European Journal of Clinical Nutrition (2011) 65, 920–928
& 2011 Macmillan Publishers Limited All rights reserved 0954-3007/11 www.nature.com/ejcn
ORIGINAL ARTICLE
Does the Mediterranean dietary pattern or the Healthy Diet Index influence the risk of breast cancer in a large British cohort of women? JE Cade1, EF Taylor1, VJ Burley1 and DC Greenwood2 1 Nutritional Epidemiology Group, School of Food Science and Nutrition, University of Leeds, Leeds, UK and 2Division of Biostatistics, Leeds Institute for Genetics, Health and Therapeutics, University of Leeds, Leeds, UK
Background/Objective: To assess the risk of developing breast cancer associated with consumption of two common dietary patterns: a Mediterranean dietary pattern and a dietary pattern, which conforms to the World Health Organization Healthy Diet Index (WHO HDI). Subjects/Methods: Dietary data from a 217-item food frequency questionnaire were used to generate two dietary patterns according to pre-defined criteria in women from the UK Women’s Cohort Study. Survival analysis using Cox regression was used to estimate hazards ratios for risk of breast cancer adjusted for known confounders. Results: This analysis included 828 incident cases of breast cancer in 33 731 women with a mean follow-up of 9 years. There were no statistically significant associations between either the Mediterranean dietary pattern or the WHO HDI and risk of breast cancer. In premenopausal women, there was a nonsignificant trend suggesting that increasing compliance with the Mediterranean diet was associated with lower risk of breast cancer. Maximal adherence to the Mediterranean diet was associated with hazards ratio ¼ 0.65 (95% confidence interval: 0.42–1.02, P trend ¼ 0.09) compared with minimal adherence. In postmenopausal women, no clear trends were observed. Conclusions: In this study, no strong association between the risk of breast cancer and the consumption of either a Mediterranean-type diet or one characterized by adherence to the WHO HDI was observed. In premenopausal, but not postmenopausal women, there was a nonsignificant inverse association with increasing adherence to the Mediterranean diet pattern.
European Journal of Clinical Nutrition (2011) 65, 920–928; doi:10.1038/ejcn.2011.69; published online 18 May 2011 Keywords: dietary pattern; breast cancer; cohort; Mediterranean
Introduction Exploration of dietary patterns may be a helpful addition to the traditional approach to studying diet and breast cancer, which has focused on single foods and nutrients (Hu, 2002). Studies linking single foods or nutrients to breast cancer risk have generated conflicting results (Gandini et al., 2000; Smith-Warner et al., 2001). Studies of single components cannot account for highly correlated food items or the
Correspondence: Professor JE Cade, Nutritional Epidemiology Group, School of Food Science and Nutrition, University of Leeds, Leeds LS2 9JT, UK. E-mail:
[email protected] Received 1 March 2010; revised 17 February 2011; accepted 18 February 2011; published online 18 May 2011
synergistic effects of food combinations; or there may be other non-nutrient phytochemicals involved (Velie et al., 2005). Major dietary patterns, including a Mediterraneantype diet, have been identified as predictors of other chronic diseases such as acute myocardial infarction (MartinezGonzalez et al., 2002), colorectal cancer (Fung et al., 2003) and all-cause mortality in women (Kant et al., 2000). In terms of breast cancer risk, dietary patterns have been less well studied than single foods or nutrients. The analytical approach that has been commonly used is factor analysis, which is population specific. In one large American study, using that approach, they failed to show an effect on risk of breast cancer with a dietary pattern characterized by high intakes of vegetables, fish, poultry and fruit (Velie et al., 2005). Another study of an Italian cohort showed that a
Dietary patterns and risk of breast cancer JE Cade et al
921 dietary pattern characterized by high intake of salad vegetables was protective against breast cancer (Sieri et al., 2004). An alternative to patterns, which have been statistically generated, would be to apply predefined criteria to describe the dietary patterns. The aim of this analysis was to explore associations between risk of breast cancer and two specific dietary patterns. One based on the degree of ‘healthiness’ of the diet as determined by extent of adherence to the World Health Organization Healthy Diet Index (WHO HDI) and the second based on the degree of compliance with components of the Mediterranean diet. The UK Women’s Cohort Study (UKWCS) was designed to include a wide range of different dietary intakes. In particular, the wide range of vegetables and fish consumed within the cohort means that we have plenty of people with higher intakes consistent with aspects of the Mediterranean diet in order to explore the impact of this dietary pattern on risk of breast cancer.
Subjects and methods Subjects The UKWCS recruited 35 372 women aged 35–69 years between 1995 and 1998. Ethical approval was obtained from 174 local research ethics committees. The cohort was selected from about 500 000 responders to a brief postal survey in order to obtain a cohort with a wide range of dietary patterns represented. All of the non-red meat eaters and vegetarians from that survey who were women aged 35–69 at survey completion were included plus a comparison group of the remaining eligible women. This was achieved by selecting, for each vegetarian, the next non-vegetarian in the survey aged within 10 years of the vegetarian. More details of the process are described elsewhere (Cade et al., 2007; Taylor et al., 2007). At baseline, the cohort had mean age of 52 years with mean body mass index 24.5 kg/m2 and 11% were current smokers. The cohort lived in all the areas of the United Kingdom with a predominance (33%) living in the south and east. The cohort was recruited from respondents to a direct mail survey from the World Cancer Research Fund with around half a million responders. Selection into the cohort was according to responses to a series of simple questions on diet, all responders who stated that they were either vegetarian or only ate fish were included plus a sample matched within 5-year age bands of meat eaters resulting in a cohort with wide range of dietary patterns with similar large numbers of women consuming a vegetarian-type diet, women who ate fish but not meat and meat eaters. This classification was only used to select the initial cohort. Subsequent analysis is dependent on response to more detailed questionnaires administered in this study. Having a cohort with a range of dietary patterns ensures adequate power while minimizing the effects of measurement error because of the the wide range of exposures included (White et al., 1994; Kaaks and Riboli, 1997; Schatzkin et al., 2001).
