Dietary intakes of ω-6 and ω-3 polyunsaturated fatty acids and the risk of breast cancer
Experimental studies suggest detrimental effects of ω-6 polyunsaturated fatty acids (PUFA), and beneficial effects of ω-3 PUFAs on mammary carcinogenesis, possibly in interaction with antioxidants. However, PUFA food sources are diverse in human diets and few epidemiologic studies have examined whether associations between dietary PUFAs and breast cancer risk vary according to food sources or antioxidant intakes. The relationship between individual PUFA intakes estimated from diet history questionnaires and breast cancer risk was examined among 56,007 French women. During 8 years of follow-up, 1,650 women developed invasive breast cancer. Breast cancer risk was not related to any dietary PUFA overall; however, opposite associations were seen according to food sources, suggesting other potential effects than PUFA per se. Breast cancer risk was inversely associated with α-linolenic acid (ALA) intake from fruit and vegetables [highest vs. lowest quintile, hazard ratio (HR) 0.74; 95% confidence interval (CI) 0.63, 0.88; p trend < 0.0001], and from vegetable oils (HR 0.83; 95% CI 0.71, 0.97; p trend 0.017). Conversely, breast cancer risk was positively related to ALA intake from nut mixes (p trend 0.004) and processed foods (p trend 0.068), as was total ALA intake among women in the highest quintile of dietary vitamin E (p trend 0.036). A significant interaction was also found between ω-6 and long-chain ω-3 PUFAs, with breast cancer risk inversely related to long-chain ω-3 PUFAs in women belonging to the highest quintile of ω-6 PUFAs (p interaction 0.042). These results emphasize the need to consider food sources, as well as interactions between fatty acids and with antioxidants, when evaluating associations between PUFA intakes and breast cancer risk. © 2008 Wiley-Liss, Inc.
Breast cancer is the most frequent malignancy among women in Western Europe, North America1 and more recently in Japan.2 In France, about 50,000 incident cases were observed and 11,200 deaths (19% of cancer mortality) were due to breast cancer in 2005.3 It has been estimated that as many as half of breast cancer deaths could be avoided through dietary modifications,4 although the available evidence is far from convincing for most dietary factors.5 The role of fat intake in breast cancer etiology has been investigated for long6 but still remains controversial.7 A meta-analysis of epidemiologic studies8 observed a significant increase in breast cancer risk with high saturated fat intake but no association with monounsaturated and polyunsaturated fatty acid (PUFA) intakes. A recent prospective study9 reported varying associations of breast cancer with dietary fat depending on its food sources, hence suggesting a possible explanation for the discrepant results across epidemiologic studies. Moreover, in a meta-analysis covering 97 studies in rodents, Fay et al.10 underlined the need to distinguish between the effects of ω-6 and ω-3 PUFAs, with ω-6 PUFAs showing strong tumor-enhancing effects and ω-3 PUFAs nonsignificant protective effects. Additional experimental studies suggest that high intakes of ω-3 PUFAs could exert inhibitory effects on mammary tumorigenesis through competition with ω-6 PUFAs11 or formation of oxidation products12 that may in turn depend on the antioxidant status.13–15 So far, such interactions have hardly been examined in epidemiologic studies.
We analyzed the relationship of breast cancer risk to PUFA intake, overall and by food sources, in a large prospective cohort of French women with a high diversity of dietary habits across the national territory.16 The analysis focused on dietary ω-3 PUFA intakes which have been hypothesized to encompass a potential for preventive strategies. Nowadays, a number of dietary fish oil supplements and foods enriched with ω-3 PUFAs are available on the market, although their benefit on cancer prevention has not yet been ascertained.17 As a secondary analysis, we investigated potential interactions of ω-3 PUFA intakes with intakes of ω-6 PUFAs and vitamin E, a liposoluble antioxidant, in relation to breast cancer risk.
Material and methods
E3N is a prospective cohort study on cancer risk factors conducted in France among women insured with the “Mutuelle Générale de l'Education Nationale” (MGEN), a national health insurance scheme covering teachers, teacher spouses and employees of the National Education System.18 Overall, 98,995 women volunteers aged 40-65 years were enrolled in 1989-1991 after replying to a baseline questionnaire and giving their informed consent. The study was approved by the French National Commission for Data Protection and Privacy.
