Abstract
Objectives: To explore associated biological outcomes and clarify the role of timing of exposure in the alcohol-breast cancer relationship.
Methods: In a population-based study of 4,575 women ages 35 to 64 years diagnosed with invasive breast cancer between 1994 and 1998 and 4,682 controls, we collected details of lifetime alcohol use and factors that could confound or modify the alcohol-breast cancer relationship. We used conditional logistic regression to compute the odds of breast cancer among drinkers relative to nondrinkers at all ages and at ages 35 to 49 and 50 to 64 years separately.
Results: Recent consumption (at reference age minus two) of ≥7 drinks per week was associated with increased risk [odds ratio (OR), 1.2; 95% CI, 1.01-1.3] and evidence of dose response was observed. Most of the excess was observed among women ages 50-64 years (OR 1.3; 95% CI, 1.1-1.6), although the test for age interaction was not statistically significant. Exposure later in life seemed more important than early exposure. Excess breast cancer associated with recent consumption was restricted to localized disease. When outcome was examined according to tumor hormone receptor status, highest risks were observed for estrogen receptor–positive/progesterone receptor–negative tumors (OR 1.6; 95% CI, 1.2-2.3).
Conclusions: The effect of timing of alcohol exposure on breast cancer risk is complicated and will require additional study focused on this one issue. Further work is needed to explain how alcohol exposure, sex hormones, and tumor receptor status interact.
Introduction
Breast cancer is the second leading cause of cancer mortality among women in the United States (1). Research has identified multiple risk factors for the disease, but its etiology remains largely unknown. Alcohol consumption was first identified as a risk factor in analyses of Third National Cancer Survey data nearly 30 years ago (2) and has since been confirmed by most published reports and reviews (3, 4). A recent pooled analysis provided strong evidence that the relationship is not confounded by cigarette smoking (5); neither is it likely to be confounded by other variables (6, 7). How alcohol may lead to occurrence of breast cancer is unclear, but various mechanisms have been proposed (8).
Experience suggests that dose, age, and timing of exposure may be related to risk (3-5, 9-18). Additional research has begun to examine the relationship between alcohol and tumor characteristics such as stage, cell type, and hormone receptor status (19-25). The National Institute of Child Health and Development recent Women's Contraceptive and Reproductive Experiences Study (Women's CARE) provided an opportunity to look at these questions further. We designed the current analysis to examine how different characteristics of alcohol exposure may affect breast cancer risk and tumor characteristics.
Materials and Methods
Women's CARE study methods have been described in detail previously (26). This large, population-based case-control study was conducted at five metropolitan sites in the United States (Atlanta, Detroit, Los Angeles, Philadelphia, and Seattle). White or Black female residents ages 35 to 64 years, without a history of breast cancer and first diagnosed with invasive breast cancer between July 1994 and April 1998, were ascertained by Surveillance, Epidemiology, and End Results registry staff at four sites and by field center staff at one (Philadelphia). Using random digit dialing techniques, control subjects were selected from among eligible women enumerated during telephone screening. Control selection fractions were designed to match case frequencies within strata of site, race, and age group. The participation rate was 76.5% among case women and 64.7% among control women. A total of 4,575 cases and 4,682 controls completed interviews.
Interviewers used standardized procedures to contact study participants and conduct in-person interviews. Although the primary purpose of the study was to examine the risk of breast cancer in relation to contraceptive and other hormone use, much of the questionnaire focused on other potential breast cancer risk factors including alcohol consumption. Investigators calculated the timing of reported alcohol exposures with reference to the date of diagnosis (cases) or telephone screening (controls). Reference age was defined as the respondent's age at the reference date.
The study questionnaire recorded alcohol consumption in detail from age of first exposure through the reference age for all women who reported having consumed ≥12 alcoholic drinks in their lifetime and ≥1 drink per month for ≥6 months. (See Appendix A, also available on-line at http://cebp.aacrjournals.org.) A drink was defined as 12 oz of beer, 4 oz of wine, or 1.5 oz of liquor. Women were asked the age at which they first consumed alcohol, the number of drinks they consumed per week or per month at that age, and the age at which the reported pattern first changed. Age at which habits changed marked the end of the first drinking interval and the start of the second. Additional intervals were recorded for each change reported. Women reported average consumption by beverage type for each interval through the reference age.
We calculated the number of drinks consumed per week for each year of age, for each beverage, and for all beverages combined. We averaged values for ages shared by two or more intervals and averaged these values across the periods of time and age used in our analyses. For women whose reference age fell in the middle of a study period, we computed an average value based on the years of data available. We used the a priori categories of 0, <7, and ≥7 drinks per week and split the latter category into ≥7 to <14 and ≥14 when numbers permitted (9, 17, 27-32). Average consumption between 0 and 7 drinks per week was denoted as <7. We calculated average consumption for each 5-year age period, starting at 10 to 14 years, and for overlapping recent time periods, including 1 to 10, 2 to 6, and 2 years and 1 year prior to the reference age. Cumulative lifetime exposure to alcohol was measured three ways: mean number of drinks per week starting at age 15 years, total number of drinking years (duration),and total number of drinks divided by 365 (drink-years).
Surveillance, Epidemiology, and End Results registries affiliated with the study sites, or collaborating hospitals and laboratories, provided information about tumor characteristics. These variables and the a priori categorizations we used to examine them include extent of disease at diagnosis (local or regional/distant; ref. 17), histologic cell type based on the first four digits of the morphology code (33; ductal; lobular and mixed lobular-ductal; papillary, tubular, mucinous, and medullary; and all other; refs. 34, 35) and hormone receptor status [estrogen receptor (ER)/progesterone receptor (PR): ER+/PR+, ER+/PR−, ER−/PR+, ER−/PR−; refs. 22-24]. Hormone receptor status was available for 83.9% of cases; we classified “borderline” ER or PR status as positive.
We tested for two-way multiplicative interaction between recent alcohol consumption (0, <7 or ≥7 drinks per week) and selected breast cancer risk factors and covariates listed in Table 1. We also tested for age interaction between age group and ≥7 drinks per week, a categorical measure (0, <7, ≥7 to <14, ≥14 drinks per week), and a continuous measure of exposure (grams per week). We used the log likelihood ratio test (36) to compare models with and without the interaction terms included. Odds ratios (OR) and 95% confidence intervals (95% CI) associated with alcohol exposure variables were obtained by fitting conditional logistic regression models to the data, using age group (six categories), race (two categories), and study site (five categories) as conditioning variables. Because previous studies have suggested that alcohol may affect younger or premenopausal women differently than older or postmenopausal women (8, 10, 11, 13, 15, 31) and because menopausal status was unknown or unclassified for more than 20% of study participants, we analyzed data for women of all ages combined and for younger women (ages 35-49 years) and older women (ages 50-64 years) separately.
