Fax: (646) 888-4555
Current or recent pregnancy is associated with adverse pathologic features but not impaired survival in early breast cancer†
Article first published online: 15 NOV 2011
Copyright © 2011 American Cancer Society
Volume 118, Issue 13, pages 3254–3259, 1 July 2012
How to Cite
Murphy, C. G., Mallam, D., Stein, S., Patil, S., Howard, J., Sklarin, N., Hudis, C. A., Gemignani, M. L. and Seidman, A. D. (2012), Current or recent pregnancy is associated with adverse pathologic features but not impaired survival in early breast cancer. Cancer, 118: 3254–3259. doi: 10.1002/cncr.26654
See editorial on pages 3226–8, this issue.
- Issue published online: 18 JUN 2012
- Article first published online: 15 NOV 2011
- Manuscript Accepted: 16 AUG 2011
- Manuscript Revised: 30 JUN 2011
- Manuscript Received: 4 APR 2011
- breast cancer;
- pathologic features
Pregnancy-associated breast cancer (PABC) may be defined as breast cancer diagnosed during pregnancy or within 1 year of giving birth. Conflicting data exist regarding the impact of pregnancy on clinical features and prognosis of breast cancer.
A single-institution retrospective chart review was performed of 99 patients identified with PABC between 1992 and 2007. Non-PABC controls were matched 2:1 to PABC cases by year of diagnosis and age. The differences in clinical features were compared between cases and controls using chi-square tests. Univariate and multivariate analyses were performed to assess the effect of PABC on survival.
Of the 99 PABC cases, breast cancer was diagnosed during pregnancy in 36 patients, and after delivery in 63. PABC cases were more likely than controls to be negative for estrogen receptor (59% vs 31%, P < .0001) and negative for progesterone receptor (72% vs 40%, P < .0001). Cases were also more likely to have advanced T class (P = .0271) and N class (P = .0104) and higher grade tumors (P = .0115). With a median follow-up of 6.3 years for cases and 4.7 years for controls, overall survival did not differ between cases and controls (P = .0787). On multivariate analysis, the independent prognostic factors for overall survival were estrogen receptor status (P = .0031) and N class (P = .0003). The diagnosis of PABC was not an independent prognostic factor (P = .1317).
PABC is associated with more adverse tumor features than non-PABC matched for age and year of diagnosis. After correcting for pathologic features, the diagnosis of PABC is not in itself an adverse prognostic factor for survival. Cancer 2011. © 2011 American Cancer Society.
Pregnancy-associated breast cancer (PABC) may be defined as breast cancer that is diagnosed during pregnancy or within 1 year after giving birth.1 With more women becoming pregnant later in their reproductive years, this diagnosis is likely to be seen more commonly in clinical practice.2 Conflicting data exist from the larger published series of PABC regarding the impact of this diagnosis on tumor features and outcomes.1, 3-12 Some authors have concluded that this diagnosis is associated with a higher risk of recurrence and death, whereas others found no evidence of impaired outcomes upon controlling for tumor pathologic features. Analysis of the outcomes of women treated for PABC at Memorial Sloan-Kettering Cancer Center, New York, NY, between 1960 and 1980 showed no difference in survival compared with age-matched controls.1 We performed a retrospective chart review of patients diagnosed with PABC at Memorial Sloan-Kettering Cancer Center between 1981 and 2007 in order to update our institutional experience and to clarify the effect of this diagnosis on prognosis and outcome in the young women affected. This modern cohort of women is likely to have been more reliably staged than earlier published series, many of which included women diagnosed as far back as the 1950s and 1960s.1, 3, 8, 10 In addition, the availability of a detailed institutional database enabled us to comprehensively review the records of cases and controls, allowing us to collect detailed data regarding treatment and outcome and preventing contamination from incorrect case or control allocation.
MATERIALS AND METHODS
Patients diagnosed with PABC between January 1, 1981, and December 31, 2007, were identified through a search of our institutional database, following approval by the MSKCC institutional review board. Because PABC is not included in the commonly used international and/or institutional medical coding systems, we used surrogate indicators to identify potentially eligible patients. We performed a free-text search of pathology reports for keywords and/or components of words, including “partum,” “pregnancy,” and “lactation.” The medical records of all patients identified through this initial search were comprehensively reviewed to confirm the diagnosis of PABC. Only patients confirmed to have received a new diagnosis of invasive breast cancer during or within 1 year after pregnancy were considered to be true cases. Details were retrieved from the medical records of these patients, including reproductive history, tumor pathologic features, treatment factors (including type of surgery and use of radiotherapy, chemotherapy, and endocrine therapy), and outcome measures including time and site of metastases and survival.
