Many reproductive factors consistently associated with breast cancer risk, such as age at menarche, parity and age at first birth, likely affect risk through a hormonal pathway.1 The peptide hormone prolactin, essential in mammary development and lactation, is associated with an increased risk of breast cancer in both premenopausal2 (Tworoger et al., 2006, submitted) and postmenopausal3, 4, 5, 6 women, likely through an increase in cell proliferation and inhibition of apoptosis.7 Thus, prolactin may be one of potentially many hormonal links between reproductive risk factors and the incidence of breast cancer.7, 8
While the association between prolactin levels and breast cancer risk is fairly consistent, whether lifestyle or reproductive factors influence prolactin levels is not well known. A number of cross-sectional studies have observed an inverse association between parity and prolactin levels in both premenopausal9, 10, 11, 12, 13 and postmenopausal9, 14 women, although whether prolactin levels decline further after the first pregnancy is unclear. Other birth-related factors, such as age at first birth, time since last birth and history of lactation, have been assessed in only a few studies, with no clear associations.11, 12, 13, 14, 15 Few studies have examined these associations in postmenopausal women.9, 14
A prior analysis of the correlations between breast cancer risk factors and prolactin levels was conducted among 217 postmenopausal women within the Nurses' Health Study (NHS).16 The current analysis expands upon this preliminary assessment and examines the associations of reproductive factors and family history of breast cancer with plasma prolactin levels in 1,089 premenopausal and 1,311 postmenopausal women in a cross-sectional study within the NHS and the NHSII.
The NHS began in 1976 when 121,701 female, married registered nurses, aged 30–55 years, completed and returned a mailed questionnaire. The NHSII follows the same format as the original cohort, and began in 1989 with 116,609 female nurses, aged 25–42 years. Both cohorts continue to be followed every 2 years by questionnaire to update exposure status and ascertain newly diagnosed disease. Data have been collected on various reproductive and lifestyle factors, including age at menarche, parity, ages at first and last birth, lactation history, postmenopausal hormone (PMH) use and family history of breast cancer.
During 1989 and 1990, blood samples were collected from 32,826 NHS cohort members who were 43–69 years of age. Details regarding the NHS blood collection methods have been published previously.17 Briefly, each woman arranged to have her blood drawn and shipped, via overnight courier with an ice-pack, to our laboratory, where it was processed and separated into plasma, red blood cell and white blood cell components. Between 1996 and 1999, blood samples were collected from 29,611 NHSII participants; details regarding the NHSII blood collection methods have been published previously.2 Briefly, participants were sent a short questionnaire and a blood collection kit containing necessary supplies to have blood samples drawn by a local laboratory or a colleague. Premenopausal women who had not taken oral contraceptives, been pregnant or breastfed within 6 months (n = 18,521) provided blood samples drawn on the 3rd to 5th day of their menstrual cycle (follicular draw) and 7–9 days before the anticipated start of their next cycle (luteal draw). Follicular plasma was aliquotted by the participants 8–24 hr after collection and frozen; luteal samples were shipped to our laboratory and processed similarly to the NHS samples. The remaining women (n = 11,090) provided a single 30-mL blood sample. These women arranged to have 1 blood sample drawn and shipped, via overnight courier and with an icepack, to our laboratory where it was processed similarly to the NHS samples. The stability of prolactin in whole blood not processed for 24–48 hr has been shown previously.18 All samples have been stored in continuously monitored liquid nitrogen freezers since collection. Informed consent was implied by receipt of completed questionnaires and blood samples. The study was approved by the Committee on the Use of Human Subjects in Research at Harvard School of Public Health and Brigham and Women's Hospital.
Women included in this analysis were controls from nested case-control studies of plasma prolactin levels and breast cancer risk (n = 2,332)2, 3 (Tworoger et al., 2006, submitted), or participants in hormone reproducibility studies (n = 191).19, 20 Postmenopausal women from the NHS cohort were defined by natural menopause (no menses for at least 12 months prior to blood collection), bilateral oophorectomy or hysterectomy with 1 or both ovaries remaining if they were at least 56 years old (if a non-smoker) or 54 years old (if a current smoker), ages at which natural menopause had occurred in 90% of these groups in the overall cohort. Premenopausal women, defined as such if the woman reported that her periods had not ceased or had a hysterectomy but had at least 1 ovary remaining and was ≤47 (for nonsmokers) or ≤45 (for smokers) years of age, were included from both cohorts. Women were excluded if they were missing any of the predictors of interest (n = 80).
