Cancer Causes & Control

, Volume 18, Issue 4, pp 361–373

Alcohol intake and breast cancer risk: the European Prospective Investigation into Cancer and Nutrition (EPIC)

Authors

    • Institute of Cancer Epidemiology, Danish Cancer Society
  • Jane Christensen
    • Institute of Cancer Epidemiology, Danish Cancer Society
  • Anja Olsen
    • Institute of Cancer Epidemiology, Danish Cancer Society
  • Connie Stripp
    • Institute of Cancer Epidemiology, Danish Cancer Society
  • Birthe L. Thomsen
    • Institute of Cancer Epidemiology, Danish Cancer Society
  • Kim Overvad
    • Department of Clinical EpidemiologyAalborg Hospital and Aarhus University Hospital
  • Petra H. M. Peeters
    • Julius Center for Health Sciences and Primary CareUniversity Medical Center
  • Carla H. van Gils
    • Julius Center for Health Sciences and Primary CareUniversity Medical Center
  • H. Bas Bueno-de-Mesquita
    • National Institute for Public Health and the Environment
  • Marga C. Ocké
    • National Institute for Public Health and the Environment
  • Anne Thiebaut
    • INSERM, ERI-20, Institut Gustave Roussy
  • Agnès Fournier
    • INSERM, ERI-20, Institut Gustave Roussy
  • Françoise Clavel-Chapelon
    • INSERM, ERI-20, Institut Gustave Roussy
  • Franco Berrino
    • Epidemiology UnitNational Cancer Institute
  • Domenico Palli
    • Molecular and Nutritional Epidemiology UnitCSPO-Scientific Institute of Tuscany
  • Rosario Tumino
    • Cancer RegistryAzienda Ospedaliera “Civile-M.P. Arezzo”
  • Salvatore Panico
    • Department of Clinical and Experimental MedicineFederico II University
  • Paolo Vineis
    • Cancer Epidemiology DepartmentUniversity of Turin
    • Imperial College
  • Antonio Agudo
    • Unit of EpidemiologyCatalan Institute of Oncology
  • Eva Ardanaz
    • Public Health Institute of Navarra
  • Carmen Martinez-Garcia
    • Andalusian School of Public Health
  • Pilar Amiano
    • Department of Public Health of Guipuzkoa
  • Carmen Navarro
    • Epidemiology DepartmentRegional Health Council
  • José R. Quirós
    • Health Information Unit, Public Health and Planning Directorate, Health and Health Services CouncilPrincipality of Asturias
  • Tim J. Key
    • Epidemiology Unit, Cancer Research UKOxford University
  • Gillian Reeves
    • Epidemiology Unit, Cancer Research UKOxford University
  • Kay-Tee Khaw
    • Department of Public Health and Primary Care, School of Clinical MedicineUniversity of Cambridge
  • Sheila Bingham
    • MRC Dunn Human Nutrition UnitWelcome Trust/MRC Building
  • Antonia Trichopoulou
    • Department of Hygiene and Epidemiology, School of MedicineUniversity of Athens
  • Dimitrios Trichopoulos
    • Department of Hygiene and Epidemiology, School of MedicineUniversity of Athens
  • Androniki Naska
    • Department of Hygiene and Epidemiology, School of MedicineUniversity of Athens
  • Gabriele Nagel
    • Division of Clinical EpidemiologyGerman Cancer Research Center
  • Jenny Chang-Claude
    • Division of Clinical EpidemiologyGerman Cancer Research Center
  • Heiner Boeing
    • Department of EpidemiologyGerman Institute of Human Nutrition Potsdam-Rehbrücke
  • Petra H. Lahmann
    • Department of EpidemiologyGerman Institute of Human Nutrition Potsdam-Rehbrücke
  • Jonas Manjer
    • Department of SurgeryMalmö University Hospital
  • Elisabet Wirfält
    • Department of Clinical SciencesLund University
  • Göran Hallmans
    • Public Health and Clinical Medicine, Nutritional ResearchUniversity Hospital of Northern Sweden
  • Ingegerd Johansson
    • Public Health and Clinical Medicine, Nutritional ResearchUniversity Hospital of Northern Sweden
  • Eiliv Lund
    • Institute of Community MedicineUniversity of Tromsø
  • Guri Skeie
    • Institute of Community MedicineUniversity of Tromsø
  • Anette Hjartåker
    • Cancer Registry of NorwayInstitute of Population-based Cancer Research
  • Pietro Ferrari
    • Nutrition and Hormones GroupInternational Agency for Research on Cancer
  • Nadia Slimani
    • Nutrition and Hormones GroupInternational Agency for Research on Cancer
  • Rudolf Kaaks
    • Nutrition and Hormones GroupInternational Agency for Research on Cancer
  • Elio Riboli
    • Department of Epidemiology & Public Health, Faculty of MedicineImperial College
Original Paper

DOI: 10.1007/s10552-006-0112-9

Cite this article as:
Tjønneland, A., Christensen, J., Olsen, A. et al. Cancer Causes Control (2007) 18: 361. doi:10.1007/s10552-006-0112-9

Abstract

Objective

Most epidemiologic studies have suggested an increased risk of breast cancer with increasing alcohol intake. Using data from 274,688 women participating in the European Prospective Investigation into Cancer and Nutrition study (EPIC), we investigated the relation between alcohol intake and the risk of breast cancer.

Methods

Incidence rate ratios (IRRs) based on Cox proportional hazard models were calculated using reported intake of alcohol, recent (at baseline) and lifetime exposure. We adjusted for known risk factors and stratified according to study center as well as potentially modifying host factors.

