-------
British Studies
Reid and Buck (1956) —
    Reid and Buck (1956) studied the causes of death of men dying while "on the
books" of the British National Coal Board coking plants during the period
1949-54.  This included both retired and actively employed workers.  Causes of
death were ascertained through the funeral fund of the National  Union of
Mineworkers or through a vital statistics search of the General  Register Office.
The authors analyzed mortality for the currently employed and retired workers
separately.
    For the actively employed, information on age and job distribution was
obtained from a special census taken in 1952.  Additional information on the
nature and duration of different jobs held in the plants was obtained from a
sample of 10% of the workers.  Total man-years of exposure over the period
1949-54 were divided proportionally according to the age and job distributions
found in the 1952 special census.  Expected deaths were derived by multiplying
the accumulated man-years in each age and job category by the comparable cause
and age-specific death rates derived from a "large industrial organization"
during the period 1950-54.  Although the authors did not disclose the identity
of this large industrial organization, they do state that the derived death
rates for this industry were similar to those of civil  servants  of the General
Post Office, 60 years of age or younger, during the same period.  Data on civil
servants were not available beyond the usual  retiring age of 60.
    The coking plant workers generally fell into four main groups.  The first
group consisted of men involved in operating  the coking ovens, driving the ram,
filling the oven, clearing the hydraulic main, etc.  The second  group was
involved in the recovery of by-products such  as tar, ammonia, and benzole.   The
third group was composed of laborers whose duties and contacts with the
                                       98

-------
processes varied greatly.  The fourth group included maintenance men and
craftsmen who occasionally were in contact with the processes.  The number of
workers falling into each of these four job groups was not reported.  Mortality
by cause and occupational exposure is reported in Table VI-17.
    There was a significant difference between observed and expected mortality
for all other cancer combined (minus respiratory cancer) for oven workers (24
observed vs. 16 expected; P < 0.05, two-tailed test).  For respiratory cancer,
however, there was no increase in observed mortality over that expected.  If the
occupational classifications are divided* according to whether the men were ever
employed at any time as oven workers or never employed as oven workers, oven
workers would have a significantly elevated risk from cancers at all sites (40
observed vs. 32 expected; P < 0.05, one-tailed test) and an elevated risk (14
observed vs. 10 expected) of respiratory cancer, which is not statistically
significant (P <0.05).  A significant excess risk of death (except respiratory
cancer) is apparent in men who never worked at the coke oven (205 observed vs.
162 expected; P < 0.01, one-tailed test).  Men employed at any time as
by-product workers do not appear to be subject to an excess risk of respiratory
cancer, "cancer all sites combined" or "deaths all  causes combined except
respiratory cancer."  By contrast, men who were never by-product workers have a
significantly elevated risk of death excluding respiratory cancer (254 observed
vs. 218 expected; P < 0.01, one-tailed test).
    Twenty workers who died from lung cancer while still  on the company payroll,
and for whom detailed occupational histories were available, spent an average of
23.0 years in the coking plants  and 16.3 as coke oven workers.  These figures
are not appreciably different from the average duration of employment for men of
    *This division was proportionally distributed according to the work
histories of the 10% random sample (800 workers).
                                       99

-------














UJ
o;
Z3
co
o
CL.
X
UJ

1

z.
o
HH
f—


0
0 ^£
O 0
«=C 3
CD
co
Oi -0
LU C
NX |^
Q;
O T3
3-r-
CL>
r-OC
•y
— 1
_J
| 	

g~-l
^r


%
|^^
r— 1
|
1 — i


UJ

CD

|__


















00

£- O
CD T3 3
> CD T3 CO
0) >> O C-
Z O £- CU
i — CL-M
C 0. 1 t-
CD E >> O
2: UJ 00 3


CD
-a o
CD CD 3
>3 ^> E "C3
f O 'r~ O
0 r- 1— £-
+J Q. Q. CO
oo E >, I s-
•r- UJ C >> CD
:n «c CQ ^
C £_
JyJ CD +J 00 O
c_ 2: « ro 3
O
3

^i CI
.£3 CD
>, t- 0
4-J CD "O
•r- > CD CD OO
r— CU >->-^ i-
ro z. o o , E CD
O T- -^
r- 1— £-
CL O
E 5*^ "^
LU C
«=C c:
C CD
CD -t-3 >
2: ro o


CD
0
c: co
ro £_
c: cu
CD ^
T3 4J £-
r— CO
CD T- 3
ar ro

o

_jj
4-» O CO
OO 3 £_
ro "O cu
	 1 0 sx
£_ £_
>5 CL O
J3 1 3

>> ca

-^
t
ro
4-> CO

o cu
•y C N^
CU £—
> O
O 3
























UJ





O







UJ





o







UJ





o







UJ





o





UJ



o





UJ




o






UJ





CD
























r-- LO co
r-l LO VO
r- 1



r^ LO cr>
r— 1 LO O1
r— )





IO CO CO
i-H LO




<^f vo tO
f— H ^^






CO t— < i— 1
f-l  O^
CM





LO tO O~l
<— < •=!•



C^l
«^- ^ CD
CM LO







H- 00
>5 CO O)
t_ i_ co
O CU 3
4-> o ro
ro co
£_ c_ ro
•r- CD 0 t-
CL 0 CD
co c i — .c

o; o et o




co
r-)
CM



«^J-
LO
CM





t— 1
r**.




CM
10

;




CM
to
t-H

t
^=
LO
o
CM





1 —
CM
r— 1



t—4
r-H
i— 1



CTl
00
1-1

•=3-
O
CM



LO
CO



co
CO





LO
to




^f-
r^








>.
C_
cn O
C 4-3
•r- ro
-a s~ s-
i— 3 •<- CD
ro i — ex o
4-> O 00 C
o x 01 ro
(— CD £- O



-a
0
r"
t_
CD
CL

CD
f~
40

£—
O
4—

i.
40
co
3
•o
C
r-

T3
CD
r—
l|—
O
CD
n.
CO
c
3

C
ro

c
o

T3
CD
CO
ro
o

^— X
CD
cn
ro

c_
0
4—
T3
O)
40
CO
3
•"-3
-o
ro

00
.(_>
ro
CD
•a

-a
CD
4-3
O
CD
a.
X
CD

II

UJ

«
CO
c~
4-3
ro
CD
TD

T3
CD
>
s_
CU
co
_a
o •
^j*
II LO
CTl
O r-1
* 1
CT^
<^J-
CTl
r-l
£_
O)
o
E

sz

cu
c~
4-3

> O
Y^ C_
£—
"O U
OJ
i % C"*
t 'I—
o
CL CO
CU -r-
£_
CU £-
_Q CD
O
O C
4-3 ro
o
C >-j
ro t-
cu o
f— j |,i^
ro
^>. <
r~ *r~
i^ f~>
ro oo
.a CD
0 £-
£_
CL cn
t~
00 -i-
ro T3
3 3

= 73
CO X
t- CD
a>
O r—
c ro
ro 4-3
o o
4-3
r— z
r—
eC CD
= _C
4~* P

t-
o
4—












































•
"^T
co
CD
4-3

-a
CD
r—
•r-
ro
^,^
1
o
3
4-3

LO
CD
CD

V
D_
N 	 *

-o
a>
o
CD
CL
X
0)

4—
O

00
00
ai
o
X
a>

c
•r"

J>^
r^
4O
c~
ro
0
•i—
4-
•c—
C
cn
•i—
CO
Cc^>
















































•
4-3
OO
cu
4-3

T3
CD
r—
•r-
ro
-]_>
CD
c
o

LO
O
o

NX
a.
^ — *

T3
CU
O
CD
CL
X
cu

4-
O

OO
00
CD
O
X
}
r^
4->
C
ro
o
•r-
4 —
•p-
C
cn
•r-
CO
»/—




100

-------
the same age included in the random sample of 800 (25.3 years in the coking
plants and 16.7 years as oven workers).  Comparison of "average" employment
duration may not reflect differences in length of employment between the lung
cancer cases and the total random group, however.  It is possible that a number
of recently exposed workers in either group may have worked for only a short
period of time, which might bias any comparison of averages.
    The number of retired workers was not known; only the number of retirees who
died was known.  Therefore, for retired workers, the proportion of respiratory
cancer deaths to all cancer deaths and the proportion of all cancer deaths to
total deaths were compared among occupational groups (oven workers, by-product
workers, laborers, maintenance workers, and foremen) by last job worked.  They
were also compared by whether or not they had ever worked as oven workers and
whether or not they had ever worked as by-product workers.  No differences were
found by either comparison.  Since the ages of death of the retired workers were
not known, mortality was not compared by age.
    The authors reported a significant (P < 0.05, one-tailed test) excess in
other than respiratory cancer mortality for workers whose last job was at the
coke ovens.  No excess in respiratory cancer was seen however.  For workers who
had ever worked at the coke ovens, there was no significant (P > 0.05) increase
in either respiratory or other cancer.  For retired workers, no difference was
seen in the proportions of cancer deaths.  The amount of confidence that can be
placed in the validity of the results of this study is in question, however,
because of the lack of details in the description given by the author regarding
the methodology and conduct of the study.  The authors fail  to adequately define
the basic study population.  It is unclear whether the study population includes
all men who ever had a record of employment in the coke plant during the period
1949-54, since the author refers to an "average" of 8,000 men employed in
                                      101

-------
National Coke Board (NCB) coking plants or just those found through the special
1952 census.  If the cohort consisted of workers  employed in 1952, and since
there was little or no follow-up of any of these members, it appears that this
study is little more than a cross-sectional study of mortality in a conglomerate
of several different coke plants.  As much as can be determined, the observed
deaths are only those deaths of members of the study group who were employed in
the period 1949-54.  Also, it should be noted that the number of lung cancer
deaths observed may have been deficient since only men dying while "on the
books" of the coking plants during the period 1949 to 1954 were included.
Lloyd (1971) reported (apparently from communication with Reid and Buck) that
men were  removed from the books after prolonged absence from work.
    Since follow-up after 1954 was nonexistent, latent effects were not
adequately considered.  Furthermore, the death rates utilized in calculating
expected deaths were those prevailing in an unknown "large industrial
organization."  It is not known how they were derived or defined.  Therefore,  it
cannot be said with any certainty that they are compatible with whatever
definition the authors utilized to derive the study population.  Regarding the
retired workers, comparison of proportionate mortality without any
age-adjustment must be viewed with some skepticism.  In short, this study leaves
many unanswered questions and is so ambivalent that it is difficult to place any
confidence  in the  study  results.

navies  (1977, 1978) —
    navies  (1977,  1978)  reported on the mortality experience from May  1954 until
June 1965 of 610 coke oven workers at two  South Wales coke works  (Works A and
B).  The  610 workers employed at the two plants had 6261.5 man-years of
follow-up;  88 had  died during the follow-up period.  Male mortality  rates for
                                       102

-------
 England and  Wales  (average  for 4 years,  1958-61) were multiplied times the
 person-years of  follow-up in  each  age  category to obtain the expected deaths for
 the coke workers.   Observed and expected deaths were for total mortality from
 malignant neoplasms  of  different sites,  cardiovascular mortality, respiratory
 disease mortality,  and  mortality from  other  causes.  The Standard Mortality
 Ratio  for the two coke  works  was 92.   There  was no significant
 (P  < 0.05) excess in mortality  for any of the diseases evaluated including
 cancer  of the lung  (8 observed  vs. 8.94  expected), cancer of the bladder and
 kidney  (3 observed  vs.  1.9  expected),  and respiratory disease (14 observed vs.
 12.6 expected).  There  was  a  significant  (P  < 0.05, one-tailed test) negative
 difference between the  observed and expected deaths from cardiovascular disease
 (29 observed  vs. 41.36  expected).
     The  follow-up period as  reported  in this study was 11 years (1954-1965).
 The follow-up period would  have been longer, of course, for those workers who
 started work  before 1954.   Without further information, however, the follow-up
 period  in the Davies study,  particularly given the small  sample size (Number =
 610), may not be adequate to detect differences in lung cancer mortality  since
 the latency period from exposure to the start of lung cancer may be  as  long  as
 20 to 30, years.