Percentage distribution among cases and controls of factors examined as modifiers of the alcohol-breast cancer relationship and drinking status (Women's CARE study)
Risk factor . | Case, n (%) . | Control, n (%) . | ||
---|---|---|---|---|
Age group (y) | n = 4,575 | n = 4,682 | ||
35-39 | 689 (15.1) | 666 (14.2) | ||
40-44 | 758 (16.6) | 832 (17.8) | ||
45-49 | 782 (17.1) | 857 (18.3) | ||
50-54 | 844 (18.5) | 825 (17.6) | ||
55-59 | 770 (16.8) | 801 (17.1) | ||
60-64 | 732 (16.0) | 701 (15.0) | ||
Race | n = 4,575 | n = 4,682 | ||
White | 2,953 (64.6) | 3,021 (64.5) | ||
Black | 1,622 (35.5) | 1,661 (35.5) | ||
Site | n = 4,575 | n = 4,682 | ||
Atlanta | 881 (19.3) | 895 (19.1) | ||
Detroit | 679 (14.8) | 779 (16.6) | ||
Los Angeles | 1,242 (27.2) | 1,255 (26.8) | ||
Philadelphia | 707 (15.5) | 736 (15.7) | ||
Seattle | 1,066 (23.3) | 1,017 (21.7) | ||
Menopausal status | n = 4,575 | n = 4,682 | ||
Premenopausal or perimenopausal | 2,116 (46.3) | 2,061 (44.0) | ||
Postmenopausal | 1,544 (33.8) | 1,595 (34.1) | ||
Unknown or unclassified | 915 (20.0) | 1,026 (21.9) | ||
Family history of breast cancer* | n = 4,575 | n = 4,682 | ||
No | 3,616 (79.0) | 4,050 (86.5) | ||
Yes | 778 (17.0) | 453 (9.7) | ||
Adopted or unknown | 181 (4.0) | 179 (3.8) | ||
Smoking status | n = 4,573 | n = 4,680 | ||
Never smoker | 2,095 (45.8) | 2,104 (45.0) | ||
Former smoker | 1,498 (32.8) | 1,469 (31.4) | ||
Current smoker | 980 (21.4) | 1,107 (23.7) | ||
Body mass index 5 y before reference date (kg/m2) | n = 4,550 | n = 4,662 | ||
<21.5 | 1,130 (24.8) | 1,104 (23.7) | ||
21.5 to <28.5 | 2,409 (53.0) | 2,413 (51.8) | ||
≥28.5 | 1,011 (22.2) | 1,145 (24.6) | ||
Exercise history (h/wk) | n = 4,566 | n = 4,678 | ||
None | 1,165 (25.5) | 1,135 (24.3) | ||
>0 to <2 | 2,016 (44.2) | 2,023 (43.2) | ||
2+ | 1,385 (30.3) | 1,520 (32.5) | ||
Use of oral contraceptives | n = 4568 | n = 4,675 | ||
Never | 1,042 (22.8) | 990 (21.2) | ||
Ever | 3,526 (77.2) | 3,685 (78.8) | ||
Never | 1,042 (22.9) | 990 (21.2) | ||
Former | 3,313 (72.7) | 3,505 (75.1) | ||
Current | 204 (4.5) | 173 (3.7) | ||
Use of hormone replacement therapy† | n = 4,575 | n = 4,682 | ||
Never | 2,837 (62.0) | 2,749 (58.7) | ||
Ever | 1,738 (38.0) | 1,933 (41.3) | ||
Never | 2,897 (63.4) | 2,838 (60.6) | ||
Former | 420 (9.2) | 540 (11.5) | ||
Current | 1,254 (27.4) | 1,303 (27.8) | ||
Use of continuous combined hormone replacement therapy‡ | n = 4,574 | n = 4,682 | ||
Never | 4,085 (89.3) | 4,233 (90.4) | ||
Former | 109 (2.4) | 147 (3.1) | ||
Current | 380 (8.3) | 302 (6.5) | ||
Drinking status during reference age | n = 4,570 | n = 4,678 | ||
Never | 1,796 (39.3) | 1,900 (40.6) | ||
Former | 822 (18.0) | 824 (17.6) | ||
Current | 1,952 (42.7) | 1,954 (41.8) |
Risk factor . | Case, n (%) . | Control, n (%) . | ||
---|---|---|---|---|
Age group (y) | n = 4,575 | n = 4,682 | ||
35-39 | 689 (15.1) | 666 (14.2) | ||
40-44 | 758 (16.6) | 832 (17.8) | ||
45-49 | 782 (17.1) | 857 (18.3) | ||
50-54 | 844 (18.5) | 825 (17.6) | ||
55-59 | 770 (16.8) | 801 (17.1) | ||
60-64 | 732 (16.0) | 701 (15.0) | ||
Race | n = 4,575 | n = 4,682 | ||
White | 2,953 (64.6) | 3,021 (64.5) | ||
Black | 1,622 (35.5) | 1,661 (35.5) | ||
Site | n = 4,575 | n = 4,682 | ||
Atlanta | 881 (19.3) | 895 (19.1) | ||
Detroit | 679 (14.8) | 779 (16.6) | ||
Los Angeles | 1,242 (27.2) | 1,255 (26.8) | ||
Philadelphia | 707 (15.5) | 736 (15.7) | ||
Seattle | 1,066 (23.3) | 1,017 (21.7) | ||
Menopausal status | n = 4,575 | n = 4,682 | ||
Premenopausal or perimenopausal | 2,116 (46.3) | 2,061 (44.0) | ||
Postmenopausal | 1,544 (33.8) | 1,595 (34.1) | ||
Unknown or unclassified | 915 (20.0) | 1,026 (21.9) | ||
Family history of breast cancer* | n = 4,575 | n = 4,682 | ||
No | 3,616 (79.0) | 4,050 (86.5) | ||
Yes | 778 (17.0) | 453 (9.7) | ||
Adopted or unknown | 181 (4.0) | 179 (3.8) | ||
Smoking status | n = 4,573 | n = 4,680 | ||
Never smoker | 2,095 (45.8) | 2,104 (45.0) | ||
Former smoker | 1,498 (32.8) | 1,469 (31.4) | ||
Current smoker | 980 (21.4) | 1,107 (23.7) | ||
Body mass index 5 y before reference date (kg/m2) | n = 4,550 | n = 4,662 | ||
<21.5 | 1,130 (24.8) | 1,104 (23.7) | ||
21.5 to <28.5 | 2,409 (53.0) | 2,413 (51.8) | ||
≥28.5 | 1,011 (22.2) | 1,145 (24.6) | ||
Exercise history (h/wk) | n = 4,566 | n = 4,678 | ||
None | 1,165 (25.5) | 1,135 (24.3) | ||
>0 to <2 | 2,016 (44.2) | 2,023 (43.2) | ||
2+ | 1,385 (30.3) | 1,520 (32.5) | ||
Use of oral contraceptives | n = 4568 | n = 4,675 | ||
Never | 1,042 (22.8) | 990 (21.2) | ||
Ever | 3,526 (77.2) | 3,685 (78.8) | ||
Never | 1,042 (22.9) | 990 (21.2) | ||
Former | 3,313 (72.7) | 3,505 (75.1) | ||
Current | 204 (4.5) | 173 (3.7) | ||
Use of hormone replacement therapy† | n = 4,575 | n = 4,682 | ||
Never | 2,837 (62.0) | 2,749 (58.7) | ||
Ever | 1,738 (38.0) | 1,933 (41.3) | ||
Never | 2,897 (63.4) | 2,838 (60.6) | ||
Former | 420 (9.2) | 540 (11.5) | ||
Current | 1,254 (27.4) | 1,303 (27.8) | ||
Use of continuous combined hormone replacement therapy‡ | n = 4,574 | n = 4,682 | ||
Never | 4,085 (89.3) | 4,233 (90.4) | ||
Former | 109 (2.4) | 147 (3.1) | ||
Current | 380 (8.3) | 302 (6.5) | ||
Drinking status during reference age | n = 4,570 | n = 4,678 | ||
Never | 1,796 (39.3) | 1,900 (40.6) | ||
Former | 822 (18.0) | 824 (17.6) | ||
Current | 1,952 (42.7) | 1,954 (41.8) |
NOTE: Percentages are calculated with nonmissing values only. Percentages do not always sum to 100.0 because of rounding.
Breast cancer in the woman's mother, sister, or daughter.
Never/ever use includes any estrogen or progestin, including pills, patches, creams, suppositories, and shots. Never/former/current use includes any estrogen pills or patches or any progestin pills.
Any combined use of estrogen pills or patches and >24 days per month of progestin pills.
To test for confounding in the risk estimates for recent alcohol exposure, we examined 20 covariates separately: education level, income index, religion, age at menarche, age at menopause, number of full-term pregnancies, age at first full-term pregnancy, menopausal status, history of tubal sterilization, lifetime exercise level, smoking status, family history of breast cancer, body mass index, number of breast biopsies, number of recent screening mammograms, history of major medical conditions, ever use of oral contraceptives, current use of oral contraceptives, ever use of contraceptive shots and implants, and ever use of hormone replacement therapy. In addition, we modeled each of the primary alcohol exposure variables with and without inclusion of a subset of these variables, selected apriori: age at menarche, age at menopause, age at first term pregnancy, number of term pregnancies, menopausal status, body mass index, family history of breast cancer, ever use of oral contraceptives, and ever use of hormone replacement therapy (37). In some models, we tested for confounding by additional variables, which we describe under Results. None of the unadjusted ORs was altered by more than 10% by any of these procedures, but we report estimates adjusted for the variables identified a priori: We also tested for increasing trends in risk for selected measures of exposure with models containing the relevant ordinal categorical variable and an ever/never indicator variable. We used log likelihood ratio tests to compare the fit of alternate models.
To compare dose response findings in this study with findings from previous studies, we converted the number of drinks per week to grams per week, assuming each serving of beer contained 13.0 g of alcohol, wine 11.0 g, and liquor 15.0 g (38), then fit linear relationships between the log odds of breast cancer and alcohol dose using conditional logistic regression. We tested for significance of the slopes of the lines using Wald tests (36).
