United States
Environmental Protection
Agency
Off iC8 Of
Research and Development
Washington, DC 20460
EP A/600/8-91/053
July 199,1
Upper-Bound Quantitative
Cancer Risk Estimate for
Populations Adjacent to
Sulfur Mustard Incineration
Facilities

-------

-------
                                               EPA/600/8-91/053
                                               July 1991
                                               Final
UPPER-BOUND QUANTITATIVE CANCER RISK ESTIMATE FOR

     POPULATIONS ADJACENT TO SULFUR MUSTARD

               INCINERATION FACILITIES
              Human Health Assessment Group
        Office of Health and Environmental Assessment
             Office of Research and Development
            U.S. Environmental Protection Agency
                    Washington, DC
                                          . rS> Printed on Recycled Paper

-------
                                   DISCLAIMER

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

-------
                                  CONTENTS

Tables	 jv

Preface		  v

Authors and Reviewers	 vi

1. INTRODUCTION	..-	 .	1

2. HAZARD IDENTIFICATION		  5

   2.1.  HUMAN STUDIES  . .	         5
   2.2.  ANIMAL STUDIES		.10
   2.3.  SUPPORTING EVIDENCE	 ...	14

3. DOSE-RESPONSE EVALUATION .  .	 .		16

   3.1.  HUMAN STUDIES		           16
   3'.2.  ANIMAL STUDIES		. .	17

        3.2.1.  McNamara et al. (1975) Rat Studies	 .	17

              3.2.1.1. Toxicity Experiment	17
              3.2.1.2. Carcinogenicity Experiment	21
              3.2.1.3. Unit Risk	 . . .	24

        3.2.2.  Relative Potency in strain A Mice	 .	25

              3.2.2.1. Heston (1950)			26
              3.2.2.2. Shimkin and McClelland (1949)	.27
              3.2.2.3. Stoner et al. (1984)	28

        3.2.3.  Relative Toxicity Approach (Watson et al., 1989)	31
        3.2.4.  Summary and Conclusions	32

4. EXPOSURE EVALUATION	 35

5. RISK CHARACTERIZATION	 .	35

   5.1.  OTHER QUANTITATIVE RISK ASSESSMENTS	'.".'..	.36

        5.1.1.  Watson et al. (1989) .,	36
        5.1.2.  EA Engineering, Science, and Technology Inc. (1987)  	36
        5.1.3  Rosenblatt (1987)	36
        5.1.4.  The Public Health Service (CDC, 1988)	  37
                                      in

-------
                               CONTENTS (continued)
                                    I             •                   -   .
                                    [                           _             •••-._
   5.2.  CURRENT ASSESSMENT	37

6. REFERENCES	40


                                      TABLES


 1-1. Identification of the chemical agents in the U.S. Army disposal program	    2

 1-2. Sites of storage and disposal of U.S. Army chemical agents .	   3

 2-1. Summary of human studies of sulfur mustard poisoning	   6

 2-2. Incidences of skin carcinomas among rats in the  McNamara et al.
     toxicJty study	 ...  12

 2-3. Incidences of skin carcinomas among rats in the  McNamara et al.
     carcinogenicity study (sexes pooled) . .	13

 3-1. Incidences of skin carcinomas among rats surviving past the discovery of the first
     tumor in the McNamara et al. toxicity study	18

 3-2. Lifetime average daily exposures to rats in the  McNamara et al.
     carcinogenicity study and associated incidences of skin carcinomas	23

 3-3. Lung adenomas in strain A mice receiving sulfur mustard
     intravenously in four injections	26

 3-4. Mean numbers of tumors per mouse observed  at various times after single
     injections of 20-methylcholanthrene	 .	27

 3-5. Lung adenomas in strain A mice induced by intraperitoneal and gavage
     administration of 3-methylcholanthrene (MC) and  benzo[a]pyrene (BaP)	.28

 3-6. Slope parameters and relative potencies obtained using a weighted leastl-squares
     procedure for the tumors per mouse data	30
                                        IV

-------
                                      PREFACE
        This document was prepared by the Human Health Assessment Group of the Office
of Health and Environmental Assessment (OHEA), Office of Research and Development, U.S.
Environmental Protection Agency, at the request of the Office of Solid Waste.  It provides an
upper-bound quantitative estimate of cancer risk for a person who might live close to (at the
fence line of) sites where the U.S. Army is planning to incinerate existing stocks of sulfur
mustard. These worst-case estimates are intended to be of use in writing permits for the
operation of incineration facilities, although the estimates are highly uncertain because of the
limited amount of available information. The literature search supporting this document was
completed in March 1988.

-------
                           AUTHORS AND REVIEWERS


      The Human Health Assessment Group of the Office of Health and Environmental
Assessment (OHEA) prepared this document.


AUTHORS
Apama M. Koppikar, M.D., Ph.D., D.P.H., D.I.H.
Robert E. McGaughy, Ph.D.
Lorenz Rhomberg, Ph.D.


PROJECT MANAGER
Robert E. McGaughy, Ph.D.
REVIEWERS

      The following individuals reviewed this document and/or earlier drafts

of this document:
Elmer Akin

Elizabeth A. Cotsworth

William G. Ewald


William R. Hartley

Robert Kainz



William E. Pepelko


Lawrence R. Valcovic


Richard Walentowicz


David Warshawsky
Region 4, U.S. EPA, Atlanta, GA

Office of Solid Waste, U.S. EPA, Washington, DC

Environmental Criteria and Assessment Office
OHEA, U.S. EPA, Research Triangle Park, NC

Office of Drinking Water, U.S. EPA, Washington, DC

Human Health Assessment Group, OHEA
U.S. EPA, Washington, DC
(on detail from the U.S. Army)
    %4
Human Health Assessment Group, OHEA
U.S. EPA, Washington, DC

Human Health Assessment Group, OHEA
U.S. EPA, Washington, DC
        i
Exposure Assessment Group, OHEA
U.S. EPA, Washington, DC

University
-------
                                   1.  INTRODUCTION
       The U.S. Environmental Protection Agency (EPA) is currently writing permits for the
operation of incineration facilities now being designed to dispose of the U.S. Army inventory of
blistering agents (sulfur mustard gas and Lewisite) and  nerve agents (VX, GA [Tabun], and
GB [Sarin]). Table 1-1 describes these agents.  Congress has mandated (Public Law 99-145
[U.S.C. 1521]) that all unitary (self-contained) lethal and chemical munitions be destroyed by
1994.
       The U.S. Army has eight locations where these agents are stored (U.S. Army,  1988),
and current plans call for incineration of the remaining stock at each location with facilities to
be constructed at the sites between 1989 and 1991. The location of the  sites and the
population within a 20-km radius of each  site are listed in Table 1 -2. All of the agents and the
munitions are at least 19 years old; some are more than 40 years  old.  The duration of the
planned incineration  period is different at each site and  varies between 1  and 3 years.
       Of these five  agents, sulfur-mustard is the only agent currently recognized to be a
carcinogen, although there is some evidence that Lewisite may be carcinogenic (Centers for
Disease Control, 1988). The question of potential cancer risks for people exposed to sulfur
mustard needs to be addressed. This report derives an upper-bound risk estimate to provide
for all parties (the EPA, the U.S. Army, and the affected public) information about the
magnitude of the potential risk to incineration of  sulfur mustard. The scope of this report will
be limited to the routine operations of the continuous burning of mustard. Specifically, the
consequences of possible accidental release of mustard are not addressed. The
Environmental Impact Statement (U.S. Army, 1988) discusses accident scenarios and the
local site characteristics in detail.  In addition, no attempt is made to evaluate the occupational
risk to people who will be handling the material during incineration.
       The Aberdeen, Maryland/site is chosen as the example for all eight sites; it has the
largest neighboring resident population, and there are estimates of ambient air concentrations
of sulfur mustard at that site (U.S. Army, 1987).  In this  report, no attempt is made to evaluate
specific factors for other sites. The worst-case exposure to a person who might live at or near
the site fence line is  used in estimating the worst-case risk potential.

-------
 GC

 g
 to
             CO
CD
                   OT

                   '8.


                   CD


                   I
                   CO
                                CD
                                          CD
 w
                                          (D
                                         CO
                        *J
                         §? -o     J2>
                        .E  o>     -=
                         5?  «>     5-
                         QJ  JSt     flj
                        4—*  W     4—•
                        ,<2  Q-     .52
                        rs  ?s     3;
                                                                     CD
                                                                  O  .52
                                                                     0)
                                                                     CO
                                                       m  CD
                                                     m
                            X     :=
                            CD     CQ
    §
s-  W
o  .52
o>  ~
~  CD
m  CO
S  O
CO  Q.
 DC
                        o     o
                             O
                                                       o
              o
                                                                            O
                                                                            CO
Ul
                                                                            cu
v>
Ul
C5
Ul

O

Ul
            CD
CO
            o
                     CD


                               CO     55-
                               ^-  e     ,t
§•£  S
CO  CD  JO

"li
                                           o>
                                           •a
                                           sc
                                                       CO
                                                                            CO
                                         cvi
                                             •«  o
        LJJ
^i  Q.
6

                                          m
                                                                O
                                                     CO
                                                                                      °P
                                                                                      en
                                                                                      S
u.
o
                    CD


                    00
                  CO

                  4
                                         cr>
                                         CO
                              -
                             s
              CO
              uo
              g
                                   CD
                                   CD
                                  CO
                                      CO
                                      o
                                  CO

                                  in
          !
          I
Ul
Ul

m
            CD

            I
           o
                                                               to
                     co
                   co
                  CO
                                         a>
    o

    c?
                                                                         t
                                                                                      SL
                                                                                      CM
                                                                                      CQ
                                                                                           CO
            CD
            O)
                  m
                  o
                                                                         -    o
                                                                         I    I

-------
                   ooooo
                   oocooo
                               -
                                   o
                                   m
V)
LU

(3
O

s
LU

O


I
cc


CO
z>
u.
o
to
o
Q.
CO

Q

Q
LU
DC



I

LL.
O


LU
CM
TABLE
                   O O


                   § ?
                           T- to in
                           CM N-
                                      CM
           O


           1
           Q.
           O
           Q.
           c

           D)
X

CO X
(5 >_


X O

Q" Q"
                              m"
                                   CD

                                   CD
                                   I-
                   Q
                   x x
                              X X Q O

                              Q" Q" ^ Q"
                              X X X X
Location
                   c

                    1
                   T3 C  .=

                      83 -8  o
                      1 1 .1 1?
                      1 * E <  ^ s °.
                      c- ? 5 f  8 s I
                              =
           .1
           CO
                                            h
                                            CQ

                                            §,
                                            §>
                                            2
                                           6
Studen
                                            o>
                                            £
                                           00

                                           .<2



                                           t
                                           5   CO
                                            _   00
                                            E   o>


                                           1   i
                                           in   g

                                           •5
                                           2?
                                           1

                                            1
                                            c

                                            CD
                                                w
                                            .2   UJ
                                           f

-------
     This evaluation will follow the format of the EPA Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 1986). The hazard identification section is brief, since the
International Agency for Research on Cancer (IARC, 1975) has already characterized sulfur
mustard as a known human carcinogen (Category 1), and no subsequently published
information contradicts that conclusion. For the dose-response assessment, several
approaches are examined and compared.  The human exposure estimate is based on
projected emission rates and estimated ambient air concentrations adjoining the Aberdeen
Proving Ground site.  The risk characterization  section summarizes the results, discusses
uncertainties in these estimates, and compares the results with other published risk
                                     i
assessments.                         !

