R E V I E W A R T I C L E – G A S T R O I N T E S T I N A L O N C O L O G Y

Use of Tumor Markers in Gastrointestinal Cancers: Surgeon
Perceptions and Cost-Benefit Trade-Off Analysis

Amish Acharya, MRCS, Sheraz R. Markar, MRCS, Michael Matar, MBBS, Melody Ni, PhD,

and George B. Hanna, PhD

Division of Surgery, Department of Surgery and Cancer, St Mary’s Hospital, Imperial College London, London, UK

ABSTRACT

Background. Gastrointestinal cancers constitute the third

most common cancers worldwide. Tumor markers have

long since been used in the postoperative surveillance of

these malignancies; however, the true value in clinical

practice remains undetermined.

Objective. This study aimed to evaluate the clinical utility

of three tumor markers in colorectal and esophagogastric

cancer.

Methods. A systematic review of the literature was

undertaken to elicit the sensitivity, specificity, statistical

heterogeneity and ability to predict recurrence and metas-

tases for carcinoembryonic antigen (CEA), cancer antigen

(CA) 19-9 and CA125. European surgeons were surveyed

to assess their current practice and the characteristics of

tumor markers they most valued. Data from the included

studies and survey were combined in a cost-benefit trade-

off analysis to assess which tumor markers are of most use

in clinical practice.

Results. Diagnostic sensitivity and specificity were ranked

the most desirable characteristics of a tumor marker by

those surveyed. Overall, 156 studies were included to

inform the cost-benefit trade-off. The cost-benefit trade-off

showed that CEA outperformed both CA19-9 and CA125,

with lower financial cost and a higher sensitivity, and

diagnostic accuracy for metastases at presentation (area

under the curve [AUC] 0.70 vs. 0.61 vs. 0.46), as well as

similar diagnostic accuracy for recurrence (AUC 0.46 vs.

0.48).

Conclusions. Cost-benefit trade-off analysis identified

CEA to be the best performing tumor marker. Further

studies should seek to evaluate new tumor markers, with

investigation tailored to factors that meet the requirements

of practicing clinicians.

Gastrointestinal cancers are the third most common

cancers worldwide, with a prevalence of 1,281,539 in the

US in 2013.
1

Survival for gastrointestinal malignancies

have been improving worldwide due to advances in mul-

timodality treatments, diagnostic strategies and expanding

the criteria for treatable disease. While these strategies

commonly involve a combination of radiological and

endoscopic techniques, some studies have shown serum

tumor markers may have a diagnostic, as well as thera-

peutic, monitoring role.
2,3

A tumor marker is defined as a compound produced by

the tumor or the host, in response to a malignancy. Tra-

ditionally, markers such as carcinoembryonic antigen

(CEA) and, to a lesser extent, cancer antigen (CA) 19-9,

have been used clinically to monitor disease response,

whereby the efficacy of treatments can be assessed by

noting a reduction in the level of a marker, which was

previously high.
4

In colorectal cancer, the use of tumor

makers in postoperative surveillance has been recom-

mended by the American Society of Clinical Oncology;

however, their use in the identification of metastasis and

diagnostic accuracy has not been established
5

despite some

clinicians incorporating them into regular practice. In

contrast, the use of tumor markers in esophagogastric

malignancies in any capacity remains controversial. A

number of studies have assessed the use of several tumor

markers in prognosis and diagnosis, however they are of

Electronic supplementary material The online version of this
article (doi:10.1245/s10434-016-5717-y) contains supplementary
material, which is available to authorized users.

� The Author(s) 2016. This article is published with open access
at Springerlink.com

First Received: 27 April 2016;

Published Online: 22 December 2016

G. B. Hanna, PhD

e-mail: g.hanna@imperial.ac.uk

Ann Surg Oncol (2017) 24:1165–1173

DOI 10.1245/s10434-016-5717-y

http://dx.doi.org/10.1245/s10434-016-5717-y
http://crossmark.crossref.org/dialog/?doi=10.1245/s10434-016-5717-y&domain=pdf
http://crossmark.crossref.org/dialog/?doi=10.1245/s10434-016-5717-y&domain=pdf


limited quality.
6,7

As such, there is no clear consensus on

the use of tumor markers, and current practice is dependent

on the individual clinician’s choice.

In both cancer types, there is hence a clear need for an

objective evaluation of common tumor markers in several

clinical scenarios. This appraisal requires a comparison of

clinical utility and costs or negatives with the utilization of

the marker; however the relative importance of these

benefits and costs, or performance characteristics, as dri-

vers to uptake has not yet been quantified by the literature

and would be dependent on how clinicians perceive the use

of tumor markers. This study aimed to critically assess the

cost-benefit trade-off of three common tumor markers in

gastrointestinal cancers by means of evaluating the per-

ceptions of surgeons to common markers.

