Introduction

The diagnosis of breast cancer for the expectant mother is certainly devastating. At a time when a woman and her family are preparing to celebrate the joy of new life, to be faced with the real possibility of death is a substantial emotional challenge. Such a situation also signifies a complex medical conundrum. The practitioner must provide best care to the mother whilst at the same time ensuring the care of the foetus is not compromised.

Pregnancy-associated breast cancer (PABC) is generally defined in the literature as breast cancer diagnosed during pregnancy or the first year postpartum. Some studies, however, have extended this definition to up to 5 years following delivery [1–3]. Whether or not such an extended definition is warranted remains unclear and will be explored in this analysis.

Breast cancer is one of the most common malignancies encountered during pregnancy. Approximately 0.2–3.8 % of all breast cancers are pregnancy related [4], whilst 10 % of breast cancers in women <40 years of age occur during pregnancy [5]. The incidence of PABC in Western countries is estimated to be 1 in 3000 to 1 in 10,000 pregnancies [6], and as women increasingly delay childbearing to older maternal ages, it is likely this incidence will rise.

The treatment of PABC requires the careful balancing of interests between the mother and foetus. Standard treatment includes surgery as well as neoadjuvant and adjuvant chemotherapy from the second trimester. Exposure during the second and third trimesters has not been associated with teratogenic effects [7–9]. Most cases during pregnancy are treated with a combination chemotherapy regimen such a FAC (5-fluorouracil, doxorubicin and cyclophosphamide); however, recent studies have also demonstrated successful outcomes with taxanes [10, 11]. Radiotherapy is contraindicated and should be delayed until after delivery.

The rarity of cases and the impracticability of conducting randomised controlled studies in this setting have limited reports on prognosis to retrospective studies with generally small cohorts. Studies that address the impact of pregnancy on the prognosis of such cases have had inconsistent results. Whilst some studies suggest outcomes for women are similar to non-pregnant patients with breast cancer [5, 12–17], others have demonstrated that pregnancy in itself is an adverse prognostic factor for survival [18–23]. A meta-analysis of 30 studies conducted by Azim et al. in 2012 concluded that PABC is associated with poorer prognosis, particularly when diagnosed postpartum [24]. Amant et al. [11] conducted the largest study to date on the prognosis of breast cancer diagnosed during pregnancy and did not find a significant difference in disease-free survival (DFS) or overall survival (OS) between 311 pregnant women with breast cancer and those of 865 non-pregnant controls matched for known prognostic factors such as stage, age, hormonal receptors and type of treatment.

A further challenge to the practitioner are women who have previously been diagnosed and treated for breast cancer and who subsequently desire to become pregnant. Such women are likely to wonder whether a future pregnancy could adversely affect their prognosis and how long they should wait to conceive following cessation of their treatment. A meta-analysis of 14 studies by Azim et al. in 2011 found that pregnancy following breast cancer was safe and in fact those who became pregnant after a diagnosis of breast cancer had improved survival outcomes compared to those who did not become pregnant [25].

It is debatable whether these favourable results are due to a selection bias rather than the actual protective effect of pregnancy. The “healthy mother effect” [26] suggests that women who have had more favourable outcomes are more likely to conceive than those who have had a recurrence after diagnosis and hence the improved outcome observed is really a reflection of selecting women who have not relapsed. In order to account for this bias, some studies have matched cases and controls according to nodal status, ER status, disease-free interval and treatment [27].

In the last 5 years, important studies have been performed in this area, significantly adding to the body of literature and understanding in the field. Since Azim’s 2012 meta-analysis [24], there have been an additional 14 studies that have examined pregnancy occurring before or concurrently with breast cancer diagnosis, whilst there have been an additional 5 studies examining pregnancy after diagnosis. Studies unique to these meta-analyses appear in bold in Tables 1 and 2. Azim’s 2012 meta-analysis defined PABC as during pregnancy or up to 1 year postpartum [24]. This study adopted a broader definition of PABC whereby cases were defined as those during pregnancy or up to 5 years postpartum. We also performed sub-analyses where cases were defined up to 2 years postpartum and up to 1 year postpartum to determine whether the extended time period postpartum provided by some studies is meaningful. It has been hypothesised that the poor prognosis seen in cases diagnosed postpartum may be responsible for the overall increased risk seen in PABC. As such we performed sub-analyses based on studies that only reported on postpartum cases, as well as those that only reported on cases during pregnancy. This study therefore aims to provide an updated and comprehensive meta-analysis on the mortality and morbidity of woman diagnosed with breast cancer before, during and after pregnancy.

