Breast Cancer Res Treat (2013) 142:143–151

DOI 10.1007/s10549-013-2714-8

EPIDEMIOLOGY

Do alternative methods of measuring tumor size, including

consideration of multicentric/multifocal disease, enhance
prognostic information beyond TNM staging in women with early
stage breast cancer: an analysis of the NCIC CTG MA.5
and MA.12 clinical trials
J. F. Hilton • N. Bouganim • B. Dong • J. W. Chapman • A. Arnaout •
F. O’Malley • K. A. Gelmon • R. Yerushalmi • M. N. Levine • V. H. C. Bramwell •

T. J. Whelan • K. I. Pritchard • L. E. Shepherd • M. Clemons

Received: 2 September 2013 / Accepted: 26 September 2013 / Published online: 11 October 2013
 Springer Science+Business Media New York 2013

Abstract The AJCC staging criteria consider tumor size
to be the largest dimension of largest tumor. Some case
series suggest using summation of all tumor dimensions in
patients with multicentric/multifocal (MC/MF) disease. We
used data from NCIC CTG MA.5 and MA.12 clinical trials
to examine alternative methods of assessing tumor size on
breast-cancer-free-interval (BCFI). The 710 MA.5 pre-/
peri-menopausal node positive and 672 MA.12 pre-meno-
pausal node-negative/-positive patients have 10-year med-
ian follow-up. All patients received adjuvant chemotherapy.
This work was presented as a poster at the CRTC-AACR San Antonio
Breast Cancer Symposium, Dec 6–10 2011, San Antonio, Texas,
USA.
Tumors were centrally reviewed for grade, hormone
receptor, and HER2 status. Continuous pathologic tumor
size was: (1) largest dimension of largest tumor (cm); (2)
tumor area (cm2); (3) volume of tumor (cm3); (4) with MC/
MF disease, summation of (1)–(3) for up to 3 foci. We
examined univariate and multivariate effects of tumor size
on BCFI utilizing (un)stratified Cox regression and the
Wald test statistic. In univariate analysis, larger tumor
dimension was significantly associated with worse BFCI in
node positive patients: p \ 0.0001 for MA.5; p = 0.01 for
MA.12. In MA.5 multivariate analysis, larger summation of
largest tumor dimensions was associated with worse BCFI
(p = 0.0003), while larger single dimension was associated
with worse BCFI (p = 0.02) for MA.12. Presence of MC/

J. F. Hilton
B. Dong
J. W. Chapman
L. E. Shepherd M. N. Levine

NCIC Clinical Trials Group, Queens University, Kingston, ON, Department of Oncology, Juravinski Cancer Centre, McMaster
Canada University, Hamilton, ON, Canada

J. F. Hilton
N. Bouganim
M. Clemons (&) V. H. C. Bramwell

Division of Medical Oncology, Department of Medicine, The
Ottawa Hospital Cancer Centre, Ottawa General Hospital,
University of Ottawa, 501 Smyth Road, Ottawa, ON, Canada
e-mail: mclemons@ottawahospital.ca
A. Arnaout
Department of Surgery, The Ottawa Hospital Cancer Centre,
University of Ottawa, Ottawa, ON, Canada
F. O’Malley
Department of Pathology, Mt. Sinai Hospital, University of
Toronto, Toronto, ON, Canada
Department of Oncology, Tom Baker Cancer Centre, University
of Calgary, Calgary, AB, Canada
T. J. Whelan
Department of Radiation Oncology, Juravinski Cancer Centre,
McMaster University, Hamilton, ON, Canada
K. I. Pritchard
Sunnybrook Odette Cancer Centre, University of Toronto,
Toronto, ON, Canada

K. A. Gelmon
R. Yerushalmi
Department of Oncology, BC Cancer Agency, University of
British Columbia, Vancouver, BC, Canada