Results

During 6.4 years of follow up, 4,285 invasive cases of breast cancer within the age group 35–75 years were identified. For all countries together the IRR per 10 g/day higher recent alcohol intake (continuous) was 1.03 (95% confidence interval (CI): 1.01–1.05). When adjusted, no association was seen between lifetime alcohol intake and risk of breast cancer. No difference in risk was shown between users and non-users of HRT, and there was no significant interaction between alcohol intake and BMI, HRT or dietary folate.

Conclusion

This large European study supports previous findings that recent alcohol intake increases the risk of breast cancer.

Keywords

AlcoholBreast neoplasmCohort studyHormone replacement therapy

Abbreviations

BMI

Body mass index

IRR

Incidence rate ratio

CI

Confidence interval

EPIC

European Prospective Investigation into Cancer and Nutrition

HRT

Hormone replacement therapy

Pre

Premenopausal

Peri

Perimenopausal

Post

Postmenopausal

PY

Person years

Introduction

Breast cancer is the most commonly diagnosed malignancy among women in Europe [1].

Alcohol consumption may be one of the few modifiable risk factors for breast cancer. Results from epidemiologic studies published during the past two decades have generally suggested that women, who regularly consume alcohol, may be at a slightly increased risk of the disease. However, the published results are not entirely consistent. Study design as well as other lifestyle habits may influence the estimated risk.

A collaborative reanalysis of individual data from 53 epidemiologic studies, including 58,515 women with breast cancer showed the relative risk of breast cancer to be 7.1% (95% CI: 5.5–8.7%) higher per 10 g/day higher intake of alcohol (ethanol). Among these, 9,693 cases were identified from cohort studies, which may be less susceptible to recall and selection bias than case–control studies. They showed a lower, but still significant 5.0% increase in relative risk per 10 g/day intake of alcohol (95% CI: 1.6–8.4%) [2].

The length of follow-up may also influence the estimated risk. This was the case in the meta-analysis by Ellison et al. [3] with an 11% higher risk (95% CI: 6%–16%) estimated from eight cohort studies with less than ten years of follow-up compared to five studies with ten or more years of follow up. In that study it was also reported that studies conducted outside the United States showed slightly higher relative risks associated with alcohol consumption than studies conducted in the United States. Limited work has been done in examining host and lifestyle factors that may modify the relationship between alcohol consumption and breast cancer risk [4]. It has been suggested that the increased risk associated with alcohol intake may be reduced by adequate intake of folate [5]. This has been supported by a few additional studies [69], while other studies did not support this hypothesis [10, 11]. Some studies have shown an additional elevation in risk of breast cancer among women both drinking alcohol and using HRT compared to non-users of HRT [4, 1214] drinking alcohol. Levels of BMI may also modify the alcohol related risk of breast cancer.

In the previously mentioned collaborative reanalysis no difference in the risk was shown in relation to BMI [2], while a pooled analysis of cohort studies found heavy drinking to increase risk to a greater extent among obese women than among leaner women [15]. Other studies, however, suggest a higher risk among leaner women [4, 13].

Increased estrogen and androgen levels in women consuming alcohol appear to be an important mechanism for the increased risk of breast cancer. Other plausible mechanism include the effect of acetaldehyde, the induction of cytochrome P450 2E1 leading to the generation of reactive oxygen species and the interaction between alcohol and the one carbon metabolism and DNA methylation [12, 16, 17].

The purpose of this paper is to describe the associations between alcohol intake and breast cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC), a large multi-center cohort study including subjects living in countries from the north to the south of Europe and thus presenting a wide range of alcohol intakes as well as other lifestyle habits, which may potentially influence breast cancer risk. We also investigated the use of HRT, the dietary intake of folate, and BMI as possible effect modifiers.

Materials and methods

The EPIC cohort is a multi-center prospective cohort study designed to primarily investigate the relation between food, nutritional status, various lifestyle and environmental factors and cancer. The EPIC cohort consists of sub cohorts recruited in 23 centers in ten European countries: Norway, Sweden, Denmark, the United Kingdom (UK), Germany, The Netherlands, France, Spain, Italy, and Greece, allowing comparisons between regions with very different incidence rates of cancer and distribution of lifestyle and food habits.

In this study we describe data for the female participants of the EPIC cohort. Many of the participants were recruited from the general population residing in a specific geographical area, a town or a province. Exceptions were most of the Oxford cohort, UK (based on vegetarian volunteers and healthy eaters living throughout the UK), the Utrecht cohort, the Netherlands and the Florence cohort, Italy (based on women attending breast cancer screening), the French cohort (based on female members of the health insurance for state school employees living throughout France), the Ragusa cohort, Italy, and most of the Spanish cohort (based on blood donors and their spouses). Eligible subjects were invited to participate in the studies, and those who accepted gave informed consent and completed questionnaires on their diet, lifestyle, and medical history. Most study participants had anthropometric measurements taken at a study center, including height and weight. The methods have been reported in full by Riboli et al. [18].

The present study was based on data from 368,010 female participants aged 35–70 years. Women with a prevalent cancer diagnosis (19,953) and with missing data for variables considered in the analysis (67,055 subjects) as well as 6,314 who were in the lowest and highest 1% of the distribution of the ratio of reported total energy intake to energy requirement were excluded, leaving 274,688 women for analysis.

The study was approved by the International Agency for Research on Cancer ethical committee and by the local ethical committees at the participating centers.