The study was registered with the National Health Service Central Register so that all incident cancer cases and deaths were notified to the study team. Incident cancers and cause of death were coded according to the International Classification of Diseases 9 and 10. The investigation censor date was 1 April 2006, with a mean follow-up of 9 years and 828 incident breast cancers recorded. A detailed diet and lifestyle questionnaire was sent to participants at baseline. Using this information, 15 952 women were classified as being premenopausal and 17 779 postmenopausal. Menopausal status was coded using specific criteria related to menstrual and obstetric history and age. If women were currently using hormone replacement therapy or had had a hysterectomy and were aged over 50 years at baseline then they were coded as postmenopausal. Women p50 years using hormone replacement therapy or having had a hysterectomy were coded as premenopausal. Women were excluded if they had extremely high (46000 kcal/day) or low (o500 kcal/day) energy intake (n ¼ 87), as were women with prevalent breast cancer (n ¼ 860).
Dietary patterns Diet was assessed at baseline using a 219-item self-administered food frequency questionnaire (FFQ). Two scores using predefined criteria, one representing a Mediterranean diet score and the other the WHO HDI, were generated. A description of the levels of foods and nutrients included in the two scores generated is presented in Table 1. Mediterranean diet score. A score indicating compliance with a traditional Mediterranean diet was based on the definition described by Trichopoulou et al. (2003), modified with respect to the lipid ratio and consideration of meat and poultry as separate categories and generated a 0–10 score. Out of 10 components selected as being significant indicators of the Mediterranean Diet, each woman from the cohort was assigned a value of 0 or 1 for 9 of these components, using the cohort median as a cutoff. Intakes above the cutoff for vegetables, legumes, fruit and nuts, cereal, fish and ratio of monounsaturated fatty acids to saturated fatty acids each increased the score by 1. Below the cutoff intake of meat, poultry and dairy products also increased the score by 1. For the 10th component, alcohol, women consuming between 5 and 25 g of alcohol per day increased their score by 1. This generated a score ranging from 0 to 10. Owing to the small numbers in the extremities of the scale, women scoring 0, 1 or 2 were combined and women scoring 7–10 were combined. World Health Organization HDI. WHO guidelines were used as a basis for developing a healthy diet indicator (WHO Technical Report Series, 2003). Use of the entire set of guidelines was not possible, as comparable dietary data for some dietary factors listed were not available from our FFQ. Dietary data were available for total fatty acids, saturated fatty acids and polyunsaturated fatty acids but n-6 European Journal of Clinical Nutrition
Dietary patterns and risk of breast cancer JE Cade et al
922 Table 1 Derivation of the dietary patterns used: Mediterranean diet score and World Health Organization Healthy Diet Index (WHO HDI) Indicator value
Mediterranean diet score Vegetables (g/day) Legumes (g/day) Fruit and nuts (g/day) Cereals (g/day) Fish (g/day) Polyunsaturated:saturated fatty acid ratio Meat (g/day) Poultry (g/day) Dairy (g/day) Alcohol (g/day) WHO HDI Total fat (% total E) Saturated fatty acids (% total E) Polyunsaturated fatty acids (% total E) Total carbohydrate (% total E) Non-milk extrinsic sugars (% total E) Non-starch polysaccharides (g) Fruit and vegetables (g) Protein (% total E) Cholesterol (g) Salt (g)
% Cohort meeting guideline
1
0
X281 X31 X271 X226 X24 X0.96
o281 o31 o271 o226 o24 o0.96
50 50 50 50 50 50
o40 o13 o97 5–25
X40 X13 X97 o5 or 425
50 50 50 50
15–30 0–10
o15 or 430 410
29 34
6–10
o6 or 410
51
55–75 0–10
o55 or 475 410
26 27
420 X400 10–15 o300 o5
o20 o400 o10 or 415 X300 X5
67 75 39 74 12
polyunsaturated fatty acids, n-3 polyunsaturated fatty acids and trans fatty acids could not be derived accurately from the FFQ without further validation. Monounsaturated fatty acids were available, however the corresponding recommendation from the WHO document is a difference between the other fats hence this component was not used as part of the score. Total carbohydrates, non-starch polysaccharides, fruit and vegetables consumption, protein, cholesterol and salt in food were readily available from the FFQ. We were able to estimate amounts of non-milk extrinsic sugars by subtracting sugar from fruit, vegetables and milk from total sugar. For salt added to food during cooking or at the table, a value was assigned based on a separate question, assuming a standard pinch of salt to be 0.25 g. Total daily salt intake was obtained by summing total salt in food, salt added during cooking and salt added at the table. The HDI is measured from 0 to 10 by assigning a score of 1 if the diet of a woman was within the recommended range for a component and 0 otherwise. Owing to the small numbers in the extremities of the scale, women scoring 0, 1 or 2 were combined and women scoring 7–10 were combined, leaving 6 categories.
Statistical analysis Cox’s proportional hazards regression was used to explore the relationship between the dietary patterns and risk of European Journal of Clinical Nutrition
breast cancer, weighted for the inverse probability of being sampled, using Stata version 9 (StataCorp, 2005). Personyears were calculated from the date the baseline questionnaire was completed until the first occurrence of either a report of incident breast cancer, death or the censor date of the analysis. Associations were estimated for pre- and postmenopausal women separately, first as a simple model adjusting for age and total energy intake by the residuals method (model 1; Willett and Stampfer, 1986), and second, as a full model adjusting for further potential risk factors and confounders. Adjustments in the full model included variables considered to be potential confounders from previous literature. These were age, total energy intake, energy adjusted fat, body mass index, physical activity (h/day sufficiently vigorous to cause sweating), oral contraceptive use, hormone replacement therapy use, smoking habit, parity, age at menarche, alcohol intake (as grams of ethanol/day), breast feeding duration, National Statistics Socio-Economic Class (Bravo et al., 2002) and level of education (model 2).
