Comparison of serum tumor markers in the diagnosis of colorectal carcinoma

Abstract

The assessment of the diagnostic power of four serum tumour markers, CEA, CA 19-9, CA 50 and CA 195 for colorectal carcinoma is described, according to recently formulated guidelines. Preoperative serum concentrations of the four markers were determined in 198 colorectal cancer patients and 57 patients with a benign colorectal disorder. The cumulative frequency distributions of the malignant and benign group show strong overlap for all markers, which indicates low diagnostic ability. This is confirmed by the Receiver Operating Characteristic curves, which have areas under the curve of 0.65 (95% confidence interval (CI) 0.58-0.73) for CA 19-9, CA 50 and CA 195 and of 0.70 (95%) CI 0.63-0.77) for CEA. The new tumour markers appear to be of slightly less diagnostic value than CEA for the primary diagnosis of colorectal cancer, although the discrepancy is not statistically significant. The low diagnostic power of CA 19-9, CA 50 and CA 195 may be due to a high proportion of colorectal cancer patients having the Lewis(a-b-) phenotype, who cannot synthesise these markers.

Full text

Br.
J.
Cancer
(1992),
66,
148-154
©
Macmillan
Press
Ltd.,
1992
Comparison
of
four
serum
tumour
markers
in
the
diagnosis
of
colorectal
carcinoma
Y.T.
van
der
Schouwl,
A.L.M.
Verbeek',
Th.
Wobbes2,
M.F.G.
Segers3
&
C.M.G.
Thomas3'4
'Department
of
Medical
Informatics
and
Epidemiology,
University
of
Nijmegen,
PO
Box
9101,
6500
HB
Nijmegen,
The
Netherlands;
2Department
of
General
Surgery,
3Laboratory
for
Endocrinology
and
Reproduction,
4Department
of
Obstetrics
and
Gynaecology,
University
Hospital
Nijmegen,
PO
Box
9101,
6500
HB
Nijmegen,
The
Netherlands.
Summary
The
assessment
of
the
diagnostic
power
of
four
serum
tumour
markers,
CEA,
CA
19-9,
CA
50
and
CA
195
for
colorectal
carcinoma
is
described,
according
to
recently
formulated
guidelines.
Preoperative
serum
concentrations
of
the
four
markers
were
determined
in
198
colorectal
cancer
patients
and
57
patients
with
a
benign
colorectal
disorder.
The
cumulative
frequency
distributions
of
the
malignant
and
benign
group
show
strong
overlap
for
all
markers,
which
indicates
low
diagnostic
ability.
This
is
confirmed
by
the
Receiver
Operating
Characteristic
curves,
which
have
areas
under
the
curve
of
0.65
(95%
confidence
interval
(CI)
0.58-0.73)
for
CA
19-9,
CA
50
and
CA
195
and
of
0.70
(95%)
CI
0.63-0.77)
for
CEA.
The
new
tumour
markers
appear
to
be
of
slightly
less
diagnostic
value
than
CEA
for
the
primary
diagnosis
of
colorectal
cancer,
although
the
discrepancy
is
not
statistically
significant.
The
low
diagnostic
power
of
CA
19-9,
CA
50
and
CA
195
may
be
due
to
a
high
proportion
of
colorectal
cancer
patients
having
the
Lewisa.lb
phenotype,
who
cannot
synthesise
these
markers.
Cancer
is
the
second
cause
of
death
in
the
USA
and
Europe,
and
colorectal
carcinoma
is
the
second
most
prevalent
malig-
nancy
in
these
continents.
The
availability
of
a
tumour
marker
detectable
in
serum
would
be-
helpful
in
confirming
the
diagnosis
of
colorectal
carcinoma.
Since
its
discovery
(Gold
&
Freedman,
1965),
the
use
of
carcinoembryonic
antigen
(CEA)
as
a
tumour
marker
has
become
widespread.
Unfortunately,
CEA
appeared
to
be
neither
organ-specific,
nor
tumour-specific
(Bates
&
Longo,
1987).
Therefore,
CEA
is
not
very
useful
in
the
primary
diagnosis
of
colorectal
carcinoma,
but
it
has
proved
to
be
an
effective
monitor
for
the
follow-up
of
these
cancers
(Fletcher,
1986).
The
search
for
new
serum
tumour
markers
has
favoured
the
development
of
monoclonal
antibodies,
which
can
be
raised
and
directed
against
circulating
tumour-associated
antigens
(TAA).
The
carbohydrate
antigen
19-9
(CA
19-9)
has
been
described
as
potentially
useful
in
the
diagnosis
of
colorectal
carcinoma
(Koprowski
et
al.,
1979;
Herlyn
et
al.,
1982).
One
year
later,
the
carbohydrate
antigen
50
(CA
50)
was
recognised
(Lindholm
et
al.,
1983)
as
a
promising
diag-
nostic
marker
for
cancers
of
colon
and
rectum.
The
monoc-
lonal
antibodies
(MAbs)
used
in
the
test
kits
of
CA19-9
have
been
shown
to
react
with
sialylated
Lacto
N-fucopentose
II
(sialyl-Lea),
a
circulating
epitope
of
the
Lewis
blood
group
antigen
(Magnani
et
al.,
1982).
The
MAbs
reactive
with
the
TAA
CA
50
react
with
two
different
carbohydrate
structures,
sialyl-Lea
and
sialosyl-lactotetraose
(Nilsson
et
al.,
1985).
More
recently,
the
TAA
195
(CA
195)
was
described
(Bray
et
al.,
1987).
The
MAbs
recognising
CA
195
have
been
shown
to
react
with
both
Lea
and
sialyl-Lea
epitopes
(Fukuta
et
al.,
1987).
