Alcohol consumption and risk of breast cancer by molecular subtype: prospective analysis of the Nurses’ Health Study after 26 years of follow-up

Abstract

Alcohol consumption is a consistent risk factor for breast cancer, although it is unclear whether the association varies by breast cancer molecular subtype. We investigated associations between cumulative average alcohol intake and risk of breast cancer by molecular subtype among 105,972 women in the prospective Nurses’ Health Study (NHS) cohort, followed from 1980 to 2006. Breast cancer molecular subtypes were defined according to estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor 2 (HER2), cytokeratin 5/6 (CK5/6) and epidermal growth factor (EGFR) status from immunostained tumor microarrays in combination with histologic grade. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI). Competing risk analyses were used to assess heterogeneity by subtype. We observed suggestive heterogeneity in associations between alcohol and breast cancer by subtype (p_het =0.06). Alcohol consumers had an increased risk of luminal A breast cancers (n=1,628 cases, per 10g/day increment HR (95%CI) =1.10(1.05–1.15)), and an increased risk that was suggestively stronger for HER2-type breast cancer (n=160 cases, HR (95%CI) =1.16(1.02–1.33)). We did not observe statistically significant associations between alcohol and risk of luminal B (n=631 cases, HR (95%CI) =1.08(0.99–1.16)), basal-like (n=254 cases, HR (95%CI) =0.90(0.77–1.04)), or unclassified (n=87 cases, HR (95%CI) =0.90(0.71–1.14)) breast cancer. Alcohol consumption was associated with increased risk of luminal A and HER2-type breast cancer, but not significantly associated with other subtypes. Given that estrogen receptors are expressed in luminal A but not in HER2-type tumors, our findings suggest that other mechanisms may play a role in the association between alcohol and breast cancer.

Introduction

Breast cancer is the most frequently diagnosed cancer in women, excluding skin cancers, and the second leading cause of cancer-related death among women in the US ¹. Alcohol is one of the most consistent dietary risk factors for breast cancer, with several large pooled analyses suggesting nearly a 50% increased risk among heavy drinkers compared to non-drinkers ^{2, 3}. Given the considerable evidence from controlled feeding ^4–6 and cross-sectional ^{7, 8} studies suggesting that alcohol influences circulating estrogens and may increase breast cancer risk via a steroid hormone pathway, stronger positive associations between alcohol and estrogen receptor (ER) positive tumors might be expected. While the bulk of evidence, including a large meta-analysis of 4 cohort and 12 case-control studies, suggest stronger positive associations of alcohol with ER positive tumors ⁹, positive associations, though potentially weaker, with ER negative breast cancer have also been observed ^9–13. Thus, alcohol may influence breast cancer risk via the estrogen pathway, or through other pathways, including increasing susceptibility to DNA damage, via reactive oxygen species and acetaldehyde ¹⁴.

Molecularly distinct subtypes of breast cancer with different gene expression profiles and clinical outcomes have been identified, and these subtypes may reflect different etiology ^15–19. The molecular subtype classification may be more important for clinical outcomes and for understanding differing etiologies within breast cancer than using single markers like ER. Several studies have observed differences in associations of traditional breast cancer risk factors by molecular subtype, such as reproductive factors ^{15, 20–23}, weight change ¹⁵, family history of breast cancer ^{15, 21}, and body size ^{21, 22, 24}. To our knowledge, only our prior prospective analysis in the Nurses’ Health Study (NHS) ¹⁵ and one case-control analysis, in the Carolina Breast Cancer Study ²², examined associations of alcohol consumption by molecular subtypes, and did not observe any significant heterogeneity. However, these studies included small numbers of cases for certain subtypes, which may have limited the ability to detect significant differences. Additionally, very few studies have evaluated markers beyond ER, PR and HER2 or used the clinical molecular subtype classification. Examining associations by molecular subtype may provide clues to the etiology of alcohol-induced breast cancer. With the addition of 8 years of follow-up and 28% more cases, and the availability of detailed repeated measures of alcohol intake, and molecular characteristics of breast tumors, we examined whether the associations of alcohol and breast cancer differed according to molecular subtype within the NHS, a large prospective cohort study.