-------
                             2.  HAZARD IDENTIFICATION

       In this section, no attempt will be made to describe the toxic responses that occur
within 24 hours of exposure to mustard gas, the noncancer health effects that have been
reported after prolonged exposure in manufacturing facilities, the vapor concentrations at
which these effects occur, or the physical properties of this agent which make it so effective in
battlefield situations.  These are discussed in several studies (Prentice, 1937; Gilchrist and
Matz, 1933; Alexander, 1947; Morgenstern et al., 1947; National Research Council [NRC],
1984).
       Since this report focuses on the carcinogenicity of the agent, a brief description of the
evidence leading to this conclusion will be given. The subject has been summarized well by
the I ARC (1975) and by the NRC (1984).

2.1. HUMAN STUDIES
       Four epidemiologic studies on Allied World War I soldiers have been conducted (Table
2-1), the first (Gilchrist and Matz, 1933) from a British clinic, the second a study of British
retired veterans (Case and Lea, 1955),  and two studies of United States soldiers (Norman,
1975; Beebe, 1960).
       No quantitative exposure analysis of these battlefield exposures has appeared in the
literature available to us, except a statement by Prentice (1937) that "about 12,000 tons of
sulfur mustard were used in the war, causing about 400,000 casualties (one per 60 pounds of
gas). Its persistence in  the field is long: 1 day in the open, 1 week in the woods in summer,
and several weeks in  winter.  It is soluble in gums (i.e., rubber) and will penetrate cloth and
leather and rubber boots.  The latent time for the appearance of toxic effects is several hours,
so that exposures go  unnoticed."  Prentice defines a casualty in military terms as a person
who is so incapacitated  that he can  no  longer be an effective  soldier.
       Gilchrist and Matz (1933) described the symptoms seen in the clinic after battlefield
exposures.  They described 27 incidents of bronchitis and pneumonia but did not mention lung
cancer.  Case and Lea (1955) described bronchitis  and lung cancer among a group of 1,267
British retired veterans of World War I who were alive in 1930 and had sulfur mustard

-------
                                                                                        0

                                                                                        c?
 C5
 CO
    8

   i
   DC
                                  in
                                  m
                                  o>


                                   CO
                                   CD
                                   CD
                                  o
                                               §
                                               CD
                          O
                          ^2
                          CD
                          CD
                          m
                                                 m
                                                             CO
                                                                        co
                                                                             CD

                                                                             o
                                                                                        CD
                                                                  8
 Q


 I
 DC
CO

U.
   w


   (U
   O
                         To
                      »
                      O
 2  8  g
 CD  O  -O
                                                             CD
                                                   O     D)
                                                   Q.     P
                                                   X
                                                   CD  CO h-


                                                  IJ3
                                                  'co T3 in
                                                      c P
                                                             CD
                                                                 8
CO
UJ
                         CO
                                               O)
CO
U.
DC
   s
xl
                     •s
              (U
GO  JT  co
~  S  1
   1  S.
                      8
                                  O O)
                                  eo CM
                                  O) u.

                                     8
                     CM
                                                  CO
                         —  C
                          CD  Q
                          CO  O
                          CO  i_
                          CD  ffi


                          I  *



                         1  1

                         co  8
                                      15
                                                 8
                                                           O)
                                                               co
                                                 o  o
                                                Z  E
CO
CM

HI

03
                       -  §
Q)  X
•—  ^)

co  o
CQ Z
                                  CO
                                     CO
                       llfi
                           03  =
                                        qj
                             £
                          -•  5
                                              CQ
                                                 -  §
                                                "O  CO
                                                "
                         CD  O  ^     CD  O  ^J
                        «c  S" co    •*"  9~  to

                        ^  CD  E    ^  CD  E
                                                  o
                                          CO
                                          CD
                                                                  CD
                                                  0>

             CO
             CL
                         O
                         CO
                        m
                                        J2
                                        (D
                                     w  'co
                                     CQ
                   8.
                                                                 CO
                                                                        co
                                                      •=     o
                                                      S     Q.
                                                                        o>  o
                                                  CO
                                                  CL
                                                  CO
                                                                CO
                                                                CO
                                                                o>

-------











f
"c
o
^
ai
_j
m
£













8
w
1
"CD
DC


CO
C
CD
O
O


aeo
JZ 0>
±± c
II


^ c
CO
1!


.0
Q.
O
a.


m
CD
m~%}

a> _.
JZ- CO
Yamada
Wada et






8
tf c
•f 8
iD c
iT^

U5
O>
^«>
i
C3>
1
*o
if
C s—
CO en
o-co


00
?

.
15
CO

_eo
1
c
CO CO
OJ 1L Jl
2 CM

C
— oo
O ^J
goo-
8 o)^*
8 ^ *=
eo c
||8
^j QJ *^^

m
o>
^•^
o>
1
"?5
I!
CO rn
Q-co
^ 0)



15

*-*
o
s
•is
.00 O>
Z 31^

8 ,
T3 <0 CD
C C3 CD
CO _ CO
15 CD 0
CD CT.S2
CO C CO
E,^CD
J5g
-J 0.5


Lung cancer: 79/26
Higher incidence with
longer exposure

m
O>

C3>
1
"rt
if
c *^
CO (/)
&s


_^
CO
•I—*
CD

£
0
CO
CD f^-
O C3>
2 C-


8
8-
0 CD
0)C
C O
— S
O CD
•z. E


10 case studies
chronic bronchitis
*-*
CO C
c E
° cS*
CD "°"
CO fl>
'eo"

15
S
{2
CO
CO
Z)
u?
C3>
CO
.to

^
•a
CO
CO
1

o
'•CO
CO
1
.CO
"5)
LU

/^
Bronchial carcinoma
was significant.
Bladder carcinoma ar

if>
O)
"^
o>
T—
1
m
German ars<














leukemia also seen









»-*.
$
^
IB
»w
03
0)
1


Q.
|i


T3
CO
X
Is
"5 55°
si
C o

in
o>
1—
o>
T—
C
O
.22 >>
i!
eo si:
|S &














Laryngeal cancer
incidence is 7/0.75













^~
00
O>
JZ
CD
S
O)
T3
CO
c
eo
+3
CD
is
< O>

CD
f
T3
CO
CO
11
800
CD
•— .£

•r-
o>
1—
i
T-


Dismantling
facilities in
Germany















E
c
"S
"Q
0)
•5
.£2
^Observed nurr

-------
  polsonfng.  Over 80% of them had bronchitis, and 29 had lung and pleural cancer, compared
  with 14 expected cases.
        A slight excess of lung cancer was noted in U.S. soldiers who served in World War I
  (Beebe, 1960), but a study by Norman (1975) of the same population after 10 additional years
  of follow-up (through 1965), did not report any lung cancer excess.
        In a series of papers describing cancer incidence in former workers of a Hiroshima
 factory manufacturing mustard and other poison gases from  1929 to 1945, Nakamura (1956),
 Wada et al. (1968), and Nishimoto et al. (1983) reported incidences of respiratory tract
 cancers. In a population of 2,620 registered as factory workers, 495 of whom reported that
 they manufactured mustard gas, Wada et al. (1968) observed 33 respiratory cancer deaths,
 compared with 0.9 expected in the male Hiroshima population. The tumor sites were along
 the main airways, but only one was in the peripheral lung. The exposure conditions for these
 workers are discussed below.
       In their study of male workers registered in  1952 at the poison-gas factory of
 Hiroshima, Nishimoto et al. (1983) found that out of 2,068 registered individuals, 761 worked
 In the poison-gas manufacturing process (Group A), 705 worked in an occupation related to
 these poison gases (Group B), and 602 worked in unrelated jobs (Group C).  By the end of
 1979, 251, 233, and 124 persons in groups^, B, and C had  died. There were 79 lung
 cancers, 17 laryngeal cancers, and 6 pharyngeal cancers in the total cohort. Groups A and B
 had three times higher incidences of lung cancer than did the Hiroshima male population
 (p < 0.01), while Group C individuals did not show any statistical difference. When the work
 history was considered, the longer the  individual worked, the  more significant was this
 difference, indirectly showing a dose-response relationship. The most frequently found lesion
 among workers was cancer of the stomach (92),  but the difference was not statistically
 significant when compared with the Hiroshima male population.
       Nakamura (1956) investigated the general conditions of the poison-gas factory of
 Hiroshima during its operation from 1929 to 1945 (refer to Yamada [1963] and NRC [1984] for
 a description in English).  He found that the work room was filled with poison gases, one of
 them  being sulfur mustard. The concentration was assumed to be 0.05 to 0.07 mg/L, as
 determined by a bioassay in which unprotected birds in the work area and exposed rabbits
 died from exposures (NRC, 1984). He reported that about 3,000 to 5,000 people worked in
the factory at a time, depending upon the needs of the production. Workers worked a 10- to

                                          8

-------
12-hour shift. One hour of work in poison-gas manufacturing was alternated with 2 hours of
work in nonpoison-gas areas.  Workers were also provided with "gas-protect" masks and "gas-
protect" rubber clothes. These devices had not been supplied in sufficient numbers. Workers
sometimes neglected to wear the protective gear and, worse still, the "gas-protect" rubber
clothes were sometimes penetrated by the vapor, resulting in stickiness of the skin.  Hence,
skin contact occurred in addition to inhalation.  The actual degree of protection afforded by
these devices is not known.
      Although several epidemiologic studies of the workers from this factory have been
conducted, the actual number of people who worked in this factory is not known because
there are no employment records. The studies have been carried out with persons who could
be identified  as former poison-gas factory workers.
       Due to irregular use of protective equipment, sporadic exposure (4 hours per day at
the maximum), the assumed presence of other poison gases in the work room  in addition to
sulfur mustard, and the lack of complete information on the number of employees, the
Japanese poison-gas factory data are not suitable for quantitative estimation.
       One study of workers in United States sulfur mustard factories (Morgenstern et al.,
1947) reported incidences of chronic bronchitis after 3 to 6 months of employment, but there
was no mention of lung cancer.
      A report of 245 workers exposed to sulfur mustard from 1935 to 1945 in a German
factory and followed from 1951 to 1974 (Weiss and Weiss, 1975) described statistically
significant increases in  malignant tumors. Incidences of bronchial carcinoma were significantly
higher in the exposed group than in the general population near the factory.  Carcinoma of the
bladder and leukemia had Standardized Mortality Ratios (SMRs) of 259 and 283, and
digestive tract tumors were slightly elevated (SMR = 135), but these incidences were not
statistically significant.  Only three cases each of  bladder cancers  and leukemia were
reported, so that the statistical confidence interval is large. The average time from the onset
of exposure to death from malignant tumors was  21.9 years.
       Manning et al. (1981) traced 428 (84%) of the 51.1 individuals who had worked in a
factory manufacturing sulfur mustard between 1940 and 1945 near Liverpool, United Kingdom.
The follow-up period was 29 years (from January 1946 to December 1974). Of the 511
persons, 9 were omitted from the cohort (they had died before 1946), 181 were dead, 6 had
emigrated, 232 were alive, and 83 were untraced. Although there was no statistically

-------
 significant increase in the total number of deaths from all neoplasms, a statistically significant
 excess was observed for incidences of cancer of the larynx and trachea (grouped together by
 the authors). The SMR was 750 (observed - 3, p < 0.02) when compared to the United
 Kingdom's general population. Seven people had developed cancer of the larynx, compared
 to an expected incidence of 0.75 in the general population (p < 0.001).
       Klehr (1984) reported that, betweeni1945 and 1951, during dismantling of the "Heeres-
 Munitlonsanstalt St. Georgen," inadequate personal protection was used. This dismantling
 resulted in contamination by and inhalation iof the  poison gases (including sulfur mustard) by
 about 400 persons. Surviving persons (number not given) had multiple skin tumors, such as
 basal cell carcinomas, Bowen's disease,  Bowen's carcinoma, and even carcinoma
 spinocellular in unexposed parts of the skin.
                                       i
       In summary, sulfur mustard has been associated with lung cancer in U.S. World War I
 veterans several years after battlefield exposure and in Japanese, German, and English
 factory workers occupationally exposed over long  periods of time.  In view of these results, the
 International Agency for Research on Cancer (IARC,  1987) classified the human evidence of
 carcinogenicity as "sufficient." The authors of this report concur with the IARC evaluation.

