Racial/Ethnic Differences in Postmenopausal Endogenous Hormones: The Multiethnic Cohort Study

Breast cancer incidence rates vary considerably across racial/ethnic groups (1). The incidence in Japanese women in Japan and in the United States has increased steadily over the years (2-6). In the Multiethnic Cohort Study (MEC), Japanese-American women have slightly higher postmenopausal breast cancer rates than White women, whereas rates among African-American and Latino women are lower (7). Native Hawaiians have the highest rate of breast cancer of any racial/ethnic group in our cohort.

A large and compelling body of epidemiologic and experimental data implicates endogenous estrogens in the etiology of breast cancer (8). Results of a pooled analysis of nine prospective studies provide evidence for an important role of estrogens and their androgen precursors in the development of breast cancer in postmenopausal women (9). The relative risks of breast cancer for women whose estrogen and androgen levels were in the top quintile compared with those whose levels were in the bottom quintile were ∼2. The results also showed that women with high circulating sex hormone–binding globulin (SHBG) levels, a protein that binds to and restricts the biological activity of estradiol and testosterone, had lower breast cancer risk (9).

Racial/ethnic differences in endogenous sex hormone levels might explain some of the variation in breast cancer incidence across racial/ethnic groups. There are only a small number of studies on differences in endogenous hormone levels between racial/ethnic groups in the United States (10-15), and comparisons with Native Hawaiians have not been reported. In the present study, we examined racial/ethnic differences in postmenopausal sex steroid hormone concentrations in a cross-sectional study of 739 postmenopausal women from the MEC (240 African Americans, 81 Native Hawaiians, 96 Japanese Americans, 231 Latinas, and 91 Whites).

Women in this study are participants in the MEC, a large prospective cohort in Hawaii and California (mainly Los Angeles). The details of the study design and baseline characteristics have been described previously (16). Briefly, the recruitment of the cohort began in 1993 and was completed in 1996. Potential participants were identified through driver's license files from the Department of Motor Vehicles, voter registration lists, and Health Care Financing Administration data files. The cohort consists of >215,000 men and women (ages 45-75 years at baseline) and comprises mainly five self-reported racial and ethnic populations: African Americans, Japanese Americans, Latinos, Native Hawaiians, and Whites living in Hawaii and California. Persons of mixed ethnicity were assigned to one of the above racial/ethnic groups according the following priority ranking: African American, Native Hawaiian, Latino, Japanese American, and White; this follows the Surveillance, Epidemiology and End Results guidelines of giving preference to non-White races for individuals of mixed ethnicity. In our cohort, >95% of every ethnic group except Hawaiians reported one race only. Each participant completed a self-administered mail questionnaire that included diet, demographic factors, anthropometric measures, other lifestyle factors, history of prior medical conditions, family history of cancer, and for women, menstrual, reproductive history, and exogenous hormone use (oral contraceptives and postmenopausal hormones).

Incident breast cancer cases were identified by record linkage to the Hawaii Tumor Registry, the Cancer Surveillance Program for Los Angeles County, and the California State Cancer Registry. All tumor registries participate in the National Cancer Institute's Surveillance, Epidemiology and End Results program of cancer registration. Deaths within the cohort were determined by annual linkage to state death certificate files in California and Hawaii and periodically to the National Death Index. Case ascertainment and death information were complete through December 31, 2002.

Beginning in 1994, blood samples were collected from a random sample of MEC participants to serve as a control pool for genetic association studies. Blood samples were collected at the participants' homes after an overnight fast, processed within 8 hours, and stored at −80°C.

Subjects of this study were drawn from the control group of the MEC nested case-control study of breast cancer. Details of this case-control study have been previously published (17). Women were selected for hormone measurements if they met all of the following criteria: ≥56 years old at the time of blood draw, reported no history of breast, endometrial or ovarian cancer on the baseline questionnaire, body weight and body mass index (BMI) information available, and not using postmenopausal hormones at baseline and at the time of blood draw. Seven hundred thirty-nine women were included in this study; of these, 240 were African Americans, 81 were Native Hawaiians, 96 were Japanese Americans, 231 were Latinas, and 91 were Whites. The mean age at blood draw was 66.9 years, ranging from 56 to 82 years. All study participants have provided informed consent, and the Institutional Review Boards at the University of Hawaii and at the University of Southern California approved the protocol.

