the 1978 NCI bioassay were 20.0 (19/96)
percent in male's, and 23.0 percent in females. The-se
*
differences suggest that the significant trend observed in the
NCI* study reflected a real increase, but additional work would
be needed to settle the matter.
Two additional tests were conducted of the data at each
subcategory, males, females, and combined sexes. The technique
used to derive both tests comes from Armitage (1955). The
first test is a test for departure from linear trend. If the
p-value for departure from linear trend is small, then the null
hypothesis of linearity (whether the association between dose
18
-------
and response is a linear one) is rejected and one concludes
that dose affects cancer incidence but in a manner more
complete than linear. Otherwise/ not significant (N.S.) is
indicated. For the NCI data only, the female data for any
maligancy is rejected at p < .05 as being linear in dose-
•response. However, small numbers of animals and respondents in
some tumor categories make interpretation of this' test
difficult. The second test determines whether the (linear)
slope of the dose-response curve is different from zero at a
one-tailed level of significance. If the p-valuo for slope is
small then the inference is that the slope is .significantly
different from zero (in a positive direction), indicating a
tendency for increasing values of dose to be associated with
increasing values of response. Not significant (N.S.) is
indicated if the p-value is greater than OT05." "The categories
of "any malignancy" for males and combined sexes are highly
significant for slope at p = 0.002 and 0.005, respectively.
For the other tumor categories small numbers of animals and
responding animals present a problem of interpretation, though
high significance was observed for slope in the data for males
with leukemias and combined sexes with malignant lymphomas.
One additional test was conducted on the dose-response
data for any malignancy to determine whether there was a
significant difference between the male and female responses
across dose levels. The method used was Cbchran's method of
19
-------
standardized difference (Gochran, 1954). No statistically
differences (at an a-level of 0.05) were found. This test was
primarily conducted to see if there would be problems in using
the combined data for any malignancy as a basis for risk
extrapolation, and it was determined there would be none from a
statistical viewpoint.
There are several reasons for using the "any malignancy"
data as a basis for risk assessment. First/ they are markedly
statistically significantly higher at the high dose, the test
for positive slope is highly significant and there is no
significant departure from a linear trend. These data also
have denominators of 48, a larger number than the 24 animals of
each sex, so that confidence limits an extrapolated dose-
response curve would be narrow. FDA used the identical
categories of total malignancies in Table 2 as a basis for part
of their risk assessment. Available risk extrapolation models
do not preclude the use of data where there are no
statistically significant increases over control at any (or
even all) dose levels, though there may be a substantial impact
on p-values and confidence limits. Better discrimination
between these models also can be obtained if there is an ample
range of doses over which the response is noted. Measures of
goodness-of-fit of the models to the available data are also
more meaningful with a larger number of different dose-response
data points.
20
-------
NCI did not analyze the data for "total malignancies" or
"any malignancy". Since NCI stated that "under the conditions
of this bioassay, Aroclor 1254 was not carcinogenic ..." the
ectablishment of a tumor category where significance occurs may
seem to be like "fishing" for statistical significance. We do
not regard this as a serious objection because there is solid
independent confirmation of carcinogenicity from the Kimbrough '
study (1975), presented in Table 5, where the high incidence of
*
hepatocellular carcinomas was observed 26/184 at 5 mg/kg/day
compared to 1/173 at control (highly significant by the Fisher
Exact Test at p < 0.0001) .
A high incidence of liver neoplastic nodules was also
observed by Kimbrough with 146/184 at 5~ mg/kg/day compared to -
0/173 at control (so obvious that no statistical test of
significance was reported) . However, like adenomas, neoplastic..
nodules are not malignant and the significance of their
appearance is not -entirely clear. (Personal conversation with
scientists in the Oncology Branch).
A question arises here:, why were the tumor types between
the NCI study and the Kimbrough study so different? The main
tumor types observed in the National Cancer Institute (NCI)
Bioassay on PCBs were, in order of decreasing incidence,
hematopoietic neoplasms (leukemias and malignant lymphomas),
liver carcinomas, stomach adenocarcinomas, and gastrointestinal
21
-------
tract malignancies. On the other hand, hepatocellnlar
carcinomas were the main tumor type identified by Kimbrough. A
possible explanation for this difference is that two different
Aroclors were tested in two different strains. The NCI studied
Aroclor 1254 in Fisher 344 rats and Kimbrough used Aroclor 1260
on Sherman rats. We do .not know what the polychlorinated
dibenzoturon concentrations were in these two studies.
(Kimbrough, 1975 and NCI, 1978) The degree of variability
between different strains within a species can be as high as
the .degree of variability between different species. Also,
while the Aroclors are mixtures they differ in both the degree
and structural location of chlorination; for Aroclor 1254 the
major component (53%) is C12HSC15 and for Aroclor 1260 the
major component (42%) is C^2H4C^6*
"•*> • . B •
.The Kimbrough study will allow only the crudest forms of
risk modelling as it has only one nonzero dose level. However,
there is no question as to the statistical significance of the
increase in hepatocellular carcinomas observed by Kimbrough.
The NCI study will allow much more mathematically sophisticated
risk modeling and measures of goodness-of-fit of the models to
the data; but, standing alone is difficult to interpret,
because of the question of statistical significance of the
various tumor types. Indeed, NCI concluded that its study did
not establish carcinogenicity. There is also some question
about the validity or advisability of using the category of
22
-------
"any malignancy". FDA combined conclusions from the two
studies to estimate cancer risks to humans. That is also the
course that this assessment follows.
Table 10 presents maximum likelihood estimates and lower
95 percent confidence limits on dose for a range of excess
risks from IxlO"1 (1 in 10) to lxlO~8 (1 in 100,000,000)
(hereafter called virtually safe dose). The data used here are
the Kimbrough data for hepatocellular carcinomas. Because
Kimbrough tested only one nonzero dose level, the only models
that could be used were one-hit or simple linear extrapolation.
As linear regression-gives nearly the same results as the one
hit models at low doses, only one-hit extrapolation was
employed. The computer program used was that for the
multistage model/ which degenerates to a one-hit model when
•only one positive dose level is available. The confidence
" j •; • •
-..-.• ...• *
limits on these virtually safe doses are- tight, roughly within
a factor of 1.5 of the point estimates at a risk of 1x10"^ (1
in 1,000,000) .
Table 11 presents a similar range of excess risks of any
malignancy over background based on the. NCI PCS Bioassay,
derived using five of the most widely known and used
extrapolation models: The multistage, logit, probit, Weibull,
and gamma-multihit. The logit and the probit ordinarily are
expected to give similar results, because of their similar
functional form. Estimates from the one-hit model were similar
23
-------
to those from the multistage model. Simple linear
extrapolation is approximated by the lower confidence limits on
virtually safe doses (that is, upper confidence limits on risk)
from the multistage model. Two types of background were
assumed for incorporation in all models except thr
multistage. Independent background implies that any ongoing
background carcinogenic processes are independent of the
carcinogenic response induced by the administration of the
carcinogenic agent under study. Additive background implies,
on the other hand, that cancers induced by administration of
the chemical under study add to already on-going carcinogenic
processes, as if arising from an effective background dose,
acting by the same mechanism and producing similar types of
tumors. It is not cltar at this time which of these two is the
more valid assumption, therefore for the four models available
**
that consider both, types of background, both.are presented.
The multistage model is consistent with both assumptions so,
traditionally, one does not speak of it as either additive or
independent. Goodness-of-fit p-values were also calculated to
give an idea of the fit of the model to the data.
It can be seen from of Table 11 that the estimates of
virtually sa-fe dose at a risk of IxlO"6 (1 in 1,000,000) from
the independent background models vary a million fold from 10~2
.for the probit to 10~8 for the gamma-multihit. The additive
background models, including the multistage model, are all in
24
-------
the area of 0.16 x 10~4 to 0.17 x 10"3 mg/kg/day. The
goodness-of-fit test results show the multistage model to be
the model with the best fit to the data (p > 0.9000) and the
additive gamma-multihit, the worst fit (although it is still.
quite adequate) at p = 0.5391.
Tables 12 and 13 present the results -of two suggested
methods for placing confidence limits on virtually safe dose,
*
one by using the variance of log-dose, the other by using the
variance of the reciprocal of dose. The confidence limits in
Table 12, based on the variance of log dose are generally much
wider than those in Table 13, based on the variance of the
reciprocal of dose. Confidence limits from the independent
models vary widely, while thos.^ from the additive models are
within factors of 5 of each other by the two methods and are
within factors of 4 -to 20"of their maximum likelihood estimates
at a risk of IxlO"6.
The confidence limits for the multistage raodel, calculated
by a method unique to that model, are within a factor of 1.8 of
its maximum likelihood estimate of virtually safe dose at a
risk of IxlO""*5. Thus the multistage confidence limits are much
more narrow or "tight-" See Figure 1, which shows upper
confidence limits on risks, rather than lower confidence limits
on dose.
25
-------
Table 14 presents, for a range of excess risks from 1x10"^
to IxlO"8, a table of maximum likelihood estimates for leukemia
alone over background based on the NCI PCB bioassay. Leukemia
alone was chosen because of the high percentage of pathology of
this type in the NCI bioassay. The risks predicted by the same
range of independent and additive background models are within
factors of 5 of similar estimates based on any malignancy.
Again estimates of confidence limits on virtually safe dose
based on the variance of log dose vary more widely than
estimates based on the variance of the reciprocal of dose. The
multistage estimates are also narrow and this model produces
the best fit among the models examined. Overall there is much
consistency between the estimates based on any malignancy and
those based on leukemia alone.
