Racial/Ethnic Differences in Sex Hormone Levels among Postmenopausal Women in the Diabetes Prevention Program

Abstract

Context:

Sex hormones may differ by race/ethnicity in postmenopausal women. Whether racial/ethnic differences also exist among those who are overweight and glucose intolerant is not clear.

Objectives:

The objective of the study was to compare sex hormones by race/ethnicity [non-Hispanic white (NHW), Hispanic, African-American (AA)] in overweight, glucose-intolerant, postmenopausal women.

Design:

This was a secondary analysis of a randomized controlled trial.

Participants:

Participants included postmenopausal glucose-intolerant women participating in the Diabetes Prevention Program.

Interventions:

Interventions included intensive lifestyle modification (consisting of diet and physical activity) or metformin 850 mg twice a day vs. placebo.

Main Outcome Measures:

Baseline levels and 1-yr intervention-related changes in SHBG, total and bioavailable estradiol (E2), total and bioavailable testosterone, and dehydroepiandrosterone were measured.

Results:

At baseline, among women not using estrogen (n = 370), NHW had higher total and bioavailable E2 and testosterone levels than Hispanics independent of age, type of menopause, waist circumference, alcohol intake, and current smoking. NHW also had higher levels of bioavailable E2 and lower levels of SHBG than AA. At baseline, among estrogen users (n = 310), NHW had higher total and bioavailable E2 than Hispanics and higher levels of SHBG than AA after adjustment. At 1 yr, among women not using estrogen, NHW had larger declines in total E2 and bioavailable E2 levels than AA after adjustment for the above covariates, changes in waist circumference, and randomization arm. At 1 yr, among estrogen users, sex hormone changes did not differ by race/ethnicity.

Conclusions:

Among postmenopausal women, there were significant race/ethnicity differences in baseline sex hormones and changes in sex hormones.

SHBG, estradiol (E2), testosterone (T), and dehydroepiandrosterone (DHEA) and its sulfate are associated with diseases in midlife women including breast cancer, (1) endometrial cancer, (2) and lipid abnormalities (3). Population-based studies conflict as to whether sex hormone levels vary between midlife women of European, African, and Hispanic race/ethnicity (4–7). In an analysis of perimenopausal women in the Study of Women's Health Across the Nation (SWAN) (4, 5), non-Hispanic white women (NHW) had higher T and dehydroepiandrosterone sulfate levels than African-American women (AA) and higher T than Hispanics, after adjustment for body size. In contrast, examinations of postmenopausal women in the Multi-Ethnic Cohort Study (MEC) (6) and the Multi-Ethnic Study of Atherosclerosis (MESA) (7, 8) did not find racial/ethnic differences in T after adjustment for body mass index (BMI). In the Women's Health Initiative (WHI) Dietary Modification Trial, NHW had lower calculated free T levels than AA (9). The MEC observed that NHW had lower bioavailable and total E2 compared with AA (6), but E2 did not differ by race/ethnicity in SWAN, MESA, or the WHI (4, 5, 7, 9, 10). Results may vary between studies due to differing definitions of menopause: in MESA, women could be in menopause naturally or due to oophorectomy (7), whereas in the MEC and WHI, women could be menopausal naturally, due to oophorectomy or hysterectomy alone (6, 9). SWAN reported on women in premenopause, adjusting for day of hormone draw in the menstrual cycle (4, 5). Examination of postmenopausal women in SWAN examined E2 and not other sex steroids; E2 levels still did not vary significantly between NHW, AA, and Hispanics (10). Differences in methods used to determine total and bioavailable hormone levels may also account for inconsistent findings.