Plasma hormone assays were done at the Reproductive Endocrine Research Laboratory at the University of Southern California, directed by one of the authors (F.Z.S.). All samples across ethnic groups were assayed together and samples from each ethnic group were included within each batch. Samples were also blinded so that laboratory personnel could not identify which ethnic group the samples were from. Plasma concentrations of androstenedione, testosterone, estrone (E₁), and estradiol (E₂) were measured by sensitive and specific immunoassays after organic solvent extraction and Celite column partition chromatography. SHBG were quantified by a solid-phase, two-site chemiluminescent immunoassay, using the Immulite analyzer (Diagnostic Products Corp., Inglewood, CA). Bioavailable (non-SHBG bound) testosterone and E₂ concentrations were calculated using a validated algorithm on the basis of total testosterone, total E₂, and SHBG measurements (18-20). Replicate-blinded quality control samples (5%) were included to check reproducibility of the hormone assays; the intraclass correlation coefficients between duplicates for androstenedione, testosterone, E₁, E₂, and SHBG were 0.89, 0.98, 0.87, 0.82, and 0.98, respectively. In addition, quality control samples with low, medium, and high concentrations (one pair per level) were included in each batch. The interassay coefficients of variation for sex steroid hormones were ≤14%, ≤12%, and ≤13% at low, medium, and high levels, respectively. The interassay coefficients of variation for SHBG were all ≤5%.

Twenty-two postmenopausal women with either E₁ values >125 pg/mL, E₂ values >75 pg/mL, or testosterone values >125 ng/dL (indicating postmenopausal hormone use) were excluded from the study. Three women (<60 years old at blood draw) who either had unknown menopausal status or were premenopausal at baseline had E₂ values >75 pg/mL and were also excluded from the study, leaving 714 women available for the analyses. Information on breast cancer risk factors and other lifestyle factors that might influence hormone levels (21-36) was obtained from the MEC baseline questionnaire. These factors were body weight, height, age at menarche, age at first birth, parity, age and type of menopause, alcohol drinking, smoking status, vigorous physical activity, calorie intake, percent calories from fat, dietary fiber intake, and soy intake. BMI was calculated as weight (kg)/height (m)². The median interval between baseline questionnaire date and blood draw was 4 years (range, 1-8 years). All hormone values were natural log-transformed to produce approximately normal distributions. Geometric mean hormone levels according to race/ethnicity were calculated using multivariate regression analysis while adjusting for age at blood draw and assay batch. Further adjustment was also done for risk/lifestyle factors that were significantly associated with a particular hormone fraction. These risk factors included BMI and soy intake that were modeled as continuous variables, and age at menarche (≤12, 13-14, and ≥15 years), age and type of menopause [natural (<44, 45-49, ≥50 years), bilateral oophorectomy (<44, ≥45 years), hysterectomy (<44, ≥45 years), and unknown], and smoking status (never, past, and current), which were modeled as categorical variables. Analysis of covariance was used to test for differences in mean hormone levels across race/ethnicity, and Ps for heterogeneity of effects across groups were also reported. For racial/ethnic differences, the main comparison group was White women, as this group had been used as the comparison group in a previous study (11). Breast cancer incidence rates in the MEC were calculated among women ages ≥55 years at baseline and included all incident cases of breast cancer through December 31, 2002. Rates were truncated to age 55 to 79 years and age-adjusted to the U.S. 1970 standard population. All Ps are two sided. The SAS statistical package version 9.0 (SAS Institute, Cary, NC) was used for all analyses.

Descriptive characteristics by racial/ethnic groups are presented in Table 1. Japanese-American women were older at blood draw (mean age = 68.8 years), whereas Latinas and Native Hawaiians were slightly younger (65.5 and 65.8 years, respectively). Body size characteristics (weight, height, and BMI) differed significantly across ethnic groups (P < 0.001). African Americans were the heaviest as a group, and despite their greater average height, they had the highest mean BMI (28.8 kg/m²). The mean BMI of Native Hawaiians and Latinas were lower at 28.2 and 28.0 kg/m², respectively, but greater than Whites (26.5 kg/m²). Japanese Americans weighed much less than any other group, and despite the fact that they were shorter on average than any other group, their mean BMI was only 23.2 kg/m². Among parous women, Latinas and Native Hawaiians had the most children, whereas Japanese had the fewest. Latinas and Native Hawaiians reported the highest average calorie intake. Overall, Japanese Americans had the lowest proportion of calories from fat, whereas African Americans had the highest. Whites (14%) were more likely to have late age at first birth than other groups (4-8%). Compared with other ethnic groups, Japanese Americans were more likely to have later age at menopause. African Americans were more likely to have undergone simple hysterectomy (21%) and bilateral oophorectomy (16%) compared with other groups. White women consumed alcohol to a greater extent than the other groups, with 17% consuming at least one drink per day. The corresponding figure for African Americans and Latinas was around 7%; this decreased to 5% for Latinas and to 2% for Japanese Americans. The highest proportion of current smokers was among African Americans (21%), and the lowest was among Japanese Americans (9%). The distribution of descriptive factors in this sample were consistent with that observed in the entire female cohort (7, 16).

