|Home | About | Journals | Submit | Contact Us | Français|
To measure the association between alcohol intake and eleven hormones and peptides in postmenopausal breast cancer survivors and to evaluate whether this association differs by tamoxifen use.
Self-reported alcohol intake was assessed via food frequency questionnaire on average 30 months post breast cancer diagnosis in 490 postmenopausal women from three western states. Concurrently, a fasting blood sample was obtained for assay of estrone, estradiol, free estradiol, testosterone, free testosterone, dehydroepiandrosterone sulfate (DHEAS), sex hormone-binding globulin (SHBG), leptin, C-peptide, insulin-like growth factor-1 (IGF1), and IGF-binding protein-3. Adjusted means of these hormones and peptides were calculated for categories of alcohol intake, overall and stratified by tamoxifen use.
The association between alcohol intake and serum hormone and peptide levels differed by tamoxifen use. We found statistically significant inverse associations between alcohol intake and both leptin and SHBG values, but only among tamoxifen users. In women not using tamoxifen we found a positive association between alcohol intake and DHEAS, but no association in tamoxifen users.
Tamoxifen may modify the association between alcohol intake and serum hormones and peptides. The significant associations found for DHEAS and SHBG are in a direction considered unfavorable for breast cancer prognosis. Postmenopausal breast cancer survivors may benefit from decreasing their alcohol intake.
Most epidemiologic studies have found a positive association between alcohol intake and breast cancer incidence [1–4]. This relationship appears to be modest in magnitude, yet consistent across a variety of populations and study designs. One possible explanation for the association between alcohol and breast cancer is that alcohol consumption increases circulating estrogen and androgen levels, which in turn promote breast cancer development . Alcohol intake may also be associated with other important biomarkers, such as leptin and insulin-like growth factors [5–8], which have proliferative effects independent of the steroid hormone mechanisms [9–11]. In addition to being implicated in breast cancer incidence, these hormone and peptide measures may also promote recurrence and poor prognosis among women diagnosed with breast cancer [12–17] however, we are not aware of published research examining the relationship between alcohol and these serum measures in women with a previous diagnosis of breast cancer.
The positive association between alcohol intake and serum hormone levels appears to be enhanced by the use of exogenous estrogens. Women taking estrogen replacement therapy are reported to have a much greater relative increase in estradiol with increased alcohol consumption compared to those not taking estrogen . Although women with a prior breast cancer diagnosis are unlikely to be taking exogenous estrogens, the use of the anti-estrogen drug tamoxifen is common . Therefore, an important clinical question is whether an exogenous hormone such as tamoxifen alters associations between alcohol and these hormones and peptides implicated in breast cancer risk and recurrence.
The purpose of this study was twofold. First, we examined the relationship between alcohol intake and serum hormone and peptide levels in postmenopausal breast cancer survivors. We hypothesized that alcohol would have the same direct association with serum hormones and peptides in these women as it does in women without breast cancer. Second, we investigated potential interactions of tamoxifen use with any observed associations. Since the presence of exogenous estrogens appears to increase the direct association between alcohol and serum estrogens, we hypothesized that any associations found between alcohol and serum hormones or peptides would be weaker in tamoxifen users than in non-users.
The Health, Eating, Activity and Lifestyle (HEAL) Study is a multi-center, multi-ethnic prospective cohort study of 1183 breast cancer survivors who are being followed to determine whether weight, physical activity, diet, sex hormones, and other exposures affect breast cancer prognosis. Women were recruited into the HEAL study through Surveillance, Epidemiology, and End Results (SEER) registries in New Mexico, Los Angeles county (CA), and western Washington. Details of the study aims, design, and recruitment procedures have been published previously [20–22]. All three study sites included women diagnosed with in situ to Stage IIIA breast cancer but the age ranges and dates of diagnosis differed by site: New Mexico included 615 women, aged 18 years or older, who were diagnosed between July 1996 and March 1999, and lived in Bernalillo, Santa Fe, Sandoval, Valencia, or Taos counties; western Washington included 202 women, ages 40 to 64 years, diagnosed between September 1997 and September 1998, and living in King, Pierce, or Snohomish counties; and in Los Angeles county the cohort consisted of 366 Black women who were diagnosed between May, 1995 and May, 1998.