Table 1 Study characteristics (pregnancy before/during diagnosis) (n = 41)
Full size table
Table 2 Study characteristics (pregnancy after diagnosis) (n = 19)
Full size table

Methods

Study protocol

We followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines [28]. A systematic search of the databases MEDLINE, EMBASE, PubMed and Google (first 20 pages) through to 31 August 2016 was conducted to identify relevant articles. The search used the terms ‘pregnant’ or ‘gestation’ and ‘breast’ and ‘cancer’ or ‘neoplasia’ or ‘carcinoma’, which were searched as text word and as exploded medical subject headings where possible. The reference lists of relevant articles were also searched for appropriate studies. No language restrictions were used in either the search or study selection. The search was limited to humans. A search for unpublished literature was performed; however, no additional studies were found.

Study selection

Pregnancy before breast cancer diagnosis

We included studies that met the following inclusion criteria: (1) PABC was recognised on clinical diagnosis and/or confirmed histologically; (2) cases were defined as those diagnosed during pregnancy or up to 5 years postpartum; (3) the study reports outcomes in terms of OS and/or DFS; (4) the risk point estimate was reported as a hazard ratio (HR), or the data were presented such that an OR could be calculated; (5) the 95 % confidence interval (CI) was reported, or the data were presented such that the CI could be calculated.

Pregnancy after breast cancer diagnosis

We included studies that met the following inclusion criteria: (1) Breast cancer was recognised on clinical diagnosis and/or confirmed histologically; (2) cases were defined as pregnancy occurring up to 5 years after diagnosis; (3) the study reports outcomes in terms of OS and/or DFS; (4) the risk point estimate was reported as a HR or OR, or the data were presented such that an OR could be calculated; (5) the 95 % CI was reported, or the data were presented such that the CI could be calculated.

Data extraction

The data extraction was performed using a standardised data extraction form, collecting information on the publication year, study design, number of cases, number of controls, total sample size, temporal direction, population type, country, continent, case control matching, number of adjusted variables, the risk estimates or data used to calculate the risk estimates, CIs and data used to calculate CIs. Authors were contacted for missing data. Adjusted ratios were extracted in preference to non-adjusted ratios; however, where ratios were not provided, unadjusted ORs and CIs were calculated from the published crude data using 2 × 2 contingency tables. Where more than one adjusted ratio was reported, we chose the ratio with the highest number of adjusted variables. Where multiple risk estimates were available in the same study, for example due to the use of different comparator groups, they were included as separate risk estimates. Where only KM curves were provided, the Parmar method [29] was used to extract the HR. The studies were then allocated into one of two groups: those for which HR was provided or extracted from KM curves, and those for which OR was calculated from crude data. In order to be statistically sound, we conducted separate analyses on these two groups. For the analysis of pregnancy occurring during or before breast cancer, there were 13 studies consisting of 488 cases for which OR had to be calculated for OS and 8 studies consisting of 439 cases for which OR had to be calculated for DFS. For the analysis of pregnancy occurring after breast cancer diagnosis, there were 6 studies consisting of 171 cases for which OR had to be calculated for OS.

Statistical analysis

Pooled OR and 95 % CI were calculated using a random effects model [30]. In our study, a HR or OR > 1 indicated an increased risk of death, or an increased risk of disease recurrence. We tested heterogeneity with Cochran’s Q statistic, with p < 0.10 indicating heterogeneity, and quantified the degree of heterogeneity using the I 2 statistic, which represents the percentage of the total variability across studies which is due to heterogeneity. I 2 values of 25, 50 and 75 % corresponded to low, moderate and high degrees of heterogeneity, respectively [31]. We quantified publication bias using the Egger’s regression model [32], with the effect of bias assessed using the fail-safe number method. The fail-safe number was the number of studies that we would need to have missed for our observed result to yield a statistically non-significant effect. Publication bias is generally regarded as a concern if the fail-safe number is less than 5n + 10, with n being the number of studies included in the meta-analysis [33]. All analyses were performed with Comprehensive Meta-analysis (version 3.0), Biostat, Englewood, NJ (2014).

