Clinical Chemistry 50:3

559 –563 (2004) Cancer Diagnostics

High Preoperative CA 15-3 Concentrations

Predict Adverse Outcome in Node-Negative and
Node-Positive Breast Cancer: Study of 600
Patients with Histologically Confirmed
Breast Cancer
Michael J. Duffy,1,2,3* Catherine Duggan,4 Rachel Keane,2,3 Arnold D.K. Hill,2,3
Enda McDermott,2,3 John Crown,5 and Niall O’Higgins2,3

Background: CA 15-3 is the most widely used serum
marker in breast cancer. Currently, its main uses are in
the surveillance of patients with diagnosed disease and
monitoring the treatment of patients with advanced
disease.
Methods: Preoperative CA 15-3 concentrations were
measured prospectively in 600 patients with histologi-
cally confirmed breast cancer. Marker concentrations
were related to patient outcome by both univariate and
multivariate analysis.
Results: After a median follow-up of 6.27 years, patients
with high preoperative concentrations of CA 15-3 (>30
units/L) had a significantly shorter overall survival
pattern than those with low concentrations. As a prog-
nostic factor, CA 15-3 was independent of tumor size,
axillary node status, and patient age. As well as being
prognostic in the total population of patients, CA 15-3
also predicted outcome in different subgroups of pa-
tients, including those with both node-negative and
node-positive disease, those who were both estrogen
receptor (ER)-negative and ER-positive, and those
younger and older that 50 years of age. CA 15-3 was also
Departments of 1 Nuclear Medicine and 5 Medical Oncology, St. Vincent’s
University Hospital, Dublin, Ireland.
2
Department of Surgery, Conway Institute of Biomolecular and Biomed-
ical Science, University College Dublin, Dublin, Ireland.
3
Dublin Institute of Molecular Medicine, Dublin, Ireland.
4
Department of Epidemiology, Mathematics and Statistics, Wolfson Insti-
tute of Preventive Medicine, Charterhouse Square, London, UK.
*Address correspondence to this author at: Department of Nuclear Medi-
cine, St. Vincent’s University Hospital, Elm Park, Dublin 4, Ireland. Fax
353-1-2696018; e-mail Michael.J.Duffy@ucd.ie.
Received August 15, 2003; accepted December 19, 2003.
Previously published online at DOI: 10.1373/clinchem.2003.025288
predictive of outcome irrespective of the type of adju-
vant therapy administered, i.e., whether adjuvant hor-
mone therapy, adjuvant chemotherapy, or radiotherapy
was administered.
Conclusion: Assay of CA 15-3 is a relatively inexpen-
sive, convenient, and noninvasive method for evaluat-
ing prognosis in newly diagnosed breast cancer patients.
© 2004 American Association for Clinical Chemistry
The CA 15-3 assay measures the protein product of the
MUC1 gene. MUC1 protein is a large transmembrane
glycosylated molecule containing three main domains,
a large extracellular region, a membrane-spanning se-
quence, and a cytoplasmic domain [for reviews, see Refs.
(1, 2 )]. Although the physiologic function of MUC1 is
unclear, the glycoprotein has been implicated in cell
adhesion, immunity, and metastasis (1, 2 ). Compared with
healthy breast tissue, MUC1 is present in higher concen-
trations but less glycosylated in breast carcinoma (2 ).
Currently, CA 15-3 is the most widely used serum
marker for breast cancer (3 ). Its main applications include
the surveillance of patients with diagnosed breast cancer
and monitoring therapy in advanced disease (1 ). In the
follow-up of patients with diagnosed breast cancer, in-
creased CA 15-3 was found either before or at the time of
recurrence in ⬃70% of cases (4 ). For monitoring the
treatment of advanced disease, CA 15-3 concentrations
were found to decrease in almost 70% of patients with
chemotherapy-induced disease regression and to increase
in 80% of patients with progressive disease (4 ).
Existing histologic (tumor size, tumor grade, and axil-
lary node status) (5 ) and biological [e.g., urokinase plas-
minogen activator, plasminogen activator inhibitor-1

