The Association of Plasma DHEA and DHEA Sulfate with Breast Cancer Risk in Predominantly Premenopausal Women

In vitro and animal evidence suggests that the adrenal androgens, DHEA and its sulfate (DHEAS), are important in the development of breast cancer; however, the mechanisms underlying this association are unclear (1). DHEA and DHEAS inhibit growth of some breast cancer cell lines while stimulating growth in others (2); DHEA can directly activate estrogen receptor α (ER-α) expression in MCF7 breast cancer cell lines (3). One hypothesis suggests that in a high estrogen setting (e.g., in premenopausal women), DHEA and DHEAS exhibit antiestrogenic effects, but in a low estrogen setting (e.g., in postmenopausal women) they exhibit weak estrogenic effects (4, 5). Epidemiologic data suggest a positive association with postmenopausal breast cancer (6), but both positive and inverse associations have been observed in younger women (7-11). Therefore, we evaluated the relationship of DHEA and DHEAS with breast cancer in predominately premenopausal women in a prospective nested case-control study within the Nurses' Health Study II cohort.

The Nurses' Health Study II was established in 1989, when 116,609 female registered nurses, ages 25 to 42 years, completed a questionnaire. The cohort has been followed biennially since inception to update exposure variables and ascertain newly diagnosed disease. The racial/ethnic breakdown is 94% White, 2% Asian, 2% African-American, and 2% Hispanic.

Between 1996 and 1999, 29,611 cohort members, ages 32 to 54 years, provided a blood sample (described in ref. 12). Briefly, premenopausal women who had not taken hormones, been pregnant, or breastfed within 6 months (n = 18,521) answered a short questionnaire and provided timed blood samples on the 3rd to 5th day of their menstrual cycle (follicular draw) and 7 to 9 days before the anticipated start of their next cycle (luteal draw). Follicular plasma was aliquoted by the participant 8 to 24 hours after collection and frozen. All other women (n = 11,090) provided a single 30 mL, untimed blood sample. Luteal and untimed samples were shipped via overnight courier, processed by our laboratory, and separated into plasma, RBC, and WBC components. Samples have been stored in continuously monitored, liquid nitrogen freezers since collection. Menopausal status determination for women providing untimed samples has been described previously (12) and is based on whether the woman's periods have ceased, ovarian status, smoking status, and age. Follow-up of the blood cohort was 98% in 2003. The study was approved by the Committee on the Use of Human Subjects in Research at the Brigham and Women's Hospital.

Cases had no previously reported cancer diagnosis and were diagnosed with breast cancer after blood collection but before June 1, 2003. Overall, 317 cases of breast cancer were reported and confirmed by medical record review (n = 298) or verbal confirmation by the nurse (n = 19). Given the 99% confirmation rate on medical record review, these latter cases were included. Mean time from blood draw to diagnosis was 31 months (range, 1-87). Each case was matched to two controls on age (±2 years), menopausal status at blood collection and diagnosis, month/year of blood draw (±2 months), ethnicity (African-American, Asian, Hispanic, Caucasian, other), luteal day of the blood collection (timed samples only, date of next period-date of luteal draw, ±1 day), and, for each blood draw, time of day (±2 hours) and fasting status (<2, 2-4, 5-7, 8-11, and 12+ hours). For each matching variable, >90% of matches were exact.

DHEA, DHEAS, and progesterone were assayed in luteal and untimed samples by two of the authors (E.F. and M.D.). DHEA was assayed by RIA (Diagnostic Systems Laboratories, Webster, TX), whereas DHEAS and progesterone were measured by chemiluminescent immunoassay using the Immulite autoanalyzer (Diagnostic Products Corp., Llanberis, United Kingdom; ref. 8). Estradiol (in timed samples only) and testosterone were assayed at Quest Diagnostic's Nichols Institute (San Juan Capistrano, CA) as previously described (13). These hormones are stable in whole blood not processed for 24 to 48 hours (14).

Case-control sets were assayed together. Samples, assayed in two batches, were ordered randomly and labeled to mask case-control status. Each batch included replicate plasma samples to assess interassay coefficient of variation (6-12%).

We considered all women (timed and untimed) together, using batch-specific quartile cutpoints for DHEA and cutpoints across batches for DHEAS (based on control distributions), because, although the correlations from a subset of 12 samples run in both batches were >0.95, the mean concentrations differed by batch for DHEA, suggesting some laboratory drift.