Diet and lifestyle questionnaires

Diet was measured by country-specific questionnaires designed to capture local dietary habits as previously described [18]. Six countries administered a diet history questionnaire, which can provide data on up to 350 food items per country. In Malmö (Sweden) a modified diet history was used, combining a 168-item questionnaire with a seven-day menu book and a structured interview. Both a food frequency questionnaire and a seven-day record were adopted in UK, but the results reported in this paper for the UK are all from the food frequency questionnaire. In Spain, Greece, and in Ragusa (Italy) a dietary questionnaire very similar in content to the above, but administered by direct interview, was used. The dietary questionnaires used in the different cohorts were all validated or are currently being validated (Norway) [19]. For most centers the reference method was 12 monthly 24-h recalls. In these studies, Pearson correlation coefficients between the mean consumption of alcohol, reported in the dietary questionnaire and the average of the 24-h recalls varied for females from 0.71 (France) to 0.91 (Germany), except for Greece where it was r = 0.58 [19]. Alcohol consumption was estimated in g/day. Study subjects reported on how many standard glasses of beer and/or cider, wine, liquor, spirits, or fortified wine they consumed per day or per week during 12 months prior to recruitment for recent intake and during the ages 20, 30, 40, 50 years for past alcohol consumption. Ethanol intake was calculated based on the average glass volume and ethanol content for each type of alcohol beverage [20]. Beverage-specific estimates for wine, beer, and spirits were computed. Abstainers were defined as participants who reported no recent intake of alcohol, so this group included mostly lifetime abstainers, but also included former drinkers.

Previous alcohol intake defined as alcohol intake before baseline was estimated as the average intake in g/day at age groups 20s, 30s, and 40s, respectively and was reported in eight countries (Denmark, UK, Germany, The Netherlands, France, Spain, Italy, and Greece). Consumption of alcoholic beverages at different ages and at recruitment was the basis to calculate mean lifetime alcohol intake. For this purpose, information on duration of alcohol consumption derived from alcohol intake at different ages was combined to the amount of alcohol consumed at these different ages. Thus, mean lifetime alcohol intake was determined as a weighted average of the intake at different ages, with weights equal to the time of individual exposure to alcohol.

Folate intake was estimated from the dietary questionnaires, as described earlier [21]. We were not able to include information on intake of folic acid from food supplements.

Lifestyle questionnaires included questions on reproductive history, education, and physical activity, history of previous illness and disorders or surgical treatment, and history of consumption of tobacco.

Height and weight was measured in all EPIC centers except for France, Norway, and Oxford for which self-reported height and weight was assessed via questionnaires [18]. In the final dataset, center-specific differences in the measurement techniques were taken into account and adjustments performed for the self-reported weight via regression equations derived from measured and self-reported data.

End-points

Incident breast cancer cases were identified by population cancer registries (Norway, Sweden, Denmark, the UK, The Netherlands, Spain, and Italy) or by active follow up (Germany, France, Greece), depending on the follow-up system in each of the participating countries. The active follow up used a combination of methods including health insurance records, cancer and pathology registries, and active follow up through participants and their next-of-kin. Each cohort member was followed up for breast cancer occurrence from the date of entry into the cohort, until the date of diagnosis of cancer, date of death, emigration, or end of the follow-up period, as reported to the common database at the International Agency for Research on Cancer (IARC), Lyon.

Mortality data were also obtained from either the cancer or the mortality registries at the regional or national level. For all centers using cancer registry data, we had complete follow-up until December 1999 (Turin); December 2000 (Asturias, Murcia, Cambridge, Bilthoven); December 2001 (Florence, Varesa, Ragusa, Naples, Granada, Navarra, San Sebastian, Oxford, Malmö, Norway); June 2002 (France); November 2002 (Greece); December 2002 (Denmark); June 2003 (Utrecht); and December 2003 (Germany, Heidelberg); March 2004 (Potsdam). Countries using individually based follow up, the end of the follow up was considered to be the last known contact, the date of diagnosis or the date of death, whichever came first. In our analysis, we included the results from all centers.

The 10th Revision of the International Statistical Classification of Diseases Injuries and Causes of Death (ICD-O-2) was used, and cancer of the breast was defined as C50.0–50.9.

Statistical methods

The analyses of the breast cancer rate were based on stratified Cox proportional hazard models (including time-dependent variables), with age as the time axis to ensure that the estimation procedure was based on comparisons of individuals of the same age [22]. This allowed for complete age adjustment in order to prevent any confounding by age. Subjects were included from their age when completing the dietary questionnaire to their age at exit, defined as date of diagnosis of cancer, date of death, date of emigration, date of lost to follow up or end of follow up, whichever came first, such that women were considered under risk from the age at enrolment. The underlying hazard was stratified according to study center. Direct standardization based on the age distribution at entry in the combined EPIC cohort in ten-years intervals was performed [23]. Instead of excluding the cases occurring the first year after entry, we allowed the rate to change with time since baseline using a linear spline with a boundary at one year after entry into the cohort. Furthermore, we examined whether the associations depended on time since entry by estimating separate associations for the first period (less than 3.4 years after entry, which was the 50th percentile of follow-up time) and the later period. The associations were very similar. Recent average daily alcohol intake (defined as the reported average intake during the last year before enrolment at baseline) was analyzed both as a linear variable and in categories, in both cases with a separate risk estimated for abstainers. First, we tested if the effect of alcohol was equal among countries by estimating separate slopes for alcohol intake for each country. This model was compared with a model estimating a single, common slope for alcohol intake by using the likelihood ratio test. Assuming a common slope for all countries we then tested whether the relation between abstainers and breast cancer was equal between countries by estimating separate values for abstainers for each country. This model was compared with a model estimating a single, common value for abstainers by using the likelihood ratio test. No statistical significant difference was shown. All the models were adjusted for some known risk factors, namely as age at menarche (<12, 13–14, ≥15), parity (yes/no), current oral contraceptive use (yes/no), current use of hormone replacement therapy (HRT) at baseline (yes/no, with no including a small percentage of previous users), menopausal status (pre-, peri-, post-menopausal), smoking status (current, former, never), education (none, primary school, technical/professional school, secondary school, university), height and weight (linear) for pre-/perimenopausal/postmenopausal.