Materials and Methods

Study Population

The Nurses’ Health Study (NHS) cohort began in 1976 when 121,700 female registered nurses ages 30–55 years responded to a baseline questionnaire. Women have been followed biennially since baseline to update risk factor information and ascertain new disease diagnoses, with over 90% follow-up in each cycle. This analysis was approved by the institutional review board of the Brigham and Women’s Hospital.

The follow-up period for this analysis began in 1980, when alcohol intake was first assessed among 92,421 women. Women with information on alcohol intake on subsequent questionnaires over follow-up (n=13,551) joined the analysis at first alcohol report. We excluded women who were diagnosed with cancer (other than non-melanoma skin) or died before the beginning of follow-up (n=4,271), and women who did not complete any alcohol assessment over follow-up (n=11,403) or who reported cumulative average alcohol consumption over 100g/day (n=54), leaving a total population of 105,972 women for analysis. Follow-up continued to 2006, the last year tissue data was available.

Exposure and covariate measurement

Alcohol consumption was ascertained from a semi-quantitative food frequency questionnaire first administered in 1980 and updated every 2–4 years. Women reported their average frequency of alcohol consumption during the past year, separately for beer, wine and liquor in nine categories ranging from “none or <1/month” to “40+/week”. Alcohol consumption in grams per day was calculated as the sum of the daily number of drinks multiplied by the average alcohol content per type of alcoholic beverage (12.8 g of alcohol per 12-oz (355 mL) serving of beer, 11.0g per 4-oz (118 mL) serving of wine, and 14.0 g per 1.5-oz (44 mL) serving of liquor) ²⁵. We categorized alcohol consumption as follows: none (reference), 0.1–4.9 g/day, 5.0–9.9 g/day, 10.0–19.9 g/day, and 20+ g/day. Alcohol intake assessed by food frequency questionnaire was highly correlated (Spearman rank-correlation coefficient, r=0.90) with alcohol calculated from dietary records, and with high-density lipoprotein levels (r=0.4) ²⁶. Cumulative average alcohol intake was calculated as an average of alcohol consumption over time for years where alcohol information was available. In addition to the cumulative average alcohol assessment, we evaluated baseline alcohol intake reported in 1980 and current alcohol intake without accounting for prior use. For analysis of most recent alcohol use, person-time for missing alcohol consumption during a specific questionnaire cycle was excluded; women re-entered the analysis when alcohol information became available.

We evaluated age, age at menarche, age at menopause, menopausal status, type of menopause, menopausal hormone (MH) use and duration of use, parity/age at first birth, breastfeeding, family history of breast cancer in a first-degree relative, personal history of benign breast disease, body mass index (BMI) at age 18, weight change since age 18, smoking status and duration, energy intake, menstrual cycle pattern and length, and oral contraceptive (OC) use and duration as potential confounders. Information on height and age at menarche was collected at study baseline in 1976, weight at age 18 was collected in 1980, and breastfeeding was collected in 1986; all other covariates were updated from biennial questionnaires. Missing indicators were created for missing responses for covariates with missing values.

Molecular Subtypes of Breast Cancer

Diagnosis of each molecular subtype of breast cancer (luminal A, luminal B, Her-2 type, basal-like, and unclassified) were the primary endpoints for this analysis. Breast cancer diagnoses and date of diagnosis were assessed on each follow-up questionnaire and confirmed through medical record review; carcinomas in situ were excluded. Pathology reports were available for 96% of the breast cancer cases in this study with 99.4% confirmation rate.

Detailed information on the breast cancer tissue collection and tissue microarray (TMA) construction has been described previously ²⁷. Briefly, archived formalin-fixed paraffin-embedded breast cancer tissue blocks were obtained for approximately 70% of incident primary breast cancer cases from 1976–2006. Women with tissue specimens were very similar with respect to breast cancer risk factors and tumor characteristics compared to the women without tissue specimens available ²⁷. TMAs were constructed in the Dana Farber Harvard Cancer Center Tissue Microarray Core Facility, Boston, MA from 4,308 breast cancers. Three 0.6mm cores from each breast cancer were used in the TMA. Immunohistochemical staining was performed for ER, PR, human epidermal growth factor 2 (HER2), cytokeratin 5/6 (CK5/6), and epidermal growth factor (EGFR) on 5-µm paraffin sections of the TMA block. Methods for performing immunostains for ER, PR, HER2, CK 5/6 and EGFR have been described previously ²⁷. Each marker was assessed in a single staining run on a Dako Autostainer (Dako Corporation, Carpinteria, CA), with appropriate positive and negative controls included in each run. Markers were selected based on evidence for their utility in the classification of molecular phenotypes ^{16, 17, 19, 28, 29}.