123
144
MF and other tumor size measurements were not associated
(p [ 0.05) with BFCI. While physicians could consider the
largest diameter of the largest focus of disease or the sum of
the largest diameters of all foci in their T-stage determi-
nation, it appears that the current method of T-staging offers
equivalent determinations of prognosis.
Keywords Breast cancer
Alternative methods of
staging
Breast cancer free interval (BCFI)
Introduction
Breast cancer staging by the AJCC tumor node metastasis
(TNM) staging system remains an important determinant
of patient prognosis and treatment [1]. Currently, the
standard method for evaluating the primary lesion is to
measure its largest unidimensional length [1]. There are,
however, theoretical reasons to consider other methods to
measure tumor size. For example, primary tumors rarely
have a uniform shape, possibly resulting in a measurement
which may over or understate the volume of disease
present. By considering the tumor as either a 2-dimensional
(i.e., area calculation) or a 3-dimensional (i.e., volume
calculation) object, it may be possible to more accurately
state disease burden.
In addition, the largest unidimensional measurement
may not be representative of the total breast tumor burden
in patients with multicentric (MC) or multifocal (MF)
disease. Although the TNM system includes the ‘‘m’’
modifier when MF disease is present, this designation
does not change the overall stage, and therefore leaves
sizable foci of disease out of the prognostic calculation.
MC/MF disease is not infrequent and occurs in around
13–21 % of cases [2–4]. Retrospective population studies
have suggested that MF disease is associated with a
reduced disease-specific survival when compared to a
single focus of disease [3, 5–7]. In these situations, a
measurement which includes all the diseases in the tumor
size evaluation may offer improved prognostic informa-
tion to guide physicians.
Using the mature survival data from the National Cancer
Institute of Canada Clinical Trials Group (NCIC CTG)
MA.5 and MA.12 trials, we examined: (1) whether alter-
native methods of estimating tumor size, such as tumor
area and volume provide a more accurate assessment of the
breast cancer free interval (BCFI) compared to unidimen-
sional measurements and (2) whether summation of max-
imal lengths, areas, or volumes for patients identified in
patients with MC or MF disease offer improved prognostic
information compared to the simple maximal length of the
largest focus.
123
Breast Cancer Res Treat (2013) 142:143–151
Materials and methods
Study population
NCIC MA.5
MA.5 was a Phase III investigation of cyclophosphamide
75 mg/m2 orally days 1 through 14, epirubicin 60 mg/m2
intravenously days 1 and 8, and fluorouracil 500 mg/m2
intravenously days 1 and 8 (CEF) versus CMF (cyclo-
phosphamide 100 mg/m2 orally days 1 through 14,
methotrexate 40 mg/m2 intravenously days 1 and 8, and
fluorouracil 600 mg/m2 intravenously days 1 and 8)
(Fig. 1a) [8]. Between 1989 and 1993, 710 pre- and peri-
menopausal women with axillary node positive breast
cancer were accrued without knowledge or treatment by
HER2 status, stratified by type of surgery (total mastec-
tomy vs. partial mastectomy), locally determined estro-
gen (ER), or progesterone receptor (PR) (ER or PR C 10,
both \10, or unknown), and number of positive axillary
nodes (1–3, 4–10, [10) (Fig. 1a). Patients were not to
receive tamoxifen. The primary endpoint was relapse-free
survival (RFS) defined as time from randomization until
recurrence; contralateral breast cancer was not an event.
At median 10 years follow-up, the hazard ratio (HR) for
CMF versus CEF was 1.31, with stratified log-rank
p = 0.007. The 10-year follow-up data set, which has
been studied in other work [9–14], was used for these
investigations.
NCIC MA.12
NCIC CTG MA.12 was a placebo-controlled Phase III trial
of tamoxifen therapy following adjuvant chemotherapy
(CEF, CMF, or AC) in pre-menopausal women with early
breast cancer (Fig. 1b) [15]. Between 1993 and 2000, 672
women were randomized to MA.12, with 338 randomized
to tamoxifen and 334 to placebo, without knowledge or
treatment by HER2 status. Local centre determination of
levels of at least one hormone receptor [ER and/or PR], by
biochemical (positive C10 fmol/mg protein) or immuno-
histochemical assay was required, but patients with any
receptor status were eligible. The stratification factors were
type of chemotherapy (CMF, CEF, AC), hormone receptor
status (ER and/or PR positive, ER and PR negative), and
nodal status (0, 1–3, 4–9, 10?). The primary endpoint was
overall survival (OS). DFS was a secondary endpoint, and
was defined as being the time from randomization to the
earliest date of recurrence or death; censoring was the last
date the patient was known to be alive. At 9.7 years median
follow-up, multivariate analysis showed a DFS benefit for
tamoxifen of borderline significance (p = 0.056) and a
trend for improved OS (p = 0.12). There was no evidence
Breast Cancer Res Treat (2013) 142:143–151
Fig. 1 Trial schema for: a MA.5 and b MA.12. CEF cyclophospha-
mide 75 mg/m2 orally days 1 through 14, epirubicin 60 mg/m2
intravenously days 1 and 8, and fluorouracil 500 mg/m2 intravenously
days 1 and 8 every 28 days for six cycles; CMF (cyclophosphamide
100 mg/m2 orally days 1 through 14, methotrexate 40 mg/m2
intravenously days 1 and 8, and fluorouracil 600 mg/m2 intravenously
days 1 and 8) every 28 days for 6 cycles; AC doxorubicin 60 mg/m2
and cyclophosphamide 600 mg/m2 intravenously day 1 every
3 weeks for 4 cycles
