We expanded the limited prior literature on endogenous hormones in pregnancy and breast cancer risk by hormone receptor status. We observed an inverse association between early pregnancy progesterone and subsequent risk of ER+/PR+ breast cancer. Further, we observed positive associations between early pregnancy testosterone and free testosterone and breast cancer, predominantly in ER+/PR+ tumors. Early pregnancy estrogens alone were not associated with breast cancer, but high estradiol in the context of low progesterone was associated with higher risk, relative to low concentrations of both hormones. We observed no significant associations between endogenous hormones and ER–/PR– or AR+ disease.
Estrogens and progesterone increase several-fold during pregnancy relative to prepregnant concentrations [15, 16]; these hormones are of placental origin. Estradiol and progesterone have well established roles in breast development [17, 18], and data from animal models suggest that mimicking the hormonal milieu of pregnancy with estradiol and progesterone confers similar protection against breast cancer as is conferred by pregnancy [2]. Progesterone is essential for normal lobular–alveolar development and differentiation in the breast [19]; our study quantified progesterone in early pregnancy, when the breast is undergoing proliferation and the early stages of pregnancy-associated differentiation.
Sex steroid hormones in pregnancy and breast cancer risk in the mother have been investigated previously in three studies nested within two populations [8,9,10]. Peck et al. investigated third-trimester hormones in the Child Health and Development Study (CHDS), observing a suggestive inverse association between progesterone and breast cancer risk (n = 194 cases; OR, extreme deciles: 0.49 (0.2–1.1); p
trend = 0.08). A positive association was observed between estrone and disease risk (OR, extreme deciles: 2.5 (1.0–6.1); p
trend = 0.12). Pregnancy estradiol and estriol were not associated with risk [8]. No association was observed between early pregnancy sex steroids and overall breast cancer in the most recent study in the FMC (n = 1199 cases) [9]. However, estradiol was positively associated with breast cancer diagnosed before age 40 (fourth vs first quartile OR: 1.60 (1.07-2.39)) and suggestively inversely associated with breast cancer diagnosis at age 40 years or older (fourth vs first quartile OR: 0.71 (0.51-1.00); p
het < 0.01). Associations among women younger than age 40 years at diagnosis were only observed for ER–/PR– tumors. In the FMC, progesterone was associated with increased risk of ER–/PR– disease among women diagnosed before age 40 years, but not associated with risk among women aged 40 years or older.
In line with findings from the CHDS, we observed an inverse association between progesterone and breast cancer risk in the current study, although this was restricted to ER+/PR+ tumors. Blood collection in the current study was at median 10 weeks GA, in contrast to mean of 34.5 weeks in the CHDS. Progesterone concentrations are modestly correlated across trimesters of a single pregnancy (Spearman correlations: first and second trimesters, r = 0.63; first and third trimesters, r = 0.39; second and third trimesters, r = 0.64) [16], suggesting that one measure in early pregnancy may not represent late pregnancy concentrations, particularly considering first and third trimester concentrations. Therefore, both early pregnancy progesterone, as measured here, and late pregnancy progesterone, as measured in the CHDS, may impact subsequent breast cancer risk.
We observed no association between estradiol and breast cancer overall risk in the current study, with the exception of an increased risk of disease in women with relatively high estradiol (above median) and low progesterone (below median), as compared to women with low concentrations of both hormones; this increase in risk was evident for both hormone receptor-positive and receptor-negative disease. Experimental data from animal models suggest that both estradiol and progesterone may be necessary to induce the long-term protective effect of pregnancy, although results differed based on the experimental model [2, 20]. The association between breast cancer and cross-classified estradiol and progesterone was not described in the previous investigations in pregnant women, nor, to our knowledge, in epidemiologic studies in premenopausal women. However, high concentrations of circulating endogenous estrogens after menopause—a period characterized by physiologically low circulating progesterone concentrations— are consistently associated with increased risk of breast cancer [5,6,7].
