Breast Cancer Res Treat (2009) 115:423–428

DOI 10.1007/s10549-008-0086-2

EPIDEMIOLOGY

Pattern of metastatic spread in triple-negative breast cancer

Rebecca Dent Æ Wedad M. Hanna Æ
Maureen Trudeau Æ Ellen Rawlinson Æ
Ping Sun Æ Steven A. Narod

Received: 29 May 2008 / Accepted: 30 May 2008 / Published online: 10 June 2008
Ó Springer Science+Business Media, LLC. 2008

Abstract Purpose The prognosis of women with triple- times more likely to experience a visceral metastasis within
negative breast cancers (defined as cancers that are estro- five years of diagnosis than those with other types of cancer
gen receptor-negative, progesterone receptor-negative and (HR 4.0; 95% CI: 2.7–5.9; P \ 0.0001). The rates of bone
HER2/neu negative) is poor, compared to women with metastases were comparable for triple-negative and for
other subtypes of breast cancer. It is proposed that the other forms of cancer in this time period (HR 0.8; 95% CI:
underlying difference in recurrence rates may be explained 0.4–1.6 P = 0.5). Conclusions The excess risk of distant
in part by different routes of metastatic spread. Experi- recurrence in triple-negative breast cancers, versus other
mental design We studied a cohort of 1608 patients forms of cancer, is attributable in large part to an excess of
diagnosed with breast cancer, diagnosed between January visceral metastases in the first five years following
1987 and December 1997 at Women’s College Hospital in diagnosis.
Toronto. Triple-negative breast cancers were defined as
those that were estrogen receptor-negative, progesterone Keywords Triple-negative
Basal-like
Breast cancer

receptor-negative and HER2/neu-negative. We compared Metastases
the incidence rates of metastatic spread to bone and to
other (non-bone) organs in women with triple-negative and Abbreviations
other forms of breast cancer. Results Of the 1,608 patients, HR Hazard ratio
180 (11.2%) had triple-negative breast cancer. The 1608 ER Estrogen receptor
women were followed for a median of 9.0 years (range PR Progesterone receptor
0.1–19 years). Compared to other patients, those with tri- HBBC Henrietta Banting database
ple-negative breast cancer had an increased likelihood of CI Confidence interval
distant recurrence over the study period (adjusted hazard
ratio (HR) 1.9; 95% CI: 1.5–2.5, P \ 0.0001). The rela-
tively poor prognosis was apparent in the five years after
diagnosis (HR 2.9; 95% CI: 2.1–3.9; P = 0.0001) but not Introduction
thereafter (HR 0.5; 95% CI: 0.2–1.1; P = 0.07). In par-
ticular, women with triple-negative breast cancer were four The use of modern genomic and immunohistochemical
techniques has enhanced our understanding of breast cancer
biology and has improved our ability to classify breast
R. Dent
W. M. Hanna
M. Trudeau cancers into distinct subsets, including hormone receptor
Department of Medical Oncology, Sunnybrook Health Sciences positive types (luminal A and luminal B); human epidermal
Center, University of Toronto, Toronto, ON, Canada
growth receptor 2 positive (HER2–positive); normal-like;
E. Rawlinson
P. Sun
S. A. Narod (&) and basal-like [1]. This final group is predominantly com-
Women’s College Research Institute, Women’s College posed of ‘‘triple-negative’’ breast cancers, because these
Hospital, University of Toronto, 790 Bay Street, cancers fail to stain for estrogen receptor (ER), progester-
7th FloorM5G 1N8 Toronto, ON, Canada
e-mail: [Link]@[Link]
one receptor (PR) or HER2. Long-term follow-up studies of