of greater efficacy for tamoxifen in the hormone receptor
positive or ER receptor-positive subgroups than in hor-
mone receptor negative or ER receptor-negative patients:
interaction test p values were 0.71 and 0.14, respectively.
The 9.7-year follow-up, which has been studied in other
work [16, 17], was used in these investigations.
145
Primary objective
The primary end point for this study was the STEEP
endpoint of BCFI, defined as the time from randomization
until first invasive or ductal carcinoma in situ (DCIS) loco-
regional or distant recurrence, contralateral invasive breast
cancer or DCIS or death from breast cancer [18]. Patients
were censored at non-breast second primary invasive can-
cer, non-breast cancer death, or longest follow-up.
Investigational factors
Tumor size evaluations
Pathology reports from each participant in NCIC MA.5 and
MA.12 were used as source documentation for centrally
extracted (JFH, NB) tumor dimension information. All
enrolled trial participants underwent chart review. For each
focus of disease up to 3 dimensions were obtained (length,
width, and height).
Three different tumor size evaluation methods were
considered (Fig. 2):
• Largest tumor length, defined as the largest diameter of
the largest tumor focus present in the pathology sample.
• Tumor area, defined by considering the tumor to be
hypothetically an ellipse or circle and calculated using
the formula 1/4p 9 length 9 width.
• Tumor volume, defined by considering the tumor to be
hypothetically a cylinder and calculated its size by the
formula 1/6p 9 length 9 width 9 height.
In participants with MC/MF disease, in addition to the
above calculations that were made on the largest focus,
aggregate sizes were also calculated with the following
definitions (Fig. 2):
• Aggregate tumor length, defined as the summation of
the largest diameters of the tumor foci present in the
pathology sample, up to a maximum of three foci.
• Aggregate tumor area, defined as the summation of the
areas of the tumor foci present in the pathology sample,
up to a maximum of three foci.
• Aggregate tumor volume, defined as the summation of
the volumes of the tumor foci present in the pathology
sample, up to a maximum of three foci.
Pathologic reporting differences from participating
institutions led to a varied number of dimensions available
by chart report; as a result, not all participants in each trial
had data for area and volume, resulting in a smaller number
of tumor sizes for these calculations. If no MC/MF disease
was present, the one tumor focus was considered the
aggregate for the statistical analysis.
123
146
Fig. 2 Illustrates the method by
which each tumor size
measurement calculated.
a Largest tumor length.
b Largest tumor area. c Largest
tumor volume. d Aggregate
tumor length. e Aggregate
tumor area. f Aggregate tumor
volume
Other investigational factors
During chart review, additional local pathologic data were
extracted which was used for evaluation of grade for all
cases in the two trials: lymphovascular invasion (LVI;
present, absent), SBR grade (1, 2, 3), DCIS (present,
absent), number of involved lymph nodes, and total num-
ber of nodes examined. Baseline patient characteristics
were obtained from the locked NCIC CTG MA.5 and
MA.12 trial databases, using a common set of factor cat-
egorizations: age (\50, C50 years), hormone receptor
status (positive, negative), nodal status (positive, negative),
ECOG performance status (0, other), T status (T1/in situ
vs. other), and menopausal status (pre-menopausal/
unknown, other). Additionally, data from a retrospective
central pathology review of grade, hormone receptor status
and HER2 overexpression for both studies were already
available, and were included in the analysis for the path-
ologically reviewed cases [13].
Statistical analysis
MA.5 and MA.12 data were not pooled due to differences
in trial therapy and stratification factors. Patient and tumor
characterization for each trial compared results for all
patients enrolled to the trial with those for patients whose
tumor had central pathology review. MA.12 results for the
different tumor size measures were classified by lymph
node status (negative, positive) to permit comparison with
123
Breast Cancer Res Treat (2013) 142:143–151
node-negative patients in MA.5. To investigate the effects
of factors on BCFI, we performed univariate and multi-
variate stratified and unstratified Cox regression analysis,
reporting HRs, 95 % confidence intervals (CI), and the
Wald test statistic. A factor was considered significant in
these exploratory analyses with a two-sided p value B0.05
that was not adjusted for multiplicity of testing. Direct
comparisons of the alternative tumor size assessments
employed forced inclusion of alternatives.
Results
Study population
Median follow-up was 10 years for MA.5 and 9.7 years for
MA.12. Central pathology review was performed for 621
(87 %) of MA.5 patients and 480 (71 %) of MA.12
patients. Baseline patient and tumor characteristics are
presented by trial in Table 1. The two trials were similar
except for design, with MA.5 having 100 % lymph node
positive patients while MA.12 had 75 % node positivity, in
addition, fewer MA.5 than MA.12 patients had T1 or in situ
tumors (37 vs. 47 %), and fewer MA.5 patients were hor-
mone receptor positive (68 vs. 75 %).
The centrally reviewed patients were similar to those in
the full trial, except for there being fewer hormone receptor
positive tumors among the reviewed patients (58 vs. 68 %
in MA.5; 69 vs. 75 % in MA.12), and MA.5 centrally
Breast Cancer Res Treat (2013) 142:143–151 147