A list of control patients matched for age and year of diagnosis was identified through the institutional database. Two controls were randomly selected for each patient from this list. If 2 controls of the same age were not available from the same year of diagnosis, controls were selected from the following year. Where 2 controls of the same age were not available from the same or following year, only 1 control was selected.
Differences in the distributions of clinical features between cases and controls were assessed, using chi-square tests. Univariate differences in overall survival (OS) for each clinical feature were assessed using the log-rank test and Kaplan-Meier methods. Univariate analysis was performed separately for the 2 groups (cases and controls). To determine whether case or control status was an independent prognostic factor, the 2 groups were combined for multivariate analysis using a Cox regression model.
A total of 296 potential PABC cases were identified through use of the free-text search of pathology records. Of these, 99 cases of PABC were confirmed following review of the medical records of these patients. The clinical and treatment features of these 99 patients are summarized in Table 1. Nine patients received chemotherapy during pregnancy, consisting of doxorubicin with or without cyclophosphamide in all cases.
|Patient Characteristics||n (%)||n (%)||Median (Range)|
|Age at diagnosis||35 (24-48)|
|Age at first delivery||31 (17-48)|
|Pregnancy status at diagnosis|
|Chemotherapy during pregnancy?||9 (25%)|
|Termination of pregnancy with or without chemotherapy||9 (25%)|
|Chemotherapy after pregnancy||17 (47%)|
|No chemotherapy||1 (3%)|
|Months from delivery||6 (1-12)|
|Died in follow-up||15 (15%)|
|Alive at last follow-up||84 (85%)|
|74 (75%)||25 (26%)|
|Chemotherapy||96 (97%)||2 (2%)||1 (1%)|
|Anthracycline||79 (80%)||18 (18%)||2 (2%)|
|Taxane||63 (64%)||34 (34%)||2 (2%)|
|Trastuzumab||9 (9%)||88 (89%)||2 (2%)|
|HER2-positive patients||9 (45%)||11 (55%)||0 (0%)|
|Radiation therapy||49 (49.5%)||49 (49.5%)||1 (1%)|
|Endocrine therapy||62 (63%)||35 (35%)||2 (2%)|
A total of 186 controls were identified: 2 controls were identified for 87 cases, 1 control was available for 10 cases, and no control was available for 1 case. Pathologic features of cases and controls are described in Table 2. The tumors of cases were significantly more likely to be negative for estrogen receptor (ER) and progesterone receptor (PR), have a higher “T” and “N” class at diagnosis (as per the TNM classification system of the American Joint Committee on Cancer), and be of higher pathologic grade. No significant differences in race were identified between cases and controls (white, 87.9% vs 80.1%; black, 5.1% vs 8.1%; Asian, 4.0% vs 8.6%; other, 1.0% vs 0.5%; unknown, 2.0% vs 2.7%). Controls were more likely than cases to be nulliparous (39% vs 6%; P < .0001, 2-sample t test). The remaining controls were parous, with 1 to 4 deliveries, but none within the previous 12 months. The age at first delivery was older among cases than controls: mean age was 31.4 years (standard deviation, 5.18 years) versus 27 years (standard deviation, 4.39 years); P < .0001, 2-sample t test.
|Feature||Cases (n = 99) n (%)||Controls (n = 186) n (%)||P|
|Negative||58 (59%)||58 (31%)|
|Positive||39 (39%)||121 (65%)||<.0001|
|Negative||57 (58%)||109 (59%)|
|Positive||20 (20%)||35 (19%)||0.74|
|Negative||71 (72%)||75 (40%)|
|Positive||26 (26%)||103 (55%)||<.0001|
|0 or 1||41 (41%)||103 (55%)|
|2||48 (48%)||64 (34%)|
|3||10 (10%)||13 (7%)|
|4||0 (0%)||5 (3%)||.0271|
|0||40 (40%)||82 (44%)|
|1||29 (29%)||75 (40%)|
|2||19 (19%)||15 (8%)|
|3||11 (11%)||11 (6%)||.0104|
|0||3 (3%)||11 (6%)|
|1||2 (2%)||7 (4%)|
|2||11 (11%)||46 (25%)|
|3||83 (84%)||120 (65%)||.0115|
To test if delayed or radiographically limited evaluation during pregnancy could be a factor in the more advanced stage at diagnosis of cases, we divided our PABC cases into 2 subgroups: those diagnosed during pregnancy and those diagnosed in the year after giving birth. We compared the T and N classes of these subgroups. No significant difference in T or N class was evident between the 2 subgroups of cases (Table 3).