123
424 Breast Cancer Res Treat (2009) 115:423–428

basal or triple-negative cohorts have demonstrated an identified a predilection for lung and brain metastases
inferior prognosis for these breast cancer patients, com- among patients with high-grade myoepithelial cancers (a
pared to other groups of women [2–4]. Given the relatively closely related group) [7]; this observation was subse-
poor outcomes for triple-negative patients, and the lack of a quently confirmed [8, 9]. Fulford and colleagues reported a
specific treatment, there is interest in elucidating the natural comparatively low proportion of bone and liver metastases
history of triple-negative disease, with the hope of devel- in women with basal-like breast cancers, compared to other
oping an effective targeted treatment (Table 1). types [10]. In general, these studies compare the relative
Breast carcinomas with particular patterns of differen- proportions of breast cancer patients who develop metas-
tiation or gene expression have characteristic sites to which tases at various sites, but do not directly compare
they metastasize [5]. For example, the bone is a common underlying recurrence rates. The aim of the current study
site of first metastasis in women with hormone receptor- was to evaluate the patterns of metastatic spread in patients
positive breast carcinomas [6]. Tsuda and colleagues with triple-negative breast cancers, and to compare these to
subgroups of patients with other types of breast cancer.
Table 1 Comparison of patients with triple-negative and other This information might help us understand mechanisms by
cancers which triple-negative breast cancer status influences
Variables (%) Other Triple-negative P-value prognosis.
n = 1428 n = 180

Year of birth, 1935.5 1939.7 \10-4

mean (range) (1899–1970) (1902–1969)
Materials and methods
-4
Age of diagnosis, 57.7 (24–93) 53.0 (22–92) \10
mean (range) Study patients
Years of follow-up, 9.2 (0.1–20) 7.9 (0.3–18.9) 0.0003
mean (range) We studied a cohort of women with invasive breast cancer
Grade who were first treated at the Henrietta Banting Breast
1 238 (19.9) 15 (9.8) Centre (HBBC) between 1987 and 1997. Details of the
2 616 (51.6) 37 (24.2) database have been published elsewhere [2, 11]. The HBBC
3 340 (28.5) 101 (66.0) \10-4 database is a hospital-based cohort of women diagnosed
Missing 224 27 with primary breast cancer at Women’s College Hospital
Nodal status (WCH). The HBBC database was established in order to
Negative 691 (56.9) 76 (45.5) systematically record data on clinical presentation, treat-
Positive 524 (43.1) 91 (54.5) 0.006 ment and outcome from women with breast cancer who
Not tested/missing 213 13 received primary surgical treatment at WCH. Patients were
Size (mm) referred from family practitioners and other physicians
B 10 318 (22.5) 11 (6.2) from the surrounding medical community to one of five
11–20 573 (40.5) 54 (30.3) teaching surgeons practicing at HBBC. Women were
21–50 461 (32.6) 99 (55.6) included in the study if they were diagnosed with invasive
[ 50 64 (4.5) 14 (7.9) \10-4
breast cancer at Women’s College Hospital from January
Missing 12 2
1987 to December 1997 and were entered into the HBBC
Size (mean) 21.5 (0–130) 29.5 (5–100) \10-4
database. Medical records and pathology reports were
reviewed. Patient information was recorded at accrual, and
Tamoxifen treatment n (%)
included details about the patient’s age at diagnosis (in
Yes 764 (54.5) 64 (36.2)
years), grade, lymph node status and pathologic tumor size
No 639 (45.6) 113 (63.8) \10-4
(mm), treatment (surgery, chemotherapy, tamoxifen ther-
Missing 24 3/0
apy and radiation treatment). For this study, ‘‘triple-
Chemotherapy
negative’’ breast cancers were defined as those that were
No 1048 (74.4) 91 (51.4)
estrogen receptor (ER) negative, progesterone receptor (PR)
Yes 360 (25.6) 86 (48.6) \10-4
negative and HER2 negative. ‘Other’ breast cancers were
Missing 20 3
defined as those that were positive for any of these markers.
Distant recurrence
No 1133 (79.3) 119 (66.1)
Immunohistochemistry
Yes 295 (20.7) 61 (33.9) \10-4
Years to recurrence, 5.0 (0.1–17.1) 2.6 (0.3–7.9) \10-4 For each patient in the database, a set of representative
mean (range)
paraffin-embedded slides was requested for antibody

