'United States
Environmental Protection
Agency
Office of Health arid
Environmental Assessment
Washington DC 20460
EPA/600/6-88/007Ab
June 1988
External Review Draft
Research and Development
A Cancer
Risk-Specific
Dose Estimate for
2,3,7,8-TCDD
Appendices A
Through F
Review
Draft
(Do Not
Cite or Quote)
NOTICE
This document is a preliminary draft. It has not been formally
released by EPA and should not at this stage be construed to
represent Agency policy. It is being circulated for eomment on its
technical accuracy and policy impticati'ons;
-------
-------
(Do Not Cite or Quote) EPA/600/6-88/007Ab
June 1988
External Review Draft
A Cancer
Risk-Specific Dose Estimate for
2,3,7,8-TCDD
NOTICE
This document is a preliminary draft. It has not been formally released by EPA and should
not at this stage be construed to represent Agency policy. It is being circulated for comment
on its technical accuracy and policy implications.
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document is an external draft for review purposes only and does
not constitute Agency policy. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
March 1988
Review Draft
APPENDIX A
QUANTITATIVE IMPLICATIONS OF THE USE OF DIFFERENT
EXTRAPOLATION PROCEDURES FOR LOW-DOSE CANCER RISK
ESTIMATES FROM EXPOSURE TO 2,3,7,8-TCDD
Steven P. Bayard, Ph.D.
Carcinogen Assessment Group
I
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
-------
-------
.. CONTENTS
I. INTRODUCTION AND DEFINITION OF TERMS. . 1
A. INTRODUCTION 1
B. DEFINITION OF TERMS . 6
1. Terms Associated with Modeling 6
2. Terms Associated with the Uncertainty Factor Approach 8
II. EPA's USE OF THE LINEARIZED MULTISTAGE MODEL FOR CARCINOGEN
RISK EXTRAPOLATION AND COMPARISON WITH OTHER MODELS 8
A. DESCRIPTION OF THE MULTISTAGE AND LINEARIZED MULTISTAGE
MODELS , 8
B. USE OF THE LINEARIZED MULTISTAGE MODEL FOR RISK EXTRAPOLATION
OF 2,3,7,8-TCDD: COMPARISON OF FOUR U.S. AGENCIES 9
C. ALLOMETRIC AND BODY BURDEN CONSIDERATIONS . .13
D. OTHER EXTRAPOLATION MODELS 18
1. Longstreth and Hushon (1984) .18
2. Sielken (1987). . .22
E. TIME-TO-TUMOR ANALYSES 26
III. TREATMENT AS A PROMOTER 28
A. UNCERTAINTY FACTOR APPROACH 28
B. MODELING AS A PROMOTER UNDER THE MOOLGAVKAR, VE-NSON,
AND KNUDSON TWO-STAGE MODEL .31
IV. COMPARISON OF ANIMAL PREDICTION WITH ACTUAL HUMAN DATA. . . 35
A. THIESS et al. (1982) 38
1. Description 38
2. Exposure Estimates 38
3. Risk Estimates • 41
-------
CONTENTS (continued)
V.
VI.
B. ZACK AND SUSKIND (1980)
1. Description . . . .
2. Risk Estimates. . .
DISCUSSION AND SUMMARY. . .
REFERENCES
.44
.44
.44
.49
.54
m
-------
I. INTRODUCTION AND DEFINITION OF TERMS
A. INTRODUCTION
2,3,7,8-Tetrachlorodibenzo-jD-dioxin (2,3,7,8-TCDD) is the most potent
animal carcinogen ever tested. It is 50 times more potent than aflatoxin Bl on
a per mole basis and 50 million times more potent than vinyl chloride. In
addition to its carcinogenic potency, 2,3,7,8-TCDD is also the most potent animal
teratogen known and it causes other reproductive and immune system effects at
extremely low doses as well.
Because of these severe toxicities, many U.S. federal and state, as well as
foreign, regulatory and health agencies have proposed or implemented regulations
or advisories based on levels of concern for 2,3,7,8-TCDD. Among these agencies,
the U.S. Environmental Protection Agency (EPA) was, to this writer's knowledge,
the first to actually produce an (upper-limit) estimate of cancer risk for
2,3,7,8-TCDD exposure to humans (U.S. EPA, 1980). This estimate was based on a
methodology that extrapolated from cancer responses at doses of 1, 10, and 100
ng/kg-day in an animal lifetime feeding study (Kociba et al., 1978; Table 1) to
humans at still lower levels. Initially, this extrapolation was based on a
simple linear extrapolation from the lowest dose to show a significant elevated
response (10 ng/kg-day caused a statistically significant increase in liver
tumors--hyperplastic nodules—versus control). Shortly afterward, however, the
methodology was modified to include all dose-response points in the extrapolation
procedure. The new methodology was based on a multistage model for carcino-
genesis due to a specific time ordering of changes, first proposed by Armitage
and Doll (1954) but modified by Crump et al.. (1977, 1979) to include dose-
response for extrapolation purposes. Often called the linearized multistage
model, the Crump model is distinguished by its approach of providing upper-limit
-------
oo
)— 4
oo
^—
_j
^c
— *~
^^
>-
CD
O
O
1—
=
~T" re
2= 1—
oo »-• oo
- «=c ai
rC
^ 00 H-
CD ZT CD
t-H f5 )_ |
oo a: cz
•^ /•"*»
'""'^ oo
OO Q
r-- i LU
cr> »— i
t-l 00 Q
>- ex. •=£
Q 3=
—^ *^- 1
T— ' ^^ r—
1— LU
OO _l OO
H— 3 «=C
—4
LU 2:
•d => «*
- _J r-4
1 «=c 2:
CM s; i— i
LU — 1
O LU LU
O ^O
I ^f
CO
«v
!**>«
r\
CO
«\
CM
•
i— t
LU
1
CQ
k
(O
-o
1
vx
^
3
*— " '
I—
cu
cu
cu
CO
0
Q
r— 1
•
O
r-l
O
CD
I— 1
CD
O
CD
1ZT"
o
o -^
c
o
o
^^^
CO
•r-
O
cn
ra
•1 —
-a
•a
cz
eo
CU
3
CO
CO
•i —
h-
-0
ja '-^
— ^ CO
«^- 1 f — N
^-^* | *~~^ CO 3^
S^ CD 5^ r-l CM J3
OO r-| r-l X CO •"-
r-H X CM WD 'CO
CO '^1- 1
. . 0 O
CD CM •* CT> =d- t-<
«3- II CM II ^~ V
^ Q. ^ Q. •=* 0.
f"— — - LO OO •> —
ja
^ — %<• — »
' — ^ r"»rr}-
5^ -4-* 1
- — ' — 00 O CD
&S. S* OO -Q r-l
CD -=d- X
"O i— 1
CO fO •
co t^- ^ jc oo
•^- CM ^.11
^- ^^ CO CM Q.
O 1-1 r-l '
Q Q T
f"*"> f~~) i(*t
*. — ^ ^-^- *^^
OO O CO
•=d- OO "^1-
: CD O CO
^~* ^-- ~^
S^ &? rM
O CM t— 1
LO <=J- LO
oo ir> oo
CD t— i cn
CU CO
•4-3 -r- O
to co •<—
•— a •»->
3 > t- O
= co cr cu fo +J
cn c) c co t_ i— ro
co c c: T- — 30.
•r- -i- T- ^3 T3 r— CU
CO N O t_ CU fO r— f"" n3
§*" — * CU "^^ E
CTi O CO O
re -r-o -r-cr-. ocuc
re cn re , re re o cr> cu ^ 3 o
Ct-r— CO"C-C_r-l >Q.-O£_
re 3 cu cu re -t-> re *^*^ *r— cu o re
r^i ..] NX (_) ^^ OO O CM — J ' g— o
• 1 —
o
0 • • «
:*: r-i CM co
^ — s
s^ o
CM ^^
coco
"* — ^ 1
o
0 r-<
•* v
— II
<3- Q.
oo — -
ja
^ — ^
*• — ^oo
^^ 1
OO o
CO t-H
1— <
00 .
«^J- CO
' — II
CO Q.
r«f x *
-7-
iS-
oo
^~
^^
CO
*^-
1— 1
r-l
LO
co
cn
'oT
>
o
o
re
00 CU
«- 0
o
4J
«\ CO
CM re
cu
r-l
^"•^ 1 ^
re
-o
cu -a
c: re
f»
E i %
o re
0 t.
*• — -.-
r— JC £-
re o o
-t-> re E
O CU 3
.
cn
CD
LO
CO
•r—
^_>
re
<-
a> •
re ~o
E ^
CU r^—
i] »r—
re
re "f^
i
M- CU
O C
O
| S s ,
f~
en •*-»
•r- CO
cu cu
5 h-
>^ 4-*
-a <->
O "3
jQ X
LU
CU
cn £-
re cu
i. JT
cu co
> -I-
re ^2
-------
. „
"O
o>
3
4J
O
O
^
r— 1
1 i 1
_J
CO
h-
"^~
03
TD
1
CO
^^
CD
13
'oi
CU
CU
0
O
, — I
CD
t — (
o
CD
,_!
O
O
CD
i _
'o
£_
O 4->
o
o
•" — '
CO
•f—
o
c;
CD
03
•r—
•o
T3
03
QJ
CO
CO
1 —
.a .a
LO CO -— »
^— -x t .* — >. i ^^
fc^ O S^ O CM -Q
CD r-l CO r-l CO • -
CM X CM X — -CO
LO «3" 1
. .00
CD VO CM i— 1 «3- r-l
«3- II CM II "•*». II
*-. Q. •*•-. d. CO Q.
co — LO — co — -
JQ
LO^
^~~ ^^ So
S^ 6* LO r-l
CD ^J" " — ' X
- — ^— • co
CO .
CO I1"-. •* r-l
«^j~ CM "^* H
"^^^ *^^^ r^* «— *
O r-l CM
. — - ' — - S^
S£ &« tO
0 O r-l
•s — s ^—^- '**~*
OO O 00
•* CO -=3-
o o oo
-— - ^— ~ en
o o ••--•
•*^-^ ^—*
LO
LO «* CO
CO LO "^^*-
O CD r-H
O)
fO 1
r- 0
to +->
O3 Q. to to
E e °-
o *o o cu
C t_ C -C
•1 — O3 T— ^^*^
o ^: o co ro
S_ ~^. £_ O) g
(X3 tO fO r™^ ^5
O CU O 3 C
+j -o .p~
r— 03 r— O O
r— sr r— cr c.
5 CU •>- d) 03
QJ O J3 O O 0
> co rs co 4-> £-
CU 3 +J CD CO to
£_ O O 03 , —
E= i — £E *- i — 3
CO CD 03 03 03 CU Q-i —
crs cozs >Oi —
cu zs cr 03 o* *»~ cu cu
c_ — 100 z; — \'z. o
cr • • •
CO r-l CM OO
^-»
^^
CM £l
oo ^->
•-co
1
0 0
•=1" r- 1
*^s. V
•=i- Q-
co «. —
J3
- — -LO
SD o
LO r-l
X
CO
oo •
<^- * — I
*"^» II
r-. ex
CM -~^
^s
to
I—I
^—^
CO
•=1-
co
cV>
r— 1
LO
CO
to
1—1
*0>
0
.a
03
oo cu
£Z
t- o
o
^_)
* CO
CM 03
cu
** r— •
r-l
*• -r^
03
-a
C 03
~E +J
O 03
O £_
r— f— t-
O3 o O
+•> 03 E
o cu 3
1 • • -1-5
CD
o
LO
CO
03
a> •
03 "O
CD r—
tf— *f—
03
03 -f->
1
q — CU
o c:
o
j % ^ ^
,c
CD +J
•r- CO
cu cu
5 ^
O 03
UJ
cu
CD £-
03 CU
t_ .c:
CU co
^C LL.
to <^
-------
estimates of risk consistent with nonthreshold low-dose linearity. This model
is presented in section 11.A.
Following EPA's efforts, two other U.S. agencies, the Food and Drug
Administration (FDA, 1983) and the Centers for Disease Control (CDC) (Kimbrough
et al., 1984) produced cancer risk estimates based on slight modifications of
EPA's methods but with similar resulting estimates. In another minor variation,
the State of California (1984) used the Crump linearized multistage model to
extrapolate the cancer response to humans from a mouse gavage study performed
by the National Cancer Institute (NTP, 1982). All efforts produced results
within a factor of 10. These are discussed in section II.B.
Within the framework of the linearized multistage model and the Kociba et
al. rat feeding study, two other efforts appearing in the literature are note-
worthy. First, Longstreth and Hushon (1984) applied several mathematical non-
threshold, nonlinear model s (Logit, Probit, Weibull, and multihit) to the
cancer response in the Kociba et al. study, and compared the extrapolated
results with those of the linearized multistage model. Second, Sielken (1987)
fit the Kociba data with the multistage model (but not the Crump linearized
version applying upper limits) and also with a modified version which allowed
the input of actual observation times. He then compared actual estimates
derived from the multistage model with those of the EPA, which used the upper
limits based on the Crump version. The Sielken paper is discussed further in
section II.D.
In contrast to all of the above attempts at extrapolating from animal data
to humans by nonthreshold models, several U.S. nonregulatory agencies have
applied safety or uncertainty factors, not models. The uncertainty factors of
between 100 and 1,000 are applied to doses that have shown no adverse effects
in animal cancer or other studies, and the resulting numbers are presumed safe
-------
for humans. This methodology has been used by EPA and many other agencies for
animal-to-human extrapolations for toxic effects other than cancer, but EPA has
never used this methodology to estimate cancer risk. The estimates of cancer
risk provided by the uncertainty factor approach are in the range of 150 to
1,500 times lower than the estimates provided by EPA's use of Crump's multistage
model. These estimates are presented in section III.A.
The differences in the magnitude of the estimates provided by these two
approaches require a closer look at the methodologies involved in each and in
the reasoning as to which is the proper one to use for cancer risk extrapo-
lation of 2,3,7,8-TCDD. The argument focuses on the model for complete carcin-
ogens versus promoters. Complete carcinogens have both initiating and promoting
ability, and it is the mechanism leading to the initiating part of carcinogenesis,
the attachment of the carcinogen to the DNA, that can be modeled on either a
linear or multistage basis. Those in favor of modeling 2,3,7,8-TCDD as a
complete carcinogen argue that 2,3,7,8-TCDD causes rare cancers of the hard
palate and nasal turbinates, tongue, (in male rats), and a rare form of lung
cancer, and that such rare tumors would be unlikely to be initiated except by,
the 2,3,7,8-TCDD in the experiment. Conversely, those in favor of the uncertainty
factor approach point to the strong evidence for the promoting effects of
2,3,7,8-TCDD in the liver where most of the tumors are occurring. They argue
that promotion is effectively a toxic reaction with a threshold and that all
promoters show not only thresholds but also reversibility upon cessation of
dosing. Treating 2,3,7,8-TCDD as a promoter and using an uncertainty factor
approach has a further advantage of comparing its cancer effects with its other
toxic effects using the same methodology.
A third approach is also possible.. This is an approach which models for
the cancer effects of 2,3,7,8-TCDD through its known mechanism of binding to a
-------
receptor. (This actual modeling and results are presented in section IV.B.)
While the basic model, the Moolgavkar-Venzon-Knudson (M-V-K) two-stage model,
with a promotion phase, has been used in the literature to explain many cancers,
and has been found to predict well the tumor promotion in mouse skin (Chu et
al., 1987), the approach is new in that modeling for promoters has never been
done before by regulatory agencies.
The purpose of the presentations that follow is to compare quantitatively
the estimates derived from each of the separate approaches and then to compare
them with each other. In order to do this, common terms must be introduced.
B. DEFINITION OF TERMS
1. Terms Associated with Modeling (defined here as an estimation of the incre-
mental cancer risk to humans arrived at by fitting a mathematical function to
animal response data)
Maximum likelihood estimate (MLE)—The statistical procedure by which the
parameters of the model are estimated. The MLE has many properties, in a sta-
tistical sense, which allow it to be referred to as a "statistical average" or
"best" estimate. In risk terms it might be thought of as a term that, if the
assumed model is true, provides overestimates and underestimates of the true
risk each 50% of the time.
Parameter—A constant in the model, associated either with the control
response, or the time or dose variable inputs. For example, in the Crump
linearized multistage model, the parameter associated with the linear dose
variable is denoted as c\i and is defined as the increase in cancer risk asso-
ciated with an incremental increase per unit of dose. For this reason qj is
expressed in units of reciprocal dose such as (ng/kg-day)'1.
Upper-confidence limit (UCL) estimates—The estimates resulting from a
-------
statistical procedure in which the upper-limit values of the parameters still
consistent with the data are estimated. In the linearized multistage model
(Crump), the upper-limit estimate associated with the linear term is designated
q^. In a statistical sense it is the 95% upper-limit estimate of the linear
term associated with the fitting of the linearized multistage model to the
animal data. In making cross-species extrapolations to humans, however, the
"95%" label is dropped, since the uncertainty associated with cross-species
extrapolations is considered far greater than the statistical uncertainty
associated with the model-fitting procedure. Also, because the linearized
multistage model becomes linear at low doses, the UCL on tf[ is the same as
the UCL on the incremental risk. This is not true of the nonlinear models such
as the Logit, Probit, and Weibull discussed in section II. Upper-limit incre-
mental risk estimates, however, are comparable, and the ratio of these estimates
from two different models can be expressed as the relative potency.
Risk specific dose (RsD)--A dose associated with a specified cancer risk.
For example, assume a linearized multistage model is fit to the data and the
parameter estimates are q-j_ = 3.0 x 10"3 (ng/kg-day)'1, qi = 0 for all i^l,
and q| = 7.5 x 10~3 (ng/kg-day)"1. Then for an incremental risk of 1 in
1,000,000, the dose would be the solution to 10~6 = l-exp(-3.0 x 10-3 d), and
RsD = 3.3 x ID"4 ng/kg-day would be called the risk specific dose. Likewise,
the RsD could be defined in terms of the lower limit of the dose corresponding
to a risk of 10~6. In this case aft would be substituted for q± and the solu-
tion would be RsD = 1.3 x 10~4 ng/kg-day.
A ratio of two RsDs can also be thought of as a measure of relative potency,
but in this case the higher the RsD, the lower the potency. The RsD thus
becomes a common unit to discuss relative potency between different approaches
and different types of toxicity.
-------
Virtually safe dose (VSD)--A dose associated with a very small risk. The
general reasoning in discussing RsDs and VSDs is identical. The only difference
is in one's definition of a "very small risk."
2. Terms Associated with the Uncertainty Factor Approach
The lowest-observed-adverse-effect-level (LOAEL) is defined as the lowest
dose in an experiment at which there is a statistically significant increase
over the control group in the proportion of animals for which adverse effects
are observed. The no-observed-adverse-effect-level (NOAEL) and no-observed-
effect-level (NOEL) are straightforward negations (Crockett and Crump, 1986).
The uncertainty or safety factor is an arbitrary factor applied to these levels
for the purpose of establishing concern or no-concern levels for humans.
II. EPA's USE OF THE LINEARIZED MULTISTAGE MODEL FOR CARCINOGEN RISK
EXTRAPOLATION AND COMPARISON WITH OTHER MODELS
A. DESCRIPTION OF THE MULTISTAGE AND LINEARIZED MULTISTAGE MODELS
EPA's reasons for using the linearized multistage model for risk extrapo-
lalation, in general, are discussed in the Guidelines for Carcinogen Risk
Assessment (U.S. EPA, 1986). For the 2,3,7,8-TCDD risk assessment, additional
discussions are presented in the Health Assessment Document (HAD) for Polychlor-
inated-Dibenzo-£-Dioxins (U.S. EPA, 1985) as well as in the document on the
cancer risk-specific dose estimate for 2,3,7,8-TCDD. Therefore, only an abbrevi-
ated review of its development will be presented here. Basically, its genesis
came from Armitage and Doll who proposed a theory that a cancer cell was generated
from a series of several heritable mutations in a specific order, the end result
of each change being termed a stage. The transition rate from one stage to the
next was hypothesized as being related to a probability of occurrence. The time
8
-------
rate of occurrence of the ith event is a-j+b-jd-j, i=l, ...... k, where a^ , b-j > 0
and d=dose. This, along with some other assumptions, leads to a dichotomous
response probability of the form
k
P(d) =l-exp-[cir (a-j+bid)] a-jbi £ 0
1 = 1 .
where a-j is the background transition rate for a progression to stage i , b-j
is the incremental increase in that rate per unit of dose, c is a function of
exposure duration, and P(d) is the probability of a tumor by some fixed age t
for a dose d. This model achieved some popularity mainly because of its success
at predicting many of the human epithelial cancers, and because the model now
presented the probability of a tumor as a function of dose. In addition, the
reparameteri zation of the individual transition rates leads to
P(d) = l-exp-(q0 + q-^d + q2d2 + ... + qkdk)
where P(d) is the lifetime probability of cancer at dose d. Since qu is the
parameter associated with the background rate, an assumption of independent
background leads to
Pt(d) = l-exp-(q1d + ... + qkdk) all qn- >. 0
where Pt(d) is the incremental (often called the extra) risk associated with
dose d. In the linearized form of this model an upper-limit estimate of the
linear term, q'J, consistent with the data, is calculated. At low doses, this
upper-limit linear term predominates, forcing the model to low-dose linearity.
B. USE OF THE LINEARIZED MULTISTAGE MODEL FOR RISK EXTRAPOLATION OF
2,3,7,8-TCDD: COMPARISON OF FOUR U.S. AGENCIES
Besides the choice of the linearized multistage model for animal-to-human
-------
risk extrapolation, the final risk estimates are dependent on a choice of
several other factors. Specifically, Table 2 presents a summary of 2,3,7,8-TCDD
cancer risk extrapolation by four U.S. agencies, all of which used the linearized
multistage model. Even with the use of the same model, however, the results
varied over 10-fold due to a selection of different factors relating both to
the animal data and to the procedure. Such factors are:
• Choice of animal bioassay
Adjustment made for differential nontumor mortality
among the animal treatment groups
• Selection of tumor types for modeling
• Animal-to-hum an dose equivalence
• Dose used for curve fit
As seen in Table 2, EPA, FDA, and CDC all used the cancer response data
from the female Sprague-Dawley rat in the 2-year feeding study conducted by
Kociba et al. (1977, 1978). In the EPA HAD for PolychloMnated'Dibenzo-j>-
Dioxins, the choice of the Kociba study was based on the female rat providing
the largest slope factor, q^, of all the available data sets. However, there
were other, unstated but probably better, reasons for the selection, such as
(1) the high quality of the study, (2) response at multiple sites, (3) more
applicable route of exposure to the human experience than the gavage study, and
(4) less controversial tumor sites than the mouse liver. The State of California
used the liver tumor response from the male mouse in the National Toxicology
Program (1982) gavage study and estimated an upper-limit incremental unit risk
estimate of qj = 1.5 x 10~7 (fg/kg-day)"1, nearly the same as that of EPA
(qj = 1.56 x 10"7). Also, in its analysis EPA contracted with an independent
pathologist, Dr. Robert Squire, to provide a second examination of the liver
slides in the Kociba study. Even though Squire's analysis indicated more liver
10
-------
CO
UJ
1—
<=c
s:
>•— 1
i—
CO
LU
i*£
CO
t—t
ce
^—
z:
j 1 1 1
i a
a: o
UJ y
CL
CL 1 1 1
=> CU
cC
ct: h-
•—i CO
1 1 1 i—<
re i—
— i
z z
"-" Q
H- LU
•=C 1-^1
_J 1-1
— — * *"*"*
CJ eC
_J LU
=C 2:
i
1—1 UJ
m
CO I—
LU
1-1 CIS
O 2T
^ t— <
LU in
cf_
O
CO Q
^ c_>
0 h-
i — i i
C£ OO
> r--
>- CO
CQ «
CM
a
LU ct:
CO O
rr> u_
CO
rv"
•o
t—
O
U-
•
CM
LU
1
CQ
CU to
cu t_
4-
4- 4->
~o cr
tl-
Cr* 4-*
.,_
; 1 £T
O •"-
<1J 1—
4- 1
4- t-
UJ CU
CL
O.
3
re
4- cr
0 l-
o
CU 4-
ro i—
CO O
1 ^
^ — ,
c ^
^
>i
-a
to
'TO
^
•r—
d
•<
cu
u
cr
or
t-
cu
4- .
n
•i—
-o
*».
CO - ,
o •-•
to cr o
t_ re ,—i
QJ jr.
cr re
o oo jr
O OO 4-J
. CU
CU ' — to
oo to
o "> cu
-a -i- — i -
cu
en
re
>.
re
en
QJ -*°j
to -a
3 3
o -t-> -=c
'E to 2:
CU
•r—
— ^
cr
CO
re
.0
O
O
en
cr
-T3
cu
cu
4- >,
"O t^~
4_) r3 4-J '
(O -M O
£_ V) CO
fO
•r-
O
O
, «•— ^
4-J
C/) ^ — -
.,_ QJ
cn t-
O *r-
' ^ S-
_C t/j
4_3 '
n3'£-
Q_i O
(U
4->
ra
E
t/) O
O
OJ ~0
E n3
4-J 4— fO
I/) AO
=S TO •
••-> c\J
x <"*>
t- +J 3
Q >r_ O
i+— , — £_
n3 O>
c: t. QJ
CIJ o f>
E e-§
t/> >> 1
3 i — -C
•«-a' t_ en
"O rti *^~
ct QJ -C
5^
CD
. — l
C
n3
4-3
t/)
t/)
QJ
1
[_
QJ
>
_1
t_
QJ
>r—
— 1
t_
CU
>
_l
en
3
, —
*
t_
QJ
>
UJ
j_
O
3
4_>
4_
0
cr
O
0
QJ
t~~
QJ
CO
>>
C
0 .
>>
t —
o
>)
cr
o
to
QJ
re
^j.^
•*-* _Q
^J £_
i — 3
fO ,j
D_
| —
"D 03
t_ (/>
n3 03
w
OJ
CL
>^
+->
to
cu
to
re
ai
c_
o
CI
•r^
QJ
a
re
c.
3
00
— ^
cu
to
o
cu
I/)
o
~C3
!_
Q)
>
>r__
; 1
CU
to
O
-o
QJ
to
O
T3
QJ
00
O
-^3
t —
re
E
1
O
-(->
1
re
_E
cr
4- O O
O 3 4->
4-> 0
t_ re
O £- 4- -QJ
+J QJ OO
o jr: re 3
re +-> o
4- Q >, E
re QJ cu
f~ *\ r~
^-t-J -i_j -t-3
JD re
t- t. i-
QJ QJ O
+J > QJ 4-
re o jr;
E i * CO
•S CO r-l
4-> CO C_
to . O "<-
QJ LO 4— O
re
QJ
re
QJ
u
re
4-
£_
OO
I cr
cr o
CU 'I-
0 4->
cr re
0 <-
0 J->
i *
(— -
en
QJ
5
>^
-a
o
r~i
re
cu
i.
re
cu
0
re
4-
L_
00
cu
o
a1 .
re
•i—
3 :
cr
QJ
QJ re
to
0 >.
-a JD .
-a cu
cu +->
t. re
QJ E.
j * -^
CO 1 s
.,- to
cr QJ CM
• r—
E *" 4-
tr> cu o
re to
re t_
en QJ o
cr t- 4->
•i- 0 0
oo cu re
rD ~o 4—
-a
cu
t_
QJ
±1
to
., —
cr
E
T3
1 —
re
cr cu
• r— (j
1 C E -r-
C o <- **-
C_ r- Q) T—
QJ O 4J 4-> i_
>• C n3 O
••- O £_ 4-> ro
__J O 4-J fO (/)
"O
QJ
^_
QJ
-I-)
to
cr
•i—
E
T3
a:
1
a;
4J
t/1
c
•r—
E
5-
QJ
t_
3
O
£_
O
4—
•a
QJ
oo
3
QJ
to +-> ,
O -i— QJ l£>
LO _cr t- E to .
. 2 cu .-a o i~~
ro - — - > re -o CM
CO
i
CD
i — i
,X
tn CM
• p~-
i — ' • LT3
' ^
1
O
1— t
X
10
Lp •*
i — 1 to
re
TTJ
|
4- a>
rH t- O ^^.
i' . o en
>j 00
03 QJ -r- 4-
-0 oo C- O
1 O
-l-> i — en ~o •*-> co
•r— re -^ •»— +J
E 4-> ~^ cu E -i-
•^ cr en u ••- cr
r- 0)4- C -- 3
IE-—- QJ I
C_ QJ £_£-•"
CU C_ 4-> QJ QJVO
O. O -i— 4- CL 1
CL cr cr QJ CuCD
=> -r- 3 (S 3 rH
.
4— >
•i —
4—
to i-
QJ O
OO O
re. CL
QJ
1- O
U 4->
QJ
-a QJ
-t-J "O
cr
QJ TD
cr QJ
-5 -a
00 3
3 ( —
•r- 3 O
•o x
ro Q)
. r. CO
tO '1—
00 QJ
, — o
.re ~a
cr
en
ro -i—
JQ -Cr
• r—
o QJ
0 I-
±£ QJ
JC
-D S
QJ
OO 'I—
3 tO
'*— > **^
"O i —
ro re
cr c:
3 re
j= QJ
-f-3 *-
'3 =
cr
~C CO
QJ
t- -a
re CU
r*» | *
F— OO
O 3
u •<->
-o
r^^ re
• cr
t— 1 3
-i-> -c
3 +->
o •!-
JT3 S
re
4- QJ
O <-
re
i- CL
O P~
4—1 O
u <->
ro
i[ t^
•
ro CM
J2J O
oo c_
QJ O
00 4->
re <->
cu re
t_ 4-
o
cr re
>^
4J JT)
cr
Q) QJ
4_j re
oo E
3 •!-
•1-3 4->
^t QJ
O3
n
-------
tumors in the control and low- and mid-dose groups, the estimates of an incre-
mental increase in cancer risk differed by less than 10%.
Another factor of concern in the extrapolation procedure was nontumor
toxicity among the female rats. In order to correct for high early mortality
in the high-dose rats, EPA's analysis eliminated all animals that died during
the first year of the study (before the appearance of the first tumor). The
elimination of nine animals in the high-dose group, one in the controls, and
one in each of the other dose groups, changed the upper-limit estimates by a
factor of either +1.7 or 1/2.5 depending on which pathologist's analysis was
used. Since EPA was the only agency to make the adjustment, its estimate
incorporating both pathologists1 analyses actually decreased by a factor of 2.7
compared with the unadjusted analysis.
In selection of tumor types, all the agencies modeled the liver tumor
response. EPA also included the cancer response in the lung and hard palate/
nasal turbinates, but this led to only a minor increase in the final estimate
since the liver produced the major response.
Animal-to-man dose equivalence factors are discussed in the HAD. Both EPA
and the State of California used dose/surface area equivalences between animal
and humans. The FDA used dose/body weight which reduces human risk estimates
compared to surface area by a factor of 5.4 for rat-to-human extrapolation.
The CDC used liver concentration at terminal sacrifice, a measure that would
be preferable if human tissue distribution was also known. In the present
case, however, the known rat liver concentration measures of dose equivalence
had to be equated back to the rat administered dose without a comparable known
relationship in humans. The dose used for the curve fit by EPA, FDA, and the
State of California was the dose actually administered to the animals. The
CDC's use of liver concentrations at terminal sacrifice resulted in the risk
12
-------
estimate being increased by a factor of 2. Species conversion factors are
discussed further in section II.C.
The end result of all of these factors in the risk extrapolation procedure
resulted in a maximum difference of a factor of 9, that being between the EPA
and the FDA. This can be seen either by comparing the upper-limit incremental
unit risk estimates or the risk specific doses (RsDs), which are just the
reciprocals xlO~6.
C. ALLOMETRIC AND BODY BURDEN CONSIDERATIONS
The dose metric or allometric equivalence for rat-to-human extrapolation
has a potentially large quantitative impact on 2,3,7,8-TCDD risk estimation
because of the large differences in half-lives in the rat and human. However,
what little attention this topic has received from regulatory agencies until
now has taken into account only the standard dose metrics. As shown in Table 2,
both EPA and the State of California used the administered dose/surface area
conversion, the FDA applied an administered dose/body weight and the CDC applied
the actual rat liver concentration at terminal sacrifice. When extrapolating
from rat to human, use of the dose/surface area allometry increases the risk
estimate by a factor of 5.4 versus either dose/body weight or liver concen-
tration. When extrapolating from the smaller mouse to human, the corresponding
use of dose/surface area allometry results in a factor of 13 greater risk
estimate.
The EPA has used the dose/surface area metric, often called a species
extrapolation correction factor, as a conservative, prudent policy. It is based
on the observations that among different mammalian species'many physiological
rates, and especially ventilation, basal metabolic, and clearance rates tend to
scale in proportion to a fractional power of body weight. It has also been found
13
-------
to hold for the acute therapeutic effects of anticancer agents. That fractional
power, often between 0.6 and 0.8 is very close to the fractional power of 2/3
relating the surface area of cylindrical or round objects to their-volume.
Since the density of most mammalian bodies is about the same, mass or body
weight can be used instead of volume, and hence the term surface area or (body
weight)2/3 correction. In simplified terms the allometry of basal metabolism
is often explained by the observation that the amount of calories a warm blooded
animal will consume is enough to maintain body temperature and that loss of
heat is related to surface area and not mass. However, the allometry of species
conversion for carcinogen risk assessment is far more complicated than simple
basal metabolism. Even assuming that the basic mechanism of the carcinogen
stays the same from high- to low doses, there are often large species differences
both in tissue distribution and in metabolic pathways to form the active carcino-
gen. Often, it is not known whether the parent compound or one (or more) of
its metabolites is the active carcinogen. Almost never is there a good under-
standing of the mechanism.
It is just because of these many unknowns that regulatory agencies have
been forced to adopt a general default position of a surface area or body weight
or parts per million (ppm) in air species conversion factor. The EPA most
often uses surface area, but sometimes uses ppm in air as a species dose
equivalence, based on the known cross-species allometry for Q2 consumption.
The FDA position is to use dose/surface area allometry when the active carcinogen
is a metabolite of the administered compound and to use dose/body weight when
the active carcinogen is thought to be the administered compound itself. Their
reasoning for the latter case, of which they consider 2,3,7,8-TCDD an example,
is that if the administered compound does not have to be metabolized to be
carcinogenic, then strict dose/body weight considerations should apply. The
14
-------
CDC apparently agrees. The EPA counters, however, that-even if the parent
compound is the active carcinogen, its activity is related to its time in the
body, which in turn is related to clearance time and hence to dose/surface area
allometry.
At this point, it may be instructive to derive the dose/surface area allo-
metry in a slightly more rigorous manner. The rat-to-human species correction
factor of 5.4 means that if the human were to receive the dose/body weight of a
compound, 5.4 times that of the rat, the different elimination capacities of the
two species would cause both species' concentration x time exposures to the
compound to be equal. In terms of first-order elimination kinetics for a
single dose of a nonmetabolized compound, a rat given concentration C0 with an
elimination constant ke would have a total area under the concentration-time
curve of C0/ke. A human given a concentration of C0/5.4 would eliminate the
material, if allometric considerations hold, at a rate of ke/5.4, so that his
total area under the concentration-time curve would be equal to that of the rat.
For continuous daily exposure, the total area under the concentration-time
curve is (C/k^) (TK - 1 + e~kT)5 where C is the daily dose/body weight and T and k
are units of days and reciprocal days, respectively. If T is large, say 730 days
V
for a 2-year rat study and k is not very small, then this area becomes approxi-
mately CT/k. Thus, in order for the total areas'to be the same for a 2-year
rat and 70-year human dosing period, the human would have to be given a concen-
tration C/(5.4x35) or 1/189 that of the rat. EPA's position in this metric is
that one rat year is equivalent to 35 human years in the cancer development
process, and that the cancer age-distributions for rats and humans are alike
when T is viewed as representing a lifetime. Therefore, over a lifetime a
human should be allowed 35 .times the C/k that of the rat as an equivalent dose.
Since rats and humans seem to follow closely enough for most compounds,
15
-------
the clearance rate dose allometry discussed above, adjusting the species
correction factor for actual clearance rates, is not often done in risk extrapola-
tion. Furthermore, consideration of all the unknowns of actual mechanism and
metabolism make it apparent that the species correction factor is only a rough
approximation, meant to somehow convey the concept of increased sensitivity of
the human compared to the smaller animals. Still, the appropriateness of the
use of the surface area correction factor is not clear in the case of 2,3,7,8-TCDD
because of its extremely long half-life in humans compared to rats. The potential
effect of this difference on quantitative risk estimation is now examined.
Rose et al. (1976), examined the fate of 2,3,7,8-TCDD following single and
repeated oral doses to Sprague-Dawl ey rats. For a single oral dose of 1.0 ug
TCDD/kg body weight, they found a half-life, assuming a one compartment open
model, of 31 +_ 6 days. For repeated oral doses of 0.01, 0.1, or 1.0 ug TCDD/kg-
day, 5 days a week for 7 weeks, they found a half-life of 23.7 days. For the
single dose, after 22 days nearly all of the compound had concentrated in
either the fat or the liver, with equal concentrations in each. For the rats
administered the repeated oral doses, the compound again concentrated mostly in
the liver and fat, with the liver concentration being three to five times as
high as that of the fat at the end of 7 weeks. This observation is consistent
with that of Kociba et al. (1978) who, in his 2-year feeding study, found liver
concentrations three to five times as high as adipose tissue when the daily dose
was at least 0.01 ug/kg-day and about the same as the adipose tissue concentration
when the daily intake was 0.001 ug/kg-day. Rose et al. estimated the elimination
constants for liver, fat, and whole body all about equal, 0.026 days"1, 0.029
days"-'-, and 0.029 days"1, respectively,, corresponding to half-lives of 24 to 27
days. The relationship between half-life and clearance times for first-order
kinetics is ti/2 = In2/ke.
16
-------
In humans the half-life of 2,3,7,8-TCDD in the body has been variously
estimated as 3 to 5 years, 6 years, 10 years, and if 2,3,7,8-TCDD acts according
to two compartment kinetics with the fat acting as a "deep" second compartment,
up to 30 years (U.S. EPA, 1988). The CDC (1987) estimates a half-life of 6 to 10
years; this estimate will be used here.
An additional complication is that nonhuman primates, unlike rats, appa-
rently accumulate a higher concentration of 2,3,7,8-TCDD in the adipose tissue
than in the liver, with ratios ranging from 10:1 to 67:1. The very limited
human data also suggest an adipose tissue to liver concentration ratio of 10:1,
with minor deposition in other organs. One experiment exposing rats and both
infant and adult monkeys to a single intraperitoneal injection (400 ug TCDD/kg
body weight) found that after 7 days the rat had concentrated 43% of the admin-
istered dose in its liver versus only 10% and 4.5% for adult and infant monkeys.
In monkeys, the larger percentages were found in adipose tissue (U.S. EPA,
1985). Thus, if the liver is the organ of primary concern, for tissue distri-
butions alone a human would have to be given anywhere from 10 to 50 times the
dose on a mg/kg-body weight basis to have the same liver concentrations as the
rat. If one is not concerned with the liver alone but with total body burden,
it is not these figures but the relative body half-lives which would apply.
Estimates of the ratio of half-lives in the human versus the rat show that
for a human half-life of 2,190 to 3,650 days (6 to 10 years) and a rat half-life
of approximately 25 days, the ratios are 88:1 to 146:1, far higher than the
5.4:1 correction used by EPA. If liver is the focus and comparative liver
tissue distributions are factored in, however, the rat-to-human correction
factor ranges from 1.8(88/50) to 36.5(146/4).
The quantitative risk implications of these adjustments for tissue distri-
bution and half-life differences between the rat and the human are presented
17
-------
in Table 3. As can be seen, if total body burden (area under the time-concentra-
tion curve) only is considered, the very long half-life in the human leads to
risk estimates between 16 and 27 times that in EPA's HAD (1985). If liver
concentrations are considered however, the relative risks range from 0.3 to 6.8
times that of EPA's estimate. All estimates are higher than the FDA's. The
limited evidence suggests that if liver tissue concentration-time species
equivalence is correct, then the EPA HAD (1985) might underpredict the upper-limit
risk by a factor of 1.6 to 6.8.
D. OTHER EXTRAPOLATION MODELS
1. Longstreth and Hushon (1984)
Alternative models have been used for extrapolating to low-dose risk.
Three of these, the one-hit, the Probit, and the Weibull, have been discussed
and modeled in Appendix C of the HAD (U.S. EPA, 1985). The latter two, plus
two others, the Logit and the multihits have been modeled by Longstreth and
Hushon (1984), but they have not adjusted their data for high early nontumor-
related mortality. This has been done in Tables 4 and 5 for the Kociba and
Squire histopathology analyses, respectively. The resulting MLE and upper-limit
risk estimates for several low-dose levels are consistent for both data sets.
For both data sets the linearized multistage and one-hit models yielded
identical results. Also, for both data sets the upper-limit estimates based on
the multistage model were consistent, while those based on the other three
models varied considerably. For example, at a dose level of 10-5 ng/kg-day the
UCLs for the multistage model were 1.5 x 10-6 and 1.6 x 10-6 for the Kociba and
Squire pathology analyses, respectively. However, at the same dose level of
10~5 ng/kg-day, the UCLs for the Logit model varied by a factor of 29, from
2.0 x 10-6 for the Kociba pathology to 5.8 x 10-5 for the Squire pathology.
18
-------
1
UJ
0
o
2:
oo
UJ -Z.
C5 o
•< (-1
1- H-
oo rD
"— < CQ
— 1 (V
2: oo
t— H
Q,— ~
U-J
UJ
f^j i , i
H— < ~~ »
rf OO
UJ I— I
^ 1—
!m Q
UJ et
— ^
t— 00
UJ
0 >
2£ 1— <
OO 1
^3 U.
1
O «<
I— z
I ^»*
CM -1-
•1— (-
03 O)
f—
>, 0) "-
O ^- ^)
c w
Q. 03 eC
^ Q
•i— LU
| j
t/) ^3
O -M
O> 1
ij r^ fQ
C '•" "O
4->
r- 0
03 -l-> O)
O 03 £_
£_ £_
0
0
CO
cu
co
o
cu -o
to 3 03
co O ni
•!~ r~
CU •*->
• r- 4-> 03
| ^ O3
03 <*_ C
r— " O
CU t- T- c
o; cu +j 03
. > 03 =
•—-(-> n:
f~
Q)
O
c:
0
0
40
OJ 03
H- ceL
•i — ^ ^
i — CO
cji o? c:
i — "O ro
03 - — - E
nr
03 03 03
CM r— I to CO OO O 00
to "^ ' — ' O O I^» to
r-\ CM
OO «d- •* •=!- LO -cJ-OO
1 1 1 1 1 II
O O O O O O O
X X X X X XX
LO •=!- LO CM r-. CM i— 1
CM to CM < — I «=d" *^~ t — '
to O1 OO •* OO O to
1^ t-< 00 •* i— « to IO
OO CM <* CO
I— 1
c: c:
OJ r-t •— 1 <— ) •— 1 CO i— 1
~^3 •• •• •• •• "^3 * •
!_ r— 1 i— 1 LO LO £_ t— 1
--i ^3
ja ja
>> >>
-0 T3
O ' O
o «*^
0 0
03 •• •• •• •• 03 ••
O O
1— 1—
LO LO LO LO LO LO LO
CM CM CM CM CM CM CM
O O O O O O O
CT» O> CTi CT» CT» LO LO
t— 1 i— 1 t-( i-H i-H to to
CM CM CM CM CM OO OO
19
03
f^ <* LO O CM
CM r-l O r-1 O
•3- ^h LO "=d- LO
III 1 |
o o o o o
XXX XX
CM CM 00 IO LO
«3- CM r-. . .
t— 1 r— 1
tO OO <* O
•^- r--. CM LO <— i
.— i
cu
c o
t— 1 r— < t— 1 O ^
•I- CU
i— t LO LO 4J i —
O "3
CU >
£_ 13
0 cr
O OJ
O O 03 >>
t— I <^t- t— I CU 03
i. -a
3
CU JQ
O
03 CO
ti _^>^
t_ ' — ,
=5 C31
to E
LO LO LO Q. Q
CM CM CM LU U_
0 O 0
LO LO LO
to to to
OO CO OO
•
CO
O
•r~
i.
f<3
C
cu
o
CO
>}
r^
CU
~2£
•r—
1
0
E
•a
CU
cu
•o
•r—
CO
c.
o
o
03
-------
o
0
o
*T
CO
t*-
co* >^
•. 1—
CM t-H
_J
0 <
0 >- S
LU o z:
CO CO =3 CO
O I— 1— _l >•
O. «£ CO UJ —1
x o: o o;
LU CD o <:
S-ZZSUJ
CO UJ r-t
-<
1— CD O r-«J—
~
ac o -a
o as Q Q <
— a o LU 2s
LU S»
U>"O
o o
— 1 E
•r— r—
1 JQ OJ
CJ> O "O
'O S- O
_J Q.E
(^M
r— JQ
.a "ai
ai o
3 E.
I %
.4. 0) r^ |£'J2.
•P cn CU: i G)
r- re -a (L-o
3 4-> O C O
E to E 0 E
CO g> 1 >>
£§-— 5*-0
p^
LO
1
O
*x
r*^.
c^
vo
i
o
"x
e
t-H
Is***
o
r-H
X
r-1
CO
O
LO
1
O
r-H
X
CO
, e
r-H
LO
1
0
r-H
X
t— 1
- r4
LO
1
O
I-H
LO
o
r—1
X
LO
g
CM
CM
i-H
1
O
i— i
X
o
ft
CO
^^
i •
O'
X
co
LO
LO
1
o
1— 1
X
LO
ft
r-H
VO
t-H
X
LO
^j!
co
r-H
i— 1
X
CM
«
r-H
«^>
O
T-H
X
r-H
ft
r-H
LO
O .
I-H
X
t-H
0
t-H
,3.
cb
•-1
«*
|
O
r-l
X
o
CO
oo
1
o
*x
LO
(
r-.
CO
i
o
r—1
X
o>
o
CM
.*>^
1
f-^
•x
LO
I-H
LO
1
O
r-H
X
LO
IO
CT^
1
O
r— 1
X
CTl
^»
^^
O
r— 1
X
l-H
f^*.
^J-
O
r-H
X
I-H
l-H
CO
1
o
r-H
CO
1
o
r-H
X
e •
CO
^^
1
CD
r-1
X
LO
ft
i»j
CM
CD
t— 1
X
LO
r-H
CO
|
CD
i-H
X
LO
r-H
^^
cb
r-1
X
^^
o?
LO
o
r—H
X
r***
•
r-H
CO
O
i— 1
X
LO
9
**j" .'
CO
§
I-H
- 9
r-1
CM
cb
t— i
CM
|
O
r— 1
X
to
9
CO
CM
1
O
1— 1
X
CO
e
CM
CM
1
O
r-H
X
CM
r-.
CM
1
o
i-H
X
LO
r-H
CM
1
O
r-1
X
CO
i-H
CO
o
r—1
X
CM
LO
CM
O
i-H
X
cr>
CM
CM
i
o
r-H
X
r-H
e
r-t
r-t
O
i-H
r-1
1
O
I-H
X
00
e
CM
i-H
1
O
I—I
X
i-H
ft
CO
r—1
|
O
1— 1
X
o
ft
co
r-l
1
o
i-H
X
LO
I—I
r-H
cb
i-H
X
r*^
l-H
i-H
cb
r-t
X
p^-
ft
i-H
r—1
cb
1— 1
X
f—.
ft
r-H
r-t
cb
r-t
X
I-H
»
r-H
r—1
< ' : ' , '
'.•"'.
• * ', ' "
vp ' • , o . s-
'e • . ^J- "r— •
, co • ••>*• ... re •
• ' r-H' '"NT . :,• . • -
' CO C
••r*. •.
vo CO >E::
00 ' •>;}* "^^*
. .. >^. ... Ol
t-H r-1 CO 3
' r-t ' '
V O .
. . ' ft •
,
tO TD ••
0) >> 1 T3 «•
s_ .0 cn E ">
o c 3 re t-
•r- •!- > X 0>
+j E at >
c t-
-O £ CLI
0) 4J 3 0
> «• c c c:
re-s<£ a) -^ cu
rj\ . r— < ^ CO r—
V- re c_ re
co o > a . >
"53 CO 333
-o 3 cr -P Cr
O r— CU O)
E > c "re c:
5 cu E '^ E
SjE 5 5 5
re ja
20
-------
O
Q
0
I —
co
^ . s
CO >-
«. h-
CM i— '
1
1— H-
d >- O
UJ Q 2:
OO CO ^> OO
O I — 1 — — 1 >-
Q_ - "Zi y UJ
OO UJ ' — '
2= _l Q h- OC
Q U- C£. „
:c i uj CD
MI • "i i -^^
LLJ • LL- * —
o zo o u_ i— i
1— CD CD I— < t—
^C Ci oo
OO Q- =C Qi "~3
>— < OO O ZD Q
o: i— ' o «*
UJ g U_
OO ear ^ 21 OO
o 2: o o i-1
Q UJ Qi OO
1 U_ 3 U, >-
30—1
CD 2: Q o -=C
-JO UJ ==
— ^ ^T= "^ ^^
LJL. tji-i —i— >— i
O UJ h- C£ U-"
oo ui o;
OO — '
1— CQ C£. 0-
«C u_ oo
i — • •
OO «C
UJ (—
Q
UJ
1
OQ
i
r^
o
•r— CO
r— ^/
•I— CO
I— *r—
f_
CU
O i —
c: ro
CU c
-0 0
C 4J
C •!-
o -a
o -a
ro
£_
CU
o.
5-
•!-> r-
•i— CU
c?^
—1 E
. %
S'aJ
en o ~a
o t- o
_l Q. E
p— '
13 i —
£-S
cu o
3 E
+J
•i- cu i— jz: •—
i — ro "O CU ~O
3 -4-> O c O
2: CO E-0 E
q
O
CO
CU
03
E
, ^
CO
CU co
•a co
o ••-
o t-
r— ro
OJ 4->
jv> SX
•r- CU
1
X.
ro
2!
4-3 t—"*
.^- CU
00
—1 E
£.-
1 JO 3J
cn o "a
o t- o
_J Q- E
_o ^
•r- -O
cu o
^J
1 ~^^ T- ro
•r- cu r-* jz: i —
r- ro T3 CU T3
3 4-> O c O
2: co E 0 E
Si O> 1 '>i
Q -—-^-0
LO
O
X
OO
LO
0
r— 1
1
0
1—1
X
«^
CO
o
t— i
X
CO
_;
IO
1
o
I— 1
X
t 1
LO
0
^.! |
X
o
I— 1
CM
1 — \
CD
i—l
X
CM
CM
«=J-
1
0
1
X
CD
CO
to
CD
r— 1
X
1—1
*
LO
o
i — i
•
o
r— )
X
Q
**•
p^
O
1-1
X
1— 1
I— 1
CO
o
1— 1
X
^
1
o
I— 1
X
10
r~H
LO
CD
,—1
X
CO
CO*
cn
o
r— i
X
10
CO
CD
,_4
X
CM
i— I
LO
cb
t~~l
X
i— i
.•
^j.
o
I— 1
co
i
o
, — 1
X
cn
t
CM
LO
1
O
I— 1
X
co
LO
CM
o
I— 1
X
LO
I— 1
1
o
r—t
X
10
1-1
1
0
X
*°"
->
4J CO ^-~. t-H
,— cu >>
3 -Q ro
CO ~O
££> cn "S c
-^ c o
re cu cn E <"
O c: 3 ro t-
•r- T- X CU
4-3 E CU >
nj CU CO £_ O
•^ 4J> O CU O
•r— CU "O -Ct
•a E cu
CU 4-> 3 0
rO -^ CU ~-^ CD
cn ^- co i—
i|- re £- ro
oo O > O >
"flj CU 3 3 3
-O 3 cr 4-3 cr
O i— CU OJ
E ro r—
> c: ro c:
r— ro E ^
+-> CD E '"- E
o -c 3 c: 3
OQ 1— . 3:
-------
Also, at the same dose level, the UCL from the Weibull model varied by a factor
of 13, while those based on the log Probit model varied by a factor of over 50
million. In general, the UCL risk estimates based on these models exhibit the
relationship:
Weibull > Logit > linearized multistage = one-hit » log Probit
All of these models fit the data satisfactorily in the animal experimental
range, yet as can be seen from Tables 4 and 5, the estimates can vary over
several orders of magnitude at lower environmental doses. The choice of the
model must rely on factors other than goodness of fit.
2. Sielken (1987)
A second risk modeling analysis of the female rat liver response in the
Kociba et al. 2-year feeding study has been published by Sielken (1987), whose
arguments have also been reproduced by Paustenbauch et al. (1986). Sielken
fits the multistage model to the data but focuses on the large difference in
low-dose behavior depending on whether or not the high experimental animal dose
of 0.1 ug/kg-day is included. Si el ken's measure of risk is the VSD which he
defines in terms of 10~6 lifetime incremental risk. The VSD is derived from an
extrapolation model of the MLE of the multistage model and not from the upper-
limit estimate. Sielken, thus, uses Crump's reparameterization of the multistage
model but not Crump's linearized form.
Si el ken's analysis is based on the argument that use of the multistage
model with the 0.1 ug/kg-day dose-response included forces the estimate of the
linear term to be positive non-zero and distorts the true shape of the dose-
response curve.
In particular, he argues, even though the multistage model fits the data
with the high-dose point included, "the (resulting) fitted models do NOT [his
22
-------
emphasis] reflect the observed behavior at the lower experimental doses." A
proper low-dose shape, he claims, is seen when the highest dose is removed. In
this case, the linear term of the multistage becomes zero and the quadratic
term becomes positive. Extrapolation with the quadratic curve goes rapidly to
zero compared to low-dose extrapolation with a linear model.
An examination of the results derived from the type of analysis suggested
by Sielken can be seen in Table 6. In this table, both the MLE and 95% lower-
bound estimates of the VSD are calculated using different permutations of the
Kociba female rat data (animals dying before the appearance of the first tumor
have been eliminated). In every case, the model with estimated parameters fit
the data satisfactorily. The first row contains the observed liver tumors for
all doses and parameter estimates, while the second and Tower rows omit the
high-dose group. The third through the seventh rows permutate the low-dose data
by increments of one tumor-bearing animal. The proportion of 6/48 represents
the 95% upper-limit on the observed proportion of 3 tumor-bearing animals out
of 48.
An examination of Table 6 shows the instability of the MLEs under this
model. As was pointed out above, omitting the high-dose group changes the form
of the model from linear to quadratic, with a corresponding 330-fold increase
in the VSD from 4.8 x 10~8 to 1.6 x 10'5. The form of the model remains quad-
ratic until the number of tumor-bearing animals is incremented by 2. However,
when the increment becomes 3, to total 6 out of 48, which is the 95% upper limit
of the actual observed response, the picture again changes. When this 95%
upper-limit of the observed low-dose response is fit, the model again incorpor-
ates a linear MLE, and the MLE of the VSD returns to 5.0 x 10"8.
In contrast to the instability of the MLEs, the 95% lower limits of the
VSDs remain quite stable over the range of permutations, varying by a factor of
23
-------
to
CO
_j
LU
Cc£ Q
O o
"J- s
to LU
a CD
CO eC
a £
-
CD
f~>
1
CD
7T"
i—
«=C
Q.
O
(—
CO
t— 4
zc
CO
o
o
03
•a
01
^^
C31
3
Q
co
to
eu
03
CO
eu
c_
eu
4-^
eu
E
C-
to
Q.
1
LU
s!
o
40
03
eu
o ••"
1 E
•i —
SS. i—
LO
en
LU
I
^r
CM
|
.^^
O >>
•r- 03
4-> -a
03 1
£_ O)
•a -^
ro "^-
cy ^5
^^_^,
V
c, f^
03 "O
eu i
c: en
Jj -^
CT>
3
N — ^
i— 1
CD
r^
0
0
t— 1
CD
O
0
0
00 O
. o
.-1 OO
CO CO
CD CD
1— ) I— 1
X X
LO 0
CO CM
CO LO
o o
T— 1 t— 1
X X
CO LO
• •
— 1
X 0
1-H
CM
-a
eu
CD 4J
«d- 4->
CO CD
co co
•=1- •*
OO CO
t— i i— <
OO CO
CO CO
LO LO
co co
en en
CD
O
en
co
CD
t— i
X
en
i— i
LO
o
.— i
X
i —
•
i— i
CO
cb
1— 1
X
r-.
co
o
•^3
eu
.f_>
4->
e§
CO
sf
OO
1
CO
*3~
LO
co
en
o
LO
en
CO
CD
i— i
X
CO
I— 1
LO
CD
r-H
X
•
<— 1
CO
CD
T— 1
X
OO
p
-a
eu
40
4->
<§
CO
•=1-
00
I— )
oo
LO
LO
co
en
CO
•
CM
CO
CD
<— 1
X
CO
T— f
CO
o
t— 1
X
o
•
LO
CO
cb
I— 1
X
t— 4
r— 1
cb
t-H
X
CD
CM
-a
eu
4^
•i-5
<§
00
«^-
co
r-H
JD
. co
LO
LO
co
^"-^
en
24
LO
•
i— 1
OO
CD
r- 1
X
t-H
co
CD
t—l
X
CO
•
CM
O
r-l
fe
i— 1
X
^
OO
•a
eu
4_)
4->
cf
co
«^-
oo
•— 1
o
co
^j-
r~.
LO
00
*^-*^
en
CO
^
CO
^~-
o
to
c:
o
Q.
d)
^>
T3
eu
^ ,
eu
to
jQ
0
el)
_£n
J "^
„ *+—
•r- 0
£_
O 4-*
4_> *r—
0 E
03 **""
4- r—
CO 1
*^> ^.
4_> eu
03 Q-
to a.
3
to
cu LO
to en
03 eu .
4*^ *^
03 4->
T3
0
eu 4-^
j^;
4J CO
-a
4-> c
•r- O
t|_ Q.
CO
to eu
. £_
0) «-
-o o
0 0
E
CO
i — *=4"
i— ""^
*-C LO
o3 f*>
CO
5!!,
CO
^t^
0
eu
to
o
a.
to
eu
t-
-o
eu
>
c_
eu
CO
-Q
O
eu
.c
4J
4-
O
4^
>r~
E
•1—
1—
1
C_
CU
OL
CX
3
0
en
en
eu
-C
4>^
0
to
•^j
c.
o
a.
to
0)
t_
o
o
OO
«^J-
"*^^
r-.
o
-------
2. The ratios of the VSDs to their lower limits reflect the instability of the
MLEs for low-dose extrapolation with the multistage model.