 2.2.  ANIMAL STUDIES
                                       I
       In the early 1950s, Hestoh published three reports of experiments in strain A  mice
 administered sulfur mustard via the intravenous (Heston, 1950), inhalation  (Heston, 1953a),
 and subcutaneous (Heston, 1953b) routes.  He observed a lung adenoma  response  to
 administration in all three studies.  The intravenous administration experiment is described in
 more detail later in this report.
       Haddow (1959) briefly described his! experiments in rats showing the induction of
tumors by sulfur mustard. However, no detailed account of his experiments could be found.
      The experiments reported in McNamara et al.  (1975) were conducted in 1970 and do
not follow current standards of good laboratory practice, quality control, or reporting.
Ordinarily, such data would not be used for quantitative risk assessment, but the shortcomings
are tolerated here out of necessity.
                                          10

-------
       McNamara et al. (1975) exposed a group of 70 male and 70 female Sprague-Dawley-
Wistar rats to vapor from distilled mustard (purity not specified) at a continuous nominal
concentration of 0.001 mg/m3. Another group of 70 males and 70 females was exposed at
0.1 mg/m3 for 6.5 hours per day, plus 0.0025 mg/m3 for the remaining 17.5 hours per day.
Fifty rats of each sex, housed in the animal colony facilities, were used as controls.  The
methods for maintaining air concentrations and the success with which the nominal
concentrations were achieved were not reported.  The two concentration regimes were not run
simultaneously, and the low concentration was run in a  chamber without temperature controls.
Dogs, rabbits, guinea pigs, and strain A/J mice were concurrently exposed in the same
chambers.
       No tumors were observed after exposure as long as  52 weeks for 10 dogs, 18 male
rabbits, 6 female rabbits, 21 male guinea pigs, and 21 female guinea pigs.  In A/J mice, 70
males and  70 females were observed for varying exposure durations up to 12 months and at
3 and 6 months after a 12-month exposure.  Pulmonary adenomas not related to
administration were observed.
       Rats were exposed to sulfur mustard in a "toxicity" experiment and a "carcinogenicity"
experiment (McNamara et al., 1975).  The toxicity experiment is closer to a long-term
carcinogenicity bioassay, since rats were exposed for up to 52 weeks and then followed for up
to 6 months. The deaths of all of the  animals in the  experiment were accounted for by a
series of planned killings (i.e., no natural deaths were included).  Killed animals were
examined for various manifestations of toxicity, and any tumors detected at necropsy were
reported. The carcinogenicity study differed in that animals were withdrawn from exposure at
various times, many of  them after quite brief exposures, and then retained for various periods
ranging from 13 to 20 months before being killed.  Aside from any tumors, no  manifestations
of toxicity were reported for these animals.  As with the  toxicity study, no results from animals
dying naturally were reported.
       The same initial pool of animals contributed subjects to both studies, and there was no
apparent pre-assignment of individuals to one or the other experiment. The reported results
for the  two  experiments together do not account for all of the 140 rats per dose (or all of the
100 controls).  Evidently, some rats were taken for other purposes not reported.  McNamara et
al. (1975) show the schedule of animals killed in their Tables A-10 through A-12.  These
tables indicate that some animals were killed very early  (before 2 months on study), and that
                                          11

-------
a few natural deaths occurred.  No results are reported for these animals.  In a telephone
conversation in early 1990, one of the co-authors of the McNamara et al. (1975) report, F.J.
VoccI, indicated that assignment of animals to one or another study or to one or another
group to be killed could have been influenced by the animal's condition —  moribund animals
could have been assigned to the toxicity study and  killed, or sick animals could have been
withdrawn from exposure and entered into the long  postexposure groups in the carcinogenicity
study.  These procedures may account for the paucity of natural deaths, even among rats
kept up to 30 months.
       In both the toxicity and carcinogenicity studies, McNamara et al. observed basal and
squamous cell carcinomas of the skin in the high exposure group.  The data are presented in
Tables 2-2 and 2-3, and the results are described in detail in section 2.1.
           TABLE 2-2.  INCIDENCES OF SKIN CARCINOMAS AMONG RATS IN THE
                            MCNAMARA ET AL. TOXICITY STUDY"
Exposure Postexposure
duration (days)
(months)

2
3
4
6
8
12
12 70
12 90
12 180
[•
Incidence of skin carcinomas
Control
0.001
mg/m3
M

0/5

0/5
0/5
0/5

0/4
0/7
F !
0/5!
0/5|

0/5
0/5
0/5|
i
0/4*
0/4b'e
M
0/5
0/5

0/5
0/5
0/5

0/4
0/6
F
0/5
0/5
0/5
0/5
0/5
0/5b

0/5
0/14'
0.1/0.0025
mg/m3
M
0/5
0/5
0/5

0/5
0/5
4/4°
0/1
0/6
F

0/5


0/5
0/5

0/5









5/13c.g,h
 Only types b and f were considered by the authors to be related to the administration of the agent and only these
 types are counted In the numerators of this table.
 One subcutaneous fibroma                   ;
°Four squamous cell carcinomas of the skin
"One squamous cell carcinoma of the uterus
'One pulmonary adenoma
"One papilloma of the skin
"One basal cell carcinoma of the skin            |
"One thyroid adenoma
SOURCE:  Adapted from Tables 4, A-30, and A-31 of McNamara et al., 1975.

                                          12

-------
       TABLE 2-3.  INCIDENCES OF SKIN CARCINOMAS AMONG RATS IN THE
          McNAMARA ET AL. CARCINOGENICITY STUDY (SEXES POOLED)1-2
Exposure
duration
(weeks)
Postexposure
 (months)
                                      Incidence of skin carcinomas
                              Control
0.001
mg/m3
0.1/0.0025
  mg/m3
 2

 4


 8



1.2


26


393

52
     13
     15
     21

     20

     16
     20

     15
     17
     18

     12
     17

     14
     18
      2
      4
      6
      7
     10
     17
     18
                                 0/4
  0/1

  0/4"

  0/5

  0/1
  0/4

  0/2
  o/i
  0/1d

  0/2
  0/3e

  0/4
                                    11
                                0/22e
                                 0/1e
                                                   0/17
   0/1a
   0/4

   0/5°

   0/1
   0/5

   0/4
   4/53f,g


   3/43


   0/3e4/44f>h
                                                                       3f
                    1/1h
                    1/1'
                    0/1
                    3/143e.2M

                    0/1e
                    4/4f
Superscripts indicate the number and types of tumors:
                                  skin, squamous cell carcinoma
                                  skin, basal cell carcinoma
                               h   skin, trichoepithelioma
                               i   skin, keratoacanthoma
  a  subcutaneous lipoma          f
  b  axillary lipoma               g
  c  subcutaneous fibroma
  d  astrocytoma
  e  skin, fibroma
2Only types f, g, h, and i were considered by the authors to be related to the administration of
 the agent and only these types are counted in the numerators of this table.
3!n the 39-week exposure group in the high-concentration group, a type h tumor co-occurred in
 one of the animals with a squamous cell carcinoma.

SOURCE: Adapted  from Table 5 of McNamara et al., 1975.
                                            13

-------
       In summary, sulfur mustard has induced lung adenomas in strain A mice via
subcutaneous and intravenous injection and via inhalation exposure in a series of experiments
reported In the early 1950s. Considering the data alone, the IARC (1982) classified the
animal evidence as "limited."  Since the original IARC evaluation (IARC, 1975), McNamara,
(1975) reported finding skin carcinomas in the high-dose group of rats exposed at vapor
concentrations of 0.1 mg/m3 but none in the group exposed at a vapor concentration that was
100 times smaller. The overall evidence of carcinogenia'ty in animals is now judged to be
                                       I
only "limited," since carcinomas were induced in only one experiment and the three
experiments in strain A mice are not standardized bioassays.  However, it is strong "limited"
evidence, since benign and malignant tumors have been induced by four different routes of
exposure in two independent laboratories in two species.

2.3. SUPPORTING EVIDENCE

       Several studies have reported that sulfur mustard causes gene mutations.  Currently,
the U.S. Army is conducting mutage.nesis tests of sulfur mustard in bacteria and mammalian
cell cultures and a dominant-lethal mutation test in mice. Rather than reviewing the
mutagenesis evidence in detail, the following excerpt from an  NRC (1984) review is quoted
here to give some idea of the types of results that have been  reported.  The references from
this quotation are not cited in this document.
       Among the first demonstrations that chemicals can induce mutations was a report by
       Auerbach and Robson (1946) that mustard gas induces mutations in Drosophila.  Over
       several years, Auerbach and her colleagues found that mustard gas causes genetic
       alterations ranging from gene mutations to chromosomal breaks and rearrangements
       (Auerbach, 1976; Fox and Scott, 1980).
                                       I
       The mechanism of mutagenesis by sulfur mustard (and other mustards) involves the
       alkylation of DMA. As a bifunctional alkylating agent, sulfur mustard causes cross-
       linkage of DNA strands, as well as monofunctional alkylation products (Fox and
       Scott, 1980). Sulfur mustard and nitrogen mustard have been used in mutation
       studies in a variety of organisms, bilt data on the relative frequencies of
       induction of different alkylation products in DNA by the two agents are limited.
       Nevertheless, sulfur mustard seems to exhibit greater SN1 character (Morrison
       and Boyd, 1966) as an alkylating agent than does nitrogen mustard (Fox and
       Scott, 1980). Because agents involved in SN1 reactions attack oxygen sites in
       DNA more readily than do agents whose reactions are almost exclusively of the
                                         14

-------
       SN2 type, sulfur mustard may differ substantially from nitrogen mustard in the
       spectrum of alkylation products formed in DNA. Such differences in
       mechanisms of alkylation and in alkylation products can lead to considerable
       differences in mutagenicity (Hoffman, 1980; Vogel and Natarajan, 1982).
       Therefore, although nitrogen mustard and sulfur mustard are both alkylating
       agents, one must be cautious in assuming they are comparably mutagenic.

       A comprehensive review of the mutagenicity of sulfur mustard and nitrogen
       mustard has been published by Fox  and Scott, 1980.  The mutagenicity data
       base on nitrogen mustard is more extensive than that on sulfur mustard.
       Nevertheless, results have been reported regarding the genetic activity of sulfur
       mustard in tests for forward mutations and reversions in bacteria (Capizzi et  al.,
       1974); differential killing of DNA-repair-deficient strains of microorganisms and
       their repair-proficient counterparts (Ichinotsubo et al., 1977; Kircher et al.,
       1979); reversions in fungi (Kircher et al., 1979; Stevens et al., 1950);
       chlorophyll mutations in vascular  plants  (Fox and Scott, 1980); gene mutations
       and chromosomal aberrations in Drosophila (Auerbach, 1976; Fox and Scott,
       1980); reversions in cultured L5178Y mouse lymphoma cells (Capizzi  et al.,
       1974; Fox and Scott, 1980); host-mediated assays involving bacteria or
       mammalian cells in mice (Capizzi et  al., 1974); and dominant lethal mutations
       in plant root tips and microsporocytes (Fox and Scott, 1980) and in cultured
       mammalian cells (Savage and Woods, 1981).

