The Effects of Aerobic Exercise on Estrogen Metabolism in Healthy Premenopausal Women

This article is featured in Highlights of This Issue, p. 745

It is well-accepted that lifetime estrogen exposure increases the risk for breast cancer as a result of cumulative stimulation of epithelial cell division by estrogen (1). It has also been suggested that some metabolites resulting from the biotransformation and inactivation of estrogen can play a significant role in breast carcinogenesis (2). Specifically, the products resulting from the oxidation of estradiol (E₂) and estrone (E₁) known as hydroxyestrogens have been shown to display varying degrees of carcinogenicity. For example, 2-hydroxyestrone (2-OHE₁) partially antagonizes the growth-stimulatory effect of E₂ in cultured human MCF-7 breast cancer cells (3) whereas 2-hydroxyestradiol (2-OHE₂) has little or no carcinogenic activity in Syrian hamsters (4, 5). In cultured mouse mammary epithelial cells, 16α-hydroxyestrone (16α-OHE₁) increases unscheduled DNA synthesis and promoted anchorage-independent growth (6, 7). The metabolite 4-hydroxyestrone (4-OHE₁) is considered genotoxic due to its redox cycling process, which generates reactive oxygen species (ROS) and highly cytotoxic semiquinone/quinone intermediates that react with DNA (2). The 2-hydroxyestrogens also undergo redox cycling but appear to lack carcinogenic activity due to a more rapid clearance in vivo (8) associated with a faster rate of inactivation through O-methylation (9, 10). Finally, one product resulting from O-methylation, namely 2-methoxyestradiol (2-MeOE₂), has been shown to be a potent inhibitor of cell proliferation and angiogenesis (11, 12).

Despite evidence suggesting the possible importance of other aspects of estrogen metabolism on breast cancer, human studies have largely focused on the ratio of 2-OHE₁ to 16α-OHE₁ (2-OHE₁/16α-OHE₁). Given the different genotoxic capacity of these metabolites, it has been hypothesized that metabolism favoring the production of 2-OHE₁ over 16α-OHE₁ may be inversely associated with breast cancer risk (13). In premenopausal women, the strongest evidence in favor of this hypothesis comes from 2 early prospective studies in which urine specimens were collected several years before diagnosis. In the Guernsey III cohort study, women in the highest tertile of urinary 2-OHE₁/16α-OHE₁ ratio had a nonsignificantly lower OR (0.75) for breast cancer than women in the lowest 2 tertiles (14). Similarly, in a study reported by Muti and colleagues, women in the highest quintile of the urinary 2-OHE₁/16α-OHE₁ ratio had an adjusted OR for breast cancer of 0.58, although again this was not statistically significant (15). In contrast, in a more recent prospective study, a higher 2-OHE₁/16α-OHE₁ ratio was associated with an increase in premenopausal estrogen receptor (ER)-positive breast cancer (16). However, the association was not statistically significant and estrogen metabolites were measured in serum and not in urine as in the 2 previous studies. Significant relationships between premenopausal breast cancer risk and urinary levels of estrogen metabolites and their ratios have been observed in some case–control studies, but findings have been inconsistent. The case–control studies of both Coker and colleagues (17) and Kabat and colleagues (18) found an increased risk in women with an increased 2-OHE₁/16α-OHE₁ ratio, but other studies did not (19–21). As for other measures, in 2 other studies, control women had significantly higher levels of 2-hydroxyestrogens, 4-hydroxyestrogens, 16α-OHE₁ (22), and 2-OHE₁/4-OHE₁ ratio (23) than women with breast cancer.