123
Breast Cancer Res Treat (2009) 115:423–428 425

staining in the reference laboratory. Staining was per- time. Mixed recurrences were classified as synchronous if
formed between 2000 and 2004. Estrogen-receptor (ER) there was evidence for both bone and visceral metastases at
and progesterone-receptor (PR) status were determined time of the initial recurrence, or non-synchronous if the
using immunohistochemistry. For ER and PR, antibodies evidence for metastasis at one site preceded the other.
were from Novocastra, with cut-off levels for receptor Relapse-free survival was defined as the time of diagnosis
positivity of greater than 10%. For 25% of the patients a to the time of the development of first evidence of clinical
paraffin-embedded slide was not available and the ER and or radiographic metastatic disease. Survival was ascer-
PR status was determined by review of the pathology tained from the HBBC database, which is maintained
records. These early assays were biochemical, employing a regularly with each patient followed up at least once per
cut-point of 10fmol/mg protein. HER2 status did not year (Table 2).
determine patient treatment, but was done for research
purposes. Over-expression of the HER2 protein was eval- Analysis
uated using the CB11 monoclonal antibody (Novocastra) in
representative paraffin sections of each tumor, using the Baseline demographic and tumor characteristics were
peroxidase-antiperoxidase technique for IHC assay. HER2- compared between the ‘triple-negative’ and ‘other’ groups,
positivity was defined as strong complete membrane using a t-test for means and chi-square statistic for fre-
staining in at least 10% of tumor cells. 1608 (80%) of 1992 quencies. We evaluated the differences in first site of
patients had sufficient details on hormone receptors and metastatic spread between the ‘‘triple-negative’’ and ‘other’
HER2 for classification and were eligible for the study. group using the chi-square statistic.
Kaplan-Meier survival analyses were carried out for
Follow-up distant recurrence-free survival. The log-rank test was used
to examine the statistical significance of the differences
Follow-up has been maintained by the database coordinator observed between the groups. A multivariate Cox regres-
through review of clinical charts and by contacting patients sion model was also employed. This was used to compute
by telephone. Local-regional relapses and the subsequent hazard ratios and 95% confidence intervals, adjusting for
surgery during the ninety-day post-surgery period were known prognostic variables (age, grade, tumor size, nodal
considered to be part of the primary management: distant status, use of adjuvant tamoxifen therapy and use of
recurrence during this time-period disqualified the patient adjuvant chemotherapy). Estimates were considered sta-
from study. Relapses after ninety days were considered as tistically significant for two-tailed values of P \ 0.05. All
events. Relapses were dated and reviewed by two medical analyses were carried out with the SAS 9.1.3 statistical
oncologists; initial inter-observer agreement was greater program. Hazard ratios were calculated for the entire fol-
than 95%. 85% of the patients had been followed for three low-up period and for periods from zero to five years after
or more years, and 75% of patients were under active diagnosis and from five years to the end of the follow-up
follow-up (or have died). For deceased patients, date and period. Patients were divided into triple-negative and
cause of death was obtained from the medical records. ‘other’ subcohorts. To calculate the rate of bone recur-
rence, patients were followed from the time of diagnosis
Outcomes until either (1) the date of first evidence of bone recurrence;
(2) date of last follow up or (3) death. To calculate the rate
All distant recurrences were classified as to site of recur- of non-bone (visceral) recurrence, patients were followed
rence. A distant recurrence was classified as a ‘‘bone’’ from the time of diagnosis until either (1) the date of first
recurrence if there was evidence of metastases to bone evidence of visceral recurrence; (2) the date of last follow
only; ‘‘non-bone’’ recurrence if there was radiological up or (3) the date of death. Patients with mixed recurrences
evidence of metastases to viscera (including lung and liver) contributed an event to the first recurrence (if they were
or brain; and ‘‘mixed’’ if patients with metastatic disease asynchronous) or to neither endpoint (if they were syn-
had evidence for both bone and visceral metastases at some chronous) (Table 3).

