UNITED
STATES
ENVIRONMENTAL
PROTECTION
AGENCY
WASHINGTON,
D.
C.
20460
OFFICE
OF
PREVENTION,
PESTICIDES
AND
TOXIC
SUBSTANCES
January
15,
2003
MEMORANDUM
SUBJECT:
Review
of
Additional
Data
on
Potential
Atrazine
Exposure
and
Review
Comments
Submitted
by
Syngenta
and
NRDC
on
Atrazine
Cancer
Epidemiology
Study:
"
Follow­
up
Study
of
Cancer
Incidence
Among
Workers
in
Triazine­
related
Operations
at
the
Novartis
St.
Gabriel
Plant"
by
Elizabeth
Delzell
et
al.
DP
Barcode
D287278,
MRID#
455184­
01,
Chemical
#
080803
FROM:
Jerome
Blondell,
Ph.
D.,
Health
Statistician
Chemistry
and
Exposure
Branch
1
Health
Effects
Division
(
7509C)

THRU:
Francis
B.
Suhre,
Senior
Scientist
Chemistry
and
Exposure
Branch
1
Health
Effects
Division
(
7509C)

TO:
Catherine
Eiden,
Senior
Scientist
Reregistration
Branch
3
Health
Effects
Division
(
7509C)

BACKGROUND
This
review
considers
additional
information
submitted
by
Syngenta
Crop
Protection,
Inc.
and
public
comment
on
the
exposure
of
workers
with
cancer
diagnosed
at
the
St.
Gabriel
Plant
in
Louisiana.
An
earlier
review
(
D281568,
March
25,
2002)
considered
the
results
of
a
cancer
incidence
study
at
this
plant.
Additional
exposure
information
has
been
provided
in
a
report
transmitted
to
the
Office
of
Pesticide
Programs
(
OPP)
on
November
1,
2002
titled
"
Summary
of
Information
on
Potential
Atrazine
Exposure
for
12
out
of
17
Prostate
Cancer
Cases
Reported
by
Delzell
et
al.
2001"
by
Charles
B.
Breckenridge.
In
addition
to
this
report,
OPP
also
received
comments
from
the
Natural
Resources
Defense
Council
(
NRDC,
June
3
and
July
30,
2002)
and
a
panel
report
titled
"
An
Evaluation
of
the
Report
by
Dr.
Delzell
et
al.
on
"`
A
Follow­
up
Study
of
Cancer
Incidence
Among
Workers
in
Triazine­
Related
Operations
at
the
Novartis
St.
Gabriel
plant'"
submitted
by
Hans­
Olov
Adami,
Graham
Colditz,
Jack
Mandel,
and
Dimitrios
Trichopoulos.
The
primary
purpose
of
this
review
is
to
consider
the
newly
submitted
exposure
information,
2
the
NRDC
comments,
and
the
panel
report.
The
following
background
information
is
quoted
from
the
March
25,
2002
review
(
Blondell
2002,
DP
Barcode
D281568):

Periodically,
Novartis
has
reported
to
EPA
on
an
ongoing
epidemiologic
study
of
workers
at
the
St.
Gabriel
plant
in
Louisiana.
The
main
product
of
this
plant
during
most
of
its
history
was
triazine
herbicides.
See
the
"
Review
of
Two
Atrazine
Epidemiology
Studies"
by
Jerome
Blondell
(
D226645,
MRID
#
s
440086­
01,
440086­
02)
which
summarizes
the
earlier
studies.
See
also
the
"
Review
of
five
atrazine
epidemiology
published
articles
for
SAP"
by
Ruth
Allen
(
D262405)
for
a
review
of
the
most
recently
published
studies
concerning
atrazine.
The
earlier
review
by
Blondell
(
September
13,
1996)
had
the
following
conclusion:
OREB
[
Occupational
and
Residential
Exposure
Branch]
concludes
that
neither
of
the
epidemiologic
studies
reviewed
here
adds
significant
new
information
concerning
adverse
health
effects
of
atrazine.
A
nonsignificant
elevation
in
non­
Hodgkin's
lymphoma
continues
to
be
observed
at
the
Louisiana
plant
among
workers
exposed
to
triazines,
including
atrazine.
By
itself
this
study
does
not
support
a
conclusion
of
increased
cancer
from
exposure
to
triazines.
However,
this
study
could
be
considered
supportive,
but
only
supportive
and
not
definitive,
if
evidence
of
an
association
between
non­
Hodgkin's
lymphoma
and
triazine
exposure
was
available
from
other
studies.
Follow­
up
by
the
National
Cancer
Institute
in
four
states
looked
specifically
to
determine
whether
earlier
associations
in
individuals
studies
could
be
attributed
to
atrazine
when
adjustment
was
made
for
exposures
to
other
pesticides.
They
concluded
that
"
detailed
analyses
suggested
that
there
was
little
or
no
increase
in
the
risk
of
NHL
attributable
to
the
agricultural
use
of
atrazine."
The
Occupational
and
Residential
Exposure
Branch
concurs
with
this
finding.

The
more
recent
studies
submitted
by
Novartis
did
not
find
additional
evidence
of
risk
of
NHL,
but
focus
principally
on
a
statistically
significant
increase
in
prostate
cancer.
The
earlier
review
concluded:
This
study
did
find
a
significant
association
between
prostate
cancer
and
working
at
the
St.
Gabriel
plant
where
triazine
herbicides,
especially
atrazine,
was
the
main
product
of
the
plant
during
most
of
its
history.
Statistically
significant
risks
were
restricted
to
the
Novartis
employees
rather
than
contract
employees.
However,
these
same
employees
were
the
most
likely
to
undergo
extensive
PSA
screening
which
likely
accounts
for
most,
if
not
all,
of
the
observed
increase,
rather
than
an
effect
of
atrazine
exposure.

A
severe
shortcoming
of
this
study
was
the
inability
to
assess
relative
exposures
among
the
workers
at
the
plant.
It
was
reported
that
eight
of
the
14
Novartis
workers
with
prostate
cancer
worked
at
jobs
that
would
have
no
more
than
"
sporadic"
exposure
to
atrazine
and,
apparently,
though
it
was
not
well
documented,
exposure
to
other
chemicals
(
e.
g.,
cyanuric
chloride,
hydrogen
cyanide)
that
might
influence
prostate
cancer
incidence.
It
is
3
difficult
to
make
any
firm
conclusions
about
the
potential
for
atrazine
to
be
the
primary
association
with
prostate
cancer
if
over
half
of
the
cases
had
questionable
exposure.
The
Health
Effects
Division
supports
the
study
authors
suggestion
that
historical
exposure
data
is
used
in
a
nested
case­
control
study.
Detailed
plant
work
histories
for
individual
subjects
could
be
developed
using
work
area,
job­
exposure,
and
information
from
biomonitoring
and
dust
sampling.

The
additional
information
submitted
by
Syngenta
Crop
Protection,
Inc.
provide
descriptive
information
about
the
exposure
of
the
prostate
cancer
cases
working
at
the
St.
Gabriel
plant.
This
information
is
reviewed
below,
followed
by
a
review
of
comments
received
from
NRDC
and
the
panel
report
prepared
by
a
separate
panel
of
epidemiologists
to
review
Dr.
Delzell's
report.

REVIEW
OF
ADDITIONAL
EXPOSURE
INFORMATION
FOR
PROSTATE
CANCER
CASES
At
EPA's
request,
Syngenta
prepared
an
atrazine
exposure
profile
of
employees
diagnosed
with
prostate
cancer.
There
were
17
cases,
including
14
Syngenta
cases
and
3
contract
employees.
Exposure
information
was
obtained
for
12
of
the
14
Syngenta
employees.
There
was
no
exposure
information
available
for
the
three
prostate
cancer
cases
among
contract
workers.
However,
contract
workers
accounted
for
62%
of
the
person
years
examined
and
their
period
of
exposure
was
a
median
of
2.6
years
compared
to
20
years
for
Syngenta
employees.
These
contract
employees
did
not
generally
receive
PSA
screening
and
their
numbers
were
not
significantly
higher
than
the
1.8
cases
expected,
based
on
using
Louisiana
as
the
comparison
population.
Using
the
local
industrial
corridor
as
a
comparison
population,
the
comparison
would
be
3
observed
prostate
cancer
cases
and
2.7
expected.
Therefore,
the
Health
Effects
Division
(
HED)
concludes
that
the
absence
of
exposure
information
for
contract
employees
is
not
of
particular
concern
because
the
observed
number
of
cases
is
close
to
expectation
and
the
duration
of
exposure
to
atrazine
is
relatively
low.

Two
of
the
14
Syngenta
cases
did
not
have
the
necessary
information
to
classify
by
exposure
but
were
concluded
to
have
low
exposure
based
upon
their
job
titles.
The
remaining
12
cases
were
classified
in
two
ways.
First,
job
titles
were
obtained
from
commencement
of
employment
until
September
2002
or
when
employment
ended.
Jobs
were
then
classified
by
their
"
proximity
to
locations
in
the
plant
where
atrazine
is
manufactured,
handled,
or
packaged."
Of
the
30
different
jobs,
5
were
classified
as
remote,
17
were
classified
as
low,
4
were
classified
as
mid,
and
4
were
classified
as
high
physical
proximity
to
atrazine
production.
For
each
prostate
cancer
case,
the
proportion
of
time
in
each
category
of
exposure
was
assessed
and
then
cumulated
up
until
the
time
of
prostate
cancer
diagnosis.
For
the
second
method,
a
relative
atrazine
proximity
scale
was
developed.
Based
on
atrazine
airborne
dust
monitoring
data,
remote,
low,
mid­
and
high
proximity
areas
were
found
to
differ
by
an
order
of
magnitude.
Thus,
each
category
could
be
assigned
a
relative
exposure
factor
of
0.1,
1,
10,
and
100.
This
value
was
multiplied
by
the
duration
at
each
type
of
location
and
cumulated
to
create
an
index
of
exposure.
This
index
of
exposure
also
adjusted
for
substantial
reductions
in
exposure
due
to
changes
at
the
plant
in
1984­
85.

Results
from
the
method
of
classifying
jobs
by
proximity
found
that
the
12
cancer
cases
spent
4
46%
of
their
plant
time
in
low
proximity
positions,
26%
in
medium
proximity,
and
28%
in
high
proximity
to
atrazine
production.
The
majority
of
the
high
proximity
time
was
due
to
three
of
the
cancer
cases
spending
the
majority
of
their
working
time
in
these
positions.

