Cancer Treatment Reviews 38 (2012) 834–842

Contents lists available at SciVerse ScienceDirect

Cancer Treatment Reviews

journal homepage: www.elsevierhealth.com/journals/ctrv

Controversy

Prognosis of pregnancy-associated breast cancer: A meta-analysis of 30 studies

Hatem A. Azim Jr. a,⇑, Luigi Santoro b, William Russell-Edu c, George Pentheroudakis d, Nicholas Pavlidis d,
Fedro A. Peccatori e
a
Breast Cancer Translational Research Laboratory (BCTL) J.C. Heuson, Université Libre de Bruxelles, Institut Jules Bordet, 1000 Brussels, Belgium
b
Department of Epidemiology and Biostatistics, European Institute of Oncology, Milan, Italy
c
Library, European Institute of Oncology, Milan, Italy
d
Department of Medical Oncology, University of Ioannina, Ioannina, Greece
e
Department of Medicine, Fertility and Procreation Unit in Oncology, European Institute of Oncology, Milan, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Background: Pregnancy-associated breast cancer (PABC) is relatively rare with considerable controversy
Received 24 April 2012 regarding its prognosis.
Received in revised form 9 June 2012 Patients & methods: Two of the authors independently performed a literature search with no date or lan-
Accepted 17 June 2012
guage restrictions. Eligible studies were control-matched, population-based and hospital-based studies
that addressed the outcome of patients diagnosed during pregnancy or 1-year afterwards. The primary
and secondary end-points were overall and disease-free survival respectively. Pooling of data was done
Keywords:
using the random effect model.
Pregnancy associated breast cancer
Prognosis
Results: 30 studies were included in this meta-analysis (3,628 cases and 37,100 controls). PABC patients
Relapse had a significantly higher risk of death compared to those with non-pregnancy-related breast cancer
Survival (pooled hazard ratio (pHR): 1.44; 95% CI [1.27–1.63]). The same results were encountered on restricting
Breast cancer during lactation the analysis to HRs of multivariate analyses (pHR: 1.40 [1.17–1.67]). A clearer trend of poorer outcome
was seen in those diagnosed postpartum (pHR: 1.84; 95% CI [1.28–2.65]) than those diagnosed during
pregnancy (pHR: 1.29; 95% CI [0.74–2.24]). DFS analysis showed a significantly higher risk of relapse
associated with PABC as well (pHR: 1.60 [1.19–2.16]).
Conclusion: Our results show that PABC is independently associated with poor survival particularly those
diagnosed shortly post-partum. This underscores a possible impact of the pregnant breast microenviron-
ment on the biology and consequently the prognosis of these tumors.
Ó 2012 Elsevier Ltd. All rights reserved.

Introduction other studies have shown an independent effect of pregnancy on

Pregnancy-associated breast cancer (PABC) is commonly de-
fined as breast cancer diagnosed during the course of pregnancy
or the one-year following delivery.1 The exact incidence is un-
known; however, it is estimated to constitute around 10% of breast
cancer cases diagnosed below the age of 40 in Western nations.2,3
This incidence is currently on the rise as more women are delaying
childbearing.4
Prognosis of PABC has been addressed in several studies with
inconsistent results.5–8 The relative rarity of the disease precludes
the conduction of large powered controlled studies to address this
question. Some studies have found that PABC is more commonly
diagnosed at an advanced stage5,6 suggesting that the poor progno-
sis is secondary to diagnostic delay rather than an inherent effect
of pregnancy or lactation on breast cancer prognosis. However,
⇑ Corresponding author. Tel.: +32 2 541 33 56; fax: +32 2 541 33 39.
E-mail address: hatem.azim@bordet.be (H.A. Azim Jr.).
outcome.9–11 Nevertheless, the small number of patients examined
in each of the individual studies has hindered the proper interpre-
tation of these data. This is of great clinical relevance as it could
guide treatment strategies for these highly challenging cases.
In this study, we aimed to perform a comprehensive analysis of
all published studies that addressed the prognosis of PABC. We
specifically wanted to clarify whether diagnosis during pregnancy
or 1-year afterwards has an impact on long-term patient outcome.
Methods
The study design was a quantitative synthesis of retrospective
control-matched, population-based and hospital-based studies
that evaluated the impact of breast cancer diagnosis during or
one year following pregnancy on survival. The primary end-point
was overall survival (OS), which was defined as time from breast
cancer diagnosis to death or last follow-up. Secondary end-points
included disease-free survival (DFS), defined as time from breast

