Tel.: 301-594-7660, Fax: +301-402-0916
Alcohol and endometrial cancer risk in the NIH-AARP diet and health study
Version of Record online: 8 OCT 2010
Copyright © 2010 UICC
International Journal of Cancer
Volume 128, Issue 12, pages 2953–2961, 15 June 2011
How to Cite
Yang, H. P., Gierach, G. L., Danforth, K. N., Sherman, M. E., Park, Y., Wentzensen, N., Hollenbeck, A., Schatzkin, A. and Brinton, L. A. (2011), Alcohol and endometrial cancer risk in the NIH-AARP diet and health study. Int. J. Cancer, 128: 2953–2961. doi: 10.1002/ijc.25623
- Issue online: 22 APR 2011
- Version of Record online: 8 OCT 2010
- Accepted manuscript online: 19 AUG 2010 01:00PM EST
- Manuscript Accepted: 5 AUG 2010
- Manuscript Received: 22 JUN 2010
- NIH (Intramural Research Program)
- National Cancer Institute
- endometrial cancer;
- prospective study
Previous investigations have provided conflicting results regarding whether alcohol consumption affects endometrial cancer risk, although in many of these studies the highest category of alcohol intake examined was limited. Further, most were unable to resolve how alcohol associations are affected by beverage type, the presence of other endometrial cancer risk factors, or tumor characteristics. To address these issues, we prospectively evaluated the association between alcohol intake and incident endometrial cancer (n = 1,491) in a cohort of 114,414 US women enrolled in the NIH-AARP Diet and Health Study. We calculated relative risks (RR) and 95% confidence intervals (CI) using Cox proportional hazards regression. After adjustment for age, body mass index (BMI), smoking and other potential confounders, the multivariable RRs (and 95% CIs) compared with nondrinkers were 0.97 (0.87–1.09) for >0–<12 g of alcohol/day, 1.06 (0.87–1.31) for 12–<24 g/day and 0.93 (0.71–1.20) for ≥24 g/day (p trend = 0.90). There was, however, some suggestion of higher risks associated with alcohol consumption among lean women (BMI, <25) and users of menopausal hormone therapy, with significant interactions with both parameters (respective interaction p-values of 0.002 and 0.005). The relationship was also enhanced, albeit nonsignificantly so, for low grade cancers. Our results do not support that alcohol is a strong contributor to endometrial cancer risk, but slight risk increases may prevail among some users or for selected tumor characteristics.
Endometrial cancer, the most common gynecological cancer in the US,1 is well recognized as being affected by hormonal risk factors2 and sex steroid hormones.3 Although alcohol consumption is known to be associated with increased levels of circulating sex steroid hormones,4–7 its relationship to endometrial cancer risk remains unclear.
The association between alcohol intake and endometrial cancer has been studied in seven cohort8–14 and numerous case-control studies,15–27 and the evidence has been summarized in two reviews28, 29 and recently a meta-analysis.30 Studies have largely reported null results, although most investigations have been limited by the highest category of alcohol intake. Among the seven prospective cohort studies, five reported no association,8–10, 13, 14 of which three8, 9, 13 examined only limited ranges of alcohol intake (highest category of alcohol intake ranged from: ≥4 g/day to ≥10 g/day). In a meta-analysis of the seven prospective studies, Friberg et al. reported on a possible J-shaped relationship between alcohol intake and endometrial cancer risk with increased risk for intakes higher than two or more alcoholic drinks per day (≥26 g/day): compared to nondrinkers, the relative risk (RR) was 1.14 (95% confidence interval (CI): 0.95–1.36) for 2–2.5 drinks per day (approximately >26–32.5 g/day) and 1.25 (95% CI: 0.98–1.58) for >2.5 drinks per day (>32.5 g/day).30 This finding is of interest given that this level of consumption is consistent with observations of significantly elevated blood hormone levels observed elsewhere.6, 7
The majority of epidemiologic studies have examined alcohol relationships by beverage type8, 10, 12, 14, 16, 17, 19, 21, 24, 26, 31 and according to established endometrial cancer risk factors.8, 10, 12, 21, 24–27, 31 However, the results of these analyses have been inconclusive. In addition, no prior study has examined whether alcohol associations differ according to tumor characteristics, including tumor grade or stage, which may explain some of the inconsistencies across studies. To further assess these relationships, we analyzed data from the large prospective NIH-AARP Diet and Health Study.