 Ceilings  (1978)--
    Collings  (1978)  conducted  a follow-up study of 2,854  male coke workers
 employed  in 14 coke plants scattered throughout Great Britain.   To be included
 in the study, these workers  had to  have attained  a  minimum  of 1.5 years
 continuous employment at the plant  ending in July  1967.   They were subsequently
 followed 9 years  from August 1, 1967  to August  1976,  and  their mortality
experience was tabulated. The cohort was derived  from  lists provided by  the
                                      103

-------
coke plants.  For each person in the cohort, a questionnaire was submitted to
the respective coke plants asking personal  information, work since joining the
coke industry, and work prior to date of entry; completion of the questionnaire
was arranged by a senior medical officer familiar with the works.  Three
distinct occupational groups (nonovens, part-ovens, and ovens) were designated.
The "non-ovens" category included 392 men who had no contact with the ovens.
"Part-ovens" included 742 men with some occasional contact with oven work.
"Ovens" was comprised of 1,615 men who had at least one specialized oven job
prior to August 1, 1967.  Length of employment was tabulated for the "ovens"
group but not for the other groups.
    For comparison, expected deaths were derived  in two separate ways.  In the
first method, the mortality experience of men in  the study cohort was compared
to that of  all men in Great Britain.   Person-years at  risk were accumulated in
the appropriate age, calendar-time period,  cancer latency period, and
occupational  groups.  Standard  population death  rates  were applied to the
comparable  person-years categories to  derive  expected  deaths and finally
Standard Mortality Ratios  (SMRs).   In  the second method, the author  derived a
"comparative  mortality  figure"  (CMF) for each occupational category  (ovens,
non-ovens,  and part-ovens).  The CMF was  derived  as the ratio of  the  sum of the
observed mortality across  all age  categories  to  that  of the  sum  of the expected
mortality.
                                       104

-------
      Expected mortality  for the CMF was derived In the following manner:
 Number  in age
 group of
 occupational
 category
Number of deaths for the
particular cause in the
age group of the occupa-
tional category
Number in age
group for all
occupational
categories
Number of deaths for the
particular cause in the
age group of the occupa-
tional category
  All  deaths for the
x particular cause
  in the age group
                                          = Expected deaths for the occupational
                                            category by age group.
     With regard to latency and specific job within the coke works, only lung
cancer mortality appears to be somewhat excessive but not significant when the
coke plant worker rates are contrasted with rates in Great Britain (45.0
observed vs. 35.7 expected, P = 0.12).  If only the coke plant workers in
England and Wales (not Scotland) are considered, and population mortality data
from those two countries are used to derive expected deaths, then lung cancer
mortality is significantly elevated (41 observed vs. 32.4 expected mortality,
P < 0.05).  However, lung cancer mortality is not significant when
manually-skilled (40.8 expected), partly-skilled (39.5 expected), or unskilled
(44.5 expected) workers in England and Wales are used to derive expected lung
cancer deaths.  The author notes that most workers in the study cohort would be
considered partly-skilled, and the lung cancer mortality in  that group is almost
identical  to that of the partly-skilled in England and Wales.  However, overall
mortality is 17% lower (254 observed vs. 306.4 expected) compared to
partly-skilled workers.  This observation led the author to  comment that the
high proportion of lung cancers in the study cohort  may be related to
occupational  factors.
     A smoking history questionnaire was submitted to a limited subgroup of
study members who were still  employed at the plants  during data collections
                                      105

-------
(1973 to 1975), and who attended the works medical  center at  that  time.   This
represented only 41% of the workforce.   Based on this  data, the  author concluded
that the 26.7% excess lung cancer mortality could not  be due  to   excessive
tobacco consumption on the part of members of the study population.   Cigarette
consumption in 4 of 14 coke plants was, according to the author, "above
average," while in the remainder it was below average.
    With respect to the three occupational groups, i.e., ovens,  part-ovens,  and
non-ovens, the comparative mortality figure computed for each occupational  group
for certain selected causes, including lung cancer, revealed no statistically
significant excesses.  The author reports that when SMRs were calculated for the
same occupational  groups  (utilizing the male population of Great Britain),  lung
cancer was excessive in all three groups, but he does not state whether these
excesses were  statistically significant nor does he provide any data to support
this statement.
    The  risk of  lung cancer as  well as the risk  of all malignant neoplasms
increased with increasing lengths of employment  for the  1471 coke oven workers.
The  population of  employees who had worked on the ovens  for more than 10 years
had  an  SMR of  127  and  a CMF of  1.24 based on  17  observed  lung cancers.
Additionally,  the  SMR  and CMF  for the  cause  "all malignant neoplasms" was 126
and  1.30,  respectively, based  upon  30  observed  cancer deaths in the  same
workers.  None of  these SMRs  or CMFs were statistically  significant, however.
The  findings  above,  the author  concludes, tend  to  support American studies that
 show an excessive  risk of lung  cancer  in  coke workers despite the fact  that none
of the findings  were statistically  significant.
     Several  comments should be made regarding the  lack of statistical
 significance.   First,  Col lings compared the  observed  lung cancer  mortality  in
 coke plant workers, which included not only  coke oven workers  (Number = 1615)
                                       106

-------
but "part-oven" workers (Number = 742) and "non-oven" workers (Number = 392)  a's
well, with the mortality expected based on national  lung cancer mortality rates.
In other words, persons not even exposed to coke oven emissions, the non-oven
workers, and persons only occasionally exposed, the  "part-oven" workers were
included in the study cohort.  The diluting effect caused by the inclusion of
persons not even exposed or seldom exposed may well  be responsible for the lack
of a statistically significant excess of lung cancer mortality.
    Second, it should be noted that the sample size  of the "nonoven" (Number  =
392) and "part-oven" groups (Number = 742) are too small for a meaningful lung
cancer mortality comparison by CMF across occupational subgroups (coke oven,
"part-oven" and "non-oven").
    Third, the analysis by length of employment does not consider a
"survivorship effect" in regard to the analysis by SMR and is misleading in
regard to the analysis by CMF.  A "survivorship effect" (Fox and Collier 1976)
means that workers who survive, for example, 10 years of employment, will have a
lower lung cancer SMR and a lower SMR for other diseases than workers who began
employment at the same time but had left the working population.  Thus the
relative risk for coke oven workers who had worked ten or more years (SMR = 127)
would likely have been even higher had a similar working population, such as
other steelworkers as was used in the Redmond et al. studies, been used to
calculate the relative risk.  The CMFs as presented  are misleading because while
they may not indicate that coke oven workers employed 10 years or more have a
significant excess of lung cancer mortality in regard to the entire group of
coke plant workers, one should understand that the difference between the CMF
for workers employed less than 5 years and the CMF for workers employed ten or
more years is significantly (P < 0.05) different.
    Fourth, the study by Collings followed workers for only 9 years.  Such a
                                      107

-------
follow-up period is usually not long enough to detect a  significant  excess  of
cancer mortality if the mortality is associated with the suspected etiologic
agent being studied.
Sakabe et al. 1975 —
    Sakabe et al. (1975) studied lung cancer mortality and cancer mortality  (all
sites) for the years 1949-1973* among retired coke oven workers  from 11
companies in Japan.  At the time of the study there were 36 companies producing
coke in Japan.  No explanation was provided as to why the other  companies  did
not participate in the study.  Mortality was ascertained by a questionnaire  to
the industrial physician or chief health inspector of each industry.  The
expected mortality was calculated from the vital  statistics for  the general
Japanese male population for the corresponding peripd of time.  Coke ovens in
Japan are categorized as those for blast furnace coke, those for casting coke,
and those for general coke, depending on the purpose for which the coke  is
manufactured.  The furnace temperature of coke ovens is about 1300°C for blast
furnace coke, 1000°C for casting coke, and 1200°C for general coke.
    The 11 companies surveyed included four iron and steel plants, four  city gas
companies, and three "coke manufacturing chemical companies and coke
manufacturing companies."  Coke ovens in the iron and steel plants in Japan  are
used solely for manufacturing blast furnace coke, and coke ovens of city gas
plants are used for manufacturing coke for blast furnaces, casting furnaces, and
general use.  No description of the purpose of the coke production was given for
the three "coke manufacturing chemical companies and coke manufacturing
    *Although 2,201 workers who retired between 1947 and 1973 were traced, only
the cancer deaths from 1949 to 1973 were included in the study.  The authors
provided no explanation for the period 1947 to 1949.
                                      108

-------
companies."  There was no statistical difference between the observed and
expected cancer (all sites) mortality or lung cancer mortality when the  study
population consisted of retired workers from all 11 companies combined.
    Sakabe et al. then compared the observed cancer mortality for the 674
retired workers from the four iron and steel plants and the 1,261 retired
workers at the four city gas companies with the expected cancer mortality for
those companies.  The number of retired workers among the "coke manufacturing
chemical companies and coke manufacturing specialized companies" was too small
for statistical comparison of observed and expected cancer mortality.  Cancer
mortality (all sites) was not significantly (P < 0.05) different from that
expected for either the iron and steel plants (36 observed, 31.87 expected)  or
the city gas companies (51 observed, 69.77 expected).  Lung cancer mortality
among the iron and steel plants coking companies was significantly greater than
expected however (8 observed, 3.38 expected, P 5 0.022).  No statistical
difference between the observed and expected lung cancer mortality was found for
the city gas companies.
     When Sakabe et al. looked at proportionate cancer mortality, the proportion
of lung cancer cases to all cancers was significantly greater (P ^ 0.05)
than expected for the iron and steel plants but not for the city gas companies.
    Sakabe et al. also studied the age of lung cancer onset and working  period
at the coke ovens.  For all coke oven (including both city gas and iron  and
steel plants) workers, lung cancer was found to occur after 5 years of working
and at the age of 50 or over (except for one individual  whose age was reported
as 44).  Among coke oven workers of the iron and steel plants, lung cancer
occurred after 10 years of working and at the age of over 50 years.
Smoking data for the lung cancer cases was incomplete.  Of the 18 coke oven
workers who died from lung cancer, 10 were smokers, 2 were nonsmokers, and no
                                      109