To our knowledge, no prior studies, either population based or among selected samples, have reported on racial/ethnic comparisons of endogenous sex hormone levels in overweight postmenopausal women with impaired glucose tolerance. These women are at increased risk of comorbid conditions, such as malignancy (11) and diabetes (12) compared with normal-weight, glucose-intolerant women. Potentially, racial/ethnic variation in sex hormones in this higher-risk population may contribute to racial/ethnic variation in the incidence and consequences of comorbid conditions linked to sex hormones. In addition, randomized trials of metabolic interventions in postmenopausal women are few and have enrolled primarily NHW (13–15), and it is unknown whether changes in sex hormones vary by race/ethnicity. In addition, racial/ethnic comparisons of sex hormones have largely focused on women not using estrogen therapy, although studies examining racial/ethnic differences in sex hormone-associated comorbidities generally include women using estrogen therapy (16). In one report of E2 levels among women using exogenous estrogen therapy, NHW required higher doses of estrogen to achieve similar levels of E2 as nonwhite women, suggesting that racial/ethnic differences in sex hormones levels could exist, even among estrogen users (17, 18). Therefore, we compared baseline and changes in sex hormone profiles in postmenopausal women participating in the Diabetes Prevention Program (DPP).

Materials and Methods

The DPP randomized nondiabetic, glucose-intolerant participants to a program of intensive lifestyle modification, metformin, or placebo (19). Characteristics of DPP participants (19) and the Sex Hormones in Postmenopausal Women ancillary study have been previously described (20). Briefly, the DPP inclusion criteria included age 25 yr or older, fasting plasma glucose of 95–125 mg/dl and 2-h plasma glucose of 140–200 mg/dl after a 75-g glucose load, and a BMI of 24 kg/m² or greater. Eligible participants recruited between 1996 and 1999 were randomly assigned to one of three interventions: 850 mg metformin twice daily, placebo twice daily, or intensive lifestyle modification. Weight and waist circumference were measured semiannually. Written informed consent was obtained from all participants before screening, consistent with the guidelines of each participating center's institutional review board.

Sex Hormones in Postmenopausal Women study participants included women who were postmenopausal at DPP randomization and who provided consent to participate in ancillary studies. Women were classified as postmenopausal if they met any of the following criteria: bilateral oophorectomy, lack of menses for at least 1 yr and retaining the uterus and at least one ovary, cessation of menses before hysterectomy, cessation of menses within the past year and age 55 yr or older, and hysterectomy at age 55 yr or older.

SHBG, FSH, total E2, total T, and DHEA were measured on heparinized plasma collected at baseline and year 1. SHBG was measured at Endoceutics (Québec City, Canada) using an ELISA (Bioline, Brussels, Belgium) with interassay coefficients of variation of 7.8 and 5.0 at 18.2 and 63.1 nmol/liter, respectively. FSH was measured at Endoceutics using an ELISA (Bioline) with interassay coefficients of variation of 3.6 and 4.4 at 27.1 and 72.9 mIU/ml, respectively. E2, T, and DHEA were analyzed using gas chromatography/mass spectrometry at Endoceutics; these methods have been previously described (Supplemental Appendix, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org) (20, 21). The limits of detection were 3.0 pg/ml for total E2; 8.0 ng/dl for total T, and 0.30 ng/ml for DHEA. For values below the detection limits, values were extrapolated below the lower limit of quantitation using Mass Hunter Workstation software (Agilent, Santa Clara, CA) (Supplemental Appendix) (20). Interassay coefficients of variation for E2 were 17.5% at 4.7 pg/ml; T, 13.0% at 14 ng/dl; and DHEA, 24.0% at 0.77 ng/ml. Bioavailable T (bioT) and E2 (bioE2) were calculated according to the method described by Södergård et al. (22) (courtesy of Frank Stanczyk, University of Southern California, Los Angeles, CA) taking the concentrations of total T, total E2, and SHBG into account and assuming a fixed albumin concentration of 4.0 g/dl.

In the DPP, participants were asked what ethnic or racial group they considered themselves as a member, and participants were asked to select only a single option. For this analysis, women who self-reported as Asian were excluded due to small numbers (n = 24 estrogen and nonestrogen users), resulting in a sample size of 370 women who did not use exogenous estrogen or testosterone at baseline and 1-yr follow-up and 310 women who used oral estrogen therapy both at baseline and at follow-up.