" "*" • • • . -
-'. We "now consider "-how estimates of virtually safe doses for
risk of hepatocellular carcinoma from the Kimbrough study
compare with estimates of virtually safe dose based on any
malignancy or leukemia alone from the NCI study. Based on the
multistage, the best fitting model, the maximum likelihood
estimates from Table .10 compare very well (within a factor of 1
to 2) with each other, the confidence limits within a factor of
1 to 3. Tt seems then that it makes little difference which
study is used, Kimbrough or NCI, to base quantitative estimates
of cancer risk to humans. The decision on which study to
choose for extrapolation might be based on the likelihood of
26
-------
exposure to a particular type of Aroclor since apparently the
tumor types were different for tests of Aroclor 1254 and 1260,
though strain of animal may have been just as important. The
extent to which one expects perfect site concordance would also
dictate the choice of a study. Remember that FDA used both
studies in its risk assessment of PCBs. . - ..
With these preliminary remarks and tabulations, we use for
*
exposures of various subpopulations provided by the Exposure
Assessment Branch, Exposure Evaluation Division of OTS to
estimate additional or excess lifetime risk. The NCI data
(used by FDA) and the Kimbrough data (used by FDA, CAG, and
OTA) are used as the bases for extrapolation.
Table 17 presents the excess or additional lifetime risk
estimates based on., total malignancies in the NCI study. The
model used to derive these excess risks was the Crump
multistage model and program (used by FDA, CAG, and OTA in
their assessments). A species conversion factor of
approximately 5.85 was used for the transformation of rat risks
to human risks; i.e., humans are presumed to be roughly six
times as.sensitive to carcinogenic effects of chemicals as
rats. The estimated exposure duration in years was used to
modify the risk estimates for the proportion of an average
human lifespan of 70 years that an individual in a given
exposure category might be exposed to PCBs. If the exposure
27
-------
duration was 38.5 years, as in the case of exposure
loading/unloading a liquid assuming PCBs are present in the
liquid at 25 mg/kg then the cancer risks were reduced by a
factor times 38.5/70 or 0.55 of their original size. (It would
be best to base such proportion of lifetime calculations on
results from a differential exposure study; however, none was .
available and this simplistic multiplication factor is
currently the state-of-the-art in this area.) Both "most
likely" or point estimates of excess cancer risk and 95 percent
upper confidence limits on excess cancer risk are presented.
The upper confidence limits from the multistage model exhibit
linearity at low exposures. They also assume additivity.
Additivity, as mentioned earlier, assumes that the effect of a
carcinogenic agent is to act through the same mechanisms as
that operating for background process.; The upper confidence
limits are not markedly more conservative .than the point
estimates of risk in the case of the NCI total malignancies
category. Many other factors mentioned in the other Agencies
^-isk assessments influence the results. Among these are the
dissimilarity between residues in fish and water and Aroclor
1245, the levels of PCBs in human adipose tissue, and the small
sample sizes in the NCI study.
The extrapolation distance between the lowest experimental
dose in the NCI study (1.25 mg/kg/day, the lowest of any level
in either the NCI dataset), and the highest and lowest exposure
28
-------
levels is on the order of from 6 to 45,000,000/000. This is a
very wide range. Keep in mind that generally the further the
distance over which risks are extrapolated the less confidence
one could have in those risk estimates, and that errors could
go in either direction.
The assumptions used'to derive the exposure levels shown
in Table 17 are discussed in the exposure assessment for
incidentially - produced PCBs (Versar 1983). One purpose of
the exposure and risk assessments is to provide guidance in
establishing a permissable level of PCBs in other chemicals.
Using conservative assumptions, the exposure assessment
estimated the upper limits of exposures that may result when
PCBs are present as impurities in a variety of chemicals and
products^ at levels of 50 ppm, 25 ppra, and 2 ppm. The exposure
levels in Table 17 correspond to the upper limits of exposure
for the various scenarios when the concentration of PCBs is
25ppm. Few, if any,. individuals will be exposed to these
relatively large amounts of PCBs. Thus, the maximum likelihood
of excess risk in Table 17 applies to highly exposed
individuals. The excess carcinogenic risks to typical members
of the exposed groups are believed to be no higher than those
in Table 17 and are probably much lower.
29
-------
References
Armitage, D. (1955). Tests for linear trends in
proportions and frequencies. Biometrics, 11, 375-385.
Bioassay of Aroclor 1254 Nd for possible carcinogenicity.
Carcin. Tech. Rep. Ser. No. 38. RHS Publ. No. (NIH) 78-
838. 1978.
Cochran,. £.G. v!954) . Some methods or strengthening the
common x tests. Biometrics 10, 417-451. -
Crump, K.S., Guess, H.A. , and Deal, K.L.. 1977. Confidence
intervals and tests of hypotheses concerning dose response
relations inferred from animal carcinogenicity data.
Biometrics 33: 437-451.
Crump, K.S., and Masterman, M. "Assessment of carcinogenic
risks from PCBs in food." April 1979. Working Paper IV.
Crump, K.S. An Improved Procedure for Low-Dose Carcinogenic
Risk Assessment from Animal Data. Jou-rnal of Environmental
Pathology and Toxicology, vol. 5, No. 2, 1981, pp.675-684.
Cordell, F. June. 1982. "Risk assessment in a Federal
regulatory agency : human consumption of some species of
fish contaminated with polychlorinated biphenyls .( PCBs)."
Environmental Health Perspectives.- Vol. 45. •/ .
Environmental contaminants in food. U.S. Off ice. of ' ,
Technology Assessment. December 1979. • '
Humphrey, H.E.B. 1977. Evaluation of changes of the level
of polychlorinated biphenyl's (PCB) in human tissue. Final
Report oh FDA Contract 233-73-2203.
Industrial Bio-Test Laboratories, Inc. 1971. Reports to
Monsanto Company. Two-year chronic toxicity with Aroclor
1242, 1254, and 1260 in albino tats. Unpublished reports.
November 12, 1971. IBT No. B7298.
John T. Gart, Kenneth C. Chu, Robert E. Tarone, "Statistical
Issues in Interpretation of Chronic Bioassay Tes^.s for
Carcinogenicity," Journal of the National Cancer Institute
Vol. 62, No. 4, April 1979.
Johnson, R.D. , and Marshe, D. D. 1977. Pesticide and other
chemical residues in total diet samples (XI). Pesticide
Monitoring Journal 11: 116-131.
30
-------
Kimbrough, R.D. , Squire, R.A. , Linde, R.E., Standbury, J.D. ,
Mental, R.J., and Burre, V.W. 1975. Induction of liver
tumor in Sherman strain female rats. J. Nat'l Cancer
Inst. 55: 1453-1459.
Mantel, N. , Bohidar, N.R., Brown, C. C. , CLminera, J.L. , and
Tukey, J.W. 1975. An improved Mantel-Bryan procedure for
the "safety" testing of carcinogens. Cancer Research 35:
865-872.
Miller, R.G. , Jr. Simultaneous statistical inference.
McGraw Hill Book Co., New York 1966, pp.6-10.
Morgan, R.W., Ward, J.M., Hartman, P. E. "Aroclor 1254-
induced intestinal carcinoma and adenocarcinoma in the
glandular stomach of F344 rats." Cancer Research, 41, pp.
5052-5059, 1981.
Rai, K. and Van Ryzin, J. Risk assessments of toxic
substances based on a generalized multihit dose response
model. Proc. SIMS Conference Alta, Utah, June 1978: 26-30.
USEPA "Ambient Water Quality Criteria for Polychlorinuted
Biphenyls" EPA 440/5-80-008. October 1980.
USEPA "Water Quality Criteria Documents; Availability" Part
V Environmental Protection Agency Friday., November 23, 1980,
pp.79339, 79353. .
Versar, "Exposure Assessment for Incidentally .Produced . -
Polychlorinated Biphenyls (PCBs)", Washington, D. C. U.S.
Environmental Protection Agency, Office of Toxic
Substances 68-01-6271, 1983. (Draft Final Report)
31
-------
!• 11.41 li IS
U)
NCI STUDY OF RATS EXPOSED TO PCB
)A~A PO NTS AND UPPER 95 CQNFlOEKCE LIMlfS FROM CRUMPS MODEL
I . 0
C
A
R
C
H
0
G
£
ty
1
C
R
E
S
P
0
N
S
e
. -
-
0.9-
'
l
0.8-
0.7-:
•
o.e-:
1
o 5-:
.
P <~
G . J-:
]
0. 2-
-
o . :-
.
o . tf-
/ '
: • '
j • v
/
' /
.
.--a
1 _. . , " — """
: T """"''
E> "r"r
_.....-'--0""""'^________ ___^ '-— ~ !
t3---"" _^-—+"~^~ " r~T~
-^~^~*^ '
,J • — . . . /
3 10 2C JO 40 CC fcC 70 BO 90 100
DOSE • K PFM
' ' *
t ,
uECEKO. MODs^ .V-^—f--*. CgS p-&I-i3 UCL
-------
l-'l fill UK 2
K
B
E
S
P
0
M
S
E
P
R
oj 0
CO
A
1
L
1
T
V
:MBROUGH HEPATOCELLULAR CARCINOMA;
DATA, MULTISTAGE MLE CURVE AND UPPER *ft PERCENT CONFIDENCE LIMIT
1 ,0-
-
0.9-i
•
0 . 8-
.
0. 7-
-
0.6-:
o.ft-:
0.*-
0.3-:
0.2-:
o. i-;
o.o-
• .
•
'
•.
• • .
/ *
— *««~**"* — **~
•
0 2ft 40 7ft 100
DOSE |M PPM
LC/*FfcJfl- IIAnCl ^ A ^ fine* .— nc* • • ff* ••••_•.. «iir AIIT
C. UC. NU • MUUtL T T-T Ofl^ Pf MS. .--.--.. y^ ^^
i ,
i
-------
Table I- C2J31 s Table for Water Quality Criterion of PC3s
PC3 Exposure Lifetime Risk Levels and Water Quality Grit*
10"6
10'
10
-5
2. liters of
drinking water
and consumption
of 6.5g contaminated
fish and shellfish
per day*
0.0079 ng/1 0.079 ng/1 0.79 ng/1
Consumption of fish
and shellfish only*
0.0079 ng/1 0.079 ng/1 0.79 ng/1
*Approximately 99% of the PCB exposure results from aquatic
organisms, which exhibit an average bioaccumulation potential
of 31,200-fold. The remaining 1% of OCB exposure results from
drinking water.
