Experimental studies have indicated that n-3 fatty acids, including alpha-linolenic acid (18:3 n-3) and long-chain n-3 polyunsaturated fatty acids inhibit mammary tumor growth and metastasis. Earlier epidemiological studies have given inconclusive results about a potential protective effect of dietary n-3 polyunsaturated fatty acids on breast cancer risk, possibly because of methodological issues inherent to nutritional epidemiology. To evaluate the hypothesis that n-3 fatty acids protect against breast cancer, we examined the fatty acid composition in adipose tissue from 241 patients with invasive, nonmetastatic breast carcinoma and from 88 patients with benign breast disease, in a case-control study in Tours, central France. Fatty acid composition in breast adipose tissue was used as a qualitative biomarker of past dietary intake of fatty acids. Biopsies of adipose tissue were obtained at the time of surgery. Individual fatty acids were measured as a percentage of total fatty acids, using capillary gas chromatography. Unconditional logistic regression modeling was used to obtain odds ratio estimates while adjusting for age, height, menopausal status and body mass index. We found inverse associations between breast cancer-risk and n-3 fatty acid levels in breast adipose tissue. Women in the highest tertile of alpha-linolenic acid (18:3 n-3) had an odds ratio of 0.39 (95% confidence intervals [CI] = 0.19–0.78) compared to women in the lowest tertile (trend p = 0.01). In a similar way, women in the highest tertile of docosahexaenoic acid (22:6 n-3) had an odds ratio of 0.31 (95% CI = 0.13-0.75) compared to women in the lowest tertile (trend p = 0.016). Women in the highest tertile of the long-chain n-3/total n-6 ratio had an odds ratio of 0.33 (95% confidence interval = 0.17–0.66) compared to women in the lowest tertile (trend p = 0.0002). In conclusion, our data based on fatty acids levels in breast adipose tissue suggest a protective effect of n-3 fatty acids on breast cancer risk and support the hypothesis that the balance between n-3 and n-6 fatty acids plays a role in breast cancer. © 2001 Wiley-Liss, Inc.
Breast cancer is the most frequent cancer among women, with an incidence rate of 88.0 per 100,000 women in France. International variation in breast cancer incidence rates and changes in incidence among migrant populations have indicated that breast cancer risk is influenced by environmental factors, in particular diet, and therefore may be preventable. Among dietary constituents, data from a number of case-control studies1–4 suggested a protective effect of a diet high in fish, but results were generally not confirmed by cohort studies.5, 6 A number of studies would indicate that a diet high in vegetables or fruits might protect against breast cancer.1, 2, 7, 8 Although a few studies9, 10 have failed to document a relation, the finding of a protective effect of vegetables or fruits is relatively consistent.
The role of specific food-related nutrients in the risk of breast cancer has yet to be determined. Most data suggest that the association between dietary fat and breast cancer may be more dependent on the type of fat consumed than on total fat intake.11–13 Most of the epidemiological studies have been concentrated on saturated fat and total polyunsaturated fatty acids (PUFA). Limited data is available on the relation of estimated dietary intake of n-3 long-chain PUFA originating from fish or alpha-linolenic acid from vegetables to the risk of breast cancer. With regard to long-chain n-3 PUFA, no association has been found between estimated dietary intake14 or energy from long-chain n-3 PUFA15 and breast cancer risk. Inconsistent data were also found for alpha-linolenic acid, with negative16 or positive association17 between dietary intake of alpha-linolenic acid and breast cancer risk. Data derived from animal experiments, however, generally showed that diets high in n-3 long-chain PUFA18 or in alpha-linolenic acid19–21 inhibit mammary tumor growth and metastasis. Given the relative consistence of experimental data on the inhibitory effect of n-3 fatty acids, the role of these nutrients in the risk of breast cancer requires further attention.
