A FT                 REVIEW DRAFT
                                             JULY 1978
             DO NOT QUOTE OR CITE
             POM EXPOSURES

        This document is a preliminary draft. It has not been
        formally released by EPA and should not at this stage be
        construed to represent Agency  policy.  It is being
        circulated for comment on its technical accuracy and
        policy implications.
               DO NOT QUOTE OR CITE
            Office of Research and Development
               Washington, D.C. 20460

                        DO NOT QUOTE OR CITE

                   PRELIMINARY REPORT ON

                       POM EXPOSURES

  This Document Is Being Released By FPA For External Review
                                   Roy/[ (I Albert, H.D,
                                   Chai rrnan
                                            00 NOT QUOTE  OR CITE
    Tnis report Is dn asi.essiiie.it or tne carcinogenic impact

o^  ciroorne polycyciic organic C'jivi;,ounus.   The polycyclic

hyorocarouns are a complex mixture OT compounds produced oy

combustion of oryanic niatter.  Toe composition ot HOf-i

mixtures undouotealy varies consi uc-rau"! y depeiioing on the

source.  The primary basis tor the assessment reported here

is an analysis of tour epi aeini al ogical stuaies which

exaiiiined the lung cancer experience in tnree occupational

circumstances involving Heavy exposure to HUM'S from various

sources and one urban-rural population study.  In all cases

oenzo(a)pyrene was useu as trie indicacor of trie amount of

exposure.  Also induced is an analysis of tr.ree animal

studies which deal with the lung cancer response to

benzol a)pyrene alone.  Tnere are no animal stuaies wnich

deal  with tne response to complex Mixtures of HUM as, for

example, coke oven, automobile, or fossil  fuel power plant

emissions as the basis tor evaluating the relative

carcinogenic activity ot PUM's ana benzo(a)pyrene.  However,

it is of interest tnat one of the aninal studies (Yanyshiva

et aj.) provided the oasis for the uenzo(a)pyrene standard

tor general  air pollution whicn is useo. in the uSSR.

    There is good consistency in the estimates ot excess

lung  cancer from PUH exposure rrodi trie epl aemlol ogical

stuaies using a linear non-thresholu extrapolation model.

Tne excess lung cancer incidence tor liretime exposure to
POM averaged u.^b* per nanogram/m3 witr. a range ot
U.3uWng/m3 to U.ko/ng/r.i-*.  Tne estimates oaseu on
animal studies compared to the epidemic!ogical stuoies
ranged from a factor ot 3 higher lor the Yanysniva chronic
benzo(a)pyrene intubation study in rats to a  tactor ot lu
lower in  the chronic ueiizo( a) pyrene intubation stuoy in
hamsters  (Feron) to a tactor ot lou lower in  a cnronic
inhalation  study to yenzo(d)pyrene ana Su^ in rats
(Laski n).
      Using  the epi aeini ol ogi cal  data as a oasis tor  the risk
estimate,  HUH is estimated to  cause about two hunareo. deaths
per year  nationwide.

                   Industrial Epidemiological Studies

        The main problem in utilizing specific industrial exposures to

POMs to predict cancer response in the environment is that the spectrum of

the POMs to which workers are exposed may be very different froca that

encountered in the general environment.  For example, the Ba? fraction of

POMs is much less in automobile exhaust than in coal combustion products.

Also, the fact that potential carcinogens other than POMs may be part of

the industrial exposures precludes an unconfounded assessment of the

effect of the POMs.  However, there is rather close agreement in the

results of the three studies discussed immediately below  regarding the

carcinogenic effect measured by the amount of the BaP indicator present.

This agreement tends to indicate  that  the estimates are reasonable.

Coke Ovens -- GAG 197R Report

        In a recent revision of a GAG  report on  the  risk  to  the U.S.

population due to coke oven  emissions, it was estimated  that  the lifetime

probability of lung cancer due  to a  lifetime exposure of  x ugm/m   of  BSO

in the  in  the  inhaled air  could be expressed as

                  _  .031521  +  .00092975X

            Q2() ~     1 + .00092975X

A close approximation of  the change  in the  background  rate is thus

 (.00092975  *  .031521)  x  100  =  2.95%/ugm  BSO.  Under  the  assumption that

 1% of  the  BSO  is BaP,  this is  a change of .295%/ngm BaP.