While the association between estrogen metabolism and breast cancer risk needs further investigation, epidemiologic evidence strongly supports the association between higher levels of aerobic exercise and reduced risk for breast cancer (24). However, whether exercise in premenopausal women results in what may be favorable effects on estrogen metabolism is not clear. For example, in one small study, highly fit women exercising strenuously for 368 min/wk had similar values of 2-OHE₁, 16α-OHE₁, and 2-OHE₁/16α-OHE₁ ratio than those of women exercising recreationally for only 60 min/wk (25). In contrast, in another small study, higher levels of self-reported physical activity were associated with higher urinary concentrations of 2-OHE₁ and a higher 2-OHE₁/16α-OHE₁ ratio (26). Recently, a large study of 603 women from the Nurses' Health Study II found high levels of physical activity not only to be correlated with a higher 2-OHE₁/16α-OHE₁ ratio but also significantly lower levels of E₂ and 16α-OHE₁ (27). In comparison, data from exercise intervention studies have been conflicting. For instance, in interventions lasting 12 weeks (28), 16 weeks (29), or even 6 months (30), moderate-to-vigorous intensity aerobic exercise in premenopausal women did not result in any significant changes in urinary concentrations of E₁, E₂, estriol (E₃), 2-hydroxyestrogens, 4-hydroxyestrogens, 16α-OHE₁, or either 2-OHE₁/16α-OHE₁ or 2-OHE₁/4-OHE₁ ratios. In 2 small exercise interventions coupled with calorie restriction lasting 4 and 6 months, there were significant increases in urinary levels of luteal phase 16α-OHE₁ and 2-OHE₁/16α-OHE₁, respectively (31, 32).

Overall, the data on the effects of aerobic exercise on premenopausal estrogen metabolism are not only conflicting but also narrow in scope. With the exception of 2 studies (27, 29), all published studies to date have focused on a limited number of estrogen metabolites, namely, 2-OHE₁ and 16α-OHE₁, and their ratio. Furthermore, no study has yet investigated the levels of the 2- and 4-methylated catecholestrogens despite their purported role in breast carcinogenesis as suggested by culture and animal studies. The WISER (Women In Steady Exercise Research) study was a large, randomized, exercise-controlled, parallel-arm, clinical study investigating the effects of 16 weeks of moderate-to-vigorous intensity aerobic exercise on several parameters pertinent to breast cancer risk in sedentary, healthy, young eumenorrheic women. Here, we report changes from baseline in urinary levels of estrogens (E₁, E₂, and E₃), 9 estrogen metabolites [2-OHE₁, 2-OHE₂, 16α-OHE₁, 4-OHE₁, 4-hydroxyestradiol [4-OHE₂,], 2-methoxyestrone, [2-MeOE₁], 2-MeOE₂, 4-methoxyestrone [4-MeOE₁], and 4-methoxyestradiol [4-MeOE₂]], and 2 estrogen metabolite ratios (2-OHE₁/16α-OHE₁ and 2-OHE₁/4-OHE₁).

The WISER study was a randomized clinical trial investigating the effects of a 16-week aerobic exercise intervention on breast cancer biomarkers of healthy, premenopausal women. All procedures were approved by the Human Subjects Review Committee at the University of Minnesota (Minneapolis, MN; Institutional Review Board; IRB ID#0505M69867). Written informed consent was obtained from each participant before participation. A complete description of the study design including participant recruitment, screening, randomization, and retention has been published (33).

Briefly, WISER study investigators emailed more than 100,000 female residents of the Minneapolis–St.Paul metropolitan area regarding participation. Women who were interested were screened online based on age (18–30 years old), physical activity (2 or less weekly sessions of moderate intensity exercise), smoking status (nonsmoking), body mass index (BMI; 18–40 kg/m² inclusive), and self-reported menstrual cycle length (24–35 days). Women who met these criteria were further screened via telephone (n = 1684) and excluded on the basis of previous hormonal contraception use (past 3 months or 12 months if depot-medroxyprogesterone acetate), gynecologic problems, metabolic or endocrine-related diseases, current or recent (past 6 months) pregnancy, nonmelanoma cancer in the past 5 years, alcohol consumption (more than 7 servings/wk), and body weight changes (more than 10% over the past year).

Of the 966 women who attended a 2-hour orientation, 391 provided written consent and were enrolled in the study. After baseline measurements, women were randomized into either an exercise intervention (n = 212) or a no-exercise, usual-lifestyle control group (n = 179) for approximately 16 weeks. Randomization was stratified on baseline BMI tertiles (≤22.8, 22.8–26.3, ≥26.3) based on the 50th and 75th percentiles from NHANES I data and age (18–24 vs. 25–30). Participants who failed to return for follow-up measures were dropped from of the study. In addition, exercisers were subject to study exclusion if they missed 15 or more exercise sessions. Figure 1 shows the recruitment, screening, randomization, retention, and completion of WISER participants.