Table 2 Site of first distant

Site of first distant Other (n = 294) Triple-negative (n = 61) P-value*
recurrence: other versus triple-
recurrence
negative breast cancer Number (%) Number (%)

*P-values were calculated with Bone only 116 39.5 10 16.4 0.0006
the use of the v2 test. One Viscera only 122 41.5 39 63.9 0.001
patient was missing data on site Bone and viscera 56 19.1 12 19.7 0.9
of recurrence

123
426 Breast Cancer Res Treat (2009) 115:423–428

Table 3 Hazard ratios in the

End point Unadjusted Adjusted
triple-negative group as
compared with the ‘‘other’’ Hazard ratio P-value* Hazard ratio P-value**
group (95% CI) (95% CI)

Risk of distant recurrence

Entire follow up period 1.9 (1.5–2.5) \0.0001 1.1 (0.9–1.5) 0.35
0–5 years follow-up 2.9 (2.1–3.9 \0.0001 1.7 (1.2–2.3) 0.02
5 years to end of follow-up 0.5 (0.2–1.1) 0.07 0.3 (0.1–0.7) 0.003
*P values were determined by
the log-rank test Risk of visceral recurrence
** A Cox model including age Entire follow up period 2.9 (2.0–4.2) \0.0001 1.7 (0.9–1.5) 0.007
at diagnosis, grade (1, 2, 3), 0–5 years follow-up 4.0 (2.7–5.9) \0.0001 2.3 (1.5–3.5) \0.0001
nodal status (negative, positive), 5 years to end of follow-up 0.5 (0.1–2.1) 0.3 0.3 (0.1–1.2) 0.09
tumor size (T1,T2, T3),
Risk of bone recurrence
chemotherapy (yes, no) and
tamoxifen therapy (yes, no) was Entire follow up period 0.8 (0.4–1.6) 0.5 0.5 (0.3–1.0) 0.06
used to estimate the adjusted 0–5 years follow-up 1.4 (0.7–2.8) 0.4 0.9 (0.4–1.8) 0.7
hazard ratios. CI denotes 5 years to end of follow-up 0.2 (0.1–1.3) 0.08 0.1 (0.1–0.8) 0.03
confidence interval

Results

One hundred eighty of 1608 patients (11.2%) were defined

as having ‘triple-negative’ breast cancers. Patients with
‘‘triple-negative’’ breast cancers were more likely to have
died than patients with ‘other’ cancers (43% vs. 30%,
respectively; P = 0.006). The median time to death was
3.5 years for patients with ‘‘triple-negative’’ breast cancer,
compared to 5.7 years for patients with other cancers
(P \ 0.0001). The mean time to distant recurrence in
patients with ‘‘triple-negative’’ cancer was 2.6 years,
compared to 5.0 years for women with other cancers. A
greater proportion of women with triple-negative breast Fig. 1 Rate of distant recurrence in viscera or brain after the
cancer presented with visceral metastases at the time of diagnosis of breast cancer. Basal versus other
first metastatic spread (84%) than did women with other
types of cancers (61%; P = 0.0003 for difference).
We calculated the actuarial rates of developing visceral
and non-visceral metastases for women with triple-negative
and other forms of cancer. The probability of developing a
visceral metastasis as the first site of distant recurrence,
within ten years of diagnosis, was significantly higher among
women with triple-negative breast cancers (23%) than for
women with other forms of breast cancer (9%: HR = 2.9;
95% CI 2.0–4.2; P \ 0.0001) (see Fig. 1). The hazard ratio
for the development of a visceral metastasis in triple-nega-
tive, versus other, breast cancers was particularly marked in
the first five years following diagnosis (HR = 4.0; 95% CI:
2.7–5.9; P \ 0.0001). After five years of diagnosis, the
actuarial rate of developing a visceral metastasis was not
Fig. 2 Rate of distance recurrence to bone after the diagnosis of
significantly different between the two groups (HR = 0.5 for
breast cancer. Basal versus other
triple-negative versus other; 95% CI 0.1–2.1; P \ 0.3).
In contrast, there was no difference in the rate of years of diagnosis was similar for women with triple-
developing bone metastases between the two groups. The negative cancer (7%) and other forms of cancer (9%;
probability of developing a bone metastasis within ten P = 0.9 for difference) (Fig. 2). The hazard rate for