-------
information was available for 6.  Reliable information concerning the amount of
smoking for each smoker could not be obtained.
    In conclusion, Sakabe et al. found a statistically significant
(P ^ 0.022) excess of lung cancer mortality among workers retired from
plants that produce coke for blast furnaces, but not among retired coke oven
workers from city gas companies.  The difference of observed to expected,
however, is of a magnitude that could be explained by differences in smoking
habits, and thus, conclusions from the study, because of the inadequate smoking
data, are somewhat limited.  An excess of lung cancer mortality among the  coke
oven workers at the city gas companies was not found perhaps because the coke
ovens at the gas companies may be operated from 100°C to 300°C lower than  the
coke ovens at the iron and steel plants. '
    The authors also found that the proportion of lung cancer mortality to all
cancer mortality was significantly (P < 0.05) in excess among retired coke oven
workers of iron and steel plants, but not among retired coke oven workers  of
city gas industries.
                                      110

-------
Summary
   Lloyd (1971) and Redmond et al. (1972, 1976, 1979) found an excess of total
cancer mortality and respiratory organ cancer mortality among workers employed
at coke works.  Both were found to be dose-related.  Workers exposed to low
(side oven), medium (side oven and topside), and high (topside only) exposure
had increasingly greater excesses of total cancer mortality rates and lung
cancer mortality rates.  Not only was there an increase in excess lung cancer
mortality by occupation site, but there was an increase by length of exposure
as well.  Workers exposed 5 years or more had a greater excess of lung cancer
than workers exposed less than 5 years.
    As indicated earlier, one criticism of the Lloyd (1971) and Redmond et al.
(1972, 1976, 1979) studies is that smoking data were not taken.  An analysis
of lung cancer mortality is generally not considered adequate without smoking
data.  However, the dose-responses seen in the Lloyd and Redmond et al.
studies is so striking that it is improbable that the excess in lung cancer
mortality could be explained by differences in smoking habits.
    An apparent discrepancy in the Lloyd and Redmond et al. studies is that
the excess lung cancer mortality among nonwhite workers in Allegheny County
was significant, while that among white workers was not significant.  Several
explanations have been offered for this phenomenon.  Perhaps the most obvious
explanation is that more nonwhites than whites were employed at the coke ovens
in Allegheny County, particularly as full-time topside workers, and the excess
among nonwhites may have been significant because of their larger sample size.
Redmond et al. (1979)  suggested that the difference may be because, within the
respective subsites at the coke ovens, whites and nonwhites may have had
different jobs and consequently different exposure to volatile hydrocarbon
effluents.  Redmond et al. (1979) also suggested that the difference may be
                                     111

-------
because mortality rates for lung cancer have been shown in other studies to be
inversely correlated with educational qualifications and occupational
category.  As Redmond et al. (1979) noted, however, the expanded studies of
non-Allegheny County coke oven workers found that the lung cancer risk was
significant for both whites and nonwhites.  In addition, Mancuso (1977) has
suggested that the difference in cancer mortality risk between the nonwhites
and whites in the Allegheny County plants may have resulted because a  majority
of the nonwhites were migrants from the South.*  Since Mancuso did not report
how many of the total steelworker population (from which the expected
mortality data were derived) were also migrants from the South, it would be
premature to conclude that being a nonwhite worker from the South predisposes
one to cancer.  Also, it would be difficult to separate the effects of place
of origin and race from the effects of industrial exposure.  Impoverished
migrant workers may well take any job that is offered including those  jobs
that place persons at an excess risk of cancer.  Finally, as noted above, the
Redmond et al. studies of non-Allegheny County coke oven workers found that
excess risk of lung cancer was significant for both whites and nonwhites.
    Prostate cancer mortality was significantly (P < 0.05) increased among all
Allegheny County coke oven workers ever employed in 1953.  For workers having
worked 5 or more years, however, the excess was not significant. Similar to
the apparent discrepancy between whites and nonwhites above, an excess of
prostate cancer mortality did exist among workers employed 5 years or  more,
but possibly because of small sample size, the excess was not significant.
Among the non-Allegheny County workers, the excess in prostate cancer
    *In 1974, Mancuso and Sterling reported that migrants in Ohio,
particularly migrants from the South, had higher death rates for various
cancer sites than did persons born in Ohio.
                                     112

-------
 prostate cancer was  significant for nonwhite workers  only.
     Kidney  cancer  mortality was significantly increased  (P  <  0.01)  among white
 workers  in  the Allegheny  County study.   Small  excesses  in kidney  cancer were
 seen  for both  whites  and  nonwhites  in the  non-Allegheny  County  plants, but
 these excesses were  not significant (P  < 0.05).
       Sakabe et al.  (1975)  found that there  was  a  significant excess of lung
 cancer mortality among  retired  Japanese coke oven  workers at  iron and steel
 coking plants  when compared to  cancer mortality  among the general population.
 The excess  found,  however,  is of a  magnitude that  could  be  explained by
 differences in  smoking habits.   Thus, because  smoking data  for the  study was
 inadequate, conclusions from the study  are somewhat limited.
      Reid and  Buck (1956) found  a significant  (P < 0.05, one-tailed test)
 difference  between the observed  and expected mortality for  cancer,  other than
 respiratory cancer, for coke plant  workers whose last job was listed as "coke
 ovens."  No excess was found for respiratory cancer deaths.  When mortality
 was analyzed by whether the workers had  ever worked at the  coke ovens, no
 significant excess in respiratory or other cancer mortality was found.  The
 Reid  and Buck study had a number of deficiencies, however.  The study
 population was  poorly defined and the "observed deaths" may not have included
 deaths of workers not "on the books."  Furthermore, the study did not
 sufficiently address the issue of a cancer latency period since little or no
 follow-up of vital  status occurred.  Analysis of mortality among retired
workers was not adequate since there was no  adjustment for age.
    navies (1977, 1978) did not find any significant difference  between  the
observed and expected cancer mortality of the coke oven workers  at two coke
works in South  Wales.  Davies's study cohort  was relatively  small  (Number =
610), however,  and  he followed his cohort for only 11  years.
                                     113

-------
    Ceilings (1978) found an excess of lung cancer among the coke plant
workers when compared to rates for Great Britain (45.0 observed vs. 35.7
expected); this excess was not significant (P > 0.05), however.  Like the
Davies study, Col lings followed his cohort for a relatively short period of
time (9 years).  Also, his coke plant study cohort included nonoven workers as,
well as coke oven workers which would have caused a diluting effect, possibly
resulting in the failure to detect a significant increase.  An internal
comparison done by Collings did not find a significant difference between coke
oven workers and the entire cohort of coke plant workers, but the sizes of the
non-oven and part-oven subgroups were not large enough to have permitted a
meaningful analysis.  It should also be noted that coke ovens are operated at
lower temperatures in Great Britain than they are in the United States
(Doherty and DeCarlo 1967), which may also contribute to the lack of positive
findings in the three British studies on coke workers (Reid and Buck 1956,
navies 1977, and Collings 1978).
    The update by Redmond et al. of the study begun by Lloyd consistently
showed a significant excess of lung, trachea, and bronchus cancer mortality.
In the Redmond et al. studies, there was a dose-response both by working area
(topside, side oven and part-time topside, and side oven) and by length of
exposure.  Prostate cancer and kidney cancer also appeared to be in excess in
the Redmond et al. (1979) update.  Because the British studies had some design
deficiencies and did not follow the workers as long as the American studies
did, and because neither the British studies nor the Sakabe et al. study
considered occupational categories within the coke works, the positive results
of the American studies are considered a better evaluation of the cancer risk
to coke oven workers.  Therefore, based on results of the American studies, it
is concluded that exposure to coke oven emissions increases the risk of cancer
                                     114

-------
of the lung, trachea, and bronchus; kidney; and prostate,  as well  as cancer at
all  sites combined.
                                     115

-------
ANIMAL STUDIES
Topside Coke Oven and Coke Oven Main
    Topside coke oven sample extract has been tested as a tumor initiator in
initiation-promotion skin treatment studies in two strains of mice (Nesnow et
al. 1981, Nesnow 1980).  An extract of material from a coke oven collecting
main has also been evaluated for activities as a whole carcinogen, an
initiator, and a promoter [in skin treatment studies with one strain of mouse
(Nesnow et al. 1981)].    ;

Initiation-Promotion Studies—
     Nesnow et al. (1981) have evaluated the effects of extracts of topside
coke oven emission and coke oven main samples in initiation-promotion  and
complete carcinogenicity studies in mice.  The methods for obtaining the test
samples from coke ovens has been described by Huisingh (1981) and Huisingh et
al. (1979).  Topside coke oven samples were collected as particulate matter
with a Massive Air Volume Sampler, and coke oven main samples were obtained
from a separator located between the gas collector and the primary coolers
within the coke oven battery.  The samples were soxhlet-extracted with
dichloromethane, which was subsequently removed by evaporation under dry
nitrogen gas.  All test materials used in this study were prepared under
yellow light immediately before application, in 0.2 ml spectral grade acetone,
onto test sites.
    Mice of the SENCAR strain, derived from mating female Charles River CD-I
mice with male skin tumor sensitive (STS) mice, were selected as the test
animals in this study.
    Each control and treatment group consisted of 40 male and 40 female mice
which were 7 to 9 weeks old at the start of the study.  Animals were caged in
                                     116

-------
 groups of 10 under yellow light.   Test  sites  on the  skin were shaved 2 days
 before initial  treatment,  and  only mice  in the resting phase of the hair cycle
 were used.   In  the initiation-promotion  experiments, initiating agents were
 applied as  single  doses except  for the  10 mg  (highest) dose which was given as
 five daily  doses of 2 mg  each.  At 1 week following application of initiator,
 2  ug of the promoter 7,12-dimethylbenz[a]anthracene-12-0-tetradecanoylphorbol-
 13-acetate  (TPA) was topically  applied twice  per week.  In tests for complete
 carcinogenicity, test material  was applied once weekly, or twice weekly at the
 highest dose, for  50 to 52 weeks.  Test  substances evaluated as promoting
 agents  were applied  to the skin once each week, or twice each week at the
 highest dose, for  34 weeks following skin treatment with a 50.5 ug dose of the
 initiator benzo[a]pyrene  (B[a]P).
     Animals were observed weekly for tumor formation, and papillomas over 2 mm
 in diameter and carcinomas were included in cumulative totals if they
 persisted for at least 1 week.  Papillomas were scored at 6 months or,  in the
 test  for promoting activity, 34 weeks, and carcinomas were totaled after  1
 year.   The  authors indicate that examination of animals by necropsy and
 tissues and tumors by histopathology was being done and that pathologic data
 would be presented in a separate forthcoming report.
     Results of initiation-promotion experiments on  topside  coke  oven  sample
 extract, coke oven main sample extract,  and B[a]P  as  initiating  agents  are
 compared in Table VI-18.   The stronger effect  of the  coke  oven main  sample
 compared to the topside coke oven  sample reflects  the greater concentration of
 ingredients contributing  to the initiating  activity of  the  former  sample.
 Responses to the coke oven main sample and  B[a]P in the study for  complete
carcinogenesis are  shown  in Table  VI-19.   Promoting activity was found  with
the coke oven main  sample and TPA  following  initiation  with  B[a]P
                                     117