Results Characteristics according to dietary pattern Of the 35 372 women in the cohort for this analysis, 860 women had a prevalent breast cancer, 695 women could not be flagged for cancer registration with the Office of National Statistics, 87 women had energy intakes outside the expected levels and 8 women had no form date. Therefore, data on 33 731 women were available for inclusion in this analysis. As only 122 subjects had the highest concordance with the Mediterranean dietary pattern and 42 subjects had the highest concordance with the WHO HDI, for further analyses the top four scores (7–10) in each pattern were grouped to ensure more equal distribution of participants by category. Lifestyle characteristics of the women at study baseline according to each dietary pattern are summarized in Tables 2 and 3. Women with a higher concordance to the dietary patterns were younger than women with lower concordance. Women with the highest scores on both patterns had a lower body mass index were less likely to smoke and took more physical activity.
Mediterranean dietary pattern and breast cancer risk The association between the Mediterranean dietary patterns and risk of breast cancer is presented in Table 4 for both model 1 and model 2 for all women combined and also split according to menopausal status at baseline. Analysis of the total cohort showed no statistically significant differences between the groups according to Mediterranean diet score for either the simple or the complex model. In premenopausal women the trend approached significance, women with the highest level of adherence to the Mediterranean diet having a hazards ratio of 0.65 (95%
Dietary patterns and risk of breast cancer JE Cade et al
923 Table 2 Characteristics of UK Women’s Cohort Study participants (n ¼ 33 731) according to Mediterranean diet score Mediterranean diet score n Age (years), mean (s.d.) Professional and managerial class (%) No education, 414 years (%) BMI (kg/m2), mean (s.d.) Physical activity (min/day), mean (s.d.) Current smoker (%) Current HRT use (%) Current OCP use (%) Nulliparous (%) Menarche at o12 years (%) Postmenopause (%) Incident breast cancer (%) Fruit and nuts (g/day), mean (s.d.) Vegetables (g/day), mean (s.d.) Fruit and vegetable (portions/day), mean (s.d.) Legumes (g/day), mean (s.d.) Cereal (g/day), mean (s.d.) Fish and fish products (g/day), mean (s.d.) Poultry and poultry products (g/day), mean (s.d.) Dairy (g/day), mean (s.d.) Meat and meat products (g/day), mean (s.d.) Total energy intake including alcohol (MJ), mean (s.d.) Ethanol (g/day), mean (s.d.) Fat intake (g/day), mean (s.d.) Monounsaturated/saturated fat ratio, mean (s.d.) Vitamin C (mg/day), mean (s.d.)
0–2
3
4
5
6
7–10
3668 54 (9) 55 23 25.6 (5.2) 12 (30) 13 21 4 17 22 60 2.56 190 (130) 190 (90) 5.1 (2.3) 21 (20) 166 (82) 20 (14) 24 (18) 122 (82) 76 (42) 8.8 (2.5) 6.5 (11.1) 81 (29) 0.8 (0.1) 118 (55)
4486 54 (10) 57 23 25.0 (4.4) 12 (24) 14 22 4 19 22 59 2.74 232 (177) 222 (112) 6.1 (3.0) 26 (24) 189 (99) 24 (18) 21 (18) 112 (80) 67 (46) 9.0 (2.7) 8.0 (11.2) 81 (30) 0.9 (0.1) 133 (62)
6008 53 (10) 61 19 24.8 (4.4) 12 (24) 11 19 4 21 21 56 2.33 272 (195) 260 (139) 7.1 (3.5) 31 (29) 215 (110) 27 (21) 20 (20) 115 (83) 60 (51) 9.3 (2.8) 8.7 (10.8) 81 (31) 0.9 (0.2) 150 (71)
6272 52 (9) 63 17 24.5 (4.3) 12 (30) 11 21 4 20 22 53 2.63 321 (224) 311 (160) 8.5 (4.1) 40 (36) 241 (119) 29 (26) 18 (20) 117 (91) 50 (54) 9.7 (3.0) 9.0 (10.7) 84 (33) 1.0 (0.2) 170 (79)
5755 52 (9) 66 14 24.2 (4.4) 18 (30) 10 19 3 21 23 51 2.15 365 (239) 363 (179) 9.8 (4.5) 50 (42) 277 (136) 30 (29) 15 (21) 118 (91) 39 (51) 10.1 (3.1) 9.1( 10.0) 86 (34) 1.0 (0.2) 188 (84)
7542 51 (9) 71 11 23.5 (3.7) 18 (30) 9 18 4 22 23 45 2.41 442 (268) 441 (203) 12.0 (5.3) 66 (50) 322 (140) 33 (33) 9 (18) 107 (90) 19 (38) 10.8 (3.0) 9.7 (8.7) 89 (33) 1.1 (0.3) 219 (97)
Abbreviations: HRT, hormone replacement therapy; OCP, oral contraceptive user.
Table 3 Characteristics of UK Women’s Cohort Study participants (n ¼ 33 731) according to WHO HDI WHO HDI n Age (years), mean (s.d.) Professional and managerial class (%) No education, 414 years (%) BMI (kg/m2), mean (s.d.) Physical activity (min/day), mean (s.d.) Current smoker (%) Current HRT use (%) Current OCP use (%) Nulliparous (%) Menarche at o12 years (%) Postmenopause (%) Incident breast cancer (%) Fruit and vegetables (g) Fruit and vegetables (portions/day), mean (s.d.) Salt (g) Total energy intake including alcohol (MJ), mean (s.d.) Fat intake (g/day), mean (s.d.) Total fat (% total E) Saturated fatty acids (% total E) Polyunsaturated fatty acids (% total E) Cholesterol (mg) Protein (% total E) Total carbohydrate (% total E) Non-milk extrinsic sugars (% total E) Dietary fiber (Englyst; g) Vitamin C (mg/day), mean (s.d.)
