Prediagnostic Circulating Levels of Sex Steroid Hormones and SHBG in Relation to Risk of Ductal Carcinoma In Situ of the Breast among UK Women

With the advent of mammographic screening, the number of women diagnosed with ductal carcinoma in situ (DCIS) of the breast has risen substantially, such that DCIS now accounts for approximately 20% of screen-detected breast cancers (1). DCIS is considered to be a nonobligate precursor of invasive breast cancer (2) and, consistent with this, existing evidence suggests that DCIS and invasive breast cancer share risk factors, including early age at menarche, late age at first live birth, nulliparity, hormone therapy use, and possibly obesity (3, 4). Given that these risk factors are largely hormone-related, it is possible that the development of DCIS may be driven by sex steroid hormones, including estrogen and testosterone, as well as the glycoprotein sex hormone–binding globulin (SHBG), but there is limited evidence to date to support their associations with risk of DCIS (2).

Evidence concerning the role of estrogen in the development of DCIS has come from studies that examined intratumoral levels of estrogen and aromatase, with one study reporting that levels of estradiol were higher in DCIS tissue than in benign breast tissue (5) and another study showing that aromatase was overexpressed in DCIS tissues (6). These findings are consistent with the observation that estrogen enhances tumor cell growth and proliferation (7, 8). Although studies have also indicated that elevated levels of testosterone (9–13) and low levels of SHBG (14, 15) may stimulate breast tumor cell growth and proliferation, it is not clear whether testosterone and SHBG influence the development of in situ disease. Epidemiologically, a small nested case–control study by Zeleniuch-Jacquotte and colleagues (16) provided no evidence to suggest that circulating levels of estradiol, testosterone, and SHBG were associated with risk of DCIS in postmenopausal women, whereas an earlier nested case–control study by Missmer and colleagues (17) suggested that total testosterone may be associated with increased risk of DCIS in postmenopausal women.

Given the paucity of epidemiologic data, the study reported here was conducted to examine the association between circulating levels of estradiol (premenopausal women only), testosterone, and SHBG and risk of DCIS among premenopausal and postmenopausal women in the large, prospective UK Biobank cohort.

The UK Biobank comprises 502,536 participants (273,402 females and 229,134 males) aged 40–69 years at recruitment between 2006 and 2010 from across 22 centers located throughout England, Wales, and Scotland (18, 19). The UK Biobank was approved by the North West Multi-Center Research Ethics Committee, the National Information Governance Board for Health and Social Care in England and Wales, and the Community Health Index Advisory Group in Scotland. An independent Ethics and Governance Council, which was established in 2004, continuously monitors UK Biobank's adherence to the Ethics and Governance Framework, which was developed for the study (http://www.ukbiobank.ac.uk/ethics/). All participants provided written informed consent.

At recruitment, information on the participants' demographic characteristics, family history of breast cancer, medical history, menstrual cycles (including time since last period and length of menstrual cycle), reproductive factors, mammogram screening history, and type and frequency of physical activity, was collected using a self-administered touchscreen questionnaire and nurse-led interviews. Regarding menopausal status, women who reported at recruitment that they had had a menstrual period in the preceding year were classified as premenopausal while they were classified as postmenopausal if they had stopped menstruating for at least 365 days. Women who had missing information on menopausal status were classified as postmenopausal if they had had a bilateral oophorectomy (with or without a hysterectomy) or were >55 years of age (with or without a hysterectomy and/or bilateral oophorectomy); women <55 years with unknown menopausal status were excluded if they had missing information on oophorectomy/hysterectomy status or if they had no information on time since last period (20). During the interviews, participants' body weight, height, and waist circumference (WC) measurements were taken by trained staff using standardized procedures. Body mass index (BMI) was calculated as weight (kg)/height (m²).

Detailed information on the collection, processing and archiving of blood samples has been published previously (21). Briefly, at baseline, blood samples were collected by trained and certified phlebotomists or nurses and stored in bar-coded vacutainer tubes (blood samples were collected from premenopausal women irrespective of the days since their last menstrual cycle). After blood collection, the tubes were scanned with a bar-code reader to link them with the participant's unique identifier number and to accurately time the blood samples for centrifugation. The blood in the serum separator tube was allowed to clot for 25–30 minutes at room temperature before centrifugation at 2,500 × g for 10 minutes. All vacutainers were then kept at 4°C until the end of day when they were packed in temperature-controlled shipping boxes and transported to a central laboratory where the samples were processed (including removal of the clot) and divided into small aliquots using a fully automated system and then transferred to long-term ultra-low temperature archives (21). Rigorous internal and external quality control measures were also implemented to ensure that high quality samples were provided.