Baseline data were collected via in-person interviews (for New Mexico and Los Angeles) and self-administered questionnaires (Washington) conducted within 1 year of diagnosis (on average 7.5 months post diagnosis). Follow-up data were collected 24 months after the baseline visit (within the third year after diagnosis; on average 31.5 months post diagnosis). Data on age, education, self-reported race/ethnicity, and measured height were collected at the baseline interview. All other data presented here were obtained at the 24-month follow-up interview.
Written informed consent was obtained from each study participant. The study was performed with the approval of the Institutional Review Boards of participating centers (University of New Mexico, Fred Hutchinson Cancer Research Center, and University of Southern California), in accord with an assurance filed with and approved by the U.S. Department of Health and Human Services.
A 30-ml fasting blood sample was collected at the 24-month follow-up interview. Blood was processed within 3 hours of collection; serum was stored in 1.8-ml aliquot tubes at −70° to −80° C until analysis. We measured estrone, estradiol, testosterone, dehydroepiandrosterone sulfate (DHEAS), sex hormone-binding globulin (SHBG), leptin, C-peptide, insulin-like growth factor-1 (IGF1), and IGF-binding protein-3 (IGFBP3). Free estradiol and free testosterone were calculated from SHBG and total estradiol and testosterone, respectively . We had insufficient blood to measure IGF1 in one subject; C-peptide in two participants; and estradiol/free estradiol in nine participants.
For California participants, all assays except testosterone were performed at the Reproductive and Endocrine Research Laboratory at the University of Southern California. Testosterone assays were performed at the University of New Mexico. For Washington and New Mexico participants, estrone and estradiol were assayed at Quest Diagnostics (San Juan Capistrano, CA) and the remaining assays were conducted at the University of New Mexico. All samples were randomly assigned to assay batches and were randomly ordered within each batch. Biorad standards were also included to monitor assay performance within and between batches at both low and high concentrations. Laboratory personnel performing the assays were blinded to patient identity and personal characteristics.
Radio-immunoassay (RIA) was utilized to measure all serum analytes . Serum extraction and chromatographic purification were performed before RIA for estrone and estradiol. The sensitivity for the estrone assay was less than 5–10pg/mL and for estradiol was less than 2 pg/ml. Serum testosterone was determined using an RIA kit (Diagnostic Products Corp, Los Angeles, CA) with a sensitivity of 40pg/mL. SHBG levels were determined using the Wein Laboratories (Succasunna, NJ) RIA kit with a sensitivity of 6nmol/L. The C-peptide of Insulin 125I RIA kit from Incstar Corp. was used to measure C-peptide levels (sensitivity of 0.1ng/mL). IGF-1 and IGFBP-3 levels were determined by 125I RIA kits supplied from Nichols Institute Diagnostics (sensitivity of 0.1ng/mL and 62.5ng/mL respectively). Leptin was assayed using Linco RIA kits (sensitivity of 0.5ng/ml).
Intra-assay variability was assessed in a reduced randomly selected sample for all hormones. The coefficients of variation (CV) were calculated to test the assay variability. In California, 24 blood samples were randomly selected for hormone assay repeats. The CV was estimated by the standard deviation of the difference of replicated measures divided by the mean of the two measures. The intra-assay CVs for estrone, estradiol, SHBG, IGF-1 and C-peptide were 26.2%, 15.4%, 9.3%, 6.2% and 10.5% respectively.
For New Mexico and western Washington, intra-assay CV's were calculated as the standard deviation of the difference between repeated measures divided by the mean of the two measures. Assays were done in batches, and duplicate aliquots of ten randomly selected participant samples were assayed per batch. The intra-assay and total CVs were 3.8% and 5.9% respectively, for SHBG, 12.0% and 14.4% for testosterone, 29.1% and 13.3% for estrone, 28.8% and 13.3% for estradiol.
