Interactions of Hormone Replacement Therapy, Body Weight, and Bilateral Oophorectomy in Breast Cancer Risk

It is widely accepted that female sex hormones, particularly estrogens, play a pivotal role in the etiology of breast cancer. Among postmenopausal women, high body adiposity, typically measured using body mass index (BMI), has been established as a risk factor for breast cancer. This positive association is due to increased endogenous estrogen synthesis in adipose tissues among postmenopausal women (1–3). Exogenous estrogen administration through hormone replacement therapy (HRT) has also been associated with elevated risk of breast cancer (1–3). Numerous studies have reported that the association of overweight/obesity with postmenopausal breast cancer risk is significantly attenuated in women who use HRT (4–11), suggesting that body weight and HRT use may interact in associations with breast cancer risk among postmenopausal women. Furthermore, a recent analysis of data from the Women's Health Initiative (WHI) randomized clinical trial found that among postmenopausal women with prior hysterectomy, ever-use of estrogen during the intervention phase, compared with the placebo group, was associated with a significantly reduced risk of breast cancer (12). It is unclear, however, if the association may differ by types of surgeries, such as simple hysterectomy or hysterectomy plus bilateral oophorectomy. Nevertheless, these recent findings may challenge existing concepts about the association between HRT use and breast cancer risk (12, 13).

Breast cancer is a complex and heterogeneous disease with a wide spectrum of clinical, histopathologic, and molecular features (14–16). Increasing evidence suggests that breast cancer subtypes defined by expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (Her2) represent distinct biologic entities with distinct clinical profiles (17–19). For example, ER⁺ tumors are associated with overexpression of genes in the ER signaling pathways and, clinically, have the most favorable prognosis; whereas triple-negative (ER⁻PR⁻Her2⁻) tumors are most likely to exhibit a basal-like pattern of gene expression and are associated with more aggressive histopathologic features and poor prognosis (16–18). Hormone-related risk factors including obesity and HRT use have been shown to be more closely related to hormone receptor–positive breast cancer; however, data are very limited about the possible interactions of HRT, body weight, and bilateral oophorectomy in the risk of breast cancer by subtypes.

The Nashville Breast Health Study (NBHS) is a large population-based case–control study of breast cancer with a primary objective to identify genetic and lifestyle risk factors for this common malignancy. Using data from the NBHS, we examined associations between ever-HRT use and breast cancer risk by subtype, according to ER, PR, and Her2 status, and further determined whether these associations may be modified by body weight and bilateral oophorectomy.

The NBHS is a population-based case–control study of incident breast cancer conducted primarily in the Nashville metropolitan area of Tennessee. Eligible cases were women newly diagnosed with primary breast cancer (invasive cancer or ductal carcinoma in situ) between 25 and 75 years of age and with no prior history of cancer other than nonmelanoma skin cancer. Most participants (92%) were residents of the Nashville eight-county metropolitan area. From February 1, 2001 through December 31, 2011, 5,078 women were recruited into the study. Breast cancer cases (n = 2,694) were identified and recruited through a rapid case-ascertainment system established across the major hospitals in Nashville and the Tennessee Cancer Registry. Information on ER, PR, and Her2 status of breast cancer tumors was obtained from pathology records. Controls (n = 2,384) were identified primarily via random-digit dialing of households in the Nashville eight-county metropolitan area, and were frequency-matched to cases on 5-year age groups, race, and county of residence. Approval for the study was obtained from the Institutional Review Boards of Vanderbilt University Medical Center and each collaborating institution. All study participants signed informed consent to participate in an epidemiologic survey to provide lifestyle and demographic data, release relevant medical information, and provide a saliva sample as a source of genomic DNA for genetic studies of breast cancer.

In the NBHS, the median interval from time of breast cancer diagnosis to study enrollment was 10.4 months. Participation rates were approximately 58% for cases and 48% for controls. Reasons for nonparticipation included refusal (n = 1,554), not completing the interview (n = 220), death (n = 206), illness (n = 13), and inability to be reached (n = 1) among cases; and refusal (n = 614), illness (n = 7), death (n = 5), and others (n = 124) among controls (Fig. 1). Information on sociodemographic characteristics and all major breast cancer risk factors was obtained through telephone interview by trained interviewers using a structured questionnaire. Reference date was defined as date of breast cancer diagnosis for cases and date of interview for controls. Menopause was defined as cessation of menstrual periods, excluding those caused by pregnancy and nursing, for at least 12 months before the reference date. For postmenopausal women, cause of menopause was further assessed, including natural menopause, surgical menopause (hysterectomy with no, one, or two ovaries surgically removed), medication-induced menopause, and unknown reasons. Information on HRT use was collected: women who indicated ever-use of HRT were asked ages at first and last use, which were used to calculate cumulative duration of HRT. BMI was defined as weight (kg)/height² (m²).