Material and Methods
The NIH-AARP Diet and Health Study design and methodology have been described in detail elsewhere.32 In brief, the NIH-AARP Diet and Health Study was established in 1995–1996 by inviting 3.5 million AARP (formerly known as the “American Association of Retired Persons”) members in six states (California, Florida, Louisiana, New Jersey, North Carolina and Pennsylvania) and two metropolitan areas (Atlanta, Georgia and Detroit, Michigan) to complete a baseline questionnaire. A total of 617,119 self-administered questionnaires were mailed back, of which 566,402 were nonduplicate and satisfactorily completed.
We excluded study participants who used a proxy respondent (n = 15,760); were male (n = 325,174); reported a previous diagnosis of cancer other than nonmelanoma skin cancer (n = 23,950), a history of hysterectomy (n = 82,107) or unknown hysterectomy status (n = 2,927), or menstrual periods that stopped due to surgery (n = 1,830) or radiation or chemotherapy (n = 117); developed nonepithelial endometrial cancer during follow-up (n =108); or died or moved out of the study area (n = 15). The resulting cohort consisted of 114,414 women. The NIH-AARP Diet and Health Study was approved by the Special Studies Institutional Review Board of the US National Cancer Institute.
Cohort members were followed through the US Postal Service national database of address changes and for updated vital status through the US Social Security Administration Death Master File and the National Death Index Plus. Incident endometrial cancers were identified by probabilistic linkages with cancer registries in the original recruitment areas and two common states of relocation (Arizona and Texas). The completeness of case ascertainment in this cohort has been reported previously, with an estimated sensitivity of ∼90% and specificity of 99.5% with respect to identification of cases by cancer registry linkage.33 Follow-up time was defined as time from study baseline (between 1995 and 1996) until diagnosis of any cancer, date of death, the date moved out of registry ascertainment area, or last follow-up (December 31, 2006). From baseline through December 31, 2006, 1,650 study subjects developed incident endometrial cancer and 1,491 are included in the analysis after the exclusions described in the previous section. Stage and grade was available for 56% (N = 831) and 93% (N = 1,384) of endometrial cases included in this analysis, respectively.
Alcohol and covariate assessment
The baseline questionnaire elicited information about demographic factors, anthropometry, reproductive factors, medical history and diet. The 124-item food-frequency questionnaire asked about a study participant's usual alcohol intake at home and in restaurants in the preceding year. This included 10 frequency categories ranging from never to ≥6 times per day and three portion sizes for beer (<12 ounces, one to two 12 ounce cans, > two 12 ounce cans), wine or wine coolers (<4 ounces, 4–8, >8) and liquor or mixed drinks (<1 shot, 1–2, >2). We converted intake frequency and portion sizes to grams/day by multiplying beverage-specific values of consumption by their respective grams of alcohol equivalents: 12 ounce beer, 12.96 g; 5 ounce wine or wine coolers, 13.72 g; and 1.5 ounce liquor, 13.93 g.34 We then summed these values to obtain total daily alcohol consumption. Total alcohol intake from alcoholic drinks was categorized into four categories: 0 g/day, >0 to <12 g/day, 12 to <24 g/day and ≥24 g/day, which approximately equates to nondrinkers, <1 drink/day, 1 to <2 drinks/day, ≥ 2 drinks/day, respectively.
Cox proportional hazards regression was used to estimate RR and 95% CIs with age as the time metric. We present the RR and 95% CI for a model adjusted for age and body mass index (BMI), two important endometrial cancer risk factors, and a multivariable model that included the following covariates: age, BMI, smoking status, “race/ethnicity, parity, oral contraceptive (OC) use, menopausal hormone therapy (MHT) use, and age at menopause. Total alcoholic beverage consumption was examined in addition to alcohol intake by beverage type (beer, wine, liquor). For covariates with missing data, women were coded into a separate category. Tests for linear trends across the alcohol categories were calculated by using a variable containing the median value of alcohol intake (grams/day) within the defined alcohol categories.
We also assessed interactions with BMI, smoking status, MHT use, age at menopause, OC use, and parity by using cross-product terms in the model as well as calculating joint effect risk estimates. The likelihood ratio test was used to determine the significance of interactions between alcohol intake and these variables. Alcohol associations were also examined by clinical characteristics of the tumor, specifically stage and grade.
For all analyses, p-values of <0.05 were considered statistically significant. All tests of statistical significance were two-tailed. Analyses were performed using SAS software release 9.1.3 (SAS Institute, Cary, NC).