Dietary intake data were collected using a self-administered food frequency questionnaire designed for the Women’s Health Initiative . Participants were asked frequency and portion size (small, medium or large) of beer, wine and liquor (1 medium serving was defined as 12 oz beer, 4 oz wine or 1.5 oz liquor). Responses were used to calculate grams per day of alcohol intake. Since our earlier work showed independent associations between fiber and fat intake and hormone levels, we also included data for these dietary variables in the analysis . The nutrient database used to convert food information into nutrients is derived from the University of Minnesota’s Nutrition Coordinating Center’s Nutrition Data Systems for Research (NDS-R, version 2005). New Mexico participants reported their usual dietary intake for the preceding year, participants at the other two sites reported intake for the preceding month. We cannot determine if these different referent periods affected the amount or type of alcohol reported by the three sites because we cannot separate referent period from study site; however, we did control for study site in our analyses.
At the time of the 24-month follow-up interview, women who were 55 years of age or older and who had not menstruated in the last year, or who did not know the date of their last menstruation but reported having had a hysterectomy, were categorized as postmenopausal. Women less than age 55 were also categorized as postmenopausal if they had not menstruated in the year prior to their interview. Ninety-five percent of women reporting natural menopause were menopausal by age 56. Women categorized as pre-menopausal or whose menopausal status was unknown were excluded from these analyses.
Trained staff measured weight and height using a standardized protocol at the clinic or home visit. Body Mass Index (BMI) was calculated as weight in kilograms divided by height in meters squared. A total of 143 participants were missing values for measured height; for these women we used self-reported height to calculate BMI. We have previously compared results of BMI calculated from a subset of women who have both measured and self-reported height, and have found high concordance: only 3 of 569 women had a change in BMI classification from overweight to healthy weight when using self-reported height rather than measured height . We classified BMI according to the World Health Organization categories: normal weight, overweight and obese (<25, 25–29.9, ≥30 kg/m2, respectively) .
We collected information on physical activity using a questionnaire administered in-person at the 24-month follow-up visit. The questionnaire was based on the Modifiable Activity Questionnaire developed by Kriska and colleagues . The type, duration, and frequency of activities performed in the past year were assessed, and for each activity, frequency was multiplied by duration in order to quantify activity level. For the present analysis, physical activity was defined as the number of metabolic equivalent (MET) hours per week in sports and recreational activities. A MET is defined as the ratio of the associated metabolic rate for a specific activity divided by the resting metabolic rate (e.g., a 2-MET activity requires two times the resting metabolic energy expenditure of sitting quietly) .
The sample size for this analysis began with 607 postmenopausal women who had serum and dietary data available from the 24-month interview. We excluded 44 women due to implausible values for dietary data (energy intake < 600 kcal/day or > 4000 kcal/day) and 47 women who reported taking hormone therapy (other than tamoxifen). At the time of the study, no participants were taking aromatase inhibitors. We also excluded 14 participants whose race/ethnicity was classified as ‘Other’ and 12 with missing values for BMI leaving a final sample of 490. We excluded 15 individual hormone/peptide values which were extremely high and deemed to be implausible outliers: estrone > 200 pg/mL; estradiol > 160 pg/mL; testosterone > 1330 pg/mL; DHEAS > 350 ug/dL; IGF1 > 600 ng/mL; C peptide > 7 ng/mL; and leptin > 100 ng/mL. For women with hormone values below the detectable limits, we assigned a value halfway between zero and the lower limit of detection such that 17 women were assigned a testosterone value of 13 pg/mL; 33 were assigned an estrone value of 5 pg/mL; and 1 woman was assigned an estradiol value of 1 pg/mL.