Among the 5,078 NBHS participants, 3,228 were postmenopausal women. Of these, 643 non-Hispanic Black women, 31 Hispanic women, and 44 women in other racial/ethnic groups were excluded from this analysis because of small sample size for a separate analysis. Thus, included in this analysis were 2,510 postmenopausal white women (1,273 breast cancer cases and 1,237 healthy controls); 1,185 with natural menopause, 1,127 with surgical menopause, 116 with medication-reduced menopause, and 82 with other reasons (not specified). Among cases, data on ER, PR, and Her2 status were available for 76.0%, 75.2%, and 61.3% of cases, respectively.

In this analysis, breast cancer subtypes were classified by hormone receptor (ER and PR) and Her2 status into the following groups and subgroups: ER status including ER⁺ and ER⁻; ER/PR status focusing on ER⁺PR⁺ and ER⁻PR⁻; and ER/PR/Her2 status including (i) luminal A (ER⁺ and/or PR⁺ and Her2⁻), (ii) luminal B (ER⁺ and/or PR⁺ and Her2⁺) and Her2 overexpressing (ER⁻PR⁻Her2⁺), and (iii) triple negative (ER⁻PR⁻Her2⁻; refs. 16, 20).

Distributions of demographic characteristics and selected risk factors between cases and controls were compared using t tests (for continuous variables) or χ² tests (for categorical variables). We used multivariable unconditional logistic regression to estimate ORs and their 95% confidence intervals (CI) for the association between HRT use (ever-use of HRT and duration of use) and overall breast cancer risk. To estimate OR and 95% CI for associations between HRT use and breast cancer subtypes (ER status, ER/PR status, ER/PR/Her2 status), we used multivariable polytomous unconditional logistic regression that enables simultaneous calculation of association results for multiple outcome categories (21, 22). Age was adjusted, along with education, to control for potential influence of socioeconomic status on study results. None of area of residence and known risk factors for breast cancer, including history of breast cancer among first-degree relatives, personal history of benign breast diseases, regular alcohol consumption, physical inactivity, early age at menarche, ever-use of oral contraceptives (OC), late age at first birth, parity, late-age menopause, and long duration of menstruation, were associated with HRT use and thus they are not confounders in this analysis. Therefore, they were not adjusted in the study.

To examine the potential modifying effects of body weight on association between HRT use and breast cancer risk, we categorized women into two groups: thin or normal weight women (BMI < 25 kg/m²) and overweight (BMI ≥ 25 kg/m²), based on World Health Organization (WHO) criterion. Among overweight women, additional analyses were performed for pre-obesity (BMI = 25–29.9 kg/m²) and obesity (BMI ≥ 30 kg/m²). For women with surgical menopause, associations of HRT use with breast cancer risk were further examined separately among women who had prior hysterectomy with bilateral oophorectomy (surgical removal of both ovaries) or prior hysterectomy without bilateral oophorectomy (no or only one ovary surgically removed). Tests for trend across categories of duration of HRT use were performed by entering categorical variables as continuous variables in the model. Interaction terms were included in the models to test for interaction between HRT use and BMI. All P values reported were two-sided. Statistical analyses were performed using SAS version 9.2 (SAS Institute).

Table 1 shows characteristics of breast cancer cases and healthy controls among postmenopausal white women in the NBHS, presented overall and by type of menopause (natural or surgical). Overall, compared with controls, cases were slightly older (59.9 years vs. 59.1 years) and were more likely to have family history of breast cancer, personal history of benign breast disease (BBD), lower annual household income, and less regular exercise. Other factors, including BMI, were generally comparable. Compared to those with natural menopause (both cases and controls), women with surgical menopause were slightly younger at study enrollment and reported age at menopause about 10 years younger, resulting in total years of menstruation about 10 years shorter. Notably, when women with surgical menopause were further stratified into two groups, prior hysterectomy with bilateral oophorectomy or without bilateral oophorectomy, age, BMI, age at menopause, and years of menstruation were also comparable between cases and controls (data not shown).