In 1993, a diet history questionnaire19 was sent to participants who had previously answered both the baseline questionnaire and a second questionnaire on reproductive history and hormonal treatments (n = 95,644). After two reminders for nonrespondents, 77,613 dietary questionnaires were collected between June 1993 and July 1997 (81.1% response rate). Of these, 2,104 were excluded because of miscoding (n = 2,050), duplicate answers (n= 46) or empty questionnaire (n = 8). In addition, we excluded 985 respondents who did not give their consent to follow-up by the health insurer (MGEN) in case of dropout. The remaining 74,524 subjects with available dietary measurements constituted the French component of the European Prospective Investigation into Cancer and Nutrition (EPIC).20
Information on educational level, reproductive history, history of benign breast diseases, familial history of breast cancer and hormonal treatments was obtained from the questionnaires preceding the dietary questionnaire and from a follow-up questionnaire accompanying the dietary questionnaire. Anthropometric measurements and smoking history were updated using all available follow-up questionnaires at the time of the analysis. Menopausal status and age at menopause were determined using information on last menstruation, hot flushes, hysterectomy, oophorectomy and use of hormonal treatments recorded in each follow-up questionnaire.
We excluded participants in the top and bottom 1% of the ratio of reported energy intake to basal metabolic rate, computed on the basis of age, height and weight at the time of the dietary survey.21 Of the remaining 73,034 subjects, 4,500 who had reported cancer diagnosis (except basal cell skin carcinoma and lobular breast carcinoma in situ) before responding to the dietary questionnaire and 901 with unavailable follow-up information after this questionnaire were excluded. Finally, to focus on dietary intakes, we excluded 11,626 women who reported use of vitamin E, C or β-carotene supplements in 1995, 2000 or 2002, leaving 56,007 subjects for the analysis.
The diet history questionnaire19 was designed to assess usual dietary intake during the past year and consisted of 2 parts. In the first part, divided into 8 meals or food occasions (from breakfast to after-dinner snacks), subjects were asked about the frequency of consumption (11 possible responses, from less than once per month to every day) of a given list of foods and food groups (e.g., fruit, meat, cheese) and the amounts usually consumed, using either natural units (number of eggs, teaspoons, etc.) or portion sizes as illustrated in a booklet of colored photos (7 possible responses).22 The second part of the questionnaire inquired the relative frequency of consumption of individual foods within food groups of the first part (e.g., oranges, apples, etc. for fruit). It also included specific questions on the usual types of fat used as a spread on bread, used for cooking, added after cooking and used in salad dressings. Overall, the questionnaire covered the daily consumption of 208 food items, beverages and recipes.
The food composition table was initially derived from the French national database.23 Moreover, specific analyses were performed to determine the fatty acid content of a number of food items, including sunflower, olive and peanut oils (3 major types of oils consumed in France), margarine, vegetables and processed meat. For the remaining foods, missing information on PUFA content [i.e., linoleic acid, arachidonic acid, α-linolenic acid (ALA), eicosapentaenoic acid (EPA), docosapentaenoic acid (DPA) and docosahexaenoic acid (DHA)] was completed using the food composition table that had been developed for another French study, based on a compilation of published data.24 Dietary intake of long-chain ω-3 PUFAs was computed as the sum of EPA, DPA and DHA intakes. We also considered the ratio of total ω-6 PUFA intake over total ω-3 PUFA intake. Other nutrients assessed included ethanol, vitamins C and E, and carotenoids expressed as β-carotene equivalents.
The reproducibility and relative validity of the diet history questionnaire was assessed in 1990-1991 using an independent sample of 115 women with an educational level comparable with that of the E3N-EPIC participants.19 The reference method for dietary assessment was based on the average of 12 24-hr dietary recalls completed monthly on varying days of the week during the year after completion of the diet history questionnaire. Spearman's correlation coefficient between the diet history questionnaire and the average of the 12 recalls was 0.49 for total fat intake, 0.48 for total PUFA intake, 0.46 for ω-6 PUFA intake and 0.35 for ω-3 PUFA intake (varying from 0.30 for EPA to 0.38 for ALA).
