Bioactive Prolactin Levels and Risk of Breast Cancer: A Nested Case–Control Study

Prolactin is a lactogenic hormone that has been associated with breast cancer risk primarily in postmenopausal women (1–3). This hormone, which is involved in many biologic processes, including breast development and lactation (4), has a range of actions in part due to substantial posttranslational modifications (5). For example, glycosylated prolactin has a higher metabolic clearance rate and lower biologic activity due to reduced binding to receptors than the nonglycosylated form (6–8), and high-molecular weight isoforms (e.g., multimers, IgG complexes) tend to have lower biologic activity than the monomer (6, 9, 10). Furthermore, in mammary epithelial cells, unmodified prolactin may promote growth, while a phosphorylated isoform seems to inhibit growth (11–13). Prior epidemiologic studies primarily have used immunoassays to measure prolactin concentrations, these studies generally observed a 30% increase in risk comparing postmenopausal women with the highest versus lowest levels. However, immunoassays assess multiple isoforms and thus may not fully reflect the biologic activity of prolactin (7) that is relevant to breast carcinogenesis. Therefore, we considered bioactive prolactin levels as measured by the Nb2 lymphoma cell bioassay, which is a sensitive measure of the somatolactogenic activity of prolactin and growth hormone in plasma. Although the assay is highly specific to prolactin and growth hormone activity, between 78% and 95% of the activity measured by the bioassay is due to prolactin and measures of prolactin activity alone and total somatolactogenic activity are highly correlated (10, 14, 15). Furthermore, there is little epidemiologic evidence to support that growth hormone is associated with breast cancer risk (16).

Although data are limited, the bioassay may better reflect the portion of prolactin that is biologically important for breast cancer risk. In several small studies, circulating measures of bioactive prolactin, but not prolactin assessed by immunoassay, were 33% higher in breast cancer cases versus controls (10), were 2- to 3-fold higher in cases of benign breast disease versus controls (17–19), or among women with versus without a family history of breast cancer (17, 20). Because no study has prospectively evaluated risk of breast cancer with bioactive prolactin levels, we conducted a nested case–control study of breast cancer within the Nurses' Health Study (NHS) and NHSII measuring bioactive prolactin levels in plasma using the Nb2 lymphoma cell bioassay, additionally considering associations between the bioassay and breast cancer risk factors among controls.

The NHS was established in 1976 when 121,700 U.S. female registered nurses, ages 30 to 55 years, completed and returned a baseline questionnaire; similarly, NHSII began in 1989 with 116,430 female registered nurses (ages 25–42 years at baseline). In both cohorts, women consented to participate and have completed additional questionnaires every 2 years since study inception to update exposure and disease status.

In 1989–1990, 32,826 NHS participants provided a heparinized blood sample with a short questionnaire; women were 43 to 70 years old at blood collection. Details have been described previously (21), but briefly, women were mailed a collection kit, collected the sample, and shipped back the specimen with an ice pack to our laboratory via overnight courier. Samples were processed into plasma, red blood cells, and white blood cell components and have been stored in alarmed liquid nitrogen freezers (<−130°C) since collection. In addition, 186 postmenopausal women not using hormone therapy (HT) provided two more samples each one year apart (22). Similarly, 29,611 NHS participants (ages 32–54 years old) provided heparinized blood samples from 1996 to 1999 (23). Briefly, those women that were still cycling and had not been pregnant, taken hormones, or lactated within the prior 6 months provided two blood samples, one on the third to fifth day of the menstrual cycle (follicular) and one on the seventh to ninth day before the anticipated start date of the next cycle (luteal). Follicular plasma was aliquoted by the participant and frozen until the luteal sample was drawn. In addition, 236 women provided two more sets of follicular and luteal samples each one year apart (24). All other women (n = 11,090) provided a single, untimed sample. Luteal and untimed samples were processed as in the NHS collection. Follow-up of the participants in these cohorts is >95%. This study was approved by the Committee on the Use of Human Subjects in Research at the Brigham and Women's Hospital (Boston, MA).