0–2
3
4
5
6
7–10
6032 54 (10) 57 21 25.3 (4.6) 12 (28) 16 21 4 18 21 60 2.50 404 (203) 5.5 (2.7) 7.8 (2.3) 9.6 (2.7) 90 (28) 36 (4) 14 (3) 6 (1) 327 (126) 17 (2) 47 (4) 14 (4) 18 (7) 12 (61)
6138 53 (10) 61 17 24.7 (4.3) 14 (27) 12 20 4 19 20 55 2.53 537 (271) 7.3 (3.7) 8.2 (2.9) 10.1 (3.2) 94 (34) 35.7(4.3) 13 (3) 6 (2) 299 (132) 16 (3) 48 (4) 13 (4) 23 (8) 155 (76)
6407 52 (10) 64 16 24.4 (4.1) 15 (30) 12 20 4 20 21 51 2.54 606 (295) 8.3 (4.1) 8.1 (2.9) 10.0 (3.2) 91 (35) 34.9(5) 12 (2) 7 (2) 260 (121) 16 (3) 49 (5) 13 (5) 25 (9) 166 (79)
5957 51 (10) 66 14 24.1 (4.4) 16 (31) 10 19 4 20 23 48 2.33 668 (305) 9.2 (4.3) 7.9 (2.6) 9.8 (2.8) 85 (31) 33.2(5.6) 11 (3) 7 (2) 210 (86) 15 (3) 52 (5) 13 (5) 27 (9) 177 (78)
4519 52 (9) 64 17 24.3 (4.4) 17 (28) 7 20 4 22 24 53 2.41 759 (362) 10.3 (4.9) 7.7 (2.6) 9.5 (2.7) 72 (27) 29.1(5.5) 9 (2) 6 (2) 177 (77) 16 (3) 55 (5) 12 (5) 29 (10) 195 (91)
4678 51 (9) 67 15 23.8 (4.1) 19 (30) 7 19 4 24 24 49 2.37 886 (429) 11.9 (5.8) 7.3 (2.7) 9.4 (2.9) 66 (23) 26.9(4) 8 (2) 6 (2) 135 (69) 15 (2) 59 (4) 12 (5) 32 (12) 215 (105)
Abbreviations: BMI, body mass index; HRT, hormone replacement therapy; OCP, oral contraceptive user; WHO HDI, World Health Organization Healthy Diet Index.
European Journal of Clinical Nutrition
Dietary patterns and risk of breast cancer JE Cade et al
924 Table 4 Mediterranean diet score and risk of breast cancer (n ¼ 33 731) Person years
Model 1a
Cases/total HR
Model 2b 95% CI
HR
95% CI
Combined analysis 0–2 (reference) 3 4 5 6 7–10
8.83 8.96 9.07 9.21 9.36 9.55
94/3668 123/4486 140/6008 165/6272 124/5755 182/7542
1.00 1.08 0.91 1.04 0.86 0.97
— (0.82, 1.41) (0.70, 1.18) (0.80, 1.35) (0.65, 1.14) (0.74, 1.26) P (trend) ¼ 0.4
1.00 1.06 0.98 0.99 0.84 0.96
— (0.77, 1.46) (0.72, 1.33) (0.73, 1.35) (0.60, 1.17) (0.70, 1.32) P (trend) ¼ 0.4
Premenopausal 0–2 (reference) 3 4 5 6 7–10
9.12 9.28 9.33 9.47 9.56 9.66
36/1485 40/1855 58/2661 63/2954 57/2830 96/4157
1.00 0.89 0.89 0.81 0.76 0.86
— (0.56, 1.40) (0.58, 1.36) (0.53, 1.24) (0.48, 1.18) (0.57, 1.30) P (trend) ¼ 0.4
1.00 0.71 0.88 0.72 0.69 0.65
— (0.43, 1.16) (0.56, 1.36) (0.46, 1.13) (0.43, 1.10) (0.42, 1.02) P (trend) ¼ 0.09
Postmenopausal 0–2 (reference) 3 4 5 6 7–10
8.64 8.74 8.86 8.97 9.17 9.42
58/2183 83/2631 82/3347 102/3318 67/2915 86/3385
1.00 1.20 0.92 1.19 0.92 1.03
— (0.85, 1.68) (0.65, 1.29) (0.86, 1.66) (0.64, 1.33) (0.73, 1.46) P (trend) ¼ 0.7
1.00 1.46 1.10 1.26 0.98 1.30
— (0.95, 2.23) (0.71, 1.69) (0.83, 1.94) (0.61, 1.58) (0.83, 2.05) P (trend) ¼ 0.9
Abbreviations: CI, confidence interval; HR, hazard ratio. a Adjusting for age, energy intake and menopausal status (combined analysis). b Adjusting for age, energy intake, menopausal status (combined analysis), calorie-adjusted fat, body mass index, physical activity, oral contraceptive use, hormone replacement therapy use, smoking status, parity, age at menarche, ethanol, total days breast feeding, socioeconomic class and level of education.
confidence interval: 0.42–1.02; P trend 0.087). However, postmenopausally, there was no association between adherence to the Mediterranean dietary pattern and risk of breast cancer.
HDI dietary pattern and breast cancer risk The association between the WHO HDI dietary pattern and risk of breast cancer is presented in Table 5 for all women combined and also split according to menopausal status at baseline. There were no statistically significant associations between the HDI and risk of breast cancer in either the simple or fully adjusted model or by menopausal status.