METHODS

Literature Search Strategy

A literature search of the PubMed, Ovid MEDLINE,

EMBASE and Google Scholar electronic databases was

conducted from January 1990 up to and including

December 2015 for studies regarding the use of tumor

markers in the diagnosis, postoperative surveillance, or

prediction of metastasis in colorectal and esophagogastric

cancer (Online Appendix 1). Search terms used included

‘colorectal neoplasms’, ‘esophageal cancer’, ‘gastric can-

cer’, ‘tumour markers’, ‘neoplasm antigens’, ‘tumour-

associated antigens’, ‘prognosis’, ‘recurrence’, ‘metastasis’

and ‘staging’ in various combinations, as well as the name

of the specific markers, relevant surgical procedures, and

alternative spellings, e.g. tumor.

Research titles were then screened for suitability, with

full-text copies retrieved. All studies that investigated the

diagnostic, prognostic, or predictive ability of a single or

multiple tumor marker in colorectal or esophagogastric

cancers that could be tested in patients were included.

Exclusion criteria involved studies with no available Eng-

lish translation, published abstracts only, and those

assessing the predictive ability for metastases in which no

diagnostic accuracy or recurrence data could be calculated.

Of those studies meeting the inclusion criteria, the stated

specificity and sensitivity were extracted. Studies that did not

explicitly state the sensitivity and specificity of the marker were

independently calculated and verified by two authors (AA and

SRM), provided sufficient data were available.

Literature Standard

The QUADAS-2
8

tool, which involves four domains,

i.e. patient selection, index test, reference standard, and

flow of subjects through the study, was used to appraise the

standard of the literature, and was implemented to assess

the quality and risk of bias of the included studies. The

reference standard was histological confirmation of

malignancy or recurrence.

Tumor Marker Survey

Surgeons affiliated with the European Association of

Endoscopic Surgery were invited to complete an anony-

mous survey regarding tumor markers. These surgeons

were asked to rank attributes of the ‘ideal marker’ in order

of their perceived importance to routine clinical practice

(Online Appendix 2). No duplication of rank was permit-

ted. These characteristics included diagnostic sensitivity,

specificity, consistency across demographics, patient

acceptability, cost, time for result, and predictive of

recurrence and metastases (defined by the AUC). A sum-

mative rank was then calculated and informed the

weighting for the cost-benefit trade-off analysis.

Statistical Methodology

For each of the assessments of cancer diagnosis, recur-

rence, and metastasis, paired sensitivity and specificity

were calculated from each eligible study, as appropriate. A

bivariate model for meta-analysis of statistical accuracy

provides more accurate results than fixed-effects modeling.

Following the validated methodology of Harbord et al.
9
,

bivariate meta-analyses were performed to generate pooled

point estimates and 95% confidence intervals (CIs) for the

sensitivity and specificity of the tumor marker under

investigation, with histopathological confirmation of

malignancy, together with hierarchical summary receiver

operating characteristic (ROC) curves. The software used

for this analysis was the custom-designed statistical pack-

age Michigan Interactive Data Analysis System

(MIDAS).
10

Areas under the hierarchical summary ROC

curves, as well as I
2

statistics, were obtained directly from

the MIDAS output (see Zhou and Tu for an in-depth

description of the statistical methods used.
11

)

Performance Characteristics

The performance of the three tumor markers, with

respect to the eight characteristics surgeons were asked to

rank in the survey, was calculated (Tables 1, 2). Cost and

speed of result were taken as the stated laboratory process

costs from St Mary’s Hospital, Paddington, UK. Sensitiv-

ity, specificity, prediction of metastases at primary

diagnosis, and recurrence following resection were calcu-

lated from the aforementioned pooled analyses, with the

1166 A. Acharya et al.



T
A
B
L
E

1
P

e
rf

o
rm

a
n

c
e

o
f

C
E

A
,

C
A

1
9

-9
a
n

d
C

A
1

2
5

w
it

h
re

sp
e
c
t

to
th

e
e
ig

h
t

p
e
rf

o
rm

a
n

c
e

c
h

a
ra

c
te

ri
st

ic
s

fo
r

c
o

lo
re

c
ta

l
c
a
n

c
e
r

C
h

a
ra

c
te

ri
st

ic
d

e
si

ra
b

il
it

y
fr

o
m

th
e

su
rv

e
y

C
h

a
ra

c
te

ri
st

ic
C

E
A

C
A

1
9

-9
C

A
1

2
5

1
D

ia
g

n
o

st
ic

se
n

si
ti

v
it

y
0

.5
3

0
.4

7
0

.2
0

2
D

ia
g

n
o

st
ic

sp
e
c
ifi

c
it

y
0

.8
6

0
.9

2
0

.9
9

3
P

re
d

ic
ti

v
e

o
f

re
c
u

rr
e
n

c
e

0
.7

8
0

.5
5

0
.5

5

4
P

re
d

ic
ti

v
e

o
f

m
e
ta

st
a
si

s
0

.7
7

0
.0

0
0

.0
0

5
L

o
w

c
o

st
(£

)
8

.0
0

1
6

.0
0

1
7

.0
0

6
S

p
e
e
d

o
f

re
su

lt
(h

o
u

rs
)

