Breast cancer is a complex and heterogeneous disease with a variety of histo-pathological and molecular subtypes with diverse clinical outcomes and relationships with established risk factors 1 2 3 . The major sub-classification of clinical breast tumors is based on the detection of estrogen (ER) and progesterone (PR) receptors and guides targeted therapies and provides important prognostic information 4 . The presence or absence of hormone receptors, along with human epidermal growth factor-2 (HER2) also broadly correspond to more detailed molecular subclassification of breast tumors, as determined by microarray-based gene expression profiling coupled to hierarchical clustering analyses 5 6 7 . In addition to the clinical use of ER and PR, epidemiological data indicate that the association of reproductive history with breast cancer differs by the expression of ER and PR receptors 2 .

Factors that influence the lifetime cumulative exposure to hormones during reproductive life, such as the age at menarche, age at first child-birth, time between age at menarche and first child birth, number of children, use of oral contraceptives (OC) and breastfeeding, have been suggested to be associated with risk of hormone receptor (HR)-positive malignancies (ER-positive or joint ER+PR+) 2 8 9 10 11 . However, distinct risk factors for HR-negative (ER-negative, or joint ER-PR) cancer are debated 2 3 8 12 and the etiologies of ER+PR+ and ER-PR- tumors remain unclear.

The incidence of HR-negative disease drops remarkably after menopause 13 suggesting that ovarian derived sex steroid hormones do have an impact on HR-negative tumors and recent studies are starting to show risk associations of reproductive factors with HR-negative malignancies 2 3 . In fact, opposite risk associations between ER-PR- and ER+PR+ tumors have been observed with parity 11 14 , age at first pregnancy 9 and breastfeeding 11 . Nonetheless, due to the rarity and heterogeneous nature of HR-negative breast tumors, epidemiological studies have been hindered by small sample sizes resulting in inconsistent risk associations between reproductive factors and HR-negative disease 8 11 15 16 .

The incidence of HR-negative disease drops remarkably after menopause 13 suggesting that ovarian derived sex steroid hormones do have an impact on HR-negative tumors and recent studies are starting to show risk associations of reproductive factors with HR-negative malignancies 2 3 . In fact, opposite risk associations between ER-PR- and ER+PR+ tumors have been observed with parity 11 14 , age at first pregnancy 9 and breastfeeding 11 . Nonetheless, due to the rarity and heterogeneous nature of HR-negative breast tumors, epidemiological studies have been hindered by small sample sizes resulting in inconsistent risk associations between reproductive factors and HR-negative disease 3 8 14 15 .

The European Prospective Investigation into Cancer and Nutrition (EPIC) is a multi-center prospective cohort study designed to investigate the relationships between diet, nutrition and metabolic factors and cancer, consisting of approximately 360,000 women and 150,000 men aged mostly between 25–70 years 16 17 . All participants were enrolled between 1992 and 2000 and came from 23 regional and national research centers located in 10 western European countries: Denmark, France, Italy, Germany, Greece Norway, Spain, Sweden, The Netherlands and the United Kingdom. Extensive details about the standardized procedures for recruitment, measuring baseline anthropometry, questionnaires on current habitual diet, reproductive and menstrual history, exogenous hormone use [OC and hormone replacement therapy (HRT) use], medical history, lifetime smoking and alcohol consumption history, occupation, level of education and physical activity and biological sample collection at study centers are given elsewhere 16 17 . All subjects gave written informed consent. The Internal Review Boards of the International Agency for Research on Cancer and the local ethics committees in participating countries approved the analyses based on EPIC participants.

A total cohort of 311,097 women was followed for a sum of 3,346,356 person years. Women were recruited into the EPIC study at the median age of 51.1 years (Table 1). At the time of recruitment, 46.5% of the women were postmenopausal and the median age of menopause was 50.0 years. For parous women, the median age at first full-term birth across the EPIC centers was 24.0 and the median time between menarche and first child birth was 11.0 years. The median age at last full-term pregnancy was 29.0 years, and a median time of 22.9 years had passed since the last childbirth. Of the women who had breastfed (83.8%), the median time of breastfeeding was 6.0 months for all pregnancies.

¹In natural and surgically menopausal women only.

²In parous women only.

³In women who breast-fed only.

⁴Both spontaneous and induced abortions.

⁵In women who ever used OC.

A statistically significant heterogeneity between the risk of ER-PR- (n = 998) and ER+PR+ (n = 3,567) tumors was observed for age at first full-term childbirth (Table 2). While Cox regression models showed no association with risk of ER-PR-malignancies, a first full-term childbirth after the age of 35 was associated with an increased risk of ER+PR+ tumors (≥35 vs. ≤19 years HR: 1.47 [95% CI 1.15-1.88] p _trend < 0.001, P _het = 0.03).