559
560 Duffy et al.: CA 15-3 in Breast Cancer

(PAI-1),6 cathepsin D, and HER-2] (6, 7 ) prognostic factors

Table 1. Pathologic, biochemical, and clinical
for breast cancer all require tumor tissue. Clearly, the
characteristics of cancers investigated.
availability of a circulating prognostic factor, especially if
n %
it provided independent data and was prognostic within
Tumor size, cm
subgroups defined by traditional criteria, would be of
0–2 208 34.7
value in breast cancer. 2–5 341 56.8
Previously, both we (8 ) and others (9 –14 ) reported that ⬎5 51 8.5
patients with increased preoperative concentrations of CA Nodal status
15-3 had a worse prognosis than those with low concen- Negative 290 48.3
trations. The aim of this investigation was to confirm and 1–3 positive 126 21.0
extend our original findings, using both a larger number 4–10 positive 122 20.3
of patients (n ⫽ 600) and longer follow-up (median, 6.27 ⬎10 positive 62 10.3
years). In this study, we also investigated the prognostic ER status
impact of CA 15-3 in different subgroups of patients with Negative 161 26.8
breast cancer. Positive 344 57.3
Unknown 95 15.8
Materials and Methods Patient age
patients ⬍50 years 209 34.8
ⱖ50 years 391 65.2
A total of 647 patients with a histologic diagnosis of breast
Patient age (subgroups), years
cancer and attending the Breast Clinic at St. Vincent’s
25–29 3 0.5
University Hospital Dublin over the period July 1981
30–34 9 1.5
through November 1999 had preoperative CA 15-3 con-
35–39 40 6.7
centrations measured. Patients were excluded if they 40–44 71 11.8
had a previous diagnosis of breast cancer or if CA 15-3 45–49 86 14.3
was not determined in the 6 weeks before surgery. Of 50–54 89 14.8
these, 47 had incomplete data on either tumor size or 55–59 74 12.3
nodal status and were excluded from the study. All 60–64 71 11.8
patients underwent either modified radical mastectomy 65–69 59 9.8
or local excision combined with external beam radiother- 70–74 60 10.0
apy. Data relating to tumor size, axillary node involve- 75–79 26 4.3
ment, estrogen receptor (ER) status, and patient age are 80–84 8 1.3
summarized in Table 1. 85–89 4 0.7
A summary of the adjuvant treatments administered is
shown in Table 2. The median follow-up time was 6.27
years. Of the 600 patients investigated, 368 were also statistics
included in our previous study (8 ). However, these pa- Nonparametric analysis was performed because CA 15-3
tients were subjected to additional follow-up, whereas 232 did not follow a gaussian distribution. The Kruskal–
new patients were included. Wallis test was used for relating CA 15-3 concentrations to
both tumor size and nodal involvement, whereas the
assays Mann–Whitney U-test was used for relating CA 15-3
Serum CA 15-3 was measured by ELISA (ES300/Elecsys concentrations to both ER status and patient age. Cuzick’s
2010; Roche Diagnostics). Between-assay variation was (9 ) test for trend was used to determine whether there
⬍6% for commercial control sera (Preci Controls; Roche was a statistically significant trend of increasing CA 15-3
Diagnostics). CA 15-3 concentrations were stratified using
a cutoff value of 30 units/L because we had previously
shown that this is the optimum cutoff point (8 ). Tumor Table 2. List of adjuvant treatments administered
concentrations of ER were also determined by ELISA Treatment n %
(Abbott Diagnostics). The cutoff point for ER was 200 No therapy 42 7.0
fmol/g of tumor tissue. All of these assays were carried Hormone therapy alone 122 20.3
out by either qualified clinical chemists or laboratory Radiotherapy alone 51 8.5
technicians. Chemotherapy alone 26 4.3
Hormone ⫹ chemotherapy 39 6.5
Hormone ⫹ radiotherapy 99 16.5
Chemotherapy ⫹ radiotherapy 45 7.5
6
Nonstandard abbreviations: PAI-1, plasminogen activator inhibitor-1; Hormone ⫹ chemotherapy ⫹ radiotherapy 88 14.7
ER, estrogen receptor; HR, hazard ratio; CI, confidence interval; and CEA,
Unknown 88 14.7
carcinoembryonic antigen.
 Clinical Chemistry 50, No. 3, 2004 561