We considered a woman to be premenopausal if (i) she gave timed samples, (ii) her periods had not ceased, or (iii) she had a hysterectomy with at least one ovary remaining and was 47 years or younger (nonsmokers) or 45 years or younger (smokers). We considered a woman to be postmenopausal if (i) her natural menstrual periods had ceased permanently, (ii) she had a bilateral oophorectomy, or (iii) she had a hysterectomy with at least one ovary remaining and was 56 years or older (nonsmokers) or 54 years or older (smokers). The remaining women, most of whom had a simple hysterectomy and were 48 to 55 years old, were considered to have an unknown menopausal status.

Cases were matched to one control on cohort (NHS and NHSII), menopausal status at baseline and diagnosis (premenopausal, postmenopausal, unknown), age (±2 years), month of blood collection (±1 month), time of day of blood draw (in 2-hour increments), fasting status (<8 vs. ≥8 hours), HT use at blood draw if postmenopausal (yes vs. no), and luteal day of blood draw (date of next period minus date of blood draw) if providing a timed sample. To assess within person stability over time, we conducted assays among 68 postmenopausal women and 105 premenopausal women with luteal samples (n = 33 had ample follicular sample for the assay) who gave three blood specimens. We also assessed the reproducibility of the assay with delayed processing, mimicking our blood collection methods, among 15 local donors, comparing blood processed immediately, versus 24 or 48 hours after collection (delayed processing samples kept with an ice pack).

For NHS, we assayed invasive cases diagnosed between June 1, 1992 and June 1, 2000—cases diagnosed within 2 years of blood collection were excluded to preserve their remaining sample volume for future early detection studies. For NHSII, we assayed invasive cases diagnosed after blood collection but before June 1, 2009. The Nb2 bioassay for biologically active somatolactogens was performed as previously described on these cases and controls (25, 26). Briefly, 24 hours before bioassay, 5 × 10⁴/cm² cells were placed into a somatolactogen-deficient medium (DMEM-ITS). The bioassay was initiated by introducing dilutions of the plasma sample in quadruplicate; 48 hours later, proliferation was assessed by 3H-thymidine incorporation. Parallel cultures were stimulated with defined concentrations of recombinant prolactin (0–5 ng/mL, NIDDK) as controls. After growth factor restriction, Nb2 cells were cultured in the presence and absence of serum to establish baselines, in addition to varying concentrations of prolactin to establish the growth curves. The sensitivity of the assay was 3 pg/mL. Case–control samples were assayed in duplicate in 104 batches (NHS: 65 batches, NHSII: 39 batches) consisting of approximately 25 to 35 women each. Average within batch coefficients of variation (CV) were calculated either using blinded quality control (QC) samples (n = 77 batches) or using the average CV of the duplicate samples within a batch if blinded QCs were not available (n = 27 batches). CVs ranged from 0.3% to 17.3% for 97 of the batches (mean = 7.0%). CVs for seven batches were above 20% (range, 21.5–26.6); thus, we conducted a sensitivity analysis excluding these batches. The between batch CV was 40%. We also measured prolactin using a microparticle enzyme immunoassay on all the samples with the Nb2 bioassay except for NHSII cases diagnosed from 2003 to 2007 and matched controls who were postmenopausal at blood draw (for cost-saving reasons, n = 65; ref. 2). The limit of detection was 0.6 ng/mL and no samples were below the limit of detection; the average CV was 7.8% (SD = 3.5%).

Case–control sets were assayed together as were follicular and luteal samples (NHSII), ordered randomly within a set, and labeled with unique IDs. Case–control sets were randomly ordered within a batch, although NHS and NHSII samples were assayed in separate batches. Among NHSII women with timed samples, we measured both the follicular and luteal sample for the prolactin immunoassay as the average of these two measures provided the highest intraclass correlation (ICC) over 3 years (24). Only the luteal sample was measured for the bioassay as plasma volume of the follicular sample was limited and averaging the follicular and luteal value did not improve the ICC (see Results).