0305-7372/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ctrv.2012.06.004
 H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842
cancer diagnosis to local or systemic relapse, secondary cancer,
death or last follow-up.
Data sources and search strategy
The search was carried out independently by two authors (H.A.
Azim Jr. and W. Russell-Edu) with no language or date restrictions
applied. The literature up to December 2011 was searched using
the MEDLINE and EMBASE bibliographic databases. Studies that
were only presented at conferences but not fully published and
available online at the time of the literature search were not con-
sidered eligible. To identify the target population ‘‘breast cancer
patients’’ and the intervention ‘‘pregnancy’’, a search was carried
out combining terminology from the keyword thesauri and free
text fields of the relevant databases: ‘‘breast neoplasms’’/‘‘breast
cancer’’ and ‘‘pregnancy’’/‘‘gestation’’. In order to ensure that all
studies dealing with PABC were captured, no selective keywords
referring to the study design were introduced in the search strat-
egy. A third author (L. Santoro) reviewed the search results and
applied the eligibility criteria to both sets of search outcomes.
Cross-referencing from relevant studies was also performed to
confirm retrieval of all possible studies.
Selection of the articles
We set minimum criteria on the basis of which studies were
deemed eligible for this meta-analysis. These were as follows:
 Presence of a ‘‘case’’ group of patients who were diagnosed dur-
ing pregnancy or one year afterwards (irrespective of pregnancy
outcome) and a ‘‘comparator’’ group of breast cancer patients
who were not diagnosed during these time periods.
Fig. 1. The PRISMA flowchart summarizing the process of identifying the eligible studies.
835
 The study should be independent from other published studies
to prevent giving double weight to estimates derived from the
same study.
 The study should compare outcomes in terms of OS, and/or DFS.
 The study should provide sufficient information to estimate
hazard ratio (HR) or odds ratio (OR) as measures of relative risk,
and 95% confidence interval (CI). This means that eligible stud-
ies had to provide either the HR or the OR or crude data and cor-
responding standard errors, variance, CIs or p-value of the
significance of the estimates. Where such information was not
reported, we estimated the HR from the published survival
curves, where available.
Authors of studies published from 1996 to June 2011 were con-
tacted to provide unpublished data within subgroups of their stud-
ies to allow more precise analysis and minimize publication bias.
Statistical analysis
Measures of association and their corresponding 95% CI (if
available) were log-transformed and the corresponding variance
and standard error were calculated using the formula proposed
by Greenland.12If estimates were not available in the paper, they
were calculated from the published crude data in terms of OR.
Woolf’s formula was used to obtain the standard error of the log
OR.13We considered a ‘‘test-based’’ estimate in case the p-value
was only provided.12Finally if only survival curves were available,
we used the Parmar method to extract the HR, its log-transforma-
tion and its standard error.14
The association between pregnancy and mortality or relapse
across the selected studies was then computed as pooled hazard
ratio (pHR) with 95% CI. The pHR was considered statistically
Table 1

836
Eligible studies.

Author Year PABC Non- Type of OS PABC Lactation HR Age Matching criteria (if case-control study) Adjusting variables (if HR from multivariate model)
PABC study definition definition estimate
Mausner17 1969 73 647 Population- 5-year Pregnancy/ <6 months Crude Mean _ _
based lactation after prg data (35)
18
Wallgren 1977 15 58 Population- 5-year Pregnancy/ <1 year after Crude <30 _ _
based lactation prg data
Nugent19 1985 19 155 Hospital- 5-year Pregnancy Crude <40 _ _
based data
Tretli 1988 20 40 Control- 5-year Pregnancy KM <45 Tumor stage at diagnosis, calendar time, Age _
(pregnant)21 matched curves at diagnosis
Tretli 1988 15 30 Control- 5-year Lactation Unspecified KM <45 Tumor stage at diagnosis, calendar time, Age _
(lactating)21 matched curves at diagnosis
Chiedozi20 1988 36 36 Control- 2-year Pregnancy/ <1 year after KM Mean Age, clinical TNM stage _
matched lactation prg curves (28)
Greene22 1988 8 36 Hospital- 14- Pregnancy Crude <35 _ _
based year data
23
Petrek 1991 56 166 Hospital- 5-year Pregnancy/ <1 year after Paper _ Pathologic LN status

H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842

based lactation prg
5
Ishida 1992 192 191 Control- 10- Pregnancy/ <2 year after KM Mean Same institution, Age at diagnosis, date of _
matched year lactation prg curves (32) beginning treatment
Zemlickis24 1992 102 269 Control- 25- Pregnancy/ Unspecified KM 23–47 Tumor stage, age at diagnosis, date of _
matched year lactation curves beginning of treatment
Guinee 1994 26 139 Hospital- 10- Pregnancy Paper <30 _ Axillary LN involvement, tumor size
(pregnant)6 based year
Guinee 1994 40 139 Hospital- 10- Lactation <1 year after Paper <30 _ Axillary LN involvement, tumor size
(lactating)6 based year prg
Chang25 1994 21 199 Hospital- 5-year Pregnancy/ <1 year after KM 24–43 _ _
based lactation prg curves
Ezzat27 1996 28 84 Control- 7-year Pregnancy Paper <45 Year of diagnosis, date of beginning No adjusting variables
matched treatment
26
Anderson 1996 22 205 Hospital- 10- Pregnancy/ <1 year after Paper <30 _ Stage at diagnosis, axillary LN involvement, Adjuvant CT, tumor size
based year lactation prg
28
Bonnier 1997 154 308 Control- 10- Pregnancy/ <6 months Paper 23–47 Same institution, age, date of beginning Clinical tumor size, LN involvement, inflammatory BC
matched year lactation after prg treatment
Ibrahim29 2000 72 216 Control- 10- Pregnancy KM Median Age at diagnosis, stage, year of diagnosis _
matched year curves
Daling30 2002 83 309 Population- 5-year Lactation <2 year after Paper <45 _ Age, diagnostic period
based prg
31
Zhang 2003 88 176 Control- 20- Pregnancy/ Unspecified KM 23–46 Age at diagnosis, treatment period, hospital, _
matched year lactation curves surgeons
Aziz34 2003 24 48 Control- 7-year Pregnancy/ <1 year after Crude <45 Age at 1st delivery, Tumor grade, Tumor size, _
matched lactation prg data LN status
Martin- 2003 14 122 Hospital- >5- Pregnancy/ Unspecified Crude <40 _ _
Canadas32 based year lactation data
Siegelmann- 2003 22 192 Hospital- 5-year Pregnancy/ <1 year after KM 25–37 _ _
Danieli33 based lactation prg curves
Whiteman35 2004 59 355 Population- 15- Lactation <1 year after Paper <45 _ Type of surgery, RT, race, OC use, education, BMI, history of benign
based year prg breast disease
Rodriguez9 2008 797 4,177 Population- 13- Pregnancy/ <1 year after Paper <55 - Race, tumor size, AJCC stage, type of surgery, ER/PgR status
based year lactation prg
Mathelin 2008 40 61 Control- 10- Pregnancy/ <1 year after Paper Mean Age at diagnosis –
matched year lactation prg (35)
Stensheim 2009 59 13,106 Population- >5 year Pregnancy Paper <50 _ Age, diagnostic period, initial extent of disease
(pregnant)7 based
Stensheim 2009 46 13,106 Population- >5 year Lactation <6 months Paper <50 _ Age, diagnostic period, initial extent of disease
(lactating)7 based after prg
 H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842 837