values across groups (tumor size and increasing nodal

involvement).
Analysis of events (death as endpoint) was performed
by Cox regression analysis (10 ). Hazard ratios (HRs) with
95% confidence intervals (CIs) were used to convey the
effects. All P values were two-sided. Multivariate analysis
was performed with Cox proportional hazard regression
(10 ). All factors that were significant at the 0.05 level in
the univariate analysis were included in the multivariate
proportional hazard model. Stepwise and backward re-
gressions were used to obtain the final model.

Results
relationship between ca 15-3 and tumor and
patient characteristics
CA 15-3 concentrations were significantly higher in pa-
tients with larger tumors (P ⫽ 0.002) and in patients with
increasing nodal burden (P ⫽ 0.004). Cuzick’s test for
Fig. 1. Overall survival according to serum CA 15-3 concentrations in
trend demonstrated a significant increase in CA 15-3
600 patients with breast cancer.
across these groups for tumor size (P ⬍0.0001) and for Thin line, CA 15-3 ⱕ30 units/L (n ⫽ 489); thick line, CA 15-3 ⬎30 units/L (n ⫽
nodal burden (P ⬍0.0001). There was no difference in CA 111). HR ⫽ 2.16 (CI, 1.55–3.03); P ⬍0.0001.
15-3 concentrations in patients who were ER positive or
negative, but concentrations were significantly higher in variate analysis, the prognostic benefit of CA 15-3 was
patients 50 years or older compared with those younger independent of tumor size, axillary node status, and
than 50 (P ⫽ 0.03). Concentrations were also higher in patient age. When used as a continuous variable, CA 15-3
patients who were axillary node positive compared with concentrations also predicted adverse outcome (P
those who were axillary node negative (P ⫽ 0.004). A ⬍0.0001).
detailed breakdown on the distribution of CA 15-3 con- The prognostic value of CA 15-3 in different subgroups
centrations in relation to tumor size, patient age, axillary of patients with breast cancer is summarized in Table 5.
nodal status, and ER status is shown in Table 3. Of particular significance was the finding that CA 15-3
predicted outcome in patients without histologic evidence
relationship between ca 15-3 and overall of metastasis to axillary nodes (Fig. 2). However, CA 15-3
survival was also prognostic in other subgroups, including node-
As shown in Fig. 1 and Table 4, patients with high CA positive patients, in those who had both ER-positive and
15-3 (⬎30 units/L) had a worse overall survival pattern -negative tumors, in those with tumors between 2 and 5
than those with low concentrations of the marker. This cm in size, and in those who were both younger and older
prognostic impact of CA 15-3 was clearly seen by both than 50 years. In contrast, CA 15-3 failed to correlate with
univariate and multivariate analysis (Table 4). In multi- outcome in patients with small tumors, i.e., ⬍2 cm in
diameter.
As shown in Table 6, CA 15-3 was also prognostic
Table 3. Median and mean CA 15-3 concentrations in irrespective of the type of adjuvant therapy administered,
different subgroups. i.e., whether patients received adjuvant hormone therapy,
CA 15-3, units /L
adjuvant chemotherapy, or radiotherapy. Although the
Variable n Median Mean number of patients not given any therapy postsurgery
Tumor size, cm was small (n ⫽ 42), high CA 15-3 concentrations also
0–2 208 20.0 20.8 appeared to be associated with shortened overall survival
⬎2–5 341 21.0 24.2 in this untreated group.
⬎5 51 26.0 37.2
Age at diagnosis Discussion
⬍50 years 209 19.0 22.8 As mentioned above, existing histologic and biological
ⱖ50 years 391 21.0 24.9 prognostic factors for breast cancer all require tumor
Axillary node status tissue. In this study, we both confirm and extend our
Negative 290 19.0 21.2 previous findings on the prognostic value of serum CA
Positive 310 21.0 26.9 15-3 in breast cancer (8 ). Our previous report included 368
ER status (n ⫽ 505) patients with a median follow-up of 3.28 years (8 ). In
Negative 161 20.0 24.1
contrast, this study includes 600 patients with a median
Positive 344 21.0 24.5
follow-up of 6.27 years. Among the most significant
562 Duffy et al.: CA 15-3 in Breast Cancer