As shown in Table 2, among women with natural menopause and BMI < 25 kg/m², ever-use of HRT was associated with significantly elevated overall risk of breast cancer (OR, 1.95; 95% CI, 1.32–2.88). Risk increased with increasing total duration of HRT use (OR, 1.67; 95% CI, 0.99–2.81, for use <5 years; OR, 1.93; 95% CI, 1.13–3.31, for use for 5–9 years; and OR, 2.13; 95% CI, 1.26–3.59, for use ≥10 years; P for trend = 0.002). In contrast, among women with BMI ≥ 25 kg/m², no association of ever-HRT use or duration of use was observed with overall risk of breast cancer (OR, 0.91; 95% CI, 0.67–1.25; P for trend = 0.706). Significant interaction between BMI and ever-HRT use (P for interaction = 0.001) or between BMI and duration of HRT use (P for interaction = 0.012) was detected. These two interaction tests remained statistically significant after adjusting for multiple comparisons (adjusted P value < 0.025 for two comparisons). Adjusting for additional variables, including area of residence and known breast cancer risk factors including history of breast cancer among first-degree relatives, personal history of benign breast diseases, regular alcohol consumption, physical inactivity, age at menarche, parity, and years of menstruation did not materially change study results (data not shown).

We further evaluated associations of ever-use of HRT and duration of use with risk of breast cancer defined by ER, PR, and Her2 status among women with natural menopause. As shown in Tables 3 and 4, among women with BMI < 25 kg/m², positive associations with HRT use were most pronounced for breast cancer defined as ER⁺, ER⁺PR⁺, and luminal A, whereas more modestly elevated risks associated with HRT use were seen for ER⁻, ER⁻PR⁻, luminal B/Her2 overexpression or triple-negative tumors. With the exception of the triple-negative subtype, interaction tests between BMI and HRT use or duration of HRT use were statistically significant at P < 0.05 and remained mostly statistically significant even after Bonferroni correction for multiple comparisons.

Table 5 shows associations between ever-use of HRT and breast cancer risk among postmenopausal women with prior hysterectomy with or without bilateral oophorectomy. Among women with prior hysterectomy without bilateral oophorectomy, ever-use of HRT was associated with a marginally significant increase in overall risk of breast cancer (OR, 1.45; 95% CI, 0.92–2.29). In contrast, among women with prior hysterectomy with bilateral oophorectomy, ever-use of HRT was associated with reduced risk of breast cancer (OR, 0.70; 95% CI, 0.38–1.31). This pattern of association was seen for virtually all breast cancer subtypes, although interaction tests were statistically significant for ER⁺ and luminal A tumors only (P = 0.034 and 0.017, respectively). No statistically significant interactions between HRT use and BMI were detected among women with prior hysterectomy with or without bilateral oophorectomy; however, sample size for these analyses was small.

It has been hypothesized that risk factors most closely associated with postmenopausal ER⁺PR⁺ breast tumors may operate through mechanisms related to estrogen and progesterone exposure, whereas the etiology of ER⁻PR⁻ breast cancer may be independent of hormonal exposure (23). Many studies have examined associations between HRT use and risk of breast cancer stratified by hormone-receptor status. Results, although not entirely consistent, suggest that positive association with HRT use is restricted to, or stronger, among women with ER⁺ or ER⁺PR⁺ tumors compared with ER⁻ or ER⁻PR⁻ tumors (3, 11, 24–31). The Nurses' Health Study (NHS) reported a stronger association between past use of HRT and ER⁺ tumors compared with ER⁻ tumors (24). Li and colleagues reported that HRT use was associated with a 2-fold increased risk of ER⁺ tumors only, with higher risk for current long-term use (OR, 2.9; 95% CI, 1.8–4.8; ref. 26). In our study, as expected, ever-HRT use showed a stronger, approximately 2.7-fold increased risk of ER⁺PR⁺ tumors, with a 3.4-fold increased risk with long-term HRT use, compared with a 1.5-fold increased risk of ER⁻PR⁻ tumors in postmenopausal women with normal weight.

Cumulative evidence suggests that HRT use may interact with adiposity for breast cancer risk (4–11, 26, 29, 32). The NHS reported a positive association of both waist-to-hip ratio and waist circumference with breast cancer risk, but only in postmenopausal women who had never received HRT (4). The WHI observational study (6) reported that compared with slimmer women (BMI < 22.6 kg/m²), heavier women (BMI > 31.1 kg/m²) had an elevated risk of postmenopausal breast cancer (RR, 2.52; 95% CI, 1.62–3.93), which was only apparent among HRT nonusers. Similar findings were reported in the Million Women Study conducted in England and Scotland (8) and the European Prospective Investigation into Cancer and Nutrition (EPIC; ref. 7). In a pooled analysis of more than 50 epidemiologic studies (Collaborative Group on Hormonal Factors in Breast Cancer), association of HRT use with breast cancer was found among women with BMI < 25 kg/m², but not among heavier women (29).