Cases were ascertained through active follow-up, with questionnaires sent in December 1994, April 1997, June 2000 and July 2002. All follow-up questionnaires asked participants whether any cancer had been diagnosed, requesting the address of their physician for confirmation. Deaths in the cohort were detected from reports by family members or the postal service, and by searching the health insurance company (MGEN) database, which contains updated information on vital status. Information on cause of death was obtained from the National Service on Causes of Deaths (INSERM CepiDC).
A total of 1,864 incident breast cancer cases were identified in the study subjects between their response to the dietary questionnaire and the mailing date of the last follow-up questionnaire in 2002. Of these, 63 were only self-reported cases, the remainder being confirmed by a pathology report (96.6%). These cases were included, as self-reporting proved to be extremely accurate in this cohort. After exclusion of 214 cases of carcinoma in situ, 1,650 cases of invasive breast cancer were available for analysis.
A total of 448,439 person-years (average follow-up 8.0 years) accumulated from the date of return of the dietary questionnaire to the date of cancer diagnosis for all cancer cases (n = 3,349), to the date of the last questionnaire returned for nonrespondents (n = 3,573) and deceased subjects (n = 504), or to the mailing date of the last follow-up questionnaire in 2002 for replies received after this date (n = 48,581). Subjects diagnosed with in situ breast cancer (n = 214) or cancer of another site (n = 1,485) were censored at their date of diagnosis.
The relationship between nutrient intakes and breast cancer risk was investigated using Cox proportional hazards regression models with age as the time scale.25 The estimated hazard ratio (HR) and the corresponding 95% confidence interval (CI) were computed after stratifying on 5-year interval birth cohorts,26 and adjusting for the following confounders: alcohol consumption (continuous), smoking history (never, ever), history of breast cancer in mother or sister(s) (never, ever), personal history of benign breast disease (never, ever), age at menarche (<12, 12-13, ≥14 years), parity (nulliparous, one or 2 children and age at first birth ≤30 years, 3 or more children and age at first birth ≤30 years, age at first birth >30 years), physical activity at baseline (in quintiles of metabolic equivalents), menopausal status (pre, post as a time-dependent covariate) and body mass index (BMI, in quintiles defined according to the baseline distribution, as a time-dependent variable). For postmenopausal women, analyses were further adjusted for age at menopause (≤45, >45 years) and current use of menopausal hormones (no, yes as a time-dependent variable). We tested for an interaction between menopausal status (as a time-dependent covariate) and PUFA intake. Missing ages at menopause were imputed to 47 years if menopause was artificial, or 51 years if menopause was natural, ages which corresponded to the medians observed among postmenopausal women with available age at menopause who reported artificial or natural menopause, respectively. For each remaining adjustment factor, missing data were imputed to the mode among the subjects with complete data, since the proportion of cases having missing data did not exceed 5%. We verified that HR estimates hardly changed after additional adjustment for educational level; however, educational level was not associated with breast cancer risk in the E3N-EPIC population that is rather homogenous with respect to this variable.
We calculated PUFA intake from the 208 food items grouped into food groups such as added fat (vegetable oils, butter, margarine, duck fat), meat (poultry, red and processed meat), marine sources (canned and fresh fish, seafood) and processed foods (food items potentially including partially hydrogenated vegetable oils, i.e., margarine, ready-to-use salad dressing, breakfast cereals, industrial bread and rusk, croissants, biscuits and cakes, chocolate bars and candies, nut mixes, pizza, tarts and sandwiches, French fries, fruit pies and dessert). In all models, PUFA intakes were expressed as percentages of nonalcohol energy (nutrient densities) and adjusted for nonalcohol energy intake.27 Alternatively, we expressed PUFA intakes as the exponentiated residuals of the linear regression of log-transformed PUFA intake on log-transformed energy intake (residual method).27 We also evaluated the addition effect of PUFA intake on breast cancer risk (as if subjects were given supplementation in PUFA) by fitting crude PUFA intake along with energy intake from other sources.28 In both instances, we generally found similar results (data not shown). Nutrient intakes were categorized into quintiles according to the distribution observed in the entire cohort. Tests for trend across quintiles were performed by fitting the median value for each quintile of nutrient intake as a continuous variable and assessing whether its coefficient equaled 0 using the Wald chi-square statistic.
