Original Article

Tumour markers: their use and misuse by clinicians

Peter J McGinley and Eric S Kilpatrick

Abstract
Address Background Several guidelines exist on the appropriate use of serum tumour
Department of Clinical Biochemistry markers in the management of patients with cancer. This study audited tumour
Hull Royal InŽ rmary
marker requesting against these guidelines in a busy teaching hospital over a 12-
Hull HU3 2JZ, UK
month period.
Correspondence Methods All marker requests from 1 April 2001 to 31 March 2002 were collected
Dr Eric S Kilpatrick using the laboratory computer. From one 24-h period, the case notes from all hospital
E-mail: eric_kilpatrick@hotmail.com requests [excluding prostate-specific antigen (PSA)] were examined.
Results Tumour marker workload increased by 125% from 1997-98 to 2001-02.
Of 27 323 tumour marker requests, 7166 were from general practice, 2312 were
requested on hospital admission before further investigation, 612/3636 of CA125 and
98/374 of CA15.3 requests were on men and 12/11 585 PSA requests on women. Of
34 case notes examined, 18 had tumour markers measured before biopsy and only
nine after. Of 19 patients with ‘normal’ markers, one had malignancy on biopsy and,
of 15 with one or more raised markers, four had normal biopsies.
Conclusions Tumour marker workload is rapidly increasing. Tumour markers are
frequently and inappropriately requested, either because they are on patients of the
wrong sex, or because they are taken before a cancer diagnosis is reached. Results
from these tests can be falsely reassuring or unduly alarming.
Ann Clin Biochem 2003; 40: 643–647

Introduction performed, as is the measurement of prostate-speci¢c

The use of serum tumour markers in the management
of malignant disease has become established prac-
tice,1^3 and their measurement forms a mainstay of
the modern pathology laboratory. However, the
speci¢c role that tumour markers may ful¢l is a subject
of deliberation, 4^7 and whether marker testing could
be utilized outside the currently held guidelines2,3,8
continues to be debated.9^11
Whilst many biological factors are known to be
produced as a result of malignant growth,5,12^15 a
select series of tumour markers are readily associated
with cancer development and thus frequently
requested by clinicians. 2^4,10,16 In particular, the
measurement of carcinoembryonic antigen (CEA) in
cases of colorectal adenocarcinoma 2,3,10,17,18 and
cancer antigen (CA) 125 during the treatment of
ovarian disease 2,9,11,16,17,19 are routinely requested. In
addition, the testing of carbohydrate antigen (CA) 19-91
and cancer antigen (CA) 15-3 for pancreatic12,17 and
breast3,17,19 cancer, respectively, are commonly
antigen (PSA) in prostate cancer.4,5,7,15,20 The
management of germ-cell malignancies1,3 also often
entails the dual measurement of a-fetoprotein (AFP)
and human chorionic gonadotrophin (hCG).1^3,18
Although increased levels of serum tumour
markers are often associated with the presence of
cancer,2^5,7,16,17,20 marker concentrations may also rise
in a number of benign conditions1^3,9,18,21 and meta-
bolic or hormonal changes.2,9,19 Conversely, a malig-
nancy may not generate elevated tumour marker
concentrations until late in its development,6,9,11,15,16 if
at all, and none of the main tumour markers are
wholly speci¢c to a single cancer type.2,3,17,18
To this end, most guidelines recommended to
clinicians 2,3,8 state that tumour markers such as CEA,
CA15-3, CA125 and CA19-9 should only be used once a
conclusive diagnosis of malignant disease has been
achieved, and that marker assays are applied speci¢-
cally to monitor response to treatment,2,3,16,17 screen
for potential reoccurrence 2 and as a prognostic
indicator.1,2,16,22