significant if the 95% CIs did not include 1.0. pHR was estimated by

PABC, pregnancy-associated breast cancer; OS, overall survival; HR, hazard ratio; prg, pregnancy; KM, Kaplan–Meier; RT, Radiation therapy; OC, Oral contraceptive; ER, Estrogen Receptor; PgR, Progesterone Receptor; CT,
pooling the study-specific estimates by random effect models fit-
ted using SAS (proc Mixed) with maximum likelihood estimate.
Histology, tumor size, metastasis, tumor grade, ER/PgR, time

These models provided estimates adjusted for the potential corre-

lation within studies as well as the heterogeneity between studies.
The homogeneity of effect across studies was assessed using the
large sample test, which is based on Cochrane’s Q statistics. A p-va-
lue <0.10 was used to indicate heterogeneity amongst effects. We
also used the I2 statistics to quantify the percentage of variation
Age at diagnosis, calendar time, education

Tumor grade, ER status, LN involvement

across studies that were attributable to heterogeneity rather than

to chance.15 The Macaskill method was used to assess publication
between symptoms and diagnosis

bias.16
Subgroup analyses and meta-regression were performed to
investigate heterogeneity between studies focusing on the study
characteristics, including year of publication, type of study, and
the time of censoring survival. Additional analyses were performed
Neoadj. CT, pN, ER

within different subgroups including time of diagnosis (i.e. during

pregnancy or 1 year afterwards), type of HR estimate (univariate or
multivariate), study size, year of publication, country and age at
diagnosis.
_

Results
Age at diagnosis, tumor size, LN status,

Same institution, age, registration year

Eligible studies

A total of 6449 articles were initially screened. After applying

Age, year of surgery, pT, pN

eligibility criteria, a total of 30 studies were deemed to be eligi-

Age, year of diagnosis

ble5–11,17–39 (PRISMA chart, Fig. 1). We were able to obtain original

data from only three of the included studies.8,11,34
A total of 3,628 patients were diagnosed with PABC and were
compared with a control group of 37,100 breast cancer patients
metastasis

(Table 1). PABC was defined as breast cancer diagnosed during

pregnancy in 5 studies11,19,22,27,29, and during lactation in 2 stud-
_

ies.30,35 In the remaining 23 studies, patients diagnosed during or

within 1 year following pregnancy (sometimes referred to as lacta-
<35

<45

<45

<45

<48

<50

tion) were included.5–10,17,18,20,21,23–26,28,31–34,36–39Of the latter

group, three studies provided the HR estimates both for patients
curves

curves
Paper

Paper

Paper

Paper

diagnosed during pregnancy and during the post-partum peri-

KM

KM

od.6,7,21 Of note, three studies included patients diagnosed up to

Chemotherapy; LN, lymph node; pT, pathological tumor size; pN, pathological nodal status.

2 years following pregnancy.5,30,38

after

after

after

after

after

Four studies17,23,26,33 did not specify how many patients, within

<1 year

<1 year

<1 year

<2 year

<1 year

the PABC, were diagnosed during pregnancy and how many post-
prg

prg

prg

prg

prg

partum. In the remaining studies, a total of 1,034 patients were

diagnosed during pregnancy while 2,416 were diagnosed within
Pregnancy/

Pregnancy/

Pregnancy/

Pregnancy/

Pregnancy/

the following year. The number of subjects analyzed in the differ-

Pregnancy
lactation

lactation

lactation

lactation

lactation

ent studies ranged from 7 to 1,110 in the PABC group and 30 to

14,611 in the control group. Fourteen studies were control-
matched studies.5,8,10,11,20,21,24,27–29,31,34,36,39 Matching criteria
6-year

were mainly according to stage, age and year of diagnosis in 5,

year

year

year

year

year
25-

15-

18-
20-

10-

11 and 6 studies respectively. The remaining studies were either

14,611 Population-

hospital-based or population-based.
564 Hospital-

matched

matched

matched

matched
32 Control-

252 Control-

186 Control-

130 Control-
based

based

Overall survival
3,628 3,7100

Of the 30 studies, 17 and 13 presented the HRs estimated from

the univariate and the multivariate adjusted Cox models respec-
104

32

87

99

65
2011 1110

tively (Table 2). Fourteen studies provided the HR directly in the

paper6,7,9–11,23,26–28,30,35,36,38,39, while in 6 studies; HR was derived
2009

2009

2010

2012

2012

from the crude data.17–19,22,32,34 In the remaining10 studies, sur-

vival was extracted from the published Kaplan–Meier
Total patients

curves.5,8,20,21,24,25,29,31,33,37
Johansson38

39
Moreira10

On considering the entire set of studies, women who were diag-

8
Beadle37

Murphy
Halaska

11

nosed with PABC had a significantly worse OS compared to breast

Azim

cancer controls (pHR: 1.44; 95% CI [1.27–1.63]). However, this was

observed with significant heterogeneity (Q test = 78.7, p < 0.0001;
 838
Table 2
Sensitivity and subgroups analyses.