Statistical analysis

Baseline characteristics were described using percentages for categorical variables and means (sd) for quantitative variables, stratified by race/ethnicity (Table 1). Due to skewed distributions, unadjusted comparisons of sex hormones used nonparametric tests. Linear regression models were used to examine the association between race/ethnicity and log-transformed baseline sex hormone measures. To determine whether associations were confounded by age, type of menopause, baseline waist circumference, alcohol use (grams per day), or current smoking (yes/no), these variables were added sequentially to the models (Table 2). We also examined whether there were associations between race/ethnicity and change in sex hormone levels. For ease of interpretability, we first examined change in sex hormone level (year 1 log hormone level minus baseline log hormone level) as the dependent variable and adjusted for randomization arm and changes in waist circumference and other covariates (Table 3). We also conducted analyses using follow-up sex hormone value as the dependent variable and adjusting for baseline values (Supplemental Appendix Table 2). For Tables 2 and 3, the β-coefficients were standardized by sd so that the strength of association between race/ethnicity with hormone level could be compared between different hormones. Correlations between baseline waist circumference and change in waist circumference were 0.16 and 0.22 for nonestrogen users and estrogen users respectively, so both baseline and change measures were included in the models (Table 3). The SAS analysis system was used for all analyses (SAS Institute, Cary, NC).

Table 1.

Baseline and baseline to year 1 change characteristics by NHW, AA, and Latina race/ethnicity and by estrogen hormone therapy (ET) use

	NHW	AA	Latina	P value
No ET (n = 370)	n = 201	n = 108	n = 61
ET users (n = 310)	n = 215	n = 53	n = 42
Age (yr), no ET	60.0 (9.6)	58.3 (8.1)	56.9 (7.7)	0.04
Age (yr), ET	57.0 (7.9)	55.0 (7.0)	55.5 (7.2)	0.16
Menopause type (%), no ET				<0.01
Bilateral oophorectomy	23	15	19
Natural menopause	68	61	70
Age ≥55 yr and hysterectomy	9	24	11
Menopause type (%), ET				0.06
Bilateral oophorectomy	40	53	33
Natural menopause	42	20	44
Age ≥55 yr and hysterectomy	18	27	22
Hysterectomy (%), no ET	36	42	36	0.58
Hysterectomy (%), ET	63	79	62	0.08
Years since final menstrual period, no ET	14.9 (9.4)	14.7 (10.3)	13.3 (10.6)	0.54
Years since final menstrual period, ET	14.1 (9.5)	15.7 (9.2)	12.3 (8.8)	0.23
Current smoking (%), no ET	6	12	2	0.03
Current smoking (%), ET	5	15	10	0.06
Current alcohol use (%), no ET				<0.01
0 g/d	53	71	76
>0 but <2 g/d	28	13	18
≥2 g/d	19	15	7
Current alcohol use (%), ET				0.12
0 g/d	46	62	56
>0 but <2 g/d	25	25	24
≥2 g/d	29	13	20
Baseline weight (kg), no ET	91.4 (18.9)	94.8 (19.5)	82.1 (18.5)	<0.01
Baseline weight (kg), ET	88.7 (18.2)	90.7 (17.1)	81.2 (16.3)	0.02
Baseline waist circumference (cm), no ET	105 (15)	105 (13)	100 (13)	0.047
Baseline waist circumference (cm), ET	101 (14)	102 (13)	98 (13)	0.28
Baseline BMI (kg/m2), no ET	34.6 (6.4)	35.4 (7.1)	33.6 (6.4)	0.21
Baseline BMI (kg/m2), ET	33.4 (6.7)	34.3 (6.3)	32.7 (6.1)	0.51
FSH (IU/liter), no ET	57.3 (28.2)	54.6 (23.8)	50.7 (23.4)	0.22
FSH (IU/liter), ET	35.2 (22.3)	32.5 (25.2)	37.0 (22.5)	0.64

Means (sd) or percentages shown unless otherwise indicated. Boldface indicates statistically significant F-test at P < 0.05.