34
-------
2: Aiimal Data Osed for Risk Extrapolation to finnans in FDA ^ssessnent
Dose of £roclor 1254 fed
Aiimal Studies
pan 0
nnAg/day 0
25
1.25
50
2.50
100
5.00
NCT aioassay - Fischer • . \
. Fischer Rats fed . .
Total Malignancies /
Males . 5/24 2/24 9/24 12/24
Females 4/24- " 13/24 8/24 9/24
Qanbined 9/48 15/48 17/48 2J/48
Liver Ccircinona & AJenonas
Males 0/24 0/24 V24 2/24
Fenales 0/24 0/24 1/24 2/24
ODtnbined 0/48 0/48 2/48 4/48
Hematopoetic Nsoplasms
Males-- " 3/.'4 2/24 5/24 - 9/24
Eenales • 4/24 , '. 6/24 ' 6/24 6/24
Oanbined 7/48 " 8/48 11/48 15/48
35
-------
Table 3: Cbper Csnfidencs Limits (99%1 on Lifetime Risks* of Oncer in Eaters of Fish Spe<
Studies
en which risks
are based
50th Percentile Eaters
901
J£suning /ssuning £ssuning £ssuning
Nb Tolerance Tolerance . Tolerance Tolerance
* 5 ps
Lake**
USA Michigan
fi-6—
1 ppa
Assuning
Ma Tolerance
Lakes* "
USA
Kimbrcugh - Pats
Liver
Nd Bioassay - Tbtal
Malignancies for Male
and Eemale
1.3 13.4 . 1.2
4.1 58.0 3.7
0.8
2.7
0.5
3.4 4.1.. 4
1.6 10.6
36
-------
>le 31 Upper 'Confidence Limits (99%) on Lifetime Risks* of Cancer in Eaters of Pish Species of Interest
imal Studies !
which risks '
) based
i
.ibrough - Rats
»er Carcinoma •
[• Bioassay - Total
i Llgnancies for Male
j Female
I Bioassay - Liver
rcinomas and Adenomas
r Male and Female
[ Bioassay ;
natopietic for
ie and Female-
Mi risks are lifetime
i
50th Percent lie Eaters
i
Ass, uning Assuning A3 sun ing"
to jlblerance Iblerance Iblerance
• Lake** = 5 ppn « 2 ppn
USA Michigan .
i • .
l i
1
1.3 18.4 1.2 0.8
1 i;
4.1 58.0 3»7 2.7
0.9 12.75 0.9 0.6
2.7 38.25 2.4 1.8 ,
risks computed as rates per 100, 000 of
90th percentile Eat
Assuning Assuning Assuning
Iblerance to "tolerance Iblerance
= 1 ppn • Lake** = 5 ppn
USA Michigan
., : .
0.5 3.4 41.4 3.1
1.6 10.6 129.2 9.8
0.4 2.5 30.5 2.3
.
1.1 7.0 85.3 6.5
the population at risk.
Hisk calculated for Lake Michigan sportsfish eaters who consume an average of 1.7 jjyAg/day pen
>r 3.9 iigAg/day at the
90th percentile. Risks in other areas
having similar sportefish consuiption
and PCB contamination are probably similar.
! i
i
! i.
i
i •
i
!
!
37,
i
i
i
•
n PDA Assesanent
rs
ssuning Assuning
blerance Iblerance
2 ppn = 1 ppn
2.3 1.4
7.2 4.4
1.7 1.0
4.7 2.9
-------
ible 4t Upper Cbnfidence Limits on Number of New Cancers per Year in Eaters of Fish Species of Interest i
n FDA Assesanent
ilmal
zudles on wliich
isks are based
50th Parcentile Eaters
90th Perce
ntile Eaters
Assuning
NJ Tolerance
Lake**
USA Michigan
Assuning Assuming Assuning
Tblerance Tblerance Tblerance
» 5 ppn = 2 ppn = 1 ppn
Assuning
No Tblerance
lake**
USA Michigan
Assuning
Tolerance
= 5
A3sunIng Assuning ,
tolerance Tblerance
3 2 ppn = 1 ppn
imbrough - Rats
Ivor Carcinoma 6.2 10.4 5.8 3.8' 2.4 16.3 23.4 14.7
Q Bloassay - Total
alignancies for Male .
nd Female 19.6 32.8 17.6 .12.9 7.6 50.6 73.1 46.8
"
Q Bioajsay - Liver
arcincmas and Adenomas
or Male and Female 4.3 7.2 4.2 2.9 2.0 12.0 17.3 10.9
Q Bloassay
smatopoietic
or Male and Female 12*9 21.6 11.4 8.6 5.3 33.4 43.3 31.0
f
All risks are the increased nunber of cancers per year for the population at risk (15.2% of U.S. populat
considering a 70 year life span.
*Dl air r»a1 /-*•! at*orl fr\tr I* nlrn Ml r*Klrtnn arwtt*l-ci f4 oH c»al-^fr-a uiKrt r*r\r\atma ark nttc\»-=u-«ci 1 f unAtr* /rlatt Dr*O
10.0 6.7
34.3 21
fl.O 4.7
22.5 13.8
on)
or 3.9 pgAg/day at tho 90th percentile. (4,000,000 people assured exposed). Risks may be similar for
sportsfish eaters in other areas) but data not available to make estLnate.
38
-------
Table 5: Data sets used by Qnjnp for CTCA in calculating virtually safe coses
Data Set No.—I Kiabrcugh-jt^aL. (1975) rat study with Acclor 1260—
fepatocellular carcinonas
Dietary level . No. of animals No., animals with fepatocellular carcinomas
(pan) (mg/kg/cay)
0 0 . - 173 . . 1
100 . 5.0 ' . 184 26 •
Data Set No. II-Kimbrough e^^ (.1975) rat study with Acoclor 1260—
Liver neoplastic nodiles
Dietary level . No. of animals No. of animals with Neoplastic nocules
(ppn) (mg/kg/cay)
0 "0 173 a
100 5.0 ' 184 146
Data Set No. ni-Incustrial Bio-Test rat experiment with Aroclor 1260—
" . liver neoplastic nocules
Dietary level - No. of animals No. of -animals with Nsoplastic nocules
(ppa). (mg/kg/day) . '
00 23 1
1 0.05 25 0
10 0.5 23 . 9
100 5.0 27 7
39
-------
le 6: Virtually safe doses ccnputed by Crunp for OTA fran data in Table 5
Data Set
Analytic Method
la Etofait
Ib Cne—Hit and
Multistage
Ila Brobit
Hb Cne-Hit and
Multistage
Ilia Probit
mb cne-Hit
Multistage
Maximum Likelihood
estimates of dose in
ppb» corresponding
to extra risk o£
extra risk of
Virtually safe doses
(lower 95% confidence
bounds for dose) in
ppb. corresponding to
KT8
ICT5
10-
lor6
10'
,-5
0069
00063
0046 .
0046
.636
.063
.465
.465
6.36
.634
4.65
4.65
1.96
.0051
.025
.00055
.176
.00235
.00205
14.2
.511
.180
.0511
1.29
.235
.205
43.6
5.11
.552
. .511
3.97
2.35
2.05
40
-------
Table 7. Estimates of lifetime extra risk io hunans of hepatccellular carcinona
based en a one-hit model (2) to the Kimbrough et'al. (1975)
rat study (appeared in OTA docunent)
Risks calculated fron converting"- honan
dose to animal dose en the basis of
Hunan PCS Dosage gen in diet mg/frg/day
3.3 u3/cay 1/328,000 1/764,000
(1976 Total Diet Study) . - .
8.7 yg/day 1/123/000 ' ]/2S8,UOO
(1975 Total Diet Study) ' ' •
127 yg/day 1/3,000 1/20,000
(A/g. intake of people consuming
more than 24 Ibs./yr. I^ke Michigan
fish, aumphrey (1977) •
41
-------
Table 3. Extra lifetime risks of cancer associated with consumption of PC2s
in food (appeared in CIA docunent)
Extra lifetime Upper limit of new
Dose (ug/day) risk/100,000 cancers/year
9.2° . . 4.4 — 21
14.9° 7.2 34
20.1C 9.8 47
coat3 - • ••"..- • . '
3.3d ' 0.13 4
8.7® 0.35 11 '
127f -5 . —
"Based on Md bioassay and total malignancies for males and females.
^Based on Kimbrough (1975) study and hepatocellular carcinomas,
C8ased on highest consuners (90th percentile) of fish species contaninatecl with
PCBs if tolerance established at 1.2 of 5 ppnu
(jBased on FDA Total Diet Study. 1976.
®Based on FDA Total Diety Study. 1975.
'Based on average intake of people consuning more than 24 Ibs/year Lake Michigan fish.
42
-------
Die 9: Ail ma I Data Used for OTS Risk Extrapolation to I lama ns*
lignancy Category
taimals Par Dose Ifivel**
•
{ Malignancy
les
nales
nblned
ukanias
les
males
nbined
lignant Lymphomas
les
males
nibined
ver Carcinoma
les
males
mblned
omach Adenocarcinoma
les
males
mbined (original report)
mbined (Morgan et al.)
PP"
in feed 0
ihgAg/day o
5/24
4/24
9/48
3/24
4/24
7/48
0/24
0/24
0/48
0/24
0/24
0/48
0/24
0/24
0/48
0/47
25
1.25 .
2/24 (N.S,)
13/24 (.0093)
15/48 (N.S.)
2/24
6/24
8/48
0/24
0/24
0/48 •
I
0/24
1/24
V48
0/24
1/24
1/48
1/48
, 50
2.50
9/24 (N.S.)