In
the
case
of
CA
50
it
was
reported
that
it
might
be
tumour-specific
(Holmgren
et
al.,
1984),
whereas
CA
19-9
and
CA
195
might
be
organ-specific
(Bhargava
et
al.,
1987;
Sundaram
et
al.,
1987).
In
comparison
with
CA
19-9,
CA
195
seems
to
be
less
often
elevated
in
benign
disease,
i.e.,
it
might
be
more
specific
for
malignancies
than
CA
19-9
(Bhargava
et
al.,
1989).
It
has
been
reported
that
individuals
with
the
Lewisa-b-
phenotype
cannot
synthesise
CA
19-9,
because
they
lack the
necessary
fucosyltransferase
enzyme
(Koprowski
et
al.,
1982;
Magnani
et
al.,
1983).
In
these
individuals,
CA
19-9
cannot
be
used
for
the
detection
of
colorectal
cancers.
The
same
applies
to
CA
50
and
CA
195,
the
production
of
which
also
depends
on
the
enzyme
fucosyltransferase.
The
lack
of
fuco-
syltransferase
concerns
approximately
10%
of
the
general
population
(Watkins,
1980).
CA
50,
however,
reacts
to
an
epitope
also
containing
sialosyl-lactotetraose,
which
can
be
produced
by
all
individuals,
irrespective
of
their
Lewis
phenotype.
It
might
therefore
be
a
better
marker
for
cancers
of
colon
and
rectum
than
CA
19-9
and
CA
195.
The
aim
of
the
present
study
was
to
compare
the
value
of
CEA,
CA
19-9,
CA
50
and
CA
195
in
the
detection
of
colorectal
carcinoma.
For
this
purpose
preoperative
levels
of
the
four
serum
tumour
markers
in
colorectal
cancer
patients
were
compared
with
marker
levels
in
patients
with
benign
colorectal
disorders.
To
complete
the
overview
of
the
value
of
the
preoperative
levels
of
the
markers,
we
also
investigated
their
prognostic
significance
for
recurrence
of
disease.
Patients
and
methods
Patients
Between
January
1985
and
June
1990
preoperative
blood
samples
were
collected
from
257
patients
who
were
going
to
have
a
curative
or
palliative
operation
for
colorectal
car-
cinoma
or
an
operation
for
a
benign
colorectal
disorder.
Follow-up
information
on
recurrence
of
disease
or
death
was
available
until
November
1991.
All
diagnoses
were
histo-
logically
confirmed
after
surgery.
For
the
patients
with
colorectal
carcinoma
the
stage
of
disease,
location
and
differentiation
of
the
tumour
were
assessed.
Tumours
were
staged
according
to
Dukes'
classification
with
Astler-Coller
modification
(Astler
&
Coller,
1954).
The
type
of
disease
was
assessed
for
the
patients
with
a
benign
colorectal
disorder.
Of
the
257
patients,
198
had
a
colorectal
carcinoma.
Stage
of
disease
is
shown
in
Table
I.
Distant
metastases
are
referred
to
as
stage
'Dukes'
D'.
Two
patients
had
a
carcinoma
of
the
prostate
and
the
stomach,
respectively,
and
were
therefore
excluded
from
the
analyses.
The
57
patients
with
a
benign
colorectal
disorder
showed
various
forms
of
pathology,
which
are
summarised
in
Table
II.
Laboratory
methods
The
blood
samples
were
taken
by
venapuncture
prior
to
cytoreductive
surgery.
After
clotting,
the
sera
were
cent-
rifuged
for
10
min
at
2000
g
and
the
serum
samples
were
stored
at
-
35'C
until
analysis.
The
immunoassays
used
were
the
immunoluminometric
assay
BeriLux
CEA
(Behringwerke
Correspondence:
Y.T.
van
der
Schouw.
Received
16
December
1991;
and
in
revised
form
27
March
1992.
'."
Macmillan
Press
Ltd.,
1992
Br.
J.
Cancer
(1992),
66,
148-154

FOUR
SERUM
TUMOUR
MARKERS
FOR
DIAGNOSING
COLORECTAL
CANCER
149
Table
I
Median
and
maximum
levels
of
CEA,
CA
19-9,
CA
50
and
CA
195
for
patients
(n
=
198)1
with
colorectal
carcinomas,
for
various
stages
of
disease,
according
to
Dukes'
classification
with
Astler-Coller
modification
Median
serum
Median
serum
Median
serum
Median
serum
Dukes'
Number
of
CEA,
ngml-t
CA
19-9,
Uml-'
CA
50,
Umt'
CA
195,
Uml-'
stage
patients
(max)
(max)
(max)
(max)
A
10
2.6
18
11
10
5%
(6.9)
(45)
(30)
(10)
B1
36
2.1
24
11
10
18%
(27) (84) (75)
(32)
B2
47
4.3
35
15
10
24%
(160) (150)
(58)
(55)
C1
9
2.8
30
11
10
5%
(200)
(65)
(210) (110)
C2
47
3
24
10
10
24%
(4,100)
(1,500)
(410)
(1,100)
'D')
47
30
73
36
29
24%
(6,600)
(45,000)
(9,300)
(28,000)
Unknown
2
4.6
20
6.9
10
1%
(5.7)
(22)
(8.1)
(5)
Total
198
3.4
30
14
5
100%
(6,600)
(45,000)
(9,300)
(28,000)
'Some
of
the
individual
parameters
have
missing
data.