Si el ken's results are consistent with the findings of Portier and Hoel
(1983), who concluded that for a multistage model fit of the data, the MLE of
the linear term is usually multimodel, the number of modes being equal to the
degree of the chosen polynomial. This instability of the MLEs from the multistage
family of models has also been confirmed by several other authors; it is further
seen in Tab! e 6.
Crump (1987) provided a critical review of Sielken's analysis and observed
that even if "the true shape of the dose response is a straight line connecting
the background response and the response at the mid-dose," the probability of an
MLE estimate of zero for the linear term in the multistage model is about 1/3.
He concluded that while the data are consistent with Sielken's interpretation
of a higher RsD, they are also consistent with those much lower, as displayed in
the confidence limits.
It was based on these types of analyses and a consideration of the typical
animal bioassay data to be fit that Crump had originally advised that an upper-
limit estimate of the low-dose risk be used for his model. Under his reparameter-
ization of the multistage model, an MLE of zero for the linear term becomes
possible, and this can cause great instability in low-dose risk estimation.
However, under the original development of the model, it is not possible to have
higher degree polynomials without having a positive linear or first-stage term.
Crump's reparameterization is necessary with quantal data in order to limit the
number of parameters. In doing this, however, the MLEs can become unstable, as
is seen above.
25
-------
E. TIME-TO-TUMOR ANALYSES
An extension of the multistage model analysis can be conducted when time
to observation of the tumors is known. This extension is modeled into the
formulation of the multistage model:
P(d,t) = l-exp-(q0
q2d
q3d3tk)
where qg, qi , q£ , Q3 » and k are parameters estimated from the data and are all
constrained to be nonnegative. This model is often called the Weibull model,
but is more appropriately described as the multistage Weibull , since it is
multistage in dose but Weibull in time. Its generalization over the multistage
model allows an estimation of the probability of cancer by a fixed age in the
absence of any competing risks. Its superiority over the quanta! or multistage
form is in its ability to adjust for treatment group differences in nontumor-
related mortality, as is seen in the Kociba study. However, the analysis
requires a pathology decision as to whether the tumors of interest were fatal
or incidental, a condition not available in the Kociba study pathology.
The results of a multistage Weibull model are presented in Table 7 and are
compared with two quanta! analyses using the linearized multistage model . In
the first quantal approach no adjustment is made for the high early mortality
in the high-dose group. In the second approach all animals dying before the
appearance of the first liver tumor were dropped from the analysis (see section
II. B). These two analyses are compared with the multistage Weibull under the
extremes of either all fatal or all incidental tumors.
As can be seen from Table 7, the largest difference in risk estimates is
between the unadjusted analysis and the fatal tumor for this analysis; 8.4
(for the MLE term) and 10 (for the upper limit). However, both these extremes
are thought to be somewhat misrepresentative of the data, and either of the
26
-------
Qi
0
Lu
CO -— »
I— OO LU
z: — I Q
=5 O
•-3 o; t—
O <£ 2Z
•=C o LU
— l
c£ a: =C
0 0 >
S •=£
3 oo =>
H- LU o-—
1 _l LU >-
o r> CD
H- Q 5 0
1 0 < — 1
LU 52; s o
s: =33:
•-1 o ac i—
h- i— i <=C
1 — 03 O-
00 OO — J
ID cC LU cc
>-< a. CD S
Qi o s: o
i— i
1— _ h-
ED — ' E>
oo i— s:
LU cC
cc o; LU
~T~
LU H-
. 1
r~^ eC Q
s: ^
LU LU «3t
— 1 U_
CQ
}
03
T3
' 1
cn
VX i—
01 _J
OTO CJ>
QL 0 13
>. -• ^
1 lOJ
cr cn
O r—
•r— H3
^ E LU
4- _J
O) 4 — y
O O
O
i_
o3
O)
C
1
1—
0)
-a
^*"
to r>»
in o
•ZL -ZL r-I r~!
^-i
LO ^f" «^J~ ^J"
II II
CD O O CD
i — 1 i — 1 t — 1 i — 1
XX XX
O LO <— " tO
•• • • •
cy* «— i co r*^.
to <=*• •* •=3-
i i ii
00 00
1— 1 t— 1 I— 1 I— 1
XX XX
tO ,-! i-H IO
to i— 1 CM tO
4.9
c: o
O) +J
E
4-5 cn
to c o
3 v- CO
•r-5 > t- •!-
"O T- o CO
§«••» >^
r— £— r~~
03 o :r> +j cu O 03
4-3 C CO ~O E ^~
CI 4-> O i3 03
03 " co co E -t-*
rs t/> i — t_ t-
cr c — ro -i— i — «-- o
1 03 g: 4_ i — 03 g
r— E »r— *"s 4«3 5
cu -t- c <4- JQ c: 4->
-o c 03 o ••- o>
O O3 CU ~O r~"
E >, CU 3 •!- 03
r- ,-? £ ' , « 4->
cu r— c •!= cu e= to
CD •-•C C3 | ^ CT) (™* 1 1
03 03
4_> 4_>
CO *> • • CO • •
•r- 'i-l CM -r- ,—1 CM
4-^ 4-^
i — r—
H3 n3
2: s:
_^*
•^
f™)
cn
c
•r-*
t
Q.
•r—
4->
^3
p:
>^
-Q
CM
OO
1
1
CQ
i — i
LU
2
E
O
4-
co
-(-•}
^3
CO
CU
£_
Q
CO "O
cu cu
CO >
To cu
C T3
O3
O)
, t_
i— cu
03 s
C- i^. |
o cr
4-
TD
OJ 03
CO
3 i-H
CT
i-_
CU 5-
-a o
0 «f-
E
c:
cu S
cn 0
03 ^:
+J CO
CO
i co
O) CU
0) =
C_ r—
-C «3
1— =»
O3 f^
•
CU
/^
03
O
t
Q.
D-
(O
_!_>
O
II
-------
middle two approaches is considered superior. While the present EPA analysis
(see section II.B) yields the lower of these estimates by a factor of 2, this
difference is considered minor.
III. TREATMENT AS A PROMOTER
Evidence for the promoting action of 2,3,7,8-TCDD in the rat liver has
been well documented in the HAD (U.S. EPA, 1985). This section compares the
quantitative cancer risk estimates under the assumption that 2,3,7,8-TCDD's
action on the liver is one of promotion. It is shown that even when treated
solely as a promoter the estimates of risk can vary greatly according to assump-
tions that one is prepared to make. In section III.A. the treatment of 2,3,7,8-
TCDD as a classical toxicant or promoter with a threshold is presented, along
with the results of Canadian and several European regulatory agencies who
estimate "virtually safe" levels by applying uncertainty factors to dose levels
at which no adverse effects are observed. In section III.B. the results of a
new approach to modeling for promoters are discussed in which cancer response
is modeled as a function of liver cell proliferation of initiated cells.
A. UNCERTAINTY FACTOR APPROACH
Several countries and the State of New York have estimated RsDs for
2,3,7,8-TCDD's potency by the application of uncertainty factors to NOELs,
NOAELs , or LOAELs . The general approach is to use uncertainty factors ranging
from 10 to 1,000 based on a rul e-of-thumb approach (Cook and Page, 1986).
• A factor of 10 where adequate chronic human toxicity data as well as
adequate chronic oral toxicity data in more than one animal species are
available;
28
-------
A factor of lO'O where adequate chronic animal toxicity data are avail-
,-
able in more than one species, but human toxicity data,are lacking; and
• A factor of 1,000 where limited chronic animal toxicity data are avail-
able (in only one species) or inconclusive results in more than one
species.
The National Academy of Sciences Safe Drinking Water Committee was more
explicit with regard to carcinogenicity studies (NAS, .1977).
"1. Valid experimental results from studies on prolonged ingestion by man
with no indication of carcinogenicity.
Uncertainty Factor = 0
2. Experimental results of studies of human ingestion not available or
scanty (e.g., acute exposure only). Valid results of long-term feeding
studies on experimental animals or in the absence of human studies,
valid animal studies on one or more species. No indication of carcino-
genicity.
Uncertainty Factor'= 100
3. No long-term or acute human data. Scanty results on experimental
animals. No indication of carcinogenicity.
Uncertainty Factor = 1,000"
Others have sought to break down uncertainty factors into components. Weil
(1972) interpreted the application of the uncertainty factor of 100 as a product
of a factor of 10 to account for differential human sensitivities and the
second factor of 10 to extrapolate results from animals to humans. When cancer
became the end point of concern, others included a third factor of 10 to raise
the total to 1,000. The uncertainty factors, the toxic end points, and the
RsDs used by several" agencies for the evaluation of 2,3,7,8-TCDD are presented
in Table 8.' The table includes results from:the agencies that have used the
29
-------
0
Q
0
•7
CO
"^
co
CM
e£ co
O UJ
II ,_,
CO
co -z.
UJ UJ
!i<
2x
CO O
UJ (_
O _]
ss
o; c£.
>• -z.
CO CD
UJ UJ
H- o;
po
O. 1 1
0 Q
rf z
u_ co
0 =3
_ CD
0 S
CO CC
»-« >
a:
J^f* I 1
§lfe
s:
o
o
CO
UJ
1
CO
«=c
\—
1
re >>
E ^3
c -P
;
re
o ~a
•i— i
^s /"^ f^ ,<^
•r_ *r^> »|— «f—
O i— < O O O
O CO O O O
re
co co co co co
s s: • s s: s
i i i i i
CO CU CU CU CO
o o o o o
c c: c c: c
re re re re re
O CO O CO O
CD VO CO «-H O
O CT> CM t— 1 «-H
t— 1
CO CO CM CM CM
II 1 1 1
O CD O O O
t— I i— ) t— 1 <— 1 r- 1
X X X X X
«3- r-~ co r^ UD
10 <& CM m 10
f—
o
c_
re
cu
cu
re . cS
•f""
(_ tJ
o c:
^— o
-i— •»—
eC r— CO •=£ +J
o_ re Q Q re
uj co «o u_ z:
^^^.^
re
re
c:
re
o
i
•r~
O
C
o
CO
-a
c=
re
re >,
.a re
o t_
0 3
>^ ^^
o
4J O
C r-H
re
i * ^
t- o
CU JJ
0 0
c: re
^D l^-
t/i
1 O
o cu
t- »»-
CO CU
c_
^-. CO
i~ >
CU T-
O +J
c o
re 3
co -a
,_)_
vo
CD
O
•
I— 1
1 >^
c:
re
E
0
T3
re
2 >>
-Q re
•r- <_
o t_
0 3
>^X "^"
, — .
>^ C^
-!-> O
«= LO
re
-!-> t-
«- o
CO 4J
0 0
C. CO
^3 tf—
e/>
1 O
o a)
t L ll
CLlt-
CO CO
t_
•~>. cu
c3>
O 4-3
c o
re 3
0 -0
CM
CO
0
o
•
CM
CU
4_3
re
co
NX
f
O
s
CO
•z.
re
,^_
o
o
^
-— ,
->> CD
4J LO
C CM
re
tf 0
CU 4J
o o
^D M —
cu
O
C
re
o
VO
i— <
C3
o
•
**
TD
c:
re
11
cu
r-
CU
ai
f—
-o
c
re
2 >>
-Q re
•r- 5_
O £_
O 3
^ S
^-,
>^ c""j
4J O
•^ d-
re
4-> t-
t- o
CU -t_>
0 0
c re
13 <)-
tn
I U
O CO
f- M-
cm_
CO CU
^ CO
CU -1-
0 4^
c o
re 3
co -o
vO
0
CD
o
•
o
,— I
^-^
10
T3
re
c:
re
"O O
c: s ^
re
cu
•p re
r— M-
cu "co
^ 2
^— ^
O
•r™
C.
re
+->
£1
O
0
10
"re
•• —
-a
c
re
re >,
f^( fO
•r™* t^ ,
O £-
O 3
v^x
CO *r-
O 4->
C O
re 3
o -a
i
i
t
i
-o
c~
re
r—
£_
CU
IS1
•r-
S
co
r—
re
c:
0
•r—
-t-5
•r—
T3
•a
re
a>
•r—
O
D.
-P
O
C
"O
•r-
"O
+_,
US
•°
"
'^
re
-a
i
cn
^
cn
c
i— {
^*~
o
I
• ' LaJ
O
. Z.
"cu re
"O
O CO
E ^
•f—
CU E
cn t~
re cu
(/) CU
•i— "O
' i— O
3 -P
E
•^
"O "O
CU 3
N 4-5
•r- IO
£_
re re
CU -Q
C •!-
•i- O
r- O
sx
CO
t- CU
4-5 c~
4->
II
T3
CO CO
s: to
_J ZD
re _o
•
L_
O
4_J
O
re
n-
>5
4-5
C.
•i —
re
-P
CO
o
3
30
-------
linearized multistage model as a comparison to those agencies using the uncertainty
factor approach. As can be seen, the RsDs generally segregate into two groups
with those in the lower potency groups all using the NOEL approach. The State
of New York, as well as the countries of Canada (plus the Province of Ontario),
Switzerland, and Germany, consider the cancer study of Kociba to establish a
NOEL at 1 ng/kg-day. The Netherlands also uses the Kociba study as its per-
tinent cancer study, but considers instead the 1 g/kg-day dose as a non-noxious
dose due to the slight increase in liver cell changes in the female rats at
this level. All the agencies, except Switzerland, chose the uncertainty factor
approach to establish an RsD. Switzerland actually estimated inhalation and
oral exposure and divided those exposures into the 1 ng/kg-day NOEL to determine
a "safety factor" attributable to those exposures.
As can be seen in Table 8, potency ratios seem to gather in factors of 10.
The RsDs for the five agencies above the dotted line span one order of magnitude
as do those for the agencies below the dotted line. Separating the higher- and
lower-potency groups is a factor of 15.6 (10 * 0.64). The total potency range
is 1,564. All but one agency uses the Kociba data.
B. MODELING AS A PROMOTER UNDER THE MOOLGAVKAR, VENZON, AND KNUDSON
TWO-STAGE MODEL •.
A variation of the multistage model has been developed by Moolgavkar,
Venzon and Knudson (M-K-V, 1979, 1981) which models cancer as a. two-stage process
with a promotion phase. This model has been shown to predict very well tumor
promotion in the mouse skin (Chu et al., 1987). A variation of this model has
been developed by Dr. Todd Thorslund to extrapolate from animals to provide
human risk estimates of liver cancer deaths (see the Appendix to this paper).
The model considers that the carcinogenic action of 2,3,7,8-TCDD is through its
31
-------
dose-response relationship on the proliferation of initiated liver cells. By
including what is known about the receptor-mediated mechanism involved, cell
proliferation is itself considered a function of 2,3,7,8-TCDD's binding capacity,
which can be shown to follow linear but saturable kinetics. When these parameters
are factored into the model, the cumulative probability function, P(x) for dose
x at a fixed time t, becomes
P(x) = 1-exp-M CexpG(x)t-l-G(x)t]/62(x)
6(x) = 6(0) + [G(co)-G(0)][l-exp-Vx]
where P(x) = the probability of a tumor with dose*,
M = the background mutation rate proportional to background tumor rate,
G(x) = the liver cell proliferation rate of initiated cells associated
with dose x,
6(0) = the background liver cell proliferation rate,
G(«>) = the maximum cell proliferation rate possible, and
V = the parameter associated with the liver saturable kinetics of
2,3,7,8-TCDD-receptor binding.
Even though this model is termed a two-stage model, conceptually it is rad-
ically different from the multistage model in several respects. The multistage
model, as computed by EPA, is a basic curve-fitting model with all parameters
estimated from the data. The two-stage model with promoti'on as constructed can
actually be fit without the estimation of any parameter from the cancer bioassay
dose-response data. For example, M and 6(0) can be estimated entirely from the
control data, V can be estimated from a separate experiment measuring 2,3,7,8-
TCDD uptake by the receptor, and 6(°°) can be estimated by the incorporation
of thymidine into the nuclei following administration of a saturation level of
2,3,7,8-TCDD. In theory, then, with the exception of the use of control animals
for the estimation of background rates, all parameters would be estimated
32
-------
separately and a goodness-of-fit test would be a good measure of how well the
model fits the actual data.
An extension of the above distinction between the multistage model and
M-V-K two-stage model is that the parameters derived from a bioassay based on
one rat strain can be used to predict the response from a second strain with
only the background rate M having to be estimated from- the control group data
of the second strain. This procedure was extended fr,om animal-to-hum an extrapo-
lation where human liver cancer death rates in the United States from 1980 were
used to estimate the background human rates for M and G(0), and the values of
the other parameters estimated from the rat data were used to provide human
risk estimates.
In addition to modeling liver cell proliferation as a nearly linear func-
tion of 2,3,7,8-TCDD receptor binding (called here, the negative exponential
form of the M-V-K model), a second form of the model 'results when the 2,3,7,8-
TCDD-induced cell proliferation rate'is assumed'proportional to the product of
cellular 2,3,7,8-TCDD levels and the number of 2,3,7,8-TCDD receptors. This
results in a model for the induced cell proliferation rate which is log-logistic
in form.
Both models are used to extrapolate from the animal to human cancer response.
The results are presented in Table 9 which is reproduced from the Appendix.
While both forms of the M-V-K model fit the observed data quite well, and both
can be justified on theoretical and some experimental ground, their use for low-
dose extrapolation leads to a wide variation in risk estimates. For a lifetime
daily dose of 0.1 ng/kg-day, or one order of magnitude below the animal experi-
mental level, the estimates vary by a factor of more than 10,000. Furthermore,
even though the negative exponential form of the model results in a .linear low
dose-response relationship, the risk estimates are 2 orders of magnitude below
33
-------
TABLE 9. ESTIMATES OF LOW-DOSE INCREMENTAL RISK TO HUMANS EXPOSED TO
2,3,7,8-TCDD BASED ON FEMALE SPRAGUE-DAWLEY RATS3
COMPARISON OF THE TWO FORMS OF THE TWO-STAGE MODEL WITH PROMOTION WITH EPA's
RISK EXTRAPOLATION USING THE LINEARIZED MULTISTAGE MODEL3
Dose
(ng/kg-day)
ID'5
10-4
lO-3
10-2
10-1
1
Promotion'3
Negative
exponential
1.7 x 10-8
1.7 x 10-7
1.7 x ID'6
1.7 x 10-5
1.8 x 10-4
2.4 x lO-3
Log-
logistic
—
--
<10-13
8.8x10-10
8.8x10-7
9.3xlO-4
Linearized multistage model
EPA
(Upper confidence limit)
1.6 x 10-6
1.6 x 10-5
1.6 x 10-4
1.6 x 10-3
1.6 x 10-2
1.6 x 10-1
aTaken from Table 11 of the Appendix.
^Squire's pathology analysis adjusting for early mortal ity.
34
-------
the upper confidence limit obtained by EPA, which modeled 2,3,7,8-TCDD as a
complete carcinogen. This difference between the estimates with the linearized
multistage model and negative exponential fo'rm of -the promoter model is due
primarily to the low human background liver cancer rates compared to those of
the female rat. However, these estimates pertain only to human liver cancer.
In its present form, the model is target organ-specific from animals to humans.
Estimates with the negative exponential form of this promoter model might
be considered as providing a conservative, on the high side, estimate of
induced liver cancer risk, since cell proliferation is modeled as a linear
function of dose. However, not enough research has been done on the low-dose
properties of these models to characterize the variability of the low-dose risk
estimates. While the upper-limit estimates should certainly be below those of
the linearized multistage model, further research needs to be done before these
models can be used for regulatory decision-making.
IV. COMPARISON OF ANIMAL PREDICTION WITH ACTUAL HUMAN DATA
With the exception of one study on 2,3,7,8-TCDD in Holmesburg, Pennsylvania,
in 1967, all human exposure data are derived from accidental exposures of
unknown quantity. In the Holmesburg study, 2,3,7,8-TCDD was topically applied
to volunteer prisoners in 'total doses ranging from 0.4 ug to 7,500 ug. According
to testimony and exhibits in EPA's 2,4,5-T cancellation hearings (Rowe, 1980),
doses below 16 ug did'not el'icft a chlo'rac'ne response, while a dose of 7,500 ug
did cause chloracne in 8 out of 10 subjects. This high dose of 0.05 mL of a 1%
2,3,7,8-TCDD solution in 50/50 alcohol chloroform solvent was applied to one
square inch of the backs every other day for a month and covered by a nonocclusive
patch. If we can assume that the subjects' average age was 35, a 25% absorption
35
-------
rate, an infinite half-life, and lifetime (70 years) average daily dose (LADD)
proportionality, one can estimate an internal dose of 2,3,7,8-TCDD that caused
chloracne to be in the 2 to 1,000 pg/kg body weight-day range. If we assume
either a 5- or 10-year half-life with first-order elimination kinetics and with
the other assumptions the same, then the chloracne causing the LADD range is
unchanged at one significant digit. These figures correspond to an external oral
dose of 4 to 2,000 pg/kg body weight-day.
Human cancer risk estimates based on the Kociba female rat feeding study
with 55% absorption yield an wpper-limit risk estimate of 1.56 x 10~4 (pg/kg-
day)~l. Multiplying this upper-limit estimate by 4 to 2,000 pg/kg-day and
adjusting for the different absorption fractions yields an upper-limit lifetime
incremental cancer risk of between 6 x 10~4 and 3 x 10~1 for hunans developing
chloracne. Only 10 of the Holmesburg prisoners were exposed to the highest
dose, and follow-up is unclear. Nevertheless, even if their lifetime projected
increnental cancer risks were as great as 0.3, and even if they were observed
for their full remaining lifetimes, less than three additional cancers would be
expected. Put another way, with three additional cancers expected, an observation
of no additional cancers would not be highly unusual (p = 0.055). Clearly,
unless specific types of cancer were to appear in these tested prisoners, no
conclusions could be drawn.
On the other hand, Tschirley (1986) in his review of the 2,3,7,8-TCDD
literature displays 9 cohorts of some 599 subjects who developed chloracne
following 2,3,7,8-TCDD and phenoxy herbicide exposure. This is reproduced as
Table 10. Of those cohorts, only the study of the 1949 accident at the Monsanto
plant in Nitro, West Virginia (Zack and Suskind, 1980), and the 1953 accident
at the BASF plant in Germany (Thiess et al., 1982), have sufficient latent
period and other information to allow a comparison to be made between observed
36
-------
TABLE 10. EXPOSURE TO 2,3,7,8-TCDD FROM INDUSTRIAL ACCIDENTS
Date
Workers
exposed
Location
of accident
Remarks
1949
250
1953
75
1956
1963
106
1964
61
1965-69
78
Monsanto plant in
Nitro, WV
BASF plant in
Ludwigshafen
Rhone-Poulenc
plant in
Grenoble
NV Philips
plant in
Amsterdam
Dow Chemical
plant in
Midland, MI
Continuing leaks
in Spolana plant
near Prague
122 cases of chloracne being
studied; 32 deaths vs. 46.4
expected; no excess deaths
from malignant neoplasms or
circulatory disease
55 cases of chloracne, 42
severe; 17 deaths vs. 11 to 25
expected (four gastrointes-
tinal cancers and two oat-cell
lung cancers); most common
injuries were impaired senses
and 1iver damage
17 cases of chloracne, also
elevated lipid and cholesterol
level s in the blood
44 chloracne cases (42 severe)
of whom 21 also had internal
damage or central nervous
system disturbances; eight
deaths (six possible myocardial
infarctions); some symptoms
of fatigue
49 cases of chloracne; 4 vs.
7.8 expected deaths; 3 cancer
deaths vs. 1.5 expected; one
a soft tissue sarcoma
78 cases of chloracne; five
deaths; many of the 50 workers
studied for more than 10 years
have hypertension, elevated
blood levels of lipid and
cholesterol, prediabetes;
significant amounts of severe
liver and neurologic damage
SOURCE: Tschirley, 1986.
37
-------
and predicted risks. These involved 1.77 chloracne cases'and e-ithe'r -a 26- or
30-year latency period. Quantitative cancer risk estimates based on these
cohort experiences are calculated below. However, the large uncertainty in the
exposure estimates should be emphasized, and the assumptions used to derive
these exposure estimates are necessary simplifications.
A. THIESS ET AL. (1982)
1. Description
At a factory in Ludwigshafen, Federal Republic of Germany, in 1953, during
the hydrolization of 1,2,4,5-tetrachlorobenzol to 2,4,5-trichlorophenol , an
accident happened exposing at least 70 persons. These 70 as well as 4 additional
persons who were only exposed "for a short time during" 1954 and 1955 were • •
included in the cohort. Of the 74 persons, 66 suffered chloracne or severe
dermatitis. All 74 persons were successfully traced through 1979. Of the-74
persons in the cohort, 21 had died during the 26 years of observation, just
slightly more than- expected in any of five different control groups. However,
there were seven cancer deaths observed versus 4.03 to 4.35 expected in these
control groups. Of these seven cancers, three were stomach (IOD 151), one was
colon (ICD 153), and three were lung (ICD 162). All seven occurred at least 10
years after the accident. A 10-year latent period will be assumed. The results
are presented in Table 11. The control group represents the expected deaths
based on the mortality rates of Rhinehessia-Palatinate 1970-75. The mortality
for stomach cancer was statistically significant (p = 0.016), while that for
lung cancer was marginally significant (p = 0.09).
2. Exposure Estimates*
Although there were no concurrent estimates or measurements of
*Contributions to this section were made by Drs. Jerry Blancato and Lorenz
Rhomberg of the Office of Health and Environmental Assessment.
38
-------
TABLE 11. OBSERVED AND EXPECTED DEATHS FROM STOMACH, COLON, AND LUNG CANCER
IN THE 74 CASES WITH CHLORACNE OR SEVERE DERMATITIS FROM THE
1953 BASF PLANT IN LUDWIGSHAFEN
(26-YEAR FOLLOW-UP WITH AT LEAST 10 YEARS' LATENCY)
Cause of death ICD No. Observed Expected SMR p-value
95% Upper
confidence
1imit based
on Poisson
distribution
Stomach cancer 151
0.52
5.76 0.016 1.15 - 16.8
Colon cancer
153
0.24
4.17 0.21
0.05 - 23.2
Lung cancer
162
1.05
2.86 0.09
0.57 - 8.3
Person-years at risk = 972..6.
39
-------
2,3,7,8-TCDD exposure, Rappe (1987) reported a mean level of 100 pg/g in adipose
tissue of four exposed BASF workers some 30 years after exposure. If we are
willing to make several simplifying assumptions, then we can estimate a range
of exposure. The assumptions are:
• The body is treated as a one-compartment open model.
• A range of half-lives is assumed to be between 5 to 30 years.
• Adipose tissue-in a 70-kg human is about 2'0% or 14 .k'g. ! .
• Absorption into the body is assumed to be: between^ 5% and 25%. This
figure is arbitrary and is chosen because there is less than 55% absorp-
tion in the rat feeding studies.
Under these assumptions the pertinent equation is:
C(t) = C0 e
-kt
where C(t) = the concentration in the fat at the time of analysis,
C0 = the concentration in the fat immediately following absorption in
1953,
k = the total body elimination rate constant, assumed to be the sum
of the elimination rate constants by various physiological
processes, and
t = the time since absorption was completed, assumed to be 30 years.
Further, we note:
k = 0.693/t0>5
where tg.5 = the half-life of elimination from the body, and
Exposure = C0V/r
where V = the weight of the fat compartment (14 kg) and r = the absorption
fraction. In this case, the weight of the fat compartment is used to calculate
the dose. The other organs may be neglected because of the propensity of
40
-------
2,3,7,8-TCDD to partition into the fat. For example, the concentration in the
fat is about 100 times that of the blood.
Based on the above assumptions and estimates, the range of estimates of
2,3,7,8-TCDD exposure in the BASF plant accident is between 10 and 1,800 ug.
Based on a LADD, these estimates range from 1.5 to 50 pg/kg body weight-day.
The calculations are shown in Table 12. Clearly, this factor of 180 in the
range of exposures and a factor of 30 in the range of LADDs creates significant
uncertainty in the risk estimate calculations. In order to narrow the range,
however, we note that in the Holmesburg study exposures below 16 ug caused no
chloracne, while exposures of 7,500 ug caused 80% chloracne. Considering the
probably greater absorption in the Holmesburg study, the exposure estimates in
the top three rows of Table 12, showing the range of 220 to 1,800 ug, seem the
most likely. We therefore adopt LADDs in the range of 6 to 5.0 pg/kg body weight-
day. The risk estimates below will be calculated on the LADD of 50 pg/kg/body
weight-day; risk estimates based on the lower end of the range would be about
eight times higher.
3. Risk Estimates
Cancer risk estimates are calculated for the stomach cancer and lung cancer
mortality presented in Table' 11. These estimates are presented in Table 13.
Two models are considered, the additive and relative risk models, which have
been used in several EPA risk assessments. They are developed and more fully
explained in the recent EPA update on dichloromethane (U.S. EPA, 1987). Both
models require estimates of LADDs, and these have been estimated above as
between 6 and 50 pg/kg body weight-day. The results show that incremental
cancer risk estimates based on these human data are higher than those based on
the animal data in every case. If the lower end of the range of LADDs had been
used, the human estimates would have been greater still. The conclusion based
41
-------
z:
LU
lJU
j
03
•o
O) 1
£0) S
•r= Cn >4»- J3
+J 03 r— OJ
• "O ~O ' —
— I 03 en
Q.
O)
C_0)
co en
O 3
r"> ^ ^
X
LU
CT3
o w>
•r- t— ^ — »
CU | ^ o3 Cn
CO 03 O)
o t_ >, en
CL+J Q.
•p— ^— CD * *
T3 CO II
«C 0 1— 0
C 0
O 4J
O 03
C
0
!Jj c:*^
03 ^ — i
C 4-> 1
•^- CO t_
E ^~ ^i
•r- o * ^
, ^J vx
LU
O SS.
1— M
CO 00
« 1
I""*- LU
«\ ^^
OO LU
•> 1
T- O
r*i ^-> 1
C_ O LO
O "3 O
CO ^- •
J3 <*- 0
Q. i «C ^^
00 (_ ' £_
LU o
!""• i^C «
)
•r- •*^^'
r—
uL "^
r— O
«H-
03
f~* (/^
O t-
•f~* fO
CD +J OJ
00 03 >^~^
c O £_ cn
O3 CX 4_> CD ^~,
cu T- c oo en
s: -a cu " CL
o 03
O O IO OO OJ i— 1 t— )
LO LO
O CD CD O O LO O
O (~O CM «— 1 CO LO r-<
CO OO OJ t-H
1 — 1
i — r — cn cn oo cn cn
•=d~ *=^" r^- oo CM < — ' i — i
to to
cn cn cr> to LO oo oo
oo oo to •=*• oo oj CM
r—l «— 1 O O CD CD CD
• ••••••
O CD O O CD CD CD
LO LO LO LO LO LO LO
O CM O CD O O CM
CD CD O CD CD O CD
LO LO O LO CD O CD
r-l i— 1 CM OO OO
O O O CD O CD O
CD CD CD CD CD CD CD
i— 1 IT 1 t— 1 i— 1 t— ) i— 1 i— 1
•
-o
0)
3
CO
CO
03
•
^--. O)
CO 03
cn - — *
pH CO -*
— CO O
> oo
. CU ••>— 1
Q. i - — -
Q. 1 LO
03 4- .0.
a: r— ox
03 1 — OJ
>•> **" ^^,
^ oo O
4-> cn cj>
-a c to
a> a> • u
£-5-0
3 cu - —
CO 4- || 4->
03 <*- -^
cu -i- -v: O
5- CD
03 J3 O "O
J %
03
o
O) i^ —
03 l^'
I — 1
• " — ^
•r~ vx*
Z5 LO
crro
0> i
0) Q-
co x
O O)
•a i
cu —
> »
!- O
13 O
01—1
C II
o
•r- Q
+J CD
03 eC
£ 1
e
01 ••
o ^~~
r~ •
o o
o .
CO i— <
cr
•r— C
+-> O
C_ •(->
cu o
T3 Ol
g~* (/)
Z3
01
03 O)
CO CO
£_
03
CO
W) •!-
CO
§03
,»
co zs
•^ "^ C-
cn o
u.
o «
O JO •
r, CO
2 -3*03
CO
X £- >,
CO
O OL.LO
CJ OO
cu
u u
0
Ol T3 4->
g c~
=i II 0)
co -a
O CD *r-
a. do
X •< O
LU — 1 03
CO 4-
42
-------
Q
a uj
O'COi
'TO.
CO Q-'
"CO U-
r~- uj to
*» f*S d"
CO po
CM"H^UJ
Q "-^ 1
1 — Q-
1 , t ^J^, V^.
a;
ID CO to
co LU Q:
O 1— UJ
Q_ "=3^ NX
x 2: a-:
uj i — i o
CQ -g
UJ CC
l t I J,
1 —
1 1 1 1 -^-
U_ eC i — i
i — i CO
_l =t >- CO
t^f ^^^ j t j^
O UJ 1 co
U_ 1— eC TD '
3H- 2^ '
a; UJ S
UJ eC O Li_ -Q
1— a- ^- eC
co re 2: 3 ex.
CD o ID O
^£ "Z. 1 LO
co <: re
•— « o 2: n
o: «<; 1—1
02! a
uj ZD i— z: •<
c_> — i co <=C — i
=C «Q CL. UJ
UJ ct — >
1 >• OO
=C i — i CD CO
1— — 1 2T -i
OO 3
» — ' f ^
LU
l i l CO
I et
CQ CQ
| —
£_
CU
CL
cu f-
00 3 CX-P
•i— CO r> T-
i_ o rs E
CL •>-
CO CO LO
O O">
C OO
CO 3
0 0
i — C
n3 ft—
c. c o
CO O LU
J— O l
CO 21
( ^^
O CO
•r— t
. en
CU \s
c «^^
•r- CD
4- r-l 3 E
•r- O •""
_J — J r—
0 *»-
•i~~ ^5 (/J
4-> 0)
O CO .p
+J O CO
Q.C IE
§,•1— ^_>
< fi c2
•
JD
t_ oo
cu co
CD CO
|.E ,
03 -P
_
CU TD
TD CO
jr 23
^^_
OO
CO
•i- X
O CO
CO co
ex
to
t— 1
1
o
CO •* CO (--. ,-H
. . . . x
CO CO r-H i— 1 «3
LO.
i— 1
•—I
i — 1 t — 1 1
1 1 O
O O ,— l
oo •=*•*-( .-i x
. . X X LO
i-l t-l CM LO i-H
4-) 3 -P 3 3 '•-
co -J co i — 1 1 —
cu
,^.
[ %
CO -O CU
o co en
Q) -r- N CO
> , "1™" 1 ^
•i- Q. f_ CO
4-> -r— O3 •!—
•t- +J CO -P
TD i — C i —
T3 3 -i-3
CU CU
£- -P
O) CO
•P E
00 ~r-
•n- 4«>
C 00
•r- CO
E
T3 CO
CO > . •
O
C Jd
CO CO
O CO
•P
C r-™
CU r—
r-
3 C
cr o
CU •!-
CO Q.
o
c: oo
o *^
•1- CO
[ ^
o >-g
• t- 0
o to
CO
o TD
CO CO
4-3
&^ CO
LO E
•i—
cu oo
+o CU
CO
•^£
to '5
CO •
c c
-COO
•P 'i— «r—
•r- +J 4.)
3 00 CO
CO CU
cu en co
•PCC
3 -r- -r-
O
t- CO CO
1 — > >
03
t_ OO CU
O •!— CO
o
TD CO T3
C t.
co 3 -a
oo cu
C 0 f_
•i— CX CU
-si X +J
oo CO co
•r—
03 C C
•r- CU •!-
>"i -i
>i CO
CO >>
TD cO CU
1 TD _C
-a s "^
-Q 0
CO +->
^sC en
-*: ^> co
CX CO N
(""> «|—
0 -o
LO O t_
O co
M- ^H TD
O C
4- CO
a o -P
Q co
«=C CU
— 1 00 CU
0 £_
•^-
-p
CO
0
•r—
r—
CX
•P
r—
3
E
CO
•p
t_
o
LU
.
IL.
03
CO
1
en
c
-^^
>^
CO
TD
CO
M-
O
CO
-p
c
3
•p—
a>
CO
OO
cu
-p
03
E
•P •
OO CO
cu c
t- ">»
CU CO
•P TD
CO |
E cn
CO -Si
c.
CO ll-
CX 0
cu oo
-c: -P
-P -T-
• c:
r— 3
O)
TD C
O -r-
E
01
cu r_
> CO
•p-
4-> OO
•r- CU
TD -P
TD CO
03 E
CO -P
JC 00
-p cu
t_ cu
O -C
U- -P
-D
**
,'co '. '.
TD •
o. • - .
E- ..-
cu
^
• r—
CO
O
•r—
r_
O.
«^-
-P
3
E
CU
•p
c_
o
f[
• n
TD
o
E
cu
•r—
• -"r^
TD
TD
03
CU
-P
c_
o
u_
•
cu
-p
CO
^—
•r—
•P
CO
a>
TD
o
o
•r—
r—
CU
^
•r—
§•
E <
•r—
X X
cO o
E Q-
II II
UJ UJ
— 1 — 1
s s:
o
•
CO
CU
3
'cO
>
.'-P '
E
"i™™
r—
t_
O)
CL
CX
3
£_
O
r—
CO
CU
O
T3
O
iL.
CX
CO
f~
I—
00
CO
o
O
•r-
-P
O
•p
CX
^,
00
03
CU
-C
4-3
C
O
TD
CU
oo
03
CU
*—
CO
co
•p
E
•r-
T
£_
cu
a.
a.
3
O)
JZI
4-5
CO
C
03
E
3
f"
L-
O
U.
TD
43
-------
on the above analysis is that the upper-limit incremental unit risk estimates
based on the animal studies do not overestimate the human risk from 2,3,7,8-TCDD
if either the lung or stomach cancer mortality response in humans is bona fide.
B. ZACK AND SUSKIND (1980)
1. Description
In Nitro, West Virginia, on March 8, 1949, excessive temperatures in an
autoclave involved in the production of 2,4,5-trichlorophenoxyacetic acid
caused a relief valve to open, allowing fumes and residues to escape into the
atmosphere and into the interior of the building. A total of 121 white males
were identified as having developed chloracne following this incident, and
these were included in the subsequent follow-up study, with vital status ascer-
tained through the last day of 1978, nearly 30 years.
There was 100% follow-up of subjects; 89 were still alive and "32 were
verified deceased by death certificate." With 46.41 expected deaths, the
standardized mortality ratio (SMR) for all causes, 69, was significantly
(p < 0.05) lower than expected. The only cause of death that displayed normal
rates was cancer which had 9 observed and 9.04 expected, SMR=99.6. Of those
cancer deaths, five were from lung (2.85 expected, SMR=175), three were from
lymphatic or henatopoietic tissue (0.88 expected, SMR=341), and one was a soft
tissue sarcoma (STS) (0.15 expected). Only one of the cancer deaths, a lung
cancer, was a nonsmoker.
2. Risk Estimates
In order to make any kind of quantitative risk estimation of the potency
of 2,3,7,8-TCDD, several assumptions must be made. These are:
(a) Exposure. As discussed above for the Holmesburg study, it is assumed
that the LADD necessary to cause chloracne was in the 2 to 1,000 pg/kg-day range,
44
-------
An estimate in the middle, or 500 pg/kg-day, appears to be a reasonable starting
point, as does a value of 150 pg/kg-day, which is close to the geometric mean.
The value of 500 pg/kg-day will be chosen as the LADD dose for those who devel-
oped chloracne, but the uncertainty of this estimate must be stressed. This
dose estimate is 10 times greater than that estimated for the BASF study; the
incremental unit risk cancer estimates will be correspondingly lower.
(b) Expected deaths from cancer. Table 14 presents the observed and
expected cancer deaths for the 29.8-year latency. However, all the cancer
deaths appeared after at least a 10-year latency, and it seems reasonable that
10 years is an appropriate latent period for any 2,3,7,8-TCDD-related cancer
to express itself. Therefore, the expected deaths presented by Zack and Suskind
must be adjusted by subtracting the first 10 years' experience. An examination
of vital statistics rates for lung cancer and STS suggests that for lung cancer
deaths approximately 20% of a 30-year death experience happens in the first 10
years; for STS the figure is approximately 30%. Based on these adjustments,
the expected deaths for this exposed cohort become 2.3 and 0.10 for lung and
STS cancers, respectively.
(c) Person-years at risk. A figure needed for the additive risk model
but not the relative risk model is the person-years at risk. Since the first
10 years are considered to be a latent or risk-free period, only the last 19.8
years are counted as person-years at risk. The follow-up was complete and
there were 32 deaths. It can be assumed that the average time until death was
20 years from first exposure (for the nine cancer deaths the average time from
the accident until death was 22 years). Therefore, the total person-years at
risk can be estimated as
P-Y = 19.8 x 89 + 10 x 32 = 2,082
45
-------
TABLE 14. FOLLOW-UP OF 121 CHLORACNE CASES AT THE MONSANTO COMPANY (NITRO,
WEST VIRGINIA) USED TO DERIVE QUANTITATIVE CANCER RISK ESTIMATES
Lung cancer
Unadjusted Adjusted
Deaths
Observed 55
Expected 2.85 2.3
Person-years
at risk — 2082
SMRs
Observed 175 217
95% Confidence 57-410 70-508
STS
Unadjusted Adjusted
1 1
0.15 0.10
2082
667 1000
9-3707 13-5560
1imits based on
Poisson
distribution
46
-------
The results of the analysis using both the additive and multiplicate models
are presented in Table 15. Based on the assumptions discussed above, both
models yield similar results. For extrapolation based on lung cancer deaths,
the MLEs for incremental risks are 9.1 x lO'2 and 4.5 x 1Q-2 (mg/kg-day)-1 for
the additive and multiplicative models, respectively. For extrapolation based
on the one STS death, the MLEs are lower than are those for lung, 3.0 x ICT2
and 8.6 x 10~3 for the additive and multiplicative models, respectively. The
95% upper-limit estimates presented in Table 15 are those based on asymptotic
normality theory and are, therefore, lower than would be produced by applying
the upper confidence limits on the SMRs from Table 14, which are based on the
Poisson distribution. Those Poisson upper-limit adjusted SMRs, for example,
would yield 95% upper-limit cancer unit risk estimates of 5.2 x 10"2 (ng/kg-day)"1
for STS, and 1.6 x 10~i for lung cancer deaths for the multiplicative model.
These values are significantly greater than those presented in Table 15 [1.4 x
10-2 and 1.1 x 10"1 (ng/kg-day)-1, respectively] and reflect the high degree of
variability due to small sample size.
Also presented in Table 15 is.the quantitative cancer risk estimate derived
from the Kociba feeding study (U.S. EPA, 1985). While the MLEs based on the
animal data are si ightly higher than those based on the Monsanto data, the
differences are no greater than a factor of 2.4 for lung cancer and 13.4 for
STS. All the 95% upper-limit estimates based on the human data are within a
factor of 2 of the upper-limit estimate based on the rat data.
The conclusion based on the above analysis is similar to that derived from
the BASF analysis--the available human cancer data on 2,3,7,8-TCDD do not
provide any evidence that the unit risk estimate based on rat data overpredict
the human experience. The information on human exposure is just too uncertain
to allow for a more definitive statement.
47
-------
o
o
O LU
I— to
1 2:
CO 0
•>CL.
1^ CO
.sLU
CO Qi
«»
CM . — .
0 2;
1— ^>.CO
CL. LU
t-U IC^LU
^> O
co co _i
0 LU Q.
CL 1 — *?•
X< LU
LU 2:
•-« O
LU CQ f—
•SLCZ Z
t— < 13 cC
1— f— CO
LU 2T
U 1 O
>-; <: 2:
—i co ^->
a; 2: CM ro
CD ~»» i— i T3
U_ LU 1
H~~ ^?f 3«
CO CC p-l JD
LU _3
I — - cn
«C Q- | — -i£
^- 1—4 *-^»
>-« Q i cn
1— c£ < Q.
CO <=C |—
LU a: ce CD
00
^£ CD. *y Lf>
co 2:
•-« =3 Q CO
1 — •— « LU <
z _J co>— -
LU ea;
S 1— CQ
LU It 1
2: LU |—
>— «4 __J ^£
. $ 3E
u- 2: i— i
0 (_
«=C CO
2: CO LU
O r-H
CO 0 31
•-« CD h-
c£ ^<£ i— <
^C ^g
O. LU
?• ; _i_
O 1—
O
2:
O
•
LO Q
i— t LU
CO
LU «=C
— 1 CQ
CQ
^t
H-
OJ
CL
CO
-ii t-
10 3
•r- tO t-
t- o co-o
Q. O.4-5
C- X CL-'-
CU CO ZD E
0 •<-
C: CO 6^1 —
ro 3 LO
o o cn
3
ro •!—
4-5 4-5
C C
CO O
£~ 5-J tJ
CO LU
t— >j I
o ro 2:
c: -o
•r- i
cn
CO v
•P CD
4-5 d C— 4-5
CO CO •*—
^~ i ^ E
•r- 0 T-
— 1 — I 1 —
0 4-
•i- O to
0 CO -(->
4-5 o ro
o. c E
E ro -r-
5-j*r— 4-5
0) £- to
cc ro co
^-o,
CO CO
4-5 4-5
|s<
ro -i-
frt t/)
Q_ QJ
£_
CO CO
O D-
ro 4-5
C_5
r—
CU T3
-a co
o to
^~ 3
"^^.
co
CO
•r- X
o a>
CO 01
a.
CO
CM
O
t— t
X
LO
*
co
CM
1
O
1 — I
X
o
t
CO
CD
1
o
.— 1
X
co
a
CM
._,.
1
o
I— 1
X
CO
•
S^J
CO
1 —
CO
CO
>
•r—
4-5
•I —
T3
"O
*3^
*-^s-
CO
c: 01
ro co
E r—
3 ro
rn 2:
i-H
t — 1 C\J i — 1 1
II 1 0
00 0 ,_)
r— t 1— « i— t X
XX X tO
CM <5f t— 1 LO
• « • •
CM r- 1 t— 1 r-<
f — I
CM CO CM 1
II 1 CD
O O CD r-4
.__[ . 1 ^x
XX X LO
t-H IO LT) r-f
• • • •
01 CO ^ r-<
O O O O
iO
1
0 ^1-
r-l CD CO
X • .
CM O O
• rH
r- 1
CO
1
CD '
r-4 O CM
X . .
CO O^ t — I
i
.— 1
H-
^^
un ^
cn Q-
o co ^r
o co
CD O •
• • «> £_
CD o cno cn CO
c co ll c: u c >'
3 I— 0 30 3 T-
—1 CO Q_ _l Q. — J i—
CO
>
• r—
4-5
ro
o -a co
•r- co cn
r— • Nl ro
CX -r- 4-3
•r- C_ CO
4-5 ro *r—
r— CO 4-5
3 C r— -
2: T- rs
— J E
CO
CO
> — r—
to ro
4-> E
ro co
48 ^LU
-a co
CO CO
co ro
4-5 E
CO •!—
•r- 4-5
•r- CO
E
-a co
ro >
o
c: -Q
ro ro
O 0)
4-5 -C
C r—
I— <
ro
•i— •
3 C
D" 0
CO T-
to CL
•r- £_
O
c to
0 JO
•i- ro
o 5^
£- 0
O LO
to
o ~O
ro co
&^ ro
LO E
-a 4->
O) CO
4-5 a>
ro
•r- 4-5
to 's
CO •
f— f—
-COO
-i— 4-5 jo
S co to
CO O)
co cn cn
4-5 C C
3 -r- .,-
o
t- ro ro
•r— .f—
"ro >
f— to CO
O v- CO
o
T3 CO -o
C £-
ro 3 -o
CO CO
c: o £_
•i— a. co
-i^ X 4-5
to co to
>r_
ro c: c:
•I- CO •!-
^ "O
>-> ro
ro >>
T3 ro co
1 TD JC
3: i 4-5
<^ S
JD o
CD 4->
^4 CD
~~~ -^ -a
-^ ^-, co
Q. cn M
Q.T-
o -a
LO O £_
o ro
t- *-<-a
O c:
**- ro
CD. O 4-5
Q to
eC CO
—J tO CO
0 £_
-=C T3 ro
ro
^_
OJ
-o
o
CO
lpw
1 \
ro
o
,
Q.
•r-
33
p—
CO
t-
£_
o
•
£_
ro
CO
i
cn
^^_
^_
ro
-a
i
cn
14.
O
CO
C
3
C^
CO
ro
to
CO
ro
E
•r—
4-5 •
co cn
co c
•^^
co ro
4-> 13
0) |
E CD
ro -*:
£_
ro 4-
Q. O
CO CO
c~ 4->
4-5 *r-'
C
i — 3
CO
T3 C
O T-
E
CO
CO L.
>• ro
4-5 CO
•r- CO
•a 4-5
-o ro
ro E
•r—
CO 4-5
-C CO
4-5 CO
C_ 0)
O -C
U. 4->
JO
v,
t
CO
"O
o
E
CO
•r—
4,j
' ro
o
_
Cu
4-5
^^
3
E
CO
-C
t-
o
• *.
^
X
o
II
LU
1
|gr
f,
CO
T3
O
E
co
>r_
4_>
T3
-o
ro
OJ
40
£_
O
u_
•
0)
ro
^
>r_
4_>
to
CO
"O
o
o
-E
•r-
t
CO
•r—
r—
c •
_E r_
_E
, —
l
co
0
0
3
£_
^
O
to
CO
o
•a
o
£_
a.
to
c-
1 —
CO
QJ
O
ro
t
ro ...
o
o
4-5
CL
to
ro
CD
JC
4-5
c~
O
T3
CO
CO
ro
CO
g
ro
to
4_)
•P-
p:
.pE
P_
1
L-
CO
Q.
a.
3
CO
4->
CO
c;
ro
jj3
t
o
Lu
-o
-------
V. DISCUSSION AND SUMMARY
This report has presented the effects on the human cancer risk estimates
from 2,3,7,8-TCDD exposure under varying assumptions involved in the animal-to-
human extrapolation procedure. It has compared EPA's risk estimates with those
of other agencies, both U.S. and foreign, discussed the rationale, used in each,
and shown the effects of slightly different assumptions on the estimates. In
general, the risk extrapolations divide roughly into two groups, those agencies
using the linearized multistage family of models for extrapolation and those
using an uncertainty factor approach. The agencies using the linearized multi-
stage model all produce RsDs within a factor of 10; similarly, the agencies
using the uncertainty factor approach are also with a factor of 10. The two
groups, however, are separated by a factor of 16, so that the lowest RsD, that
of EPA which used the linearized multistage model, is 1,600 times lower than
that of Health and Welfare Canada which used an uncertainty factor of 100.
While EPA's cancer risk estimates were the highest of the agencies pre-
sented, other methodologies consistent with the data have yielded still higher
estimates. For example, fitting the data with both the Logit and the Weibull
models would have produced significantly higher estimates—the Logit by roughly
one order of magnitude and the Weibull by 2 orders of magnitude. In addition,
even extending the multistage model to the Weibull-in-time model under a time-
to-tumor analysis would have increased the upper-limit estimates by as much as
a factor of 5. Of even greater uncertainty is the extremely long half-life of
2,3,7,8-TCDD in the human compared to the rat. If half-life is related to species
sensitivity as implied by the cross-species extrapolation factor, then recent
estimates of human half-life of 6 to 10 years implies that rat-to-human extrapo-
lation estimates should be significantly higher, probably by a factor of 2 to 7.
49
-------
Finally, a new methodology has been presented that models 2,3,7,8-TCDD as
a carcinogen promoter only. The methodology models the promotion (liver cell
proliferation) phase of the M-V-K two-stage model as a linear saturable function
of 2,3,7,8-TCDD-receptor binding. The risk estimates of induced human liver
cancer are 2 orders of magnitude less than those using EPA's current method-
ology. However, research on this new methodology is ongoing and, at this time,
not enough is known about the low-dose estimation properties to make definitive .
statements.
Which of these is the "correct answer"? Probably "none of the above."
What the above analyses show are that all the answers are consistent with the
observed d-ata, and all have some credence depending upon the be! ievabil ity of
the assumptions used. The most pertinent fact is that 2,3,7,8-TCDD causes
liver, tongue, hard palate/nasal turbinates, and lung tumors in rats at doses
and conditions to which humans would never be exposed. As such, even with
animal bioassays, as well-conducted as were those for 2,3,7,8-TCDD, the informa-
tion they contain for low-dose extrapolation i,s very limited. Within a 100-
fold decrease from experimental dose levels, the range of estimates predicted by
models that fit the data well, rapidly diverge to a "pay your money, take your
choice" level of 3 orders of magnitude. Below that, divergence is even more
rapid (see Tables 4 and 9). Furthermore, when extrapolation is made from
animals to humans, the uncertainty about the effect of the extremely, 1 ong half-
life in hunans gives concern about the conservativeness of the upper-limit based
on the surface area correction for extrapolation.
Use of human data for risk assessment purposes is also impossible with
2,3,7,8-TCDD. First, the evidence for human carcinogenicity of 2,3,7,8-TCDU
alone is judged inadequate (see Appendix B). Second, the studies providing posi-
tive evidence for carcinogenicity are of a case-control design; these lack both
50-
-------
a population base and exposure estimates. The few cohort studies available
lack sensitivity because of a combination of factors including exceptionally
small cohorts, insufficient latency, and in some instances, little evidence of
exposure. Only with one study was there even an estimate of 2,3,7,8-TCDD
exposure, and that study lacked power to discredit any of the predictions
provided by the animal data.
Well, the next question is whether any of these estimates can be con-
sidered superior to the others. To answer this, one must first presume that
mathematical models can be used for prediction, and then decide on whether to
model 2,3,7,8-TCDD as a complete carcinogen or as a promoter only. Modeling as
a complete carcinogen, only the one-hit and multistage models have a theoretical
backing in carcinogenesis; the other models presented are merely well-known
tolerance distribution models. The EPA position is that use of the linearized
multistage model for extrapolating upper-limits of incremental risk is both
prudent and protective. When both animal and human data have been available
for risk estimation, the linearized multistage model's use with animal data has
provided estimates comparable with those derived from human data. Furthermore,
low-dose supralinearity is rarely seen, so that using the linearized multistage
model for extrapolating from experimental levels probably represents a protective
level for humans.
When one models 2,3,7,8-TCDD as a promoter only, many additional uncertain-
ties arise. The two most important for modeling are those associated with
reversibility and threshold. Classical promoters are known to show reversi-
bility of lesions when the promoter is no longer administered and cleared from
the system. Furthermore, large doses are typically required for promotion,
indicative of a toxicity effect either leading directly to cell damage and
regenerat-ion, or overwhelming the cell's ability to prevent the promoter from
51
-------
reaching its site of action. In the case of 2,3,7,8-TCDD, the evidence for
liver cancer in animals points to a mechanism of promotion via receptor binding,
yielding a very strongly bound complex. Even if this complex is broken down,
the persistence of 2,3,7,8-TCDD allows it to bind with another receptor. In
terms of mathematical modeling, this information translates to a dose-response
function for cell proliferation which excludes threshold and reversibility, but
otherwise can be defined by Michaelis-Menten type kinetics. When this function
is substituted into the promoter form of the M-V-K model, the results yield
low-dose estimates of liver cancer for humans which are 2 orders of magnitude
lower than the upper-limit estimates provided by the one-hit and linearized
multistage models.
While the upper-limit estimates can be considered upper-limits for total
human 2,3,7,8-TCDD-induced cancer, the predictions with the promoter form of
the M-V-K model require further examination. First, they apply only to human
liver cancer, a condition reported only in the cohort exposed to dibenzofuran-
contaminated PCBs in Yusho, Japan (Amand et a!., 1984). They do not include
estimates for STS or non-Hodgkins lymphomas. Second, they are considered
"best" estimates compared with the upper-limit estimates calculated from the
linearized multistage model. Third, the form of the M-V-K model used predicts
carcinogenic response only on the basis of promotion. If a 2,3,7,8-TCDD-induced
initiation stage had been incorporated (as is suggested by Holder and Rosenthal ,
1987) the low-dose cancer predictions would have been higher.
For these reasons, the estimates provided by the promoter form of the
M-V-K model might be considered prudent lower bounds on cancer risk, while those
upper-limit estimates provided by EPA's current methodology are to be considered
upper bounds. While any number of assumptions could produce either lower or
higher risk predictions, the biological facts incorporated into both models
52
-------
provide what should be considered reasonable working limits. In addition,
2,3,7,8-TCDD1s many other toxicities should also be a consideration in setting
a lower limit. Any criteria level lower than that provided by the M-V-K model
would be at a level where these concerns would prevail. A final caveat remains,
however, on the impact of the extremely long half-life of 2,3,7,8-TCDD in man.
If 2,3,7,8-TCDD is not sequestered in the fat, but is bioavailable, the
quantitative risk estimates would be considerably larger.
53
-------
VI. REFERENCES'
Anand, M.; Yagi, K.; Nakajima H.; Takehora R.; Sakai H.; Umed, 6.; Labor K.
(1984) Statistical observations about the causes,of the death of patients
with oil poisoning. Japan Hygiene 39(1): (translation) ' '
Armitage, P.; Doll, R. (1954) The age distribution of cancer and a'multistage
theory of carcinogenesis. Br. J. Cancer 8:1-12.
Barnes, D. (1987) Holmesburg prison-based assessment. Unpublished notes.
March 10, 1987. , '....••••
California. (1985, April 19) Health effects of 2,3,7,8-TCDD and related
compounds (draft). Department of Health Services. Berkeley, California.
Unpublished.
Centers for Disease Control (CDC) (1980) Serum dioxin in Vietnam-era veterans.
Preliminary report. Morbidity and Mortality Weekly Report 36(28):470-475.
Chu, K.C.; Brown, C.C; Tarone, R.E.; Tan, W.Y. (1987) Differentiating among
proposed mechanism for tumor promotion in mouse skin with the use of the
multievent model for cancer. J. Natl. Cancer Inst. 79(4)789-796.
Crump, K.S. (1983) How to handle animal survival data in quantitative risk
assessment: a discussion paper. Prepared for Research Triangle Institute,
EPA Contract No. 68-01-6826, Delivery Order 28.