Trial No. of studies/strata No. of patients PABC/non PABC PRR (95%CI) P-value [BG]a I2parameter, % P-value [WG]b
Sensitivity analysis: lactation period as ‘‘<1 year after delivery’’
Exclusion of studies5,30,38 (<2 years after delivery) 27/30 2243/21989 1.37 (1.21–1.55) 53.5 0.0003
Exclusion of studies5,30,38 (<2 years after delivery) and studies21,24,31,32 24/26 2024/21392 1.40 (1.19–1.63) 58.6 0.0001
(lactation period unspecified)
Sensitivity analysis: published HR or published crude data
Exclusion of studies5,8,20,21,24,25,29,31,33,37 (survival estimated from the 20/22 2924/35155 1.43 (1.23–1.67) 62.9 <0.0001
published survival curves)

H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842

Subgroup analyses
-Type of estimate
HR from univariate model (crude estimate/survival curves) 17/18 897/3072 1.48 (1.22–1.79) 56.0 0.002
HR from multivariate model 13/15 2731/34028 1.40 (1.17–1.67) 0.71 65.0 0.0003
Diagnosis of BC during pregnancy 4 178/13459 1.29 (0.74–2.24) 57.4 0.07
Diagnosis of BC during lactation 4 228/13909 1.84 (1.28–2.65) 0.242 0 0.61
-Country
Western (Europe + North America) 25 3080/35898 1.43 (1.22–1.67) 62.1 <0.0001
Other (Africa + Asia + South America) 8 548/1202 1.46 (1.14–1.88) 0.86 52.9 0.04
-Year of publication
<1990 7 186/1002 1.45 (1.15–1.81) 1.5 0.41
1990–1999 9 641/1561 1.47 (1.03–2.10) 66.0 0.003
2000–2005 7 362/1418 1.50 (1.20–1.87) 15.0 0.31
P2005 10 2439/33119 1.36 (1.02–1.82) 77.9 <0.0001
-Type of studies
Hospital based 10 332/1778 1.44 (1.00–2.08) 0.852 51.1 0.03
Matched controls 15 1054/2059 1.45 (1.14–1.84) 65.7 0.0002
Population based 8 2242/33263 1.42 (1.20–1.67) 0.98 64.0 0.007
-Age at diagnosis (upper limit)
<40 years 9 270/1471 1.49 (1.03–2.14) 44.9 0.07
P40 years 18 2889/34312 1.38 (1.21–1.57) 0.82 54.4 0.003
-Study size
<200 14 382/971 1.61 (1.19–2.17) 55.2 0.007
P200 19 3246/36129 1.38 (1.21–1.58) 0.40 62.8 0.0001

CI = confidence interval; PRR = Pooled Relative Risk.

The I2 represents the percentage of total variation across studies that are attributable to heterogeneity rather than chance.
a
Heterogeneity between groups.
b
Heterogeneity within groups.
 H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842 839

I2 = 59.4) (Fig. 2). On excluding studies, which included patients
who were diagnosed with breast cancer after 1 year of preg-
nancy5,30,38, we obtained the same results (pHR: 1.37; 95% CI
[1.21–1.55]) with significant heterogeneity as well (Q test = 62.4,
p = 0.0003; I2 = 53.5) (Table 2). To increase the estimate accuracy,
a sensitivity analysis was performed by excluding the 10 stud-
ies5,8,20,21,24,25,29,31,33,37 in which the HR was estimated from the
published Kaplan–Meier curves. However, the pHR of the other
20 studies remained unchanged (pHR: 1.43; 95% CI [1.23–1.67]),
as well as the heterogeneity (I2 = 62.9, p < 0.0001) (Table 2).
To examine the independent effect of pregnancy on outcome,
we performed a sensitivity analysis including 13 studies in which
the HRs from multivariate analyses were available. Adjustments
varied according to the individual study (Table 1). In 9 of these
studies, adjustment for a staging parameter (tumor size, nodal sta-
tus or both) was performed. We found that PABC appeared to be
independently associated with worse survival (pHR: 1.40; 95% CI
[1.17–1.67]). High heterogeneity was found among these studies
(I2 = 65.0, p = 0.0003) and no difference was found in comparison
to the group of 17 studies providing only univariate HR (pHR:
1.48; 95% CI [1.22–1.79]) (Table 2).
Disease-free survival
Only 10 studies provided information on DFS (PABC = 531;
controls = 1,842), and hence were included in this analy-
sis.8,11,26–29,32,33,36,37 The HR was provided in the paper in 4 stud-
ies11,28,29,32, estimated from the crude data in two27,36 and
extracted from the published DFS curves in the remaining 4 stud-
ies.8,26,33,37 In an unadjusted analysis, PABC was also associated
with a significant risk of relapse compared to controls (pHR:
1.60; 95% CI [1.19–2.16]) (Fig. 3).Limited information was avail-
able to allow pooling HRs from multivariate models, or to investi-
gate difference according to time of diagnosis (i.e. during
pregnancy or postpartum).
Only two studies investigated the difference in local relapse
rates. Beadle et al. found no differences between patients with
PABC and matched controls.37 A separate analysis according to
time of diagnosis (i.e. during pregnancy or postpartum) showed
the same result. The same findings were reported by Azim et
al., in which there was no difference in local relapse rates be-
tween patients diagnosed during pregnancy and breast cancer
controls.11