Table 2.

The association (standardized β-coefficient) between race/ethnicity and baseline sex hormone value

	Among women not using estrogen therapy		Among women using estrogen therapy
	Hispanics vs. NHW	AA vs. NHW	Hispanics vs. NHW	AA vs. NHW
Unadjusted
Log SHBG	0.02	0.09	−0.11	−0.26a
Log total E2	−0.36a	−0.15	−0.21	−0.14
Log bioE2	−0.37a	−0.17	−0.20	−0.04
Log total T	−0.38a	0.03	0.15	−0.02
Log bioT	−0.38a	−0.02	0.20	0.12
Log DHEA	0.10	−0.12	0.25b	0
Adjusted for age
Log SHBG	0.08	0.11b	−0.06	−0.21b
Log total E2	−0.48a	−0.24a	−0.25	−0.18
Log bioE2	−0.52a	−0.28a	−0.28b	−0.09
Log total T	−0.40a	0.01	0.14	−0.03
Log bioT	−0.44a	−0.05	0.16	0.09
Log DHEA	0.03	−0.16b	0.20	−0.05
Adjusted for age and type of menopause
Log SHBG	0.11	0.15a	−0.06	−0.29a
Log total E2	−0.46a	−.13	−0.26	−0.23
Log bioE2	−0.50a	−0.21b	−0.31b	−0.12
Log total T	−0.36b	−0.26a	0.18	0.01
Log bioT	−0.41a	0.06	0.18	0.17
Log DHEA	0.01	−0.03	0.18	0.03
Adjusted for above, and baseline waist circumference
Log SHBG	0.05	0.13b	−0.09	−0.29a
Log total E2	−0.38a	−0.18	−0.26	−0.23
Log bioE2	−0.41a	−0.22b	−0.29b	−0.11
Log total T	−0.37b	0.06	0.19	0.01
Log bioT	−0.39b	−0.02	0.22	0.19
Log DHEA	−0.03	−0.14	0.17	0.03
Adjusted for above, and alcohol use and current smoking
Log SHBG	0.07	0.13b	−0.10	−0.26b
Log total E2	−0.45a	−0.18	−0.30b	−0.22
Log bioE2	−0.49a	−0.22b	−0.33b	−0.12
Log total T	−0.39b	0.10	0.21	0.06
Log bioT	−0.43a	0.02	0.24	0.21
Log DHEA	−0.06	−0.13	0.18	0.04

Reference group is NHW. A β-coefficient less than 0 indicates that sex hormone levels in that racial/ethnic group are lower compared with NHW.

^aP ≤ 0.01.

^bP ≤ 0.05.

Table 3.

The association (standardized β-coefficient) between race/ethnicity and changes in sex hormone value between baseline and year 1 follow-up