Crimp, K.S. (1988) A critical evaluation of a dose-response assessment for
TCDD. Food. Chem. Toxicol. In press.
Crump, K.S.; Watson, W.W. (1979) GLOBAL 79: A fortran program to extrapolate
dichotomous animal carcinogenicity data to low dose. Prepared for the
National Institute of Environmental Health Sciences, Contract No. I-ES-2123.
Crump, K.S.; Guess, H.A.; Deal, L.L. (1977) Confidence intervals and test of
hypothesis concerning dose-response relations inferred from animal car-
cinogenicity data. Biometrics 33:437-451.
Cook, B.T.; Page, N.P. (1986, February 3) Comparison of carcinogenicity risk
assessment of 2,3,7,8-TCDD (draft). EPA Contract No. 68-02-4131, Work
Order No. 3.
Crockett, P.W.; Crump, K.S. (1986, February) Methods for assessment of non-
cancer health risks. Prepared by K.S. Crump and Co. for the Electric
Power Research Institute, Contract No. RP 1826-17.
Environmental Canada. (1984) Chlorophenol s and their impurities in the Canadian
environment. 1983 Supplement. Economic and technical review report.
EPS3-EP-84-3. Environmental Protection Programs, Directorate, Canada.
54
i'
-------
Food and Drug Administration (FDA). (1983, June 30) Statement by S.A. Miller,
Director, Bureau of Foods, FDA, before the Subcommittee on Natural Resources
Agriculture Research and Environment, U.S. House of Representatives.
Germany. (1984) Report on dioxins. Update to Nov. 1984'. Federal Environ-
mental Agency.
Howe, R.B.; Crump, K.S. (1982) WEIBULL82: a FORTRAN program for low-dose
extrapolation using time to tumor data. K.S. Crump and Company, Inc.,
Ruston, Louisiana.
Holder, J.M.; Rosenthal, S. (1987, February 13) Issue paper concerning the
mechanism of 2,3,7,8-TCDD carcinogenicity. Office of Health and Environ-
mental Assessment, U.S. Environmental Protection Agency, Washington, DC.
Unpublished draft report.
Kimbrough, R.D.; Falk, H.; Stehr, P.; Fries, 6. (1984) Health implications
of 2,3,7,8-tetrachlorodibenzo-_p_-dioxin (2,3,7,8-TCDD) contamination of
residential soil. J. Toxicol. Environ. Health 14:47-93.
Kociba, R.J.; Keyes, D.G.; Beyer, J.E.; et al . (1977) Results of a two-year
chronic toxicity and oncogenicity study of 2,3,7,8-tetrachlorodibenzo-_p_-
dioxin in rats. Submitted to the U.S. Environmental Protection Agency.
Unpublished.
Kociba, R.J.; Keyes, D.G.; Beyer, J.E.; et al. (1978) Results of a two-year
chronic toxicity and oncogenicity study of 2,3,7,8-tetrachlorodibenzo-_p_-
dioxin in rats. Toxicol. Appl. Pharmacol. 46(2): 279-303.
Longstreth, J.D.; Hushon, J.M. (1983) Risk assessment for 2,3,7,8-TCDD.
In: Tucker, R.E.; Young, A.L.; Gray, A.P., eds. Human and environmental
risk of chlorinated dioxins and related compounds. New York, NY:
Plenum, pp. 639-664.
Moolgavkar, S.H.; Venzon, D.J. (1979) Two-event models for carcinogenesis.
Incidence curves for childhood ad adult tumors. Math Biosci. 47:55-77.
Moolgavkar S.H.; Knudson, A.G. (1981) Mutation and cancer: a model for
human carcinogenesis. J. Natl. Cancer Inst. 66:1073-1052.
Nati.onal Academy of Sciences (NAS). (1977) Drinking water and health. Safe
Drinking Water Committee, National Academy of Sciences. Washington, DC.
National Research Council of Canada (NRCC). (1981) Polychlorinated dibenzo-^-
dioxins: criteria for their effects on man and his environment. Pub!. No.
18474, ISSN 0316-0114. NRCC/CNRC Associate Committee on Scientific
Criteria for Environmental Quality, Ottawa, Canada. 251 p.
National Toxicology Program (NTP). (1982) Carcinogenesis bioassay of 2,3,7,8-
tetrachlorodibenzo-jD-dioxin in Osborne-Mendel rats and B6C3F1 mice.
Technical Report No. 209. Research Triangle Park, NC.
Ontario, Canada, Ministry of the Environment. (1985) Scientific criteria
document for standard development. No. 4-84. PCDDs and PCDFs.
55
-------
Paustenbauch, D.J.; Shu, H.P.; Murray, F.J. (1986) A critical examination of
assumptions used in risk assessment of dioxin contaminated soil. Regul.
Toxicol. Pharmacol. 6:284-307.
Portier, C.; Heel, D. (1983) Low-dose-rate extrapolation using the multistage
model. Biometrics 39:897-906.
Rappe C.; Aldersson, R.; Bergquist, P. (1987, January) Sources and relative
importance of PCDD and PCDF emissions. Presented at the ISWA/WHO-EURO/DAKOFA
Specialized Seminar on Emissions of Trace Organics from Municipal Solid
Waste Incinerators, Copenhagen, Denmark.
Rose, J.Q.; Ramsey, J.C. ; Wentzlin, T.H,,; Hummel, R.A.; Gehring, P.J. (1976)
The fate of 2,3,7,8-TCDD following single and repeated oral doses to the rat.
Toxicol. Appl. Pharmacol. 36:209-226.
Rowe, V.K. (1980) Direct testimony at U.S. EPA's cancellation hearings of
2,4,5-T. Exhibit 865. FIFRA Docket Nos. 415 et al.
Sielken, R.L. (1987) Quantitative cancer risk assessments for TCDD. Food
Chem. Toxicol. 25(3):257-267.
Sielken, R.L.; Carlborg, F.W.; Paustenbach, D.J.; Shu, H.P.; Murray, F.J.
(1986) Alternative approaches to mathematically analyzing the bioassay
data for 2,3,7,8-TCDD (Abstract 1133.) Presented at the 25th Annual
Meeting of the Society of Toxicology, New Orleans, LA.
Squire, R.A. (1980) Pathologic evaluations of selected tissues from the
Dow Chemical TCDD and 2,4,5-T rat studies. Prepared for the Carcinogen
Assessment Group, U.S. Environmental Protection Agency, Contract No.
68-01-5092.
Switzerland. (1982) Environmental pollution due to dimins and furans from
communal rubbish incineration plants. Federal Office of Environmental
Protection, Bern.
Thiess, A.M.; Frentzel-Beyme, R.; Link, R. (1982) Mortality study of persons
exposed to dioxin in a trichlorophenol-process accident that occurred in
the BASF on November 17, 1983. Am. J. Ind. Med. 3:179-189.
Tschirley, F. (1986) Dioxin. Scientific American 25(2):29-35.
U.S. Environmental Protection Agency (EPA). (1980) Risk assessment on (2,4,5-
trichlorophenoxy)acetic acid [2,4,5-T], (2,4,5-trichlorophenoxy) propionic
acid, and 2,3,7,8-tetrachlorodibenzo-jD-dioxin [TCDD], Office of Health and
Environmental Assessment, Washington, DC. EPA 600/6-81-003. NTIS
PB81-234825.
U.S. Environmental Protection Agency. (1985) Health assessment document for
polychlorinated dibenzo-_p_-dioxins. Office of Health and Environmental
Assessment, Washington, DC. EPA/600/8-84/014F. NTIS PB86-122546/AS.
U.S. Environmental Protection Agency. (1986) Guidelines for carcinogen risk
assessment. Federal Register 51(185)33992-34003.
56
-------
U.S. Environmental Protection Agency. (1987) Chapters. Epidemiology: recent
Kodak study. In: Technical analysis of ne^ methods and data regarding
dichloromethane hazard assessments. Office of Health and Environmental
Assessment, Washington, DC. External Review Draft. EPA/600/8-87/029A.
U.S. Environmental Protection Agency. (1988) Estimating exposures to 2,3,7,8-
TCDD. Exposure Assessment Group, Office of Health and Environmental
Assessment, Washington, DC. External Review Draft. EPA/600/6-88/005A.
Weil E.S. (1972) Statistics vs. safety factors and scientific judgement in
evaluation of safety for man. Toxicol. Appl. Pharmacol. 21:254-463.
Zack J.; Suskind, R. (1980) The mortality experience of workers exposed to
TCDD in a trichlorophenol process accident, J. Occup. Med. 22(l):ll-rl4.
57
-------
-------
APPENDIX
QUANTITATIVE DOSE-RESPONSE MODEL
FOR THE TUMOR PROMOTING ACTIVITY OF TCDD
Prepared by:
Todd W. Thorslund, Sc.D.
ICF-Clement Associates
1850 K Street, N.W., Suite 450
Washington, D.C. 20006
For: The Carcinogen Assessment Group
(Dr. Steven Bayard Project Manager)
Research and Development
U.S. Environmental Protection Agency
EPA Contract No. 68-01-6939
February 17, 1987
-------
-------
1. TECHNICAL SUMMARY
A method is developed in this report for the estimation of
cancer risk associated with exposures to TCDD under the assump-
tion that TCDD acts exclusively as a tumor promoter. It is
recognized that TCDD may have other potential effects on the
mechanisms of carcinogenesis; however, evidence suggests that
stimulation of cell growth by direct or indirect mechanisms could
play an important role in TCDD's carcinogenic effect.
TCDD's ability to affect cell proliferation can be described
using a mathematical dose-response model. A very explicit defi-r
nition of promotional activity can be used within the context of
the Moolgavkar-Knudson-Venzon two-stage model to accomplish this.
It is assumed that a promoter can act by increasing the net
growth rate of preneoplastic, initiated stem cells and has the
ability to increase the growth rate to a fixed upper bound. The
difference between this upper bound and the background growth
rate is the maximum increase possible. The mathematical function
that defines the fractional part of the change in the maximum
increases in the growth rate due to a given exposure level of an
agent is the critical element in the derivation of a dose-response
relationship. Information that can'be used to elucidate the form
of the critical function that defines the exposure-dependent cell
growth rate comes from the three sources: (1) theoretical biolog-
ical arguments, (2) the shape of tumor dose-response relation-
ships from TCDD carcinogenesis bioassays, and (3) studies of
-------
TCDD-induced cell proliferation rates as direct or indirect
measures of cell growth in vivo or in vitro.
In this report, two separate parametric models describing
the dose-dependent changes in cell growth are postulated. Each
model is consistent with a series of studies on the mechanisms of
action of TCDD. The first growth rate model assumes that the
changes follow first order kinetics. This assumption results in
a negative exponential model that contains only one parameter
requiring estimation. This model is fitted to the most extensive
data set available (liver tumors in female Sprague-Dawley rats)
in order to estimate the three unknown parameters in the tumor
dose-response model. The validity of the resulting model is
tested in four different ways: (1) its goodness-of-fit to the
tumor data, (2) its consistency with TCDD-induced proliferation
data, (3) its ability to predict the TCDD-induced tumor response
in other sexes and;strains of the same species (rats) using
adjustments only for background tumor rates, and (4) its ability
to predict the TCDD-induced tumor rates in another species
(mice), making adjustments for background rates and estimating a
separate growth rate parameter for that species.
A second model is investigated that assumes the rate of
change of the TCDD-induced growth rate is proportional to the
product of cellular TCDD levels and the number of TCDD receptors.
The resulting model for the increased cell growth rate is log-
logistic in form. This model has two parameters, the slope and
-------
intercept; only the latter can be estimated with the available
tumor dose-response .data. In order to use this model, the slope
parameter needs to be specified. This is done by specifying
slopes that span the range of those determined from other types
of biological systems that are log-logistic in form. The
intercept associated with a slope is then estimated using the
bioassay data.
Both models are used to extrapolate from observed animal
tumor rates to expected human response rates. The age-specific
human death rates due to liver cancers are used to estimate two
of the parameters in the human model. The use of human data for
this purpose is mandated by two factors. Under the assumed
model, the agent acts on initiated, preneoplastic cells to
increase their number. As a result, the increase in the tumor
rate should be proportional to the background number of preneo-
plastic cells, which in turn is proportional to the background
tumor rate. This theoretical prediction is confirmed by the
strong dependence of the sensitivity of the tumor response to the
background tumor rates observed in five separate animal bio-
assavs. The usual assumption of the equivalent dose between
species being proportional to the cube root of the ratio of the
species weight is also employed in the development of the human
dose response model. Using the developed models it is
demonstrated that low-dose linearity can result from either form
of the models under certain circumstances. In contrast, is is
also shown that the log-logistic model with a slope equal to 3
gives prediction of risk that decrease 3 orders of magnitude for
each order of magnitude of reduction of exposure.
-------
• - 4 -
While both forms of the M-V-K model fit the observed data
quite well, and both can be justified on theoretical and some
experimental ground, their use for low-dose extrapolation leads
to a wide variation in risk estimates. For a lifetime daily dose
of 0.1 ng/kg-day, or one order of magnitude below the animal
experimental level, the estimates vary by a factor of more than
10,000.
Although the negative exponential form of the model results
in a linear low-dose response relationship, the risk estimates
derived from it are two orders of magnitude below the upper bound
values obtained by EPA when TCDD was modeled as a complete carci-
nogen. This difference is primarily due to the low human back-
ground liver cancer rates as compared to those of the female rat.
Thus, treating TCDD as a promoter could have a strong impact on
any regulatory decision. However, the model is based upon the
assumption that the site of cancer induction in humans would also
be in the liver. An analysis of tissue dose distribution and cell
turnover rates and receptor protein levels for other organs or
tissues should be conducted before the equivalence of target site
between- humans and rodents is accepted unequivocally.
2.
MATHEMATICAL DOSE RESPONSE MODEL FOR PROMOTERS
A mathematical model has been developed by Moolgavkar,
Venzon and Knudson (1979, 1981) that is based on a two-stage
model for carcinogenesis. This model is depicted in Figure 1.
-------
w
Q
O
S
W
Q
3
W
a.
w
Q
I
w
s
H
EH
Q
u
EH
CO
o
H
o
H
CQ
W
2
W
JJ
te
X
1
W
en
o
X
H
O
&
H
>l
l^
OH
1
4J — 1
03 rH
^ OJ
-H U
t.
(i)
73 CT
(0 (0
Q) 4J
Q W
0)
c
o
.H
o
(0
TJ
C
(0
X!
-P
•rH L-l
rH • -H
(C S-l £!
f i
ti ^
a-p
Q)
4J
(0
(U
•o
c
03
£3
-P
^i
•H
(C
Q)
T5
0) n}
D Q) 0)
g (0 rH
H
d g 3
tn r< in
o o o
C^4 CH
X X
0) 0) 0)
0) 4J (1)
[ j ti I)
O
<1) IH 0)
• .p
H c
-H (0
Q) -P
o in
c
Q) O
tn o
(0
-P a)
tn ja
i
-P O
tfl -p
^4
•H ^
Q)
%
tn
tn
(0
0}
o tn
tit ,_j
rc tn
Q)
-^.p
4J (0
x
*>*• c
oo
T3 <0
C -P
(0 3
g
tn to in ^
- a o
rH O ^
(0 Q) Q)
g c o
r4 O C
O M fO
c a o
nj (0 (0
tn tn w
.. -H -H -H
PH
W O »H O
x" c
^ o
O-H
CQ 4->
-rH
- tn
4j" (0
- M
X -P
o
Q
03 tfl XI
Q) U
Z>* ro XJ
x"
•— c -
H O rH
S -H Q)
-P >
TJ rd Q)
C -P rH
(0 3
g Q)
*—^ SH 3
-P o tn
o
x c a
•—ox
rH-H Q)
ffl -P
-H OJ
- tn xl
— c -P
^ ro
^ M tn
x -P -H
-p
O
0)
tn
0)
-P
<*H
o
0)
Cn
ro
d)
-P
X -P
-------
According to the model, the population of cells at risk is proli-
ferating cells, often referred to as stem cells. Stem cells are
those from which most other cells in an organ arise; once a cell
has differentiated and left the pool of proliferating cells it is
no longer susceptible to heritable alterations of its DNA. A
normal, susceptible stem cell may do one of several things. It
may divided into two daughter stem cells, terminally differen-
tiate, die, or undergo mutation at a critical site that results
in formation of a preneoplastic or intermediate cell. A preneo-
plastic cell has undergone one of the changes necessary to become
a cancer cell but is not yet cancerous. The cancerous cell will,
after a sufficient length of time, divide into enough cells that
it becomes a detectable tumor. All of these processes can be
described mathematically by postulating specific rates for the
cell changes. Moolgavkar and Knudson (1981) showed that to a close
approximation, the age-specific tumor rate at age t for their two-
stage model may be expressed as follows:
I(t) = MjjMj J"G0(v)|exp [(B-D) (t-v) ]] dv
f"i *• '
(i:
where
Kt)
M,
0
M.
C0(v)
B
D
age-specific cancer incidence at age t?
transition rate from stem to preneoplastic
cell;
transition rate from preneoplastic to
cancerous cell;
number of susceptible stem cells at age v;
birth rate or rate of cell proliferation of
preneoplastic cells; and
death rate of preneoplastic cells.
-------
- 6 -
Essentially, the equation describes the progression from a
normal stem cell to a cancerous cell under the assumptions of the
two-stage model. This model is a combination of deterministic
and stochastic components. The numbers of preneoplastic and
fully malignant cells at any time are assumed to be random
variables that are dependent upon these event rates, while the
number of normal cells at risk of transformation at time t,
denoted by C0 (t) , is assumed to be deterministic and known.
The biological processes described by the equation 1 can be
affected by the level of exposure to carcinogenic agents, and are
more likely to occur as the length of exposure increases; as a
result, they are exposure and time dependent. Thorslund et al.,
(1987) have described an exposure- and time-dependent version of
the model. In the context of biological mechanisms of carcino-
genesis, the model can be used to predict the risk of agents that
exert their effects in a number of different ways. Mutation-
inducing initiating agents could affect the transition rates
between cell stages (MQ or F^), while promoting agents may
increase the proliferation rate of preneoplastic cells (increas-
ing B without affecting D) . Cocarcinogens may increase the
proliferation rate of normal stem cells (CQ), thereby increasing
the size of the target for initiating agents. Inhibitors could
remove cells from the populations of susceptible stem or preneo-
plastic cells through toxicity (e.g., increasing D without
affecting B) or by inducing differentiation.
-------
- 7 -
For the purposes of the model, a mechanism of action for
TCDD has been postulated that is similar to that of a promoting
agent. It has been assumed that TCDD exerts its carcinogenic
effect by increasing the preneoplastic cell birth rate. The
difference between the birth and death rates is the net preneo-
plastic cell growth rate and is dose-dependent. This rate is
denoted as:
G(x) = B + B (x) - D
(2)
where
G(x) = dose-dependent preneoplastic cell growth rate
B (x) = the increase in the birth rate due to TCDD
B,D are as defined in equation (1)
Upon maturity, the number of stem cells in the liver may be
viewed as relatively constant. Under this assumption, the term
C (v) may be taken to be equal to a non-time-dependent constant,
C . Substituting these definitions for G(x) and C (v) into
o o
equation (1) and integrating yields:
G(x)t-l]/G(x) (3)
which is the age-specific tumor rate expressed in a dose and time
dependent form. The probability of a tumor by time t in the
absence of competing mortality is obtained from the well-known
relationship:
P(x,t) = 1-exp- £ I(x,v)dv (4)
Substituting equation (3) into equation (4) and integrating
yields the expression:
-------
- 8 -
P(x,t) = 1-exp- {M[expG(x)t-l-G(x) t]/G(x)2) (5)
where M = M M-C is a composite parameter that is proportional to
the background tumor rate. The growth rate has a normal value
G(o) = B-D in the absence of exposure and has a finite upper
bound G () that is determined by how rapidly a cell can go
through its normal proliferative cycle. Using this notation, the
exposure-dependent growth rate of preneoplastic cells may be
expressed as:
G(x) = G(o) + [G(«o)-G(o)]R(x) (6)
where R(x) is the fractional amount of the maximal increase in
the growth rate that can be induced by exposure at level x. The
function R(x) is dependent upon the specific mechanism by which
TCDD induces cell proliferation. A specific form for R(x) can be
derived for each hypothesized mechanism of action. The next
section discusses the factors that need to be considered in the
selection of the R(x) to be used in equation (6) .
3 . MECHANISMS OF THE ACTION
Carcinogenesis is a multistep process which displays at
least two distinct steps. External agents (carcinogens) can act
to augment the process at each of these steps (Weinstein, 1981) .
For simplicity, carcinogens are often divided into two classes:
"initiators" (those that act at the first stage of cancer devel-
opment) and "promoters" (those that act at later stages and hence
-------
- 9 -
"promote" the action of initiators) (Weisburger and Williams,
1983; Weinstein, 1984) . However, recent studies have shown that
both initiation and promotion are more complex phenomena than is
reflected in this dichotomous classification (Becker, 1984;
Upton, et al., 1985; Gallagher, 1986). Promotion frequently
involves the action of promoters on cell membranes, including
binding to protein kinases, inhibition of intercellular communi-
cation, and other effects on membrane structure and function, but
also may involve genetic changes, including methylation of DNA,
modulation of expression of genes involved in cell differentia-
tion, activation of oncogenes, and chromosome damage (Shank,
1984; Yamasaki and Weinstein, 1985; Weinstein, 1984; Jones, 1986),
It is unlikely that TCDD is a typical initiator. There is
little evidence that it causes point mutations in bacteria (e.g.
in the Ames test) but it has some genotoxic potential, as
reflected by a clear positive result in the mouse lymphoma assay
(Rogers, et al., 1982). TCDD has little propensity for covalent
binding to DNA (Poland and Glover, 1979); however, it is known
that after binding of TCDD to a cytosol "receptor" protein, the
TCDD-receptor complex is translocated to the nucleus of the cell
where it interacts with different sites on DNA and can affect
various functions. These include regulating transcription of the
cytochrome P-448 gene (Israel and Whitlock, 1984; Whitlock, et
al., 1984; Jones, et al., 1985); regulating the levels of other
enzymes such as ornithine decarboxylase, DT-diaphorase, and UDP-
glucuronyl transferase; and affecting the regulation of cell
proliferation and differentiation by interfering with the ability
-------
- 10 -
of other sites on DNA to bind receptor complexes such as those of
epidermal growth factor and glucocorticoids (Poland and Knutson
1982) .
The latter two functions may be those by which TCDD affects
tumor promotion. Agents that increase the number of proliferat-
ing cells by increasing cell proliferation rates or decreasing
the number of cells that terminally differentiate could increase
the likelihood of mutational events by tumor initiators or permit
clonal expansion of initiated cells. The mechanism of enhanced
cell proliferation may involve the interaction of a single TCDD-
receptor protein complex with a single site on DNA (such as that
of a regulator gene), which would be consistent with a fractional
change in a growth rate function of the form R(x) = 1-exp-Vx.
Alternatively, TCDD's action may require the formation of
multiple complexes with roles at multiple sites on DNA, which
would suggest a log-logistic relationship for
R(x) = [l+exp-(I+S In x)]'1.
Studies performed to date evaluating the ability of TCDD to
cause hepatic cell proliferation are inconclusive. Conaway and
Matsumura (1975) , for example, found an increase in hepatic
nuclear 3H-thymidine activity as compared to controls of 56% or
94%, depending on the method of preparation, ten days after adminis-
tration of a single oral dose of 5 ug/kg TCDD to male Sprague-
Dawley rats. Dickens et al. (1981) saw an increase of 94% in
3H-thymidine activity associated with hepatic DNA five days after
-------
- 11 -
a similar dose although the difference was not statistically
significant due to the small number of animals used. These authors
also noted significant increases in liver net weight and relative
liver weight compared to body weight in treated animals. In a
similar experiment, however, Christian and Peterson (1983) saw no
such increases. Further studies are clearly needed to resolve
the discrepancies along with studies of the dose-response and
time-course relationships of any increases observed.
In the absence of detailed, reliable data on the dose-
related effects of TCDD on cell growth rates, we shall assume two
functional forms for R(x):
(1) R(x) = 1-exp-Vx which would be appropriate if the
proliferation effect was due to the interaction of a single TCDD-
receptor protein complex with a single site on DNA, or
(2) R(x) = [l+exp-(I+S In x)] which would be appropriate
if the interaction of multiple TCDD-protein receptor complexes at
multiple sites were required for stimulation of the growth rate.
4. RATIONALE FOR THE SELECTION OF THE DATA USED TO DEVELOP
TUMOR DOSE RESPONSE MODEL
A number of decisions regarding the most appropriate study,
tumor end point, time frame, pathological diagnosis, and exposure
parameter must be made in order to obtain the specific data set
that is used to estimate the parameters in the postulated dose
response models. The rationale for those decisions is discussed
in this section.
-------
- 12 -
4.1 Selection of Animal Tumor Data
As discussed in the Health Effects Assessment Document for
Polychlorinated Dibenzo-p-Dioxins (USEPA, 1985), several chronic
animal studies of TCDD have demonstrated enhanced liver tumor
rates. In its extrapolation from animal data to human cancer
risk estimates, EPA used the hepatic tumor rates in female
Sprague-Dawley rats from the Dow 2-year feeding bioassay (Kociba
et al., 1978). These data are discussed in some detail in both
the HAD and in the issue paper on the quantitative implication of
the use of different models. Every U.S. and foreign agency
except one (California) has used the DOW study as its pertinent
cancer study for extrapolating to humans. In this report, only
the liver response will be modeled. While the female rats in
this study also developed rare tumors of the hard palate/nasal
turbinates as well as infrequent lung tumors, neither of these
tumor types has been shown to be the result of promotion. On the
other hand, TCDD has been shown to be a potent promoter in the
rat liver (Pitot et al., 1980).
In its extrapolation procedure for TCDD, EPA uses the
pathology analysis results of two independent pathologists, Dr.
R. Kociba and Dr. R. Squire. Even though Squire's analysis
included more liver tumors in the control, low- and mid-dose
groups, the estimates of incremental increase in cancer risk
differed by less than 10%. This report will use the Squire
pathology analysis of the data. Although the higher background
-------
- 13 -
rates based on the Squire analysis lead to higher incremental
cancer risks for animals under a promoter model, the human back-
ground rates are substituted when the results are extrapolated to
humans. The final estimates derived for humans should be
similar, therefore, since they depend only on differences in
response between the animal dose groups, which were about the
same according to both pathologists.
In an attempt to adjust for high early non-tumor related
mortality in the high dose group, EPA eliminated from considera-
tion all animals dying prior to the 13th month, when the first
liver tumor was observed. This censored data set also will be
used here.
4.2 Selection of Exposure Variable
Rose et al. (1976) applied a simple first-order (i.e., one
compartment open) model to estimate the steady-state concentra-
tions of TCDD in the liver of female Sprague-Dawley Spartan rats.
Based upon their analysis, it was concluded that the administered
dose following oral exposure was proportional to the target organ
dose (i.e., steady-state liver dose) in the oral dose range of
0.01-1.0 ug/kg/day TCDD; however, actual steady-state doses were
not measured. Kociba et al. (1978) measured the levels of TCDD
at the end of a two-year carcinogenesis bioassay in five rats
from each exposure group receiving 0.001, 0.001 or 0.1 ug/kg/day.
Levels of 0.54, 5.1, or 24 parts per billion, respectively, were
observed. Portier et al. (1984) used these exposure levels to
-------
- 14 -
fit dose-response models to the tumor data. To extrapolate expo-
sure levels from those received in the bioassay to the lower levels
encountered environmentally, the relationship between administered
and target organ dose was assumed to be linear from 0.01 ug/kg/day
to zero. This approach has several problems. The results are
based on small sample sizes with considerable variability between
measurements. Potential biases may also have been introduced by
obtaining observations only at terminal sacrifice and by having
high tumor rates and liver weight increases at the highest dose
level, which distort the levels of TCDD per cell'when measured as
proportional to liver weight. As a result of these problems, the
average liver exposure level data will not be used to .derive dose-
response models. Nevertheless, since the value of the highest dose
used in the bioassay is the only one that would be altered by using
these data, it can be shown that they have virtually no influence
on the estimate that results from the models employed in the
subsequent analysis. Furthermore, the use of administered doses
is consistent with EPA's previous approaches so that direct
comparisons are more meaningful.
5. PARAMETER ESTIMATION
The two models that will be used to predict risk may be
expressed as:
P(x,t) = 1-exp-MJ[expG(x)t-l-G(x)t]/G(x)2 I (8)
G(x) = G(o) + [G(«o)-G(o) ]R(x)
_, v _ l~exp-Vx case 1
' ' ~ _ i
[H-exp-d+S In x) ] case 2
-------
- 15 -
The number of parameters to be estimated is 4 (i.e., M, G(o),
G(^, V) for the negative exponential preneoplastic cell growth
model (case 1) , and 5 (i.e., M, G(o), G («o) , I, S) for the log-
logistic preneoplastic cell growth model (case 2). This section
explains how the parameters were estimated and presents the results
obtained for the most informative dose response relationship.
As discussed in the previous section, the most reliable and
biologically meaningful data set that can be used to obtain a
dose response relationship for TCDD is liver tumors in female
rats surviving one year in the DOW study using the Squire
pathology analysis. This data set is shown in Table 1.
Of the four or five unknown parameters in the models, M and
G(o) do not depend on exposure. The parameter M is proportional
to the product of the background cell transition rates and G(o)
is proportional to the preneoplastic cell growth rates. If
time-to-tumor data were available, these parameters could be
estimated from control data. When x=o the exposure-time model
has the form:
P(o,t) = l-exp[-MG(t) ] (9)
where
G(t) = [exp(G(o)t)-l-G(o)
In the absence of reliable time-to-tumor data, G(o) may be
specified based upon knowledge of cell turnover rates or time
-------
- 16 -
dependent tumor occurrence. The approach taken here was to find
a range for G(o) that is consistent with the age-specific tumor
rate increasing from the 3rd to the 5th power of age. This range
was found by Cook et al. (1969) to be consistent with the tumor
registry data for most tumor sites in eleven different popula-
tions. Using the multistage model, the probability of a tumor
occurrence by time t may be expressed as
P(o,t) = l-exp-Atk where 4^k^6 (10)
An estimate of A can be obtained by fixing k and and substituting
a background rate estimate at a fixed time in equation (10),
yielding
-ln[l-P(o,t)]/tk = A (11)
The background rate is obtained from the vehicle control, which
results in an estimate of
P(o,t) = 16/85 = 0.1882
Substituting t=104, P (o, t) =0 . 1882 , and k=4 or 6 into equation
-9 -13
(11) gives the results A=l. 78x10 when k=4 and A=l. 65x10 when
k=6. To estimate G(o), we transform equation (9) to the form
= ln(M/G(o)2)+ln[expG(o)t-l-G(o)t] (12)
-------
- 17 -
which for G(o)t^3 is approximated closely by the simple rela-
tionship
= ln(M/G(o))+G(o) t
(13)
This equation is then equated to numerical values obtained from
equation (10) at t=52,104, which gives two linear equations and
2
two unknowns for each k that can be used to solve for ln(M/G(o) )
and G(o). These values are used in turn to estimate M. Follow-
ing this approach the values shown in Table 2 were obtained. The
purpose of this manipulation is simply to obtain a time-to-tumor
relationship for the M-K-V model that corresponds to that which
has been previously observed often in terms of the multistage model,
Taking the range of the parameters is done in order to investi-
gate the sensitivity of the assumed parameters in the final risk
estimates. The observed tumor data are not used to estimate G(o)
so that a valid goodness-of-fit test can be obtained from the
four data points, which is an added advantage of the approach.
The remaining two parameters, G («o) and V, are obtained by equating
the parametric form of the model to the observed rates at the two
highest doses and solving the resulting simultaneous non-linear
equations for two unknowns. This is done for both values of k
but since the resulting models give virtually identical final
results, only the values for k=4 are shown in Table 3, along with
the corresponding expected values under the model.
The log-logistic cell growth model is also fitted to the
data. This is done by assuming the slope is equal to 1,2, or 3
-------
- 18 -
common values in many biological systems and estimating the
intercept corresponding to each assumed slope value. As can be
seen in Tables 1 and 3, values of the parameters that provide the
best log-logistic fit to the data are S=3 and 1=14.5. The
results of this analysis are displayed in Table 3. The estimates
for G (°o) and I were obtained from the tumor response data using
the same methodology as was done for the negative exponential
model.
In the next section the validity of the fitted models is
tested in a variety of different ways.
6. EVIDENCE FOR THE VALIDITY OF THE OBTAINED DOSE RESPONSE
MODELS
The validity of the dose response models that were obtained
in the previous section can be evaluated in a number of ways.
Since the models are a subset of the M-V-K model, which has been
shown to have a remarkable ability to describe a variety of
different carcinogenic phenomena and age-specific cancer rates in
humans, a degree of acceptance for their application to TCDD
should be accorded on purely theoretical grounds.
In addition, the predictions of the model can be evaluated
using (1) the goodness-of-fit of the model to the tumor data from
which it was derived, (2) its consistency with TCDD-induced
proliferation data from separate experiments, (3) its ability to
predict the TCDD-induced tumor responses in other sexes and
-------
- 19 -
strains of the same species (rats) using adjustments only for
background tumor rates, (4) its ability to predict the TCDD-
induced tumor rates in another species (mice), making adjustments
for background rates and estimating a separate growth rate para-
meter for that species, and (5) its consistency with the predic-
tion that the slope of the linear relationship between the log of
the age-specific tumor rate regressed against age will be a
raonotonically increasing function of exposure.
In this section, the models are evaluated using the first
four of these criteria. The consistency of the dose-dependent
slope (i.e., criterion (5)) was not attempted due to time and
resource constraints.
6.1 Goodness-of-Fit of Models to DOW Female Rat Data
The goodness-of-fits of the hypothesized models to the data
set from which they were derived are shown in Table 1. As is
2
indicated by the X values, both models fit the data adequatelv.
A better fit could have been obtained by letting the background
rate parameter deviate from the observed background; however,
considerable attention has been paid to the fact that the low-
dose tumor response is lower than the control tumor response,
which has been suggested to imply that some type of unspecified
compensatory low-dose mechanism is operating (see Sielken, 1987).
The goodness-of-fit test for the model fitted using only three
data points has a more specific meaning and greater power than
-------
- 20 -
using all the exposure levels to obtain parameter estimates. The
null hypothesis in this case is that the low-dose information is
consistent with our hypothesized tumor dose response model. As
indicated previously, there is no evidence that is inconsistent
with this hypothesis, which is a stronger criterion than a
general goodness-of-fit model. The better fit of the log-
logistic model compared to the negative exponential model should
not be interpreted as suggesting that the former model is more
valid, since both models fit the-data adequately. Equivalently,
the slight improvement in fit due to increasing the slope in the
logistic model should not be viewed as strong evidence for a
large slope. We note in Table 3 that the log-logistic model fits
the data adequately for slope values from 1 to 3, even though, as
is indicated by the estimated rat cancer risks, the implications
for low dose extrapolation are very different. To illustrate the
type of results one might obtain with a log-logistic model, we
use the case with a slope of 3 in the final evaluation in addition
to the negative-exponential model.
6.2 Consistency with Cell Growth Rate Data
The form of R(x) (i.e., dose-dependent change in growth
rate) is obtained from theoretical considerations and the shape
of the liver tumor dose-response model. Ideally, the growth
rates of hepatocellular adenomas would be used to estimate the
parameters in R(x). The hepatocellular adenomas induced by TCDD
are thought to be a preneoplastic stage of hepatocellular
-------
- 21 -
carcinomas. A more practical but less direct alternative would
be to measure the TCDD-induced growth or turnover rates in normal
liver cells in the same species for which TCDD-induced tumor dose
response information exists. One measure that can be used in
this regard is the H thymidine levels incorporated in liver cell
nuclei after TCDD exposure. As cited by EPA (1985), page 8-20,
it was shown that the H-thymidine activity at control and a 5
ug/kg TCDD exposure levels were 29 and 45 cpm/mg liver,
respectively. We assume that the growth rate is proportional to
this index and that the maximum growth rate is obtained at the
5 ug/kg exposure level. If multiple exposures had been used, it
might have been possible to define the shape of R(x), as is
indicated in Figure 1. However, it is possible to obtain an
estimate of G( ) from the limited data available. The ratio of
the maximum to minimum growth rates (G( )/G(o)) can be estimated
by taking the ratio of the H thymidine counts. Multiplying this
ratio (45/29) by the assumed G(o)=0.0533 gives the value G( ) =
(45/29)xO.0533=0.0827, which compares closely with the value of
0.0817 obtained from the shape of the bioassay curve.
6.3 Consistency of Model With Other Rat Dose-Response Data
Under the assumed M-V-K model, a promoting agent acts on
initiated, preneoplas'tic cells to increase their number. As a
result, the increase in tumor rate should be proportional to the
background number of preneoplastic cells, which in turn is
proportional to the background tumor rate. The background tumor
-------
- - 22 -
rate parameter M should be different, therefore, for strains and
sexes that have varying background liver tumor rates. To test
the validity of the derived model, all parameters except M were
assumed to be known and the resulting model was fitted to the NTP
bioassay data for male and female Osborne-Mendel rats. The
results of this analysis are shown in Tables 4 and 5 for female
and male rats respectively. We note that the resulting model
gives an adequate fit to both data sets even though the non-
background tumor rates have little weight in the parameter
estimation.
6.4 Consistency of Model With Mice Dose-Response Data
Due to biological differences, we would expect that enzyme .
and receptors levels would vary between species. As a result, in
addition to background levels, a parameter for the exposure-
induced change in the growth rate V would also have to be
estimated when using a different species. For the male mice data
in the NTP study, two parameters (M and V) were estimated and the
other parameters taken from the previous model. For the female
data, only one parameter (M) was estimated while V was obtained
from the male data. The results of this analysis are shown in
Tables 6 and 7 for male and female mice, respectively. We note
that the resulting models also are consistent with the observed
data.
-------
- 23 -
7.
HUMAN DOSE RESPONSE MODEL
In this final section, human age-specific cancer death rate
data and parameters estimated from the animal bioassays are
combined to obtain a human dose-response model. The rationale
for this approach is that the background tumor rate parameter is
the critical factor in determining species sensitivity. This
rationale is supported by the results in Table 8, which demon-
strate the consistency of risks among species after adjusting for
background rates.
The human age-specific liver cancer death rates can be used
to obtain estimates of the human parameters M and G(o) under the
assumptions that most liver cancers are fatal and that these
rates represent actual background rates. Under the model, the
age-specific background tumor rates may be expressed as
=M[exp(G(o) t)-
(14)
Since we expect G(o)t>3 for most of the observed age range,
taking the natural log of eq. 14 yields
£Jln [M/G(o)]+G(o)t
(15)
Values for I(t) are derived from Table 1-25, Section 1, Vol. II,
Part A, of the Vital Statistics of the United States 1980 (U.S.
HHS, 1985) and the 1980 U.S. Census.
-------
- 24 -
Using these data, a simple linear regression of the form
In[I(t)]=A+Bt yields a correlation coefficient r=0.99 with values
A=-15.6870 and B=0.09381(years)~ (see table 9). M and G(o) are
then estimated by setting the linear regression equation equal to
eq.(15). Since A and B are now known, the intercept A=ln[M/G(o)]
and B=G(o). Estimates of the parameters are thus
M=B(expA)=1.44xlO~8 and G(o)=.09381 (years)'1 (16!
The maximum growth rate G (<*9 is assumed to be the same per-
cent change over background as was observed in the DOW female rat
data. Thus, for the human model G(«°) =0 . 09381x (0 . 0817/0 . 0533) =0 .1438 .
Finally, the parameters V and I for humans are obtained from the
DOW female rat study by using the usual surface area correction
(70/0.45) =5.38. The value of the human parameters so obtained
are shown in Table 10. The resulting mathematical extrapolation
model using these parameters and risks at various low environ-
mental exposure levels are shown in Table 11.
-------
Table 1 - Comparison of Observed and Predicted Liver Tumor
Rates in the DOW TCDD Feeding Study Using Female Rats
Exposure
ug/Kg/day
(X)
0.0
0.001
0.01
0.1
Number of
Animals
Surviving
First Year
85
48
48
40
i • 4
Number
Observed
16 (19%)
8 (16%)
27 (56%)
33 (82%)
2
X
d.f.
P value
(%) of Animals With Liver Tumors |
Predicted
1st order Equilibrium
16.0 (19%)
10.8 (23%)
27.0 (56%)
33.0 (82%)
0.96
1
0.32
Log-Logistic S»3
16.0 (19%)
9.4 (20%)
27.0 (56%)
33.0 (82%)
0.16
1
0.78
Parameters
Symbols
G(o)
G(<~)
M
V
I
S
Estimates
1st order] Log-Logistics
0.0533
0.0817 ,
2.3798x 10
109.51
14.50
3.00
Values Derived From
See text
Estimated From Data
Estimated From Data
Estimated from Data
Estimated from Data
Assumed
P(xft) - 1-exp-M [expG(x)t-l-G(x)t]/G (x) , t=104
G(x) * G(o) + [G(o*)-G
-------
Table 2 - Parameter Estimates For The Moolgavkar-Knudson
Model That Simulate Range of Multistage Time
to Tumor Model for Usual Range of Human Data
Assumed Value
of Time
Parameter
4
6
1980 U.S.
Vital Statistics
ICD 155.
Liver Cancer
Deaths
Resulting Parameter
Estimates
G
0.0533
0.0800
0.0631a
M
2.15xlO~6
3.03xlO~7
1.44xlO~8
Value adjusted by .0938(years)~ x(70 years/104 weeks)
-------
Table 3 - Effects of Varying Slope Parameter in Log-
Logistic Model on Goodness-of-Fit and Low-Dose
Risk Estimates Based on DOW Study
Parameter
Svmbol
M
G(o)
G(°o)
I
S
Augmented
Rat Risk at
x=lxlO~3
x=lxlO~4
x2
P
Parameter Estimate
2.3798 x 10~6
0.0533
0.0817
5.29
1
6.3xlO~2
6.6xlO~3
2.066
0.18
9.90
2
6.6xlO~3
6.6xlO~5
0.245
0.62
14.50
3
6.6xlO~4
6.6xlO~7
0.155
0.72
oo
0
0
0.146
0.73
I
I
-------
Table 4 -. Comparison of Observed and Predicted TCDD-Induced
Liver Tumor Rates In the Osborne-Mendel Female
Rats From the NCI Gavage Study..
Exposure
ug/kg/day
(X)
Historical
Control
0
0.0014
0.0071
0.0714
Number of
Animals
Exposed
970
75
49
50
49
Number (%)
Observed
21 (2.2)
5 (6.7)
1 (2.0)
3 (6.0)
14 (28.6)
2
X
d.f.
P value
of Animals with Liver Tumors
Predicted
2.8 (3.7)
2.4 (4.9)
5.4 (10.8)
13.2 (26.9)
3.99
3
0-.28
Values of Parameters (see text)
Symbols
G(o)
G(o«)
V
M
Biological
Interpretation Estimates
background cell
growth rate 0.0533
maximum cell
growth rate 0.0817
incremental rate
change per
unit dose 109.51
background _
mutation rate 4.3036 x 10~
Values drived from
Assumed (A priori)
Dow Study
Dow Study
Estimated From
Data
P(x) - 1-exp-M •[ [expG(x) t-l-G(x)t] /G
G(x) = G(o') + [G(<*>)-G(o) ] [1-exp-Vx]
-------
Table 5 - Comparison of Observed and Predicted TCDD-Induced
Liver Tumor Rates in Osborne-Mendel Male Rats from
the NCI Gavage Study.
Exposure
ug/kg/day
(X)
Historical
Control
0
0.0014
0.0071
0.0714
Number of
Animals
Exposed
975
74
50
50
50
Number (%)
Observed
9 (0.9)
0 (0.0)
0 (0.0)
0 (0.0)
3 (6.0)
x~
d.f.
P value
of Animals With Liver Tumors
Predicted
0.3 ( .4)
0.2 ( .5)
0.6 (1.2)
1.6 (3.2)
2.37
3
0.58
Values of Parameters (see text)
Symbols
G(o)
G(oo)
V
M
Biological
Interpretation
background cell
growth rate
maximum cell
growth rate
incremental rate
change per
unit dose
background
mutation rate
Estimates
0.0533
0.0817
109.51
4.420 x 10~8
Values drived from
Assumed (A priori)
Dow Study
Dow Study
Estimated
From Data
P(x) = 1-exp-M |[expG(x)t-l-G(x)t]/G2(x)]
G(x) = G(o) + fG( «=>)-G(o) ] [1-exp-Vx]
-------
Table 6 - Comparison of Observed and Predicted TCDD-Induced
Liver Tumor Rates in B6C3F1 Male Mice from the NCI
Gavage Study.
Exposure
ug/kg/day
(X)
0
0.00143
0.00714
0.07143
Number of
Animals
Exposed
73
49
49
50
Number (%) o
Observed
15 (20.5)
12 (24.5)
13 (26.5)
27 (54.0)
2
X
d.f.
P value
f Animals With Liver Tumors
Predicted
15.00 (20.5)
10.40 (21.2)
11.77 (24.0)
27.77 (55.5)
0.53
2
0.77
Values of Parameters (see text)
Symbols
G(o)
G(«o)
V
M
Biological
Interpretation
background cell
growth rate
maximum cell
growth rate
incremental rate
change per
unit dose
background
mutation rate
Estimates
0.0533
0.0817
13.19
2.6248 x 10~7
Values drived from
Assumed (A priori)
Dow Study
Estimated
From Data
Estimated '
From Data
P(x) = 1-exp-M
G(x) = G(o)
>-M | [expG(x)t-l-G(x)t]/G2(x) |
+ . [G(oo) -G(o) ] [1-exp-Vxl
-------
Table 7 - Comparison of Observed and Predicted TCDD-Induced
Liver Tumor Rates in B6C3F1 Female Mice from the
NCI Gavage Study.
Exposure
ug/kg/day
(X)
0
0.006
0.028
0.286
Number of
Animals
Exposed
73
50
48
47
Number (%) c
Observed
3 ( 4.1)
6 (12.0)
6 (12.5)
11 (23.4) .
2
X
d.f .
P value
if Animals With Liver Tumors
Predicted
.3.7 ( 5.1)
2.9 ( 5.9)
4.5 ( 9.4)
16.0 (34.0)
6.46
3
0.09
Values of Parameters (see text)
Symbols
G(o)
G(«<>)
V
M
Biological
Interpretation
background cell
growth rate
maximum cell
growth rate
incremental rate
change per
unit dose
background
mutation rate
Estimates
0.0533
0.0817
13.19
5.9884 x 10~7
Values drived from
Assumed (A priori)
Dow Study
NCI-Male Mice
Estimated
From Data
P(x) = 1-exp-M | [expG(x)t-l-G(x)t]/G2(x)
G(x) = G(o) + [GC*3 )-G(o) ] [1-exp-Vx]
-------
IQ
to
en
>,
m
(0
(0
(0
o
•H cn
PQ Q)
-P
C (0
IU «
OJ
CQ O
M
to tn
^ X
w o
•H (0
cc; CQ
u-4 m
o o
o o
(0 -H
-H 4J
M O
(0 td
(X M
Es fe
o
oo
a)
rH
JQ
It
EH
T~f
M
s
•H
£j
«•
CO
•H
T3
n
3
0
M
60
0
cd
PQ
>tf*
CO
•H
PJ
w
•H
Pi
3
O
M
61
O
cd
01
4J
cd
Pi
"O
C
3
O
M
60
O
Cd
«
<
u
cd
S
,
0)
.H
Jj
cd
o
i
<
OO
cr.
i — i
1
1
1
o r
1
0
— 1
X
1 — 1
o
CM
"d"
•
"*
o
•
o
") ~;
0
i — i
X
•
oo
0)
C!
CU
2
I
m
ON
o
i-H
»^-
O
•
o
r ^
0
^-<
X
CO
«
1— (
1 — 1
fa
CO
u
vo
PQ
0)
O
-H
S
01
"«
e
CU
•o
o
"x
oo
~-<
e»s
CO
CM
D
1
0
r— 1
X
oo
v^j-
CM
vO
•
CM
in
i^.
sf
m
o
CM
*
o
t-
0
t— J
X
«
"*
fa
CO
CJ
vO
PQ
CU
O
•H
cu
t — 1
S
-------
TABLE 9
Estimation of Background Human Lifetime Liver Cancer
Rate (M) and Background Human Liver Cell Proliferation
Rates G(o) by Fitting In[I(t)]=A+Bt=ln(M/G(o))+G(o)t
1980 U.S.
Liver Cancer
1980 U.S.
Age
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-74
75+
(Mid-point)
27.5
32.5
37.5
42.5
47.5
52.5
57.5
62.5
70
80
Deaths
ICD (155)a
36
55
53
95
173
289
522
699
1795
1805
Population
(thousands)
19,518
17,558
13,963
11,668
11,088
11,709
11,614
10,086
15,578
9,973
Rates/
100,000
0.1844
0.3132
0.3726
0.8142
1.5602
2.4682
4.4946
6.9304
11.5227
18.0970
All races, both sexes
In I(t)=-15.687+0.098381t
-------
Table 10 - Parameters in Human Dose Response Model
Parameters
Symbols
G(o)
G(oo)
M
V
I
S
Estimates
1st order I Log-Logistics
0.0938
0.1438
1.44 x 10"8
589.16
19.55
3.00
Values Derived From
Human Mortality Data
Proportional to DOW Female
Rats
Human Mortality Data
Surface Area x DOW Female
DOW Female + 3 x Log
Surface Area
Assumed
-------
Table 11 - Estimates of Low-Dose Incremental Ri
Exposed to 2,3,7,8-TCDD a'D'C'Q°
Risk To Humans
MODEL
Promotion
Dose
ng/kg-day
io-5
1C'4
io-3
1(T2
io-1
1
Negative
Exponential
1.7 x 10~8
1.7 x IO"7
1.7 x 10"6
1.7 x 10~"5
1.8 x 10~4
2.4 x 10"~3
Log-
Logistic
_—
—
-------
P60
T© "TCOD
TV*io*s
1O4
8
.1
M
0.0,
10
-------
2.—
Oft
M/CC
I - exp - M
wlnere -c » 104
-I -G(M
PROBABIUT/OF
0*01
s-.f? x 10
*7 ivrp
-2-
-------
Ho:
(|)
a
(2)
3— Ofos«rved *n<£ Postulated
H3"Tfi«f»*>«J'l>ve Co u 10^5. in, Rat*
E-gposove "ha ~T"GPD.,n ^ :
count" is Surro^«de -fov-
1 °
-fr*e.t/o«*l
Cells is
|4
H
in nuclei rat
TCPD
- 1-533
-------
Figure 4 "•
of
of
.7
.3
.2
.I
0.0 .001 .004 .006 .008.0(0 ,OIZ •0(4*0(6 -O(8
-------
U-S- L
ISS and
2S 30 35
40 45 50 55
*80 fy *
-------
References
Becker, F.F. The Direct and Indirect Effects of Promoters May
Depend upon the Nature of the Initiated Cell. In: Fujiki et
al. eds. Cellular Interactions by Environmental Tumor
Promoters. Jap Sci. Soc. Press. Tokyo, pp. 349-349. 1984.
Christian, B.J., and Peterson, R.E. 1983. Effects of
2,3,7,8-tetrachlorodibenzo-p-dioxin on [3H]thymidine
incorporation into rat liver deoxyribonucleic acid.
Toxicol. 8:133-146.
Cook, P., Doll, R., and Fellingham, S.A. (1969). A mathematical
model for the age distribution of cancer in man. Int. J
Cancer, 4, 93-112.
Dickens, M., Seefeld, M.D., and Peterson, R.E. 1981. Enhanced
liver DNA^Synthesis in partially hepatectomized rats pre-
treated with 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol.
Appl. Pharmacol. 58:389-398.
Gallagher, R.E. Biochemistry of Neoplasia. In Comprehensive
Textbook of Oncology, A.R, Roosa, M'.C. Robson, and S.C.
Schimpff, editors. Williams and Wilkins, Baltimore.
Chapter 3. 1986.
Israel, D.I., and J.P. Whitlock. Regulation of Cytochrome P-450
Gene Transcription by 2,3,7,8-Tetrachlorodibenzo-p-dioxin in
Wild Type and Variant Mouse Hepatoma Cells. J. Biol. Chem.
259:5400-5402. 1984.
Jones, B.L., D.R. Galeazzi, J.M. Fisher, and J.P. Whitlock.
Control of Cytochrome P-450 Gene Expression by Dioxin.
Science 227:1499-1502. 1981.
Jones, P.A. DNA Methlation and Cancer.
February, 1986.
Cancer Res. 46, 461-466,
Kociba, R.J., D.G. Keyes, J.E. Beyer, et al. 1978. Results of a
two~year chronic toxicity and oncogenicity study of 2.3,7,8-
tetrachlorodibenzo-p-dioxin in rats. Toxicol.' Appl.
Pharmacol. 46(2): 279-303.
Moolgavkar S.H. and Venzon, D.J., "Two-Event Models for
Carcinogenesis:Incidence Curves for Childhood and Adult
Tumors," Math. Biosci. 47, 55-77 (1979).
Moolgavkar S.H. and Knudson, Jr., A.G., "Mutation and Cancer: A
Model for Human Carcinogenesis," JNCI 66, 1037-1052 (1981).
Pitot, H.C., T. GoIdsworthy and H. Poland. 1980. Promotion by
2,3,7,8-tetrachlorodibenzo-p-dioxin of hepatocarcinogenesis
from diethylnitrosamine. Cancer'Res. 40 : 3616-3620.~
-------
Poland, A. and Glover, E. An Estimate of the Maximum in vivo
Covalent Bonding of 2,3,7,8-Tetrachlorodibenzo-p-dioxin to
Rat Liver Protein, ribosomal RNA and DNA. Cancer Res. 39,
3341-3344. 1979.
Poland, A., and Kuntson, J.C. (1982). 2,3,7,8-Tetrachlorodibenzo-
p-dioxin and related halogenated aromatic hydrocarbons: Exami-
nation of the mechanisms of toxicity. Ann. Rev. Pharraacol.
Toxicol. 22, 527-554.
Portier, C.J., Hoel, D.G., van Ryzin, J. 1983. Statistical
analysis of the carcinogenesis bioassay data relating to the
risks from exposure to 2,3,7,8-Tetrachlorodibenzo-p-dioxin.
From Public Health Risk of the Dioxins, proceedings of a
symposium held October 19-20, 1983 at the Rockefeller
University, New York City.
Rogers, A.M., M.E. Anderson, and K.C. Back. Mutagenicity of
2,3,7,8-Tetrachlorodibenzo-p-dioxin and Perfludro-n-decanoic
acid in L5178Y Mouse Lymphoma Cells. Mutat. Res.
105:445-449. 1982.
Rose, J.Q., Ramsey, J.C., Wentzler, T.H., Hummel, R.A., and
Gering, P.J., 1976. The fate of 2,3,7,8-Tetrachlorodibenzo-
p-dioxin following single and repeated oral doses to the
rat. Toxicol. Appl. Pharmacol. 36: 209-226.
Shank, P.C. Toxicity-induced Aberrant Methylation of DNA and Its
Repair. Pharmacological Reviews 36(2): 195-245. 1984.
Sielken, R.L. 1986. Quantitative cancer risk assessments for
TCDD. Fd..Chem. Toxicol. (in press).
Thorslund, T.W., Brown, C.C., and Charnley, G., 1987.
Biologically Motivated Cancer Risk Models. Risk Analysis
7:109-119.
Upton, A.C., D.B. Clayson, J.D. Jansen, H.S. Rosenkranz, G.M.
Williams. Task Group 5 Final Report: Report of the ICPEMC
Task Group on the Differentiation Between Genotoxic and
Non-genotoxic Carcinogens. International Commission for
Protection Against Environmental Mutagens and Carcinogens
(ICPEMC) Biol.. Zbl. 104 417-453. 1985.
U.S. Environmental Protection Agency (USEPA). 1985. Health
Assessment Document for Polychlorinated Dibenzo-p-Dioxins.
EPA/600/8-84/014F.
Weinstein, I.E. Current Concepts and Controversies in Chemical
Carcinogenesis. Journal of Supramolecular Structure and
Cellular Biochemistry. 17:99-120. 1981.
-------
Weinstein, I.E. The Relevance of Tumor Production and Multistage
Carcinogenesis to Risk Assessment. In D.G., Hoel, R.A.
Merril, and R..F. Perera, eds. Risk Quantitation and
Regulatory Policy. Cold Spring Harbor, New York. 1984.
Weisburger, J.H., and G.M. Williams. The Distinct Health Risk
Analyses Required for Genotoxic Carcinogens and Promoting
Agents. Environ. Health Perspect. 50:233-245. 1983.
Whitlock, J.P., D.I. Israel, D.R. Galeazzi, and A.G. Miller.
2,3,7,8-Tetrachlorodibenzo-p-dioxin Regulates Cytochrome
P.-450, Gene Expression. In Banbury Report 18. Biological
Mechanisms of Dioxin Action. A. Poland, and R.D. Kimbrough,
editors. Cold Spring Harbour, N.Y. 1984.
Yamasaki, J., and I.E. Weinstein. Cellular and Molecular
Mechanisms of Tumor Promotion and Their Implications for
Risk Assessment. In Methods for Estimating Risk of Chemical
Injury: Human and Non-Human Biota and Ecosystems, V.B. Vouk,
G.G. Butler, D.G. Hoel, and D.B. Peakall, editors. John
Wiley and Sons, New York. 1985.