2
4

.0
0

2
4

.0
0

2
4

.0
0

7
P

a
ti

e
n

t
a
c
c
e
p

ta
b

il
it

y
G

o
o

d
G

o
o

d
G

o
o

d

8
C

o
n

si
st

e
n

c
y

0
.8

6
0

.9
6

0
.9

6

C
E
A

c
a
rc

in
o

e
m

b
ry

o
n

ic
a
n

ti
g

e
n

,
C
A

c
a
n

c
e
r

a
n

ti
g

e
n

T
A
B
L
E

2
P

e
rf

o
rm

a
n

c
e

o
f

C
E

A
,

C
A

1
9

-9
a
n

d
C

A
1

2
5

w
it

h
re

sp
e
c
t

to
th

e
e
ig

h
t

p
e
rf

o
rm

a
n

c
e

c
h

a
ra

c
te

ri
st

ic
s

fo
r

e
so

p
h

a
g

o
g

a
st

ri
c

c
a
n

c
e
r

C
h

a
ra

c
te

ri
st

ic
d

e
si

ra
b

il
it

y
fr

o
m

th
e

su
rv

e
y

C
h

a
ra

c
te

ri
st

ic
C

E
A

C
A

1
9

-9
C

A
1

2
5

1
D

ia
g

n
o

st
ic

se
n

si
ti

v
it

y
0

.3
4

0
.3

0
0

.3
1

2
D

ia
g

n
o

st
ic

sp
e
c
ifi

c
it

y
0

.8
5

0
.8

1
0

.9
4

3
P

re
d

ic
ti

v
e

o
f

re
c
u

rr
e
n

c
e

0
.4

0
0

.4
8

0
.0

0

4
P

re
d

ic
ti

v
e

o
f

m
e
ta

st
a
si

s
0

.7
0

0
.6

1
0

.4
6

5
L

o
w

c
o

st
(£

)
8

.0
0

1
6

.0
0

1
7

.0
0

6
S

p
e
e
d

o
f

re
su

lt
(h

o
u

rs
)

2
4

.0
0

2
4

.0
0

2
4

.0
0

7
P

a
ti

e
n

t
a
c
c
e
p

ta
b

il
it

y
G

o
o

d
G

o
o

d
G

o
o

d

8
C

o
n

si
st

e
n

c
y

0
.9

3
0

.9
8

0
.9

6

C
E
A

c
a
rc

in
o

e
m

b
ry

o
n

ic
a
n

ti
g

e
n

,
C
A

c
a
n

c
e
r

a
n

ti
g

e
n

Use of Tumor Markers in Gastrointestinal Cancers 1167



latter two represented by the area under the curve (AUC).

Consistency was calculated from the I
2

heterogeneity

statistic from the included studies, representing the per-

centage of total variation across the studies, with a higher

number meaning lower consistency.

Cost-Benefit Trade-Off Analysis

The eight performance characteristics were broadly

divided into either costs (time for result and financial cost)

or benefits (sensitivity, specificity, predictive ability for

recurrence or metastases and consistency). To assess trade-

offs between costs and benefits among the tumor markers,

we employed Multi-criteria decision analysis (MCDA)

methods.
12–15

To achieve this, we rated each tumor marker

on all the performance characteristics (criteria ratings),

and assessed the relative importance of performance

characteristics (criteria weights) based on the average

rankings retrieved from the tumor marker survey (see

above). Using a weighted average model, we then com-

bined the ratings to produce an overall benefit score for

each tumor marker, and then contrasted the benefit scores

against the scores on costs and time criteria, respectively.

Criteria Ratings

Within MCDA, the performance of each tumor marker

(so-called a criteria rating), with respect to each charac-

teristic, is bounded between 0, assigned to the worst-

performing tumor marker (e.g. most expensive or least

sensitive), and 100 for the best-performing marker (e.g.

cheapest or most sensitive). We assumed linearity between

performance and rating, using linear interpolation to assess

criteria ratings of any intermediate performance. For

instance, if the costs of one tumor marker were halfway

between the most expensive and the least expensive

options, then this tumor marker received a criteria rating of

50 with respect to cost.

Criteria Weights

Trade-offs between the criteria are achieved through

‘criteria weights’, which capture the relative importance of

the eight performance characteristics. From the tumor

marker survey, respondents provided rankings as to the

desirability of each of the characteristics. We converted the

average rankings into numerical weights by assigning a

criteria weight of 100 to the highest ranked performance

characteristic, 90 to the second highest ranked, and so

forth. We then normalized all the weightings so that they

totaled 1. For example, if sensitivity received a criteria

weight of 100, whereas specificity received a weight of 50,

this would imply that the surgeons considered a difference

of 10%, with respect to the diagnostic sensitivity between

two tumor markers, to be equivocal to a difference of 20%

in their relative diagnostic specificities.

Overall Benefit Scores

Our aim was to assess trade-offs between the costs and

benefits of using a tumor marker. We therefore combined

the ratings of the characteristics previously designated as

benefits into an overall score. Under the assumption that

these characteristics are independent of each other, we can

assess the overall score by a weighted average model:
14

Benefit ¼
X

k

WkRk

where Rk is the rating on the kth benefit criterion and Wk is

the weight assigned to that criterion.