Data on whether adiposity modifies the association of HRT use with risk of breast cancer by hormone receptor subtypes are very limited. Recently, the EPIC study reported that current use of HRT, compared with never-use of HRT, was significantly associated with increased risk of both ER⁺ and ER⁻ tumors, although more strongly with the former. Associations with HRT were significantly stronger in leaner women (BMI ≤ 22.5 kg/m²) than overweight women (BMI ≥ 25.9 kg/m²), and HRs were statistically significant in leaner women for both ER⁺PR⁺ (HR, 2.33; 95% CI, 1.84–2.92) and ER⁻PR⁻ (HR, 1.74; 95% CI, 1.15–2.63) breast cancer (11). These estimates are remarkably similar to the corresponding ORs for HRT use among lean women in our study. Results from a case–control study in the Seattle–Puget Sound metropolitan area (26) were not entirely consistent. A positive association between estrogen plus progestin hormone therapy and ER⁺PR⁺ breast cancer was observed regardless of BMI. It was strongest among lean women (BMI < 25 kg/m²), although the interaction was not statistically significant. Moreover, in contrast with the results of our study and the EPIC report, HRT was not associated with the risk of ER⁻PR⁻ breast cancer, regardless of BMI. Our study has extended previous results by including for the first time Her2 status in the classification of breast cancer subtypes. When Her2 status was considered, associations between HRT use and risk of luminal A (ER⁺ and/or PR⁺ and Her2⁻), and luminal B (ER⁺ and/or PR⁺ and Her2⁺) and Her2 overexpressing (ER⁻PR⁻Her2⁺) subtypes were much stronger than the association between HRT use and risk of triple-negative breast cancer (ER⁻PR⁻Her2⁻). Such associations were only seen in women with BMI < 25 kg/m², and not in overweight women (BMI ≥ 25 kg/m²). Thus, our results provide additional evidence that the relationship between ever-use of HRT and risk of breast cancer subtypes defined by HR and Her2 status is modified by body weight.

The biologic mechanisms underlying this interaction have been postulated but remain unclear. The positive relationship between body weight and breast cancer risk among postmenopausal women is generally attributed to an increased conversion of adrenal androgens to estrogens by aromatase within the larger adipose stores of overweight women, as well as decreased circulating sex hormone–binding globulin (SHBG), which leads to more bioavailable estrogens (33–35). Thus, in general, overweight postmenopausal women have higher circulating estrogen levels than normal-weight women. Among HRT users, however, the proportional increase in circulating estrogen levels from exogenous estrogen among postmenopausal women may be relatively smaller in overweight/obese women compared with normal-weight women (36, 37). In addition, overweight women tend to have a higher prevalence of insulin resistance and hyperinsulinemia. Insulin itself is a mitogenic agent, and chronic hyperinsulinemia is also associated with downregulation of SHBG and insulin-like growth factor binding protein-1 and 2, leading to increased levels of bioavailable estrogen, insulin-like growth factor-1 (IGF-1), and testosterone (38). Long-term treatment with low estrogen doses among overweight women may improve insulin resistance and reduce elevated insulin levels, thereby attenuating the stimulatory effects of insulin on tumor growth and resulting in a reduction of breast cancer risk compared with HRT treatment in leaner women.

A few studies have found that HRT use was associated with reduced breast cancer risk in women with surgically induced menopause (12, 39, 40). In a case–control study of 472 postmenopausal women with a BRCA1 mutation, Eisen and colleagues found that among women with surgically induced menopause, ever-HRT users had a lower breast cancer risk compared with nonusers (OR, 0.48; 95% CI, 0.19–1.21; ref. 39). The WHI study recently reported that among postmenopausal women with prior hysterectomy, ever-use of estrogen during the 5.9-year intervention phase was associated with decreased breast cancer risk after follow-up of more than 10 years (HR, 0.77; 95% CI, 0.62–0.95; ref. 12). In line with these previous findings, our study also observed significantly reduced breast cancer risk associated with HRT use among women with surgically induced menopause, but only among women who had a hysterectomy with both ovaries removed. Among those who had a hysterectomy with no or one ovary removed, ever-use of HRT was associated with a nonsignificant increase in breast cancer risk.

