Breast cancer screening programmes: The development of a monitoring and evaluation system

Abstract

It is important that the introduction of breast screening is closely monitored. The anticipated effect on breast cancer mortality will take 10 years or more fully to emerge, and will only occur if a succession of more short-term end points are met. Data from the Swedish two-county randomised trial provide targets that should be achieved, following a logical progression of compliance with the initial invitation, prevalence and stage distribution at the prevalence screen, the rate of interval cancers after the initial screen, the pick-up rate and stage distribution at later screening tests, the rate of interval cancers after later tests, the absolute rate of advanced cancer and finally the breast cancer mortality rate. For evaluation purposes, historical data on stage at diagnosis is desirable; it is suggested that tumour size is probably the most relevant variable available in most cases.

Full text

Br.
J.
Cancer
(1989),
59,
954-958
©
The
Macmillan
Press
Ltd.,
1989
Breast
cancer
screening
programmes:
the
development
of
a
monitoring
and
evaluation
system
N.E.
Day',
D.R.R.
Williams2
&
K.T.
Khaw2
1MRC
Biostatistics,
5
Shaftesbury
Road,
Cambridge
CB2
2BW
and
2Department
of
Community
Medicine,
Cambridge,
UK.
Summary
It
is
important
that
the
introduction
of
breast
screening
is
closely
monitored.
The
anticipated
effect
on
breast
cancer
mortality
will
take
10
years
or
more
fully
to
emerge,
and
will
only
occur
if
a
succession
of
more
short-term
end
points
are
met.
Data
from
the
Swedish
two-county
randomised
trial
provide
targets
that
should
be
achieved,
following
a
logical
progression
of
compliance
with
the
initial
invitation,
prevalence
and
stage
distribution
at
the
prevalence
screen,
the
rate
of
interval
cancers
after
the
initial
screen,
the
pick-up
rate
and
stage
distribution
at
later
screening
tests,
the
rate
of
interval
cancers
after
later
tests,
the
absolute
rate
of
advanced
cancer
and
finally
the
breast
cancer
mortality
rate.
For
evaluation
purposes,
historical
data
on
stage
at
diagnosis
is
desirable;
it
is
suggested
that
tumour
size
is
probably
the
most
relevant
variable
available
in
most
cases.
Screening
for
breast
cancer
is
being
introduced
in
Britain
on
the
recommendation
of
the
Forrest
Report.
Experience
from
the
cervical
cancer
screening
programme
has
demonstrated
that
the
performance
of
a
national
programme
may
fall
below
expectation
based
on
experience
from
specialist
UK
centres,
or
from
other
countries.
It
is
therefore
important
to
monitor
the
performance
of
the
national
breast
cancer
screening
programme
from
its
inception,
to
determine
how
closely
the
benefits
it
achieves
approach
the
benefits
seen
in
the
randomised
trials
and
population
demonstration
projects,
the
results
of
which
formed
the
basis
for
the
Forrest
Report's
recommendation.
The
most
relevant
of
the
trials
which
have
so
far
reported
results
is
the
Swedish
two-county
study
(Tabar
et
al.,
1985),
for
a
number
of
reasons.
It
used
mammography
as
the
sole
screening
modality,
and
for
the
age
group
of
relevance,
women
aged
50-64
years,
the
average
inter-screening
interval
(33
months)
was
similar
to
the
3-year
interval
to
be
adopted
in
Britain.
It
also
screened
over
30,000
women
in
this
age
group,
compared
to
the
15,000
or
fewer
screened
in
Florence
(Palli
et
al.,
1986),
Nijmegen
(Verbeek
et
al.,
1984)
or
Utrecht
(de
Waard
et
al.,
1984).
It
is
thus
of
interest
to
examine
the
different
evaluation
measures
that
emerged
from
the
Swedish
study,
to
identify
the
information
on
which
these
measures
are
based
and
at
what
stage
in
the
trial
this
information
became
available.
The
emphasis
in
this
paper
is
on
the
fundamental
effect
measure,
breast
cancer
mortality.
Measures
relating
to
other
important
aspects
of
the
screening
programme,
for
example
costs,
quality
of
care
or
the
value
of
different
diagnostic
procedures,
are
not
considered.
