Anthropometric measures and serum estrogen metabolism in postmenopausal women: the Women’s Health Initiative Observational Study

Background Several anthropometric measures have been associated with hormone-related cancers. However, it is unknown whether estrogen metabolism plays an important role in these relationships. We examined whether measured current body mass index (BMI), waist-to-hip ratio (WHR), height, and self-reported BMI at age 18 years were associated with serum estrogens/estrogen metabolites using baseline, cross-sectional data from 1835 postmenopausal women enrolled in the Women’s Health Initiative Observational Study. Methods Fifteen estrogens/estrogen metabolites were quantified using liquid chromatography-tandem mass spectrometry. Geometric means (GMs) of estrogens/estrogen metabolites (in picomoles per liter) were estimated using inverse probability weighted linear regression, adjusting for potential confounders and stratified on menopausal hormone therapy (MHT) use. Results Among never or former MHT users, current BMI (≥30 vs. <25 kg/m2) was positively associated with parent estrogens (multivariable adjusted GM 432 vs. 239 pmol/L for estrone, 74 vs. 46 pmol/L for estradiol; p-trend < 0.001 for both) and all of the 2-, 4-, and 16-pathway estrogen metabolites evaluated (all p-trend ≤ 0.02). After additional adjustment for estradiol, unconjugated methylated 2-catechols were inversely associated (e.g., 2-methoxyestrone multivariable GM 9.3 vs. 12.0 pmol/L; p-trend < 0.001). Among current MHT users, current BMI was not associated with parent estrogens but was inversely associated with methylated catechols (e.g., 2-methoxyestrone multivariable GM 216 vs. 280 pmol/L; p-trend = 0.008). Similar patterns of association were found with WHR; however, the associations were not independent of BMI. Height and BMI at age 18 years were not associated with postmenopausal estrogens/estrogen metabolite levels. Conclusions Our data suggest that postmenopausal BMI is associated with increased circulating levels of parent estrogens and reduced methylation of catechol estrogen metabolites, the estrogen metabolism patterns that have previously been associated with higher breast cancer risk. Electronic supplementary material The online version of this article (doi:10.1186/s13058-017-0810-0) contains supplementary material, which is available to authorized users.


Background
Several anthropometric measures have been associated with increased risk of various hormone-related cancers (endometrial, ovarian, and postmenopausal breast cancers) [1][2][3][4][5][6][7]. Authors of meta-analyses have estimated a 54% increased risk of endometrial cancer [1] and a 12% increased risk of postmenopausal breast cancer [6] associated with every 5 kg/m 2 increase in body mass index (BMI). Abdominal adiposity estimated by waist-to-hip ratio (WHR) has also shown strong positive associations with endometrial and postmenopausal breast cancer risk [1,4], although it is unclear whether the associations are independent of BMI. Further, every 10-cm increase in height has been associated with a 15% higher risk of endometrial [1] and ovarian cancers [3]. One of the hypothesized mechanisms for the BMI associations is that overweight and obese postmenopausal women have elevated levels of circulating estrogens [8], because adipocytes produce estrogens via aromatase activity [9]. Height may indicate early-life nutritional status and exposure to high levels of endogenous proliferative hormones such as estrogens in preadolescence [10].
Estrogens play a key role in endometrial, ovarian, and breast carcinogenesis [11,12]. Parent estrogens (estradiol and estrone) stimulate cell proliferation largely via estrogen receptor-mediated mechanisms [13]. Parent estrogens are metabolized via the 2-, 4-, and 16-hydroxylation pathways, leading to an array of metabolites in each pathway. Experimental studies have shown that these metabolites can also stimulate cell proliferation via estrogen receptor-mediated mechanisms, and catechols of 2-and 4-pathways, if not methylated, can induce DNA damage directly by forming quinone DNA adducts or indirectly via redox cycling [14][15][16]. In recent epidemiologic studies, higher levels of parent estrogens and most estrogen metabolites have consistently been associated with higher risk of endometrial and postmenopausal breast cancers [17][18][19][20].
Parent estrogens and estrogen metabolites may play an important role in the associations between anthropometric measures and cancer risk. However, beyond parent estrogens [8,[21][22][23], little is known about the associations between anthropometric measures and circulating estrogen metabolites in postmenopausal women. Further, although current menopausal hormone therapy (MHT) users have higher circulating estrogen/estrogen metabolite levels [24,25], it is unknown whether anthropometrics are associated with further differences in circulating levels. Using cross-sectional data from the Women's Health Initiative (WHI) Observational Study (OS), we examined whether current BMI, WHR, height, and BMI at age 18 years were associated with circulating levels of 15 estrogens/estrogen metabolites in postmenopausal women by MHT use.