|Status at Diagnosis|
|All (n = 99)||Pregnant (n = 36)||Postpartum (n = 63)||P|
|T0||1 (1%)||1 (3%)||0 (0%)|
|T1||40 (40%)||14 (39%)||26 (41%)|
|T2||48 (48%)||20 (56%)||28 (44%)|
|T3||10 (10%)||1 (3%)||9 (14%)||0.17|
|N0||40 (40%)||12 (33%)||28 (44%)|
|N1||29 (29%)||9 (25%)||20 (32%)|
|N2||19 (19%)||8 (22%)||11 (17%)|
|N3||11 (11%)||7 (19%)||4 (6%)||0.18|
Among patients with breast cancer positive for hormone receptor, similar percentages of cases and controls had documented adjuvant endocrine therapy (72.1% vs 67.4%), defined as tamoxifen and/or aromatase inhibitor treatment and not including oophorectomy. More cases than controls had documented chemotherapy (96% vs 79%); this may reflect the more advanced stage of breast cancer diagnosed in cases. Among patients who received chemotherapy, similar proportions received taxanes (66% vs 65%) and anthracyclines (83% vs 80%). Among patients with breast cancer positive for human epidermal growth factor receptor 2 (HER2), more cases than controls received trastuzumab (45% vs 29%).
On univariate analysis, axillary lymph node involvement was a significant prognostic factor among cases, whereas tumor ER status and T class were significant prognostic factors among controls. On multivariate analysis of all patients, ER status and lymph node involvement emerged as independent factors, whereas case or control status was not an independent prognostic variable (Table 4).
|Case vs control||0.59||0.29, 1.17||0.1317|
|ER-negative vs ER-positive||0.35||0.18, 0.70||0.0031|
|N3 vs N0||6.70||2.65, 17.43|
|N2 vs N0||1.24||0.38, 4.07||0.0003|
|N1 vs N0||1.38||0.62, 3.07|
At a median follow-up of 6.3 years for cases and 4.7 years for controls, median OS has not been reached. There is no evidence of a difference in OS between cases and controls (Fig. 1). Similarly, analysis of distant disease-free survival did not detect a significant difference between cases and controls.
We found a preponderance of adverse pathologic features in our patients who had PABC, including more advanced T and N class, high histologic grade, and hormone receptor negativity. Similar findings have previously been described in PABC by others.13 Such findings may be attributed to the young age of women with PABC, because young women in general are more likely to develop tumors with these pathologic characteristics.14, 15 Our matching strategy allowed us to correct for the impact of age, however, and we found that recent or current pregnancy is independently associated with more adverse pathologic features in young women.
On multivariate analysis, we found that diagnosis during or within a year of pregnancy was not an independent prognostic variable for OS. In other words, although PABC is associated with tumors that are more likely to be of higher grade, negative for hormone receptor, and locally advanced at diagnosis, our data suggest that the diagnosis of PABC is not itself a negative prognostic factor after controlling for these features.
Multiple other case-control studies have attempted to answer this question of whether pregnancy is a negative prognostic factor in young women with breast cancer.1, 3-12 Some of these studies found no impact on survival when pathologic features were corrected for,1, 3-6 some had conflicting results among different subgroups of women with PABC,7, 8 whereas others suggested a worse outcome for women with PABC.9-12 Most of these studies were hampered by very small sample sizes, with many containing 40 or fewer cases.1, 4, 6-9, 11 A multicenter French study of comparable size to ours found a similar preponderance of hormone receptor–negative and larger tumors among PABC cases.10 However, this study indicated a worse outcome for PABC cases than controls matched for age and year of diagnosis. Notably, the impact of pregnancy on OS did not reach statistical significance on multivariate analysis (P = .08), and hormone receptor status was not included in multivariate analysis. A single-institution study from the University of Texas MD Anderson Cancer Center compared the outcomes of 104 patients with PABC to those of 564 unmatched patients without PABC.16 No difference in locoregional recurrence, distant metastases, or OS was found, although multivariate analysis was not performed. The largest study of its kind used registry information to compare tumor features and outcomes of 797 PABC cases and 4177 age-matched controls and did find a small but significant effect of pregnancy on outcome on multivariate analysis, with a 14% increased risk of death.12 Although we did not find an independent effect of pregnancy on prognosis, it is possible that a small effect of a similar magnitude to that described in the registry study truly exists, which cannot be detected by our study due to size limitations.