Information on physical activity was available for only 241,649 women. Adjustment for physical activity did not change the risk estimates, where this information was available. We therefore chose, in order not to exclude too many women, not to include physical activity in the final model. Information on family history of breast cancer and history on breast biopsy was not available in the database.

We investigated possible effect modification of the relationship between alcohol intake and breast cancer by use of HRT (yes/no), BMI (<18.5/18.5–25/25–30/>30 kg/m2), and dietary folate (≤200/200–300/300–400/400+ mcg/day). As a test for interaction, we assessed whether the slopes for total alcohol intake at each level of the potential effect modifier differed significantly using the likelihood test statistics calculated from Cox’s partial likelihood.

All quantitative variables were considered as continuous variables in the model, because this is biologically more reasonable than the step functions corresponding to categorization and furthermore increases the power of the analyses [24]. The linearity of the associations was evaluated graphically by linear splines with three boundaries placed at the quartile cut points according to exposure distribution among the cases [25].

None of the associations showed signs of inflection or threshold values; therefore, all quantitative variables were entered linearly in the model. The PHREG procedure in SAS release 9.1 (SAS Institute, Inc., Cary, North Carolina, USA) was used for the statistical analyses.

Results

The cohort members (274,688) accrued a total of 1,695,876 person years (PY) of follow up, during which time 5,054 cases (1,114 pre-, 969 peri-, and 2,962 post-menopausal) were diagnosed.

Of these, 4,285 were invasive tumors and 433 in situ tumors. No information on morphology characteristics was available for the remaining 336 breast cancers. Only verified invasive cases were retained in the analysis. The median length of follow-up was 6.4 years (1st to 99th percentiles: 1.8–9.4).

Table 1 shows the number of breast cancers included in the analyses, according to country, age at baseline and attained age, together with the corresponding number of cohort members, person years and crude/age-standardized incidence rates. The age-standardized incidence rate was highest in Sweden (425 per 100,000 PY), while the lowest rates were seen in Greece (87 per 100,000 PY) and Spain (147 per 100,000 PY).
Table 1

Description with number of cases and incidence rates of invasive breast cancer in the EPIC cohort

 

Breast cancer cases

Cohort numbers

Person years

Incidence rate per 100,000 PY

Pre

Peri

Post

Total

Crude

Age-standardizeda

Country

Norway

42

75

62

179

31,156

95,482

187

181

Sweden

5

80

220

305

12,732

93,077

328

425

Denmark

46

108

509

663

26,985

180,094

368

318

UK

128

70

275

473

35,977

195,070

242

257

Germany

74

41

174

289

27,671

161,148

179

202

Netherlands

65

70

282

417

22,499

153,728

271

253

France

286

255

706

1,247

52,438

438,616

284

274

Spain

97

16

85

198

22,379

147,328

134

147

Italy

164

85

224

473

30,010

183,951

257

269

Greece

12

2

27

41

12,841

47,383

87

87

Age at baseline

35–44

396

12

6

414

57,981

324,377

128

 

45–54

518

736

678

1,932

119,018

732,418

264

 

55–64

5

54

1,585

1,644

79,986

530,557

310

 

>65

0

0

295

295

17,703

108,524

272

 

Attained age

35–39

24

0

0

24

 

49,779

48

 

40–49

497

108

23

628

 

384,425

163

 

50–59

392

656

918

1,966

 

710,042

281

 

60–69

6

38

1,387

1,431

 

464,035

308

 

70–79

0

0

236

236

 

87,384

270

 

>80

0

0

0

0

 

211

0

 

Total

919

802

2,564

4,285

274,688

1,695,876

253

 

a According to age at baseline

Due to the relatively low number of cases and short follow-up time in Norway and Greece, results for these countries should be interpreted with caution.

The reported intake of alcohol is presented in Table 2. The recent median baseline consumption in the total cohort among drinkers was 5.7 g/day, with the lowest intake reported among women in Greece of 2.2 g/day and the highest intake among women in Denmark of 10.2 g/day. The percentage of abstainers among cases varied from 2% in Denmark to 49% in Spain.
Table 2

Percentage of drinkers and intake of alcohol among drinkers in different exposure periods for the cases and the total EPIC cohort

Country

Median intake (5–95 percentiles)

20s

30s

Abstainers (%)

Total alcohola,b

Abstainers (%)

Total alcohola,b

Cases

Total cohort

Cases

Total cohort

Cases

Total cohort

Cases

Total cohort

Norway

Sweden

Denmark

29

32

4.23 (1.35–16.96)

4.23 (1.35–16.90)

16

20

6.90 (2.11–21.22)

6.34 (1.49–24.23)

UK

48

44

6.34 (1.17–27.38)

7.20 (1.41–33.85)

36

36

7.23 (1.41–37.12)

7.96 (2.13–36.58)

Germany

22

19

2.09 (0.23–12.91)

2.03 (0.24–14.83)

7

6

3.71 (0.46–21.30)

3.56 (0.46–20.53)

Netherlands

48

44

3.26 (1.48–16.90)

3.26 (1.48–14.81)

France

71

73

3.95 (0.85–24.99)

4.19 (0.68–21.67)

41

43

7.09 (1.25–27.91)

7.53 (1.25–31.18)

Spain

55

53

8.98 (1.28–29.08)

8.99 (0.79–29.08)

52

51

8.99 (1.28–26.97)

8.99 (0.79–27.96)

Italy

39

37

5.60 (0.56–19.12)

2.74 (0.56–18.92)

23

22

6.48 (0.62–30.49)

5.91 (0.62–23.44)

Greece

87

75

7.32 (2.60–17.25)