	Among women not using estrogen therapy		Among women using estrogen therapy
	Hispanics vs. NHW	AA vs. NHW	Hispanics vs. NHW	AA vs. NHW
Unadjusted
Log SHBG	0.08	−0.04	−0.01	0
Log total E2	0.04	0.13	−0.21	0.01
Log bioE2	0.02	0.13	−0.15	0.01
Log total T	0.17	0.07	−0.07	0.06
Log bioT	0.14	0.08	−0.05	0.06
Log DHEA	0.06	0.08	−0.06	0.04
Adjusted for randomization arm
Log SHBG	0.07	−0.04	−0.01	0.01
Log total E2	0.05	0.13	−0.20	−0.02
Log bioE2	0.03	0.14	−0.15	−0.02
Log total T	0.16	0.06	−0.07	0.06
Log bioT	0.14	0.07	−0.04	0.05
Log DHEA	0.06	0.09	−0.06	0.06
Adjusted for above, age, and type of menopause
Log SHBG	0.08	−0.04	−0.05	−0.01
Log total E2	0.11	0.20	−0.18	−0.02
Log bioE2	0.09	0.22a	−0.08	0
Log total T	0.21	0.05	−0.16	0.01
Log bioT	0.18	0.08	−0.10	0.02
Log DHEA	0.05	0.08	−0.02	0.02
Adjusted for above and baseline and change in waist circumference
Log SHBG	0.07	−0.05	−0.05	−0.03
Log total E2	0.15	0.22a	−0.16	0
Log bioE2	0.14	0.24a	−0.05	0.03
Log total T	0.23	0.06	−0.18	0.02
Log bioT	0.20	0.09	−0.13	0.03
Log DHEA	0.06	0.08	0	0.02
Adjusted for above and alcohol use and current smoking
Log SHBG	0.07	−0.02	−0.06	−0.05
Log total E2	0.15	0.29b	−0.12	0.04
Log bioE2	0.14	0.29b	0	0.08
Log total T	0.22	−0.02	−0.23	0
Log bioT	0.22	0.07	−0.17	0.03
Log DHEA	0.03	0.08	−0.04	0.01

Reference group is NHW. A β-coefficient greater than 0 indicates that the changes in sex hormone are smaller in that racial/ethnic group compared with NHW.

^aP ≤ 0.05.

^bP ≤ 0.01.

Results

Characteristics of the cohort by race/ethnicity and stratified by estrogen use are presented in Table 1. Among women who did not use estrogen, NHW were slightly older than Hispanics and AA, and NHW were more likely to have undergone bilateral oophorectomy. NHW were less likely to smoke than AA but more likely to drink some alcohol. NHW had baseline weights that were higher than Hispanics and lower than AA. Among women using oral estrogen at baseline and 1-yr follow-up, racial/ethnic differences at baseline were less marked (Table 1), although NHW still had baseline weights higher than Hispanics and lower than AA. The majority of oral estrogen use consisted of conjugated equine estrogen and type of estrogen was similar by race/ethnicity (P = 0.82).

Unadjusted baseline sex hormone levels by race/ethnicity are shown in Fig. 1. Table 2 shows the strength of the associations between race/ethnicity and unadjusted log-transformed hormone levels in Hispanics and AA compared with NHW. Among women who did not use estrogen, NHW had higher log total and bioE2 and higher total and bioT than Hispanics (Table 2). NHW had similar sex hormone levels compared with AA. Among estrogen users, NHW had higher SHBG than AA and lower DHEA than Hispanics (Table 2).

Fig. 1.

Unadjusted baseline levels of sex hormone levels by race/ethnicity. A, SHBG (nm/liter); B, DHEA (ng/ml); C, total E2 (pg/ml); D, bioavailable E2 (pg/ml); E, total T (ng/dl); F, bioavailable T (ng/dl). Box plots illustrate medians and interquartile ratios (25th percentile and 75th percentile), and whiskers represent 10th and 90th percentiles. *, Significant differences between groups on nonparametric tests among nonestrogen users; there were no significant differences between racial/ethnic groups among estrogen users.

Table 2 also shows the cross-sectional association between race/ethnicity and baseline log sex hormone levels in Hispanics and AA compared with NHW, after adjustment for covariates. Among women who did not use estrogen, the differences between NHW and Hispanics persisted, that is, NHW had higher total and bioT and bioE2 than Hispanics. NHW had lower SHBG than AA and higher bioE2 than AA; differences in DHEA from Hispanics were no longer significant. Among estrogen users (Table 2), NHW also had higher log total E2 and higher log bioE2 than Hispanics, and NHW still had higher SHBG than AA. Racial/ethnic differences in T or DHEA were not observed after adjustment.