8/24 (N.S.)
17/48 (N.S.)
5/24
6/24
1V48
0/24
0/24
0/48
J/24
0/24
1/48
1/24
1/24
2/48
3/48
100
5.00
12/24 (N.S.)
9/24 (N.G)
21/48 (.0074)
8/24 (N.S.)
4/24 (N.S.)
12/48 (N.S.)
1/24 (N.S.)
2/24 (N.S.)
3/48 (N.S.)
2/24 (N.S.)
0/24 (N.S.)
2/48
0/24 (N.S.)
0/24 (N.S.)
0/48 (N.S.)
2/48 (N.S.)
Risitiye
Slope
0.002
(N.S.)
0.005
0.014
N.S.
0.080
, 0.063
0.015
. 0.004
0.030
N.S.
0.016
N.S.
N.S.
N.S.
N.S.
P-Value
p-va
frc
N.
P
N.
N
N.
N
N
N
N
N
N
N
N
N
N
N
***
lue for departure
n linear trend
S.
< .05
S.
S.
S.
S.
S.
S.
S.
S.
S.
S.
S.
S.
S.
S.
43
-------
Sable 9« /ftimal Data Used tor OTS Risk Extrapulation to Hunana* (Continued)
lalignancy Category
Jastrointestinal Tract
Malignancies (including
Jejunum, Cecun, and
Stomach
1ales
females
Xxnbined (original report)
Ddmbined (Morgan et al.)
in feed 0
mgAg/day 0
0/24
0/24
0/48
0/48
25
1.25
I
0/24
1/24
1/48
1/48
taimals Per
50
2.50
1 !
2/24
1/24
3/48
4/48
Dose Level**
100
5.00
1/24 (N.S.)
0/24 (N.S.)
1/48 (N.S.)
3/48 (N.S.)
P~
Positive p
Slope
N.S.
N.S.
N.S.
0.044 1
/alue***
value for departure
from linear trend
J.S.
M.S.
M.S.
*From the National Cancer Institute bioassay of Aroclor 1254
except as noted. Morgan et al. ( Cancer Research 41, 5052-5039,
December 1981) re-examined the whole-tissue specimans of startach
taken frcm the NCI bioassay and presented malignancy data per combined
sexes only.
**Mjmbers in parenthesis are Fisher Exact Test p-values for statistical
significance over control. Hie Bonferroni ineojuality was employed to correct
for lowered p-values due to multiple comparisons with the same control group.
N.S. indicates not significant at thea = 0.0500 level.
***1he technique used to derive these numbers canes from (bchran (1954) and Annltage
(1955). If the p value for positive slope is small the Inference is that the slope
is significantly different frcm zero (in a positive direction), indicating that there
is a tendency for dose to be associated with increasing values of response. Otherwise
N.S. (not significant) is indicated if the p-value is greater than 0.10. If the p-value
for departure from linear trend is small the null hypothesis of
linearity (or whether the association between dose and response is a
linear one) is rejected. Otherwise N.S. is indicated. .
44
-------
labl« 10
Exoit Rlik to R*l» ol lUpatocal lular Carclnoo* Ovor Uackyround
(0«t*4 on Klabrough PCB Bloattay)
Mod* I
Multlttag*
(On*-tt*g*l*
|.0006-OI t.O.IOt-02 I.OOOE-0) I.OUOE-04 I.OOOC-OJ I.OOOE-06 |.i)OOE-OI I.UlOE-Oa
LlkcllbooJ E»tl«*l«t ol Virtually Sal* Oat* tor Halt
0.miOOE«02 0.6«9»ilE»OI 0.6t6TO}E«00 0.6B6192E-OI 0.666J6U-02 0.6B6J3Bt-0> 0.6061 J8t-04 0.6 16 J)B£-01
tit On*-$|il*d lo.ar Coaf| an Virtually Sol* Ooiat lor Rait (oy/kg/dayl
Multlttag*
lUna-ttagal*
0.324'i«2Ei02 O.J0029J£»OI 0.4»B026E«00 0.4»J80|E-OI 0.4»7)I»E-02 0.497JI6E-OJ 0.487IJ6E-04 0.
9II76E-OS
•III* •ultlttay* nodol dtgnnoral of to Ilio tpoclllc <:a>* ol th* one-hit «odul ahen only on* potlllv* 4as« lav*l It avallabla.
45
-------
labl* II
I
Exceii Hltk |o >ia It o4 Any Malignancy Over Background
(Bated on NCI fCU Bloattay)
Hod* I
I.OOOE-OI
I.OOOE-02
.OOOE-03
I.OOQE-04
I 030E-03
I.OOOF-06
I.OOOE-OI
I.OOOE-08 '
likelihood Eftlaat* ol Virtually Sal* Dot* for Rail lag/kg/day)
Goodn**t ol III
!•»* p-value'
Multlttag*
0.338381E402 ,3373694E»OI 0.333443E«00 0.333234E-OI 0.333133E-02 0.333233E-03 0.333233E-04 u. 333233E-103 >0.»000
Independent ' .
Problt" O.I88339E«02 O.II|OI7E«OI O.I43637E400 0.276096E-OI 0.638313E-02 O.I82634E-02 0. 367* 12E-03 O.I9260IE-
i
Independent
loglt" 0. I836|)E«02 0.486|19E«00 0. I3I306E-OI 0.47I073E-01 O.I3I28IE-04 0.478783E-06 O.I3I330E-07 0.479380E-
Independent
Nalbull" 0. I79320E«02 0.327304EIOO * 0.634606E-02 0. I3I939E-03 0.266236E-03 0.337I92E-07 0. I06392E-08 0.2I8707E
lnd*p*nd*nt GBMM*
Hultlhlt" 0.939I99EIOI O.I13339E-OI 0.338I4U-04 0.494380E-OI O.I34404E-09 0.260I33E-I2 0.303551E-I3 0.97467JE
Additive
Problt" 0. I93480E402 0. I23692E40I 0. I20492E400 0.1I9983E-OI 0. II9934E-02 0. II9929E-03 0. II9929E-04 0. II9928E-
Addltlv* '
Loglt" O.I92963EI02 O.I23398E»OI O.II7993E»00 0.|U«A3E-OI O.II74IJE-02 O.II7407E-03 O.II7407E-04 O.II7407C
Additive
Hclbull" O.I90640E<02 O.II3363E«OI O.I09303E«00 O.I0893IE-OI 0. I08874E-02 0. IOR869E-03 O.I0886UE-04 0. 1088601-
03 0.8632
09 0.1801
10 0.8937
18 0.6784
1)3 0.8336
1)3 0.8349
03 0.8632
\
Additive Gao*i* . . ,
Multlhlt" 0.2I3479E«02 O.I66643E«OI O.I620IOEtOO O.I6I348E-OI O.IAI302E-02 O.I6I498E-03 O.I6I497E-04 O.I6I499E-03 0.3391
? « .
'Any p-yalua yr«al*r than 0.30 Indicates an adequate lit ol the node I lo III* data,
"In nany experiments the res^onte ol lotereil alto occurs »pontaneouiIy In control anlaol*. lull background >ay b* aisuaiad lo b* ellhor lnd«|>*ndenl ol Hi*
Induced raiponiei or additive In • nechanlitlc Banner. II In* apontanfoua and Induced r**pon»*» *r* **iu*)*4 lo b* Independent, than Ih* |>robablllly ol
ob»*rvlng • retponi* ol either typ* at dot* d It given byi PMdl - q » (I - q) Plill, xh*r* 0 < q < I donate* th* tponlaneuut background rale. Under Hi*
addltlvlty *»u*ptlun, th* background r*tpont* *>*y b* contld*r*d •» arising Iron an *ll*ctlv* background dos* * > 0, «l|h P*ld) • P(d * «l. Although Ih*
• anner In uhlch background r*tpont* It accommodated It crucial, th* extent to «h^^ Independence or *ddltlvlty It Indicated by •llhor biological lefliVv or
CKpor l»»n|^^^data It aomewliot unclear at th 11 tin*.
46
-------
l.bl. 12
9).Of Ona-Sldad Loxor Conlldanc* Lloltt on VIrtuaI Iy SaI a Oo«*» lor fUtt In ig/kg/day In lob I a II ttatad on th* Variance ol Loj-Uoni
(Baiad on MCI PC6 Qloatiayl ' !
Nod* I
I.OOOE-OI
I.OOOE-02 I.OOOE-OJ
I.OOOE-04
I.OOOE-OJ
I.OOOt-06
t.OOOE-07
I.OOOE-08
Multlttay*
0.20I9J6E«02 O.I90a09E40l .0. l897e>E«00 0. U»68)E-OI 0. I8967JE-02 O.I89672E-0) O..I89678E-04 0. I896I2E-0)
lnd*p*nd*n|
Problt1 - 0.220746E40I
Loylt* C.II9026E«OI
Independent
Welbull* . O.U3I92E40I
Independent Oammt
Nultlhlt* 0. I6914IEIOO
Additive
Problt* (i.260»OE«OI
Addltlv*
Loylt1 (I.274J34E10I
Addltlv*
Melbull* (l.2)49)a£«OI
Additive Go««a
MulllMI* 0.497889E 1 1 72E-1 1 0.649296E-19 O.I27II3E-46 0.248641E-S4
1
; . i
0.6J4469E-OI 0.34347U-02 0.3)72»3E-0) 0.3J6484E-04 0.3J640JE-05 0.5J6J96E-06 0.5>6^3£-07
' -1
..