Results
By design, cases and controls had similar age and race distributions (Table 1). Cases differed from controls with respect to most known risk factors for breast cancer, but had comparable educational backgrounds and incomes, as was reported in an earlier article that described this population (37). Although we found no evidence of interaction between any of the variables in Table 1 and the effect of recent alcohol exposure on overall breast cancer risk, we report ORs for women of all ages, as well as for women ages 35 to 49 and 50 to 64 years separately.
More than 99.5% of study participants provided complete alcohol histories. Lifetime drinking patterns of cases and controls were similar. Proportions of cases and controls who ever consumed alcohol were 60.8% and 59.4%, respectively. The OR of breast cancer among ever drinkers was 1.0 (95% CI, 0.9-1.1). The case group and control group each consumed 2.4 drinks per week on average from age 15 years through the reference age. Among ever drinkers, cases consumed an average of 3.9 drinks per week and controls 4.1. Average lifetime consumption of ≥7 drinks per week was not associated with increased risk of breast cancer (OR, 1.1; 95% CI, 0.9-1.3). Cases and controls also did not differ in duration or drink-years of exposure (data not shown).
Recency of Exposure and Dose
We explored the importance of recent alcohol exposure by comparing the number of drinks consumed by cases and controls within 10 years of the reference age (Table 2). We show results for two measures of recent exposure only, 1 to 10 years and 2 years before reference age (reference age−2) because results for other measures were similar to those seen at reference age−2. Among all women studied at reference age−2, the odds of breast cancer increased slightly (OR, 1.1; 95% CI, 1.00-1.1) with increasing level of exposure (<7, 7-14, and ≥14 drinks per week). Excess risk observed at reference age−2 seemed restricted to older women, although tests for interaction (age group by ≥7 drinks per week and by the categorical variable) did not reach statistical significance (P < 0.10 and <0.12).
Risk of breast cancer by period of alcohol use, number of drinks per week, and age group (Women's CARE study)
Period of alcohol use . | Overall (N = 9,257) . | . | . | Ages 35-49 y (n = 4,584) . | . | . | Ages 50-64 y (n = 4,673) . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case (n = 4,575) . | Control (n = 4,682) . | OR* (95%percnt; CI) . | Case (n = 2,229) . | Control (n = 2,355) . | OR* (95%percnt; CI) . | Case (n = 2,346) . | Control (n = 2,327) . | OR* (95%percnt; CI) . | |||||||||
1-10 y before reference age | ||||||||||||||||||
None | 2,161 | 2,278 | 1.0 | 949 | 1,015 | 1.0 | 1,212 | 1,263 | 1.0 | |||||||||
<7 | 1,811 | 1,852 | 1.0 (0.9-1.1) | 993 | 1,049 | 1.0 (0.8-1.1) | 818 | 803 | 1.0 (0.9-1.1) | |||||||||
≥7 | 550 | 506 | 1.1 (1.0-1.3) | 260 | 266 | 1.0 (0.8-1.2) | 290 | 240 | 1.3 (1.04-1.5) | |||||||||
7 to <14 | 345 | 312 | 1.1 (1.0-1.3) | 168 | 165 | 1.0 (0.8-1.3) | 177 | 147 | 1.2 (1.0-1.6) | |||||||||
≥14 | 205 | 194 | 1.1 (0.9-1.4) | 92 | 101 | 0.9 (0.7-1.3) | 113 | 93 | 1.3 (1.0-1.7) | |||||||||
OR (95% CI) for trend† | 1.0 (1.0-1.1) | 1.0 (0.9-1.1) | 1.1 (1.0-1.2) | |||||||||||||||
2 y before reference age | ||||||||||||||||||
None | 2,503 | 2,658 | 1.0 | 1,133 | 1,231 | 1.0 | 1,370 | 1,427 | 1.0 | |||||||||
<7 | 1,490 | 1,511 | 1.0 (0.9-1.1) | 824 | 852 | 1.0 (0.9-1.1) | 666 | 659 | 1.0 (0.9-1.1) | |||||||||
≥7 | 534 | 470 | 1.2 (1.01-1.3) | 249 | 248 | 1.0 (0.9-1.3) | 285 | 222 | 1.3 (1.1-1.6) | |||||||||
7 to <14 | 322 | 285 | 1.2 (1.0-1.4) | 154 | 146 | 1.1 (0.8-1.4) | 168 | 139 | 1.2 (1.0-1.6) | |||||||||
≥14 | 212 | 185 | 1.2 (1.0-1.5) | 95 | 102 | 1.0 (0.7-1.3) | 117 | 83 | 1.5 (1.1-2.0) | |||||||||
OR (95%percnt; CI) for trend† | 1.1 (1.0-1.1) | 1.0 (0.9-1.1) | 1.1 (1.02-1.2) |
Period of alcohol use . | Overall (N = 9,257) . | . | . | Ages 35-49 y (n = 4,584) . | . | . | Ages 50-64 y (n = 4,673) . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case (n = 4,575) . | Control (n = 4,682) . | OR* (95%percnt; CI) . | Case (n = 2,229) . | Control (n = 2,355) . | OR* (95%percnt; CI) . | Case (n = 2,346) . | Control (n = 2,327) . | OR* (95%percnt; CI) . | |||||||||
1-10 y before reference age | ||||||||||||||||||
None | 2,161 | 2,278 | 1.0 | 949 | 1,015 | 1.0 | 1,212 | 1,263 | 1.0 | |||||||||
<7 | 1,811 | 1,852 | 1.0 (0.9-1.1) | 993 | 1,049 | 1.0 (0.8-1.1) | 818 | 803 | 1.0 (0.9-1.1) | |||||||||
≥7 | 550 | 506 | 1.1 (1.0-1.3) | 260 | 266 | 1.0 (0.8-1.2) | 290 | 240 | 1.3 (1.04-1.5) | |||||||||
7 to <14 | 345 | 312 | 1.1 (1.0-1.3) | 168 | 165 | 1.0 (0.8-1.3) | 177 | 147 | 1.2 (1.0-1.6) | |||||||||
≥14 | 205 | 194 | 1.1 (0.9-1.4) | 92 | 101 | 0.9 (0.7-1.3) | 113 | 93 | 1.3 (1.0-1.7) | |||||||||
OR (95% CI) for trend† | 1.0 (1.0-1.1) | 1.0 (0.9-1.1) | 1.1 (1.0-1.2) | |||||||||||||||
2 y before reference age | ||||||||||||||||||
None | 2,503 | 2,658 | 1.0 | 1,133 | 1,231 | 1.0 | 1,370 | 1,427 | 1.0 | |||||||||
<7 | 1,490 | 1,511 | 1.0 (0.9-1.1) | 824 | 852 | 1.0 (0.9-1.1) | 666 | 659 | 1.0 (0.9-1.1) | |||||||||
≥7 | 534 | 470 | 1.2 (1.01-1.3) | 249 | 248 | 1.0 (0.9-1.3) | 285 | 222 | 1.3 (1.1-1.6) | |||||||||
7 to <14 | 322 | 285 | 1.2 (1.0-1.4) | 154 | 146 | 1.1 (0.8-1.4) | 168 | 139 | 1.2 (1.0-1.6) | |||||||||
≥14 | 212 | 185 | 1.2 (1.0-1.5) | 95 | 102 | 1.0 (0.7-1.3) | 117 | 83 | 1.5 (1.1-2.0) | |||||||||
OR (95%percnt; CI) for trend† | 1.1 (1.0-1.1) | 1.0 (0.9-1.1) | 1.1 (1.02-1.2) |
ORs (versus none) were derived by conditional logistic regression (with study site, race, and 5-year age group as conditioning variables) and were adjusted for menopausal status, age at menarche, age at menopause, number of term pregnancies, age at first term pregnancy, body mass index, family history of breast cancer, use of hormone replacement therapy, and use of oral contraceptives. Women with missing values for any of the specified variables were excluded.
Trend tests were conducted in models containing the relevant four-level ordinal exposure variable (none, <7, 7 to <14, ≥14) and the addition of an ever/never exposure variable.
To examine the relationship between alcohol dose and breast cancer risk more closely, we converted the average number of drinks per week at reference age −2 to grams and modeled the log odds of breast cancer both as a linear function and for each 60-g increment up to 300 g/wk (data not shown). For women overall, a single, continuous measure for dose fit the data better and the slope of the fitted line showed an overall increase of 1.4% (95% CI, 0.3-2.5%) for each additional 10 g/wk consumed. The point estimates, however, suggested no increased risk with the first two 60-g increments of exposure followed by a step-up in risk with the third (120-180 g/wk, or ∼9-14 drinks per week) and a leveling off of risk at higher doses. Although age interaction was not present (P < 0.13), this pattern was more evident among older women, in whom the nonlinear model offered a better fit (P < 0.025). At consumption of ≥300 g/wk, the slopes of the fitted lines were negative and not statistically significant, although fewer than 2% of cases and controls consumed ≥300 g/wk.