 Gas Workers  -- Doll  (1972)

         It was estimated that  the British gas carbonization workers were

 exposed to 2 ygm/m3  of  BaP in their working environment for 22% of the

year.  If we make the additional assumption that the workers spent their

entire working lifetimes in this job, we estimate a lifetime average

exposure of

            x = 2 x IQ3 x .22 x 45 T 70 = 283 ngm/m3 BaP.

This exposure was estimated to have produced an increase in the yearly

age-adjusted lung cancer rate of 1.6 x 10  .  The age-adjusted lung cancer

rate in England and Wales during the study period was about 2.0 x 10

Thus the percentage increase in the lung cancer rate per pgm/m  of BaP is

estimated to be (1.6 - 10~3) * (283 x 2.0 x 1Q~3) x 100 = .283%/ngm BaP.

Roofers — Hammond et al. (1976)

        The mortality experience of 5,939 roofers and waterproof ers

exposed to hot pitch fumes for at least nine years was studied over a two-

year interval.  Estimates of exposure were obtained from breathing samples

of dust collected on masks worn during various job tasks.  It was esti-

mated that the average worker inhaled 16.7 pgm BaP per seven-hour working


        The SMR for lung cancer in this sample was found to be 1.52 for

20 to 29 years of exposure and 1.61 for 30 to 39 years of exposure.  The

lifetime air concentration x in ugm/m  required to obtain the same life-

time exposure as that of workers, assuming a 240-day working year and a

breathing volume of 24 m  per day, is
                16.7 x 240

            X ~
                 24 x 365 x 70

where £ = the number of years of on-the-job exposure and 70 is the average


                                                             DO  NOT QUOTE Gf. C
         From  these data we can obtain two points that can in turn be used
 to estimate the increase in the lung cancer rate per change in -gm of 3aP
 exposure.  In  the GAG  (1978) report on risk due to arsenic exposures, it
 was shown  that, under  a set ot simplifying assumptions, the slope of the
 line through  the origin of the relationship between the standard  mortality
 ratio minus one and the exposure is an estimate of the change in  the rate
 per change of unit exposure.  Assuming the average of the exposure dura-
 tions to be the midpoint of the exposure duration intervals,  namely 25  and
 35 years respectively, it follows that ,
               y = SMR - 1        x = exposure ygm BaP/m
                   • 52                     163.4
                   •61                     228.8
so that an estimate of the change in the rate  per change of unit  exposure  is

          79 QA9  =  -00284,  or  a  change of  .284%  per  ngm BaP/m3.
                 Environmental Epidemiological Study —
                         Carnow and Meier  (1973)

         In  a very detailed  study  that  took into account smoking

 and  other epidemiological factors, Carnow and Meier estimated that an

 increase of 5%  in the  lung  cancer rate could be expected for a change of
 1  ngm/m of BaP in  the atmosphere.  This study has been criticized (in

 our  view somewhat unjustly) for the necessity of utilizing crude measures

 of effect and exposure such as county-adjusted BaP levels and statewide

 tobacco sales and lung cancer rates.  Crudeness in data invalidates only a

 negative result.  A statistically highly significant result was found for

 the  association between lung cancer death rates and BaP measured which

 was  consistent  for  sex and  race,  and thus should be looked at with some

        However, as noted by Carnow and Meier, the use of 1967-69 BaP

levels may not reflect at all accurately the lifetime average exposure of

those people who died in the 1959-61 period.  For example, as discussed by

Pike et al. (1972), the BaP levels in downtown Los Angeles changed from

        3                        3
31 ngm/m  in 1952-53 to 1.6 ngm/m  in 1959 as a result of controls imposed

primarily on refuse burning.  Making the rather tenuous assumption that

the Los Angeles situation is typical nationwide, we have a 19.4-fold drop

in exposure, which would change Carrow and Meier's estimate to an increase

of 5% T 19.4 = .258% in the lung cancer rate per increase of 1 ngm BaP in

the atmosphere.