A small number of women in our study (n = 9) were treated with cytotoxic chemotherapy during pregnancy, all of whom received anthracycline-based treatment. The vast majority of experience with cytotoxic therapy in pregnancy involves anthracycline-based regimens, whereas experience with newer agents such as taxanes and the biologic agent trastuzumab is largely limited to case reports.17 No immediate complications were reported in our patients, and all infants were delivered healthy. Several case series have reported on the use of cytotoxic therapy in pregnancy, with no apparent increase in the risk of congenital malformations seen when therapy was initiated after the first trimester in the largest series.18-20 Despite these reassuring findings, a series of women receiving treatment during pregnancy for various malignancies (including breast cancer) found a high rate of preterm deliveries (54.2%).21 Although this figure was largely composed of iatrogenic inductions, the spontaneous preterm labor rate does appear to be higher in patients exposed to chemotherapy than in the general population (12.9% vs 4%, P = .006, binomial test). In order to improve understanding of breast cancer diagnosed during pregnancy, a voluntary national cancer registry is collecting prospective, long-term data.22 As of 2010, 130 patients had been enrolled (99 prospectively). Among 104 patients treated with chemotherapy during pregnancy, including 11 with taxanes, there was no evidence of an increased rate of fetal malformations.
Pregnancy has a complex bimodal effect on breast cancer risk, with an initial transient increase in risk, especially for pregnancies at older maternal ages, ultimately followed by a long-term protective effect.23, 24 In addition, population-based data suggests that the interval between last pregnancy and breast cancer diagnosis may be linked to the risk of death, with the highest risk for breast cancers diagnosed in the first 12 months after childbirth.25 With increasing age at first delivery, the negative effect may not be outweighed by the protective effect within a woman's lifetime.24 Pregnancy clearly has multiple complex effects on breast tissue, which may have profound effects on both normal tissue and developing tumors. A potential mechanism for increased aggressiveness of breast cancer diagnosed or treated in the postpartum period is facilitation of metastasis through the wound-healing and/or proinflammatory microenvironment of the involuting breast.24 In fact, gene expression analysis of nontumorous breast tissue from nulliparous and recently pregnant (<2 years) women indicates differential expression of multiple genes, including inflammation-associated genes, ER alpha, PR, and HER2.26 The use of such molecular profiling techniques could help to clarify if there is a true biologic distinction between pregnancy-associated and non–pregnancy-associated breast cancer in young women.
Despite the availability of a comprehensive electronic database, we still encountered difficulty identifying women with the diagnosis of PABC. Of the 296 potential cases identified by our initial database search, two-thirds were shown not to be true cases of PABC on chart review (our search terms were admittedly broad in order to “capture” as many cases as possible). In addition, some patients in the control group were replaced when chart review revealed that they did in fact have PABC. Accordingly, the real incidence of PABC at our institution is likely much higher than that suggested by the number of cases we describe here. As a noncoded, nonidentified disease, PABC remains a shadowy entity, typically described in retrospective case series and therefore limited by all of the potential biases inherent to such studies. The solution to this problem is the establishment of prospective databases of such patients. Thankfully, this is being performed, both within institutions and as part of large national and international collaborations.22, 27
In conclusion, we found that PABC is a clinically and biologically distinct entity, with differing pathologic findings to breast cancer diagnosed in other young women of the same age at diagnosis. Nevertheless, our data suggest that regardless of biologic differences, PABC is itself not an independent prognostic factor and the risk/benefit analysis of adjuvant therapy should focus on the same clinicopathologic characteristics as in other women.
Dr Murphy received funding support from the Irish Society of Medical Oncology.
CONFLICT OF INTEREST DISCLOSURE
The authors made no disclosure.