Author, year HR (95% CI)

Mausner, 1969
Wallgren, 1977
Nugent, 1985
Tretli, 1988 (pregnancy)
Tretli, 1988 (lactation)
Chiedozi, 1988
Greene, 1988
Petrek, 1991
Ishida, 1992
Zemlickis, 1992
Guinee,1994 (pregnancy)
Guinee, 1994 (lactation)
Chang, 1994
Ezzat, 1996
Anderson, 1996
Bonnier, 1997
Ibrahim, 2000
Daling, 2002
Zhang, 2003
Aziz, 2003
Canadas, 2003
Siegelmann, 2003
Whiteman, 2004
Rodriguez, 2008
Mathelin, 2008
Stensheim, 2009 (pregnancy)
Stensheim, 2009 (lactation)
Beadle, 2009
Halaska, 2009
Moreira, 2010
Azim, 2012
Johansson, 2011
Murphy, 2012 HR (95% CI)

Pooled Hazard Ratio* 1.44 (1.27 –1.63)

2
Q test for Heterogeneity=78.7 (p<0.0001), df=32, I =59.4

0.5 1.0 2.0 5.0 10.0

*Mixed effect model: estimates adjusted for the correlation within studies and heterogeneity between studies
No publication bias.Macaskilltest p-value=0.33

Fig. 2. A forest plot showing the difference in overall survival between PABC patients and controls. The pooled hazard ratio (pHR), the 95% confidence interval are shown, p-
values for heterogeneity and publication bias are shown.
840 H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842

Subgroup and sensitivity analyses
Several factors that may have induced differences on outcome
were investigated with subgroup analyses. These include year of
publication (<1990, 1990–1999, 2000–2005, >2005), type of study
(case-control, hospital-based, population-based), time of diagnosis
(during pregnancy, post-partum) and study sample size (<200,
>200) (Table 2). PABC patients were consistently associated with
poor survival across the different subgroups. No heterogeneity be-
tween strata was observed, however significant heterogeneity
within strata was encountered except for the analysis based on
time of diagnosis (i.e. during pregnancy or within the year follow-
ing pregnancy). Patients diagnosed in the postpartum period had a
clear trend of poorer OS (pHR: 1.81; 95% CI [1.34–2.46]), which was
less apparent in those diagnosed during pregnancy (pHR: 1.30; 95%
CI [0.95–1.79]) (Fig. 4) with a borderline interaction test (p = 0.12).
The analysis performed on the subset of studies providing HRs
from multivariate models showed the same result; pHR: 1.84;
95% CI [1.2–2.65] for patients diagnosed in the postpartum period,
and pHR: 1.29; 95% CI [0.74–2.24] for patients diagnosed during
pregnancy.
Discussion
To the best of our knowledge, this is the largest work to inves-
tigate the prognosis of PABC. Notably, we found that PABC is asso-
ciated with poor prognosis, even after adjustment for confounding
factors. We also found that PABC has a higher risk of recurrence
compared to controls. We did not find any impact on the type, time
and size of the study on survival outcome.
It is known that pregnancy exert a bi-directional effect on
breast cancer development with a short-term increase in risk (up
to 5 or 10 years), and a somewhat protective effect afterwards.40
Hence, this has resulted in considerable inconsistencies in the def-
inition of PABC, with some studies referring to patients diagnosed
with breast cancer up to 5 years or even sometimes 10 years fol-
lowing pregnancy as ‘‘PABC41’’. In our study, we opted to restrict
the analysis to a rather more conventional definition. In addition,
we performed a sensitivity analysis according to the time of diag-
nosis (i.e. during pregnancy or one year post-partum) in order to
work on a more homogenous patient population and elucidate
whether this has an impact on outcome.
One explanation for why PABC has a poor prognosis is that diag-
nosis is delayed in new mothers.5,6 Pregnancy and breastfeeding
increase breast density, making clinical examinations and mam-
mography more difficult to interpret. In the current study, we
found that prognosis remains poor even after adjusting for staging
parameters. An alternative hypothesis is that pregnancy has an
independent effect on prognosis via influencing the biology of
breast cancer. Indeed breast cancer arising at a young age is asso-
ciated with unique biology and poor prognosis.42 Whether preg-
nancy adds a layer of biological complexity is not very clear. A
recent preclinical study has shown that tumors arising shortly
post-partum are more aggressive than tumors arising in nullipa-
rous mice and this is mediated via collagen and Cox-2.43 However,

PABC Controls HR (95% CI)

Author, year Events^/N (%) Events^/N (%)