Unadjusted changes in sex hormone levels by race/ethnicity are shown in Fig. 2. Table 3 shows the strength of the associations between race/ethnicity and unadjusted changes in log-transformed hormone levels in Hispanics and AA compared with NHW. There were no significant associations between race/ethnicity and changes in unadjusted hormone levels, regardless of estrogen therapy use.

Fig. 2.

Unadjusted changes in sex hormone levels (year 1 measure to baseline measure) by race/ethnicity. A, Year 1-baseline SHBG (nm/liter); B, year 1-baseline DHEA (ng/ml); C, year 1-baseline total E2 (pg/ml); D, year 1-baseline bioavailable E2 (pg/ml); E, year 1-baseline total T (ng/dl); F, year 1-baseline bioavailable T (ng/dl). Box plots illustrate medians and interquartile ratios (25th percentile and 75th percentile), and whiskers represent 10th and 90th percentiles. There were no significant differences between racial/ethnic groups; for Hispanics not using estrogen, the 10th percentile of estradiol change was −30.9 pg/ml.

Table 3 also shows the association between race/ethnicity and change in log sex hormone levels in Hispanics and AA compared with NHW, after adjustment for covariates. Among women who did not use estrogen therapy, NHW had larger reductions in E2 and bioE2 than AA. Race/ethnicity was not associated with changes in other sex hormones in adjusted analysis. Among estrogen users, there were no significant racial/ethnic associations with sex hormone changes. When we examined unadjusted changes in sex hormones by race/ethnicity and intervention arm, we did not see statistically significant differences by race/ethnicity with the exception of a greater increase in SHBG among Hispanic women randomized to metformin compared with NHW (Supplemental Appendix Table 1). When we examined follow-up sex hormone values as the dependent variable (and including baseline levels as a covariate), we did not see changes in this pattern of results with one exception (Supplemental Appendix Table 2). Among estrogen users, Hispanic women had lower total E2 levels than NHW before and after adjustment for covariates, including baseline sex hormone values.

Discussion

Using data from a large multicenter randomized trial, we examined sex hormone levels and changes in these levels in a population at high risk for morbidity, i.e. postmenopausal women who are overweight and glucose intolerant. To our knowledge, no other studies have examined racial/ethnic differences in sex hormone levels in this population, nor have they examined changes in sex hormone levels and sex hormone levels among exogenous estrogen users. Among women who were not using estrogen therapy, NHW had higher total and bioE2 and bioT levels than Hispanics and lower SHBG and higher bioE2 levels than AA. These racial/ethnic differences were significant, even after adjustment for potential confounders including age, waist circumference, menopause type, and health behaviors. Among women using oral estrogen therapy, NHW still had higher baseline levels of E2 than Hispanics and higher levels of SHBG than AA. However, among women not using estrogen therapy, levels of total E2, bioE2, and bioT between Hispanics and NHW were significant at the P < 0.01 level, whereas among women using estrogen therapy, racial/ethnic differences in total E2 and bioavailable E2 were significant at the P < 0.05 level and were of smaller magnitude. In addition, there were also race/ethnicity differences in sex hormones changes at 1 yr among nonusers, in which NHW had greater reductions in total E2 and bioE2 compared with AA.