Study population
This study includes participants in a case-control study of ovarian and endometrial cancers nested within the WHI-OS [20,26]. The WHI-OS is a prospective cohort study in which researchers recruited 93,676 postmenopausal women aged 50-79 years at 40 clinical centers in the United States between 1993 and 1998 [27,28]. At the baseline clinic visit, anthropometric measures (height, weight, and waist and hip circumferences) were measured and blood samples were collected. Additionally, baseline self-administered questionnaires were used to collect information on participants' demographics, medical history, and health behaviors.
Details of the nested case-control study are described elsewhere [20,26]. In brief, cases were women with ovarian or endometrial cancer diagnosed between study enrollment and 2012. Control subjects were selected from among those women who remained cancer-free at the date of case diagnosis and were frequency-matched to cases on the basis of age at baseline (5-year categories), year of blood draw (1993)(1994)(1995)(1996)(1997)(1998), race/ ethnicity (white, black, Hispanic, other/unknown), hysterectomy at baseline or during follow-up prior to the index date (for ovarian control subjects only), and MHT use (never, ≤1 year since last MHT use, >1 year since last MHT use, current). Participants had no history of cancer (except nonmelanoma skin cancer), bilateral oophorectomy, or hysterectomy (for endometrial control subjects only), and had ≥1.1 ml of serum available.
A total of 1835 women (507 cases and 450 control subjects among never/former MHT users, 457 cases and 421 control subjects among current MHT users), representing 56,109 women when weighted back to the entire cohort, were included in the study. Because all serum samples were collected at baseline prior to cancer diagnosis (mean 6.7 years from baseline to cancer diagnosis), we included both cases and control subjects in this cross-sectional analysis.

Exposure assessment
Baseline measured height and weight as well as waist and hip circumferences were used to calculate current BMI (in kilograms per meter squared) and WHR, respectively. BMI at age 18 years was estimated using height and weight at age 18 years recalled via baseline questionnaire. Current BMI was categorized into three groups using the World Health Organization classification cutpoints for overweight and obesity: <25.0 kg/m 2 (normal), 25.0-29.9 kg/m 2 (overweight), and ≥30 kg/m 2 (obese) [29]. In secondary analyses, we subdivided the obesity category by class (class 1: 30.0-34.9 kg/m 2 , class 2: 35.0-39.9 kg/m 2 , class 3: ≥40 kg/m 2 ) [29] to further examine the potential dose-response relationship between current BMI and estrogens/estrogen metabolites. WHR, adult height, and BMI at age 18 years were categorized into tertile groups based on the distributions in the entire study population as follows: WHR <0.76, 0.76-0.82, ≥0.83; height <160 cm, 160-164.9 cm, ≥165 cm; and BMI at age 18 years <20 kg/m 2 , 20-21.9 kg/m 2 , ≥22 kg/m 2 .