Because of the assay variation between laboratory batches over time, we recalibrated both immunoassay and bioassay levels from all batches to have a comparable distribution to an average batch (27). We assumed that all batches combined represented an average batch. We then regressed prolactin immunoassay or bioassay levels on indicator variables for batch as well as potential predictors of prolactin levels including age, case–control status, menopausal status, family history of breast cancer, age at menopause, early life body size, body mass index (BMI) at blood draw, fasting status, time of blood collection, parity, history of HT use, and use of antidepressants and steroids at the time of blood collection. Levels were then recalibrated based on the coefficients for each batch and the average of the batch coefficients. We classified the recalibrated values into quartiles based on the overall distribution in controls.

ICCs in the pilot studies were calculated by dividing the between-person variance component by the sum of the between- and within-person variance components obtained from a mixed model, with participant ID as the random effect. ICCs were assessed on ln-transformed assay values.

We used linear regression adjusting for matching factors among controls only to evaluate the association of prolactin bioassay levels with reproductive and hormonal factors previously associated with breast cancer risk. We evaluated the following factors assessed at or near blood collection in a multivariate model: age (<45, 45–59, 50–54, 55–59, 60+ years), BMI (<22, 22–24.9, 25–29.9, 30+ kg/m²), parity (0, 1, 2, 3, 4+ children), age at first birth among parous women (<25, 25–29, 30+ years), breastfeeding among parous women (never, 1–11, 12+ months), family history of breast cancer (yes, no), personal history of benign breast disease (yes, no), antidepressant use (yes, no), alcohol intake (0, <5, 5+ g/day), age at menarche (<12, 12, 13, 14+ years), average childhood body size at ages 5 and 10 years based on a 9-level pictogram (ref. 28; 1, >1–2, >2–3, >3), menopausal status/HT use (premenopausal, postmenopausal/current HT use, postmenopausal/never or past HT use), and age at menopause among postmenopausal women (<48, 48–50, 51–52, 53+ years). These variables were assessed on the questionnaire completed at the time of blood collection or on the biennial questionnaires.

Relative risks (RR) and 95% confidence intervals (CI) of breast cancer were determined using conditional logistic regression comparing quartiles of prolactin bioassay concentrations (29). Tests for trend were modeled using quartile medians and assessed using the Wald statistic. Models were adjusted for BMI (continuous), childhood body size (continuous), age at menarche (<12, 12, 13, 14+ years), history of benign breast disease (yes, no), family history of breast cancer (yes, no), antidepressant use (yes, no), age at menopause (continuous, with an indicator for premenopausal), and alcohol intake (0, <5, 5+ g/day). We considered separate adjustment for parity/age at first birth (nulliparous, 1–2 children/<25, 1–2 children/25–29, 1–2 children/30+, 3+ children/<25, 3+ children 25+) and breastfeeding (none, <12 months, 12+ months) as this may be part of the biologic pathway through which prolactin affects breast cancer (30).

On the basis of our prior results using the immunoassay (2), we investigated whether the bioassay/breast cancer association varied by menopausal status at blood collection by including a multiplicative interaction term between menopausal status and the prolactin quartile median variable, using the Wald test as well as assessing associations by time between blood draw and diagnoses/reference date (<4, 4–7.9, 8+ years). We additionally assessed associations by tumor characteristics differentially associated with the immunoassay using polytomous logistic regression including estrogen (ER) and progesterone (PR) receptor status, ductal versus lobular, HER2⁺ tumors, tumor size, and lymph node metastasis status, adjusting for covariates as described above and matching factors [age at blood draw (continuous), date of blood draw (continuous), time since last meal (≥8 vs. <8 hours), time of day of blood draw (midnight–6 am, 6–8 am, 8 am–midnight), menopausal/HT status (premenopausal, postmenopausal/current HT use, postmenopausal/never or past HT use), and cohort (NHS, NHSII)]. We also considered whether three other metrics combining information about the bioassay and immunoassay were associated with breast cancer risk, including the ratio of the bioassay to immunoassay, a score that summed the decile ranking for each of the assays, or the cross-classification of the two assays, as prior studies of the bioassay suggested that this may reflect the nonmonomeric prolactin forms (15, 17, 19) and in one study, the bioassay:immunoassay ratio was higher in breast cancer cases versus controls (10).