Ezzat, 1996 16/28 57.1% 46/84 54.8%

Anderson, 1996 12/22 54.5% 56/205 27.3%

Bonnier, 1997 ** 76/154 49.3% 120/308 39.0%

Ibrahim, 2000 (stage I/II) 4/23 17.4% 18/70 25.7%

Ibrahim, 2000 (Stage III) 10/29 34.4% 46/85 54.1%

Canadas, 2003 7/14 50% 46/122 37.4%

Siegelmann, 2003 7/20 35% 28/181 15.5%%

Mathelin, 2008 13/40 32.5% 7/61 11.5%

Beadle, 2009 23/104 22.1% 132/564 23.4%

Halaska, 2009 12/32 37.5% 6/32 18.7%

Azim, 2012 19/65 29.2% 27/130 20.8%

HR (95% CI)

Pooled HR* 199/531 37.5% 532/1842 28.9% 1.60 (1.19 – 2.16)

Q test for Heterogeneity=20.7 (p=0.023), df=10, I2=51.7

0.1 0.5 1.0 2.0 5.0 10.0

Lower DFS in Controls Lower DFS in PABC

*Mixed effect model: estimates adjusted for the correlation within studies and heterogeneity between studies
No publication bias. Macaskill test p-value=0.69
^No. Events (disease): recurrence, progression or death. ** metastasis-free survival

Fig. 3. A forest plot showing the difference in disease-free survival between PABC patients and controls. The pooled hazard ratio (pHR), the 95% confidence interval are
shown, p-values for heterogeneity and publication bias are shown.
 H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842 841

(A) Only BC during Pregnancy HR (95% CI) HR (95% CI)

Nugent, 1985 0.96 (0.55-1.67)
Tretli, 1988 (pregnancy) 2.41 (1.32-4.37)
Greene, 1988 1.50 (0.18-12.6)
Guinee,1994 (pregnancy) 2.83 (1.24-6.45)
Ezzat, 1996 0.90 (0.60-1.30)
Ibrahim, 2000 1.06 (0.69-1.62)
Stensheim, 2009 (pregnancy) 1.23 (0.83-1.81)
Azim, 2012 1.70 (0.80-3.90)

Pooled Hazard Ratio 1.30 (0.95-1.79)

2
Q test for Heterogeneity=13.2 (p=0.07), df=7, I =46.9

(B) Only BC during Lactation

Tretli, 1988 (lactation) 1.47 (0.66-3.27)
Guinee, 1994 (lactation) 1.88 (0.88-3.98)
Daling, 2002 2.30 (1.40-3.90)
Whiteman, 2004 1.51 (1.02-2.23)
Stensheim, 2009 (lactation)
1.95 (1.36-2.78)

Pooled Hazard Ratio 1.81 (1.34-2.46)

Q test for Heterogeneity=2.1 (p=0.71), df=4, I 2=0

0.5 1.0 2.0 5.0 10.0

Lower OS in controls Lower OS in PABC

Heterogeneity Between groups, p=0.12

Fig. 4. (A) A forest plot showing the impact of breast cancer diagnosis during pregnancy on overall survival, (B) A forest plot showing the impact of breast cancer diagnosis
within 1 year following pregnancy on overall survival. The pooled hazard ratio (pHR), the 95% confidence interval are shown, p-values for heterogeneity within and between
strata are shown.