Statistical analyses
Because estrogen/estrogen metabolite levels vary by MHT use [24,25], all analyses were stratified on MHT use (n = 957 never/former users, n = 878 current users). Inverse probability sampling weights were used to adjust the data to represent the population in the entire cohort using methods for secondary phenotype analysis of case-control data described by Li and Gail [30]. Sampling weights were calculated by taking the inverse of sampling fractions: 1 for all cases, and varying weights for control subjects, depending on their strata as defined by matching factors. After log transformation of data to improve normality, geometric means (GMs) in picomoles per liter of individual serum estrogen/estrogen metabolite concentration by exposure categories were estimated using inverse probability weighted linear regression, adjusting for potential confounders: age at blood draw, blood draw year, race, smoking status (never, former, current), and time since menopause (<10 years, 10-19 years, ≥20 years, missing). Additional adjustments for soy and alcohol intake did not change the results and thus were not included in the final models. For current BMI and WHR, we additionally adjusted for physical activity (0, 0.1-9.9, ≥10.0 metabolic equivalents of task-h/week). We performed a test for trend by including the exposure in the model as a continuous variable. The percentage change in GMs between the highest and the lowest categories was estimated by taking the ratio of the GM difference between the two categories over the GM of the lowest category, multiplied by 100. We statistically tested for the difference using a Wald test.
Several secondary analyses were performed. First, for WHR and BMI at age 18 years (Spearman's r ≤ 0.48 with current BMI), we estimated the associations after additional adjustment for current BMI to examine their independent associations. Second, for current BMI among never/former MHT users, we additionally adjusted for unconjugated estradiol to examine whether the associations with individual estrogen metabolites were driven by their correlations with unconjugated estradiol, the estrogen/estrogen metabolite most strongly correlated with current BMI (Spearman's r = 0.46). Next, using pathway groups, we investigated whether current BMI was associated with altered patterns of estrogen metabolism. We compared the mean proportions of parent estrogens out of summed estrogens/estrogen metabolites across BMI categories, with adjustment for the summed concentration of estrogens/estrogen metabolites. Further, because 2-, 4-, and 16-pathway metabolites ("child metabolites") are metabolized from a limited pool of shared precursors (parent estrogens), an increase in the level of one downstream pathway indicates a reduction in levels of other competing pathways. To address this, we modeled proportions of each child metabolite pathway group (2- Finally, because subclinical or underlying diseases may influence the associations, we performed sensitivity analyses after excluding endometrial and ovarian cancer cases (n = 507 never/former MHT users, n = 457 current MHT users); excluding women with a history of diabetes at baseline (n = 57 never/former MHT users, n = 28 current MHT users); excluding women with low BMI (<18.5 kg/m 2 ; n = 14 never/former MHT users, n = 14 current MHT users); and, among never/former MHT users, excluding women using any potential hormonerelated medications (e.g., antiestrogens, glucocorticoid steroids; n = 303). To test the robustness of our results, we also repeated analyses after excluding outliers (≤10 values for never/former MHT users, ≤10 values for current MHT users) identified using the extreme Studentized deviate many-outlier procedure [31] and after further stratifying by never vs. former MHT users (n = 645 never MHT users, n = 312 former MHT users).
All statistical tests were two-sided with a 5% type I error rate. Q values reflecting the false discovery rates (FDRs) were calculated to account for multiple comparisons (25 tests per exposure). Analyses were conducted with SAS version 9 software (SAS Institute, Cary, NC, USA).

Study population characteristics
The mean age was 64.6 years for never/former MHT users and 61.3 years for current MHT users (Table 1). Most women were white (89% among never/former users, 94% among current users). Compared with never/ former MHT users, current users were more likely to have been postmenopausal for <10 years (44% vs. 32%). Mean current BMI, BMI at age 18 years, height, and WHR were similar between the two groups. As expected, serum concentrations of all evaluated estrogens/ estrogen metabolites were higher in current MHT users than in never/former users (Tables 2 and 3).