All P values were two sided and were considered statistically significant if <0.05. All analyses were conducted using SAS, version 9.3 (SAS Institute Inc.).

We conducted pilot studies to assess the stability of prolactin bioactivity with delayed processing of the sample for up to 48 hours as well as the within person stability of bioactive prolactin over 3 years in premenopausal and postmenopausal women separately. The ICC comparing samples processed immediately versus 24 or 48 hours after collection was 0.87, suggesting that bioactivity of prolactin is not substantially altered by delayed processing. When examining within person stability over 3 years, we observed an ICC of 0.63 for postmenopausal women, and in premenopausal women, an ICC of 0.66 for the follicular phase, 0.63 for the luteal phase, and 0.64 for the average of the follicular and luteal phases. The correlation between the bioassay and immunoassay was 0.57.

We had assay results available on 1,329 cases and 1,329 controls, about 59% of which came from the NHS (Table 1). Cases were more likely to be nulliparous, have a family history of breast cancer and a personal history of benign breast disease than controls.

Among controls, parity was inversely associated with bioactive prolactin levels (P = 0.01; Table 2); nulliparous women had the highest levels (least square mean = 10.8), those with one child had intermediate levels (mean = 9.5), those with two or more children had the lowest levels (mean ranged from 8.8–8.9). Those with 4+ children had 18% lower bioactive prolactin than nulliparous women. In addition, women with a family history of breast cancer had 14% higher bioactive prolactin levels than those without a family history (P < 0.01). Both age at menopause (P = 0.03; percent difference comparing 53+ with <48 years = −18%) and BMI (P < 0.01, percent difference comparing 30+ with <22 kg/m² = −16%) were inversely associated with bioactive prolactin levels. Premenopausal women had suggestively higher bioactive prolactin levels (least squared mean = 9.6, P = 0.43) than postmenopausal women using HT at blood draw (mean = 8.7) or postmenopausal women not using HT (mean = 8.4). Other factors, including age, age at first birth, breast feeding, history of benign breast disease, antidepressant use, alcohol use, age at menarche, and early life body size were not significantly associated with bioactive prolactin levels.

In age-adjusted analyses, bioactive prolactin was suggestively positively associated with breast cancer risk overall (RR, >11.2 vs. <5.7 ng/mL = 1.25; 95% CI, 0.97–1.53; P_trend = 0.12); however, the association was slightly attenuated in multivariate adjustment (comparable RR, 1.19; 95% CI, 0.94–1.51; P_trend = 0.18; Table 3). The association was possibly stronger (P_{heterogeneity} = 0.55) for postmenopausal (comparable multivariate RR, 1.36; 95% CI, 0.93–1.98) than for premenopausal (comparable multivariate RR = 0.99; 95% CI, 0.71–1.37) women. Among postmenopausal women, current HT users had a suggestively stronger association in the top quartile versus bottom than those not using HT at blood draw, but the risk estimates for quartiles 2 and 3 were similar by HT use. Within the dataset on which we had bioassay results, the corresponding RRs for the prolactin immunoassay were 1.01 (95% CI, 0.79–1.30) for all women, 1.18 (95% CI, 0.80–1.72) for postmenopausal women, and 0.76 (95% CI, 0.51–1.12) for premenopausal women; these results are generally consistent with that observed in a much larger dataset from our prior analysis of the immunoassay (2), although not statistically significant due to the smaller sample size.