Women randomized to the exercise intervention trained aerobically 5 times a week for 30 minutes on a treadmill, stair-stepper, or elliptical machine, at a specified intensity based on age-predicted maximal heart rate (max HR) for 16 weeks (±2 weeks). The exercise intensity was initially set at 65% to 70% of the age-predicted max HR and was gradually increased by 5% every 4 weeks until 80%–85% of age-predicted max HR was reached (stage 1 = 65%–70%; stage 2 = 70%–75%; stage 3 = 75%–80%; stage 4 = 80%–85%).

All training sessions took place at the University of Minnesota's Recreation Center. At the first training session, a certified personal trainer provided instruction on the proper use of the exercise machines, heart rate monitor and watch, and completion of an exercise log after each workout. Trainers supervised exercise sessions and reviewed the exercise logs at least once weekly to monitor adherence and safety. When not meeting with a trainer, participants were expected to complete the remaining of the workout sessions unsupervised. Exercise adherence was assessed using the data from the heart rate monitor (Polar Electro Inc.) and exercise logs.

Any physical activity conducted after randomization and outside the prescribed exercise intervention was assessed at the end of the study with a physical activity questionnaire administered by a research staff member. All participants, regardless of randomization outcome, were asked to maintain their baseline body weight. Control participants were asked to not only to maintain their usual level of physical activity but also to not change their eating habits.

Body mass was measured to the nearest 0.1 kg using an electronic scale (Scale Tronix) 4 times throughout the study (baseline, intervention weeks 4 and 8, and follow-up). Height was measured without shoes to the nearest 0.1 cm (Scale Tronix) by a stadiometer at baseline. BMI was calculated by dividing body mass in kg by height in meters squared (kg/m²). Body composition was assessed at baseline and follow-up by dual-energy X-ray absorptiometry (DXA) using a Lunar Prodigy DXA apparatus (Lunar Radiation Corp.).

Aerobic fitness was assessed at baseline and immediately after the intervention with a submaximal treadmill test described previously (33). This workload was then converted to metabolic equivalents (MET) by using a standard conversion formula (34). Self-reported physical activity conducted a year before the study and during the 4-month follow-up period was assessed by a research staff using a modified version of the Modifiable Activity Questionnaire (35). This information was transformed into MET-hours per week (MET-h/wk) using commonly accepted MET values (36).

Usual dietary intake was assessed through self-reported, 3-day food records completed concomitantly with the urine collections at baseline and follow-up. Nutrient intake was determined using The Food Processor SQL by ESHA Research.

Forty-eight hours before the urine collection, participants were asked to avoid moderate or vigorous exercise and abstain from alcohol. Urine was collected for 3 consecutive 24-hour periods in the midfollicular phase (follicular days 7–9 of baseline menstrual cycle 2 and follow-up menstrual cycle 6). Throughout each day, urine was collected in a 1-L bottle and kept cold with ice packs inside an insulated bag. At the end of each collection day, urine was transferred into a 3-L bottle containing ascorbic acid (1 mg/mL) to prevent oxidation and stored in a home refrigerator or cooler provided by the study. Once the urine collection was completed, collection bottles were retrieved by a research staff member and brought to the General Clinical Research Center at the University of Minnesota for processing. Urine was refrigerated and 0.1% sodium azide was added before the three 24-hour collections were pooled. Aliquots were taken and stored at −20°C until analysis.

Urinary estrogens (E₁, E₂, and E₃) and their metabolites (2-OHE₁, 2-OHE₂, 16α-OHE₁, 4-OHE₁, 4-OHE₂, 2-MeOE₁, 2-MeOE₂, 4-MeOE₁, and 4-MeOE₂) were analyzed in the midfollicular phase of baseline and follow-up cycles by liquid chromatography/tandem mass spectrometry (LC/MS-MS) conducted using a Thermo Electron Quantum Discovery Max Triple Quadrupole LC-MS/MS instrument (37). Quantitative analysis was conducted using Thermo Electron Xcalibur proprietary software.