Table
II
Diagnoses
and
median
and
maximum
levels
of
CEA,
CA
19-9,
CA
50
and
CA
195
for
patients
with
a
benign
colorectal
disorder
(n
=
57)
Median
serum
Median
serum
Median
serum
Median
serum
Number
of
CEA,
ng
ml-'
CA
19-9,
U
ml-'
CA
50,
U
ml-'
CA
195,
U
ml-'
Diagnosis
patients
(max)
(max)
(max)
(max)
Polyps
or
polyposis
11
2.4
24
13
10
21%
(18)
(49)
(25)
(12)
Diverticular
disease
12
0.7
6.5
4.4
10
21%
(1
1)
(340)
(140)
(58)
Crohn's
disease
16
2.1
16
8.3
10
28%
(5)
(36)
(51)
(15)
Ulcerous
colitis
5
1.5
11
5.9
10
7%
(4.9)
(39)
(18)
(10)
Othera
8
1.3
18
9.6
10
14%
(5)
(92)
(46)
(20)
Unknown
5
2.3
27
14
10
9%
(4.3)
(50) (17) (10)
Total
57
1.9
19
9.5
10
100%
(18)
(340) (140)
(58)
'Other
diseases
comprise
fat
necrosis;
lipoma;
appendicular
infiltrate;
endometriotic
colon;
fibrotic
lumen
stricture;
nonspecified
inflammation;
pancreatic
pseudocyste;
perianal
fistula.
AG,
Marburg,
Germany),
the
Tandem-R
CA
195
immuno-
radiometric
assay
(Hybritech
Inc.,
Dan
Diego,
CA,
USA),
the
microparticle
enzyme
immunoassay
IMx
CA
19-9
(Abbott
Laboratories,
Abbott
Park,
IL,
USA)
and
the
Canag
Delfia
CA
50
time-resolved
fluoroimmunoassay
(Wallac
Oy,
Turku,
Finland).
The
performance
and
characteristics
of
these
methods
have
been
described
previously
(Van
der
Schouw
et
al.,
submitted;
Wobbes
et
al.,
1992).
The
precision
of
the
assays
was
calculated
for
the
means
of
duplicate
determinations
of
several
different
serum
pools
in
terms
of
within-assay
and
between-assay
coefficients
of
varia-
tion
(CVW
and
CVb,
respectively)
as
described
by
Rodbard
(1974).
The
CV,'s
ranged
from
2.0%
(CA
50)
to
5.1%
(CEA),
whereas
the
CVb's
ranged
between
6.2%
(CA
195)
and
11.7%
(CA
50).
Statistical
methods
The
usefulness
of
the
serum
markers
for
the
primary
diag-
nosis
of
colorectal
carcinoma
was
assessed
by
cumulative
frequency
distributions
and
Receiver
Operating
Characteris-
tic
(ROC)
curves.
The
cumulative
frequency
distributions
display
the
cumulative
percentage
of
colorectal
cancer
patients
as
well
as
of
patients
with
benign
colorectal
dis-
orders
against
the
serum
marker
concentration.
The
resulting
figures
allow
the
reading
of
sensitivity
and
specificity
at
any
requested
cut-off
level
for
test
positivity.
Furthermore,
it
shows
the
extent
of
overlap
of
the
marker
distribution
of
carcinoma
patients
with
that
of
the
patients
with
benign
disorders.
ROC
curves
plot
the
sensitivity
against
one
minus
specificity
at
various
cut-off
levels
of
the
diagnostic
test.
A
non-discriminating
test
will
have
an
ROC
curve
which
coin-
cides
with
the
diagonal.
A
perfect
test
will
have
an
ROC
curve
in
the
upper
left
corner
of
the
diagram
(Metz,
1978;
Swets,
1973;
Weinstein
&
Feinberg,
1980).
The
area
under
the
curve
(AUC),
ranging
from
0.5
for
a
non-discriminating
test
to
1.0
for
a
perfect
test,
is
a
measure
for
the
diagnostic
ability
of
a
test
(Hanley
&
McNeil,
1982).
The
usefulness
of
combinations
of
markers
was
assessed
by
ROC
curves
as
well.
Combinations
of
the
markers
were
made
by
adding
and
multiplying,
respectively,
the
concentrations
of
markers
for
individual
patients.
The
prognostic
value
of
the
markers
with
respect
to
first
recurrence
of
disease
was
assessed
by
fitting
a
Cox'
propor-
tional
hazards
model
for
each
marker,
with
time
from
surgery
to
first
recurrence
of
disease
in
months
as
the
depen-
dent
variable
(Cox,
1972).
Tumour-free
status
at
the
end
of
the
study
and
death
were
considered
censored.
Results
Table
I
shows
the
median and
maximum
serum
concentra-
tions
of
the
four
individual
tumour
markers
for
carcinoma
patients
for
the
different
stages
of
disease.
Means
are
not
presented,
due
to
the
skew
distributions
of
the
four
markers.

150
Y.T.
VAN
DER
SCHOUW
et
al.
To
improve
the
clarity
of
the
Tables,
minimum
levels
of
marker
concentration
are
not
displayed
either.
These
minima
approximate
the
lowest
detectable
concentration
for
all
disease
stages,
locations
and
grades
of
differentiation
and
no
increasing
trend
could
be
observed
in
the
minima.
Maximum
concentrations
of
CEA,
CA
19-9
CA
50
and
CA
195
increase
with
increasing
extent
of
Eisease.
In
the
case
of
median
levels
this
trend
cannot
be
observed;
they
are
approx-
imately
similar
in
all
stages
of
disease,
except
for
'Dukes'
D'
(Table
I).
The
various
tumour
locations
comprised
coecum,
ascending
coecum,
hepatic
flexure,
transverse
colon,
lienalic
flexure,
descending
colon,
sigmoid,
recto-sigmoid
and
rectum.