For instance, suppose the benefit of tumor markers

comprises their sensitivity and specificity alone. Assuming

that a tumor marker receives ratings of 100 and 50 on

sensitivity and specificity, respectively, which have nor-

malized weights of 0.75 and 0.25, it follows that the benefit

score of this tumor marker is 87.5 (=100 9 0.75 ?

50 ? 0.25).

Trade-Offs

Using normalized weights, the benefit scores are boun-

ded between 0 and 100; a hypothetical tumor marker that

has the worst performances on any criteria is scored 0,

whereas a tumor marker that has the best performances is

scored 100. The higher the score, the more beneficial the

tumor marker is and vice versa. We plotted two-way cost-

benefit maps to compare benefit against financial costs, and

separately from time to results. The analysis was performed

using the decision analytic software HiView, version

3.2.0.7 (educational copy).

RESULTS

Literature Search

A total of 151 full-text articles met the inclusion

criteria and were appraised following the literature

search (Online Appendix 2), constituting a total of

39,857 patients (19, 634 with colorectal malignancy and

20, 223 with esophagogastric malignancy). Overall, 102

articles assessed colorectal tumor markers (42 on diag-

nostic ability, 28 on prediction of recurrence, and 32 on

prediction of metastases), and 112 articles assessed

esophagogastric tumor markers (71 on diagnostic ability,

15 on prediction of recurrence, and 26 on prediction of

metastases).

1168 A. Acharya et al.



QUADAS-2 Evaluation

The results of the QUADAS-2 evaluation are shown in

Fig. 1. Of the 151 studies included, 40% had a ‘high risk’

of bias with respect to patient selection, while a further

12% did not provide sufficient detail on exclusion criteria

for patients enrolled in the study. With respect to the ref-

erence standard, 32% had a ‘high risk’ of bias due to their

retrospective nature.

Pooled Analysis for Diagnostic Sensitivity

and Specificity

The pooled diagnostic sensitivity and specificity of CEA

in colorectal cancer was 0.53 (95% CI 0.50–0.56) and 0.86

(95% CI 0.85–0.88), respectively (Table 3). For CA19-9,

sensitivity was 0.47 (95% CI 0.44–0.51) and specificity

was 0.92 (95% CI 0.91–0.94) and, for CA125, pooled

diagnostic sensitivity was 0.20 (95% CI 0.15–0.26) and

specificity was 0.99 (95% CI 0.98–1.00). Heterogeneity for

diagnosis determined using the I
2

statistic, which repre-

sented the consistency, was 0.86 for CEA and 0.96 for

CA19-9 and CA125. Similar results were found for

esophagogastric cancer (Table 4).