Discussion The UKWCS was designed to maximize the potential to compare between women consuming different dietary patterns (Schatzkin et al., 2001; Cade et al., 2010). The study design resulted in a higher power to detect differences because of selection of women with greater contrast in dietary patterns and reducing the effect of measurement error (Schatzkin et al., 2001). The cohort analysis undertaken European Journal of Clinical Nutrition
provides a representative and unbiased result, as the sampling method was taken into account, being weighted for the inverse probability of being sampled. In this cohort, the women who had the highest scores for both the Mediterranean pattern and the WHO HDI had healthier lifestyle characteristics, with lower body mass index, were less likely to smoke and took more exercise compared with women with the lowest scores. Similar results were seen in a cohort from Greece which showed that those with a higher Mediterranean diet score reported healthier characteristics, being less likely to smoke, more likely to have a higher level of physical activity and a lower body mass index (Trichopoulou et al., 2003). A random sample of subjects from Mediterranean Southern France found that those with a lower Mediterranean diet score were more likely to be smokers (Scali et al., 2001), although in this study they also tended to be younger, unlike those in the UKWCS. Dietary patterns have the potential to be a useful tool for health promotion activity by reflecting the whole diet as opposed to single nutrients. Numerous scores exist that vary according to the choice of components in the score, the foods allocated to the specific food groups, cutoff values chosen, adjustment (or not) for energy and weighting of individual components to the total score (Waijers et al.,
Dietary patterns and risk of breast cancer JE Cade et al
925 Table 5 World Health Organization Healthy Diet Index and risk of breast cancer (n33 731) Person years
Model 1a
Cases/total HR
Model 2b 95% CI
HR
95% CI
Combined analysis 0–2 (reference) 3 4 5 6 7–10
8.84 9.06 9.19 9.37 9.40 9.53
151/6032 155/6138 163/6407 139/5957 109/4519 111/4678
1.00 1.04 1.05 0.87 0.99 0.95
— (0.83, 1.31) (0.84, 1.32) (0.68, 1.12) (0.76, 1.28) (0.73, 1.24) P (trend) ¼ 0.4
1.00 0.95 1.08 0.94 0.95 0.94
— (0.72, 1.24) (0.83, 1.41) (0.70, 1.24) (0.68, 1.34) (0.67, 1.32) P (trend) ¼ 0.8
Premenopausal 0–2 (reference) 3 4 5 6 7–10
9.18 9.33 9.41 9.56 9.61 9.67
55/2435 61/2776 69/3127 69/3107 43/2131 53/2376
1.00 0.98 0.97 0.88 0.90 0.84
— (0.67, 1.43) (0.67, 1.40) (0.60, 1.30) (0.58, 1.38) (0.56, 1.27) P (trend) ¼ 0.3
1.00 0.97 1.11 1.02 0.89 0.83
— (0.64, 1.48) (0.74, 1.66) (0.67, 1.54) (0.53, 1.49) (0.50, 1.39) P (trend) ¼ 0.6
Postmenopausal 0–2 (reference) 3 4 5 6 7–10
8.61 8.84 8.99 9.15 9.22 9.39
96/3597 94/3362 94/3280 70/2850 66/2388 58/2302
1.00 1.07 1.10 0.85 1.03 1.00
— (0.80, 1.42) (0.82, 1.46) (0.62, 1.17) (0.74, 1.43) (0.71, 1.41) P (trend) ¼ 0.7
1.00 0.92 1.05 0.85 0.96 0.99
— (0.64, 1.31) (0.74, 1.49) (0.57, 1.27) (0.61, 1.50) (0.63, 1.55) P (trend) ¼ 0.9
Abbreviations: CI, confidence interval; HR, hazard ratio. a Adjusting for age, energy intake and menopausal status (combined analysis). b Adjusting for age, energy intake, menopausal status (combined analysis), calorie-adjusted fat, body mass index, physical activity, oral contraceptive use, hormone replacement therapy use, smoking status, parity, age at menarche, ethanol, total days breast feeding, socioeconomic class and level of education.
2007). The two patterns considered in this analysis provide a means of describing the healthiness of the diet. However, the definition of these patterns is not straightforward. Development of an index requires many arbitrary choices to be made. In addition, correlation between dietary components may not be adequately taken into account. As a result, dietary indexes may not predict disease or mortality better than individual dietary factors. A number of different Mediterranean diet definitions have been developed (Trichopoulou et al., 1995, 2003; Scali et al., 2001; Martinez-Gonzalez et al., 2002; Knoops et al., 2004), each of which have different constructs. Medians (Trichopoulou et al., 2003), quintiles (Martinez-Gonzalez et al., 2002) or specific cutoffs (Scali et al., 2001) are used as boundaries. Using the group median as a cutoff value may not be related to a healthy level of intake per se, and this will differ between populations (Waijers et al., 2007). Differences in definitions could lead to differences in associations between the dietary pattern and health outcomes under consideration. In order to have adequate numbers at the extremes of the score distribution, we collapsed the top and bottom end of the scores. Analysis that used all score levels did not result in different findings (data not shown).
