Urinary Hydroxyestrogens and Breast Cancer Risk among Postmenopausal Women: A Prospective Study

Exposure to endogenous or exogenous estrogens is seemingly an important risk factor for development of breast cancer (1, 2). Prevailing theories propose that the carcinogenic properties of estrogens involve increased cell proliferation by stimulation of the estrogen receptor (3) and that metabolic activation of estrogens to quinone derivatives (4) causes DNA changes in terms of oxidative damage (5) and DNA adducts followed by depurination (6, 7). Two main pathways are involved in estrogen metabolism in humans (8), which result in either hydroxylation of the A-ring (2- or 4-hydroxy derivatives) or hydroxylation of the D-ring (at 16α-position). Estrogens hydroxylated in 2 position (2-OHE; 2-hydroxyestrone and 2-hydroxyestradiol) bind to the estrogen receptor with reduced affinity compared with the parent agent and make 2-OHE less mitogenic, whereas C-16α hydroxylated estrone (16α-OHE1) is a potent estrogen receptor agonist (9). The C-2 hydroxylated metabolite has no uterotrophic activity in female rats; it suppresses growth and proliferation of MCF-7 cells in culture (10) and it exerts opposite effects compared with estrogen on both hormone-sensitive breast cancer cell growth and cell differentiation (11). However, during further metabolism, 2-OHE may redox cycle, which will form reactive oxygen species capable of inducing oxidative DNA damage (12); furthermore, 2-OHE quinone-derived DNA adducts have been found mutagenic in mammalian cells (13). Conversely, 16α-OHE1 has been reported to be elevated in women with breast cancer (14) and it has previously been positively correlated with murine mammary tumor incidences (15). In addition, 16α-OHE1 has significant uterotrophic activity and low binding affinity for sex hormone binding globulin (16). Although 2-OHE and 16α-OHE1 seem to have different effects, they have often been presented as a ratio in epidemiologic studies with conflicting results (17-24).

Hormone replacement therapy (HRT) is a well-known risk factor for breast cancer among postmenopausal women (25) and it has been shown that estradiol-based HRT use increases the formation and urinary excretion of both 2-OHE and 16α-OHE1 more than ten-fold (26). Thus, when evaluating the risk of developing breast cancer, it may be particularly important to investigate the associations with the hydroxyestrogen metabolites in postmenopausal women who use HRT and have high exogenous but low endogenous estrogen levels when evaluating the risk of developing breast cancer. Moreover, breast cancer can be classified as estrogen receptor–positive or estrogen receptor–negative, and it has been speculated whether the potential effects of the estrogen metabolites vary on the development of estrogen receptor–positive or estrogen receptor–negative breast cancers.

The effect of HRT use on the association between the two hydroxyestrogens and estrogen receptor–specific breast cancer among postmenopausal women has never been evaluated. Therefore, we analyzed the association between estrogen receptor–specific breast cancer and the estrogen metabolites 2-OHE and 16α-OHE1 in a case-control study (426 cases and 426 controls), nested within a cohort comprising 24,697 postmenopausal women, to investigate whether the risk of developing breast cancer differed with estradiol-based HRT use.

From December 1993 to May 1997, 79,729 women of ages 50 to 64 years were invited to participate in the prospective study “Diet, Cancer and Health.” The women were all born in Denmark and lived in the greater Copenhagen or Aarhus areas; none was previously registered with cancer in the Danish Cancer Registry (27). A total of 29,875 women accepted the invitation. All participants visited one of two established study centers where urinary samples were collected. In addition, a food frequency questionnaire and a lifestyle questionnaire were completed by the women. The lifestyle questionnaire included questions about reproductive factors, health status, social factors, and lifestyle habits. From this questionnaire, we obtained information about years of school education (short: ≤7 years, medium: 8 to 10 years, or long: ≥10 years), parity (parous/nulliparous, number of births, and age at first birth), use of HRT (never, past, current), and duration of HRT use. Health professionals obtained anthropometric measurements. Body mass index was calculated as weight (in kilograms) per height (in meters) squared.