Age of Exposure
We calculated breast cancer risk in relation to age at first use of alcohol and level of exposure at different age periods of use (Table 3). Regardless of reference age, cases were 1.2 times (95% CI, 1.03-1.3) as likely as controls to have delayed first alcohol use until after age 24 years. We also present results for analyses of exposure at each 10-year age period beginning at age 15 years. Positive associations with risk are not apparent for alcohol use prior to age 45 years. None of these estimates were confounded by recent drinking. Among women ages 50 to 64 years, cases consumed more alcohol than controls beginning at age 35 years and consumed significantly more so at ages 45 years and greater. Whereas controls decreased consumption after age 45 years, cases did not (Fig. 1).
Risk of breast cancer, by age at first alcohol use, number of drinks per week according to age period of use, and age group (Women's CARE study)
. | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* (95% CI) . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
Age at first alcohol use | ||||||||||||||||||
Never drinker | 1,779 | 1,883 | 1.0 | 798 | 855 | 1.0 | 981 | 1,028 | 1.0 | |||||||||
≤17 y | 440 | 446 | 1.0 (0.9-1.2) | 287 | 281 | 1.0 (0.8-1.3) | 153 | 165 | 1.0 (0.8-1.3) | |||||||||
18-24 y | 1,539 | 1,628 | 1.0 (0.9-1.1) | 840 | 950 | 0.9 (0.8-1.0) | 699 | 678 | 1.0 (0.9-1.2) | |||||||||
≥25 y | 776 | 686 | 1.2 (1.03-1.3) | 283 | 247 | 1.2 (1.0-1.5) | 493 | 439 | 1.2 (1.0-1.4) | |||||||||
Age period of alcohol use | ||||||||||||||||||
15-24 y | ||||||||||||||||||
None | 2,556 | 2,569 | 1.0 | 1,082 | 1,102 | 1.0 | 1,474 | 1,467 | 1.0 | |||||||||
<7 | 1,797 | 1,872 | 0.9 (0.8-1.0) | 1,003 | 1,097 | 0.9 (0.8-0.99) | 794 | 775 | 1.0 (0.9-1.1) | |||||||||
≥7 | 169 | 193 | 0.9 (0.7-1.1) | 116 | 129 | 0.9 (0.7-1.1) | 53 | 64 | 0.9 (0.6-1.2) | |||||||||
7 to <14 | 106 | 123 | 0.8 (0.6-1.1) | 74 | 81 | 0.9 (0.6-1.2) | 32 | 42 | 0.8 (0.5-1.2) | |||||||||
≥14 | 63 | 70 | 0.9 (0.6-1.3) | 42 | 48 | 0.8 (0.5-1.3) | 21 | 22 | 1.0 (0.6-1.9) | |||||||||
25-34 y | ||||||||||||||||||
None | 2,119 | 2,208 | 1.0 | 911 | 982 | 1.0 | 1,208 | 1,226 | 1.0 | |||||||||
<7 | 1,966 | 1,982 | 1.0 (0.9-1.1) | 1,036 | 1,102 | 1.0 (0.8-1.1) | 930 | 880 | 1.0 (0.9-1.2) | |||||||||
≥7 | 435 | 447 | 1.0 (0.9-1.2) | 252 | 246 | 1.1 (0.9-1.3) | 183 | 201 | 0.9 (0.7-1.2) | |||||||||
7 to <14 | 269 | 262 | 1.0 (0.9-1.3) | 148 | 152 | 1.0 (0.8-1.3) | 121 | 110 | 1.1 (0.8-1.5) | |||||||||
≥14 | 166 | 185 | 0.9 (0.7-1.2) | 104 | 94 | 1.1 (0.8-1.5) | 62 | 91 | 0.7 (0.5-0.99) | |||||||||
35-44 y | ||||||||||||||||||
None | 2,125 | 2,249 | 1.0 | 1,010 | 1,080 | 1.0 | 1,115 | 1,169 | 1.0 | |||||||||
<7 | 1,860 | 1,870 | 1.0 (0.9-1.1) | 947 | 1,001 | 1.0 (0.8-1.1) | 913 | 869 | 1.1 (0.9-1.2) | |||||||||
≥7 | 540 | 516 | 1.1 (0.9-1.2) | 247 | 249 | 1.0 (0.8-1.3) | 293 | 267 | 1.1 (0.9-1.4) | |||||||||
7 to <14 | 321 | 298 | 1.1 (0.9-1.3) | 146 | 159 | 1.0 (0.7-1.2) | 175 | 139 | 1.3 (1.0-1.6) | |||||||||
≥14 | 219 | 218 | 1.1 (0.9-1.3) | 101 | 90 | 1.2 (0.9-1.6) | 118 | 128 | 1.0 (0.8-1.3) | |||||||||
45-54 y | ||||||||||||||||||
None | 1,584 | 1,711 | 1.0 | 401 | 455 | 1.0 | 1,183 | 1,256 | 1.0 | |||||||||
<7 | 1,128 | 1,119 | 1.1 (0.9-1.2) | 288 | 314 | 1.0 (0.8-1.3) | 840 | 805 | 1.1 (0.9-1.2) | |||||||||
≥7 | 384 | 329 | 1.2 (1.04-1.5) | 87 | 85 | 1.1 (0.8-1.6) | 297 | 244 | 1.3 (1.05-1.5) | |||||||||
7 to <14 | 233 | 203 | 1.2 (0.96-1.5) | 58 | 57 | 1.1 (0.7-1.7) | 175 | 146 | 1.2 (0.96-1.6) | |||||||||
≥14 | 151 | 126 | 1.3 (1.01-1.7) | 29 | 28 | 1.1 (0.7-2.0) | 122 | 98 | 1.4 (1.02-1.8) | |||||||||
55-64 y | ||||||||||||||||||
None | 863 | 910 | 1.0 | |||||||||||||||
<7 | 444 | 445 | 1.0 (0.8-1.2) | |||||||||||||||
≥7 | 179 | 135 | 1.4 (1.05-1.7) | |||||||||||||||
7 to <14 | 105 | 83 | 1.3 (0.9-1.7) | |||||||||||||||
≥14 | 74 | 52 | 1.5 (1.01-2.2) |
. | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* (95% CI) . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
Age at first alcohol use | ||||||||||||||||||
Never drinker | 1,779 | 1,883 | 1.0 | 798 | 855 | 1.0 | 981 | 1,028 | 1.0 | |||||||||
≤17 y | 440 | 446 | 1.0 (0.9-1.2) | 287 | 281 | 1.0 (0.8-1.3) | 153 | 165 | 1.0 (0.8-1.3) | |||||||||
18-24 y | 1,539 | 1,628 | 1.0 (0.9-1.1) | 840 | 950 | 0.9 (0.8-1.0) | 699 | 678 | 1.0 (0.9-1.2) | |||||||||
≥25 y | 776 | 686 | 1.2 (1.03-1.3) | 283 | 247 | 1.2 (1.0-1.5) | 493 | 439 | 1.2 (1.0-1.4) | |||||||||
Age period of alcohol use | ||||||||||||||||||
15-24 y | ||||||||||||||||||
None | 2,556 | 2,569 | 1.0 | 1,082 | 1,102 | 1.0 | 1,474 | 1,467 | 1.0 | |||||||||
<7 | 1,797 | 1,872 | 0.9 (0.8-1.0) | 1,003 | 1,097 | 0.9 (0.8-0.99) | 794 | 775 | 1.0 (0.9-1.1) | |||||||||
≥7 | 169 | 193 | 0.9 (0.7-1.1) | 116 | 129 | 0.9 (0.7-1.1) | 53 | 64 | 0.9 (0.6-1.2) | |||||||||
7 to <14 | 106 | 123 | 0.8 (0.6-1.1) | 74 | 81 | 0.9 (0.6-1.2) | 32 | 42 | 0.8 (0.5-1.2) | |||||||||
≥14 | 63 | 70 | 0.9 (0.6-1.3) | 42 | 48 | 0.8 (0.5-1.3) | 21 | 22 | 1.0 (0.6-1.9) | |||||||||
25-34 y | ||||||||||||||||||
None | 2,119 | 2,208 | 1.0 | 911 | 982 | 1.0 | 1,208 | 1,226 | 1.0 | |||||||||
<7 | 1,966 | 1,982 | 1.0 (0.9-1.1) | 1,036 | 1,102 | 1.0 (0.8-1.1) | 930 | 880 | 1.0 (0.9-1.2) | |||||||||
≥7 | 435 | 447 | 1.0 (0.9-1.2) | 252 | 246 | 1.1 (0.9-1.3) | 183 | 201 | 0.9 (0.7-1.2) | |||||||||