Results
of
the
Swedish
two-county
study
The
three
main
criteria
one
can
use
in
evaluating
the
effect
of
the
screening
programme
are:
(1)
changes
in
mortality;
(2)
changes
in
the
absolute
rate
of
advanced
disease;
(3)
parameters
of
the
screening
process,
comprising
both
the
screening
test
and
the
further
diagnostic
procedures
applied
to
women
positive
on
the
screening
test
-
these
parameters
include
sensitivity,
specificity,
the
distribution
of
lead
time
(the
length
of
time
diagnosis
is
advanced
by
screening)
and
sojourn
time
(the
length
of
time
preclinical
lesions
are
detectable
by
screening),
and
the
predictive
value
for
malignancy.
Mortality
of
course
is
the
basic
evaluation
measure.
In
the
Swedish
study,
however,
no
difference
between
the
study
and
Correspondence:
N.E.
Day.
Received
21
November
1988,
and
in
revised
form,
23
January
1989.
control
group
was
seen
until
the
fourth
year.
It
was
not
until
the
end
of
the
seventh
year
that
the
gap
had
widened
sufficiently,
and
adequate
numbers
accrued,
for
one
to
be
satisfied
that
breast
cancer
mortality
had
been
reduced.
Similarly,
in
Utrecht
(Collette
et
al.,
1984)
and
Nijmegen
(Verbeek
et
al.,
1984),
at
least
7
years
elapsed
after
the
start
of
screening
before
the
effect
on
mortality
was
able
to
be
assessed.
Although
similar
evaluation
will
be
necessary
in
this
country,
7
years
or
more
is
a
long
time
to
wait
before
one
can
determine
whether
the
programme
is
effective.
Earlier
measures
are
required.
The
effect
on
mortality
is
the
result
of
earlier
diagnosis,
which
is
seen
in
the
reduction
in
4he
rate
of
advanced
disease.
This
reduction
in
advanced
disease,
if
it
occurs,
will
be
detectable
earlier
than
the
reduction
in
mortality.
Figure
1
gives
the
corresponding
results
for
advanced
disease
(Figure
la)
and
breast
cancer
deaths
(Figure
lb) for
the
group
aged
50-59
years
at
study
entry
from
the
two-county
study.
One
can
see
that
the
gap
between
the
two
curves
begins
to
appear
some
2
years
earlier
for
the
advanced
cancers
than
for
the
deaths.
The
reduction
in
advanced
disease
results
from
earlier
diagnosis
and
so
depends
on
the
lead
time
distribution
of
cases
diagnosed
by
screening.
This
distribution
expresses
quantitatively
the
length
of
time
by
which
diagnosis
has
been
advanced.
It
is
reflected
in
the
incidence
of
interval
cancers
among
screened
women
in
the
years
following
a
negative
screening
test
(Day
&
Walter,
1984).
To
be
informative,
the
incidence
of
interval
cancers
needs
to
be
expressed
as
a
proportion
of
the
incidence
that
would
have
been
expected
in
the
absence
of
screening,
as
shown
in
Figure
2
from
the
two-county
study
(Tabar
et
al.,
1987a).
The
difference
between
the
incidence
rate
of
interval
cancers
and
the
rate
expected
in
the
absence of
screening
reflects
the
number
of
cancers
with
a
diagnosis
that
was
advanced
to
the
previous
screening
test.
An
initial
indication
of
the incidence
rate
of
interval
cancers
(i.e.
as
in
Figure
2)
is
therefore
given
by
the
prevalence
rate
of
cancers
detected
at
the
first
screen.
As
with
interval
cancers,
this
rate
is
more
informatively
expressed
if
divided
by
the
incidence
rate
expected
in
the
absence of
screening
in
women
presenting
for
screening.
Results
from
the
two-county
study
are
given
in
Table
I.
Since,
however,
some
of
these
cancers
may
not
have
been
destined
to
surface
clinically
until
much
later,
if
at
all,
and
may
have
low
malignant
potential,
this
prevalence
is
not
an
adequate
surrogate
measure
of
the
rate
of
interval
cancers.
Both
need
to
be
considered.
The
more
favourable
stage
distribution
obtained
in
the
group
invited
for
screening
arises
because
the
cancers
whose
diagnosis
was
brought
forward
in
time
by
early
detection,
Br.
J.