123
Breast Cancer Res Treat (2009) 115:423–428 427

developing bone metastases was 0.8 (95% CI 0.4–1.6; phenotypes) and brain (18% vs. 2%). Hicks and colleagues
P \ 0.5) over the entire follow-up period. found the basal phenotpye to be over-represented among
Among the 295 women with triple-negative breast women who presented with brain metastases after a breast
cancer, the median duration from first distal recurrence cancer diagnosis [8].
until death was 2.4 months. The time from recurrence to We observed a marginal reduction in the rate of bone
death was similar if the first recurrence was visceral metastases in women with triple-negative breast cancers
(17 months) or was in bone (26 months; P = 0.33 for after five years from diagnosis. Rodriguez-Pinila et al. also
difference). reported a low frequency of bone metastases in women
with basal-like breast cancers [20]. It has been suggested
that triple-negative cancers tend to disseminate via hema-
Discussion
togenous spread, rather than through the lymphatics.
However, lymph node metastases are not uncommon in
The triple-negative phenotype is easily defined by the use
triple-negative cancers. In our earlier study, we reported
of three commonly used immunhistochemical markers. We
that 40% of 180 women with triple-negative breast cancers
have shown that a high risk of developing visceral metas-
had positive lymph nodes at diagnosis, compared to 32% of
tases shortly after diagnosis is responsible in large part for
women with other types of cancer. However, in the triple-
the adverse prognosis associated with the phenotype.
negative group, the prevalence of positive lymph nodes did
‘‘Triple-negative’’ is often used as a surrogate for the
not correlate strongly with tumour size.
‘‘basal-like’’ group, which is defined either by immuno-
Molecular studies of triple-negative or basal breast
histochemistry or by gene expression profiling, but the two
cancers have identified proteins in addition to ER, PR and
subgroups are not synonymous [12, 13]. If the basal-like
HER2 which are characteristic of the phenotype. It is
category is defined by expression profiling, then a panel
possible that the expression of some of these proteins (or
composed of ER, HER2, and cytokeratin 5/6, and epider-
other proteins) underlie the tendency towards visceral
mal growth factor receptor (EGFR) can identify these
metastases, or perhaps make the cells less likely to adhere
tumors with a high degree of accuracy [14]. Triple-negative
to bone. Examples of proteins that are often expressed in
breast cancers are over-represented in African women [15–
triple-negative cancers include cytokeratins 5/6, cytokera-
17] and in those with BRCA1 mutations [18, 19]. The high
tin 14, cytokeratin 17, vimentin, laminin, fascin and
proportion of basal-type breast cancers among women of
epidermal growth factor receptor. Triple-negative cancers
African descent explains to some extent the relatively
also exhibit features of rapidly-dividing cancers, including
young age of onset and the relatively poor outcome in this
a high histologic grade, a high mitotic index and the
group of patients. It has been proposed that BRCA1
expression of proteins which are associated with cellular
function may be impaired in basal tumours which lack
proliferation [1, 21–24]. Also critical in the establishment
identifiable inherited BRCA1 mutations [12].
of metastases is the development of a local blood supply.
Our data are descriptive and do not suggest a mechanism
We followed the majority of the women in this study for
to explain the early tendency for visceral spread. It is likely
ten years or more, and this allowed us to look at the effects
that the factors which lead to early distal recurrence are
of prognostic variables at various time periods over the
those which also lead to rapid progression. In a previous
course of the disease. It should be noted that the adverse
study of this cohort, we reported that all distal recurrences
effect of the triple-negative status is transient; i.e. it is
in the triple-negative subgroup occurred within eight years
predictive of distal recurrence only during the five years
of diagnosis and that all deaths due to breast cancer in
after diagnosis. The time-dependence of an effect will be
patients with ‘‘triple-negative’’ cancer occurred within
overlooked in a study if the majority of patients have only a
10 years of diagnosis. In contrast, deaths from breast
relatively short period of follow-up, or if the proportional
cancer in patients with ER-positive breast cancer continue
hazards model is applied (this model assumes that the
to accrue up to 18 years after diagnosis. Furthermore, the
impact of a particular variable is constant throughout the
time from distant recurrence to death is much shorter in
course of the disease). Future studies of the natural history
triple-negative tumours than in other tumours; this is con-
of breast cancer should be attentive to the possibility of
sistent with the relatively high mortality associated with
different effects of prognostic factors at different times, and
visceral (versus bone) metastases [2, 5].
that particular factors may be relevant only among certain
Other authors have reported an increased proportion of
histologically-defined subgroups.
basal-like or triple-negative cancers among women who
develop visceral metastases. Luck et al. reported that the Acknowledgments We thank the Canadian Breast Cancer Foun-
main sites of metastases for patients with basal breast dation (Ontario Chapter) for support of the Henrietta Banting Data
cancer were the lung (52% in basal vs. 23% in other Base and Tumour Bank and for their contribution to this research.