This
analysis
was
supported
by
the
cumulative
index
of
exposure
method.
Three
cases
had
high
proximity
to
atrazine
production
throughout
their
working
at
the
plant
with
a
cumulative
index
greater
than
10,000.
Four
cases
had
a
medium
exposure
with
a
cumulative
index
greater
than
1,000
and
less
than
10,000.
The
remaining
five
cases
had
a
low
exposure
index
(
less
than
1,000).
As
noted
above,
the
two
unassessed
cases
were
likely
to
be
low
proximity
based
on
their
job
titles.

Further
analysis
did
not
find
any
relationship
between
age
at
diagnosis
and
proximity
to
atrazine.
Had
their
been
such
a
relationship,
it
would
have
supported
the
possibility
that
atrazine
was
a
causative
factor
in
the
subsequent
diagnosis.
All
12
of
the
employees
with
prostate
cancer
participated
in
the
prostate
cancer
screening
program
and
at
least
10
of
them
were
initially
detected
due
to
the
PSA
screening.
Of
the
total
14
Syngenta
cases,
12
had
early
stage
localized
prostate
cancer
and
2
had
regional
cancer
within
the
prostate.
"
No
distant,
advanced
stage,
metastasized
cancer
was
detected
in
Syngenta
employees."
Together
these
results
are
consistent
with
the
conclusion
that
the
observed
significant
excess
in
prostate
cancer
at
the
Syngenta
plant
in
Louisiana
was
more
likely
an
artifact
of
the
extensive
PSA
screening
program
than
a
result
of
exposure
to
atrazine.

Additional
Exposure
Information
not
available
to
external
peer
reviewers:

The
information
provided
above
did
not
provide
comparative
information
on
non­
exposed
workers.
This
is
a
key
shortcoming
because
if
only
a
very
small
proportion
of
all
workers
were
involved
in
the
high
exposure
work
areas,
it
could
mean
that
the
incidence
of
prostate
cancer
was
much
higher
in
these
areas.
Therefore,
the
following
request
was
sent
to
Syngenta
at
the
end
of
December
2002.

Can
you
let
us
know
what
proportion
of
the
workforce
at
the
St.
Gabriel
plant
work
in
the
different
buildings?
Figure
1
from
your
report
shows
18
buildings
and
only
4
of
these
were
classified
as
medium
or
high
proximity
for
atrazine
exposure.
It
would
be
helpful
to
know
even
a
rough
average
number
of
workers
that
spent
time
in
the
buildings
classified
as
high,
medium
or
low.
Just
having
number
for
the
prostate
cancer
cases
makes
it
difficult
to
make
any
comparison
or
conclusion.

In
response,
Syngenta
supplied
the
following
information
on
January
16,
2003:

Table
1
in
the
attached
file
now
provides
estimates
of
the
number
of
male
Syngenta
employees
at
St.
Gabriel
excluding
contract
manufacturers
as
you
requested.
We
have
performed
chi
square
tests
of
the
observed
and
expected
incidences
of
prostate
cancer
case
based
on
12
or
14
cases
found
in
the
St.
Gabriel
work
force
up
to
the
end
of
1999
.
.
.

The
results
of
the
chi
square
analysis
presented
in
Table
2
indicate
that
a
higher
than
expected
incidence
of
prostate
cancer
cases
was
distributed
to
the
moderate
proximity
subgroup.
Table
3
shows
that
when
the
low
and
moderate
proximity
subgroups
are
combined,
the
observed
and
expected
incidence
of
prostate
cancer
cases
are
comparable.
5
These
results
support
the
interpretation
that
there
is
no
association
between
proximity
to
atrazine
manufacture
and
the
occurrence
of
prostate
cancer
among
Syngenta
employees
at
St.
Gabriel.

A
copy
of
Table
1
is
reproduced
below:

Table
1.
Estimated
Number
(%)
of
Male
Employees
at
St.
Gabriel
at
the
End
of
Each
Year
(
Syngenta
Male
Employees
Only)

Proximity
to
Atrazine
Manufacturing
Year
1977
1986
1996
2002
Low
205
(
72%)
275
(
77%)
306
(
79%)
291
(
80%)

Moderate
20
(
7%)
20
(
6%)
20
(
5%)
19
(
5%)

High
60
(
21%)
60
(
17%)
63
(
16%)
55
(
15%)

Total
Number
285
355
389
365
Based
on
this
Table,
an
average
of
77%
of
Syngenta
employees
had
low
proximity
to
atrazine
manufacturing;
an
average
of
6%
had
moderate
proximity;
and
17%
had
high
proximity
to
atrazine
manufacturing.
Of
the
14
prostate
cancer
cases,
50%
were
classified
as
low
proximity
to
atrazine
manufacturing,
28%
were
classified
as
moderate
proximity,
and
21%
were
classified
as
high
proximity.
It
appears
that
there
would
be
no
strong
evidence
of
dose­
response,
although
a
higher
proportion
of
workers
(
50%
versus
23%)
were
involved
in
jobs
with
moderate
or
high
proximity
to
atrazine
manufacturing.
However,
no
strong
conclusions
should
be
drawn
from
this
crude
comparison.
A
proper
comparison
would
require
measuring
the
exposure
of
cases
and
non­
cases
in
the
same
manner
and
taking
into
account
confounders
such
as
age
and
person­
years
of
exposure.
Syngenta
acknowledges
this
shortcoming
and
is
planning
a
case­
control
study
within
the
cohort
to
address
this
issue.
6
REVIEW
OF
COMMENTS
FROM
THE
NATURAL
RESOURCES
DEFENSE
COUNCIL
The
Natural
Resources
Defense
Council
(
NRDC)
sent
EPA
two
sets
of
comments
concerning
the
prostate
cancer
in
workers
exposed
to
atrazine.
NRDC
noted
that
"
these
cancers
appeared
commonly
in
younger
workers
(
almost
every
cancer
case
was
a
man
under
age
55),
and
are
most
common
in
those
workers
who
spent
the
most
time
at
the
facility."
However,
as
discussed
above
and
in
the
earlier
review
(
Blondell
2002),
the
PSA
screening
at
the
plant
was
offered
to
relatively
young
workers.
According
to
Table
19
of
the
Delzell
et
al.
report,
among
active
Novartis
workers
40
years
of
age
and
over,
just
7­
9%
were
55
years
or
older
from
1989
through
1999.
Thus,
there
was
a
very
small
number
of
workers
over
age
55
who
participated
in
the
PSA
screening.
Table
9
of
the
report
generally
found
higher
or
roughly
equal
incidence
ratios
among
workers
who
worked
less
than
5
years
compared
to
workers
who
worked
5
years
or
more
after
stratifying
for
years
since
hire.
One
would
have
expected
higher
ratios
among
those
with
5
or
more
years
employment
at
the
plant,
if
dose
of
atrazine
was
a
factor
contributing
to
prostate
cancer.
There
was
some
evidence
of
a
consistent
pattern
between
duration
of
exposure
since
hire
and
higher
prostate
cancer
incidence
ratios,
but
this
was
not
consistent
across
subgroups
and,
in
any
case,
was
likely
correlated
to
opportunity
for
and
number
of
PSA
screening
tests.

NRDC
criticized
the
Syngenta
study
for
the
lack
of
exposure
information
based
on
specific
job
descriptions
which
was
crucial
to
examining
whether
chemical
exposure
was
associated
with
the
reported
illnesses.
This
comment
was
echoed
by
the
HED
and
other
reviewers.

NRDC
noted
concern
that
"
more
than
200
workers
were
excluded
from
the
study"
who
left
Louisiana
before
1988
because
they
might
be
more
likely
to
develop
cancer
due
to
longer
time
since
exposure.
However,
the
study
depends
on
the
Louisiana
Tumor
registry
for
both
the
source
of
cases
and
controls
and
in
such
studies,
it
is
appropriate
to
limit
subjects
to
those
that
could
be
captured
by
the
Tumor
registry.

NRDC
was
concerned
that
the
study
population
was
over
60
percent
contract
workers
who
had
relatively
short
exposure
times
and
"
were
therefore
far
less
likely
to
be
at
risk
from
chronic
exposures".
However,
it
should
be
acknowledged
that
the
Syngenta
study
stratified
the
analysis
and
examined
the
standardized
risk
incidence
ratios
for
just
the
workers
employed
at
the
plant
with
much
longer
exposures.
Therefore,
HED
disagrees
with
the
conclusion
that
"
the
company
likely
diluted
the
apparent
frequency
that
exposed
individuals
develop
cancer".
Further
statements
by
NRDC
concerning
the
ability
"
to
extract
relevant
information"
and
to
show
cancers
were
concentrated
among
plant
employees
who
worked
longer
[
and,
therefore,
had
much
more
access
to
PSA
screening]
do
not
support
a
finding
that
results
were
"
diluted"
by
including
low
exposure
employees.
NRDC
notes
prostate
cancer
cases
had
a
median
length
of
employment
of
20
years,
but
should
also
have
noted
that
this
association
applies
to
age
and
opportunity
for
PSA
screening,
two
primary
confounders
of
concern.

NRDC
states
that
"
Even
if
the
study
is
correct
to
suggest
that
improved
screening
may
account
for
some
increase
in
cancer
incidence,
the
number
of
cancers
at
the
St.
Gabriel
facility
cannot
be
explained
merely
by
rigorous
testing."
The
question
of
the
effects
of
screening
is
an
important
7
one.
Other
studies
(
Mettlin
2000,
Roberts
et
al.
1999)
have
noted
a
two­
to
three­
fold
increase
in
age­
adjusted
prostate
cancer
incidence,
while
mortality
increased
by
20­
50%.
Both
incidence
and
mortality
were
reported
to
decline
after
1992
in
these
studies.
Howe
et
al.
(
2001)
commented:

Following
large
increases
in
prostate
cancer
incidence
due
to
the
introduction
of
PSA
screening
in
the
late
1980s,
the
incidence
trend
has
been
stable
from
1995
through
1998,
while
death
rates
have
continued
to
decline
for
both
white
males
and
black
males.
Much
of
the
geographic
variation
in
prostate
cancer
incidence
reflects
differences
in
PSA
screening,
with
regions
of
high
PSA
screening
penetration
having
higher
incidence
rates,
often
because
of
the
discovery
of
clinically
insignificant
tumors.

How
much
of
the
national
prostate
cancer
incidence
increase
is
due
to
PSA
screening
is
unclear.
The
study
in
a
Minnesota
county
(
Roberts
et
al.
1999)
found
a
three­
fold
increase
in
incidence
among
men,
over
50%
of
whom,
in
their
sixties
and
seventies,
received
PSA
screening,
a
proportion
characterized
as
"
extremely
high".
The
six­
fold
increase
reported
at
the
Syngenta
plant
where
nearly
100%
received
screening
is
not
inconsistent
with
a
conclusion
that
screening
rather
than
some
environmental
factor
is
a
major
cause,
and
perhaps,
the
only
cause
for
the
jump
in
the
incidence
rate.