Previous reports in perimenopausal and postmenopausal populations have conflicted regarding racial/ethnic differences in sex hormones among women not using oral estrogen (4–7). Our findings that T was higher in NHW than AA and Hispanics are similar to the perimenopausal population of SWAN (5), although not MEC (6) or MESA (7). It is possible that racial/ethnic differences in androgen levels differ in women closer to the menopausal transition because the SWAN women were perimenopausal, and the postmenopausal women in the DPP were younger than postmenopausal women in the MEC (mean age 68 yr) (6) and in MESA (mean age 65 yr) (7). In the WHI (9) and the Rancho Bernardo Study, (23), SHBG and T increased slightly with age and time from menopause. It is also possible that such racial/ethnic differences are accentuated among women who have higher body mass because the DPP population was overweight as enrollment criteria, and postmenopausal DPP participants had average BMI exceeding 30 kg/m². The cohorts previously mentioned had BMI averaging approximately 28 kg/m² (4–7). The sex hormone differences between NHW and Hispanics tended to be more pronounced than between NHW and AA, suggesting that the more pronounced differences in age and weight in NHW vs. Hispanics may have contributed to sex hormone differences. Different types of menopause, particularly the lower prevalence of oophorectomy in AA, might have contributed to racial/ethnic differences. However, the more frequent bilateral oophorectomy in NHW would have been expected to lead to lower androgen levels in NHW because oophorectomy is associated with lower androgen levels than natural menopause (21, 23). It is possible that dietary differences in racial/ethnic groups between studies may have also contributed to differences between our results and those of other studies; in the DPP, NHW, AA, and Hispanic women were similar in dietary intake and changes in dietary intake as would be expected from their phenotype, (24), but this may not have been the case in other cohorts. Finally, although the racial/ethnic differences persisted after adjustment for age and waist circumference, residual confounding is possible. Increased adiposity may decrease hepatic production of SHBG in vitro (25, 26). Adiposity has also been associated with increased serum total E2 and T levels (27), and visceral adipose tissue may produce and metabolize T (28) and also E2 (29).

Adjustment for factors that differed between racial/ethnic groups at baseline tended to have the largest impact on racial/ethnic differences in sex hormones. Among nonestrogen users, adjustments for age and health habits (alcohol use and smoking) increased NHW/Hispanic differences in baseline E2 and T levels, whereas adjustment for waist circumference tended to decrease them, reflecting differences in the racial/ethnic distribution of these factors as well as their variable influence on sex hormone levels. After consideration of the combined influence of all of these factors, NHW and Hispanic E2 and T differences were actually larger than unadjusted differences. Adjustment for baseline factors had a similar but smaller effect on NHW/Hispanic differences among estrogen users, perhaps reflecting the predominating influence of exogenous estrogen. Among NHW and AA who were nonusers, adjustment for age and type of menopause increased NHW/AA differences in E2 and SHBG. Among NHW and AA who were estrogen users, adjustment for type of menopause increased racial/ethnic differences in SHBG but age did not, perhaps reflecting the similar ages of NHW and AA women who used estrogen. As would be expected, when we examined changes in sex hormones rather than baseline differences, adjustment for factors that differed at baseline had minimal impact, and the magnitude of change was similar by racial/ethnic groups. The exception was that NHW had slightly larger changes in E2 levels compared with AA after consideration of the type of menopause.

Apart from our study, no other studies have found that NHW have higher E2 than AA or Hispanics after consideration of body size. Of note, the association between race/ethnicity and E2 levels was generally slightly more robust than the associations between race/ethnicity with androgen levels, suggesting that there may be greater racial/ethnic variation for estrogens compared with androgens, at least among overweight, glucose-intolerant women. The aforementioned differences in cohorts regarding age, weight, and menopause type might partially explain the racial/ethnic differences in E2. Insulin sensitivity may influence sex hormone levels (30), and this may have further accentuated racial/ethnic differences in E2 among DPP women. We also found that, among nonestrogen users, NHW had greater treatment-related declines in E2 than AA even after consideration of declines in weight circumference. We have previously reported that among nonestrogen users, randomization to lifestyle change led to increases in SHBG and DHEA levels compared with placebo, independent of key covariates including age, race/ethnicity, changes in waist circumference, and changes in insulin levels. Differences in changes in E2 and T were not observed, and randomization to metformin was not associated with changes in sex hormone levels (20). Although our results need to be replicated, our findings suggest that in this subset of glucose-intolerant women, weight changes may have weaker associations with E2 changes in AA than in other racial/ethnic groups.