Results

From 6492 citations screened by our search, we identified a total of 60 studies that met our inclusion criteria (Fig. 1). 41 of these studies examined pregnancy occurring before or concurrently with breast cancer diagnosis up to 5 years postpartum, whilst 19 examined pregnancy followed a breast cancer diagnosis. Tables 1 and 2 show selected characteristics of the identified studies. In terms of study design, 44 were case control studies and 16 were cohort studies. 27 studies examined populations from Europe, 22 from North America, 6 from Asia, 3 from Oceania, 1 from South America and 1 was a global multicentre study. In the data collection process, adjusted and calculated HR were used and where they were not provided nor could be extracted from KM survival curves, OR were manually calculated by 2 × 2 contingency tables. Our results are summarised in Table 3.

Fig. 1
figure 1

Flowchart of the search strategy

Full size image
Table 3 Meta-analyses conducted
Full size table

Pregnancy occurring before breast cancer diagnosis or concurrently with breast cancer diagnosis

41 studies (34 case control and 7 cohort) comprising of 4929 cases and a total of 65,970 individuals were identified for the meta-analysis. 18 studies had data available for both OS and DFS, whilst 18 had data on OS only and 4 had data on DFS only.

Overall survival (OS)

37 studies (32 case control and 5 cohort) comprising of 5583 cases and a total of 124,277 individuals were identified for the meta-analysis on OS. There was no publication bias as analysed by funnel plot based on Egger’s regression (p = 0.19) (Fig. 2). There was an overall increased risk of death for patients compared to controls with a pooled hazard ratio (pHR) of 1.57 (95 % CI 1.35–1.82) (Fig. 4). There was significant heterogeneity (I 2 = 84.90, p = 0.001). When cases were limited to those diagnosed up to 2 years postpartum, the pHR increased significantly to 1.96 (95 % CI 1.87–2.05) but did not change when the definition was limited to up to 1 year postpartum, HR 1.97 (95 % CI 1.88–2.06) (Table 3). There was however, significant heterogeneity up to 1 year postpartum (I 2 = 84.31, p = 0.001), as well as up to 2 years postpartum (I 2 = 83.97, p = 0.001).

Fig. 2
figure 2

Funnel plot to assess publication bias. Circles indicate individual studies. Diamond indicates summary estimates. SE standard error

Full size image
PABC (pregnant and postpartum)

23 studies comprising of 3150 cases with a total of 34,445 individuals were identified for the PABC sub-analysis. This sub-analysis examined where diagnosis occurred either during pregnancy or up to 5 years postpartum, and excluded studies that exclusively examined the postpartum period. PABC individuals were at an increased risk of death, with a pHR of 1.46 (95 % CI 1.17–1.82) (Fig. 4). There was significant heterogeneity (I 2 = 75.45, p < 0.001). When cases were defined as up to 1 year postpartum, the pHR increased marginally to 1.53 (95 % CI 1.11–2.11) (Table 3). The pooled OR of studies where OR was calculated provided similar results [OR 1.56; 95 % CI 1.10–2.21) (Table 3) with mild heterogeneity that was not significant (I 2 = 37.61, p = 0.11).

Pregnancy

13 studies comprising of 906 cases with a total of 45,082 individuals were identified for the pregnancy only sub-analysis. This included studies where diagnosis of breast cancer occurred exclusively during pregnancy. There was an increased risk of death for patients diagnosed during pregnancy compared to non-pregnant controls, with a pHR of 1.47 (95 % CI 1.04–2.08) (Table 3). There was substantial heterogeneity (I 2 = 83.11, p < 0.001).