© 2003 The Association of Clinical Biochemists 643

644 McGinley and Kilpatrick
Notwithstanding these recommendations, extending
the use of tumour marker testing to diagnose or case
¢nd malignant disease has generated considerable
interest 4^7,9^12,15,23 with the instigation of studies to
assess the merits of inclusion in the diagnostic
process.23 Indeed, in some areas of cancer diagnosis
the use of tumour markers is already advocated by
some,7,20 despite evidence relating to the lack of
speci¢city and sensitivity of such testing,2,3,9,17^19 and
thereby the ability to satisfactorily di¡erentiate
between benign and malignant disease.1^3,9,18,21
While this contentious issue continues to be
debated,6,9,11 there are concerns as to whether tumour
markers are already being used by clinicians as part of
a diagnostic strategy for their patients, which may in
turn partly explain why many hospital laboratories
have been receiving increasing numbers of tumour
marker requests despite a lack of an obvious change in
cancer incidence or treatment during the same period.
This study was therefore undertaken with two main
aims. The ¢rst was to identify the changes in tumour
marker workload in a busy teaching hospital since
1997/98 and to identify the source of current
requesting. The second aim was to take a sample of
recent requests and to audit from case notes and
reports whether tumour markers are being utilized in
accordance with published clinical guidelines, or
whether marker assays are also being performed in a
screening capacity for suspected malignancies.
Methods
Patient population
All tumour marker requests sent to the Department of
Clinical Biochemistry, Hull Royal In¢rmary, during
the period 1 April 2001 to 31 March 2002 were
included in the study. The markers routinely
measured, together with their quoted reference
intervals, are shown in Table 1. Data collection of
tumour marker results was performed by extracting
them from the laboratory’s pathology computer
system (Masterlab, Torex Health Systems, She¤eld,
UK) along with demographic details of age, sex,
Table 1. Tumour markers investigated by Hull and East
Yorkshire Hospitals NHS Trust, and reference intervals issued to
clinicians
Tumour marker Reference interval
Carcinoembryonic antigen (CEA) 0–5 ng/mL
Carbohydrate antigen 19-9 (CA19-9) 0–37 U/mL
Cancer antigen 125 (CA125) 0–35 U/mL
Cancer antigen 15-3 (CA15-3) 0–28 U/mL
Prostate-speciŽ c antigen (PSA) 0–4 ng/mL
a-Fetoprotein (AFP) 0–10 kU/L
Human chorionic gonadotrophin (hCG) 0–5 mIU/mL
Ann Clin Biochem 2003; 40: 643–647
hospital number and the source of the request. To
gauge the change in workload, the same data were
gathered for the period 1 April 1997 to 31 March 1998.
From the 2001/02 data collection, a more detailed
analysis was undertaken of individual cases identi¢ed
from specimens sent on a single weekday (randomly
chosen as 18 September 2001). These cases did not
include any samples sent from general practitioners
(GPs) in primary care because of the likely di¤culty in
obtaining case records, so only requests from the Hull
and East Yorkshire Hospitals NHS Trust were included.
In addition, the decision was taken not to include
patients with prostate-speci¢c antigen (PSA) requests
in this part of the study because of the lack of
consensus as to whether the test should be used in
screening and/or case ¢nding for prostate cancer, with
some authorities advocating its use 3,4,7,20 and others
not.2,3
The detailed analysis took two forms. First, the case
notes for each patient were examined and a ques-
tionnaire completed to establish details surrounding
the timing and reasons for tumour marker requesting.
Then, for each individual, the Masterlab laboratory
computer system was used to identify (or con¢rm)
the temporal relationship between tumour marker
request(s) and biopsy diagnosis, or otherwise, of a
malignancy.
Approval for the audit was gained from the Hull and
East Yorkshire Hospitals NHS Trust prior to the data
collection.
Laboratory assays
All tumour marker analyses were performed on a DPC
Immulite 2000 immunoassay analyser (Diagnostics
Products Corporation UK, Glyn Rhonwy, UK) with the
exception of PSA testing which was conducted using
an Abbott Architect i4000 immunoassay analyser
(Abbott Diagnostics Division, Maidenhead, UK). All
assays achieved satisfactory internal and external
quality assessment during the study period.
Statistical analysis
Statistical analysis was performed using Microsoft
Excel throughout.
Results
The Hull and East Yorkshire Hospitals NHS Trust
carried out a total of 27 323 tumour marker tests on
13218 patients (9726 men, 3213 women, 279 sex
unspeci¢ed; median age 67 years (interquartile range
57^76) between 1 April 2001 and 31 March 2002.
Table 2 details the number of tumour marker requests
(including and excluding PSA) from general practice
and from secondary care outpatient clinics and ward
areas. In relation to the change in tumour marker
requesting since 1997/98, Fig. 1 shows the change in
 Tumour markers: their use and misuse by clinicians 645