“Diet, Cancer and Health” and the substudy reported here were approved by the regional ethical committees on human studies in Copenhagen and Aarhus and by the Danish Data Protection Agency.

Of the 29,875 women enrolled in the study, 326 (∼1%) were later reported to the Danish Cancer Registry with a cancer diagnosed before their baseline visit and were therefore excluded. In addition, eight women were excluded from the study because they did not complete the lifestyle questionnaire. This study was aimed at only postmenopausal women due to the limited number of premenopausal women in this cohort, and 4,844 women were excluded as they did not fulfill this criteria: 4,798 who were considered premenopausal, with at least one menstruation within the 12 months before study entry and no use of HRT; 9 with no lifetime history of menstruation; and 37 who did not answer the questions about current or previous use of HRT. Known postmenopausal women were (a) nonhysterectomized and reported no menstruation during the 12 months before inclusion, (b) reported bilateral oophorectomy, or (c) reported age at last menstruation lower than age at hysterectomy. Probable postmenopausal women (a) reported menstruation during the 12 months before inclusion and current use of HRT (we assumed the bleeding were caused by HRT), (b) reported hysterectomy with a unilateral oophorectomy or an oophorectomy of unknown laterality, or (c) reported last menstruation at the same age as that at hysterectomy. Accordingly, 24,697 postmenopausal women remained in the cohort. Members of the cohort were identified by a personal identification number, which is allocated to every Danish citizen by the Central Population Registry, and after linkage to this registry, information on vital status and emigration was available. Through record linkage to the Danish Cancer Registry via the personal identification number, it was possible to gain information on cancer occurrence among the cohort members. Follow-up for breast cancer was done on each woman from the date of entry (date of visit to the study center) until diagnosis of cancer (all except nonmelanoma skin cancer), date of death, date of emigration, or 31 December 2000. During the follow-up period, 434 women from the cohort were diagnosed with breast cancer; of these, 84 were diagnosed within the first year of follow-up. The median (5-95 percentiles) period from collection of the urinary samples to diagnosis was 2.4 (0.2-4.9) years.

In addition to the Danish Cancer Registry, a registry only on breast cancer has been established in Denmark: the Danish Breast Cancer Cooperative Group. This registry has information on ∼90% of all Danish breast cancer cases and includes information on estrogen receptor status (28). In the present study, 24 (6%) of the breast cancer cases found in the Danish Cancer Registry could not be found in the Danish Breast Cancer Cooperative Group. Therefore, information on estrogen receptor status was not available for these 24 women. Linkage to Danish Breast Cancer Cooperative Group was also done using the personal identification number. Although several medical centers were involved, a standardized immunohistochemical method was used. The cutoff level used to define positive receptor status was ≥10% positive cells.

Women who were using HRT at urinary sampling were defined as HRT users, and those not using HRT at urinary sampling (including former HRT users) were defined as HRT nonusers.

Due to the large number of cohort members, it was not feasible to determine urinary levels of 2-OHE and 16α-OHE1 in all; therefore, we used a nested case-control design. One control was selected for each of the 434 cases. The control was cancer-free at the exact age at diagnosis of the case and was further matched on certainty of postmenopausal status (known/probably menopausal), use of HRT on inclusion into the cohort (current/former/never), and age on entry into the cohort (6-month intervals). Selection of controls was done by incidence density sampling.

Of the 434 pairs (866 women: 434 cases and 434 controls, including two cases), 5 pairs were excluded due to the lack of a urine sample and 3 pairs due to nondetectable levels of 2-OHE and/or 16α-OHE1 in either the case or control. This left 426 pairs for study.

Spot urine samples were collected at entry of the study during the visit to the study center, were frozen at −20°C within 2 hours, and were transferred into liquid nitrogen vapor (max −150°C) by the end of the day. Approximately 1 month before analysis, the samples were transferred to a −80°C freezer.