-------
March 1988
Review Draft
APPENDIX B
EPIDEMIOLOGIC CANCER STUDIES ON
POLYCHLORINATED DIBENZO-£-DIOXINS,
PARTICULARLY 2,3,7,8-TCDD
David L. Bayliss
Carcinogen Assessment Group
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
-------
-------
INTRODUCTION
Since the publication of the Health Assessment Document (HAD) for Poly-
chlorinated Dibenzo-p-Dioxins (U.S. EPA, 1985), many new epidemiologic studies
have been reported in the literature dealing with the risk of soft tissue
sarcoma (STS) and/or non-Hodgkin's lymphoma in users and producers of phenoxy-
acetic acid herbicides and chlorophenols. Furthermore, the major positive
studies have come under intense criticism. The earlier HAD document concluded
that there was limited epidemiologic data regarding the carcinogenicity of
polychlorinated dibenzo-p-dioxin-contaminated phenoxyacetic acid herbicides
and/or chlorophenols based mainly on the Swedish case control studies, but the
evidence regarding 2,3,7,8-tetrachloro-dibenzo-p_-dioxin (2,3,7,8-TCDD) was
judged inadequate due to the inability ,of any of the data up to that time to
demonstrate that the risk was due to 2,3,7,8-TCDD alone and not to one or more
of the other 74 isomers of polychlorinated dibenzo-_p_-dioxins found in the
phenoxyacetic acid herbicides and/or chlorophenols or the phenoxyacetic acid
herbicides or chlorophenols free of 2,3,7,8-TCDD.
This report is a review and brief analysis of the epidemiologic evidence
to date. . . . .
SOFT TISSUE SARCOMA
The main studies supporting the finding of an excess risk of STS were the
two independent Swedish case control studies of Harden and Sandstrom (1979) and
Eriksson et al. (1979, 1981). These investigators reported statisticially sig-
nificant (five- to sevenfold) elevated risks of STS from occupational exposure
to phenoxyacetic acid herbicides and/or chlorophenols either alone or separately,
-------
some of which are known to contain 2,3,7,8-TCDD as an impurity. Eriksson et
'
al. further subdivided their cases and controls into two categories based on
expected presence or absence of 2,3,7,8-TCDD among the phenoxyacetic acids
after removing from the group those persons that had been exposed to chlorophenol
and then recalculating the risk of STS. The relative risks were 17 in the
group exposed to 2,3,7,8-TCDD-contaminated phenoxyacetic acids [2,4,5-trichloro-
phenoxyacetic acid (2,4,5-T)] versus 4.2 in the group exposed to other phenoxy-
acetic acid herbicides [(2,4-dichlorophenoxyacetic acid (2,4-D) and 4-chloro-
2-methyl-phenoxyacetic acid (MCPA)] not believed to contain 2,3,7,8-TCDD. This
seemed to imply that either 2,4,5-T appears to be associated with a high risk
of STS or the polychlorinated dibenzo-p_-dioxin contaminants (such as 2,3,7,8-
TCDD) within are contributing to the high risk of STS.
The Swedish studies have been severely criticized in the scientific liter-
ature because of known or alleged methodology flaws and biases. These criticisms,
which are outlined in a report provided by the Australian Royal Commission (1985)
on the Use and Effects of Chemical Agents on Australian Personnel in Vietnam
include the following: recall bias, unreliability of the exposure data, infor-
mation bias, observation bias, absence of a significant risk at any single
specific cancer site, significant confounding factors, and lack of support from
other studies. (Although not specifically mentioned in the report, the Eriksson
et al. case-control study is included in this critique by inference.)
Some of the observations of the Australian Royal Commission have validity.
Recall bias certainly may have been present. Persons who have been diagnosed
with a health problem hypothesized to be associated with exposure to a given
agent such as a phenoxyacetic acid herbicide are more likely to be reminded of
a possible link with that herbicide than is someone diagnosed with a different
health problem not hypothesized to be connected with that agent. Hence, they
-------
are more likely to "recall" that exposure when questioned about it. However,
given the rather exceptionally high risk ratios found by both investigators in
their separate studies, it does not seem likely that recall bias is the explana-
tion. At the risk ratios calculated the power exceeds 99% in both the Harden
study as well as the Eriksson study. Even if the positive responses given by
the authors to the possibility of exposure were incorrect on 33% and were
discarded, these studies would still retain 80% power i.e., confidence that the
recalculated significant Odds Ratio (OR) measures a true association. Recall
bias was discussed somewhat in the HAD for polychlorinated dibenzo-p-dioxins
(U.S. EPA, |1985).
Also, !there is a lack of substantiation of quantity of exposure to the
phenoxyacetic acid herbicides as well as to the chlorophenols. Harden described,
in his doctoral dissertation dealing with the same cases as in his case-control
study, which herbicides andichlorophenols his cases and controls were exposed to,
for how long, and when they were exposed; actual quantities were not provided
although latency was determined. Unfortunately, neither investigator saw a need
to provide|a differential analysis by latency or by quantity or length of exposure
to the herbicides and/or chlorophenols. But, of course, at the time, the authors
probably did not think such an analysis would be important. They did not have
the vision of hindsight. Harden and Sandstrom (1979) and Eriksson;et al.
(1979, 1981) contacted employers to verify exposures reported in responses to
questionnaires sent to study members and survivors. The response from employers
regarding use of phenoxyacetic acid herbicides was, "uncertain and difficult to
evaluate" due to poor record keeping by the employer. But for chlorophenol
there was "good agreement" with the statements given by examined persons.
With respect to the possibility of information bias being present, this
could certainly lead to an enhancement of the risk ratios as persons with STS's
-------
would more likely assume their disease was connected with exposure to herbi-
cides because of intense media publicity. However, in a rebuttal to the con-
clusions of the Australian Royal Commission, Hardell reminded the Commission
that debate about phenoxyacetic acid herbicides had abated with the banning of
2,4,5-T in Sweden in 1977, thus reducing media coverage (Axelson, 1986).
Hardell pointed out that the studies were undertaken during the period from
1978 to 1981.
For the purpose of assessing the possible presence of observation bias in
his study, in 1981 Hardell repeated his analysis using colon cancer cases but
with the same controls as in the first study. Hardell found no association
between exposure to phenoxyacetic acid herbicides and/or chlorophenols and the
risk of colon cancer. He pointed out that this finding is inconsistent with the
occurrence of observational bias in the assessment of exposure (Hardell, 1981).
The issue of lack of site specificity with respect to the occurrence of
STS is clearly not an unexpected phenomenon. Carcinogens are not always site-
specific. There are many examples of multiple site carcinogens (1,3-butadiene,
vinyl chloride, and arsenic to name a few). Furthermore, the human body con-
sists of two main types of tissue, i.e., epithelial and connective. Epithelial
tissue constitutes all the duct glands, skin, the GI tract, neural tube, the
respiratory tract, etc. Malignant tumors originating in these tissues are called
carcinomas. Connective tissue, on the other hand, consists of bone, cartilage,
fat, muscle, and subcutaneous tissue. Malignant tumors originating in connec-
tive tissues are called sarcomas. With respect to appearance, a carcinoma is a
hard, adhesive mass with a well-defined, advancing front. Microscopically, a
carcinoma of one organ is distinguishable from that of another. In contrast,
sarcomas are soft, jellylike, and fall apart easily. Microscopically, they are
not clearly distinguishable by site. Corresponding to the diverse nature of
-------
connective tissue throughout the body, sarcomas could be found anywhere in the
body. They have poorly demarcated anatomic boundaries in contrast to carcino-
mas, and they contain large quantities of intercellular materials. As such,
they are subject to the same kind of insult as is the intercellular material of
epithelial tissue.
With respect to other significant confounding factors in these studies,
there appears to be little substantiation of that criticism. Harden certainly
controlled for age, sex, place of birth, and place of death. He further analyzed
smoking habits and found them not to be different between the cases and controls.
Hardell does state, however, that other pesticides might constitute confounding
factors. The Commission's statement regarding the presence of "other signifi-
cant confounding factors" is a quantum leap from Hardell's commentary about his
own data. This appears to consist mostly of inuendo rather than substantive
criticism.
The term STS is a convenient rubric under which all sarcomas arising out
of connective tissue have been classified. Individual tumor types, i.e.,
rhabdomyosarcoma, fibrosarcoma, liposarcoma, etc., are subtypes that identify
a specialized connective tissue tumor. In the Ninth Revision of the International
Classification of_ Diseases and Causes of_ Death, STS's are coded mainly to
category 171x; however-, under certain circumstances, they are coded to several
other diffuse categories as well. If the STS is not classified to a specific
site, it is given an ICD code of 171.9. If the STS is coded to a site of the
body other than certain organ sites, it is still coded to the 171x category.
However certain STS's of specific organ sites are coded to.cancer of that site.
This could serve to reduce observed STS deaths that may be due to exposure to
polychlorinated dibenzo-p_-dioxin and spread them out over several site-specific
categories. However, since expected deaths based on population death rates in
-------
category 171x would be subject to the same degree of difference as observed
deaths, one would not expect that the calculations of relative risks would be
affected although the ability to detect such elevated risks as significant
would be reduced. Clearly, electron microscopy or use of immunohistological
techniques are needed to distinguish a sarcoma at one site from a sarcoma at
another site, but these techniques are rarely used. This difficulty in diagnosis
is compounded by the fact that even collectively STS's are an exceedingly rare
form of cancer in the population. Age-adjusted (1970) standard incidence rates
based on category 171x during the period from 1973 to 1977 were estimated by
the Surveillance, Epidemiology and End Results (SEER) group of the National
Cancer Institute to be 1.9 per 100,000 in the United States (Young et al.,
1982). The corresponding age-adjusted mortality rate for ICD category 171x was
0.9 per 100,000 (Table 1). STS rates generally follow the pattern of other
site-specific cancer rates by age, starting off low in younger age groups (less
than 2 per 100,000 in persons under 50) and increasing with age to a maximum of
12.2 per 100,000 in persons aged 85 and older. Mortality rates remain under 1
per 100,000 up to age 50 but increase gradually to a high of 7.7 in persons age
85 and older.
Additionally, the literature is replete with studies of carcinogens that
cause rare cancers. These did not require confirmation before the scientific
community was convinced of their authenticity (e.g., DES and vaginal clear-cell
adenocarcinoma; vinyl chloride and angiosarcoma of the liver and glioblastoma
multiforme of the brain).
Furthermore, although the Ninth Revision of the International Classification
of Diseases and Causes of Death has assigned a specific category (171x) to this
cause, it is not always certain that the correct diagnosis will be made. This
was ably pointed out by Fingerhut et al. (1984) in a discussion of seven case
-------
TABLE 1. AVERAGE ANNUAL CRUDE, AGE-SPECIFIC, AND CUMULATIVE (age 0-74)
INCIDENCE AND MORTALITY RATES FOR MALIGNANT CANCER
INCLUDING in situ CASES BY PRIMARY SITE, ALL RACES, BOTH SEXES, ALL AREAS
(EXCLUDING PUERTO RICO),
Age group
Crude
Adjusted
Cum (0-74)
<5
5-9
10-14
15-19
20-24
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
75-79
80-84
85+
Soft tissues
Incidence
1.9
1.9
0.2
1.2
0.4
0.5
0.8
0.9
1.1
1.3
1.5
1.5
1.9
2.7
3.5
4.2
4.3
7.0
8.0
9.3
12.2
(including heart)3
Mortality
1.0
0.9
0.1
1.1
0.5
0.2
0.3
0.3
0.3
0.4
0.6
0.7
0.7
1.4
1.8
2.0
2.9
3.5
4.1
4.7
7.7
1973-1977
Non-Hodgkin's
Incidence
8.9
9.0
0.8
0.6
0.9
1.0
1.0
1.6
2.3
2.7
3.6
6.1
8.9
13.6
18.9
27.3
34.8
45.8
51.3
53.2
51.7
lymphoma3
Mortality
4.9
4.8
0.4
0.2
0.4
0.4
0.6
0.5 ,
0.7
1.0
1.5
2.2
4.2
6.4
9.9
13.6
19.0
27.3
34.0
39.5
40.8
aper 100,000 population
SOURCE: Young et al., 1982.
-------
reports of STS. Two of the seven were shown to be carcinomas. This difficulty
and the exceptionally rare nature of this tumor creates methodological problems
in the assessment of risk. As most trained epidemiologists are aware, the use
of the cohort technique to estimate risk of rare cancers is not recommended.
Even the detection of not-so-rare cancer risks requires the generation of thou-
sands of person-years following the completion of a suitable latent period.
Hence, cohort analyses of rare cancers are inappropriate vehicles for the
assessment of risk. They lack power and are insensitive.
The issue of latency is of importance here. Several studies have been
published that purport to show no association between exposure to 2,3,7,8-TCDO
and the risk of STS as well as to most site-specific cancers. Part of the
explanation for this may be the short period of time that has elapsed between
onset of exposure and clinical manifestation of disease, in this case, cancer.
In most nonpositive epidemiologic studies, this period has been under 10 years.
This short period of time is probably inadequate to assess the risk of most
site-specific cancers including STS. Few cancers have been shown to have a
latent period under 10 years. Hueper and Conway (1964) suggested a latent
period for the development of sarcomas of between 15 and 30 years. In the
positive epidemiologic studies of Hardell and Sandstrom and Eriksson et al.,
the latent period has been found to range from 9 to 27 years after initial
exposure to the phenoxyacetic acid herbicides and/or chlorophenols with a
median time lapse of 20 years. In the Lynge (1985) study, the lapse of time
from start of employment (exposure) to diagnosis ranged from 14 to 26 years if
one excludes the one STS in that study with a lapsed period of time from initial
employment to diagnosis of only 5 years. That is not likely to be due to
exposure to 2,3,7,8-TCDD-contaminated chemicals. If one assumes that the
latent period based on these positive studies is between 14 and 27 years, then
8
-------
it is not likely that an excess risk of STS will appear until at least the 15th
year after initial exposure. This risk should increase until around the 20th
year. Probably 20 years should be considered the latent period for the develop-
ment of STS from exposure to 2,3,7,8-TCDD-contaminated chemicals if one accepts
the hypothesis of a causal relationship which by no means is a certainty.
It is thus remarkable that several cohort studies of individuals exposed
to 2,3,7,8-TCDD-contaminated herbicides and chlorophenols have produced any
STS's. And, of course, the expectation of their appearance is so small that
significance tests cannot be done. The method of choice then is the case-
control technique. However, the case-control technique is also frought with
pitfalls, many of which are apparent in the evaluation of the epidemiology data
on 2,3,7,8-TCDD. With this technique, controls are matched with actual cases
usually on the basis of known confounders (i.e., age, race, sex) or suspected
confounders (i.e., time of death, length of employment, and so on). Uncontrolled
confounding, known or unknown, can play a significant role in the determination
of a risk, as well as other factors such as latency or selection of an appropriate
index of exposure if exposure cannot be directly measured. The net result can
lead the researcher to attribute a significant risk, if found, to the exposure
being measured instead of to a confounder which may be the real culprit.
Second, if an index of exposure is used that does not truly measure the actual
f
exposure, misclassification will result. This will force the risk estimate
toward the null. Hence, if a true risk is present, it will not be found.
Persons with no actual exposure could be classified as "exposed" while persons
with actual exposure could be classified as "unexposed." Hence, it is always
better to choose a surrogate that is as close as possible to the target organ
/
dose. For example, human tissue levels of 2,3,7,8-TCDD would be a far better
surrogate of exposure than would information that a person was located in or
-------
near a spot where 2,3,7,8-TCDD was present.
Biases can also increase or reduce the risk calculation depending on their
nature. There is really no way to control completely for some biases, i.e.,
recall bias and, to some extent, even information bias when dealing with ques-
tionnaire data no matter how hard the researcher might try. And these will
always loom as potential problems in any case-control study, no matter how well
conducted and designed it is. On the other hand, some biases could be elimina-
ted by a suitable selection of controls. For example, if cancer controls are
used, the researcher is obligated to eliminate those cancer controls where it
is suspected that the cancer might be associated with the exposure under inves-
tigation. This is definitely within the power of the investigator. Based on
the previous arguments there is no real basis upon which to conclude that the
Hardell and Sandstrom (1979) and Eriksson et al. (1979, 1981) studies are not
good scientific studies of the risk of STS.
Several other studies provided some support to the finding of an excess
risk of STS among individuals exposed to 2,4,6-trichlorophenol (TCP) and/or
2,4,5-T. Zack and Suskind (1980), in a small cohort study, noted that one
worker among 121 workers accidentially exposed to TCP (contaminated with 2,3,7,8-
TCDD) in 1949 developed STS's. All 121 were chosen for this study because they
developed chloracne which indicates substantial exposure to 2,3,7,8-TCDD.
No STS's would have been expected.
Lynge (1985), in an incidence study of 3,390 males employed in two factories
manufacturing phenoxyacetic acid herbicides, chiefly 2,4-D and MCPA, found a
nonsignificant excess risk of STS in male employees. The author stated that
these results supported the Swedish observation of an increased risk of STS
following exposure to phenoxyacetic acid herbicides "unlikely to be contaminated
with 2,3,7,8-TCDD." However, after a 10-year latency, the excess of STS was
10
-------
significant (4 observed vs. 1.00 expected; p < 0.05) in male employees of the
single factory where 2,4,5-T had been produced and used and where all five
STS's arose. However, the author cautions that because of the limited amount
of 2,4,5-T processed at that factory, exposure is unlikely although not impossible.
In a study of 2,189 manufacturing employees of a chemical plant that
produced TCP, 2,4,5-T, and other chlorinated phenols contaminated with hexa-
chlorinated to octachlorinated dioxins, Cook et al. (1986) reported one STS
which was previously found by Fingerhut et al. (1984) in a case review, through
pathology evaluation, not to be an STS. However, in a 1985 letter to the
Carcinogen Assessment Group of the U.S. Environmental Protection Agency,
Cook reported that another member of the cohort was found to be suffering from
a confirmed STS. This case was not reported as an STS because the person died
after the end of the follow-up period. Cook et al. did not estimate how many
deaths should be expected. Latency was not considered although it appears that
a sufficient follow-up had been achieved in which latency could have been
evaluated for the risk of cancer at other sites. The authors believe that
their data, provide little evidence of a cancer risk from exposure to 2,3,7,8-
TCDD especially STS.
In a later update of this same cohort by Ott et al. (1987), the authors
employed a mathematical analytical technique known as a "serially additive
expected dose model" designed by Smith et al. (1980). This statistical device
has several limitations just as does the commonly accepted method of simply
analyzing for latent effects by calculating observed and expected mortality
after a lapse of a sufficient number of years from initial exposure. Despite
the subjective designation of jobs according to an "intensity of exposure"
scale from 0 to 4, which is not based on actual exposure and is subject to
misclassification bias, it still takes 15 to 20 years for the development and
11
-------
detection of a malignancy! If few persons in the cohort have achieved that
latent period, no excess cancer risk will be detected no matter what the
"intensity" of exposure. This method also assumes that one year of exposure at
a high intensity score is equivalent to 10 years of exposure at the next lower
score in its ability to produce cancer. This is not a valid assumption because
such an equivalency fails to consider that a brief excursion at a very intense
exposure could affect the metabolism differently from that of a long continuous
(or discontinuous exposure). This device, although limited in some respects,
allows for the assessment of latent effects for certain sites. Only total cancer,
stomach cancer and lymphomas were analyzed for latent effects in this manner
due to the "paucity of deaths through 1987." Except for total malignant neo-
plasms, only six stomach cancer deaths and six lymphomas were separately analyzed.
The authors saw no trend of increasing risk of cancer at the few sites examined.
However, it must be kept in mind that with a long latent period required for
the manifestation of cancer, the power to detect a significant risk of cancer
at these two sites must of necessity be small, even for this technique, although
the authors did not estimate it. Of course, no other cancer sites were analyzed
for latent effects by this method. Of interest is the persistently high signi-
ficant risk of "other and unspecified" cancer (12 observed vs. 4.6 expected,
C.I. 135-456) for which the authors could offer no explanation. This cohort or
some portion of it wil be included in the National Institute for Occupational
Safety and Health (NIOSH) registry of exposed workers that will be analyzed by
NIOSH in the near future. One question remains, however, and that is, if
2,3,7,8-TCDD tissue levels in a substantial portion of these workers are similar
to 2,3,7,8-TCDD tissue levels in the background referent population and always
have been (thus indicating little evidence of differential exposure), an elevated
risk of cancer would not be expected to occur. It is possible that due to the
12
-------
7-year half-life of 2,3,7S8-TCDD (Pirkle et al., 1987), tissue levels that were
elevated in the past from early exposure to 2,3,7,8-TCDD have returned to
background levels, but of course, there is no evidence that tissue levels of
2,3,7,8-TCDD in the past were ever elevated (not always the same as they are
today), and that possibly no elevation ever occurred even during times when
potential exposure was hypothesized.
Hence, because of this possibility, and the knowledge that the presence of
chloracne at some time in the past more than likely is proof that a substantial
exposure to 2,3,7,8-TCDD occurred, it seems likely that a cohort consisting en-
tirely of chloracne persons having known contact with 2,3,7,8-TCDD in the past
would provide the most ideal cohort for assessing the long-term health effects of
exposure to TCDD! This is not the same as saying that the presence of chloracne
is.necessary in order to have exposure. The major problem with current efforts
to determine tissue levels of 2,3,7,8-TCDD is that they don't reveal what the
tissue levels were in the past.
Smith et al. (1984) conducted a case-control study of STS in New Zealand
and found a significantly high risk of STS among railroad workers. Specific
exposure to phenoxyacetic acid could not be identified. Tannery and meat
workers were also found to be at a significantly high risk of STS. In those
jobs among meat workers where pelts are treated with chemicals such as TCP
(2,4,6-trichlorophenol) that contain 2,3,7,8-TCDD, the risk ratio increases but
is marginally significant because of small numbers. Although Smith et al. found
excess nonsignificant risks in applicators and users who were exposed to only
phenoxyacetic acid herbicides and/or chlorophenols, their personal exposure
information is not substantiated, similar to the Swedish studies. In a later
update of this study (Smith and Pearce, 1986), which included more cases and
controls that had been diagnosed, the excess was reduced to the null. The
13
-------
authors have been criticized by Axel son (1986) for failing to exclude from the
referent group cancers that may be related to the exposure in question (i.e.,
lymphomas), thus potentially underestimating the true risk. Smith et al.
concluded that at the levels of exposure experienced by individuals in contact
with ground spraying, it is unlikely that 2,3,7,8-TCDD could cause STS.
Other studies that are consistent or are not inconsistent with the finding
of an elevated risk of STS among persons potentially exposed to polychlorinated
dibenzo-£-dioxin-contaminated herbicides and/or chlorophenols are Balarajan and
Acheson (1984), Cantor (1982), Milham (1982), Kogan and Clapp (1985), Michigan
Department of Public Health (1983a, b), Fett et al. (1984), and most recently,
Puntoni et al. (1986), Merlo and Putoni (1986), Woods et al. (1987), and Kang
et al. (1987). However, in many but not all of these studies, definite evidence
of exposure to 2,3,7,8-TCDD is not established. Furthermore, the risk estimates
are based upon generally small numbers as one might expect.
In a case-control study based on data provided by the National Cancer
Register of England and Wales, Balarajan and Acheson (1984) found a significant
risk of STS among farmers, farm managers, and market gardeners, but not among
agricultural workers, gardeners, and groundsmen. The authors hypothesize exposure
to 2,3,7,8-TCDD-containing herbicides in high risk occupational groups as a
possible cause, although actual evidence of exposure is not provided.
Cantor (1982), in a case-control study, noted a significant risk of reti-
culum cell sarcoma among farmers under age 65 in Wisconsin in counties that
were high in suronary measures of General Agricultural Activity based on county-
wide measures of farm-related exposures. However, occupations had been clas-
sified based on what was recorded on the death certificates. Often, the infor-
mation is absent, however, or if an occupation is mentioned, it is frequently
the last job held or the job of longest duration or simply the .word "retired"
14
-------
will appear. Such information must be considered unreliable at best and could
lead to misclassification. Furthermore, the use of 2,3,7,8-TCDD-containing
chemicals is only suggested by the authors.
Mil ham (1982), in a proportionate mortality rate (PMR) study, noted a
significantly high risk of STS among farmers in Washington State. The author
states that exposure would most probably have been to phenoxyacetic acid herbi-
cides (chiefly 2,4-D) and chlorophenols. There are two major problems with
this study: (1) no actual evidence of exposure is provided, and (2) the source
of death information is not given. Additionally, PVR studies are inherently
inferior to cohort and case-control designs.
Kogan and Clapp (1985), in a second PMR study, found a significantly high
rate of connective tissue cancer (STS) among Vietnam veterans in Massachusetts
as opposed to non-Vietnam veterans in Massachusetts. Again, service in Vietnam
appears to be the surrogate of exposure and as such may bear no relationship to
actual exposure to 2,3,7,8-TCDD-containing chemicals. It is entirely possible
that many Vietnam veterans in Massachusetts were never exposed to large quanti-
ties of 2,3,7,8-TCDD-containing chemicals, while there is no reason to think that
some non-Vietnam Massachusetts veterans may have been exposed to 2,3,7,8-TCDD-
containing chemicals in other settings such as occupational.
The Michigan Department of Public Health (1983a, b) found significantly
high death rates among females of Midland County, Michigan, compared to the
national average for the period 1960 through 1978. In contrast, STS death
rates for males of Midland County were not elevated during the same period.
The Dow Chemical Company produced phenoxyacetic acid herbicides and/or chloro-
phenols in this county. Since this is an ecological study, it will of necessity
include in both the numerator and denominator individuals who may or may not
have received exposure to 2,3,7,8-TCDD or polychlorinated dibenzo-p_-dioxin-
15
-------
contaminated chemicals. Cook and Cartmill (1984), in a discussion of the soft
tissue sarcoma issue, maintained that the Michigan Department of Public Health
found no commonalities among the women that would suggest any single agent
including exposure to 2,3,7,8-TCDD-contaminated chemicals as the cause except
that several of the women moved to Midland County with pre-existing diseases.
Whether the diseases they brought with them had anything to do with STS was not
established. The authors also suggested that several women may have been
incorrectly classified to ICD category 171x since the site of the cancer was
not specifically mentioned. However, there is no reason to suspect that the
Michigan Department of Public Health coded their death certificates any dif-
ferently than did the National Center for Health Statistics, the U.S. agency
responsible for generating U.S. and state vital statistics. These death rates
provide the basis for calculating the expected deaths in that study.
A study of Vietnam veterans in Australia (Fett et al., 1984) essentially
reported nonpositive findings, but the veterans also exhibited an excess of STS,
albeit small (i.e., 2 observed vs. 0.64 expected). In the comparison group of
non-Vietnam veterans no STS's were observed versus 0.84 expected. This study
again used "service in Vietnam" as a surrogate for exposure to Agent Orange.
No actual proof of exposure is provided, and one would expect many if not most
of the Vietnam veterans to have had little exposure to Agent Orange.
A recent study (Puntoni et al., 1986; Merlo and Puntoni, 1986) reports a
high incidence of STS in residents of the region around Seveso, Italy, where an
industrial explosion at a factory that produced 2,3,7,8-TCDD-contaminated
2,4,5-T occurred on July 10, 1976. Few specifics are available regarding this
study, but it is apparent that the rate of STS's was higher in residents of the
Seveso region as contrasted with those in the Varese region (2 per 100,000) and
tended to increase with time (i.e., 4.35 in 1981) when so-called "polluted"
16
-------
areas are lumped with "non-polluted" areas. The authors suggested that these
two areas be combined since chloracne rates calculated for each area separately
were inconsistent with the defined soil levels of 2S3,7,8-TCDD in each area,
the surrogate measure used for defining a "polluted" area as contrasted with a
"non-polluted" area. Misclassification of exposure was suggested by the authors
as the explanation for the lack of a positive correlation with the incidence of
STS in so-called "polluted" areas. Hardell and Eriksson (1986) suggested that
exposure to the chemicals produced by this factory prior to the accident
accounted for the high rates befores during, and after the accident in the
Seveso region since latent factors argue against the accident being the cause
of the excess.
The remaining nonpositive studies are even less remarkable than those
already mentioned. Thiess et al. (1982) examined a cohort of 74 employees of
the Badische Anil in und Soda Fabrik (BASF) who were accidentally exposed to
trichlorophenol in 1953, 66 of whom suffered chloracne; after a lengthy follow-
up, no STS's appeared. However, the study has little power to detect a sig-
nificant risk of cancer.
Axel son et al. (1980) studied 348 railroad workers who sprayed 2,4-D and
2,4,5-T herbicides from 1957 to 1972 and found no STS. Again, this study has
limited power to detect a risk of STS as well as little evidence to substantiate
exposure.
Riihimaki et al. (1982) studied 1,926 Finnish applicators of 2,4-D and
2,4,5-T from 1955 to 1971 and found no STS's (0 observed vs. 0.1 expected).
However, the authors suggested major difficulties with their paper as follows
". . . an observation period under 10 years, the presence of selection bias,
and low or little exposure." They conclude that it "cannot be regarded as a
negative study." This study also suffers from survivorship bias.
17
-------
Wiklund and Holm (1986) recently studied a massive cohort of 354,620
Swedish men who were recorded as having an agriculture or forestry job according
to the census of 1960 versus 1,725,845 Swedish men in all other industries.
The primary exposure in those jobs was postulated to be MCPA; 2,4-D and 2,4,5-T
were also used to a lesser extent. The authors found that the relative risk of
STS was only 0.9. This study has several problems: (1) a lack of individualized
exposure data; (2) only 15% of Swedish agricultural and forestry workers were
estimated to be exposed to phenoxyacetic acids and 2% to chlorophenols; (3)
Swedish agricultural workers are known to have a decreased cancer risk and use
health services less frequently; (4) classifying workers according to a 1-year
employment status presents the possibility of misclassification; and (5) the
crude rate of STS in agricultural and forestry workers based on data in the
study is 5.45 per 100,000 person-years, and in the remaining workers it is 5.00
per 100,000 person-years. Both rates are high compared to rates from other
nations (1 to 3 per 100,000 person-years).
The authors of the previous study, Wiklund et al. (1987), recently evaluated
the relative risks of Hodgkin's disease and non-Hodgkin's lymphoma in a cohort
of 20,245 Swedish pesticide applicators who were licensed anytime between 1965
and 1976. The authors estimate that 72% had had an opportunity for contact
with phenoxyacetic acid herbicides (chiefly MCPA, mecoprop, and dichloprop, and
to a lesser extent, 2,4,5-T and 2,4-D). This estimate was based on responses
to a questionnaire that had been mailed to a random sample of 273 persons in
the cohort. No actual measurements of exposure were taken. Overall, based on
a rather short follow-up averaging 12.2 years, no significant excess risk of
either disease was found. However, the authors noted a trend of increasing
risks for both diseases with lapsed time since licensing. Further follow-up of
this cohort is needed since presumably the latent period for lymphoma is in
18
-------
excess of the observation period of this study.
Another nonpositive study (Lathrop et al., 1984a, b; Wolfe et a!., 1984,
1985) of a group of young Air Force officers and skilled and unskilled enlisted
men (called Ranch Handers) who presumably were involved with the aerial spraying
of herbicides containing 2,3,7,8-TCDD revealed no excessive cancer risk. How-
ever, only 6 of 1,256 Ranch Handers actually developed cancer in the short
period of time they were followed. Although this study offers little evidence
to substantiate exposure to Agent Orange by itself, recently Pirkle et al.
(1987) reported that some 75 Ranch Handers (6% of the total) had reported a
range of exposure to 2,3,7,8-TCDD (from 16 to 423 ppt) based on adipose tissue
samples collected by the Centers for Disease Control (CDC) and the Air Force.
Furthermore, the study suffers from a lack of power and short latency.
Another nonpositive study by Greenwald et al. (1984) of 281 STS cases from
the New York State Cancer Registry matched with 281 referents taken from New
York driver's license files and 130 from New York death certificate files
exhibited only a nonsignificant excess cancer risk for workers in chemical
manufacturing and highway construction. Based on questionnaire responses, the
risk from exposure to 2,3,7,8-TCDD-contaminated Agent Orange was only 0.7 while
that of herbicide and/or pesticide use was 1.0. The risk for farming was 0.79.
Two major problems are evident in this study. First, "service in Vietnam" was
used as a surrogate for exposure to Agent Orange. This potentially can lead
to a problem similar to that of other studies where exposure to Agent Orange
could not be substantiated, and secondly, those who did not serve in Vietnam
could have received exposure to 2,3,7,8-TCDD through other means. Furthermore,
the average time lapse from Vietnam service to a diagnosis of STS in this study
was only 11 years, a short latent period. Many additional years of observation
must be added before the latency question can be put to rest.
19 /
-------
Kang et al. (1986) group-matched 234 Vietnam-era veterans who were STS
patients and had served in the U.S. military between 1964 and 1975 with 13,496
patients sytematically sampled from the same Vietnam-era patient population
without a diagnosis of STS. Service in Vietnam (a surrogate for exposure) did
not appear to be associated with STS. However, the authors reported that more
than likely most patients (and controls) had not achieved a follow-up time
sufficient for latent effects to become manifest. Furthermore, actual exposure
as evidenced by in vivo measurements of 2,3,7,8-TCDD were not available or, as
the authors pointed out, Vietnam veterans may have been exposed to such small
amounts that standard epidemiologic designs could not detect an excess risk.
In a second case-control study, Kang et al. (1987) selected 217 STS
patients from the Armed Forces Institute of Pathology and matched them to 599
matched controls for service in Vietnam, exposure to chemicals, medical history,
and life style. Military service was verified by the patients' military per-
sonnel records. Vietnam veterans, in general, did not exhibit increased risks
of STS. However, the authors found that the risk of STS increased in veterans
with combat experience versus those without (OR = 2.57, nonsignificant) and
increased even more for those same combat veterans who were assigned to the
region where most spraying of Agent Orange took place (OR = 8.64, nonsignificant).
The authors noted that their study had "very low power" to detect enhanced
risks in subgroups of veterans who had greater opportunities for exposure to
Agent Orange. The authors concluded that a possiblity exists that certain
subgroups of Vietnam veterans may be subject to "modestly increased risks of
STS" from exposure to Agent Orange, but it can neither be confirmed nor ruled
out.
Karon et al. (1987, unpublished), at a recent meeting on dioxin, reported
only one army veteran with a current serum 2,3,7,8-TCDD level in excess of
20 -~
-------
20 ppt out of some 775 veterans who participated in the study. About 675 of
these served as ground combat troops in Vietnam and included some of the
troops who presumably had had a high likelihood of exposure to Agent Orange.
These levels today are not different from those found in background populations
of industrialized nations. At first glance this seems to argue that veterans
in Vietnam may not have been heavily exposed to Agent Orange, and hence findings
based on the surrogate experience in Vietnam may not be relevant. Such may
not be the case since actual time served in Vietnam occurred a considerable num-
ber of years ago. Elevated serum 2,3,7,8-TCDD levels brought on by massive
exposure to Agent Orange at that time may have dissipated and returned to
background levels during the period following removal of troops from Vietnam.
The Hoar et al. (1986) study is not contradictory to the hypothesis that
2,3,7,8-TCDD is the contaminant responsible for the development of STS. The
authors found that 2,4-D, which does not contain 2,3,7,8-TCDD, was identified
by most members of the cohort as being the herbicide to which they were
exposed. Again, the results of this study are based on questionnaire data and
may suffer from recall bias. This study supports only the hypothesis that the
herbicide 2,4-D appears not to be a credible candidate for the carcinogenic
agent responsible for STS.
NON-HODGKIN'S LYMPHOMA
Hardell et al. (1981), in a case-control study of lymphomas in Swedish
workers, found a statistically significant risk of lymphoma in agricultural,
forestry, and woodworking employees from exposure to phenoxyacetic acid herbi-
cides and/or chlorophenols. Risk ratios ranged from 4.3 to 6.0 for both classes
of compounds together as well as separately. However, exposure to organic
21
-------
solvents such as benzene, trichloroethylene, and styrene was also found to be a
risk factor. Unfortunately, the data were based on answers derived from adminis-
tered questionnaires, and the exact extent of exposure to 2,3,7,8-TCDD is
difficult to determine. As such, this study is subject to some recall bias and
confounding, but the risk ratios are sufficiently high that problems with the
study probably could not account for them entirely. This is another one of the
Swedish studies that had been severely criticized for alleged biases, confound-
ing, and distortions. The discussion following the Harden and Sandstrom
(1979) and Eriksson et al. (1979, 1981) studies regarding those criticisms
pertains to this study as well.
Hoar et al. (1986), in a similar study, found significantly high rates of
non-Hodgkin's lymphoma (NHL) in farmers in Kansas who use herbicides, particu-
larly 2,4-D and triazines. Few respondents could remember exposure to 2,4,5-T
Which contains 2,3,7,8-TCDD. 2,4-D is not believed to contain 2,3,7,8-TCDD but
does contain other polychlorinated dibenzo-p_-dioxin impurities. The risk was
found to increase with increasing frequency and duration of herbicide usage.
Although "herbicide usage" could mean any of the herbicides identified by Hoar,
she wrote that this is "essentially synonymous" with use of 2,4-D. The next
most used herbicides (i.e., triazines and uracils) are nonsignificant when
exposures to phenoxyacetic acids are controlled for. However, this study has
problems similar to the Harden et al. (1981) study in that there is a lack
of substantiation of exposure, and the information is based on questionnaire
responses that are subject to some recall bias. Moreover, there is a statisti-
cally significant risk associated with the use of other herbicides as well,
i.e., triazines, amides, and trifluralin.
A population-based case-control study of the relationship of occupational
exposure to the risk of STS and NHL was recently completed by Woods et al.
22
-------
(1987). A total of 128 STS cases and 576 NHL cases diagnosed between 1981 and
1984 were matched with 694 randomly selected controls without cancer. The
authors reported no overall increased risk of either disease from past
occupational exposure to phenoxy herbicides or chlorophenols based on personal
interviews. However, significant elevated risks of NHL were observed in certain
subgroups thought to have somewhat heavier exposure to herbicides, i.e., farmers
(OR = 1.33, CI 1.03-1.7), forestry herbicide applicators (OR = 4.8, CI 1.2-19.4),
and persons potentially exposed to phenoxy herbicides for 15 or more years
during the period 15 years prior to cancer diagnosis (OR = 1.71, CI 1.06-1.7).
Of interest in this study is the finding that presence of chloracne is associated
with a high (borderline significance, p < 0.075) risk of STS and less so to a
risk of NHL. In so far as the information on exposure, based on questionnaire
data, is somewhat suspect because actual confirmation was not obtained (i.e.,
adipose tissue specimens with confirmed presence of 2,3,7,8-TCDD), the presence
of chloracne in conjunction with known contact with chemicals containing 2,3,7,8-
TCDD could have meant a massive exposure to 2,3,7,8-TCDD above and beyond
background levels. The chloracne was not clinically confirmed. Because the
use of questionnaires to elicit past-history of exposure to specific phenoxy
herbicides and/or chlorophenols has a potential for misclassification among
cases and controls (because of memory problems and lack of confirmation of
exposure), this study seems to have reported paradoxically conflicting results.
Buesching and Wollstadt (1984), in a case-control study, found that white
male farmers of Winnebago County, Illinois, had a statistically significant
risk of 2.65 for NHL as contrasted with all white males of that county. Although
the author presents no evidence of exposure to any particular herbicide or
chlorophenol, he suggested that phenoxyacetic acid herbicides were probably
used. Of major concern in this study is the use of occupation as coded on the
/•"
23
-------
death certificates to classify the deceased by exposure. As mentioned pre-
viously, use of such information may lead to misclassification.
Cantor (1982), in a case-control study of NHL, found a significantly high
risk of NHL among farmers under age 65 in counties of Wisconsin known to be
high in insecticide and herbicide use, and having small grain acreage, wheat
acreage, and dairy sales. The risk of reticulurn cell sarcoma is higher at 3.2
among younger farmers in these same counties. For small grain acreage and
acres treated with insecticides, the risk is 6.6, and for wheat acreage, it is
4.4; all are significant. Unfortunately, information on exposure was derived
based on occupations listed on death certificates similar to that of Buesching
and Wollstadt (1984) and as such cannot be considered reliable. The potential
for misclassification of exposure among the cases and controls is great.
Additionally, no particular herbicide, insecticide, or pesticide was mentioned.
They were only surmised to be in use by the author.
Burmeister et al. (1983), in another case-control study of NHL in Iowa
farmers, found a statistically significantly high risk of NHL associated with
farming. It was found to be elevated in those born before 1901 and was associ-
ated with jobs involving egg-laying chickens, solid milk products, hog produc-
tion, and herbicide use. This study, which also depended on occupation as
recorded on death certificates for exposure, may be very unreliable, and it is
likely that misclassification could have resulted. The authors suggested that
herbicides and/or insecticides, such as those mentioned in the Hardell and
Sandstrom (1979) study, may be carcinogenic.
Milham (1982), in a proportionate mortality ratio study of deceased workers
in the agriculture, forestry, and wood products industries of Washington State,
noted a significant excess of Hodgkin's disease and a high but nonsignificant
risk of NHL in paper and pulp mill workers. Again, there is no evidence of
24
-------
actual exposure to phenoxyacetic acid herbicides and/or chlorophenols cited by
the authors. Furthermore, the proportionate mortality design is inherently
inferior to the cohort or case-control study design because the former forces
interdependence of the risk estimates.
Cook et al. (1986), in a cohort study of 2,189 manufacturing employees of
a chemical plant that produced TCP, 2,4,5-T, and other chlorinated phenols
found a nonsignificant increased risk of NHL (5 observed deaths versus 2.1
exposed). However, the authors reported that these results do not support a
causal association between chronic disease and exposure to the chemicals in
question. The problems with this study involve lack of evidence of exposure
to most members of the cohort (only 15% had chloracne), and no effort was made
to assess latent effects.
Ott et al. (1987) updated the Cook et al. (1986) study by using a
statistical technique known as a "serially additive expected dose model" that
assesses latent effects. Based on six lymphomas, no latent trend was found.
The problems with this design were discussed previously.
Other studies, such as those of Lynge (1985), lack and Suskind (1980),
Thiess et al. (1982), Axel son et al. (1980), Riihimaki et al. (1982), Lathrop
et al. (1984a, b), Wolfe et al. (1984), Fett et al. (1984), and Kogan and Clapp
(1985), did not find an excess risk of NHL in individuals who were potentially
exposed to 2,3,7,8-TCDD-contaminated chemicals. All of these studies have
methodological flaws that make it unlikely that they could detect a significant
risk if one were present. The Lynge (1985) study noted only a slight excess of
lymphomas in males after a lapse of 10 years from initial exposure; however,
sensitivity was somewhat reduced. The Zack and Suskind (1980), Thiess et al.
(1982), and Axelson et al. (1980) studies were exceptionally small studies and
hence lacked sensitivity because of their size despite a lengthy follow-up.
25 /,''. •
-------
The Riihimaki et al. (1982) study, although seemingly large, had many methodo-
logical deficiencies as described in the section under soft tissue sarcomas
that precluded detection of a risk of NHL. The study by Lathrop et al. (1984a,
b) and Wolfe et al. (1984) of Ranch Handers also lacked sufficient statistical
power to detect a risk as being significant. Furthermore, there was a lack of
evidence of substantiation of differential exposure to Agent Orange in this
study, a short latent period in which to expect the development of NHL, and a
lack of sensitivity for the detection of NHL. The Australian Veterans Health
Studies (Fett et al., 1984) are much like the Ranch Handers studies in that they
lack sensitivity, provide little evidence of differential exposure, and have a
short latent period. Kogan and Clapp (1985) did not evaluate NHL in their
study of Massachusetts veterans. This study also had several limitations which
were discussed in the section on soft tissue sarcomas.
STOMACH CANCER
Two small cohort studies by Thiess et al. (1982) and Axel son et al. (1980)
reported statistically significant excesses of stomach cancer (albeit three
cases in each study), while a third smaller study (121 persons) by Zack and
Suskind (1980) reported no excess risk of stomach cancer. This exceptionally
small cohort lacks sufficient power. Only 0.5 stomach cancer deaths would have
been expected to have occurred by the end of the cut-off date in that study.
Lynge (1985) reported a nonsignificant excess risk of stomach cancer in chemical
workers in Denmark. Cook et al. (1986) noted a slightly elevated risk of stomach
cancer in his cohort which was also nonsignificant. Cook et al., however, did
not analyze their data for latent effects. The Riihimaki et al. (1983) cohort
study is another nonpositive study with large deficits of mortality. This
26
-------
study suffers from major problems not the least of which is a survivorship
phenomenon. The members of this cohort had to live long enough to be included
in the cohort. In the Wolfe et al. (1984) and Lathrop et al. (1984a, b) study
of Ranch Handers, no excesses of any cancer of any kind were found. Again,
this mortality study of mostly young men produced only six cancer deaths to
date, and has several problems as mentioned previously. Burmeister et al.
(1983), in a case-control study, noted a significant risk of stomach cancer in
individuals who had jobs in "milk products, cattle production, and/or jobs
involving high yields of corn per acre." The authors suggested as a cause
possible exposure to herbicides and/or insecticides such as those reported in
the Hardell and Sandstrom (1979) study. However, occupation as recorded on
death certificates was the surrogate used by the authors for exposure, and as
such is a very unreliable vehicle for the determination of exposure.
ALL OTHER CANCERS
Other cancer sites have been found to be significantly elevated among
persons or groups potentially exposed to 2,3,7,8-TCDD-contaminated chemicals.
Buesching and Wollstadt (1984) and Burmeister et al. (1983) reported significant
elevated risks of prostate cancer in farmers in Winnebago County, Illinois, and
agricultural workers in Iowa, respectively. Both used death certificate desig-
nation of occupation as their surrogate for exposure. As was pointed out
earlier, occupation as recorded on death certificates is very unreliable. No
substantiation of exposure to any herbicides or pesticides was done.
Milham (1982) found a significant excess of Hodgkin's disease in persons
employed in the agriculture, forestry, and wood products industry of Washington
State. His source of data was not revealed. However, there is again no evidence
27
-------
actual exposure to 2,3,7,8-TCDD-containing phenoxyacetic acid herbicides and/or
chlorophenols in this proportionate mortality study.
Kogan and Clapp (1985), in his proportionate mortality study of Vietnam
veterans versus non-Vietnam veterans in Massachusetts, found a significant
excess of kidney cancer. Again, no evidence of actual exposure to Agent Orange
was given. This study and its methodological flaws were discussed previously.
In short, service in Vietnam is the surrogate for exposure.
Finally, Cook et al. (1986) found a statistically significant excess of
other and unspecified cancers in his study of employees of a chemical plant
that produced TCP, 2,4,5-T, and other chlorinated phenols. Again, the flaws
in this design were discussed previously.
. SUMMARY AND CONCLUSIONS
The evidence that persons who are exposed to phenoxyacetic acid herbicides
and/or chlorophenols are subject to an elevated risk of STS's is based primarily
on two independent case-control studies by Harden and Sandstrom (1979) and
Eriksson et al. (1979, 1981). The problems with these studies, as outlined in
the section on STS, are not sufficient to explain the highly significant risks
of STS found in workers exposed to these chemicals. Power considerations alone
indicate that as many as 1/3 of the positive responses in the cases could be
reversed without a major loss of power in both studies. Additionally, Eriksson's
data further suggest that the risk is greater from exposure to 2,3,7,8-TCDD-
contaminated herbicides such as 2,4,5-T rather than from those herbicides free
of 2,3,7,8-TCDD contamination. Several additional studies tended to support or
were consistent with the Hardell and Sandstrom (1979) and Eriksson et al.
(1979, 1981) findings of an excess risk of STS in certain subgroups of the study
28 "-
-------
groups thought to be more highly exposed to 2,3,7,8-TCDD-contaminated chemicals
(Zack and Suskind, 1980; Lynge, 1985; Balarajan and Acheson, 1984; Cantor,
1982; Mil ham, 1982; Kogan and Clapp, 1983; the Michigan Department of Public
Health, 1983a, b; Puntoni et al., 1986; Merlo and Puntoni, 1986; Wood et al.,
1987; and Kang et al., 1987). However, except for the Zack and Suskind study,
where proof of substantial exposure to 2,3,7,8-TCDD-contaminated trichlorophenol
is evident by the occurrence of chloracne in all subjects, differential exposure
to 2,3,7,8-TCDD, sufficient to produce a detectable change in risk estimates,
can only be assumed to have occurred where proof of exposure is available. In
addition, authors of several studies have maintained that their own results do
not support the results of the Swedish studies. However, many of these studies
exhibit small nonsignificant risks of STS when it could be measured. For
example, the Cook et al. (1986) study produced one STS where none were expected,
while the Smith et al. (1984) study produced a significant risk of STS in
workers involved in treating pelts with chlorophenols. Even the Fett et al.
(1984) study of Australian Vietnam veterans produced two STS's where only 0.67
were expected. On the other hand, in studies where no risk of STS was found,
such as Thiess et al. (1982) and Axelson et al. (1980), the cohorts are exception-
ally small, and one would not expect to see any of the exceedingly rare STS's
even though many years have passed. Even the defective study of Riihimaki et
al. Il982) anticipated only 0.1 STS. Wolfe and Lathrop's Ranch Handers exhibited
little cancer mortality, and never achieved a latent period long enough to
detect any STS's. Wiklund and Holm (1986), on the other hand, in their massive
study of STS in Sweden, although nonpositive, provided enough data to calculate
crude STS rates that appear unusually high compared to that of other countries.
Woods et al. (1987) found a borderline significant risk of chloracne although
clinically unconfirmed with risk of STS. Kang et al. (1987) found a high risk
29
-------
of STS in Vietnam veterans most likely to have been exposed to Agent Orange,
although power was very low. Greenwald et al. (1984) used "service in Vietnam"
as a surrogate for exposure. Only five of his nine "exposed" Vietnam service
cases actually thought they were exposed to Agent Orange, while four "non-Vietnam
service" controls claimed exposure to phenoxyacetic acid herbicides in occupational
settings. The Hoar et al. (1986) study seems to imply only that 2S4-D and
triazines could be ruled out as a cause of STS, since 2,3,7,8-TCDD-contaminated
2,4,5-T was believed not be present. Hence, this study does not really contradict
the Swedish studies because 2,4,5-T was not considered. Most of these studies
suffer from one or more methodological flaws that preclude any conclusion that
they are nonsupportive of a finding of a risk of STS.
The net result is that basically the Swedish studies of Hardell and
Sandstrom and Eriksson et al. provide the best evidence of a causal relationship
from exposure to the phenoxyacetic acid herbicides and/or chlorophenols. There
is some additional support from some small cohort studies and a few poorly and
not so poorly designed case-control and proportionate mortality studies. However,
these two Swedish case-control studies, contrary to the criticism leveled at
them, appear to have been constructed and executed along classic lines as
outlined in any good epidemiology course on methodology. The authors appear to
have been concerned about possible bias, confounders, and other problems in
their analyses as well as their critics. Harden and Sandstrom re-examined their
data by using colon cancer cases and matching them with the same controls in
order to assess the effect of certain biases with which they were concerned.
Under any other circumstances these studies would be considered reasonably good
studies minus certain analyses relating to dose response that probably were not
envisioned at the time by the authors. Largely on the basis of these Swedish
studies, the International Agency for Research on Cancer decided to reclassify
30
-------
the phenoxyacetic acid herbicides and/or chlorophenols in 1986 as probable
human carcinogens based on 1 invited evidence of their carcinogem'city in humans.
Although the epidemiologic data appear to provide limited evidence that
exposure to phenoxyacetic acid herbicides and/or chlorophenols is causally
related to the risks of STS, none of these studies could be said to incriminate
2,3,7,8-TCDD directly as the agent solely responsible for the excess of STS's.
The question remains that the risk of STS may be in part due to the presence of
other polychlorinated dibenzo-p_-dioxin contaminants as well or even to the
phenoxyacetic acid herbicides and/or chlorophenols themselves, or perhaps some
unknown confounder not heretofore discovered. Hence, based on the epidemiologic
data, the evidence still must be considered inadequate but suggestive that
2,3,7,8-TCDD is carcinogenic to humans.
With respect to non-Hodgkin's lymphoma (NHL), a similar situation exists.
Data from Hardell et al. (1981) and more recently Hoar et al, (1986) support
the findings of an excess risk of NHL from exposure to phenoxyacetic acid
herbicides and/or chlorophenols. However, this risk is confounded by the
presence of organic compounds also shown to be carcinogenic on their own in the
Hardell et al. study. In the Hoar et al. study, although 2,3,7,8-TCDD-contam-
inated 2,4,5-T was not identified as being present, other herbicides and insec-
ticides as well as 2,4-D were present and were significantly associated with
exposure. The contaminant 2,3,7,8-TCDD appears not to be present in any of
these compounds in the Hoar et al. study although other isomers of polychlori-
nated dibenzo-jD-dioxin were present. . Wiklund et al. (1987) noted a trend of
increasing risks for Hodgkin's disease and NHL with lapsed time since licensing
of herbicide applicators although nonsignificant. Woods et al. (1987) reported
significantly increased risks of NHL in certain occupational subgroups known to
have a high likelihood of contact with herbicides. Buesching and Wollstadt
31 S'
-------
(1984) found a significant excess of NHL in deceased farmers in Winnebago
County, Illinois, but this study used occupations as listed on the death certifi-
cates as the basis for determining exposure. No particular herbicide or other
chemical was identified however. Cantor (1982) noted a similar significant
excess of NHL (reticulum cell sarcoma) in young Wisconsin farmers in counties
with a high summary measure of agricultural activity, such as number of acres
in small grain acreage, number of acres treated with insecticides, and number
of acres in wheat acreage. Again, no particular herbicide, pesticide, or
chlorophenol was identified. Burmeister et al. (1983) also noted a significant
excess of NHL in Iowa farmers who were involved with "egg-laying chickens,
solid milk production, hog products, and herbicide use." Again, no particular
herbicide was identified. Similarly, Milham (1982) noted, in a proportionate
mortality study, a higher nonsignificant risk of NHL in Washington State paper
and pulp mill workers. Cantor (1982) stated that 2,4-D and/or chlorophenols
were chiefly used in these industries.
Cook et al. (1986), in his cohort mortality study, also reported a non-
significant excess risk of NHL in chemical workers exposed to trichlorophenols,
2,4,5-T, and 2,4,5-T formulations.
Other studies that report slight excesses of NHL or none at all include:
Zack and Suskind, Thiess et al., Lynge, Axelson et al., Riihimaki et al., Fett
et al., Wolfe et al., and Kogan and Clapp. However, these studies cannot be
said to support the argument that there is no risk because of problems connected
with each of them, i.e., exceptionally small cohorts, insufficient latency,
insensitivity, and, in some instances, little evidence of exposure. These
problems were discussed previously.
The net overall thrust of the data supports the judgment that there is
limited human epidemiologic evidence that the phenoxyacetic acid herbicides
32
-------
and/or chlorophenols are causally related to NHL. However, with respect to
identifying any particular isomer, such as 2,3,7,8-TCDD, it must be regarded as
inadequate.
With respect to stomaqh cancer, Thiess et al. (1982) and Axel son et al.
(1980), in small cohort studies, found significantly high stomach cancer risks
based on three cases in each study. Zack and Suskind (1980), on the other hand,
did not find any in another small study. Only one other study produced a
significant excess of stomach cancer, and that is the case-control study by
Burmeister et al. (1983) in which the surrogate for exposure was occupation as
given on the death certificates. No other study reported a signficant excess
of stomach cancer, although most studies have methodological flaws. At this
time, the data must be considered only suggestive at best that phenoxyacetic
acid herbicides and/or chlorophenols cause stomach cancer in humans based on
the epidemiologic data. Again, with respect to the potential carcinogenicity
of 2,3,7,8-TCDD, the data must be regarded as inadequate.
Other excess cancer risks by site occur sporatically at best in a few of
these epidemiologic studies. They must be regarded as inconclusive. A summary
of the preceding text will be found in tabular form in Tables 2 through 8
following the section on Future Expectations.
ONGOING RESEARCH
Several agencies of the U.S. government are conducting ongoing research
that may help to resolve some of the issues raised regarding interpretation of
the human health studies on 2,3,7,8-TCDD. NIOSH is conducting both a morbidity
and mortality study of workers potentially exposed to TCP, 2,4,5-T, and penta-
chlorophenol, and therefore 2,3,7,8-TCDD. Using NIOSH-registry data, which
33
-------
includes demographic and work history information on some 7,000 employees of 14
different production plants in the United States where potential exposure to
2,3,7,8-TCDD may have occurred, NIOSH is conducting a historic prospective
mortality study. The results will be reported in the near future.
The second major use of the registry data will be to conduct a hands-on
medical study of some 450 workers at two of the 14 plants. Some 450 referents
were age, sex, race, and neighborhood-matched based on Census block information.
Noninvasive medical tests (including blood specimens) will be conducted in
order to evaluate presence or absence of health conditions hypothesized to be
associated with exposure to TCP, 2,4,5-T, and pentachlorophenol and consequen-
tially to 2,3,7,8-TCDD. Only the first phase of this study is complete. That
phase included complete evaluations of approximately 80 cases and their refer-
ents. Two papers are in preparation at this time; they include an assessment
of chloracne in pentachlorophenol workers as well as a description of medical
results.
The IARC and National Institute of Environmental Health Sciences are jointly
maintaining an international registry of persons occupationally exposed to
phenoxy acids, chlorophenols, and contaminants during their manufacture and
use. This study will pool together cohorts from over 12 countries. The same
protocol will be used, and the data will be analyzed as one big study. The
NIOSH registry is included as part of this study.
The CDC is further conducting pilot exposure studies on former Vietnam
veterans in conjunction with the Veterans Administration (VA). Currently,
blood is being drawn from some 600 Vietnam and non-Vietnam veterans and analyzed
for the levels of 2,3,7,8-TCDD. This information will be used to compare and
validate categories of exposure that have been defined in advance regarding the
degree of exposure to 2,3,7,8-TCDD sustained by these veterans.
34
-------
Additionally, CDC and the VA are .corraborating on a study of about 40
"heavily" exposed Ranch Hand veterans to determine the half-life of 2,3,7,8-TCDD
in blood. Apparently, new blood will be drawn and the levels of 2,3,7,8-TCDD
will be determined in that new blood and contrasted with the levels of 2,3,7,8-
TCDD found in old blood drawn from the same persons at an earlier time. This
project was expected to be complete in September of 1987.
CDC is also planning a series of case-control studies of liver cancer,
nasal cancer, STS, and lymphomas in Vietnam veterans to be completed in 1989.
The National Cancer Institute (NCI) is planning a case-control study of
Nebraska farmers exposed to phenoxyacetic acid herbicides similar to the Hoar
et al. (1986) study of Kansas farmers. Results will be available in 1989.