Table 1 Baseline patient characteristics for NCIC MA.5 and MA.12 Alternative measurements of tumor size
Factors NCIC MA.5 total NCIC MA.12 total
number (%) number (%) The alternative measurements of tumor size are compared
in Table 3. For node-positive patients, larger single
All cases Cases with All cases Cases with
central central dimension size was associated with worse BCFI:
pathology pathology p \ 0.0001, for all MA.5 cases; p = 0.01 for 75 % of
review review MA.12 cases. None of the other simultaneously examined
tumor size measures were significantly associated with
Total 710 (100) 621 (100) 672 (100) 480 (100)
BCFI, nor was there an association with BCFI for any
Age \50 593 (84) 514 (83) 553 (82) 401 (84)
measure in node negative MA.12 patients.
Hormone 482 (68) 360 (58) 505 (75) 330 (69)
receptor Larger aggregate size, the sum of largest single dimen-
positive sions, was associated with worse BCFI (p = 0.003) in
Lymph node 710 (100) 621 (100) 505 (75) 365 (76) MA.5 in the multivariate analysis, while larger single
positive dimension was associated with worse BCFI in MA.12
ECOG 572 (81) 504 (81) 440 (65) 322 (67) (p = 0.02) (Table 4). Area and volume assessments, or the
performance 0 presence of MF disease, were not associated with BCFI for
Menopausal 710 (100) 621 (100) 672 (100) 480 (100) either trial. A greater number of involved lymph nodes
status: pre/
uncertain were associated with worse BCFI in both trials
T1/in situ 265 (37) 234 (38) 314 (47) 219 (46)
(p \ 0.0001). Higher grade also led to worse BCFI in
Chemotherapy 710 (100) 621 (100) 672 (100) 480 (100)
MA.5 (p = 0.0005 for local grade; p = 0.005 for centrally
reviewed grade).
Anthracyclines 351 (49) 310 (50) 368 (55) 268 (56)
MF 104 (15) 89 (14) 98 (15) 63 (13)
DCIS 310 (44) 267 (43) 368 (55) 261 (54)
Discussion
LVI 243 (34) 217 (35) 219 (33) 147 (31)
HER2 positive N/A 83 (13) N/A 83 (17)
AJCC TNM staging is an internationally accepted system
Grade
used to standardize discussions with regards to prognosis
1 62 (9) 77 (12) 79 (12) 125 (26)
and treatment. It is based on a retrospective evaluation of a
2 224 (32) 204 (33) 254 (38) 203 (42)
large population of patients representing all classified
3 180 (27) 340 (55) 232 (35) 152 (32)
stages of disease [1]. The 7th edition of the TNM system
Unknown 244 (34) N/A 107 (16) N/A
represents the most updated staging system for breast
ECOG Eastern Cooperative Oncology Group, DCIS ductal carcinoma cancer and classifies the primary tumor size (T-stage) as
in situ, LVI lymphovascular invasion the largest tumor diameter of the largest tumor focus and,
other than the designation of ‘‘m’’, does not add any
additional consideration for the presence of MC/MF
reviewed patients having proportionately more grade 3 disease.
tumors (55 vs. 27 %). MA.5 had measurements in 2 and 3 There are theoretical reasons to consider alternative
dimensions for 59.2 and 39.2 %, respectively; while methods to evaluate tumor size. Primary tumors rarely have
MA.12 had measurements in 2 and 3 dimensions for 50.7 a uniform shape, possibly resulting in a measurement
and 41.5 %. In both studies, 15 % of the study population which may over or understate the volume of disease
had identified MC/MF disease. present. By considering the tumor as either a 2-dimensional
A number of factors could have influenced disease or a 3-dimensional structure, it may be possible to more
outcome; the results of the unstratified univariate analysis accurately state disease burden. In addition, summation of
evaluating each one for their effect on BCFI is shown in the tumors present when MC/MF disease is diagnosed may
Table 2. Higher grade (p \ 0.0001 for MA.5), T-stage of better address the burden of disease in this situation. The
T2 or higher (p \ 0.0001 for MA.5; p = 0.04 for MA.12), focus here was on simple determinations with easily
number of positive lymph nodes (p \ 0.0001 for MA.5 and available pathology reporting.
MA.12), MC/MF disease (p = 0.02 for MA.5), and LVI Utilizing data from the NCIC MA.5 and MA.12 clinical
(p = 0.005 for both MA.5 and MA.12) had significantly trials databases, we hypothesized that the alternative
worse BFCI. In MA.5, randomization to anthracycline- determinations of tumor size described in this paper would
based chemotherapy, CEF, led to better BCFI (p = 0.005), yield additional prognostic information for clinicians and,
while clinician-directed receipt of anthracycline therapy in therefore, could potentially influence treatment decisions.
MA.12 was associated with worse BCFI (p = 0.004). Although previous work in this area has utilized population