Postpartum

8 studies comprising of 1439 cases with a total of 44,662 individuals were identified for the postpartum only sub-analysis. Postpartum individuals were at an increased risk of death up to 5 years postpartum, with a pHR of 1.79 (95 % CI 1.39–2.29) (Fig. 4). There was significant heterogeneity (I 2 = 83.67, p < 0.001). When cases were defined as up to 1 year postpartum the pHR increased further to 1.99 (95 % CI 1.90–2.09) (Table 3) with no significant heterogeneity (I 2 = 0.00, p = 0.64).

Disease-free survival (DFS)

23 studies comprising of 1857 cases and a total of 9798 individuals were identified for the meta-analysis on DFS. There was no publication bias as analysed by funnel plot based on Egger’s regression (p = 0.07) (Fig. 3). Patients were at an increased risk of relapse or progression up to 5 years postpartum, with a pHR of 1.51 (1.22–1.88) (Fig. 5). There was significant heterogeneity (I 2 = 60.75, p = 0.01). When the definition of PABC was narrowed to up to 2 years postpartum, the pHR did not differ, 1.52 (95 % CI 1.23–1.88) (Table 3). The pooled OR provided similar results [OR 1.66; 95 % CI 1.09–2.52] (Table 3) with substantial heterogeneity (I 2 = 72.41, p = 0.001).

Fig. 3
figure 3

Funnel plot to assess publication bias. Circles indicate individual studies. Diamond indicates summary estimates. SE standard error

Full size image
Fig. 4
figure 4

Meta-analysis of the association between breast cancer and pregnancy diagnosed during pregnancy or up to 5 years postpartum on OS. Test for heterogeneity Overall: I 2 = 84.90, p = 0.001, PABC: I 2 = 75.45, p < 0.001, Postpartum: I 2 = 83.67, p < 0.001, Pregnancy: I 2 = 83.11, p ≤ 0.001. Each study is shown by a hazard ratio estimate with the corresponding 95 % CI

Full size image
PABC (pregnant and postpartum)

16 studies comprising of 1111 cases with a total of 6257 individuals were identified for the PABC sub-analysis. This sub-analysis examined where diagnosis occurred either during pregnancy or up to 5 years postpartum. PABC individuals had a worse DFS with a pHR of 1.54 (95 % CI 1.16–2.04) (Fig. 5). There was moderate heterogeneity; however, it was not significant (I 2 = 46.61, p = 0.06). When cases were defined as up to 2 years postpartum, the pHR increased marginally to 1.62 (95 % CI 1.29–2.05) (Table 3). The pooled OR indicated a worse DFS 2.84 (1.70–4.73) with moderate heterogeneity that was not significant (I 2 = 52.38, p = 0.05).

Fig. 5
figure 5

Meta-analysis of the association between breast cancer and pregnancy diagnosed during pregnancy or up to 5 years postpartum on DFS. Test for heterogeneity Overall: I 2 = 60.75, p < 0.001, PABC: I 2 = 46.61, p = 0.06, Postpartum: I 2 = 13.86, p = 0.28. Each study is shown by a hazard ratio estimate with the corresponding 95 % CI

Full size image
Pregnancy

Five studies comprising of 551 cases with a total of 1996 individuals were identified for the pregnancy only sub-analysis. This included studies where diagnosis of breast cancer occurred during pregnancy. There was an increased risk of relapse or disease progression with a pHR of 1.13 (0.69–1.85) (Fig. 5); however, the result was not significant. There was no heterogeneity (I 2 = 0.00, p = 1.00).

Postpartum

Two studies comprising of 195 cases and 1 545 individuals were identified for this sub-analysis. Those diagnosed during the postpartum period up to 5 years following childbirth had an increased risk of relapse or disease progression, HR 1.86 (95 % CI 1.17–2.93) (Fig. 5). There was no heterogeneity (I 2 = 13.86, p = 0.28).

Pregnancy after breast cancer diagnosis

Nineteen studies (10 case control and 9 cohort) comprising of 1828 cases and a total of 23,736 individuals were identified for the meta-analysis. Four studies had data available for both OS and DFS, whilst 14 had data on OS only and 1 had data on DFS only. There was no publication bias as analysed by funnel plot based on Egger’s regression (p = 0.53).