Each core on the immunostained TMA slides was evaluated manually for ER, PR, HER2, CK5/6 and EGFR expression. Cases with any nuclear staining for ER or PR in any of the three tissue cores were considered positive; all ER or PR-negative cases had complete absence of tumor cell staining in all tissue cores for that case. HER2 protein overexpression was defined as greater than 10% of cells showing moderate (2+) or strong (3+) membrane staining in any of the tissue cores. Based on prior scoring criteria ^{19, 28, 29}, cases were considered basal CK-positive or EGFR-positive if any cytoplasmic or membranous staining was detected in tumor cores.

Breast cancer molecular subtypes were defined according to ER, PR, HER2, CK5/6 and EGFR status from immunostained tumor microarrays in combination with histologic grade, used as a surrogate for Ki67 proliferation index. Luminal A cases were ER-positive and/or PR-positive and HER2-negative and grade 1 or 2; luminal B cases were either ER-positive and/or PR-positive and HER2-positive or ER-positive and/or PR-positive and HER2-negative with grade 3; HER2-type cases were ER-negative, PR-negative and HER2-positive; basal-like cases were negative for ER and PR, and HER2 and positive for CK 5/6 and/or EGFR; Unclassified tumors lacked expression of all five markers.

Statistical Analyses

Cox proportional hazards models with age in months and calendar year as the underlying time metric were used to estimate hazard ratios (HR) and 95% confidence intervals for the association between alcohol consumption and the risk of breast cancer molecular subtype. Women contributed person-time from the return date of the baseline questionnaire in 1980 until the first diagnosis of any type of cancer (except non-melanoma skin cancer), death, or the end of follow-up (June 1, 2006). We evaluated interaction terms between exposure and time period in our model to assess the proportionality assumptions of the Cox models. We adjusted for several breast cancer risk factors, categorized as shown in the Table 2 footnote.

Table 2.

Cumulative Average Alcohol Consumption and Risk of Breast Cancer by Molecular Subtype

	Alcohol consumption, grams per day					p for trend	RR per 10g/day	p-hetb
	0	0.1–4.9	5.0–9.9	10.0–19.9	20+
Luminal A (n=1,628)
# cases/ person-years	363/560,555	639/941,068	214/295,621	255/291,582	157/179,545			0.06
Age-adjusted HR (95% CI)	1.00 (ref)	1.07 (0.94, 1.21)	1.12 (0.94, 1.33)	1.35 (1.15, 1.58)	1.29 (1.07, 1.55)	0.001	1.11 (1.06, 1.16)
Multivariate HRa (95% CI)	1.00 (ref)	1.03 (0.90, 1.17)	1.07 (0.90, 1.27)	1.29 (1.10, 1.52)	1.24 (1.03, 1.50)	0.001	1.10 (1.05, 1.15)
Luminal B (n=631)
# cases/ person-years	157/560,760	254/941,410	77/295,739	86/291,728	57/179,619
Age-adjusted HR (95% CI)	1.00 (ref)	0.91 (0.74, 1.11)	0.88 (0.67, 1.15)	1.06 (0.81, 1.38)	1.10 (0.81, 1.49)	0.24	1.07 (0.99, 1.15)
Multivariate HRa (95% CI)	1.00 (ref)	0.86 (0.71, 1.06)	0.84 (0.63, 1.10)	1.03 (0.79, 1.34)	1.08 (0.79, 1.47)	0.21	1.08 (0.99, 1.16)
Her2 type (n=160)
# cases/ person-years	33/560,875	69/941,592	20/295,800	20/291,790	18/179,651
Age-adjusted HR (95% CI)	1.00 (ref)	1.24 (0.81, 1.88)	1.18 (0.68, 2.06)	1.20 (0.69, 2.10)	1.68 (0.94, 2.99)	0.15	1.17 (1.03, 1.33)
Multivariate HRa (95% CI)	1.00 (ref)	1.23 (0.80, 1.87)	1.16 (0.66, 2.03)	1.17 (0.66, 2.05)	1.63 (0.91, 2.91)	0.20	1.16 (1.02, 1.33)
Basal-like (n=254)
# cases/ person-years	65/560,836	110/941,558	37/295,778	28/291,777	14/179,657
Age-adjusted HR (95% CI)	1.00 (ref)	1.05 (0.77, 1.43)	1.11 (0.74, 1.66)	0.82 (0.52, 1.28)	0.66 (0.37, 1.18)	0.08	0.90 (0.77, 1.04)
Multivariate HRa (95% CI)	1.00 (ref)	1.04 (0.76, 1.41)	1.09 (0.72, 1.64)	0.81 (0.52, 1.27)	0.66 (0.37, 1.18)	0.08	0.90 (0.77, 1.04)
Unclassified (n=87)
# cases/ person-years	29/560,880	29/941,640	9/295,804	14/291,792	6/179,665
Age-adjusted HR (95% CI)	1.00 (ref)	0.69 (0.41, 1.15)	0.65 (0.31, 1.38)	0.97 (0.51, 1.84)	0.60 (0.25, 1.46)	0.59	0.93 (0.73, 1.17)
Multivariate HRa (95% CI)	1.00 (ref)	0.66 (0.39, 1.11)	0.59 (0.28, 1.26)	0.91 (0.48, 1.75)	0.54 (0.22, 1.32)	0.45	0.90 (0.71, 1.14)