                 Discussion of Results of Industrial and

                  Environmental Epidemiological Studies

        Even though these studies were based on very different sources of

exposure to POMs and different methods were employed to obtain the esti-

mates, the end results are remarkably consistent.  The geometric mean of

these four studies is
            x  - 4/(.284)(.283)(.295)(.258) = .2797
and an estimate of the standard deviation of the log x  is
            3..   -  = .01236 .
             log x
These values will be utilized in the exposure section to give a crude esti-

mate of the variability of the health hazard.

                             Animal Studies

        The main problem in utilizing animal experiments to quantify the

carcinogenic effects of POMs is that no study of carcinogenic effect has

been conducted based on what might be called a "typical" POM mixture.  If

we utilize the studies employing BaP exposures we are still forced to make

an additional assumption specifying what percentage of the total carcino-

genic activity of a "typical" POM mixture is caused by BaP.  As a result,

the findings in the animal studies are viewed as having only a very limited


        The geometric average of the estimates of carcinogenic potency

derived from the animal studies discussed in the next section will be

obtained.  This average will then be compared to the geometric average

obtained from the human epidemiological studies to estimate a potency of

BaP relative to a POM mixture indexed by BaP.  If this estimate appears to

be reasonable then we will say  that the animal data is consistent with the

human data, thus giving us greater confidence in the epidemiological evi-


Intratracheal Intubation Studies

        In order to utilize an  intratracheal intubation animal study to

predict human risk it is necessary to equate the animal exposure usually

given in total lifetime mgs to  the exposure level in air breathed by man.

We assume that an equivalent exposure between man and animal that results

in the same lifetime probability of tumor death is the daily average life-

time exposure on a mg-per-surface-area basis.  If an animal is given a

total of x mg of a chemical during its lifetime and the air exposure to man


is u mg/m , then the equivalent dose is

                    x           24u
            W 2/i x a  x 365   W
             a       a          m

where W  and W  are the weights of man and animal respectively and I  ±s
       ma                                                     a

the lifespan in years of the animal.  This equivalence is based on the

following assumptions:

   (1)  All the chemical breathed into the lungs is absorbed

   (2)  Man breathes in 24 m  of air per day

   (3)  The breathing rate for all animals is proportional to their surface

area, which is taken to be proportional to the 2/3 power of their weight.

Solving for u we have that

            u = 	

                24 x 365 x i  x w  '
                            a    a

Thus a dose-response relationship for an animal experiment based on x mg

can be converted to a human dose-response relationship by using the equiva-

lent human doses, u, instead of x in the estimate of  the parameters of  the

dose-response curve.

Feron et al. (1973)

        A clear positive respiratory tract tumor dose-response relationship

was obtained in Syrian golden hamsters by repeated intratrachael instilla-

tion of a benzo(a) pyrene-ferric oxide mixture suspended in .9% NaCl solu-

tion.  The lowest dose, 3.25 mg total per lifetime, induced a 10% response

in the experimental animals while none was observed in the controls.

        The equivalent human air exposure in this case is
            u	(3.25)(70;2/3        - .0149

                24 x 365 x 1.5 x (.15) '

where:      W  = 70 kgm,
                    DO NOT  v^-i

            1  =1.5 years, and

            x  = 3.25 mg.

The estimated slope for humans from the one-hit curve is thus

            3 = -ln[ln -  .1] T .0149 = 7.071


where exposure is in mg/m in air.  To express this rate as percentage


increase per ngm/m  air exposure, we assume a  .0315 probability of

respiratory trace cancer  death in the absence  of BaP, giving a value

[7.071xlO~6T  .0315] x 102 = .0224% increase per change of r,gm/m3 of BaP.

Yanysheva and Antomonov (1976)

        In this article little information was given about the experimen-

tal animal.  We make the assumptions that the rats were W    .400 kgm in

weight, a common rat size, and their lifespan was equivalent to the longest

time noted to tumor appearance, 2.  =2.25 years.