-------
           TABLE VI-18.
SENCAR MOUSE SKIN TUMORIGENESIS
(Nesnow et al.  1981)
Dose
(ug/mouse)

0
0
2.52
2.52
12.6
12.6
50.5
50.5
101
101

100
100
500
500
1000
1000
2000
2000
10,000
10,000

100
100
500
500
1000
1000
2000
2000
10,000
10,000
Mice with
No. Mice ! Papillomas*
Surviving (%)

(M)
(F)
(M)
(F)
(M)
(F)
M
(F)
(M)
(F)

(M)
(F)
(M)
(F)
(M)
(F)
(M)
(F)
(M)
(F)

(M)
(F)
(M)
(F)
(M)
(F)
(M)
(F)
(M)
(F)

37
39
40
39
40
37
39
40
38
38

40
40
40
40
37
39
39
38
39
40

38
39
39
39
39
39
40
40
38
37
BENZO[A]PYRENE -
8
5
45
31
73
57
100
75
95
97
TOPSIDE COKE OVEN
13
10
73
70
95
72
95
90
100
100
COKE OVEN MAIN -
50
31
90
82
87
90
78
100
100
100
Mice with
Papillomas Carcinomast
per mouse* (%)
TUMOR INITIATION
0.08
0.05
0.50
0.44
1.8
1.1
5.8
2.8
10.2
7.9
- TUMOR INITIATION
0.13
0.20
1.6
1.8
2.6
2.0
4.0
3.5
7.1
7.7
TUMOR INITIATION
0.63
0.38
3.7
2.2
3.3
3.1
3.1
5.3
8.9
8.1

5
0
5
5
20
23
25
20
30
25

0
8
5
15
15
3
13
10
13
20

10
25
54
54
53
48
48
45
55
65
Carcinomas
per Mouset

0.05
0
0.07
0.05
0.20
0.23
0.25
0.20
0.33
0.25

0
0.08
0.05
0.15
0.15
0.03
0.13
0.10
0.15
0.23

0.10
0.25
0.59
0.54
0.53
0.48
0.48
0.45
0.55
0.65
*Scored at 6 months.
tCumulative score after one year.
                                 118

-------
            TABLE VI-19.  SENCAR MOUSE SKIN TUMORIGENESIS
                  (adapted from Nesnow et al. 1981)
Dose Mice with Carcinomas*
(ug/mouse/week) (%)

12.6
12.6
25.2
25.2
50.5
50.5
101
101
202
202
0
0

100
100
500
500
1000
1000
2000
2000
4000
4000

(M)
(F)
(M)
(F)
(M)
(F)
(M)
(F)
(M)
.(F)
(M)
(F)

(M)
(F)
(M)
(F)
(M)
(F)
(M)
(F)
(M)
(F)
BENZO[A]PYRENE - COMPLETE CARCINOGENESIS
10
8
63
43
93
98
80
90
80
93
0
0
COKE OVEN MAIN - COMPLETE CARCINOGENESIS
5
5
36
30
48
60
82
78
98
75
Carcinomas
per Mouse*

0.10
0.08
0.63
0.43
0.93
0.98
0.83
0.98
0.80
0.98
0
0

0.05
0.05
0.36
0.30
0.55
0.60
1.00
0.78
0.98
0.85
Cumulative score after one year.
                                 119

-------
(Table VI-20).  Spontaneous tumor formation in the control  groups  was  not
evident in the studies for complete carcinogenesis and  promoting activity
(Tables VI-19 and VI-20) and was below 10% incidence for papillomas  and  was  5%
(males) or 0% (females) incidence for carcinomas  in the experiment for
initiating activity (Table VI-18).
    Results of the study by Nesnow et al. (1981)  show that  coke oven main
sample extract contained ingredients capable of producing skin tumors  in
SENCAR mice either as an initiator, a promoter, or a complete carcinogen.
Topside coke oven sample extract was also active as an  initiating  agent;
however, according to Nesnow et al. (1981), an unknown  portion of  the  topside
sample was contaminated with particulate matter from ambient air due to  the
location of the Massive Air Volume Sampler and local wind conditions (this
issue is further discussed on pages 39, 41, and 44 of the mutagenicity section
herein).  Hence, the extent to which the topside coke oven  sample  extract  used
in the initiation-promotion experiment is representative of topside  coke oven
sample per se appears uncertain.
    Data in Tables VI-18 and VI-19 show that the tumorigenic responses to the
coke oven sample extracts and B[a]P tended to be constant at all doses above
the lowest dose in the  dose ranges used.  The nature of these dose-responses
indicates that the doses used were in the range capable of producing maximal
effects in relation to  the sensitivity of the SENCAR strain to the initiating
and complete  carcinogenic properties of these test materials.  The authors
proposed that a lack  of a monotonic dose-response across a dose range  may be
due to a toxic effect of the test material being tested which damages  the
epidermis to  yield a  reduced tumorigenic  response.   Forthcoming results of the
histopathologic examination of  skin test  sites may  provide evidence in favor
of this possibility.   However,  although not identified as an experimental
                                      120

-------
                 TABLE VI-20.   SENCAR MOUSE SKIN TUMORIGENESIS
                        COKE OVEN MAIN -  TUMOR PROMOTION
                       (adapted from Nesnow et al.  1981)













TPA
TPA
Dose
(ug/mouse)
0 (M)t
0 (F)
100 (M)§
100 (F)
500 (M)
500 (F)
1000 (M)
1000 (F)
2000 (M)
2000 (F)
4000 (M)^l
4000 (F)
, 4 ug (M)#
, 4 ug (F)
Mice with Papillomas*
0
0
3
10
26
38
53
68
84
85
100
100
86
97
Papillomas per mouse*
0 -
0
0.02
0.10
0.44
0.83
1.2
1.2
2.5
3.1 .'
8.2
8.8
3.1
5.9
    tMice initiated with 50.5 ug benzo[a]pyrene (B[a]P) and subsequently
treated weekly with acetone.


    §Mice initiated with 50.5 ug (B[a]P) and subsequently treated weekly with
coke oven main.


    UMice initiated with 50.5 ug (B[a]P) and subsequently treated twice weekly
with 2 mg coke oven main.


    #Mice initiated with 50.5 ug (B[a]P) and subsequently treated twice weekly
with 2 ug TPA.
                                     121

-------
problem in the study report, it is possible that the rather constant  response
over the dose ranges used may be due to incomplete solubility  of  the  test
samples in acetone, which in turn actually might have resulted in the
application of rather similar doses throughout the dose ranges.   Nonetheless,
although the responses in Tables VI-18 and VI-19 were generally not monotonic
throughout the entire dose ranges used, the data clearly show  positive
activity for the indicated test materials.  As shown in Table  VI-20,  a  clearer
indication of a dose-related effect was obtained in the evaluation of coke
oven main sample extract as a promoter.
    In summary, coke oven main sample extract was positive as  a complete
carcinogen, an initiator, and a promoter on the skin of SENCAR mice,  and
topside coke oven sample extract was positive as an initiator  on  the  skin of
SENCAR mice.

    Nesnow (1980) reported results of an additional initiation-promotion
experiment with the topside coke oven extract on C57BL/6 mice  done for
comparison with the experiment on SENCAR mice.  Similar protocols were  used
for the two studies except that the C57BL/6 mice were on study for 52 weeks.
Tumor-initiating activity at the application site was not observed with coke
oven emission sample extract at doses as high as 10 mg/mouse;  however,
tumor-initiating activity was also not demonstrated with the positive control
chemical, benzo[a]pyrene, at doses as high as 403.68 ug/mouse.  Thus, results
obtained in the experiment on C57BL/6 mice are considered inconclusive  as
indicated by the resistance of this mouse strain to tumor-initiating  activity
by the positive control agent.
                                     122

-------
 Coal Tar
     Cardnogenicity studies on aerosols of coal  tar and coal  tar fractions  in
 laboratory animals were reported by Morton (1961),  Morton et  al. (1963),  Tye
 and Stemmer (1967), MacEwen and Vernot (1972-1976), Kinkead  (1973),  McConnell
 and Specht (1973), and MacEwen et al.  (1976).   These studies  provide evidence
 for a carcinogenic effect  of coal  tar  aerosol  test  samples as  discussed
 herein.
     Numerous  carcinogenicity studies on  coal tar  samples  applied topically to
 the skin of laboratory animals have been  reported.   Studies discussed herein,
 which show an  ability  of coal  tar  samples  to produce  local tumors following
 skin treatment, include those  reported by  Bonser  and  Manch (1932), Hueper and
 Payne (1960),  Morton  (1961), and Wai leave  et al.  (1971).  Morton (1961) and
 Wallcave et al. (1971)  tested  coal  tar samples from coking operations.