7 to <14 | 269 | 262 | 1.0 (0.9-1.3) | 148 | 152 | 1.0 (0.8-1.3) | 121 | 110 | 1.1 (0.8-1.5) | |||||||||
≥14 | 166 | 185 | 0.9 (0.7-1.2) | 104 | 94 | 1.1 (0.8-1.5) | 62 | 91 | 0.7 (0.5-0.99) | |||||||||
35-44 y | ||||||||||||||||||
None | 2,125 | 2,249 | 1.0 | 1,010 | 1,080 | 1.0 | 1,115 | 1,169 | 1.0 | |||||||||
<7 | 1,860 | 1,870 | 1.0 (0.9-1.1) | 947 | 1,001 | 1.0 (0.8-1.1) | 913 | 869 | 1.1 (0.9-1.2) | |||||||||
≥7 | 540 | 516 | 1.1 (0.9-1.2) | 247 | 249 | 1.0 (0.8-1.3) | 293 | 267 | 1.1 (0.9-1.4) | |||||||||
7 to <14 | 321 | 298 | 1.1 (0.9-1.3) | 146 | 159 | 1.0 (0.7-1.2) | 175 | 139 | 1.3 (1.0-1.6) | |||||||||
≥14 | 219 | 218 | 1.1 (0.9-1.3) | 101 | 90 | 1.2 (0.9-1.6) | 118 | 128 | 1.0 (0.8-1.3) | |||||||||
45-54 y | ||||||||||||||||||
None | 1,584 | 1,711 | 1.0 | 401 | 455 | 1.0 | 1,183 | 1,256 | 1.0 | |||||||||
<7 | 1,128 | 1,119 | 1.1 (0.9-1.2) | 288 | 314 | 1.0 (0.8-1.3) | 840 | 805 | 1.1 (0.9-1.2) | |||||||||
≥7 | 384 | 329 | 1.2 (1.04-1.5) | 87 | 85 | 1.1 (0.8-1.6) | 297 | 244 | 1.3 (1.05-1.5) | |||||||||
7 to <14 | 233 | 203 | 1.2 (0.96-1.5) | 58 | 57 | 1.1 (0.7-1.7) | 175 | 146 | 1.2 (0.96-1.6) | |||||||||
≥14 | 151 | 126 | 1.3 (1.01-1.7) | 29 | 28 | 1.1 (0.7-2.0) | 122 | 98 | 1.4 (1.02-1.8) | |||||||||
55-64 y | ||||||||||||||||||
None | 863 | 910 | 1.0 | |||||||||||||||
<7 | 444 | 445 | 1.0 (0.8-1.2) | |||||||||||||||
≥7 | 179 | 135 | 1.4 (1.05-1.7) | |||||||||||||||
7 to <14 | 105 | 83 | 1.3 (0.9-1.7) | |||||||||||||||
≥14 | 74 | 52 | 1.5 (1.01-2.2) |
ORs (versus never drinker or none) were derived by conditional logistic regression (with study site, race, and 5-year age group as conditioning variables) and were adjusted for menopausal status, age at menarche, age at menopause, number of term pregnancies, age at first term pregnancy, body mass index, family history of breast cancer, use of hormone replacement therapy, and use of oral contraceptives. Women with missing values for any of the specified variables were excluded.
Percentage distribution of cases and controls ages 50 to 64 years who consumed ≥7 drinks per week by age period of alcohol use.
Percentage distribution of cases and controls ages 50 to 64 years who consumed ≥7 drinks per week by age period of alcohol use.
Type of Alcoholic Beverage
At reference age−2, 2.5% of cases consumed ≥7 drinks per week of beer, 5.1% consumed this much wine, and 2.8% consumed this much liquor (Table 4). Wine consumption was significantly more common among cases than controls. ORs were higher among older women than younger women, increasing to 2.2 (95% CI, 1.4-3.7) at ≥14 drinks per week (data not shown). Statistical evidence of age interaction, however, was borderline (P < 0.06).
Risk of breast cancer by type of alcoholic beverage, number of drinks per week 2 years before reference age, and age group (Women's CARE study)
Type of alcoholic beverage . | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
Beer† | ||||||||||||||||||
None | 3,659 | 3,748 | 1.0 | 1,684 | 1,764 | 1.0 | 1,975 | 1,984 | 1.0 | |||||||||
<7 | 754 | 779 | 0.9 (0.8-1.1) | 446 | 494 | 0.9 (0.7-1.0) | 308 | 285 | 1.0 (0.9-1.3) | |||||||||
≥7 | 114 | 112 | 1.0 (0.8-1.4) | 76 | 73 | 1.1 (0.8-1.6) | 38 | 39 | 1.0 (0.6-1.6) | |||||||||
Wine† | ||||||||||||||||||
None | 3,065 | 3,210 | 1.0 | 1,414 | 1,538 | 1.0 | 1,651 | 1,672 | 1.0 | |||||||||
<7 | 1,232 | 1,254 | 1.0 (0.9-1.1) | 698 | 704 | 1.1 (0.9-1.2) | 534 | 550 | 0.9 (0.7-1.0) | |||||||||
≥7 | 230 | 175 | 1.3 (1.1-1.6) | 94 | 89 | 1.1 (0.8-1.6) | 136 | 86 | 1.5 (1.1-2.0) | |||||||||
Liquor† | ||||||||||||||||||
None | 3,413 | 3,543 | 1.0 | 1,642 | 1,733 | 1.0 | 1,771 | 1,810 | ||||||||||
<7 | 986 | 978 | 1.0 (0.9-1.2) | 518 | 543 | 1.0 (0.8-1.1) | 468 | 435 | 1.1 (0.9-1.3) | |||||||||
≥7 | 128 | 118 | 1.1 (0.8-1.4) | 46 | 55 | 0.8 (0.5-1.3) | 82 | 63 | 1.3 (0.9-1.9) |
Type of alcoholic beverage . | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
Beer† | ||||||||||||||||||
None | 3,659 | 3,748 | 1.0 | 1,684 | 1,764 | 1.0 | 1,975 | 1,984 | 1.0 | |||||||||
<7 | 754 | 779 | 0.9 (0.8-1.1) | 446 | 494 | 0.9 (0.7-1.0) | 308 | 285 | 1.0 (0.9-1.3) | |||||||||
≥7 | 114 | 112 | 1.0 (0.8-1.4) | 76 | 73 | 1.1 (0.8-1.6) | 38 | 39 | 1.0 (0.6-1.6) | |||||||||
Wine† | ||||||||||||||||||
None | 3,065 | 3,210 | 1.0 | 1,414 | 1,538 | 1.0 | 1,651 | 1,672 | 1.0 | |||||||||
<7 | 1,232 | 1,254 | 1.0 (0.9-1.1) | 698 | 704 | 1.1 (0.9-1.2) | 534 | 550 | 0.9 (0.7-1.0) | |||||||||
≥7 | 230 | 175 | 1.3 (1.1-1.6) | 94 | 89 | 1.1 (0.8-1.6) | 136 | 86 | 1.5 (1.1-2.0) | |||||||||
Liquor† | ||||||||||||||||||
None | 3,413 | 3,543 | 1.0 | 1,642 | 1,733 | 1.0 | 1,771 | 1,810 | ||||||||||
<7 | 986 | 978 | 1.0 (0.9-1.2) | 518 | 543 | 1.0 (0.8-1.1) | 468 | 435 | 1.1 (0.9-1.3) | |||||||||
≥7 | 128 | 118 | 1.1 (0.8-1.4) | 46 | 55 | 0.8 (0.5-1.3) | 82 | 63 | 1.3 (0.9-1.9) |
ORs (versus none) were derived by conditional logistic regression (with study site, race, and 5-year age group as conditioning variables) and were adjusted for menopausal status, age at menarche, age at menopause, number of term pregnancies, age at first term pregnancy, body mass index, family history of breast cancer, use of hormone replacement therapy, and use of oral contraceptives. Women with missing values for any of the specified variables were excluded.
Estimates are computed separately for beer, wine, and liquor, with all three factors in the model.