Tumor Marker Survey

The survey was distributed online from 1 August to 1

September 2015, and a total of 443 responses were T
A
B
L
E
3

R
e
su

lt
s

o
f

th
e

p
o

o
le

d
se

n
si

ti
v

it
y

,
sp

e
c
ifi

c
it

y
a
n

d
A

U
C

fr
o

m
th

e
li

te
ra

tu
re

se
a
rc

h
fo

r
d

ia
g

n
o

st
ic

a
b

il
it

y
,

p
re

d
ic

ti
v

e
a
b

il
it

y
o

f
re

c
u

rr
e
n

c
e

a
n

d
m

e
ta

st
a
se

s
fo

r
C

E
A

,
C

A
1

9
-9

a
n

d
C

A
1

2
5

,

w
it

h
re

sp
e
c
t

to
c
o

lo
re

c
ta

l
c
a
n

c
e
r

T
u

m
o

r
m

a
rk

e
r

D
ia

g
n

o
si

s
D

ia
g

n
o

si
s

o
f

re
c
u

rr
e
n

c
e

D
ia

g
n

o
si

s
o

f
m

e
ta

st
a
se

s

S
e
n

si
ti

v
it

y
S

p
e
c
ifi

c
it

y
A

U
C

S
e
n

si
ti

v
it

y
S

p
e
c
ifi

c
it

y
A

U
C

S
e
n

si
ti

v
it

y
S

p
e
c
ifi

c
it

y
A

U
C

C
E

A
0

.5
3

3
(0

.5
0

2
–

0
.5

6
3

)
0

.8
6

4
(0

.8
4

8
–

0
.8

8
1

)
0

.7
0

4
0

.5
5

0
(0

.5
2

7
–

0
.5

7
1

)
0

.8
0

8
(0

.7
9

6
–

0
.8

2
0

)
0

.7
7

9
0

.6
3

2
(0

.6
0

0
–

0
.6

6
4

)
0

.7
7

1
(0

.7
5

3
–

0
.7

8
9

)
0

.7
7

1

C
A

1
9

-9
0

.4
7

1
(0

.4
3

6
–

0
.5

0
6

)
0

.9
2

4
(0

.9
0

7
–

0
.9

3
9

)
0

.6
7

4
0

.3
2

5
(0

.2
8

9
–

0
.3

6
3

)
0

.7
5

1
(0

.7
2

1
–

0
.7

8
9

0
.5

4
9

0
.0

0
0

*
0

.0
0

0
*

0
.0

0
0

*

C
A

1
2

5
0

.1
9

8
(0

.1
4

8
–

0
.2

5
7

)
0

.9
9

3
(0

.9
7

7
–

0
.9

9
9

)
0

.9
9

3
0

.2
4

5
(0

.1
9

8
–

0
.2

8
7

)
0

.6
3

9
(0

.5
9

3
–

0
.6

8
3

)
0

.5
5

0
0

.0
0

0
*

0
.0

0
0

*
0

.0
0

0
*

V
a
lu

e
s

in
p

a
re

n
th

e
se

s
re

p
re

se
n

t
9

5
%

c
o

n
fi

d
e
n

c
e

in
te

rv
a
ls

A
U
C

a
re

a
u

n
d

e
r

th
e

c
u

rv
e
,
C
E
A

c
a
rc

in
o

e
m

b
ry

o
n

ic
a
n

ti
g

e
n

,
C
A

c
a
n

c
e
r

a
n

ti
g

e
n

*
In

su
ffi

c
ie

n
t

d
a
ta

a
v

a
il

a
b

le

Risk of Bias

Applicability

Flow

Reference Standard

Reference Standard

Index Selection

Index Selection

Patients Selection

Patients Selection

0 20 40

Percentage of Studies

Percentage of Studies

60 80 100

0 20 40 60 80 100

High

High

Low

Low

Unclear

Unclear

FIG. 1 Results of the QUADAS-2 evaluation of the included studies
with respect to the risk of bias and applicability

Use of Tumor Markers in Gastrointestinal Cancers 1169



collected, representing a response rate of 8.1%. There

were 273 respondents with an interest in colorectal

disease (200 consultants, 59 registrars, and 14 who

were primarily academic), and who had completed a

median of more than 100 cancer operations. CEA was

the most commonly utilized tumor marker in col-

orectal cancer (Fig. 2a), with surveillance for

recurrence the most common indication (Fig. 2c).

With respect to upper gastrointestinal surgery, 170

respondents (131 consultants, 29 registrars, and 10

primarily academic) had completed a median of more

than 100 cancer operations. CEA was the most

commonly employed tumor marker (Fig. 2b), with

assessment for recurrence being the most common

indication (Fig. 2d).

Of the eight performance characteristics, the ideal

tumor marker would have, diagnostic sensitivity

ranked the overall highest (most desirable), followed

by diagnostic specificity. Consistency across

demographics was considered the least desirable

(Tables 1, 2).

Cost-Benefit Trade-Off

All tumor markers had identical performances with

regard to patient acceptability and speed for result. As

all were derived from serum, with a result returned

within 24 h, these characteristics were removed from

the analysis. Tables 1 and 2 display the performance

of each tumor marker with respect to the eight char-

acteristics and the associated importance ranking

derived from the survey.

Figure 3 displays the trade-offs between the ben-

efits (high diagnostic sensitivity, specificity,

consistency, predictive ability for recurrence and

metastases) and costs (financial). CEA outperformed

both CA19-9 and CA125 with respect to overall

utility; it had lower associated financial costs and

higher benefits, as weighted by the importance placed

on the characteristics. This pattern was seen with both

colorectal and esophagogastric cancer.

DISCUSSION

The present study has highlighted the variable

performance of common tumor markers in the

assessment of gastrointestinal cancers. Despite this

variability, the majority of surgeons who were sur-

veyed utilize tumor markers in their practice. CEA

was found to have high utility, with a high ability to

predict recurrences and metastases, and was also

associated with the lowest costs, primarily financial,T
A
B
L
E
4

R
e
su

lt
s

o
f

th
e

p
o

o
le

d
se

n
si

ti
v

it
y

,
sp

e
c
ifi

c
it

y
a
n

d
A

U
C

fr
o

m
th

e
li

te
ra

tu
re

se
a
rc

h
fo

r
d

ia
g

n
o

st
ic

a
b

il
it

y
,

p
re

d
ic

ti
v

e
a
b

il
it

y
o

f
re

c
u

rr
e
n

c
e

a
n

d
m

e
ta

st
a
se

s
fo

r
C

E
A

,
C

A
1

9
-9

a
n

d
C

A
1

2
5

,

w
it

h
re

sp
e
c
t

to
e
so

p
h

a
g

o
g

a
st

ri
c

c
a
n

c
e
r

T
u

m
o

r
m

a
rk

e
r

D
ia

g
n

o
si

s
P

re
d

ic
ti

o
n

o
f

re
c
u

rr
e
n

c
e

P
re

d
ic

ti
o

n
o

f
m

e
ta

st
a
se

s

S
e
n

si
ti

v
it

y
S

p
e
c
ifi

c
it

y
A

U
C

S
e
n

si
ti

v
it

y
S

p
e
c
ifi

c
it

y
A

U
C

S
e
n

si
ti

v
it

y
S

p
e
c
ifi

c
it

y
A

U
C

C
E

A
0

.3
4

3
(0

.3
2

2
–

0
.3

6
5

)
0

.8
5

3
(0

.8
4

0
–

0
.8

6
6

)
0

.5
2

3
0

.3
3

5
(0

.3
0

3
–

0
.3

6
8

)
0

.8
6

0
(0

.8
4

8
–

0
.8

7
2

)
0

.4
6

4
0

.2
4

1
(0

.2
2

1
–

0
.2

5
7

)
0

.2
4

1
(0

.2
2

1
–

0
.2

5
7

)
0

.6
9

9

C
A

1
9

-9
0

.2
9

8
(0

.2
7

8
–

0
.3

1
8

)
0

.8
1

2
(0

.7
9

1
–

0
.8

3
1

)
0

.6
6

3
0

.2
5

8
(0

.2
2

5
–

0
.2

9
9

)
0

.8
5

4
(0

.8
3

9
–

0
.8

6
8

)
0

.4
7

8
0

.1
9

5
(0

.8
0

–
0

.2
1

0
)