Prolactin was measured using a microparticle enzyme immunoassay. Samples were initially assayed in 4 batches between 1993 and 1997 at the University of Massachusetts Medical Center (Worcester) using IMx System (Abbott Laboratory, Abbott Park, IL). The remaining samples were assayed in 7 batches between 2001 and 2006 at the Reproductive Endocrinology Unit Laboratory at the Massachusetts General Hospital, using the AxSYM Immunoassay system (Abbott Diagnostics, Chicago, IL). A subset of 60 samples was assayed at both laboratories; the correlation between the 2 laboratories was 0.91. The limit of detection (for both laboratories) was 0.6 ng/mL. Because of low blood volume or technical difficulties with the assay, some prolactin values were missing (n = 35). Intra-assay coefficients of variation all were less than 15%.
Predictors and covariates
Information on predictors and covariates were obtained from biennial questionnaires within the 2 cohorts. In the NHS, age at menarche and height were queried at baseline in 1976. Age at first birth and parity were assessed in 1976 and updated biennially until 1984. Lactation history was queried in 1986. Time since last birth was calculated as the difference between age at blood draw and age at last birth, which was calculated from reported births. Family history of breast cancer (defined as breast cancer in the woman's mother or any of her sisters) was queried in 1976, 1982 and 1988. History of benign breast disease, age at menopause and postmenopausal hormone (PMH) use were assessed biennially. Current weight and medication use were collected on the blood questionnaire. For NHSII participants, information on age at menarche and height were collected at baseline in 1989. Parity and history of benign breast disease were assessed biennially; ages at first and last birth were reported in 1995. Lactation history was queried in 1997. Family history of breast cancer was assessed in 1989 and 1997. Current weight and medication use were collected on the blood questionnaire.
For premenopausal women with a follicular and a luteal blood sample (n = 479), we used the average of the 2 phases in our analyses because prolactin levels do not vary substantially by menstrual phase, and the average of follicular and luteal samples better represents long-term levels.20, 21 We identified statistical outliers using the generalized extreme studentized deviate many-outlier approach22; premenopausal women in the NHS with prolactin >99 ng/mL (n = 2) and postmenopausal women with prolactin <1 or >83 ng/mL were excluded (n = 6); no outliers were detected in the NHSII premenopausal women. Our final sample size included 1,089 premenopausal (528 from NHS and 561 from NHSII) and 1,311 postmenopausal (all from NHS) women.
We calculated adjusted geometric mean hormone levels across categories of predictive factors using generalized linear models, regressing natural log-transformed prolactin levels on exposures and covariates and adding the mean hormone level to the average of the residuals; results were then exponentiated and reported on the raw scale. p-values were calculated by the F-test, using continuous variables (e.g., parity) or the contrast between 2 categories (e.g., family history yes/no).
Mean plasma prolactin concentrations from quality control samples run in multiple batches differed by batch, indicating that there was some laboratory drift over time. Therefore, an indicator for laboratory batch was included in all analyses. In addition, all analyses were adjusted for age, date, time of day and fasting status at blood collection. Multivariate analyses included all of the predictors of interest; analyses among postmenopausal women also adjusted for PMH use and age at menopause. All analyses were conducted using SAS software, version 9 (SAS Institute, Cary, NC).
Mean prolactin levels were higher among premenopausal than postmenopausal women (Table I). Premenopausal women had fewer children and a slightly later age at first birth, and were more likely to have breastfed their children longer, all likely due to secular trends in reproductive practices given that the 2 studies represent different birth cohorts (birth years 1921–1946 in the NHS, 1946–1964 in the NHSII). Age at menarche and body mass index were similar between the 2 groups, while postmenopausal women were more likely to have a history of benign breast disease and a family history of breast cancer.