9.09 (1.30–27.26)

69

62

7.39 (1.71–17.25)

7.79 (1.30–27.26)

All

50

49

4.23 (0.64–20.30)

4.23 (0.62–22.91)

28

31

6.48 (1.10–26.72)

6.35 (0.93–28.75)

Country

Median intake (5–95 percentiles)

40s

Recentc

Abstainers (%)

Total alcohola,b

Abstainers (%)

Total alcohol (g/day)b

Cases

Total cohort

Cases

Total cohort

Cases

Total cohort

Cases

Total cohort

Norway

18

20

3.10 (0.40–12.30)

2.38 (0.40–11.83)

Sweden

21

20

7.66 (1.13–28.86)

7.96 (1.13–25.94)

Denmark

11

14

9.92 (2.11–34.32)

8.45 (2.11–33.26)

2

3

11.98 (1.18–45.64)

10.22 (0.74–45.33)

UK

8

7

4.47 (0.36–31.49)

5.30 (0.36–30.61)

Germany

4

5

4.72 (0.65–26.37)

4.41 (0.62–25.06)

6

4

6.40 (0.36–34.35)

5.54 (0.37–34.03)

Netherlands

19

16

8.53 (1.78–36.82)

6.67 (1.78–31.91)

19

16

7.03 (0.16–36.71)

5.73 (0.20–35.18)

France

33

36

8.47 (1.25–29.24)

8.75 (1.25–34.28)

12

14

8.35 (0.52–38.66)

7.66 (0.45–40.05)

Spain

56

53

6.62 (0.79–22.30)

7.62 (0.79–26.97)

49

51

5.33 (0.18–26.57)

5.38 (0.17–32.22)

Italy

23

22

6.48 (0.62–30.49)

5.91 (0.62–30.49)

23

22

8.28 (0.18–36.93)

5.70 (0.19–35.24)

Greece

69

60

5.78 (1.30–16.27)

7.08 (1.30–27.26)

39

37

1.92 (0.63–14.21)

2.24 (0.51–17.61)

All

21

23

8.18 (1.19–31.06)

6.91 (0.93–30.80)

14

17

7.25 (0.36–37.82)

5.71 (0.36–34.77)

Country

Median intake (5–95 percentiles)

Lifetimec

    

Abstainers (%)

Total alcohola,b

    

Cases

Total cohort

Cases

Total cohort

    

Norway

        

Sweden

        

Denmark

1

1

8.07 (0.52–27.03)

7.16 (0.15–25.77)

    

UK

1

2

4.46 (0.12–21.62)

4.92 (0.12–25.33)

    

Germany

2

1

4.76 (0.51–20.54)

4.38 (0.52–20.93)

    

Netherlands

10

7

5.87 (0.24–26.76)

5.33 (0.85–23.48)

    

France

9

11

4.71 (0.12–22.14)

4.46 (0.11–23.45)

    

Spain

30

34

4.29 (0.05–22.30)

4.46 (0.06–23.38)

    

Italy

12

10

4.97 (0.05–23.01)

4.15 (0.05–21.04)

    

Greece

39

30

1.82 (0.08–9.60)

2.17 (0.09–17.93)

    

All

7

8

5.37 (0.13–23.77)

4.85 (0.13–23.46)

    

a Among those who drank in the period

b Grams ethanol per day

c Weighted average of the intakes at different ages with weights equal to the total subject-specific time of investigation

Information on alcohol intake in previous lifetime periods was not available for all countries. The percentage of participants with information on previous alcohol intake among cases varied from 86% in Italy to 100% in Denmark. For the total cohort it varied from 75% in Italy (where only four out of five study centers collected this information) to 100% in Denmark.

For most countries where information about previous alcohol intake was available, the percentage of reported abstainers decreased with increasing age. When restricting the analysis to women above 50 years at baseline, it was shown, taking all countries together, that a slightly higher intake were reported for women in their 40s compared to recent intake (results not shown).

Figure 1 shows the adjusted IRRs for breast cancer in relation to recent alcohol intake (g/day) in individual countries, and for all ten countries together. Unadjusted values were very close to adjusted values. The difference between the country-specific associations was not larger than could be expected by pure random (p = 0.30). For all countries together the IRR was 1.03 (95% CI: 1.01–1.05) for a 10 g/day increase in alcohol intake. In Table 3 the association of alcohol intake and invasive breast cancer in categories based on quintiles is shown. Women with a recent daily intake above 19 g ethanol per day had an adjusted IRR = 1.13 (95% CI: 1.01–1.25) compared with women with a very low intake of alcohol. To further clarify the association within the highest quintile of intake, quartiles within this group are presented using the same comparison group (Table 3). We examined the relationship between intake of specific types of alcoholic beverages and breast cancer. The beverage-specific intakes of alcohol from wine (IRR: = 1.02; 95% CI: 0.99–1.05), beer (IRR = 1.05; 95% CI: 0.98–1.12), and spirits (IRR = 1.09; 95% CI: 0.99–1.21) were all adversely associated with breast cancer, although not significant, and the associations were not significantly different (p = 0.39 for heterogeneity).
https://static-content.springer.com/image/art%3A10.1007%2Fs10552-006-0112-9/MediaObjects/10552_2006_19R1_f1.jpg
Fig. 1

IRRs (95% confidence intervals) of invasive breast cancer (n = 4,285) according to alcohol baseline average daily intake corresponding to 10 g/day increase in the EPIC cohort. aAdjusted for height (linear), weight (linear), age at menarche (≤12, 13–14, ≥15), parity (yes/no), current oral contraceptive use (yes/no), current use of hormone replacement therapy (yes/no), menopausal status (pre-, peri-, post-menopausal), smoking status (current, former, never), and education (none, primary school, technical/professional school, secondary school, university)