we lack any solid preclinical or clinical evidence on the impact of
pregnancy on the biology of breast cancer arising during the course
of gestation. In the current study, the prognosis of breast cancer
arising postpartum was significantly associated with poor OS, on
considering the pooled analyses of the univariate as well as the
multivariate models. In addition, this was observed with no heter-
ogeneity (I2 in the multivariate model: 0%). On the other hand, the
trend was not as substantial in patients diagnosed during preg-
nancy, in which the HR was lower, not statistically significant with
wide confidence interval and high heterogeneity (I2 in the multi-
variate model: 57.4%). There was no interaction between the re-
sults obtained during pregnancy and postpartum, acknowledging
that the test was of borderline significance (p = 0.12, significant
level <0.10). However, in our opinion, the trend observed in those
diagnosed postpartum is more convincing and less likely to be
confounded by other factors that were not adjusted for, such as
systemic therapy. The latter point is highly relevant as patients
diagnosed during pregnancy could be offered ‘‘non-standard’’ sys-
temic therapy, which could be suboptimal. Among the eligible
studies, Azim et al. have shown that patients diagnosed during
pregnancy had significantly poorer DFS and a trend of a worse
OS even on adjusting for therapy.11 However, treatment was sel-
dom adjusted for in the other studies. Recently, Litton et al. pre-
sented a study showing that women treated with a standard
anthracycline-based regimen [FAC (5-fluorouracil, doxorubicin,
and cyclophosphamide)] during pregnancy had similar DFS and
OS to those treated with the same regimen outside of pregnancy.44
This study is not yet published and included only 54 breast cancer
patients diagnosed during pregnancy, making it hard to draw solid
conclusions. Nevertheless, taking all these facts into account, we
believe that the lack of proper documentation of systemic therapy
in a large fraction of the analyzed studies could partly explain the
trend of poor outcome that we observed in patients diagnosed dur-
ing pregnancy.
However, this does not rule out completely that pregnancy
could impact the biology and prognosis of patients diagnosed with
breast cancer during the course of gestation. Recently, it has been
shown that pregnancy induces a transient 11-fold increase in
mammary stem cell (MaSC) number, which is mediated via
RANK-ligand.45 In humans, a recent in silico analysis has further
shown that both MaSC and RANKL are highly expressed and highly
correlated in breast cancer arising in young women (<40 years).42
MaSCs are also known to express high levels of growth hormone
(GH) receptors.46 Given that GH is highly expressed during preg-
nancy47, it is plausible to assume that GH acts on the GH receptors
on the MaSCs, resulting in an increase in their number and poten-
tially promoting the aggressiveness of the disease. However, this
remains to be explored and confirmed in the setting of breast can-
cer arising during pregnancy.
Our study has some limitations that should be taken into ac-
count. This analysis is based on published data rather than original
patient data. Although we tried to collect original information, we
842 H.A. Azim Jr. et al. / Cancer Treatment Reviews 38 (2012) 834–842
were only successful in obtaining data from three studies. Further-
more, the included studies were mostly old, or included patients
diagnosed over long periods, and hence relevant information
regarding pathological staging, tumor features and treatment
strategies were not always precisely reported. It could have been
valuable to perform sensitivity analysis according to estrogen
receptor status; however this information was also seldom re-
ported in the analyzed trials. We endeavored to take all these
points into consideration on interpreting the results.
In conclusion, patients diagnosed with PABC are independently
associated with poor overall survival. This is particularly obvious in
patients diagnosed in the 1-year post-partum period than those
diagnosed during pregnancy. These results call for an interrogation
of the biology of these tumors beyond phenotyping. Furthermore,
we believe that further studies should not consider patients diag-
nosed during pregnancy or shortly post-partum as one entity. This
will guide further refinement of our understanding to their progno-
sis and biology, which would consequently impact the clinical
management of these patients.
Conflict of interest
The authors have declared no conflicts of interest.
Acknowledgments
Part of this work was presented as a poster presentation in the
2011 San Antonio Breast Cancer Symposium. Hatem A. Azim Jr. and
Luigi Santoro contributed equally to this work. Hatem A. Azim Jr. is
supported by an ESMO research grant. Fedro Peccatori is supported
by the IEO and Avon foundations. We would like to thank Dr. Sayed
A. Aziz for providing original patient information for his study.34
References
1. Psyrri A, Burtness B. Pregnancy-associated breast cancer. Cancer J
2005;11:83–95.
2. Anders CK, Johnson R, Litton J, Phillips M, Bleyer A. Breast cancer before age 40
years. Semin Oncol 2009;36:237–49.
3. Ives AD, Saunders CM, Semmens JB. The Western Australian gestational breast
cancer project: a population-based study of the incidence, management and
outcomes. Breast 2005;14:276–82.
4. Andersson TM, Johansson AL, Hsieh CC, Cnattingius S, Lambe M. Increasing
incidence of pregnancy-associated breast cancer in Sweden. Obstet Gynecol
2009;114:568–72.
5. Ishida T, Yokoe T, Kasumi F, et al. Clinicopathologic characteristics and
prognosis of breast cancer patients associated with pregnancy and lactation:
analysis of case-control study in Japan. Jpn J Cancer Res 1992;83:1143–9.
6. Guinee VF, Olsson H, Moller T, et al. Effect of pregnancy on prognosis for young
women with breast cancer. Lancet 1994;343:1587–9.
7. Stensheim H, Moller B, van Dijk T, Fossa SD. Cause-specific survival for women
diagnosed with cancer during pregnancy or lactation: a registry-based cohort
study. J Clin Oncol 2009;27:45–51.
8. Halaska MJ, Pentheroudakis G, Strnad P, et al. Presentation, management and