However,
NRDC
points
out
and
HED
agrees
that
one
should
be
cautious
at
attributing
all
of
the
observed
excess
to
screening
rather
than
exposure
to
atrazine
without
further
investigation.
The
report
on
exposure,
subsequently
received
from
Syngenta,
found
more
prostate
cancers
among
those
with
low
or
moderate
exposure.
This
would
seem
difficult
to
attribute
to
atrazine
unless
there
were
some
threshold
effect,
such
that
the
risk
of
cancer
did
not
follow
a
dose­
response
relationship,
or
unless
there
were
errors
in
the
measurement
of
exposure.
Possible
errors
would
mean
that
those
in
the
low
or
moderate
category,
characterized
as
having
exposures
that
were
"
low
or
below
detection
limits",
actually
did
have
high
exposures
that
were
missed.
For
example,
there
might
have
been
an
unforeseen
release
leading
to
short­
period
of
high
levels
of
atrazine
in
the
breathing
space
of
employees
otherwise
exposed
to
only
very
low
levels.
Another
possibility
might
be
a
temporary
transfer
of
atrazine
on
dust
particles
that
were
transferred
to
low
exposure
workers,
their
food
or
water.
These
possibilities
are
at
best
speculative
and
given
the
limitation
of
significant
exposure
to
only
4
of
18
buildings,
they
appear
unlikely.

NRDC
notes
that
one
subset
of
workers,
actively
working
at
diagnosis
had
a
nine­
fold
increase
(
11
cases,
1.2
expected).
However,
subsets
of
data
should
be
expected
to
exhibit
greater
variation.
The
95
percent
confidence
interval
on
the
nine­
fold
increase
shows
the
odds
ratio
may
be
expected
to
fall
anywhere
between
4.6
and
16.4.
A
weight­
of­
evidence
determination
would
require
much
more
consistency
among
various
subsets
with
higher
exposed
subgroups
exhibiting
higher
estimated
rates
of
risk.
Earlier,
NRDC
had
claimed
that
those
leaving
the
state
were
"
those
most
likely
to
develop
work­
related
cancer",
but
this
would
be
inconsistent
with
active
workers,
with
a
relatively
shorter
time
since
exposure,
having
the
highest
risk.

NRDC
notes
that
a
number
of
other
cancers
exhibited
excesses,
including
buccal
cavity
(
3
observed,
2.1
expected),
esophagus
(
2
observed,
0.7
expected),
stomach
(
2
observed,
0.9
expected),
bladder
(
3
observed,
1.6
expected),
thyroid
(
2
observed,
0.6
expected)
and
leukemia/
lymphomas
(
7
8
observed,
4.5
expected).
These
data
are
based
on
Table
7
of
the
Syngenta
report
by
Delzell
et
al.
This
table
shows
9
different
estimates
of
risk,
not
counting
prostate
(
already
discussed)
and
certain
grouped
categories.
Therefore,
6
of
9
categories
exhibited
an
excess,
though
statistically
insignificant
risk.
Chance
alone
is
a
possible
explanation
for
such
findings.
In
addition,
bias
and
confounding
could
produce
such
results.
Therefore,
these
elevated,
nonsignficant
incidence
ratios
must
be
considered
preliminary
findings.
Until
these
findings
are
replicated
in
other
studies
that
address
the
serious
methodological
limitations
(
especially
the
low
statistical
power)
of
the
present
study,
they
should
be
regarded
as
spurious
or
suggestive
at
best.

Finally,
NRDC
charged
that
EPA
disregarded
other
"
epidemiologic
studies
linking
atrazine
to
ovarian,
prostate,
testicular,
and
breast
cancer,
and
to
non­
Hodgkin's
lymphoma."
Although
it
is
true
that
the
later
EPA
reviews
had
not
addressed
these
specific
studies
in
detail,
earlier
reviews
examined
the
more
important
studies.
The
question
of
non­
Hodgkin's
lymphoma
(
NHL)
was
addressed
in
the
October
2000
review
"
Review
of
Atrazine
Incident
Reports,
DP
Barcode
D270014".
One
of
the
most
comprehensive
reviews
of
atrazine
was
the
study
by
Zahm
et
al.
(
1993)
reviewed
by
HED
as
follows:

Zahm
et
al.
(
1993)
combined
three
population­
based
casereferent
studies
from
Nebraska,
Iowa­
Minnesota,
and
Kansas
to
evaluate
the
role
of
atrazine
in
the
development
of
NHL.
The
studies
in
Iowa­
Minnesota
and
Kansas
included
white
men,
and
in
Nebraska,
the
study
included
both
white
men
and
women.
In
Nebraska,
the
study
involved
227
white
men,
twenty­
one
years
or
older
who
had
been
diagnosed
with
NHL
between
July
1,
1983
and
June
30,
1986.
In
Iowa­
Minnesota,
the
study
involved
780
white
men,
thirty
years
or
older
who
were
newly
diagnosed
cases
with
the
disease
between
March
1981
and
October
1983,
and
in
Minnesota
between
October
1980
and
September
1982.
In
Kansas,
the
study
involved
200
white
men,
twenty­
one
years
or
older
between
1979
and
1981.
Controls
were
randomly
selected
from
the
same
geographic
areas
as
the
cases
and
were
matched
by
race,
gender,
five­
year
age
groups,
and
vital
status.
An
odds
ratio
of
1.4
was
determined
for
the
three
studies
combined
for
101
NHL
and
214
controls
where
atrazine
was
used
on
farms
where
they
worked
or
lived.
The
odds
ratio
ranged
from
1.2
in
Iowa
to
2.7
in
Kansas.
In
two
of
the
states
and
in
all
states
combined,
the
risks
were
higher
for
farmers
who
used
atrazine
in
their
farming
operations
but
did
not
handle
the
chemical
than
among
farmers
who
handled
atrazine.
Other
than
atrazine,
the
farmers
could
have
been
exposed
to
other
herbicides
and
insecticides
which
could
have
increased
their
chances
of
experiencing
NHL.
The
study
concluded
that
there
"
was
little
or
no
increase
in
the
risk
of
NHL
attributable
to
the
agricultural
use
of
atrazine".

To
conclude
that
the
relationship
between
NHL
and
Atrazine
might
be
confounded
by
exposure
to
other
herbicides
such
as
2,4­
D
is
not
the
same
as
concluding
that
2,4­
D
is
known
to
be
the
risk
factor.
9
This
association
can
be
confounded
by
a
number
of
other
associations
and,
in
any
case,
a
full
and
separate
weight­
of­
evidence
determination
based
on
all
the
studies
would
be
needed
to
assess
the
carcinogenicity
of
2,4­
D.

Studies
in
Kentucky
have
suggested
an
association
between
county
cancer
rates
of
breast
cancer
and
ovarian
cancer
and
triazine
exposure
(
Kettles
et
al.
1997,
Hopenhayn­
Rich
et
al.
2002).
A
similar
study
in
California
found
various
associations
between
atrazine
and
leukemia,
brain
cancer,
testicular
cancer,
and
prostate
cancer
(
Mills
1998).
However,
these
studies
are
subject
to
aggregation
bias
because
the
actual
exposures
of
individuals
in
the
county
or
how
long
they
resided
there
is
not
known.
Such
studies
are
considered
useful
for
hypothesis
generating
but
not
for
drawing
conclusions.
The
authors
themselves
warn
"
conclusions
concerning
causality
cannot
be
drawn"
(
Kettles
et
al.
1997).
Having
conducted
an
ecologic
study
in
Kentucky
of
similar
design
to
those
above,
the
current
reviewer
is
very
aware
that
agricultural
pesticide
use
correlates
with
a
whole
host
of
factors
that
vary
across
an
urban­
rural
gradient
(
Blondell
JM.
Urban­
rural
factors
affecting
cancer
mortality
in
Kentucky,
1950­
1969.
Cancer
Detection
and
Prevention
11:
209­
223,
1988).
Persons
living
in
rural
areas
differ
not
only
in
terms
of
pesticide
exposures,
but
also
diet,
parity,
physical
activity,
exposure
to
viruses,
and
other
lifestyle
factors.
Appropriate
controls
are
critical
when
studying
the
relationship
between
pesticide
exposure
and
cancer
(
Blondell
1990).
10
REVIEW
OF
PANEL
REPORT
TO
EVALUATE
THE
REPORT
BY
DELZELL
ET
AL.

In
February
2002,
Syngenta
asked
Dr.
Jack
Mandel
(
Vice
President
of
Health
and
Environmental
Groups,
Exponent,
Inc.)
to
convene
a
separate
panel
of
epidemiologists
to
review
the
report
"
A
follow­
up
study
of
cancer
incidence
among
workers
in
triazine
operations
at
the
Novartis
St.
Gabriel
plant".
The
panel
was
convened
in
early
March
and
included
Dr.
Mandel
and
the
following:
Hans­
Olov
Adami,
M.
D.,
Ph.
D.
­
Professor
and
Chairman,
Dept.
of
Medical
Epidemiology,
Karolinaka
Institutet,
Sweden.
Graham
Colditz,
M.
D.,
Dr.
P.
H.
­
Professor,
Dept.
of
Epidemiology,
School
of
Public
Health,
Harvard
University.
Dimitrios
Trichopoulos,
M.
D.
­
Professor
of
Epidemiology,
Dept.
of
Epidemiology,
School
of
Public
Health,
Harvard
University.
Vincent
L.
Gregory,
Professor
of
Cancer
Epidemiology,
Dept.
of
Epidemiology,
School
of
Public
Health,
Harvard
University.

This
panel
report
starts
out
by
summarizing
the
prostate
cancer
results
as
follows:

The
excess
prostate
cancer
incidence
was
restricted
to
men
younger
than
60
years
of
age
and
Novartis
employees;
indeed,
among
contract
employees
and
people
older
than
60
years,
there
was
no
excess
incidence
of
prostate
cancer.
Only
18%
of
the
prostate
cancer
cases
diagnosed
among
plant
employees
were
not
localized,
compared
to
twice
as
many
(
37%)
in
the
general
population
of
LA
state.
Even
in
a
large
prostate
cancer
screening
trial,
the
proportion
of
clinically
advanced
cancers
was
higher
(
30%)
than
among
the
workers
in
the
Novartis
St.
Gabriel
Plant
(
Koning
et
al.
[
International
Journal
of
Cancer
97:
237­
244],
2000)
­
a
reflection
of
the
high
intensity
of
PSA
screening
in
the
latter
group.