Serum levels of estradiol and testosterone were quantified using two-step and one-step competitive analysis, respectively (Beckman Coulter Unicel Dxl 800), while SHBG was measured using two-step sandwich immunoassay analysis (Beckman Coulter Unicel Dxl 800; https://biobank.ndph.ox.ac.uk/showcase/showcase/docs/serum_biochemistry.pdf; about 96% of postmenopausal women had undetectable estradiol levels, and therefore we could not evaluate the association between estradiol and DCIS among these women). The coefficients of variation for estradiol, testosterone, and SHBG can be found in Supplementary Table S1. The midpoints between 0 and the lowest measured values for estradiol (premenopausal women only; N = 11,772), testosterone (N = 26,729), and SHBG (N = 2) were used as the values for those with undetectable levels of these hormones (i.e., levels that were too low to be detected by the assay). Free estradiol (FE2) concentration (premenopausal women only) was calculated from total estradiol, SHBG, and albumin, and free testosterone (FT) was calculated from total testosterone, SHBG, and albumin, both using mass action equations as described previously (22, 23).

DCIS was ascertained within the UK Biobank cohort through linkage to cancer registries in England, Wales, and Scotland. DCIS was defined using the 10th Revision of the International Classification of Diseases (ICD-10: D051). Complete follow-up was available through 31 March 2016 for England and Wales and 31 October 2015 for Scotland. Death was ascertained via linkage to death registries.

An overview of the selection process (including enumeration of the exclusions) for the analytic cohort can be found in Supplementary Fig. S1. We excluded 15,019 women with no exposure information; 29,785 women <55 years with unknown menopausal status whose hysterectomy or oophorectomy status was unknown or who had missing information on days since last period; 29,350 women who were either current hormone replacement therapy (HRT)/oral contraceptive (OC) users or had unknown HRT/OC use; 15,150 prevalent cancer cases (except nonmelanoma skin cancer; ICD-10 C44) and163 with lobular carcinoma in situ. After exclusions, our analytic cohort comprised 182,935 women.

Our main analyses were performed separately for premenopausal and postmenopausal women. Estradiol (for premenopausal women only), testosterone, and SHBG were analyzed both as continuous variables and as quantiles based on the distribution of the noncases (tertiles for premenopausal women, due to the small number of premenopausal DCIS cases, and quartiles for postmenopausal women). Serum levels of the hormones among premenopausal women typically fluctuate throughout the menstrual cycle. Thus, to account for the cyclical nature of the hormone levels, among these women, the residuals from linear regression models of the sex hormones and SHBG on phase of menstrual cycle were used to create the tertiles (24). Information on days since last period (forward dating) was used to categorize the menstrual cycle phase as follows: early follicular (days 1–5 forwards), late follicular (6–10 forwards), mid-cycle (11–14 forwards), early luteal (15–18 forwards), mid luteal (19–24 forwards), late luteal (25+ forwards; ref. 24).

Multivariable Cox proportional hazards regression models were used to estimate HRs and 95% confidence intervals (CI) for the associations of estradiol, testosterone, and SHBG with risk of DCIS. For these analyses, time-to-event was defined from date of enrollment until date of diagnosis of DCIS, or censored at the occurrence of any invasive cancer [except nonmelanoma skin cancer (ICD-10 C44)], date of death, date of withdrawal from the study, or the end of follow-up (March 31, 2016 for England and Wales and October 31, 2015 for Scotland), whichever came first. Covariates were selected a priori or if their inclusion in the models resulted in a 10% change in the HR estimates. The following variables were included in the models: age at recruitment; socioeconomic status (based on quintiles of Townsend deprivation index); age at menarche; parity/age at first live birth; history of mammographic screening; family history of breast cancer in a first-degree relative; BMI; alcohol consumption; hormone therapy use (postmenopausal women only); age at menopause (postmenopausal women only); menstrual cycle phase (premenopausal women only); and physical activity. A missing value indicator was included for covariates with missing information. The inclusion of smoking, oral contraceptive use and history of diabetes did not change the HRs, and therefore these variables were not included in the models. Tests for linear trend were conducted by modeling the ordinal variables as continuous variables and performing Wald tests. Linear trend was also assessed using the continuous exposures. Given that the hormones and SHBG were skewed, a log base 2 transformation was used both to improve normality and to assist with interpretation of the coefficient (i.e., the HR reflects a doubling in the concentrations of the exposures). The proportional hazards assumption was confirmed using Schoenfeld residuals.