123
148 Breast Cancer Res Treat (2013) 142:143–151

Table 2 NCIC MA.5 and MA.12 univariate results for breast cancer free interval
Factors NCIC MA.5 HR (95 % CI) p valuea NCIC MA.12 HR (95 % CI) p valuea
All cases Cases with central All cases Cases with central
pathology review pathology review

Age C50 vs. \50 0.96 0.97 0.71 0.78

(0.73–1.26) (0.72–1.30) (0.49–1.03) (0.51–1.20)
0.79 0.85 0.07 0.26
Hormone receptor status (positive vs. negative) 1.08 0.87 0.96 0.89
(0.86–.37) (0.69–1.10) (0.72–1.29) (0.65–1.21)
0.50 0.24 0.79 0.45
ECOG performance [0 vs. 0 1.14 1.1 0.91 0.87
(0.88–1.47) (0.84–1.45) (0.69–1.19) (0.64–1.19)
0.33 0.48 0.48 0.40
Grade (1, 2, 3) 1.47 1.4 1.17 1.17
(1.20–1.80) (1.20–1.71) (0.96–1.43) (0.96–1.43)
\0.0001 \0.0001 0.12 0.11
Path CT2 vs. T1/in situ 1.75 1.8 1.32 1.17
(1.38–2.21) (1.40–2.33) (1.03–1.71) (0.87–1.57)
\0.0001 \0.001 0.04 0.30
Anthracyclines (Y/N) 0.74 0.72 1.47 1.24
(0.61–0.90) (0.57–0.91) (1.14–1.91) (0.92–1.66)
0.005 0.005 0.004 0.17
MF (Y/N) 1.38 1.41 0.92 0.86
(1.06–1.83) (1.05–1.89) (0.63–1.33) (0.55–1.35)
0.02 0.03 0.64 0.50
DCIS (Y/N) 1.04 1.02 1.12 1.08
(0.84–1.29) (0.81–1.29) (0.87–1.44) (0.81–1.45)
0.69 0.85 0.40 0.61
LVI (Y/N) 1.36 1.40 1.45 1.22
(1.10–1.69) (1.11–1.78) (1.12–1.87) (0.89–1.67)
0.005 0.005 0.005 0.21
HER2 (positive vs. negative) N/A 1.65 N/A 1.26
(1.22–2.24) (0.86–1.83)
0.001 0.24
No. of involved lymph nodes 1.11 1.11 1.15 1.15
(1.08–1.13) (1.09–1.14) (1.11–1.20) (1.11–1.20)
\0.0001 \0.0001 \0.0001 \0.0001
Total no. of nodes examined 0.996 0.995 1.002 1.01
(0.98–1.02) (0.98–1.02) (0.98–1.03) (0.98–1.04)
0.64 0.64 0.85 0.48
a
p value is Wald test statistic for unstratified test
HR hormone receptor, ECOG Eastern Cooperative Oncology Group, DCIS ductal carcinoma in situ, LVI lymphovascular invasion

databases to address the question of MC/MF disease [5–7];
as far as we are aware, this is the first evaluation that is
evaluating the above questions using large clinical trial
datasets. The advantage of performing this evaluation on
participants of MA.5 and MA.12 compared to population
databases is that the survival data has matured, participants
were randomized, and therapy was standardized.
Our evaluation found that alternative considerations of
tumor size (tumor area and volume) for patients with a
single focus of disease and similar calculations of aggre-
gate sizes when MC/MF disease is present, does not con-
sistently add additional prognostic information beyond
what is offered by 7th edition AJCC staging. One reason
why no additional prognostic information was obtained is

123
Breast Cancer Res Treat (2013) 142:143–151 149

Table 3 Assessment method of tumor size and effect on breast cancer-free interval
Tumor measurement NCIC MA.5 All patients node positive NCIC MA.12 Node-negative patients NCIC MA.12 Node-positive patients
method HR (95 % CI) p valuea HR (95 % CI) p valuea HR (95 % CI) p valuea
All continuous All cases Cases with central All cases Cases with central All cases Cases with central
variables (n = 710) pathology review (n = 167) pathology review (n = 505) pathology review
(n = 621) (n = 115) (n = 365)