This
summary
points
out
the
importance
of
understanding
the
effect
of
intensive
PSA
screening
on
the
pattern
of
prostate
cancers
that
was
observed.
After
providing
this
summary
the
panel
provided
the
following
"
minor
criticisms":
1.
Expected
prostate
cancers
were
calculated
on
the
basis
of
5
year
intervals
rather
than
1
year
which
could
lead
to
some
underestimation
of
expected
cases
when
the
incidence
of
the
disease
was
rising
relatively
quickly,
as
was
occurring
in
Louisiana
up
to
1996.
2.
Prostate
cancer
incidence
rates
were
unavailable
for
1998­
1999
and
this
could
introduce
more
underestimation
of
expected
prostate
cancers.
3.
The
panel
argues
that
the
authors
should
have
given
more
prominence
to
the
comparison
with
the
industrial
corridor
than
with
the
State,
because
this
area
would
be
more
comparable
in
terms
of
socioeconomic
status
and
health
practices.
4.
Reporting
of
cancer
cases
for
the
entire
state
was
incomplete
during
the
earliest
part
of
the
followup
period
and
this
could
have
led
to
underestimation
of
prostate
cancer
cases
in
the
State.
5.
As
acknowledged
by
the
study
authors,
incomplete
residence
histories
may
have
led
to
a
relative
underestimation
of
expected
prostate
cancer
cases.
6.
PSA
screening
provided
by
outside
medical
care
providers
was
not
evaluated.
And
the
reliance
on
hard­
copy
only
data
for
the
1989­
1993
period
may
not
have
been
complete.
11
7.
The
study
authors
reference
to
slight
increase
in
several
other
forms
of
cancer
is
considered
to
be
misleading
because,
while
true,
the
opposite
was
true
for
other
cancers.
Therefore,
the
increases
and
decreases
combined
would
be
what
one
would
expect
to
occur
by
chance.
8.
The
study
authors
noted
the
increase
in
all
cancers
in
the
subgroup
with
relatively
long
potential
induction
time
and
long
duration
employment.
The
panel
stated
this
was
misleading
because
it
was
due
to
prostate
cancer
that
was
heavily
concentrated
in
the
Novartis
employees
who
worked
much
longer
and
had
more
follow­
up
than
the
contract
workers.

HED
agrees
that
these
criticisms
could
be
characterized
as
minor.
The
first
point,
about
the
use
of
a
five
year
interval,
could
result
in
some
underestimation
of
expected
rates,
but
this
effect
was
likely
very
small.
HED
disagrees
with
the
second
point.
Given
that
rates
were
unavailable
for
1998­
99,
it
could
mean
either
an
underestimation
or
overestimation
of
expected
cases.
The
third
point
about
socioeconomic
status
and
health
practices
appears
logical
on
a
superficial
basis
but
is
not
backed
up
by
any
hard
data.
As
the
earlier
review
by
Blondell
(
2002)
noted
there
needs
to
be
a
more
careful
justification
for
choosing
the
industrial
corridor
which
is
likely
to
overlap
the
known
"
cancer
alley"
in
that
part
of
the
country.
The
effects
of
points
4
and
5
would
be
expected
to
be
slight
at
the
most.
In
point
6,
the
mention
of
incomplete
screening
is
likely
to
have
at
the
most
a
minor
effect
given
that
PSA
screening
among
workers
already
approached
100%.
Points
7
and
8
are
agreed
upon
by
HED
as
noted
above.

The
rest
of
the
panel
review
focuses
on
the
implications
of
PSA
screening.
Autopsy
studies
have
shown
a
relatively
high
prevalence
of
small
prostate
cancers,
even
among
middle­
aged
men.
Just
how
many
of
these
"
latent"
cancers
would
progress
to
clinical
disease
remains
uncertain,
but
they
can
be
detected
by
PSA
testing.
This
means
that
PSA
likely
introduces
an
over­
diagnosis
bias
by
detecting
some
latent
cancers
that
would
not
have
progressed
to
clinical
disease.
The
panel
suggests
that
the
relatively
late
introduction
of
therapeutic
interventions
are
unlikely
to
have
influenced
mortality
statistics.
Therefore,
unlike
most
other
cancers,
prostate
cancer
mortality
data
may
be
a
better
indicator
of
incidence
trends
than
incidence
rates
which
are
confounded
by
increased
screening.
According
to
the
panel:
differences
in
diagnostic
intensity
can
create
substantial
differences
in
the
recorded
incidence
of
the
disease.
These
differences
would
be
largely
limited
to
early
stage
disease
with
a
low
fatality
rate
and
they
would
influence
mortality
rates
little,
if
at
all.
While
these
features
of
prostate
cancer
have
long
been
known,
PSA
testing
as
a
screening
tool
has
amplified
the
consequences
of
differences
in
the
intensity
of
surveillance.

In
particular,
PSA
screening
provides
substantially
longer
lead
time
than
other
cancer
screening
tests.
A
detailed
analysis
of
a
large
prostate
cancer
screening
trial
(
Auvinen
et
al.
2002)
suggests
that
PSA
testing
add
5­
7
years
lead
time
which
could
advance
the
time
of
diagnosis
by
up
to
10­
14
years
compared
with
clinical
detection.
As
a
result,
newly
detected
cases
could
exceed
the
expected
numbers
by
a
factor
of
5
or
more.
In
support
of
this
effect,
prostate
cancer
incidence
in
the
US
(
SEER
data)
more
than
doubled
after
its
introduction
in
the
1980s.
Then
the
incidence
declined
to
a
level
about
65%
higher
than
during
the
pre­
PSA
screening
era,
as
would
be
expected
if
the
test
were
responsible
for
"
identification
of
a
substantial
fraction
of
long­
standing,
indolent
sub­
clinical
cases."
The
panel
acknowledges
that
it
is
difficult
to
quantify
the
effects
of
PSA
screening
on
12
incidence
due
to
fragmentary
data
over
varying
time
periods
and
the
practice
of
measuring
only
single
screenings
but
not
the
effects
of
multiple
measurements
for
an
individual.
Despite
these
limitations
and
based
on
data
from
multiple
sources,
the
panel
infers
that
PSA
screening
has
increased
recorded
incidence
2.3
times
and
perhaps
as
much
as
3.5
times
if
those
of
younger
age
with
repeated
screenings
are
considered.
Later,
the
panel
supports
the
plausibility
of
a
5­
fold
increase
by
citing
the
Olmstead
County,
Minnesota
study
(
Roberts
et
al.
1999)
which
found
a
3.5­
fold
increase.
HED
used
this
same
study
to
arrive
at
a
similar
conclusion.

The
panel
presents
the
following
points
as
arguing
against
occupational
exposure
as
explanation
for
the
increase
incidence
in
prostate
cancer
at
the
Novartis
St.
Gabriel
plant:
1.
There
is
no
prior
biological
or
epidemiological
evidence
that
atrazine
is
a
human
carcinogen.
HED
agrees
with
this
point
for
prostate
cancer,
but
it
should
not
be
extended
to
all
cancer
sites.
2.
"
There
is
no
established
environmental
risk
factor
that
could
double
the
incidence
of
prostate
cancer."
HED
disagrees
with
this
point.
Just
because
an
environmental
factor
hasn't
been
found
before
doesn't
mean
that
the
next
study
conducted
won't
find
one.
This
is
especially
true
for
prostate
cancer
given
the
lack
of
careful
study
of
environmental
factors.
3.
Environmental
factors
are
known
to
exist
based
on
studies
of
Japanese
migrants,
but
the
effects
on
incidence
require
the
passage
of
at
least
a
generation.
This
is
a
weak
argument
at
best.
Although
some
other
known
factor
operates
early
in
life
and
takes
a
very
long
time
to
manifest,
does
not
preclude
factors
that
occur
later
in
life.
Latency
periods
for
most
cancers,
including
prostate,
are
not
well
known
with
any
degree
of
precision.
4.
There
is
no
established
or
suspected
non­
genetic
risk
factor
that
affects
incidence
differentially
at
young
ages.
HED
agrees
that
one
would
have
expected
an
increase
at
all
ages.
However,
it
is
also
true
that
the
population
available
to
study
for
older
age
groups
was
limited.
Therefore,
the
power
of
the
study
to
detect
such
an
effect
was
similarly
limited.
5.
The
cancers
detected
at
St.
Gabriel
were
almost
exclusively
localized
disease.
There
is
increasing
evidence
that
environmental
factors
are
more
likely
to
influence
tumor
progression
rather
than
initiation.
Therefore,
if
an
effect
were
observed
it
would
be
less
likely
to
be
confined
to
localized
disease.
HED
partly
agrees,
although
given
the
absence
of
knowledge
about
initiators
or
promotors,
far
more
research
is
needed
to
confirm
this
supposition.
6.
"
There
was
no
excess
of
incidence
from
all
other
forms
of
cancer
combined
among
the
workers,
reducing
the
likelihood
that
a
carcinogenic
factor
was
operating
in
the
working
environment".
HED
disagrees
with
this
point.
Many
carcinogens
are
site
specific
and
only
a
few
are
known
to
effect
multiple
sites
(
e.
g.,
smoking
and
arsenic).
The
panel
conclusion
reads
as
follows:
Following
the
introduction
of
systematic
widespread
PSA
testing
in
the
Novartis
St.
Gabriel
Plant,
prostate
cancer
incidence
has
increased
as
much
as,
but
no
more
than,
would
have
been
expected
on
the
basis
of
empirical
evidence
and
biological
considerations
concerning
the
consequences
of
PSA
screening
on
reported
prostate
cancer
incidence.
There
is
neither
a
need
to
invoke,
nor
evidence
to
support,
the
contribution
of
environmental
factors
in
the
particular
occupational
setting
on
prostate
carcinogenesis.
13
SUMMARY
OF
EXTERNAL
PEER
REVIEW
COMMENTS
Each
reviewer
was
supplied
a
copy
of
the
original
study
at
the
St.
Gabriel
Plant
in
Louisiana,
comments
from
the
Natural
Resources
Defense
Council
(
NRDC,
June
3
and
July
30,
2002)
and
a
the
panel
report
submitted
by
Hans­
Olov
Adami,
Graham
Colditz,
Jack
Mandel,
and
Dimitrios
Trichopoulos.

1.
Dr.
Howard
Morrison,
Health
and
Welfare
Canada
Dr.
Morrison's
main
conclusion
was
"
I
don't
think
you
can
conclude
much
from
the
study,
what
little
you
could
would
be
that
while
there
was
almost
definitely
some
increased
prostate
cancer
case
finding
because
of
increased
PSA
screening,
there
was
a
suggestion
that
this
might
not
be
the
entire
explanation."

While
generally
agreeing
with
the
panel
report
by
Adami
et
al.,
Dr.
Morrison
expressed
concern
that
a
single,
underpowered
epidemiologic
study
was
not
likely
to
find
compelling
evidence
of
an
association
between
atrazine
and
prostate
cancer.
He
cautioned
that
saying
that
PSA
screening
could
be
an
explanation
for
the
excess
in
prostate
cancer
was
not
the
same
as
saying
that
PSA
screening
had
been
demonstrated
to
be
the
full
explanation
for
the
increase.