1
i
0.370630E-OI 0.4B4I97E*02 0.476910E-OS 0. 476 1 50E-04 0. 41607 2E -05 0. 4 V6064E -06 0.4J6U6i£-07
i
i
0.44I988E-OI O.J67162E-02 0.)606I9E-01 O.J39932E-04 0. J39883E-05 0.)i9878E-06 O.J5«J)uE-07
i
0. I9091JEIUO 0. I7I334E-OI O.I69708E-02 0. I69323E-OJ 0. I69306E-04 0. I6950JE-05 0. I6»:07E-06
•la B>ny *xparla*nlr. In* r*ipont* ol Interctt alia occur • >pont*n*outl y In control «nl«ol$. Ihli background *ay ba a>»un*i| to b* allhar'
lnd*|i*nd«nt ol th* Inducad r«sponsas .or addltlv* In a nachanlatlc aannaro M tno §pontan*out and Inducad rosponi*s aro tstuaad to bu
Indapandont, lh*n vlic probability ol obturvlng a ratponta ol allhar I ypa at do*« d l> glvan byi PMdt • q « (I - q> P(dl, uliar* 0 < o, < 1
danol*> Iho aponlaiiaou* bacfcyround ral*. Uiidar Ilia addltlvlty oiuaptlon, th* background roiponi* o>ay b« con>ld«r«d at arltlmj Iron an
allacllva backyround dot* • > 0, ullli P*(dl - Pld « «». Although th* aannar In uhlch background ratponsa It accoanudat *d It crucial, II «
axlont to ulilch I mlapandanc* or addltlvlly It Indlcatad by althar biological Ihaory or oxpar I nan I al data It ton««ltal unclaar al Ihlk llua.
47
-------
table II
91.Of One-'iMed Lover Conlldonce Llalts on Virtually Sale Doses, lor Rats In a>g/kg/day In Table II Bated on the Variance ol the Heclprocal ol Uute
(Bated oa NCI PCB Bloasaayl
Model
I.OOOE-OI
I.OOOE-02 I.OOOE-OJ
i.oooE-04 I.OOOE-OI I.OOOE-O& I.OOOE-OI i.oooc-uo
Mulllatage
0.20I916C«02 O.I90809E«OI 0. I89785E«00 0. U9eiJ£-OI 0. I8967JE-02 0. I89672E-0) 0. I096I2E-04 0. lt9672E-01
Independent
Problt* 0.999|22E»OI
Independent
Loglt' 0.)60906E«OI
Independent
Welbull* 0.129801EIOI
Independent Oarja
Hultlhll* 0. I90484E«OI
Additive
Problt* 0.6J99J8E«OI
Additive
Loglt* 0.634I26E«OI
Additive
Wolbull* 0.6J2956EIOI
Additive Gaoma
Multlhlt* 0.874I28E«OI
•
0. I6J994E»00 0. I3JI9JE-OI 0.231194E-02 0.4800J7E-0) O.II807IE-OJ 0. J125J6E-04 O.I0187SE-04
< '• . !
0.)98872C-OI 0. II7040E-02 0.269J04E-04 0.6707IIE-06 O.I7490JE-07 0.470661E-09 0. I295J7E-IO
t
0.1701IIE-OI 0.436419E-01 0.6647181-0} 0. I049J9E-06 O.I7J944E-08 0.297762E-|0 0.}2IIOU-I2
i
0. II1728E-02 0. I29I98E-0) 0. I76I77E-OB 0.244809E-II 0.4I7882E-I4 0.68272IE-I) 0. II41I9E-I9
D.)I))I6E«00 0.2942J4E-OI 0.2922I9E-02 ' 0.2920I9E-OJ 0.29I999E-04 0.29I997E-OJ 0. ^91 99SE -06
i
O.J0291IE400 0.28I468E-OI . 0.2794I7E-02 0.2792UE-0) 0.279I9JE-04 O.J79I90E-OJ 0. 2JI9 1 90E-06
'..•'' i
0.270681E«00 0.249I66E-OI 0.247II8E-02 0.2469I4E-OJ 0.246894E-04 0.246892t-05 0. 246n92E-06
.,
0.)2627IE«00 0.499J07E-OI 0.496S67E-02 0.496J01E-OJ 0.4»621JE-04 0.446279E-OS 0. ^9620 IE-06
•In ntny •xparlBtnt* lh» rctpont* ol lnt«ratl •) >o occurs tpont *n«ou« I y In control anlatli. Ihl» background nay b« at»u*«d to b« •Illior
lnd*p«ndanl ol Ib* Induced roponto or eddltlv* In • ••cbanlatlc m»aa»r, II lha ipontoneoua and Induced retuontoi are astuved lo b«
Independent, Ihen the probability ol ob«ervln|| • reaponae ol either type et doae d la given byi P'ld) • q » II - a,-) PUodll, «hore 0 <-i| < I
denotes Hie tpontaneou* beckground rate. Under tbe oddltlvlly aiauaptlon, the background retpont* *ay be con>lder«d aa arltlng lro« an
ellectlve background dose >• > 0. «l Ih P'(d) - Pld « x). Although the' Manner In uhlch background rasponsv Is accoainodalad It crucial, the
extent to «h I ch Independence or eddlllvlly It Indicated by either biological theory or exper laental dale Is iononhal unclear al thlt Una.
48
-------
table 14
Excess Risk fo Hats at leukeale Over Background
(Based on NCI PCD Bloassayl
Hod* I
I.OOOE-OI
.OOOE-02
I.OOOE-01
I.OOOE-84
I.OOOE-0)
I.OOOE-06
i.uooE-oJ
.003E-08
Goodness el lit
test p-velue*
Maxliuai Likelihood EitUate ol Virtually Sale Dot* to Rats |pg/kg/dayl
Multistage
0.873338E»02 0.827988E40I 0.t2)601E«00 0.02)I7|C-0| 0.82JI27E-02 0.021I21E-03 0.e2JI22E-Q4 0.62JI22E-OJ >0.1»00
lnd«B*nd*ot
Pr ob11• •
0.«J»am«02 0.»26J4U»OI O.I2444aEtQ| 0.291421EIOO 0.6J99JU-OI 0.274622E-OI 0.»»I062£-02 0. 186n2t-02 0.6768
lnd>p*ndont
Log It"
0.a41)IOE<02 0.»I62IJE»OI O.J»J86IE«00 0.241262E-OI O.I10I18E-02 0. IH09JE-OJ O.SI»JOJ£-05 0. »68692t-06 0.6616
In •pendent
M.lbull"
O.B)IUJE«02 U.48I62«E«OI 0.289M4E»00 0. U1108E-OI O.I06IJ3E-02 0.642J9JE-04 0.58906IE-OJ 0.2J55J9E-U6 0.6)1)
Independent Oe
Nultlhlt"
0.8«86i2E»02 0.441I>4E*OI 0.243)22E»00 O.I12JIIE-OI O.J2J29JE-OJ 0.)8420oE-04 0.2I«8JO£-OS 0.|IIO»?E-06 0.6S62
Additive
Problt'"
0.8J7JJJE«02 0.566187E«OI O.J44626E»00 0.)4249)E-OI 0.342282E-02 0.34226IE-OJ 0.542218E-04 0. »4,22)SE-OS 0.67J9
Additive
. Loglt"
0.816617E«02 O.J6482JE»OI 0.}4207IE«00 0.»98)JE-OI O.JJ96I2E-02 O.JJ9389E-OJ 0.514S87E-04 O.SJ938JE-05 0.6111
Additive
Helbull"
0.8403'IE«02 0.)6I226E«0| O.J17097E«00 O.U4II9E-OI 0.334482E-02 0.3J4438E-01 0.314436E-04 0.».'44J34-0) 0.6718
Additive Gtmmm
Multlhll"
0.8JI>30El02 0.6I)6IOE«OI O.J9497IE400 0.39292JE-OI 0.3»27I)E-02 0.392697E-03 0.592694E-04 0. )<>2690E-Oi 0.6610
•Any p-velue grenter thin 0.30 Indlcetet en adequete lit ol the eiodel to tne dele.
••In aany experl>i*nl> th* response ol Interest elso occurs spontaneously In'control anlaials. Ihls background aay ba assumed to be eltlior Independent o| the
Induced responses or acdltlve In a Mechanistic Banner. II Ihe spontaneous and Induced responses ore assumed to be Independent. th«n the probability ol
observing a response of either type «f dose d Is given byl P*ldl • n, 4 11 - 4! Pld). Nhere 0 < q < I denotes the spontaneous backyruund rate. Under Ihe
addltlvlty assunptlon, Ihe background response nay be considered as arising from an ellectlve background dose » > 0, with P*(d) - f*('l « •). Although (lie
•anner In Hhlch beckgruund response Is acconuodated Is crucial, the extent to uhlch Independence or addltlvlty 8> (nd'lcaled by either biological) theory or
experlaental dale Is sctaevhat unclear at this tlee. i
49
-------
labl
91. Ill On*-Slded Loiter Confidence Haiti an Virtually Sal* Doses lor lUtt la »g/kg/dey la table |4 Uase4 on the yarlencu al toy-Hose
(Based on NCI PCS Bloassay)
Had. I
I.OOOE-OI
I.OOOE-02
I.OOOE-0) I.OOOE-04. I.OOOE-0) I.OOOE-06
I.OOOE-0?