Ours is the first report to examine racial/ethnic differences in sex hormones among oral estrogen users. Among women using oral estrogen therapy both at baseline and at follow-up, we have previously reported that randomization to either lifestyle change or metformin (vs. placebo) led to decreases in bioE2 levels but not changes in other hormone levels including SHBG (32). In the current report, we found that despite the use of exogenous estrogen, which would be expected to diminish racial/ethnic differences, NHW still had significantly higher circulating total and bioE2 levels than Hispanics. In contrast to findings in women who did not use estrogen therapy, NHW had higher SHBG levels on hormone therapy than AA. Although explanations are speculative, it is possible that exogenous estrogen is metabolized differently by race/ethnicity. These findings demonstrate that exogenous estrogen may not completely mask other racial/ethnic determinants of sex hormone levels. Among men, 5α-reductase levels may differ in AA men than NHW men and result in different levels of dihydrotesterone and total T (33), and it is possible that there are racial/ethnic differences in the relative activity of the oxidase vs. reductase of the 3α-hydroxysteroid dehydrogenases, the amount and specificity of steroid coactivator and corepressor activity in tissues, and finally steroid receptor specificity and activity in tissues (34), although racial/ethnic differences have not been reported.

Our findings of racial/ethnic differences in endogenous sex hormones may partially explain racial/ethnic differences in the incidence of several hormone-sensitive comorbidities. Among postmenopausal women, NHW are more likely to develop breast cancer than AA (even though AA have higher breast cancer mortality) (1). Even after declines in the use of estrogen therapy associated with the WHI trial results, NHW still had a higher incidence of invasive and in situ hormone receptor-positive breast cancer than other race/ethnicities (1). In the MEC, NHW had higher risk for endometrial cancer, another estrogen-sensitive cancer, than AA or Hispanics, even after adjustment for age, estrogen use, smoking, and other risk factors (2). Our findings regarding E2 are also consistent with studies that found that higher endogenous E2 levels were positively associated with high-density lipoprotein cholesterol (3), and studies that reported that NHW had higher with high-density lipoprotein cholesterol levels than AA (20) and Hispanics (3, 35). Our findings do not correspond with known differences in bone mineral density (BMD), which tends to be higher in AA (36). However, others have found that E2 and T levels were not associated with BMD or BMD loss after consideration of weight and race/ethnicity (36) and FSH (31), suggesting that other race/ethnicity-specific factors other than E2 may be more powerful determinants of BMD.

Limitations included our inability to examine differences in Asian-Americans in the DPP, the relatively small number of Hispanics, and multiple comparisons. It is possible that our use of proxies for body fat in the DPP may have increased or decreased the racial/ethnic differences in sex hormones compared with direct measures of fat. The DPP was a randomized trial, and thus, the population is not necessarily representative of the general population of women with glucose intolerance. Our results apply to women who are overweight and glucose intolerant and do not necessarily reflect racial/ethnic differences among women who are not of this phenotype. To our knowledge, no population-based studies of women with impaired glucose tolerance exist that examine racial/ethnic differences in sex hormones, and thus, the findings are of value despite the selected nature of the population. Due to small numbers, we were not able to examine race/ethnicity differences in the response of sex hormones stratified by the individual DPP interventions. Strengths of our report include our adjustment for menopause type, measurement of sex hormones using mass spectrometric techniques, and a multicenter sample. We observed differences in sex hormone levels among estrogen users by race/ethnicity, a comparison that has not been previously reported.

In conclusion, we found that NHW had higher endogenous T and E2 compared with Hispanics and higher bioE2 than AA independent of ethnic differences in body size and weight distribution. These results suggest that racial/ethnic differences in sex hormone levels may differ more among glucose-intolerant women than what has previously been found in women without this phenotype. Racial/ethnic differences in sex hormones diminished but persisted despite the use of estrogen therapy. Our report suggests that race/ethnicity might influence sex hormone changes in response to weight-loss interventions as well. The clinical significance of these racial/ethnic differences in sex hormones remains to be established.