Cancer
(1989),
59,
954-958
"-.
The
Macmillan
Press
Ltd.,
1989

BREAST
CANCER
SCREENING
PROGRAMMES
955
*---
-
Control
*
*
Study
b
Years
after
randomisation
Figure
1
(a)
Cumulative
rates
per
104
women
of
advanced
cancers
and
(b)
cumulative
mortality
rates
per
105
women
in
the
Swedish
2-county
study.
100-
0
C)
-c
4,0
C
0
0
.-
C
0
c
'a
._
C)
Cu
U)
Cu
0
0)
.0
0
a)
0-
Time
since
previous
screening
(months)
Figure
2
Incidence
of
breast
cancer
among
screeped
women.
are
diagnosed
at
an
earlier
stage.
Table
II
gives
the
proportion
of
stage
II
or
worse
of
cancers
detected
at
the
first
screen,
of
interval
cancers,
of
cancers
detected
at
the
second
or
later
screen,
and
of
cancers
seen
in
the
control
group
(Tabar
et
al.,
1987).
The
similarity
of
the
interval
and
the
control
group
cancers
is
striking.
Table
II
also
gives
the
size
distribution
of
screen-detected
cancers
and
of
cancers
diagnosed
in
the
control
group.
That
the
screen-detected
Table
I
Swedish
two-county
study:
prevalence
per
1,000
women
of
breast
cancer
detected
at
the
initial
screening
test
Underlying
incidence
Prevalence
rate
per
1,000
Prevalencel
Age
group
per
1,000
person
years
incidence
50-59
4.63
1.50
3.09
60-69
9.08
1.98
4.59
Table
II
Stage
distribution
by
means
of
detection:
age
group
50-69
Swedish
two-county
study
Initial
Second
or
Interval
Control
Age
group
screen
later
screen
cancers
group
Proportion
of
stage
II
or
worse
cancers
50-59
33.3
25.0
55.3
58.7
60-69
34.3
16.9
58.5
58.4
Tumour
size
(mm)
1-9
10-14
15-19
>
20
Total
Distribution
(%)
of
tumour
size,
invasive
cancers
only
Screen-detected
23
29
22
26
414
Control
group
7
15
18
59
461
20

956
N.E.
DAY
et
al.
cancers
should
have
a
more
favourable
stage
distribution
and
be of
smaller
size
is
a
prerequisite
for
the
subsequent
deficit
of
advanced
cancers
in
the
group
allocated
to
screening,
necessary
but
not
sufficient.
Finally,
the
effect
of
the
programme
on
the
subsequent
rates
of
advanced
disease
and
mortality
will
depend
directly
on
the
proportion
of
the
target
population
who
present
for
screening.
Compliance
rate
is
clearly
an
important
initial
measure
-
necessary
but
not
sufficient
-
of
programme
effectiveness.
In
the
Swedish
two-county
study,
compliance
in
the
age
group
50-64
was
of
the
order
of
90%.
Implications
for
the
information
requirements
of
a
regional
and
national
evaluation
system
The
foregoing
description
of
the
process
whereby
screening
leads
to
a
reduction
in
breast
cancer
mortality
pinpoints
the
information
required
to
determine
whether
the
programme
is
on
course.
Table
III
summarises
a
minimum
set
of
measures
that
an
information
system
should
monitor
to
evaluate
the
effectiveness
of
the
programme
in
reducing
severity
of
and
mortality
from
the
disease.
The
first
three
measures
(com-
pliance,
screening
characteristics
and
rate
of
advanced
cancers)
are
not
in
themselves
sufficient
to
demonstrate
a
reduction
in
mortality.
A
favourable
value
for
each
of
these
measures
is
necessary,
however,
if
an
acceptable
effect
on
breast
cancer
mortality
is
to
be
achieved.
Poor
performance
indicates
where
remedial
action
is
required.
The
information
required
to
monitor
these
performance
measures
is
described
below.
Compliance
rate
It
is
important
that
the
real
compliance
rate
is
measured,
i.e.
the
proportion
who
present
for
screening
among
the
women
invited
who
are
both
alive
and
resident
in
the
catchment
population.
One
needs
to
ascertain
the
accuracy
of
the
population
lists
that
are
used.