A recent comparison of dietary patterns generated by three different methods: cluster analysis, factor analysis and index score analysis found similar beneficial health characteristics in those with the highest quintiles of factor and index scores, suggesting some similarities across the approaches (Reedy et al., 2010). However, no comparison was made across patterns with health outcomes. Other approaches to describe dietary patterns, such as reduced rank regression, may be helpful in explaining variation in disease rates via intermediary nutritional biomarker/nutrients. This approach has shown that similar patterns can account for relatively large amounts of variation across European countries (Kroger et al., 2009). The Mediterranean diet score has been constructed in a number of ways and components may vary depending on the population and outcome under examination. In general, characteristics of the Mediterranean diet are high intakes of vegetables, fruits, legumes, cereals, fish and a moderate intake of red wine during meals (Sofi et al., 2008). The EPIC Greece cohort found that the dominant components of the Mediterranean diet score as a predictor of lower mortality were moderate consumption of ethanol, low consumption of meat and meat products, and high consumption European Journal of Clinical Nutrition
Dietary patterns and risk of breast cancer JE Cade et al
926 of vegetables, fruits and nuts, olive oil and legumes (Trichopoulou et al., 2009). However, in Northern European populations, such as the UKWCS, these essential components may be consumed in different forms and proportions. Rumawas et al. (2009) have recently described an alternative approach for the derivation of a Mediterranean pattern score for use in non-Mediterranean populations. Jensen et al. (2008) used the Mediterranean diet score developed by Trichopoulou et al. (2005) to estimate adherence to a hearthealthy diet in a Danish population. A modified version of this score, in which monounsaturated fats are replaced by all unsaturated fats, has been suggested to make its application more suitable to countries in which olive oil is not the main source of unsaturated fatty acids. Healthy dietary patterns, and in particular the Mediterranean diet, has been explored with regard to cardiovascular disease outcomes (Martinez-Gonzalez et al., 2002; WHO Technical Report Series, 2003; Panagiotakos et al., 2006; Willett, 2006). Less is known about how these patterns relate to risk of cancer (de Lorgeril et al., 1998; Fung et al., 2005; Cottet et al., 2009). A meta-analysis of six cohort studies found that a Mediterranean dietary pattern significantly reduced the incidence or mortality from cancer, relative risk 0.94, 0.92–0.96; Po0.0001 for a two point increase in adherence (Sofi et al., 2008). In our study, we were not able to find any statistically significant associations between either the Mediterranean diet score or the WHO HDI and risk of breast cancer. There is some evidence that a Mediterranean diet could impact on diabetes incidence rates (Martinez-Gonzalez et al., 2008), a disease which is likely on the causal pathway for chronic heart disease, but potentially also for breast cancer (Kabat et al., 2009). A recent analysis of the 14 807 women in the Greek EPIC cohort also found no overall association with the Mediterranean diet score and risk of developing breast cancer, however they did find a marginally significant inverse association among postmenopausal women (Trichopoulou et al., 2010). Studies which have explored the relationship between the Mediterranean dietary pattern and breast cancer risk have tended to use factor analysis to assign the dietary pattern (Sieri et al., 2004; Fung et al., 2005; Velie et al., 2005; Ronco et al., 2006; Murtaugh et al., 2008; Cottet et al., 2009). This approach is very different from using scores, as it groups correlated food types into patterns. Patterns identified in one population may well be different from patterns identified in another, making the transferability of population-specific patterns potentially problematic. Factor analysis if used on nationally representative samples may, however, lead to identification of dietary patterns, which may be more suitable for local dietary advice. Our results are not conclusive with regard to the patterns described. Other studies have used different dietary patterns to describe diet. The EPIC cohort explored risk of breast cancer with fruit and vegetable consumption and found no associations (van Gils et al., 2005), however, the whole dietary pattern was not considered in that analysis. This may be important, as European Journal of Clinical Nutrition
another analysis of the EPIC data found that although adherence to the traditional Mediterranean diet was associated with a significant reduction in total mortality, the associations between each individual food group contributing to the Mediterranean diet score and total mortality were not all individually significant (Trichopoulou et al., 2005). A large French cohort with 2381 postmenopausal invasive breast cancer cases also described dietary patterns by factor analysis rather than using a score. The ‘healthy/Mediterranean’ dietary pattern was negatively associated with breast cancer risk, particularly in women who were estrogen receptor positive and progesterone receptor negative (Cottet et al., 2009). The Nurses Health cohort also used factor analysis to identify two main dietary patterns: the prudent diet, similar to a Mediterranean-type diet and the Western diet. They did not observe an overall association between dietary pattern and risk of postmenopausal breast cancer. Unlike the French cohort, amongst the 17% of women who were estrogen receptor negative, there was a reduced risk of breast cancer with higher prudent diet score (Fung et al., 2005). Hormone receptor status is not currently available for the UKWCS. Putative mechanisms that relate dietary patterns to the development of cancer do exist and may include rates of growth and development. For example, a high fat, low fiber dietary pattern may advance the onset of puberty, resulting in earlier menarche, earlier onset of breast development, and an earlier growth spurt. Both earlier menarche and adult tallness are markers of increased risk of breast cancer (Key et al., 2001). To date, no studies have been designed with a long enough dietary follow-up throughout the life course to confirm this theory. In some populations (Mikkila et al., 2005; Lake et al., 2006), though not all (Gallagher et al., 2006), adult diets have been correlated with childhood diets suggesting that diet in adulthood may be a reasonable proxy for diet earlier in life. Dietary patterns in adulthood may also be associated with risk of breast cancer development possibly through an effect on hormone levels. For example, vegetarian women who have higher levels of fiber and lower fat intakes have lower blood levels and reduced urinary excretion of estrogens (Goldin et al., 1982). As this is a prospective study, recall bias is unlikely. However, accurate measurement of food intake is important for studies of dietary pattern. The FFQ used in this cohort has been validated against biomarkers (Spence et al., 2002) and follows recommendations for good design (Cade et al., 2002). The dietary patterns described reflect existing predefined scores and may not necessarily be those that are optimal for breast cancer prevention. The UKWCS has a health conscious outlook with relatively low smoking rates and low body mass index (Cade et al., 2004). It is possible that less healthy dietary patterns were underrepresented in our cohort. A further weakness of this study was that we did not have information on hormone receptor status of the tumor. Other studies have shown a possible link between dietary pattern and hormone receptor status although
Dietary patterns and risk of breast cancer JE Cade et al
927 findings are not consistent (Fung et al., 2005; Cottet et al., 2009). In conclusion, no statistically significant associations were seen between two common healthy dietary patterns, the Mediterranean diet and the WHO HDI, assessed by a standard scoring method and risk of development of breast cancer. In premenopausal, but not postmenopausal women, there was a nonsignificant inverse association with increasing adherence to the Mediterranean diet pattern.