Table I. Characteristics of Pre- and Postmenopausal Women in the Nurses' Health Study and Nurses' Health Study II
|Prolactin (mean, SD) (ng/mL)1||16.3 (9.3)||10.3 (5.9)|
|Age (mean, SD) (yr)||46.0 (4.4)||60.8 (5.0)|
|Age at menarche (mean, SD) (yrs)||12.4 (1.3)||12.6 (1.4)|
|Parity (mean, SD)2||2.6 (1.1)||3.4 (1.6)|
|Age at first birth (mean, SD) (yrs)2||25.5 (4.0)||25.1 (3.3)|
|Age at last birth (mean, SD) (yrs)2||30.1 (4.3)||31.6 (4.5)|
|Time since first birth (mean, SD) (yrs)2||20.7 (6.4)||35.7 (5.5)|
|BMI (mean, SD) (kg/m2)||25.3 (5.3)||25.6 (4.6)|
|BMI at age 18 (mean, SD) (kg/m2)||21.3 (2.7)||21.5 (2.9)|
|Breast fed infant ≥12 months (%)2||35.6||18.3|
|Age at menopause (mean, SD) (yrs)3||–||49.1 (4.8)|
|Current PMH use (%)||–||34.9|
|Family history of breast cancer (%)||8.6||12.4|
|History of benign breast disease (%)||28.3||34.0|
Premenopausal parous women had significantly lower prolactin levels than nulliparous women (multivariate-adjusted geometric means = 14.1 ng/mL vs. 16.6 ng/mL, p < 0.001) (Table II). The association appeared to be driven by having at least 1 pregnancy, as prolactin levels did not decrease with increasing parity (p = 0.23). Parity was modestly but significantly associated with lower prolactin levels among postmenopausal women (nulliparous vs. parous mean = 10.1 ng/mL vs. 9.1 ng/mL, p = 0.04). Although levels appeared to decline with increasing parity up to ≥4 children (p = 0.07), a linear trend was not observed when parity was extended to ≥6 children (data not shown).
Table II. Geometric Mean Prolactin Levels (ng/mL) Across Birth Variables, According to Menopausal Status
|Parity|| || || || || || |
|p-value|| ||<0.001||<0.001|| ||0.04||0.04|
|p-value|| ||0.001||0.002|| ||0.01||0.01|
|p-value (parous)|| ||0.34||0.23|| ||0.11||0.07|
|Age at first birth|| || || || || || |
| <25 years||439||14.5||14.6||628||8.9||8.9|
| 25–29 years||372||13.8||13.7||474||9.4||9.4|
| ≥30 years||147||13.8||13.9||124||9.1||9.1|
|p-value (parous)|| ||0.41||0.28|| ||0.50||0.69|
|Time since first birth|| || || || || || |
| <10 years||62||12.8||12.7|| || || |
| 10–14 years||124||13.4||13.4|| || || |
| 15–19 years||214||13.8||13.7||6||9.2||9.1|
| 20–24 years||315||14.7||14.8||44||9.9||10.0|
| 25–29 years||184||14.3||14.4||141||9.2||9.3|
| 30–34 years||55||14.8||14.8||317||9.1||9.1|
| 35–39 years||4||14.5||15.1||416||8.9||8.9|
| ≥40 years|| || || ||302||9.1||9.1|
|p-value (parous)|| ||0.19||0.12|| ||0.55||0.88|
Prolactin levels did not vary by age at first birth among parous premenopausal (p = 0.28) or postmenopausal (p = 0.69) women (Table II). Prolactin levels among premenopausal women tended to increase with increasing time since first birth (p = 0.12); after 20 years since first birth, the difference in prolactin levels between parous and nulliparous women was smaller and closer to the differences observed among postmenopausal women. Similar, though weaker, associations were observed across years since last birth (data not shown). Compared with premenopausal women, postmenopausal women generally were further from their first birth (mean 20.7 vs. 35.7 years) and prolactin levels did not vary by time since first birth (p = 0.88) or time since last birth (data not shown).
Prolactin levels did not vary significantly by age at menarche among either premenopausal or postmenopausal women (p = 0.46, 0.15, respectively) (Table III). Duration of lactation also was not associated with prolactin levels among parous premenopausal (p = 0.95) or postmenopausal (p = 0.31) women. Similarly, prolactin levels did not vary by history of benign breast disease in either group (premenopausal p = 0.77; postmenopausal p = 0.99). Among premenopausal women, those with a family history of breast cancer had significantly higher prolactin levels than those without (15.9 ng/mL vs. 14.3 ng/mL, p = 0.04). However, prolactin levels were similar among postmenopausal women with and without a family history (9.1 ng/mL vs. 9.3 ng/mL, p = 0.73).