Tumor Characteristics
Relatively light (<7 drinks per week) at reference age−2, was not associated with localized or metastatic disease at diagnosis (Table 5). In contrast, higher consumption (≥7 drinks per week) was associated with localized disease, particularly among older women. Effect modification by age was present (P < 0.04). We found similarly increased risk of ductal cancer (OR, 1.2; 95% CI, 1.01-1.4) and lobular cancer (OR, 1.2; 95% CI, 0.9-1.6). Excess risk seemed limited to older women, especially in relation to ductal histology, in which evidence of age interaction was also found (P < 0.05). Although the numbers of cases with lobular and mixed lobular-ductal histology who drank ≥7drinks per week were small (n = 38 and 33, respectively), the ORs for the two groups were similarly positive (data not shown).
Risk of breast cancer by tumor characteristics, number of drinks per week 2 years before reference age, and age group (Women's CARE study)
Type of alcoholic beverage . | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
Extent of disease | ||||||||||||||||||
Local | ||||||||||||||||||
None | 1,464 | 2,658 | 1.0 | 628 | 1,231 | 1.0 | 836 | 1,427 | 1.0 | |||||||||
<7 | 923 | 1,511 | 1.0 (0.9-1.1) | 482 | 852 | 1.0 (0.9-1.2) | 441 | 659 | 1.0 (0.9-1.2) | |||||||||
≥7 | 345 | 470 | 1.2 (1.1-1.5) | 143 | 248 | 1.0 (0.8-1.3) | 202 | 222 | 1.5 (1.2-1.8) | |||||||||
Regional or distant | ||||||||||||||||||
None | 1,010 | 2,658 | 1.0 | 489 | 1,231 | 1.0 | 521 | 1,427 | 1.0 | |||||||||
<7 | 547 | 1,511 | 1.0 (0.8-1.1) | 329 | 852 | 1.0 (0.8-1.2) | 218 | 659 | 0.9 (0.8-1.1) | |||||||||
≥7 | 183 | 470 | 1.0 (0.9-1.3) | 104 | 248 | 1.0 (0.8-1.3) | 79 | 222 | 1.0 (0.8-1.4) | |||||||||
Tumor cell type | ||||||||||||||||||
Ductal† | ||||||||||||||||||
None | 1,896 | 2,658 | 1.0 | 876 | 1,231 | 1.0 | 1,020 | 1,427 | 1.0 | |||||||||
<7 | 1,123 | 1,511 | 1.0 (0.9-1.1) | 655 | 852 | 1.0 (0.9-1.2) | 468 | 659 | 1.0 (0.8-1.1) | |||||||||
≥7 | 403 | 470 | 1.2 (1.01-1.4) | 191 | 248 | 1.0 (0.8-1.3) | 212 | 222 | 1.4 (1.1-1.7) | |||||||||
Lobular‡ | ||||||||||||||||||
None | 279 | 2,658 | 1.0 | 106 | 1,231 | 1.0 | 173 | 1,427 | 1.0 | |||||||||
<7 | 181 | 1,511 | 1.0 (0.8-1.3) | 73 | 852 | 0.8 (0.6-1.2) | 108 | 659 | 1.2 (0.9-1.5) | |||||||||
≥7 | 71 | 470 | 1.2 (0.9-1.6) | 23 | 248 | 0.9 (0.5-1.4) | 48 | 222 | 1.5 (1.0-2.1) | |||||||||
Papillary, tubular, mucinous, or medullary§ | ||||||||||||||||||
None | 133 | 2,658 | 1.0 | 59 | 1,231 | 1.0 | 74 | 1,427 | 1.0 | |||||||||
<7 | 86 | 1,511 | 1.2 (0.9-1.6) | 37 | 852 | 0.9 (0.6-1.5) | 49 | 659 | 1.5 (1.0-2.2) | |||||||||
≥7 | 28 | 470 | 1.1 (0.7-1.7) | 15 | 248 | 1.2 (0.6-2.1) | 13 | 222 | 1.1 (0.6-2.0) | |||||||||
Other subtypes∥ | ||||||||||||||||||
None | 195 | 2,658 | 1.0 | 92 | 1,231 | 1.0 | 103 | 1,427 | 1.0 | |||||||||
<7 | 100 | 1,511 | 0.9 (0.7-1.2) | 59 | 852 | 0.9 (0.7-1.3) | 41 | 659 | 0.9 (0.6-1.3) | |||||||||
≥7 | 32 | 470 | 0.9 (0.6-1.3) | 20 | 248 | 1.0 (0.6-1.7) | 12 | 222 | 0.7 (0.4-1.4) |
Type of alcoholic beverage . | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
Extent of disease | ||||||||||||||||||
Local | ||||||||||||||||||
None | 1,464 | 2,658 | 1.0 | 628 | 1,231 | 1.0 | 836 | 1,427 | 1.0 | |||||||||
<7 | 923 | 1,511 | 1.0 (0.9-1.1) | 482 | 852 | 1.0 (0.9-1.2) | 441 | 659 | 1.0 (0.9-1.2) | |||||||||
≥7 | 345 | 470 | 1.2 (1.1-1.5) | 143 | 248 | 1.0 (0.8-1.3) | 202 | 222 | 1.5 (1.2-1.8) | |||||||||
Regional or distant | ||||||||||||||||||
None | 1,010 | 2,658 | 1.0 | 489 | 1,231 | 1.0 | 521 | 1,427 | 1.0 | |||||||||
<7 | 547 | 1,511 | 1.0 (0.8-1.1) | 329 | 852 | 1.0 (0.8-1.2) | 218 | 659 | 0.9 (0.8-1.1) | |||||||||
≥7 | 183 | 470 | 1.0 (0.9-1.3) | 104 | 248 | 1.0 (0.8-1.3) | 79 | 222 | 1.0 (0.8-1.4) | |||||||||
Tumor cell type | ||||||||||||||||||
Ductal† | ||||||||||||||||||
None | 1,896 | 2,658 | 1.0 | 876 | 1,231 | 1.0 | 1,020 | 1,427 | 1.0 | |||||||||
<7 | 1,123 | 1,511 | 1.0 (0.9-1.1) | 655 | 852 | 1.0 (0.9-1.2) | 468 | 659 | 1.0 (0.8-1.1) | |||||||||
≥7 | 403 | 470 | 1.2 (1.01-1.4) | 191 | 248 | 1.0 (0.8-1.3) | 212 | 222 | 1.4 (1.1-1.7) | |||||||||
Lobular‡ | ||||||||||||||||||
None | 279 | 2,658 | 1.0 | 106 | 1,231 | 1.0 | 173 | 1,427 | 1.0 | |||||||||
<7 | 181 | 1,511 | 1.0 (0.8-1.3) | 73 | 852 | 0.8 (0.6-1.2) | 108 | 659 | 1.2 (0.9-1.5) | |||||||||
≥7 | 71 | 470 | 1.2 (0.9-1.6) | 23 | 248 | 0.9 (0.5-1.4) | 48 | 222 | 1.5 (1.0-2.1) | |||||||||
Papillary, tubular, mucinous, or medullary§ | ||||||||||||||||||
None | 133 | 2,658 | 1.0 | 59 | 1,231 | 1.0 | 74 | 1,427 | 1.0 | |||||||||
<7 | 86 | 1,511 | 1.2 (0.9-1.6) | 37 | 852 | 0.9 (0.6-1.5) | 49 | 659 | 1.5 (1.0-2.2) | |||||||||
≥7 | 28 | 470 | 1.1 (0.7-1.7) | 15 | 248 | 1.2 (0.6-2.1) | 13 | 222 | 1.1 (0.6-2.0) | |||||||||
Other subtypes∥ | ||||||||||||||||||
None | 195 | 2,658 | 1.0 | 92 | 1,231 | 1.0 | 103 | 1,427 | 1.0 | |||||||||
<7 | 100 | 1,511 | 0.9 (0.7-1.2) | 59 | 852 | 0.9 (0.7-1.3) | 41 | 659 | 0.9 (0.6-1.3) | |||||||||
≥7 | 32 | 470 | 0.9 (0.6-1.3) | 20 | 248 | 1.0 (0.6-1.7) | 12 | 222 | 0.7 (0.4-1.4) |
ORs (versus none) were derived by conditional logistic regression (with study site, race, and 5-year age group as conditioning variables) and were adjusted for menopausal status, age at menarche, age at menopause, number of term pregnancies, age at first term pregnancy, body mass index, family history of breast cancer, use of hormone replacement therapy, and use of oral contraceptives. Women with missing values for any of the specified variables were excluded.
International Classification of Diseases for Oncology (ICD-O) code 8500.
Lobular (ICD-O code 8520) and mixed lobular and ductal tumors (ICD-O code 8522).