0
.1

9
5

(0
.1

8
0

–
0

.2
1

0
)

0
.6

0
6

C
A

1
2

5
0

.3
1

3
(0

.2
6

4
–

0
.3

6
5

)
0

.9
3

5
(0

.9
1

0
–

0
.9

5
4

)
0

.7
4

9
0

.0
0

0
*

0
.0

0
0

*
0

.0
0

0
*

0
.3

3
2

(0
.2

8
5

–
0

.3
8

1
)

0
.9

1
3

(0
.8

9
5

–
0

.9
2

9
)

0
.4

6
0

V
a
lu

e
s

in
p

a
re

n
th

e
se

s
re

p
re

se
n

t
9

5
%

c
o

n
fi

d
e
n

c
e

in
te

rv
a
ls

A
U
C

a
re

a
u

n
d

e
r

th
e

c
u

rv
e
,
C
E
A

c
a
rc

in
o

e
m

b
ry

o
n

ic
a
n

ti
g

e
n

,
C
A

c
a
n

c
e
r

a
n

ti
g

e
n

*
In

su
ffi

c
ie

n
t

d
a
ta

a
v

a
il

a
b

le

1170 A. Acharya et al.



as all three tumor markers took the same timeframe to

attain the result. Therefore, CEA outperformed CA19-9

and CA125 in cost-benefit trade-off with respect to both

colorectal and esophagogastric cancer. However, all

markers had poor sensitivity, which would suggest their

use in diagnosis is significantly limited.
16,17

CEA is a glycoprotein produced in minimal amounts

after fetal development, and which is involved in cell

adhesion.
18

The use of CEA in postoperative surveillance

is well established, with the majority of surgeons surveyed

using tumor markers for monitoring recurrence. We have

shown that CEA could be of use in the assessment of

metastasis, having a predictive ability of 77%, due to its

association with the spread of cancer and increasing tumor

burden.

However, the use of CEA in diagnosis is not widely

endorsed. Its use is confounded by its association with

smokers, and the need for repeated measurements to mit-

igate limited sensitivity. Despite this, our survey showed

11% of surgeons still utilize tumor markers for diagnosis,

whereas we have shown CEA performs similarly to chance,

with a diagnostic sensitivity of 50%. Moreover, high

diagnostic sensitivity and specificity were the two most

desired characteristics of the ‘ideal tumor marker’, despite

monitoring of recurrence being the primary role for CEA.

Hence, there is a clinical need for an improved diagnostic

tumor marker for colorectal cancer. While utilizing a single

diagnostic marker is challenging, a diagnostic tumor mar-

ker that could be used in conjunction, or even triage further

more specific testing, would be of use. Tumor markers are

most commonly used in clinical practice in combination

rather than in isolation; however, it was not possible to test

the combination of tumor markers in this current study due

to limitations in the data. None of the traditional tumor

markers would satisfy the criteria for use alone clinically,

and, as such, research should focus on novel markers with

diagnostic ability.
19,20

In contrast to colorectal cancer, there remains no con-

sensus for the use of tumor markers in esophagogastric

cancer. Despite this, they appear widely used in clinical

practice, with CEA and CA19-9 the most often employed,

and monitoring for recurrence the most common indica-

tion, but, in as many as 20% of cases, surgeons admitted to

using tumor markers for diagnostic investigation. While

CEA again outperformed CA19-9 and CA125 overall, all

three tumor markers were found to have low diagnostic

accuracy, as determined by the AUC. Unlike in colorectal

cancer, all three were also found to have a low capacity to

predict recurrence, which should preclude their use in

widespread clinical practice.

(a) Colorectal cancer

CEA

Monitor Recurrence

Predict Metastases Assess Surgical Difficulty

Other

Diagnosis Monitor Recurrence

Predict Metastases Assess Surgical Difficulty

Other

Diagnosis

Ca19-9 Ca125 Other None CEA Ca19-9 Ca125 Other None

1%

39 %
32%

15%
20%

3%

14%

14%

31%

82%
71%

20%

6%

1%

2%
0%

4%

11%

3%

31%

(c) Colorectal Cancer

(b) Oesophago-gastric cancer

(d) Oesophago-gastric cancer

FIG. 2 Survey results detailing the most commonly used tumor markers and the most commonly cited indications in (a, c) colorectal cancer and
(b, d) esophagogastric cancer. CEA carcinoembryonic antigen, CA cancer antigen

Use of Tumor Markers in Gastrointestinal Cancers 1171



Discrepancy with the use of tumor markers in esopha-

gogastric cancer would suggest further prospective

evaluation is warranted, with a degree of discrimination

with respect to their clinical interpretation. Furthermore,

given the sensitivity of upper gastrointestinal endoscopy of

95%, the capacity for histological evaluation, and surveil-

lance for Barrett’s esophagus,
21

the pragmatic value of

traditional tumor markers for esophagogastric cancer is

restricted.