Largest size 1.18 1.17 1.11 0.99 1.14 1.14

(1.13–1.22) (1.13–1.22) (0.80–1.56) (0.70–1.41) (1.03–1.26) (1.01–1.28)
\0.0001 \0.0001 0.53 0.97 0.01 0.04
Aggregate size 1.08 1.08 1.11 0.92 0.97 0.89
(0.96–1.22) (0.95–1.21) (0.80–1.56) (0.64–1.30) (0.75–1.25) (0.64–1.25)
0.15 0.21 0.52 0.62 0.84 0.51
Largest Area 1.00 1.00 1.05 1.00 0.98 0.95
(0.99–1.01) (0.99–1.01) (0.97–1.14) (0.91–1.11) (0.94–1.02) (0.90–1.01)
0.45 0.48 0.23 0.97 0.33 0.10
Aggregate Area 1.00 1.00 1.05 0.99 0.98 0.95
(0.99–1.01) (0.99–1.01) (0.97–1.14) (0.90–1.10) (0.94–1.02) (0.90–1.01)
0.57 0.60 0.24 0.89 0.36 0.11
Largest volume 1.00 1.00 1.02 1.00 1.00 0.99
(0.997–1.001) (0.997–1.001) (0.99–1.04) (0.96–1.04) (0.99–1.01) (0.98–1.004)
0.34 0.34 0.19 0.88 0.51 0.24
Aggregate volume 1.00 1.00 1.02 1.00 1.00 0.99
(0.997–1.001) (0.997–1.001) (0.99–1.05) (0.96–1.03) (0.99–1.01) (0.98–1.004)
0.36 0.36 0.19 0.81 0.52 0.24
a
p value is Wald test statistic for unstratified test

because staging based on size does not take into account
the biologic characteristics of the tumor. The importance of
tumor biology in prognosis and influencing treatment
decisions is demonstrated by the 2011 St. Gallen Interna-
tional Expert Consensus on primary therapy for early
breast cancer. In contrast to the 2009 recommendations
where treatment decisions were significantly influenced by
tumor size, the latest treatment guidelines are based on
tumor biology, with Ki67, hormone receptor and HER2
status forming the basis for recommended therapies [19,
20]. In addition, the 21-gene Recurrence Score assay
(Oncotype Dx assay) provides further evidence that tumor
biology contributes significantly to prognosis [21]. Even
when taking ER and HER2 status into account, our alter-
native methods for evaluating tumor burden did not add
any additional prognostic information. This adds further
support to our conclusion that the biology of the tumor is
the important prognostic variable for physicians to
consider.
Previous work has suggested that using alternative
methods for tumor size evaluation may be prognostic,
especially for patients with MC/MF disease; however, the
populations evaluated in these studies received much lower
rates of chemotherapy compared to the MA.5 and MA.12
study population. For example, Weissenbacher and
colleagues reported that MC/MF disease was a poor
prognostic factor; however, 48 % of the patients with MC/
MF disease did not receive chemotherapy as the standard
of care in their study was to administer adjuvant chemo-
therapy in node positive disease only [7]. In contrast, in
MA.5 and MA.12, all participants received chemotherapy,
although in MA.5, where 68 % of the patients were hor-
mone receptor positive, adjuvant endocrine therapy was not
administered. Interestingly, the summation of the largest
tumor dimension was correlated with a worse BFCI in
MA.5 but not MA.12, suggesting that the prognostic ben-
efit from greater precision in tumor size measurements may
be more pronounced in situations where patients have
received what could be perceived as inadequate systemic
therapy.
There are some limitations to our analysis. First, our
work was retrospective, although these issues are some-
what addressed by the data being extracted from two ran-
domized clinical trial populations. The reporting of second
and third dimensions was center dependent, and could have
down-sized the estimates of area and volume which would
have the greatest impact in multi-focal disease which was
indicated here for only 15 % of patients, on the low side of
the range seen in other series [2–4]. Second, the maximal 3
measurements were obtained with visual microscopy which

123
150 Breast Cancer Res Treat (2013) 142:143–151

Table 4 Multivariate effects of factors on breast cancer free interval

Factors NCIC MA.5 NCIC MA.12
HR HR
(95 % CI) (95 % CI)
p valuea p valuea
All cases Cases with central All cases Cases with central
pathology review pathology review