Dr.
Morrison
questioned
the
value
of
further
studies
of
this
cohort.
"
The
proposed
nested
case­
control
study
lacks
any
discussion
of
the
abysmal
lack
of
power
the
study
will
realistically
have."

2.
Dr.
Edward
Giovannucci,
Harvard
School
of
Public
Health
Dr.
Giovannucci
limited
his
comments
"
to
the
question
of
what
is
the
likely
cause
of
this
apparent
excess
of
prostate
cancer."
His
conclusion,
in
part,
was
"
In
my
opinion,
the
magnitude
of
the
increase
is
compatible
with
PSA
screening
as
being
the
explanation.
PSA
screening
advances
the
diagnosis
time
of
prostate
cancer
by
5­
10
years.
Prostate
cancer
incidence
increases
sharply
with
age,
more
so
than
any
other
cancer.
Thus,
by
advancing
the
diagnosis
of
prostate
cancer
by
a
number
of
years,
there
will
be
an
apparent
increase
in
the
incidence
of
this
cancer."
He
goes
on
to
cite
two
studies
which
demonstrate
that
sharp
increases
in
incidence
can
result
from
PSA
screening.
He
felt
this
finding
was
further
supported
by
the
following
evidence
"
The
excess
in
prostate
cancer
was
observed
in
active
employees
who
received
intensive
PSA
screening.
Screening
detected
cancers
would
be
expected
to
be
asymptomatic
and
localized,
as
the
vast
majority
were
in
this
study.
There
was
no
indication
of
an
increase
in
aggressive
or
advanced
prostate
cancers,
or
an
increase
in
prostate
cancer
mortality."

Regarding
the
additional
information
about
the
exposure
of
the
prostate
cancer
cases,
submitted
by
Syngenta,
Dr.
Giovannucci
stated
that
this
information
further
supported
the
hypothesis
that
PSA
screening
accounted
for
the
increase
in
prostate
cancer.
Quoting
from
his
review:
The
additional
data
provided
by
Syngenta
of
the
12
prostate
cancer
cases
in
regards
to
cumulative
index
of
proximity
to
atrazine
manufacturing
and
on
the
prevalence
of
PSA
screening
further
support
the
conclusions
that
the
apparent
increase
in
prostate
14
cancer
in
this
population
is
due
primarily
if
not
entirely
to
intensive
PSA
screening.
Under
the
intensive
screening
for
men
of
this
age
group,
a
large
increase
in
the
diagnosis
of
asymptomatic,
localized
prostate
cancer
is
inevitable.
The
exact
magnitude
of
increase
expected
is
uncertain,
but
appear
within
the
realms
of
the
available
data.
The
fact
that
there
is
no
apparent
increase
in
advanced
tumors,
mortality,
that
proximity
of
atrazine
manufacturing
did
not
appear
to
be
correlated
with
risk,
and
the
increase
was
during
the
years
of
PSA
screening
strongly
supports
the
hypothesis
that
the
excess
of
prostate
cancer
is
related
to
PSA
screening
in
this
population.

3.
Dr.
Richard
Hayes,
National
Cancer
Institute
Dr.
Hayes
stated
his
principle
conclusion
as
follows
"
While
PSA
screening
may
account
for
much
of
the
excess
of
prostate
cancer
in
this
Triazine
manufacturing
facility,
it
would
be
premature
to
reject
a
potential
role
of
occupational
exposure
to
triazines
as
a
contributing
factor
to
the
observed
excess
of
this
disease."
Regarding
the
excess
observed
for
other
cancers,
he
commented
"
The
study
sample
size
is
small,
so
suggestive
excesses
for
a
number
of
suspect
cancers,
such
as
NHL,
can
not
be
adequately
investigated."

Dr.
Hayes
expressed
concern
that
atrazine
might
effect
hormonal
levels
in
men
which
are
thought
to
have
a
role
in
prostate
cancer,
but
acknowledged
that
nothing
was
"
known
about
intraprostatic
action
of
atrazine".
Regarding
future
examination
of
this
cohort,
Dr.
Hayes
recommended
"
comparisons
of
exposure
history
of
prostate
cancer
cases
and
non­
cases­­
coupled
with
individual
data
on
PSA
screening­­
could
provide
insight
about
the
reasons
for
the
prostate
cancer
excess
in
this
factory.

4.
Dr.
Aaron
Blair,
National
Cancer
Institute
Dr.
Blair
stated
his
principle
conclusion
as
follows
"
To
clearly
understand
the
issue
of
prostate
cancer
and
atrazine
exposure
in
this
cohort
it
is
essential
that
a
quantitative
exposure
assessment
be
added.
The
approach
described
by
Breckenridge
for
a
few
of
the
cases
is
a
reasonable
starting
point.
A
few
additional
analyses
of
prostate
cancer
risk
before
and
after
initiation
of
screening
would
also
provide
direct
information
on
the
impact
of
the
screening."

Dr.
Blair
agreed
that
the
report
by
Adami
et
al.
(
2002)
"
suggest
that
PSA
screening
may
well
explain
the
excess
incidence
of
prostate
cancer
in
this
cohort.
It
would
be
helpful,
however,
to
have
more
information
supporting
the
selection
of
multiplication
factor
regarding
the
impact
of
age
and
other
factors
that
might
differ
between
a
cohort
of
working
individuals
and
the
general
population."

The
additional
exposure
information
provided
by
Breckenridge
was
criticized:
"
This
report
provides
no
useful
information
regarding
the
issue
of
atrazine
exposure
and
prostate
cancer
because
it
only
includes
cases.
To
effectively
use
information
about
amount
of
exposure
among
the
cases
it
is
necessary
to
have
a
noncase
group
for
comparison."
15
OVERALL
CONCLUSION
It
appears
that
most
of
the
increase
in
prostate
cancer
incidence
at
the
St.
Gabriel
plant
in
Louisiana
is
likely
due
to
intensive
PSA
screening.
The
study
was
insufficiently
large
and
suffered
from
other
limitations
that
prevent
ruling
out
atrazine
as
a
potential
contributor
to
the
increase
observed.
On
balance,
however,
a
role
for
atrazine
seems
unlikely
because
prostate
cancer
was
found
primarily
in
active
employees
who
received
intensive
PSA
screening,
there
was
no
increase
in
advanced
tumors
or
mortality,
and
proximity
to
atrazine
manufacturing
did
not
appear
to
be
correlated
with
risk.
Atrazine
has
been
tied
to
inflammation
of
the
prostate
in
laboratory
animals
and
changes
in
testosterone
levels
at
high
doses.
However,
neither
condition
has
been
tied
to
the
increased
risk
of
prostate
cancer
and
HED
concludes
the
animal
data
do
not
provide
biologically
plausible
evidence
to
support
atrazine
as
a
cause
of
prostate
cancer.

Other
cancers
besides
prostate
were
found
to
have
an
elevated,
though
not
statistically
significant,
increase
in
risk
at
the
St.
Gabriel
plant.
Other
studies
have
suggested
an
increased
risk
for
ovarian,
breast,
and
other
cancers,
including
non­
Hodgkin's
lymphoma.
However,
these
studies
are
at
best
preliminary
and
should
not
serve
as
a
basis
for
implicating
atrazine
as
a
human
carcinogen
due
to
their
methodological
limitations.

HED
recommends
continued
monitoring
of
all
cancers
in
this
cohort,
but
does
not
make
a
recommendation
regarding
the
special
nested
case­
control
study
suggested
by
Breckenridge.
External
peer
reviewers
were
evenly
split:
further
study
of
this
cohort
was
supported
by
Blair
and
Hayes
but
not
by
Giovannucci
and
Morrison.

References
Auvinen
A,
Maattanen
L,
Stenman
UH,
Tammela
T,
Rannikko
S,
Aro
J,
Juusela
H,
Hakama
M.
2002.
Lead­
time
in
prostate
cancer
screening
(
Finland).
Cancer
Causes
and
Control
13:
279­
285.

Blondell
JM.
1988.
Urban­
rural
factors
affecting
cancer
mortality
in
Kentucky,
1950­
1969.
Cancer
Detection
and
Prevention
11:
209­
223.

Blondell
JM.
1990.
Problems
encountered
in
the
design
of
epidemiologic
studies
of
cancer
in
pesticide
users.
La
Medicina
del
Lavoro
81:
524­
529.

Blondell
J.
March
25,
2002.
Review
of
Atrazine
Cancer
Epidemiology
Studies:
"
A
Follow­
up
Study
of
Mortality
Among
Workers
at
the
Novartis
St.
Gabriel
Plant"
and
"
Follow­
up
Study
of
Cancer
Incidence
Among
Workers
in
Triazine­
related
Operations
at
the
Novartis
St.
Gabriel
Plant"
both
by
Elizabeth
Delzell
et
al.
DP
Barcode
D281568,
MRID#
451521­
01
and
455184­
01,
Chemical
#
080803.

Blondell
J,
Spann
MF.
October
31,
2000.
Review
of
Atrazine
Incident
Reports,
DP
Barcode
D270014,
Chemical
#
080803.
16
Hopenhayn­
Rich
C,
Stump
ML,
Browning
SR.
2002.
Regional
assessment
of
atrazine
exposure
and
incidence
of
breast
and
ovarian
cancers
in
Kentucky.
Archives
of
Environmental
Contamination
and
Toxicology
42:
127­
136.

Howe
HL,
Wingo
PA,
Thun
MJ,
Ries
LAG,
Rosenberg
HM,
Feigal
EG,
Edwards
BK.
2001.
Annual
report
to
the
nation
on
the
status
of
cancer
(
1973
through
1998),
featuring
cancers
with
recent
increasing
trends.
Journal
of
the
National
Cancer
Institute
93:
824­
842.

Kettles
MA,
Browning
SR,
Prince
TS,
Horstman
SW.
1997.
Triazine
herbicide
exposure
and
breast
cancer
incidence:
an
ecologic
study
of
Kentucky
counties.
Environmental
Health
Perspectives
105:
1222­
1227.

Mettlin
C.
2000.
Impact
of
screening
on
prostate
cancer
rates
and
trends.
Microscopy
Research
and
Technique
51:
415­
418.

Mills
PK.
1998.
Correlation
analysis
of
pesticide
use
data
and
cancer
incidence
rates
in
California
counties.
Archives
of
Environmental
Health
53:
410­
413.