 Inhalation  Exposure Studies--
     Morton  et al.  (1963) examined  C3H  mice  (a strain that was reported to have
 a low historical incidence  of  spontaneous pulmonary adenomas) for lung tumors
 following  inhalation exposure to coal  tar aerosol, gaseous formaldehyde,  or
 gaseous  formaldehyde followed  by coal  tar aerosol.  In the first part of  the
 experiment, groups of 60, 60, and 42 mice were exposed to concentrations  of
 0.5,  0.10,  or 0.20 mg/liter, respectively, of gaseous formaldehyde for three
 1-hour periods  per week.  The control  group consisted of 59 untreated mice.
After 35 weeks, none of the animals that were sectioned of those that died
 (118 of 221) during the period had developed lung  tumors.   The surviving
animals were used to conduct further experiments  with coal  tar and
formaldehyde.   The surviving 33 mice from the control  group  in  the first  part
of the experiment and the  surviving 26  mice from the group that had been
                                     123

-------
 exposed to 0.10 mg/liter of gaseous formaldehyde in  the  first  part  of the
 experiment were exposed to 0.30 mg/liter of coal  tar aerosol for three  2-hour
 periods per week for up to 36 weeks.   The surviving  36 mice  from the group
 that had been exposed to 0.05 mg/liter of formaldehyde in the  first part of
 the experiment were exposed to 0.15 mg/liter of  formaldehyde for three  1-hour
 periods each week for up to 35 weeks.   Also,  the  untreated control group* was
 observed for 82 weeks.
     The test animals were exposed  to the test substances until death.   The
 first death occurred 1 to 11 weeks after exposure and the longest time  until
 death was 36 weeks.   Serial  sections of the  trachea, large bronchi, and lung
 Of the exposed animals and sections of the  lung of 30 unexposed mice were
 examined (Table VI-21).
     Five mice inhaling coal  tar aerosol  and  one mouse inhaling formaldehyde
 followed by coal  tar developed  squamous  cell tumors  in the periphery of the
 lung,  involving one-third to one-half  of the lobe.   In two mice from the
 former group,  several  lobes  were involved.  A sixth mouse in the former group
 that  died  after 20 weeks  of  exposure had  an invasive squamous cell  carcinoma,
 which  was  described  as  "unquestionably a  squamous cell carcinoma,  whereas,
 those  occurring in the other  five  animals probably represented an  earlier
 stage  of development  at the time of death."  One mouse in each  group had
 adenoma  of the  lung.   Tumors  of the lung were not observed in mice  breathing
 formaldehyde only or  in untreated controls.
    There  were  other  changes  produced in the tracheobronchial  epithelium as
the result of the inhalation of coal tar.  The most  striking  was a  necrotizing
tracheobronchitis in the majority of mice; the incidence  was  not reported.   In
    *Tne initial size of the untreated group was not reported.   At  the
termination of the experiment at 82 weeks,  the group consisted  of 30 mice.
                                     124

-------
TABLE VI-21.  TUMORS OF THE LUNG IN MICE INHALING FORMALDEHYDE
                AND/OR AEROSOL OF COAL TAR
            (adapted from Horton et al. 1963)
Treatment
Untreated
Controls
Coal Tar
Formaldehyde
and Coal Tar
Formaldehyde
Squamous Cell
Tumors
0/30
6/33*
1/26
0/36
(0%)
(18%)
(4%)
(0%)
Adenomas
0/30
1/33
1/26
0/36
(0%)
(3%)
(4%)
(0%)
Total
0/30
7/33
2/26
0/36
(0%)
(21%)
(8*)
(0%)
                          125

-------
addition, squamous cell  metaplasia extended into  the  smaller  bronchi.
Hyperplasia of the bronchial  epithelium occurred  frequently,  sometimes with
papillary infolding.  The epithelium of untreated mice was  normal,  showing
neither metaplasia nor hyperplasia.
    Epithelial changes in mice inhaling formaldehyde involved mostly the
trachea; extension into the major bronchi was infrequent and did not occur  at
all in the  smaller bronchi.   In general, the inhalation of formaldehyde
resulted in an acute tracheobronchitis  ranging from slightly to severely
necrotizing, or  developing into a chronic  type with proliferation of fibrous
tissue.  This was sometimes  complicated by bronchopneumonia.  In summary, mice
 inhaling coal tar aerosol  developed squamous  cell  carcinomas  of the lung, as
 well  as  hyperplastic and metaplastic epithelial  changes.

     Tye and Stemmer (1967) separated two different coal  tars into  phenolic
 (P-tar) and nonphenolic (N-tar) fractions and exposed mice by inhalation to
 various blends of the coal tar fractions and to one of the original  tars.   The
 same coal  tar (T-l) (specific gravity  1.17; 4.5% tar acid, 0.7%
 benzo[a]pyrene,  and 67% Diels-Adler compounds*) that was used in the
 experiments by  Horton, Tye,  and  Stemmer  (1963)  and a second, somewhat
 different  tar (T-2)  (specific  gravity  1.24;  1.4%  tar acid, 1.1%
 benzoCa]pyrene, and 2% Diels-Adler compounds*), were the two tars from which
 the  phenolic (P-tar)  and nonphenolic (N-tar)  fractions were  separated.
      Fifty male C3H/HeJ mice, 3 to 5 months old, were in each test  group.   The
  tests groups consisted of untreated, Tar-1,  N-Tar-1, N-Tar-1 plus  P-Tar-1,
  N-Tar-1 plus P-Tar-2, and N-Tar-2 plus P-Tar-1.  Mice were exposed for 2  hours
  	*'As indicative oflmthracene and  polycyclic aromatic hydrocarbons  with
  three  linear aromatic rings with a free meso position.
                                       126

-------
 every  3 weeks.   During the  first  8 weeks, the exposure was at a concentration
 of  0.20 mg/liter,  but this  was  reduced to 0.12 mg/liter because so many mice
 died.
     Three mice  from each group  were killed after 4 weeks, and five mice were
 killed after  31  weeks.  Surviving mice were killed at the end of 55 weeks.
 Mortality from  exposure was high  in all groups of treated mice.  At the end of
 the experiment,  there were  31/50  (62%), 11/50 (22%), 11/50 (22%), 10/50 (20%),
 21/50  (42%),  and 21/50 (42%) mice alive in the control, Tar-1, N-Tar-1,
 N-Tar-1 plus  P-Tar-1, N-Tar-1 plus P-Tar-2, and N-Tar-2 plus P-Tar-1 groups,
 respectively.  Tumor response is  recorded in Table VI-22.
     The most  prominent lesions  were intrabronchial adenomas and
 adenocarcinomas, occurring  anywhere in the bronchial  tree.  Multiple tumors
 were frequently  seen.  The  intrabronchial adenomas were papillary.  There also
 were alveolar adenomas which were peripheral.  Tumors of the lung were
 diagnosed as  adenocarcinomas only if there was invasion or if metastases were
 observed.
    Adenomas  and adenocarcinomas of the lung were observed in 60% to 100% of
 the mice inhaling aerosols of coal tars, whereas tumors were not seen in any
 of  the control mice.  Incidences of squamous metaplasia varied from 10% to 38%
 in  treated mice and were absent in control  mice.  "Alveolar epithelization"
 was also observed, but less often than squamous  metaplasia.   Areas of squamous
 and alveolar metaplasia were not considered as tumors,  even  when they occupied
 relatively large spaces.

    MacEwen and Vernot (1972-1974),  Kinkead (1973),  and McConnell  and Specht
 (1973)  reported on a study in  which  mice,  rats,  hamsters,  and rabbits were
exposed to a coal tar aerosol  from which the light  oil  and solid fraction  was
                                     127

-------
TABLE VI-22.  INCIDENCE OF LUNG TUMORS IN MICE INHALING  AEROSOLS  OF  COAL TARS*
                     (adapted from Tye and Stemmer  1967)
Treatment
Untreated
Controls
Tar-1
N-Tar-1
N-Tar-lt
P-Tar-1
N-Tar-lt
P-Tar-2
N-Tar-2t
P-Tar-1
Metaplasia
0/32
5/13
2/20
5/19
7/25
4/23
(0%)
(38%)
(10%)
(26%)
(28%)
(17%)
Adenomast
0/32
12/13
16/20
14/19
14/25
14/23
(0%)
(92%)
(80%)
(74%)
(56%)
(61%)
Adenocarcinomas
0/32
3/13
0/20
1/19
1/25
0/23
(0%)
(23%)
(0%)
(5%)
(4%)
(0%)
Adenomas and
Carcinomas
0/32
13/13
16/20
15/19
15/25
14/23
(0%)
(100%)
(80%)
(79%)
(60%)
(61%)
    *Mice surviving for 46 weeks or longer.

    tlncludes intrabronchial  and alveolar adenomas.
                                     128

-------
 removed.   Gross  skin  pathology  for the mice was reported; any other tumor
 response  in the  mice  and  in the other animals was not reported.*
     Groups of  64 female yearling and 64 weanling  (32 of each sex)
 Sprague-Dawley rats,  50 male JAX-CAF1 mice, and 50 male ICR-CF1 mice were
 exposed continuously  for  90 days (except for 15 minutes a day to allow for
 animal maintenance) to concentrations of 0.2, 2.0, and 10.0 mg/m3 of coal
 tar  aerosol.   Ninety-two  female yearling Sprague-Dawley rats, 82 weanling
 Sprague-Dawley rats (73 female and 9 male), 75 male JAX-CAF1 mice, 75 male
 ICR-CF1 mice,  100 male golden Syrian hamsters, and 24 New Zealand white
 rabbits were exposed  continuously, as above, for the same 90-day period to a
 concentration  of 20 mg/m3.  The control animals consisted of 41 female and
 41 male Sprague-Dawley weanling rats, 82 female Sprague-Dawley yearling rats,
 75 male JAX-CAF1 mice, 75 male ICR-CF1 mice, 24 female New Zealand white
 rabbits, and 100 male golden Syrian hamsters (MacEwen and Vernot 1972).  Many
 of the mice contracted a streptococcus infection and died before 93 days
 postexposure.  Skin tumor response for the mice is found in  Table VI-23.
     Tumor responses of 28% (10 of 36), 38% (3 of 8), and 8%  (2 of 25)  were
 seen in the three highest dose groups of the ICR-CF1 mice; no tumors (0 of 62)
 were found in  the controls.  A tumor response of 37% (10 of  27)  was found in
 the  highest dose group JAX-CAF1 mice; no tumors (0 of 74)  were found in the
 JAX-CAF1 controls.  McConnell  and Specht (1973) examined some of the skin
 tumors histologically and concluded that a whole spectrum  of epithelial
 tumors, from squamous cell papilloma  to keratoacanthoma  to "frankly
 aggressive" appearing squamous cell  carcinoma  are  stimulated by  the coal  tar
    *Per contractual  agreement, Sasmore performed internal  and  skin
histopathology for the study and reported his results  (Sasmore  1976),  but
because information in the Sasmore report is  incomplete,  no conclusions  can be
made about the report.
                                     129

-------
         TABLE VI-23.  TUMOR RESPONSE IN MALE ICR-CF1 AND JAX-CAF1  MICE
                    FOLLOWING EXPOSURE TO COAL TAR AEROSOL
                   (adapted from McConnell and Specht 1973)
Dose (rng/rn^)
20.0
10.0
2.0
0.2
0.0
ICR-CF1*
10/36
3/8
2/25
0/2
0/62
(28%)t
(38%)§
(8%)§
(o«)§
(0%)t
JAX-CAF1*
10/27
0/12
0/47
0/47
0/74
(37%)t
(0%)§
(0%)§
(0%)§
(0%)t
    *The numerator is the number of animals with tumors  at  415  days
postexposure.  The denominator is the number of animals  that  were  alive  at  93
days postexposure.

    tThis dose group began with 75 animals.

    §This dose group began with 50 animals.
                                     130

-------
 aerosol,  although  the  majority  of  these  tumors  fall  in the  squamous  cell

 category.   McConnell and  Specht  also  found  a time-to-tumor  dose-response for

 the  coal  tar  aerosol.   This  dose-response is shown  in Table VI-24.   As  stated

 above, tumor  response  was not reported for  the  rats, hamsters, or rabbits.