123
428 Breast Cancer Res Treat (2009) 115:423–428

References 12. Reis-Filho JS, Tutt ANJ (2008) Triple-negative tumur, a critical
review. Histopathology 52:101–118. doi:10.1111/j.1365-2559.
1. Perou CM, Sorlie T, Eisen MB et al (2000) Molecular portraits of 2008.03046.x
human breast tumours. Nature 406:747–752. doi:10.1038/ 13. Cheang MC, Voduc D, Bajdik C et al (2008) Basal-like breast
35021093 cancer defined by five biomarkers has superior prognostic value
2. Dent R, Trudeau M, Pritchard KI et al (2007) Triple negative than triple-negative phenotype. Clin Cancer Res 14:1367–1376.
breast cancer. Clinical features and patterns of recurrence. Clin doi:10.1158/[Link]-07-1658
Cancer Res 13:4429–4434. doi:10.1158/[Link]-06- 14. Nielsen TO, Hsu FD, Jensen K et al (2004) Immunohistochemical
3045 and clinical characterization of the basal-like subtype of invasive
3. Haffty BG, Yang Q, Reiss M et al (2006) Locoregional relapse berast carcinoma. Clin Cancer Res 10:5367–5374. doi:10.1158/
and distant metastasis in conservatively managed triple negative [Link]-04-0220
early-stage breast cancer. J Clin Oncol 24:5652–5657. 15. Carey LA, Perou CM, Livsay CA et al (2006) Race, breast cancer
doi:10.1200/JCO.2006.06.5664 subtypes and survival in the Carolina Breast Cancer Study.
4. Liedtke C, Mazouni C, Hess KR et al (2008) Response to neo- JAMA 295:2492–2502. doi:10.1001/jama.295.21.2492
adjuvant therapy and long-term survival in patients with triple- 16. Bowen RL, Duffy SW, Ryan DA, Hart IR, Jones JL (2008) Early-
negative breast cancer. J Clin Oncol 26:1–9. doi:10.1200/ onset breast cancer in a group of British black women. Br J
JCO.2007.14.4147 Cancer 98:277–281. doi:10.1038/[Link].6604174
5. Luck AA, Evans AJ, Green AR, Rakha EA, Paish C, Ellis IO 17. Chukwuemeka U, Naab TJ, Mezghebe HM et al (2008) Basal
(2008) The influence of basal phenotype on the metastatic pattern cell-like (triple-negative) breast cancer, a predictor of distant
of breast cancer. Clin Oncol 29:40–45. doi:10.1016/[Link]. metastasis in African American women. Am J Surg 195:153–158.
2007.10.002 doi:10.1016/[Link].2007.09.033
6. Maki D, Grossman R (2000) Patterns of disease spread in met- 18. Foulkes WD, Stefansson IM, Chappuis PO et al (2003) Germline
astatic breast carcinoma: influence of estrogen and progesterone BRCA1 mutations and a basal epithelial phenotype in breast
receptor status. AJNR Am J Neuroradiol 21:1064–1066 cancer. J Natl Cancer Inst 95:1482–1485