^aMultivariate model adjusted for BMI at age 18 (continuous), weight change since age 18 in kgs (continuous) physical activity in MET hours/week (<3, 3-<9, 9-<18, 18-<27, 27+), parity/age at first birth (nulliparous, 1–2 children <25 years, 1–2 children 25–29 years, 1–2 children 30+ years, 3+ children <25 years, 3+ children 25–29 years, 3+ children 30+ years), menopausal hormone use (premenopausal, postmenopausal never/ever users), oral contraceptive use (never/ever), age at menarche (<12, 12, 13, 14, 15+ years), family history of breast cancer (yes/no), and benign breast disease diagnosis (yes/no).

^bp-heterogeneity was based on the medians of cumulative average alcohol intake categories across subtypes

We used Cox proportional hazards competing risk analysis, using data duplication methods to assess heterogeneity by subtype ³⁰. In these models, the estimates for alcohol intake and covariates with reported heterogeneity by subtype were allowed to vary between breast cancer subtypes, while estimates for other covariates were constrained to a single effect estimate. We compared these models to a model that held all alcohol associations constant between the subtypes using a likelihood ratio test to evaluate differences across subtype.

We used the Wald test to examine the linear trend across categories of alcohol consumption, using the median of each intake category as a continuous variable. Stratified models were examined to assess heterogeneity by BMI (<25/≥25 kg/m²). Test for interaction were conducted using the Wald test for the cross-product interaction term. Because of the strong association between alcohol and tobacco consumption, we conducted a priori sensitivity analyses among non-smokers only and never smokers only, and also examined models additionally adjusted for smoking status. Given that breast cancer risk associated with alcohol intake may be greater in women with low intakes of folate³¹, we examined models additionally adjusted for folate consumption and models stratified by total folate intake. All statistical tests were two-sided and were considered statistically significant at p-value<0.05. All analyses were conducted using SAS software, version 9.3 (SAS Institute, Inc., Cary, North Carolina).

Results

A total of 2,760 cases of invasive breast cancer that could be classified by molecular subtype were diagnosed in our study population over 2.3 million years of follow-up (luminal A (n=1,628), luminal B (n=631), HER2-type (n=160), basal-like (n=254), and unclassified (n=87)). Compared to non-drinkers, women who consumed ≥20 g/day of alcohol were slightly older, leaner, more physically active, more likely to be nulliparous, have past OC and MH use, have a history of benign breast disease, and family history of breast cancer, and were more likely to be current smokers (Table 1). The reported alcohol consumption among women in this study ranged from 0 to 99 g/day, with an average among drinkers of 8.0 g/day.

Table 1.