To investigate interactions between ω-3 and ω-6 PUFA and vitamin E intakes, HRs of breast cancer associated with increasing ω-3 PUFA intake were calculated per quintile of ω-6 PUFA (expressed as nutrient density) or vitamin E intakes. As they showed similar results, the second to fourth quintiles of dietary vitamin E intakes were grouped together. We performed tests for interaction by applying the likelihood ratio test to PUFA trend variable (2 degrees of freedom). As there was no clear evidence for an interaction between PUFA and carotenoid intakes (expressed as β-carotene equivalents) with respect to breast cancer, results are not included here. We verified that the proportional hazards assumption was not violated for our main exposure and other fixed covariates by testing interactions between these and functions of age. Moreover, we conducted a lag analysis by excluding the first year of follow-up for all subjects and 168 cases of invasive breast cancer among them. All p values were 2-sided. Statistical analyses were performed using SAS statistical software (release 9.1, 2004; SAS Institute Inc, Cary, NC).
At baseline, total fat intake contributed an average 38.3% of nonalcohol energy intake, including 6.7% from PUFA intake. Linoleic acid was the predominantly consumed PUFA (5.9% energy), followed by ALA (0.43% energy). Among long-chain ω-3 PUFAs (0.22% energy), DHA was the major fatty acid (0.13% energy). The main food sources for linoleic acid intake were added fat (59.3% total intake) including sunflower oil (34.7%), processed foods (24.8%) including nut mixes (7.2%), and meat (10.7%). Arachidonic acid originated mostly from meat (70.9%), followed by marine sources (14.7%) and processed foods (11.8%), whereas long-chain ω-3 PUFAs originated mostly from marine sources (80.3%, including 32.8% from fatty fish), followed by meat (15.8%). In contrast, a variety of foods contributed to ALA intake: processed foods (28.2%) including nut mixes (12.1%), fruit and vegetables (16.4%), cheese (15.2%), butter and margarine (11.4%), meat (11.1%) and vegetable oils (8.7%). Table I shows the characteristics of subjects at inclusion according to quintiles of ALA and long-chain ω-3 PUFA intake (as nutrient density). Increasing quintiles of ALA intake were associated with increasing total and saturated fat intakes as percentages of nonalcohol energy, increasing alcohol consumption, a greater likelihood of smoking and increasing BMI. The same held for long-chain ω-3 PUFAs, except for the absence of a trend with saturated fat intake.
Table I. Baseline Characteristics (Means and Proportions) By Intakes of α-Linolenic Acid and Long-Chain ω-3 Polyunsaturated Fatty Acids (PUFA), E3N-Epic Cohort (n = 56,007), 1993-2002
|Age at entry (years)||53.4||52.8||52.5||52.8||52.5||53.6|
|Education level (%)1|
| <12 years||12.6||11.3||12.4||13.3||11.8||11.5|
| 12-14 years||48.8||50.9||50.1||48.7||50.9||48.1|
| 15-16 years||17.1||18.1||17.6||17.7||17.6||17.8|
| 17+ years||17.3||15.9||16.0||16.3||15.7||18.2|
|Nonalcohol energy intake (kcal/d)||2,125||2,097||2,057||2,138||2,134||1,939|
|Total fat intake (% energy)||32.7||38.8||42.7||36.7||38.4||39.7|
|Saturated fat intake (% energy)||13.1||15.5||16.2||15.3||15.3||14.7|
|Ratio ω-6/ω-3 PUFA||11.1||9.6||8.5||11.7||9.8||7.5|
|Alcohol intake (g/d)||8.4||11.0||14.3||8.9||11.4||13.0|
|Dietary vitamin E intake (mg/d)||12.9||14.2||15.6||13.4||14.2||15.0|
|Mother/sister(s) ever diagnosed with breast cancer (%)||7.4||7.1||7.4||7.4||7.4||7.5|
|History of benign breast disease (%)||27.5||28.3||28.2||27.5||28.1||27.5|
|Ever smoked (%)||42.1||47.7||53.2||44.6||47.3||52.0|
|Body mass index (kg/m2)||22.7||23.2||23.6||22.6||23.1||23.8|