¹Stratified by age at recruitment and center and further adjusted for BMI, height, menopausal status at enrolment, HRT use, physical activity, smoking status, alcohol consumption and attained level of education; ²heterogeneity between ER+PR+ and ER-PR- tumors was assessed on the trend score using the data augmentation method as described by Lunn and McNeil; ³in postmenopausal women only; ⁴in parous women only; ⁵in women who breast-fed only; ⁶in both spontaneous and induced abortions; ⁷heterogeneity between ER+PR+ and ER-PR- tumors was assessed on the unordered categorical variable of never, past and current OC use using the data augmentation method as described by Lunn and McNeil; ⁸in women who ever used OC.

For ER-PR- breast tumors, similar risk associations to ER+PR+ breast tumors were observed with increasing menarcheal age (≥15 vs. ≤ 13 years ER-PR- HR: 0.84 [95% CI 0.69-1.03], P _trend =0.17; ER+PR+ HR: 0.76 [95% CI 0.68-0.85], P _trend <0.001; P _het = 0.48), and among parous women, with longer time between menarche and first full-term childbirth (≥10 vs. < 10 years ER-PR- HR: 1.15 [95% CI 0.99-1.34], P _trend =0.09; ER+PR+ HR: 1.22 [95% CI 1.12-1.33], P _trend <0.001; P _het = 0.52). Although the relative risk estimates for ER-PR- breast malignancies were in a similar direction to the ER+PR+ tumors and no statistical heterogeneity between the breast cancer subtypes was observed, it should be noted that risk associations for ER-PR- malignancies were weaker in magnitude and failed to reach statistical significance.

Ever a full-term childbirth, number of childbirths, age- and time since last full-term childbirth were generally not associated with the risk of ER-PR- malignancies but were associated with ER+PR+ tumors, however no statistical heterogeneity between the breast cancer subtypes was observed. Moreover, OC use, duration of OC use, breast feeding, ever having an abortion (spontaneous or induced) and age at menopause showed no significant associations with either ER-PR- or ER+PR+ breast cancer.

When all of the pregnancy related variables were examined together in the same model, risk associations for an increased number of full-term births, and later first and last full-term childbirth with ER-PR- tumors appeared to slightly strengthen; however they remained statistically not significant and had wider confidence intervals (Table 3). When analyses were restricted to postmenopausal women, the risk estimates for a later age at first full-term childbirth with ER+PR+ tumors were stronger (≥35 vs. ≤19 years HR: 1.64 [95% CI 1.05-2.56] p _trend = 0.002) compared to the estimates for all the women combined (≥35 vs. ≤19 years HR: 1.30 [95% CI 0.95-1.79] p _trend = 0.02).

¹Stratified by age at recruitment and center and further adjusted for BMI, height, menopausal status at enrolment, HRT use, physical activity, smoking status, alcohol consumption and attained level of education; ²stratified by age at recruitment and center and further adjusted for BMI, height, HRT use, physical activity, smoking status, alcohol consumption and attained level of education; ³mutually adjusted for the pregnancy related variables in this table; ⁴heterogeneity between ER+PR+ and ER-PR- tumors was assessed on the trend score using the data augmentation method as described by Lunn and McNeil.

Subgroup analyses showed no heterogeneity in most of the risk estimates of the reproductive variables for ER-PR- and ER+PR+ tumors (data not shown), with the exception of the duration of breastfeeding by age at first birth among parous women (P_{heterogeneity} = 0.002). Among women who had their first full-term childbirth before the age of 25, a longer duration of total breastfeeding showed an indication of a decreased risk association with ER+PR+ tumors (≥18vs. ≤ 1 month HR: 0.86 [95% CI 0.67-1.10], P_trend = 0.01). In contrast, among women who had their first full-term childbirth after the age of 25, a longer cumulative duration of breastfeeding was associated with a particular increased risk of ER+PR+ tumors (≥18vs. ≤ 1 month HR: 1.50 [95% CI 1. 13–1.99], P_trend = 0.005). For ER-PR- tumors, risk estimates in the similar direction to the ER+PR+ tumors were observed, however statistically not significant (first childbirth before the age of 25, ≥18vs. ≤ 1 month of breast feeding HR: 0.90 [95% CI 0.58-1.39], P_trend = 0.89; first childbirth after the age of 25, ≥18vs. ≤ 1 month of breast feeding HR: 1.31 [95% CI 0.72-2.39], P_trend = 0.41).