Roberts
RO,
Bergstralh
EJ,
Katusic
SK,
Lieber
MM,
Jacobsen
SJ.
1999.
Decline
in
prostate
cancer
mortality
from
1980
to
1997,
and
an
update
on
incidence
trends
in
Olmsted
County,
Minnesota.
Journal
of
Urology
161:
529­
533.

cc:
atrazine
file
(
080803)
Kimberly
Nesci
(
7508C)
17
COPIES
OF
EXTERNAL
PEER
REVIEWS
Electronic
copies
provided
by
each
of
the
external
peer
reviewers
are
reproduced
below:

Email
from
Dr.
Howard
Morrison,
Health
and
Welfare,
Canada,
December
23,
2002.

A
few
thoughts
on
an
entertaining
collection
of
reviews.
You
didn't
have
to
wonder
who
had
written
what.
Let
me
know
if
you
would
like
clarifications.
I
don't
think
you
can
conclude
much
from
the
study,
what
little
you
could
would
be
that
while
there
was
almost
definitely
some
increased
prostate
cancer
case
finding
because
of
increased
PSA
screening,
there
was
a
suggestion
that
this
might
not
be
the
entire
explanation.

Review
of
the
evaluation
by
Adami
et
al.

The
reviewer's
conclusions
are
generally
sound:
the
paper
by
Delzell
et
al.
does
not
provide
compelling
evidence
of
an
association
between
atrazine
and
prostate
cancer.
How
likely
was
it
for
a
single,
underpowered
epidemiologic
study
to
do
so?
A
perfectly
reasonable
explanation
for
the
excess
noted
is
increased
PSA
screening.
This
is
not,
however,
to
say
that
the
study
provides
a
"
clean
bill
of
health"
for
atrazine.
There
is
a
difference
between
suggesting
a
reasonable
alternate
explanation,
and
demonstrating
that
this
explanation
is
the
full
story.

The
evaluation
has
problems:

?
"
While
the
excess
risk
was
confined
to
men
below
60
years
of
age,
there
is
no
established
or
suspected
non­
genetic
risk
factor
for
prostate
cancer
that
affects
incidence
differentially
at
younger
ages."

A
specious
argument.
Few
epidemiologic
studies
report
relative
risks
by
age.
Those
that
do
typically
report
higher
risks
for
younger
ages,
a
reflection
of
the
nature
of
the
risk
measure
(
multiplicative)
and
genetic
susceptibility.
See
the
paper:
Morrison
H,
Savitz
D,
Semenciw
R,
Hulka
B,
Mao
Y,
Morison
D,
Wigle
D.
Farming
and
prostate
cancer
mortality.
Am
J
Epidemiol.
1993
Feb
1;
137(
3):
270­
80
for
an
example
of
a
study
which
linked
herbicide
exposure
with
prostate
cancer
(
mortality),
with
a
stronger
effect
noted
among
younger
farmers.

?
There
is
no
established
environmental
risk
factor
that
could
double
the
incidence
of
prostate
cancer.

True,
but
so
what?
There
is
a
huge
difference
between
saying
we
haven't
identified
strong
risk
factors
to
saying
that
there
aren't
strong
risk
factors.

?
It
isn't
really
fair
to
say
that
"
evaluation
of
the
prostate
cancer
risk
by
duration
of
employment
and
empirical
induction
period
did
not
reveal
any
clear
pattern...."
Give
the
number
of
cases,
it
is
unlikely
to
have
had
the
power
to.

?
The
review
ignores
the
role
of
TURPs
[
Transurethral
resection
of
the
prostate]
in
the
run­
up
in
prostate
cancer
incidence
rates
in
the
1980'
s.
18
?
That
IARC
has
modified
it's
classification
of
atrazine
to
group
3
hasn't
been
without
controversy.
See:
Huff
J.
IARC
monographs,
industry
influence,
and
upgrading,
downgrading,
and
under­
grading
chemicals:
a
personal
point
of
view.
International
Agency
for
Research
on
Cancer.
Int
J
Occup
Environ
Health.
2002
Jul­
Sep;
8(
3):
249­
70.

?
To
say
that
there
is
no
prior
biological
or
epidemiological
evidence
that
atrazine
is
a
human
carcinogen
is
misleading.
For
an
example
of
the
non­
existent
epidemiological
evidence,
see:
Schroeder
JC,
Olshan
AF,
Baric
R,
Dent
GA,
Weinberg
CR,
Yount
B,
Cerhan
JR,
Lynch
CF,
Schuman
LM,
Tolbert
PE,
Rothman
N,
Cantor
KP,
Blair
A.
Agricultural
risk
factors
for
t(
14;
18)
subtypes
of
non­
Hodgkin's
lymphoma.
Epidemiology.
2001
Nov;
12(
6):
701­
9.
Many
other
examples
of
positive
epidemiologic
literature
are
cited
by
the
review
by
the
Natural
Resources
Defence
Council.

Review
of
the
evaluation
by
the
National
resources
defence
council
This
evaluation
identifies
limitations
to
the
data
provided
by
Syngenta
in
their
report
to
the
EPA.
The
National
Resources
Defence
Council
has
correctly
identified
the
lack
of
exposure
histories
of
workers
as
both
significant
and
curious.
It
is
a
fair
criticism
to
note
the
failure
of
Syngenta
to
model
the
effects
of
screening
on
incidence
rates,
and
to
wonder
how
Syngenta
could
conclude
that
prostate
cancer
cases
tended
to
be
only
mildly
exposed
when
they
also
claimed
that
they
lacked
information
on
individual
employee's
actual
exposures.

Some
of
the
criticisms
miss
the
mark.
It
is
unfortunate
that
the
study
needed
to
exclude
more
than
200
workers
because
they
were
known
or
presumed
to
have
left
Louisiana
before
1988,
however,
what
would
the
reviewer
suggest?
Such
individuals
are
"
lost
to
follow­
up"
and
should
have
been
censored
upon
being
lost
to
follow­
up.

It
isn't
quite
fair
to
say
that
"
exposure
misclassification
is
known
to
bias
results
towards
the
null
hypothesis"
(
page
21)
­
rather,
non­
differential
exposure
misclassification
tends
to
bias
towards
the
null
hypothesis.

It
is
fair
to
point
out
that
excesses
were
noted
from
cancers
other
than
prostate.
However,
the
numbers
involved
are
small;
I'm
not
sure
what
to
make
of
finding
3
cases
of
buccal
cavity
cancer
instead
of
the
2.1
expected,
but
it
isn't
much.
There
were
cancer
sites
with
deficits
(
none
statistically
significant),
but
these
aren't
given
equal
attention.
This
is
a
standard
problem
in
epidemiology
­
focus
on
the
not­
statistically
significantly
increased
associations,
but
ignore
the
not­
statistically
decreased
associations.

The
review
makes
much
of
the
finding
of
11
cases,
when
1.2
were
expected,
among
company
employees
actively
working
at
diagnosis.
This
is
a
curious
and
suggestive
finding,
which
isn't
adequately
dealt
with
in
the
paper
by
MacLennan
et
al.

Review
of
the
paper:
"
Cancer
incidence
among
triazine
herbicide
manufacturing
workers."
The
paper
is
well
done,
and
fair,
albeit
somewhat
conservative,
in
its'
analysis
of
the
data.
Perhaps
not
surprisingly,
the
authors
have
not
dealt
19
with
their
failure
to
do
an
adequate
job
in
assessing
exposure
histories.
Nor
have
they
discussed
what
is
in
many
ways
the
biggest
limitation
of
the
study,
that
it
is
underpowered.
Just
how
much
could
you
say
about
dose­
response
when
you
study
only
has
11
cases?
It
also
needs
to
be
remembered
that
to
conclude
"
that
medical
surveillance
is
a
plausible
alternative
explanation
of
the
results,"
while
clearly
true,
isn't
the
same
as
concluding
that
this
is
what
happened
Review
of
the
Syngenta
submission
This
report
is
problematic.

Having
an
alternate
explanation
for
the
increased
number
of
prostate
cancer
cases
isn't
the
same
as
indicating
that
all
of
the
increase
was
from
this
cause.

Why
does
Breckenridge
say
that
there
was
a
300%
risk
in
prostate
cancer
incidence
reported
in
SEER
for
the
US
population
after
the
introduction
of
PSA
testing?
Increases
were
not
nearly
this
large.
If
it
is
for
a
specific
sub­
group,
this
needs
to
be
identified.

Although
it
is
interesting
to
know
about
the
likely
exposures
of
the
prostate
cancer
cases
are,
by
itself
it
is
relatively
uninformative
­
there
are
no
denominator
data
given
(
and
the
numbers
are
so
small...).
Case
in
point
­
figure
2.
How
would
this
compare
to
those
who
weren't
diagnosed
with
prostate
cancer?

What
is
the
point
of
figure
3?

Figure
6
­
what
does
it
mean
to
compare
cumulative
incidence
in
plant
workers
to
SEER
incidence
data?
It
isn't
a
valid
comparison.

The
proposed
nested
case­
control
study
lacks
any
discussion
of
the
abysmal
lack
of
power
the
study
will
realistically
have.
It
would
be
a
great
study
for
the
company
to
have
conducted,
given
that
it
has
little
likelihood
of
observing
a
statistically
significant
effect
because
it
will
be
underpowered.
20
Dr.
Edward
Giovannucci,
Harvard
School
of
Public
Health,
December
31,
2002
To
Dr.
Jerome
Blondell.
From
Dr.
Edward
Giovannucci
Re:
Review
of
follow­
up
study
of
prostate
cancer
at
the
Novartis
St.
Gabriel
Plant
December
31,
2002
As
requested,
the
following
is
a
review
of
the
Syngenta
report
on
the
epidemiologic
study
of
triazine
workers
at
the
Novartis
St.
Gabriel
Plant.
The
study
design
is
for
the
most
part
appropriate.
A
major
limitation
for
such
a
study
design
is
determining
the
most
appropriate
population
rates
for
comparison.
Two
groups
are
used
for
comparison.
The
results
suggest
an
increase
in
cancer,
due
mostly
to
prostate
cancer.
My
comments
here
are
limited
to
the
question
of
what
is
the
likely
cause
of
this
apparent
excess
of
prostate
cancer.
I
do
not
consider
other
potentially
relevant
data
such
as
animal
or
mechanistic
studies
in
this
review.

The
major
question
is
whether
the
excess
rates
are
due
to
an
occupational
exposure,
are
compatible
with
chance,
or
are
due
to
another
factor.
These
findings
are
not
likely
due
to
chance.
Besides
occupational
exposure,
the
most
likely
candidate
for
an
alternative
explanation
is
a
high
frequency
of
PSA
screening,
the
diagnostic
test
that
detects
prostate
cancer.
The
notable
fact
about
this
study
is
that
the
time
period
overlaps
with
the
availability
of
the
PSA
test
in
the
United
States.
Prostate
cancer
is
unique
in
that
the
vast
majority
of
pathologic
cancers
would
not
come
to
clinical
notice
except
for
sensitive
diagnostic
tests,
such
as
PSA.
When
PSA
began
to
be
used
in
the
United
(
and
other
areas),
the
diagnosis
of
prostate
cancer
increased
dramatically.
There
is
no
credible
evidence
that
there
was
a
true
increase
in
incidence
as
mortality
has
been
relatively
stable
or
even
decreased.
Thus,
the
use
of
the
PSA
test
can
definitely
cause
a
remarkable
substantial
increase
in
prostate
cancer.