Conflict of interest The authors declare no conflict of interest.
Acknowledgements This analysis was funded by a grant from the World Cancer Research Fund. We thank James Thomas for his work on the cohort database.
References Bravo Y, Greenwood DC, Cade JE (2002). The impact of social class on a healthy diet: analysis from the U.K. Women’s Cohort Study. Proc Nutr Soc 61, 142A. Cade J, Thompson R, Burley V, Warm D (2002). Development, validation and utilisation of food-frequency questionnaires—a review. Public Health Nutr 5, 567–587. Cade JE, Burley VJ, Greenwood DC (2004). The UK Women’s Cohort Study: comparison of vegetarians, fish-eaters and meat-eaters. Public Health Nutr 7, 871–878. Cade JE, Burley VJ, Greenwood DC (2007). Dietary fibre and risk of breast cancer in the UK Women’s Cohort Study. Int J Epidemiol 36, 431–438. Cade JE, Taylor EF, Burley VJ, Greenwood DC (2010). Common dietary patterns and risk of breast cancer: analysis from the United Kingdom Women’s Cohort Study. Nutr Cancer 62, 300–306. Cottet V, Touvier M, Fournier A, Touillaud MS, Lafay L, ClavelChapelon F et al. (2009). Postmenopausal breast cancer risk and dietary patterns in the E3N-EPIC prospective cohort study. Am J Epidemiol 170, 1257–1267. de Lorgeril M, Salen P, Martin JL, Monjaud I, Boucher P, Mamelle N (1998). Mediterranean dietary pattern in a randomized trial: prolonged survival and possible reduced cancer rate. [Comment]. Arch Intern Med 158, 1181–1187. Fung T, Hu FB, Fuchs C, Giovannucci E, Hunter DJ, Stampfer MJ et al. (2003). Major dietary patterns and the risk of colorectal cancer in women. Arch Intern Med 163, 309–314. Fung TT, Hu FB, Holmes MD, Rosner BA, Hunter DJ, Colditz GA et al. (2005). Dietary patterns and the risk of postmenopausal breast cancer. Int J Cancer 116, 116–121. Gallagher AM, Robson PJ, Livingstone MBE, Cran GW, Strain JJ, Murray LJ et al. (2006). Tracking of energy and nutrient intakes from adolescence to young adulthood: the experiences of the Young Hearts Project, Northern Ireland. Public Health Nutr 9, 1027–1034. Gandini S, Merzenich H, Robertson C, Boyle P (2000). Meta-analysis of studies on breast cancer risk and diet: the role of fruit and vegetable consumption and the intake of associated micronutrients. Eur J Cancer 36, 636–646.
Goldin BR, Adlercreutz H, Gorbach SL, Warram JH, Dwyer JT, Swenson L et al. (1982). Estrogen excretion patterns and plasma levels in vegetarian and omnivorous women. N Engl J Med 307, 1542–1547. Hu FB (2002). Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 13, 3–9. Jensen MK, Chiuve SE, Rimm EB, Dethlefsen C, Tjonneland A, Joensen AM et al. (2008). Obesity, behavioral lifestyle factors, and risk of acute coronary events. Circulation 117, 3062–3069. Kaaks R, Riboli E (1997). Validation and calibration of dietary intake measurements in the EPIC project: methodological considerations. European Prospective Investigation into Cancer and Nutrition. Int J Epidemiol 26(Suppl 1), S15–S25. Kabat GC, Kim M, Caan BJ, Chlebowski RT, Gunter MJ, Ho GY et al. (2009). Repeated measures of serum glucose and insulin in relation to postmenopausal breast cancer. Int J Cancer 125, 2704–2710. Kant AK, Schatzkin A, Graubard BI, Schairer C (2000). A prospective study of diet quality and mortality in women. JAMA 283, 2109–2115. Key TJ, Verkasalo PK, Banks E (2001). Epidemiology of breast cancer. Lancet Oncol 2, 133–140. Knoops KT, de Groot LC, Kromhout D, Perrin AE, Moreiras-Varela O, Menotti A et al. (2004). Mediterranean diet, lifestyle factors, and 10-year mortality in elderly European men and women: the HALE project. JAMA 292, 1433–1439. Kroger J, Ferrari P, Jenab M, Bamia C, Touvier M, Bueno-de-Mesquita HB et al. (2009). Specific food group combinations explaining the variation in intakes of nutrients and other important food components in the European Prospective Investigation into Cancer and Nutrition: an application of the reduced rank regression method. Eur J Clin Nutr 63(Suppl 4), S263–S274. Lake AA, Mathers JC, Rugg-Gunn AJ, Adamson AJ (2006). Longitudinal change in food habits between adolescence (11–12 years) and adulthood (32–33 years): the ASH30 study. J Public Health 28, 10–16. Martinez-Gonzalez MA, Fernandez-Jarne E, Serrano-Martinez M, Marti A, Martinez JA, Martin-Moreno JM (2002). Mediterranean diet and reduction in the risk of a first acute myocardial infarction: an operational healthy dietary score. Eur J Nutr 41, 153–160. Martinez-Gonzalez MA, Fuente-Arrillaga C, Nunez-Cordoba JM, Basterra-Gortari FJ, Beunza JJ, Vazquez Z et al. (2008). Adherence to Mediterranean diet and risk of developing diabetes: Prospective cohort study. BMJ 336, 1348–1351. Mikkila V, Rasanen L, Raitakari OT, Pietinen P, Viikari J (2005). Consistent dietary patterns identified from childhood to adulthood: the cardiovascular risk in young