       Data from a mouse dominant-lethal test suggest that sulfur mustard reaches
       germinal tissue and induces dominant lethal  mutations (Rozmiarek et al.,
       1973). However, the data are inadequate for prediction of genetic risk to
       human germ cells. Uncertainties stem from the lack of data on defined genetic
       events induced by sulfur mustard in mammalian spermatogbnia or oocytes and
       from the variation in mutagenic potency  that has been observed for the
       mustards in various assay systems.  Nevertheless, the possibility that sulfur
       mustard is a human germ cell mutagen  cannot be disregarded, particularly
       because it is mutagenic in diverse assays, including tests for germ cell
       mutations in Drosophila and dominant lethal  mutations in mice; moreover, other
       direct-acting alkylating agents are known to induce mutations (Russell et al.,
       1981) and chromosomal alterations (Generoso et al., 1980) in mammalian
       germ cells.
       In summary, there is strong evidence that sulfur mustard is a genotoxic agent.  It is a

DNA alkylating agent and it is mutagenic in mouse lymphoma cells, bacteria, and fungi,  In

Drosophila, it induces aneuploidy, heritable translocations, dominant lethal mutations, and sex-
linked recessive mutations.
                                          15

-------
                           3.  DOSE-RESPONSE EVALUATION
 3.1.  HUMAN STUDIES
        Several case reports and cohort studies have shown that mustard gas is carcinogenic
 In humans, but the data are inadequate for deriving dose-response relationships. One attempt
 has been made to use this information.
                                       i
        Information on the World War I battlefield exposures poses two problems.  First, there
 is no way to reconstruct believable quantitative estimates of battlefield exposures.  One
 possible solution to this problem is to assume that the soldiers who survived to be veterans
 were exposed to no more than the lethal dose, for which there are estimates (Prentice, 1937).
 This assumption would lead to a lower bound on the potency of mustard (which is the risk per
 unit dose).  The second problem is that no exposed battlefield cohort with a corresponding
 nonexposed population has been defined and followed to determine mortality in an
 epidemioiogic study.
        For the  Japanese factory workers, the starting point for the  definition of the cohort
 studied was the people who were registered with the government as poison-gas workers,
 residents of the area in which the factory operated, and friends and relatives of former
 employees.  This method of cohort selection has inherent biases that will over-represent those
 with health problems resulting from employment.  Another problem  is that workers were
 exposed to other toxic agents along with  sulfur mustard.  There are no employment records
 for this factory,  and the actual number of  people who worked there  is unknown. There  was
 Inconsistent use of protective equipment and sporadic exposure (maximum of 4 hours per
 day), making any exposure evaluation a matter of guesswork.
       For the  U.S. manufacturing operations, only one study is available (Morgenstern et al.,
 1947), and it describes case reports only. The German arsenal study (Weiss and Weiss,
 1975) has only  limited usefulness for quantitative risk estimates because no exposure
 estimates are available.
      The report by Klehr (1984) of accidents in the handling of poison-gas reserves after
World War II  in  Germany also has no defined worker population, and exposure estimates
cannot be made. Because of the paucity of information needed to evaluate risks
quantitatively, this report will not attempt to evaluate human data.

                                          16

-------
       Rosenblatt (1987) has made a quantitative assessment of the data of Wada et al.
(1968).  This is one of the series of papers described in section 2.1 . which reports the
incidence of lung cancer in Japanese factory workers. Rosenblatt obtains a slope factor of
0.16 per (mg/kg)/day which he increases by a safety factor of 1 0 to get 1.6 per (mg/kg)/day.
The corresponding air unit risk is 1 .6 per (mg/kg)/day x 0.286 x 1 0"3 (mg/kg)/day per ng/m3
4.6x lO^per ng/rh3.
3.2.  ANIMAL STUDIES

3.2.1.  McNamara et al. (1975) Rat Studies

       These studies were described in section 2.2.  The unusual experimental design makes
quantitative analysis difficult. Marked bias in assignment of animals for various exposures is
possible and  perhaps even likely.  Many of the exposures in the carcinogenicity study are very
brief (as short as one week), and there were over 30 distinct exposure and follow-up regimens
employed, most with only a few animals each. Although exposure patterns were more regular
in the toxicity study, most animals were killed well before their capacity to develop late-
appearing tumors was fully tested.  Despite the shortcomings, the McNamara et al. data are
the best available for use in estimating the carcinogenic potency of mustard vapor.  In the
analysis that  follows, the two data sets (from the toxicity and carcinogenicity experiments) are
taken at face value; the problems associated with lack of random allocation of animals to
experimental  groups are ignored.

3.2.1.1. Toxicity Experiment
       The tumor incidences from the toxicity experiment (as reported in McNamara et al.
[1975], Tables 4, A-30, and A-31) are shown in Table 2-2. Only those tumors identified by
McNamara et al. as being agent-related are included; these tumors comprise eight squamous
cell carcinomas of the skin and one basal ceil carcinoma of the skin, all appearing in rats
administered sulfur mustard at high concentrations.  One could argue for the further inclusion
of a skin papilloma appearing in a rat administered the lower concentration, but exclusion of
this tumor makes little difference in the resulting estimates.  Excluded were two subcutaneous
                                          17

-------
 fibromas, a uterine squamous cell carcinoma, a pulmonary adenoma, and a thyroid adenoma,
 which were scattered among dose groups. Superscripts in Table 2-2 indicate the groups in
 which these tumors appeared.
       The striking pattern in Table 2-2 is that tumors appear only in the high-dose group and
 only late in the experiment.  Most of the animals were killed rather early, before the
 appearance of the first tumor (which occurred in an animal kept 70 days after the termination
 of 52 weeks of exposure).  Some of the animals that were killed may have developed tumors
 had they been retained on test for a full lifetime; the lack of response among them cannot be
 given full weight in the calculation of final tumor incidence. One solution to this problem is to
 eliminate from consideration all animals dying (or, in this case, killed) before the appearance
 of the first tumor of interest. That is, only those animals tested long enough to exceed the
 demonstrated minimum latency period are included.  In the present case, this approach is
 rather rigid; only 77 of the original 222 rats remain in the analysis. The approach is also
 approximate; some rats killed just before the  appearance of the first tumor are probably
 potential responders and are unfairly discounted from the denominator, while other rats killed
 just after the appearance of the first .tumor (and hence counted as nonresponders) may have
 developed tumors if they had been kept on test a little longer. Table 3-1 shows the lifetime
 tumor incidences with the denominators corrected to  include only animals surviving past the
 appearance of the first tumor.  Because of the low sample sizes, and because there is no
 apparent difference between the sexes, it is preferable to pool the data for males and females
 into a single set of incidence data.
   TABLE 3-1.  INCIDENCES OF SKIN CARCINOMAS AMONG RATS SURVIVING PAST
 THE DISCOVERY OF THE FIRST TUMOR IN THE McNAMARA ET AL. TOXICITY STUDY

Males
Females
Sexes Pooled
Control
0/11
0/8
0/19
0.001 mg/m3
0/10
0/19
0/29
0.1/0.0025 mg/m3
4/11
5/18
9/29
SOURCE Adapted from Tables 4, A-30, and A-31 of McNamara at a!., 1975.
                                         18

-------
       Exposure was terminated after 52 weeks.  Since the majority of the rats still included in
the analysis were followed for an additional 180 days (6 months), a reasonable exposure
measure is the daily average concentration over 18 months, or two-thirds of the concentration
during the 12-month exposure period. This approach assumes that the effect of a varying
dose rate is the same as its long-term average; i.e., the appropriate measure of total exposure
is the total concentration-time product (mg/m3 x days). For the low-exposure group, 0.001
mg/m3 for 12 months followed by 6 months of no exposure averages out to 0.00067 mg/m3, or
more conveniently, 0.67 jig/m3.  For the high-exposure group, the daily average during the
exposure period is the time-weighted average of 6.5 hours at 0.1 mg/m3 and 17.5 hours at
0.0025 mg/m3, or 29 ng/m3, and the 18-month average is two-thirds of this, or 19.3 ng/m3.
       These exposure measures  and the incidences of tumors pooled across sexes were
analyzed via the GLOBAL 86 computer program (Howe et al., 1986) to estimate a multistage
model dose-response curve. The  linearized upper bound on the low-dose slope of the
estimated curve, q*, is 2.9 x 10"2 per ng/m3 of lifetime exposure.  (Separate analysis
of each sex alone produces slightly higher numbers, owing to the effect of the low sample
sizes on the calculation of the upper bound.)  The above potency estimate is the result of an
experiment in which animals were  observed for about 18 months (the actual average is just
under 17 months). If rats are assumed to live 24 months, the potency estimate should be
corrected to reflect additional tumors that would be expected to appear after 18 months, had
the experiment been conducted for a fulllifetime.  This correction can be made by multiplying
the potency estimated from the experiment by the ratio of lifetime to experimental duration,
raised to the third power. This correction, which amounts to (24/18)3 = 2.37 in this case,
reflects the tendency of age-specific tumor incidence rates to rise as a power of age.  The
power of three is somewhat arbitrary but is conventionally considered typical.  (Further
analysis, shown below,  supports this choice.) The correction results in an estimated unit risk
of 6.8 x 10'* per ng/m3.
       Since mustard vapor is a well-absorbed, reactive, direct-acting agent, the effective
dose for producing skin tumors is considered to be the air concentration in contact with the
skin.  The simplifying assumption is made that fully clothed humans exposed to sulfur mustard
at a certain concentration have the same skin tumor response as rats exposed at the same
concentration in the exposure chamber.  Undoubtedly, inhalation exposure also occurred in
the experiment, but this would not  contribute to skin cancer response.
                                          19

-------
       This analysis makes crude corrections for the effect of observing different animals at
 different ages. Since the lifetime probability of a tumorigenic response to exposure develops
 over time, individuals examined early (because they were killed or died from another cause)
 will not have manifested their full probability of response. When data are available on the
 times of examination of individual animals, a more sophisticated analysis is  possible that
 estimates not only the shape of the dose-response function but also the shape of the time-
 response function, i.e., the  pattern of increase in tumor risk with age. In such a ''time-to-
 tumor'1 analysis, animals contribute information to tumor risk at the age at which they are
 examined, and not to later ages.  Thus, all animals can be included in the analysis, regardless
 of the time at which they were killed (although the risks at later ages continue to be estimated
 only from animals alive at those times).  Moreover, the effects of intercurrent mortality or
 killing of the animals can be corrected for; once the dose-time-response  relationship is
 estimated, the risks at any specific age, including a full lifetime, can be calculated, just as
 though all animals had reached that age.
       The data on tumor incidences at the specific times the animals were  killed, taken from
 Table 2-2, were used in the WEIBULL 82 computer program (Howe and Crump,  1983).  This
 program fits an equation of the form
                     P(d,t) - 1-exp[ (-qO-qid-q2d2-...-qndn) x (t-gk ],
                                        i
where P(d,t) is the probability of bearing a tumor after time t of continuous exposure at dose
rate d. The parameters estimated by maximum likelihood are the q's, the non-negative
coefficients of a polynomial in dose; t,, is an estimated latency time between tumor initiation
and its detection; and k is an estimated exponent of time. As with the case of GLOBAL 86,
the method assumes continuous dosing at a fixed rate, and the use of a time-averaged dose
rate Is an approximation. Empirically, the latency time (t,,) estimated from the Weibull 82
program is about 2 months, so the tumors appearing at the  end of the experiment were
induced somewhat earlier, when exposure was still occurring.  Thus, one can use the nominal
concentration during the 12 months of exposure (1.0 and 29.0 ng/m3) rather than the
exposures prorated over 18 months. (The choice among these expressions of exposure  leads
to only small differences in the final result.)
                                          20