Overall survival (OS)

Seventeen studies (10 case control and 7 cohort) comprising of 1707 cases and a total of 21,448 individuals were identified for sub-analysis. Women who became pregnant after a diagnosis of breast cancer had a significantly reduced risk of death compared to those who did not become pregnant, pHR 0.63 (95 % CI 0.51–0.79) (Fig. 6). There was significant (I 2 = 50.52, p = 0.02). One cohort study [34] with a relatively small sample size was not included in the analysis because calculation of the OR resulted in a zero in a single cell. The pooled OR similarly indicated an improved prognosis for such women, OR 0.65 (95 % CI 0.46–0.92) (Table 3) with no significant heterogeneity (I 2 = 12.98, p = 0.33).

Fig. 6
figure 6

Meta-analysis of the association between pregnancy and breast cancer diagnosed before pregnancy on OS. Test for heterogeneity: Overall: I 2 = 50.52, p = 0.02, Matching: I 2 = 44.54, p = 0.06, Present: I 2 = 73.54, p = 0.05. Each study is shown by a hazard ratio estimate with the corresponding 95 % CI

Full size image
Controlling for the “healthy mother effect”

The “healthy mother effect” [26] is a selection bias where only women who have had favourable outcomes following diagnosis are likely to conceive. Studies control for this bias by matching for nodal status, ER status, disease-free interval and treatment. 12 studies (10 case control and 2 cohort) comprising of 1 387 cases and a total of 17 049 individuals were identified for sub-analysis. When including only those studies that accounted for the “healthy mother effect” bias, women who become pregnant after a diagnosis of breast cancer had a reduced risk of death, pHR 0.65 (95 % CI 0.52–0.81) (Table 3). There was no significant heterogeneity (I 2 = 44.54, p = 0.06). The pooled OR indicated similar findings, although the result was not significant (Table 3).

Disease-free survival (DFS)

Five studies (2 case control and 3 cohort) comprising of 655 cases and a total of 4372 individuals were identified for sub-analysis. There was a decreased risk of recurrence or disease progression amongst women who became pregnant following a diagnosis of breast cancer compared to those who did not become pregnant [pHR 0.84; 95 % CI 0.69–1.02] (Table 3); however, the result was not significant. There was no significant heterogeneity (I 2 = 0.00, p = 0.41).

Controlling for the “healthy mother effect”

Two studies (2 case controls) comprising of 440 cases and a total of 4909 individuals were identified for sub-analysis. There was a decreased risk of recurrence and disease progression for women who became pregnant following diagnosis of breast cancer [pHR 0.93; 95 % CI 0.68–1.28] (Table 3). However, these results were not significant and data were only pooled across two studies. There was no significant heterogeneity (I 2 = 35.12, p = 0.21). There is a large difference in HR between this subgroup and the overall group. The protective effect of pregnancy appears less pronounced in studies that have accounted for the “healthy mother effect” bias as would be expected.

Discussion

Pregnancy before or during breast cancer diagnosis

Our meta-analysis shows that women with PABC up to 5 years postpartum are at an increased risk of mortality and a substantially increased risk of disease recurrence compared to controls. Our subgroup analysis showed that those diagnosed in the postpartum period had a worse prognosis compared to those diagnosed during either pregnancy alone or in the pregnancy or postpartum (PABC) subgroup. Cases diagnosed postpartum are therefore likely to be driving the overall increased risk. Our results are consistent with a previous meta-analysis of 30 studies conducted in 2012 [24]; however, the negative effect of pregnancy on OS appears to be more pronounced in our study overall and within each subgroup. Similarly, we demonstrated that those diagnosed postpartum had poorer outcomes compared to those diagnosed during pregnancy. We demonstrated this same trend for postpartum cases on DFS.

In our study, the results for both overall survival and disease-free survival were highly heterogeneous. This was likely due to significant differences in sample size, the definition of PABC, the definition of a pregnancy, treatment interventions and the follow-up length. Additionally, the studies used different criteria in the matching of cases for case control studies, and different variables were adjusted for where multivariate analysis was completed.