Age and age-standardized characteristics of 92,421 women in the Nurses’ Health Study in 1980 by alcohol consumption category

	Alcohol Consumption in 1980, grams/day
	0	0.1–4.9	5–9.9	10–19.9	≥20
	(n=29,693)	(n=30,943)	(n=10,008)	(n=14,009)	(n=7,768)
Age in years (mean, SD)	47.0 (7.3)	46.2 (7.3)	46.6 (7.2)	47.4 (7.0)	48.0 (6.7)
BMI at age 18 years in kg/m2 (mean, SD)	21.7 (3.3)	21.4 (3.0)	21.2 (2.7)	21.1 (2.6)	21.0 (2.6)
BMI in kg/m2 (mean, SD)	25.4 (5.1)	24.5 (4.4)	23.8 (3.8)	23.4 (3.6)	23.4 (3.6)
Weight change since age 18 years in kilograms (mean, SD)	9.9 (11.8)	8.4 (10.5)	7.1 (9.5)	6.3 (9.1)	6.4 (9.1)
Height in inches (mean, SD)	64.4 (2.5)	64.5 (2.4)	64.6 (2.4)	64.7 (2.4)	64.7 (2.4)
Physical Activity in MET-hrs/week (mean, SD)	12.1 (18.4)	14.2 (20.8)	15.7 (20.9)	16.0 (22.3)	14.5 (19.0)
Folate in µg/day (mean, sd)	367 (294)	370 (278)	365 (262)	365 (255)	351 (246)
Current Smokers (%)	22.5%	27.0%	30.3%	35.2%	46.7%
Past Smokers (%)	20.3%	28.4%	33.0%	35.4%	31.4%
Age at menarche (mean, SD)	12.4 (1.8)	12.4 (1.8)	12.5 (1.8)	12.5 (1.7)	12.5 (1.8)
Premenopausal (%)	65.1%	66.0%	66.4%	65.8%	64.9%
Nulliparous (%)	5.5%	5.4%	6.0%	6.8%	8.3%
Age at first birtha (mean, sd)	25.2 (3.4)	25.1 (3.3)	25.0 (3.2)	25.2 (3.3)	25.1 (3.3)
Ever breastfeda (%)	60.2%	59.3%	60.7%	60.5%	59.4%
Past OC use (%)	45.4%	48.8%	50.3%	51.5%	52.9%
History of benign breast disease (%)	23.4%	24.5%	26.3%	26.3%	27.2%
Family history of breast cancer (%)	6.1%	6.1%	6.4%	6.6%	6.7%
Ever PMH use (%)	15.5%	15.4%	16.1%	16.1%	16.8%
Age at Menopause (mean, SD)b	47.5 (5.2)	47.7 (5.3)	47.9 (5.1)	47.9 (4.9)	48.1 (4.6)

^aAmong parous women only

^bAmong women with natural menopause or bilateral oophorectomy

Results were similar between the age-adjusted and the multivariate-adjusted models (Table 2); therefore, results from the multivariate-adjusted models are presented here. There was suggestive heterogeneity in associations between alcohol and breast cancer by subtype (p_het =0.06). Alcohol consumers had a significantly increased risk of luminal A (20+g/day vs. non-drinkers HR (95%CI)=1.24 (1.03–1.50), p-trend=0.001, per 10g/day increment HR (95%CI) =1.10(1.05–1.15)), but not luminal B (20+g/day vs. non-drinkers HR (95%CI) =1.08 (0.79–1.47), p-trend=0.21, per 10 g/day HR (95%CI) =1.08(0.99–1.16)) breast cancer (Table 2). A suggestively stronger increased risk was observed for HER2-type (20+g/day vs. non-drinkers HR (95%CI) =1.63 (0.91–2.91), p-trend=0.20, per 10 g/day HR (95%CI) =1.16(1.02–1.33)) breast cancer, though the difference between HER2-type and luminal A was not significant (p_het=0.78). We did not observe statistically significant associations between alcohol and risk of basal-like (20+g/day vs. non-drinkers HR (95%CI) =0.66 (0.37–1.18), p-trend=0.08), per 10 g/day HR (95%CI) =0.90(0.77–1.04)or unclassified (20+g/day vs. non-drinkers HR (95%CI) =0.54 (0.22–1.32), p-trend=0.45, per 10 g/day HR (95%CI) =0.90(0.71–1.14)) breast cancer. Associations with baseline or current alcohol use were similar to the cumulative average results (Supplementary Table 1).

Although we were limited by sample size in stratified analyses for certain subtypes, the observed associations did not differ substantially by BMI and there was no evidence of statistical interaction by age or BMI (data not shown). Further, the results were similar when we restricted to never smokers or non-current smokers, and in models adjusted for smoking (Supplemental Table 2). For example, alcohol consumers had increased risk of luminal A, per 10g/day increment HR (95%CI) =1.09 (1.04–1.15) and HER2-type 1.16 (1.01–1.32) breast cancer after controlling for smoking status. Associations also were similar adjusted for and stratified by folate intake.