     MacEwen and Vernot  (1975 and 1976) and  MacEwen  et al. (1976) reported on

 two  studies of the tumor  response  of  mice,  rats, rabbits, and monkeys

 following exposure to  coal tar  aerosol.  In the first study, 80 female

 Sprague-Dawley yearling rats, 80 Sprague-Dawley weanling rats (40 males and 40

 females), 75  JAX-CAF1  male mice, 75 ICR-CF1 male mice, and 100 male  Syrian

 golden hamsters were exposed continuously for 90 days (except for 15 minutes a

 day  to allow  for animal maintenance)  to  concentrations of 0.2, 2.0, and 10.0
     o
 mg/m-3 of  coal tar aerosol.  An equal  number of each species were used for

 controls.  The coal tar used to generate the aerosol in this study was:

     a composite mixture collected  from multiple coking ovens around
     the greater Pittsburgh area.   The coking ovens were of several
     different types and used different coal  sources for their starting
     materials.  The coke oven effluents were collected in air collec-
     tion devices using a chilled water spray to condense the higher
     boiling distillate fractions.  After settling and separation of the
     liquid phase, the  various coal  tar samples were blended together
     with a 20% by volume amount of the BTA  (benzene, toluene, xylene)
     fraction  of the coke oven distillate.

 An aerosol particle size determination in the exposure chambers  was performed,

 and  it was found that a minimum of 97% of all droplets were in a respirable

 range of 5 microns or less in diameter.  Only skin tumor response for the mice

was  reported  (Table VI-25).   Tumor response was not reported for the hamsters

or rats.

     In the second study, 75 female  and 100 male ICR-CF1  mice (described as

tumor susceptible), 50 female JAX-CAF1 mice  (described as a  tumor-resistant

hybrid strain), 40 male and 40 female CFN strain Sprague-Dawley  weanling rats,
                                     131

-------
     TABLE VI-24.   LATENT PERIOD  OF  FIRST  TUMOR  INDUCTION  IN CTV-I EXPOSED
                                 ICR-CF1 MICE
                          (McConnell  and Specht  1973)
        Dose (mg/m3)
Time of Tumor Appearance (Days)
            20

            10

             2
            < 93

             128

             142
    TABLE VI-25.   SKIN TUMOR RESPONSE IN ICR-CF1  AND  JAX-CAF1  MICE  FOLLOWING
                         EXPOSURE TO COAL TAR AEROSOL
                          (MacEwen and Vernot 1976)
Cumulative Number of Tumors*
Dose
(mg/m3)
10
2
0.2
Week of
Observationt
100
103
101
Exposed
44/75 (59%)
14/75 (19%)
1/75 (1%)
ICR-CF1
Control
3/75 (4%)
0/75 (0%)
0/75 (0%)
Exposed
18/75 (24%)
3/75 (4%)
1/75 (1%)
JAX-CAF1
Control
1/75 (1%)
0/75 (0%)
1/75 (1%)
    *The numerator is the number of animals with tumors;  the denominator  is
the number of animals exposed.

    tlncludes the 90-day exposure period.
                                         132

-------
 18  New Zealand albino rabbits, and 5 male and 9 female Macaca mulatta monkeys
 were exposed to 10 mg/m^ of coal tar aerosol for 6 hours each day, 5 days
 per week, for 18 months.  The coal tar used to generate the aerosols in this
 study was the same as that of the first study.  Aerosol particle size was
 determined monthly in the exposure chambers.  A minimum of 99% of the total
 droplets in both chambers were 5 microns or less in diameter and were thus
 within a respirable size range for rodents.
     Exposure to the coal tar at 10 mg/m^ significantly reduced the body
 weight of rabbits and rats compared with the controls, whereas monkeys showed
 no  significant change in body weight.  Sixteen of 18 rabbits and 6 control
 mice died during the test period.  These deaths were attributed to a chronic
 respiratory infection which caused debilitation and dehydration.  At the
 conclusion of the exposure period, the test monkeys and the surviving test
 rabbits along with the unexposed controls were delivered to the National
 Institute for Occupational  Safety and Health (NIOSH) Laboratories in
 Cincinnati, Ohio.   Since the number of surviving rabbits (2 of 18) was too few
 for statistical  comparison, and those animals were sacrificed (Gibb 1978a),  no
tumor response was found in the sacrificed rabbits (Gibb 1978b).  The monkeys
were kept for observation at the NIOSH Laboratories until  1979 when they were
moved to Gulf South Research in New Iberia,  Louisiana, where they are
currently being maintained.  One of the dosed monkeys died in 1981; results  of
the autopsy are not yet available (Gibb 1981).
    Alveolargenic  [sic] carcinomas were produced in 26 of 61 (43%) ICR-CF1
mice and in 27 of  50 (54%)  JAX-CAF1 mice.   The  number of tumors in the ICR-CF1
and the JAX-CAF1 control  mice were 3 of 68 (4%)  and 8 of 48 (17%),
respectively.   The exposed  and control  groups did  not differ in the incidence
of other types of  tumors  including squamous  cell  carcinomas, lymphosarcomas,
                                     133

-------
subcutaneous sarcomas, alveolargenic adenomas,  bronchiogenic  carcinomas,
reticulum cell  sarcomas, hemangiosarcomas,  and  hemopoietic tumors.
    Skin tumors were produced in 5 of 75 (7%) of the  ICR-CF1  mice and 2 of 50
(4%) of the JAX-CAF1 mice as compared to 3  of 75 (3%)  and 1 of 50 (2%) in the
ICR-CF1 and JAX-CAF1 controls, respectively.  The criterion for counting a
1'esion as a skin tumor was a growth greater than 1 mm  in diameter and in
height.  Each tumor was ultimately confirmed by histologic examination.
MacEwen et al.  compared the lack of skin tumor  response in the second study to
the tumor response of the 10 mg/m3 dose group of the  first study.   As stated
previously, the first study found a skin tumor  incidence of 14 of 75  (59%) in
the treated ICR-CF1 controls and 18 of 75 (24%) in the treated JAX-CAF1 mice
as opposed to only 3 of 75 (4%) in the ICR-CF1  controls and 1 of 75  (1.3%) in
the JAX-CAF1 controls, respectively.  A calculation of total  exposure time
(MacEwen et al. 1976) revealed that the same amount of coal tar aerosol
reached the skin of mice in the second study as in the first  study.   MacEwen
et al. suggested that the 18-month intermittent exposure of the animals in
their study allowed the animals enough time each day  to permit normal cleaning
of the fur.
    The incidence of coal tar tumorigenesis in  rats  is reported in  Table
VI-26.  The incidence of squamous cell carcinomas in  the lungs was  100%
(38/38) in exposed males and 82% (31/38) in exposed  females as opposed to 0%
(0/36) in male controls and 0% (0/37) in female controls.
    A dose-related tumor response was observed  for both the ICR-CF1  and the
JAX-CF1 mice.
                                     134

-------
                   TABLE VI-26.  COAL TAR TUMORIGENESIS  IN RATS
                              (MacEwen et al. 1976)
                                         Controls
                                     Males     Females
    Exposed
Males     Females
Number Examined Histologically*
Number of Rats with Tumors:
Squamous Cell Carcinoma, Lung
Squamous Cell Carcinoma
Intra-abdominal Carcinoma
Mammary Fibroadenoma
Mammary Adenocarcinoma
Other Tumors
Overall Tumor Incidence (%)
30

0
0
0
0
0
0
0
37

0
1
1
1
1
1
13
38

38
0
o
0
0
8
100
38

31
0
n
\j
3
n
\J
2
82
             -1     	•  ~ •  •«~-' f'*- i  ;jivv*t-' »*«a -J TU «   i IWWG V C I $  UC*_aUoC Ul
 autolysis  and/or cannibalization,  a few animals were unsuited for
 histopathological  examinations.


 Topical  Application  Studies--

     Bonser and  Manch (1932)  studied the tumor response  from application  to

 mouse  skin of three  samples  of Scottish blast-furnace tar, one sample  of

 English  crude tar, and  an  ether  extract of the  latter.   The three samples  of

 Scottish tar  (I,  n,  in)  were made from coke oven  charges which  contained in

 addition to the  coal, 15 to  17%, 25%, and  10% coke,  respectively;  the  English

 crude  tar  was made from a  charge containing 75%  coal and 25%  coke.   Sixty  mice

 were used  for testing each sample of tar.  There were no negative  and  positive

 control groups.  The hair was  clipped away from  a small area  of skin in  the

 region between the shoulder blades.  The tar  was applied biweekly  for  the

 first  14 weeks, and thereafter once weekly because of marked  ulceration  of the

 skin of many mice.  Tar samples were used without indication  of further

preparation in solvent.   The study was-continued for 56 weeks, by which  time

all  the mice had died.  Fifty-seven tumors were grossly identified.  Thirty-

one of the total 57 tumors that had developed were confirmed  historically.
                                     135

-------
    Tumor findings are described in Table VI-27.   In mice treated with  the
three Scottish samples, the first tumors appeared at the 18th  week.   The
Scottish I, II, and III tar samples produced a tumor incidence of 7/60  (12%),;
10/60 (17%), and 8/60 (13%), respectively.  The tumors were malignant in  three
mice.  The first tumor appeared at the 21st week when an English crude  tar  was
used.  Eight mice (13%) treated with the English crude tar developed tumors as
did 24 mice (40%) treated with an ether extract of the English crude tar.
Nine tumors in mice given the ether extract were malignant.
    The tumors were papillomas or squamous cell carcinomas of the skin.  The
carcinomas invaded the muscle.  One malignant tumor, seen after 47 weeks  of
application of ether extract of English tar, consisted of a mass of
"mononuclear round cells" invading the adjacent muscle and fat and metas-
tasizing to the lymph nodes.
TABLE VI-27.   INCIDENCE OF SKIN TUMORS IN MICE TREATED WITH BLAST FURNACE TARS
                      (adapted from Bonser and Manch 1932)

Number Of
Animals
with Tumors/
Tar Sample
Scottish I
Scottish II
Scottish III
English Crude
Number of
7/60
10/60
8/60
8/60
Animals
(12%)
(17%)
(13%)
(13%)
Appearance of
First Tumor
(weeks)
16
16
16
21

Malignant
Tumors
0/60 (0%)
2/60 (3%)
1/60 (2%)
0/60 (0%)
 Ether Extract
 of English Crude
24/60    (40%)
12
9/60    (15%)
                                      136

-------
     Hueper and Payne (1960)  found  that  skin  tumors  were  produced  in mice
 following  the  application  of coal  tar.   Coal tar, four petroleum  road asphalts
 (Venezuelan, Mississippian,  Oklahoman,  and Californi.an), one  petroleum  roofing
 tar,  and paraffin  oil were applied to the napes of  the necks  of groups  of 50
 black C57  mice (25 of each sex)  for  2 years.   An untreated control group
 consisted  of 200 mice.  A  positive control group was not used in this study.
 So  that the materials, could  be applied  as droplets, the coal  tar and roofing
 asphalt were heated  to make  them liquid, and the road asphalts were diluted
 with  a sufficient  amount of  acetone.  The paraffin  oil was painted on the
 skin.  Post-mortem examinations were performed on all mice, and histological
 examinations were  made of  all tissues which exhibited gross abnormalities.
 The results are found in Table VI-28.
    Carcinomas of  the skin were found in 22 of 50 (44%) and papillomas  in four
 of  50  (8%) mice receiving  dermal applications of coal tar, whereas control
 mice  did not develop tumors of the skin.
    Hueper and Payne also  administered some of the substances via inhalation
 and intramuscular  injection.  Daily volatilization of 10 to 30 g of coal tar
 did not produce lung tumors in female Bethesda black rats or strain 13 guinea
 pigs  inhaling the  fumes 5  hours daily*, 4 days per week, for periods up to 2
years.  However, coal tar distillate produced muscle sarcomas in 50 of 100
mice given 6 biweekly intramuscular injections and observed for a duration of
 2 years.