Selected characteristics of analyzed cohort
A total of 114,414 women contributed 1,066,722.6 person-years, with average follow-up of 5.2 years for cases and 9.4 years for noncases. The mean ± standard deviation ages for entry were 62.3 ± 5.3 years for cases vs. 61.6 ± 5.5 for noncases; for ages at exit comparable values were 67.4 ± 5.8 years for cases and 71.0 ± 5.9 for noncases. Most women were White (90%) and postmenopausal (90%). Women who were overweight (BMI = 25–29.9 kg/m2) or obese (BMI ≥30 kg/m2) at baseline contributed 31% and 21% of the total person years, respectively.
Baseline characteristics of the women included in our analyses are shown in Table 1 according to categories of alcohol intake. The majority (57%) were light alcohol consumers (>0–<12 g/day), while 9% were moderate alcohol consumers (12–<24 g/day) and 6% heavy alcohol consumers (≥24 g/day). Compared to the nondrinkers, alcohol consumers were slightly more likely to have education beyond the high school level, and to be ever smokers and OC or MHT users. They were also less likely to be obese. The distributions of race or ethnicity, age at menarche, age at menopause and parity were similar across alcoholic beverage intake categories.
As previously described in this cohort, endometrial cancer was positively associated with BMI and later age at menopause, and inversely associated with duration of OC use, parity, cigarette smoking, later age at menarche and non-White races.35
Associations using baseline alcohol intake
Overall, baseline alcohol consumption was not statistically significantly associated with endometrial cancer risk (Table 2). No linear trends were observed between alcohol consumption and endometrial cancer in age and BMI-adjusted analyses (p trend = 0.66) or in analyses further adjusted for additional confounders (p trend = 0.90). Three percent of the cohort reported consuming 24-≥36 g/day, while another three percent reported more excessive drinking (≥36 g/day). There was no evidence of any alteration in endometrial cancer risk among either of these groups.
We also examined endometrial cancer risk in relation to intake of specific alcoholic beverages. Among those reporting alcohol consumption at baseline, 36%, 62% and 49% of the cohort consumed beer, wine and liquor, respectively. The mean (standard deviation) of any alcohol intake among beverage-specific nondrinkers were 3.2 g/day (12.1 g/day), 2.7 g/day (16.8 g/day) and 2.3 g/day (10.6 g/day) for beer, wine and liquor nondrinkers, respectively. No clear associations were found between these beverages and endometrial cancer risk before or after adjustment for the other alcoholic beverage types (Table 2). Given the strong inverse associations between cigarette smoking and endometrial cancer in this and other studies, we also examined the association only among never smokers (n = 50,118 women), and results remained essentially the same.
Interactions with alcohol intake
Table 3 summarizes the joint associations on endometrial cancer risk of alcohol consumption and selected endometrial cancer risk factors. Alcohol relationships were not modified by smoking status, OC use, or parity. However, alcohol associations were significantly modified by BMI, MHT use and age at menopause (BMI p interaction = 0.002; MHT use p interaction = 0.005; age at menopause p interaction = 0.004). There was some suggestion that alcohol consumption was positively associated with risk among lean women, MHT users, and women with menopause onset at 55 years old or greater, although trends were not statistically significant. In contrast, significant inverse trends were observed among heavier women (p trend = 0.04). In addition, the interaction between alcohol intake and age at menopause was no longer statistically significant among nonsmokers (p interaction = 0.120). We also examined a cross-tabulation of BMI and MHT use (data not shown). We found that alcohol intake was most clearly associated with increased endometrial cancer risk among MHT users in lean women: compared to nondrinkers, increased risk was observed for >0–12 g/day (RR = 1.33; 95% CI: 0.95–1.87), 12-<24 g/day (RR = 1.60; 95% CI: 1.05–2.45) and ≥24 g/day (RR = 1.28; 95% CI: 0.73–2.23).
Alcohol relationships did not appear to vary by stage at diagnosis (Table 4), but there was some indication of a slight increase in risk for lower grade tumors (p trend = 0.09) and a reverse trend for higher grade (III-IV) tumors (p trend = 0.04). This latter relationship was based on small numbers of endometrial cancer cases. Given the small numbers, we did not attempt to examine relationships among nonsmokers.