Given the observed positive associations of alcohol with HER2-type, and because the luminal B subtype includes both HER2 positive and negative cases, we additionally examined luminal B subtypes by HER2 status. Interestingly, we observed suggestive positive associations for luminal B tumors that were HER2 positive (n=435 cases, per 10g/day increment HR (95%CI) =1.10 (1.00–1.21)), and no evidence for an association between alcohol and HER2 negative luminal B tumors (n=196 cases, per 10g/day increment HR (95%CI) =1.03(0.89–1.19)).

Discussion

In this large, prospective cohort study, we observed suggestive heterogeneity in associations of alcohol and breast cancer molecular subtype, supporting the notion of different etiologies across subtypes. Significant positive associations of alcohol consumption were observed with luminal A and HER2-type, but not with luminal B, basal-like, or unclassified breast cancer.

Given that estrogen and progesterone receptors are negative in HER2- type, our findings suggest that other mechanisms may play a role in the association between alcohol and breast cancer. Alcohol is metabolized to acetaldehyde, which can directly induce DNA damage ^{14, 32}, and may also be broken down to reactive oxygen species (ROS), which could contribute to breast malignancies via DNA mutation, base deletion, and single and double strand breaks ^{14, 32}. Further, alcohol has been shown to downregulate E-cadherin expression, an important tumor suppressor, and to stimulate cell adhesion, migration and invasion in human breast cancer cells ^{33, 34}. Moreover, alcohol consumption may activate phosphatidylinositol 3-kinase (PI3K) pathways ³⁵, which are frequently aberrantly activated in breast carcinogenesis, with mutations occurring in up to 25% of breast cancers ³⁶. Further, overexpression of HER2 was shown to activate the PI3K signaling pathway in human breast cancer cell lines ³⁵ and was associated with an enhanced response to ethanol-stimulated cell proliferation, suggesting that overexpression of HER2 may amplify ethanol-induced signaling and promote alcohol-induced breast tumor metastasis ³⁷.

Alternatively, hormonal mechanisms in alcohol-induced breast carcinogenesis are also possible. Alcohol intake has been associated with increased concentrations of estrogens among premenopausal and postmenopausal women ^4–8, and has been shown to stimulate expression of ER ^38–41, which may influence breast tissue’s sensitivity to estrogens and preferentially enhance proliferation in ER-positive tumors ⁴². Further, positive associations between hormonal factors (e.g. age at menopause ¹⁵ and age at menarche ²⁰) and HER2-type cancers have been observed in prior studies, indicating that hormonal pathways may also be important in HR-negative breast cancer. Moreover, in mouse models, alcohol has been shown to promote HER2-positive tumors only in the presence of estrogens, suggesting that alcohol-induced HER2-type breast cancer may require the presence of circulating estrogen ³⁹. Together, this evidence suggests that alcohol-induced effects on estrogen levels may contribute to the development of HR-positive and HR-negative breast cancer.

The increased risk of luminal A tumors observed in our study is consistent with our prior analysis in the NHS, the only prior prospective analysis to date, in which we observed a significant positive trend between alcohol consumption and luminal A tumors (p=0.04) ¹⁵. The positive associations of alcohol and luminal A breast tumors could be partially explained by hormonal mechanisms given the HR status of luminal tumors ^40–43. However, it is unclear why alcohol appears to increase risk of luminal A but not all luminal B tumors, which are also ER positive. Findings from a population-based case-control study of 1,616 breast cancer cases indicate that alcohol consumption may be associated with advanced, but not early stage breast tumors ⁴⁴, which is somewhat inconsistent with our findings. Thus, future studies may be necessary to understand if there are important differences in associations of alcohol and luminal A and B subtypes.