-------
       In time-to-tumor analysis, one must specify whether the killing of animals (and hence
the search for tumors) is independent of the occurrence of tumors.  If so, the tumors are
discovered incidentally; i.e., they are taken from animals killed periodically, and the prevalence
of tumor-bearing at that time is estimated from the necropsy results obtained.  If, on the other
hand, animals were killed and necropsies were prompted by the presence of a tumor, the
pattern of occurrence dictates the pattern of discovery of tumors over time, and this must be
accounted for in the analysis. In the present case, the schedule for killing animals was
predetermined,  and it seems appropriate to treat the discovery of tumors as incidental.  Since
the tumors in question may be apparent in living animals, one cannot discount the possibility
that rats were killed because they were believed to be bearing tumors. Since tumors
appeared only near the end of the experiment, the effects of such bias are probably small and
are ignored here.
       An upper bound on the unit risk can be determined by calculating the linearized upper
bound on estimated risk at a single unit of dose and at an appropriate age.  At 18 months (the
duration corresponding to the post-first-tumor GLOBAL 86 analysis discussed  previously), the
upper-bound  risk from continuous exposure at 1 jag/m3 is 2.0 x 10"2, when estimated from the
time-to-tumor data (duration plus postexposure times in Table 2-2) on males and females
combined. (This agrees closely with the GLOBAL 86 result of 2.9 x 1O'2.) As with this
previous  analysis, one wishes to estimate the risk, not after the experimental period,  but after
a full lifetime  of exposure.  Rather than using a correction factor, the time-to-tumor analysis
can make this adjustment according to the rise in age-specific incidence estimated from the
actual bioassay data.  After a full lifetime of 24 .months, the time-to-tumor analysis estimates
the upper-bound risk as 8.5 x 10"2 per ng/m3. (This is also in close accord with the lifetime-
corrected GLOBAL result of 6.8 x 10"2.  The empirically derived power of time  is 2.68, which is
close to the assumed value of 3 used in the age correction of the previous analysis.)

3.2.1,2. Carcinogen/city Experiment
       The previous analysis used data from the so-called "toxicity study" of McNamara etal.
(1975).  These authors also conducted a "carcinogenicity study" in which rats were withdrawn
from exposure at various times and followed for up to 20 months before being killed and
necropsied.  Analyzing these data is difficult,  not because the animals were killed early but
                                          21

-------
because they were killed after brief and varying exposure times.  Several rats with longer
exposure times (and hence greater cumulative doses) developed skin carcinomas, and so the
carcinogenicity study should also be examined as a possible basis for potency calculation.
McNamara et al. (1975) present data on exposure patterns, follow-up times, and tumor
incidences in their Tables 6, A-38,  and A-39.  The data shown in Table 2-3 of this document
are abstracted from these sources. As before, the data for males and females are pooled,
owing to the low sample sizes.
       Table 2-3 shows that the experimental design included over 30 distinct patterns of
exposure and follow-up time.  Time-to-tumor methods that allow for the variety of
noncontinuous dosing patterns exist, but no current computer implementation can
accommodate so many patterns. In any case, the issue of animals being killed ©arly in the
study, which dominates the analysis of the toxicity study, is not so problematic in the present
case. Although some exposures were very short, all animals were followed for at least 13
                                       I
months after the beginning of  exposure, and rriost were kept for 20 months or longer before
being killed. The first tumors of  interest appeared in animals exposed for 12 weeks and
followed for 12 months before being killed.  Only  one control animal was killed earlier, which
was after 13 months. Although there was a range of times at which animals were killed
(between 13 and 30 months),  75% (99 of 132) of the animals were killed between 20 and 22
months following onset of exposure. Thus, virtually all of the animals in the study can be
considered to be on test for close to a full lifetime. The small differences in the ages when
they were killed can be ignored, and the data may be analyzed, using GLOBAL 86 and the
standard multistage model procedure, as a set of 17 lifetime incidences (two doses times
eight exposure durations plus  a control group) following various levels of dosing.
       The large number of short exposures raises two problems.  First, many of the 17  •
resulting total cumulative exposures are very small, and the lack of tumor response is not very
Informative.  Second, as noted previously, the multistage model assumes continuous dosing
throughout the experiment. Averaging short exposures over a lifetime puts great weight on
the assumption that only the total cumulative exposure, and not its pattern of delivery over
time, matters in comparing toxic effects. At the extreme, for example, we assume that one
week of exposure to mustard vapor according to  the high-concentration regime (0.1  mg/m3 for
6.5 hours per day and 0.0025 mg/m3 otherwise) early in life is equivalent in carcinogenic effect
to a continuous lifetime exposure to 0.000279 mg/m3, since the two share the same lifetime
                                          22

-------
average daily concentration. The rendering of the 17 exposures into their lifetime average
daily equivalents is shown in Table 3-2.
        TABLE 3-2.  LIFETIME AVERAGE DAILY EXPOSURES TO RATS IN THE
    McNAMARA ET AL. CARC1NOGENICITY STUDY AND ASSOCIATED INCIDENCES
                               OF SKIN CARCINOMAS
Exposure
duration
(weeks)
Control
1
2
4
8
12
26
1
39
52
2
4
8
12
26
39
52
Exposure
concen-
tration
....
Low
Low
Low
Low
Low
Low
High
Low
Low
High
High
High
High
High
High
High
Lifetime average
daily exposure
(ug/m3)
0.0
0.0096
0.0192
0.0385
0.0769
0.115
0.250
0.279
0.375
0.500
0.558
1.12
i
2.23
3.35
7.25
10.9
14.5
Incidence of
skin
carcinomas
0/27
0/5
0/5
0/5
0/4
0/5
0/4
0/5
0/3
0/17
0/5
0/6
0/4
4/5
4/5
4/4
10/23
Note:  Low-exposure concentration was 0.001 mg/m3, experienced continuously. High-exposure concentration was
      0.1 mg/m3 for 6.5 hours daily, and 0.0025 mg/m3 for the remaining 17.5 hours. A 2-year lifetime was
      assumed.

SOURCE: Adapted from Table 5 of McNamara et al., 1975.
                                        23

-------
       The linearized upper bound on the low-dose slope (q?), estimated from these data by
GLOBAL 86, is 9.4 x 1 0"2 per ng/m3.  This result agrees well with the results of the analyses of
the toxicity study.

3.2.1.3. Unit Risk
       The various analyses considered above are presented to ensure that the various ways
of examining the data give compatible results.  The most reliable estimate is probably the
timQ-to-tumor analysis of the data from the toxicity experiment.  The exposures were long-
term, the effect of killing animals before a full lifetime is explicitly accommodated, and the
sample sizes are the largest obtainable from the McNamara et al. (1 975) study.  It is
reassuring that separate analyses of males and females (where possible), use of a quantal
model  (GLOBAL 86) versus a time-to-tumor model (WEIBULL 82), and use of data from the
carcinogenicity study or the toxicity study did not alter the risk estimates greatly. The
recommended upper-bound unit risk for lifetime inhalation exposure  to mustard vapor is thus
8.5x10'2per ng/m3.
       This estimate of mustard vapor's carcinogenic potency should be regarded  as an
upper bound.  It is calculated from a 95% upper confidence limit on  the linear component of
the fitted dose-response curve; that is,  the curve is chosen that has  the largest linear
component at low doses and still retains enough curvature to adequately fit the data on tumor
incidences at higher doses. This practice reflects the difficulty of determining the true shape
of the dose-response curve at low doses, where expected incidences are too low to examine
experimentally. Depending on the unknown true shape of the dose-response curve at low
doses, actual risks may be anywhere from this upper bound down to zero.
       It may be noteworthy that malignant tumors appeared in the  McNamara et al. (1975)
experiments only at the higher concentration of mustard,  and then only late in life.  Perhaps it
may exert its carcinogenic activity secondarily through  lifelong exposure to its cytotoxic or
irritating effects.  In such a case, cancer risks from exposures low enough to avoid these toxic
effects may be much less than proportional to exposure.  Under such circumstances, human
exposures at low concentrations for limited times may entail much less risk than implied by a
unit risk factor estimated from lifetime effects at higher doses.  On the other hand,  the lack of
                                          24

-------
 low-dose responses and early-appearing tumors in the McNamara data may be due simply to
 the inherent difficulty of detecting low-risk levels in experiments of reasonable size.  It is
 therefore prudent to use the upper bound provided by the unit risk in estimating the potential
 harm to human health from low exposures to mustard vapor.

 3.2.2. Relative Potency in Strain A Mice
       This approach uses the strain A mouse lung adenoma response to intravenous (IV)
 injection of sulfur mustard as described by Heston (1950) in combination with the response in
 the same strain to the IV injection of 20:methylcholanthrene (MC) as described by Shimkin
 and McClelland (1949).  These two responses are used to derive a relative potency ratio of
 sulfur mustard compared to MC.
       MC was chosen for comparison for two reasons. First, it is a member of a well-studied
 class of carcinogens (the polyaromatic hydrocarbons) for which relative potency estimates are
 available. The U.S. EPA has derived a carcinogenic potency factor for benzo[a]pyrene (BaP),
 a commonly occurring member of the class, and that potency value can be used together with
 the relative potency of MC and BaP.  Second, MC was tested in the same strain of animals in
 the same time period by the same route of exposure as sulfur mustard, although in a different
 laboratory. Therefore, the test conditions are comparable in many respects.
       In the second step of this  approach, the Stoner et al. (1984) paper reporting the strain
 A mouse lung adenoma response to both  MC and BaP is used to derive a relative potency
 ratio of MC to BaP.  The product  of these  two ratios, which is the ratio of sulfur mustard to
 BaP potencies, is then multiplied  by the BaP oral and/or inhalation potency to derive an
 estimate for the sulfur mustard potency. The latter value is combined with the ambient
 concentration estimate to derive a risk estimate.
      There are  conceptual difficulties with  this approach. BaP and MC are both
 metabolically transformed to epoxides which bind to  DNA, whereas sulfur mustard is a direct
 alkylating agent.  Therefore, the potency ratio of sulfur mustard to MC might not be constant
 across species, routes of exposure, and doses. The approach specifically requires the
 assumptions that (1) the animals  in the two laboratories (Heston, 1950; Shimkin and
 McClelland, 1949) had equal sensitivities at the doses used, and (2) the ratio of responses to
the two compounds is invariant to a common scale change in their concentrations at all doses
                                          25

-------
from the experimentally used doses to negligibly small doses.  These assumptions are
probably valid for compounds with similar mechanisms of action (e.g., BaP and MC) but may
not be valid for sulfur mustard and MC.    ,
      The relative potency approach would have less uncertainty if comparison with a third
agent could be avoided. For example, Heston (1950) tested nitrogen mustard using the same
bioassay, but we do not have a human potency value for nitrogen mustard.  A somewhat less
desirable approach would be to find a strain A mouse study of BaP published near 1950, but
these data were not found.