Characteristics
of
the
screening
procedures
Prevalence
rate
at
the
first
screen
This
measure
should
be
by
5-year
age
group,
since
rates
increase
rapidly
with
age.
One
has
the
approximate
relationship:
Prevalence
rate
at
first
screen
sensitivity
x
average
sojourn
Expected
incidence
rate
time
This
expression
indicates
that
to
be
informative
in
terms
of
the
underlying
screening
parameters,
and
so
for
comparison
with
other
programmes,
the
prevalence
rate
needs
to
be
expressed
as
a
multiple
of
the
expected
annual
incidence
rate
in
screened
women
(i.e.
the
rate
one
would
have
seen
in
the
Table
III
Measure
of
performance,
satisfactory
values
of
which
are
necessary
to
obtain
an
acceptable
reduction
in
breast
cancer
mortality
Measure
Rationale
Compliance
rate
Mortality
reduction
in
target
population
directly
related
to
%
compliance
Screening
characteristics,
in
To
achieve
results
comparable
to
particular:
the
Swedish
study,
the
screening
test
sensitivity
as
performed
should
be
comparable
lead
time
distribution
in
its
characteristics
sojourn
time
distribution
stage
distribution
among
screen-detected
cancers
compared
to
that
among
a
comparable
group
of
clinically
diagnosed
cancers
Rate
of
advanced
cancers
Earlier
surrogate
of
mortality
rates
Breast
cancer
mortality
absence
of
screening).
This
incidence
rate
is
not
directly
observable,
but
it
can
be
derived
from
the
expected
rate
in
the
total
population
and
the
rate
in
non-attenders.
The
rate
in
non-attenders
is
directly
observable
provided
that
the
population
is
covered
by
cancer
registration
of
high
quality,
and
that
the
non-attenders
are
well
identified.
For
the
latter
one
needs
to
know,
among
the
women
who
were
invited
but
did
not
attend,
the
proportion
alive
and
living
in
the
catchment
area
(as
for
the
assessment
of
compliance).
Estimates
of
the
rate
in
the
total
population,
which
is
not
directly
observable,
can
be
obtained
either
from
rates
in
comparable,
neighbouring
unscreened
populations
or
from
historical
incidence
data.
Both
require
cancer
registration
and
the
latter
requires
the
existence
of
good
quality
cancer
registration
in
previous
years.
The
expected
rate
among
the
attenders
is
then
obtained
from
the
identity:
Incidence
rate
in
total
population
=Pxincidence
rate
in
attenders
+
(1
-
P)
x
incidence
rate
in
refusers
where
P
is
the
real
compliance
rate
(expressed
as
a
proportion).
Incidence
of
interval
cancers
Registration
of
interval
cancers
requires
coverage
of
the
population
by
good
cancer
regist-
ration.
As
noted
before,
it
is
important
to
express
the
rate
of
interval
cancers
as
a
proportion
of
the
expected
rate
in
the
screened
group,
which
requires
the
expected
incidence
rate
in
the
total
population
and
the
incidence
rate
among
non-attenders.
Comparison
of
the
interval
cancer
rates
and
the
initial
prevalence
rates
with
those
seen
in
the
two-county
study
will
indicate
whether
the
following
parameters
of
the
screening
process
(i.e.
screening
test
and
associated
diagnostic
pro-
cedures)
are
comparable
to
those
seen
in
an
effective
pro-
gramme:
(1)
sensitivity;
(2)
distribution
of
sojourn
time
and
lead
time;
(3)
'overdiagnosis'
of
breast
cancer
-
this
appeared
to
be
absent
from
the
two-county
study
(Day
et
al.,
1988),
but
has
been
suspected
elsewhere.
It
would
be
surprising
if
comparable
values
for
sensitivity
and
the
sojourn
time
distribution
did
not
lead
to
comparable
effects
on
mortality
and
advanced
disease.
Stage
distribution
of
screen
detected
cancers
As
can
be
seen
from
Table
II,
the
stage
distribution
of
cancers
detected
at
the
first
screen
may
differ
from
that
seen
at
later
screens.
The
definition
of
stage
needs
to
take
account
of
the
infor-
mation
likely
to
be
available
in
the
majority
of
cases.
Tumour
size
may
be
an
acceptable
substitute
for
stage,
as
discussed
in
the
next
section.