Table 2. Distribution of requests issued for tumour marker From the patients tested on 18 September 2001, 34
testing, 1 April 2001 to 31 March 2002 ful¢lled the criteria of having been hospital inpatients
or outpatients who were tested for one or more tumour
Total tumour Tumour marker markers other than PSA (median 2 markers, range
marker requests excluding
1^6). Figure 2 shows the number of cases in which
Requesting location requests PSA
tumour markers were tested either before or following
GP surgeries 7166 1329 a biopsy/cytology examination, as well as the number
OPD/clinics/day surgeries 10 335 9100 for which a biopsy was never performed. The results of
Wards/admissions 8215 6463 the initial tumour marker requests for 19 of the study
Miscellaneous sources 1495 1039 group members returned all results within stated
No location/source 112 83 reference intervals and 15 subjects showed one or
PSA, prostate-speciŽ c antigen; GP, general practitioner; OPD, out- more initial tumour marker results exceeding these
patient departments. ranges.
From the 19 cases in which tumour marker results
the average number of requests per month for all
returned within the reference intervals, Fig. 3 shows
markers.
one subject diagnosed with malignant disease upon
During the period 1 April 2001 to 31 March 2002, a
total of 2312 requests for tumour markers were issued
from the acute admitting wards and the accident and
emergency department of the Hull and East Yorkshire
Hospitals NHS Trust. This ¢gure includes 120 requests
for marker testing by accident and emergency (57 of
which were for PSA), 19 tests performed on behalf of
the short stay ward (two for PSA), 1054 tests requested
by the medical assessment unit (264 for PSA) and 1119
from the surgical admissions ward (57 for PSA).
During the 12-month study period, 612 of the 3616
requests for CA125 were on male subjects, a further 98
out of 374 tests for CA15-3 were also on men, and 12 of
the 11585 requests for PSA were on women. Sixty-six Figure 2. Temporal relationship between tumour marker and
of the 98 men who had CA15-3 measured had biopsies biopsy requests (n=34). Hatched area, tumour markers
examined in the year of the study or in the 2 years requested before biopsy; open area, tumour markers requested
prior. None of these were on breast tissue. after biopsy; Ž lled area, biopsy not performed.

Figure 1. Change in average monthly requests for tumour markers in Hull and East Yorkshire. CEA, carcinoembryonic antigen; CA,
carbohydrate/cancer antigen; AFP, a-fetoprotein; hCG, human chorionic gonadotrophin; PSA, prostate-speciŽc antigen. Open columns,
1997/98; hatched columns, 2001/02.