The
tumour
location
does
not
show
any
relationship
with
the
marker
concentrations.
None
of
the
markers
show
clear
rela-
tions
with
the
grade
of
differentiation
of
the
tumours.
It
is
observed
that
the
highest
level
of
the
marker
occurs
in
tumours
of
which
the
grade
of
differentiation
is
unknown,
but
this
can
probably
be
explained
by
the
stage
of
disease
of
100
GL)
0)
these
tumours,
which
were
all
'Dukes'
D'.
Apparently,
in
clinical
practice
the
grade
of
differentiation
is
frequently
not
established
in
patients
with
distant
metastases.
Table
II
pre-
sents
median
and
maximum
observed
concentrations
of
the
markers
in
patients
with
benign
colorectal
disorders.
It
is
noted
that
the
maximum
concentration
for
all
markers
is
found
in
patients
with
diverticular
disease.
Figures
1
through
4
display
the
cumulative
frequency
distributions
for
the
markers.
They
present
a
rather
similar
picture;
the
distribution
of
the
carcinoma
patients
shows
an
80-90%
overlap
with
that
of
the
benign
colorectal
disorder
patients,
but
for
all
markers
a
cut-off
point
can
be
deter-
mined
above
which
patients
almost
certainly
have
carcin-
omas.
This
point
is
indicated
in
each
figure
and
varies
from
18
ng
ml-'
for
CEA
to
340
arbitrary
U
ml-',
140
arbitrary
U
ml-'
and
58arbitraryUml'
for
CA
19-9,
CA
50,
and
CA
195,
respectively.
Figure
5
presents the
ROC
curves
and
the
corresponding
Benign
colorectal
90-
disorder
patients
n
=
57
80
-
70
-
60
-
50
-
40
-
30
20
10
0
0.1
1
10
Colorectal
cancer
patients
n
=
198
Max.
of
patients
with
benign
disorders
I
I
I
I
l
l
I
I
Il
lI
I1
II1
I
100
1000
10
000
SerumCEA(ngml-
1)
Figure
1
Cumulative
frequency
distribution
of
colorectal
cancer
patients
(198)
and
patients
with
a
benign
colorectal
disorder
(57)
for
CEA.
Benign
colorectal
disorder
patients
n
=
57
Colorectal
cancer
patients
n
=
198
Max.
of
patients
with
I
benign
disorders
100
Serum
CA
19-9
(U
ml-')
Figure
2
Cumulative
frequency
distribution
of
colorectal
cancer
patients
(198)
and
patients
with
a
benign
colorectal
disorder
(57)
for
CA
19-9.
100
-
90
-
80
-
70
-
60
-
50
-
40
-
30
-
-'g
C.)
c
a)
03
cr
160)
01)
'._
CD
E
0
20
-
10
-

FOUR
SERUM
TUMOUR
MARKERS
FOR
DIAGNOSING
COLORECTAL
CANCER
151
Benign
colorectal
disorder
patients
n
=
57
Colorectal
cancer
patients
n
=
198
Max.
of
patients
with
benign
disorders
10
100
Serum
CA-50
(U
ml-
1)
Figure
3
Cumulative
frequency
distribution
of
colorectal
cancer
patients
(198)
and
patients
with
a
benign
colorectal
disorder
(57)
for
CA
50.
100-
90
-
Benign
colorectal
disorder
patients
80
-
=
Colorectal
cancer
patients
>
770|
n
=198
70
-
C._
60
50-
40-
E
30
Max.
of
patients
with
benign
disorders
0.1
1
10
100
1000
10
000
100
000
Serum
CA
195
(U
mlV-1)
Figure
4
Cumulative
frequency
distribution
of
colorectal
cancer
patients
(198)
and
patients
with
a
benign
colorectal
disorder
(57)
for
CA
195.
AUC's
for
the
four
tumour
markers.
The
ROC
curves
of
the
newer
markers
all
have
an
AUC
of
0.65,
with
a
95%
confi-
dence
interval
(95%
Cl)
of
0.58-0.73,
which
is
rather
low
and,
moreover,
even
lower
than
that
of
CEA
(AUC
0.70,
95%
Cl
0.63-0.77),
although
the
difference
is
very
small
and
not
statistically
significant.
Various
combinations
of
the
serum
tumour
markers
did
not
result
in
a
better
discri-
minative
ability
(Figure
6).
Figure
7
shows
tumour-free
survival
functions
for
two
categories
of
CA
50
(CA
50
<
13
Uml-I/CA
50
>
13
U
ml-'),
adjusted
for
stage
of
disease
(two
categories;
Dukes'
A,
B1,
B2/Dukes'
C1,
C2,
'D').
Due
to
the
low
number
of
recurrences
(16),
division
into
more
categories
led
to
empty
cells.
The
other
markers
showed
very
similar
pictures
and
are
therefore
not
shown.
In
a
Cox'
proportional
hazards
model
marker
concentration
was
held
continuous
to
investigate
whether
a
monotonous
relationship
with
the
risk
of
recur-
rence
exists,
but
this
could
not
be
found
at
all,
adjustment
for
age
(continuous)
and
stage
of
disease
(two
categories;
Dukes'
A,
Bi,
B2/Dukes'
Cl,
C2,
'D')
did
not
reveal
any
association
either
(P
=
0.5-0.9).
Discussion
Although
earlier
investigations
indicated
very
promising
results
for
the
serum
tumour
markers
CA
19-9,
CA
50
and
CA
195,
these
markers
showed
disappointingly
low
diagnos-
tic
power
in
the
present
study.