Our results show that PABC individuals are at an increased risk of death. Of the 37 studies included, there were 21 studies that found a negative [35] or null association [2, 5, 11–17, 21, 36–46] between PABC and mortality. In a recent study, Litton et al. [35] showed that women treated with FAC chemotherapy during pregnancy had OS and DFS comparable to, if not better than controls who received the same treatment. This promising result, however, may be due to the small sample size of the study. Callihan et al. [2] did not find a significant result when cases were defined as during pregnancy or up to 1 year postpartum. When cases were defined within 5 years postpartum however, there was an increased risk of both death [HR 2.65; 95 % CI 1.09–6.42] and distant recurrence [HR 2.80; 95 % CI 1.12–6.57] compared to nulliparous controls when adjusted for tumour biological subtype, stage and year of diagnosis. The authors suggest that pregnant and postpartum cases are distinct subsets of PABC and further research should treat them as such in order to clarify whether events subsequent to pregnancy such as breast involution impact breast cancer. They also proposed the definition of PABC should be expanded, with further studies needed to delineate the exact outer limit of the relevant postpartum timeframe. In our analysis, we found there was a significantly increased risk of death in all time frames, but the risk was worse when a more conservative definition of the postpartum period was used. The DFS on the other hand did not meaningfully change with variation in the postpartum period. As such, we found the most relevant period for increased risk for patients is during pregnancy and up to 1 year postpartum; however, an increased risk is evident up to 5 years postpartum.

In contrast, Dimitrakakis et al. [47] found an increased risk of death amongst women with PABC. The study compared 39 cases of PABC up to 1 year postpartum with 39 controls that were matched for stage, age, year of diagnosis and treatment with a 5-year follow-up. The study resulted with a hazard ratio (HR) of 9.28 (95 % CI 2.94–29.27) when adjusted for stage, ER status, grade and age at diagnosis. A well-designed case–control study by Azim et al. [21] reported no difference in OS survival but an inferior DFS (HR 2.3; 95 % CI 1.3–4.2) when comparing 65 pregnant patients with non-pregnant controls after adjustment for age, pT, pN, neoadjuvant chemotherapy, ki-67, HER2 and perivascular invasion. Whilst the exact mechanism is unknown the authors hypothesise the hormonal environment characterising pregnancy may be responsible for pregnancy being a poor prognostic factor for breast cancer. Specifically, it has been postulated that the high levels of growth hormone present in pregnancy may exert an effect on the mammary stem cells that have been shown to transiently overexpress GH receptors during pregnancy [21].

Pregnancy after breast cancer diagnosis

Our meta-analysis shows that women who have had a previous diagnosis of breast cancer and who subsequently became pregnant have a significantly reduced risk of death compared to those who do not become pregnant following diagnosis. The time between diagnosis and pregnancy is clearly a critical issue however, and future studies should determine this time period. These results are consistent with the previous meta-analysis of 14 studies conducted by Azim et al. [25]. There was moderate heterogeneity in our study which resolved after we accounted for the healthy mother effect bias. Our study showed that there was also a decreased risk of recurrence; however, this result was not significant, most likely due to the small sample size.

Our results demonstrate that pregnancy is safe following a breast cancer diagnosis, and indeed associated with an improved prognosis. This result is reassuring for women who have received treatment for breast cancer and are concerned that a pregnancy may worsen their chance of survival. Of the 17 studies included in the analysis on OS, there were 9 studies that found a positive or null association [40, 48–55] between pregnancy following breast cancer and mortality. Kranick et al. [55] found no significant prognostic difference between women who had a pregnancy subsequent to diagnosis and those who did not. A small non-significant adverse effect was found for women who conceived within 12 months of diagnosis. The result was likely due to small sample size. Similarly, a recent study by Valentini et al. [40] of 53 women with BRACA 1/2 mutation did not find that pregnancy following diagnosis adversely affected survival.