I.OGOE-00
Multistage
0.)7i)l)E«02 0.1)ie84E«OI 0. J14006E 400 0.11K07E-OI 0. 111B80E-02 O.JJJ888E-OJ 0.»>afaE-04 0. :i J J888E -0}
Independent
Problt* 0. |7tt40IE<02 0.2407J7E-OI 0.9974S6E-04 0. IO)32lC-01 O.I96042E-01 0.36J3JiE-0» 0.2J097JE-IO
Independent
Loglt* O.ia«7»l£t02 O.S27032E-02 0.773747E-06 0. I|2»9E-09 0. I6I647E-I1 0.2J0974E-I7 0.129II2E-2I
Independent • •
Melbull* 0. ia't64IE«02 0. J67066E-02 O.JI5628E-06 0.2«}»'i£-|0 0.2I78»E-I4 O.IIt||lE-l« O.I46879E-J2
Multlhlt* 0. ial4»E\»02 0.2J77IJE-02 0. I267B1E-06 0.387124E-II 0.248JJOE-IJ 0.|2I977E-|9 0. JSJ026E-24
Additive • ;
Problt* 0. |3l94JE«02 0.947604E-OI 0.661969E-02 0.64I149E-OJ 0.6J89J6E-04 0.6J8696E-OJ 0.6JV67IE-06
Additive
Loglt* 0. (6»))E-07
0. 31J617E-07
i
0.72|3a»E-07
•In >aay *«p«rli*nt» th'» ropant* ol Inlarxt «lia occur! tpont«n»oi(t I y |n control •nlaalt. Ih(» bickgrouad my b* a*a»4 lo b* •llhar
lnd*p*nd«nt ol III* |qdL^»d r«>pan>*t or addltlv* In a ••chanlttlc •*na«r . II th» (pontaavous and Induod rciponiot *r* a»ua*d to b»
Independent, then the probability ol abiorvlng a raipont* ol either type at du*e d It given byi P'ldl > n «'(! - i)l Pldt, »h«r« 0 < <| < I
denote! the ipcntaneouk background rate. Under the addltlvlty ettunptlon, the background response nay be considered al arltlng |roi» an
ellectlve background dbie » > 0, •llh P'(d> - P(d * v). Although 'the oannar In "Men background response Is accovaodal ed Is crucl.il, (he
extent to oh I clt Independence or addltlvtty Is Indicated by either biological theory or experimental data Is souenlial unclear at this ll»o.
50
-------
fable 16
91.01 Onu-SI
0. I2Ollt-07
0.49J586E-08
0. 2)(<066E-Oa
0.99<ny exparlaent* the reiponte ol Intereit alia occurs *pont encoutl y In control anlvalt. Tblt background »«y bo a»uaed to be either
Independent ol thu Induced r*ipon>«> or additive In • •echenlttlc Banner. II the »pont«noou» end Induced retpontei are attuaxl to b«
Independent, then the probability ol obtorvlng a retponto ol either type at dote d U given byi P'ldt - 4 • IT- <)1 riiodll, oltare 0 <
ellecllve backgrotnd dote » > 0, ullh P*ld) • Pld * •). Although lh» aanner In. iitiicb background response Is accoonodalad Is crucial, gi>«
extent to uhlch I idupandance or addlllvlty Is Indicated by either biological tlmory or exparlnental data Is somexliat unclear at this I l«a.
51
-------
' .
Table 17i Excess Lifetime Cancer Risk From PCI) Exposure Derived Using the Multistage Model.
Daalsi NCI Total Malignancies Category
Estimated Individual exposure* Duration of Haxlmtm
Bxpoaure tyi>e Exposure .icenarlo During period or exposure Cl let lino average exposure** Likelihood
|mg>yl| (mg/kg/aay) |mg/kg/da9l (years) Estimates of
Excess Risk
\ •.
Ambient
Inhalation o deference scenarios '
o Exposure at the average urban atmospheric level I'.SxlO"1 1.4x10"' 1.4(10'' 70 6.2xlO~7
(5 ng/m1)
o Exposure at the average rural atmospheric level J.5xlO~* 1.4xlO~e 1.4x10"" 70 6.2x'iO~*
(0.05 ng/mj)
o Exposure, ot the PCB level of quant Itatlon for air 69 2. 7x10" J 2.7xlO"J 70 I.2XIO"1
(1ft ug/mj) '
o Exposure t>t a distance of 800 m (0.5 mile) down-
wind of a large capacity chemicals manufacturing
plant with PC Be present In the process stream at ...
25 mg/kg 2.9«10~J ' l.lulO"6 l.lxlO'* 70 4.9x10"'
o exposure at a distance of 600 m (0.5 mile) downwind
of a largt capacity Industrial Incinerator burning , ,
< wastes containing 50 mg/kg PCOa <2.4«10~2 • <9.4xlO"7 <9.4xlO"' 70' <4.2xlO~'
Ambient
Ingest Ion o Reference scenarios ' '
o Averag< adult Intake of PC Da via food during 197Q . '
as reportud by FDA <6,.9ilO~i <2.7xlO"5 <2.7xlO~5 70 <1.2xlO"5
o Ingest Ion of fish containing 2 ppm of PCBs (I.e.,
the 1977 proposed FUA tolerance level for PC 8s In
the edible portion of fish) 4.75 1.9x10"* 1.9x10** 70 8.4»10~5
\ . " '
951 Upper
Confidence Limit
on
Excusa Hick
I.OxlO'6
1.0xlO"B
2.0«10"J
a.ixio"1
<6.9xIO"7
2.0xlO"5
I.4xl0"4
52
-------
Table 17 (Continued)
Bxpoaure type Kxpoeura ccenarlb
Amblant ,
Ingeatlon o Ingestlon cf flah or water obtained from water
bod 1 oo down atrekm of chemical plants discharging
waatewater containing 100 ug/1 of PC Da
o Ingbatlm of groundwater drawn from walla located
down gradient from a landfill receiving waatea
containln) 50 mg/kg PC Do
o Consumers of water contaminated by dlachargca from
a typical aluminum forming plant with 66 general
hydraulic ayatema aeaumlng
- All ayatema contain SO mg/kg PCOa In July 1984
and thereafter
- All ayatvmu contain 1,752 mg/kg' PCBa from July
1984 to July 198S and contain SO ng/kg thereafter
-July 1984 to July 85
-after July 85
o Conaumere of water contaminated by dlachargera
from a typical petroleum refinery with 8 heat
tranafur ayatema aaaumlng
- All ay at fma contain 50 mg/kg PCOa July 1984
to July 1985 and contain 50 mg/kg^ thereafter
- All aysteua contain 176 my/kg PC Da from July 1984
to July 1985 and contain 50 mg/kg thereafter
-July 19B4 to July 85
-after July 85
- All aystena contain 441 mg/kg PCBa from July 1984'
to July 1985 and contain 50 mg/kg thereafter '
-July 1984 to July 85
-after July 85
Eatlmated Individual exposure*
During pe Clod 01! 'ekpoBUre llfefTino average
|mg/yf|
2.7«10"4
to 110
<7.7xlO"6
4,J»10"6 -
to 4.S»10~J
1.U10"4
to 1.1x10"*
4»5«10"' -
to 4.5xlO"J
' t
\
l.OxlO"7 ,
to l.lxlO~J
S.OxlO"7
to 2.1xlO"J
l.OxlO"7 -
to l.lxlO~J
i.2«10~' ,
to S.6xlO"J
l.OxlO"7 ^
to l.lxlfl"1
(ag/kgYdly)
l.lxIO"8 ,
to S.lxlO"1
O.OxlO"10
to l.SxlO"7
to 4.1xlO"6
to l.BxlO"7
to S.lxlO"8
J.OxlO"11
to 9.0xlO~8
to S.lxlO"8
to 2.2xlO"7
to S.lxlO"8
img/icg/
l.lxlO'8 to
S.lxlO-1
<1.0xlO"»°
•
1.8K10"10 _
to 1.8x10"'
to 2.1xlO~7
2.1«10"l° ',
to 2.1x10"'
1.2xlO"U
to 5.1x10""
l.JxlO"11
to S.lxlO"8
to S.lxlO"8
1.2xlO~U „
to S.lxlO"8
to S.lxlO"8
Duration of Maximum
expoauro l.lkellhooil
(yeara) '• Batliuatea of
Ktceai Rlak
70 4.9x10"' ,
to 1.1x10"'
70
-------
Table 17 (Continued)
Eat 1 nated
Exposure type Exposure acenarlo i>urln<| faflod
1
Occupational
Inhalation o lluforonce scenarloa
o Exposure at. the O3HA atandard for VCUn in air .
(1,000 ug/a') 2.1xlO~J
o exposure at the. level of quantl tat Ion' for PCDa
In air (10 ug/n1) 2}
o Expoaura at the HIOSII reconnended atandard for
PCDa In air (1 ug/a1) 2.1
1
o Transfer and handling operation acenarloa • .
o Loading/unloading a liquid aaaunlng PCBa are preaent ,
In the liquid at 25 ag/kg. 6.0«10~2
o Load InQ/unloid Ing a powder aaaualng PCBa ace preaent ,
In the powder -i 25 ag/kg • 2.9x10"'
o Loading/unloading a powder aauunlng conpllance with '
the rtSI'A nuleanr: duet atandard and aaaunlng PCBa .