Given
the
extraordinarily
high
prevalence
of
PSA
screening
in
the
active
Novartis
employees
(
98%
had
at
least
one
PSA
test
by
age
45
years),
an
excess
of
prostate
cancer
would
be
expected.
Several
facts
indicate
that
the
excess
of
prostate
cancer
was
related
to
PSA
screening:


The
excess
in
prostate
cancer
was
observed
in
active
employees
who
received
intensive
PSA
screening.


Screening
detected
cancers
would
be
expected
to
be
asymptomatic
and
localized,
as
the
vast
majority
were
in
this
study.


There
was
no
indication
of
an
increase
in
aggressive
or
advanced
prostate
cancers,
or
an
increase
in
prostate
cancer
mortality.

Given
the
localized,
asymptomatic
nature
of
the
tumors,
the
majority
were
clearly
detected
through
PSA
screening.
The
remaining
question
is
whether
this
excess
is
consistent
with
what
would
be
expected
through
PSA
screening
alone,
or
if
the
data
suggests
an
additional
etiologic
factor
accounting
for
an
increase
in
risk.
In
my
opinion,
the
magnitude
of
the
increase
is
compatible
with
PSA
screening
as
being
the
explanation.
PSA
screening
advances
the
diagnosis
time
of
prostate
cancer
by
5­
10
years.
Prostate
cancer
incidence
increases
sharply
with
age,
more
so
than
any
other
cancer.
Thus,
by
advancing
the
diagnosis
of
prostate
cancer
by
a
number
of
years,
there
will
be
an
21
apparent
increase
in
the
incidence
of
this
cancer.
For
example,
in
one
study,
the
ratio
of
prostate
cancer
incidence
in
men
who
were
screened
with
PSA
was
6.5
times
higher
than
the
control
group.
In
essence,
there
were
6.5
times
more
prostate
cancers
diagnosed
due
solely
to
PSA
screening
(
BJU
International
2001;
88:
811­
17).
In
a
screening
trial
in
a
Finnish
population,
the
ratio
of
the
number
of
cases
detected
through
PSA
screening
in
the
first
year
relative
to
the
number
expected
based
on
age­
specific
incidence
rate
in
Finland
was
14.4
for
men
aged
55
years
(
Cancer
Causes
and
Control
2002;
13:
279­
285).
This
ratio
of
screened
detected
cases
to
unscreened
population
incidence
increased
with
age
so
the
potential
bias
in
men
aged
younger
than
55
years
would
be
even
greater
based
on
these
data.
Thus,
the
increased
excess
of
prostate
cancer
observed
in
the
Novartis
study
is
compatible
with
increases
expected
in
a
population
that
is
receiving
intensive
PSA
screening.

The
additional
data
provided
by
Syngenta
of
the
12
prostate
cancer
cases
in
regards
to
cumulative
index
of
proximity
to
atrazine
manufacturing
and
on
the
prevalence
of
PSA
screening
further
support
the
conclusions
that
the
apparent
increase
in
prostate
cancer
in
this
population
is
due
primarily
if
not
entirely
to
intensive
PSA
screening.
Under
the
intensive
screening
for
men
of
this
age
group,
a
large
increase
in
the
diagnosis
of
asymptomatic,
localized
prostate
cancer
is
inevitable.
The
exact
magnitude
of
increase
expected
is
uncertain,
but
appear
within
the
realms
of
the
available
data.
The
fact
that
there
is
no
apparent
increase
in
advanced
tumors,
mortality,
that
proximity
of
atrazine
manufacturing
did
not
appear
to
be
correlated
with
risk,
and
the
increase
was
during
the
years
of
PSA
screening
strongly
supports
the
hypothesis
that
the
excess
of
prostate
cancer
is
related
to
PSA
screening
in
this
population.

Please
contact
me
if
you
have
any
questions.

Sincerely.

Edward
Giovannucci,
M.
D.
Sc.
D.
Associate
Professor
of
Medicine
Harvard
Medical
School
Associate
Professor
of
Nutrition
and
Epidemiology
Harvard
School
of
Public
Health
22
Dr.
Richard
Hayes,
National
Cancer
Institute
National
Institutes
of
Health
National
Cancer
Institute
Executive
Plaza
South
8118
6120
Executive
Blvd
MSC
7240
Bethesda
MD
20892­
7240
January
8,
2003
Jerome
Blondell,
Ph.
D.,
M.
P.
H.
Health
Effects
Division
(
7509C)
U.
S.
Environmental
Protection
Agency
Washington,
D.
C.
20460
Dear
Dr.
Blondell,

Please
find
attached
my
review
of
the
materials
related
to
the
epidemiologic
study
of
triazine
workers
at
the
St.
Gabriel
plant
in
Louisiana.

If
you
need
further
information,
please
do
not
hesitate
to
ask.

Sincerely
yours,
Richard
B.
Hayes,
DDS,
PhD
EPS
8114
Division
of
Cancer
Epidemiology
and
Genetics
[
attachment]

The
"
Follow­
up
Study
of
Cancer
Incidence
among
Workers
in
Triazine­
related
Operations
at
the
Novartis
St.
Gabriel
Plant"
included
2,045
subjects
identified
from
Novartis
employment
records
at
the
St.
Gabriel
plant,
including
for
study
workers
who
had
accrued
at
least
six
months
of
employment
in
the
plant
by
1992
and
had
worked
at
anytime
since
1970
in
jobs
with
potential
exposure
to
triazines
(
based
on
job
location).
Study
subjects
included
three
broad
categories
of
worers:
company
employees
working
at
anytime
since
1970,
contract
production
employees
working
anytime
since
1977,
and
contract
maintenance
employees
working
anytime
since
1983
(
i.
e.,
earliest
years
for
which
work
records
were
available
for
individual
employees).
Exit
date
from
Louisiana
(
LA)
was
determined
through
tracing
for
almost
all
subjects.
Cancer
incidence
was
based
on
linkage,
using
appropriate
identifiers,
to
LA
cancer
incidence
and
mortality
files,
and
was
calculated
to
date
of
specific
cancer
occurrence,
date
of
death,
to
end­
of­
follow­
up,
1997,
or
to
the
estimated
date
of
emigration
from
the
State,
whichever
came
first.

Cancer
incidence
was
evaluated
during
1985­
97
showing
no
overall
increase
in
cancer,
however,
a
statistically
significant
excess
of
prostate
cancer
was
found
(
11
observed,
4­
6
expected);
the
excess
23
was
restricted
to
men
less
than
60
years
of
age
and
was
found
mostly
among
active
Novartis
employees.
Further
follow­
up
through
1999
found
17
cases,
yielding
a
6
to
9­
fold
increase
above
the
expected
prostate
cancer
rate.
It
is
noteworthy
that
prostate
cancer
cases
were
largely
limited
to
workers
who
had
been
hired
at
least
10
years
prior
to
disease
development,
allowing
for
a
potential
latency
time,
however,
duration
of
work
was
not
clearly
related
to
excess
risk.
Active
PSA
testing
of
non­
symptomatic
men
began
about
1992.
Five
of
the
prostate
cancer
cases
occurred
before
this
date
and
12
afterwards.
Fourteen
cases
were
localized,
3
were
not
(
by
time
period?)
Mortality
data
showed
1
prostate
cancer
death
and
0.5
expected.

The
study
was
well
designed.
Appropriate
procedures
were
used
to
identify
study
subjects.
The
follow­
up
procedures
for
determining
vital
status
and
cancer
occurrence
while
resident
in
LA
are
adequate.
The
use
of
the
LA
cancer
registry
and
death
certificates
for
case
identification
is
appropriate.
The
use
of
local
population
rates
of
cancer
is
a
standard
approach.
The
decision
to
censor
follow­
up
at
time
of
emigration
from
the
State
is
defensible;
the
proportion
of
subjects
censored
due
to
this
is
relatively
small
and
it
is
unlikely
that
risk
estimates
were
substantially
biased
by
this
procedure.

The
study
does
have
limitations.
The
study
sample
size
is
small,
so
suggestive
excesses
for
a
number
of
suspect
cancers,
such
as
NHL,
can
not
be
adequately
investigated.
Exposure
data
is
inadequate
to
assess
a
relationship
between
occupational
exposure
to
triazines
(
or
other
agents
in
the
study
facility)
and
cancer
risk.
Relationships
with
duration
of
employment
are
difficult
to
interpret
because
it
is
unknown
how
exposures
varied
over
time.
While
a
high
rate
of
PSA
screening
since
1993
among
active
Novartis
employees
may
account
for
much
of
the
prostate
cancer
excess,
the
study
does
not
provide
quantitative
estimates
of
such
an
impact.
Further
analysis
of
risk
by
calendar
period,
comparing
PSA
screening
vs.
non­
screening
time­
periods,
could
provide
some
insight
into
this
important
issue.
Risks
are
unknown
for
the
time
period
prior
to
1985.

Aside
from
increased
risk
associated
with
increasing
age,
race,
and
family
history
of
prostate
cancer,
the
causes
of
this
disease
are
uncertain.
Steroidal
hormones
are
believed
to
play
a
role
because
of
their
importance
in
prostate
development,
prostate
cancer
management,
and
their
successful
use
in
experimental
disease
induction.
While
testosterone
and
its
metabolites
are
the
prime
suspects,
interrelationships
in
prostate
carcinogenesis
with
estrogenic
compounds
may
also
be
important.
Atrazine
and
related
compounds
have
profound
estrogen
disrupting
capacity
in
amphibians
at
a
very
low
dose
(
Hayes
TB
et
al.
PNAS
99:
5476­
5480,
2002).
IARC
judged
the
evidence
on
the
carcinogenicity
of
atrazine
to
be
sufficient
in
animals
but
inadequate
in
humans
(
IARC
Monographs,
Vol.
73,
1999).
There
is
virtually
nothing
known
about
intraprostatic
action
of
atrazine
and
related
compounds
and
the
few
other
epidemiologic
studies
on
triazine­
related
prostate
cancer
risk
in
humans
are
not
informative,
due
to
study
design
limitations.