Table 2 shows the associations between plasma hormone concentrations and breast cancer risk factors or lifestyle factors that have been found to be associated with hormone levels in previous studies. As expected, we observed a large increase in E₁ and E₂ levels and a large decrease in SHBG levels with increasing BMI (Ps <0.001). We observed an increase in bioavailable testosterone with increasing BMI, but this increase in bioavailable testosterone is likely to be explained by the substantial decrease in SHBG concentrations because no association was observed with total testosterone. Although the number of children and age at first birth was not associated with significant differences in hormone levels, late age at menarche was modestly associated with lower estrogen concentrations (Ps ≤ 0.05). Age and type of menopause was associated with difference in testosterone levels (Ps <0.001). We observed that smoking was associated with differences in androgen levels but not with estrogen or SHBG levels. Increasing soy intake was modestly related with lower testosterone levels (P = 0.04). No statistically significant relationships were observed between hormone levels and vigorous physical activity, alcohol drinking, calorie intake, and proportion calories from fat or fiber intake. Ethnic-specific associations between selected risk/lifestyle factors and hormone levels are provided in Supplementary Table S1.

In Table 3, we provide the updated incidence rates of breast cancer and geometric means hormone levels by racial/ethnic group. As we have shown previously (7), Native Hawaiians had the highest incidence rates of breast cancer followed Japanese Americans, Whites, African Americans, and Latinas. These lower breast cancer rates observed in African Americans and Latinas than in White women in our cohort reflect the pattern previously reported for women in the Surveillance, Epidemiology and End Results program (37). In age-adjusted analyses, hormone levels varied significantly across racial/ethnic groups for all hormones (Ps <0.001) except for androstenedione (P = 0.10) and SHBG (P = 0.07). Further adjustment for risk/lifestyle factors yielded significant racial/ethnic variation in SHBG levels (P = 0.004).

Across racial/ethnic groups, Native Hawaiian women had the highest age-adjusted mean levels of androstenedione, total and bioavailable testosterone, and total and bioavailable E₂ and the lowest age-adjusted mean levels SHBG. Although further adjustment for risk factors somewhat influenced mean hormone levels in Native Hawaiians, it did not explain the observed differences between Native Hawaiian and White women. In the multivariate analyses, compared with White women, Native Hawaiians had 22% higher androstenedione (P = 0.017), 26% higher testosterone (P = 0.013), 33% higher bioavailable testosterone (P = 0.002), 21% higher E₁ (P = 0.009), 26% higher E₂ (P = 0.001), 31% higher bioavailable E₂ (P < 0.001), and 12% lower SHBG (P = 0.07) levels.

African Americans in our study had higher age-adjusted mean levels of E₁ (+24% P = 0.002), E₂ (+25% P = 0.002), and bioavailable E₂ (+26% P = 0.005) compared with Whites; further adjustment for risk factors (BMI and age at menarche) reduced these differences, but they remained statistically significant. In the multivariate models, African-American women had 21% higher E₁ (P = 0.004), 20% higher E₂ (P = 0.007), and 20% higher bioavailable E₂ (P = 0.015) levels compared with White women. Across ethnic groups, African Americans had the highest levels of SHBG despite being the heaviest group. Compared with Native Hawaiians, they had 22% higher SHBG levels (P = 0.001), but compared with Whites, the difference in SHBG levels was not significant (+7%, P = 0.224).

In age-adjusted analyses, Japanese-American women had similar mean hormone levels to that of Whites. However, after adjustment for BMI and age at menarche, Japanese Americans were found to have 15% higher E₂ (P = 0.036) and 18% higher bioavailable E₂ (P = 0.024) levels than Whites.

There were no significant differences in plasma hormone concentrations between Latina and White women.

In our cohort, the prevalence of bilateral oophorectomy varied across ethnic groups, and this might influence ethnic differences in testosterone levels; we, thus, repeated our analyses while excluding women who had reported bilateral oophorectomy on the questionnaire (n = 72). We observed similar results before and after the exclusion of women with a bilateral oophorectomy. We also repeated our analyses while restricting the analyses to women with covariate data collected within 4 or 5 years of blood draw, and our results did not change.

In our study, 13% of the Japanese-American and 46% of Latina women were born outside the United Stated. Compared with U.S. born Japanese, Japanese women born outside the United States had lower levels of androgens and estrogens and higher levels of SHBG; but these differences were not statistically significant (data not shown). U.S. born and foreign-born Latina women had similar hormone profiles. Adjusting for birthplace in the multivariate models did not alter our results (data not shown).