are pri aunt In the powder at 25 ag/kg 8.8x10"*
Individual ex
ot fiKp6surfi
|B9/K9/nayl
9.0xlO"4
9. OxlO"5
2.1x10"'
1.1x10"'
J.«xlO-*
poaure*
LUe'tlue average
|ag/Kg/aayi
5. OxlO"2
5. OxlO'4
5.0xlO"S
1.1x10"'
6.2xlO"7
1.9xlO"5
Duration of
expoauto
(yearat ••
18.5
18.5
1U..5
18.5
ia.s
IB. 5
Haxliiua
Likelihood
Kal liiatea of
Exceau Rlak
2.2
-------
Tabla 17 (Continued)
Exposure typu
Occupat lonal
Inhalation
Exposure ticenarlo
o Exposure to fugitive emissions for « worker
stationed sla meters downwind of leaking
equipment assuming PCBs are present In the
emitted chemical at 25 mg/kg
o Expouurd during open-eurface tank operations
(e.g., decreasing tank) assuming tank liquid
temperature of 7S*C and assuming PCBs are
present In the liquid a 25 mg/kg
o Exposure to evaporative emissions during
non-spray coating operations assuming a
coating temperature of 75*C and assuming PCBs
are present In the coating at 25 mg/kg-
o Exposure to paint mists during spray painting
assuming PCBs are present In the binder at
25 mg/kg
o Exposure to paint mists during spray painting
assuming PCBs are present In the solvent at
25 mg/kg
o Exposure to paint mists during spray painting
assuming PCBs are present In the pigment toner
at 25 og/kg
o Exposure to evaporation emissions during liquid
product formulation assuming a liquid temperature
of 25*C. open formulation tanks, and PCB concen-
trations In the liquid at 25 mg/kg
Estimated Individual exposure* Duration of
OurlnQ' pflflod 'of Mpndlire "Lifetime average exposure
Img/yrj • (mg/Kg/aay) |mg/Kg/aayi . (year a) *•
9.2*10"^ 1.6xlO"5 ?.OxlO"* 18.5
6.5xlO~2 2.5xlO"6 1.4x10"' 18.5
4.6xlO"! 1.8xlO"5 9.9x10"' 18.5
S.Sillu"11 2.2-10"* 1.2*10"* 18.5
«.4xlO"1 2.5xlO"5 1.4x10"* 18.5
tt.2xlO"2 1.1x10"* 6.9xlO"7 18.5
2.5«10~J 9.axlO"fl 5.4xlO"8 18.5
Hi I Imt.m
Likelihood
Eatlmetes of
Excess Risk
B.exjO"'
«.2xlO"7
4.4x10"'
5.1x10"'
6.2x10"'
2.4x10""
951 Upp»(
Confldunce Limit
on
Exueus Hlsk
1.0x10''
7.1x10"'
8.8x10"'
1.0x10"*
S.lxlO"7
4.0xlO"8
55
-------
Table 17 (Continued)
Exposure type
Exposure scenario
Estimated Individual exposure* Duration of Haxlaja 95% Upper
Pur ImT period 'ot''a«pBBP?f Lifetime average exposure Likelihood Confidence Lluli
I«g7ylr'• lag/It<]/dBy| |ng/kg/day| (years) •• Estlaites of on
. . Excess Risk Excess Risk
Occupational
Innalat Ion
Exposure to alsts during air-blast pesticide
spraying assuming PCBa are present In the active
peatlcldo Ingredient at 25 ag/kg
o Exposure to evaporative ealaalons during grain
fumigation assunlng PCDa are present In the
fumlgant at 25 ag/kg
o Expoi:ui:e to oil mists during operations such as
printing and netalworkIng assunlng compliance
with the OSIIA standard for alneral oil nlat and
assuming pCBs are present In the oil at 25 ag/kg
o Exposure to evaporative ealsslons during foamed
plast'.cs manufacturing operations assuming PCDs
are present In the blowing agent (which con-
stitutes approximately 17 percent by weight of
the foam formula't Ion) at 25. mg/kg
8.6x10"* J.JxlO"8 I.8xlO"8
<6.5xlO'6 <1.6xlO"6
l.lxlO"5 6.2x10"*
1.9x10"' J.JxlO"6
2.9xlO~l
,
l.OxlO"1
18. S
18. &
18.S
18.5
"9
B.OxlO
-------
Table 17 (Continued)
. estimated Individual exposure'
Bxpoaure type Kxpouuce acenarl'b During patioa
img/yri
Occupational
Inhalation o Bxpoaure to evaporative emissions during
plaatlc manufacturing operations assuming PCBs .
are preunnt In the plaatlc at 25 mg/kg 7;1«10"*
o K.poauie during manufacture of aaphalt roofing <2.1xlO"l
* products to (.4
o Exposure to evaporative emissions during paper
manufacturing assuming PCua are present In waste- •
paper furnish ati
- 12 mg/kg 9.2*10"}
- 5 mg/kg 1.9*10"i
- 2.5 ng/kg . -2. 0*10 .
o Exposure to evaporative emlsalona during paper
manufacturing assuming PCua are present In the
printed Ink of tJaetepaper furnish ati
2 mg/kg 6.0x10"'
- 25 mg/kg 7.5*10~i
- 50 mg/kg 1.5*10"*
o Exposure durlilg rereflnlng of waste oil assuming
PCDs «r« present. In the waste oil ati
2 mg/kg 7.8*10"'
- 25 mg/kg 9.7*10"*
50 mg/kg 1.9*10-*
oc exposure I. iieflme average
img/Kg/tiay!
2.8xlO"5
<9.0xlO"' .
to 2.5x10"*
7'.8*10"7
2.4.10'*
2.9*10";
5.9»10"7
l.OxlO"7
1.8*10"'
7.4*10"'
l«g/icg/aay»
1.5«10"5
<5.0*10"' to
2.0*10"!
8.4*10"'
4.1»10"7
l.'6*10"7
1.2*10"7
1.7*10"7
2.1*10"'
4.1*10"'
Duration of
exposure
(years) ••
18.5
18.5
18.5
18.5
18.5
18.5
18.5
18.5
18.5
18.5
18.5
Max Inurn
Likelihood
Kat lua tea of
Bxcesi- Risk
6.6i!lO"6
to li.2xlO"J
i.'7i!io;«
7.'l'lO"?
1.4xlO"7
7.5«10~§
9.1«10"7
1.8.10"'
95t Upper
Cunfldunce l.lml
on
B.cuas Hlak
l.lxlO"5
to'l.OxlO"4
6i2xlO''
1.2xlO'7
9.6.10'^
1.2x10';
2.4*10'7
1.1x10'?
1.5*10''
1.0x10''
57
-------
Table IT (Continued)
Exposure^ type Exposure scenario
Occupational .
Inhalation o Exposure during garment dry cleaning operations
assuming ?CI»a Ire present In the cleaning fluid
at 25 ng/kg
o Sampling and Maintenance operation acenarloa
o Exposure during on-line repair of equipment
leaking gas/fluid containing PCOa at 25 ng/kg
o Kxpoauie during aampllng assuming the proceaa
stream contains PC Da at 25 ng/kgi
o Expoauie while cleaning equipment containing a
fluid In which PC Da are assumed to be present at
25 ng/kg
o Expoauie while repairing equipment off-line
assumlrg the equipment contains a fluid In
which IClla are present at 25 ng/kg
o Bxpoauie during filter removal assuming PCBs
are present at 25 mg/kg
o Expoauie during removal of atlll hot tons
assuming PCBa are present In the stilt
bottoms ati
200 nig/kg
2500 mg/kg
- 5000 tug/kg
o Exposure during cleaning of spilled liquids
assuming PCOa are present In the liquid at
25 mg/».g
Estimated Individual
During perioa or exposure
T9/y 1 l-9/ig/a.,
1.6xlO"2 6.1xlO"7
. 1.7xlO~l 6.6x10"'
1.6 l.4xlO~*
S.OxlO"1 2.0xlO"5
2.5x10"' 9.8x10"'
l.lxlO"1 5.1*10*'
9.8xlO~l 1.8x10"!
25 9*.8xlO"*
i.OxlO"1 7.8x10"*
exposure*
LlieTTmu average
rl (ng/kg/dSJi
1.4xlO"7
1.6x10"'
7.7xlO"5
l.lxlO*5
5.4x10"'
2.8x10"'
2.1x10"?
2.8(10**
S.4xlO"*
4.1xlO"8
Duration of Maximum
exposure Likelihood
(years) •• Estimates of
Excess Disk
18.5 l.Sxlft"7
18.5 1.6.IO"6
18.5 1.4»:10"5
18.5 4.9>:10~'
18.5 2.4iilO~'
18.5 1.2x10"'
18.5 9.1alO*J
18.5 1.21.10"*,
18.5 2.4:MO~*
18.5 1.9:
-------
Table 17 (Contlimed)
Exposure type Bxpouure scenario
Occupat lonal
Inhalation o Exposure to evaporative emissions for a worker
stationed one meter downwind of a leaking
hydraulic uyatea operating at 75 C assuming
- All aiatnns contain SO ng/kg PCBa In July 1984
and tl.eriiafter
- All aystona contain 1.752 ng/kg PCBa from
July 1964 to July 1985 and contain SO ng/kg
thereafter
-July 84 to July 85
-after July 85
o Exposure to evaporative ewlsalona for a worker
atatlonod throe netera downwind of a leaking heat
tranafei ayatea assuming
- All systems contain SO mg/kg PCBa July 1984
to July 1985 and contain SO ng/kg thereafter
- All syatons contain 176 ng/kg PCBa fron July 1984 •
to July 1985 and contain SO ng/kg thereafter
-July 84 to July 85
-after. July 85
- All uyst-jma contain 441 ng/kg PCUa from July 1984
to July 1985 and contain SO ng/kg thereafter
-July 84 to July 85
-after July 85
Eatlnated Individual exposure* '
Dur lnj~p"6r loo or exposure |. IfBlTlna <.vora<
"^•g/yri img/icg/aTy) . |IgVxg/dayj
* i "
l.SxlO"1 5.9«10"' 1.2x10"'
5.J 2.1x10"* 6.2x10*'
l.SxlO"1 S.9xlO"6 6.2x10"'
•'
S.8xlO~l 2.1xlO"5 l.lxlO"5
2.0 7.8xlO"5 1.1x10"*
5.8x10"' 2.1xlO"5 l.lxlO"5
5.1 2.0xlO"4 l.SxlO"5
S.BxlO"1 2.1xlO"5 l.SxlO"5
Duration of
ue expoaura
*~ (yeara) •«
18.5
1
17.5
.18.5
1
17.5
1
17.5
Haxlnum
Llkellhoud
Estimates of
Excesii Rlak
1.4x10"'
2.7x10"'
2.1x10"'
5.7x10"'
5.7x10"'
5.7x10"'
6.6x10"'
6.6x10"'
951 Upper
Confidence LI*
on
Kxcuaa Hlak
2.4x10"'
4.6x10"'
4.6x10"'
9.6x10"'
9.6x10"'
9.6x10"'
l.lxlO"5
l.lxlO"5
59
-------
Table 17 (Continued)
Exposure type Bxpoauru acenarlo
Occupational
dermal o Exposure to hydraulic system operators and
malntcnancn workers assuming
- Ml systems contain 50 mg/kg PC Da In
July 198-1 and thereafter
- Ml systems contain 1,752 ng/kg PCBa from
July 1984 to July 1985 and contain SO
mij/kg thereafter
-July 64 to July 85
-after July 85
o Exposure to heat transfer system operators and
maintenance workers assuming
- Ml systems contain 50 mg/kg PCBa July 1984
to July 1985 and contain 50 mg/kg thereafter
- Ml ayatoma contain 176 mg/kg PCBa from July 1984
to July 1985 and contain 50 mg/kg thereafter
-July 84 to July 85
-after July 85
- Ml systems contain 441 ma/kg PC Be from July 1984
to July 1965 and contain SO mg/kg thereafter
-July 84 to July 85
-after July 85
'.. Eat (mated
• Our Ind per loo c
|mg/yr| '
17..7
620
17.7
'" ' !