Conclusion:
While
PSA
screening
may
account
for
much
of
the
excess
of
prostate
cancer
in
this
Triazine
manufacturing
facility,
it
would
be
premature
to
reject
a
potential
role
of
occupational
exposure
to
triazines
as
a
contributing
factor
to
the
observed
excess
of
this
disease.
Future
studies:
Recently,
Syngenta
carried
out
a
limited
exposure
assessment
for
most
of
the
prostate
cancer
cases.
Did
this
assessment
cover
worker
exposures
from
the
beginning
of
24
employment
 
prior
to
1970?
Quantitative
exposure
information
was
not
used
(
and
is
apparently
not
available),
so
the
exposure
scoring
is
on
a
qualitative
scale.
Although
limited,
comparisons
of
exposure
history
of
prostate
cancer
cases
and
non­
cases­­
coupled
with
individual
data
on
PSA
screening­­
could
provide
insight
about
the
reasons
for
the
prostate
cancer
excess
in
this
factory.
In
designing
such
a
study,
I
would
not
match
at
a
4:
1
ratio
on
PSA
and
other
factors,
as
suggested
in
the
Exponent,
Inc.
proposal,
as
questions
will
always
remain
about
the
suitability
of
the
matching
choice
in
this
small
study.
With
only
2,045
subjects,
all
could
be
included
or
a
random
sample
of
some
20%,
allowing
for
reconstruction
of
risks
to
the
full
cohort.
Page
25
Dr.
Aaron
Blair,
National
Cancer
Institute
National
Institutes
of
Health
National
Cancer
Institute
Executive
Plaza
South
8118
6120
Executive
Blvd
MSC
7240
Bethesda
MD
20892­
7240
December
23,
2002
Jerome
Blondell,
Ph.
D.
Health
Effects
Division
(
7509C)
U.
S.
Environmental
Protection
Agency
1200
Pennsylvania
Avenue,
N.
W.
Washington,
DC
20460
Dear
Dr.
Blondell:

I
have
reviewed
the
additional
documents
regarding
the
"
Follow­
up
study
of
cancer
incidence
among
workers
in
triazine­
related
operations
in
the
Novartis
St.
Gabriel
plant"
as
you
requested.
My
comments
are
below:

Evaluation
by
Adami,
Colditz,
Mandel,
and
Trichopoulos:
These
scientists
were
hired
by
Syngenta
to
evaluate
the
Delzell
report
with
specific
instructions
to
"
focus
on
understanding
the
impact
of
screening
associated
with
prostate
specific
antigen
testing."
My
comments
regarding
their
review
are
below.

Page
3
­
The
authors
note
that
there
"
may
have
been
biases
that
could
have
led
to
underestimation
of
expected
number
of
prostate
cancer
cases".
This
is
correct,
but
there
are
also
a
number
of
biases
that
might
lead
to
an
underestimation
of
relative
risk,
particularly
in
the
area
of
exposure
misclassification.
This
can
have
profound
effects
on
estimates
of
relative
risks.

­
I
do
not
think
the
comment
that
"
the
change
in
incidence
requires
the
passage
of
at
least
one
generation"
regarding
migration
is
correct.
For
prostate
and
a
number
of
other
cancers,
the
rates
change
within
the
migrant's
lifetime.

­
It
is
true
that
there
is
no
established
prostate
cancer
risk
factor
that
is
known
to
affect
incidence
differentially
by
age.
However,
there
are
very
few
risk
factors
known,
so
this
is
not
surprising.

Page
6
­
It
seems
to
me
the
incidence
pattern
is
a
bit
more
complex
than
described
here.
The
authors
say
that
the
entire
cancer
excess
(
5
cases
from
46
observed
versus
41
expected)
is
entirely
explained
by
the
prostate
cancer
excess
(
which
is
11
­
4.5
=
6.5).
It
is
also
correct
to
say
that
all
cancers
except
colon,
lung
and
breast
in
Table
7
have
a
larger
observed
than
expected
number.
Prostate
could
account
for
the
small
excess,
but
so
could
other
combinations
of
sites.
In
Table
7
Page
26
if
you
add
the
observed
and
expected
number
for
the
specific
cancers
listed,
i.
e.,
everything
but
"
Other"
you
get
37
observed
and
29.7
expected.
This
means
that
for
the
"
Other"
there
are
4
observed
and
11.3
expected.
This
is
about
what
one
would
expect
for
a
cohort
where
the
overall
cancer
rate
is
about
1.0,
i.
e.,
some
sites
with
SIRs
above
1.0
and
some
below,
but
without
other
information
it
is
not
appropriate
to
single
out
one
cancer
as
explaining
the
excess
or
deficit.

­
The
proportion
of
localized
prostate
cancers
is
an
important
indicator
of
the
effect
of
screening.
The
higher
rate
in
the
cohort
than
the
general
LA
population
suggests
a
screening
effect.
There
is
some
difference
in
the
proportion
of
prostate
cancer
that
is
localized
by
age,
and
it
was
not
clear
if
the
authors
took
this
into
account
in
this
comparison.

Page
7
­
It
is
suggested
that
use
of
5­
year
age
groupings
would
underestimate
the
expected
number
of
cases.
This
is
probably
true,
but
without
some
quantification
of
the
magnitude
of
this
effect,
it
is
difficult
to
evaluate
its
impact.
It
is
likely
to
be
small.

­
Incidence
rates
were
not
available
for
the
years
1998
and
1999.
With
rising
prostate
cancer
rates
this
could
also
underestimate
the
expected
number.
Again
the
impact
is
likely
to
be
small
because
the
greatest
affect
of
screening
on
prostate
cancer
was
in
the
early
1990s.
This
problem
can
be
eliminated
through
analysis
of
cancer
incidence
in
the
cohort
only
through
1997.

Page
8
­
The
authors
state
that
since
the
prostate
cancer
excess
can
account
for
the
total
cancer
excess
and,
therefore,
the
comment
on
excesses
for
other
cancers
by
Dr.
Delzell
is
"
unintentionally
misleading."
It
would
seem
that
this
comparison
could
also
be
turned
around.
The
excesses
for
some
other
cancers
can
account
for
the
total
cancer
excess.
It
is
unclear
why
it
should
only
be
viewed
from
the
prostate
cancer
perspective.

Page
9
­
The
authors
say
that
"
small,
histopathologically
malignant,
lesions
in
the
prostate
among
middle­
aged
and
particularly
elderly
men"
are
very
high.
It
was
my
impression
that
this
applied
only
to
elderly
men.
Nonetheless,
it
underscores
the
reason
why
screening
for
prostate
cancer
may
uncover
malignant
lesions
that
would
not
necessarily
progress
to
diagnosis
under
normal
conditions.

Page
20
­
The
authors
provide
a
useful
procedure
to
estimate
the
number
of
prostate
cancer
cases
that
might
be
due
to
the
company
PSA
screening.
It
seems
reasonable
that
repeat
screening
would
increase
the
number
of
cases
identified,
but
this
would
surely
have
diminishing
returns.
It
was
my
impression
that
Auvinen
recommended
that
screening
did
not
need
to
take
place
every
year.
This
would
indicate
that
each
repeat
year
identifies
fewer
cases.
The
authors
appear
to
have
taken
this
into
account,
but
it
was
not
entirely
clear
to
me.
More
information
on
how
the
authors
arrived
at
the
value
for
X
of
3.5
would
be
helpful,
since
it
is
this
value
that
is
critical
for
determining
if
screening
can
explain
the
excess
cases
at
the
plant.

Page
21
­
The
authors
state
that
environmental
factors
may
have
profound
effects
on
prostate
cancer
incidence,
but
these
"
factors
are
likely
to
operate
early
in
life."
However,
later
they
say
that
"
there
is
growing
evidence
that
external
factors
may
have
a
stronger
influence
on
advanced,
Page
27
rather
than
earlier,
stages
of
the
natural
history
of
prostate
cancer."
These
points
seem
contradictory.

The
review
by
Adami
et
al.
provides
useful
information
to
evaluate
the
impact
of
company
PSA
screening
on
the
prostate
cancer
incidence
rates
in
the
cohort.
They
estimated
the
number
of
cases
likely
to
occur
from
the
PSA
screening
using
information
from
the
literature.
This
is
an
appropriate
approach.
The
results
suggest
that
PSA
screening
may
well
explain
the
excess
incidence
of
prostate
cancer
in
this
cohort.
It
would
be
helpful,
however,
to
have
more
information
supporting
the
selection
of
multiplication
factor
regarding
the
impact
of
age
and
other
factors
that
might
differ
between
a
cohort
of
working
individuals
and
the
general
population.

Summary
Information
by
Charles
Breckenridge:
This
document
provides
information
on
atrazine
exposure
for
some
of
the
prostate
cancer
cases
in
the
Delzell
report.
This
report
provides
no
useful
information
regarding
the
issue
of
atrazine
exposure
and
prostate
cancer
because
it
only
includes
cases.
To
effectively
use
information
about
amount
of
exposure
among
the
cases
it
is
necessary
to
have
a
noncase
group
for
comparison.
If
it
is
possible
to
reconstruct
possible
atrazine
exposure
for
cases,
as
was
done
for
this
report,
it
could
be
done
for
controls
also.
This
would
be
helpful,
although
the
small
number
of
cases
would
be
a
severe
limitation.

National
Resources
Defense
Council
(
NRDC)
Comments:
The
NRDC
raises
a
couple
of
methodologic
issues
about
the
cohort.
They
note
that
more
than
200
workers
were
excluded
from
the
study
because
they
presumedly
left
Louisiana
before
1988.
This
is
appropriate
because
the
time
frame
for
the
cohort
follow­
up
was
1988
through
1997.
They
are
correct
that
if
these
workers
had
longer
or
heavier
exposure
the
results
might
have
been
different,
but
this
is
always
the
situation.
A
cohort
can
only
make
clear
statements
about
the
exposures
being
studied.
A
second
point
was
that
contract
workers
with
lower
levels
of
exposure
were
included
and
this
could
lower
relative
risks
because
of
exposure
dilution.
This
could
also
be
true,
but
it
is
not
a
bias
in
the
study,
however
it
does
limit
extrapolation
of
study
results
to
other
scenerios.
Both
of
the
NRDC
points
underscores
the
absolute
necessity
of
performing
an
atrazine
exposure
assessment
in
the
study.

Conclusion:
To
clearly
understand
the
issue
of
prostate
cancer
and
atrazine
exposure
in
this
cohort
it
is
essential
that
a
quantitative
exposure
assessment
be
added.
The
approach
described
by
Breckenridge
for
a
few
of
the
cases
is
a
reasonable
starting
point.
A
few
additional
analyses
of
prostate
cancer
risk
before
and
after
initiation
of
screening
would
also
provide
direct
information
on
the
impact
of
the
screening.

Please
feel
free
to
contact
me
if
you
have
questions.

Sincerely,
Aaron
Blair,
Ph.
D.
Chief,
Occupational
Epidemiology
Branch