A commercially available enzyme immunoassay kit (ESTRAMET 2/16 Enzyme Immunoassay Kit, IMMUNACARE Corporation, Bethlehem, PA; ref. 29) was used to analyze the urine samples for 2-OHE and 16α-OHE1. The assays were done in random order and cases and controls were handled in pairs and identically. The laboratory that analyzed the samples was blinded about whether the samples were from cases or controls. It has previously been shown that oral hormone therapy increases both the 2-OHE and the 16α-OHE1 excretion in urine (26). Therefore, HRT status at urinary collection for cases and controls was elucidated before the samples were run to enhance the chances of correct dilution, as the sample values should be within the standard curve. The basic protocol for the ESTRAMET 2/16 Kit was followed. The metabolites, 2-OHE and 16α-OHE1, are mainly found as glucuronides in the urine. These require removal of the sugar moiety before recognition by the monoclonal antibodies of the assay. To remove any precipitate, the urine samples were centrifuged and then incubated with deconjugating enzymes. After neutralization and wash, the enzyme immunoassay plates were incubated for 3 hours with the substrate and read kinetically using an Ascent Multiscanner (Labsystems Oy, Helsinki, Finland). Data were reduced using Ascent Software Version 2.4.1. All samples were run in triplicate and averaged (geometric) to reduce measurement error. 2-OHE and 16α-OHE1 for the same person were always run in the same batch.

The cross-reactivity of the antibody to 2-OHE was as follows: 2-hydroxyestrone, 100%; 2-hydroxyestradiol, 100%; 2-hydroxyestriol, 68%; 4-hydroxyestrone, >2.1%; and for other metabolites, ≤0.2%. For 16α-OHE1, the cross-reactivity of the antibody was 100% for 16α-hydroxyestrone, 3.3% for 5-androsten-3β,16α-diol-17-one, 3.1% for 5α-androstan-3β,16α-diol-17-one, and ≤0.2% for other metabolites. Within assay coefficients of variances were 10.8% for both 2-OHE and 16α-OHE1, whereas the between assay coefficients of variance were 11.8% and 13.2% for the 2-OHE and 16α-OHE1 metabolite, respectively. We tested whether time from sample collection to assay date affected the level of 2-OHE/creatinine or 16α-OHE1/creatinine, and found no associations (P = 0.65 and P = 0.62, respectively).

Due to the sampling design with perfect match on age at cancer diagnosis, we used conditional logistic regression analyses to estimate the breast cancer incidence rate ratios (IRRs; without the rare disease assumption; ref. 30). The estrogen metabolites were log 2 transformed so that the estimated rate ratios corresponded to doubling of the concentrations. This allowed comparison of associations with levels of each metabolite and with the ratio between them.

We evaluated the potentially confounding effects of a set of baseline values of established risk factors for breast cancer: parity (yes/no), number of births (linear), age at first birth (linear), length of school education (short, medium, long), duration of HRT (linear), body mass index (linear), and alcohol intake (linear). The choice to constrain our set of potential confounders to those mentioned was based on a combination of the information obtainable from the questionnaire and a literature-based assessment on which variables were the most important. None of these potential confounders was significantly associated with the breast cancer incidence, and the adjustment for these variables did not affect the associations. Thus, the variables were not confounders in the present study and were therefore not included in the final analyses.

All quantitative variables including the log 2–transformed metabolite concentrations were entered linearly in the model (31). The linearity of the associations was evaluated graphically using linear splines with three boundaries placed at the quartile cut points according to the exposure distribution among cases (32). None of the associations showed signs of inflection or threshold values. For the metabolites, this indicates that the IRR associated with a doubling of the concentration was independent of where on the scale the woman was tested. Present users of HRT were compared with never/past users of HRT with respect to the distributions of ng 2-OHE/mg creatinine, ng 16α-OHE1/mg creatinine, and the 2-/16α-OHE ratio using the Wilcoxon two-sample test.

SAS, release 6.12 (SAS Institute, Inc., Cary, NC) was used for the analyses.