We did not identify any statistical outliers (15). Some samples had missing hormone values related to technical difficulties or low volume; the sample size varied by hormone (see Table 1; estradiol: 185 cases/368 controls, testosterone: 302 cases/605 controls). We used mixed-effects regression to test log-transformed hormone differences between matched cases and controls.

We used conditional logistic regression to estimate relative risks (RR) and 95% confidence intervals (95% CI; ref. 16). To maximize statistical power, RRs and 95% CIs for various case groups (invasive only, estrogen receptor (ER) positive/progesterone receptor (PR) positive time between blood draw and diagnosis) and stratified by age at blood (<45 versus ≥45 years), body mass index (BMI) at blood (<24 versus ≥24 kg/m²), family history of breast cancer, history of benign breast disease, and oral contraceptive use history (never/<2 years versus ≥2 years) used unconditional logistic regression adjusting for matching factors. Only 14 postmenopausal breast cancer cases were premenopausal at blood collection; therefore, we could not examine this group separately. Secondary, a priori analyses were limited to premenopausal women with an ovulatory cycle (progesterone levels ≥400 ng/dL) or 3 to 21 days between blood collection and date of next period, because the luteal levels of DHEAS and DHEAS may be more reflective of long-term exposure in this population (17). We adjusted for BMI at age 18 years, family history of breast cancer, age at menarche, history of benign breast disease, and parity/age at first birth. Further adjustment for lactation history and past oral contraceptive use did not substantially alter the results. Secondarily, we considered adjustment for estradiol (follicular and luteal) and testosterone (average of follicular and luteal for women with timed samples) to assess whether these hormones mediated the associations of DHEA and DHEAS with breast cancer risk. Tests for trend were conducted by modeling quartile medians and calculating the Wald statistic (18). All P values were two-sided.

A higher proportion of cases versus controls had a family history of breast cancer and a history of benign breast disease (Table 1). Differences for other risk factors were generally small, but in the expected direction. Median DHEAS levels were modestly higher for cases than controls (P = 0.08), but DHEA levels were similar between groups (P = 0.46).

Modest, nonsignificant associations were observed between DHEA or DHEAS and breast cancer risk among all women or premenopausal women with an ovulatory cycle and a normal luteal phase length (Table 2). Both hormones were positively associated with ER+/PR+ breast cancer (DHEA RR, top versus bottom quartile, 1.6; 95% CI, 0.9-2.8; P_trend = 0.09; DHEAS RR, 1.9; 95% CI, 1.1-3.3; P_trend = 0.02). Neither hormone seemed to be associated with ER−/PR− tumors (n = 34 cases, comparable RR = 1.1 for both, P_trend > 0.40). Adjustment for estradiol did not change the results (data not shown); however, adjustment for testosterone attenuated the RRs for ER+/PR+ tumors (DHEA RR, 1.2; 95% CI, 0.7-2.2; P_trend = 0.60 and DHEAS RR, 1.4; 95% CI, 0.8-2.6; P_trend = 0.21).

The associations significantly differed by age (P_interaction, DHEA = 0.02 and DHEAS = 0.04; Table 3). For DHEA, the RR was 0.6 (top versus bottom quartile, 95% CI, 0.3-1.3; P_trend = 0.16) among women <45 years and was 1.7 (95% CI, 1.0-3.0; P_trend = 0.05) among women ≥45 years. For DHEAS, the comparable results were 0.8 (95% CI, 0.4-1.5; P_trend = 0.62) and 2.0 (95% CI, 1.2-3.5; P_trend = 0.003), respectively. Adjusting for estradiol or testosterone levels did not substantially alter the results (data not shown). The association did not vary by time between blood draw and diagnosis, family history of breast cancer, history of benign breast disease, oral contraceptive history, or BMI (data not shown).

We found that DHEA and DHEAS were modestly, but not significantly, associated with breast cancer in predominantly premenopausal women; both hormones were positively associated with risk of ER+/PR+ tumors. We also observed a possible trend toward an inverse association in women <45 years old and a significant positive association among women ≥45 years old.