17.7
62.1
. 17.7
1S6
17.7
Individual exposure*
>f etpgBOTi CTTBfTisa average
(SgYkg/aay) [•g/tg/aayi
6.9x10** 1.8x10'*
2.«xlO"? 7.1x10'*.
6.9x10"* 7.1x10'*
6.9x10"* 1.8x10'*
2.4x10"? 4.0x10'*,
6.9x10'* 4.0x10'*
6.1x10'? 4.6x10"*
6.9x10"* 4.6x10"*
Uuratlo-i of
exposure
lyenrs) ••
18.5
17.5
18. 5
1
17.5
1
17.5
Max liium
Likelihood
Estimates of
Bxcoa.i Risk
1 .7x10"*
1.1x10"*
1.1x10"*
1.7x10"*
1.8x10"*
1.8x10"*
2.90x10"'
1. 09x10"*
951 Upper
Confidence Llml
on
Excuss Nlsk
2.8x10"* •
5.2x10'*.
5.2x10"*
2.8x10'*
2.9x10'*.
2.9x10"*
4. 8)xlO~'
1.81x10*4
60
-------
Table 17 (Continued)
,
Exposure type Bipoaura scenario
Occupational
dermal
o
i o
0
0
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
.
-
-
o
Transfer and handling operations assuming PC Be
are present at 25 mq/kq
Loading/unloading liquid
Loading/unloading powder
Processing operations asuuilnq PCBs are
present At 2S mg/kg
Closed process operations
Open surface tank operations
Spray painting operations
Grain tunlgatlon operations
Air-blast pesticide spraying operations
Non-spray coatlna operations
Product formulation operations
Product fabrication operations
HetaKorklng operations
Newspaper production
Plastics manufacture
Dry cleaning of garments
Sampling and maintenance operations assuming PCBs
are present at 2S ng/kg In the process stream
On- 11 nd repair of leaking equipment
Sampling process stream
Cleaning equipment
Off-line repair of equipment
Removing filters
Removing still bottoms asuumlng PCBa are present
In still bottoms ati . j
200 me/kg
2500 i.g/Kg
SOOO >ig/kg
Spill cleanup
Estimated Individual exposure'
During* "fter lod"bf e«poatir»"
img/yrj
6.0
5.8
1.0
IS
IS
H
?s"
6.0
•• .'
IS
IS
15
IS
•
0.2S
IS
1.0
l.S
1.0
24
loo
600
0.75
(ag/kg/aay|
2
2
1
S
S
1
S
2
5
5
S
5
9
S
1
S
I
9
1
2
2
.4x10"*,
.4x10"*
.2x10"*
.9x10-*.
.9x10'}
.9x10-*
.6x10"?
.9x10"*
.4x10"*
.9xlO"f
.9x10"*.
.9x10"*
.9x10"*
.8x10"$
.9x10"*
.2x10"*
.9x10"*
.2x10"*
.1x10"*
Ixlfl
.IxlO'2
.9xlO"5
Liretlme average
(•g/kg/day|
1.1x10"*,
1.2x10'*
.4x10"*
.2x10"*.
.2x10'
.2x10"
.'2xlO~
.1x10"
.2x10"
.2x10"
.2x10"
.2rlO"*
.4x10"?
.2x10"*
.4x10"?
.2xlO"{
.4x10"*
5.1x10"*
6.0x10"*
l.lxlO"2
1.6xlO"b
.Duration of
exposure
(years) ••
18
18
18
18
18
. 18
18
18
18
18
18
ia
18
18
. 18
18
18
18
ia
18
18
.5
.5
.5
.5
.'s
.'s
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
.5
Hax Inum
LlkdllhmxJ
Estliiatua of
Kxcean Risk
S.
S.
2.
•
•
.
2.
1.
2.
1.
2.
2.
2.
S.
7.
IxlO"5
8x10"?
4xlO~*.
4x10"'
4x10:*
4ltO"*
1xlO"s
4x10"*
4xlO~*.
4x10"'
4x10"*
4xlO"f
4x103*
4x10"^
dxlO"5
1x10"*
6x10"'
IxlO"1
IxlO"6
95t Upper
Confidence Mi
on
BNUUUS Hlek
9
8
4
2
4
2
4
1
I
9
1
!BXIO"S
.7x10"*
.4x10"!
.4x10"*
.4x10"*
.sxifl-;
.4x10"*
.6xlO"5
.4x10"*.
.4X10"J
.4x10"*
.4x10"*
.OxlO'J
.4x10"*
.1x10"?
.4xlO"|
.7xlO"5
.8x10"*
'.SxlO"1
.2xlO"5
61
-------
Table 17 (Continued)
Exposure type Exposure ucanarlo
Conauraer
Inhalation o Exposure resulting from uae of apace deodoranta
aaaumlng PCDa are present In the product at
25 mg/kg
o Exposure resulting from uae of moth control
products as iuiilng PCua are present In the
product at 25 mg/kg
o Exposures resulting from painting the Interior
of a houaa la sum Ing CB are preaent In the
pigment at 25 mg/kg
o Exposures resulting from painting the Interior
of a houaa aaaumlng PCBa are present In a realn
Intermediate at 25 mg/kg
o Exposures resulting from Inhabiting a home with
a newly painted Inferior aaaumlng PCBa are
preaent In the p^.nt pigment at 25 ng/kg
o Exposures resulting from Inhabiting a home with
a newly painted Interior aasunlng PCBa are
preaent In a realn Intermediate at 25 mg/kg
o Exposures resulting from uae of spray paint
aaaumlng PCBa are preaent In a realn or
aolvent at 25 mg/kg
o Exposures resulting from uae of apray palnta
aasunlng PCBa are preaent In the pigment at
25 mg/kg
Eat !•« ted
Pur Ing perioa
TBgTFtr
1 7xlO'1 •
J.2xlO"2
1.7*10"*
7.7«10"*
2.6«10"2
1.2*10"1
2,9xlO"2
2.2xlO"J
Individual exposure1
ot exposure LuBfTme average
l»g/»g/:10'9
1.8KIO"7
8.4H10'7
3. HitlO'7
J.OalO'8
95% Upper
Confidence Llml
on
Excuaa Hlak
.,.„.
4.6x10"'
».6«lO-»°
4.4xlO"9
J.OxlO"7
1.4x10''
6.5xlO'7
5.0«IO'a
62
-------
fable 11 (Continued)
Estimated Individual eiipoaiite*
During pecioa 01 e
• ' " • • j
Duration of
exposure
(years) • »
xpoaure typo
Exposure scenario
During
• i^i- B^J y •
rr (.ireciue average
|mgvicg/
-------
Exposure type
Consumer
Inhalation
Conaumer
dermal
j Estimated Individual exposure* Duration of Maximum
Exposure acenarlo OUr(n4 peflbd At exposure LlfeTTno avurayo exposure Likelihood
I"5 'V *) l»g/Kg/"ay) (•g/xg/tiy) lyearat ** eatlmatoa oi
txceaa Disk
o Expoauie resulting from use of peatlclde apray*
aaaumlng PCBa are preaent In the active Ingredient * i .1 i
at 25 mg/kg • 2.2x10"* B.CxlO"' i.8x|0~' 55 ).0»IU~'
o Expoauie resulting from uae of peatlclde sprays <
assuming PCBa afe present In the Inert Ingredients ! i « t c,
at 21 mg/kg l.lulO"1 4.1x10"° l.'xlO • 55 1.5»:10'e
o Expoauiea resulting from uae of apray
cleunlng/dtalnfactanl producta aaaumlng that
PCBa are preaent In a conatltuent (that accounts
for 50 pnrcent of the weight of the product) at , , , ,
25 mg/kg 2.1x10"* 9.0x10" 7.1x10" 55 ' l.l:ilO"°
o Expoaurea resulting from Inhabiting a new home
containing plaallc building materials which are . , .
aeaumed to contain PCBa at 25 mg/kg i.l 1.2x10"* 2.7x10"' 6 l.2clO~&
o Expoaure:! resulting from uae of deodorant soaps ,
aaaumlng PCBa ate preaent In the surfactant at ' , _ , • _
25 mg/kg J.lxlO"1 8.2x10"' 8.2«10"7 70 1.6«10"7 .
to 19 .to 7.4x10"* to 7.4x10"* to 1.1x10"*
o Expoaurea resulting from use of akin lotlona aaaumlng ...
PCIla are preaenl In the surfactant at 25 mg/kg 6.4 2.5x10"* 2.5x10"* 70 1.1x10"*
o Expoaurea reaultlng from handling of printed matter
assuming PCbs ate present In the Ink pigment at . , '
25 mg/kg 5.9x10"^ 2.1x10"' 2.0x10"* 60 B.ttxlO'1
o Bxpoauceu raujiltlng from painting the Interior of
house aa lulling PCBa are preaont In the pigment at
25 mg/kg 7.0X10"1 2.7xlO'7 5.5xlO"B 14 2.4xlO"8
951 Upper
Confidence Urn
on
Exui-aa HUk
5.0«10"7
2.5x10"'
5.2x10"'
2.0xlO~*
6.0x10"] to
5.4xlO~*
1.8x10"*
1.4x10"'
4.0x10'"
------- |