Information about estrogen receptor status of the tumors was obtained for 393 (92%) cases of breast cancer, with 302 of the observed tumors reported to be estrogen receptor–positive and 91 tumors estrogen receptor–negative. Information about estrogen receptor status was not obtained for the remaining 33 cases. Of these, it was not possible to determine estrogen receptor status on 9 in situ tumors, and the remaining 24 could not be found in the Danish Breast Cancer Cooperative Group registry.

Baseline characteristics of the study population are presented in Table 1. Cases had a longer duration of HRT use, a higher alcohol intake, and more often had a long school education than controls, although none of these factors were significantly associated with breast cancer. Age at baseline and present/ever use of HRT were identical among cases and controls due to the matching procedure.

Table 2 shows the 2-OHE and 16α-OHE1 levels in urine stratified according to HRT use among cases and controls. The excretion of 2-OHE and 16α-OHE1 was ∼12- and 9-fold increased in HRT users compared with nonusers, respectively. The 2-/16α-OHE ratio was also significantly increased in HRT users. Cases, in general, had higher urinary 2-OHE/mg creatinine, 16α-OHE1/mg creatinine, and 2-/16α-OHE ratio than controls.

Table 3 shows the IRRs for total breast cancer, estrogen receptor–positive breast cancer, and estrogen receptor–negative breast cancer corresponding to a doubling of the creatinine adjusted 2-OHE concentration (the numerator of the 2-/16α-OHE ratio) and a halving of creatinine adjusted 16α-OHE1 concentration (the denominator of the 2-/16α-OHE ratio) mutually adjusted, together with the unadjusted estimate corresponding to the 2-/16α-OHE ratio. The analyses are stratified according to HRT use at baseline.

A significant adverse association was seen between total breast cancer and the creatinine-adjusted 2-OHE concentration among baseline users of HRT (P = 0.02), whereas no sign of adverse association was seen among nonusers of HRT. No significant associations were found between the creatinine-adjusted 16α-OHE1 concentration ratio and breast cancer risk either among all women or when stratified according to use of HRT. To further illustrate these associations, we present the IRRs for quartiles of 2-OHE and 16α-OHE1 concentrations (Table 4).

The adverse association with the creatinine-adjusted 2-OHE concentration was observed only for estrogen receptor–positive breast cancer. No associations were seen between estrogen receptor–negative breast cancer and the 2-/16α-OHE or the creatinine-adjusted 2-OHE and 16α-OHE1 concentrations. We tested whether the associations between receptor-specific breast cancer and the two metabolites depended on baseline use of HRT and whether the associations differed significantly for receptor-negative and receptor-positive breast cancer; we found no significant differences (all P > 0.26). To evaluate whether the established risk factors for breast cancer (presented in Table 1) confounded the associations, we assessed the associations between the 2-/16α-OHE ratio and total breast cancer and estrogen receptor–positive breast cancer among current HRT users in adjusted models. Due to missing information on the established risk factors, we had to exclude 43 pairs from these analyses. The IRR for total breast cancer among current HRT users in the restricted data set was 1.25 [95% confidence interval (95% CI), 0.99-1.56] per doubling of the ratio in the unadjusted analyses, whereas it was 1.28 (1.00-1.63) when adjusted for baseline values of school education, intake of alcohol, parity (parous/nulliparous, number of births, and age at first birth), duration of HRT use, and body mass index. With estrogen receptor–positive breast cancer as the outcome, the unadjusted and adjusted IRRs (95% CI) were 1.26 (0.96-1.65) and 1.26 (0.95-1.68) per doubling of the 2-/16α-OHE ratio. Thus, adjustment for the established risk factors did not affect the estimates. This indicates that these risk factors did not act as confounders in the present study. To evaluate whether the associations between the metabolites and breast cancer differed for women diagnosed within the first year of urinary sampling, we excluded these women (n = 84) from the analyses. This did not change the results and the women were reentered. Because the association between the 2-/16α-OHE ratio and breast cancer has previously been shown to differ between pre- and postmenopausal women (18, 19), it was also evaluated whether the effect differed for known and probable postmenopausal women. We did not find differences between these two groups (all P < 0.40).