This study has also highlighted the importance of a

holistic evaluation of tumor markers prior to incorporating

them into clinical practice. While ‘benefits’ such as diag-

nostic sensitivity and specificity are key characteristics, as

seen by their relative importance denoted in the survey,

there must be an appreciation of costs. A marker that has

high sensitivity but has a high financial cost or requires

excessive processing time may be practically precluded

from widespread use, which may explain the fact that

despite poor diagnostic accuracy and only relatively high

predictive ability, the tumor markers this study has

appraised are still in widespread use. While novel markers

may offer greater diagnostic use, they are also likely to

require more esoteric, and therefore expensive, assays.

The limitations of this study are as a result of the pub-

lished studies included to inform the cost-benefit trade-off.

QUADAS evaluation revealed a ‘high’ level of bias in 32%

of the studies due to their retrospective design, limiting the

reliability of the results obtained from the pooled analyses.

Moreover, the potential for bias in some of these studies

was high due to their case-control design and restrictive

exclusion criteria, as was seen by the ‘high’ risk of bias

with respect to patient selection.

As only a few studies assessed the use of CA19-9 and

CA125 for the prediction of recurrence or metastases, there

were insufficient data to undertake the analyses, and an

assumption was therefore required, with the marker scoring

zero for that performance characteristic. This would lead to

CEA spuriously appearing to have higher benefits, simply

as a product of it being more extensively investigated. The

trade-off analysis is also informed by the survey, which

would only represent the views of members of the Euro-

pean society who responded. This suggests some positive

selection bias, especially given the response rate of

approximately 8%.

CONCLUSIONS

Tumor markers have long been utilized in the moni-

toring of gastrointestinal cancers, with variable success.

While traditional markers have a use in colorectal cancer

surveillance, their use in esophagogastric malignancies is

somewhat less defined and requires clarification. In both

cases, there appears to be a need for a tumor marker with

higher diagnostic accuracy. This would suggest that further

areas of research should focus on the search for new novel

biomarkers for diagnosis and therapeutic monitoring (See

Online Appendix 3 for References of included papers).

FUNDING Mr. Sheraz Markar is funded by the National Institute
for Health Research (NIHR). In addition, this research was supported

by the NIHR Diagnostic Evidence Co-operative London at the

Imperial College Healthcare NHS Trust. The views expressed are

those of the authors and are not necessarily those of the NIHR or the

Department of Health.

DISCLOSURE None.

OPEN ACCESS This article is distributed under the terms of the
Creative Commons Attribution 4.0 International License (http://

creativecommons.org/licenses/by/4.0/), which permits unrestricted

use, distribution, and reproduction in any medium, provided you give

appropriate credit to the original author(s) and the source, provide a

link to the Creative Commons license, and indicate if changes were

made.

100

100

90

80

80

70

60

60

50

40

Summation of Benefit Scores

S
u

m
m

at
io

n
 o

f 
C

os
t 

S
co

re
s

30

20

20 40

CA125
32,0

CA199
41,11

CEA
68,100

10

0

0

100

100

90

80

80

70

60

60

50

40

Summation of Benefit Scores

S
u

m
m

at
io

n
 o

f 
C

os
t 

S
co

re
s

30

20

20 40

CA125
36, 0

CA199
42,11

CEA
71,100

10

0

0

(a)

(b)

FIG. 3 Cost-benefit trade-off for use of the tumor markers in
(a) colorectal cancer and (b) esophagogastric cancer. Higher
summated cost score represents lower costs, while higher summated

benefit score represents higher benefits. CEA carcinoembryonic

antigen, CA cancer antigen

1172 A. Acharya et al.

http://creativecommons.org/licenses/by/4.0/
http://creativecommons.org/licenses/by/4.0/


REFERENCES

1. National Cancer Institute. SEER stat fact sheets. http://seer.

cancer.gov/statfacts/html/colorect.html. Accessed 1 Nov 2015.

2. Bagaria B, Sood S, Sharma R, Lalwani S. Comparative study of

CEA and CA19-9 in esophageal, gastric and colon cancers

individually and in combination (ROC curve analysis). Cancer

Biol Med. 2013;10(3):148–157.

3. Carpelan-Holmstrom M, Louhimo J, Stenman UH, Alfthan H,

Haglund C. CEA, CA 19-9 and CA 72-4 improve the diagnostic

accuracy in gastrointestinal cancers. Anticancer Res.

2002;22(4):2311–2316.

4. Yang XQ, Chen C, Wang FB, Peng CW, Li Y. Preoperative

serum carcinoembryonic antigen, carbohydrate antigen19-9 and

carbohydrate antigen 125 as prognostic factors for recurrence-

free survival in colorectal cancer. Asian Pac J Cancer Prev.