The
very
low
median
concen-
trations
alone,
presented
in
Tables
I
and
II,
point
to
the
poor
discriminative
ability
of
all
markers
tested.
The
ROC
curves
are
in
accordance
with
this
finding.
The
three
newer
markers
all
have
an
almost
identical
ROC
curve
with
an
AUC
of
0.65
(95%
Cl
0.58-0.73).
The
ROC
curve
of
CEA
was
even
100
90
80
-
70
-
C.)
C7
a)
C.)
60
-
50
-
40-
30-
20
10
Il
I
I
Ili
ITll
IX
t
II
I1111
li

152
Y.T.
VAN
DER
SCHOUW
et
al.
slightly
better,
having
an
AUC
of
0.70
(95%
Cl
0.63-0.77),
which,
however,
is
not
statistically
significant.
Organ-
specificity
was
not
investigated
in
the
present
study,
but
the
tumour-specificity
is
disappointing.
Even
CA
50,
which
has
been
reported
to
be
tumour-specific
(Holmgren
et
al.,
1984)
does
not
show
a
better
discriminative
ability
than
CEA,
which
is
known
to
be
increased
in
nonmalignant
disorders
and
healthy
smokers
(Moore
et
al.,
1989).
However,
in
the
U
_91
a)
o-
co
en1
0(0
o_
._
0.
0
X
0
90
80
70
60
50
40
30
20
10
0
10
20 30
40
50
60
70
100
90
80
70
C')
.)
(I)
40
30
20
0
10
20
30
40
50
60
70
80
90
100
100
minus
Specificity
(%)
--CEA
CA
19-9
-*-CA
50
9
CA
195
Figure
5
Receiver
Operating
Characteristic
curves
of
CEA,
CA
19-9,
CA
50
and
CA
195
for
colorectal
cancer
patients
(198)
and
patients
a
with
benign
colorectal
disorder
(57).
AUC
=
Area
under
the
curve;
CEA:
AUC
=
0.70,
95%
Cl
0.63-0.77;
CA
19-9:
AUC
=
0.65,
95%
Cl
0.58-0.73;
CA
50:
AUC
=
0.65,
95%
Cl
0.58-0.73;
CA
195:
AUC
=
0.65,
95%
Cl
0.58-0.73.
At
an
arbitrarily
selected
high
specificity
rate
of
95%
all
markers
had
low
sensitivity
rates,
varying
from
27%
(CA
50),
28%
(CA
19-9)
and
34%
(CA
195)
to
39%
CEA),
indicating
high
numbers
of
false
negative
test
results
at
high
levels
of
specificity.
100
90
80
/
70-
60
n
4
50
o40
C,)
0
10
20 30
40
50
60
70
80
90
100
100
minus
Specificity
(%)
Sum
of
markers
Product
of
markers
Figure
6
Receiver
Operating
Characteristic
curves
for
sum
and
product
of
CEA,
CA
19-9,
CA
50
and
CA
195
for
colorectal
cancer
patients
(198)
and
patients
with
a
benign
colorectal
disorder
(57).
AUC
=
area
under
the
curve;
Sum:
AUC
=
0.71,
95%
Cl
0.64-0.79;
Product:
AUC
=
0.73,
95%
Cl
0.66-0.80.
Survival,
recurrence-free
(months)
Low
CA
50/low
stage
Low
CA
50/high
stage
High
CA
50/low
stage
-
High
CA
50/high
stage
Figure
7
Tumour-free
survival
of
colorectal
cancer
patients
(198)
in
months
according
to
CA
50-concentration
and
stage
of
disease.
Low
CA
50:
CA
50
<
13Uml-';
High
CA
50:
CA
50>
13
U
ml-;
Low
stage:
Dukes'
A,
BI
or
B2;
High
stage:
Dukes'
C1,
C2
or
'D'.
study
of
Holmgren
et
al.
(1984),
58%
of
the
carcinoma
patient
group
had
disseminated
metastases.
Furthermore,
19%
of
the
control
group
were
patients
with
pneumonia
and
68%
were
even
healthy
blood
donors.
The
use
of
these
groups,
with
serum
marker
concentrations
on
both
extreme
ends
of
the
marker
distribution,
probably
masked
the
fact
that
patients
with
early
stages
of
malignant
disease
have
CA
50
concentrations
comparable
with
those
of
patients
with
benign
colorectal
disorders.
To
investigate
the
similarity
in
the
performance
of
the
markers
further,
Pearson
correlation
coefficients
were
cal-
culated
for
all
markers
as
presented
in
Table
III.
CA
19-9,
CA
50
and
CA
195
appeared
to
have
a
correlation
of
about
0.55-0.60
with
CEA
but,
more
interestingly,
the
new
mark-
ers
showed
a
very
high
mutual
correlation
with
correlation
coefficients
ranging
from
0.91
to
0.99.
Accordingly,
it
is
not
surprising
that
they
showed
comparable
diagnostic
power
for
colorectal
carcinoma.
Probably,
these
high
correlations
can
in
part
be
explained
by
the
reactivity
of
all
three
markers
with
sialyl-Lea.
However,
some
reports
indicate
enhanced,
i.e.
more
specific
assay
performance
using
MAbs
that
react
with
both
the
Lea
and
the
sialyl-Lea
epitopes
as
is
the
case
with
CA
195
(Fukuta
et
al.,
1987).
This
is
not
confirmed
by
the
present
study.
CA
19-9
and
CA
50
have
been
reported
to
show
identical
diagnostic
results
(Roberts,
1988),
although
CA
50
reacts
with
an
epitope
also
containing
sialosyl-lacto-
tetraose
(Nilsson
et
al.,
1985).
These
findings
are
in
accor-
dance
with
our
data.