Body mass index at age 18 years
Among never/former MHT users, BMI at age 18 years (≥22 vs. <20 kg/m 2 ) was positively associated with estradiol (64.1 vs. 49.8 pmol/L; p-trend = 0.04) but was not associated with other estrogens/estrogen metabolites (Additional file 5: Table S3). The association with estradiol did not remain after adjustment for current BMI (data not shown), suggesting that the association may be accounted for by current BMI. Among current MHT users, BMI at age 18 years was not associated with estrogens/estrogen metabolites (Additional file 5: Table S3).
There was no statistically significant interaction between BMI at age 18 years and current BMI (p-interaction >0.05) (data not shown).
Most associations remained at 5% FDR (Tables 2, 3, 4 and 5, Additional file 1: Table S1, Additional file 4: Table S2, Additional file 5: Table S3). Results were similar after excluding endometrial or ovarian cancer cases at the time of last follow-up (Additional file 6: Table  S4), after excluding women who reported a history of diabetes at baseline (data not shown), after excluding women with low BMI (<18.5 kg/m 2 ) (data not shown), after excluding women using any potential hormonerelated medications (data not shown), and after further stratifying by never vs. former MHT users (data not  shown). Results also did not change after excluding outliers (data not shown).

Discussion
To the best of our knowledge, this study is the first to examine the relationships of anthropometric measures with detailed serum estrogen/estrogen metabolite measures in postmenopausal women by MHT use. In this cross-sectional analysis, current measured BMI was positively associated with parent estrogens and all estrogen metabolites evaluated among never/former MHT users. Adjusted for unconjugated estradiol, current BMI was inversely associated with methylated 2-catechols but not with the other estrogen metabolites measured, suggesting that associations with individual estrogen metabolites among never/former MHT users were driven largely by their correlations with unconjugated estradiol. Associations with measured WHR were not independent of current BMI, suggesting that the associations were driven by overall adiposity rather than by fat distribution per se. BMI at age 18 years and height were not associated with postmenopausal estrogen/estrogen metabolite levels.
Our findings of positive associations between current BMI and parent estrogens are consistent with those from previous studies [8,[21][22][23][32][33][34]. A pooled analysis of eight studies estimated 83% higher levels of circulating total estradiol and 60% higher levels of total estrone in postmenopausal obese women (BMI ≥30 kg/m 2 ) compared with lean women (BMI <22.5 kg/m 2 ) [8]. These findings are also in line with biological evidence that supports the major source of estrogens in postmenopausal women being via aromatization of androgens in adipose tissue [35,36]. Further, although few studies have examined the associations in current MHT users, our findings are consistent with data from the Nurses' Health Study regarding unconjugated estradiol BMI (≥30 vs. <25 kg/m 2 ) associations in both never/former and current MHT users and the finding of stronger positive associations in never/former users (62% vs. 22% change in estradiol levels) [23]. In the present study, unconjugated estradiol was consistently positively associated with current BMI (≥30 vs. <25 kg/m 2 ) among both never/former (130% change) and current MHT users (20% change), although the association was not statistically significant in current MHT users. The differential associations between BMI and parent estrogens by MHT use mirror the similar differential associations of BMI with cancer risk [37,38], further supporting the notion that endogenous estrogens may mediate the BMI-cancer association, although alternative mechanisms may also exist [24,39,40]. The positive associations of BMI with postmenopausal breast cancer [37,38] and endometrial cancer [41,42] risk have been found consistently in women not using MHT; however, the associations have been weakly positive or near null among current MHT users [37,38,41,42], possibly owing to a threshold effect of circulating estrogens in current MHT users. Current MHT users have higher circulating levels of estrogens; thus, increased estrogen production by adipose tissue may not contribute to further increase in cancer risk. Moreover, given the high estrogen levels in current MHT users, the association with BMI on the relative scale may be masked, whereas an absolute difference in estrogen concentration may be apparent, as in never/former users.  (1993)(1994)(1995)(1996)(1997)(1998), race (white, non-white), smoking status (never, former, current), time since menopause (<10, 10-19, ≥20 years, missing), moderate-to vigorous-intensity physical activity (0, 0.1-9.9, ≥10 MET-hr/wk), summed concentration of estrogens/estrogen metabolites (continuous, pmol/L). Note: Summed estrogens/estrogen metabolites include the summed concentration of parent estrogens (estrone, estradiol) and child metabolites (2-hydroxyestrone, 2-hydroxyestradiol, 2-methoxyestrone, 2-methoxyestradiol, 2-hyroxyestrone-3-methyl ether, 4-hydroxyestrone, 4-methoxyestrone, 4-methoxyestradiol, 16α-hydroxyestrone, estriol, 16-ketoestradiol, 16-epiestriol, 17-epiestriol). a Overall F-test p=0.02. b P-value for comparing percent parent estrogens between women with current BMI 25-29.9 vs. <25 kg/m 2 was 0.11. c P-value for comparing percent parent estrogens between women with current BMI ≥30 vs. 25-29.9 kg/m 2 was 0.10. P-value for comparing percent parent estrogens between women with current BMI ≥30 vs. <25 kg/m 2 was 0.006.
In the present study, we did not observe associations of WHR with estrogens/estrogen metabolites after adjustment for current BMI. Similarly, in studies where body fat distribution was measured by dual-energy X-ray absorptiometry [32] or measured WHR [33], central obesity was not associated with circulating unconjugated estradiol independent of BMI among postmenopausal women not using MHT, suggesting that body fat distribution does not provide additional information about circulating estradiol beyond what overall adiposity (as measured by BMI) provides.
In the present study, adult height and BMI at age 18 years were not associated with estrogens/estrogen metabolites after adjustment for current BMI. BMI at age 18 years and height may indicate early-life nutritional status, which may not influence estrogen metabolism in postmenopausal women. In a study of premenopausal women, BMI at age 18 years was not associated with urinary estrogen/estrogen metabolite levels; however, contrary to our findings, height was inversely associated with urinary parent estrogens and 2-pathway metabolites in the same study [48]. The difference in results may be due to the variation in study populations (e.g., age, menopausal status) or biospecimen used (circulating vs. excreted levels).
We acknowledge several limitations of this study. We measured circulating estrogens/estrogen metabolites in a single baseline serum sample. However, in a previous study using the same assay we used, researchers showed moderate to high 1-year ICCs in postmenopausal women for parent estrogens (0.72 for estrone, 0.65 for estradiol) and most estrogen metabolites of 2-, 4-, and 16-pathways (range 0.35-0.53), with the exception of 2methoxyestrone (0.10) and 2-hydroxyestrone-3-methyl ether (0.14) [17], suggesting that measured serum estrogens/estrogen metabolites may also adequately represent postmenopausal levels over at least 1 year. BMI at age 18 years was based on self-reported height and weight. However, measurement error in this context is unlikely to be related to serum estrogen levels and, if present, would likely underestimate the associations. Although we cannot exclude the possibility of false-positives, most associations with current BMI remained at 5% FDR.
Despite these limitations, this study has important strengths. Whereas most epidemiologic studies have used self-reported measures, measurement error in the present study was reduced by using measured height, weight, and waist and hip circumferences. Use of a high-performance LC-MS/MS assay also allowed comprehensive evaluation of individual estrogens/estrogen metabolites with high reliability, sensitivity, and specificity. By stratifying the analysis by MHT use, we were able to take into account variations in estrogen/estrogen metabolite levels by exogenous hormone use. Further, use of a large sample size and careful adjustment for potential confounders assessed at blood collection increased the validity of the results.

Conclusions
In this comprehensive analysis of measured anthropometrics and serum estrogens/estrogen metabolites, we observed strong positive associations between current BMI and parent estrogens in postmenopausal women not using MHT. Adjusted for unconjugated estradiol, current BMI was also associated with lower levels of methylated catechols among both never/former and current MHT users. Similar estrogen metabolism patterns have previously been associated with higher breast cancer risk; thus, these findings further support the need to prospectively evaluate the roles of endogenous estrogens/estrogen metabolites in the BMI-breast cancer risk association.