In this prospective study about the associations between urinary estrogen metabolites and breast cancer risk, we found an adverse association between breast cancer risk in postmenopausal women and the 2-OHE metabolites. The association was confined to HRT users and was the strongest on the incidence of estrogen receptor–positive tumors.

The present study is the largest prospective study done to date, and is based on a cohort with almost complete follow-up (99.8%). Due to the nested case-control design within the cohort, selection bias is probably a minor problem. Nevertheless, generalization of the results outside the cohort must be with caution. Further, the limited power in the subgroup analyses should be considered when evaluating the results.

In previous studies on the association between estrogen metabolites and breast cancer, current HRT users were either excluded or HRT use was not stated. Our study seems to be the first to address this issue. HRT use in postmenopausal women is widespread (33) despite growing evidence suggesting an association between HRT use and development of breast cancer (34). Postmenopausal women using estradiol-based HRT had much higher excretion of the urinary metabolites in this and one other study (26). In a recent study including 310 postmenopausal women, the use of non-estradiol-based estrogen replacement therapy/HRT also significantly increased the level of both 2-OHE and 16α-OHE1 (35), whereas this was not confirmed in another small study (36). The elevated level of estrogens may be of importance in breast cancer development, particularly in postmenopausal women, who normally have low endogenous estrogen concentrations. HRT use could enhance the risk of breast cancer through several mechanisms. The mother compound and some main metabolites, particularly 16α-OHE1, could act as promoters by activation of the estrogen receptor. The 2- and 4-hydroxylated metabolites could, during further metabolism, form semiquinones and quinones, which alkylate and may generate DNA damage and promote reactive oxygen species via redox cycling (12, 37-39). Estradiol, estrone, and 16α-OHE1, however, have no such properties (40). Estrogens hydroxylated in C-4 position are proven carcinogens in the Syrian hamster model (41, 42), whereas there is no evidence for a direct carcinogenic effect of 2-OHE in animal models. This is possibly due to (a) a faster oxidation of the former to the quinone form (41, 42), (b) a faster methylation of the 2-OHE compared with the 4-hydroxylated metabolites (43), and (c) a reduced dissociation rate from the estrogen receptor of 4-hydroxylated estrogens compared with both 2-OHE and estradiol (44). 2-Hydroxylations of estrogens are principally produced by CYP1A1/2, whereas CYP1B1 preferentially hydroxylates in 4-position (45). However, the enzymes are not specific and a considerable amount of the other metabolite is produced regardless of the enzyme (46). For example, recent measurements show that using the CYP1A1 enzyme, the ratio between 4- and 2-hydroxylation is 7% and 19% when the substrates were E2 and E1, respectively (45). The present data suggest that the 2-OHE is more important than 16α-OHE1 for the risk of breast cancer, indicating a possible formation of 4-hydroxylated estrogens and the generation of (semi)quinones and reactive oxygen species may be of importance. Previous studies suggest that exposure to estrogens may increase the formation of oxidative DNA damage in several systems both in vitro (40, 47, 48) and in vivo (48, 49). Further, in a mouse-model, catechol estrogens showed far more potent carcinogenic effects than estradiol (50). Intratissue concentration measurements of estrogens, hydroxyestrogens, and methoxy derivates showed significantly higher levels of 2- and 4-hydroxyestradiol in malignant breast tissue compared with normal breast tissue, whereas the 16α-OHE1 metabolite and essentially all other native estrogens and other estrogen metabolites were unchanged (51). When evaluating the tissue ratio according to patient survival, fatal cases had a significantly higher ratio compared with surviving patients (51). This supports the suggestion that 2-OHE is important in the carcinogenic process.