Ann Clin Biochem 2003; 40: 643–647

646 McGinley and Kilpatrick
the completion of a biopsy study. The results of biopsies
for 14 remaining subjects showed no indications of
malignancy, whilst biopsies were not performed in
four cases. From the 15 cases in which initial tumour
marker testing indicated one or more result exceeding
the stated reference intervals, Figure 3 also shows
three did not include a biopsy study whilst 12
proceeded to undergo such a procedure. Of these 12
individuals, eight were diagnosed with malignant
disease as a result of the biopsy report and four
individuals showed no indications of malignancy.
Discussion
This study has provided an opportunity to assess the
role that tumour markers are now ful¢lling in clinical
practice, and to compare this role to that promoted by
the published guidelines and recommendations.2,3,8 It
has also given a degree of insight into the place that
tumour markers now have in the investigation and
management of patients with cancer or suspected
malignant disease.
The data show that there has been a marked
increase in the requesting of tumour markers in recent
years, with a 125% rise overall between 1997/98 and
2001/02 (see Fig. 1). As cancer diagnoses have not
risen by a similar proportion, it would be reassuring to
suggest that this increase is either as a result of clin-
icians being more aware of the existence of tumour
marker testing and/or because they are su¤ciently
comfortable with the application and signi¢cance of
these tests to use them more readily in appropriate
patients.
Unfortunately, the ¢ndings of this study would
suggest that this is not the only reason for the rise in
requests. One other cause appears to be a degree of
ignorance amongst healthcare sta¡ about which
tumour marker is appropriate for which tumour type.
For example, 17% of all CA125 requests (a marker
typically associated with ovarian disease) and 26% of
all CA15-3 tests (a common marker of breast malig-
nancy) were performed on men. Despite the fact that
Ann Clin Biochem 2003; 40: 643–647
Figure 3. Outcome of tissue biopsies in patients
with tumour marker results within and outside
stated reference intervals (n=34). Hatched areas,
malignancy; open areas, no malignancy; Ž lled
areas, no biopsy.
these markers can be raised in other malignancies, it
seems likely that many of these requests are inap-
propriate. Male breast cancer exists, but none of the
men tested for CA15-3 had breast biopsies in the year
of the study or the 2 years prior. Even for PSA, there
was still a small (0.1%) number of cases where the test
was also performed on the wrong sex of patient. This
information would therefore suggest that there is not
only a need to promote published guidelines but also a
requirement to give more basic information regarding
the use of these markers. The laboratory itself also has
a responsibility in setting up systems that will high-
light these requests before they are analysed.
As well as demonstrating a lack of knowledge, the
use of tumour marker testing on the wrong sex of
patient may also be indicative of the use of one or more
of these tests to `screen’ for cancer in someone not
known to have malignant disease. The use of testing in
this way in Hull (in discordance with the majority of
guidelines relating to the issue) is corroborated by the
fact that 1932 non-PSA results for tumour markers
were issued from the accident and emergency and
acute medical admission wards in the year studied. As
one of the fundamental roles of these units is to
acutely diagnose cases before forwarding them to
other departments or wards, the necessity to perform
tumour markers at this early stage is probably open to
question.
The suspicion that tumour markers are being
requested before a diagnosis is made was con¢rmed by
the 34 cases studied at length. In the majority of
instances a request for one or more tumour markers
was issued before obtaining an appropriate diagnosis.
As these tests were performed and their results
returned to the clinicians before a conclusive tissue
study was reported, it would seem proof that the
¢ndings of these tumour marker requests were being
taken into account before a ¢nal diagnosis was
reached.
Use of tumour markers in this manner would still be
of value if they gave an accurate indication of whether
the patient had malignancy or not, but, even in the