Table 3

IRRs (95% confidence intervals) of invasive breast cancer according to average daily alcohol intake in the EPIC cohort

Alcohol (g/day)

       

IRR’s (95% confidence intervals) of invasive cancer in quartiles for women drinking more than 19 g/day with women drinking >0–1.5 g/day as the reference group

 

Abstainers

>0–1.5

>1.5–4.7

>4.7–10

>10–19

>19 in total

>19–23.6

23.6–29.9

29.9–37.1

>37.1

Cases/ cohort

612/46,939

701/50,979

723/51,087

731/48,585

759/40,931

765/36,167

211/10,724

154/8,156

194/7,795

206/9,492

IRRa

0.99 (0.89–1.11)

1.00

0.99 (0.90–1.10)

0.99 (0.89–1.10)

1.10 (0.99–1.22)

1.18 (1.06–1.31)

1.11 (0.95–1.29)

1.08 (0.90–1.28)

1.43 (1.21–1.68)

1.14 (0.97–1.34)

IRRb

1.01 (0.91–1.13)

1.00

0.98 (0.89–1.09)

0.97 (0.88–1.08)

1.07 (0.96–1.19)

1.13 (1.01–1.25)

1.08 (0.92–1.26)

1.03 (0.86–1.23)

1.36 (1.15–1.60)

1.09 (0.93–1.28)

a Crude

b Adjusted for height (linear), weight (linear), age at menarche (≤12, 13–14, ≥15), parity (yes/no), current oral contraceptive use (yes/no), current use of hormone replacement therapy (yes/no), menopausal status (pre-, peri-, post-menopausal), smoking status (current, former, never) and education (none, primary school, technical/professional school, secondary school, university)

Figure 2 shows the IRR for the reported alcohol intake (g/day) during the 20s, 30s, and 40s mutually adjusted and adjusted for known confounders as well as recent alcohol intake for all countries together. Most values were very close to one. Lifetime intake (data for Denmark, Germany, France, Spain, and Greece only) was adjusted only for known confounders and showed a weak non-significant positive association, IRR = 1.02 (95% CI: 0.99–1.06). Furthermore, in a sub-analysis we restricted to women who had a daily intake of alcohol between 10 and 30 g/day at recruitment, in order to test whether previous consumption added to the current risk induced by current consumption. IRRs for the 20s, 30s, and 40s were not statistically different and very close to the results for the total time period (results not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs10552-006-0112-9/MediaObjects/10552_2006_19R1_f2.gif
Fig. 2

IRR (95% confidence intervals) of invasive breast cancer according to previous and lifetime alcohol intake corresponding to 10 g/day increase in Denmark, Germany, France, Spain, Italy, and Greece. aAdjusted for height (linear), weight (linear), age at menarche (≤12, 13–14, ≥15), parity (yes/no), current oral contraceptive use (yes/no), current use of hormone replacement therapy (yes/no), menopausal status (pre-, peri-, post-menopausal), smoking status (current, former, never), and education (none, primary school, technical/professional school, secondary school, university). Mutually adjusted within the age groups: 20s, 30s, and 40s. bAdjusted for recent alcohol intake. cOnly for Denmark, Germany, France, Spain, Italy, and Greece. dLifetime intake was not mutually adjusted for intake in earlier age groups or recent alcohol intake

We examined the association between alcohol intake and breast cancer risk by level of dietary folate intake, because earlier studies have shown that adequate folate intake may diminish the effect of alcohol intake (Table 4). In this study no significant interaction was shown, neither at the country-level (results not shown) nor for the total EPIC-cohort (p = 0.59).
Table 4

IRRs (95% confidence intervals) of invasive breast cancer per each additional 10 g/day average daily alcohol intake according to categories for average dietary intake of folate in the EPIC cohort

Dietary folate intake (mcg/day)

 

≤200

>200–≤300

>300–≤400

>400

p for interaction

Cases/cohort

711/53,264

876/49,184

944/49,184

957/43,506

 

IRRa

1.02 (0.97–1.08)

1.06 (1.02–1.12)

1.05 (1.00–1.10)

1.02 (0.98–1.07)

0.58

IRRb

1.01 (0.96–1.07)

1.05 (1.01–1.11)

1.04 (1.00–1.09)

1.02 (0.97–1.06)

0.59

a Crude

b Adjusted for height (linear), weight (linear), age at menarche (≤12, 13–14, ≥15), parity (yes/no), current oral contraceptive use (yes/no), current use of hormone replacement therapy (yes/no), menopausal status (pre-, peri-, post-menopausal), smoking status (current, former, never), and education (none, primary school, technical/professional school, secondary school, university), food supplements at baseline (yes/no)

When the effect of alcohol was analyzed according to levels of BMI (Table 5), some limited evidence of a higher risk among lean or normal weight women were shown, although the interaction was not significant.
Table 5

IRR (95% confidence intervals) for the association between breast cancer and recent alcohol intake per average intake of 10 g /day by level of BMI and by current use of hormone replacement therapy (HRT) in the EPIC cohort, stratified according to menopausal status

 

Cases/cohort

 

IRR (95% CI) a

p-valuec

IRR (95% CI)b

p-valuec

Premenopausal

BMI

14/1,514

<18.5

1.07 (0.81–1.42)

0.80

1.05 (0.78–1.40)

0.93

533/46,848

18.5–25

1.05 (0.99–1.12)

1.04 (0.98–1.11)

 

178/17,668

25–30

1.04 (0.95–1.14)

1.04 (0.95–1.14)

 

51/6,302

>30

0.96 (0.79–1.15)

0.97 (0.81–1.17)

 

All

  

1.04 (0.99–1.10)

1.04 (0.98–1.09)