|Age at menarche (%)|
| <12 years||17.5||20.3||22.8||18.5||20.0||23.7|
| 12-13 years||51.4||51.7||50.7||51.4||51.1||50.7|
| 14+ years||31.1||28.0||26.5||30.2||28.9||25.6|
|Combined age at first birth and parity (%)|
| First birth before 30, 1-2 children||47.5||49.7||50.7||48.9||49.8||49.8|
| First birth before 30, 3+ children||27.7||27.5||27.1||27.1||27.9||26.0|
| First birth at 30+||8.7||7.9||6.7||8.7||7.9||6.8|
|Age at menopause (%)|
| ≤45 years||7.5||7.3||7.4||7.5||6.8||8.1|
| 46-54 years||46.8||45.3||44.4||44.1||43.3||49.6|
|Current use of menopausal hormones (%)||31.2||33.1||33.0||29.1||31.0||36.9|
Overall, breast cancer risk was not related to ω-6 (total, linoleic acid, arachidonic acid) or ω-3 (total, ALA, long-chain) PUFA intake, or the ratio of ω-6 to ω-3 PUFAs (Table II). Interaction tests with menopausal status or hormone treatment use among postmenopausal women were not statistically significant. When we stratified on total ω-6 PUFA intake, we found a decreased risk of breast cancer associated with high long-chain ω-3 PUFA intake confined to the highest quintile of ω-6 PUFA intakes (highest vs. lowest quintile of long-chain ω-3 PUFA, HR 0.62; 95%CI 0.44, 0.86; p trend 0.021; p interaction 0.042). Breast cancer risk was not related to long-chain ω-3 PUFA intake in any but the highest quintile of dietary ω-6 PUFAs, and remained unrelated to total ω-3 or ALA intake regardless of ω-6 PUFA consumption.
Table II. Hazard Ratio (HR) and 95% Confidence Interval (CI) for Breast Cancer According to Quintiles of Polyunsaturated Fatty Acid (PUFA) Intakes, E3N-EPIC Cohort (n = 56,007), 1993-2002
|Total ω-6 PUFA (% energy)||I||3.41||337||89,107||1.00|| || |
|II||4.63||341||89,739||1.02||0.87, 1.18|| |
|III||5.63||329||90,177||0.97||0.84, 1.13|| |
|IV||6.82||330||89,964||0.98||0.84, 1.14|| |
|Linoleic acid (% energy)||I||3.33||338||89,114||1.00|| || |
|II||4.54||340||89,711||1.01||0.87, 1.18|| |
|III||5.53||324||90,219||0.96||0.82, 1.11|| |
|IV||6.72||339||89,219||1.01||0.86, 1.17|| |
|Arachidonic acid (% energy)||I||0.05||326||88,897||1.00|| || |
|II||0.07||346||89,865||1.04||0.89, 1.21|| |
|III||0.09||328||89,825||0.98||0.84, 1.15|| |
|IV||0.11||310||90,254||0.92||0.79, 1.08|| |
|Total ω-3 PUFA (% energy)||I||0.44||324||90,687||1.00|| || |
|II||0.54||330||90,490||0.99||0.85, 1.16|| |
|III||0.62||344||89,738||1.03||0.89, 1.20|| |
|IV||0.71||318||89,389||0.94||0.80, 1.10|| |
|α-linolenic acid (% energy)||I||0.32||328||89,605||1.00|| || |
|II||0.37||321||89,940||0.96||0.83, 1.12|| |
|III||0.41||342||89,880||1.01||0.87, 1.18|| |
|IV||0.46||301||89,768||0.89||0.76, 1.04|| |
|Long-chain ω-3 PUFA (% energy)||I||0.08||332||90,892||1.00|| || |
|II||0.13||345||90,353||1.04||0.89, 1.21|| |
|III||0.18||323||90,200||0.96||0.82, 1.12|| |
|IV||0.25||329||89,220||0.96||0.82, 1.12|| |
|Ratio ω-6/ω-3 PUFA||I||5.48||335||88,401||1.00|| || |
|II||7.33||352||89,463||1.06||0.91, 1.23|| |
|III||8.95||343||89,823||1.04||0.89, 1.21|| |
|IV||10.91||305||90,363||0.93||0.80, 1.09|| |
Analyzing the relationship between breast cancer risk and individual PUFA intake from their main food sources (Table III), we found decreased breast cancer risk associated with high linoleic acid or ALA intake from vegetable oils, as well as with high ALA intake from fruit and vegetables. Conversely, an increase in breast cancer risk was associated with high linoleic acid or ALA intake from processed foods, including nut mixes. Breast cancer risk was related to neither ALA intake from other food sources (cheese, butter, margarine or meat) nor longer chain ω-6 (arachidonic acid) and ω-3 PUFA intake from any source. No test for interaction with menopausal status was statistically significant.