Our data also indicated that higher 2-OHE was associated with higher risks of estrogen receptor–positive breast cancer. A previous small case-control study found a higher 2-/16α-OHE ratio in estrogen receptor–positive cases than in estrogen receptor–negative cases; however, the results were not significant and the metabolites were not tested separately due to the small number of participants (17). The mechanism behind the apparent receptor specificity is unknown, although several findings suggest the involvement of oxidative DNA damage. Oxidative DNA damage was increased in breast cancer tissue compared with normal breast tissue, with strong correlation to estrogen receptor status (52). Furthermore, in vitro catechol estrogens induced oxidative DNA damage with a higher frequency in estrogen receptor–positive breast cancer cell lines compared with cells that did not express estrogen receptor (53). Two main pathways for the involvement of estrogen receptor in estrogen-induced oxidative DNA damage have been suggested: (a) Catechol estrogens may bind to the estrogen receptor, which internalizes the complex and carries it to estrogen sensitive sites in the nucleus where redox cycling and reactive oxygen species generation take place causing immediate DNA damage (53). The estrogen receptor–positive breast cancer cells would therefore have a higher intranuclear concentration of catechol estrogen compared with estrogen receptor–negative cells and, consequently, a higher level of DNA damage and possibly a higher risk of breast cancer. (b) Estrogens are capable of increasing the formation of reactive oxygen species through an estrogen receptor–mediated regulation of the antioxidant genes (5). Both pathways might lead to exaggerated levels of DNA damage in estrogen receptor–positive cells induced by estrogens.

Estrogen metabolites in spot urine samples from postmenopausal women were studied. Previous findings have shown little variation of the urinary excretion of estrogen metabolites over the day (29) and that spot urine is representative of 24-hour urine (54). The intraindividual variation of the urinary excretion of estrogen metabolites over time is limited. A good correlation was found between the enzyme immunoassay used in this study and gas chromatography-mass spectroscopy measurements (55). The calibration curve of the present immunoassay is relatively narrow and requires dilution of the urine samples, particularly from HRT users.

Most previous breast cancer studies evaluated the importance of estrogen metabolism using a ratio between the estrogen metabolites as the exposure variable. This assumes that the effect of a doubling of the 2-OHE numerator corresponds to a halving of the 16α-OHE1 denominator. If this is not the case, the use of the ratio instead of mutually adjusted separate hydroxyestrogens leads to an underestimation of the association with one of the hydroxyestrogens and an overestimation of the other. Although we cannot exclude that the joint effect of 2-OHE and 16α-OHE1 can be expressed by the ratio analysis of associations between breast cancer and the concentration of the mutually adjusted metabolites, it is suggested that 2-OHE may be the more important predictor for the higher risk with higher 2-/16α-OHE ratio among current HRT users.

Some small case-control studies of postmenopausal women have reported a lower 2-/16α-OHE ratio in incident breast cancer cases compared with controls (17, 21, 22, 56), whereas one study comprising 66 cases and 76 controls found no association between the urinary 2-/16α-OHE ratio and breast cancer (20), supported by a recent study measuring the serum ratio (57). Another study suggested that a high urinary 2-/16α-OHE ratio may be associated with high-density Wolfe mammographic parenchymal patterns, a recognized indicator of risk of breast cancer (58). In one of the two previously done small prospective studies, the risk of developing breast cancer was insignificantly lower among postmenopausal women in the highest tertile of the ratio compared with the lowest tertile (18). A similar tendency was found among premenopausal women in the same study. In the more recent of the two prospective studies, an inverted U-shaped association was found between the 2-/16α-OHE ratio and the risk of developing breast cancer among postmenopausal women (19). Both studies had a limited number of postmenopausal participants (42 and 71 cases, respectively), excluded HRT users, and lacked power to establish a significant risk increase or risk reduction. Taken together with the present data, it would seem that the 2-/16α-OHE ratio may not be associated with the risk of breast cancer among postmenopausal women not using HRT.

In conclusion, we found that a high urinary 2-OHE excretion was associated with a higher risk of breast cancer, particularly of the estrogen receptor–positive type among postmenopausal women currently using estradiol-based HRT, whereas no effect was seen in nonusers. The 16α-OHE1 was not significantly associated with cancer risk. The 2-OHE metabolites have little estrogen receptor activation capacity, whereas they can redox cycle, which generates reactive oxygen species and alkylating compounds. This suggests involvement of this mechanism in breast cancer development related to HRT.