IGFBP3 and C-peptide concentrations were normally distributed. The remaining hormone and peptide concentrations were log-normally distributed and were log-transformed; geometric means are presented for these variables. Alcohol intake was divided into the following categories in order to make them comparable to other similar studies : <1; 1 – 4.9; 5 – 14.9; and 15 or more grams per day. There were three race/ethnicity categories in the HEAL study: Hispanic, Black, and non-Hispanic white. All but three of 53 Hispanics were from New Mexico; all but 1 Black was from Los Angeles; non-Hispanic whites came from both western Washington and New Mexico. We found important differences in hormone levels between non-Hispanic whites depending on which study site the participant was from. For this reason, we present data for non-Hispanic white women by study site. We used least squares regression analysis to estimate the association of alcohol intake with hormone and peptide concentrations. We included in the model any of the demographic or health-related characteristics that were significantly associated with at least some of the hormone/peptide measures and with alcohol use. Because our analysis evaluates the associations between hormones and alcohol stratified by tamoxifen use, we decided to keep tamoxifen use in the final model regardless of whether it was significantly associated with serum hormone values or alcohol intake. The final model included age (continuous), BMI (continuous), education (high school or less, some college, college graduate or higher), activity level (tertiles), tamoxifen use (Yes/No), and a combined variable for race/ethnicity and study site (Hispanic, Black, non-Hispanic white from New Mexico, and non-Hispanic white from western Washington) and number of ovaries (0,1,2). We then ran the models separately by tamoxifen use to investigate possible differential effects. We first examined the distributions of the stratified results to determine whether outlying values may have contributed to the occasional spike or dip in adjusted means and found no evidence of obvious outlying values. The data were analyzed using SAS software (version 9.1.3, SAS Institute, Cary NC).
Table 1 describes the study participants at the 24-month follow-up interview. Their average age was 60.7 years. Almost half (45.1%) of study participants were non-Hispanic whites from New Mexico; 18.8% were non-Hispanic whites from western Washington; 10.8% were Hispanic, and 25.3% Black. Participants were about equally distributed among the three BMI categories of normal weight, overweight, and obese, and almost half of the participants reported use of tamoxifen at the 24-month follow-up interview. Over half (53.5%) reported no use of alcohol; another 7% reported intakes less than 1 g/day so that over 60% of the study group fell into the lowest category of alcohol intake: < 1 g/day. Sixteen percent reported daily intakes of between 1.0 and 5.0 grams; 14% consumed 5.0 – 14.9 grams daily; and less than 10% reported drinking 15.0 or more grams/day.
The associations between various demographic and health-related characteristics and the hormone and peptide concentrations are shown in Table 2. We also present alcohol intake (% drinkers) by each of these variables. There were statistically significant differences between most variables by race/ethnicity. The observed differences between hormone levels for non-Hispanic white women from western Washington and from New Mexico were mainly due to age differences in the two groups: women from western Washington were all under age 65 on entry to the study compared to the women from New Mexico who had a maximum age of 91.
Increasing age was significantly associated with decreasing DHEAS, IGF1, IGFBP3, and alcohol use. More years of education were significantly associated with increasing IGF1, IGFBP3, and alcohol use, and decreasing circulating leptin concentrations. BMI was significantly associated with nine of the 11 hormone/peptide measures and with alcohol use (p for trend <= 0.05); this association was positive for the estrogens, androgens, leptin and C-peptide, and negative for SHBG, IGF1 and alcohol. Conversely, physical activity was inversely associated with estrone, leptin, C-peptide, but positively associated with IGF1 and alcohol use. Women taking tamoxifen had significantly lower levels of free estradiol, free testosterone, and IGF1 but higher levels of SHBG than did women not taking the drug. Intake of dietary fat was directly associated with leptin and C-peptide and negatively associated with IGF1 whereas the associations were reversed for dietary fiber intake with leptin and IGF1.
We investigated the association between prior chemotherapy and hormone levels and alcohol intake (data not shown). We found that estradiol, free estradiol, DHEAS and leptin differed significantly by prior chemotherapy; but there was no association between alcohol intake and chemotherapy (p=0.53). We also found dietary fat and fiber intake to correlate with some of the serum measures but not with alcohol intake (p = 0.18 and 0.51 respectively). Because of our decision to use in the final model only those variables that were significantly associated with both hormone levels and alcohol intake, we did not include chemotherapy, dietary fat or dietary fiber in our final analytic model.
Adjusted means of each of the 11 hormone/peptide measures by alcohol intake are shown in Table 3 for all subjects combined, and separately by tamoxifen use. We observed statistically significant (p < 0.05), positive associations between alcohol intake and testosterone, free testosterone, and DHEAS for the combined study sample (N = 490); we also observed a negative association between alcohol and leptin (p = 0.06) in this group.