The
stage
(or
tumour
size)
distribution
of
screen
detected
cancers
needs
to
be
compared
to
the
stage
distribution
one
would
have
expected
in
the
absence
of
screening
among
women
who
presented
for
screening.
This
latter
distribution
can
be
obtained
from
the
stage
distribution
of
cancers
among
non-attenders
and
that
of
cancers
in
the
total
population
before
the
start
of
the
programme.
Absolute
rates
of
advanced
cancers
There
are
two
problems
in
the
use
for
evaluation
of
the
rate
of
advanced
cancers.
First
is
the
definition
of
an
advanced
cancer.
Second
is
the
choice
of
comparison
groups.
Definition
of
an
advanced
cancer
In
the
Swedish
two-county
study,
'advanced'
meant
stage
II
or
worse
with
histological
examination
of
the
nodes.
Thus
a
stage
I
cancer
had
to
be
less
than
20
mm
diameter,
and
no
involvement
of
the
nodes,
with
an
adequate
number
examined.
Although
screen-
detected
cancers
may
be
sufficiently
investigated
to
give
acceptable
stage
information,
many
cancers
diagnosed
clini-
cally
will
not
be.
Any
comparison
group
will
clearly
be
formed
of
the
latter,
and
cancer
registry
information
will
be

BREAST
CANCER
SCREENING
PROGRAMMES
957
Table
IV
Monitoring
measures
and
the
associated
information
requirements
Measure
Qualifying
comments
Additional
information
required
Type
of
evaluation
provided
Compliance
rate
Validation
of
population
list
Identification
of
real
non-compliance
Indicates
potential
for
effectiveness
of
the
overall
programme
Prevalence
rate
at
Expressed
as
multiple
of
expected
Incidence
rates
in
non-compliers
and)
initial
screening
test
incidence
rate
in
screened
women
in
a
comparable
unscreened
group,
Provide
estimates
of
sensitivity,
lead
e.g.
historical
rates
tIme
sojoum
etimaes
and
predictivlea
Rate
of
interval
Expressed
as
a
proportion
of
Accurate
identification
of
interval
time,
sojourn
time
and
predictive
cancers
expected
incidence
rate
in
screened
cancers,
and
calculation
of
J
value
women,
and
by
time
since
the
last
additional
incidence
rates
as
above
screening
test
Stage
(or
size)
Compared
to
expected
stage
Stage
(or
size)
distribution
in
non-
Indicates
potential
for
reduction
in
distribution
of
screen-
distribution
in
the
absence
of
compliers
and
in
total
population
absolute
rate
of
advanced
cancer
detected
cancers:
screening
before
screening
started
(1)
at
initial
screen;
(2)
at
subsequent
screen
Rate
of
advanced
Need
for
a
definition
of
'advanced'
Stage
(or
tumour
size)
information
Earlier
surrogate
of
mortality
cancers
which
can
be
used
for
the
great
needed
historically,
and
on
cancers
majority
of
cases
given
the
among
non-compliers
information
available.
Probably
based
on
tumour
size
Breast
cancer
death
Breast
cancer deaths
linked
to
date
Final
evaluation
rate
of
diagnosis
needed
for
staging.
Examination
of
retrospective
data
in
the
East
Anglian
cancer
registry
(de
Bono
&
Kingsley-Pillers,
1978),
indicates
that
the
most
frequent
information
missing
for
staging
purposes
is
status
of
the
nodes,
which
makes
the
important
distinction
between
stage
I
and
stage
II
imposs-
ible.
Tumour
size
is
frequently
present,
and
when
absent
there
are
often
clear
indications
that
the
tumour
is
inoper-
able
and
advanced.
That
is,
for
cancers
small
enough
to
be
stage
I,
information
on
tumour
size
is
only
rarely
absent.
Use
of
tumour
size
as
a
surrogate
for
stage
can
be
argued
for
since:
(1)
tumour
size
is
of
strong
prognostic
value
in
its
own
right;
(2)
the
proportion
with
positive
nodes
is
strongly
related
to
tumour
size
and
this
relationship
is
the
same
in
screen
detected
as
in
clinically
detected
cases
(Tabar
et
al.,
1987b).
This
last
point
suggests
that
screen
detected
small
cancers
are
similar
in
behaviour
to
clinically
detected
small
cancers.