2011;12(5):1251–1256.

5. Locker GY, Hamilton S, Harris J, Jessup JM, Kemeny N, Mac-

donald JS, et al. ASCO 2006 update of recommendations for the

use of tumor markers in gastrointestinal cancer. J Clin Oncol.

2006;24(33):5313–5327.

6. Tas F, Faruk Aykan N, Aydiner A, Yasasever V, Topuz E.

Measurement of serum CA 19-9 may be more valuable than CEA

in prediction of recurrence in patients with gastric cancer. Am J

Clin Oncol. 2001;24(2):148–149.

7. Tian SB, Yu JC, Kang WM, Ma ZQ, Ye X, Cao ZJ, et al.

Combined detection of CEA, CA 19-9, CA 242 and CA 50 in the

diagnosis and prognosis of resectable gastric cancer. Asian Pac J

Cancer Prev. 2014;15(15):6295–6300.

8. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ,

Reitsma JB, et al. QUADAS-2: a revised tool for the quality

assessment of diagnostic accuracy studies. Ann Intern Med.

2011;155(8):529–536.

9. Harbord RM, Whiting P, Sterne JA, Egger M, Deeks JJ,

Shang A, et al. An empirical comparison of methods for

meta-analysis of diagnostic accuracy showed hierarchical

models are necessary. J Clin Epidemiol. 2008;61(11):

1095–1103.

10. Fox DJ, Guire KE. MIDAS. University of Michigan; Aug 1974.

https://babel.hathitrust.org/cgi/pt?id=mdp.39015000964760;view=

1up;seq=7.

11. Zhou XH, Tu W. Confidence intervals for the mean of diagnostic test

charge data containing zeros. Biometrics. 2000;56(4):1118–1125.

12. Keeney RL, Raiffa H. Decisions with multiple objectives: pref-

erences and value trade-offs. 2nd ed. Cambridge: Cambridge

University Press; 1993.

13. Muennig P. Cost effectiveness analysis in health: a practical

approach. 2nd ed. John Wiley & Sons; 2007.

14. Thokala P. The multiple criteria decision analysis for health

technology assessment. National Institute for Health and Clinical

Excellence; 2011.

15. Dolan JG. Multi-criteria clinical decision support: a primer on the use

of multiple criteria decision making methods to promote evidence-

based, patient-centered healthcare. Patient. 2010;3(4):229–248.

16. Chester SJ, Maimonis P, Vanzuiden P, Finkelstein M, Bookout J,

Vezeridis MP. A new radioimmunoassay detecting early stages of

colon cancer: a comparison with CEA, AFP, and Ca 19-9. Dis

Markers. 1991;9(5):265–271.

17. Feng B, Zheng MH, Zheng YF, Lu AG, Li JW, Wang ML, et al.

Normal and modified urinary nucleosides represent novel

biomarkers for colorectal cancer diagnosis and surgery monitor-

ing. J Gastroenterol Hepatol. 2005;20(12):1913–1919.

18. Lee JC, Lee SY, Kim CY, Yang DH. Clinical utility of tumor

marker cutoff ratio and a combination scoring system of preop-

erative carcinoembryonic antigen, carbohydrate antigen 19-9,

carbohydrate antigen 72-4 levels in gastric cancer. J Korean Surg

Soc. 2013;85(6):283–289.

19. Narita M, Oussoultzoglou E, Chenard MP, Fuchshuber P,

Yamamoto T, Addeo P, et al. Predicting early intrahepatic

recurrence after curative resection of colorectal liver metastases

with molecular markers. World J Surg. 2015;39(5):1167–1176.

20. Gold P, Freedman SO. Specific carcinoembryonic antigens of the

human digestive system. J Exp Med. 1965;122(3):467–481.

21. Choi SE, Hur C. Screening and surveillance for Barrett’s

esophagus: current issues and future directions. Curr Opin Gas-

troenterol. 2012;28(4):377–381.

Use of Tumor Markers in Gastrointestinal Cancers 1173

http://seer.cancer.gov/statfacts/html/colorect.html
http://seer.cancer.gov/statfacts/html/colorect.html
https://babel.hathitrust.org/cgi/pt%3fid%3dmdp.39015000964760%3bview%3d1up%3bseq%3d7
https://babel.hathitrust.org/cgi/pt%3fid%3dmdp.39015000964760%3bview%3d1up%3bseq%3d7

	Use of Tumor Markers in Gastrointestinal Cancers: Surgeon Perceptions and Cost-Benefit Trade-Off Analysis
	Abstract
	Background
	Objective
	Methods
	Results
	Conclusions

	Methods
	Literature Search Strategy
	Literature Standard
	Tumor Marker Survey
	Statistical Methodology
	Performance Characteristics
	Cost-Benefit Trade-Off Analysis
	Criteria Ratings
	Criteria Weights
	Overall Benefit Scores
	Trade-Offs

	Results
	Literature Search
	QUADAS-2 Evaluation
	Pooled Analysis for Diagnostic Sensitivity and Specificity
	Tumor Marker Survey
	Cost-Benefit Trade-Off

	Discussion
	Conclusions
	References