 

Perimenopausal

BMI

8/694

<18.5

0.99 (0.64–1.54)

0.49

0.96 (0.61–1.50)

0.61

442/25,905

18.5–25

1.02 (0.95–1.09)

1.00 (0.94–1.07)

 

168/12,198

25–30

0.93 (0.84–1.04)

0.93 (0.83–1.03)

 

77/4,475

>30

0.97 (0.82–1.14)

0.99 (0.83–1.17)

 

All

  

0.99 (0.94–1.05)

0.98 (0.93–1.04)

 

Postmenopausald,e

BMI

33/1,707

<18.5

1.13 (0.94–1.36)

0.76

1.13 (0.94–1.36)

0.68

188/58,182

18.5–25

1.06 (1.02–1.10)

1.05 (1.01–1.09)

 

723/36,949

25–30

1.04 (0.98–1.09)

1.02 (0.97–1.08)

 

258/15,307

>30

1.06 (0.97–1.15)

1.04 (0.96–1.30)

 

All

  

1.05 (1.02–1.08)

1.04 (1.01–1.07)

 

Postmenopausald

HRT

1,229/75,685

No

1.05 (1.01–1.10)

0.48

1.06 (1.01–1.10)

0.49

973/36,460

Yes

1.03 (0.99–1.08)

1.03 (0.99–1.08)

 

a Crude

b Adjusted for height (linear), weight (linear), age at menarche (≤12, 13–14, ≥15), parity (yes/no), current oral contraceptive use (yes/no), smoking status (current, former, never) and education (none, primary school, technical/professional school, secondary school, university)

cp-value for interaction

d Postmenopausal women include 2,618 cases among 139,316 cohort members

eP-value for interaction of menopausal status, p = 0.21

In most countries, the intake of alcohol was slightly higher among current users of HRT (results not shown). The association between alcohol intake and the risk of breast cancer did, however, not differ significantly between current users of HRT and previous/never users of HRT (p = 0.49) (Table 5).

Discussion

In this large prospective study including women from all over Western Europe with a large variation in alcohol drinking habits, we found a significant adverse association between recent average daily alcohol intake and breast cancer risk. The study supports previous evidence of a modest relationship between alcohol intake and the risk of breast cancer. Most countries showed either an adverse association between alcohol intake and breast cancer or no association. The association varied in the different cohorts included but this apparent variation was not statistically significant when testing of heterogeneity and could be due to random variation. There was no significant difference between the associations for the different types of beverage. Alcohol intake in previous lifetime periods did not show any association with breast cancer risk when adjusted for recent alcohol intake. Our results on the association between alcohol and breast cancer do not appear to be materially confounded by the effect of the other risk factors available for the total EPIC-cohort, since adjustment for possible confounders did not substantially alter the estimated IRR. We found no strong signs of effect modification by the level of dietary intake of folate, BMI or the current use of HRT.

The major strength of this large multi-center study is the size with 4,285 incident cases of invasive breast cancer and the large variation in alcohol intake, caused by inclusion of subjects living in countries from all over Europe. Median recent daily alcohol intake among alcohol drinkers varied from 2.2 g/day in Greece to 10.2 g/day in Denmark. In most countries the main contributor to average daily alcohol intake was wine except for one Spanish center (Murcia) where it was beer [26]. The differences between beverage-specific associations were insignificant supporting previously published Danish data that did not show any statistical difference in the beverage-specific estimates [27], and in accordance with the literature where no convincing evidence has been shown for a different effect on breast cancer risk for the different types of beverage [2, 12, 15].

Information on alcohol intake was self-reported. If self-reported information were underestimating the true consumption, this would result in an overestimation of the relative risk of breast cancer for a given difference in alcohol consumption. By contrast, random misclassification among both cases and non-cases may have the opposite effect, possibly resulting in an underestimation of the relative risk of breast cancer. However, the information of alcohol consumption was demonstrated to be valid and reliable in validation studies [19] conducted in the included country-specific studies.

The overall results of the current study are in agreement with the large reanalysis from the collaborative Group on Hormonal Factors in Breast Cancer [2], where a small, but significant 5% increase in risk was shown per 10 g/day intake of alcohol among women in ten pooled cohort studies.

In our study, the intake of alcohol in previous lifetime periods did not influence the current risk of breast cancer when adjusted for average intake at baseline in the EPIC cohorts. The importance of recent average alcohol intake was supported by other prospective studies by Holmberg et al. [28] and Horn-Ross [4] in which drinking later in life appeared to have greater effect than drinking earlier in life. In contrast, a recent case–control study showed a stronger association with lifetime alcohol intake than current alcohol intake [29].

It has been proposed that alcohol intake increases endogenous levels of estrogen. Reichman et al. [30] showed that among premenopausal women in a controlled diet study, alcohol consumption was associated with statistically significant increases in levels of several hormones and an elevated absolute amount of bioavailable estrodiol. In an intervention study, postmenopausal women who consumed 15 or 30 g of alcohol per day for eight weeks had levels of estrone sulfate concentrations increased by 7.5% (95% CI: 0.3–15.9%) and 10.7% (95% CI: 2.7–19.3%), respectively, compared with levels when the women consumed placebo. DHEAS concentration also increased by 5.1% (95% CI: 1.4–9.0%) and 7.5% (95% CI: 3.7–11.5%), respectively, compared with placebo levels [31]. Two other randomized trials have been published [32, 33] supporting an increase in circulating estradiol among women using HRT after alcohol ingestion. In a recent observational study, women who were non-users of HRT and consumed more than 25 g of alcohol per day had higher levels of estrone, estradiol and DHEAS compared to non-drinkers [34]. New data from the EPIC study among 790 pre- and 1291 postmenopausal women, comparing women drinking more than 25 g/day with non-consumers, showed higher levels of DHEAS, testosterone, androstenedione, and estrone among alcohol consumers [35], but no increase in estradiol was shown.