In contrast, several studies have demonstrated improved survival outcomes for women conceiving after treatment for breast cancer [26, 27, 56–60]. These findings, however, may be a result of the “healthy mother” effect [26], a selection bias whereby women who have had favourable outcomes are more likely to conceive than those who have relapsed thereby skewing the true effect. To overcome this bias, a study by Azim et al. [27] ensured cases and controls were matched according to disease-free interval, as well as ER status, nodal status, treatment, age and year of diagnosis. The study consisted of a large cohort of 333 patients and found that women who became pregnant after diagnosis experienced better OS (HR 0.72; 95 % CI 0.54–0.97) than the controls who did not, with no interaction according to ER status. Additionally, women who became pregnant within 2 years of diagnosis had increased DFS compared to matched controls, whilst those who became pregnant more than 2 years after diagnosis had comparable outcomes. Interestingly, there was no statistical difference in DFS between patients who had an abortion or miscarriage with their controls. We conducted a subgroup analysis of the studies that accounted for the “healthy mother effect” bias and produced similar results to Azim et al. [25]; however, the protective effect of pregnancy in our study was significant and more pronounced.

In another large study, Mueller et al. [58] found that there was a decreased risk of mortality for women who gave birth 10 months or more after diagnosis compared to controls. Furthermore, Kroman et al. [60] found that women who had a full-time pregnancy after a breast cancer diagnosis had a reduced risk of dying. This result was not modified by age at diagnosis, tumour size, nodal status or pregnancy history before diagnosis. Whilst the exact mechanism of the apparent protective effect of pregnancy is unclear, several hypotheses have been proposed. Janerich [61] suggests that in pregnancies following breast cancer an alloimmunisation against the cancer occurs. This hypothesis speculates that because breast cancer cells and foetal cells share common antigens [62], a mother’s immune system is activated during pregnancy and eliminates not only circulating foetal cells but also quiescent tumour cells, resulting in improved prognosis. A second hypothesis suggests that the substantial increase in oestrogen levels in pregnancy after deprivation may induce apoptosis in oestrogen-responsive breast cancer cells [63].

Strengths and limitations

Our study has several strengths. This is the largest and most current meta-analysis conducted in this field. We performed a comprehensive literature search involving multiple databases, which was not limited by language. We searched unpublished studies and scanned reference lists for further studies. Only studies with a recognised clinical diagnosis of breast cancer were included. There were a large number of studies and there was no publication bias. Study populations were derived from a wide range of countries, increasing the relevance of this study to most populations. Whilst many of the individual studies were unable to detect a significant result due to small sample size, the large number of patients provided us with adequate statistical power to detect an effect. Azim et al. earlier meta-analysis on PABC was restricted to cases occurring up to 1 year postpartum, whilst their study on pregnancy following breast cancer was limited to studies up until 2009 and had a substantially smaller sample size than our study. In contrast, our study adopted a broad definition of the postpartum period up to 5 years following delivery, reported on both groups of women, and contained more recent studies as well as those omitted by previous meta-analyses. We also performed separate analyses based on whether HR was provided or whether OR was calculated from crude data. It therefore includes all the available evidence for women with pregnancy-related breast cancer.

There were also a number of limitations that should be taken into consideration when interpreting our results. Many OR had to be manually calculated, as they were not provided in the study and thus not adjusted for co-variants. Our study was also heterogeneous with regard to study design, particularly in regard to the definition of PABC, the definition of pregnancy, the follow-up time and treatment. Lastly, data were based on published rather than original patient data. It would be useful for future studies to specifically research each subset of PABC, that is, those diagnosed during pregnancy and those diagnosed postpartum in greater detail. Further studies should also ascertain the ‘cut-off’ time of the increased risk that exists for postpartum cases. Previous studies have been unable to assess any difference in survival for women according to oestrogen receptor status with pregnancy. Future studies should perform sensitivity analysis according to oestrogen receptor status to ascertain whether there are any differences in outcomes for those who become pregnant with endocrine-responsive tumours compared to non-responsive tumours. Further investigation is also needed regarding the interval between diagnosis and pregnancy to ascertain the safest period in which to conceive following treatment. It would also be interesting to examine the effect of assisted reproductive technology (ART) on patients with breast cancer diagnosed before and after pregnancy.

Conclusion

Our meta-analyses demonstrated that women who are diagnosed with breast cancer during or after pregnancy are at an increased risk of both death and recurrence compared to those diagnosed with non-pregnancy-related breast cancer. On the other hand, women with a history of breast cancer who subsequently become pregnant have improved survival rates compared to those who do not become pregnant.