    Horton (1961),  in several experiments,  tested  a  number of crude coal tars,
coal tar distillates, and fractions of coal  tar for  skin  tumor response  in C3M
mice.   In  the first part of the study,  five  coal tars (four from the  coking  of
bituminous coal and one from the coking  of  lignite coal),  a mixture of one
                                     137

-------
    TABLE VI-28.  SKIN TUMORS IN MICE GIVEN DERMAL APPLICATIONS OF COAL TAR,
       PETROLEUM ROOFING TAR, PARRAFIN OIL, OR PETROLEUM ROAD ASPHALTS
                    (adapted from Heuper and Payne 1960)
Treatment
   Skin
Carcinomas
   Skin
Papillomas
   Total
Control
Coal Tar
Petroleum
Paraffin


Roofing Tar
Oil
0/200
22/50
1/50
1/50
(0%)
(44%)
(2%)
(2%)
0/200
4/50
0/50
1/50
(0%)
(8%)
(0%)
(2%)
0/200
23/50
1/50
2/50
(0%)
(46%)
(2%)
(4%)
Petroleum Road Asphalt
  Venezuelan               0/50     (0%)
  Mississippian            1/50     (2%)
  Oklahoman                0/50     (0%)
  Californian              1/50     (2%)
                   0/50
                   1/50
                   1/50
                   0/50
       (0%)
       (2%)
       (2%)
       (0%)
0/50
2/50
1/50
1/50
(0%)
(4%)
(2%)
(2%)
                                      138

-------
 of the bituminous coal  tars in  50% benzene,  and  a  benzo[a]pyrene  mixture were
 tested.   The authors  did not report using  a  control  group.   No  data were
 provided on the number  of mice  tested  nor  on the length  of time the animals
 were treated;  however,  the time-to-tumor for each  group  was  reported.  The
 incidence of tumors was reported  to be greater than  75%  (only a percentage was
 reported) for  each test group.  Morton developed a numerical index designed to
 grade the various tars  and tar  fractions for relative carcinogenic potency.
 This index was  referred to as the  potency  for a  minimum  concentration of
 material  (PMC).   A high PMC value  was  meant  to indicate  a greater carcinogenic
 potency.   For tars D-l  and 0-613,  for  which  multiple doses were applied, a
 dose-response was evident.   The mean time-to-tumor (in weeks),  the schedule of
 application, and  the  PMC  values for each of  the  tars, the tar solution, and
 the  benzo[a]pyrene solution  are reported in  Table VI-29.
     Two tars from the previous group (D-l and D-8)  were chosen to test the
 effect of skin washing  with  a detergent in water 5  to 60 minutes after tar
 application.  Tars D-l  and  D-8 had the highest (0.8)  and lowest (0.1)
 benzene-insoluble content,  respectively.  Washing delayed tumor development,
 but  the final tumor incidence was not significantly changed.   The delay was
 greater in the animals washed 5 minutes after dermal  application.
     Norton also determined the relationship between the amount  of
 benzo[a]pyrene in distillates of coal tar and the carcinogenic  potency of
those distillates.  Tar D-l, a distillate oil of  D-l  (the first  9  to 13.5%  of
the  distillation), a proportionate reblend  of nine  distillate fractions  of  D-l
and two distillate fractions (a  carbolic oil  and  a  light  creosote  oil)  of a
coal  tar  (D-9) not previously used in the experiments,  were tested for BaP
content and carcinogenic potency (PMC)  to the skin  of mice.   With  the
exception of Tar D-l,  all test materials were applied to  mice (strain
                                     139

-------
TABLE VI-29.  MEAN TIME-TO-TUMOR AND PMC VALUES FOR FOUR BITUMINOUS  TARS,
              ONE LIGNITE TAR, AND ONE SOLUTION OF BENZO[A]PYRENE
                        (adapted from Morton 1961)
Schedule of
Application
Treatment (Doses/week - mg/Dose)
D-l

n-4
D-5
D-5A


D-8
D-12
D-613



- bituminous tar

- bituminous tar
- bituminous tar
- 50% dilution
by weight of
D-5 tar
- bituminous tar
- lignite tar
- benzo[a]pyrene
in 85% beta-
methyl naphthalene
and 15% benzene
solution
2-1
2-50
3-100
2-10
2-10
2-10


3-50
3-50
2-15
2-50



Mean Time-to- PMC*
Tumor (weeks)
15. 6t
12. 6t
7.0t
24.8
23.6
25.1


21.9
17.1
33. Ot
30. 6t


iT* i n -t-vio Dwr
0.27t
0.37t
0.63t
0.13
0.14
0.13


0.11
0.16
0.08t
O.lOt


•inrre3<;p<; as
carcinogenic potency increases.

   tThe multiple doses for Tars D-l and D-613 demonstrated a mean
tumor and a PMC dose-response.
                                                                 time-to-
                                    140

-------
 unspecified)  in  10 mg doses.  Tar D-l was applied in 20 mg doses.  The number

 of applications  was described as "repeated," but neither the frequency nor the

 duration was  specified.  The PMC values and benzo[a]pyrene content of the test

 substances are reported in Table VI-30.

    Comparison of the benzo[a]pyrene content with the carcinogenic potencies

 of various fractions showed that no tumors were produced by those fractions in

 which no benzo[a]pyrene could be detected, while the carcinogenic potency of

 the test materials that contained benzo[a]pyrene was correlated with their

 content by weight of this carcinogen.  Despite this observation, the authors

 did caution that these results do not imply that benzo[a]pyrene is the only

 carcinogen in these substances.
TABLE VI-30.  PMC VALUES AND BENZO[A]PYRENE CONTENT FOR TWO COAL TARS, SEVERAL
DISTILLATES OF THOSE COAL TARS, AND A PROPORTIONATE REBLEND OF THE DISTILLATES
                            FROM ONE OF THE TARS
                          (adapted from Horton 1961)
Test Material
Tar D-l
Distillate Oil of
Doses (mg)
20
10
Content of
Benzo[a]pyrene (%
0.74
0.01
Relative Carcinogeni
) Potency (PMC)
0.27
0.01
c


  Tar D-l

Proportionate Reblend    10
  the Nine Cuts of
  Tar D-l

Carbolic Oil  of Tar
  D-9                    10

Light Creosote Oil        10
  of Tar D-9
0.08




0.00

0.00
0.11




0.00

0.00
                                     141

-------
    Wallcave et al. (1971) prepared benzene extracts of coal  tar pitches
obtained from coke ovens and tested them for carcinogenic activity  on  mouse
skin.  Equal numbers of male and female Swiss albino mice received  twice
weekly applications of 1.7 mg of coal  tar pitch in 25 ul  of benzene.   Exposed
animals survived for an average of 31 weeks.  Among 58 treated mice,  53
developed skin tumors, of which 31 were carcinomas.  Although tumors  at other
sites were present, the incidence in the control and experimental  groups  were
similar.  No carcinomas and only one papilloma on the skin were found in  26
control mice painted with benzene alone.  Wallcave et al. (1971) identified
several polycyclic hydrocarbons, including benzo[a]pyrene (0.84 and 1.25% of
undiluted coal tar pitch in 2 samples), in the pitch samples and concluded
that they were responsible for the tumorigenic effects observed.
                                      142

-------
 CARCINOGENICITY OF COKE OVEN EMISSION COMPONENTS
 Polycyclic Organic Matter (POM)
     Numerous polycyclic aromatic compounds are distinctive in their ability to
 produce tumors in skin and most1epithelial tissues of practically all  species
 tested.  Malignancies are often induced by acute exposures to microgram
 quantities of POM (for a review, see U.S. EPA 1979).  Latency periods  can  be
 short (4 to 8 weeks)  and the tumors produced may resemble human carcinomas.
 Carcinogenesis studies involving POM have historically involved primarily
 effects on the skin  or lungs.   In addition,  subcutaneous  or intramuscular
 injections are frequently employed to produce sarcomas at the injection  site.
 Ingestion  has not  been a preferred route  of  administration for  the  bioassay  of
 POM.   A listing of POM found in  coke oven emissions  is presented  in  Table
 VI-31  along with an indication of carcinogenic  activity.

 Other  Carcinogens  Identified in  Coke Oven Emissions
    The  contribution  of  compounds  other than  POM to  the carcinogenic activity
 of  coal  combustion products  has  received  little attention.  Other constituents
 of  coke  oven  emissions that  have  been found to be carcinogenic include
 arsenic, lead,  beryllium, chromium,  nickel, 2-naphthylamine, and benzene (U.S.
 EPA 1977a;  1978b, c;  1980n;  IARC  1973b; 1974; 1976;  1979).

j)ocarcinogens
    Numerous compounds, which by themselves display no carcinogenic activity,
are known to enhance the tumorigenic activity of B[a]P when applied together
to the skin of mice (Hoffman et al. 1978,  Van Duuren  and Goldschmidt 1976).
These so-called cocarcinogens include certain PAH-containing fractions  of
tobacco tar, and several  structurally diverse compounds (catechol,  pyrogallol,
                                     143

-------
TABLE VI-31.
                       POLYCYCLIC ORGANIC MATTER (POM)  IDENTIFIED IN
                            COKE OVEN EMISSIONS*
Compound
                                 Animal Carcinogenicityt
                              IARC                    CAG
Anthracene
Benz[a]anthracene
Oi benz[a ,c]anthracene
Methylphenanthrene
Phenanthrene
BenzoCc]phenanthrene
Benzo[a]fluorene
Benzo[b]fl uorene
Dihydrobenzo[a]fl uorene
Dihydrobenzo[b]fluorene
Dihydrobenzo[c]fluorene §
Fluoranthene
Benzo[c]fluorene
Benzo[b]f1uoranthene
Benzo[j]f1uoranthene
Benzo[k]f1uoranthene
Benzo[ghi]f1uoranthene
Pyrene
Methylpyrene
Benzo[a]pyrene
Benzo[e]pyrene
Dibenzopyrenes
Chrysene §
Triphenylene §
Perylene
Benzo[ghi]perylene §
Anthanthrene §
Coronene
Acridine
Benzoquinoline
Octahydrophenanthrene
Octahyd roanth racene
Dihydrofluorene
Benzindene
Fluorene
Dihydrophenanthrene
Dihydroanthracene
Methylfluorenes
Fluorene  Carbonitrile
Methyl anthracene
Ethylphenanthrene
Ethyl anthracene
                                              (continued on the following  page)
                                      144