FUTURE EXPECTATIONS
The NIOSH studies should provide answers to whether U.S. workers involved
in the production of herbicides and chlorophenols contaminated with 2,3,7,8-
TCDD suffer from increased risks of site-specific cancer mortality as well as
other long-term health effects a priori found to be associated with exposure to
2,3,7,8-TCDD. The CDC studies will help to substantiate or invalidate subjec-
tively defined 2,3,7,8-TCDD exposure gradients in Vietnam veterans as well as
provide estimates on the half-life of 2,3,7,8-TCDD in blood. This information
will be useful in estimating quantitatively the probable dose received by
Vietnam veterans at the time of exposure to Agent Orange. A series of case-
control studies planned by CDC on Vietnam veterans of cancer sites shown
previously to be associated with a higher risk of exposure to Agent Orange may
provide additional evidence to either refute or substantiate a higher risk of
cancer, especially if it can be determined what the likely dose was at the time
35 i . /
-------
of exposure based on the half-life estimates. The NCI case-control study of
Nebraska farmers should serve as a check on the results of the earlier Hoar et
al. (1986) study of Kansas farmers.
36
-------
CO
c
(_
0)
u
c
o
o
o
¥
OJ
CZ
u
ro
t-
o
t-
o
r^
1 &•»
00 O)
« 0
1^. CZ
« re
CO O
»
CM
1
to
UJ
o
00 X
UJ
0
o -a
CM 0)
-Si£
UJ C_
1 §
J— co
ro
^
Exposure
co
*^^
O
1
/
+J t/> •
Z3 •" *-O
u "o cR
5 ° =
c_ O -g
a> , 3 =
•fj Qj | * rjj
4- +J CO S
ra "3
-o t- CTI
13 O) OO
OJ t- 4J i-l
•i- >»- (O «
Q O i— CM
CO
>~
CO
. 3
O
jQ ro
iH §
^•^
C 0
ro O
C '!-
•? CO
I— *t~
s:
c-
CM
OO
o^ ^^^
O re
>1
Lf) O*l
O^ i— 1
.„
^™*" 1
Trichlorophenol (TCP
operator, plant
mechanic (TCP depart
ment) .
-^^*
1-4
oo
Ji CTl
O 1-H
8-
•o o o
o -P iS
•o *> i • M-
-o a> = £= t. a) S co o <1> X :. co c
°- = >> O 4J O O
XOfZ o CLOre-i— O. O
C 0) 1
&5lf*S ." S°°.sS2 .t:°°.=
oo a, i °- ro t! = ^ % g °- = ^ aj
•r->« p-« o-«
CO
3 *»<
O u P—
-t-> ro ra . ct. &CO
= 2 " J; ,
Q)O rOCO^J 3*r— COO rO_C S
pp C0p0 p4J rO«- £-0.0
t- *- OC-CO t- CO -r- J3 3re*r-
3 rO Q. .O Qj JU *r- "rj *r— QJ f QJ
sU"* Ij'lZ6 iZ"*1 • S "^ •— T3-—
•" A- U_
O O O CM o» ^
00 00 00 r>. |i. 00
en 222*" ^
i- 1 T3
4J^}-r_r— <4- vo -r— i-«|«> CTl
_. >- O^»-Q'Oa)C-oCr>Srjc IO i— 1
oro •• oo o"1 ~-- > aJ^oof_a.+j(y)f^t_coK -o
co -c*-co^OCCTlXCO ro+JO<— !OJ>>.4->''-rO'-'OE (UCre -C
•-i^-* u. rTi u. 10 (—
T3
•a OO) «|+J -a
*- »-• id o co QJ>S r— i-< 4J 0)«-U. .C-r-O.i-0 CSS
aj t. -r- o ••- . .. QJ= •• JQ «a; • co ••- i— s -r- « .£:
O D ro «- 3 «.)- I— 10 J2 D. coec-ooco . o-«S oo Eu»
c-ir: 3-0 i -uio -i-> (D ai ^> 4-> =3-oi»--o-o-tJ .TaSjlj
o -"ir> ctr_ cr=4j> c atatojcu =Scc
C0-r-0t_t_« ruco at = O.t»_ ro .. C -g to J2 « C-Uro
W * *O CX • *3r *r— e? r-- "r— • • O O "^^ T3 O 3rO4) ••OflJ "^
+JCOM- io«. +JCOJEZ -M ^1 o •— £ T- -o .?:.,_ _
caJt_34j«cMCT> re o re to •>— 4^0 o j-> r_ aj co c t.jj:>o
•^-Ore=^^og; ^^Q Q.C.U-.1-*. oiorEtlrrS .^"rrjoS
^ > >> E 5 CM +j rt CM ro ° co iZ|.c: § -a co to "§ ie ^"o £ to 'o -a
tf-
O-~ •• C «— » co^-^
j*: f— 4 o c^ i— i •-> ,_ [
CO -i- 00 V) ro CO IOOOO
OJ"Or-cn CCTI E-QOCTI
cocai^i JC^^H Ccro<-<
O ro OO •— • o OJ~_^ ro ro "^ "— '
s: ^> oo
37
-------
T3
CO
3
C
•r-
C
O
u
•
CM
Ul
CO
S
.
CO
c
CO
0
o
o
c_
o
•<•»
£
cn cn
i— i i— i
1 C C
0 -i- v-
t- £-
03 C_
i— 4J 0
-CO 4J
o re re
10 >4- t.
4-> 3 CU
c c a.
co re CD
o- e
• • r— -a
4->|O C
COIC O
t-l CO O
•r-LC 0>
U.C,
a. co f-i
o co oo
.c -o ecn
co c: o I-H
•r- re "3^—
oa
co
cn c
c o
•1- _Q
•a t_
3 re •
r— O CO
cn o
C Q.
CO X
—1 CO
•
C
re
"a.
CO
re
CO
CO
£
>^ >j
r— 3
re 4_> re • ~o
^ ^- 5 z o
co *o -i- 14-
+-> O CO
•o o -H ^H
CO c *" X1^ °
re re -c o r— co o
+J +•> C O r— T3 V
4->3'^co<«-recS'ri a.
re a. co o re ~~^
"co i co +-> ^ 6 E o "^ +j C >, "". **i
co re •-< co
•i— o O CO
P re o t_ i—
Cj—
O
CO CO
CO 3 3
O o <" O
C CO C CU
CO C 0 C
co re -i- re
(U 4J CO 4J
t- 3 CO CM C
co o O
t_ Q.-0 0
•r- X C 4->
— j 3
0 0
1
1
1
|
CO CO 1
c "c _j
•r- •!— T-
re .^ re j^ re ^
E cnE en E
o 13 o -a o
•§. S-i. S-i.
i, if, ci,
r^> ^^
in co
CO O
<- CM
re
CO O CO
re
o co >
v> •
• U Q
|
J
3
5
3
>
3
i
5
-------
-a
O
CL
2
^
^_
C_
o
x:
3
1- wlS
0.0 +J >
3 o QJ
1 C O> "B •
*0 = C *»
E*" 'c
+J O +» co
x:
C «- t- CO
0) O IB CB c
CO
cu
o
c
2i
0 •*->
co O
x: a.
+j 0)
c
a. c i -r- 4J .
O t- <8 X C CO CO
t— * «- o a>«o i (.
a> o >4_> M— -53 en a> ra
o3^- ** o "o >»
•a o NO • a»
a> CM a> t-i
I. (U J3 c 2
0) c: a> ••— IB i— T)
J* O l_ TJ •>- co 0>
(->B3Oo<-(-e: •
§t. V) t- +J Q. d) •>- (_
o o o •« ^ g a>
i— Q.I— (U C. <8 +J
enx:x-c=3CoxiB
r~«o»o-oo3BOj^-
<— «
CM
CO
|"Cl
•a
n_ w a)
Q. O +-> >
3 0 •!—
S o i- x:
O •!- 1- 0
f— IB >>
o c o c: -a
(4- .,_ .r- O> 3
§C OJ -4->
OJ JQ CO
O a) cn
+i O 4-> <"
x:
-a c. >r. c
V - ^- O
U C o "> CO
o a> co o ' *r»
-4-> O O] 1- 10
.
»—
*
M •
1 O) oco c.
co c Cn r- O g o <*-
J- o •-« i— r— =IB CO.
<-o>c.Se:
o = % 5 "^ * -c
o a. •*-*
t- =1— i-H
0) ,— IB OO
IB > .»
<= Ol2
co o
>>-(-> O
39
-------
CO
X
o
i— i
* '
CJ
0
s
LU
CO
0 ol
,-. "J
CD NX
£1
3S
gg
_ i i-i
0 5
>- ac
-J 0
s
t*
feg
c/>£
LU 1
i— i 00
a •
j2 ^
CO CO
•*
Q£ CM
LU
gg
_j
U =3
sa
O h-
i-l Ou
LU
a
a.
LU
•
CO
3
CQ
t—
to
t_
O)
u
c
o
o
c.
0
?
z
to
en
c
•1—
-O
•i—
C.
O
z:
to
•I—
o
CL
0)
•r-
i— •
(O
CO
0)
to
o
o.
LU
I/I
*t"
O
JC
4J)
4E
to
•1—
JC
+J
<*_
O
^>^
+J
•r—
I— •
•r—
jQ
ce
01
i5
to
f*
4J)
O)
*«
•*->
,0
i—
^>
^~
03 |
u a> a>
C -t- JC t-
n3 -l-> 4-> CL
•p -^- <»- >>
U 4-> -i- i—
0) « (-
O) to +J -M
T3 C
to ia to
O -i—
>, O -i- CJ
T3 Q) C O)
3 ft- 01 <*-
to ai 'to O)
a.
X
ai
to
jd
CVJ °
CO
>— 1
>
o -a
CO fO 3
•i-r— •<->
to
' >>
4J T3 l/l
C 3 T-
^
^_
C- 3
0) •(-> t_ >,
S O +J i—
0 O) r—
CLM- to K3
4- .^ -i-
r— O) U
O C (J Q.
•r- rt3 <1) to
4-> 4- ai
CO +•> <•_
•r- O O) "
+i 0) -(J
ro 4-> O) C
-!-> Ol jc O)
tO "O 4J CO
0)
CO
ai-i-» Jc-aoit-c-H-
to to 1— O) •»->•>- OLCO
£i<
LO ^H ^CZ
^_>
OJ
15
^j
.0
h-
O>
C 00
3£
a.
X
0)
to
-Q
O
t— 1
CO
^
ai
•r-
co
^_^
un
o
•
o
,3
CM
in
£3^
CO
JC
o
«J
o
in
>, 14-' • "§ C
0)
§
c
ai
O CO g to
to *~~> fn f-
03 Si
t- E •" t-
(O O JC 0
C. U +jj $
>, (t) c <*-
r_ CO -e- O
0)
SO) ^-"^ - TH O
OJ 4J — . O
e i
«*"* •(— ff
Q- 3 cr>
•i- -a **•
+j (O -r- •— 1
•o^
•^ o°
J!^ CO
Ul O>
3 i-H
CO *-^
2?« g
C t- O)
0) 3 4->
4Jt to (O
J28.
X
4JI 0) to
C; • C.
CD **-+-> no
•r- O c: ai
O *^ >i
•e- t- 4J
C C tO 4-
O)
c
o o
« -4->
O O)
,-. c_
JC 3
o o
O
•a ex
a> x
CL O)
O
r— ^~
0) fo
0) C
5
tO 'I-
t- U
O) Q
J3 •<
to
to
0)
•r—
40 if
o jo o *
<- en
to nj O)
to Ol to c.
O) O- a> 3
O Q.-O W>
X "3 T- O
a> > CL
t- 0 X
4J O) &. 0)
e o a.
nj c <*-
O 1C >, O
•r- O T3
<4- 30)
'c o to c
Ol A3 0^
•r- E to "O
to O -i- -r-
4-> JC >
ns to P 0)
I to
CO >> =
. t_ « o
r». o ts t/i
CO* O U. 0)
•
^^
c
0)
ia
r™"
C
0)
•r"
<(-
O-
3
to
0) 0)
» t_
• co Q. O) O)
CM H- •> U- >
oiu. a; i~-
CCO (»_
•r- «C IO t|_
ceo j= o
•r- tO
fO *_^
-------
in
c
O)
c:
o
u
I ,
in t_ T-
.
04->T30
CLtflCdljC
xre!o
01
Ol -S O
EM— O
o +•>•(-> o
3 c ,«4->'-io
re 4-> 3
i- .
^
o>>
cxc
Q)
-
oo*
0)
in
en
o
cn
coo
« O
in in
01 01
CO IO
£ E
CD
1/1 l/> i—
ns
oo r— •
CO CM
LO r-l
uno
O
CO <«f
cn t—•
CM «-l
ti
o •— <
oor~-
o re ~o t_
O >>T °
o
u
O t_ +J OJ
O QJ C c
. . . +J 0) O
r-H jvj ^H l(- -4->
-—IO ^" '—*p~" IO
oo
o
u
CO
o
d.
-u
<0
4->
in
•r- O •-
O -r-
£- 01— +J to
O -r- CO <1)
TS C^
c CD
+->•— o cy i— 3 i—
C V i— I O
ITS oo o. cnir> w 10 »—i
0 •— c ^T ° ~
•r- CO •«•'- »*• C71I»-
l«- Q. I/I «t3 C O
(/> -I— o r~-
o. t- co"
•r- J>C O *J t- T3 i— -I-
(/)(/) O tt) O r— t_
«*: 'Z i-4 I T3 -r-
O
[ * ^ (_ • «^B OQ j * (JO ^1* ^}
o<->4-><7ic: »«c
U_ O O O «
>i— IO O CO «+J
CSJ .C X JT >,4-> O) «£-«}•
C I O '-
CM
in
01 Oi "t- IO I
j?>°-^
h^_
•i.
r-^ 00
c .
-r- +J CM
IT) ••- (U
(U oo u u
cner* = =
41
-------
Major concerns
3
| «
5 -M
+j
_eu
u
•f—
3
ro
-
oajoo
•— T3
O •>-
I •!-
to -o
l_ t-
+-> O
CO U
c j=
O+J
0)
> _
+» M-
4J
oi>
o-r-r
ai-a
on3
aj
t/i -a
-r-(l)
oaj
t-
a)
Ol
•r- +J w
> C 4J
CT
CO CU C_ (_
ia (/> eu o
J3 4-» >4-
CO IO 0)
: O -g 4J «
; c £ -r- -o
I Ol to
.to >
I CO "O I T3
.CO CU 4J 3
: 3 to "O o <"
Z cu
*« u
in
o ia
Q.JZ
X -(-)
CU
o
•a a
|ff
5*.
to r~
3 «
CO CO
CU CM
<8 O
•C 4^
•t-> E
C O
CU u
CU 0)
ts
•r-
-O *4-
cu°
.r— 4J
•»-> 5
10 nj
eu
3
••OH)
C CU 4->
O V) (J
•r-
<8 t_ cu
r— O "O
O.T- CU
o .a
a. o
c >>
c 10
O « £
CO O)
•r- (J Jrf
t- C I/)
1C -C
CM
01
oo
CM
(U
r— X3
<8 c eu
4J CU -r- •£
•r- > n to -a <4-
C -i- Ol S C ••-
J JZ CU
eo a: Q.I_ Q.
T3
I C
i +•> (8
CO
UJ
•r-
§.
OJ
j=.ot_-'-
j-> ^ Q..r- W) 0 4f
•
<" O ••- O ff- O
tJ°° '-'^
<"uioo e m c +j -a
-(/)-O t_Ul+
3 > -i-
Ot-Vl+jO)
x 01 oi oif—
o» a> c. 10 10 01 aj
"
i «- eu ^
(/> cu a. i/i
IO J^ O. «r-
O '•""
T3 -O ••-> —
<1) r— O CU
J3 r— 3 CU O
0) O d. c
c «- S x to
IO CU CU (J
O > -P
CU •r- O
i— +j cu •— eu
co +j .a m
"O -f- -r- UJ
m co a>
o T- 3 t.
C <«- 4-> "8 U
•r- -I— c P— C
CO co CU Q.T-
^S'c
•r- c CU
CU CU Q.
J3 +J C. .
P~ 18 2 "1
<8 —i x: cu
CO +J
* 18
co . -i-> e
CU CU -C -r=
•* r— 01+J
. ja ••- co
<" <8 E 0)
4->
eu o "> j<
co cu O C -I— IQ
eu
co
g.
iS
CO
eu
— — — ~O+Jr— t. •'-•'— Or— rtj
S O.CM t_ t- o r— COL.U
O O ^3 JZ to 'r— CU O
2>r- ^»r- r- f8 • O -U T3 a.^! CO
•r- O O 3 -4-> Q. t_
. _ h. -. <" <.="->>
C 3 4J t/j f— O t/) -C >« O
(U O ^
r^.
C •
O co
01
C +J
0---
co
check latency
three cancer
00
CO
IO
JC
u
IO
#
•O CO
c c
10 •!-
IO
Ql V
J= t-
O "O
0 0.
c to
10 C
0 3
o> .
c o
•r- CO
= ^
{_
O
fQ S
•t- •(->
Q. CO
10 >,
A
T3
Ol
IO
5
"oJ
r—
+J
C
CO
0
•r—
•i_
C
0)
CO
<4-
o
Ol
CO
3
IO
O
£
•
•a
a
10
IO
00
^
00
CO
t—
IO
4J
•r«
s
en
(U
r— •
z:
4
i
c
CO
o
t(_
°-5
t_ CO
OJ E
3 CO
i— O
F- <+-
CO
4->
01 10
jc 0)
+J TJ
..J.
o^
I— 1
•o
c
«3
Ol VO
•r™ OO
U t-
O) Ol
C *T
c +^
s°
3
i
O <4—
-C O
0.0 ^
i— t- CO
0 0
CO
liir
IO 4->
IO
- ~ QJ
*- ••->
OU *w QJ
0 E "O
•0
CM
ir>
CO
c
•r™
J«£ IO -55
O) E C
-00 «
3= 0. {_
1 FT Ol
C >> JC
O ^" -f ^
?^ f^
^
•(-> i i i
1 T- C <- 1 « 0
IO CO o O Ol •(-> Ol
i— i— C O Cn -4J CO -r-j
f— 0) Ol IO 0) jQ
IOC7>4->OI-I->OX:3 •
Scci— co en co c:
CO O *— ' IO O -r- O
r— O • -C CO ••—
Ol CO Ol LO -U
ja 10 • > i co >, 10
T3 O) H- o *"- t_ C
• C. O t~ t- Q) O>
>> Ol -r- 3 T3 Ol > -r-
+J .(_> 1— l/l <*- CO
•p-M_ to Q. Q. ia(t_.,_-o
co x "> t.
O)c+JOI-l-> >>COO)
UOIC COcOt--r->
0)>oi o co -i- o>H- -p
ir>
^f. Q
V
CM a.
rH ^.^
•o
0)
•r"
<»-
O
Ol
a.
co
c
i
!J
1
Q.
O)
;
:
!
>
3
5
^
J
j
3
5
-------
_>,
w
3
C
4J
§
U
*""*
•
co
LLl
•J
§
CO
C
c_
Ol
o
o
u
o
•1-3
CO
cn
C
•r—
"O
C
•f—
*4—
t_
O
'at
£
(/)
C
i.
.jj
C
CO
•f—
r—
OI
(O U T3
> as T-
ns u
£_ n3
0 0
•r- +J Ol
**— O -C
0 <£
Ol 4->
0.40 co
OO CO •"-"*
40 0
13 (IS 0
OJ t_ o
40 •
to 01 O
3 o O
**"O C r*H
•o co ,
(O ™CJ ™
AT- Ol
O Q.
5>.c*~'
C
00 T- O
I—
00 C Ol
o >
<<- T- CO
o cnoo
CO
CO t_ to
o at
01 *0 C
•o s
•r- U» O
O 40 C
1= C J<£
•r- oi
O) to fl3
E2!S
4-* 4J
«S ns 1
f Q
C 4J O
o cj
1/i t'hr
0 000
i— 4J •
O. o ^>
X ns «
CO q-CO
*« nl00
'E ;z"S
1/1 4J 3
3 1-HT3
TJ O
C C «-
"-« ns Q.
C r— 10
O ns CO
C! % .
4-* rH
Q-
"O «" 1
*O CO "3 >,
<= Cl 4->4->OllOI 4-> r-
'-•'- 3'i— cn 4->l.4J3C04JCO=OICOr— CO Ol
-O "O (O fl3 CO r—* ^- 4O 40 nS O. 4-) C -O O
CO U C i-Cj-oOOC+JOlO'— 4JOCO CO-1-T3
CCO 4J (L.
O'f-T- 'SWT3I 2 Ol CO O Ol -l-> IB Cn3T3O
CL O) =OOIOC2 3cO-l.n34J 3,— OI-r-
(/)*CICO T3-C+JT-OO(1JCOO|£-O CO r— +J t_
tu a> t- comnst-cjro •r-co-'-eo onset.
t. 4-> L.OI= COQ.C
nSO 3O CX Q,) d *r— C C T— nS L. *r— C
CnOlv) i — t/>cO a)l»*oOca O>Cg-r-OIO4-><4-'OOinS>> +J .r- r— O
OJCOI 0-cOOin3j-rt_c3CO = «--— ns 3t_.C'i-
^•t-OO = (O-ain3nS.^eC-onSnsOE «=CCXO-o
= >>
T3 CO i-HOO"*VOCn*a-tr5 •S' Olr—
C0»- 4.) r~-cOCOcOCMoCO CO i;(OCM
3T3 OS CMCMCMCOCM-*^- CO ** i* tn -i— L.I-H
r-^ t. 01 t-
cn >» ns x: 01
r-H OQ S» +J Q.
1 1
C trt I i t
O O >, I -i-
1— "o i^E E
1 OJ o, O CO *o •'-
"«r— o c in (5 4J
"^" Q. i * CO | * r"- CO *
CM t> CO •>- ns 4-> C
IO L. "O 4-> I/I Ol
•0 P£ 3 3 ••- C -0
oicn co ~o ,— .c ns-r-
4Jt— 1 O CU ,— +-> .C O
ns O.4J O +J O
C •> X ns o. E ns
i- O Ol C O t_
C r-l T- c 1- Ol Ol
S3 41 C_ .r- <4_ .,_ .E
4J >>r— OX i— 4J
C i— .Q r— O ~O {_
O 3 T- J= T- 0) ns "»-
orj(/> CO
-o f- t- cn
w-gfiSCi 44
Ol
1 JZ
o
o. cn
Ol T- 1—
t- CO O
Ol Ol f
(J 0)
ns £ 40
E
4-> O Ol
0) O C
T3 01 0
•r- r— CO T-
O .Q 4O CO
O T- nS CO
nS co t. t-
-------
A
l-H
X
o
>— <
Q
1
CLJ
o
f*«^J
SE 00
uj o;
03 O
^"™* t"1"
Q — i
<*V Q
O SE
O OO
>- Q;
— 1 UJ
U.
o ..
o
00 Q
UJ o
Q "7
.mi co
r— «
oo r-»
0£. Co"
Ul «
C-J CVI
*^
O «J
o;
o «c
^^ _J
^•3 ~ ^
O o
0 J-
l-H ft*
UJ Q.
o
CL
UJ
e
**
UJ
— 1
3
in
c
t_
O)
(J
c
o
o
c.
o
•r^
c
01
•f—
,^
IB
OO
1
Ul
8.
2
2
t_
O
JC
^j
"*
T£J
Ol
•1—
E
•r-
01
IB
f
t/)
t/J
*^*
£
Wl
JC
•fj
IB
Ol
**^
F™
IB
+J
0
1—
01 IB
IB S <*-
t- o <"
i. O 0)
Ol **— w> "O
3 C Ol ••-
a T- Q. o
, >>••-
•sl!!!
•r- O
+J UJ JQ >»
(rt fw (O ^
•r- O
4-* *"" C C
to p~ o Q^
t! !'£ Q.
CL
X
01
"*^^
U)
-Q
U> °
«^-
^*^
oo
CVJ
CO
O)
4J>
•r™
00
1 1
IB •!- >, «
C- > t» aj in
4J -r- • O I. IB +->
c -a -D +J 3 2 £-
.CD
C JC C C 3
aiujui-i->oi "o> o
ui c n— o -a •(/> jc
Ol O IB *F~ Ol O) 4**
cl'-J3 -g 5 5 .4 a. 5
o
a
0
V
CL
*«*
t-H
1 — Ul
O 4->
-••. IB
CO Ol
a
r—
IB
JC +J
O O
IB f—
1 %
00
\
Ol r^ C ,
J3 O "~ r—
o 01 c:
«> c o
Ol IB -r-
JC 3 C "
-u c t-
T3 -r= 01
• O 0> >
m 'r- QJ Ol
o> £- ja is
t_ 0) 0
IB CL «3T
i— Ul
C {_
Ul O IB •
S ••- oi cvi
cD +J >,r--
in > co r-<
a.
x
Ol
WJ
«5f -Q
•sf 0
^^.
10
CVI
cr>
rH
Ol
tfff
oo
(JJ «
t- 4J T3 |
O O) «- IB 1 I X
3 -C >* '^ E o ^1" Ol -^
CVJ O *4— >> E d ) % «p- Ol (B ^
oil — O-r-OI Ul t— T3 t_ O
t-Ol r— JC IB UOO OIU1JC
Or~*OIIB-lJ' lOin O3'r-O
« c5 o-c'SluS'm'icvi ui°
O -r- O -^ p~. •!— CL
r_OI-i-+JinO C+Jr— i-lt—JC
CL > 3 3 cvi •<— IB ul
Q.-I- -0 JC t— . U< 3"O • t-
IB ^3 3 C > O ^3 «i^ *n >
Ol 3 O *r" *^5 • > ^- w **"
in Ol r— jc j=EOJ--
O1-O4JC. SEf^TS'^SS-
JCOICr— OOE'r-Cf'Ct-r— 3
— re 4- +•» •!— +J <-H -r- CL O Ol
CO <-l
• »
O O
0 0
Ul
re
o
JC
CL
E oo
-J*00
>><*- T3
U
C
Ol
IB
r—
_
Ul
IB
**
Ul
&_
jS
£_
i
•o
IB
O
r—
IB
£_
JC
vt
•r-
•o
C
o
"5
^
<
O Ol
Ul IB
>> c 01 i—
S 0 IB IB
•sS
2
01 — •
f
•a
c.
IB
Ol
£_
Ul
C
o
in
3
r—
U
o
u
z.
a t-
1 • 0
« O C Ul
Cvl -i- o t.
OJ S- C — 0
^J » fl) IB | ^ 4*3
•r- i— JC CJ) IB IB
O O +J t- t_ U
9
t- «- T3
O 01 Ol
o >
Ol C -r-
O "9 S~
c o oi
Ol "O
in IB
C_ <*- ja
0.0
c
Ol Ol IB
JC O o
•U «=
Ol JX
O> Ul y^
C -Q •!-
•I- IB t_
T3
in o I
«_ Q X t-
Ol S Ol O
r— 3 r™ C
CL I_ <4- (0
JD4J CO. r— £ "r-O
t,-r-u5T-ai Q. .C i.1
Ol E 3 <- CL_C: 4J CO «*•
a: ?g r_ T3 a. IB
CL C
• C. * -Q -a
cvi ai i— o -i- •>—
t^. S 1 T- o <->
cr> if> c •»- -r-
•— i T3 • IB ja ja
Ol «S- O) L. t_
o u> • t- 01 01
+J 3 CVI O JC 3=
•1—
IB
E
•r—
.C
•|F~
45 *.
Ul 'r- *
•r- 5 CVI
C •
c T? t/i q
•r- 0) r— 4->
U. > O
i— t_ T3
o
in
in
t— i
E
O
r-
1
LA •
C_ O *J Ol «r-i
3 > c ui «er r-»
o c o o «CT>
<»- •!- O CLCVl r->
IB 00 CO
Ch co
4-J t— H C^
Ol •—'l-H
)
en
re
a.
?
>
5
J
j
-
O
11
J
5
«— X
-------
"O
O)
3
C
•r*
§
U
^**
e
Kt*
LU
3
i—
en
c
t_
0)
o
c
o
c_
o
••-3
en
en
c
•r*
•a
c
•r»
t-
(.
O
•r-J
£
CO
4->
C
I
4>3
<=
•f—
,0
O)
c«
3
CO
g.
i2
^_^
en
«h^
^
O
JM
•U
3
U, 1 tt) -U 0)
C) O C | ^-^•^ Is C 3 <~ rt) o -^ ^™ O "O
>C«J O C Q) 3 • 3: CO X: 0> 4J C <«-
r— 0) +•> Q. O "O r- >, CO Of—
St_ (C +J Ot3 T--O O t- -O O
CX'^" IT3 2 t_ C +J O Ul QC'>- CO'r-3fl3'r-
JC ^ r— O 3 O O CO cnO -d^O •"OC'^'XZCOOJ+J
•MQ> ^— -a O C -i— i— co -Q ai "3 co > a. u
w+jocTJ+JtniocnT- ,i-i-<-cn -a ,•!-
w «r— Q}Ct.t-O*O4J Ot^-CO OJ3C(U4->CUflJr— **-
O-Q+J-r-flJOOs « C->- «4^X: O-QlOr— O+J^^'P- i^:eom
4-5 4J OJ -r- • CO 4J -U QJ O> T3C « >>r— +J O •*- «* •»- f- CO t-
3 CO r— ej_ ig 3T-T3C ••- O)COjrX^3'OOlt- -r- -f- i— «J3
"~ o QJ o u c. i_ oj IZ co a.
^enO.^a.OTfSfJti'cn z"SS5"a.cIo!So§i2'5'Ea1
tr> tn"
O o en
• . c-
Qj 0 O OJ
en x v v u
-C 0) 0.0. C
RJ CO O O)
a> -Q to m o 4_>
•a o lotno jc«j«
ro • . • u 3 co
t- CO CVJrHCM n3 r_ t-
-c
C ,-1 4J 0> 4->
10 ^. en 3
•^ +j -i- -a o
•— 10 E c c a)
+J (V CO -^ (U 4->
O 4J — 1 O 3 CO (_
J— -r- 31 I_ a; H- 0) >»
CO Z Q-_J CO 3 J3
en
en 0)
C- ^£ t-
•0 0 en 3
Q) co -d .,- co
4_> •• 4^ | o
lO C 3 CL
>••-<: c x
CU \x o Q) '
f-^.
^3
4-> 31 _J O +->
c i a: 10
IO C Z n) r—
o ox-' > 3 O
«C t. r- CO •!->
1 CO
•— i ~o en
i— t- •»- "a i O nj
M-I-JQJOCX: 1 0) , § * 'S -t §>
<•— oiaUcO)«~ *r- ^ *•••••
E •— esj 4J x: en ^j-
•r- «J OO C O •— CO
-c en o eo -o i— o?
•r-+j.-io ojcoi-i
•r- Q)*~" — .c 3 (0 36 — •
o; 40 oa
>
)
;
(
1
I
c
t
)
3
i
5
o
*^r
-------
concer
in
o>
•O
•o
QJ
3
0)
(_
1
IO
Ol
" O) I
-. -. C jz V .> * CD »r- 'I— QJ 3
O C i * (O »r~* '^ QJ 4-^ t/1 C- O I * fi™ *X3 O
i \ QJ Q ^ (E ^" /rt f rtl j ^ » » rT» * .p
O) C > Q. O
O>
c
CU
O
U>
I-o
Q-r— i—
O
o
Q. C
o uj
OO
4-> O)
IO O
Q.
co
to
3
"c
Q. • C.
-aOt_o
O CD !_
C3 C c
to .r- i—I —^ t- O>
(/)
IO
t.
« E
I
3
CD
•r- £-
Q)
-r-3
o: +j
0)
O
II
O
CD
oc re
c -,- c -o Q)
at o C-*:
C >> >--r- «3 Q) Q) Q) E
•<- a-
_
c -r- (/) 0) C V)
«3
0)
Q
cu
01
oo
X
UJ
O CO 10
^O IQ Q
CM CM O
O
CM
CTl
IT)
0»
10
CM
C
Q)
•(-> I O)
41 C *>
£ £ K
O^ CD
« e
CM CO
CO
«3 OO
e •
O O
OJ O
o ^o
e e
CM CM
ts
fS
en
<=
o
•o
QJ
C
o
o
E
c
Q)
a>
C
at
•)-) t
(DC
•Pi -PI
00 O> C _J QJ C
I— •£- O IE £• O
oo > 2: z > s
o
o
LLJ
_l
CO
I/I
*->
O
Q.
C>>— O-
3 QJ Q.OO 0) *O X
r— IO O O>J3 "O i—
3 r- E- C £_ C 4-> S-
o t_ a» o ~a ojie-f-o
•i— e
» ..o « g __
t- t. 10 S c
O)
C O)
i- O
. (.
o
ex
X
at c
O CU
IO QJ
CD OJ
O = =
S- •
t- v> E
3 .,- 10
in c
QJ -U
QJ {_ QJ
J= 3 r-
F- O) 5>
IO C
IO «*•
t- atoo
*"~ C ^Ht
00
47
.^•51
^
•(-> ai"—
u.
o>
-------
jor concerns
1 0>
o .c
•i-rex x: re i— x: re re (.01 •.f_4jn
3 i— x:_c> X: 4-) l_ x: C to 4_J O-.r- U i—I
c/) -i^
• I_ . O
>> O r- O
,
.t- £_ +j +J ^- OJ 3
1^1— (aa>o.t/>co B o -r- re oi o s
B o -r
5>o+
oior-
t-'i-»-o ooooirec c cnc+Jo
t--xg-r-fl3-C: CUnJ-C: O>"O OJ-OW1O 10
,s
3 »uo.r-a)a) a.,— <»-x:t.-:3oocc:
----
re
CL
X
Q.coc+J rewi
E i— O) COS-i-
S re
o
V
a>
3
C
C
o
o
LO
_J
CO
co
a>
c
o
d)
CO
CO
o
CO
CO
re
CJ
r~~
re
o
0)
re ••-
>— ce.
£
CO
CO
o
CM
CO
O
+J
CO) i—I
-Q CO
O| CO
CO
CO
CT)
c ro ai O O c _?
O3C04J+J OC.OIO
5 •—i o '"~ corex:^.
c:o
<1) to OJ -r- 0"»-3-C:
L. '
a..
. (/) co en Q} re 01 O)
aioc •>«- +j 4-> -o _c
W)> o=3O">yOT-+J
S-yt-OJt-eea. OT3
o
t_
CO
a
.2
i >> oo tn
r— 3 r- | | . « C_
re Q. ta -rj • .^- ^_ -p c r» <4- o;
CC.MO oi cu tn t_ >, o o " «to
^-jtfoi'r- re i— t-'i-0)o<"a)CM-or—
x: 3 c. x: co « c- cx+-> t_ E re •»- ja c
i-'I-*_0«Pc:^S3 CM U
"Sx: 3t!.
-(OSO-03
o « 14- , j= T3
j=«3-o o j_> c: a; +-> ai
ot«t_ore
— — ~ .£> o
ni y -r^i
O O) T3
re
X
O)
OCO +J Cn c: ><- 0) CM = OO O)T3 «-O
o O)cocreot.p».c.
O £- 0) i- o> «x: O) £_ -C >, • oi 4-J co •
_
5co s o
oo -
C
oT--oaio)
c e o t- +j
oreoo)x
u -4-> H- s a)
(-
o
3 EcS
i— T3 •— O>
48
-------
,. — „
•a
ai
c
-p
c
0
o
*^__^
.
n^-
UJ
ca
•=c
01 4_>
1 g= c:
3: CM •"— a>
0 . -M 4J
,_ r— cvj to -a
<4- fO r- ,-H -C ,— ' CD
o -.- o en to rs
+J 4- >, =3 l»_ t- C
to c: i — o o o •!—
4-> <1) O) C .C -C 4-)
c~ t % ^n O ^ c~ | ' r~
ai o +•> T-«CO
O) (U 03 O +-> •
t_ «>sz:o -ocx.
3 cu «f cu • c: p
tO £_ CU 4-) I/) OJ 1
a)to'---to-aogo
EEO:>>4-)t- CUO' —
CLI — (C J_ t- u i —
O X c"" i~> CU O *^~" CU O
2: ai o rs >>*-
0)
•r- ^ ^H CD
co +•»«/> o {vi
co i — o: "-^ - <-5
cu
M- ce:
o
O-*
in —
CU IT> to •— I i— 1
to *^J" ^> cvj t — i
ta cvi CD
o «
i— CVJ
C
c - cu ••- J= -i-
C 4J t- jj tO
2 £ 0 ••- c
W Q C71 £Z O
"O x> '•" to ••—
to _i£ o •>— */* CU < —
O -^ CU ra <•-
.c: £_ c- >, t- cu o
4-* O CU <-> tO
ZJ 40 _C C -r- •' O E
. (0 Z -r-
cn-I^ aT o 4_>
P **— t- to O "O
O *^ CU rtj "O ^ CU
-C c: -c cu c: to
+-> O14-> tO CU t- CL
^_ -r- .^- .,_ t_ o "3
«C to cu -a -*-> u— < —
Q
>> >1 1 1
i— x >, t- c_ «3- to
c >> o r— o cu •cu<»-
TOO+JCM— r — t/)CVJt-O
a> 't~aJai~Fjt"-r.j'""to
c: • zj d.jc •"— i — cct-
CU fl> ^O 40 O "O flO C CU
•i- ••- 01 o +-> to •> 1— ••- E
O i— i-H CX-r- CU Q- O 1 4-> 35
•i- 1 Q.ZTJO4-JLOW1E
oCL^^-3CO •
CU O {^* CI O *^ O ^3 * O~^^ | »
Q.-I- t— c IT3 03 J3 O C {VI CVJt-
+j 4J t_ cu 03 c: o
•i— i— {U -C O "Q.CT3 U
*O ^— S nn *^ O {^ ^J ^*
CU f^ 4~* v^ L« *^ *^ e 40 V) C CU
•o •— -
C i — t^-
3
-K: 4-> •-<
5«— 49
_,
ai
en
3
Q.
CD
—
~£
O
D
CU
J
c
o
•a
cu
5
o
-------
ts
ai
§
o
UJ
CO
VI
c
Ol
o
o
U
c.
o
V>
O)
C
c
•r"
<*-
01
<»
x o+Jt-ororo-M-r- >(— as a.a>
0«/> -l->3(OUt. 4J«3=in-oa>>T-oCQ.l-
a. a> • >,•!- o -r-occ+J t- ajojia O>^— 'f-(OO.C4l -t-> ^-»r~- <»-«/> i— L. CO
+JXC3OO-OJ: 4J3S_i^ oo cvia) •
c» air— o4-» <«- c-i-'t- a) «ja-(-> cu 4J o oo +-> a..c a ^ w> j-
0) OlOt-UlOO Cja «t-r-4J4JCOt.
-r-c-r- ajreX'i-cnOJsccri't-cOx •OJ'i-^w+jO^H-Cio'aT-uo^'p-
o j= c\j x: =« +J JD o — ^•>--nj sst—^j+jj:
O)
-a
O 10
c j=
ae to
;8^
I LT5
LO
O)
C C=
o o
•r- C/l
in ai
•r- 0.
O)
s
O) JQ
T3 '<-
to •»-> f—
1C 4-> -r-
I
c
O
8^T
o t- m
0> O £
Q.4J
en TO Q. co
•r- U 3 O)
£*. W> 13
CO 1-H
o.r~. -a in
5 «
X
Ol
J& •
+J VI
• *w
T3 0) >-
C -O Ol
§M-5
^00
t- 8)
to
£ re o
in 04o
- +J ro
OO tOJ=
I— O> •!->
CO 0>
t_ CO
I
CO
a
^
g
"aj a) i— i -o 01 o en
MO aiaic'CocaG
t- c o co CO Ol
O) s t_ en
•r- C
-------
CU
o
X
o
t—1
Q
k\
LU O
CO C£
I—I a^
O CO
O LU
UJ =5
I— CO
< CO
01-
5*0-
o to
So
O
LL. t_3
0 i
coco
LU •>
=> CO
I— «
CO CM
O
CC
O
V
LU
CO
O
10
LU
CO
CU
U
O
O
t_
O
to
c
•r—
•o
O
CL
C
- C i—
O M— CU 4J o ^ *—
(COt—<£~t_OjQrt3 CUCUO
C CO 4-> J= >>r- CU e
>
i— t)_+ja.ucutO(o
—».—.10
LO LO O
O O •
• • o
O O V
v v a.
Q. n - j
O
CO
4->
CU
CU
CO *-
t— > >> CO
i— c— C
C c f-
o
c_>
I I I
CO ID l~*
to » X
C4_r- 0)
O •—
1C (_
Jki C CU
to O S
•i— 'i- O
a: +-> o.
« is)
^ >j CU
>> <«- 5? t_
cu t- o o •(-
I J= 4-> I I r- T3 (O
>, to +j to cu >, i co co o c: c
O "O c t. o IT-., cu c: c o
c
to
cu
to M- C
CU «r- CU
&T >>'
dkw • VI ••— fc, " ^ '
a.jp •— -a c •(-> 3
ii_a)cioo3'^->>4-> +j o"
O-CCU C TO"— CEC1*-
-
i— c
<»>> rs
o e +•> t— to o
>>>"S
o o .a
:c—i o
Q.O O
I I I
CO CO 1^
...
VO CO «3-
IS) tO IS)
§11
to to to
£ £ c£
ck
cu
o
X
ai
i »
E y «
i
cu
£
to
o
to
O)
cu
le_£S~
c o o o to _
o "*- « ra Qi CU
CU O CO
o ^ o
^* T3 ^~^l^» o ^3 t.
;>} «cyT3O «C»r-QJ
» (— CO -M -P" o CO CU O 5
IS) I
O I O C. to
to co cu « -i- to I—
I CO
f— I*-* O IS)
t- i— O O
O >> «3 +J Q.
,- CL t- •!- x
L. e *t— CU 4^ (J *4" CU ^*~
-PcS£_tOQI»» O
.
>»-Pc
W>TJOL«J
10
I O •!-> -t- C rtj
i— (— t_ -t- C
.
0.0 C
I IS)
L "O
eg o cu o
•1C £_ 4-> d)
CL. ^7 1C *f~ 4^
» (!) Q. CO CO
: +J cu £- -r-
t- cu c «_ ••- co a>
t—I o OJ CU
r"
O :
"o. I
i— to er>
cu -a ^~
-o -a c ai
(. C
«
-------
^>
^3
CO
3
•r—
•P
0
0
•
10
LU
—I
CO
£
r—
Major concerns
t/1
o>
c
•1—
•a
c
•i—
o
'<~3
to
s:
VI
4-9
C
•r-
o
CL
4J
C
CU
•^*
^—
c55
cu
VI
o
CL
X
Lu
in
*tT
0
3
4J
C
CU
en
03
cu
4->
O
4J
CU
3
V)
O
a.
X
LJJ
in
03
E
O
U
s_
03
«n
cu
VI
VI
•i—
4J>
**—
O
to
1
en
in
VI
S
00
1—
oo
t^-
o
VI
•t—
cc
c_
•o
cu
4-»
O
•r- 07 « 01 y m
>
t— -PICO -pcucra
.— CU£=4-> ' ot--r-E
« -o c in >
O O"O4-> T3VJO4J
CUVI-i-CUC O> (X) £ fO
t. ra ^j c_ oj m 3 J3
in CO c- 3 vi £_ 4J
C CU CU 4-5 CU CU
O«-3int!- viv>4jvi
3 CT'i- O) S_ IO 3 S_
•OV) Ci- OOOCO
CU O g i- -C JC O
vid-SS-o 4->£-4JC
(OXI.T3C 301.,- 1C
jacuH-fOoo =coso
^_^ ^ ^ ^2*
p— £_ CU
, <8 >j CU >,
, O VI o I/I £— o S—
^— ^ L« £_ 03 R3
Z2Za3,»Jn32C CO»J CU
VI Q. CU O CO Q. >,O >>
C 1 ^ 1 *> , 0 . 0
CU lf> Lf) <— ^ »— ^
£- IO lf> CO IO
*4— i-nret— 4rt3r— V)4-»VI4Jvi4->
R3 O (O (13
CM 3CU3CU3CU3CO
r-l £_ £_ S_ S_
§-^ C3 jj^; O ji^ O
vi E vi S in E
o: cs: cu cu
WHO
"> £- Oi
viScnra «< co s v— - t.
r_ -r- CO , — ,J= 3
«34->-CECQ+J4J4-) VI
o *— * » res c o
•i— S-jCCMCCUCUQ.
§^43'"" u '"" s J iu
OCMn3*-c£«-— orcJcs
« O CU O •— CL O
•oro-i--^' — -co >,cu t.
CU <*- £- -i— t_ T3 It-
>> it T- Q m o cu
(O c5ja!CUs:4J^-».
t-oscn u a. c i— CM a.
Q.O -f-CovinsCU oO
vi — m •!- 3 CU *j Q.f^. I—
•a
CO 1 «
£_ VI -O •!- y!
CU CO C CU JQ r—
• g 4J O > t- O
"O ^- ® B VI •!•" f— CU C
ccumc:.i-4->oj=cu
fiS JC O C "3 > JC
r— 4-» SS -r- .,- Q. C >>Q.
fl3OO4-> g3.r-XO
O) S. VI -Q O O t.
tx! 5l_Cl.o
s: m co t- +j
V>«3 • 3 T- VI J +J3OST
O CI fi- C -T3 03
CU t_ 4-> in CU c CU
>,(-CUVI cU"O->-viin
t. CU O '"" t- •>" 3 Qi
4-> t- c cn .,r- -a -a
incuraQJroOcCT-
•r-O!O£-C:4-'«3'<-O
52
c
CO O
-P -p
V)
O CO
40 cr
2 =
*«
£3-
V)
£= 0
QJ Q.
OJ X
_Q
t_
5
•p
«3
CO
risk estimates.
— J
o
1
CM
1^
CO
^s
V)
Q£
in
c_
§
J <
03
CO
>-
c_
c
c
4_3
<•-""•' ci) ' • - "
O 3
One-to-one matching
cases with referent
is no longer valid i
to loss of 30 cases
and 20 referents.
^.^
•}->
S
t
a.
CO
•o
f %
^
CO
CL
cu
cn
03
Q.
i
V
3
5
1)
3
J
3
•— "
-------
T3O « O> i— c C t- = t-O
0) -I- -C >n3a>T-£-Et-4->
+j O (8 O en 3"33
T3 C E
« OJ >> 0) O
C S_ O 1
£ J= O
*
O.
(!)
o
tr
o
u
ai
o • c: = Oco-r-nj
i * *ii m ti\ m yli i * *"\ fll ^-^ g t rtl «*t
s u/ us +j vi «t— ^j .f^ iv **j
CO > T3 r— C O .Q i— "OST3 • i—
VI ^K* V^ '^ U/ ^» ^ ^
-OB OJJ=OO t. JQ x **~o XJ
'i™* ^ *^ r*" 3 liO Q) QJ QJ QJ
>I3 4J
OJ
« .,-
U r-
OJ
T3
I O
I CO
(U
u
en
c
(_
3
3
C
03
a>—.
CO
11 §
on o
• o
CO
• 0)
a.
o
o
O Ol •>
C CM
II «3
0£ O O
•o
: ia
o
u
a:
-o
ai
3
O
O
re c:
(U
•a
n QJ -rj
O +-> O)
LO
LU
co
C
O
CL
C
a>
• <^- c a»'i— o aj
:"«>o>Tg .f-ooS+j
•r- GJ «r- *<~ O ^- 3 *— "O O3
rt33'r-Q>CO»r— rt3cO rt3C
O > CO O T3 t- 4) >><»- O
j \ t. ^ fs — *~ ^ • »
O ,5,
Qi V) t/> t. QJ **f~* t-
t. «j « Ol I O
\V *^^ ^"~ *^
Q.I— 3 O
uco
u a> c
t. t. • -r- •»-> O -r" O t. E ai
0> Tg O tt) O
CO • •
•t- 0) en -<-> c
t. t- c a> •»-
t- re at -
•a a> o t- o a)
C ^" C «r™ 'i
IB 4^ (U 3J 1* r-* u)
0) 4-> CO)
_+•> US OO +j_i--—
COC «OO)-'-1r- COCOl.
Ctt»>-'t— t_4-> >C 3O3
.
c x o c
CO
o
X
LU
> a
"^ii .
.C U) TO
c •*-* aj 3
CO
at
(o
o
t- Q.
d) o
•i— ro O>
E S rt>
o
•p
C HO 00
0) O!
4) +J ,-1
-------
0)
o
O
U
o
•I-J
in
o>
C
•r-
•a
•a
v
o
u
eo
co
«u
C CO -U
•r- O\ O
,-4 CO
-a^-in
CO 00
o ,—.c ~c
•r- to 4-> -i—
<*> "} 3 I S
"£>,§-=
l«l = |
CO
7)
a.
in
CO
in
(O
O
to
O
CM
t.
o
^? c Ja
CO 3 t-
o I
in a>" u. E o s cri1*- 40 c <*~ ^~
in -I— • E in o C3 £- -c Q. .
c O> 0)coov>
en
54
-------
^
T3
OJ
3
•r™
O
U
Nn^
Lf)
UJ
CO
^_
CO
C
c_
OJ
0
o
u
c_
o
•i-J
g
CO
en
c
•5
<•—
j_
o
a
to
c
•F—
O
CL
C
Ol
•1—
5
OJ
im
3
V)
o
Q.
X
LU
CO
O
-C
•M
3
•a:
<4—
0
CO
•!->
C
O)
cD
t.
3
CO
re
O)
S
re
o
u
re
OJ
CO
•«-
^
o
CO
o
•SX
•r—
t_
en
e~
CO
re
OJ
i;
C
*
0)
C*>
3
0
Q.
X
0)
f—
fO
c
o
•^
4-)
re
Q.
3
O
fj
V)
T3
O
3
0) 1
o> o
3 ••-
t! a]
o
Cu^—
eg re
3 0)
?1
| —
OO
CM
J=
.JjJ
•r-
_- Q
2 =
re o
O •4-J
to t-
cn 3
3 "re
CO 0)
.IS
> °-
X 0
0 t-
c o
0
're oo
CM
-t-) l-l
4J *^
^
, O)
co re
O) Q.
CL V)
CO
C_ M
il
>»
X
0
co c
4_> Ol
0) Q.
0) O
t- O
S re
vo^S
•r—
Q£
Ol
S_
3
V)
r- O
re CL
,J£ Jj
S^z
JC •»->
O -g
"E°
4J 3
C CO
4-> O.
•
AI
*r—
t
OJ
•M
• plH
^^
re
c
o
t_
O)
CL
a
0
L.
oji^ s
co • re
(j O O
SB*
jc "o
=> •(->
re en u
CO
o
c
eu 01
•§.5
0 u
0 S
0 JC
r-~
<0 O >— I
•o
'o re re
J3 "*
t-^£ ^£
j=.£ .J2
C£ Q£
f
,C -n-
O en
C ••!-
jc a3 *"
Q. > OJ
0) C
0 3 T-
<(— :n t_
O T3
Ol • t-
co "o ja
>,CJ (0
c ••- co
«- 0
"o
0)
-u o o
O> -t-> -^£
t- 3 0
C C" >>
re I—
3 ra t-
'H.^'^
re -r- 3
0£.
CO
I
OJ co
O (1)
2 "if
0 0
O OJ •
CO CO
°.|k
-^ en
•r"~ «f^ "r"
•a t-
C 3 Q.
*
55
en
|
o
u
re
CO
en
re
t^
OJ
JQ
^3
1
en
O
r-»
CO
OO
•=J-
re
3
•"
Si
,
°
i °
re to
3
X3
en
a> c
•i- 3
> lo
£_
QJ
J3
r—
VO
t- CSJ
0
tl re
a.
.^ .J2
z.
aj
a3
"s -1^
Q.JC
3|— .
I- 3
5- CO -C
>, §12
o
Q- +J JD cr S.
re"o
^ §
•r— 0—
a:
^^
u
0
a.
E
3
U
3
5
u
3
™
j
—
3
«_ ^
-------
^^
«^J
OJ
3
C
•r™
•P
C
O
o
v
in
UJ
§
h-
CO
c.
CU
o
c
o
u
c.
0
CO
01
c:
•r-
c
<4~
£_
O
re
°5?
CO
•P
•r-
a
•p
CU
,»
re
oo
CU
t_
3
CO
1
LLJ
3
"tT
0
•P
3
C 1 1
o c- • c.
•i- O CU CU
-P .p M— r— to >»
C co -r- re >» E
-P t- -P r— t-
re oj o oj •"— to e
r— CO CU t- 18 -p (0
J3 1- 3 > •!- C
•P 01- CO « i— +->
C OJ O -i- CU
CU C_ O.-P E •<-
•i- o -P X O 5»
O <4- C CU C CO
•i- CU 3 C
4- -rj -P i— C O •'-
«t- 5 CU
CO (_ +j +J CU O
c: 0) **— o re t. fr~
>-5 Q. O «I S Q- >
^
^t—
co re
C CU
CU >
t- 0
O)
<«- CO
. cu oo
col t- •
I— 1 CO O
Eo| f— 10
GO CT> CO
**• to
•* •> s
co co
CM ,-H Ji
CO
5
c
o
-[-<
re
o
O CO C
CO 3 -i—
CO o
re -r- cu
> u
C £_ >
re D. t-
u cu
•r- -n CO
•t^ c
•r- t— -t-1 c:
to co •>- -P
r— OJ
5°E>
T3 0
C -P
co eo
i c >> c*
•r- i— (O
co .a c_
3 cu re o
0 0 g
•r- «r- 3 +J
> > co c
cu c_ cu cu
±88L*
•rel
Ol C71
C ^J t-H
re co-— '
a.
r- CU
CO C-
C 3
•? 0
+j a.
fO ^
r- CU
cu
re
0 3
C -P
u
c_ re
re
OJ O
ja -p
56
T3 CO
i cu re *P
J3 CO S C
3 o 1 CO
co a.' -r- -a 4J
<)- 4J x <*- o re
o o c cu M- •!- .—
• CU •*— ° 3 t-
t/1 OJ i * >% OJ t/) CU OJ
.pC- CU CO i— Ol C Q» >
C3"O-2"*-QC*r— £-
oj co co »r- re -P cu
SOO'T-cot-cCco
Q.^C- cocorecu^
t. x o -p o
3 cu t_ -a Q.-P >, re •
CO CUOJ C ^. P— O 40
re i— s +J co cu ^3 -P-P
aj re o re Q. oi-p- -P o
•P CUo COCUOJ**—
greg-5 fe,S£o S'cS
.p ai
i re i 01
IJI E 41 a)
c«j c co O
r— Ol Ol C
r— •(— .C -r- re ^J
re co jj co f- c
£- c -r- cr cu cu
CU O 3 O -P Ol
w > c c » **c
+J O CO >
C « C A oj
fc- co co L. ^o re cu ~&
ooj » ja jr aj
col i— cu cu c o re
re re •'— re cu Q,
f^* C^ 3 3 ** t» 3 * [ * t/1
^-i en +j co -p re
csjco^;^^ c OL^^C o co
•r- -i- 0 CU .1- O r— S
Q£ as ac
15 -r- 0 £-
t- ~o ~a cu c Q- cu
OJCUCU9 > 3 X .p
ccot-co re4jCJre
55 re re c -c c. cu
oicucxrec>. ocuc_
£_ Et.oOQ.OlOl
c o o co ^ a. c
•r— C O "^ CO S O "^ 4^
•f- cu o. t- re
co c > 3 co -a o
c c cu o c cu cu
t_ 3 re 01 *- re £
QJCU Cjacuccuo •
+J >CO*J 3-P-r- 13) -P OO
curei-cucocu.pct^i—
>^:co-r- >co gco
> « oj t- 8)
E -P c>« 1 ^. S O CU 1*~
re o o G- cu re L> **— co o
c c o > c cu
4J ^j£CZCUXJi.-P >J=+JCOt_
O C >>
"cu o
1 CO T- 4->
c t_ => co
o = ••-
c. co «jc
O CO
T3 0.,— r-
c x re re .
re cu o o cu
•r- •*— r—
r— I— • E "O | >>
re re
o o re cu "K
c5 o > co re
're oo
Ol Ol
C -P ^H
re co — •
-------
CO
»—i
X
a
LU
co oo
*-.
a>
u
o
o
o
•r-3
ro
01
o
ro
c
o
00
O)
a
X
LU
CO
o
O CO « O X (- l_ CO
I CO-OT-+J I £- O I O I T3 rorO
C|p— 4)t-O^'-OC'^(*-s OD4->flJ -r-
OdJ^fOo roOJOr— «^ £Z C • Q} t/v O rO o O •* 3E • *r~ rjj o o
o-c:o>> •"" ot_+j~~».(/)1»-.s t^-rjroiia).,-
_Or—+j-a^ <"^ ^ "! 35 c +j t. jj
a> ai
c o ^ ^
o > qj - - - - - ' - - -- - =- ° *" ° +*
Q.I— c
C.O£3s i— -r-fOc
' t. C Sj ro O O 4J
3 'r- p— 4J •?— C QJ
O CD ro C O> O
Ul
us
10 CO
^ ^^
ss
J= t-
Q. Q)
e "*-
5, oo
^o co
.-I <*>
LO
o
•
o
V
a.
oo
^!
CO
s
•
o
V
a.
o
i
CO
V)
(/)
£
I/)
o
c
O)
Q.
O
V
Q.
"8
o
o
I
o
CO
5
«3 r- 3C
S I -C Z
•O J= -O M- <»-
ro CL O 3C o
4J (O - o
: O re
°-% *2 S
t- o-r
3 ro O
c/) c
OL
O
OJ
<*- O • <4-
O 31 •—» <" -i—
I _l - C E rb O
»« c a; c o> o >>r—
JS IS'^^i: § o
* -*-> «h/
>, c o
>> I T3 -C. U> -rj > CO i
O C +J
O HI C O >, «4-> _
o> •>- i— i— ^ -i- a»ooo"os»o>>
^ri_jr>iUL^e:a.»-t_ a_ -o flJ'r- co^i"O
-C o ^ O O.3 4->-0(_CroUr—
-------
^"
3
c
•£>
o
u
vo
LU
C-J
CQ
^C
h-
VI
C
c.
4)
O
c
o
o
o
«r^
S
vi
Ol
c
T3
£
c_
o
•<->
*
VI
4J>
c
•r—
o
CL
-M
c
4)
•*"•
A
0)
c_
3
in
O
CL
>^
LU
in
*i^
o
JC
i
(I)
t_
(O
VI
0)
4J>
ra
0
•^
I|K
•P-
4«i
^
41
U
J=
ro
nj
Cj
ra
e
o
JC
Q.