Given the divergent association between estradiol and breast cancer risk in analyses stratified by age at diagnosis in the FMC, we evaluated risk stratified by age at diagnosis in this study (<45 vs ≥45 years). Results were similar in both age groups for estrogens and progesterone, while testosterone was more strongly associated with breast cancer diagnosed at age 45 years or older. Given the age distribution in our cohort, we were unable to evaluate risk using the same age thresholds as the FMC (i.e., only 30 cases in our population were diagnosed prior to age 40 years). The FMC population was somewhat older at first birth than the NSMC study population and due to technical considerations (i.e., restriction of the study population to women providing blood samples prior to 1988 and longer follow-up), women in the NSMC were diagnosed at an older age (median age at diagnosis: FMC = 41.2 years; NSMC = 46.7 years) and after longer lag-time between pregnancy and cancer diagnosis (median lag-time: FMC = 10.9 years; NSMC = 19.8 years). Therefore, our findings from the NSMC may pertain to the long-term impact of early pregnancy hormones and breast cancer risk, whereas the FMC results may better describe risk associated with more proximate exposure to pregnancy hormones.
In contrast to estrogens and progesterone, androgens increase gradually across pregnancy, approximately doubling from preconception to the third trimester [15, 16]. Androgens are produced by the ovary and maternal adrenal cortex as well as the adrenal glands and liver of the fetus [21]. Androgens are relatively stable from prepregnancy to early pregnancy, and the androgens quantified in our study are likely representative of circulating premenopausal androgen concentrations. Epidemiologic data consistently show a positive association between androgens and breast cancer risk, in both premenopausal [3, 4] and postmenopausal [5,6,7] women. This may be due to a direct androgen effect, or may be a result of conversion of androgens to estrogens in breast tissue via aromatase; aromatase is expressed in both normal and malignant tissue [22].
We observed higher risk of ER+/PR+ breast cancer risk with higher circulating testosterone concentrations in the current study. In the only prior study on androgens in pregnancy and breast cancer risk, in the FMC [9], testosterone was positively associated with risk of ER–/PR– tumors in the subgroups of women diagnosed younger than age 40 or with first birth younger than age 30; testosterone was not associated with breast cancer overall or in the ER+/PR+ subgroup. As with estrogens and progesterone, the divergent findings between the current study and results from the FMC may be due to the different age distributions and interval between pregnancy and breast cancer diagnosis in the two study populations.
To our knowledge, our study is the first investigation of early pregnancy hormones and maternal breast cancer risk by androgen receptor status. Experimental data suggest that crosstalk between the ER and AR results in impeded receptor signaling of both receptors, thus inhibiting hormone-related growth and proliferation [23]. Further, epidemiologic data show ER+/AR+ tumors have better prognosis that ER+/AR– tumors [23]. We observed a positive association between testosterone and ER+/PR+ disease in all cases. The associations between testosterone and ER+/PR+/AR+ and ER+/PR+ breast cancer risk, among the subset of women with AR data, were similar (among women with AR data: ER+/PR+ ORlog2: 1.32; ER+/PR+/AR+ ORlog2: 1.36), suggesting the AR may not play an important role in this context.
Our study has important strengths and limitations. Blood samples were collected and stored using standardized procedures. However, samples are stored at a relatively warm temperature (−20 °C). Estradiol and SHBG concentrations were weakly correlated with storage time (estradiol: r = 0.19, p < 0.01; SHBG: r = −0.20, p < 0.01), as were free estradiol (r = 0.19, p < 0.01) and free testosterone (r = 0.13, p = 0.01). Cases and controls were carefully matched for date of blood collection, therefore the weak correlations observed between hormone concentrations and storage time should not impact our results. Hormones change systematically in early pregnancy with GA. We accounted for these changes by adjusting for GA in regression models. An alternative approach would be to use regression residuals. In our study, results adjusting for GA were similar to those using regression residuals. Finally, sample size was limited for analyses of ER–/PR– tumors, and AR data were only available for a subset of cases.