In the sensitivity analysis restricted to cases with any indication of a positive ER and PR expression versus a complete absence of ER and PR expression with reproductive factors showed similar patterns with risk of joint ER+PR+ and ER-PR- breast cancer subtypes (data not shown).

Our results showed a significantly heterogeneous risk association for age at first full-term pregnancy by receptor status of the tumor, where later first full-term childbirth was associated with increased risk of ER+PR+ tumors but not with risk of ER-PR- tumors. Menarcheal age and time period between menarche and first full-term childbirth showed suggestively similar risk associations with both ER-PR- and ER+PR+ tumors, however the risk estimates for ER-PR- tumors were generally weaker than for their ER+PR+ counterparts and only borderline significant. Although the heterogeneity between breast cancer subtypes was not statistically significant, other parity related factors (such as ever having a full-term childbirth, number of full-term childbirths, age- and time since last full-term childbirth) were associated only with ER+PR+ malignancies. Finally, the factors related to breastfeeding and OC use were generally not associated with HR-positive or HR-negative breast cancer risk.

Previous prospective studies investigating the association of reproductive factors with HR-positive breast cancer have shown relatively consistent inverse risk associations with increasing menarcheal age, ever having a full-term childbirth and particularly a full-term childbirth at an early age 3 9 14 25 . However, consensus on the associations with HR-negative tumors has not been reached because previous prospective studies have lacked sufficient sample sizes 3 9 26 and because of the heterogeneous nature of HR-negative subtypes 1 . A more recent study within the Women’s Health Initiative 14 showed that ever having a full-term childbirth was associated with an increased risk of triple-negative breast cancer (ER-, PR- and HER2-) (n = 307) and the positive association was strengthened with an increasing number of full-term births. We were unable to confirm the positive risk association with ever having a full-term childbirth and increasing number of full-term pregnancies with HR-negative tumors, however we did not have information on triple negative tumors.

We observed for both ER-PR- and ER+PR+ tumors, similar risk associations with increasing menarcheal age and a longer time between menarche and first full-term childbirth. A recent study within the EPIC cohort previously reported on the association of menarche with both ER-PR- and ER+PR+ tumors within the context of childhood growth and earlier sexual maturity 27 . This study extends onto this study of menarche and growth and focuses on the time period between sexual maturity and first full-term pregnancy, thus illustrating the complex and entwined nature of endocrine-sensitive tumors with hormones during different life phases of growth, sexual maturity and reproduction. The inverse association of menarcheal age with increased breast cancer risk is thought to be resultant of a longer exposure to estrogens during a women’s reproductive life 2 but may also reflect early pubertal years characterized by more intensive and increased exposure to estrogen 27 28 . Estrogens have been long established to have a late-stage growth promoting effect on estrogen sensitive tumors 29 , however, evidence suggests that estrogens may also play an important role in earlier developmental stages of both HR-positive and -negative tumor types 30 . Mammary stem cells have been shown to respond to sex steroid hormones without having a clear expression of an ER or PR 31 . Further, the EPIC cohort, showed that pre-diagnostic levels of estrogens were associated with both HR-negative and HR-positive postmenopausal breast cancer 20 . The longer time between menarche and a women’s first full-term childbirth would equate to a longer period of time with undifferentiated breast epithelial tissue and a shorter period of time that the breast is resistant to malignant transformation 26 and thus may have etiological importance in the formation of ER-PR- tumors as well.

We observed a significantly different risk association for a later age at first full-term childbirth with risk of HR-negative and HR-positive tumors, whereas the associations for parity related factors (such as ever having a full-term childbirth, age a last full-term childbirth and time since last full-term childbirth) appeared to aggregate around HR-positive tumors. The role of a pregnancy with the risk of breast cancer is thought to stem from two major avenues, firstly, hormonal changes before and after pregnancy and secondly 26 , dramatic structural changes in the ductal system of the breast after pregnancy 26 32 . A full-term childbirth is associated with a long term post-pregnancy reduction in levels of circulating hormones 26 . Before a women’s first pregnancy the breast contains a high proportion of undifferentiated ducts and associated alveolar buds. Complete differentiation only occurs during pregnancy and lactation via complex morphological, physiological, and molecular changes 32 . Terminally differentiated epithelial cells are more resistant to carcinogenic influences because of lower proliferation rates and longer DNA repair phases 26 . The distinct inverse risk association for an earlier age at first full-term childbirth with ER+PR+ disease could be due to a shorter exposure to higher levels of ovarian estrogens and a shorter period of time of undifferentiated breast epithelial cells.