The association between alcohol intake and serum measures differed by tamoxifen use in that significant negative associations between alcohol intake and leptin and SHBG values were seen in tamoxifen users but not in non-users. There was also a significant positive association between alcohol intake and DHEAS only among those women who did not report tamoxifen use. The association between alcohol intake and SHBG in tamoxifen users is notable because the range of SHBG differs considerably from that in non-users: 57.5 to 71.0 nmol/L v 36.9 to 41.0 nmol/L respectively. We tested for interaction by creating a cross-product term for alcohol category by tamoxifen use; this interaction term was statistically significant in the model with SHBG as the dependent variable (p = 0.01, data not shown). A similar interaction variable was not significant in models predicting DHEAS and leptin (p = 0.44 and 0.51 respectively). We performed 33 tests of significance in Table 3; the six statistically significant associations found in the table are more than the 1 or 2 that would be expected by chance.
We next evaluated whether the association between alcohol use and SHBG in tamoxifen users was influenced by either estrogen receptor status or body mass index and stratified the analyses by these two factors (data not shown). For estrogen receptor status we were missing data for 25% of the participants and number of women with estrogen receptor negative tumors (59 taking tamoxifen and 15 not taking tamoxifen) was too small to draw meaningful conclusions. For participants with estrogen receptor positive tumors we found no statistically significant association between alcohol intake and SHBG in either category of tamoxifen use. For BMI, we found no association between alcohol intake and SHBG in non-users of tamoxifen regardless of whether subjects had a normal BMI or a BMI > 25 kg/m2; for tamoxifen users there was a significant decrease in SHBG by alcohol intake that was greater for women whose BMI was > 25 kg/m2 (SHBG equal to 65.4, 58.6, and 42.7 nmol/L for alcohol categories < 1gm, 1 to 14.9 gm, and >= 15 gm respectively; p = 0.02) compared to those with a BMI ≤ 25 kg/m2 (SHBG equal to 78.7, 74.2, and 68.0 nmol/L; p = 0.09).
In this cross-sectional study of postmenopausal breast cancer survivors, we found significant associations between alcohol intake and serum hormone and peptide values that differed by tamoxifen use. These associations were strong and significant despite the relatively low range of alcohol intake in study participants. In tamoxifen users we found statistically significant inverse associations between alcohol intake and both leptin and SHBG values which, to our knowledge, have not previously been reported. The positive association we found between alcohol intake and DHEAS has been reported by other researchers; however, in the HEAL study this association exists only in women not using tamoxifen.
We had hypothesized a positive association between alcohol intake and serum DHEAS in our study of breast cancer survivors because this positive association has been found fairly consistently in postmenopausal women without breast cancer. This association has been shown in controlled feeding studies [32–34] as well as cross-sectional analysis [31, 35]. The mechanism by which alcohol influences DHEAS is not completely understood; however, DHEAS is produced by the adrenal glands  and, at least in animals, alcohol is known to stimulate the hypothalamic-pituitary-adrenal axis . Most studies have found high DHEAS levels to be associated with an increased risk of breast cancer [38–40] and with a poorer outcome after diagnosis [41, 42].
Our second hypothesis, that tamoxifen use would dampen any associations between alcohol intake and the serum measures, was supported by the DHEAS results. But we found unexpected negative associations between alcohol intake and both SHBG and leptin in tamoxifen users. Both SHBG and leptin have been implicated in breast cancer incidence and prognosis but the evidence is stronger for SHBG. SHBG binds estradiol reducing the bioavailability of this hormone , and post-menopausal women with low serum SHBG appear to have increased risk of breast cancer [44–46]. The higher levels of bioavailable estrogen associated with low serum SHBG may also affect breast cancer prognosis since high serum estrogen concentrations have been associated with increased recurrence of breast cancer  and also because anti-estrogen therapy is an established treatment for the prevention of breast cancer recurrence [15, 16]. The implications of low leptin levels are less clear those of SHBG. A recent review summarizes ten case-control studies of the association between leptin and breast cancer: three found high leptin levels associated with increased risk of breast cancer and seven found no association . Similarly, studies investigating leptin’s association with breast cancer prognosis are inconsistent [47, 48].