Body fat is known to alter circulating sex steroid hormone and SHBG levels (25, 26) and breast cancer risk (25, 26). Hence, we also assessed whether BMI and WC (a measure of central adiposity) were effect modifiers of the associations of sex steroid hormones and SHBG with risk of DCIS (these analyses were restricted to postmenopausal women given the small number of DCIS cases among premenopausal women).

We performed two sensitivity analyses. First, given that DCIS is typically detected during mammographic screening (1), we conducted an analysis in which we excluded noncases without mammography screening or with mammography screening 2 years or more before recruitment to address the possibility of under-ascertainment of DCIS mdash;this analysis was restricted to postmenopausal women since mammographic screening in the UK is mainly recommended for women aged 50 or older (27). Second, to assess the possibility of reverse causation, we excluded women who developed DCIS within 2 years of enrolment.

All statistical analyses were performed using Stata 14.1 (StataCorp). All P values are two-sided.

This study was conducted using data from the UK Biobank study. Information on data availability can be obtained via the UK Biobank website (http://www.ukbiobank.ac.uk).

A total of 717 DCIS cases (n = 186 for premenopausal women and 531 for postmenopausal women) were ascertained during a median follow-up of 7.1 years (interquartile range: 6.4–7.7 years). Baseline characteristics of the study population are shown in Table 1. Compared with premenopausal and postmenopausal women without DCIS, a higher proportion of cases had a family history of breast cancer in a first-degree relative and had had mammographic screening. Among postmenopausal women, cases were also less likely to be nulliparous.

Table 2 shows the median concentrations for estradiol, FE2, testosterone, FT, and SHBG. Among premenopausal women, median levels of FE2 were somewhat higher among cases than noncases, but median levels of total estradiol, testosterone, and SHBG did not differ significantly between cases and noncases. For postmenopausal women, median levels of total and free testosterone were higher among cases than noncases while median levels of SHBG were lower.

Figures 1 and 2 present the associations of the sex steroid hormones and SHBG with risk of DCIS among premenopausal and postmenopausal women, respectively. In premenopausal women, analyses of the exposures by tertiles showed that total and free estradiol in the highest tertile were positively associated with risk of DCIS (HR: 1.54; 95% CI, 1.06–2.23 and HR: 1.72; 95% CI, 1.15–2.57, respectively; Fig. 1). The HRs for doubling in the concentrations of total and free estradiol were 1.37 (95% CI, 1.05–1.80) and 1.51 (95% CI, 1.12–2.04), respectively (Fig. 1). However, neither testosterone nor SHBG was associated with risk of DCIS among premenopausal women. With respect to postmenopausal women, the risk of DCIS increased with increasing levels of FT (HR: 1.07, 95% CI, 0.82–1.41; HR: 1.28 95% CI, 0.98–1.66; and HR:1.42, 95% CI, 1.09–1.85 for the second, third and fourth quartiles, respectively, compared with the lowest quartile; Fig. 2). The HRs per doubling in concentrations of total and free testosterone were 1.24 (95% CI, 1.04–1.50) and 1.29 (95% CI, 1.07–1.54), respectively. Serum SHBG levels were inversely associated with risk of DCIS in postmenopausal women (Fig. 2). The HR per doubling in concentration of SHBG was 0.89 (95% CI, 0.77–1.03) and for the highest versus the lowest quartile of SHBG, the HR was 0.75 (95% CI, 0.56–0.99).

We assessed the modifying effect of body fat measures on the associations of the sex steroid hormones and SHBG with risk of DCIS in postmenopausal women. These analyses did not show any evidence of heterogeneity by BMI or WC strata (Supplementary Tables S2 and S3).