In this study conducted among postmenopausal women in the MEC, after adjusting for risk/lifestyle factors, we found significant differences in the endogenous hormonal milieu across five racial/ethnic groups. Native Hawaiian women, who have the highest risk of breast cancer in our cohort, had the highest levels of androgens, estrogens and lowest levels of SHBG. We also found African Americans and Japanese Americans to have higher estrogen levels than White women, whereas levels among Latinas were similar to those of Whites.

Compared with Whites, Native Hawaiians were found to have higher circulating levels of androgens and estrogens; differences that were not accounted for by the racial/ethnic differences in the prevalence of breast cancer risk factors and other lifestyle factors. We have previously shown that after adjustment for seven known breast risk factors (age at menarche and first birth, parity, age and type of menopause, weight, postmenopausal hormone use, and alcohol intake), the breast cancer risk for Native Hawaiians was 65% greater than that of Whites (7). The Native Hawaiians' “high-risk” hormonal profile is consistent with their high rates of breast cancer and suggests that the excess breast cancer risk in Native Hawaiians may be due to their having high plasma androgen and estrogen levels and low SHBG levels.

The low postmenopausal estrogen levels of Japanese women living in Japan (12) were not observed in the Japanese-American women in this study, most of whom (87%) were U.S. born. In our study, Japanese Americans had adjusted mean levels of E₂ that were significantly higher than those of Whites. In a previous smaller study conducted among 193 postmenopausal women in the MEC, Japanese-American women (n = 30), despite their low body weight, had 32% higher E₁ and E₂ levels as high as White women (n = 39; ref. 11). Although the magnitude of mean differences in hormone levels between these two ethnic groups was much smaller in the current study, our findings provide support for the previous observation in the initial report. In the MEC, the risk factor adjusted incidence of breast cancer among Japanese Americans was slightly higher than Whites (7). The elevated E₂ levels in Japanese Americans may provide an explanation for their increase in incidence of breast cancer, and that increases in estrogen levels and breast cancer rates may both be determined by long-term exposure to a western diet and other lifestyle factors.

It is well documented that African-American women have higher premenopausal breast cancer rates and lower postmenopausal rates relative to White women (38). Most (10, 13, 15, 39) but not all (40, 41) studies in premenopausal women have reported that African Americans had higher E₂ levels than Whites. In the MEC, the incidence of postmenopausal breast cancer in African Americans was also lower than in Whites; however, their endogenous estrogen levels were found to be significantly higher than those of Whites, confirming findings from an earlier and smaller study in the MEC (11). If having elevated estrogen levels contribute to their higher rates during the premenopausal period, then it is interesting to find out why this elevation does not persist into the postmenopausal period when their estrogen levels are also elevated. Clearly, more work is needed to identify the factors that contribute to these racial/ethnic differences in risk that exist in the premenopausal and postmenopausal periods.

Latinas had sex steroid hormone profiles that were very similar to Whites, in accord with our previous results from a smaller study of 58 Latina postmenopausal women (11). In the MEC, postmenopausal breast cancer risk among U.S. born Latinas was found to be similar to Whites (relative risk, 0.95; 95% confidence interval, 0.75-1.20), whereas migrant Latinas had 16% lower risk than Whites (relative risk, 0.84; 95% confidence interval, 0.64-1.10; ref. 7). We observed that U.S. born and migrant Latinas had similar hormone profiles.

The factors underlying racial/ethnic variation in postmenopausal sex steroid hormone levels are largely unknown; other than body weight or BMI, little is known about their determinants. Variation in key genes involved in estrogen biosynthesis pathway have been associated with differences in circulating hormone levels (42-49); thus, some of the observed ethnic differences in hormone levels may be ascribed to the differences in the distributions of polymorphisms in these genes. Additional research, however, will be needed to test this hypothesis.

Low response rates in certain ethnic groups may affect generalizability of our results to the general population. The response rates in our cohort ranged from 20% in Latinos to 49% in Japanese Americans. As previously shown, however, the distributions of education level in our cohort generally resemble those reported by the U.S. Census in Hawaii and Los Angeles County for the same ethnic and age groups; thus, we believe that findings from this cohort can be compared across ethnic and social strata and are broadly generalizable (16).

In summary, this study confirms the existence of racial/ethnic differences in endogenous sex hormone levels and adds to the sparse data available in the Latino and Native Hawaiian population. Future research should be aimed to elucidate the determinants of Native Hawaiians' high-risk hormonal profiles and increasing E₂ levels in Japanese Americans and to clarify the relationship between sex steroid hormones and breast cancer risk in postmenopausal African-American women.

We thank Xia Zhang for her assistance in performing the hormone assays; Dr. Kristine Monroe, Peggy Wan, and Hank Huang for their assistance in data management and analysis; Dr. Malcolm Pike for his invaluable advice on the analysis and article; and the participants of the Multiethnic Cohort Study for their participation and commitment to the study.