>j
I/I
C
•r-
01
-o
o
3C
i
.c
•r™
2
4)
VI
ra
4)
c_
o
c
I/I
en
Ol
•r-
c
ra
**•?
<«-
o
>)
o
c
CO
3
CT
4)
t-
LU
t_
O
ro
U
i
ro
C
•r-
I3E
**—
+j
41 0)
^•-o
J3
nj r—
•r- lO
r- C
> =
(- 0
a> o
> o
j=
•+-*
ro
TJ
a:
"^
r*^
'"*•
VI
• O
Q.'"-
3 +J
1 ro
3 >
o a>
^ "cu
o
c
i*_ ra
o o
JC M-
• \ ff~
01 =
C 01
41 •!—
r- «0
TJ
41
in ai
o Ja
Q.
LU +J
•o
• 0>
c c
o ••-
•I- 1-
4J 4)
ro T3
Q.
3 OJ
O *-
o flj
o *
CVI
co
d.
•a
0)
N
• fH>
O l~
c ra
c
41 O
t_ t/l
ra t-
4)
OJ Q.
f_
aii—
J= ro1
h— 3
•a
•r-
C ••"
o -a
•i- C
•M *f~
E
VI <*-
,|J
c m
25
IH
r*.
fO
u
t_ ra
o vi
i]- .
in ^~
o -a
c.
c o
01
> T3
ai -i-
o
Q
s. o
•r™
4JI
• ai
VI U
41 ra
t- >>
3 X
VI O
o c
CL 41
X JC
0) CL
^ • .»
in
VI
3
ra
u
{_
41
+J
O
ur>
Ol IO
c
E en
ra ra
4) 0)
> "g
at =
VI
° 0)
41 E
«» *
aj'c'
c c
o ra
(U C
« a)
•r- 4J
Ol
41 C
(_ •!-
ai a>
S J3
VI VI
r— "
t_ -u
o c
JC TJ
o •<-
• •
^_
ra
c.
C CM
41 CM
Ot •
t— 1
C
•r- II
c" m
•i— «r-
ro •
LL.
VI.
•P" O
^, >f^
3 0
S •£
C 1)
'vi c
C C
0 £
10 >
'lO ••!-
M- ^*<4-
0 t! 0
O
ro +J re
c. ia r—
r^
(- J3 t.
O O
' > • - . • .
t_
C ro I •«
E JCr— 01 J= lO II
ro -i— t- +j at •>- vo cu
c > o it ra
I- JC 4> •!— ra -a a)
riJ .$J> Ol 4»* CU VI c.
Ol *r- O C 4J> CU fj
C »>CU CU ro Oi+j*r- ra
•i- t. t- •*->V}OjC
C c VI ^.j ^^ Q) QJ j%
O 3 ra r^" ra c» VI
Eo4)3Eoc:
-------
"-o
CU
"I
o
o
•*_x
.
*o
LU
03
t—
CO
C
^_
CU
o
0
o
C-
o
•r- 5
2
CO
en
c
•r—
-a
c
•r-
it-
(_
o
•1—3
£
CO
+"*
£Z
•r-
O
Q.
C
OJ
•r~
r—
*s
v)
CU
L.
3
(/)
0
a.
X
LU
CO
s^_^
t_
O
.c
3
CO
0)
+0
O
-r"
c_
CU
o
^
<8
CU
o
CO
CU
CO
to
o
18
E
o
JZ
a.
. p£
5)
CO
"c
'C
01
T3
O
31
1
i
ai
c
•p~
i-
T3
CU
t-
cu
3s
•o
CU
CO
o
a.
.2
CU
+J
CO
«r-
CU
E
CO
•" 1 -O
•r- (o i a) >, -Q ra -C
t- O E £_ '"~ •!- -r- ^i >,
CO QJ3CU-|JO4->COjQ
D cu ••- >t--o JS+JCUQ.-!-!!!.
fe£^c+JS^«j^o «-t->co-cja t_ xi ex cua^+j
f r*^ >rM {Q Q 3 ^rf t^> P" "^ -^
fO rtj ^> C E CO CU O «-^ CO re
__^^^. ,^^
Lf> Lf> Ln LO
o o o o
O C3 CD O
Or— 1 l~~ CM VVVV
un o CM 1-1 a. a. a. a.
LO rH CO CO • ' — ' '*~^'
M M A
«-« «3" rH CO "P CTl CM
«*• CM rH CO
1 1 1 1 ....
r-l r- 4 rH t— t
CO fO i- t- CU
§E « CU !/) 1 1 1 1
O O O 18
r- jrcrc o >gO(J C. J^ OOOO
ai +Jo o: >>EOO
i— CO 03 (O CM E ^>
Q.- +J g | r- CU JK
•f C CO S CU 4J O
+-> -r- O 4J •• 01 r-
O £- C ,_ -* I- tO
I C U. S T3
c o o
S " f
£
I CO t. T3 1 T3
cn i 18 cu cu i E T-
en O > -o
CU ~^ &• *^J CU CO CU 9 *r~ C CU r^ n 9
'""" Cl."'"" £H ^^ r^ C flj W) &_ CO i ^ *r— Qi ^^ ^i j« (rt
.C'r- OCUi— -«J E- CJ1 ^ O OJ qi e f g m m t •»-
-*->ECn"r™O'r"-i^Orj>(8 T3 C+J ->r"OE
•r- O^3CE*8W3OO*r- c3 CUCE^ICU
3A^^-<8 O CO BO O^-*4— r— •!— CU -M -^
QJCU'O -C*(— •On3&--OCMO O{8CU-*->"OCU r—
4-» -^ i— "O O 5 i— O'>— CU u CU CU i— O >~ CU
ns o o c: BS o " JE cooco s-.t_3:£ ^-i-»
o-^ oo ozso ••— -f-J — ' c c:
CO ^JCIOIB*J4J t-r— -r- e— .C CO c O)O
«/>Cn(jO -i— o O CU3I «3cnt- «3O
(8 C 3 "r"" O 2 3 . C (/) CO CU O 3 CU ** *8 Q r—
•^" "0+° »O-g"OCUOo * t-'i-oOfOE-C
(8'8C-3cOCOt_t-OtOx * J=O^:l8<5.Er— O
S r— Q.T3 3rtSQ.Q. O i— O. n3 C
C
1 CU
to J3 -A CU
<- o «- 3:
eu >8 g>-j •!- jj
+j ai •»-> oj
•r- >, E <— C «r J3
r- O *8 O CU .
3 o JE t- o *J_LJ tj o oj c ™o
4J O •*- O "O 'I— C
OJ t_ O t3
CU *" O5 O> d. t, VO
Of**- C
;
»
3
)
3
3
o
o
}
3
j
~
o
o
ta_^
-------
x .-.
^-*»
-o
OJ
•r"
o
O
,
iO
UJ
_J
S
..._;
VI
c
£_
Ol
u
c
o
(J
o
•>-3
s
1/3
en
c
•^
c:
•r—
L.
o
•O
U)
•r"
R
+f
Ol
to
to
0)
t_
I/)
0
CL
x
UJ
C/l
^7**
o
J=
4J
3
C 1 QJ
II 0 " JD -C
cotu -r- « >> 3 -tJ o>
CO CL C— -(-> QJ ^ CO O L. "O
] * X tO in O5 ** £ CU OJ
U) Q) OJCjr "0 3 •"—
_a • oj I*~*
u> O O) cu i- « co Q. c c CL
5^55SS£5«5i i
n o •— » co -si- vo o)
i— 1 CM in O <- < i— 1 3
ii i i i -a
i_ -i—
fO CO QJ ^ ^ 'O O
s e cj. e e e -r-
jj M £ ^ .i-e
CL CL " — • CL Q. CL fl)
i.!. 1.1. i.-^
^> ^> Lf> >) >5 >»**-*•
r— i— O i— • r-- r—
O
v> en * t/i to t/i
tf) • . i-H CO- -
(U C C *£ C C C
yi *i— «j— ( • -CO *i^ *r~* *r~*
ro ^ ^: co -r- i*. j* a
•a -a c "O ~o "a
O O O, ' _Q O O
a: 3= ^ a: a: a:
C •] „ C C
£ c "<- c :s>
CL -i- o W) O o i—
to | S -i- .,- -i- 0) +J
OJ 1— 4J ° QJ +J '" "O O
£« Q (O (O QJ (A V) ro M '^^ (~
1 C S t_ C3>.r- 3 -r- 1= 0 J=
QJ O 3 — <" -t-3
+J -C O -^ J3 ••— -r-
OoCL 3 4-^(/)O)fc. S
"O S"O £•!— V)"OOJQ «
C jC >r~ i— V O -C «4JO
Oc-0'1- O O-r- =J.
•t— o c c JD m e s T-CO
<«-.i-T-aj*-u5^«to «g «
C ro O> CTJ= J= O 1 IO g CO +J •
en ^-» *T3 OJ [^ c ^^ o *o c ^^
•i- OJ O t- t- 3 >> «•»- JC 01 O •
CO 1.3:14- O«C">CM O CL3 (JCM
OJ E
« s t- <" o i
(nOQJi &-c: co+Jt-W
Ul Q) T LU g 0 S 0)<*- 4-> C <»_ r-
fO"O^t* QJ4J 3OrO^ 3
C o CM • V QJ QJ ^ ^> QJ QJ O (/)
(O .,- |^(-<_l/>J=4->C-<<-T-CLra
t-r— CWwO !/i o Qi -r- •— C
•r-j^cONCCL QjQJro Ct-i^
r— CO *• X • -t- > O 'QjQJ
t_ to U t. C T.^-cinO-r-QJo
£= -M CLQJoc OJ C
U. Z CLCO (_ j3 CM -i- 4-> ii OO S +-> 4->
i— VO
CO CO
t- CS1
« -P i-l
o ai«— ' en
a: oU
,'-" • ' ••?••-•*:••:••.? -.-p
Q)
CO
CO
CL
CO
e~
•r-
X
o
o
14-
QJ
J=
C
O
"C V) T3
O >, O 1- "> OJ
. QJ -^. « QJ QJ 3
VO !_ > =••- '<-
E <0 0 *£
I ~U 1 •(-> -i— C
"O C JO O
00
« ••
c e
•^ *r~
-^ ^^
cn en
T3 T3
5 5
1 1
C C
o 6
3Z S*
-------
TJ
0)
3
C
+J
O
O
*•— -
• c
t_
Ol
0
c: '
o
u-
£_ '<
O
•r- 3
S
cn
c:
•r—
-o
c
•r—
O
co
C
•f—
§.
•M
C
0)
•r-
ro
CO
0)
t-
3
CO
O
Q.
£
O
-C
^o
5
CO (P (/) T~ •(- O
C 1 -O +J <<- r— Q.
••— o -t— c: -i— oj a; x
C! (O •*" > CO 3 o O
O t- U O i— O U Q. l/> OJ
,c in tn x => t- >>
O O C o '•- ••-•!- S <4- 03 O
C | — 0) O ••-
-Q 3'"- t-jQ CCO-r-JT
••"tOt-O'l^O^^ e
CO O O CO *p— t1^ GO -P ^)
<" CLr— ~a CO 4J O Ol CO C.
a)
•o t-
t_ 0 O
- — * - — * o o «i— **—
o LA o cu -O c:
O «T3 Ol-.,-
V O'f V3T3o»'-+J
a. vo a.co-'-cnJm
**— «^ ^i *^~ *••—•* O O *r™ ft) t_
~, ^-'•P °-'« *- >>+*
CO CO w ^-4 >^ j^ ^3 yj
*l~* C*") OO ^ trt f""** QJ t™ TJ LO «i^
(O soj3(^-i-sco>st_Qj
E ZC- E 0),— t_ cn •!-
•r- c <3 3 -r- Ol -r- O t-
> OJ ^ 5_ *Z S CO
i— J3 3 -r- • T3°<1J
CD O O > V) C_lf.J!£
-C (0 flu +-> OJ "O -Q p
(Q QJ (/} O V) t^
<->3Q_ >,'i— CO .r- C
•r- 3 ,_?J3 0) e -I-
t- —1 O .Q t. O
•r- a: t. .,- a) x +j T3
czcncoJr dice s
•r- « Q. f> ? O 3
•f™ o ^* c Q)
-l-> C O O •<-
Q-JS -c -a t_
3 a. u a!
-
-------
A
OO
jE;
X
o
o
•H
o
z
UJ
CO
J— 4
a
0
LU
r~*
^C •
zz
H~4
g
-J Q
3= 0
O O
j ^7
2 **£.
r--
O CO*
00 CM*
UJ
3 ij
^3 i">*
' — <;
00 _l
tZ
^H |.IB
£2
g
z:
LL)
H~*
^C
0
1—4
g
Q.
O
C£.
a.
.
f*^
UJ
_i
CQ
Q
C
CO
•r-
0)
t.
1
CO
0)
CO
re
0
CTl
2.
OO
OO
CO
CO
o
eo
J^»
c:
u
•r—
«i—
C
O1
•i—
CO
«C
CO
0)
•o
O
•r~
C.
CO
JE
>3
O
c
CO
°~
g
re
i
i— CO
4j
O O
re j->
<<— O)
o u
a> co
o o
§ 5"
T3 CO
f — «
co
CO
CO
re
u
o
t— i
CO
CO
re
CO
•r*
-a
CO
e—
-*:
en
•a
^
c!
2
C
O
CM C
CO
>>J=
fj- Q.
CO °
!c °
• — • o
CM
CO
i— 1
1
2
^
CO
>^
x
o
c
Q.
CO
CO
(J
CO
.|_>
C
o
•r™
M-
•i»
C
ai
•r"
CO
««
*
>.
•4-*
CO
CO
C.
O
t.
Q
•~»
C
re
CO
CO
«^-
o
17
CO
CO
0
co
CO
t-H
""•'
g
f
a.
i,
t/>
M
C
•r»
"en
T3
o
I
^
CO
CO
CO
T3
co
~c
•r"
cn
•a
3^
it-
IS
c
CO
+0
O
<<- 3
•"§
CO (-
<_
3
4->
U
•r-
£_
M
U
a.
T3
o
§
•a
re
,_
Q
C
.C
a.
0
t.
o
o
1
£
ai
£
^
^i
T3
QJ
^*
~o
•p- O
^r
ex"S
i_
CO
ex
(O
a.
c
•r-
•
CO
•p
5
c
Q
-)->
Ol
•r~
J=
CO
3
10
CO
c.
CO
Jrf
i.
n
§
+J
C
re
•u
«a
^™
o
s
(_
3
«8
-r-
O
C
o.
o
A(
CO
(-
1
cu
«J
VI
+J
(O
c
,®
. %
£_
o
Q.
cl
cc.
^
^>^
IT)
O
•
0
V
Q.
^««
CM
CO
•
rH
II
00
OO
ex
3
Q.
CO
Q.
re
^^
•
•r- S_
re^
E L.
o a
f> r —
c ,_
•— E
62
CO
CO
re ^
C» dj
c_
•M^
•r" c
r— *r>
re
4^ QJ
o re
EJ=
cu
«?•
2.
a.
,_^
en
o
o
3.
CXI
,_^
II
CO
CO
re
CO
•r—
T3
CO
C
•r—
C7>
-o
£
.J;
CO
CO
CO •
CO
c
^^
s.
ae
E
CO
CO
•
t— 1
11
g
_c
Q.
5
Ul
c
•r™
O>
1
c
S
o i
I— I
CO ^ C o •
u >«- -J- 1_
«— t- CO ^J
CO co^-.^
t_ 10 to
-(«.;i—
j=T-'-(D4-)t_t-T-ait_r-i
'OO
o
•a
i
-------
!/)
C
CU
o
o
o
&-
O
•"-3
£
to
en
c
>r~
c
*^*
(_
O
•1-3
rg
2£
•o
CU
3
C
C
0
u
.
I**** ro -C i .r- m
' CU 3 -M -r- CU 4-> 4J
_ 4J -d 4-> -a -t-> ro <=
o o o £- c cu d) 3. o >> cu
>•— — - S +J -i— C
3 x o) iB t_ ^_ .;:
t- 0 O C OJ ?•(-> CX^ • V 4J^-. . 0)
a. r-< o o
> . Q. IO ro
£ 00 T- 0 — ^0) 0
<-> « c_) o.r~> "S >i •
-i- CO
w> if> 'in j: <~*
'*- o
t/) (/j c ,j i * [, > (/) ^
•^- O^ *r" 2^ QJ QJ «p- Q.
o; — a:to>in ce^ — •
* > a> i
-^43 1o <" o <3
O o 4) > 4-> O C
L. <1) t_ d) 4J 40
4J C fl) r- OJ W) Ol
W C 2 O) W) c -i-
o ° t. ^ w >
•)-> ° O
OJ c c c "i a; c
3 •
^3 O O fl3 O t/l
°* I CU
EE d) jj
C "° « S L " fO ft- »p- -g*
ai o 4-> Q-in oja; o4->
*3 *^« p- fm l\
*-* <"^ ^ w ^J
*•"• *O O *^- * t/i t» c/>
S *o o c* ra ra t/)
>OZ-'»->3 WOJ
-a
c
rt3 *^^
c §Ico
fO ^ CTl
I°" fi3
(/J • 1
CtU Ol O tO 'r- •!— > £_
VJ <1J C ro ^— OJ i/i f-
-i— o "O CU t- C -r- O4->-r^-a
S-c-tl'o c g o i— f—
-------
^*
QJ
C
+J
o
u
"~*
•
r^
a
S
oncerns
u
£_
O
'*->
£
to
o>
c
•r—
TO
r*
•r™
C_
O
3
j-
to
4J
c
o
CL
4«i
c
0)
"^
5
QJ
L.
to
i.
X
LU
"to"
c_
O
4j
=3
1
•r- ,— O) c to
> TO JC | .r- QJ
GJ 3 4-» *r- QJ 4J +*>
O O O S- C 3 O >, Ol
to u_ QJ rf • .(— . %' r-
•r- 0 «- >>v^^-* E •!-> -T- C
3 X HI TO t- r- •!-
t- o o c Q> c +•> Q.+J at
c Q. aj -a
f 0) X f r- C. .r- t_O«
H--OQJQ.OO> (1. E -C
1
«- «- S =>
QJ • ni m i—
O OS O C o
<= g c -g 10
*•> CO O T- to IO
^* to •
QJ »— » QJ i s; (*i
3 i-H 3 C
to o to o -~~ u
tO • (/) C i-H
•*- 0 ." o S-
4-> V [ i «-^ • QJ
Q. t-H O O
> • a. to to
•f™ CO •!•• ^3 «-«x QJ cj
4-> P«^ ^_> V f™
QJ «?J* QJ ^~^OO E QJ ctl
£= c • c 5
CM c vo LO QJ -a £
O O --I 4J -r- CO
O — > O • •••>- -^
•*- 0) <4- C S 1- i-H
o to o H to en — .
QJ -a
•> > QJ 1
•^ 4^ (O to O TO
O o QJ > •*-> O C
t- QJ £- QJ +-> +->
+J c gj r- Q) to QJ
to c S QJ to c •!-
O 3 10 >
out- >,x: {_ i
4-> c
^3 CO 03 CO */) > •
"O O o m O (/>
jw >! aj *JI « S
aSsfis&S
• •a +-> i o»
E QJ en c 4-> +j
TO -a c Q» QJ -i-
C •!- tO (O «- •!- JC
Q» O +•> OJ .c
O +j 10 QJ o QJo •
'> o ^ -J^ '> §~ > to QJ
COO.OOCO ocS E
•a
c:
TO&1
a>,2 ,_|
o o **^
^ 64
(/> • i
1 •!— C QJ -r- C i— QJ *f- tO
CO) Q)'^-'*-> t-
•i- Q. C. T3 C >> I O OJ QJ
tO OJ C TO r— QJ W>JC
C/) 1 CJ QJ d^ TO W ^ C TO XJ
ajdica. ciaQJtoQJo
QJ 5 i— ,_
QJ-C-r- 0 <»- C C = t- r- ^_
S 4-> O CO'1-^OQ.TOTO
-------
REFERENCES
Axelson, 0. (1986, Jan. 21) Rebuttals of the final report on cancer by
the Royal Commission on the use and effects of chemical agents on
Australian personnel in Vietnam. Department of Occupational Medicine,
University Hospital, Universitetet Linkoping, Linkoping, Sweden.
Axelson, 0.; Sundell, L.; Anderson, K.; Edling, C.; Hogstedt, C.; Kling, H.
(1980) Herbicide exposure and tumor mortality: an updated epidemiologic
investigation of Swedish railroad workers. Scand. J. Work Environ. Health
6:73-79.
Australian Royal Commission. (1985) Cancer, Vol. 4. In: Final report—Royal
Commission on the use and effects of chemical agents on Australian personnel
in Vietnam. Canberra, Australia: Australian Government Publishing Service,
pp. VIII-166.
Balarajan, R.; Acheson, E.D. (1984) Soft tissue sarcomas in agriculture and
forestry workers. J. Epidemiol. Community Health 38:113-116,
Bishop, C.M.; Jones, A.H. (1981) Non-Hodgkin's lymphoma of the scalp in workers
exposed to dioxins. The Lancet 2(8242):369.
Buesching, D.P.; Wollstadt, L. (1984) Cancer mortality among farmers.
J. Natl. Cancer Inst. 72(3):503.
Burmeister, L.F.; Everett, G.D.; Van Lier, S.F.; Isacson, P. (1983) Selected
cancer mortality and farm practices in Iowa. Am. J. Epidemiol. 118(1):
72-77.
Cantor, K.P. (1982) Farming and mortality from non-Hodgkin's lymphoma: a
case-control study. Int. J. Cancer 29:239-247.
Cook, R.R. (1981) Dioxin, chloracne and soft tissue sarcoma. The Lancet
1:618-619.
Cook, R.R.; Cartmill, J.B. (1984) Dioxin: comparing apples and oranges.
Chem. Tech. :534-537.
Cook, R.R.; Bond. G.G.; 01 sen, R.A.; Ott9 M.G.; Gondek, M.R. (1986) Evaluation
of the mortality experience of workers exposed to the chlorinated dioxins.
Chemosphere 15(9-12):1769-1776.
Eriksson, M.; Hardell, L.; 0;Berg, N.; Moller, T.; Axelson, 0. (1979) Case-
control study on malignant mesenchymal tumors of the soft tissue and
exposure to chemical substances. Lakartidningen 76:3872-3875.
(translation)
Eriksson, M.; Hardell, L.; O'Berg, N.; Moller, T.; Axelson, 0. (1981) Soft
tissue sarcomas and exposure to chemical substances: a case-referent
study. Br. J. Ind. Med. 38:27-33.
65
-------
Fett, M.J.; Dunn, M.; Adena, M.A.; O'Toole, B.I.; Forcier, L. (1984)
Australian veterans health studies: the mortality report. Part I: A
retrospective cohort study of mortality among Australian national
servicemen of the Vietnam conflict era, and an executive summary of the
mortality report. Canberra, Australia: Australian Government Publishing
Service.
Fingerhut, M.; Halperin, W.; Honchar, P.; Smith, A.B.; Groth, D.; Russell, W.O.
(1984) An evaluation of reports of dioxin exposure and soft tissue sarcoma
pathology in U.S. chemical workers. Scan. J. Work. Environ. Health
10:299-303.
Greenwald, P.; Kovasznay, B.; Collins, D.: Therriault, G. (1984) Sarcomas of
soft tissues after Vietnam service. J. Natl. Cancer Inst. 73(5):1107-1109.
Hardell, L. (1981) Relation of soft tissue sarcoma, malignant lymphoma and
colon cancer to phenoxy acids, chlorophenols and other agents. Scand. J.
Work Environ. Health 7:119-130.
Hardell, L.; Eriksson, M. (1986) Soft tissue sarcoma and exposure to dioxins.,
The Lancet p. 868.
Hardell, L.; Sandstrom, A. (1979) Case-control study: soft tissue sarcomas
and exposure to phenoxyacetic acids or chlorophenols. Br. J. Cancer
39:711-717.
Hardell, L., Eriksson, M.; Lenner, P.; Lundgren, E. (1981) Malignant lymphoma
and exposure to chemicals, especially organic solvents, chlorophenols and
phenoxy acids: a case-control study. Br. J. Cancer 43:169-176.
Hoar, S.K.; Blair, A.; Holmes, F.F.; Boysen, C.D.; Robel, R.J.; Hoover, R.;
Fraumeni, J.F., Jr. (1986) Agricultural herbicide use and risk of lymphoma
and soft tissue sarcoma. J. Am. Med. Assoc. 256(9):1141-1147.
Hueper, W.C.; Conway, W.D. (1964) Chemical carcinogenesis and cancers.
Springfield, MA: C.A. Thomas.
Johnson, F.E.; Kugler, M.A.; Brown, S.M. (1981) Soft tissue sarcomas and
chlorinated phenols. The Lancet 2(8236):40.
Kang, H.K.; Weatherbee, L.; Breslin, P.P.; Lee, Y.; Shepard, B.M. (1986)
Soft tissue sarcomas and military service in Vietnam: a case comparison
group analysis of hospital patients. J. Occup. Med. 28(12):1215-1218.
Kang, H.K.; Enziger, F.; Breslin, P.; Feil, M.; Lee, Y.; Shepard, B. (1987)
Soft tissue sarcoma and military service in Vietnam: a case-control study,,
J. Natl. Cancer Inst. 79(4)693-699.
Karon, J.M.; Flanders, W.D.; Devine, O.J.; Needham, L.L.; Patterson, D.G. Jr.
(1987) Serum 2,3,7,8-tetrachlorodibenzo-p-dioxin levels in U.S. Army Vietnam
and non-Vietnam veterans. Presented at the 7th International Symposium
on Chlorinated Dioxins and Related Compounds; Oct. 4-9; Univ. of Nevada,
Las Vegas, NV.
66
-------
Kogan, M.D.; Clapp, R.W. (1985) Mortality among Vietnam veterans in Massa-
chusetts, 1972-1983. Massachusetts Office of the Commissioner of Veterans
Services, Agent Orange Program, Massachusetts Department of Public,
Division of Health Statistics and Research, Boston, MA.
Lathrop, 6.D.; Wolfe, W.H.; Albanese, R.A.; Moynahan, P.M. (1984a) An
epidemiologic investigation of health effects in Air Force personnel
following exposure to herbicides: baseline morbidity study results. USAF
School of Aerospace Medicine, Brooks Air Force Base, Texas. 19 chapters.
Lathrop, 6.D.; Wolfe, W.H.; Albanese, R.A.; Moynahan, P.M. (1984b) An
epidemiologic investigation of health effects in Air Force personnel
following exposure to herbicides and associated dioxins. Presented at the
Fourth International Symposium on Chlorinated Dioxins and Related
Compounds; October 16-18; Ottawa, Ontario (abstract).
Lynge, E. (1985) A follow-up study of cancer incidence among workers in
manufacture of phenoxy herbicides in Denmark. Br. J. Cancer 52:259-270.
May, G. (1982) Tetrachlorodibenzodioxin: a survey of subjects ten years
after exposure. Br. J. Ind. Med. 39(2):128-135.
Merlo, F.; Puntoni, R. (1986) Soft tissue sarcomas, malignant lymphomas,
and 2,3,7,8-TCDD exposure in Seveso. The Lancet p. 1455.
Michigan Department of Public Health. (1983a, May 4) Evaluation of congenital
malformation rates for midland and other selected Michigan counties
compared nationally and statewide 1970-1981. Michigan Department of
Public Health.
Michigan Department of Public Health. (1983b, May 4) Evaluation of soft and
connective tissue cancer mortality rates for Midland and other selected
Michigan counties compared nationally and statewide. Michigan Department of
Public Health.
Mil ham, S., Jr. (1982) Herbicides, occupation, and cancer. The Lancet
pp. 1464-1465.
Moses, M.; Selikoff, I.J. (1981) Soft tissue sarcomas, phenoxy herbicides and
chlorinated phenols. The Lancet 1:(8234):1370.
Olsson, H.; Brandt, L. (1981) Non-Hodgkin1s lymphoma of the skin and occupa-
tional exposure to herbicides. The Lancet p. 579.
Ott, M.G.; Olson, R.A.; Cook, R.R.; Bond, G.G. (1987) Cohort mortality study
of chemical workers with potential exposure to the higher chlorinated
dioxins. J. Occup. Med. 29(5):422-429.
Pazderova-Vejlupkova, J.; Nemcova, M.; Pickova, J.; Jirasek, L.; Lukas, E.
(1981) The development and prognosis of chronic intoxication by tetra-
chlorodibenzo-p-dioxin in men. Arch. Environ. Health 36(1):5-11.
67
-------
Pirkle, J.L.; Wolff, W.H.; Patterson, D.6.; Needham, L.L.; Michael, J.E.; Miner,
J.C.; Peterson, M.R. (1987) Estimates of the half-life of 2,3,7,8-TCDD in
ranch hand veterans. Presented at the 7th International Symposium on
Chlorinated Dioxins and Related Compounds; Oct. 4-9; Univ. of Nevada,
Las Vegas, NV.
Puntoni, R.; Merlo, F.; Fini, A.; Meazza, L.; Santi, L. (1986) Soft tissue
sarcomas in Seveso. The Lancet p. 525.
Riihimaki, V.; Sisko, A.; Hernberg, S. (1982) Mortality of 2,4-dichloro-
phenoxyacetic acid and 2,4,5-trichlorophenoxyacetic acid herbicide
applicators in Finland. Scand. J. Work Environ. Health 8:37-42.
Riihimaki, V.; Asp, S.; Pukkala, E.; Hernberg, S. (1983) Mortality and
cancer morbidity among chlorinated phenoxy acid applicators in Finland.
Chemosphere 12(4/5):779-784.
Sarma, P.R.; Jacobs, J. (1981) Thoracic soft tissue sarcoma in Vietnam
veterans exposed to Agent Orange. The Lancet 306(18):1109.
Smith, A.H.; Pearce, N.E. (1986) Update on soft tissue sarcoma and phenoxy-
herbicides in New Zealand. Chemosphere 15(9-12):1795-1798.
Smith, A.H.; Waxweiler, R.J.; Tyroler, H.A. (1980) Epidemiologic investigation
of occupational carcinogenesis using a serially additive expected dose model
Am. J. Epidemiol. 112:787-797.
Smith, A.H.; Pearce, N.E.; Fisher, D.O.; Giles, H.J.; Teague, C.A.; Howard, J.K.
(1984) Soft tissue sarcoma and exposure to phenoxyherbicides and chloro-
phenols in New Zealand. J. Nat!. Cancer Inst. 73(5):1111-1117.
Thiess, A.M.; Frentzel-Beyme, R.; Link, R. (1982) Mortality study of persons
exposed to dioxin in a trichlorophenol-process accident that occurred in
the BASF AG on November 17, 1953. Am. J. Ind. Med. 3:179-189.
U.S. Environmental Protection Agency (1985). Health assessment document
for polychlorinated dibenzo-j3-dioxins. Office of Health and Environmental
Assessment. Washington, DC. EPA/600/8-84/014F. NTIS PB86-122546/AS.
Wiklund, K.; Holm, L.E. (1986) Soft tissue sarcoma risk in Swedish agricultural
and forestry workers. J. Nat!. Cancer Inst. 76(2):229-234.
Wiklund, K.; Dich, J.; Holm, L.E. (1987) Risk of malignant lymphoma in Swedish
pesticide appliers. Br. J. Cancer 56:505-508.
Wolfe, W.H.; Michalek, U.E.; Albanese, R.A.; Lathrop, G.D.; Moynahan, P.M.
(1984) An epidemiologic investigation of health effects in Air Force
personnel following exposure to herbicides: mortality update, 1984.
USAF School of Aerospace Medicine, Brooks Air Force Base, Texas.
Wolfe, W.H.; Lathrop, G.D.; Albanese, R.A.; Moynahan, P.M. (1985) An
epidemiologic investigation of health effects in Air Force personnel
following exposure to herbicides and associated dioxins. Chemosphere
14(6/7):707-716.
68
-------
Woods, J.S.; Polissar, L.; Severson, R.K.; Heuser, US.; Kulander, B.6. (1987)
Soft tissue sarcoma and non-Hodgkin's lymphoma in relation to phenoxyher-
bicide and chlorinated phenol exposure in Western Washington. J. Natl.
Cancer Inst. 78(5):899-910.
Young I.U; Racy, C.L.; Asire, A.J., eds. (1982) Surveillance, epidemiology and
end results: incidence and mortality data, 1973-77. Natl. Cancer Inst.
Monogr. 57:14, 32, 72, 84.
Zack., J.A.; Suskind, R.R. (1980) The mortality experience of workers exposed
to tetrachlorodibenzodioxin in a trichlorophenol process accident.
J. Occup. Med. 22(1):11-14.
69
-------
-------
February 1988
Review Draft
APPENDIX C
REPRODUCTIVE AND DEVELOPMENTAL TOXICITY OF 2,3,7,8-TCDD
G. L. Kimmel, Ph.D.
Reproductive Effects Assessment Group
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
-------
-------
EXECUTIVE SUMMARY
There is not sufficient evidence to link 2,3,7,8-tetrachlorodibenzo-
p_-dioxin (2,3,7,8-TCDD) to human reproductive or developmental toxicity;
however, it has been shown to be a reproductive and developmental toxicant in
animal studies. Among the effects that have been reported are reduced
fertility, litter size, postnatal survival, and offspring body weight, as well
as an increase in structural malformations. Effects on the male and female
gonads and on the female menstrual/estrous cycle have also been reported.
Reproductive and developmental effects have been observed in a variety of
species, indicating that the toxicity is not a species-specific event.
The studies on reproductive function and fertility remain the basis for
establishing the lowest effective (toxic) exposure level. It appears that a
0.01 ug/kg/day exposure is the lowest effect level that can be supported by the
data, although further analysis of the data may provide some support for a
lower effect level of 0.001 ug/kg/day. In addition, a detailed analysis of
studies in the subhuman primate may also provide support for a lower effect
level of 0.001 ug/kg/day. In relation to developmental toxicity, a large
number of studies in a variety of species has demonstrated that 2,3,7,8-TCDD is
a developmental toxicant. Collectively, these studies indicate that long-term,
low-dose exposure is of concern relative to the potential for altering
reproductive function and fertility. The results also demonstrate that acute
and short-term exposures are effective in causing altered development, and
therefore, should also be of concern.
-------
INTRODUCTION
2,3,7,8-TCDD has been shown to be a reproductive and developmental
toxicant in animal studies. Among the effects that have been reported are
reduced fertility, litter size, postnatal survival, and offspring body weight,
as well as an increase in structural malformations. 2,3,7,8-TCDD also affects
the male and female reproductive systems. Gonadal dysfunction has been
demonstrated in both sexes, and alterations of normal reproductive cycles have
been reported in the female. Although there have been accidental exposures of
humans to mixtures containing 2,3,7,8-TCDD, there is not sufficient evidence
from the case reports and epidemiologic studies that have been carried out to
date to link 2,3,7,8-TCDD to human reproductive or developmental toxicity (U.S..
EPA, 1986a). Much of the information on human exposure is covered in Appendix
D.
In line with the original request for the development of this review, this
appendix covers the effects of 2,3,7,8-TCDD on the integrity of the
reproductive system and fertility and on prenatal and early postnatal
development. The appendix is not inclusive of all studies on these effects.
Rather, it focuses on the key studies and issues that may go into an overall
risk characterization. The present reviewer has tried to present a balanced
view of the studies and the uncertainties inherent in the data and the
analysis. Studies on other congeners of polychlorinated dibenzo-p_-dioxins or
of 2,3,7,8-TCDD as a mixture or contaminant of other agents are not included,
except as support where appropriate. A more comprehensive review of
2,3,7,8-TCDD's toxicity and its relation to risk characterization can be found
in the Health Assessment Document (HAD) for Polychlorinated Dibenzo-p_-Dioxins
-------
(U.S. EPA, 1985).
ANIMAL STUDIES
REPRODUCTIVE FUNCTION/FERTILITY-
The study by Murray et'al. (1979) continues to be the guidepost for
setting standards of exposure relative to reproduction. The study employs a
multigeneration approach and examines the exposure of male and female rats over
three generations to relatively low levels of 2,3,7,8-TCDD (0, 0.001, 0.01, and
0.1 ug 2,3,7,8-TCDD/kg body weight/day). The analysis of the data is made
somewhat difficult by considerable variation in the fertility index in both
control and exposed groups. In addition, the number of impregnated animals in
the exposed groups was lower than desirable (Palmer, 1981). However, there
were effects that cannot be automatically associated with the variation in the
fertility index, including an increased time between first cohabitation and
delivery, a decrease in litter size, a decrease in the gestational survival
index, and a decrease in postnatal body weight. Specifically, Murray et al.
reported statistically significant changes in several of the measured
parameters, and these are outlined in Table 1.
While there is no dispute over the reproductive toxicity seen in this
study, there is some disagreement over the appropriate effect levels. Murray
et al. (1979) indicated that the lowest statistically significant adverse
effect was observed at 0.01 ug/kg/day and that a no-effect level could be
established at 0.001 ug/kg/day. In a reanalysis of this study, however, Nisbet
and Paxton (1982) argued that the analysis of Murray et al. (1979) was limited
by the statistical approach used. Nisbet and Paxton applied a different
-------
TABLE 1. SUMMARY OF EFFECTS OF 2,3,7,8-TCDD ON REPRODUCTION
Parameter
Fertility
Litter size
Gestation Survival
Postnatal survival
Postnatal body
weight
Generation
fo
fo
fl
fz
fla
fib
f3
f
fib
fs
fla
fib
?3
fla
fib
•fz
ug TCDD/kg/daya
0.001 0.001
_..
. — —
dec
— dec
• — —
— . — .
— dec
dec
._.
-__ ..
dec dec
— dec
dec dec
inc —
— dec
— ' —
— —
— , —
— dec
— dec
0.1
dec
dec
dec
dec
decb
c
b
c
b
c
a (—), unaffected; dec, decreased; inc, increased.
b No liveborn offspring.
c One litter only.
SOURCE: Murray et al., 1979.
-------
statistical approach which included the pooling of data across all generations,
and their reanalysis indicated that 0.001 ug/kg/day (the lowest dose used) was
an effect level and that a no-effect level could not be set. The authors of the
HAD for Polychlorinated Dibenzo-fi-Dioxins (U.S. EPA, 1985) accepted this
argument and used the Nisbet and Paxton reanalysis to establish a lowest
observed adverse effect level (LOAEL) of 0.001 ug/kg/day. However, the HAD
(U.S. EPA, 1985) also noted that the FIFRA Scientific Advisory Panel (SAP) did
not feel that the effects were consistent enough at 0.001 ug/kg/day and that
this would have to be considered a no observed effect level (NOEL). Since this
latter decision by the SAP was made before the Nisbet and Paxton reanalysis, it
is not possible to know if the decision would have been different in light of
the reanalysis. However, the present reviewer feels that effect levels should
not be set on the basis of the Nisbet and Paxton reanalysis. While it-appears
that Nisbet and Paxton's approach for increasing the limited statistical power
of the Murray et al. (1979) study is appropriate statistically, it is difficult
to see the biological rationale for pooling the data. Litters from different
generations (or from subsequent matings within a generation) are not the same.
They have different histories of exposure and each is ti'ed to the effect of the
agent on its parental generation. Thus, as a general rule, pooling of data
from different generations would seem biologically inappropriate. Unless some
specific exception can be identified, it is not clear how pooling can be
biologically justified in this case.
A limited review of a report by Murray et al. (1978), which served as
Exhibit 77 in a 1980 EPA hearing, has raised some questions relative to the
offspring survival. The Murray et al. (1979) paper included the standard
parameters of the Gestation Survival Index and Postnatal Survival Index as
-------
measures of offspring survival (Table 1). These parameters showed significant
changes, but not always in a dose-related fashion. This may be because
offspring viability is examined during discrete periods of offspring
development and the investigators do not report viability over the entire early
postnatal period. Additional data from the 1978 report by Murray et al. has
been summarily reviewed on the basis of overall offspring survival, i.e., not
separating the Gestation Survival Index and Postnatal Survival Index. There
appears to be a general pattern of decreased survival even at 0.001 ug/kg/day,
if one assumes a survival rate of control offspring of 90%. Appropriate
analytical techniques would have to be applied to confirm this. Two points
should be raised regarding the data. The first is that the number of offspring
used in the calculations of Murray et al. (1978) varied considerably among the
control and two exposure groups. How this could affect the parameters is not
entirely clear to this reviewer, but Bailar (1981) spoke to similar issues in
his testimony and noted that he found the data suggestive of an effect at the
0.001 ug/kg/day level. The second point is the survival of the control
population of offspring. In both the fjjj and the f3 litters, survival of the
controls by postnatal day 21 ranged from 70% to 80%. Although viability varies
within any laboratory animal population, this figure seems low and may account
for there not being an established decrease in offspring viability in these two
groups at 0.001 ug/kg/day. A more detailed analysis of this data base may
provide a clearer indication of the potential for decreased postnatal survival
at the 0.001 level.
In addition to the data on offspring survival, the Murray et al. (1978)
report summarizes their observations on renal pathology. When all observed
effects on the kidney (i.e., "slightly dilated" and "dilated") are combined,
-------
there appears to be an increase at the 0.001 and 0.01 ug/kg/day in the fja and
f}k litters. As has been pointed out, it is not entirely appropriate to
combine these two end points, since slight dilation may be due to delayed
development which may be transient in nature. Nevertheless, the kidney is a
recognized end organ for 2,3,7,8-TCDD effects, and the findings of Moore et al.
(1973) in the mouse indicate that the continuous exposure that is found in a
three-generation study may be more likely to lead to the most obvious effects
on kidney development. Research in this particular area is continuing. Abbott
et al. (1987a, b) recently reported that the kidney alterations that occur in
the mouse following a single, prenatal 12 ug/kg dose of 2,3,7,8-TCDD are
consistent with true hydronephrosis.
Allen and his colleagues examined 2,3,7,8-TCDD effects on reproduction in
the monkey (Allen et al., 1977; Allen et al., 1979; Barsotti et al., 1979;
Schantz et al., 1979). In a series of studies, female rhesus monkeys were fed
50 or 500 ppt 2,3,7,8-TCDD for up to 9 months. Menstrual cycles and serum
steroid levels were examined. Following 7 months of exposure, the females were
bred. Females exposed to 500 ppt showed obvious clinical signs of 2,3,7,8-TCDD
toxicity and lost weight throughout the study. Five of the eight monkeys died
within one year after exposure was initiated. In a summary of the reproductive
function and fertility of these animals, Allen et al. (1979) reported that
although the menstrual cycle and menstruation were normal, there was a decrease
in serum estradiol and progesterone in five of eight monkeys. Only three of
the animals conceived, and only one was able to carry the pregnancy to term.
Females exposed to 50 ppt 2,3,7,8-TCDD in the diet (Schantz et al., 1979)
showed normal menstrual cycles and serum estradiol and progesterone through 6
months of exposure. When they were bred at. 7 months, four of eight females did
-------
not conceive and two of four that did could not carry the pregnancies to term.
Only two conceptions resulted in normal births.
The results of this series of studies could potentially support a lower
LOAEL than that reported by Murray et al. (1979) in the rat (i.e., 0.01
ug/kg/day). The high dose (500 ppt) resulted in considerable maternal toxicity
and reproductive dysfunction, while a comparable exposure level (0.01
ug/kg/day) in the rat did not produce any significant clinical signs in the
parental generation. This could indicate that the rhesus monkey is more
sensitive to 2,3,7,8-TCDD when exposure occurs over long periods and when
reproductive parameters are the critical end points. The low dose (50 ppt) was
reported as resulting in specific reproductive dysfunction in the absence of
maternal toxicity. Since this exposure level is calculated to be approximately
0.002 ug/kg/day, the report suggests that even lower doses are required for
effects on reproductive function than are required in the rat and would support
a lower adverse effect level. Unfortunately, much of the data on the monkeys
has been presented in abstract form or as part of a review, and consequently, a
critical analysis of the data is impossible. There has been some indication
that studies of even lower levels (i.e., 5 and 25 ppt) showed signs of
reproductive toxicity, and the data are now beginning to appear in the
literature. Schantz et al. (1986) reported altered maternal care of offspring
in monkeys exposed to 5 and 25 ppt for 45 to 49 months, and Bowman et al.
(1987a, b) reported on altered maternal-infant interaction and other
reproductive parameters at the Seventh International Symposium on Chlorinated
Dioxins and Related Compounds. These reports are now being evaluated and, if
supported, could significantly affect the adverse effect levels calculated for
reproductive and developmental toxicity.
8
-------
It is important to note that none of these findings establish an
unequivocal effect at the 0.001 ug/kg/day or below level. However, the
evidence is suggestive enough and the uncertainties are great.enough that it
would seem prudent to consider the 0.001 level as highly suspect. ,
DEVELOPMENT
Numerous studies have been done on the developmental toxicity of 2,3,7,8-
TCDD, many of which have been summarized in the HAD for Polychlorinated
Dibenzo-£-Dioxins (U.S. EPA, 1985). Of particular interest to this review are
those studies which present data that may factor into an overall risk
characterization: Courtney and Moore (1971), Giavini et al. (1982a), Khera and
Ruddick (1973), McNulty (1980), Moore et al. (1973), Smith et al. (1976), and
Sparschu et al. (1971).
Developmental toxicity following exposure to 2,3,7,8-TCDD has been
demonstrated in different species, including the chicken, mouse, rat, rabbit,
ferret, and monkey. Thus, developmental toxicity of 2,3,7,8-TCDD does not
appear to be related to a species-specific metabolic or physiological response
to exposure. While specific responses and effective doses do vary among
species and among strains within a species (Courtney and Moore, 1971; Poland
and Glover, 1980), developmental toxicity in response to 2,3,7,8-TCDD exposure
can be expected to occur in all species.
The exposure range at which developmental toxicity first becomes apparent
is 0.125 to 1.0 ug/kg/day when exposure occurs over a major period of
organogenesis. The no-effect level appears to be approximately 0.1 ug/kg/day
in rats, mice, and rabbits. Giavini et al. (1982a) did note an increase in
extra ribs at 0.1 ug/kg/day in the rabbit. However, the number of ribs
9
-------
normally varies between 12 and 13 in rabbits, and it would require a more
detailed analysis of the data to establish this as a true effect level.
Several laboratories have examined the effects of more acute exposures during
organogenesis. Moore et al. (1973) demonstrated that a single oral dose of
1 ug/kg given to mice on gestation day 10 produced hydronephrosis. In an
interesting extension of this finding, they investigated the postnatal
development of the kidney in cross-fostered offspring following prenatal
exposure. The frequency of hydronephrosis seen postnatally was largely
dependent on whether the offspring were nursed by a dam that had also been
prenatally exposed to 2,3,7,8-TCDD. Thus, it appeared that continued exposure
during the lactation period was required to produce or maintain the greatest
effect on kidney development during the postnatal period studied.
Except for the multigeneration studies, which tend not to critically
evaluate many developmental end points, few studies have been carried out on
developmental periods other than the period of organogenesis. Giavini et al.
(1982b) did examine 2,3,7,8-TCDD exposure on gestation days 1-3 in the rat and
reported possible delays in implantation and some effect on fetal weight and
the kidney. There appear to be no studies on exposure during late prenatal
development. As noted above, Moore et al. (1973) examined effects on the
kidney postnatally following prenatal exposure of the dams, and demonstrated
that transfer in the milk is a likely contributing factor to the developmental
toxicity of 2,3,7,8-TCDD. There have also been reports describing the effects
of 2,3,7,8-TCDD exposure on the developing immune system (see Appendix E).
However, carefully designed studies on postnatal exposures or on changes in
postnatal function following prenatal exposure have generally not been carried
out, making it impossible to evaluate the potential effect of 2,3,7,8-TCDD on.
10
-------
the developing young animal.
In summary, studies on the developmental toxicity of 2,3,7,8-TCDD have
clearly demonstrated that it is a developmental toxicant in a wide variety of
species at relatively low doses. These studies have also shown that extended
periods of exposure are not necessary for developmental toxicity to result.
Thus, there is a need for concern over acute or short-term exposure to 2,3,7,8-
TCDD. A composite review and summary of the studies on the developmental
toxicity of 2,3,7,8-TCDD is limited by factors such as the simultaneous
occurrence of maternal toxicity, the use of different animal strains and
exposure routes, and in many studies the small number of animals per treatment
group that are included in the final data analysis. In addition, there are
differences in study designs and approaches to data analysis which must be
considered in comparing the studies. These factors do not alter the finding
that 2,3,7,8-TCDD is a developmental toxicant at very low exposure levels.
However, they can potentially affect the final assessment of exposure levels
that can be considered toxic.
MALE AND FEMALE REPRODUCTIVE SYSTEM
Specific components of the male and female reproductive systems are
affected by 2,3,7,8-TCDD exposure. In the male, exposures above 1 ug/kg/day
resulted in evidence of testicular atrophy with destruction of the seminiferous
tubules and spermatogenic cells (Kociba et al., 1976; Norback and Allen, 1973;
McConnell et al., 1978). However, following exposures of 0.001 to 0.1
ug/kg/day over a 2-year period, Kociba et al. (1978) reported that the male
reproductive organs appeared to be unaffected, relative to the controls. In a
study of offspring of male mice exposed to 0.16 to 2.4 ug/kg/day of 2,3,7,8-
11
-------
TCDD combined with higher'doses of 2,4-dichlorophenoxyacetic acid and
2,4,5-trichlorophenoxyacetic acid, there did not appear to be any effect on
fetal or neonatal development or viability (Lamb et al., 1981). In the female,
exposure to 1 to 2 ug/kg/day for 13 weeks resulted in changes in estrous
cycles, anovulation, and signs of ovarian dysfunction (Kociba et al., 1976).
At exposures of 0.001 to 0.01 ug/kg/day in a 2-year study, Kociba et al. (1978)
reported no effects on the female reproductive system. At 0.1 ug/kg/day, a
decrease in uterine changes such as endometrial hyperplasia were reported. As
noted previously in the Allen et al. studies, female monkeys exposed to 500 ppt
2,3,7,8-TCDD were reported to exhibit changes in their serum steroid levels.
SUMMARY
2,3,7,8-TCDD has been shown to be a reproductive and developmental
toxicant in animal studies. Among the effects that have been reported are
reduced fertility, litter size, postnatal survival, and offspring body weight,
as well as an increase in structural malformations. Effects on the male and
female gonads and on the female menstrual/estrous cycle have also been
reported. The effects have been observed in a variety of species, indicating
that the toxicity is not a species-specific event and can be expected to occur
in all species, including the human.
The studies on reproductive function and fertility remain the basis for
establishing the lowest effective (toxic) exposure level. There is some
disagreement, centered on the appropriate approach for data analysis, over the
effect level based on the Murray et al. (1979) study. Based on the current
information from this study, a 0.01 ug/kg/day level is the lowest effect level
12
-------
that can be supported by the data. However, there is enough suggestive
evidence to indicate a real potential for an effect at 0.001 ug/kg/day.
Further analysis of this study and of the subhuman primate studies by Allen
and his colleagues may provide support for a lower effect level of at least
0.001 ug/kg/day.
In relation to developmental toxicity, a large number of studies in a
variety of species demonstrated that 2,3,7,8-TCDD is a developmental toxicant.
Although a longer term exposure appears to cause effects at slightly lower
doses, acute and short-term exposures are effective in causing altered
development, and therefore, should also be of concern. When exposure occurs
during the prenatal period of organogenesis, the lowest effect level is in the
range of 0.125 to 1.0 ug/kg/day and the no-effect level is approximately 0.1
ug/kg/day. Studies focusing on other periods of development are limited, but
they do indicate that exposure at any time during prenatal and early postnatal
life must be considered a potential threat to normal development.
2,3,7,8-TCDD also affects the male and female reproductive systems.
Unfortunately, the amount of attention that has been given, to these areas of
investigation has not been as great as that given to reproductive function and
development. Gonadal dysfunction has been demonstrated in both sexes, and
alterations of normal reproductive cycles have been demonstrated in the female.
Continuing investigations of the effect of 2,3,7,8-TCDD and related agents on
reproductive physiology and cellular events should increase our understanding
of the potential effect of 2,3,7,8-TCDD on the male and female reproductive
systems.
The uncertainties that arise from this data base are many, largely because
the area of reproductive and developmental toxicology covers a wide range of
13
-------
potential exposure-response scenarios, and because the reproductive and
developing systems present constantly changing targets with which a toxicant
may interact. In the case of 2,3,7,8-TCDD exposure, many of these
uncertainties have been discussed in this review and should be taken into
account in carrying out the final risk characterization. The laboratory data
establishes 0.01 ug/kg/day as a lowest observed adverse effect level (LOAEL)
when exposure is chronic. Standard approaches for applying uncertainty/
modifying factors and determining a reference dose (RfD) have been used (U.S.
EPA, 1986b; 1987). An unequivocal no observed adverse effect level (NOAEL) can
be questioned. A 1000-fold uncertainty factor (UF) can be applied to account
for variation in sensitivity within the human population, uncertainty in
extrapolating animal data to the human situation, and the use of a LOAEL
instead of a NOAEL in calculating the reference dose. An additional modifying
factor seems unnecessary, since the uncertainty factor accounts for many of the
concerns of this reviewer, i.e., the true LOAEL/NOAEL and the potential for the
pregnant woman and her offspring to be more sensitive than the average healthy
adult. An exception to this position would arise if an actual effect level
much below 0.001 ug/kg/day was established, and as noted above, suggestive
evidence is accumulating that would support a lower NOAEL/LOAEL. The
calculation of a reference dose (RfD) for reproductive and developmental
toxicity, based on the current data and literature base, is as follows:
RfD = LOAEL/UF
= (0.01 ug/kg/day)/1000
= 1 x 10~5 ug/kg day..-.
14
-------
This does not account for potential or actual human exposures which would have
to be factored into the final risk characterization.
Future investigations should be encouraged to more clearly define the
substantial data base that already exists. In the area of reproductive
function and fertility, it would be helpful if a more critical review of the
data of the three-generation rat study and the monkey studies was carried out
to determine if a lower effective dose can be established. Additionally, a
carefully designed multigeneration study could address some of the limitations
of previous studies and fill certain data gaps. However, multigeneration
studies are not easily designed, executed, or evaluated, and this step should
only be taken when it is obvious that these data are necessary. In the area of
developmental toxicology, the most pressing needs seem to be the evaluation of
toxicity during periods of development that have not been adequately assessed,
i.e., the late prenatal and early postnatal periods. With regard to the
postnatal period, the potential for childhood exposure from breast feeding and
from other environmental sources (e.g., ingestion of soil) seems considerable,
and it has been suggested that the child may be particularly sensitive to
exposure to polychlorinated dibenzo-p_-dioxins. As relates to reproductive and
developmental toxicity in general, a greater effort should be directed at
identifying the effects of polychlorinated dibenzo-£-dioxins on hormonal
regulation and normal cellular and tissue functions. There is evidence that
2,3,7,8-TCDD influences steroid metabolism and may be associated with steroid
action at the cellular receptor level. Pratt and his colleagues (Dencker and
Pratt, 1981; Pratt et al., 1984) have also shown a correlation in various mouse
strains between the susceptibility to induction of cleft palate and the
occurrence of the 2,3,7,8-TCDD receptor, and have proposed that 2,3,7,8-TCDD
15
-------
exerts its teratogenic effect on the palate directly through this receptor.
A considerable amount of literature has developed on the mechanisms of cellular
interaction and action aspects of 2,3,7,8-TCDD, as well as on the
structure-activity relationships of 2,3,7,8-TCDD with other dioxins and related
compounds (recent reviews include: Safe, 1986; Silbergeld and Mattison, 1987).
Efforts should be made to incorporate this information and to evaluate
2,3,7,8-TCDD within the context of the larger family of related agents. As
additional information becomes available, this review will be updated and
reconsidered relative to its appropriateness in defining critical studies and
issues related to the reproductive and developmental toxicity of 2,3,7,8-TCDD.
16
-------
REFERENCES
Abbott, B.D.; Birnbaum, L.S.; Pratt, R.M. (1987a) TCDD produces
hydronephrosis in fetal mice by inducing hyperplasia of the ureteric
epithelium. The Toxicologist 7:124. Abstract 496.
Abbott, B.D.; Birnbaum, L.S.; Pratt, R.M. (1987b) TCDD effects the basal
lamina and extracellular matrix of the fetal mouse kidney. The Toxicolo-
gist 7:124, Abstract 495.
Allen, J.R.; Barsotti, D.A.; Lambrecht, L.K.; Van Miller, J.P. (1979)
Reproductive effects of halogenated aromatic hydrocarbons on nonhuman
primates. Ann. NY Acad. Sci. 320:419-425.
Allen, J.R.; Barsotti, D.A.; Van Miller, J.P.; Abrahamson, L.J.; Lalich, J.J.
(1977) Morphological changes in monkeys consuming a diet containing low
levels of 2,3,7,8-tetrachlorodibenzo-p-dioxin. Food Cosmet. Toxicol.
15:401-410.
Bailar, J.C., III. (1981) Direct testimony of Dr. John C. Bailar, III. FIFRA
Docket No. 415, et al., Exhibit 936. U.S. Environmental Protection
Agency, Washington, DC. -
Barsotti, D.A.; Abrahamson, L.J.; Allen, J.R. (1979) Hormonal alterations in
female rhesus monkeys fed a diet containing 2,3,7,8-tetrachlorodibenzo-
jD-dioxin. Bull. Environ. Contam. Toxicol. 21:463-469.
Bowman, R.E.; Schantz, S.L.; Gross, M.L.; Ferguson, S. (1987a) Behavioral
effects in monkeys exposed to 2,3,7,8-TCDD transmitted maternally during
gestation and for four months of nursing. Presented at the Seventh
International Symposium on Chlorinated Dioxins and Related Compounds, Las
Vegas, Nevada. Abstract AT02.
Bowman, R.E.; Schantz, S.L.; Weerasinghe, N.C.A.; Gross, M.L.; Barsotti, D,A.
(1987b) Chronic dietary intake of 2,3,7,8-tetrachlorodibenzo-j)-d;ioxin
(TCDD) at 5 and 25 parts per trillion in the monkey: TCDD kinetics and
dose-effect estimate of reproductive toxicity. Presented at the Seventh
International Symposium on Chlorinated Dioxins and Related Compounds, Las
Vegas, Nevada. Abstract AT03.