Age C50 vs. \50 1.09 1.19 1.10 0.97

(0.78–1.51) (0.86–1.65) (0.82–1.48) (0.67–1.40)
0.64 0.29 0.54 0.85
Grade (1, 2, 3) 1. 51 1.35 1.23 1.18
(1.19–1.91) (1.10–1.66) (0.98–1.56) (0.91–1.51)
0.0005 0.005 0.09 0.21
No. of involved lymph nodes 1.16 1.14 1.10 1.15
(1.12–1.21) (1.10–1.18) (1.05–1.16) (1.08–1.21)
\0.0001 \0.0001 \0.0001 \0.0001
Total No. of nodes examined 0.96 0.95 0.98 0.99
(0.93–0.99) (0.93–0.98) (0.95–1.01) (0.96–1.02)
0.004 0.0002 0.23 0.51
LVI (Y/N) 1.10 1.11 1.08 0.88
(0.82–1.47) (0.85–1.46) (0.80–1.44) (0.61–1.28)
0.55 0.44 0.63 0.50
MF (Y/N) 1.09 1.35 0.78 0.85
(0.75–1.59) (0.85–2.14) (0.53–1.15) (0.45–1.61)
0.66 0.20 0.20 0.62
HER2 (positive vs. negative) N/A 1.47 N/A 1.25
(1.02–2.11) (0.79–2.00)
0.04 0.34
Largest size (continuous) 1.02 1.14 1.14 2.78
(0.89–1.17) (0.72–1.82) (1.01–1.28) (0.87–8.87)
0.73 0.57 0.02 0.08
Aggregate size (continuous) 1.09 0.94 0.95 0.49
(1.03–1.16) (0.59–1.46) (0.74–1.23) (0.16–1.48)
0.003 0.77 0.68 0.20
Largest area (continuous) 1.001 1.11 0.99 0.43
(0.99–1.01) (0.94–1.30) (0.95–1.03) (0.08–2.43)
0.88 0.22 0.73 0.34
Aggregate area (continuous) 1.00 0.92 0.99 2.22
(0.99–1.01) (0.78–1.08) (0.96–1.03) (0.39–12.54)
0.97 0.29 0.78 0.37
Largest volume (continuous) 1.00 0.93 1.00 1.36
(0.999–1.001) (0.84–1.04) (0.99–1.01) (0.41–4.50)
0.90 0.20 0.85 0.61
Aggregate volume (continuous) 1.00 1.07 1.00 0.74
(0.999–1.001) (0.96–1.20) (0.99–1.01) (0.22–2.43)
0.93 0.21 0.86 0.61
a
p value is Wald test statistic for unstratified test
HR hormone receptor, ECOG Eastern Cooperative Oncology Group, DCIS ductal carcinoma in situ, LVI lymphovascular invasion