Samples with nondetectable levels were assigned values of the lowest detectable standard (0.014 ng/mL urine). Concentrations were expressed both as nanomol per day (nmol/d) and nanograms per milligram of creatinine (ng/mg Cr). Urinary creatinine was analyzed at the Fairview University Diagnostic Laboratories.

Samples were run in duplicate and in batches such that each batch contained both baseline and follow-up samples from each participant and an equal number of exercise and control participants. One quality control sample was included in each batch. The mean intra- and interassay coefficient of variations (CV) were 5.1% and 13.4% for E₁; 5.2% and 16.0% for E₂; 5.6% and 11.4% for E₃; 4.2% and 12.3% for 2-OHE₁; 7.7% and 10.8% for 2-OHE₂; 6.2% and 18.7% for 16α-OHE₁; 4.3% and 12.2% for 4-OHE₁; 14.0% and 51.2% for 4-OHE₂; 7.0% and 11.3% for 2-MeOE₁; 5.8% and 10.0% for 2-MeOE₂; 7.4% and 10.3% for 4-MeOE₁; and 6.9% and 7.9% for 4-MeOE₂.

Unadjusted comparisons of baseline characteristics were conducted using Student t tests for continuous variables and χ² tests for categorical variables.

Two estrogen metabolite ratios of interest, namely, the 2-OHE₁/16α-OHE₁ ratio and 2-OHE₁/4-OHE₁ ratio, were calculated by dividing the concentration of 2-OHE₁ by either 16α-OHE₁ or 4-OHE₁, respectively.

Baseline associations between urinary estrogens, their metabolites, and metabolite ratios and measures of body composition, adiposity, fitness, reproductive characteristics, and diet were determined using Spearman correlation coefficients.

The main study analysis assessed the intervention effects using only data from participants who completed the baseline and follow-up urine collection regardless of compliance level. Baseline and follow-up comparisons were conducted using log-transformed values and results are presented as geometric means with 95% confidence intervals (CI). Changes from baseline comparisons were compared on the original scale. All comparisons were adjusted for study design age and BMI strata with a general linear model. When there were significant differences at baseline in an outcome, follow-up and change from baseline comparisons were additionally adjusted for baseline values. Linear models were calculated using SAS software, version 9.2 (SAS institute Inc.). P < 0.05 was considered statistically significant.

Estrogen metabolite data were analyzed as both nmol/d and ng/mg Cr. Results from the 2 analyses did not differ significantly, and we report results as nmol/d.

Of the 212 and 179 women randomized into the exercise and control groups, 165 (77.8%) and 153 (85.5%), respectively, completed the WISER study. With the exception of education (47% of dropouts had some college education vs. 27% of study completers, P = 0.002), dropouts were no different from women who completed the study in any of the baseline demographic characteristic measured (data not shown). Also, there were no significant differences between exercisers and controls in baseline demographic characteristics (Table 1). In general, women who completed the study were mostly Caucasian (72%), single (82%), nulliparous (93%), had education beyond high school (96%), and had no first-degree relatives with breast cancer (97%).

With the exception of 2-OHE₁ (P = 0.084) and 2-OHE₁/16α-OHE₁ ratio (P = 0.044), exercisers had similar levels of urinary estrogens, estrogen metabolites, and 2-OHE₁/4-OHE₁ ratio than control participants at baseline (Table 2). No significant baseline associations between any of the urinary endpoints and measures of body composition, adiposity, fitness, reproductive characteristics, and diet were found.

Overall, the concentration of estrone and its metabolites were higher than their estradiol counterparts, especially for E₁, 2-OHE₁, 4-OHE₁, and 4-MeOE₁ as compared with E₂, 2-OHE₂, 4-OHE₂, and 4-MeOE₂, respectively. Estrogen hydroxylation showed an isomeric preference for the C-2 position. Specifically, concentrations of 2-OHE₁ were about 14- and 20-fold higher than those of 16α-OHE₁ and 4-OHE₁, respectively, and concentration of 2-OHE₂ about 40-fold those of 4-OHE₂.