If
tumour
size
is
used,
it
can
either
be
dichotomised
at
say
15
or
20mm,
to
give
'early'
and
'advanced'
cancers,
or
subdivided
more
finely.
The
first
approach
is
simpler
and
may
prove
equally
informative.
Table
V
Suggested
levels
beyond
which
corrective
action
is
strongly
indicated
Measure
Acceptable
level
Compliance
rate
No
less
than
60%
Prevalance
rate
at
first
No
less
than
three
times
the
underlying
screening
test
incidence
rate
Rate
of
interval
cancers
No
more
than
25%
of
expected
incidence
in
first
2
years
after
a
negative
test,
and
no
more
than
60%
of
expected
incidence
in
the
third
year
Stage
distribution
of
screen-detected
cancers
At
first
test
No
more
than
40%
stage
II
or
more
advanced
At
subsequent
tests
No
more
than
30%
stage
II
or
more
advanced
Reduction
in
rate
of
No
less
than
30%
in
target
population,
advanced
cancers
seven
years
after
first
invitation
sent
Reduction
in
breast
No
less
than
25%
in
target
population
cancer
mortality
rates
free
from
breast
cancer
when
first
invitation
sent,
10
years
after
programme
starts
Choice
of
comparison
groups
Three
approaches
are
possible:
(1)
The
target
population
can
be
compared
with
historical
data
covering
the
same
age
group
and
geographic
area.
A
figure
comparable
to
Figure
la
would
result,
where
the
'control'
group
data
are
replaced
by
expected
numbers
for
the
target
population
based
on
historical
cancer
registry
data.
There
is
the
possibility
of
confounding
with
secular
change
in
stage
of
presentation,
but
the
historical
data
can
be
examined
for
such
changes.
One
can
also
examine
current
data
in
age
groups
outside
those
targeted
for
screening,
for
any
indication
of
secular
trends.
(2)
The
target
population
can
be
compared
with
a
geographically
neighbouring
popu-
lation.
This
comparison
can
only
be
made
while
the
programme
is
being
introduced,
and
is
thus
of
limited
usefulness.
It
would
require
incidence
rates
by
small
geo-
graphic
area.
(3)
A
comparison
of
screened
with
unscreened
women
using
either
a
case
control
approach
or
data
from
the
entire
cohort.
This
comparison
evidently
runs
a
serious
risk
of
bias.
One
would
need
to
compare
rates
in
the
unscreened
women
with
historical
data,
to
assess
selection
biases
affecting
both
underlying
incidence
rates
and
stage
at
presentation.
In
the
Utrecht
case-control
study
of
breast
cancer
deaths,
these
comparisons
were
made
and
bias
was
thought
to
be
small.
In
the
HIP
study
and
the
two-county
study,
selection
bias
was
strong
but
acted
in
different
directions.
In
New
York,
unscreened
women
were
at
low
risk
for
breast
cancer,
in
Sweden
unscreened
women
presented
with
particularly
late
stage
cancers.
This
approach
avoids
problems
due
to
secular
trends,
so
that
combining
it
with
the
comparison
with
historical
data
strengthens
both.
Such
a
combined
approach
has
recently
been
adopted
in
a
further
analysis
of
the
Utrecht
study.
Breast
cancer
mortality
Evaluation
of
the
effect
on
mortality
can
take
the
three
approaches
described
in
the
previous
section,
but
concentrat-
ing
on
breast
cancer
deaths
occurring
among
breast
cancer
cases
diagnosed
after
the
start
of
the
screening
programme.
For
this
purpose,
date
of
diagnosis
will
be
required
for
all
breast
cancer
deaths
in
the
region
for
a
number
of
years
before
the
trial
starts.
Mortality
comparison
can
then
be
constructed
equivalent
to
Figure
lb
using
historical
infor-
mation
for
the
controls.

958
N.E.
DAY
et
al.
Conclusions
The
scheme
for
an
information
system
described
above
is
shown
in
Table
IV
with
the
time
sequence
in
Figure
3.
It
follows
the
process
of
screening
from
the
start,
the
identifica-
tion
of
the
target
population,
to
the
final
evaluation
mea-
sure,
the
effect
on
breast
cancer
mortality.