Although we found no published studies on the relationship between alcohol and serum hormones and peptides in breast cancer survivors, one paper examined risk factors for change in estrogen metabolites in breast cancer patients . Specifically, the study measured change in the ratio of 2-hydroxyestrogen (2OHE) to 16-hydroxyestrone (16OHE1) in breast cancer patients pre and post surgery. A high ratio is believed to be associated with reduced risk for breast cancer and better prognosis because 2OHE acts as a weak estrogen or as an anti-estrogen whereas 16OHE1 is pro-carcinogenic. The authors did not look at the effect of alcohol separately by tamoxifen use; however, they reported results from a multivariable model showing the strongest factors associated with a favorable shift in this estrogen ratio were tamoxifen combined with radiation therapy, increasing alcohol intake, and high coffee consumption. The effect of alcohol was significant only for women in the highest category of alcohol intake (at least 4 drinks per week). Estrogen metabolites are different from serum hormone levels and we would not necessarily expect these two measures to be related to alcohol intake in a similar manner. The fact that findings differ depending upon which outcome measures are evaluated suggests more studies of alcohol and hormone levels in breast cancer survivors are needed.
Tamoxifen use has been shown to increase SHBG levels [50, 51], but it is unclear why this drug, an exogenous hormone, would influence the association between alcohol and endogenous hormones. There is, however, evidence that exogenous estrogen use influences the association between alcohol intake and endogenous estradiol levels. A randomized, placebo-controlled study found a significant increase (300% above baseline) in estradiol levels in women using estrogen replacement therapy (ERT) following alcohol ingestion whereas a placebo resulted in minimal increases in estradiol, and alcohol did not change plasma estradiol levels in women not using ERT . The authors indicate that the mechanism behind this is unclear, but other studies have found an increased risk of breast cancer for women on ERT who also drink alcohol [52, 53], above that of either variable alone.
A limitation to this study is the small range of alcohol reportedly consumed. A majority of participants reported drinking less than one gram of alcohol per day. Less than ten percent drank more than 15 grams, or the equivalent of one and a half drinks per day. It is not clear if breast cancer survivors under-report amount of alcohol use, as is common in other populations . The small number of women reporting high alcohol intake also prevented us from investigating associations by type of alcoholic beverage (wine, beer, liquor). This is an observational study that uses self-report of alcohol intake, as opposed to a controlled drinking study where alcohol intake is known. Self-reports generally lead to underestimates of total alcohol consumed , however there is no reason to suspect this underestimate would differ by any of the outcome measures.
Our findings suggest that, in postmenopausal women with a previous diagnosis of breast cancer, the use of tamoxifen modifies the association between alcohol and hormones and peptides implicated in breast cancer prognosis. The associations we report – a negative association between alcohol intake and SHBG and leptin in tamoxifen users and a positive association with DHEAS in non-users – occurred in women whose alcohol intake could be considered light to moderate. The significant associations found for DHEAS and SHBG are in a direction considered unfavorable for breast cancer prognosis. These preliminary findings suggest breast cancer survivors may benefit from moderating their alcohol intake.
National Cancer Institute contracts N01-CN-75036-20, N01-CN-05228, N01-PC-67010/N01-PC-35139, N01-PC-67007/N01-PC-35138 and N01-PC-67009/N01-PC-35142, and training grant T32 CA09661. A portion of this work was conducted through the Clinical Research Center at the University of Washington and supported by the NIH grant M01-RR-00037. Data collection for the Women’s Contraceptive and Reproductive Experiences Study at the University of Southern California was supported by the National Institute of Child Health and Human Development contract N01-HD-3-3175. Patient identification was supported in part by the California Department of Health Services grant 050Q-8709-S1528. Preparation of this manuscript was supported, in part, by the University of New Mexico Cancer Center, a recipient of NCI Cancer Support Grant P30-CA118100.