In sensitivity analyses, among postmenopausal women, the observed associations were still apparent after excluding noncases who did not have mammographic screening within the two years prior to recruitment (Table 3) and after excluding those who had a DCIS diagnosis within two years of recruitment (Supplementary Tables S4 and S5). The observed association between FE2 and DCIS was also still evident, among premenopausal women, after excluding those who had a DCIS diagnosis within two years of recruitment.

In this large prospective study, we observed that elevated total and free estradiol levels were positively associated with risk of DCIS among premenopausal women. Regarding postmenopausal women, relatively high circulating levels of total testosterone, and particularly FT, were associated with increased risk of DCIS while high circulating levels of SHBG were inversely associated with risk.

We are not aware of any previous study which has assessed the associations of sex steroid hormones and SHBG with risk of DCIS among premenopausal women. Our findings of positive associations between total and free estradiol with risk of DCIS are in accord with those of a pooled analysis of prospective studies which demonstrated that total and free estradiol were positively associated with risk of invasive breast cancer before age 50 (24). Subsequent findings from the European Prospective Investigation into Cancer and Nutrition, which to date is the largest individual prospective study on the association between sex steroid hormones and risk of invasive breast cancer among premenopausal, also provided some suggestion of a positive association between total estradiol and risk of invasive breast cancer before age 50 (28). Several epidemiologic studies have also provided evidence for positive associations between testosterone (total and free) levels and risk of invasive breast cancer (24, 28–32). We observed a doubling in risk of DCIS in association with testosterone in premenopausal women, although the results did not reach statistical significance, possibly due to the small number of DCIS cases in this stratum.

In accordance with the findings of studies relating to invasive breast cancer, we observed positive associations between testosterone, and particularly FT, with risk of DCIS among postmenopausal women (17, 33–36). However, the magnitude of associations that we observed were somewhat weaker than those for invasive breast cancer, for which most studies have found at least a 2-fold increase in risk when comparing the highest versus the lowest categories of total and free testosterone (17, 33–36). One previous nested case–control study (n = 34 cases) also found that total, but not FT was positively associated with risk of DCIS (17), whereas another nested case–control study (N = 69 cases) did not observe associations for testosterone (16).

Biologically, estrogen is thought to contribute to breast tumor development through several processes. First, estrogen has a proliferative effect on mammary epithelial cells which can result in random genetic errors leading to DNA mutations (8, 37). Second, estrogen and its metabolites (e.g., semi-quinones and quinones) may enhance the production of reactive oxidative species, which can generate oxidative DNA damage, and they may also suppress processes which sense and repair DNA damage, thereby inducing breast tumorigenesis (37). Third, histopathologic studies have shown that the estrogen-synthesizing enzyme, aromatase, is overexpressed in DCIS tissues (6). These findings align with the observed positive association between estradiol and risk of DCIS among premenopausal women in the current study. It is biologically plausible that estradiol may also contribute to DCIS among postmenopausal women. However, as most of the postmenopausal women had estradiol levels below the limit of detection (i.e., <175 pmol/L; http://biobank.ndph.ox.ac.uk/showcase/showcase/docs/serum_biochemistry.pdf), we were not able to assess the association in this subgroup.

The mechanisms linking testosterone to breast cancer remain unclear. Evidence from some clinical and histopathologic studies suggests that testosterone can have an antiproliferative effect on epithelial tissues by inhibiting estrogen-induced cell proliferation and other processes that are known to induce cell growth and proliferation (10, 38–40). In contrast, other studies have suggested that testosterone may contribute to breast carcinogenesis independently of estrogen. In this regard, androgen receptors (AR) have been observed to be highly expressed in noninvasive and invasive breast tumors (9, 10, 12). Experimentally, one study demonstrated that androgens have a proliferative effect on breast cancer cell lines in an androgen receptor–dependent and estrogen receptor (ER)-independent manner (41). Studies have also suggested that testosterone can indirectly contribute to breast carcinogenesis through the conversion of androgens to estrogen in mammary tissues (42). While there is some experimental evidence to support a role for testosterone in the etiology of breast cancer among postmenopausal women (10), we are not aware of any experimental studies that have specifically examined the influence of androgens on breast carcinogenesis among premenopausal women.