Recent prospective studies have reported reduced risk associations with breastfeeding with both ER-PR- and ER+PR+ breast cancer 3 14 . In the current analysis, we also observed an inverse risk association for both ER-PR- and ER+PR+ malignancies with a longer cumulative duration of breastfeeding however, this was restricted to women who had an early full-term childbirth. In contrast to the recent studies, among women who had a later first full-term childbirth, an increased risk with ER-PR- and ER+PR+ breast tumors with a longer total duration of breastfeeding was observed. Breastfeeding is thought to protect a woman from developing breast cancer by increasing breast differentiation, postponing the return of the ovulatory menstrual cycle post-pregnancy, and/or changing the hormonal environment of the breast 9 26 32 . The inverse risk association of HR-negative and HR-positive tumors with breastfeeding coupled with a longer duration among young first time mothers could convey a similar protection. In addition, lactation at a younger age would also mean a shorter period of undifferentiated breast epithelial tissue 26 .

OC use has been extensively studied by many epidemiological studies and most studies have found either no association or a moderate increased risk of overall breast cancer, particularly among very young women and recent OC users 26 . More recent case–control studies investigating the risk of HR-defined breast cancer have started showing relationships of OC use with HR-negative breast cancer 33 34 35 36 37 . In the current study, we were unable to confirm significant risk associations of ever OC use and a longer duration of OC use with HR-negative tumors. There were only a small number of baseline OC users within this analytical cohort, and unfortunately within the EPIC cohort, information of dose or type of OC used, date of last use and information on changes to OC use after baseline are not available. Similarly, the estrogens and progestins in oral contraceptives differ in type and concentration 37 and this could be hiding a risk association.

Major strengths of this study are its prospective design and large number of incident cases with receptor information. The large case numbers allowed an in-depth analysis of reproductive-related relative risks, describing risk associations among women of predominantly premenopausal and postmenopausal age. The large case numbers also enabled us to examine subgroup effects such as age at first birth, age at diagnosis, BMI and breastfeeding. To our knowledge, this analysis uses the largest number of incident cases of ER-PR- malignancies, although future prospective studies with a greater number of ER-PR- cases are necessary to characterize the associations, which are of substantially smaller magnitude when compared to their HR-positive counterparts. Our study does have its limitations. The determination of ER and PR status in breast tumors has become a standard part of breast cancer diagnosis and is used to predict endocrine therapy response 4 . While a number of studies have shown that the classification of the ER and PR in tumors is relatively robust 38 39 , the accuracy of classifying an ER or PR-negative tumor remains controversial 40 41 . In the analysis that compared women with a complete absence of ER and PR expression to women with any indication of an ER and PR positive expression, ER-PR- risk estimates remained unchanged. Furthermore, proportions of ER and PR–negative tumors in the EPIC cohort are in line with previous reports 13 42 . In addition, the inclusion of PR provides an indication of a functional estrogen pathway 2 and thus a joint ER-PR- may be more reflective of a true ER-negative tumor. At the time of this study, additional information on HER2 expression to determine breast cancer subtypes into more detailed molecular sub-classifications could not be completed because of insufficient information on HER2. However, as the routine assessment of HER2 is relatively more recent than ER and PR assessment, future cohort analyses will be able to include HER2.

In conclusion, our study provides evidence that later age at first full-term childbirth is associated with an increased risk of ER+PR+ tumors but not with ER-PR- tumors. Moreover, age at menarche and time between menarche and first full-term childbirth may be associated with the etiology of both HR-negative and HR-positive malignancies, although associations were only borderline significant for HR-negative tumors. Further studies with more incident cases of ER-PR- tumors are needed to provide more precise risk estimation for reproductive factors with HR-negative tumors.

EPIC: European Prospective Investigation into Cancer and Nutrition; ER: Estrogen receptor; PR: Progesterone receptor; HR-positive: Hormone receptor-positive; HR-negative: Hormone receptor-negative; OC: Oral contraception; HR: Hazard ratio; 95% CI: 95% confidence interval; HER2: Human epidermal growth factor-2; BMI: Body mass index.

The authors declare they have no competing interests.

RR, AL, and RK contributed to the conception of the current analysis and all authors were involved in the design and acquisition of data from the EPIC cohort. RR, AL and RK contributed to the analysis and all authors contributed to the interpretation of the data. RR, KT, AL and RK drafted the manuscript and all authors revised the final draft critically for important critical content. All authors have given final approval of the version to be published.