Table III. Geometric Mean Prolactin Levels (ng/mL) Across Menarche, Lactation, Benign Breast Disease and Family History, According to Menopausal Status
|Age at menarche|| || || || || || |
| <12 years||253||14.7||14.8||273||9.0||8.9|
| 12 years||312||13.6||13.6||350||9.1||9.1|
| 13 years||337||14.7||14.7||400||9.1||9.1|
| 14+ years||187||14.9||14.9||288||9.4||9.5|
|p-value|| ||0.40||0.46|| ||0.21||0.15|
|Lactation history|| || || || || || |
| Never (parous)||179||14.9||15.0||407||9.0||9.0|
| ≤3 months||163||14.7||14.7||346||9.0||9.0|
| 4–11 months||228||13.9||13.9||249||9.0||9.0|
| 12–23 months||195||13.1||13.2||158||9.6||9.6|
| 24+ months||193||14.1||14.1||66||9.0||9.0|
|p-value (parous)|| ||0.46||0.95|| ||0.53||0.31|
|Benign breast disease|| || || || || || |
|p-value|| ||0.65||0.77|| ||0.64||0.99|
|Family history|| || || || || || |
|p-value|| ||0.02||0.04|| ||0.92||0.73|
Because prolactin levels have a diurnal variation and increase within 30 min of a meal,23, 24 we repeated our analyses restricted to women who gave morning, fasting blood samples; results were essentially unchanged. Our results also did not change substantially when we excluded women using antidepressants or thyroid hormones, which can alter prolactin levels, at the time of blood collection (n = 141 premenopausal, 232 postmenopausal). Among premenopausal women with timed samples, associations were similar when analyzed separately by follicular and luteal phase. Results also were essentially unchanged when nulliparous women reporting female-factor infertility were excluded (n = 21 premenopausal, 35 postmenopausal).
In this large, cross-sectional study we observed higher prolactin levels among nulliparous women compared with parous women in both premenopausal and postmenopausal women, although differences were more marked among premenopausal women. Number of children and age at first birth were not associated with prolactin levels among parous women, regardless of menopausal status, but levels appeared to increase with time since first birth among premenopausal women. Age at menarche, duration of lactation and benign breast disease were not associated with prolactin levels. Family history of breast cancer was associated with higher prolactin levels among premenopausal women, but not postmenopausal women.
Our observation of significantly lower prolactin levels among parous compared with nulliparous premenopausal women is consistent with several prior studies.9, 10, 11, 12, 13 Parous postmenopausal women also had significantly lower prolactin levels than nulliparous women, consistent with 2 prior studies,9, 14 though the difference was not as large as among premenopausal women. We did not observe a significant gradient in prolactin concentration with increasing number of children, similar to the results of some10, 11, 13 but not all9, 12, 14 studies. In one of the studies with an observed gradient, the difference between nulliparous and uniparous was greater than between subsequent number of children.9 In the other 2 studies, the change in prolactin levels with each child was relatively small (<1 ng/mL), similar to what we observed with a continuous variable, and may have been statistically significant given the large sample sizes (n = 4,550 and 2,119).12, 14 While pregnancy appears to decrease prolactin levels, an additional benefit of earlier age at first birth has not been shown in the few prior studies of this association11, 12, 13, 14 or in the current study.
Our results suggest that the difference between nulliparous and parous prolactin levels may be more pronounced among premenopausal women, and stronger with less time since first birth. Although the absolute difference in prolactin levels between premenopausal and postmenopausal women in our study may be due, in part, to laboratory batch variation, prolactin levels decline after menopause.23 Thus, it is possible that the decline after menopause, regardless of parity, attenuates differences in prolactin levels between nulliparous and parous women. However, we observed that the reduction in prolactin levels appeared to attenuate with time since first birth among premenopausal women. Thus, after 20 years since first birth, the difference between parous and nulliparous women more closely resembled the difference by parity among postmenopausal women. It is possible that the attenuated association with parity in postmenopausal women was due to greater time since first birth, regardless of menopausal status. To our knowledge, only 1 small study has evaluated time since birth,11 with no differences observed within 12 years of last birth.