Elevated levels of SHBG may potentially reduce a woman's risk of breast cancer by suppressing the action of estradiol-controlled genes, such as BCL-2, progesterone receptor, and ERα, which regulate cell growth, apoptosis, and cell estrogen dependence (43, 44). Moreover, SHBG can potentially minimize the carcinogenic effects of estrogen and testosterone by binding to them in plasma, thus reducing the concentration of free circulating levels of these hormones (44, 45). These observations are in line with the results of this study where SHBG was found to be inversely associated with risk of DCIS among postmenopausal women.

In a recent publication, we showed that obese postmenopausal women who were not current users of HRT at baseline had increased risk of DCIS (46). Obesity is known to alter levels of testosterone and SHBG (as well as estrogen) in women (25, 26). Therefore, it is possible that the observed associations between these markers and risk of DCIS among postmenopausal women were partly influenced by body fat levels. However, in this study, the associations of testosterone and SHBG with DCIS risk were adjusted for BMI. Furthermore, we did not observe any heterogeneity in the associations between the sex steroid hormones/SHBG and DCIS by BMI or waist circumference categories among postmenopausal women.

Our study had several limitations. First, we lacked follow-up information on mammographic screening. Therefore, it is likely that some of the noncases were misclassified. However, the UK has a rigorous breast cancer screening program, particularly for women aged 50–70 years, and cancer diagnoses are routinely reported to the cancer registries (47). Therefore, particularly among postmenopausal women, it is likely that most of the DCIS cases were ascertained. Second, there was a relatively small number of DCIS cases among premenopausal women, and thus, limited statistical power to detect associations between the exposures of interest and the outcome. Third, in this study, the assay that was used was not sufficiently sensitive to measure the low estradiol levels typically observed in postmenopausal women (most women had estradiol levels below the assay LOD of 175 pmol/L) and, therefore, we were unable to assess the association between estradiol and risk of DCIS among postmenopausal women. This is not surprising as secretion of estradiol decreases significantly in postmenopausal women due to loss of ovarian function (48). Consistent with this, in studies of invasive breast cancer, the reported levels for estradiol among postmenopausal women have typically fallen below 175 pmol/L (49–51), and, therefore, an LOD of 175 pmol/L may have been too high for postmenopausal women in the UK Biobank and may have contributed to most of these women having undetectable levels of estradiol. Fourth, the observed associations between the hormones and DCIS in this study should be interpreted with caution as the study was based on a single measure of the exposures and therefore the findings may not reflect the impact of longitudinal changes in the hormone levels on risk of DCIS, particularly among premenopausal women who experience cyclical changes in hormone concentrations. Furthermore, the premenopausal group may have included perimenopausal women, as the median age (46 years) among these women was within the age range when women typically become perimenopausal (i.e., mid to late forties; ref. 52). During the perimenopausal phase, there is a continuous decline in the levels of estradiol (53, 54), and hence the inclusion of potentially perimenopausal women may have led to underestimation of the associations between sex hormone levels and risk of DCIS among premenopausal women. Fifth, only a small proportion of non-European women were included in this study (∼5%), and therefore, our findings may not truly represent non-European women. Finally, some breast cancer subtypes are more hormone-sensitive than others. However, we were unable to investigate differences in the association between the exposures and various hormone-receptor subtypes of DCIS because information on receptor subtype is currently unavailable in the UK Biobank.

In conclusion, in this large prospective study, estradiol was positively associated with risk of DCIS among premenopausal women. Furthermore, relatively high testosterone levels were positively associated with risk of DCIS among postmenopausal women while SHBG was inversely associated with risk in this subgroup. These findings may provide valuable information for developing risk prediction models as well as prevention strategies aimed at reducing risk of breast cancer among premenopausal and postmenopausal women.

No potential conflicts of interest were disclosed.

Conception and design: R.S. Arthur, T.E. Rohan

Development of methodology: R.S. Arthur, T.E. Rohan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.E. Rohan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.S. Arthur, X. Xue, T.E. Rohan

Writing, review, and/or revision of the manuscript: R.S. Arthur, T.E. Rohan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.S. Arthur

Study supervision: T.E. Rohan

This work was supported by the Breast Cancer Research Foundation (BCRF-16-137; to T.E. Rohan). This research was conducted using the UK Biobank Resource (project ID: 30247).

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.