On average, exercise participants completed 127 min/wk of the assigned 150 minutes of exercise intervention. Details about exercise adherence and compliance can be found elsewhere (38).

The exercise intervention resulted in significant improvements in body composition and aerobic fitness without changes in body weight. As previously reported, exercisers experienced significant increases in aerobic fitness (0.90 METs reached at 85% of max HR vs. 0.12 METs in controls) and lean body mass (0.55 vs. 0.07 kg), as well as significant decreases in fat mass (0.57 vs. 0.04 kg) and percent body fat (0.95% vs. 0.09%). In contrast, control participants experienced no changes in body composition, aerobic fitness, and body weight despite a significant reduction in daily caloric intake (−224 kcal/d). Exercisers also reduced their food consumption, but only by 18 kcal/d (P > 0.05). Details on the effects of this intervention on body composition, body weight, aerobic fitness, and energy intake have been published previously (39).

As previously reported, the exercise intervention resulted in no significant changes in endogenous levels of E₂, estrone sulfate, progesterone, T, or SHBG (38). Exercisers, however, did experience a significant increase in urinary 2-OHE₁/16α-OHE₁ ratio (P = 0.043), whereas controls had a nonsignificant decrease in 2-OHE₁/16α-OHE₁ ratio. The difference in the change from baseline in 2-OHE₁/16α-OHE₁ ratio between groups was significant (P = 0.045), even after adjustment for baseline values. Figure 2 shows that many, but not all, of the participants who experienced an absolute change in 2-OHE₁/16α-OHE₁ ratio greater than 100 had a high baseline 2-OHE₁/16α-OHE₁ ratio. Levels of E₁ remained unchanged in exercisers but decreased in controls resulting in a statistically significant change from baseline between the groups (P = 0.042). No significant within-group changes or between-group differences at follow-up were observed for other estrogens, estrogen metabolites, or ratios.

We found that in healthy premenopausal women, an exercise regimen of 150 minutes of moderate-to-vigorous aerobic exercise per week for 16 weeks resulted in significant changes in estrogen metabolism in a direction consistent with reduction of breast cancer risk. Specifically, exercise participants experienced a significant increase in urinary levels of 2-OHE₁ and a small nonsignificant decrease in 16α-OHE₁ levels. These changes resulted in a significant increase in the 2-OHE₁/16α-OHE₁ ratio. In contrast, women in the control group had a nonsignificant decrease in 2-OHE₁/16α-OHE₁ ratio largely attributable to those few controls with large baseline ratio having large decreases in the ratio. Controls also had an unexplained significant decrease in E₁. We did not find evidence for exercise resulting in changes in the 4-hydroxylation pathway or other differences that conceivably could have been found.

Overall, our results differ from those of other exercise intervention studies investigating the effects of aerobic exercise on premenopausal estrogen metabolism. For instance, in a small 5-month weight loss clinical trial involving moderate-intensity exercise, both controls and exercisers had significant increases in the urinary 2-OHE₁/16α-OHE₁ ratio, but in contrast to our results, the change in ratio between the groups was not statistically significant (32). In a small pre-post design study, 4 months of moderate exercise coupled with calorie restriction resulted in nonsignificantly higher urinary levels of 2-OHE₁ and 16α-OHE₁ and the ratio (31). In our study, both 2-OHE₁ and the 2-OHE₁/16α-OHE₁ ratio increased significantly whereas 16α-OHE₁ decreased nonsignificantly. Both of these studies differed from our study in that aerobic exercise was coupled with significant calorie restriction making it impossible to discern whether the changes reported were the result of the exercise or diet. When compared with exercise-only interventions, our study remains the only one to report significant changes in estrogen metabolism. For example, in a moderate-to-vigorous aerobic exercise intervention lasting 12 weeks (30–45 minutes, 4 d/wk), Campbell and colleagues reported no significant changes in urinary premenopausal levels of 2-OHE₁, 16α-OHE₁, or 2-OHE₁/16α-OHE₁ ratio (28). Similarly, in a pilot study conducted by our research group, total levels of 2-OHE (2-OHE₁ + 2-OHE₂), 4-OHE (4-OHE₁ + 4-OHE₂), and the 2-OHE₁/16α-OHE₁ ratio remained unchanged in 15 young women exercising aerobically for 30 min/d, 5 times a week for 16 weeks (29). Similarly, Robles-Gil and colleagues did not find significant changes in E₁, E₂, or E₃ in 20 premenopausal women after 6 months of 60 minutes of moderate-intensity exercise 3 ds/wk (30).