123
Breast Cancer Res Treat (2013) 142:143–151
would not represent the granularity or heterogeneity of true
tumor burden. Higher levels of virtual in situ slicing of
tumors would provide more accurate true volumetric esti-
mates, and such methods are currently under development
in multiple disease types [22, 23]. Finally, we were unable
to stratify important tumor biology factors, such as Ki67
status, into the multivariate analysis. It is possible that our
conclusions might be different if we were able to do so.
In conclusion, in the MA.5 multivariate analysis, larger
summation of the largest tumor dimensions was associated
with worse BCFI (p = 0.0003), while larger single dimen-
sion was associated with worse BCFI (p = 0.02) for MA.12.
It does not appear that utilization of other alternative
methods of tumor size, such as calculated area or volume or
summation of MC/MF disease in tumor size evaluations
adds additional prognostic information for physicians
treating early breast cancer patients. When faced with the
situation of MC/MF disease, physicians could consider the
largest diameter of the largest focus of disease or the sum of
the largest diameters of all foci in their T-stage determina-
tion, but it appears that the current method of T-staging as
outlined in the 7th edition of the AJCC guidelines for cancer
staging offers equivalent determinations of prognosis.
Acknowledgments Dr. Hilton’s Fellowship is sponsored by Astra-
Zeneca and the University of Ottawa. Dr. Dong’s post-doctoral Fellow-
ship was sponsored by Astra-Zeneca. Dr. Bouganim’s Research Fel-
lowship was sponsored by The Canadian Breast Cancer Foundation—
Ontario Chapter. The NCIC CTG is supported by the Canadian Cancer
Society through a grant from the Canadian Cancer Research Institute
Conflict of interest The authors would like to report no relevant
conflicts of interest for this manuscript.
References
1. (2010) Part VII—Breast. In: Edge SB, Byrd DR, Compton CC,
Fritz AG, Greene FL, Trotti A, (eds) American Joint Committee
on Cancer: Cancer Staging Manual, 7th edn. Springer, New York,
p 347–377
2. Holland R, Veling SH, Mravunac M, Hendriks JH (1985) Histo-
logic multifocality of tis, T1-2 breast carcinomas. Implications for
clinical trials of breast-conserving surgery. Cancer 56:979–990
3. Fish EB, Chapman JA, Link MA (1998) Assessment of tumor size
for multifocal primary breast cancer. Ann Surg Oncol 5:442–446
4. Joergensen LE, Gunnarsdottir KA, Lanng C, Moeller S, Ras-
mussen BB (2008) Multifocality as a prognostic factor in breast
cancer patients registered in Danish Breast Cancer Cooperative
Group (DBCG) 1996–2001. Breast 17:587–591
5. Tot T, Gere M, Pekár G et al (2011) Breast cancer multifocality,
disease extent, and survival. Hum Pathol 42:1761–1769
6. Yerushalmi R, Kennecke H, Woods R, Olivotto IA, Speers C, Gel-
mon KA (2009) Does multicentric/multifocal breast cancer differ
from unifocal breast cancer? An analysis of survival and contralateral
breast cancer incidence. Breast Cancer Res Treat 117:365–370
7. Weissenbacher TM, Zschage M, Janni W et al (2010) Multi-
centric and multifocal versus unifocal breast cancer: is the
151
tumor-node-metastasis classification justified? Breast Cancer Res
Treat 122:27–34
8. Levine MN, Pritchard KI, Bramwell VH et al (2005) Randomized
trial comparing cyclophosphamide, epirubicin, and fluorouracil
with cyclophosphamide, methotrexate, and fluorouracil in pre-
menopausal women with node-positive breast cancer: update of
National Cancer Institute of Canada Clinical Trials Group Trial
MA5. J Clin Oncol 23:5166–5170
9. O’Malley FP, Chia S, Tu D et al (2009) Topoisomerase II alpha
and responsiveness of breast cancer to adjuvant chemotherapy.
J Natl Cancer Inst 101:644–650
10. O’Malley FP, Chia S, Tu D et al (2011) Topoisomerase II alpha
protein and responsiveness of breast cancer to adjuvant chemo-
therapy with CEF compared to CMF in the NCIC CTG random-
ized MA.5 adjuvant trial. Breast Cancer Res Treat 128:401–409
11. Pritchard KI, Munro A, O’Malley FP et al (2012) Chromosome
17 centromere (CEP17) duplication as a predictor of anthracy-
cline response: evidence from the NCIC Clinical Trials Group
(NCIC CTG) MA.5 Trial. Breast Cancer Res Treat 131:541–551
12. Hertel PB, Tu D, Ejlertsen B et al (2012) TIMP-1 in combination
with HER2 and TOP2A for prediction of benefit from adjuvant
anthracyclines in high-risk breast cancer patients. Breast Cancer
Res Treat 132:225–234
13. Pritchard KI, Shepherd LE, O’Malley FP et al (2006) HER2 and
responsiveness of breast cancer to adjuvant chemotherapy.
N Engl J Med 354:2103–2111
14. Cheang MC, Voduc KD, Tu D et al (2012) Responsiveness of
intrinsic subtypes to adjuvant anthracycline substitution in the
NCIC.CTG MA.5 randomized trial. Clin Cancer Res 18:2402–2412
15. Bramwell VH, Pritchard KI, Tu D et al (2010) A randomized
placebo-controlled study of tamoxifen after adjuvant chemo-
therapy in premenopausal women with early breast cancer
(National Cancer Institute of Canada–Clinical Trials Group Trial,
MA.12). Ann Oncol 21:283–290
16. Chia SK, Bramwell VH, Tu D et al (2012) A 50-gene intrinsic
subtype classifier for prognosis and prediction of benefit from
adjuvant tamoxifen. Clin Cancer Res 18:4465–4472
17. Yan Y, Li X, Blanchard A, Bramwell VH et al (2013) Expression
of both estrogen receptor-beta 1 (ER-b1) and its co-regulator
steroid receptor RNA activator protein (SRAP) are predictive for
benefit from tamoxifen therapy in patients with estrogen receptor-
alpha (ER-a)-negative early breast cancer (EBC). Ann Oncol
24:1986–1993
18. Hudis CA, Barlow WE, Costantino JP et al (2007) Proposal for
standardized definitions for efficacy end points in adjuvant breast
cancer trials: the STEEP system. J Clin Oncol 25:2127–2132
19. Goldhirsch A, Ingle JN, Gelber RD et al (2009) Thresholds for
therapies: highlights of the St. Gallen International Expert Con-
sensus on the primary therapy of early breast cancer 2009. Ann
Oncol 20:1319–1329
20. Goldhirsch A, Wood WC, Coates AS et al (2011) Strategies for
subtypes–dealing with the diversity of breast cancer: highlights of
the St. Gallen International Expert Consensus on the Primary
Therapy of Early Breast Cancer 2011. Ann Oncol 22:1736–1747
21. Paik S, Shak S, Tang G et al (2004) A multigene assay to predict
recurrence of tamoxifen-treated, node-negative breast cancer.
N Engl J Med 351:2817–2826
22. Gutman DA, Cooper LA, Hwang SN et al (2013) MR imaging
predictors of molecular profile and survival: multi-institutional
study of the TCGA glioblastoma data set. Radiology 267:560–569
23. Clarke GM, Eidt S, Sun L, Mawdsley G, Zubovits JT, Yaffe MJ
(2007) Whole-specimen histopathology: a method to produce
whole-mount breast serial sections for 3-D digital histopathology
imaging. Histopathology 50:232–242
123