finns study. Br J Nutr 93, 923–931. Murtaugh MA, Sweeney C, Giuliano AR, Herrick JS, Hines L, Byers T et al. (2008). Diet patterns and breast cancer risk in Hispanic and non-Hispanic white women: the Four-Corners Breast Cancer Study. Am J Clin Nutr 87, 978–984. Panagiotakos DB, Arapi S, Pitsavos C, Antonoulas A, Mantas Y, Zombolos S et al. (2006). The relationship between adherence to the Mediterranean diet and the severity and short-term prognosis of acute coronary syndromes (ACS): the Greek Study of ACS (The GREECS). Nutrition 22, 722–730. Reedy J, Wirfalt E, Flood A, Mitrou PN, Krebs-Smith SM, Kipnis V et al. (2010). Comparing 3 dietary pattern methods—cluster analysis, factor analysis, and index analysis—with colorectal cancer risk: the NIH-AARP Diet and Health Study. Am J Epidemiol 171, 479–487. Ronco AL, De SE, Boffetta P, Deneo-Pellegrini H, Acosta G, Mendilaharsu M (2006). Food patterns and risk of breast cancer: a factor analysis study in Uruguay. Int J Cancer 119, 1672–1678. Rumawas ME, Dwyer JT, Mckeown NM, Meigs JB, Rogers G, Jacques PF (2009). The development of the Mediterranean-Style Dietary Pattern Score and its application to the American diet in the Framingham Offspring Cohort. J Nutr 139, 1150–1156. Scali J, Richard A, Gerber M (2001). Diet profiles in a population sample from Mediterranean southern France. Public Health Nutr 4, 173–182.
European Journal of Clinical Nutrition
Dietary patterns and risk of breast cancer JE Cade et al
928 Schatzkin A, Subar AF, Thompson FE, Harlan LC, Tangrea J, Hollenbeck AR et al. (2001). Design and serendipity in establishing a large cohort with wide dietary intake distributions: the National Institutes of Health-American Association of Retired Persons Diet and Health Study. Am J Epidemiol 154, 1119–1125. Sieri S, Krogh V, Pala V, Muti P, Micheli A, Evangelista A et al. (2004). Dietary patterns and risk of breast cancer in the ORDET cohort. Cancer Epidemiol Biomarkers Prev 13, 567–572. Smith-Warner SA, Spiegelman D, Yaun SS, Adami HO, Beeson WL, van den Brandt PA et al. (2001). Intake of fruits and vegetables and risk of breast cancer: a pooled analysis of cohort studies. JAMA 285, 769–776. Sofi F, Cesari F, Abbate R, Gensini GF, Casini A (2008). Adherence to Mediterranean diet and health status: meta-analysis. BMJ 337, a1344. Spence M, Cade JE, Burley VJ, Greenwood DC (2002). Ability of the UK Women’s Cohort Food Frequency Questionnaire to rank dietary intakes: a preliminary validation study. Proc Nutr Soc 61, 117A. StataCorp (2005). Stata Statistical Software: Release 9.1. Stata Corporation: College Station, TX. Taylor EF, Burley VJ, Greenwood DC, Cade JE (2007). Meat consumption and risk of breast cancer in the UK Women’s Cohort Study. Br J Cancer 96, 1139–1146. Trichopoulou A, Bamia C, Lagiou P, Trichopoulos D (2010). Conformity to traditional Mediterranean diet and breast cancer risk in the Greek EPIC (European Prospective Investigation into Cancer and Nutrition) cohort. Am J Clin Nutr 92, 620–625. Trichopoulou A, Bamia C, Trichopoulos D (2009). Anatomy of health effects of Mediterranean diet: Greek EPIC prospective cohort study. BMJ 339, 26–28.
European Journal of Clinical Nutrition
Trichopoulou A, Costacou T, Bamia C, Trichopoulos D (2003). Adherence to a Mediterranean diet and survival in a Greek population. N Engl J Med 348, 2599–2608. Trichopoulou A, Kourisblazos A, Wahlqvist ML, Gnardellis C, Lagiou P, Polychronopoulos E et al. (1995). Diet and overall survival in elderly people. Br Med J 311, 1457–1460. Trichopoulou A, Orfanos P, Norat T, Bueno-de-Mesquita B, Ocke´ MC, Peeters PH et al. (2005). Modified Mediterranean diet and survival: EPIC-elderly prospective cohort study. BMJ 330, 991. van Gils CH, Peeters PH, Bueno-de-Mesquita HB, Boshuizen HC, Lahmann PH, Clavel-Chapelon F et al. (2005). Consumption of vegetables and fruits and risk of breast cancer (see comment). JAMA 293, 183–193. Velie EM, Schairer C, Flood A, He JP, Khattree R, Schatzkin A (2005). Empirically derived dietary patterns and risk of postmenopausal breast cancer in a large prospective cohort study. Am J Clin Nutr 82, 1308–1319. Waijers PM, Feskens EJ, Ocke MC (2007). A critical review of predefined diet quality scores. Br J Nutr 97, 219–231. White E, Kushi LH, Pepe MS (1994). The effect of exposure variance and exposure measurement error on study sample size: implications for the design of epidemiologic studies. J Clin Epidemiol 47, 873–880. WHO Technical Report Series (2003). Diet, Nutrition and the Prevention of Chronic Diseases. Report of a Joint WHO/FAO Expert Consultation. World Health Organization: Geneva. Willett W, Stampfer MJ (1986). Total energy intake: implications for epidemiologic analyses. Am J Epidemiol 124, 17–27. Willett WC (2006). The Mediterranean diet: science and practice. Public Health Nutr 9, 105–110.