A possible explanation for the disparity between the results reported by these exercise intervention studies, and our study may be found in the choice of study design and methodology. The WISER study had many methodologic advantages over previously published research. First, the sample size in our study (n = 318) was an order of magnitude or more large than those of the other three studies. Second, our study design used randomized controls (only Campbell and colleagues study was randomized). Third, unlike the studies of Campbell and colleagues and Robles-Gil and colleagues, in which first morning urine samples were used, WISER participants collected three 24-hour urine collections allowing for a more robust and representative analysis of the chronic effect of aerobic exercise on estrogen metabolism. Finally, our study provides the most comprehensive analysis on the effects of exercise on estrogen metabolism to date. We have analyzed urinary levels of the major parent estrogens (E₁, E₂, and E₃) and 9 of their estrogen metabolites by LC/MS-MS. This newer methodology is considered to be superior not only to the ELISA methods used by these studies but also to the current gold standard gas chromatography/mass spectroscopy (GC/MS) due to its increased sensitivity and sample throughput (37, 40). Unlike Xu and colleagues, we were able to quantify and report 4-OHE₂, concentrations, although its lack of detection in 53.4% of our samples resulted in a higher-than-expected interassay CV.

Altogether, the findings of the WISER study are significant because they provide the first clinical evidence that aerobic exercise can significantly change estrogen metabolism in premenopausal women. Specifically, our results show that such an exercise intervention can lead to increases in 2-OHE₁ and possible decreases in 16α-OHE₁ ultimately resulting in significant increases in the 2-OHE₁/16α-OHE₁ ratio. Importantly, increases in this ratio have been associated with a significant reduction in breast cancer risk. From a clinical point of view, the assessment of urinary 2-OHE₁/16α-OHE₁ ratio is also relevant as it has been found to be a good approximation to the 2-OHE₁/16α-OHE₁ ratio of breast tissue (41). Perhaps one mechanism by which exercise mediates estrogen metabolism is through the regulation of P450 cytochrome enzymes responsible for controlling estrogen hydroxylation and catecholestrogen methylation. Given the implication these results have for breast cancer prevention efforts, future studies should not only attempt to corroborate our results but also investigate the exact mechanisms by which exercise leads to these favorable estrogen metabolism changes.

No potential conflicts of interest were disclosed.

Conception and design: A.J. Smith, W.R. Phipps, K.H. Schmitz, M.S. Kurzer

Development of methodology: A.J. Smith, M.S. Kurzer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.J. Smith, M.S. Kurzer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.J. Smith, W.R. Phipps, W. Thomas, M.S. Kurzer

Writing, review, and/or revision of the manuscript: A.J. Smith, W.R. Phipps, W. Thomas, K.H. Schmitz, M.S. Kurzer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.J. Smith, M.S. Kurzer

Study supervision: A.J. Smith, M.S. Kurzer

The authors thank the General Clinical Research Center at the University of Minnesota, the Minneapolis YWCAs, and the study participants. The authors also thank the WISER administrative and laboratory staff, especially Mike Wachter and Steve McColley.

The project was funded by the NIH/National Cancer Institute grant 1U54CA116849-010003, the Department of Defense/U.S. Army Medical Research and Materiel Command Congressionally Directed Medical Research Programs award #W81XWH-08-1-0301, and the NIH/National Center for Research Resources grant M01-RR00400.

Clinical Trial Registration: clinicaltrials.gov identifier: NCT00393172.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.