Table III. Hazard Ratio (HR) and 95% Confidence Interval (CI) for Breast Cancer According to Quintiles of Polyunsaturated Fatty Acid (PUFA) Intake Divided into Selected Food Sources, E3N-EPIC Cohort (n = 56,007), 1993-2002
|Vegetable oils||50.0% linoleic acid||I||0.71||354||88,764||1.00|| || |
|8.7% α-linolenic acid||I||0.007||343||90,632||1.00|| || |
|Fruit and vegetables||16.4% α-linolenic acid||I||0.037||360||89,758||1.00|| || |
|Processed foods3||24.8% linoleic acid||I||0.46||298||89,559||1.00|| || |
|28.2% α-linolenic acid||I||0.033||305||89,086||1.00|| || |
|Of which nut mixes4||7.2% linoleic acid||I||0.00||373||105,643||1.00|| || |
|12.1% α-linolenic acid||I||0.000||373||105,643||1.00|| || |
Analyses stratified by vitamin E intake are presented in Table IV; results are not shown for intermediate quintiles of dietary vitamin E where null associations were observed. A significant trend of increasing breast cancer risk with increasing ALA intake was found in women in the top quintile of dietary vitamin E (p 0.036); however, the test for interaction was not statistically significant. When we considered ALA intake from various food sources (data not shown), the positive associations with processed foods and nut mixes and the inverse associations with vegetable oils, fruit and vegetables were independent of dietary vitamin E intake. There was no association with longer chain PUFAs, neither arachidonic acid nor ω-3 PUFAs, regardless of dietary vitamin E intake.
Table IV. Hazard Ratio (HR) and 95% Confidence Interval (CI) for Breast Cancer According to Quintiles of Polyunsaturated Fatty Acid (PUFA) Intake Stratified on Quintiles of Dietary Vitamin E Intake, E3N-EPIC Cohort (n = 56,007), 1993-2002
|Lowest quintile of dietary vitamin E intake (median 8.5 mg/d)|
| α-linolenic acid (% energy)||I||0.32||108||27,217||1.00|| || || |
| Long-chain ω-3 PUFA (% energy)||I||0.08||103||23,925||1.00|| || || |
| Arachidonic acid (% energy)||I||0.05||92||21,132||1.00|| || || |
|Highest quintile of dietary vitamin E intake (median 21.1 mg/d)|
| α-linolenic acid (% energy)||I||0.32||46||13,831||1.00|| || || |
| Long-chain ω-3 PUFA (% energy)||I||0.08||59||15,190||1.00|| || || |
| Arachidonic acid (% energy)||I||0.05||65||16,242||1.00|| || || |
The statistically significant associations of breast cancer risk with linoleic acid and ALA intakes from vegetable oils, ALA intake from fruit and vegetables, and linoleic acid and ALA intakes from nut mixes remained when we excluded the first year of follow-up (data not shown). The inverse association between long-chain ω-3 PUFA intake and breast cancer risk among women in the highest quintile of ω-6 PUFA intake also remained statistically significant (p trend 0.018); however, the test for interaction became borderline significant (p 0.061).