Combinations
of
the
four
markers
did
not
improve
diag-
nostic
performance
significantly,
as
is
clear
from
Figure
6.
This
was
to
be
expected
from
the
high
correlation
between
the
markers.
Apparently,
there
is
still
discussion
on
the
diag-
nostic
value
of
combinations
of
markers.
Some
authors
report
improved
diagnostic
power
for
a
combination
markers,
others
report
approximately
equal
diagnostic
power
(Kuusela
et
al.,
1984;
Bray
&
Gaur,
1988;
Bhargava
et
al.,
1989).
Probably,
the
poor
diagnostic
power
of
all
three
new
markers
can
at
least
partly
be
explained
by
the
lack
of
the
enzyme
fucosyltransferase
in
approximately
10%
of
the
population
(Watkins,
1980),
who
have
a
Lea4b
phenotype
and
Table
III
Pearson's
correlation
coefficients
(P-value)
for
CEA,
CA
19-9,
CA
50
and
CA
195
CEA
CA
19-9
CA
50
CA
195
CEA
1
0.57
0.59
0.59
(0.0001)
(0.0001)
(0.0001)
CA
19-9
1
0.93
0.91
(0.0001)
(0.0001)
CA
50
1
0.99
(0.0001)
CA
195
1
80
U
4
n.

FOUR
SERUM
TUMOUR
MARKERS
FOR
DIAGNOSING
COLORECTAL
CANCER
153
hence
cannot
synthesise
CA
19-9,
CA
50
and
CA
195
(Koprowski
et
al.,
1982;
Magnani
et
al.,
1983).
Recently
it
was
suggested
that
the
Leab
phenotype
is
more
frequent
in
patients
with
urinary
bladder
(12.2%)
and
colorectal
(23.8%)
carcinoma
(Langkilde
et
al.,
1991).
However,
in
this
study
the
Lewis
phenotypes
were
determined
on
the
erythrocytes,
and
it
has
been
demonstrated
that
their
phenotype
can
con-
vert
from
Lewis-positive
to
Lewis-negative
(Hirano
et
al.,
1987).
Therefore,
as
far
as
studies
based
on
Lewis
phenotype
determination
are
concerned,
the
hypothesis
is
yet
to
be
investigated
in
serum,
which
does
not
allow
conversion
(Hirano
et
al.,
1987)
and
on
tumour
tissue.
Only
CEA
has
some
relationship
with
the
extent
of
the
disease,
as
can
be
concluded
from
the
higher
CEA
concentra-
tions
in
patients
with
more
extensive
disease.
Therefore,
unlike
CEA,
the
markers
CA
19-9,
CA
50
and
CA
195
are
probably
neither
useful
for
the
primary
diagnosis
nor
for
the
staging
of
colorectal
carcinoma.
Although
they
were
develop-
ed
from
colorectal
cell
lines
(Koprowski
et
al.,
1979;
Schwartz,
1990),
there
are
indications
that
some
of
these
markers
could
play
a
role
in
the
diagnosis
of
pancreatic
cancer
(Staab
et
al.,
1985;
Paganuzzi
et
al.,
1988;
Bhargava
et
al.,
1988).
In
the
case
of
CA
19-9
a
sensitivity
of
89%
is
reported
at
a
specificity
level
of
95%
(Staab
et
al.,
1985),
whereas
a
sensitivity
of
81%
at
a
specificity
of
89%
is
described
for
CA
50 (Paganuzzi
et
al.,
1988)
and
a
sensitivity
of
64%
at
a
specificity
of
94%
for
CA
195
(Bhargava
et
al.,
1988).
Our
data
show
that
none
of
the
tumour
markers
had
prognostic
value,
that
is,
none
of
the
markers
could
predict
recurrence
of
disease
within
34
months
after
diagnosis
(median
follow-up,
maximum
follow-up
is
81
months).
The
three
new
markers
were
evaluated
in
accordance
with
a
so-called
first
phase
of
diagnostic
marker
assessment
as
was
described
recently
(Van
der
Schouw
et
al.,
submitted).
In
that
paper
it
was
indicated
that
the
spectrum
of
participating
patients
must
represent
the
spectrum
of
patients
that
is
seen
in
clinical
practice.
The
colorectal
cancer
patients
as
well
as
the
benign
colorectal
disease
patients
in
the
present
paper
are
a
representation
of
the
patients
presenting
to
the
out-patient
Department
of
General
Surgery
of
a
university
hospital.
The
promising
diagnostic
power
of
the
serum
tumour
markers
as
described
in
literature
probably
results
from
comparisons
of
serum
marker
concentrations
of
colorectal
cancer
patients
with
those
of
healthy
individuals
and
patients
with
non-
colorectal
benign
disorders
(Holmgren
et
al.,
1984;
Bhargava
et
al.,
1987).
The
methods
of
statistical
analysis
used
in
this
paper
are
also
put
forward
in
those
recently
proposed
guidelines
(Van
der
Schouw
et
al.,
submitted).
They
form
a
convenient
way
of
expressing
the
diagnostic
power
of
a
test,
mainly
because
they
are
independent
of
cut-off
levels
for
test
positivity.
Such
an
analysis
shows
sensitivities
and
specificities
at
all
possible
cut-off
points
simultaneously.
Cumulative
frequency
distribu-
tions
show
the
relationship
of
these
test
characteristics
for
the
particular
serum marker
concentrations.
Finally,
ROC
curves
provide
one
summary
measure
of
performance,
i.e.
the
AUC,
rather
than
two
separate
measures,
i.e.
sensitivity
and
specificity,
which
have
to
be
considered
simultaneously.