7. Tsuda H, Takarabe T, Hasegawa T et al (2000) Large, central 19. Lakhani SR, Reis Filho JS, Fulford L et al (2005) Prediction of
acellular zones indicating myoepithelial tumor differentiation in BRCA1 status in patients with breast cancer using estrogen
high-grade invasive ductal carcinomas as markers of predispo- receptor and basal phenotype. Clin Cancer Res 11:5175–5180.
sition to lung and brain metastases. Am J Surg Pathol 24:197– doi:10.1158/[Link]-04-2424
202. doi:10.1097/00000478-200002000-00005 20. Ridriguez-Pinilla SM, Sarrio D, Honrado E et al (2006) Prog-
8. Hicks DG, Sm Short, Prescott NL et al (2006) Breast cancers with nostic significance of basal-like phenotype and fascin expression
brain metastases are more likely to be estrogen receptor negative, in node-negative invasive breast cancers. Clin Cancer Res
express the basal cytokeratin 5/6, and overexpress HER2 or 12:1533–1539. doi:10.1158/[Link]-05-2281
EGFR. Am J Surg Pathol 30:1097–1104. doi:10.1097/[Link]. 21. Sorle T, Perou CM, Tibshirani R et al (2001) Gene expression
0000213306.05811.b9 patterns of breast carcinomas distinguish tumor subclasses with
9. Albiges L, Andre F, Balleyguier C et al (2005) Spectrum of clinical implications. Proc Natl Acad Sci USA 98:10869–10874.
breast cancer metastasis in BRCA1 mutation carriers; Highly doi:10.1073/pnas.191367098
increased incidence of brain metastases. Ann Oncol 16:1846– 22. Sorlie T, Tibshirani R, Parker J et al (2003) Repeated observation
1847. doi:10.1093/annonc/mdi351 of breast tumor subtypes in independent gene expression data
10. Fulford LG, Reis-Filho JS, Ryder K et al (2007) Basal-like grade sets. Proc Natl Acad Sci USA 100:8418–8423. doi:10.1073/
III invasive ductal carcinoma of the breast: patterns of metastasis pnas.0932692100
and long-term survival. Breast Cancer Res 9:R4. doi:10.1186/ 23. Sotiriou C, Neo SY, McShane LM et al (2003) Breast cancer
bcr1636 classification and prognosis based on gene expression profiles
11. Sawka CA, Pritchard KI, Lickley HLA et al (1995) The Henriette from a population based study. Proc Natl Acad Sci USA
Banting breast centre database: a model for clinical research 100:10393–10398. doi:10.1073/pnas.1732912100
utilizing a hospital-based inception cohort. J Clin Epidemiol 24. Seewaldt VL, Scott V (2007) Images in clinical medicine. Rapid
48:779–786. doi:10.1016/0895-4356(94)00176-Q progression of basal-type breast cancer. N Engl J Med 356:e12.
doi:10.1056/NEJMicm063760

123