Courtney, K.D.; Moore, J.A. (1971) Teratology studies with 2,4,5-trichloro-
phenoxyacetic acid and 2,3,7,8-tetrach1orodibenzo-£-dioxin. Toxicol.
Appl. Pharmacol. 20:396-403. , ,
Dencker, L.; Pratt, R.M. (1981) Association between the presence of the Ah
receptor in embryonic murine tissues and sensitivity to TCDD-induced cleft
palate. Teratogenesis Carcinog. Mutagen. 1:399-406.
Giavini, E.; Prai, M.; Vismara, C;- (1982a) Rabbit teratology study with
2,3,7,8-tetrachlorodibenzo-fi-dibxin. Environ. Res. 27:74-78.
17
-------
Giavini, E.; Prai, M.; Vismara, C. (1982b) Effects of 2,3,7,8-tetrachloro-
dibenzo-fi-dioxin administered to pregnant rats during the preimplantation
period. Environ. Res. 29:185-189.
Khera, K.S.; Ruddick, J.A. (1973) Polychlorodibenzo-£-dioxins: perinatal
effects and the dominant lethal test in Wistar rats. In: Blair, E.H., ed.
Chlorodioxins-origin and fate. Adv. in Chemistry Series 120. Washington,
DC.: American Chemical Society.
Kociba, R.J.; Keeler, P.A.; Park, C.N.; Gehring, P.O. (1976) 2,3,7,8-
tetrachlorodibenzo-£-dioxin (TCDD): results of a 13-week oral toxicity
study in rats. Toxicol. Appl. Pharmacol. 35:553-574.
Kociba, R.J.; Keyes, D.G.; Beyer, J.E.; Carreon, R.M.; Wade, C.E.; Dittenber,
D.A.; Kalnins, R.P.; Frauson, I.E.; Park, C.N.; Barnard, S.D.; Hummel,
R.A.; Humiston, C.G. (1978) Results of a two-year chronic toxicity and
oncongenicity study of 2,3,7,8-tetrachlorodibenzo-p_-dioxin in rats.
Toxicol. Appl. Pharmacol. 46:279-303.
Lamb, J.C., IV.; Moore, J.A.; Marks, T.A.; Haseman, O.K. (1981) Development
and viability of offspring of male mice treated with chlorinated phenoxy
acids and 2,3,7,8-tetrachlorodibenzo-fi-dioxin. J. Toxicol. Environ.
Health 8:835-844.
McConnell, E.E.; Moore, J.A.; Haseman, J.K.; Harris, M.W. (1978) The
comparative toxicity of chlorinated dibenzo-p_-dioxins in mice and guinea
pigs. Toxicol. Appl. Pharmacol. 44:335-356.
McNulty, W.P. (1980) Direct testimony of Dr. Wilbur P. McNulty, FIFRA Docket
Nos. 415 et a!., Exhibit No. 106. U.S. Environmental Protection Agency,
Washington, DC.
Moore, J.A.; Gupta, B.N.; Zinkl, J.G.; Vos, J.G. (1973) Postnatal effects of
maternal exposure to 2,3,7,8-tetrachlorodibenzo-£-dioxin (TCDD). Environ.
Health Perspect. 5:81-85.
Murray, F.J.; Smith, F.A.; Nitschke, K.D.; Humiston, C.G.; Kociba, R.J.;
Schwetz, B.A. (1979) Three-generation reproduction study of rats given
2,3,7,8-tetrachlorodibenzo-jD-dioxin (TCDD) in the diet. Toxicol. Appl.
Pharmacol. 50:241-252.
Murray, F.J.; Smith, F.A.; Nitschke, K.D.; Humiston, C.G.; Kociba, R.J.; John,
J.A.; Blogg, C.D.; Schwetz, B.A. (1978) Three-generation study of rats
ingesting 2,3,7,8-tetrachlorodibenzo-j3-dioxin (TCDD). Report of the Dow
Chemical Company, Exhibit 77, FIFRA Docket Nos. 415 et al. U.S.
Environmental Protection Agency, Washington, DC.
Nisbet, I.C.T.; Paxton, M.B. (1982) Statistical aspects of three-generation
studies of the reproductive toxicity of 2,3,7,8-TCDD and 2,4,5-T. The
American Statistician 36:290-298.
18
-------
Norback, D.H.; Allen, J.R. (1973) Biological responses of the nonhuman
primate, chicken, and rat to chlorinated dibenzo-p_-dioxin ingestion.
Environ. Health Perspect. 5:233-240.
Palmer, A.K. (1981) Regulatory requirements for reproductive toxicology:
theory and practice. In: Kimmel, C.; Buelke-Sam, J., eds. Developmental
Toxicology. New York: Raven Press.
Poland, A.; Glover, E. (1980) 2,3,7,8-tetrachlorodibenzo-p_-dioxin:
segregation of toxicity with the Ah locus. Mol. Pharmacol. 17:86-94.
Pratt, R.M.; Grove, R.I.; Kim, C.S.; Dencker, L.; Diewert, V.M. (1984)
Mechanism of TCDD-induced cleft palate in the mouse. In: Poland, A.;
Kimbrough, R.D., eds. Biological mechanisms of dioxin action. Banbury
Report No. 18. Cold Spring Harbor Lab.
Safe, S.H. (1986) Comparative toxicology and mechanism of action of
polychlorinated dibenzo-p_-dioxins and dibenzofurans. Annu. Rev. .
Pharmacol. Toxicol. 26:371-399.
Schantz, S.L.; Barsotti, D.A.; Allen, J.R. (1979) lexicological effects
produced in nonhuman primates chronically exposed to fifty parts per
trillion 2,3,7,8-tetrachlorodibenzo-p_-dioxin (TCD.D). Toxicol. Appl.
Pharmacol. 48:(2)A180.
Schantz, S.L.; Laughlin, N.K.; Van Valkenberg, H.C.; Bowman, R.E. (1986)
Maternal care by rhesus monkeys of infant monkeys exposed to either lead
or 2,3,7,8-tetrachlorodibenzo-p_-dioxin. Neurotoxicology 7:637-650.
Silbergeld, E.K.; Mattison, D.R. (1987) Experimental and clinical studies on
the reproductive toxicology of 2,3,7,8-tetrachlorodibenzo-p_-dioxin. Am.
J. Ind. Med. 11:131-144.
Smith, F.A.; Schwetz, B.A.; Nitschke, K.D. (1976) Teratogenicity of 2,3,7,8-
tetrachlorodibenzo-p_-dioxin in CF-1'Mice. Toxicol. Appl. Pharmacol.
38:517-523.
Sparschu, G.L.; Dunn, F.L.; Rowe, V.K. (1971) Study of the teratogenicity of
2,3,7,8-tetrachlorodibenzo-p_-dioxin in the rat. Food Cosmet. Toxicol.
9:405-412.
U.S. Environmental Protection Agency (EPA) (1985) Health assessment document
for polychlorinated dibenzo-p_-dioxins. Office of Health and Environmental
Assessment. EPA/600/8-84/014F. NTIS PB86-122546/AS.
U.S. Environmental Protection Agency (EPA) (1986a) Teratology and
reproduction studies with TCDD. In: Pitot, H., ed. The report of the
"dioxin" Update Committee. Office of Pesticides and Toxic Substances,
Washington, DC.
19
-------
U.S. Environmental Protection Agency (EPA) (1986b)
assessment of suspect developmental toxicants.
51:34028-34040.
Guidelines for the health
Federal Register
U.S. Environmental Protection Agency (EPA) (1987) Reference dose (RfD):
description and use in health risk assessments. Integrated Risk
Information System (IRIS): Appendix A. Online. Intra-Agency Reference
Dose Workgroup, Office of Health and Environmental Assessment.
Environmental Criteria and Assessment Office, Cincinnati, OH.
EPA/600/8-86/032a.
20
-------
February 1988
Review Draft
APPENDIX D
EPIDEMIOLOGIC DATA ON REPRODUCTION AND
EXPOSURE TO 2,3,7,8-TCDD:
ITS USEFULNESS IN QUANTITATIVE RISK ASSESSMENT
Sherry G. Selevan, Ph.D.
Reproductive Effects Assessment Group
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
-------
-------
INTRODUCTION '
The following is a discussion of epidemiologic data on the reproductive
effects of 2,3,7,8-tetrachlorodibenzo-p_-dioxin (2,3,7,8-TCDD) and their
usefulness in qualitative and quantitative risk assessment. Several factors
affect the usefulness of these studies. First, and probably most important, is
the assessment of exposure. Other important factors include the biologic
plausibility of the effects of exposure, given the time frames of exposure and
its association with relevant outcomes, and the difficulty of attributing the
effects of combined exposure to polychlorinated dibenzo-p_-dioxins (hereafter
referred to as "dioxin"), a contaminant in herbicides or pesticides.
Due to the narrow focus of this report, that is, the use of existing
reproductive data in the risk assessment of dioxin, certain restrictions will
be made on the data discussed: The usefulness of data for risk assessment
depends upon the manner in which the probabilities of exposure for the study
members are determined. The quality of exposure data may range from very
indirect data to detailed industrial hygiene or environmental monitoring.
These indirect data are less useful and result from assumptions that
individuals were exposed due to his/her presence in a potentially exposed
region/plant during a specific time period. More useful data would include
measurement of actual levels of dioxin in air, soil, or water with the most
useful data describing the individuals' levels of exposure. Studies with only
indirect, assumed exposure data contribute little to a risk assessment because
of limited confidence in the ability to determine whether a given individual
was actually exposed. Misclassification of exposure will inevitably occur,
typically resulting in a reduction of the estimate of risk (if such a risk
1
-------
truly exists). Studies with more detailed exposure data will contribute more
specific information to such a risk assessment.
ROUTES OF EXPOSURE
Humans may be exposed to 2,3,7,8-TCDD through several routes: dermal,
ingestion, and inhalation. The rates of absorption through these routes vary,
as do the important routes for each individual. The routes for each individual
can be affected by such characteristics as location during spraying (e.g.,
Vietnam/Oregon forests), diet (e.g., consumption of contaminated water/fish),
and work practices for those occupationally exposed. As these characteristics
vary, so will each person's effective dose. Table 4 (U.S. EPA, 1988) describes
the absorption fraction for dioxin from several routes: soil ingestion - 0.3,
dermal exposure to soil - 0.005, vapor inhalation - 0.75, fat ingestion from
dairy products or beef - 0.68, dust inhalation - 0.27, fish ingestion - 0.68,
and surface water ingestion - 0.5. For example, an individual exposed to equal
amounts of 2,3,7,8-TCDD from different routes may have radically different
internal doses (e.g., during the ICMESA plant explosion in Seveso, Italy, an
individual would have a different potential exposure than.she/he would have if
exposed to soil dust later on). Consequently, each individual's actual dose is
difficult to estimate accurately in an epidemiologic study. Another potential
source of 2,3,7,8-TCDD exposure of potential reproductive importance may be the
infant's exposure through human breast milk (Rappe, 1985; Schecter et al.,
1987; van den Berg et al., 1986). ,
The epidemiologic literature, while limited at this point in time, seems
to cover two broad categories: In occupational settings, paternal effects have
-------
been examined (e.g., aerial sprayers in Vietnam, Vietnam veterans, agricultural
sprayers, and employees of manufacturers of 2,4,5-trichlorophenoxyacetic acid
[2,4,5-T]). Birth defects and fetal loss have been examined through
environmental exposures (e.g., aerial spraying of 2,4,5-T in Oregon, New
Zealand, and Vietnam, and the ICMESA plant accident in Seveso, Italy). In
these settings, separation of maternal and paternal effects is very difficult.
The only exception may be in such settings as the plant accident in Seveso
where the woman may already be pregnant and In utero exposures are the ones of
importance. None of these exposures are "clean;" that is, of dioxih alone.
All the human exposures to dioxin have been as a contaminant of some other
agent (e.g., 2,4,5-T), and typically, only sketchy qualitative and quantitative
exposure data were available. Therefore, it is difficult to separate out the
effects of dioxin from the agent it contaminates. Only some very general
conclusions may be drawn from the relative toxicities of the associated
compounds.
A number of reviews have discussed the human data in detail (Kimbrough et
a!., 1984; Constable and Hatch, 1985; Friedman, 1984; Hatch, 1984; Hatch and
Stein, 1986; U.S. EPA, 1985). These efforts will not be repeated here, but the
discussion will be limited to several key areas of study and certain studies of
special interest to help define the limits of our knowledge on the human
reproductive effects of dioxin.
STUDIES OF VIETNAM VETERANS
Two studies have examined the reproductive effects of 2,3,7,8-TCDD and
defoliants in U.S. Vietnam veterans. The primary exposure was to Agent Orange
-------
which consists of defoliants 2,4-dichlorophenoxyacetic acid (2,4-D) and
2,4,5-T, the latter contaminated with 2,3,7,8-TCDD. (Agents Purple, Pink, and
Green were primarily used prior to July 1965.) These two studies are the Ranch
Hand study (Lathrop et al., 1984) and the Center for Disease Control's (CDC)
study of birth defects (Erickson et al., 1984). A third study examines the
birth defects in children born to Australian Vietnam veterans (Donovan et al.,
1983).
The Ranch Hand study (Lathrop et al., 1984) compared airmen who had flown
in the aerial spraying of defoliants (in fixed-wing aircraft) to those who
belonged to flying organizations responsible for transporting cargo. Other
types of spraying (helicopters and backpacks) were not included in the Ranch
Hand operations. Hhile the data in this report have not yet been verified
(through birth registration or medical records), preliminary analyses have
examined various measures of fertility and reproductive success (through
pregnancy outcomes, sperm count, and morphology). The researchers are
currently validating the interview data and will subsequently (they state) do
more detailed analyses. Exposure to 2,3,7,8-TCDD was estimated by calculating
the amount of 2,3,7,8-TCDD (in herbicides) used during each airman's tour and
dividing this number by the total number of airmen with the same
responsibilities during the subject's tour.
The Ranch Hand study individually matched comparisons to each exposed man,
put in replacements for refusals, and did not maintain the matches in the
analyses. Due to these procedures, determination of the total study population
and response rate is difficult; 1,174 Ranch Handers were included in the
analysis of reproductive-data, as were 1,531 comparisons. Both the man and his
current or former spouse were interviewed on reproductive history; research has
-------
shown such interview data on reproductive events.to have greater validity when
collected from the wife. However, the researchers combined pregnancies from
both spouses (but not necessarily reported by both), so it is impossible to
examine the women's data separately. No differences were observed in early or
late fetal loss, induced abortion, or live births in a crude analysis of the
two groups. More detailed analyses examined enlisted men and officers
separately, thus reducing the power of the analyses. Some important
independent risk factors were not adjusted for in the analysis (e.g., prior
fetal loss in analyses of current fetal loss). It appears that multiple
pregnancies per family group were all analyzed; however, there was no
discussion of the problems associated with the analysis of nonindependent
events. A statistically significant excess was found for birth defects,
controlling for parental age and for maternal smoking and drinking; however,
the reproductive outcome of "birth defects" is probably the one which most
needs validation against medical records to assure accurate classification. No
differences were observed in sperm count or morphology in the two groups. The
more detailed analysis planned for the future could result in useful
information since the potential for misclassification of exposure in this study
appears to be less than for the other veterans studies.
Another major study of U.S. Vietnam veterans was a case-referent study of
birth defects done by the CDC (Erickson et al., 1984). The cases and referents
were drawn from births occurring in metropolitan Atlanta from 1968 through
1980. The 7,133 eligible cases consisted of live and still births with serious
or major defects identified using the 8th revision of the International
Classification of Diseases (ICD-8). The referent group was selected from
metropolitan Atlanta live births during the same years; the referents were
-------
frequency-matched on race, year of birth, and hospital of birth. Power
calculations suggested that a referent group of 3,000 was sufficient; above
that no substantial increase in power would be found for "all defects." The
authors also stated that power for odds ratios greater than 3.0 for specific
birth defects also would not be substantially affected by using more
comparisons. To allow for a 70% response rate, 4,246 referent births were
selected. Seventy percent of the women responded, but only 56% of the men did.
The potential for exposure for the CDC study was based on three
definitions: First, a man was considered exposed if he had served in any
capacity at any time in Vietnam before the conception of the infant. Second,
each veteran was queried about his exposure to Agent Orange. Third, an
Exposure Opportunity Index (EOI) score was (subjectively) developed by a panel
of specialists who evaluated the veterans' duties, location, and time spent in
Vietnam. Veteran status was examined to determine whether any military service
might be associated with the 96 birth defects examined. Veteran status was not
associated and subsequently dropped from further analyses.
These data were analyzed a number of ways, but for all, the data were
stratified on the three variables used for the frequency matching: (1) without
potentially confounding characteristics; (2) with key characteristics
identified a priori (maternal age, education, and alcohol consumption, and the
presence of birth defects in close (first-degree) relatives of the child
studied; and (3) with a. posteriori testing of 108 -other characteristics. In
addition, certain groupings of the 96 birth defect categories, thought to be
related, were also examined. A limited number of elevated findings were
reported out of almost 400 analyses. Spina bifida was associated with both
levels of the EOI indexes; cleft lip without cleft palate was associated with
-------
veteran status and high EOI; specified anomalies of nails was reported in
Vietnam veterans; and "other neoplasms" were associated with high EOI. Due to
the large number of analyses and hypotheses tested, one would expect
approximately 20 significant associations (elevated or decreased odds ratios)
with a significance level of p = 0.05 by chance alone. Only 11 statistically
significant findings were reported, less than that expected by chance. The
authors appropriately concluded that "... the data collected contain no
evidence to support the position that Vietnam veterans have had a greater risk
than other men for fathering babies with all types of serious structural birth
defects combined."
The Australian government examined the association between military
service in Vietnam and birth defects in their offspring (Donovan et al., 1983).
A case-referent study compared births occurring from 1966 through 1979, of
8,517 children with defects recognizable at birth to an equal number of live
born children without birth defects. The referents were matched by time of
birth and maternal age. Exposure was defined as presence of, the father in
Vietnam; no specific index of exposure was developed, but the investigators
stated that such exposure was probably low for the Australian troops in
Vietnam. No difference lwas observed in exposure patterns between cases and
referents. " ,
For veteran studies, the exposures are to the fathers, typically many
years before the pregnancy under study was conceived. These studies have used
different methods to assign veterans into different exposure groups, based on
service in Vietnam (Erickson et al., 1984; Donovan et al., 1983), matching of
troop movement with aerial spraying (Erickson et al.,' 1984), or participation
in the Ranch Hand operations (Lathrop et al., 1984). At the time of these
-------
studies, a suitably sensitive assay for 2,3,7,8-TCDD in serum did not exist.
The CDC recently reported on a new assay they developed in 1986; this assay
yielded serum 2,3,7,8-TCDD levels which correlated well (r = 0.98) with
2,3,7,8-TCDD levels in the adipose tissue of the same individuals (CDC, 1987).
Paired samples collected from the same individuals in 1982 and 1987 were used
to derive the half-life of the 2,3,7,8-TCDD body burden (6 to 10 years). In
this preliminary report, CDC compared the assayed levels of 2,3,7,8-TCDD to the
exposure categories they had established using the EOI and interview data: no
association was found between the three methods of scoring for potential
exposure and the serum levels observed. Furthermore, in their examination of
the sera of Vietnam era veterans (444 Vietnam veterans versus 75 non-Vietnam
veterans), the median sera levels did not differ for these two subgroups.
These data raise a number of issues in the consideration of this body of
research:
(1) The relatively long half-life broadens the range of potential
mechanisms that could occur if an association exists between paternal
2,3,7,8-TCDD exposure and birth defects. Friedman (1984) and Hatch and
Stein (1986) have suggested that such an exposure would cause birth
defects in offspring through gene or chromosomal mutation of
spermatogonial stem cells (premeiotic effects). However, the long half-
life observed means that postmeiotic effects on sperm are also possible.
The long half-life also lends more support to Friedman's suggestion of
maternal/fetal exposures from 2,3,7,8-TCDD in seminal fluid.
(2) These data suggest that a great deal of misclassification of exposure
occurred in the assignment of veteran's to their exposure categories for
the CDC study. Plus use of this assay would improve exposure estimation
-------
in the other studies.
(3) Finally, the levels and similarity of median 2,3,7,8-TCDD sera values
in both Vietnam and non-Vietnam veterans raises questions about the
magnitude of exposure of Vietnam veterans. Other data suggest that the
exposures of Vietnam veterans were less than originally thought: the
median levels of 2,3,7,8-TCDD in both groups are below those found in the
adipose tissue of the control group of a study in eastern Missouri
(Patterson et al., 1986).
OCCUPATIONAL STUDIES
Occupational studies tend to be more useful in the evaluation of a
potentially toxic exposure in risk assessment. This is especially true for
qualitative risk assessment, since one can be fairly certain that exposure to
the worker did occur. If good historical industrial hygiene data are
available, these can also be of use in quantitative risk assessment. As
described in the Vietnam veterans studies, the exposures evaluated here are
primarily to male workers and the studies have been restricted to examination
of paternal effects; however, wives may have been exposed indirectly, through
handling their husband's clothes, etc. Unlike the veterans studies, the
exposures may be occurring concurrently with conception, thus increasing the
potential exposure- level at the time of the pregnancy.
Smith et al. (1982) identified 616 male chemical applicators from a list
maintained by New Zealand's agricultural Chemicals Board and 531 comparison
workers at small agricultural contracting companies. A total of 89% of the
chemical applicators and 83% of the contractual workers responded to a mailed
9
-------
questionnaire requesting occupational histories from the men and reproductive
histories from their wives. Exposure to 2,4,5-T was defined as having sprayed
this pesticide during the year of the pregnancy studied or during the year
preceding that pregnancy. Group 1 (not exposed to any spraying) consisted of
392 pregnancies; group 2 (sprayed other chemicals than 2,4,5-T) had 109
pregnancies; and group 3 (sprayed 2,4,5-T) had 473 pregnancies. Comparisons of
group 3 to group 1 found no statistically significant differences in fetal loss
or birth defects. Other risk factors did not appear to be controlled in the
analysis of these data: for example, the analysis of miscarriages did not
appear to control for maternal age or occurrence of prior fetal loss. Multiple
pregnancies per family unit were studied; however, the authors did not address
the problem of nonindependent events. Additionally, the power of this study
was very low for birth defects.
Dow Chemical studied 930 male workers potentially exposed to dioxins,
through work with chlorophenol processes, for at least one month from January
1939 through December 1975 (Townsend et al.', 1982). The reproductive
experience of their wives, obtained by interview, was compared to the wives of
an equal number of unexposed male employees. For the 930 exposed workers, 586
wives were identified and 370 responded (63%); for the comparison group, 559
wives were identified and 345 responded (62%). Exposure potentials from low to
high were assigned to jobs using historic surface contamination data by an
industrial hygienist familiar with the history of the facility. A pregnancy
was considered exposed if the employee had worked in an exposed area for at
least one month at any time prior to conception. Thus, this study, in its
analysis, has similar problems to those described above for the study of
Vietnam veterans concerning the time delay between exposure and conception.
10
-------
The outcomes examined included miscarriages (< 20 weeks gestation),
stillbirths, and birth defects identified (or potentially .identified) before
the child's first birthday. The relationships of these outcomes were compared
to several classifications of exposure: any dioxin, only 2,3,7,8-TCDD,
moderate and higher levels of 2,3,7,8-TCDD only, other dioxins only. The "any
dioxin" and "only 2,3,7,8-TCDD" categories were done as broad groups, plus
split into two groups each, based on duration of exposure (< 12 months, or > 12
months) during the entire employment period preceding conception. Over two
thousand pregnancies were in the "non-dioxin" group, while the "dioxin" group
consisted of 737 pregnancies. The "non-dioxin" group included pregnancies
occurring before exposure for exposed workers. No differences were found in
the two groups. The power was low for the examination of birth defects. As
noted above, the assumption of any exposure greater than one month at any time
prior to conception could obscure a true effect in this population. A
reanalysis of these data, looking at possible associations with paternal
exposure during the 3 to 4 months preceding conception might give more insight
into potential paternal effects associated with this exposure.
In a clinical, cross-sectional study, reproductive characteristics in
current (N = 131) and retired (N = 161) men who had worked with 2,4,5-T were
selected for comparison to workers without exposure (N = 133). Of the workers
selected, 55% responded. Historic exposure data were not available, and job
titles were not sufficient to estimate exposure; therefore, interview exposure
history was used to group workers into a probable exposure category.
Contamination of the plant with 2,4,5-T occurred in 1949. For the purpose of
comparing reproductive histories, reported chloracne was used to distinguish
groups. No differences were found for fetal loss or birth defects between the
11
-------
workers with chloracne (N = 107/number of pregnancies = 235) and those without
(N = 91/number of pregnancies = 203) during or after 1948 (when 2,4,5-T was
first handled at the plant under study). Differences were observed in reports
of decreased libido and difficulty with erection or ejaculation. Due to the
very indirect nature of exposure assessment, the probability of
misclassification of the exposure category is great.
AGRICULTURAL AND FORESTRY SPRAYING
A number of studies have examined the relationship between reproductive
effects and aerial spraying of pesticides (Field and Kerr, 1979; Hanify et al.,
1981; Nelson et al., 1979; Thomas, 1980). In these studies, exposures could be
to either or both parents. These are all ecologic studies, in which exposure
of study members is assumed by their presence in an area (e.g., by their
residence), with a specified likelihood of exposure to a given agent. These
studies are heir to the "ecologic fallacy" in that individuals defined by
residence or some other factor as being in a certain exposure category, may not
in fact belong in that category. Thus, such studies are of very limited value
in either qualitative or quantitative risk assessments; therefore, only brief
discussions of published reports of these follow.
The earliest of these studies (Nelson et al., 1979) examined the
association between cleft lip and/or cleft palate (CL/P) in live births
occurring during the 32-year period beginning in 1943 and the spraying of
2,4,5-T in Arkansas. Approximately 1,200 cases of CL/P Were identified using
both birth certificate data and records of the Crippled Children's Services of
Arkansas Social and Rehabilitative Services. 2,4,5-T was primarily used on
12
-------
rice; therefore exposure was determined by estimating the proportion of rice
acreage to total acreage in each county from data supplied by the Arkansas
State Plant Board (1970-1974). (The exposure definition does not address the
use of 2,4,5-T in forestry.) The 75 counties were then divided into categories
of low, medium, and high potential for 2,4,5-T exposure. An increase was
observed in facial clefts over time, which the authors suggested were related
to improved recording and case ascertainment and not to an association with
2,4,5-T spraying.
A letter to the editor of Lancet at approximately the same time (Field and
Kerr, 1979) discussed the relationship between 2,4,5-T and 2,3,7,8-TCDD and
neural-tube defects in New South Wales, Australia. Only limited data were
presented in this letter: in New South Wales, annual rates for neural-tube
defects (anencephaly and meningomyelocele) were compared to annual usage
figures for 2,4,5-T for all of Australia. Seasonal variation of 2,4,5-T usage
was obtained by questionnaire of local governments from the years 1965-1976.
2,4,5-T usage increased from 90 tonnes (1 tonne = 1,000 kg) in 1965 to 482
tonnes in 1976. The authors reported a linear correlation between the previous
years' 2,4,5-T usage and neural-tube defects. In addition, a seasonal pattern
was noted, with the highest rates in conceptions occurring during the summer
months (December, January, February). The authors also noted that in the
Northern Hemisphere, the highest rates in conceptions also occurred during the
summer months. In addition to the limitations present in ecologic studies, no
comparison rates of neural-tube defects were reported in communities without
such exposure, nor were high versus low exposure communities compared.
Another letter to the editor of Lancet the next year (Thomas, 1980)
discussed a comparison of selected birth defects (cleft lip, cleft palate,
13
-------
spina bifida, anencephalus, and cystic kidney disease) and use of 2,4,5-T in
Hungary. The birth defects were identified using Hungary's malformation
registry, which records malformations recognized at up to one year of age. In
Hungary, the use of 2,4,5-T rose from 46 tonnes in 1969 to over 1,200 tonnes in
1975. The rates of some birth defects (spina bifida and anencephalus)
decreased over the time period examined, while others (cleft lip, cleft palate,
and cystic kidney disease) remained essentially unchanged. In a comparison of
this report with the New South Wales letter from Field and Kerr (1979), Thomas
noted that: (1) the population examined in Australia was spread over a much
greater geographic area (thus reducing the probability for exposure), (2) over
half of the population in Hungary lived in rural areas, while only 15% of the
population in Australia did, and (3) 24.6% of the population in Hungary was
actively involved in forestry and agriculture as compared with only 7.4% of the
Australian population. While these data suggest no association between the
defects and 2,4,5-T, this report also did not attempt to compare exposed
communities with less exposed (or unexposed) groups.
Hanify et al. (1981) compared rates of diagnosed birth defects in
stillbirths (> 28 weeks gestation) and live births in Northland, New Zealand,
to densities of aerial 2,4,5-T spray application (1960-1977).' The reproductive
events were identified through seven regional hospital records: 3.7,751 births
which included 436 stillbirths, 264 neonatal deaths, and 510 with recorded
birth defects. The location, data, and quantity of 2,4,5-T sprayed was
obtained from company records, and monthly estimates were made for each of.the
seven regions. Environmental levels were estimated modeling both the new
applications plus that fraction of the previous applications thought to still
be present. 2,4,5-T was not sprayed from 1959-1965; thus, for this time
• i-''
14
-------
period, exposure was considered to be zero. The zero time period (1960-1965)
was compared to years thought to be representative of "spraying years"
(1972-1976). Associations were found for all birth defects and for talipes
under certain assumptions of environmental persistence of 2,4,5-T. The authors
did not discuss the possibility of secular trends in the data, due to other
medical or environmental factors, or differences in the
identification/recording of malformations over time.
OTHER ENVIRONMENTAL EXPOSURES -- SEVESO, ITALY
This category of "other environmental exposures" includes unplanned
exposures such as exposures from accidental emissions from industrial
facilities. With an immediately recognizable environmental incident, exposure
of interest for reproduction could occur for both parents, or just for the
mother, if she is already pregnant at the time of first exposure (unless there
could be in utero exposure to dioxin in seminal fluid). In 1976, during the
production of trichlorophenol at the ICMESA plant in Seveso, Italy, a runaway
reaction resulted in an explosion that ultimately contaminated 700 acres in the
surrounding community. Environmental levels of 2,3,7,8-TCDD were determined
using wipe tests, evaluating toxic effects in small animals, and analyzing
grass samples. Approximately 2 weeks later, over 200 families were evacuated
from high contamination areas. Exposures were sufficiently high for chloracne
to be observed in this environmentally exposed population. Several reports
(Bisanti et al., 1979; Homberger et al., 1979; Pocchiari, 1980; Pocchiari et
al., 1980; Reggiani, 1978; Rehder et al., 1978; Tuchmann- Duplessis, 1977)
reported on comparisons of four potentially affected communities (Seveso, Meda,
15
-------
Cesano, and Desio) to nearby, unexposed commun'ities. Changes in fetal loss
rates occurring during the last quarter of 1976 and first quarter of 1977 were
found in both exposed and unexposed communities. An estimated 150 women were
in the first trimester of pregnancy at the time of the accident (Rehder et al.,
1978); of these, 125 women wished therapeutic abortions by October 1976.
Therapeutic abortions were approved for 30. Another estimate
(Tuchmann-Duplessis, 1980) reports a total of 108 (50 in 1976 and 58 in 1977)
therapeutic abortions in the four affected communities. Several reports
(Bisanti et al., 1979; Pocchiari, 1980; Pocchiari et al., 1980) suggested that
a large number of women obtained unapproved, and therefore not reported,
therapeutic abortions. This supposition was supported by a steep decrease in
birth rates in the first 6 months of 1977, primarily observed in the exposed
communities. (All communities had had decreases over time; however, the
decrease in exposed communities at this key time was much larger.) Marked
increases in the number of birth defects were noted in 1977. These have been
attributed to changes in reporting. Prior to this time, only certain birth
defects were required to be reported to Italian Health Officers (Reggiani,
1980). In addition to the limitations due to legal requirements, there were
other reasons for the underreporting of malformations: "Traditionally, Italian
physicians have under-reported congenital malformations because of their severe
negative social implications" (Tuchmann-Duplessis, 1980a, b). Although
physicians and midwives were encouraged to do more complete recording, Reggiani
(1978) concluded that the malformation data were "missing" for 1976 and
"incomplete" for the first trimester of 1977. In general, the researchers
cited above felt that, in spite of the problems associated with the data,
adverse reproductive effects did not occur in this population. However, the
16
-------
data do not appear to be complete enough to make definitive conclusions. In
addition, these reports have the same problems concerning exposure assessment
as those described previously in the Agricultural/Forestry Spraying section.
STUDIES OF THE VIETNAMESE
While studies of the Vietnamese have not been published in Western
journals, a review has been presented by Constable and Hatch (1985). Two types
of studies have been executed: In the South, the reproductive experience of
exposed and unexposed couples was compared; in this case, effects of maternal
and paternal exposures cannot be separated. In the North, reproductive
experience was compared for families of men who had served in the South (thus,
assuming paternal exposure) to families of men remaining in the North (and
therefore assuming no exposure of the father). Exposures for these studies
have been defined by residence, using historical data on spraying and/or
determined by the evidence of destruction of vegetation. Thus, these studies
have the same limitations for use in qualitative and quantitative risk
assessment as discussed previously.
Only one study (Lang et al., 1983 described in Constable and Hatch, 1985)
has attempted to determine types of exposure (e.g., exposed during spraying
versus the more indirect exposure due to dust or diet) and assign exposure
scores. In this study of veterans in the North, exposure was restricted to the
father at some time prior to conception, thus resulting in the same problems
encountered in the American and Australian studies of veterans.
17
-------
SUMMARY
All three Vietnam veterans studies examined the occurrence of birth
defects. The Ranch Hand study found an excess for "all birth defects" (these
were not validated); the CDC study found a difference in exposure for cases
with spina bifida, cleft lip without palate, specified anomalies of the nails,
and "other neoplasms"; the third study (Donovan et.al., 1983) found no
differences. No associations with other reproductive outcomes were found in
the Ranch Hand study (the only study of the three that looked at the other
outcomes). No differences were found in reproductive outcome in the three
workplace studies (Moses et al., 1984; Smith et a!., 1982; Townsend et al.,
1982), however, Moses et al. did report increased impotence and sexual
dysfunction in 2,4,5-T workers with chloracne (assumed to result from 2,4,5-T
and/or 2,3,7,8-TCDD). Several of the ecologic studies (Nelson et al., 1979;
Field and Kerr, 1979; Hanify et al., 1981) reported associations with some
birth defects and volume of 2,4,5-T application; these data were contested by
Thomas (1980). Birth defects in Seveso increased after the plant accident,
potentially due to changes in reporting (Reggiarii, 1980). As presented in this
paragraph, one might conclude that 2,3,7,8-TCDD/2,4,5-T is related to adverse
reproductive outcomes, but as discussed previously, limitations in the design
of the studies neither allow nor rule out such an interpretation.
Due to the very nature of;the exposures to 2,3,7,8-TCDD, obtaining
suitable data for either a qualitative or quantitative risk assessment is
difficult, if not impossible. First, the exposures are primarily in general
environmental settings (wartime exposures are, for these purposes, being
classified as environmental) where it is difficult to define whether an
18
-------
individual is even exposed and nearly impossible.to quantitate that exposure.
In occupational settings,, both qualitative and quantitative exposure data may
exist, but small population sizes limit the power of the studies. A nationwide
study planned by the National Institute for Occupational Safety and Health
(NIOSH) on their Dioxin Registry could overcome the problems of limited power.
All of the studies described above contained a major limitation:
imprecise definitions of exposure of the subjects. In many studies, the
exposure was assumed by place of residence or presence in a particular area.
Even where more descriptive exposure data are available, such as the CDC and
Ranch Hand studies, misclas.sification may be great (CDC, 1987). Other studies,
such as the ecologic studies, compared the same area over time rather than to a
comparison area (e.g., Field and Kerr, 1979; Thomas, 1980); thus, differences
in these studies may be due to differences in secular trends rather than to
changes in potential exposure to 2,3,7,8-TCDD. The evidence from all of these
studies, therefore, is open to question.
The reports with more detailed exposure information include those
examining paternal exposure. Kimmel has reviewed the animal literature in
Appendix C and reported that 2,3,7,8-TCDD resulted in testicular atrophy and
reduced sperm count, but .a dominant lethal study found no effects (Kimmel,
1988). The animal data are too limited, at this time, to suggest the presence
or absence of an effect in the offspring in studies with human male exposure.
The large differences in time frame between male exposure and conception in the
epidemiologic studies probably reduces the chance of identifying any paternally
mediated effect that is potentially occurring. However, new data on the half-
life of 2,3,7,8-TCDD in humans (CDC, 1987) suggests some exposure may be
occurring many years later, depending on the level of initial exposure.
19
-------
Reanalyses of occupational studies, with stricter exposure definitions, might
yield useful information.
The only studies that address the possibility of female exposure are the
ecologic/environmental studies and the studies of Seveso, In his review of the
animal'literature, Kimmel (1988) found evidence of reproductive and
developmental effects of dioxin. The human studies have less informative
exposure data, and so are less useful in drawing conclusions concerning the
reproductive effects of 2,3,7,8-TCDD. Additionally, in both types of studies,
males may also be exposed, thus clouding the interpretation of such data.
20
-------
REFERENCES
Bisanti, L.; Bonetti, F.; Caramaschi, F.; Del Corno, G.; Favaretti, C.;
Giambelluca, S.; Marni, E.; Montesarchio, E.; Puccinelli, V.; Remotti, G.;
Volpato, C.; Zambrelli, E. (1979) Experience of the accident of Seveso.
Proceedings of the 6th European Teratology Conference, Akademiai Kiado.
Pub.
Centers for Disease Control (CDC). (1987) Serum dioxin in Vietnam-era
veterans - preliminary report. Morbidity and Mortality Weekly Report.
36(28):470-475.
Constable, J.D.; Hatch, M.C. (1985) Reproductive effects of herbicide
exposure in Vietnam: recent studies by the Vietnamese and others.
Teratogenesis Carcinog. Mutagen. 5(4):231-250.
Donovan, J.W.; Adena, M.A.; Rose, G.; Batistutta, D. (1983) Case-control
study of congenital anomalies and Vietnam service. Canberra: Australian
Government Publishing Services. (As described in Hatch and Stein, 1986).
Erickson, J.D.; Mulinare, J.; McClain, P.W.; Fitch, T.G.; James, L.M.;
McClearn, A.B.; Adams, M.J. (1984) Vietnam veterans' risks for fathering
babies with birth defects. JAMA 252(7):903-912.
Field, B.; Kerr, C. (1979) Herbicide use and incidence of neural-tube
defects. Lancet 1(8130):1341-1342.
Friedman, J.M. (1984) Does agent orange cause birth defects? , Teratology
29:193-221.
Hanify, J.A.; Metcalf, P.; Nobbs, C.L.; Worsley, R.J. (1981) Aerial spraying
of 2,4,5-T and human birth malformations: an epidemiological investiga-
tion. Science 212:349-351.
Hatch, M.C. (1984) Reproductive effects of the dioxins. In: Lowrance, W.W.,
ed. Public health risks of the dioxins. Proceedings of a symposium held
at The Rockefeller University in New York City, Oct. 19-20, 1983. Los
Altos, CA: William Kaufmann.
Hatch, M.C.; Stein, Z.A. (1986) Agent orange and risks to reproduction: the
limits of epidemiology. Teratogenesis Carcinog. Mutagen. 6(3):185-202.
Homberger, E.; Reggiani, G.; Sambeth, J.; Wipf, H. .(1979) The Seveso
accident: its nature, extent and consequences. Report from Givaudan
Research Company Ltd. and F. Hoffmann-La Roche & Co. Ltd.
Kimbrough, R.D.; Falk, H.; Stehr, P.; Fries, G. (1984) Health implications of
2,3,7,8-tetrachlorodibenzodioxin (TCDD) contamination of residential soil.
J. Toxicol. Environ. Health 14(l):47-93.
21
-------
Kimmel, G.L. (1988) Reproductive and developmental toxicity of 2,3,7,8-TCDD.
In: A cancer risk-specific dose estimate for 2,3,7,8-TCDD, Appendix C.
External Review Draft. U.S. Environmental Protection Agency, Washington,
DC.
Lathrop, 6.D.; Wolfe, W.H.; Albanese, R.A.; Moynahan, P.M. (1984) An
epidemiologic investigation of health effects in air force personnel
following exposure to herbicides: baseline morbidity study results.
USAF School of Aerospace Medicine, Brooks Air Force Base, Texas.
Moses, M.; Lillis, R.; Crow, K.D.; Thornton, J.; Fischbein, A.; Anderson, H.A.;
Selikoff, I.J. (1984) Health status of workers with past exposure to
2,3,7,8-tetrachlorodibenzo-j)-dioxin in the manufacture of 2,4,5-
trichlorophenoxyacetic acid: comparison of findings with and without
chloracne. Am. J. Ind. Med. 5:161-182.
Nelson, C.J.; Holson, J.F.; Green, H.G.; Gaylor, D.W. (1979) Retrospective
study of the relationship between agricultural use of 2,4,5-T and cleft
palate occurrence in Arkansas. Teratology 19:377-384.
Patterson, D.G.; Hoffman, R.E.; Needham, L.L.; Roberts, D.W.; Bagby, J.R.;
Pirkle, J.L.; Falk, H.; Sampson, E.J.; Houk, V.N. (1986) 2,3,7,8-Tetra-
chlorodibenzo-p-dioxin levels in adipose tissues of exposed and control
persons in Missouri: an interim report. JAMA 256(19):2683-2686.
Pocchiari, F. (1980) Accidental TCDD contamination in Seveso (Italy):
epidemiological aspects. FIFRA Docket No. 415, Exhibit 1469. U.S.
Environmental Protection Agency, Washington, DC.
Pocchiari, F.; Silano, V.; Zampieri, A. (1980) Human health effects from
accidental release of TCDD at Seveso (Italy). FIFRA Docket No. 415,
Exhibit 1470. U.S. Environmental Protection Agency, Washington, DC.
Rappe, C. (1985) Problems in analysis of PCDDs and PCDFs and presence of
these compounds in human milk'. , Consultation on organohalogen compounds in
human milk and related hazards. Geneva: World Health Organization.
Reggiani, G. (1978) Medical problems raised by the TCDD contamination in
Seveso, Italy. Arch. Toxicol. 40:161-188.
Reggiani, G. (1980) Direct testimony before the U.S. Environmental Protection
Agency. FIFRA Docket No. 415, Exhibit 861. U.S. EPA, Washington, DC.
Rehder, H.; Sanchioni, L.; Cefis, F.; Gropp, A. (1978) Pathological-
embryological investigations in cases of abortion related to the Seveso
accident. Journal of Swiss Medicine 108(42):1617-1625.
Schecter, A.; Ryan, J.J.; Constable, J.D. (1987) Polychlorinated dibenzo-£-
dioxin and poiychlorinated dibenzofuran levels in human breast milk from
Vietnam compared with cow's milk and human breast milk from the North
American continent. Chemosphere 16(8/9):2003-2016.
22
-------
Smith, A.H.; Fisher, D.O.; Pearce, N.; Chapman, C.J. (1982) Congenital
defects and miscarriages among New Zealand 2,4,5-T sprayers. Arch.
Environ. Health 37(4):197-200.
Thomas, H.F. (1980) 2,4,5-T use and congenital malformation rates in Hungary.
Lancet ii:214-5.
Townsend, J.C.; Bodner, K.M.; Van Peenen, P.P.; Olsen, R.D.; Cook, R.R. (1982)
Survey of reproductive events of wives of employees exposed to chlorinated
dioxins. Am. J. Epidemiol. 115(5):695-713.
Tuchmann-Duplessis, H. (1977) Embryo problems posed by the Seveso accident.
Le Concours Medical No. 44.
Tuchmann-Duplessis, H. (1980a) Direct testimony before the U.S. Environmental
Protection Agency. FIFRA Docket No. 415, Exhibit 864. U.S. EPA,
Washington, DC.
Tuchmann-Duplessis, H. (1980b) Tables in direct testimony before the U.S.
Environmental Protection Agency. FIFRA Docket No. 415, Exhibit 864a.
U.S. EPA, Washington, DC.
U.S. Environmental Protection Agency. (1985) Health assessment document for
polychlorinated dibenzo-p_-dioxins. Office of Health and Environmental
Assessment, Washington, DC. EPA/600/8-84/014F. NTIS PB86-122546/AS.
U.S. Environmental Protection Agency. (1988) Estimating exposures to 2,3,7,8-
TCDD. Exposure Assessment Group, Office of Health and Environmental
Assessment, Washington, DC. External Review Draft. EPA/600/6-88/005A.
van den Berg, M.; van der Wielen, F.W.M.; Olle, K.; Van Boxtel, C.J. (1986)
The presence of PCDDs and PCDFs in human breast milk from the Netherlands.
Chemosphere 15(6):693-706.
23
-------
-------
March 1988
Review Draft
APPENDIX E
IMMUNOTOXICITY OF 2,3,7,8-TCDD: REVIEW, ISSUES, AND UNCERTAINTIES
Babasaheb R. Sonawane, Ph.D.
Reproductive Effects Assessment Group
Office of Health and Environmental Assessment
Ralph J. Smialowicz, Ph.D.
Health Effects Research Laboratory
Office of Health Research
Robert W. Luebke, Ph.D.
Health Effects Research Laboratory
Office of Health Research
Office o.f Research and Development
U.S. Environmental Protection Agency
-------
-------
EXECUTIVE SUMMARY - - . ,
2,3,7,8-Tetrachlorodibenzo-jp_-dioxin (2,3,7,8-TCDD) is a potent immuno-
suppressant in the laboratory animal species studied; however, the immunologi-
cal effects are apparent at exposure levels that also produce other discernible
pathological lesions and reproductive/developmental effects. There are no
unequivocal cases of significant immune function alterations in humans following
exposure to 2,3,7,8-TCDD. The cellular and molecular mechanism(s) of 2,3,7,8-
TCDD-induced immunotoxicity is unknown. Significant data gaps and uncertainties
exist to prevent an immunotoxicity-based health hazard evaluation.
INTRODUCTION
This document discusses the relevant scientific literature on the immuno-
toxicity of 2,3,7,8-tetrachlorodibenzo-j3-dioxin (2,3,7,8-TCDD) in laboratory
animals and humans. The document does not provide a comprehensive literature
review of the 2,3,7,8-TCDD-induced immune effects; however, attempts were made
to identify and discuss critical studies and issues, strengths and weaknesses
of the data, and significant uncertainties associated with the evaluation and
interpretation of the data. Furthermore', significant data gaps in knowledge
are recognized as they may relate to potential risk to the immune system of
humans upon exposure to 2,3,7,8-TCDD.
The immunotoxic effects of 2,3,7,8-TCDD have been studied by numerous
investigators for over a decade. Several recent reviews have been published
that provide a very good overview of the animal and human data dealing with
the immune alterations following exposure to 2,3,7,8-TCDD (Dean and Lauer, 1984;
Dean and Kimbrough, 1986; Thomas and Faith, 1985). The U.S. Environmental
1
-------
Protection Agency also reviewed the immunoloyical effects of 2,3,7,8-TCDD in
the Health Assessment Document for Polychlorinated Dibenzo-j3-Dioxins (U.S. EPA,
1985). The present review updates the immunotoxicity literature on 2,3,7,8-
TCDD and presents issues of concern and uncertainties in risk assessment.
ANIMAL STUDIES
Early experimental work with 2,3,7,8-TCDD revealed that the thymus is a
target organ of toxicity. All animal species that have been studied have con-
sistently displayed involution of the thymus and loss of cortical thymocytes
following acute or chronic exposure to 2,3,7,8-TCDD, as well as lymphocyte
depletion of T-cell areas of the lymph nodes and the spleen (Harris et al.,
1973; Zinkl et al., 1973; Gupta et al., 1973; Vos et al., 1973; 1974; Luster et
al., 1979; 1982). While these changes in the lymphoid tissues were similar to
those produced by glucocorticosteroids, adrenalectomy failed to prevent 2,3,7,8-
TCDD-induced thymic atrophy or hepatotoxicity (van Logten et al., 1980). The
administration of thymosin to mice exposed to 2,3,7,8-TCDD by postnatal maternal
treatment did not affect thymus atrophy or decrease mitogen responses (Vos et
al., 1978). Thyroidectomy, however, was found to protect rats from the immuno-
suppressive effects of 2,3,7,8-TCDD (Pazdernik and Rozman, 1984). Vos and
co-workers (1973) reported the effects of 2,3,7,8-TCDD on the immune system of
guinea pigs, rats, and mice. Several tests were employed in these three species,
and the authors concluded that the effect of 2,3,7,8-TCDD was primarily on the
cell-mediated immune function of these animals. Guinea pigs treated with
1 ug/kg of 2,3,7,8-TCDD per week either died or were moribund after four doses.
At this dose and also at lower doses (0.008, 0.04, and 0.2 ug/kg), guinea pigs
showed severe loss of body weight, depletion of the lymphoid organs, particularly
-------
the thymus, and lymphopenia. At sublethal doses, the effects were primarily on
the thymus. The weights of other lymphoid organs, such as the spleen, cervical
lymph nodes, and popliteal lymph nodes, were not affected. Adrenals showed a
slight enlargement at the 0.2 ug/kg dose level. Serum cortisol and corticoste-
rone values were not influenced by the 2,3,7,8-TCDD treatment. No microscopic
lesions were apparent in any of the lymphoid organs upon dosing with up to
0.2 ug/kg of 2,3,7,8-TCDD per week other than atrophy of the thymic cortex (Vos
et al., 1973).
In selected experiments, the treated and control animals were sensitized
to tetanus toxoid or killed Mycobacterium tuberculosis. Seven days after the
tetanus toxoid injection, there was a small but significant increase in the
serum, antitoxin levels at 0.008 and 0.04 ug/kg dose levels. The animals were •
challenged by a second dose of tetanus toxoid 2 weeks after'the first challenge
to evaluate the secondary response. One and 2 weeks after the second dose, the
guinea pigs treated with 0.2 ug/kg of 2,3,7,8-TCDD per week had significantly
lower tetanus antitoxin titers. The animals challenged with IM. tuberculosis
were tested for their skin reactivity 12 and 19 days later by intradermal
tuberculin injections. Both skin thickness and the diameter of the reaction
site were significantly decreased, the former being significant at the
0.04 ug/kg dose level (Z.inkl et al., 1973; Vos et al., 1973).
Female albino rats were treated orally with 0.2, 1, and 5 ug/kg of
2,3,7,8-TCDD once a week for 6 weeks. The body weights and thymus weights were
reduced at the highest dose level and adrenal weights were reduced at the ,1 and
5 ug/kg dose level. The relative thymus weight (organ/body weight ratio) was
approximately half that of the control animals at the 5 ug/kg dose level,
whereas at the same dose level, the relative spleen weights were significantly
increased. The total peripheral blood leukocyte count and lymphocyte numbers
-------
did not show a significant 2,3,7,8-TCDD-related effect as observed in guinea
pigs. Skin reactivity of rats challenged to tuberculin was not affected at any
2,3,7,8-TCDD dose levels (Zinkl et al., 1973; Vos et al., 1973).
Graft versus host (GVH) activity of mice was evaluated after 0.2, 1, 5,
and 25 ug/kg of 2,3,7,8-TCDD treatment per week for 4 weeks. The absolute
and relative thymus weights were decreased in groups treated with 5 ug/kg of
2,3,7,8-TCDD. Dose-related reduction in GVH activity was observed (Vos et al.,
1973).
Mice treated orally with 2,3,7,8-TCDD were challenged with either Salmonel-
la bern or Herpesvirus suis (Thigpen et al., 1975). In two separate experiments
the animals were exposed to 2,3,7,8-TCDD levels ranging from 0.5 to 20 ug/kg
once every week for 4 weeks. The animals were challenged with the infectious
agents one day after the fourth dose of 2,3,7,8-TCDD. Treatment of mice with
2,3,7,8-TCDD reduced the time of death after the bacterial challenge at the
5 ug/kg dose level whereas the increased mortality was obvious at 1 ug/kg. The
challenge with H. suis influenced neither the period of time to death nor the
mortality rate.
Vos and co-workers (1978) reported the response of young mice .to
Escherichia coli endotoxin after 2,3,7,8-TCDD treatment. There was a dose-
related increase in mortality, both with respect to 2,3,7,8-TCDD and the endo-
toxin. While the administration of 250 ug of endotoxin produced no mortality
in the control group, as little as 10 ug of endotoxin produced death in mice
treated with 15 or 50 ug 2,3,7,8-TCDD. A similar increase in endotoxin sensi-
tivity was reported when mice were treated with a single oral dose of 100 ug/kg
of 2,3,7,8-TCDD and challenged with 20 ug of endotoxin. In these experiments,
a 2,3,7,8-TCDD dose-related decrease in body, thymus, and spleen weights was
observed.
-------
The increased susceptibility to S_. bern reported in mice exposed to
2,3,7,8-TCDD (Thigpen et al., 1975) has been postulated to be due to an in-
creased sensitivity of 2,3,7,8-TCDD-exposed mice to bacterial endotoxin (Vos
et al., 1978), although variable effects have been reported in host resistance
following 2,3,7,8-TCDD exposure (Dean and Lauer, 1984; Vos et.al., 1978; Luster
et al., 1980). Other studies have shown a correlation between increased mortal-
ity from bacterial or parasitic infections and decreased serum complement levels
(White et al., 1986) or decreased B-cell mediated responses (Tucker et al., 1986)
in 2,3,7,8-TCDD-exposed mice, respectively.
Clark et al., (1981) have shown that low doses of 2,3,7,8-TCDD (4 ng/kg to
0.4 ug/kg) administered intraperitoneally once a week for 4 weeks suppress the
generation of cytotoxic T-cells by lymph node and spleen cells in male C57BL/6
mice but have little effect on delayed hypersensitivity, antibody responses, or
thymic cellularity. Suppressor cells capable of blocking the cytotoxic T-cell
response were found in the thymus of 2,3,7,8-TCDO-treated mice following cumula-
tive doses of 2,3,7,8-TCDD as low as 4 ng/kg. The immunosuppressive effects
observed at 4 ng/kg were significantly less than the 4 ug/kg required to induce
thymic atrophy. In 1983, the same investigators (Clark et al., 1983) reported
that susceptibility to 2,3,7,8-TCDD-induced immunosuppression in mice is strain-
dependent, and occurs at doses that have little effect on hepatic microsomal
enzymes. Clark et al. also reported that other types of haloaromatics can
induce similar immunosuppression provided they possess sufficient binding
affinity for the genetically controlled 2,3,7,8-TCDD-receptor protein. These
observations suggest a receptor-dependent mechanism for the stimulation of
suppressor T-cell activity by haloaromatic hydrocarbons.
Examination of T-cell-mediated responses in 2,3,7,8-TCDD-exposed animals
has generally demonstrated a correlation between thymic atrophy and impaired
-------
cell-mediated immunity. Exposure of mice, rats, and/or guinea pigs to
2,3,7,8-TCDD has been reported to result in depressed delayed hypersensitivity
(Vos et al., 1973; Faith and Moore, 1977; Clark et al., 1981), prolonged allo-
graft rejection (Vos and Moore, 1974), depressed GVH response (Vos and Moore
1974), decreased responses to T-cell mitogens and/or allogeneic cells in vitro
(Dean and Lauer, 1984; Vos and Moore, 1974; Luster et al., 1980), and decreased
generation of cytotoxic T lymphocytes (CTL) (Clark et al., 1981, 1983).
Depressed CTL activity in adult mice following exposure to 2,3,7,8-TCDD has
been reported to be associated with an increase in suppressor T-cells, but not
associated with a reduction in the frequency of CTL precursors (Clark et al.,
1981; 1983; Nagarkatti et al., 1984).
Nonspecific immune responses affected by natural killer cells (Dean and
Lauer, 1984; Mantovani et al., 1980) and jnacrophages (Dean and Lauer, 1984; Vos
et al., 1978; Mantovani et al., 1980) have not been observed to be affected by
2,3,7,8-TCDD exposure. Both of these cell types possess tumoricidal ,
bactericidal, and virucidal activity.
2,3,7,8-TCDD has also been reported to affect bone marrow and humoral
immune responses of experimental animals. Exposure of mice to 2,3,7,8-TCDD
resulted in myelotoxicity and suppression of bone marrow progenitor cells
(Tucker et al., 1986; Luster et al., 1985; Chastain and Pazdernik, 1985).
Studies in mice exposed to 0, 1.0, 5.0, or 15 ug/kg body weight of 2,3,7,8-TCDD
pre- and postnatally by maternal dosing indicated that both 5 and lb ug/kg
dosage groups had a significant reduction in bone marrow cellularity, (colony-
forming unit-spleen (CFU-S) or pluripotent stem cells and colony-forming unit-
granulocyte/macrophage (CFU-GM). He'matology profiles and blood smears
revealed a normocytic anemia in; these mice (Luster et al., 1980). Bone marrow
toxicity correlated with depressed immunologic and host-resistance responses.
-------
The hematopoietic stem cells have a limited renewal capacity, and damage to
these cells can induce a permanent decrease in their pro!iferative capacity;
however, this limited evidence of bone marrow toxicity, as observed in mice,
is difficult to evaluate due to the lack of additional significant studies in
other species or humans.
The antibody response of 2,3,7,8-TCDD-exposed mice, as measured by the
plaque-forming cell (PFC) assay, was suppressed following immunization with
T-cell dependent and/or T-cell independent antigens (Tucker et al., 1986;
Vecchi et al., 1983; Holsapple et al., 1986). Furthermore, these effects on
B-cell-mediated responses were observed to occur at 2,3,7,8-TCDD doses below
those that cause thyrnic atrophy (van Logten et al., 1980; Tucker et al., 1986;
Luster et al., 1985; Vecchi et al., 1983; Holsapple et al., 1986).