-------
                             TABLE  VI-31.   (continued)
Compound
   Animal
I ARC
Carci nogeni ci tyt
                                                                     CAG
Octahydrofluoranthene §                      ?
Octahydropyrene §                            ?
Indeno[l,2,3-cd]pyrene                       +                         +
Dibenz[a]anthracene                          +
Benz[c]acridine                              +                         +
Dibenz[a,h]acridine                          +                         +
Dibenz[a,j]acridine                          +                         +
Dihydrofluoranthene'                         ?
Dihydropyrene                                ?
Methylfluoranthene                          _+
Dihydrobenz[a]anthracene §                  7
Dihydrochrysene §                            ?
Dihydrotriphenylene §                        ?
Dihydromethylbenz[a]anthracene §             ?
Dihydromethylchrysene §                      ?
Dihydromethyltriphenyl ene §                  ?
Methyl benz[a]anthracene                      4^
Methyltriphenylene                          T
Methyl chrysene                            .   +_
Dihydromethylbenzo[k and b]-                ~
 fluoranthenes§                              ?
Dihydromethylbenzo[a and e]pyrenes §         ?
Dimethylbenz[a]anthracene §                  +                         +
Dimethyltriphenyl ene §
Dimethylchrysene §                           +
Methylbenzo[k]fluoranthene §                 ?
Methylbenzo[b]fluoranthene §                 ?
Methylbenzo[a]pyrene                         +
Dimethylbenzo[k and b]-
 fluoranthenes                               ?
Dimethylbenzo[a]pyrene                       +
o-Phenylenepyrene                            ?
Methyldibenzanthracene                       +
Methylbenzo[ghi]peryl ene                     ?

    *The POM's were identified in coke oven emissions by Lao et al. 1975 or
Bjorseth et al. 1978.   The data on carcinogenicity is taken from CAG  (1980b)
and IARC (U.S. EPA 1979).

    tSymbols:  + complete carcinogen or. tumor initiator
               - negative
               ? activity not known
               +_may be positive or negative depending on the isomer tested

    Confirmation  of chemical structure questionable in Lao et al. (1975).
                                     145

-------
anthralin, decane, undecane, tetradecane).  Since many of these compounds  may
occur in coke oven emissions, the possibility arises that they may contribute
to carcinogenic risk.  However, the mechanism of cocarcinogenesis is not
understood, and its relevance to tumor formation in tissues other than  mouse
skin is not known.  Thus, we can only conclude that the presence of
cocarcinogens in complex mixtures such as coke oven effluents  may pose  an
additional risk for humans beyond that attributable to recognized carcinogens
such as benzo[a]pyrene.
                                     146

-------
                          VII.  QUANTITATIVE ESTIMATION
 INTRODUCTION
      This quantitative section deals with the unit risk for coke oven emissions
 in air and the potency of coke oven emissions relative to other carcinogens  that
 the CAG has evaluated.  The unit risk estimate for an air pollutant is defined as
 the lifetime cancer risk occurring in a population in which  all  individuals
 are exposed continuously from birth throughout their lifetimes  to a concentration
 of 1 ug/m3 of the agent in  the air they breathe.   This  calculation  is  done to
 estimate in quantitative terms the impact of the  agent  as  a  carcinogen.   Unit
 risk estimates are used for two purposes:  1)  to  compare the carcinogenic
 potencies  of several  agents with  each other,  and  2)  to  give  a crude  indication of
 the population risk that  might be  associated with  exposure to these  agents, if
 the actual  exposures  are  known.
      The data  used for the  quantitative  estimate  can  be of two types:  1) life-
 time  animal  studies,  and  2)  human  studies where excess cancer risk has been
 associated with exposure  to the agent.   It is  assumed, unless evidence exists
 to  the contrary, that  if  a  carcinogenic  response occurs at the dose levels used
 in  a  study,  then  responses  will occur at  all lower doses with an incidence
 determined by the extrapolation model.
      There is  no solid scientific basis  for any mathematical  extrapolation
model that relates carcinogen exposure to cancer risks at  the extremely low
concentrations which must be dealt with  in evaluating environmental  hazards.
For practical reasons, such  low levels of risk cannot be measured directly
either by animal experiments or by epidemiologic studies.   We must,  therefore,
depend on our current  understanding of the mechanisms of carcinogenesis  for
guidance as to which risk model to use.   At  the present  time, the dominant
                                      147

-------
view of the carcinogenic process involves the concept that most agents  that
cause cancer also cause irreversible damage to DNA.   This  position  is reflected
by the fact that a very large proportion of agents that cause  cancer are  also
mutagenic.  There is reason to expect that the quanta!  type of biological
response, which is characteristic ,of mutagenesis,  is associated with a  linear
non-threshold dose-response relationship.  Indeed, there is substantial evidence
from mutagenesis studies with both ionizing radiation and  a wide variety  of
chemicals that this type of dose-response model  is the appropriate  one  to use.
This is particularly true at the lower end of the  dose-response curve.  At higher
doses, there can be an upward curvature probably reflecting the effects of
multistage processes on the mutagenic response.   The linear non-threshold dose-
response relationship is also consistent with the  relatively few epidemiologic
studies of cancer responses to specific agents that  contain enough  information
to make the evaluation possible (e.g., radiation-induced leukemia,  breast and
thyroid cancer, skin cancer induced by arsenic in  drinking water, liver cancer
induced by aflatoxins in the diet).  There is also some evidence from animal
experiments that is consistent with the linear non-threshold model  (e.g., liver
tumors induced in mice by 2-acetylaminofluorene  in the large scale  EDgi study
at the National Center for Toxicological  Research  and the  initiation stage of
the two-stage carcinogenesis model  in rat liver  and  mouse  skin).
     Because of these facts, the linear non-threshold model  is considered to be
a viable model  for any carcinogen, and unless there  is direct  evidence  to the
contrary, it is used as the primary basis for risk extrapolation to  low levels
of exposure.
     The quantitative aspect of carcinogen risk  assessment is  included  here
because it may be of use in the regulatory decision-making process,  e.g., in
setting regulatory priorities, evaluating the adequacy of  technology-based

                                      148

-------
controls, etc.  However, it should be recognized that the estimation of cancer
                                                                           •a
risks to humans at low levels of exposure is uncertain.  At best, the linear
extrapolation model used here provides a rough but plausible estimate of the
upper limit of risk, i.e., it is not likely that the true risk would be much
more than the estimated risk, but it could very well be considerably lower.
The risk estimates presented below should not be regarded as accurate
representations of the true cancer risks even when the exposures are
accurately defined.  The estimates presented may, however, be factored into
regulatory decisions to the extent that the concept of upper risk limits is
found to be useful.
ESTIMATION OF THE UNIT RISK—CONSIDERATIONS
The Need to Employ Mathematical Models, and the Errors They Introduce
     To estimate a unit risk directly, one would need a cohort of individuals
exposed from birth to death to 1 ug/m3 of coke oven emissions.  The estimate
of the unit risk would then be the observed number of respiratory cancer deaths
in the cohort minus the expected number of respiratory cancer deaths divided by
the number of individuals who were originally in the cohort.   The expected
number of respiratory cancer deaths would be based upon an  equivalent nonexposed
population.  Obviously, no such ideal  situation will  ever exist for a human
population.
     To estimate the unit risk from human data, we must make  use  of the  information
contained in epidemiol ogic studies.  Typically, epidemic! ogic studies involve
workers whose exposure started and stopped at different ages  and  fluctuated  over
time due to changes in job classification and working conditions.  Also,  workers
enter the cohort under observation for cancer mortality at  different ages and are
followed for different lengths of time (usually less  than their full  lifetimes).

                                      149

-------
     In order to use this type of fragmented information to predict lifetime
cancer mortality due to lifetime environmental  exposures, it is necessary to
postulate a mathematical  relationship between the future age-specific cancer
death rate and the past exposure pattern.  Once this relationship is established,
it can then be used to estimate the unit risk by a well-known mathematical
technique.
     The use of a mathematical relationship or model, although necessary, introduces
several major sources of potential  error in the estimation of risk.  All  such
models contain unknown parameters that must be estimated from the data obtained
from epidemiologic studies.  The parameter estimates are variables that may
differ from the true unknown parameter values by amounts that are large enough
to alter estimates of risk considerably for low levels of exposure.  To express
the plausible range of risk associated with an assumed mathematical model, con-
fidence bounds on the risk are obtained.  These bounds are values between which
we have a specified degree of confidence that the true unknown risk lies.  However,
the accuracy of these derived confidence bounds depends upon how well the assumed
model corresponds to the true dose-response relationship.
     The mechanisms by which a chemical induces cancer are not presently understood.
As a result, it is not possible to derive any mathematical model  that relates
exposure to cancer rate with a high degree of plausibility.
     The difference between the assumed dose-response relationship and the true
unknown response constitutes an additional source of error in estimating risk.
This source of error, however, is not very great if the levels of exposures on which
a risk estimate is desired are near the levels of exposure which existed in the
epidemiologic study from which the dose-response model is derived.  In such a
case, the confidence bounds constitute a reasonable approximation of the total
uncertainty.  However, as the difference between the two exposures increases, the

                                      150

-------
 uncertainty  associated  with  the  difference  between the  assumed and actual dose-
 response model  increases  rapidly, and eventually exceeds all other sources
 of  uncertainty.
      The next  section discusses  some of the general factors that should be
 considered in  selecting a form for the mathematical model to be used to
 generate unit  risk  estimates.

 Selecting a  Form of a Model  for  the Age-Specific Exposure-Induced Respiratory
 Cancer Death Rate"       ~~~                                     ~~
      A number  of general principles, some of which stem from the logical
 basis of scientific method,  are  relevant in deciding how to select a model to
 describe the age-specific exposure-induced cancer death rate.

 The Model Should Conform to the  Observed Data Set--
      For any set of data there exists an infinite number of mathematical models
 that  can give predictions which  fall within the limits of the random error
 associated with the data.   Some  of these models fit the data better than others.
 Hence, there is a temptation to  select the model that  gives the "best fit" as
 the one to use in making risk predictions.  The trouble with this approach is
 that the "best-fitting model" selected in this manner  is itself a random variable.
 A small change in the data could alter what was defined as  the "best-fitting
model."  If the predictions of risk relate to exposures that fall  within the
 range of the experimental  data used to derive the model, it is of little practical
consequence which of the adequately fitting models  is  used.   However,  the  choice
of model  becomes critical  when one attempts to extrapolate  far beyond the  range
of the experimental  data.   Two models  that give an  adequate fit to  the data  in
the experimental range can give answers  that differ by many orders  of magnitude
for environmental  levels of exposure.
                                      151