ICD-O codes 8050, 8260, and 8503 (papillary); ICD-O code 8211 (tubular); ICD-O codes 8480 and 8481 (mucinous); and ICD-O codes 8510 and 8512 (medullary).
ICD-O codes 8010, 8020, 8022, 8032, 8033, 8070, 8074, 8140, 8141, 8200, 8201, 8230, 8310, 8401, 8501, 8502, 8504, 8521, 8530, 8541, 8550, 8560, 8562, and 8570.
Cases with discordant hormone receptor status (ER+/PR− or ER−/PR+) were less common than cases with concordant receptor status (ER+/PR+ or ER−/PR−; Table 6). Even so, the highest ORs were detected for these relatively uncommon presentations among the older women. Among younger women, increased risk is suggested only for the ER+/PR− subgroup, that with the highest risk among the older women. Results of tests for age interaction did not reveal significant differences between younger and older women, however.
Risk of breast cancer, according to tumor hormone receptor status, number of drinks of alcohol consumed per week, and age group (Women's CARE study)
Hormone receptor status . | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
ER+/PR+ | ||||||||||||||||||
None | 1,151 | 2,658 | 1.0 | 493 | 1,231 | 1.0 | 658 | 1,427 | 1.0 | |||||||||
<7 | 734 | 1,511 | 1.0 (0.9-1.1) | 390 | 852 | 1.0 (0.8-1.2) | 344 | 659 | 0.9 (0.8-1.1) | |||||||||
≥7 | 270 | 470 | 1.2 (1.0-1.4) | 123 | 248 | 1.1 (0.9-1.4) | 147 | 222 | 1.2 (1.0-1.6) | |||||||||
ER+/PR− | ||||||||||||||||||
None | 179 | 2,658 | 1.0 | 56 | 1,231 | 1.0 | 123 | 1,427 | 1.0 | |||||||||
<7 | 137 | 1,511 | 1.3 (1.04-1.7) | 64 | 852 | 1.5 (1.04-2.3) | 73 | 659 | 1.2 (0.9-1.7) | |||||||||
≥7 | 54 | 470 | 1.6 (1.2-2.3) | 17 | 248 | 1.4 (0.8-2.4) | 37 | 222 | 1.8 (1.2-2.7) | |||||||||
ER−/PR+ | ||||||||||||||||||
None | 120 | 2,658 | 1.0 | 64 | 1,231 | 1.0 | 56 | 1,427 | 1.0 | |||||||||
<7 | 54 | 1,511 | 0.8 (0.5-1.1) | 40 | 852 | 0.9 (0.6-1.3) | 14 | 659 | 0.5 (0.3-1.0) | |||||||||
≥7 | 28 | 470 | 1.4 (0.9-2.1) | 15 | 248 | 1.0 (0.6-1.9) | 13 | 222 | 1.7 (0.9-3.2) | |||||||||
ER−/PR− | ||||||||||||||||||
None | 626 | 2,658 | 1.0 | 338 | 1,231 | 1.0 | 288 | 1,427 | 1.0 | |||||||||
<7 | 336 | 1,511 | 0.9 (0.8-1.1) | 219 | 852 | 0.9 (0.8-1.2) | 117 | 659 | 0.9 (0.7-1.2) | |||||||||
≥7 | 109 | 470 | 1.0 (0.8-1.2) | 65 | 248 | 0.9 (0.7-1.3) | 44 | 222 | 1.0 (0.7-1.5) |
Hormone receptor status . | Overall . | . | . | Ages 35-49 y . | . | . | Ages 50-64 y . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | Case . | Control . | OR* (95%percnt; CI) . | |||||||||
ER+/PR+ | ||||||||||||||||||
None | 1,151 | 2,658 | 1.0 | 493 | 1,231 | 1.0 | 658 | 1,427 | 1.0 | |||||||||
<7 | 734 | 1,511 | 1.0 (0.9-1.1) | 390 | 852 | 1.0 (0.8-1.2) | 344 | 659 | 0.9 (0.8-1.1) | |||||||||
≥7 | 270 | 470 | 1.2 (1.0-1.4) | 123 | 248 | 1.1 (0.9-1.4) | 147 | 222 | 1.2 (1.0-1.6) | |||||||||
ER+/PR− | ||||||||||||||||||
None | 179 | 2,658 | 1.0 | 56 | 1,231 | 1.0 | 123 | 1,427 | 1.0 | |||||||||
<7 | 137 | 1,511 | 1.3 (1.04-1.7) | 64 | 852 | 1.5 (1.04-2.3) | 73 | 659 | 1.2 (0.9-1.7) | |||||||||
≥7 | 54 | 470 | 1.6 (1.2-2.3) | 17 | 248 | 1.4 (0.8-2.4) | 37 | 222 | 1.8 (1.2-2.7) | |||||||||
ER−/PR+ | ||||||||||||||||||
None | 120 | 2,658 | 1.0 | 64 | 1,231 | 1.0 | 56 | 1,427 | 1.0 | |||||||||
<7 | 54 | 1,511 | 0.8 (0.5-1.1) | 40 | 852 | 0.9 (0.6-1.3) | 14 | 659 | 0.5 (0.3-1.0) | |||||||||
≥7 | 28 | 470 | 1.4 (0.9-2.1) | 15 | 248 | 1.0 (0.6-1.9) | 13 | 222 | 1.7 (0.9-3.2) | |||||||||
ER−/PR− | ||||||||||||||||||
None | 626 | 2,658 | 1.0 | 338 | 1,231 | 1.0 | 288 | 1,427 | 1.0 | |||||||||
<7 | 336 | 1,511 | 0.9 (0.8-1.1) | 219 | 852 | 0.9 (0.8-1.2) | 117 | 659 | 0.9 (0.7-1.2) | |||||||||
≥7 | 109 | 470 | 1.0 (0.8-1.2) | 65 | 248 | 0.9 (0.7-1.3) | 44 | 222 | 1.0 (0.7-1.5) |
ORs (versus none) were derived by conditional logistic regression (with study site, race, and 5-year age group as conditioning variables) and were adjusted for menopausal status, age at menarche, age at menopause, number of term pregnancies, age at first term pregnancy, body mass index, family history of breast cancer, use of hormone replacement therapy, and use of oral contraceptives. Women with missing values for any of the specified variables were excluded.
Discussion
Using data from Women's CARE, a large, population-based case-control study, we found a small association between alcohol and breast cancer after adjustment for potentially confounding factors. Several statistically significant effects were observed, including evidence of dose response in relation to levels of recent drinking. Excess risk was most obvious among recent consumers of wine. Although tests for age interaction showed mixed results, excess risk seemed generally restricted to women ages 50 to 64 years and was most apparent in relation to localized disease, ductal or lobular histology, and ER+/PR− receptor profile. Although these results are statistically significant, the magnitude of the association is small and bias cannot be ruled out as a possible explanation.
The strength of the association we observed varies with the measure used for alcohol consumption. Lifetime consumption showed no association, which is consistent with results of most studies that have used this measure (6, 17). Recent consumption, however, showed ORs and dose-response relationships similar to those reported in reviews, meta-analyses, and pooled data sets (3-6). Data from U.S. population surveys and a longitudinal study indicate that drinking in women declines markedly over a lifetime (39-41), a pattern reflected by control women in our study. As a result, average lifetime exposure probably classifies a large number of women as drinkers who have become nondrinkers or less frequent drinkers. This drop-off in alcohol use may explain why studies with the longest periods of elapsed time since assessment of alcohol intake show the weakest alcohol effects (3, 41). The importance of recent drinking is consistent with the importance of current and recent exposure to hormonal contraceptives (42) and hormone replacement therapy (43-48). In our study, the relationship between recent drinking and risk of invasive breast cancer was not modified by measures of exogenous hormone exposure, as some previous reports have indicated (41, 49).
Our results implied that <7 drinks per week conveyed no risk for breast cancer. This observation may reflect the low mean consumption of 2.3 drinks per week among cases and controls in this category. Other studies have failed to find increased risk at similar levels of exposure (9, 14, 17, 28, 4149-51). Average recent consumption among cases and controls who consumed ≥7 drinks per week was much greater: 16.0 drinks per week. Because average consumption among lighter drinkers was much closer to that of nondrinkers than of drinkers of ≥7 drinks per week, it is perhaps not surprising that increased breast cancer risk is not apparent until higher categorical levels of alcohol exposure. At least among women ages 50 to 64 years, gradually increasing levels of exposure may not be associated with gradually increasing levels of risk. Similar results have been reported elsewhere (4, 17, 52).