Circulating prolactin increases gradually during pregnancy, with levels at the end of pregnancy approximately 10 times higher than prepregnancy levels.23 Prolactin levels remain elevated throughout the early postpartum lactation period, and increase acutely with suckling episodes.23 However, both animal and clinical data suggest that pituitary prolactin secretion is reduced after a pregnancy. In clinical studies, the administration of dopamine receptor antagonists resulted in a diminished prolactin response in parous compared with nulliparous women.11, 25 Compared to nulliparous rats, parous rats had blunted prolactin responses to both dopamine antagonists26 and physiologic doses of estrogen.27 To our knowledge, no studies have examined whether these effects are attenuated with greater time since first birth. Although breast cancer risk increases transiently after pregnancy, parity is associated with long-term reductions in risk.28 While higher levels of hormones, including prolactin, during pregnancy may increase risk through the likelihood a genetic mutation will be replicated, lower levels of prolactin after birth may contribute to long-term reduction.1 Even with attenuation of the association over time since first birth, alterations in the postpregnancy prolactin response resulting in lower circulating prolactin may contribute to the pregnancy-associated reduction in breast cancer risk over a woman's lifetime.1
Prolactin is important in mammary gland development, which begins in utero, continues during puberty and reaches maturation with pregnancy.23, 29 Consistent with the few other studies of this association,12, 14, 15 we did not observe an association between age at menarche and prolactin levels. Prolactin is also an essential hormone for lactation,8 thus it is plausible that breastfeeding may alter prolactin levels beyond the effect of pregnancy. However, we observed that parous women who never breastfed had similar prolactin levels as parous women who breastfed for various amounts of time. To our knowledge, the only prior study of breastfeeding and prolactin levels is the Guernsey study with 2 subcohort components; a significant association between ever and never lactation was observed in 1 subset, but not the other.12
Benign breast disease, particularly atypical hyperplasia, is associated with increased breast cancer risk.30, 31 Given that prolactin is produced in normal and malignant breast tissue,7, 32 prolactin levels may differ in women with a history of benign breast disease, compared with those without such history. To our knowledge, we are the first to examine this association. The lack of correlation with prolactin suggests that benign breast disease acts to increase breast cancer risk through a different, although potentially still hormonal, pathway.
Women with a family history of breast cancer have a two- to fourfold increased risk of breast cancer, compared with women with no family history.33 It is possible that heritable hormonal factors may account for some of this association. Several prior studies of the association between family history of breast cancer and circulating prolactin levels have been conducted, with inconsistent results.9, 10, 12, 14, 34, 35 No association was observed with family history of breast cancer in the largest studies of premenopausal12 and postmenopausal14 women, as well as in 2 small studies of premenopausal women.10, 34 However, significantly higher prolactin levels were observed in premenopausal women with a family history of breast cancer in 1 small study,35 and Ingram et al.9 observed an association between family history and prolactin levels among premenopausal, but not postmenopausal, women, similar to our results. Family history in younger, premenopausal women may reflect stronger, more likely heritable, familial associations, with relatives more likely diagnosed at younger ages. In contrast, family history among older, postmenopausal women may include sporadic cases that occur within 1 family by chance, but we did not have enough power to evaluate a family history of 2 or more first-degree relatives with breast cancer.
There are several limitations to our analysis. Given the cross-sectional nature of the study, it is not possible to decipher whether factors associated with prolactin levels, such as parity, determined the prolactin levels, or if prolactin levels contributed to whether a woman had children. However, analyses of prolactin levels in women before and after pregnancy suggest that the lower prolactin levels we observed among parous women are likely a result of parity.11 The nature of prolactin, with multiple forms and varying biological activity,8 poses another limitation, as it is possible that one of the specific isoforms is more strongly associated with the factors studied. The assay we used identifies most prolactin forms, but does not distinguish between them.36 Laboratory drift between batches may have introduced some random error, but the correlation between batches was high and we controlled for batch in all analyses. Finally, the circadian variation in prolactin23 and increase after noon meal24 is a potential limitation, but we controlled for these variables and our results were unchanged when we restricted our analyses to women who gave fasting, morning blood samples.
The large sample size of our study is a strength, as most studies to date of women without breast cancer have been small (n = 26–178).9-11, 13, 15, 34, 35 To our knowledge, this is the second largest study of prolactin levels and reproductive and family history correlates, after the Guernsey studies.12, 14 The large number of both premenopausal and postmenopausal women allowed us to examine associations by menopausal status. Given the extensive information on reproductive factors, we were able to isolate the impact of parity from other birth-related factors on prolactin levels.
In summary, we observed that prolactin levels were lower in premenopausal and postmenopausal parous women, compared with their nulliparous counterparts, and that an increasing number of children did not further decrease prolactin levels. We also observed that premenopausal women with a family history of breast cancer had higher prolactin levels than women without; this association was not apparent among postmenopausal women. These findings suggest that prolactin may be one of many hormones that are part of the mechanism through which these 2 established risk factors influence breast cancer risk.