Table 4. Comparative prognostic value of CA 15-3, nodal status, tumor size, patient age, and ER status.
Univariate analysis Multivariate analysis

Variable HR (95% CI) P HR (95% CI) P

CA 15-3 2.16 (1.55–3.03) ⬍0.0001 1.95 (1.2–3.07) 0.004
Nodal statusa 1.79 (1.31–2.44) ⬍0.0001 1.54 (1.04–2.77) 0.003
Tumor size 1.99 (1.46–2.08) ⬍0.0001 2.05b (1.31–3.21) 0.002
3.97c (1.96–8.0) ⬍0.0001
ER statusd 0.66 (0.47–0.91) ⬍0.03
Patient agee 1.11 (1.04–1.19) 0.001 1.17 (1.08–1.26) 0.0001
a
For nodal status, node negative was compared with node positive. For tumor size, patients were divided as follows: 0 –2 cm, 2–5 cm, and ⬎5 cm.
b
Patients with tumors 2–5 cm in size relative to those with tumors 0 –2 cm in size.
c
Patients with tumors ⬎5 cm in size relative to those with tumors 0 –2 cm size.
d
ER status was not included in the multivariate analysis because status was known in only 505 of 600 patients investigated.
e
Patient age was treated as a continuous variable.

differences between this and our previous report are the
inclusion of more patients and longer follow-up, CA 15-3
was prognostic in the node-negative subgroup of patients.
Another difference is that in the present study, CA 15-3
predicted outcome in ER-negative patients, whereas in
the earlier investigation it failed to do so (8 ). Furthermore,
we report for the first time that preoperative CA 15-3
concentrations predict outcome irrespective of the type of
adjuvant therapy administered.
Other groups have also found that high preoperative
CA 15-3 predicts adverse outcome in patients with breast
cancer (11–16 ). In two of these studies, only small num-
bers of patients were investigated (⬍100) and multivariate
analysis was not used (11, 12 ). Ebeling et al. (14 ), how-
ever, studied 1046 patients and found that preoperative
concentrations of both CA 15-3 and carcinoembryonic
antigen (CEA) were prognostic in breast cancer. In mul-
tivariate analysis, CEA retained its prognostic impact, but
CA 15-3 lost its value. Similarly, Canizares et al. (13 )
found that CA 15-3 was prognostic by univariate analysis
but not by multivariate analysis. In contrast to these
findings (13, 14 ) and in agreement with our study,
Kumpulainen et al. (16 ) recently reported that CA 15-3 is
an independent prognostic factor in breast cancer. It is of
interest that the two studies reporting a prognostic value
for CA 15-3 based on multivariate analysis both used 30
units/L as the cutoff point, whereas the reports failing to
find an independent prognostic impact used 40 (13 ) and
25 units/L (14 ).
In our study, CA 15-3 was also prognostic when we
used a cutoff point of 25 units/L, but at this lower cutoff
concentration, the prognostic impact was less than at 30
units/L [HR ⫽ 1.45 (P ⫽ 0.03) vs 2.16 (P ⬍0.0001)].
Similarly, in this investigation, CA 15-3 was prognostic
when we used a cutoff point of 40 units/L (HR ⫽ 2.65; P
⬍0.0001), but at this high cutoff, only 8% of the patients
would be regarded to have a poor outcome.
The most important group of patients with breast
cancer for which new prognostic factors are required is
the axillary node-negative subgroup. Currently, uroki-
nase plasminogen activator and PAI-1 are the only vali-
dated biological prognostic factors for this subgroup (17 ).
However, unlike CA 15-3, which can be measured in
serum, assays of urokinase plasminogen activator and