3.2.2.1.  Heston (1950)

      Strain A mice approximately 2 months old received IV injections at 2-day intervals (for
a total of four injections) of 0.25 cc of a 1:1.0 dilution of a saturated (0.06% to 0.07%) solution
of sulfur mustard.  The total amount injected was 0.06 to 0.07 mg. A total of 24 males and 24
females were in the dosed group, and an equal number  were in the control group.  At 16
weeks after the first injection, the animals were killed and the lungs were examined for
tumors.
      The results are  shown in Table 3-3.  In the control group, 6 of the 46 animals had a
single tumor, for an average of 0.13 tumors per mouse, whereas in the dosed group, 32 of the
47 animals had a range of 1 to 3  tumors per mouse, for  a group average of 1.09 tumors per
                                      i
mouse.
   TABLE 3-3.  LUNG ADENOMAS IN STRAIN A MICE RECEIVING SULFUR MUSTARD
                       INTRAVENOUSLY IN FOUR INJECTIONS
Total dose
 (mg)
Fraction (percent) of
 mice with tumors
Tumors/mouse
(group mean)
 0
 0.065
  6/46 (13%)
  32/47 (68%)
  0.13
  1.09
SOURCE: Heston, 1950.
                                        26

-------
   The authors concluded that sulfur mustard was carcinogenic, showing a response similar
to nitrogen mustard, which was also studied in the same paper.

3.2.2.2. Shimkin and McClelland (1949)

   Male strain A mice 2.5 to 3.5 months old received single IV injections of 0.0625, 0.125,
0.25, or 0.5 mg of MC (also called 3-methylcholanthrene). Each dosed group had 90 mice,
and the control group had 60 mice.  The material was prepared as serial dilutions using
cholesterol colloid.  Equal numbers of mice in each dose group were examined at 8, 13, and
18 weeks for the presence of white nodules (tumors) on the  lung surface. The results of
these counts are shown in Table 3-4.
   TABLE 3-4.  MEAN NUMBERS OF TUMORS PER MOUSE OBSERVED AT VARIOUS
         TIMES AFTER SINGLE INJECTIONS OF 20-METHYLCHOLANTHRENEa
Tumors/mouse (± standard deviations)
Dose (mg)
0
0.0625
0.125
0.25
0.50
n
30
29
29
30
28
8 weeks
0.10 ±0.31
0.66 ± 0.80
0.62 ± 0.96
2.03 ± 1 .58
4.39 ± 3.46
n

30
30
27
29
13 weeks

1.23 ±1.38
2.40 ± 1 .76
5.37 ±2.86
13.55 ± 10.08
n
30
30
28
29
27
1 8 weeks
0.30 ± 0.52
1 .27 ± 1 .66
2.61 ± 1 .87
5.55 ± 2.88
19.3± 11.83
an is the number of animals observed in each group.

SOURCE: Shimkin and McClelland, 1949.
      The authors concluded that this experimental system is sufficiently sensitive to
distinguish doses differing by a factor of two for compounds like MC.  They then proceeded to
calculate how many animals were needed to distinguish responses at the three times that
animals  were killed (e.g., 8 vs. 13 weeks or 8 vs. 18 weeks).
      For this report, we are interested in finding what dose of MC is necessary to give a
response similar to that induced by the sulfur mustard at 0.065 mg in  the Heston (1950)
experiment.  Although a numerical interpolation could be done with respect to the duration and
                                         27

-------
dose variables, inspection of the data in Table 3-5 shows that a MC dose of somewhat less
than 0.0625 mg for 18 weeks would produce a response similar to the sulfur mustard dose of
0.065 mg after 16 weeks in the Heston (1950) study. Therefore, their potencies are nearly
equal, as Heston stated in a later paper (Heston, 1953a).      *

3.2.2.3.  Stonor et al. (1984)
                                     \
      These authors tested the strain A/J mouse lung adenoma response to 11 compounds,
including BaP and MC, via both intraperitoneal (IP) and gavage administration. The animals
were 6 to 8 weeks old at the start of the tests and were killed 24 weeks after exposure began.
The IP doses were given in a single injection, and the gavage doses were given as 24
administrations. Table 3-5 shows the results in terms of the average number of tumors per
mouse and the tumor incidence among survivors. Note that for both agents the response is
larger with the single dose IP administration than with gavage. The data also show that the
response to MC is larger than that to BaP via both routes.
  TABLE 3-5.  LUNG ADENOMAS IN STRAIN A MICE INDUCED BY INTRAPERITONEAL
           AND GAVAGE ADMINISTRATION OF 3-METHYLCHOLANTHREME
                         (MC) AND BENZO[a]PYRENE (BaP)
Dose Tumors per mouse (± S.E.) Incidence (percent) in survivors
(mg/kg) MC BaP
Intraperitoneal (single injection)
0 0.28 ± 0.08
- MC

8/32 (25)
20 57 ±3.09 1±0.16 28/28(100)
50 139 ±9.3 5 ± 0.6
31/31 (100)
100 199 13 ±1.06 2/2(100)
Gavage (24 administrations)
0 0.28 ± 0.06
20 2 ± 0.26 0.45 ± 0
50 12 ±1.14 0.71 ±0

23/98 (23)
.13 28/31 (90)
.16 32/32(100)
1 00 6 ± 0.98 1 .0 ± 0.24 1 8/18 (1 00)
BaP


22/33 (67)
31/31 (100)
32/32(100)


11/32(35)
14/31 (45)
22/32 (69)
SOURCE: Sterner et ah, 1984.
                                        28

-------
       To estimate the relative potency of MC and BaP, four sets of data could be used; for
each route, both tumors per mouse and incidence data could be used.  One of
these (IP incidence data) is not usable because all doses gave 100% incidence for MC. For
the same reason, only the lowest dose group of the gavage incidence data could be used.
The gavage tumors per mouse data for MC show at the high dose an abnormally low
response and reduced survival, but the two lower doses could be used.  Only the IP tumors
per mouse data show a dose-response trend over the entire range of doses,  but the accuracy
of counting 199 lung tumors in a mouse lung is questionable.
       For the IP tumors per mouse data, after subtracting background, it can be seen that
the MC response increases more slowly as a function of increasing dose than it does for BaP,
so that the MC/BaP ratio decreases from 79 at 20 mg/kg to 16 at 100 mg/kg.  Since the MC
dose-response curve is beginning to show high-dose  saturation even at the two lowest doses
tested, the relative potency of MC/BaP at low doses is likely to be higher than that indicated
by these data.
       Three  approaches could be used to derive relative potency values from this data:
       1)  Fit reasonable models to the IP tumors per mouse data.
       2)  Fit reasonable models to the lowest two dose groups of the MC gavage tumors per
          mouse data and all dose groups of the BaP gavage tumors per mouse data.
       3)  Fit the linearized multistage model to the gavage incidence data.  The gavage data
          are more reliable than the IP data since the responses there were much lower.
3.2.2.3.1.  Tumors per mouse data (approaches a and b).  To derive the relative potency of
MC compared to BaP from the tumors per animal data, a dose-response function, T(d), is
chosen which fits the data reasonably well and is linear at low doses.  The ratio of the low-
dose slopes is then defined as the relative potency.  Since the standard errors (S.E.) of the
data in Table 3-5 generally increase with dose, a weighted least squares procedure is
appropriate, with the weights chosen to be the inverse of the variance (i.e., 1/[S.E.]2). This
was done using the BMDP programs P1R, P2R,  and PAR (BMDP, 1985).
      Two reasonable candidates for the mathematical form of T(d) are:
          Linear T(d) = A + Bd; slope parameter is B
          Linear-quadratic:  T(d) = E (1 + Fd)2; slope parameter is 2EF.
                                         29

-------
       When these two functions are used to fit the tumors per mouse data in Table 3-5, the
results in Table 3-6 are obtained.  Note that the linear function does not fit either the complete
MC data or the complete BaP data, whereas the linear-quadratic function fits the gavage data
but not the IP data.  Since the linear quadratic model with the gavage data is the only
combination of model and data set that gives a good fit, the value of MC/BaP potency of 9,7, -
or nearly 10, is taken to be the most appropriate description of the tumors per animal data.
  TABLE 3-6.  SLOPE PARAMETERS AND RELATIVE POTENCIES OBTAINED USING A
  WEIGHTED LEAST-SQUARES PROCEDURE FOR THE TUMORS PER MOUSE DATA

Exposure
group
IP-MC

IP-BaP
GAV-MC0

GAV-BaP
Linear11
Slope Relative
potency0
i •
2.81
35.5
(0.0791 )d
(0.123)d
16.0
0.00769
Linear -
Slope

(0.31 1)d

0.0294
0.0524

0.00541
Quadratic11
Relative
potency0

10.6


9.7

*Tha value of B In the function T(d) - A + Bd.
bThe value of 2EF In the function T(d) - E (1 + Fd)2.
cThe ratio of slope parameters, MC/BaP.
dCurve fit is visually poor for values in parentheses.  :
"Highest dose group deleted.
3.2.2.3.2.  Gavage incidence data (approach c).  The linearized multistage model (GLOBAL
86), fit to both the MC and the BaP data, gives q? values of 0.16 and 0.012, respectively, for a
ratio of qf(MC)/q?(BaP) - 13.3. This estimate is considered comparable in accuracy to the
gavage tumors per mouse data.          '
      The two estimates from the gavage data are considered to be the most appropriate for
low-dose extrapolation for reasons given above.  Therefore, the ratio of MC to BaP potencies
                                         30

-------
is estimated to range from about 10 to 13. The potency range of sulfur mustard relative to
BaP is estimated by taking the product

                       (mustard/Me) x (Mc/BaP) = 1 x 10 to 1 x 13
                                mustard/BaP = 10 to 13.

       If the potency of mustard compared to BaP is the same for inhalation exposure in
humans as it is for strain A mice, the inhalation potency for mustard can be estimated using
that ratio and the inhalation unit risk for BaP in humans. The current value of the latter factor
is 1 1 .5 per (mg/kg)/day (U.S. EPA, 1 980) based on a dietary study of BaP.  Assuming the
standard human daily air inhalation volume is 20 m3 per day for a 70*kg person and assuming
100% absorption, a BaP concentration of 1 mg/m3 results in a BaP absorption of 1 mg/m3 x 20
m3 per day x 1/70 kg - 0.286 (mg/kg) per day. Therefore, the inhalation unit risk for BaP is:

          1 1.5 per (mg/kg)/day x (0.286 (mg/kg)/day per mg/m3) = 3.29 per mg/m3.
The unit risk of sulfur mustard in the ambient air has a range of:
           3.29 x (10 to 13) - (33 to 43) per mg/m3 = (0.033 to 0.043) per
In this calculation, the human inhalation potency estimate for BaP is derived from a dietary
study in rats rather than from the inhalation study of hamsters developed by the U.S. EPA in
1984 (U.S. EPA, 1984) because the hamster data, although published, are considered
incomplete.  The U.S. EPA is currently analyzing more complete data from the authors that
were not published.