Our results indicating the importance of recent alcohol exposure and those suggesting exposure later in life conveys a higher risk of breast cancer support the possibility that alcohol acts as a breast cancer promoter rather than initiator (52). This is in contrast to the opinion that alcohol acts at an early stage of breast carcinogenesis and therefore early-life exposure is most important (53). Consistent with some previous findings (31, 52) but not others (9-12, we did not find an increased risk among women who started drinking before age 25 years or who drank alcohol at early ages. Moreover, risk seemed to increase monotonically with increasing age of consumption beginning in the fifth decade of life, which may partially explain why little increased risk was observed among women under age 50 years: the women may not have yet reached the age at which alcohol becomes a risk factor. This finding is not due to confounding by recency of exposure because adjustment for recency did not change the ORs. Levels of consumption not risky at midlife may become increasingly so with advancing age (Fig. 1). Results of individual studies have been mixed (3), but pooled analyses have found a lower risk from alcohol exposure in premenopausal women (4, 5). Furthermore, drinking among younger women may be more sporadic than the drinking among older women (40), and sporadic or binge drinking may not have the same effect on risk as steady drinking (54). Our study measured average consumption over a week or a month but not the pattern of exposure during that period. Studies that cannot distinguish between binge drinking and steady drinking may fail to identify risk associated with only one pattern.
Previous literature has not consistently linked one beverage type to breast cancer risk. However, in several studies the ORs observed for wine were higher than those associated with other alcoholic beverages (2, 55, 56). Additional studies of French and Italian women have also tied wine drinking to increased odds of breast cancer (57-61), although effects of other ethanol-containing beverages were often not considered. In our study, moderate drinkers consumed wine more often than beer or liquor. Such a pattern may implicate wine when, in fact, risk may be associated with moderate consumption of any alcoholic beverage (15, 16). Alternatively, trends in ethanol content and serving size of alcoholic beverages could have contributed to our findings (62); the volume of a typical serving of wine may exceed 4 oz now in the United States, whereas the alcohol content of a typical serving of beer may have declined with the advent of “light” beer. The preference for beverage we observed is not likely due to confounding by socioeconomic status, because cases and controls did not differ on this parameter and because younger women did not show the same pattern; residual confounding by unmeasured life-style factors, however, cannot be excluded.
All observed alcohol-associated risk for breast cancer in our study can be attributed to increased occurrence of localized disease among women ages 50 to 64 years. This finding contrasts with results from a study of women ages 20 to 44 years that found strongest evidence of an association among consumers of ≥14 drinks per week for regional/distant disease (OR, 2.4; 95% CI, 1.6-3.8; ref. 17). Perhaps more intensive screening and follow-up among older women who drink may lead to earlier diagnosis of breast cancer. A history of screening mammography within 2 years of the reference date did not confound the observed relationship in our data, however: among women ages 50-64 years who had ≥7 drinks per week, the proportion of cases with localized disease who had mammography screening (78%) was similar to the proportion of screened controls (76%).
Breast cancer histology has been examined infrequently in relation to alcohol exposure. In two previous studies of alcohol in relation to tumor cell type, higher ORs were observed for lobular cancer than for ductal cancer (19, 25). However, ORs were elevated for both cell types in both studies and no evidence of statistically significant differences between lobular and ductal cancer was reported. We saw no suggestion of increased risk for tumors of other histologies in our data, and our findings suggest that alcohol may increase risk of ductal and lobular tumors similarly.
We noted ORs >1.0 for all hormone receptor subtypes except ER−/PR−, but only in relation to ER+/PR− tumors did the association reach statistical significance (OR, 1.6; 95% CI, 1.2-2.3). ER−/PR+ tumors may in fact be false-negative ER+/PR+ tumors because the gene that codes for the PR receptor is turned on by the estrogen-ER complex (63, 64). If this is true, then all subtypes showing any suggestion of increased risk in this study are in fact ER+. Results from three recent studies found the strongest alcohol associations for ER+/PR+ tumors (22, 25) and ER+ tumors (19). Alcohol use and breast cancer risk factors associated with elevated levels of sex hormones may be more strongly associated with receptor-positive than receptor-negative tumors (19, 65, 66).
A number of our results seem to be biologically plausible if the relationship between breast cancer and alcohol use is mediated through changes in circulating sex hormone levels, the dominant explanatory model at present (41). Other breast cancer risk factors associated with endogenous hormone exposure, such as age at menarche, age at menopause, and parity, suggest sex hormones are involved in the etiology of the disease. Recent evidence from physiologic studies has made this explanation even more likely by demonstrating that alcohol consumption can increase sex hormone levels (67) and that women with breast cancer tend to have higher levels of some sex hormones (68, 69). The differences in risk related to age, histology, and receptor characteristics seen in our study may then be related to different effects on sex hormone levels of different age-specific patterns of alcohol consumption. For example, binge drinking favored by younger women (40) may have less effect on estrogen levels than does steady light drinking. Similarly, the differences by cell type and receptor profile we noted may be related to variation in the etiologic contribution of elevated sex hormone levels to different breast cancer histology and receptor profiles.
The primary strength of our study is its large size, which allowed us to examine more precisely questions about levels of risk and age at exposure as well as some of the more uncommon cell types and receptor combinations. Another strength is the use of in-person interviews and life events calendars, which enabled detailed accountings of alcohol consumption by beverage type for multiple periods within a woman's life. The validity and reliability of consumption histories obtained by retrospective interview have been established previously (70-72), and any underreporting by cases relative to controls would have pushed ORs toward 1.0. The extensive information on cell type and receptor status available in our study has been infrequently reported but may ultimately contribute to our understanding of causal mechanisms.
We did not attempt to blind interviewers to the case-control status of respondents, which is a concern if interviewers had knowledge of the hypotheses under study. The broad content of the questionnaire, however, and the remoteness of alcohol use from the primary exposures of interest minimize the likelihood that interviewer bias could have contributed to our findings. Selection bias is a potential concern because the participation rate among controls was relatively low and lower than that of the cases. Failure to find a difference in socioeconomic status between cases and controls further suggests that more affluent women may have been overrepresented in the control group (73). The percentage of controls living at less than twice the poverty threshold in this study is in fact smaller than that in the general U.S. adult female population (74). In 1996, 24.4% of White women ages 35 to 44 years and 25.8% of those ages 55 to 64 years had household incomes at this level, whereas the corresponding proportions among White controls in our study were 11.6 and 15.0. However, if such selection bias occurred, it would have pushed ORs toward the null, so it cannot account for the excess risk we noted.
Our study also lacked dietary information. Results from other studies of breast cancer risk, however, provide little evidence of interaction between alcohol and nutritional variables (75-77).
This study, based on Women's CARE data, is useful because it confirms an important association between alcohol use and breast cancer and because it is large enough to add to our knowledge of the characteristics of that relationship.
Appendix A. Alcohol history questions, Women's Contraceptive and Reproductive Experiences study
I1. Before ____________ (REFERENCE DATE), did you ever have more than 12 drinks of alcoholic beverages such as beer, wine, or liquor over your entire life?
YES ............................................................1
NO ......................................2 (SECTION J)
I2. Before ____________ (REFERENCE DATE), had you ever drunk alcoholic beverages such as beer, wine, or liquor at least once a month for 6 months or more?
YES ............................................................1
NO ......................................2 (SECTION J)
People's drinking habits often change quite a bit at different times in their lives. I now have some questions about your usual drinking patterns and how your patterns may have changed throughout your life. I am going to ask about beer, wine, and liquor use at various ages. Looking at the calendar may help you remember times in your life when your drinking habits have changed.
Grant support: National Institute of Child Health and Human Development, with additional support from the National Cancer Institute, through contracts with Emory University (N01-HD-3-3168), the Fred Hutchinson Cancer Research Center (N01-HD-2-3166), the Karmanos Cancer Institute at Wayne State University (N01-HD-3-3174), the University of Pennsylvania (N01-HD-3-3176), and the University of Southern California (N01-HD-3-3175) and through an intra-agency agreement with the Centers for Disease Control and Prevention (Y01-HD-7022). The Centers for Disease Control and Prevention contributed additional staff and computer support. The Surveillance, Epidemiology, and End Results Program of the National Cancer Institute provided assistance for study sites in Atlanta (N01-PC-67006), Detroit (N01-CN-65064), Los Angeles (N01-PC-67010), and Seattle (N01-CN-0532).
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.
Acknowledgments
We thank Hoyt G. Wilson for important contributions, other past and present members of the Women's CARE study team, and the women who participated as respondents in the Women's CARE study.