Some in vitro experiments have shown that estrogen deprivation of hormone-dependent MCF-7 breast cancer cells causes them to undergo adaptive changes in which estrogen switches from being a proliferative agent to paradoxically inhibiting growth and inducing apoptosis (42). Thus, it has been postulated that administration of estrogen through HRT use may induce apoptosis of breast cancer cells that are present among women with bilateral oophorectomy, and thus reduce breast cancer risk. However, it is possible that HRT use among women with bilateral oophorectomy is an indicator of early age at oophorectomy, and thus shorter duration of endogenous estrogen exposure, as postmenopausal women with bilateral oophorectomy typically do not use HRT. Indeed, in our study, women with bilateral oophorectomy had an early age of menopause. Furthermore, bilateral oophorectomy itself is also reported to be associated with reduced breast cancer risk (41). Therefore, the reduced risk associated with HRT use among women with bilateral oophorectomy may not be due to a protective effect of exogenous estrogen exposure among these women, as suggested by in vitro experiments. Instead, this association could be due to confounding effects that cannot be controlled in our study.

Several limitations of our study should be acknowledged. First, approximately 30% of breast cancer cases in our study did not have information available about receptor status, which might introduce selection bias. However, in our study sample, prevalence rates of ER⁺, PR⁺, and Her2⁺ were 77.4%, 62.6%, and 23.8%, respectively, consistent with many previous large-scale studies conducted among white women (16, 20, 24). Second, as with all case–control studies, our study is subject to recall bias, especially as we relied on self-reported information on reproductive factors and HRT use. Several studies, however, have shown generally consistent agreement between self-report and medical records about reproductive factors and hormone use in postmenopausal women (46–48). Furthermore, the stronger association of exogenous estrogen use with ER⁺ tumors observed in this study argues against the effect of recall bias on our study results. Third, breast cancer intrinsic subtypes in our study were defined on the basis of ER, PR, and HER2 status without information on Ki-67 (a proliferation marker). Because luminal B tumors include two subgroups (ER⁺ and/or PR⁺, Her2⁺, any Ki-67 and ER⁺ and/or PR⁺, Her2⁻, Ki-67 high), it is likely that some cases included in the luminal A group (ER⁺ and/or PR⁺, Her2⁻, Ki-67 low) may actually represent luminal B tumors (ER⁺ and/or PR⁺, Her2⁻, Ki-67 high). Fourth, multiple ORs were estimated in each analysis of interaction between HRT use (or duration of HRT use) and breast cancer risk, overall or by subtypes. Some ORs were statistically significant, perhaps due to multiple comparisons. However, the focus of this study is to identify potential interaction, for which only a limited number of tests were performed, and virtually all significant interactions identified in this study remained significant after adjusting for multiple comparisons.

In summary, our study clearly shows that the relationship between HRT use and breast cancer risk among postmenopausal women is modified by body weight and prior bilateral oophorectomy. Among women having natural menopause, ever-HRT use was significantly associated with increased risk of breast cancer, especially ER⁺, ER⁺PR⁺, luminal A, and luminal B and Her2 overexpression subtypes in women of normal weight (BMI < 25 kg/m²). Such associations, however, were not seen among overweight women (BMI ≥ 25 kg/m²). Among women with prior bilateral oophorectomy, ever-HRT use was associated with reduced breast cancer risk for ER⁺, ER⁺PR⁺, and luminal A tumors. These results may be helpful in recommending HRT use among postmenopausal women and identifying high-risk women among HRT users. Further investigation and clarification of underlying mechanisms contributing to these effects are warranted.

No potential conflicts of interest were disclosed.

Conception and design: M.J. Shrubsole, W. Zheng

Development of methodology: M.J. Shrubsole, X.-O. Shu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.L. Deming-Halverson, M.J. Shrubsole, X.-O. Shu, W. Zheng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Cui, S.L. Deming-Halverson, A. Beeghly-Fadiel, L. Lipworth, W. Zheng

Writing, review, and/or revision of the manuscript: Y. Cui, S.L. Deming-Halverson, A. Beeghly-Fadiel, L. Lipworth, M.J. Shrubsole, A.M.Fair, X.-O. Shu, W. Zheng

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Beeghly-Fadiel, M.J. Shrubsole

Study supervision: M.J. Shrubsole, W. Zheng

The authors thank the study participants and research staff of the NBHS for their support of this research, and Mary Jo Daly for technical assistance in article preparation.

This work was supported by a research grant (R01CA100374) from the U.S. National Cancer Institute. Surveys for this study were conducted by the Biospecimen and Survey Shared Resource, supported in part by P30CA68485.

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.