We also observed a difference (P_{heterogeneity} = 0.02) in the strength of the association by time between blood draw and diagnosis. A significant positive association was observed for cases diagnosed within 4 years of blood draw (comparable RR; 2.66, 95% CI, 1.45–4.89; P_trend < 0.01), but bioactive prolactin was not associated with cases diagnosed 4 to 8 (P_trend = 0.78) or 8+ years (P_trend = 0.74) after blood draw. The association differed neither by estrogen receptor (ER) status (P_{heterogeneity} = 0.69) nor by other tumor characteristics including progesterone receptor status, HER2 status, invasive versus in situ, ductal versus lobular, tumor size, or lymph node status (data not shown). Additional adjustment for parity/age at first birth, and breastfeeding slightly attenuated the results (data not shown). Furthermore, the ratio of bioassay:immunoassay prolactin levels, a score summing the deciles of each assay, or cross-classified bioassay/immunoassay levels were not associated with breast cancer risk (data not shown). Results were similar when excluding the seven batches with CVs > 20% (data not shown).

This is the first prospective study examining the relationship of bioactive prolactin levels as assessed by the Nb2 bioassay and risk of breast cancer in over 1,300 cases and controls. Overall, we did not observe a clear relationship between prolactin measured using this assay and breast cancer risk, although there was a suggestive association for postmenopausal women and a significant positive relationship for women diagnosed within 4 years of blood draw. Furthermore, bioactive prolactin was associated in cross-sectional analyses with several known breast cancer risk factors, including parity, family history of breast cancer, age at menopause, and BMI.

Similar to what has been previously observed by ourselves and others in analyses of the prolactin immunoassay and breast cancer risk (1–3), there was about a 30% and 40% increase in risk of invasive postmenopausal breast cancer for women in the top versus bottom quartile of bioactive prolactin, but no association for premenopausal women. The lack of statistical significance for the postmenopausal group in this analysis likely reflects the smaller sample size than in our immunoassay study.

Our hypothesis was that the association would be substantially stronger for bioactive prolactin than for levels assessed by immunoassay because of the heterogeneity of circulating prolactin (6) and because the bioassay could more directly reflect the hormone's effect on cell proliferation. The observed similar or only slightly stronger association for prolactin measured via the bioassay versus the immunoassay may be due to several reasons. First, the somatolactogenic receptors on the Nb2 rat lymphoma cell line may not reflect the receptors expressed in human breast tissue. Prolactin receptor has multiple (4) isoforms and the particular forms of prolactin that activate the receptors in breast tissue may differ from those in the Nb2 cell line. One study examined the correlation between the Nb2 assay and a similar assay using human prolactin receptor (31). When the predominant form of prolactin was monomeric, the two assays performed similarly; however, when high levels of macroprolactin were present, the Nb2 assay had higher levels than the human-based assay. Second, although the bioactivity assay is primarily influenced by prolactin, proliferation also can be stimulated by growth hormone (15, 26), although the concentrations of growth hormone must be very high to stimulate Nb2 cells and over 80% of the stimulation is due to prolactin (10, 14, 15). Notably, this assay is specific to prolactin and growth hormone as there is no cell proliferation observed when including antiprolactin and antigrowth hormone antibodies to the assay (15). We observed a correlation of 0.57 between the immunoassay and bioassay, which is slightly lower than that observed in prior studies (r = ∼0.73–0.95; refs. 15, 17, 19). This may reflect that we used a microparticle immunoassay as opposed to a radioimmunoassay in the other studies. Third, while the within assay CV was very good for the bioassay, in part because we ran samples in quadruplicate, the between assay CV was higher (∼40%). This may be due to the use of live cells, which cannot be standardized in the same way as chemical reagents. We corrected for this statistically using a batch recalibration technique designed specifically to address this type of heterogeneity (27), but some amount of measurement error may have attenuated the results.