We performed several sensitivity analyses for the association between alcohol intake and endometrial cancer. We also adjusted individually for calendar time, calories, red meat, dietary fiber, coffee and several endometrial cancer risk factors, including education, age at menarche, self-reported diabetes, self-rated health quality and physical activity; results were essentially the same and are not shown here. In addition, given the positive correlation between alcohol intake and smoking (heavy drinkers are often heavy smokers), we adjusted for smoking dose and smoking status (never; former ≤20 cigs/day; former >20 cigs/day; current ≤20 cigs/day; current >20 cigs/day; unknown), and results remained essentially the same. The results were also similar when we restricted the analysis to postmenopausal women.
The overall null association was similarly observed regardless of whether the reference was those who were nondrinkers, those who consumed the lowest category of alcohol (>0–<12 g/day), or those who consumed less than the median of the lowest category of alcohol (<1 g/day). We also examined the risk associations with beverage-specific alcohol intake and exclusive beverage-specific alcohol intake compared to nondrinkers of any alcohol, and observed no statistically significant associations. In addition, we also excluded cases (N = 278) identified in the first 2 years of follow-up to account for any preclinical symptoms that may have led women to stop drinking, and found similar associations as presented in Table 2.
The NIH-AARP Diet and Health Study provided a unique opportunity to prospectively examine the association of alcohol with endometrial cancer risk in a large cohort of women. Endometrial cancer risk was unaffected by amounts or types of alcoholic beverages consumed.
Five prospective studies8–10, 13, 14 and 10 case-control studies15, 17, 19, 20, 22–24, 26, 27, 31 also observed overall null associations, while a few others reported a positive or negative association between alcohol and endometrial cancer risk.12, 16, 21, 25 We were particularly interested in examining the association with higher amounts of consumption (notably ≥24 g/day), but found no evidence of increased risk even among these heaviest consumers. Similar to our null results, no association was observed in the Million Women Study14 and in the National Breast Screening Study11 with highest alcohol consumers, ≥15 drinks/week (i.e. ≥21 g/day) and ≥30 g/day, respectively. Our results contrast with those of Setiawan et al. who had a comparable number of cases in the highest category of alcohol intake (≥24 g/day) similarly based on intake during the year preceding the baseline questionnaire, but found a statistically significant twofold increased risk.12 Another smaller cohort study similarly found an increased risk (RR = 1.78) only among the highest alcohol consumers (>30 g/day), but this was not statistically significant.10
We found some suggestion of higher risks associated with alcohol consumption among lean women, MHT users, and those with older ages at menopause with significant interactions with each parameter. Given the positive association between alcohol intake and smoking (heavy drinkers are often heavy smokers), we examined relationships among never smokers only. While interactions with obesity and MHT use remained statistically significant, the interaction for age at menopause was no longer significant (p interaction = 0.120). The one cohort study that examined age at menopause reported observing no interaction with alcohol intake (p interaction = 0.39).10 Similar to our data which suggest an increased endometrial cancer risk associated with alcohol intake among lean women, a previous study reported a stronger positive association in lean women (BMI <25 kg/m2),12 however others have found an inverse association16 or stronger positive association21 in heavier women. Earlier studies examining possible interactions between alcohol intake and MHT have not been consistent in their observations,28 but our results align with reports that alcohol consumption is more strongly related to increased estradiol levels among MHT users than nonusers.36, 37 We also examined the joint association between BMI and MHT in its alcohol-endometrial cancer association, and our data suggest that moderate alcohol intake among lean women using MHT had the strongest increased risk. A limitation of the baseline questionnaire is that MHT formulation was not captured. While formulation was captured on the follow-up questionnaire, case numbers were too small to examine formula-specific interactions. Since estrogens alone are associated with much higher RRs than estrogen plus progestin formulations, future work needs to address the relationship between alcohol intake and MHT formulations.
The mechanisms by which alcohol, obesity and estrogen influence endometrial cancer risk are not well understood. Adipose tissue is a significant site of endogenous estrogen production particularly among postmenopausal women,38 a mechanism hypothesized to underlie the high risks of endometrial cancer observed among obese women. Alcohol consumption has been related to elevated circulating estrogen levels,4–7 but an increased risk associated with alcohol may be undetectable among heavier women because of their generally higher estrogen levels. This notion was supported by a previous analysis that showed that most endometrial cancers among MHT users occurred in lean and moderately overweight women (BMI <30 kg/m2) and most endometrial cancers among nonusers occurred in obese women (BMI ≥30 kg/m2).35 We did not see this compensative effect of higher endogenous levels among MHT users. On the other hand, the modest levels of estrogen in MHT users may be more sensitive to the synergistic effect of increased estrogen associated with MHT and alcohol intake. However, we cannot rule out the possibility that we observed the effect modifications by BMI and MHT use by chance alone. Further studies are required to establish a relationship between alcohol intake and blood estradiol levels in women who are lean and type and use of MHT.