The HER2 molecular subtype of breast cancer accounted for 5.8% of all breast cancers in our study and these tumors are generally considered more aggressive ⁴⁵ with higher rates of recurrence and overall poorer prognosis, in the absence of HER2 directed therapy ^46–48. There is limited evidence for distinct risk factors for HER2-type breast cancer. In our prior study of traditional risk factors and breast cancer molecular subtypes in NHS, positive associations between alcohol and HER2-type tumors (n=113) were somewhat suggestive but confidence intervals were wide and non-significant ¹⁵. With the additional number of breast cancer cases included in this study, we detected a significant positive association of alcohol intake with HER2-type tumors. Further, our findings suggest that the positive associations with alcohol may be stronger for HER2-type tumors, although not significantly different from associations with luminal A tumors. In the Carolina Breast Cancer case-control study (n=1,424 breast cancer cases), associations of alcohol and HER2-type cancers (n=116) were also similar to those with luminal A tumors; however, luminal B cases showed suggestively stronger positive association with alcohol use relative to the other subtypes ²², which is inconsistent with our findings. However, this study defined luminal B tumors as those that were ER-positive, PR-positive, and HER2-positive, and did not include hormone receptor (HR)-positive, HER2-negative, high grade tumors in their definition of luminal B. In sub-analyses, we observed suggestive associations of alcohol with HER2-positive luminal B tumors, but not among HER2-negative high grade luminal B tumors, suggesting potential heterogeneity within the traditional luminal B subtype. Although based on single marker status from pathology records, and not the HER2 molecular subtype, the European Prospective Investigation into Cancer and Nutrition (EPIC) study recently reported positive associations of alcohol for HER2 negative but not for HER2 positive breast cancer, which is inconsistent with our findings ¹³. Discrepancies in findings between studies could be partially explained by variations in the definition of breast cancer subtypes. The role of alcohol in HER2-type breast cancer is somewhat unclear, though both hormonal ^{15, 20, 39} and non-hormonal mechanisms are possible ^{35, 37}. Future studies are needed to replicate this unique finding of an increased risk of HER2-type breast cancer in relation to alcohol consumption and to examine potential mechanisms underlying this association.

To our knowledge, this study is the largest to evaluate associations between alcohol, a well-established breast cancer risk factor, and breast cancer molecular subtype. Additional strengths of this study include the use of a well-described, large, prospective cohort study with detailed comprehensive assessment of alcohol intake and the consideration of a large number of potential covariates. Potential limitations include the use of self-reported alcohol intake; although, the utility of this measure has been previously demonstrated ²⁶. While alcohol consumption in our study population was rather moderate, we were able to detect significant associations, suggesting adverse effects even with low consumption, as has been reported with overall breast cancer ¹⁰. Small numbers of the rare subtypes (e.g., basal-like) may have limited our ability to detect significant associations in these subgroups, particularly in analyses stratified by menopausal status. Future pooled studies may be necessary to better understand associations with these rare subtypes of breast cancer and to assess whether associations vary across subgroups. Further, our study population included primarily white women, although associations of alcohol and breast cancer overall have not been shown to differ substantially by race ⁴⁹, and it is unlikely that the observed associations by molecular subtype would differ by race.

In this large prospective study, we observed suggestive heterogeneity in the association of alcohol and breast cancer molecular subtypes, emphasizing the importance of considering molecular subtype in etiologic studies of breast cancer. Further, our findings suggest that both hormonal and non-hormonal mechanisms may play a role in the association between alcohol and breast cancer. Given that survival rates for HR-positive tumors are higher than those for HR-negative tumors ⁵⁰ and that most of the known risk factors for breast cancer are associated with HR-positive tumors, identifying modifiable risk factors for the more aggressive HER2 positive tumors is of critical importance.

Brief description of novelty and impact.

It is unclear whether the association of alcohol and breast cancer varies by molecular subtype. In this prospective analysis, we observed suggestive heterogeneity in the association of alcohol and breast cancer subtype, with an increased risk of Luminal A tumors and an increased risk that was suggestively stronger for HER2-type breast cancer. These findings suggest that mechanisms other than estrogenic pathways may play a role in the association between alcohol and breast cancer risk.

Abbreviations

NHS

Nurses’ Health Study

ER

estrogen receptor

PR

progesterone receptor

HER2

human epidermal growth factor 2

CK5/6

cytokeratin 5/6

EGFR

epidermal growth factor

HR

hazard ratios

CI

confidence intervals

MH

menopausal hormone

BMI

body mass index

TMA

tissue microarray

OC

oral contraceptive

ROS

reactive oxygen species

PI3K

phosphatidylinositol 3-kinase

EPIC

European Prospective Investigation into Cancer and Nutrition study