3.2.3. Relative Toxicity Approach (Watson et-al., 1989)
       Watson et al. (1989) estimated the carcinogenic potency of sulfur mustard using a
relative toxicity approach described earlier by Jones et al. (1985).
       The dose of sulfur mustard that causes certain toxic effects was compared with the
doses of other chemicals that cause the same toxic effects, preferably in the same study with
                                          31

-------
the same animal system using the same route of exposure. Several standard chemicals were
                                      I
chosen from the literature as a basis for comparison, both with respect to the other standard
chemicals and with respect to sulfur mustard.  The entire toxicity literature (as summarized by
RTECS files, Lewis and Sweet, 1985) was scanned to tabulate these relative potencies, and
average relative potencies were calculated for each route of administration, animal system,
and toxic end point for which data were available.  Watson et al. chose BaP as their reference
compound and, after calculating the potency of sulfur mustard relative to BaP, used the
cancer unit risk for BaP, determined by the U.S. EPA, to derive the risk estimates for sulfur
mustard. Their end result was that mustard is 1.3 times more potent than BaP, with a range
of 0.6 to 2.9. Using a value for BaP potency  of 6.1  per (mg/kg) per day (derived from a
hamster inhalation study in U.S. EPA, 1984),  along with the ambient air concentration at
Aberdeen Proving Ground (which are the same values we use in the following exposure
section of this report), they obtained lifetime risk estimates ranging from 3 x 10"8 to 1  x 10"7 for
a lifetime exposure.   Later, Watson et al. (1989) re-calculated  these lifetime risk estimates as
5.0 x 10* to 1.4. x 10"7. This minor change has negligible effects on subsequent comparisons
made in this report.  This is equivalent to a lifetime risk from 3-year exposures of 1.3 x 10'9 to
4.3 x10'9.
                                      I
3.2.4.  Summary and Conclusions
       In this report, two approaches have been taken to estimate the dose-response
relationship for sulfur mustard. The first, based on long-term exposure of Sprague-Dawley/
Wistar rats to sulfur mustard vapor (McNamara et al., 1975), concluded that the most
appropriate estimate of the upper-bound lifetime incremental unit cancer risk is 8.5 x 10"2 per
pg/m3, although an analysis of a different set of data in that study yielded a value of
9.4 x 10"2 per ng/m3. The carcinogenic responses observed in this study at a vapor
concentration of 0.1 mg/m3 were squamous cell and basal cell carcinomas of the skin.
       Although one might question the relevance of rat skin cancer to human lung cancer,
sulfur mustard is extremely reactive to any tissue of contact, and the trapping of the material
by the fur is expected to produce conditions for persistent skin contact. The bronchial
epithelium is also exposed to constant doses under conditions of continuous inhalation. The
measure of the effective dose of sulfur mustard to both rat skin epithelium and human lung
                                          32

-------
 epithelium is assumed to be the air concentration (ppm).  This is based on the expectation
 that the carcinogenic sensitivity of the two tissues to airborne contaminants is equal, which is
 a reasonable but as yet untested assumption.
        The second approach compared the strain A mouse lung adenoma response to
 intravenous administration of sulfur  mustard (Heston, 1950), intravenous .administration of 20-
 methylcholanthrene (Shimkin and McClelland, 1949), and intraperitoneal and gavage
 administration of 20-methylcholanthrene and BaP (Stoner et al.,  1984).  From these studies a
 plausible range of the potency ratio of sulfur mustard to BaP was estimated to be 10 to 13.
 This ratio was"multiplied  by the BaP slope factor derived from an ingestion study of rats (U.S.
 EPA, 1980), which is converted to an inhalation risk estimate,  to derive the inhalation potency
 for sulfur mustard. The result of this process is a range from 3.3 x 10'2 to 4.3 x 10'2per ng/m3.
        Since the results of the relative potency approach are close (within a factor of three) to
 the inhalation bioassay in rats of sulfur mustard vapor, the value of 8.5 x 10'2 per ng/m3
 derived from the direct bioassay is considered to be confirmed by the relative potency
 approach and is considered the most appropriate estimate.
       The unit risk for sulfur mustard from the Watson et al. (1989)  results, as discussed in
 section 3.2.3., can be derived as follows:

       sulfur mustard unit risk      = BaP inhalation unit risk x relative potency

                                  = 6.1 per (mg/kg)/day x 1.3
                                      x 0.286 x 10'3 (mg/kg)/day per ng/m3

                                  = 2.3x 10'3  per |ag/m3.

This has a range of (1.0 to 5.1) x 10'3 per jag/m3.

       These dose-response estimates are considered to be highly uncertain, but to a  degree
that cannot be measured. The McNamara et al. (1975) study was not of a standard design,
and the assumptions required to do  a quantitative analysis are considered to be relatively
tenuous. Separate estimates of the  "toxicity" and the "carcinogenicity" sections of that  report
were made in order to check for consistency. Their close agreement does show that our
assumptions about the conduct of the experiment and about the accounting for the  unreported
animals are internally consistent,  but too  few animals were exposed and followed for a  lifetime

                                          33

-------
to give adequate sensitivity for detecting long-term effects, and the uncertainty about the
experimental concentrations is too great to allow confidence about the absolute potency value.
The relative potency argument is conceptually weak in its foundation, but it was attempted for
at least a partial verification of our analysis of the inadequate animal experiment.  The degree
of agreement between the two completely different sets of data and assumptions is
remarkably close, but it is unclear whether the agreement is fortuitous or whether it is some
indication of overall accuracy in the conclusions.
                                          34

-------
                             4. EXPOSURE EVALUATION
                                           >
       Since the incinerator facilities are just now being designed, no monitoring data exist.
 The incinerator operations are being designed so that no detectable concentrations of sulfur
 mustard occur in the incinerator stack, which is continuously monitored with instrumentation
 that can detect a concentration of 0.03 mg/m3. The U.S. Army has made an estimate of the
 ambient concentrations at the boundary of the Aberdeen Proving Ground, assuming
 continuous operation at the designed incinerator feed rate (U.S. Army, 1987). The results of
 air dispersion modeling at this facility show that the annual average concentration at various
 points adjacent to the property fence line ranges from 1.3 x 10"5 to 3.7 x 10'5 jig/m3.  The
 expected duration of the entire national disposal program is 1 to 3 years, with the actual
 duration at each site depending on the amount of material to be processed.  As Table 1 -2
 shows, the population residing near the other sites is much smaller than that near Aberdeen,
 Maryland, so the analysis of this site gives an idea of the largest and worst-case number of
 estimated cases around any of the facilities.  Since the authors (U.S.  Army, 1987) did not
 attempt to produce a population-weighted average air concentration, the Aberdeen exposures
 are expressed here as a maximum concentration (3.7 x 10'5 ng/m3) and a total number of
 people (200,000) within 20 km of the stack emissions.  Presumably the other facilities will
 have comparable ambient concentrations but fewer people than at Aberdeen.
       The Exposure Assessment Group of the Office of Health and Environmental
Assessment has not attempted to verify the dispersion modeling carried out under the
sponsorship of the Army.  The most obvious  refinement to the Army assessment would be to
produce a population-weighted air concentration estimate so that a more comprehensive
upper-bound estimate of the total number of cancer cases could be made.
                                         35

-------
                            5.  RISK CHARACTERIZATION

5.1.  OTHER QUANTITATIVE RISK ESTIMATES

       Other groups have investigated the safety of the expected maximum emissions of
sulfur mustard at Aberdeen Proving Ground.  They are discussed below.

5.1.1.  Watson etal. (1989)

       This approach, discussed in section 3.2.3. and modified to account for a 3-year
exposure rather than a lifetime exposure, results in an upper-bound individual risk range of
1.3 x10'9to4.3x10-9.

5.1.2.  EA Engineering, Science, and Technology Inc. (1987)

       EA Engineering used the same approach as Watson et al. (1989), adopting their higher
relative potency estimate (mustard/BaP) of 3.2 (from an earlier draft of their paper) and two
estimates of the BaP carcinogen potency of 6.1 and 11.5 per (mg/kg) per day. They
calculated a range of 5.9 x 10"9 to 1.1 x 10* for a 2-year exposure period.  This is equivalent
to a range of 8.9 x 10"9 to 1.7 x 10"8 for a 3-year exposure period.
                              *
5.1.3.  Rosenblatt (1987)

       Rosenblatt (1987) quantitatively estimated lung cancer (air unit risk) from the Wada et
al. (1968) report of lung cancer incidence in Japanese factory workers. From this estimate
(see section 3.1.), the upper-bound incremental individual risk from a 3-year exposure to the
maximum (fence line) air concentration of sulfur mustard is

              R - 3.7 x 10-5 ng/m3 x (3/70) x 4.6 x 10"4 per (4/m3) = 7.3 x 10'10.
                                          36

-------
 5.1 A The Public Health Service (CDC, 1988)

       The Public Health Service (PHS), based on an expert panel review and evaluation,
 recommended that the ambient concentration of mustard gas be kept as low as possible
 during incineration to ensure the safety of the general population. A control limit of 1 x 10"*
 mg/m3 (or 0.1 >g/m3), which is the lowest detectable level, is considered by the PHS to be
 protective of health for the general population for the limited duration of the incineration
 program. Their conclusion was not based on a quantitative risk estimate, which they clearly
 consider inappropriate, considering the poor quality of the data. Rather the PHS limit was
 developed on the basis of an overall judgment that the ambient levels expected in the
 incineration program are  not dangerously high. The U.S. Army can monitor for mustard
 agents at concentrations  as low as 0.1 jag/m3 using a 12-hour sampling time. This capability
 has been shown at only one site, and the PHS recommends (CDC, 1988) that it be developed
 at all sites of mustard destruction. The current limit for reliable detection of sulfur mustard in
 the workplace air is 0.003 mg/m3 when the detector is operated with an automatic cycle time
 of 8 minutes.
       When the cancer  unit risk information developed in the current report (section 3.2.4.) is
 used, the upper-bound incremental lifetime  cancer  risk (R) that a person would have if
 exposed for 3 years to 0.1 jag/m3 is:

                 R = 0.1 fig/m3 x 8.5 x 10'2 per ng/m3 x 3/70 = 3.6 x10'4.

 This upper-bound risk value is not extremely low.  If our analysis is correct, then any amount
 of mustard detectable (above 0.1 jag/m3) over a long-term average may be of some concern.
      ,The lower population risk estimates, which we characterize in the next section, are
 based on ambient concentrations that occur after the  stack emissions that are already below
 detectable concentrations are diluted by the ambient air.

5.2.  CURRENT ASSESSMENT
       At the Aberdeen site, a resident hypothetically living at the point of highest ambient
concentration at the property boundary and exposed 24 hours per day for the maximum
                                         37

-------
expected program duration of 3 years (3/70 - 0.0429 of a 70-year lifetime) would have an
upper-bound incremental lifetime cancer risk due to sulfur mustard exposure of:
               R - 3.7 x 1 0-6 uglm3 x 8.5 x 1 0'2 (ng/m3)'1 x 0.0429 - 1 .4 x 1 0'7

       Using this upper-bound individual lifetime risk value, along with the estimate of 200,000
people residing within 20 km of the Aberdeen site, one could estimate the number of cases
(N) over a 70-year period (the standard human lifetime) under the extremely unlikely worst-
case assumption that all 200,000 people actually breathed the same maximum fence line
concentration for 24 hours per day for a 3-year incineration period. This is:

        N » 1 .4 x 1 0'7 x 200,000 - 0.028 per 70 years - 0.028/70 = 4 x 1 0"4 per year.