Table 5. Prognostic value of CA 15-3 in different subgroups

of patients with breast cancer.
Subgroup HR (95% CI) P
Node negative 2.43 (1.36–4.33) 0.003
Node positive 1.81 (1.20–2.76) 0.004
Tumor size 0–2 cm 1.51 (0.62–3.66) NSa
Tumor size ⬎2–5 cm 1.93 (1.26–2.94) 0.002
Tumor size ⬎5 cm 2.14 (0.98–4.70) 0.06
ER negative 1.88 (1.04–3.41) 0.03
ER positive 2.28 (1.43–3.64) 0.007
Age ⬍50 years 2.98 (1.64–5.46) 0.001 Fig. 2. Overall survival according to serum CA 15-3 concentrations in
Age ⱖ50 years 1.89 (1.26–2.84) 0.005 290 patients with axillary node-negative breast cancer.
a Thin line, CA 15-3 ⱕ30 units/L (n ⫽ 252); thick line, CA 15-3 ⬎30 units/L (n ⫽
NS, not significant
38). HR ⫽ 2.42 (CI, 1.35– 4.34); P ⫽ 0.003.
 Clinical Chemistry 50, No. 3, 2004
Table 6. Prognostic value of CA 15-3 in different subgroups
of patients with breast cancer based on adjuvant
therapy administered.
Adjuvant treatment HR (95% CI) P 5.
Hormone therapy 2.25 (1.50–3.38) 0.0004
Chemotherapy 1.8 (1.11–3.08) 0.03 6.
Radiotherapy 2.61 (1.71–3.99) ⬍0.0001
No therapya 6.63 (1.47–29.70) 0.03 7.
a Clin Lab Sci 2001;38:225– 62.
Results from the group receiving no adjuvant therapy should be interpreted
with caution because of small numbers (n ⫽ 42). 8.
PAI-1 require tumor tissue. In this investigation, we 9.
showed that preoperative CA 15-3 concentrations are a 10.
significant prognostic factor in this subset of patients.
Similar findings were reported recently by Gion et al. (15 ). 11.
Although histologic factors such as tumor size, tumor
grade, and lymph node status have been the cornerstone
of assessing cancer prognosis for decades, emerging data
12.
suggest that circulating markers can provide additional or
independent data. Thus, tumor markers such as ␣-feto-
protein, human chorionic gonadotropin, and lactate de-
hydrogenase were recently incorporated into the Union
Internationale Contre le Cancer (UICC) staging system for 13.
germ-cell tumor of the testis (18 ). Furthermore, a consen-
sus conference of the American Joint Committee on Can-
cer (AJCC) recently recommended that preoperative CEA
14.
concentrations should be added to the staging system for
colorectal cancer (19, 20 ).
15.
In conclusion, our study shows that CA 15-3 is both an
independent prognostic factor and prognostic in different
subgroups of patients with breast cancer. For node-nega-
tive patients, CA 15-3 concentrations could be of use in 16.
combination with other factors in deciding whether adju-
vant chemotherapy should be administered. In node-
positive patients, those with high CA 15-3 could be
17.
considered for more aggressive treatments. Clearly, CA
15-3 has the potential to help with individualized therapy
in patients with breast cancer. 18.
References
1. Duffy MJ. CA 15-3 and related mucins as circulating markers in
breast cancer. Ann Clin Biochem 1999;36:579 – 86. 19.
2. Duffy MJ, Shering S, Sherry F, McDermott E, O’Higgins N. CA 15-3:
a prognostic marker in breast cancer. Int J Biol Markers 2000;15:
330 –3.
3. Duffy MJ. Biochemical markers in breast cancer: which ones are 20.
clinically useful. Clin Biochem 2001;34:347–52.
4. Bast RC, Ravdin P, Hayes DF, Bates S, Fritsche H, Jessup JM, et
563
al. 2000 update of recommendations for the use of tumor markers
in breast and colorectal cancer: clinical practice guidelines of the
American Society of Clinical Oncology. J Clin Oncol 2001;19:
1865–78.
Elston CW, Ellis IO, Pinder SE. Pathological prognostic factors in
breast cancer. Crit Rev Oncol Hematol 1999;31:209 –23.
Duffy MJ. Biochemical markers as prognostic indices in cancer.
Clin Chem 1990;36:189 –91.
Duffy MJ. Clinical uses of tumor markers: a critical review. Crit Rev
Shering S, Sherry F, McDermott E, O’Higgins N, Duffy MJ. Preop-
erative CA 15-3 concentrations predict outcome in breast cancer.
Cancer 1998;83:2521–7.
Cuzick J. A Wilcoxon-type test for trend. Stat Med 1985;4:87–90.
Cox DR. Regression models and life tables. J Royal Stat Soc B
1972;34:187–220.
Berruti A, Tampellini M, Torta M, Buniva T, Gorzegno G, Dogliotti L.
Prognostic value in predicting overall survival of two mucinous
markers: CA 15-3 and CA 125 in breast cancer patients at first
relapse of disease. Eur J Cancer 1994;30A:2082– 4.
Tampellini M, Berutti A, Gerbino A, Buniva T, Torta M, Gorzegno G,
et al. Relationship between CA 15-3 serum levels and disease
extent in predicting overall survival of breast cancer patients with
newly diagnosed metastatic disease. Br J Cancer 1997;75:698 –
702.
Canizares F, Sola J, Perez M, Tovar I, De Las Heras M, Salinas J,
et al. Preoperative values of CA 15-3 and CEA as prognostic
factors in breast cancer: a multivariate analysis. Tumour Biol
2002;22:273– 81.
Ebeling FG, Stieber P, Untch M, Nagel D, Konecny GE, Schmitt UM,
et al. Serum CEA and CA 15-3 as prognostic factors in primary
breast cancer. Br J Cancer 2002;22:1217–22.
Gion M, Boracchi P, Dittadi R, Biganzoli E, Peloso L, Mione R, et
al. Prognostic role of serum CA 15.3 in 362 node-negative breast
cancer. An old player for a new game. Eur J Cancer 2002;38:
1181– 8.
Kumpulainen EJ, Keskikuru R, Johansson RT. Serum tumor
marker CA 15.3 and stage are the two most important predictors
of survival in primary breast cancer. Breast Cancer Res Treat
2002;76:95–102.
Duffy MJ. Urokinase plasminogen activator and its inhibitor, PAI-1,
as prognostic markers in breast cancer: from pilot to level 1
evidence studies. Clin Chem 2002;48:1194 –7.
International Germ Cell Collaboration Group. International germ
cell consensus classification: a prognostic factor-based staging
system for metastatic germ cell cancers. J Clin Oncol 1997;15:
594 – 603.
Compton C, Fenoglio-Preiser CM, Pettigrew N, Fielding LP. Ameri-
can Joint Committee on Cancer Prognostic Factors Consensus
Conference: Colorectal Working Group. Cancer 2000;88:1739 –
57.
Compton CC, Fielding LP, Burgart LJ, Conley B, Cooper HS,
Hamilton SR, et al. Prognostic factors in colorectal cancer. Arch
Pathol Lab Med 2000;124:979 –94.