This prospective cohort study showed no evidence of an association between estimated ω-6 or ω-3 PUFA intakes and breast cancer risk, in agreement with former meta-analyses of prospective studies for linoleic acid29 and total ω-3 PUFA17 intakes. However, our study provided some indication that dietary PUFA may play opposite roles in breast cancer risk depending on their food sources and that ω-6 and ω-3 PUFA intakes may interact in their association with breast cancer. Whether dietary vitamin E further affects dietary PUFA and breast cancer association was not clearly demonstrated from our study.
The epidemiologic evidence regarding ALA and breast cancer relationship is conflicting. Only one case-control study in Uruguay30 found a positive association between ALA intake and breast cancer risk. In contrast, the Netherlands Cohort Study,31 a case-control study in Italy32 and a case-control study using adipose breast tissue33 showed decreased breast cancer risks associated with high ALA levels. Other case-control studies reported no association of breast cancer risk with dietary ALA intakes,34 or plasma35, 36 and erythrocyte membrane37 levels. The discrepancy across studies seems unlikely to be explained by different levels of consumption: for instance, median ALA intake in the E3N-EPIC cohort was 0.9 g/d, compared with 1.0 g/d in the Uruguayan controls.30
Instead, another possible explanation for the lack of consistency among studies may be the different dietary sources of ALA. In the Uruguayan study,30 meat was its major dietary source, whereas oils and raw vegetables were the main contributors in the Italian study finding an inverse association.32 The diversity of food sources that we observed for ALA intake in our study is consistent with another French study.24 In line with previous studies, we found decreased breast cancer risk associated with high consumption of ALA from fruit and vegetables, and vegetable oils, and increased breast cancer risk associated with high consumption of nut mixes and processed meat. Because nut mixes were proposed as a single item eaten with aperitifs, we cannot exclude residual confounding by alcohol consumption; we note, however, that another French study reported a positive association between breast cancer risk and ALA intake originating mainly from nuts.38 Moreover, the stronger association that we observed between ALA intake and breast cancer risk in the highest quintile of dietary vitamin E intake favors a direct association with nuts, a major dietary source for both these nutrients. Altogether, these results suggest that ALA intake may not have a direct biological effect on breast cancer occurrence but may reflect different food patterns involving other nutrients that could affect breast cancer risk. Thus, the inverse association of breast cancer risk with ALA intake from vegetables and vegetable oils, as with linoleic acid, may indeed result from other compounds, such as folate or lignan that have been related to a reduced risk in this cohort.39, 40
Alternatively, whether all cis vs. cis/trans-ALA isomers originating from different food sources may exert different effects on carcinogenesis remains to be determined. It is of interest that linoleic acid intake also showed opposite associations with breast cancer depending on whether it came from (unaltered) vegetable oils or from processed foods. Partially hydrogenated vegetable oils used in the preparation of margarines and a number of industrial foods are a source of trans fatty acids including isomers of ALA and linoleic acid41, 42 that have been associated with an increased risk of breast cancer in biomarker studies,43, 44 including the present cohort.45
Overall, we found no association between long-chain ω-3 PUFA intake and breast cancer risk, in agreement with most epidemiologic cohort studies.17, 46–48 As exceptions, inverse associations have been reported in Chinese and Japanese women having intakes up to 40 times greater than Western ones.49–51 We note that, like in other Western populations,24, 52 meat consumption sensibly contributed to long-chain ω-3 PUFA intake in this cohort, in addition to marine sources. Although null associations persisted regardless of food sources, we did observe a statistically significant interaction with ω-6 PUFAs, suggesting that women with lowest long-chain ω-3 PUFA intake but highest ω-6 PUFA intake (therefore very low long-chain ω-3 to ω-6 PUFA ratio) could benefit from increasing their long-chain ω-3 PUFA intake. Similarly, some epidemiologic studies have shown decreased breast cancer risks associated with high ratios of (long-chain) ω-3 to ω-6 PUFA intake, based on dietary questionnaires53, 54 or biomarkers.33, 55, 56 In the same line, a Chinese prospective study49 found a direct association between ω-6 PUFA intakes and breast cancer risk confined to women having the lowest intakes of long-chain ω-3 PUFAs. Such observation is consistent with the competition between ω-6 and ω-3 PUFAs for eicosanoid production as an underlying mechanism.11, 57