outcome of 32 patients with pregnancy-associated breast cancer: a matched
controlled study. Breast J 2009;15:461–7.
9. Rodriguez AO, Chew H, Cress R, et al. Evidence of poorer survival in pregnancy-
associated breast cancer. Obstet Gynecol 2008;112:71–8.
10. Moreira WB, Brandao EC, Soares AN, Lucena CE, Antunes CM. Prognosis for
patients diagnosed with pregnancy-associated breast cancer: a paired case-
control study. Sao Paulo Med J 2010;128:119–24.
11. Azim Jr HA, Botteri E, Renne G, et al. The biological features and prognosis of
breast cancer diagnosed during pregnancy: a case-control study. Acta Oncol
2012;51:653–61.
12. Greenland S. Quantitative methods in the review of epidemiologic literature.
Epidemiol Rev 1987;9:1–30.
13. Woolf B. On estimating the relation between blood group and disease. Ann Hum
Genet 1955;19:251–3.
14. Parmar MK, Torri V, Stewart L. Extracting summary statistics to perform meta-
analyses of the published literature for survival endpoints. Stat Med
1998;17:2815–34.
15. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat
Med 2002;21:1539–58.
16. Macaskill P, Walter SD, Irwig L. A comparison of methods to detect publication
bias in meta-analysis. Stat Med 2001;20:641–54.
17. Mausner JS, Shimkin MB, Moss NH, Rosemond GP. Cancer of the breast in
Philadelphia hospitals 1951–1964. Cancer 1969;23:260–74.
18. Wallgren A, Silfversward C, Hultborn A. Carcinoma of the breast in women
under 30 years of age: a clinical and histopathological study of all cases
reported as carcinoma to the Swedish Cancer Registry, 1958–1968. Cancer
1977;40:916–23.
19. Nugent P, O’Connell TX. Breast cancer and pregnancy. Arch Surg
1985;120:1221–4.
20. Chiedozi LC, Iweze FI, Aboh IF, Ajabor LN. Breast cancer in pregnancy and
lactation. Trop Geogr Med 1988;40:26–30.
21. Tretli S, Kvalheim G, Thoresen S, Host H. Survival of breast cancer patients
diagnosed during pregnancy or lactation. Br J Cancer 1988;58:382–4.
22. Greene FL. Gestational breast cancer: a ten-year experience. South Med J
1988;81:1509–11.
23. Petrek JA, Dukoff R, Rogatko A. Prognosis of pregnancy-associated breast
cancer. Cancer 1991;67:869–72.
24. Zemlickis D, Lishner M, Degendorfer P, et al. Maternal and fetal outcome after
breast cancer in pregnancy. Am J Obstet Gynecol 1992;166:781–7.
25. Chang YT, Loong CC, Wang HC, Jwo SC, Lui WY. Breast cancer and pregnancy.
Zhonghua Yi Xue Za Zhi (Taipei) 1994;54:223–9.
26. Anderson BO, Petrek JA, Byrd DR, Senie RT, Borgen PI. Pregnancy influences
breast cancer stage at diagnosis in women 30 years of age and younger. Ann
Surg Oncol 1996;3:204–11.
27. Ezzat A, Raja MA, Berry J, et al. Impact of pregnancy on non-metastatic breast
cancer: a case control study. Clin Oncol (R Coll Radiol) 1996;8:367–70.
28. Bonnier P, Romain S, Dilhuydy JM, et al. Influence of pregnancy on the outcome
of breast cancer: a case-control study. Societe Francaise de Senologie et de
Pathologie Mammaire Study Group. Int J Cancer 1997;72:720–7.
29. Ibrahim EM, Ezzat AA, Baloush A, Hussain ZH, Mohammed GH. Pregnancy-
associated breast cancer: a case-control study in a young population with a
high-fertility rate. Med Oncol 2000;17:293–300.
30. Daling JR, Malone KE, Doody DR, Anderson BO, Porter PL. The relation of
reproductive factors to mortality from breast cancer. Cancer Epidemiol
Biomarkers Prev 2002;11:235–41.
31. Zhang J, Liu G, Wu J, et al. Pregnancy-associated breast cancer: a case control
and long-term follow-up study in China. J Exp Clin Cancer Res 2003;22:23–7.
32. Martin Canadas F, MorilloConejo M, Diaz M, et al. Breast cancer and pregnancy.
Analysis of our casuistry and literature review. Ginecol Obstet Mex
2003;71:387–93.
33. Siegelmann-Danieli N, Tamir A, Zohar H, et al. Breast cancer in women with
recent exposure to fertility medications is associated with poor prognostic
features. Ann Surg Oncol 2003;10:1031–8.
34. Aziz S, Pervez S, Khan S, et al. Case control study of novel prognostic markers
and disease outcome in pregnancy/lactation-associated breast carcinoma.
Pathol Res Pract 2003;199:15–21.
35. Whiteman MK, Hillis SD, Curtis KM, et al. Reproductive history and mortality
after breast cancer diagnosis. Obstet Gynecol 2004;104:146–54.
36. Mathelin C, Annane K, Treisser A, et al. Pregnancy and post-partum breast
cancer: a prospective study. Anticancer Res 2008;28:2447–52.
37. Beadle BM, Woodward WA, Middleton LP, et al. The impact of pregnancy on
breast cancer outcomes in women<or=35 years. Cancer 2009;115:1174–84.
38. Johansson AL, Andersson TM, Hsieh CC, Cnattingius S, Lambe M. Increased
mortality in women with breast cancer detected during pregnancy and
different periods postpartum. Cancer Epidemiol Biomarkers Prev
2011;20:1865–72.
39. Murphy CG, Mallam D, Stein S, et al. Current or recent pregnancy is associated
with adverse pathologic features but not impaired survival in early breast
cancer. Cancer 2012;118:3254–9.
40. Lambe M, Hsieh C, Trichopoulos D, Ekbom A, Pavia M, Adami HO. Transient
increase in the risk of breast cancer after giving birth. N Engl J Med
1994;331:5–9.
41. Borges VF, Schedin PJ. Pregnancy-associated breast cancer: an entity needing
refinement of the definition. Cancer 2012;118:3226–8.
42. Azim Jr HA, Michiels S, Bedard P, et al. Elucidating prognosis and biology of
breast cancer arising in young women using gene expression profiling. Clin
Cancer Res 2012;18:1241–51.
43. Lyons TR, O’Brien J, Borges VF, et al. Postpartum mammary gland involution
drives progression of ductal carcinoma in situ through collagen and COX-2. Nat
Med 2011;17:1109–15.
44. Litton JK, Warneke CL, Hahn K, et al. Case-control analysis of patients (pts)
treated with chemotherapy during pregnancy for breast cancer (BC). J Clin
Oncol 2010 Abstract # 105.
45. Asselin-Labat ML, Vaillant F, Sheridan JM, et al. Control of mammary stem cell
function by steroid hormone signalling. Nature 2010;465:798–802.
46. Dontu G, Abdallah WM, Foley JM, et al. In vitro propagation and transcriptional
profiling of human mammary stem/progenitor cells. Genes Dev
2003;17:1253–70.
47. Laban C, Bustin SA, Jenkins PJ. The GH-IGF-I axis and breast cancer. Trends
Endocrinol Metab 2003;14:28–34.