We also attempted to assess whether the alcohol-endometrial cancer association differed by endometrial cancer tumor characteristics and observed a slight increase in risk for lower grade tumors. Our finding, however, needs to be cautiously interpreted given that it was based on small numbers and no other published studies have reported this finding. We were not able to classify cases according to histologic type of tumor because of incomplete information among those selected for this alcohol analysis.
The major strengths of our study include the large, prospective evaluation of alcohol consumption on incident risk of endometrial cancer. In addition, the detailed NIH-AARP study questionnaire allowed for examination of the association between a wider range of total and beverage-specific alcohol intakes with endometrial cancer risk and provided information on potential confounders and effect modifiers, which allowed for a thorough assessment of the independence of alcohol from other related factors and the joint effects between alcohol and these lifestyle factors, and tumor characteristics. However, the questionnaire asked about average drinking intake in the year before questionnaire completion date and included both drinking and nondrinking days, which did not allow for investigation of patterns of drinking (regular/irregular, binge/not binge),39 duration of drinking, or of changes in lifetime drinking pattern. In addition, we could not separately assess cancer risk for nondrinkers and former drinkers. If nondrinkers are former drinkers who stopped drinking due to their illness,40 our estimates might be biased towards the null.
In conclusion, our results do not support alcohol as a strong contributor to endometrial cancer risk, even with moderate or greater alcohol intake, except possibly an increased risk among women who are lean and are MHT users, or for low grade cancers. Future studies examining the association between alcohol and endometrial cancer should be conducted in large studies with lifetime alcohol consumption history to assess associations by beverage type, potential effect modifiers, and clinical characteristics of the tumors.
Cancer incidence data from the Atlanta metropolitan area were collected by the Georgia Center for Cancer Statistics, Department of Epidemiology, Rollins School of Public Health, Emory University. Cancer incidence data from California were collected by the California Department of Health Services, Cancer Surveillance Section. Cancer incidence data from the Detroit metropolitan area were collected by the Michigan Cancer Surveillance Program, Community Health Administration, State of Michigan. The Florida cancer incidence data used in this report were collected by the Florida Cancer Data System under contract to the Department of Health (DOH). The views expressed herein are solely those of the authors and do not necessarily reflect those of the contractor or DOH. Cancer incidence data from Louisiana were collected by the Louisiana Tumor Registry, Louisiana State University Medical Center in New Orleans. Cancer incidence data from New Jersey were collected by the New Jersey State Cancer Registry, Cancer Epidemiology Services, New Jersey State Department of Health and Senior Services. Cancer incidence data from North Carolina were collected by the North Carolina Central Cancer Registry. Cancer incidence data from Pennsylvania were supplied by the Division of Health Statistics and Research, Pennsylvania Department of Health, Harrisburg, Pennsylvania. The Pennsylvania Department of Health specifically disclaims responsibility for any analyses, interpretations or conclusions. Cancer incidence data from Arizona were collected by the Arizona Cancer Registry, Division of Public Health Services, Arizona Department of Health Services. Cancer incidence data from Texas were collected by the Texas Cancer Registry, Cancer Epidemiology and Surveillance Branch, Texas Department of State Health Services. We are indebted to the participants in the NIH-AARP Diet and Health Study for their outstanding cooperation. We also thank Mr. Sigurd Hermansen and Ms. Kerry Grace Morrissey from Westat for study outcomes ascertainment and management and Ms. Leslie Carroll at Information Management Services for data support and analysis.
- 7The association of plasma androgen levels with breast, ovarian, and endometrial cancer risk factors among postmenopausal women. Int J Cancer 2009., , , , , , , .
- 32The National Institutes of Health-American association of retired persons (NIH-AARP) diet and health study: power to detect diet-cancer associations after adjusting for measurement error. Am J Epidemiol 2001; 153: S259., , , , .
- 33Comparison of cancer registry case ascertainment with SEER estimates and self-reporting in a subset of the NIH-AARP Diet and Health Study. J Registry Manag 2005; 32: 70–5., , , , , , .
- 34MyPyramid equivalent database for USDA survey food codes, 1994–2002, version 1.0. USDA, USA, 2006., .