        A total of 5 malignant epithelial tumors of the lung were observed

in a group of 32 animals exposed to a total dose of .5 mg in 10 periods  of

administration.  The human- equivalent dose is thus estimated to be
            u =
                24 x 365 x 2.25 x (.4)
                -4    3
-JTJ =  7.935 x  10  mg/m ,
the one-hit slope estimate is
            3 = -ln[l - 5/32] / (7.935 x 10"4) = 2.1406 * 10~,
and the percentage change per ngm/m  BaP is

             (2.1406 x 102 T .0315) x 10~4 = .680%.

Inhalation Study — Laskins et al.  (1970)

        It is assumed that air exposures result in equivalent carcinogenic

responses for all species.  Laskins, exposing rats to 10 rag/ra  of BaP for

1 hour a day 5 times per week for their lifetimes and an additional SO

exposure, produced 5 tumors in 21 rats.  An SO  control group and a BaP-

exposed group in the absence of S0? did not show any tumors.

        The equivalent lifetime average exposure to man that would result

in the estimated resoonse rate of 5/21 = .238 is
            u = 10 x 5/7 x 1/24 =  .2976 mg/m3.
Thus an estimate of the one-hit slope for man is

            B = -ln[l - .238] * .2976 = .9133
and the percentage change in the rate per ngm/m  of BaP is
            [.9133 v .0315] * io"4 = .0029% .
Discussion of Animal Studies

        The three estimates derived from the animal experiments varied 235-

fold.  However, taking into account the differences in species utilized and

in exposure routes, differences of this magnitude are not really surprising.

The geometric mean of these estimates is
            x  = 3»(.0224)(.680)(.0029) = .03535 .

Compared to the geometric mean of x  = .2797 for the human epidemiological

experiments, we have an estimate of the relative potency
            o = x /-  = .03535 T .2797 = .1264 .
                 a x


Thus 12.6% of the carcinogenic activity of  the POMs is estimated to be

attributable to BaP.  A value of this magnitude would not appear to be

grossly inconsistent with intuitive estimates.  As a result the animal

data do not give any reason to doubt the results obtained from the epi-

demiological studies.

                 Risk to U.S. Population Exposed to PQMs

        Energy and Environmental Analysis,  Inc., estimates the exposure,

D, in units of 1,000 people  * ngm/m  for each state and the nation as a

whole (table E-2, Atmospheric POM:  Sources and Population Exposure Esti-

mate, Draft).

        Assuming that, in the absence of POM exposure, the lifetime

probability of death due to respiratory cancer is  .0315 and the increase

in this rate is .2797% per ngm/m  of BaP (the geometric mean of the epi-

demiological studies), an approximate estimate of  the number of respira-

tory cancers caused per year by POMs is

            ND = .0315 x .002797 x D x 1Q3  T 70.96 =  .1242 x 10~2D

where D is the exposure in the given units  and 70.96 is the average U.S.


        For the entire nation, D is estimated to be 170,000 so that

N  = 211 deaths/year due to POM exposure.

        An estimate of the number of deaths per year in any individual

state is made by assuming that that state has the  same proportion of

deaths as its exposure, D , is to the total U.S. exposure, D.  Thus

                 D                                     -

            {I  = _§. M= D  211 i 170,000 = 1.241  x 10  D  .
             s    D  D    s                              s



                                                          DO IW i yw~.-. viv ui»

For example, the expected number of respiratory deaths in the state of

Washington due to POM exposure is estimated to be

            N  = 1.241 x 10~3 x 1,700 = 2.11 deaths per year.

Confidence Limits on Numberof Lung Cancer

Deaths per Year Due to POMs

        If we are willing to accept the assumption that each of the epi-

demiological studies gives an unbiased estimate of the percentage increase

in the lung cancer rate per ngm/m  of BaP and that each study is equally

valid, then it is possible to obtain approximate confidence limits for the

estimated total number of respiratory tumor deaths per year caused by the

POMs in the atmosphere.

        The total number of respiratory tumor deaths per year may be writ-

ten in the form
            ND = Ki3D

where K = (.0315 v 70.96) x 10  = .44391 is a constant derived from U.S.

health statistics; 3 is the change in the mortality rate per change of

         3                                                                3
ngm BaP/m  exposure; and D is the exposure in units of 1,000 people x ngm/m

of atmospheric BaP.