In conclusion, these studies do provide evidence, although not entirely consistent, that current alcohol intake may increase sex steroid levels in women, and thereby increase the risk of breast cancer.

In this large prospective study we found limited evidence for effect modification with regard to BMI-level, with slightly higher risk among lean and normal weight women. Earlier studies have been inconsistent, with some studies supporting our findings [4, 13], others showed increased risk among obese women [15], and yet others suggested no difference in the risk according to BMI [36]; but for most studies like in our study the interaction was insignificant.

Some studies have suggested that use of HRT and ingestion of alcohol may synergistically enhance the risk [4, 13, 14]. Horn-Ross et al. [4] found that among users of combination HRT, risk was increased for alcohol consumers of ≥20 g/day RR = 1.51 (95% CI: 1.13–2.03) (relative to non-drinkers) but not among women who never used HRT, RR = 0.98 (95% CI: 0.55–1.73). In the present study, we did not find any systematic signs of a synergistic effect between alcohol intake and current use of HRT, since the overall estimated association with average alcohol intake was almost similar for current users and non-users of HRT. Our results on the total EPIC cohort are in agreement with the recent pooled analysis of both cohort and case–control studies [2] in supporting the hypothesis that the breast cancer risk associated with the combination of heavy drinking and HRT is not multiplicative. In a recent Swedish cohort, a statistically significant interaction between alcohol and HRT was restricted only to estrogen receptor positive (ER+) tumors [36]. Unfortunately, we did not have information on ER status in the common EPIC database.

Other mechanisms might also be important for the increased risk of breast cancer. It has been suggested that women with low intake of folate might have the highest alcohol associated risk of breast cancer [59]. Folate plays a major role in the formation of S-adenosylmethionine, an important methyl group donor. Folate depletion by low intake in association with high alcohol intake may lead to DNA hypomethylation, disruption of DNA integrity and DNA repair. These aberrations in DNA may enhance carcinogenesis by altering the expression of tumor suppressor genes and proto-oncogene [37]. No clear interaction was shown between alcohol consumption and the intake of dietary folate in the present study. A weakness of our study was that we were not able to include the intake of folic acid from food supplements, but only adjusted for the overall general use of food supplements.

Acetaldehyde, rather than alcohol itself, is responsible for the carcinogenic or the co-carcinogenic effect of alcohol [17]. Acetaldehyde is highly mutagenic and carcinogenic, and interferes at many sites with DNA synthesis and repair and consequently tumor development [38], acetaldehyde induces inflammation and metaplasia in the mammary gland [39], and may thereby influence breast cancer risk.

The duration of follow up in EPIC is still relatively short. Therefore, the information on alcohol intake given by the participants at baseline is likely to apply to their behavior during follow up. In the meta-analysis of Ellison et al. [3] the pooled RR per 12 g alcohol per day were 11% higher when cohort studies with fewer than ten years of follow up (pooled RR, 1.15) were compared to studies with longer follow up (pooled RR, 1.04). Based on those findings, a stronger association between alcohol intake and breast cancer might have been expected in this cohort. However, we did not detect any significant difference in the association between recent alcohol intake and breast cancer when estimating separate associations for less than 3.4 years of follow up and the later period of follow up (results not shown).

In conclusion, this large European study supports previous findings that recent average alcohol intake, irrespectively of beverage type, increases the risk of breast cancer.

More studies are needed to determine whether the adverse association between alcohol intake and breast cancer is modifiable by other lifestyle and host factors. Likewise the specific effect of alcohol in relation to ER and PR-receptor expression should be investigated.

Acknowledgments

We would like to thank all the participants in the EPIC study. The authors acknowledge the important technical support given by Programmer Katja Boll and secretarial help by Jytte Fogh Larsen. Financial support: Grant sponsor: Europe Against Cancer programme of the European Commission (SANCO); Grant sponsor: Deutsche Krebshilfe; Grant sponsor: German Cancer Research Center; Grant sponsor: German Federal Ministry of Education and Research; Grant Sponsor: Danish Cancer Society; Grant sponsor: Health Research Fund of the Spanish Ministry of Health; Grant sponsor: Spanish regional governments of Andalusia, Asturia, Basque County, Murcia and Navarra; Grant sponsor: Cancer Research UK; Grant sponsor: Medical Research Council, UK; Grant sponsor: Stroke Association, UK; Grant sponsor: British Heart Foundation; Grant sponsor: Department off Health, UK; Grant sponsor: Food Standards Agency, UK; Grant sponsor: Wellcome Trust, UK; Greek Ministry of Health; Grant sponsor: Greek Ministry of Education; Grant sponsor: Italian Association for Research on Cancer; Grant sponsor: Italian National Research Council; Grant sponsor: Dutch Ministry of Public Health, Welfare and Sports; Grant sponsor: National Cancer Registry and regional cancer registries of Amsterdam, East and Maastricht, the Netherlands; Grant sponsor: World Cancer Research Fund; Grant sponsor: Swedish Cancer Society; Grant sponsor: Swedish Scientific Council; Grant sponsor: Regional Government of Skåne, Sweden; Grant sponsor: The Norwegian Cancer Society, The Norwegian Research Counsil; Grant sponsor: French League Against Cancer, 3 M Company, Mutuelle Générale de l’Education Nationale. French Institute of Health and Medical Research (INSERM), Gustave Roussey Institute, “Fondation de France”, several departmental councils in France; Grant sponsor: the ISC III Network RCESP (C03/09), Spain.

Copyright information

© Springer Science+Business Media B.V. 2007