The
information
measures
described
plot
the
course
that
the
programme
has
to
follow
to
achieve
the
results
on
breast
cancer
mortality
expected
from
the
Swedish
randomised
trial.
Table
V
pro-
poses
minimum
levels
of
performance
for
each
of
these
measures.
The
only
aspect
not
considered
is
treatment;
it
is
clear
that
a
reduction
in
mortality
will
result
from
the
achievement
of
earlier
diagnosis
only
if
the
early
lesions
are
adequately
treated.
Years
since
start
of
programme
0
1
2
3
4
5
6 7
8
9
10
First
Screening
Round
Second
Screening
Round
Third
Screening
Round
Compliance
at
1
st
round.
Compliance
at
2nd
round.
Prevalence
and
stage
Prevalence
and
stage
distribution
distribution
at
first
at
second
round.
screening
test.
Incidence
rate
of
Incidence
rate of
interval
cancers
interval
cancers
after
first
round.
after
second
round.
Evaluation
in
terms
of
Evaluation
in
terms
rate
of
advanced
cancers.
of
mortality
Figure
3
References
DE
BONO,
A.M.
&
KINGSLEY-PILLERS,
E.M.
(1978).
Carcinoma
of
the
breast
in
East
Anglia
1960-75:
a
changing
pattern
of
presentation.
J.
Epidemiol.
Commun.
Health,
32,
178.
COLETTE,
H.J.A.,
ROMBACH,
J.J.,
DAY,
N.E.
&
DE
WAARD,
F.
(1984).
Evaluation
of
screening
for
breast
cancer
in
a
non-randomized
study
(the
DOM
project)
by
means
of
a
case-control
study.
Lancet,
i,
1224.
DAY,
N.E.
&
WALTER,
S.D.
(1984).
Simplified
models
for
screening:
estimation
procedures
from
mass
screening
programmes.
Bio-
metrics,
40,
1.
DAY,
N.E.,
WALTER,
S.D.,
TABAR,
L.,
FAGERBERG,
C.J.G.
&
COLLETTE,
H.J.A.
(1988).
The
sensitivity
and
lead
time
of
breast
cancer
screening:
a
comparison
of
the
results
of
different
studies.
In
Screening
for
Breast
Cancer,
Day,
N.E.
&
Miller,
A.B.
(eds).
Hans
Huber:
Toronto.
PALLI,
D.,
ROSELLI
DEL
TURCO,
M.,
BUIATTI,
E.
and
4
others
(1986).
A
case-control
study
of
the
efficacy
of
a
non-randomized
breast
screening
programme
in
Florence
(Italy).
Int.
J.
Cancer,
38,
501.
TABAR,
L.,
GAD,
A.,
HOLMBERG,
L.H.
and
9
others
(1985).
Reduc-
tion
in
breast
cancer
mortality
by
mass
screening
with
mammo-
graphy:
first
results
of
a
randomised
trial
in
two
Swedish
countries.
Lancet,
i,
829.
TABAR,
L.,
FAGERBERG,
C.J.G.,
DAY,
N.E.
&
HOLMBERG,
L.
(1987a).
The
Swedish
two-county
breast
cancer
screening
trial:
update
and
initial
results
on
the
screening
interval.
Br.
J.
Cancer,
55,
547.
TABAR,
L.,
DUFFY,
S.W,
&
KRUSEMO,
U.B.
(1987b).
Detection
method,
tumour
size
and
node
metastases
in
breast
cancers
diagnosed
during
a
trial
of
breast
cancer
screening.
Eur.
J.
Cancer
Clin.
Oncol.,
23,
959.
VERBEEK,
A.L.M.,
HENDRICKS,
J.H.C.L.,
HOLLAND,
R.,
MRAVUNAC,
M.,
STURMENS,
F.
&
DAY,
N.E.
(1984).
Reduction
of
breast
cancer
mortality
through
mass
screening
with
modem
mammography:
first
results
of
the
Nijmegen
project
1975-81.
Lancet,
i,
1222.
DE
WAARD,
F.,
COLLETTE,
H.J.A.,
ROMBACH,
J.J.,
BANDERS
VAN
HALEWIJN,
E.A.
&
HONING,
C.
(1984).
The
DOM
project
for
the
early
detection
of
breast
cancer,4"'Utrecht,
the
Netherlands.
J.
Chronic
Dis.,
37,
1.