small study sample, this was not the case. Instead, the
data here con¢rm larger studies in showing that
`normal’ tumour marker values can be falsely reas-
suring, while four of the 12 individuals with initial
marker results exceeding the stated local reference
intervals did not demonstrate any malignancy on
biopsy. It is not possible to know from the current
study whether the raised markers in these four
individuals led to any undue distress or needless
investigation on the part of the patient, as has been
reported previously.21
In summary, this study has demonstrated a
substantial rise in serum tumour marker requests in a
busy teaching hospital over a 4-year period. Some of
this rise appears to be due to a lack of knowledge
amongst some clinicians as to what malignancies a
particular marker may be of value in. In other
instances there is good evidence that, despite
published guidelines, these tests are being used to
`screen’ for malignancy before a tissue diagnosis is
made. In some of these cases, the results gained from
the screen will have been falsely reassuring, while in
others they are likely to have been unduly alarming. At
the very least, these ¢ndings highlight the need for
further educating clinical sta¡ in the use and misuse
of tumour marker testing.
References
1 Hanning I, Dow E. Tumour markers. Surgery 1998; 16: iii–v
2 Duffy M, McGing P, McSweeny J. Guidelines for the Use of Tumour
Markers. Association of Clinical Biochemists in Ireland, 2000
www.iol.ie/*deskenny/acbi-tmk.html (last accessed September
2003)
3 Sturgeon C. Practice guidelines for tumour marker use in the clinic.
Clin Chem 2002; 48: 1151–9
4 Gorgoulis V, Tiniakos D, Mariatos G, Manolis E, Mitropoulos D, Kittas
C. Recent advances in early prostatic cancer. Anticancer Res 1999;
19: 2327–48
5 Blijenberg B, Storm B, Van Zelst B, Boeken Kruger A, Schroder F.
New developments in the standardisation of prostate speciŽ c
antigen. Clin Biochem 1999; 32: 627–34
6 Farley K. Ovarian cancer early detection program. Oncol Nurs
Forum 1999; 26: 833–4
7 Kroll M. Prostate cancer: free PSA adds value to PSA testing. Clin
Lab News 2002; 4: 10–1
Tumour markers: their use and misuse by clinicians 647
8 European Society for Medical Oncology. ESMO minimum clinical
recommendations for diagnosis, adjuvant treatment and follow-up of
colon cancer. Ann Oncol 2001; 12: 1053–4
9 Bell R. Screening for ovarian cancer: is it worth the risk? Nurs
Times 1999; 95: 52–3
10 Maxwell P. Carcinoembryonic antigen: cell adhesion molecule and
useful diagnostic marker. Br J Biomed Sci 1999; 56: 209–14
11 Urban N. Screening for ovarian cancer: we now need a deŽ nitive
randomised trial. BMJ 1999; 319: 1317–8
12 Rhodes J. Usefulness of novel tumour markers. Ann Oncol 1999; 4:
S118–S121
13 Konety B, Getzenberg R. Urine based markers of urological
malignancy. J Urol 2001; 165: 600–11
14 Kim J-H, Skates S, Uede T, Wong K-W, Schorge J, Feltmate C.
Osteopontin as a potential biomarker for ovarian cancer. JAMA
2002; 287: 1671–9
15 Ikihara O, Cheli C, Partin A, Fritche H, Chan D, Sokoll L.
Comparative analysis of complexed prostate-speciŽ c antigen, free
prostate-speciŽ c antigen and their ratio in detecting prostate cancer.
J Urol 2002; 167: 2017–24
16 Rustin G, Nelstrop A, Betzen S, Piccart M, Bertelsen K. Use of
tumour markers in monitoring the course of ovarian cancer. Ann
Oncol 1999; 10: S21–S27
17 Erbagci B, Yilmaz N, Kutlar I. Menstrual cycle dependent variability
for serum tumour markers CEA, AFP, CA19-9, CA125 and CA15-3
in healthy women. Dis Markers 1999; 15: 259–67
18 Khalifa A, Mady E, Abadeer N, Kamal A. Differential tumour
markers and hepatitis markers proŽ le in liver tumours. Anticancer
Res 1999; 19: 2495–500
19 Bon G, Kenemans P, Dekker J, Hompes P, Verstraeten R, van
Kamp G. Fluctuations in CA125 and CA15-3 serum concentrations
during spontaneous ovulatory cycles. Hum Reprod 1999; 14: 566–70
20 Brett J, Watson E, Bukah C, Austoker J. PSA Testing for Prostate
Cancer. An information sheet for men considering a PSA test.
Cancer Research UK Primary Care Education Research Group,
2002 www.dphpc.ox.ac.uk/crcpcerg/research/resprostate.htm (last
accessed September 2003)
21 Krishnan S, Philipose Z, Rayman G. Hypothyroidism mimicking
intra-abdominal malignancy. BMJ 2002; 325: 946–7
22 Lamm D. The challenge of improved diagnostic and prognostic
indicators of superŽ cial bladder cancer. J Urol 2000; 164: 690–1
23 Jacobs I, Skates S, MacDonald N, Menon U, Rosenthal A, Prys
Davis A. Outcome of a pilot randomised controlled trial of ovarian
cancer screening. Lancet 1999; 253: 1207–10
Accepted for publication 23 June 2003
Ann Clin Biochem 2003; 40: 643–647