        Under the additional set of assumptions that:

        (1)  Our estimate of 6 is lognormally distributed so that
            log 3 = log x  = log (.2797) = -.553308

                                                      2        *       2
is normally distributed with mean log g and variance j    ~ = (a..   — )

= .15277 x 1Q~3


        (2)  Our estimate of D is lognormally distributed so that

            log D = log (170,000) = 5.23045

                                                      O    1     O

is normally distributed with mean log D and variance a  = [~Tq7~] » where

the value of u is a "guesstimate" of the level of precision of our estimate

of D expressed in terms of being 95% confident that our estimate is between

[(1/u - 1) to u - 1] x 100% of the true value

        The estimate of the total number of respiratory tumor deaths per

year is


            ND   (.44391)(.002797)(170,000)   211.07  .

Thus log ND = 2.32444 is normally distributed with mean log N  and

          2   ,  .  2
variance af"   „ + o_
          log 6    D
so that a 95% confidence interval  for  log ND given  u  is
            log  (211) ± 1.96/.153  x  10~3 +  (\0
                                     =  log  211  ±  /.588    10"3 +  Iog2u  .
        For various values of u  the upper and lower 95% confidence limit


of the estimate  N  = 211 is given below.

- 1) x 100%
lower limit
upper limit
                 *Assumes  no  error  in  exposures.

         We note chat due to the highly consistent independent values

obtained from the epidemiological studies,  most of the variability is due

to lack of precision in the exposure estimates.  However it is felt that

the variability estimate associated with the epidemiological studies most

likely underestimates the true variability.


                                                       DO NOT QUOTE OR CITE


Carnow, B. W.,  and P.  Meier.   "Air Pollution and  Pulmonary Cancer."
        Arch. Environ. Health 27 (3):207-218,  1973.

Doll, R., M. P. Vessey, R.  W. R. Beasley,  A. R.  Buckley,  E. C.  Fear,
        R. E. W. Fisher, E. J. Gammon,  W.  Gunn,  G.  0.  Hughes, K. Lee,
        and B.  Norman-Smith.   "Mortality of Gasworkers:   Final  Report
        of a Prospective Study." Brit.  J.  Ind.  Mrd. ^9:394-406, 1972.

Feron, V. J., D. De Jong, and P. Emmelot.   "Dose-Response Correlation for
        the Induction of Respiratory-Tract Tumours  in  Syrian Golden
        Hamsters by Intratrachael Instillations of  Benzo(a)pyrene."
        Europ.  J. Cancer. 9_:387-390, 1973.

Hammond, E. C., I. J.  Selikoff, P. L. Lawther,  and  H.  Seidman.  "Inhala-
        tion of Benzopyrene and Cancer in Man."  Ann.  N.Y. Acad. Sci.
        271.: 116-124, 1976.

Laskin, S., M.  Kuschner, and R. T. Drew.  "Studies  in  Pulmonary Carcino-
        genesis."  In Inhalation Carcinogenesis,  pp.  321-350.   Edited by
        M. G. Hanna, P. Nettesheim, and J. R. Gilbert.  AEC Symposium
        Series no. 18.  Washington, D.C.:   U.S. Atomic Energy Commission,

Pike, M. C., R. J. Gordon,  B. E. Henderson, H.  R. Mench,  and J. SooHoo.
        "Air Pollution."  In Persons at High Risk of  Cancer:  An Approach
        to Cancer Etiology and Control.  Edited by  J.  F.  Fraumeni,  Jr.
        New York:  Academic Press, 1975.

U.S. Environmental Protection Agency.  Carcinogen Assessment Group.
        Preliminary Report on Population Risk to Ambient Coke Oven
        Exposures.  Washington, D.C., 1978.

Yanyshiva, N. Ya, and Yu G. Antomonov.  "Predicting the Risk of Tumor
        Occurrence under the Effect of Small Doses  of  Carcinogens."
        Envir. Health Perspect. 13:95-99, 1976.