Fur-
thermore,
ROC
curves
provide
the
possibility
of
comparing
multiple
tests
of
which
the
results
are
expressed
on
different
scales,
such
as
ng
ml-'
or
arbitrary
U
ml-'
as
is
the
case
in
the
present
paper.
It
can
be
concluded
that
CA
19-9,
CA
50
and
CA
195
do
not
appear
to
be
very
useful
in
the
primary
diagnosis
of
colorectal
carcinoma.
Probably,
they
are
not
of
value
in
staging
the
disease
and
in
prognosis
either.
Investigation
into
the
value
of
the
markers
in
the
monitoring of
colorectal
carcinoma
and
the
diagnosis,
staging
and
monitoring
of
pan-
creatic
carcinoma
is
necessary
and,
indeed,
in
progress.
References
ASTLER,
V.B.
&
COLLER,
F.A.
(1954).
The
prognostic
significance
of
direct
extension
of
carcinoma
of
the
colon
and
rectum.
Ann.
Surg.,
139,
846-851.
BATES,
S.E.
&
LONGO,
D.L.
(1987).
Use
of
serum
tumor
markers
in
cancer
diagnosis
and
management.
Semin.
Oncol.,
14,
102-138.
BHARGAVA,
A.K.,
PETRELLI,
N.J.,
KARNA,
A.
&
5
others
(1987).
Circulating
CA-195
in
colorectal
cancer.
J.
Tumor
Marker
Oncol.,
2,
319-327.
BHARGAVA,
A.K.,
PETRELLI,
N.J.,
KARNA,
A.
&
4
others
(1988).
Use
of
CA
195
and
CA
19-9
in
gastrointestinal
cancers.
Clin.
Chem.,
34,
1297.
BHARGAVA,
A.K.,
PETRELLI,
N.J.,
KARNA,
A.
&
6
others
(1989).
Serum
levels
of
cancer-associated
antigen
CA-195
in
gastrointes-
tinal
cancers
and
its
comparison
with
CA
19-9.
J.
Clin.
Lab.
Anal.,
3,
370-377.
BRAY,
K.R.,
GAUR,
P.K.,
STONE,
M.R.,
LEUNG,
J.P.
(1987).
A
monoclonal
antibody
that
detects
a
tumor
associated
antigen
in
the
sera
of
patients
with
colon
cancer.
Fed.
Proc.,
46,
1059.
BRAY,
K.R.
&
GAUR,
P.K.
(1988).
Serum
levels
of
cancer-associated
antigen
195,
a
circulating
marker
for
colon
cancer,
and
its
rela-
tionship
to
carcinoembryonic
antigen.
J.
Clin.
Lab.
Anal.,
2,
187-93.
COX,
D.R.
(1972).
Regression
models
and
life-tables.
J.
R.
Stat.
Soc.,
34,
187-220.
FLETCHER,
R.H.
(1986).
Carcinoembryonic
antigen.
Ann.
Intern.
Med.,
104,
66-73.
FUKUTA,
S.,
MAGNANI,
J.L.,
GAUR,
P.K.
&
GINSBURG,
V.
(1987).
Monoclonal
antibody
CC3C
195,
which
detects
cancer-associated
antigens
in
serum,
binds
to
human
Lea
blood
group
antigen
and
to
its
sialyated
derivative.
Arch.
Biochem.
Biophys.,
255,
214-216.
GOLD,
P.
&
FREEDMAN,
S.O.
(1965).
Demonstration
of
tumor-
specific
antigens
in
human
colonic
carcinoma
by
immunological
tolerance
and
absorption
techniques.
J.
Exp.
Med.,
121,
439-462.
HANLEY,
J.A.
&
MCNEIL,
B.J.
(1982).
The
meaning
and
use
of
the
area
under
a
Receiver
Operating
Characteristic
(ROC)
curve.
Radiology,
143,
29-36.
HERLYN,
M.,
SEARS,
H.F.,
STEPLEWSKI,
Z.
&
KOPROWSKI,
H.
(1982).
Monoclonal
antibody
detection
of
a
circulating
tumor-
associated
antigen.
I.
Presence
of
antigen
in
sera
of
patients
with
colorectal,
gastric,
and
pancreatic
carcinoma.
J.
Clin.
Immunol.,
2,
135-40.
HIRANO,
K.,
KAWA,
S.,
OGUCHI,
H.
&
4
others
(1987).
Loss
of
Lewis
antigen
expression
on
erythrocytes
in
some
cancer
patients
with
high
serum
CA
19-9
levels.
J.
Natl
Cancer
Inst.,
79,
1261-1267.
HOLMGREN,
J.,
LINDHOLM,
L.,
PERSSON,
B.
&
8
others
(1984).
Detection
by
monoclonal
antibody
of
carbohydrate
antigen
CA
50
in
serum
of
patients
with
carcinoma.
Br.
Med.
J.,
288,
1479-1482.
KOPROWSKI,
H.,
STEPLEWSKI,
Z.,
MITCHELL,
K.,
HERLYN,
M.
&
FUHRER,
P.
(1979).
Colorectal
carcinoma
antigens
detected
by
hybridoma
antibodies.
Somatic
Cell.
Genet.,
5,
957-972.
KOPROWSKI,
H.,
BLASZCZYK,
M.,
STELEWSKI,
Z.,
BROCKHAUS,
M.,
MAGNANI,
J.L.
&
GINSBURG,
V.
(1982).
Lewis blood
type
may
affect
the
incidence
of
gastrointestinal
cancer.
Lancet,
1,
1332-1333.
KUUSELA,
P.,
JALANKO,
H.,
ROBERTS,
P.
&
4
others
(1984).