We did not find any evidence of an interaction between long-chain ω-6 or ω-3 PUFA and dietary vitamin E with respect to breast cancer risk. Long-chain PUFA contribute to the production of peroxides57 which have been hypothesized to play a role in breast cancer etiology.58 Although peroxides have a proven genotoxic and cytotoxic action on normal cells, they might favor apoptosis in tumor cells.59 Experimental evidence indicates that antioxidants, notably the lipid-soluble vitamin E, may interact with the effect of long-chain PUFAs on tumor growth by preventing the formation of peroxides.12, 13, 60 So far, a limited number of epidemiologic studies have been published, and their results were divergent: whereas 2 prospective studies suggested an inverse association of breast cancer risk with combined high intakes of vitamin E and PUFA,61, 62 a case-control study reported a decreased risk of breast cancer associated with high arachidonic acid intakes among women with low vitamin E intakes, but an increased risk among women with both high arachidonic acid and vitamin E intakes.34
Our prospective analysis of the relationship between dietary PUFA intake and breast cancer risk had several strengths and limitations. The strengths of our study included the minimal loss to follow-up (only 6.4% nonrespondents), the high degree of reliability of the dietary questionnaire comparable to other studies,19 its detailed list of foods including 14 types of fish, allowing the distinction between leaner and fattier fishes and an accurate estimation of long-chain ω-3 PUFA intake,57 high variability of dietary patterns across France16 and of ALA food sources24 and a sufficient sample size to allow stratified analyses. In contrast, our study was limited by a single dietary measurement, and the inability to assess other liposoluble antioxidants (e.g., carotenoids) or to distinguish between vitamin E compounds (e.g., α-tocopherol). Changes in dietary habits or food composition may have occurred before the onset of breast cancer, further enhancing measurement error inherent to the use of dietary questionnaire.63, 64 Moreover, we could not assess consumption of ω-3 supplements during the study period although it was likely to be marginal.24 How resulting misclassification between categories of fatty acid or vitamin E consumption65 may affect estimated associations between PUFA intakes and breast cancer risk is unclear. At least, misclassification is likely to be nondifferential with respect to breast cancer due to the prospective design. Finally, we cannot exclude results due to chance alone owing to the high number of tests performed.
In conclusion, findings from this large French cohort indicate that the association of breast cancer risk with ω-6 and ω-3 PUFA intake may differ according to food sources, either reflecting different food patterns or geometrical isomers. The study also provided further support to the hypothesis of an interaction between ω-6 and ω-3 PUFAs with respect to breast cancer risk. Altogether, these results emphasize the need to distinguish among food sources of PUFAs, e.g., fish, vegetables, vegetable oils vs. processed foods and to consider interactions between ω-6 and ω-3 PUFAs when evaluating potential beneficial effects of high ω-3 PUFA intakes for breast cancer prevention. The hypothesis of an effectmodification by antioxidant intakes deserves further consideration.
Dr. Anne Thiébaut is grateful to “Fondation de France” in the French League Against Cancer for their financial support. Dr. Véronique Chajès is the recipient of a grant from Carrefour Foundation. The authors also wish to thank Ms. Rafika Chaït and Mr. Yann Follain (E3N-EPIC) for collecting and entering the pathology report data, Ms. Maryvonne Niravong (E3N-EPIC), Dr. Jayne Ireland (CIQUAL), Dr. Serge Hercberg and Dr. Pierre Astorg (SU.VI.MAX) for their help in drawing up the food composition table, Dr. Nicole Combe (ITERG) for performing the fatty acid analysis of selected foods, Mr. Baptiste Prestel (E3N-EPIC) for updating the analysis dataset, Dr. Marina Touillaud (E3N-EPIC) for her help with the revisions, and Mr. Garth Evans for his assistance with the English. The study of individual polyunsaturated fatty acid intakes and breast cancer risk was part of a collaborative project (Nutrialis) funded by the French Ministry of Research. The E3N study is being carried out with the financial support of the French League Against Cancer, the European Community, 3M Company, “Mutuelle Générale de l'Education Nationale”, the French Institute of Health and Medical Research (INSERM), Gustave Roussy Institute and several General Councils in France. The authors are indebted to women participants and practitioners for their active collaboration to the E3N study.