Biological data of DHEA and DHEAS effects on breast tumors are conflicting. DHEA and DHEAS are primarily produced by the adrenal gland, although DHEA can be produced in the ovaries (19), and are androgen and estrogen precursors (20, 21). DHEA and DHEAS have independent androgenic properties via the androgen receptor and their metabolite 5-androstenediol has weak estrogenic properties through the ER (22-25). In vitro data suggest that DHEA and DHEAS inhibit breast cell proliferation possibly through epidermal growth factor (26), although they can stimulate proliferation, especially in a low-estrogen environment (1, 3, 22, 24, 25). Thus, DHEA and DHEAS may act as estrogen antagonists in a high estrogen environment, but act as weak estrogen agonists in a low-estrogen environment (1, 2, 27). A second hypothesis posits that women who develop breast cancer have a blunted age-related decline in DHEA and DHEAS levels (1, 17, 28, 29); that is, compared with healthy controls, women who develop breast cancer in early- to mid-premenopausal years have similar or lower DHEA or DHEAS levels but those diagnosed in late-premenopausal and postmenopausal years have higher levels (1).

Epidemiologic data generally do not support the hypothesis of different associations by estrogen levels. Case-control studies in premenopausal women in which blood samples were collected after breast cancer diagnosis generally reported no associations (30, 31); however, major surgery can alter levels (32) possibly explaining these results. One case-control study collected serum several weeks after surgery and reported a significant positive association for DHEA (odds ratio, top versus bottom quartile, 2.7; ref. 32).

Results from prospective studies in premenopausal women seem contradictory. Three studies (n = 15-116 cases) reported inverse (10, 11) or nonsignificant positive associations (9). Two larger studies (n > 300 cases) reported no association in older premenopausal women (8) and a significant positive association for DHEAS that was stronger in women <49 years (7). In contrast, we observed a positive association among women ≥45 years old and a suggestion of an inverse association among those <45 years. This was not related to different proportions of ER+/PR+ tumors (<45 years: 67%, ≥45 years: 65%) or follicular/luteal estradiol concentrations (<45 years: 45/117 pg/mL, ≥45 years: 43/114 pg/mL).

Although the reasons for the conflicting results are unclear, large prospective studies generally observed a positive association between DHEA and DHEAS levels and breast cancer in premenopausal women, although this population has high estrogen levels. This suggests that these hormones may not act as an estrogen antagonist in premenopausal women. Conversely, our observed interaction with age supports the hypothesis that women who develop breast cancer have a blunted DHEA and DHEAS decline, such that levels in cases are lower at younger ages and higher at older ages than controls. These results should be interpreted with caution because no similar interaction was observed in other studies (7, 8). Additional studies are needed to address this issue in detail.

We also observed a stronger association among women with ER+/PR+ tumors, suggesting that DHEA and DHEAS may act directly or indirectly through the ER (33). In vitro data suggest that DHEA can directly activate ERα expression in MCF7 cells (3). Adjustment for testosterone attenuated the observed association of DHEA and DHEAS with ER+/PR+ tumors, suggesting that the association may in part be due to conversion of these hormones to testosterone. However, given the sample size and correlation between DHEA or DHEAS and testosterone (Spearman r ∼ 0.5), the independent influences of individual hormones need to be reevaluated with further follow-up.

One limitation of this study is the laboratory drift over time for DHEA, which precluded us from considering absolute DHEA concentrations and risk; however, there was no laboratory drift for DHEAS. Although this study had a relatively large sample size, we had limited power for some subgroup analyses, requiring further follow-up and reevaluation. Although we only had one blood sample, DHEA and DHEAS are relatively stable over time (intraclass correlation = 0.66 and 0.81, respectively), so there is little misclassification (34). Additionally, plasma DHEA levels may reasonably reflect breast tissue levels (35).

In this study of predominantly premenopausal women, we found that DHEA and DHEAS were positively associated with breast cancer risk, particularly for ER+/PR+ tumors and in older women, supporting the hypothesis that breast cancer patients have a blunted age-related decline in these hormones. However, other studies have reported conflicting results about an age interaction (7, 8). Given the positive association of DHEA and DHEAS with postmenopausal breast cancer risk (6), it is important to further explore these relationships in premenopausal women with continued follow-up in this and other cohorts.

We thank Drs. Graham Colditz, David Hunter, and Walter Willett for their helpful comments on the manuscript.