Com-
parison
of
CA
19-9
and
carcinoembryonic
antigen
(CEA)
levels
in
the
serum
of
patients
with
colorectal
diseases.
Br.
J.
Cancer.,
49,
135-139.
LANGKILDE,
N.C.,
WOLF,
H.,
MELDGARD,
P.
&
0RNTOFT,
T.F.
(1991).
Frequency
and
mechanism
of
Lewis
antigen
expression
in
human
urinary
bladder
and
colon
carcinoma
patients.
Br.
J.
Cancer,
63,
583-586.
LINDHOLM,
L.,
HOLMGREN,
J.,
SVENNERHOLM,
L.
&
5
others
(1983).
Monoclonal
antibodies
against
gastrointestinal
tumour-
associated
antigens
isolated
as
monosialogangliosides.
Int.
Arch.
Allergy
Appl.
Immunol.,
71,
178-181.
MAGNANI,
J.L.,
NILSSON,
B.,
BROCKHAUS,
M.
&
6
others
(1982).
A
monoclonal
antibody
defined
antigen
associated
with
gastrointes-
tinal
cancer
is
a
ganglioside
containing
sialylated
lacto-N-
fucopentaose
II.
J.
Biol.
Chem.,
257,
14365-14369.

154
Y.T.
VAN
DER
SCHOUW
et
al.
MAGNANI,
J.L.,
NILSSON,
B.,
BROCKHAUS,
M.
&
6
others
(1982).
A
monoclonal
antibody
defined
antigen
associated
with
gastrointes-
tinal
cancer
is
a
ganglioside
containing
sialylated
lacto-N-
fucopentaose
II.
J.
Biol.
Chem.,
257,
14365-14369.
MAGNANI,
J.L.,
STEPLEWSKI,
Z.,
KOPROWSKI,
H.
&
GINSBURG,
V.
(1983).
Identification
of
the
gastrointestinal
and
pancreatic
cancer-associated
antigen
detected
by
monoclonal
antibody
19-9
in
the
sera
of
patients
as
a
mucin.
Cancer
Res.,
43,
5489-5492.
METZ,
C.E.
(1978).
Basic
principles
of
ROC
analysis.
Semin.
Nucl.
Med.,
8,
283-98.
MOORE,
M.,
JONES,
D.J.,
SCHOFIELD,
P.F.
&
HARNDEN,
D.G.
(1989).
Current
status
of
tumor
markers
in
large
bowel
cancer.
World
J.
Surg.,
13,
52-59.
NILSSON,
O.,
MANSSON,
J.E.,
LINDHOLM,
L.,
HOLMGREN,
J.
&
SVENNERHOLM,
L.
(1985).
Sialosyllactotetraosylceramide,
a
novel
ganglioside
antigen
detected
in
human
carcinomas
by
a
monoclonal
antibody.
FEBS
Lett.,
182,
398-402.
PAGANUZZI,
M.,
ONETTO,
M.,
MARRONI,
P.
&
4
others
(1988).
CA
19-9
and
CA
50
in
benign
and
malignant
pancreatic
and
biliary
disease.
Cancer,
61,
2100-2108.
ROBERTS,
P.J.
(1988).
Tumour
markers
in
colerectal
cancer.
Scand.
J.
Gastroenterol.,
149,
suppl
50-58.
RODBARD,
D.
(1974).
Statistical
quality
control
and
routine
data
processing
for
radioimmunoassay
and
immunoradiometric
assays.
Clin.
Chem.,
20,
1255-1270.
SCHOUW,
Y.T.,
VAN
DER,
RUIJS,
J.H.J.,
SEGERS,
M.F.G.
&
4
others
(1991).
Initial
assessment
of
diagnostic
markers,
illustrated
by
CEA
for
colorectal
carcinoma.
Submitted.
SCHWARTZ,
M.K.
(1990).
Advances
in
the
use
of
tumor
markers.
Compr.
Ther.,
16,
51-7.
STAAB,
H.J.,
BROMMENDORF,
T.,
HORNUNG,
A.,
ANDERER,
F.A.
&
KIENIGER,
G.
(1985).
The
clinical
validity
of
circulating
tumor-
associated
antigens
CEA
and
CA
19-9
in
primary
diagnosis
and
follow-up
of
patients
with
gastrointestinal
malignancies.
Klin.
Wochenschr.,
63,
106-115.
SWETS,
J.A.
(1973).
The
relative
operating
characteristic
in
psychology.
Science,
182,
990-1000.
SUNDARAM,
S.G.,
MANIMEKALAI,
S.,
UNNI,
S.
&
4
others
(1987).
Evaluation
of
antigen
CA-195
for
the
detection
of
colon
cancer.
J.
Tumor
Marker
Oncol.,
2,
315-318.
WATKINS,
W.M.
(1980).
Biochemistry
and
genetics
of
the
ABO,
Lewis,
and
P
blood
group
systems.
In
Harris,
H.
&
Hirschhorn,
K.
Advances
in
Human
Genetics.
Plenum
Press:
New
York.
WEINSTEIN,
M.C.
&
FEINBERG,
H.V.
(1980).
Clinical
Decision
Analysis.
W.B.
Saunders
Company:
Philadelphia.
WOBBES,
TH.,
THOMAS,
C.M.G.,
SEGERS,
M.F.G.
&
NAGENGAST,
F.M.
(1992).
Evaluation
of
seven
tumour
markers
(CA-50,
CA-
19-9,
CA-19-9
TruQuant,
CA-72.4,
CA-195,
CEA
and
TPA)
in
pretreatment
sera
of
patients
with
gastric
ulcer.
Cancer,
69,
2036-2041.