Susceptibility to suppression of humoral immune responses and 2,3,7,8-TCDD
induction of the aryl hydrocarbon hydroxylase (AHH) system were found to corre-
late (Luster et al., 1980; Tucker et al., 1986; Chastain and Pazdernik, 1985;
Vecchi et al., 1983; Holsapple et al., 1986), as does thymic atrophy (Poland and
Glover, 1980), T-cel1-mediated responses (Nagarkatti et al., 1984), and serum
complement levels (White et al., 1986). A good correlation between the degree
of AHH inducibility and immunosuppression was observed in Fj crosses of
"responsive" (Ahb/Ahb) and "nonresponsive" (AhdAhd) mice (Nagarkatti et al.,
1984; Vecchi et al., 1983). These results, taken together, indicate that there
is a strong association between the presence of the Ah receptor and the induc-
tion of immune effects following 2,3,7,8-TCDD exposure in experimental animals,
and susceptibility to 2,3,7,8-TCDD-induced immune effects in the mouse may be
under genetic influences.
The immunological reactivity of young or/suckling rats and mice has been
evaluated using a variety of experimental protocols .employing prenatal or post-
-------
natal exposure of mothers or a continued prenatal through postnatal exposure
with 2,3,7,8-TCDD. One such study reported by Vos and Moore (1974) is summa-
rized below. Pregnant rats were treated orally with 1 or 5 ug/kg body weight
of 2,3,7,8-TCDD on day 11 and 18 of gestation. The weights of 1-day-old pups
from mothers treated with 5 ug/kg dose level were significantly reduced. At
this dose level, reduced weights of the thymus and spleen were also observed.
Most of the pups in the high-treatment group died within 25 days after birth.
The newborn animals from mothers given a 1 ug/kg dose level were further
treated with the same dose of 2,3,7,8-TCDD administered to mothers on day 4,
11, and 18. Additional groups of suckling rats were exposed to 2,3,7,8-TCDD
postnatally by oral treatment of mothers with 5 ug/kg. Reduction in body,
thymus, spleen, and adrenal weights were observed in different groups at 25
days of age. Splenic lymphocytes of rats exposed postnatally to the 5 ug/kg
level showed a decrease in phytohemagglutinin (PHA)-induced DNA synthesis.
This effect was not observed in animals exposed pre- and postnatally to 1
ug/kg of 2,3,7,8-TCDD. Thymocytes cultured from postnatally exposed male rats
to 5 ug/kg of 2,3,7,8-TCDD showed a significant reduction of thymidine incor-
poration in the presence of PHA. DNA synthesis in response to concanavaltn A
(Con-A), however, was not reduced in thymocytes. The authors (Vos and Moore,
1974) reported the insensitivity of 4-month-old mice to 2,3,7,8-TCDD treatment.
One-month-old mice treated with four weekly doses of 25 ug/kg of 2,3,7,8-TCDD
showed a decreased responsiveness of their splenic lymphocytes to PHA, whereas
5-month-old animals failed to show this effect after six weekly doses.
Similar effects were observed on GVH activity of spleen cells from 25-day-
old pre- and/or postnatally exposed rats. Only those animals treated with 5
ug/kg postnatally showed a significant decrease in GVH activity of spleen cells,
Reaction times of heterologous skin grafts was prolonged in rats exposed in
-------
utero and in mice exposed pre- and postnatally to 2,3,7,8-TCDD (Vos and Moore,
1974).
In summary, exposure of mice, rats, and/or guinea pigs (as described
earlier) to 2,3,7,8-TCDD during the perinatal period resulted in thymic atrophy
(Faith and Moore, 1977; Vos and Moore, 1974; Luster et al., 1980; Moore and
Faith, 1976'; Faith et al., 1978). These animals exhibited a "wasting syndrome"
that is similar to that seen in neonatally thymectomized animals that have been
treated with corticosteroids (Thomas and Faith, 1985). 2,3,7,8-TCDD adminis-
tered during immune ontogenesis has been observed to affect immune responses
more dramatically than when administered to adults. Generally, thymic atrophy,
suppressed T-cell mediated responses, and bone marrow toxicity have been report-
ed to be more profound following pre- and/or postnatal exposure to 2,3,7,8-TCDD
than with adult exposure (Dean and Lauer, 1984; Dean and Kimbrough, 1986;
Thomas and Faith, 1985; Luster et al., 1979; 1980; Faith and Moore, 1977; Vos
and Moore, 1974). These results suggest that the developing immune system is
more susceptible to 2,3,7,8-TCDD-induced alterations and, consequently, that
the very young may be at a higher risk than adults to the immunotoxic effects
of 2,3,7,8-TCDD.
Exposure of animals to 2,3,7,8-TCDD has been shown to decrease the respon-
siveness of lymphocytes to various mitogens in culture. For example, rabbits
exposed for 8 weeks to 2,3,7,8-TCDD at 0.01 to 10 ug/kg/week and challenged
with tetanus toxoid had a decreased PHA response at the highest dose (Sharma
et al., 1979). Sharma et al. (1979) also reported that immunologic effects of
a single dose of 2,3,7,8-TCDD in mice were reversible during an 8-week period
while the hepatic lesions persisted. Mice were treated with a single oral
dose of 10 ug/kg of 2,3,7,8-TCDD, and selected animals were sacrificed at 2, 4,
and 8 weeks after this dosing. The spontaneous increase of DNA synthesis in
-------
splenic cultures and a decreased responsiveness of splenic lymphocytes to phy-
tohemagglutinin and pokeweed mitogen was apparent at 2 weeks after the treat-
ment. The effects persisted up to 4 weeks after the administration of
2,3,7,8-TCDD, but were not noticed when the spleens were obtained from animals
at 8 weeks after the 2,3,7,8-TCDD dosing. Lymphocyte depletion in the thymus
and reduced thymus weights showed a partial recovery at this time.
The influence of direct addition of 2,3,7,8-TCDD to mouse-splenic cultures
was reported by Sharma and Gehring (1979). 2,3,7,8-TCDD decreased the unstimu-
1 ated DNA synthesis in lymphocytes at concentrations as low as 10~9 M; however,
no effects were observed in mitogen-stimulated cultures. Vos and Moore (1974)
reported that the addition of 2,3,7,8-TCDD up to 0.01 ug/0.5 mL in culture
medium did not alter the DNA synthesis in mouse spleen or rat thymus cells,
either with or without the presence of phytohemagglutinin or Con-A. Luster and
co-workers (1979) exposed the spleens from mice to 2,3,7,8-TCDD in dimethylsul-
foxide. DNA, RNA, and protein synthesis in the spleens were inhibited at
2,3,7,8-TCDD concentrations of 10~7 M. The ability of lymphocytes to bind with
mitogens was not influenced by 2,3,7,8-TCDD.
The mechanism(s) for 2,3,7,8-TCDD-induced immunosuppression is not fully
known. However, genetic and structure-activity relationship studies have
provided evidence that 2,3,7,8-TCDD-induced immunosuppression in mice is asso-
ciated with the presence of the Ah locus and is mediated through a 2,3,7,8-TCDD
cytosol receptor (Nagarkatti et al., 1984; White et al., 1986; Tucker et al.,
1986; Luster et al., 1985; Vecchi et al., 1983; Holsapple et al., 1986; Poland •
and Glover, 1980; Dencker et al., 1985; Kouri and Ratrie, 1975). Some evidence
suggests that thymic epithelial cells may be the principal target for 2,3,7,8-
TCDD-induced immunotoxicity (Clark et al., 1983; Greenlee et al., 1985). Bind-
ing of 2,3,7,8-TCDD to receptors in the thymus may promote altered T-cell matu-
10
-------
ration and differentiation, and it may be the molecular basis for the observed
thymic atrophy and immunotoxicity. Other work, however, suggests that hemato-
poietic stem cells and B-cells may also be targets for 2,3,7,8-TCDD-mediated
immune effects. For example, it has recently been shown that 2,3,7,8-TCDD selec-
tively inhibits the differentiation of B-cells into antibody-secreting cells in
vitro (Tucker et al., 1986) and inhibits bone marrow stem cell colony growth in
vivo and in vitro (Luster et al., 1985). The binding of 2,3,7,8-TCDD to recep-
tors on lymphocytes, thymus epithelial cells, and/or hematopoietic precursor
cells may cause alterations in maturation and differentiation that may result
in the immune alterations observed in animals following in vivo exposure to
2,3,7,8-TCDD, Further work is clearly needed to investigate the mechanism(s)
for 2,3,7,8-TCDD-induced immune effects in order to make better estimates of
the potential risks associated with exposure of humans to 2,3,7,8-TCDD.
HUMAN STUDIES '..'-.
For a variety of reasons, humans have been exposed to 2,3,7,8-TCDD in the
environment. The immune function has been examined in individuals with pro-
bable or known exposure to 2,3,7,8-TCDD using assays designed to evaluate the
component parts of the immune system. Several accounts of immune functions in
humans exposed to 2,3,7,8-TCDD have been referred to in summary-type articles
reviewed in the preparation of this paper. Some of these reports have not
been published in the scientific literature or are anecdotal. However, none
of these reports, in the opinion of the reviewers (Dean and Lauer, 1984; Dean
and Kimbrough, 1986; Marshall, 1986), presented convincing evidence for altered
immune function in the exposed populations. In one study, no abnormalities in
measured immune parameters were observed in military personnel who had been
11
-------
involved in the spraying of 2,3,7,8-TCDD-contaminated Agent Orange during
operation Ranch Hand in Vietnam (conversation between J. Silva, CHHS, Test and
Evaluation Activity, Bethesda, MD, and Dr. Ralph Smialowicz, U.S. Environmental
Protection Agency, February 6, 1987).
Attempts have been made to study immune functions in populations that have
been exposed to 2,3,7,8-TCDD. All of these attempts have been complicated by
technical difficulties in both design and execution, in spite of efforts by
the researchers to control for all possible variables. Three such studies are
summarized.
In July of 1976, an uncontrolled chemical reaction at a chemical plant in
Seveso, Italy, resulted in the release of an estimated 300 g of 2,3,7,8-TCDD
mainly into an uninhabited area. However, a significant number of people were
potentially exposed, and thus, a major health effects surveillance effort was
initiated. The results of this study, including immunologic assessment, have
been published (Homberger et al., 1979; Reggiani, 1980). A group of 45
2,3,7,8-TCDD-exposed children, 20 of whom had chloracne, and 44 children without
2,3,7,8-TCDD exposure were evaluated immunologically every 4 months for approxi-
mately 1.5 years. These studies revealed no differences between the two groups
in serum immunoglobulin or complement levels or in the ability of their T- and
B-cells to respond to mitogens in vitro. Critical evaluation of these data are
not possible, however, since quantitative data were not presented. It should
be noted that a review (Tognoni and Bonaccorsi, 1982) of the Seveso incident,
published 5 years after the fact, reported increased serum complement hemolytic
activity and significantly higher lymphocyte proliferative,responses in exposed
subjects. However, no quantitative data were presented, the patient population
was not identified per se, and no reference was made as to how or when these
data were collected. Thus, critical review and interpretation of the reported
12
-------
findings are both impossible and unadvisable.
The town of Times Beach, Missouri, was contaminated with 2,3,7,8-TCDD in
1972 and 1973 when waste from a chemical plant was sprayed on roadways to
control dust. Soil samples were tested for 2,3,7,8-TCDD 10 years later, and
levels of contamination were so great that the entire town was purchased under
the provisions of the Superfund law (Powell, 1984). Selected residents were
subsequently classified as having had high or low risk of 2,3,7,8-TCDD exposure
and were evaluated immunologically (Knutsen, 1984). Tests of delayed type
hypersensitivity (DTH) to a standard battery of skin test antigens, lymphocyte
blastogenic responses to mitogens, and T-cell subset analysis revealed no sig-
nificant differences between the high- and low-risk groups. However, there was
a tendency among members of the high-risk group to respond to fewer skin test
antigens and to have slightly different T-cell subset profiles than low-risk
group members. Lymphocytes from children'in the high-risk group also had a
decreased proliferative response to tetanus toxoid compared with those from
children in the low-risk group. It should be noted, however, that in a prelim-
inary study of an unspecified population of Missouri residents exposed to
2,3,7,8-TCDD, no differences were detected in T-cell subset profiles, skin test
responses, or lymphocyte proliferation responses (Stehr et al., 1986).
Residents of the Quail Run Mobile Home Park in Gray Summit, Missouri, were
exposed for various lengths of time to 2,3,7,8-TCDD-contaminated soil. Contami-
nation was the result of dust control efforts using waste oil containing chemi-
cal, sludge from the same source as in Times Beach. Soil sampling studies
revealed levels of 2,3,7,8-TCDD ranging from 39 to 2,200 parts per billion.
Immunologic testing was performed on residents who had lived in the park for
at least 6 months (N = 154), and their results were compared to residents
(N = 155) of other mobile parks in areas where no evidence of 2,3,7,8-TCDD
13
-------
contamination was found by soil sampling (Hoffman et al., 1986). DTH to skin
test antigens, antigen- and mitogen-stimulated lymphocyte proliferation
responses, lymphocyte subset analysis, cytotoxic T-lymphocyte responses, and
serum immunoglobulin levels were evaluated in both populations, Some members
of the high-risk group did not respond to skin test antigens (anergy) and, of
those that did respond, positive reactions were obtained for fewer antigens
than in the unexposed group (relative anergy). The exposed group had an in-
creased frequency of anergy (11.8% vs. 1.1%) and relative anergy (35.3% vs.
11.8%). It must be pointed out that there were significant technical problems
with the interpretation of the skin test responses and, as a'result, nearly 50%
of the data had to be discarded. Several skin test readers, with various
amounts of experience, were employed in this investigation. In addition, two
of the four readers who had received special training in the interpretation of
DTH skin tests recorded anergy in 15% or 40% of the control group, a rate 75 or
200 times the expected rate. Although allowances were made for'this by the
investigators, the DTH data in this study are thus questionable. The mean
ratio of T helper/inducer cells to T suppressor/cytotoxic cells was similar in
both groups, although there was a nonsignificant proportion of exposed group
members with a T4/T8 ratio less than 1.0 (8.1% vs. 6.4%). The report likewise
states that 12.6% of the exposed group versus 8.5% of the low-risk group had
abnormal in vitro T-cell functions, although examination of the tabular data
provides no evidence whatsoever for a difference between levels of immunocom-
petence in the two groups. The authors concluded that long-term exposure to
2,3,7,8-TCDD is associated with depressed cell-mediated immunity, although the
effects have not resulted in an excess of clinical illness. Individuals from
this initial study (Hoffman et al,, 1986) have been re-evaluated, and the
results of this follow-up study, as reported by Evans et al. (1987) at the
14
-------
International Conference on Dioxin, failed to corroborate the report of anergy
in the 2,3,7,8-TCDD exposed cohort of the initial study (Hoffman et al., 1986).
UNCERTAINTIES, DATA GAPS, AND RESEARCH NEEDS
The immune system, unlike many organ systems, is self-renewing from a
pool of pluripotent stem cells. Functional cells are generally end-stage
cells that have a limited lifetime and are replaced on a regular basis.
Therefore, unless the store of stem cells is destroyed or unless there is a
permanent blockade of cellular differentiation, acute chemical exposure is
unlikely to cause long-term suppression of the host-defense system. However,
the uncertainty in the premise of self-limited chemical immunotoxicity resides
in the unknown effects of chronic or repeated acute, exposure to immunotoxic
agents on the regenerative capacity of the immune system.
Although a great amount of time and effort has gone into evaluating the
immunotoxic effects of 2,3,7,8-TCDD in experimental animals, there,are a
number of uncertainties that remain to be resolved. These include but are
not limited to the following: (1) the lack of a strong association between
2,3,7,8-TCDD exposure and decreased host resistance, (2) the different suscep-
tibility of species and strains to 2,3,7,8-TCDD-induced immunosuppression, (3)
the transient and reversible nature of 2,3,7,8-TCDD-induced immune effects,
(4) the apparent but not proven increased susceptibility of very young animals
to 2,3,7,8-TCDD-induced immune effects, (5) the uncertainty of the mechanism(s)
of 2,3,7,8-TCDD-induced immune effects, and (6) the lack of evidence between
observed immunological effects in animals and the questionable immune altera-
tions reported in human populations inadvertantly exposed to 2,3,7,8-TCDD.
These areas of uncertainty are briefly discussed.
15
-------
Inconsistent results have been reported regarding the susceptibility of
2,3,7,8-TCDD-exposed animals to tumor development and infection (Dean and
Lauer, 1984; Faith and Moore, 1977; Luster et al., 1980; Thigpen et al., 1975;
White et al., 1986; Tucker et al., 1986). Despite the fact that exposure to
2,3,7,8-TCDD results in depressed T- and B-cell responses, further research is
necessary to define the conditions under which host resistance is adversely
affected by 2,3,7,8-TCDD.
In general, the experimental animal data suggest that species differences
exist in 2,3,7,8-TCDD sensitivity. With the exception of the earliest work,
the mouse has been the predominant species for studying the immunotoxic effects
of 2,3,7,8-TCDD. This is most probably due to the fact that more is known
about the immune system of the mouse than any other animal species, as well as
the fact that there are more validated methods available for examining the
immune system in this species. Nevertheless, extensive interspecies comparisons
among several animal species are warranted considering their pharmacokinetic
differences. These studies are needed not only to substantiate and corroborate
the results of studies with the mouse but also to provide the framework to
extrapolate the potential for 2,3,7,8-TCDD to affect the human immune system.
Interspecies studies are also necessary in order to determine if an association
exists between the Ah locus and 2,3,7,8-TCDD-induced immune effects in species
other than the mouse. This includes extension of in vitro studies which have
demonstrated the presence and inducibility of AHH in human lymphoid tissue
(Kouri and Ratrie, 1975). Hopefully, these studies will provide useful infor-
mation about the role that the Ah locus may play in the susceptibility of the
human population to 2,3,7,8-TCDD.
The immune effects that have been observed following 2,3,7,8-TCDD exposure
have, in all cases where it has been examined, returned to normal levels over
16
-------
a period of time after the cessation of exposure to 2,3,7,8-TCDD. This is
probably a result of the plastic nature of the immune system, >as wel 1 as the
fact that chronic dosing with 2,3,7,8-TCDD has not been performed. It is not
clear whether and how repeated doses might affect the immune, system or whether
short-term exposure could result in irreversible effects. Low-dose chronic
studies in animals are needed to determine whether such exposures not only
produce immune alterations but also to determine if long-lasting impairment is
produced. Low-dose chronic studies are important to extrapolate data from the
high dosages used under experimental situations because they are most likely
to mimic the form of human exposure to 2,3,7,8-TCDD in the environment.
Exposure to 2,3,7,8-TCDD during immune system development has been shown
to affect the immune system of experimental animals (Faith and Moore, 1977;
Vos and Moore, 1974; Luster et al., 1980; Faith et al., 1978). The effects
produced following perinatal exposure have been reported to be more profound
than those produced following adult exposure, although in some cases a high
degree of fetal toxicity has been reported (Vos and Moore, 1974). While this
may be true, no attempt has been made to test this hypothesis by making a
direct comparison between perinatal and adult exposure to 2,3,,7,8-TCDD using
identical exposure regimens (i.e., B- and T-cell function, host resistance
models, natural killer cell activity, etc.). Work is warranted in this area
in order to provide evidence of an increased risk of the developing immune
system to 2,3,7,8-TCDD exposure. This is necessary so that an informed judg-
ment can be made as to the potential increased relative risk to infants and
children exposed to 2,3,7,8-TCDD.
There is still a great deal of uncertainty about the mechanism(s) by which
2,3,7,8-TCDD causes immune alterations in animals. Work with other animal
species and in vitro work with human tissues will hopefully provide new insights
17
-------
In this area. A clearer understanding of the mechanisms of 2,3,7,8-TCDD-
induced immunosuppression in animals will be invaluable for extrapolation of
the potential health risk in humans.
Laboratory studies have clearly demonstrated the immunotoxic effects of
2,3,7,8-TCDD in a variety of animal models. Extrapolation of these data to
predict effects of human 2,3,7,8-TCDD exposure are complicated at best, since
actual exposure levels, route of exposure, and even comparability of human and
animal susceptibility to the toxic effects of 2,3,7,8-TCDD are unknown. Results
of rodent studies suggest that the immune responses of children should be more
sensitive to the immunotoxic effects of 2,3,7,8-TCDD than that of adults (Faith
and Moore, 1978). However, immune function was followed in children from the
most heavily contaminated area of Seveso for 18 months after the accident and
appeared to be normal (Homberger et al., 1979; Reggiani, 1980). Furthermore,
baseline data for the immune system of children is not readily available, and
the normal response in children of various ages is not well defined. Further-
more, altered immunocompetence has been reported in 2,3,7,8-TCDD-exposed
residents of Missouri (Knutsen 1984; Hoffman et al., 1986), although the differ-
ences between mean values for control and exposed populations were not statis-
tically significant. Suppression was not detected in functional parameters;
rather, trends in the distribution of exposed group members into "normal" and
"abnormal" response categories were cited as indicative of immune dysfunction.
While these trends may or may not be related to 2,3,7,8-TCDD exposure, there has
been no report of an increase in clinical illness attributable to suppressed
immune function. Thus, it appears to be that there are no unequivocal cases of
significant immunotoxicity in humans following 2,3,7,8-TCDD exposure (Dean and
Lauer, 1984; Dean and Kimbrough, 1986; Marshall, 1986; Evans et al., 1987).
18
-------
The available animal data reviewed above suggest that 2,3,7,8-TCDD-induced
thymus atrophy and immune alterations may result from direct actions on peri-
pheral lymphocytes or progenitor lymphoid cells in the bone marrow and through
altered differentiation of intrathymic precursor cells, specifically by a direct
action on thymic epithelium. The induction of T-cells, cytotoxic for tumor
target cells, was found to be impaired in mouse studies conducted by Clark et
al. (1981) following total exposure to a 2,3,7,8-TCDD dose of 4 ng/kg over a
4-week period. These dosages are below those levels significantly altering
other cell-mediated immune responses (Clark et al., 1981); however, these
results are unconfirmed and questionable. For example, the kinetics of CTL
suppression was not defined in the Clark et al. studies (1981, 1983), and the
effect of 2,3,7,8-TCDD on CTL-mediated tumor resistance remains to be resolved.
Additional concern is raised about Clark's findings (Clark et al., 1981) by
researchers at the Chemical Industry Institute of Toxicology, (CUT) who claim
that the CTL response was not suppressed at a dose of 4 ng/kg (conversation
between Jack Dean, CUT, Research Triangle Park, NC, and Bob Sonawane, U.S.
Environmental Protection Agency, February 17, 1987). The £059 (dose producing
50% maximal response) for the induction of thymic atrophy in sensitive mouse
strains is approximately 10 umol/kg (Poland and Glover, 1980) and for antibody
plaque-forming cell (PFC) suppression varies by 30-fold (1 ug/kg to 30 ug/kg)
between C57BL/6 and DBA/2 mice (Dean and Lauer, 1984). Furthermore, the issue
of host-resistance effects following exposure to 2,3,7,8-TCDD is unresolved.
In summary, it may be inappropriate to derive an immunotoxicity-based
hazard assessment for 2,3,7,8-TCDD from mostly acute and/or subchronic type
studies. A three-generation reproductive study by Murray et al. (1979) demon-
strated the critical noncancer end point for adverse effects at 1 ng/kg/day
compared to questionable immunosuppressive effects observed in mice by Clark
19
-------
et al. (1981) at 4 ng/kg by parental administration. It seems that the level
of concern identified in either of these studies or in chronic bioassays for
carcinogenicity is essentially the same. The magnitude of differences rein-
forces a common conclusion that the biological significance of any immunological
effects observed in laboratory animals is not adequately established to support
its use as the critical end point in hazard evaluation of 2,3,7,8-TCDD exposure
to humans.
20
-------
REFERENCES
Chastain, J.E.; Pazdernik, T.L. (1985) 2,3,7,8-tetrachlorodibenzo-jJ-dioxin
(TCDD)-induced immunotoxicity. Int. J. Immunopharmacol. 7:849-856.
Clark, D.A.; Gauldie, J.; Szewczuk, M.R.; Sweeney, G. (1981) Enhanced sup-
pressor cell activity as a mechanism of immunosuppression by 2,3,7>8-
tetrachloro-dibenzo-jj-dioxin. Proc. Soc. Exp. Biol. Med. 168:290-299.
Clark, D.A.; Sweeney, G.; Safe, S.; Hancock, E.; Kilburn, D.G.; Gauldie, J.
(1983) Cellular and genetic basis for suppression of cytotoxic T-cell
generation by haloaromatic hydrocarbons. Immunopharmacology 6:143-153.
Dean, J.H.; Kimbrough, R.D. (1986) Immunotoxicity of the chlorinated dibenzo-
jj-dioxins and dibenzofurans. Presented at the Dioxin Update Conference;
July 1-2; Alexandria, VA. Office of Pesticides and Toxic Substances, U.S.
Environmental Protection Agency.
Dean, J.H.; Lauer, L.D. (1984) Immunological effects following exposure to
2,3,7,8-tetrachlorodibenzo-j>-dioxin: a review. In: Lowrence, W.W., ed.
Public health risk of the dioxins. Los Altos, CA: W. Kaufmann,
pp. 275-294.
Dencker, L.; Hassoun, E.; Dar-gy, R.; Aim, G. (1985) Fetal thymus organ cul-
ture as an in vitro model for the toxicity of 2,3,7,8-tetrachlorodibenzo-
£-dioxin and its congeners. Mol. Pharmacol. 27:133-140.
Evans, R.G.; Webb, K.B.; Knutsen, A.P.; Roodman, S.T.; Roberts, D.; Bagby, J.;
Garrett, W.A., Jr.; Andrews, J.S. (1987) A medical follow-up of the
health effects of long-term exposure to 2,3,7,8-TCDD. 7th International
Conference on Dioxin; October; Las Vegas, NV, pp. EPOS.
Faith, R.E.; Moore, J.A. (1977) Impairment of thymus-dependent immune func-
tions by exposure of the developing immune system to 2,3,7,8-tetrachloro-
dibenzo-jj-dioxin (TCDD). J. Toxicol. Environ. Health 3:451-464.
Faith, R.E.; Luster, M.I.; Moore, J.A. (1978) Chemical separation of helper
cell function and delayed hypersensitivity responses. Cell Immunol.
40:275-284. . .
Greenlee, W.F.; Dold, K.M.; Irons, R.D.; Osborne, R. (1985) Evidence of direct
action of 2,3,7,8-tetrachlorodibenzo-£-dioxin (TCDD) on thymic epithelium.
Toxicol. Appl. Pharmacol. 79:112-120.
Gupta, B.N.; Vos, J.G.; Moore, J.A.; Zinkl, J.G.; Bullock, B.C. (1973) Patho-
logic effects of 2,3,7,8-tetrachlorodibenzo-p_-dioxin in laboratory animals.
Environ. Health Perspect. 5:125-140.
21
-------
Harris, H.W.; Moore, J.A.; Vos, J.G.; Gupta, B.N. (1973) General biological
effects of TCDD in laboratory animals. Environ. Health Perspect.
5:101-109.
Hoffman, R.E.; Stehr-Green, P.A.; Webb, K.B.; Evans, G.; Knutsen, A.P.; Schramm,
W.F.; Staake, J.L.; Gibson, B.A.; Steinberg, K.K. (1986) Health effects
of long-term exposure to 2,3,7,8-tetrachlorodibenzo-£-dioxin.
JAMA 255:2031-2038.
Holsapple, M.P.; McCay, J.A.; Barnes, D.E. (1986) Immunosuppression without
liver induction by subchronic exposure to 2,3,7,8-tetrachlorodibenzo-jD-
dioxin in adult female B6C3F1 mice. Toxicol. Appl. Pharmacol. 83:445-455.
Homberger, E.; Reggiani, G.; Sambeth, J.; Wipf, H.K. (1979) The Seveso acci-
dent: its nature, extent, and consequences. Ann. Occup. Hyg. 22:327-367.
Knutsen, A.P. (1984) Immunologic effects of TCDD exposure in humans. Bull.
Environ. Health 6:27-43.
Kouri, R.E.; Ratrie, H., III; Atlas, S.A.; Niwa, A.; Nebert, D.W. (1975)
Aryl hydrocarbon hydroxylase induction in human lymphocyte cultures by
2,3,7,8-tetrachlorodibenzo-_p_-dioxin. Life Sci. 15:1585-1595.
Luster, M. I.; Faith, R.E.; Clark, G. (1979) Laboratory studies on the immune
effects of halogenated aromatics. Ann. N.Y. Acad. Sci. 320:473-486.
Luster, M.I.; Boorman, G.A.; Dean, J.H.; Harris, M.W.; Luebke, R.W.; Padarath-
singh, M.L.; Moore, O.A. (1980) Examination of bone marrow, immunologic
parameters and host susceptibility following pre- and postnatal exposure
to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). Int. J. Immunopharmacol.
2:301-310.
Luster, M.I.; Dean, J.H.; Boorman, G.A. (1982) Altered immune functions in
rodents treated with 2,3,7,8-tetrachlorodibenzo-£-dioxin, phorbal-12-my-
ristate-13-acetate, and benzo(a)pyrene. In: Hunt, V-R-; Smith, M.K.;
Worth, D., eds. Environmental factors in'human growth and development.
Banbury Report No. 11. Banbury, NY: Cold ,Spring Harbor Laboratory,
pp. 199-215.
Luster, M.I.; Hong, L.H.; Boorman, G.A.; Clark, G.; Hayes, H.T.; Greenlee, W.F.;
Dold, K.; Tucker, A.N. (1985) Acute myelotoxic responses in mice exposed
to 2,3,7,8-tetrachlorodibenzo-jJ-dioxin (TCDD). Toxicol. Appl. Pharmacol.
81:156-165.
Mantovani, A.; Vecchi, A.; Luini, W.; Sironi, M.; Candiani, G.P.; Spreafico,
F.; Garattini, S. (1980) Effect of 2,3,7,8-tetrachlorodibenzo-j)-dioxin
on macrophage and natural killer cell-mediated cytotoxicity in mice.
Biomedicine 32(4):200-204.
Marshall, E. (1986) Immune system theories on trial. Science 234:1490-1492.
22
-------
Moore, J.A.; Faith, R.E. (1976) Immunologic response and factors affecting
its assessment. Environ. Health Perspect. 18:125-131.
Murray, E.O.; Smith, F.A.; Nitschke, K.D.; Humiston, C.G.; Kociba, R.J.;
Schwetz, B.A. (1979) Three-generation reproduction study of rats given
2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in the diet. Toxicol. Appl.
Pharmacol. 50:241-251.
Nagarkatti, P.S.; Sweeney, G.D.; Gauldie, J.; Clark, D.A. (1984) Sensitivity
to suppression of cytotoxic T-cell generation by 2,3,7,8-tetrachlorodiben-
zo-jD-dioxin (TCDD) is dependent on the Ah genotype of the murine host.
Toxicol. Appl. Pharmacol. 72:169-176.
Pazdernik, T.L.; Rozman, K.K. (1984) Effect of thyroidectomy and thyroxine
on 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced immunotoxicity. Life Sci.
36:695-703.
Poland, A.; Glover, E. (1980) 2,3,7,8-tetrachlorodibenzo-£-dioxin: segre-
gation of toxicity with the Ah locus. Mol. Pharmacol. 17:86-94.
Powell, R.L. (1984) Dioxin in Missouri: 1971-1983. Bull. Environ. Contam.
Toxicol. 33:648-654.
Reggiani, G. (1980) Acute human exposure to TCDD in Seveso, Italy.
J. Toxicol. Environ. Health 6:27-43.
Sharma, R.P.; Gehring, P.J. (1979) Effects of 2,3,7,8-tetrachlorodibenzo-£-
dioxin (TCDD) on splenic lymphocyte transformation in mice after single
and repeated exposures. Ann. N.Y. Acad. Sci. 320:487.
Sharma, R.P.; Kociba, R.J.; Gehring, P.J. (1979) Reversal of immunologic and
"toxicologic effects of a single exposure of 2,3,7,8-tetrachlorodibenzo-£-
dioxin in mice. Trace Substances Environ. Health 12:473.
Stehr, P.A.; Stein, G,; Falke, H.; Sampson, E.; Smith, S.J.; Steinberg, K.;
Webb, K.; Ayres, S.; Schramm, W.; Donnell, H.D.; Gedney, W.B. (1986) A
pilot epidemiological study of possible health effects associated with
2,3,7,8-tetrachlorodibenzo-jJ-dioxin contaminants in Missouri. Arch.
Environ. Health 41:16-22.
Thigpen, J.E.; Faith, R.E.; McConnel, E.E.; Moore, J.A. (1975) Increased
susceptibility to bacterial infection as a sequela of exposure to 2,3,7,8-
tetrachlorodibenzo-p_-dioxin. Infect. Immun. 12:1319-1324.
Thomas, P.T.; Faith, R.E. (1985) Adult and perinatal immunotoxicity induced
by halogenated aromatic hydrocarbons. In: Dean, J.; Luster, M.; Munson,
A.; Amos, H.E., eds. Toxicology of the immune system. New York, NY:
Raven Press, pp. 305-313.
Tognoni, G.; Bonaccorsi, A. (1982) Epidemiological problems with TCDD: a
critical review. Drug Metab. Rev. 13:447-469.
23
-------
Tucker, A.N.; Vore, S.J.; Luster, M.I. (1986) Suppression of B cell differen-
tiation by 2,3,7,8-tetrachlorodibenzo-£-dioxin. Mol. Pharmacol. 29:372-377.
U.S. Environmental Protection Agency (1985) Health assessment document for
Dolychlorinated dibenzo-p-dioxins. Office of Health and Environmental
Assessment, Washington, DC. EPA/600/8-84/014F. NTIS PB86-122546/AS.
van Logten, M.J.; Gupta, B.N.; McConnel, E.E.; Moore, J.A. (1980) Role of
the endocrine system in the action of 2,3,7,8-tetrachlorodibenzo-£-dioxin
(TCDD) on the thymus. Toxicology 15(2):135-144.
Vecchi, A.M.; Sironi, M.; Canegrati, M.A.; Recchia, M.; Garattini, S. (1983)
Immunosuppressive effects of 2,3,7,8-tetrachlorodibenzo-£-dioxin in strains
of mice with different susceptibility to induction of aryl hydrocarbon
hydroxylase. Toxicol. Appl. Pharmacol. 68:434-441.
Vos, J.G.; Moore, J.A. (1974) Suppression of cellular immunity in rats and
' mice by maternal treatment with 2,3,7,8-tetrachlorodibenzo-£-dioxin.
Int. Arch. Allergy Appl. Immunol. 47:777-790.
Vos J.G.; Moore, J.A.; Zinkl, J.G. (1973) Effect of 2,3,7,8-tetrachloro-
dibenzo-£-dioxin on the immune system of laboratory animals. Environ.
Health Perspect. 5:149-162.
Vos, J.G.; Moore, J.A.; Zinkl, J.G. (1974) Toxicity of 2,3,7,8-tetrachloro-
dibenzo-£-dioxin (TCDD) in C57BL/6 mice. Toxicol. Appl. Pharmacol.
29:229-241.
Vos, J.G.; Kreeftenberg, J.G.; Engel, H.W.; Miderhoud, A.; van Noorle, J.L.M.
(1978) Studies on 2,3,7,8-tetrachlorodibenzo-j)-dioxin induced immune
suppression and decreased resistance to infection: endotoxin hypersensi-
tivity, serum zinc concentrations and effect of thymosin treatment.
Toxicology 9(l-2):75-86.
White, K.L.; Lysy, H.H.; McCay, A.; Anderson, A.C. (1986) Modulation of serum
complement levels following exposure to polychlorinated dibenzo-_p_-dioxins.
Toxicol. Appl. Pharmacol. 84:209-219.
Zinkl, J.G.; Vos, J.G.; Moore, J.A-.; Gupta, B.N. (1973) Hematologic and
clinical chemistry effects of 2,3,7,8-tetrachlorodibenzo-£-dioxin.
Environ. Health Perspect. 5:111-118.
24
-------
June 1988
Review Draft
APPENDIX F
RATIONALE FOR A HORMONE-LIKE MECHANISM
OF 2,3,7,8-TCDD FOR USE IN RISK ASSESSMENT
Michael A. Gallo
Professor and Chief, Division of Toxicology
Department of Environmental and Commmunity Medicine
University of Medicine and Dentistry of New Jersey
Robert Wood Johnson Medical School
675 Hoes Lane
Piscataway, NJ 08854-5635
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
-------
-------
The mechanisms of action of 2,3,7,8-tetrachlorodibenzo-|)-dioxin (2,3,7,8-
TCDD) have been intensely investigated since the pioneering work of Kimmig and
Schultz (1957) in elucidating the chloracnegen in the chlorophenol processes.
The most complete and thought-provoking treatise on the subject is that of
Poland and Knutson (1982). These authors and others in Poland's laboratory
drew upon their work and others to present a unified hypothesis to account for
the varied responses in animals exposed to 2,3,7,8-TCDD. In essence, this
hypothesis, which is partly based on the classic receptor theories invoked for
steroid action, suggests that (1) there is a cytosolic receptor for
arylhydrocarbons (the Ah receptor) that binds several compounds and then
translocates to the nucleus,' and (2) there is a second stage to the toxic
reaction(s) that is related to, but not congruent with, the induction of
cytochrome Pi-450. Activation of this receptor leads to a cascade of reactions
culminating with the association of the receptor-2,3,7,8-TCDD complex with
nuclear DNA. This association leads to the synthesis of a specific microsomal
protein designated as cytochrome Pj-450. To date, the chemical that binds this
putative receptor with the greatest avidity is 2,3,7,8-TCDD. However, many
other halogenated and non-halogenated compounds also bind to this cytosolic
protein (Nebert et a!., 1972). The xenobiotics with the greatest affinity are
those that are planar, with two phenyl rings and contain substitutions in the
lateral positions. Several investigative teams have examined the structure
activity relationships (SARs) among the polyhalogenated biphenyls (PCBs, PBBs),
polychlorinated dibenzo-j)-dioxins (PCDDs), and polychlorinated dibenzofurans
(PCDFs) (Knutson and Poland, 1980). Safe and his co-workers have synthesized
several of the highly active PCBs and have demonstrated remarkable SARs for
several biological end points (Mason et al., 1987). Excellent reviews on the
1
-------
SAR for PCDDs, PCDFs, and PCBs can be found in recent issues of the Annual
Reviews in Pharmacology and Toxicology (Vols. 22 [Poland and Knutson, 1982] and
26 [Safe, 1986], respectively). The hypothesis also states that ". . .there is
a second stage to the toxic reactions of 2,3,7,8-TCDD that are related to, but
not congruent with, the induction of cytochrome Pj-450." This portion of the
hypothesis is supported by the reports of Poland and his co-workers with XB
cells in culture (Knutson and Poland, 1980), Safe and co-workers' findings of
PCB inhibition of 2,3,7,8-TCDD toxicity (Haake et al., 1987), and the Umbreit
et al. report of apparent maximal induction of arylhydrocarbon hydroxylase (the
major system affected by cytochrome Pj-450) (Umbreit et al., 1987) by complex
mixtures of polyaromatic hydrocarbons and very low bioavailability of
PCDDs/PCDFs as measured by tissue levels and signs of toxicity.
At this point in time, it appears clear that 2,3,7,8-TCDD is working
through the proposed receptor mechanism for the first phase of its activity
(i.e., binding to a putative cytosolic receptor with subsequent induction of
P-450). However, all of the biological effects of this molecule cannot be
explained by simple receptor binding and induction of cytochrome PI-450. The
recent evidence from several laboratories has expanded on the initial studies
of Poland and Knutson (1982), Neal et al. (1982), and Barsotti et al. (1979) to
show quite dramatically that 2,3,7,8-TCDD markedly affects the interaction of
steroids with their respective receptors (Romkes et al., 1987; Gallo et al.,
1986) and 2,3,7,8-TCDD alters the number of Epidermal Growth Factor (EGF)
receptors in susceptible cell lines (Matsumura et al., 1984). Molloy et al.
(1987) recently reported the alteration of specific epidermal keratins in the
HRJ/S strain of mice after treatment with 2,3,7,8-TCDD. This finding is
especially relevant to the database since it was in this strain of mice that
-------
Poland and Knutson (1982) reported the model for chloracne. Matsumura et al.
(1984) studied the role of 2,3,7,8-TCDD and EGF receptors, while several
laboratories (Safe, 1986; Gierthy et al., 1987; Umbreit and Gallo, 1988;
Goldstein et al., 1987) have been pursuing the interactions of 2,3,7,8-TCDD
with steroids, primarily estrogen-sensitive tissues. The interactions with
glucocorticoids has been studied extensively by several laboratories (Luster et
al., 1984; Sunihara et al., 1987). In general, the response can be summarized
as a decrease in the number of available cytosolic receptors for EGF or the
steroids without a decrease in the affinity for the respective ligand. This
phenomenon is termed "down-regulation" of the cytosolic receptor. The
measurement of receptor binding is biochemical. The estimate of affinity and
binding site number is by extrapolation of the response curve(s) by Scatchard
analysis (1949). The strengths of this analysis are obvious, but the
weaknesses are difficult to reconcile. The major weakness is the lack of
direct binding information; this does not allow segregation of the receptors by
tertiary structure nor does it completely account for nonspecific binding. The
second weakness of ligand binding experiments is the inability of the analysis
to shed any light on the reason for the down-regulation. It must be emphasized
at this point that none of the steroid receptor research has demonstrated an
antagonism between 2,3,7,8-TCDD and the endogenous steroid for the respective
steroid receptor, nor has any competitive binding of steroids by the Ah
receptor been demonstrated. However, the steroid receptors are products of a
supergene family which is responsible for the protein synthesis of all these
receptors (Nebert et al., 1972), and the Ah receptor has many of the structural
and functional characteristics of the steroid receptors (Poellinger et al.,
1986).
-------
To better understand the role of 2,3,7,8-TCDD in cellular function (or
dysfunction), one must look to the results of the laboratories working on the
mechanisms of action of 2,3,7,8-TCDD at the molecular level. The major groups
involved in this research are Poland, Nebert's group at the National Institutes
of Health, and Whitlock's laboratory at Stanford University. As stated above,
Poland established the role of the Ah receptor in some of the actions of
2,3,7,8-TCDD. Whitlock (1987) has summarized his data and that of other
investigators regarding the regulation of the cytochrome P-450 gene family,
along with the data supporting the hypothesis that the Ah locus is part of a
super gene family responsible for the metabolism of xenobiotics and endogenous
compounds. Recent work in this laboratory has also elucidated a region on DNA,
which is sensitive to the 2,3,7,8-TCDD-cytosolic receptor complex, upstream
from the cytochrome Pj-450 gene (Neuhold et al., 1986). These findings are
critical in light of the findings of the 2,3,7,8-TCDD- responsive gene
expression enhancer system (region) described by Whitlock (Jones et al:, 1986).
Hence, the two laboratories have defined the regulatory mechanisms by which
2,3,7,8-TCDD controls gene expression of the cytochrome Pj-450 (Whitlock,
1987). The significance of these findings for 2,3,7,8-TCDD risk analysis is
the congruency between gene regulation for the Ah receptor, the glucocorticoid
receptor, and the estrogen receptor (Becker et al., 1986). The importance of
these findings cannot be underestimated. There is direct analogy with the
steroid receptor mechanisms and the control of the steroid receptor messenger
RNA (Yamamoto, 1985). The role of the estrogen receptor (ER), and other
steroid receptors, is understood to a greater extent than the Ah receptor
probably because of the greater emphasis on the physiology of steroids. The
analogy between the receptor complexes and DNA leads to the obvious comparison
-------
of effects of 2,3,7,8-TCDD and steroids. Many of the changes seen in animals
after dosing with 2,3,7,8-TCDD mimic estrogen or antiestrogen effects. Umbreit
and Gallo (1988) reviewed these findings, which are presented in Table 1.
Kociba et al. (1978) demonstrated the hepatocarcinogenic effect of orally
administered 2,3,7,8-TCDD, but in the same study there was a marked
dose-dependent decrease in tumors of the mammary glands and uteri which are
estrogen-sensitive organs. These are highly significant observations which
have been pursued by some laboratories. If 2,3,7,8-TCDD is acting through
hormonal (estrogen) mechanisms, then alteration of ovarian function, exogenous
estrogens, or antiestrogens should modify the response(s) to 2,3,7,8-TCDD.
Recent results have shown that 2,3,7,8-TCDD effects can be overridden by
exogenous estradiol (Gallo et al., 1986) and the down-regulation of the
estrogen receptor is also antagonized by estradiol (Romkes et al., 1987). The
significance of these findings are amplified if one couples the reports of
regulation of the EGF receptor by estrogens (Mukku and Stance!, 1985; Madhukar
et al., 1984) along with the consistent observation that the lowest doses in
the lifetime bioassays of 2,3,7,8-TCDD decrease tumor yield in rodent livers
but do not affect the backround levels of breast or uterine tumors (Kociba et
al., 1978). 2,3,7,8-TCDD inhibition of tumor growth at low doses and
enhancement at higher levels (in the bioassays) is supported by the recent
report of a marked decrease in tumorigenesis in the two-stage liver model at
the lowest dose of 2,3,7,8-TCDD after diethylnitrosamine (DEN) initiation
(Pitot et al., 1987). These findings are consistent with the hypothesis that
2,3,7,8-TCDD may be working through an endocrine-sensitive mechanism to yield
its toxic effects. If one accepts this premise, then it is reasonable to
assume that the actions of 2,3,7,8-TCDD can be explained using a
-------
TABLE 1. ASSOCIATION OF EXTROGENS WITH 2,3,7,8-TCDD TOXICITY
Many of the toxic effects of 2,3,7,8-TCDD are similar to effects of
estrogens in non-2,3,7,8-TCDD treated animals.
1. Some effects of 2,3,7,8-TCDD resemble effects of elevated
estrogens.
2. Other effect of 2,3,7,8-TCDD resemble antiestrogenic effects
[most antiestrogens have estrogenic effects at different doses].
3. Some 2,3,7,8-TCDD effects are not straightforward estrogenic or
antiestrogenic effects. For some of these, an influence of
estrogen on the effect is known.
4. Other signs of 2,3,7,8-TCDD toxicity may be related to cholesterol
mobilization for estrogen synthesis.
-------
TABLE 1. (continued)
2,3,7,8-TCDD EFFECTS
Fat loss 1,4
Wasting 1,4
Changes in serum lipids: 1,3,4
increased cholesterol
increased LDL, VLDL
Anorexia 1
Hypophagia 1
Hypoinsulinemia 1
Altered serum fatty acids 1,4
Hypoglycemia 1,4
Lowered Q£ consumption
Immunosuppression 1
Thymic involution 1
Decreased thymic
cellularity
Hirsutism 1
Chloracne 3
Skin keratinization 1
Membrane damage 1,4
Stimulates differentiation 1
in certain epithelial
cells
Lowered T4 in serum 3
Increased:
thyroid weight 3
serum TSH
T4 excretion as
glucuronide
Lower serum testosterone 3
Uterine suppression 2
Reproductive failure 1,2,3
Blockage of E2 uterotrophism 1
Terata 1,2,3
-------
TABLE 1. (continued)
Lowered serum corticoids 3,4
Blockage of ACTH stimulation of corticosteroid synthesis 3,4
Downregulates:
E6F receptor 1
Prolactin receptor 1
Glucocorticoid receptor 1
LDL receptor 1
Estrogen receptor 1
Ascites 1,4
Hepatocyte membrane damage 1,4
Hepatocyte membrane cAMP reduced
Enzyme inductions 3
AHH (EROD, P-448, P-450c, and d)
Ornithine decarboxylase
UDP-GTs
ALA synthetase
Others
Anemia
Porphyria cutanea tarda 1
Iron accumulation in liver
Altered iron transport in gut
8
-------
physiologically based model. The physiological implications of an endocrine
mechanism can explain many of the responses seen in animals after exposure to
2,3,7,8-TCDD (see Table 1) since these responses are similar to hyper- and
hypo-hormonal states (O'Malley and Buller, 1977; Potter et al., 1986; Jones et
al., 1986; Gustafsson et al., 1987). As stated above, the analogy to the
estrogen system is arguably the strongest to 2,3,7,8-TCDD effects (both
hyperplastic and dysplastic responses in endocrine sensitive organs), and the
similarity between the cytosolic receptors and their stabilization by molybdate
(Denison et al., 1986), activation of a nuclear site, and the anti-2,3,7,8-TCDD
effect of estradiol strengthens the analogy. The effects of estrogens are
widespread throughout the body. Some of these effects may not be receptor-
mediated, but the majority of the effects are directly attributable to receptor
binding. The toxic effects of estrogens have recently been summarized (Umbreit
and Gallo, 1988) and include thymic involution and decreased response to septic
challenge (Luster et al., 1984; Grossman, 1984), a wasting syndrome
characterized by weight loss, hirsutism, and epidermal lesions. As a note,
there are recent reports of estrogens enhancing or causing cachexia and wasting
which are major effects of 2,3,7,8-TCDD seen in intoxicated animals. The role
of estrogens as immunosuppressives is not well understood, but it is
hypothesized that the putative suppressant is either an excess of circulating
estradiol or perhaps an excess of trophic hormones. Estrogens also play a role
in the action of other hormones and trophic factors such as EGF (Kirk!and et
al., 1981; Gardner et al., 1978; Mukku, 1984; Mukku and Stance!, 1985; Hsueh et
al., 1981; Gonzales et al., 1984; Dickson and Lippman, 1987). These findings
lead to the conclusion that the multiple effects of TCDD could be mediated
through an endocrine mechanism. The weakness of this assumption is that
-------
2,3,7,8-TCDD causes effects that appear similar to both hyperestrogenemia and
hypoestrogenemia. It has been hypothesized that this apparent paradox is the
result of 2,3,7,8-TCDD or the ligand complex preventing the endogenous
substrate from interacting "correctly" with both the active site and a
secondary binding site (Umbreit and Gallo, 1988; Umbreit et al *, 1988).
Pleiotropism is not an uncommon finding with molecules such as hormones or
in this case 2,3,7,8-TCDD. One has only to review the early experiments on
multistage mouse skin carcinogenesis of 2,3,7,8-TCDD to see that in some cases
it inhibited tumor formation by PAH initiators (DiGiovanni et al., 1977; Berry
et al., 1978). It must be emphasized that the responses in multistage models
are dependent on time, sequence of administration, dose, and species. Hence,
inhibition under some conditions might have been predictable. This is
juxtaposed to the two-stage liver model (Pitot et al., 1980) in which it has
been shown that orally administered 2,3,7,8-TCDD enhances the tumorigenic
action of DEN. However, in subsequent experiments at lower doses of 2,3,7,8-
TCDD, a parabolic dose-response curve has been reported in the DEN/2,3,7,8-
TCDD initiation-promotion protocol (Pitot et al., 1987). This paradoxical
effect is not well understood, but it does not appear to be solely the function
of enhanced metabolism, or Ah receptor binding (Mason et al., 1987). Perhaps
it is the result of alteration of EGF receptors at low doses (Madhukar et al.,
1984) which displays a commonality with several steroid hormone receptors.
The importance of these findings to the approximation of human and animal
health risks from exposure to PCDDs and related molecules cannot be overstated.
Mathematical modeling of physiological phenomena, especially those related to
receptor function, is conducted using the Michaelis-Menten equation (1913) as
modified by Clark for the "classical" receptor model (1933). The weight of
10
-------
evidence for the most prevalent 2,3,7,8-TCDD effects falls into the category of
the receptor model (Poland and Knutson, 1982). The recent finding that the
hepatocarcinogenesis is related to estrogen levels or to the presence of
functional ovaries (Goldstein et al., 1987), and that DEN hepatocarcinogenesis,
in partially hepatectomized rats, is first inhibited then promoted by 2,3,7,8-
TCDD (Pitot et al., 1980, 1987) indicate that 2,3,7,8-TCDD is not causing its
myriad of effects in liver by a simple one-step event such as binding to the Ah
receptor and subsequent induction of cytochrome Pj-450. However, operationally
2,3,7,8-TCDD is a potent hepatocarcinogen in some species and strains of
rodents. ,
Risk modeling for carcinogenic xenobiotics has recently been segregated
into three classes or types of models: physiologically based pharmacokinetic
(PBPK) models in which the body is considered to be a small group of
physiological compartments (Hoel et al., 1983; Krewski et al., 1986; Bischoff,
1987); biologically motivated models of carcinogenesis (BMMC) in which the
carcinogenic process is considered to occur through a series of linked
reactions that result from two or more molecular events followed by cellular
amplification by "promoter" molecules (Moolgavkar, 1986; Thorslund et al.,
1987; Krewski et al., 1987); and the linearized multistage model (IMS) of
Armitage-Doll as modified by Crump and Howe (1984) in which it is assumed that
a sequence of mutational events occur within a single cell leading to the
neoplastic change (Armitage, 1985).
The model that appears to accommodate most of the critical components from
the biological data base on 2,3,7,8-TCDD is the BMMC model, which is generally
referred to as the Moolgavkar-Venzon-Knudson (M-V-K) model (Moolgavkar and
Venzon, 1979; Moolgavkar and Knudson, 1981). This model allows for several of
11
-------
the concepts of initiation-promotion-progression, along with the growth-
stimulating role of endogenous substrates such as hormones (Moolgavkar, 1986).
Incorporation of some of the factors necessary for the PBPK model can also be
done using the M-V-K model as modified, or, more correctly, expanded by
Thorslund et al. (1987). These expansions of the M-V-K model give the risk
assessor a powerful tool for looking at cancer risk mechanistically.
This option is not available with the IMS model as originally proposed.
The use of the IMS model may not be appropriate for the 2,3,7,8-TCDD data set
since this model assumes an initiating event, such as a point mutation, to
start the process. However, the IMS model can be accommodated if one
hypothesizes that the initiating event: (1) is the result of an indirect
action of 2,3,7,8-TCDD through modification of exogenous or endogenous
compounds, (2) that a population of initiated cells exists, or (3) 2,3,7,8-
TCDD leads to focal necrosis which serves as a mitogenic stimuli.
Recent reports have shown that 2,3,7,8-TCDD and other promoters in liver
enhance stimulation of DNA synthesis in situ, and stimulate repair of
0-6-methylguanine in liver DNA (Busser and Lutz, 1987; Den Engelse et al.,
1986). Lutz et al. (1984) presented a scheme for promoter potency based on
stimulation of DNA synthesis and the assumption that cell division is a
prerequisite for several stages in the carcinogenesis process. These reports
indicate that 2,3,7,8-TCDD can act as a complete-indirect-carcinogen, including
promoter activity, despite the lack of DNA binding or direct mutagenesis. The
sum of all these findings, along with the myriad of toxic responses, suggests a
model for 2,3,7,8-TCDD carcinogenesis in rodent liver as shown in Figure 1.
This model can account for the dose-response data in the bioassays and the
multistage promotion experiments, as well as allow for incorporation into
12
-------
existing risk models, and the scheme is not incongruent with the reports of
decreased tumor formation in some tissues. If pathway (A) can be verified by
demonstration of reactive intermediates after 2,3,7,8-TCDD treatment, then the
LMS model, slightly modified, can be used. The preponderance of evidence at
the moment supports a mechanistic model(s) which is at variance with the LMS
model. However, Figure 1 presents possibilities that are not mutually
exclusive for the existing models. The scheme also presents several testable
hypotheses which should be examined.
13
-------
Q
Q
O .
I—
I
00
CO
A
CM
O
rtj
O
Ol
CD
O
C
(Q
O
10
E
-C
o
CJ
ns
T3
O
O
Ol
CO
s_
14
-------
REFERENCES
Armitage, P. (1985) Environ. Health Perspect. 63:195-201.
Barsotti, D.A.; Abrahamson, L.J.; Allen, J.R. (1979) Bull. Environ. Contam.
Toxicol. 21:463-469.
Becker, P.B.; Gloss, B.; Schmid, W.; Strahle, U.; Schultz, G. (1986) Nature
(London) 324:686-688.
Berry, D.L.; DiGiovanni, J.; Juchau, M.R.; Bracken, W.M.; Gleason, G.L.;
Slaga, T.J. (1978) Res. Commun. Chem. Pathol. Pharmacol. 20:101-107.
Bischoff, K.B. (1987) In: Drinking water and health 8:36-64. Washington,
DC: National Academy Press.
Busser, M-T.; Lutz, W.K. (1987) Carcinogenesis 8:1433-1437.
Clark, A.J. (1933) The mode of action of drugs on cells. Baltimore, MD:
Williams and Wilkins.
Crump, K.S.; Howe, R..B. (1984) Risk Analysis 4:163-176.
Den Engelse, L.; Floot, B.G.J.; Menkveld, G.J.; Tates, A.D. (1986)
Carcinogenesis 7:1941-1947.
Denison, M.S.; Vella, L.M.; Okey, A.B. (1986) J. Biol. Chem. 261:3987-3995.
Dickson, R.B.; Lippman, M.E. (1987) Endocr. Rev. 8:29-43.
DiGiovanni, J.; Viaje, A.; Berry, D.L.; Slaga, T.J.; Juchau, M.R. (1977)
Bull. Environ. Contam. Toxicol. 18:552-557.
Gallo, M.A.; Hesse, E.J.; Macdonald, G.J.; Umbreit, T.H. (1986) Toxicol.
Lett. 32:123-132.
Gardner, R.M.; Kirkland, J.L.; Ireland, J.S.; Stance!, G.M. (1978)
Endocrinology 103:1164-1172.
Gierthy, J.F.; Dickerman, H.W.; Seeger, J.L.; Lincoln, D.W.; Martirez, H.;
Kumar, S.A. (1987) Toxicologist 7:159.
Goldstein, J.A.; Graham, M.J.; Sloop, T.; Maronpot, R.; Goodrow, T.; Lucier,
G.W. (1987) Dioxin 87, Abstract # MA-07.
Gonzales, F.; Lakshman, S.; Hoath, S.; Fisher, D.A. (1984) Acta Endocr.
105:425-428.
Grossman, C.J. (1984) Endocr. Rev. 5:435-455.
15
-------
Q) O
3
cn
8
m
o
o
CO
CO
oo
o
o
•o 2 < *• 2 m
»gi =rS8
^ !$• _ m -r m
HI
' § S
of
s a
a 0
».S
§1 s
|? £
r||
53-
-* Q)
Q. O
S =
s I
3 Q)
5?
°> S
c ffl,'
si
5 2.
> me
ID 3 3
n> <. s-
3 i a>
M
3 J?
(D CD
5 »
5T en
TJ
1
§
O
-
-5 3 S?
2 £ •*
2» ? o
—.0-1
O 3 rn
•
Ol O
N> 3
o> 3
00 (D
f
y^
------- |