Interestingly, as with the immunoassay (2), we observed a significantly stronger association between the bioassay and risk of disease occurring shortly after blood collection, with over a 2-fold risk of breast cancer comparing women with the highest versus lowest bioactive prolactin levels. This may reflect production of biologically active prolactin by the tumor itself, increasing circulating levels (32–34). Prolactin also may be involved in processes important for late-stage carcinogenesis, such as proliferation, cell migration, and angiogenesis (4, 35–40), and high levels, particularly of bioactive prolactin, could enhance those processes in vivo. Furthermore, prolactin has some within person variability over time suggesting that the association could attenuate with increasing time between blood draw and diagnosis. The immunoassay has an ICC of about 0.4 to 0.55 over 3 years (24, 41), but the bioassay has a higher ICC in the same datasets of around 0.63 to 0.66, which is similar to that for blood pressure and cholesterol (42). Unlike the immunoassay results, we did not observe stronger associations for ER⁺ tumors either overall or in postmenopausal women (data not shown). The relatively small number of postmenopausal women with ER⁻ (n = 98) disease may have limited our ability to observe a differential relationship by ER status.

Similar to prior studies of the immunoassay (30, 43–46), we observed that parity, but not age at first birth or breast feeding, was associated with bioactive prolactin levels. Nulliparous women had significantly higher bioactive prolactin than parous women. Several lines of evidence suggest that pituitary prolactin secretion diminishes after pregnancy (47, 48). We also observed that women with a family history of breast cancer had higher bioactive prolactin levels; this is consistent with both previous bioassay (17, 20) and immunoassay (30, 43) results. It is possible that some heritable component of breast cancer risk also influences hormone levels (49). Prior small studies (N < 52) of bioactive prolactin in both blood and nipple aspirate reported higher levels in benign breast disease patients than controls; (17–19, 50) however, our much larger study did not observe such an association. Interestingly, we noted an inverse association between current BMI and bioactive prolactin, which is inconsistent with previous studies of the immunoassay (45, 51, 52) that have either observed positive or no association with BMI. The reason for this association is unclear, but it is possible that this relationship may be driven in part by the influence of growth hormone on the bioassay. In one previous analysis in the NHSII, we observed that growth hormone was strongly inversely associated with BMI (53). Another possibility is that recent evidence suggests that prolactin can be produced and released by adipose tissue; however, cell culture studies of adipocytes from obese versus nonobese individuals suggest that prolactin release is substantially lower in obese women (54).

Overall, the results of this study are consistent with the hypothesis that prolactin is a risk factor for breast cancer. However, given that the association was largely similar for bioactive prolactin and prolactin measured by immunoassay and the lower inter-assay CVs for the immunoassay, future epidemiologic studies should consider using the immunoassay, which is less expensive and more high throughput. Additional research understanding the etiology of prolactin in breast carcinogenesis, particularly on late-stage effects, is warranted.

No potential conflicts of interest were disclosed.

Conception and design: S.S. Tworoger, S.E. Hankinson

Development of methodology: S.S. Tworoger, C.V. Clevenger, S.E. Hankinson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.S. Tworoger, Y.B. Feeney, C.V. Clevenger, S.E. Hankinson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.S. Tworoger, M.S. Rice, B.A. Rosner, S.E. Hankinson

Writing, review, and/or revision of the manuscript: S.S. Tworoger, M.S. Rice, B.A. Rosner, C.V. Clevenger, S.E. Hankinson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.V. Clevenger

Study supervision: S.S. Tworoger, S.E. Hankinson

The authors thank the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, and WY.

This work was supported by the NIH [R01 CA49449 (PI: S.E. Hankinson), R01 CA138580 (PI: S.S. Tworoger), P01 CA87969 (PI: Stampfer), R01 CA67262 (PI: S.E. Hankinson), R01 CA50385 (PI: Willett), and T32 CA09001 (PI: Stampfer)]. Also, this study was approved by the Connecticut Department of Public Health (DPH) Human Investigations Committee. Certain data used in this publication were obtained from the DPH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.