The occurrence of less than 0.03 cases in the entire population over a 70-year period would
be considered negligible by most program offices in the U.S. EPA. If this were the average
rate of appearance of cases in the population, we would  have to wait for 100 x 70 = 7,000
years to get three occurrences, or 7,000/3 = 2,300 years to get one incidence due to the
inhalation of sulfur mustard from the incineration.  This is clearly a negligible population risk,
which becomes even more insignificant if one considers the implausibly overstated
assumption that all 200,000 people experience the maximum concentration.
       In our judgment, it is not necessary to refine the exposure estimate by calculating the
population-weighted average exposure, since that refinement would only lower an already
negligible upper-bound population risk.     I
       Both  the individual risk to a maximally exposed person (about 1  x 10'7) and the total
Incidences per year in the population under worst-case assumptions (about 0.0004) would be
considered negligible by most program offices in the U.S. EPA.  Therefore, even though sulfur
mustard is a known human carcinogen, the expected ambient concentrations  under routine
operating conditions during the incineration program are so low that no alterations in the
current incinerator design are necessary for further protection of public  health.
       The conclusions reached in this document are dependent on all of the assumptions
                                        i
made in this report. Of particular importance is the assumption that all eight facilities will  be
operated at similar efficiencies and feed rates to that specified in the Aberdeen 1987 report

                                           38

-------
(U.S. Army, 1987), and that no accidents will occur to cause any of the facilities to generate
emissions in excess of the plant design.  In the dose-response section of the report, the most
significant assumptions are:  (1) that human cancer risks are linearly related to the product of
air concentration and duration, even in the low-concentration range expected in the ambient
air; and (2^ that the human dose-response relationship is the same as that for animals, when
doses are expressed as lifetime average air concentrations.
       In the hazard identification section, the  major conclusion, namely that sulfur mustard is
classifiable as a category A human carcinogen using the U.S. EPA guidelines (U.S. EPA,
1986),  is not significantly in doubt.  Lung cancer appeared in veterans several years after
battlefield exposures and in factory workers occupationally exposed over longer periods of
time. The evidence showing induction of skin  cancer in rats exposed to sulfur mustard vapors
and lung adenomas in mice after intravenous injection supports the human findings.  Evidence
showing genotoxicity of mustard and its chemical characteristics as an alkylating  agent (IARC,
1975; NRC, 1984)  also supports the categorization of sulfur mustard as a known human
carcinogen.
                                          39

-------
                                  6.  REFERENCES
Alexander, S.F. (1947) Medical report of the Bar! Harbor mustard casualties. The Military
      Surgeon 101:1-17.

Beebe, G.W.  (1960)  Lung cancer in World!War I veterans: possible relation to mustard-gas
      injury and 1918 influenza epidemic. J. Natl. Cancer Inst. 25:1231-1251.

Biomedical Data Processing (BMDP) (1985)  BMDP Statistical Software.  Berkeley, CA:
      University of California Press.

Case, R.A.M.; Lea, A.J. (1955)  Mustard gas poisoning, chronic bronchitis, and lung cancer.
      Br. J. Prev. Med. 9:62-72.

Centers for Disease Control (CDC)  (1988) Final recommendations for protecting the health
      and safety against potential adverse effects of long-term exposure to low doses of
      agents:  GA, GB, VX, mustard (H,HD,HT) and lewisite (L).  Federal Register 53:8504-
      8507.
                                       i
EA Engineering, Science, and Technology Inc. (EST Inc)  (1987)  Community review support
      study: Aberdeen area.  Sparks, MD; Report no. USA72A.
                                       i
Gilchrist, H.L.; Matz, P.B.  (1933) The residual effects of warfare gases: II mustard. Available
      from: Government Printing Office, Washington, DC; pp. 43-93.

Haddow, A.  (1959)  The chemical and genetic mechanisms of carcinogenesis: II. Biologic
      alkylating agents.  In: Hamburger, F., ed.  Physiopathology of cancer. 2nd ed. New
      York, NY: Hoeber-Harper, pp. 602-685.

Heston,  W.E. (1950) Carcinogenic  action of the mustards. J. Natl. Cancer Inst. 11:415-423.

Heston,  W.E. (1953a) Pulmonary tumors ii) strain A mice exposed to mustard gas. Proc.
      Soc. Exp. Biol. Med. 82:457-460.

Heston,  W.E. (1953b) Occurrence of tumors in mice injected subcutaneously with sulfur
      mustard and nitrogen mustard. J. Natl. Cancer Inst. 14:131-140.

Howe, R.B.; Crump,  K.S. (1983) WEIBULL 82.  K.S. Crump and Company, Inc., Ruston, LA.

Howe, R.B.; Crump,  K.S.; Van Landingham, C.  (1986)  Global 86: A computer program to
      extrapolate quantal animal toxicity to low doses. Washington, DC: U.S. Environmental
      Protection Agency; EPA subcontract no. EPA-2-251U-2745, EPA contract no.
      EPA-68-01-6826.

International Agency for Research on Cancer (IARC) (1975) I ARC monographs on the
      evaluation of the carcinogenic risk of chemicals to man: some aziridines, N-, S  and O-
      mustards and selenium, vol. 9. Lyon, France: IARC, pp. 181-192.

                                         40

-------
 International Agency for Research on Cancer (IARC)  (1982) IARC monographs on the
       evaluation of carcinogenic risk to humans:  chemicals, industrial processes and
       industries associated with cancer in humans: IARC monographs, volumes 1  to 29.
       supplement 4. Lyon, France: IARC, 163.

 International Agency for Research on Cancer (IARC)  (1987) IARC monographs on the
       evaluation of carcinogenic risk to humans:  overall evaluations of carcinogenicity:  an
       updating of IARC monographs volumes 1 to 42. supplement 7.  Lyon, France: IARC,
       41.        '

 Jones, T.D.; Walsh, P.J.; Zeighami, E.A.  (1985)  Permissible concentrations of chemicals in
       air and water derived from RTECS entries: a "RASH" chemical  scoring system.
       Toxicol. Ind. Health 1:213-234.

 Klehr, N.W.  (1984) Cutaneous late manifestations in former mustard gas workers.  Z. Hautkr
       59(17):1161-1170.

 Lewis, R.J.; Sweet, D.V. (1985)  Registry of toxic effects of chemical substances. U.S.
       Department of Health and Human Services; DHHS report no. 87-114.  Available from:
       Government Printing Office, Washington, DC; S/N 17-33-00431-5.

 McNamara, B.P.; Owens, E.J.; Christensen, M.K.;  Vocci, F.J.; Ford, D.F.; Rozimarek, H.
       (1975)  lexicological basis for controlling levels of mustard in the environment.
       Edgewood Arsenal Special Publication EB-SP-74030.  U.S.  Army.  Aberdeen Proving
       Ground, MD.

 Manning, K.P.; Skegg, D.C.G.; Stell, P.M.; Doll, R.  (1981)  Cancer of the larynx and other
       occupational hazards of mustard gas workers.  Clin. Otolaryngol. 6:165-170.

 Morgenstern, P.; Kass,  F.R.; Alexander, W.W.  (1947) Residual mustard gas bronchitis:
       effects of prolonged exposure to low concentrations  of mustard  gas. Ann. Intern. Med
       26:27-40.

 Nakamura, T.  (1956) Studies on the warfare gas  injury in Japan.  Report 1. On the general
       condition of the poison gas island.  Hiroshima Med. J. 4:1141-1149. (In Japanese.)

 National Research Council (NRC)  (1984)  Possible long-term health effects of short term
       exposures to chemical agents,  vol. 2, cholinesterase reactivators, psychochemicals,
       and irritants and vesicants.  Washington, DC: National Academy Press.

 Nishimoto, Y.; Yamakido, M.; Shigenobu, T.; Onari, K.; Yukutake, M. (1983) Long-term
       observation of poison gas workers with special reference to respiratory cancers.
       langyo Ika Siagaki Zasske  March 20:(5 suppl.)89-94.

Norman, J.E.  (1975) Lung cancer mortality in World War I  veterans with mustard-gas injury:
       1919-1965.  J. Natl. Cancer Inst. 54:311-317.

Prentice, A.M.  (1937)  Chemicals in war.  New York, NY: McGraw-Hill,
                                         41

-------
Rosenblatt, D.H.  (1987) Recalculation of general population exposure limits and stock
      detection levels for mustard [unpublished]. Fort Detrick, MD: U.S. Army Biomedical
      Research and Development Laboratory.

Shimkin, M.B.; McClelland, J.N.  (1949)  Induced pulmonary tumors in mice.  IV. Analysis of
      dose-response data with methylcholanthrene. J. Natl. Cancer Inst. 10:597-603.

Stoner, G.D.; Greisiger, E.A.; Schut, H.A.; Pereira, M.A.; Loeb, T.R.; Klaunig, J.E.; Branstetter,
      D.G. (1984) A comparison of the lung adenoma response in strain A/J mice after
      intraperitoneal and oral administration of carcinogens.  Toxicol. Appl.  Phairmacol.
      72:313-323.

U.S. Army (1987) Chemical stockpile disposal program: evaluation of multiple incinerator air
      quality impacts Edgewood area.  Program Executive Officer—Program Manager for
      Chemical Demilitarization. Aberdeen Proving Ground, MO, Report no. SAPEO-CDE-
      IS-87004.

U.S. Army (1988) Chemical stockpile disposal program: final programmatic environmental
      Impact statement.  Program Executive Officer—Program Manager for Chemical
      Demilitarization, Aberdeen Proving Ground, MD, Volumes 1,2,3.
                                       i
U.S. Environmental Protection Agency (U.S. EPA) (1980) Ambient water quality criteria
      document for polynuclear aromatic hydrocarbons.  Prepared by the Office of Health
      and Environmental Assessment, Environmental Criteria and Assessment Office,
      Cincinnati, OH, for the Office of Water Regulations and Standards, Washington, DC;
       EPA report no. EPA-440/5-80-069. Available from: National Technical Information
      Service, Springfield, VA;  PB-81 -117806.

U.S. Environmental Protection Agency (U.S. EPA) (1984) Health effects assessment for
       benzo[a]pyrene. Prepared by the Office, of Health and Environmental Assessment,
       Environmental Criteria and Assessment Office, Cincinnati, OH, for the Office of
       Emergency and Remedial Response, Washington, DC; EPA report no.
       EPA-540/1 -86-022.  Available from: National Technical Information Service,
       Springfield, VA; PB-134335/AS.
 U.S. Environmental Protection Agency (U.S. EPA) (1986)
       assessment.  Federal Register 51:33992-34003.
Guidelines for carcinogen risk
 Wada, S.; Nishimoto, Y.; Miyanishi, M.; Kambe, J.; Miller, R.W.  (1968)  Mustard gas as a
       cause of respiratory neoplasia in man.  Lancet June: 1161 -1163.

 Wada, S.; Yamada, A.; Nishimoto, Y.; Tokuoko, S.; Miyanishi, M; Katsuta, S.; Umisa, H.
       (1963) Neoplasms of the respiratory tract among poison gas workers.  J. Hiroshima
       Med. Assoc. 16(8):56-68.

 Watson, A.; Jones, T.D.; Griffen, G.D.  (1989)  Sulfur mustard as a carcinogen: application of
       relative potency analysis to the chemical warfare agents H, HD, HT.  Regul. Toxicol.
       Pharmacol. 10:1-25.
                                          42

-------
Weiss, A.; Weiss, B.  (1975)  Carcinogenesis due to mustard gas exposure in humans: an
       important indicator for therapy using alkylating agents.  Dtsch. Med. Wochenschr
       100:919-923.

Yamada, A. (1963)  On the late injuries following occupational inhalation of mustard gas, with
       special reference to carcinoma of the respiratory tract. Acta Pathol. Jpn. 13:131-155.
                                          43
                                                       &U.S. GOVERNMENT PRINTING OFFICE: 1991 - 548-187/40593

-------

-------

-------

-------

-------
m
     3
I   ?
 I
en   —
CO
     8
                      > me
(f  Q)
5  »
£  (A
                     Q5TO
                     ?f?I
                      "
                           rn
                     1
              m
              3)
                 m
                     BO
                     Tl
                     m
              P     jn
                     com

-------