Conclusive evidence of a role for individual n-3 fatty acids in breast cancer risk may be precluded by the many methodological limitations in measurements of dietary intake of individual fatty acids.22 In this regard, the use of reliable markers of relatively stable metabolic characteristics related to diet is of major interest. Among all biological markers of qualitative composition of dietary fatty acids, adipose tissue fatty acid composition is particularly advantageous because it reflects qualitative dietary intake of fatty acids on a long-term basis,23–25 thereby avoiding the potential bias derived from an effect of the disease on the measured biochemical parameters. We have previously examined adipose tissue fatty acid levels and the risk of metastasis occurrence in patients treated for breast cancer. We reported an inverse association between the level of alpha-linolenic acid in adipose tissue and the development of metastases, suggesting a favorable effect of alpha-linolenic acid on breast cancer prognosis.26 Furthermore, when we compared these patients to a population of control patients, we found an inverse association between alpha-linolenic acid in adipose fat and breast cancer, suggesting a protective effect of this fatty acid on the risk of the disease.27 This finding, however, remained to be confirmed, and additional data on the relation of long-chain n-3 PUFA (originating from fish) to the risk of breast cancer are needed.
To examine the role of n-3 fatty acids (alpha-linolenic acid and long-chain n-3 PUFA) in the risk of breast cancer, we therefore conducted a new, independent case-control study among 329 women treated for breast cancer or benign breast disease at the University Hospital of Tours in France.
MATERIAL AND METHODS
A total of 329 patients were selected for our study at the University Hospital of Tours, central France, between 1992 and 1996. They included 241 patients presenting with non-metastatic invasive breast carcinoma (cases) and 88 patients with benign breast diseases (controls). The population of control and case patients was ethnically homogeneous (all were Caucasian women) and lived all in a limited area of central France. Patients had surgery as first treatment step, during which a specimen of adipose tissue was retained. The following information, collected at baseline at the University Hospital, was recorded: age at diagnosis, height, weight, body mass index (BMI), menopausal status, as well as age at menarche and age at first pregnancy (Table I).
Table I. Descriptive Characteristics of Study Population
| ≤ 37.3||29 (32.9)||22 (9.1)|
| 37.4–49||29 (32.9)||265 (27.0)|
| > 49||30 (34.2)||154 (63.9)|
| Age at menarche||0.3|
| ≤ 12||28 (31.8)||95 (39.4)|
| 13||17 (19.3)||55 (22.8)|
| ≥ 14||36 (40.9)||80 (33.2)|
| Unknown||7 (8.0)||11 (4.6)|
| Age at first pregnancy||0.051|
| ≤ 22||22 (33.8)||87 (41.4)|
| 23–25||23 (35.4)||58 (27.6)|
| ≥ 26||17 (26.2)||54 (25.7)|
| Unknown||3 (4.6)||11 (5.3)|
| Menopausal status||0.001|
| Pre-menopausal||66 (75.0)||105 (43.6)|
| Post-menopausal||22 (25.0)||136 (56.4)|
| Height (cm)||0.01|
| ≤ 157||15 (17.4)||72 (30.3)|
| 158–160||17 (19.8)||61 (25.6)|
| 161–165||29 (33.7)||67 (28.2)|
| ≥ 166||25 (29.1)||38 (15.9)|
| Body mass index (kg/m2)||0.0009|
| ≤ 20.4||29 (32.9)||36 (14.4)|
| 20.5–22.6||29 (32.9)||62 (25.7)|
| > 22.6||30 (34.2)||141 (58.5)|
| Unknown||0||2 (0.8)|
Clinical characteristics of cases' carcinomas included histologic type, tumor size, nodal status, estrogen and progesterone receptors (Table II). Benign breast tumors among the controls included dystrophy (43.2%), inflammatory pathologies (4.6%), papilloma (10.2%), fibroadenoma (27.3%), phylloid tumors (2.3%) and other types of benign pathologies (12.4%).
Table II. Clinical Characteristics of Breast Carcinoma
| Invasive ductal carcinoma||205 (85.1)|
| Invasive lobular carcinoma||26 (10.8)|
| Other type||9 (3.7)|
| Unknown||1 (0.4)|
| T0||14 (5.8)|
| T1 (< 20 mm)||74 (30.7)|
| T2 (20–50 mm)||131 (54.4)|
| T3 (> 50 mm)||18 (7.5)|
| T4||2 (0.8)|
| Unknown||2 (0.8)|
| N0||184 (76.3)|
| N1 (1 or more suspected mobile lymph nodes)||24 (10.0)|
| N2 (fixed lymph nodes)||31 (12.9)|
| Unknown||2 (0.8)|
|Estrogen receptor (fmol/mg)|
| ≤ 15||26 (10.8)|
| > 15||196 (81.3)|
| Non available||19 (7.9)|
|Progesterone receptor (fmol/mg)|
| ≤ 15||58 (24.0)|
| > 15||165 (68.5)|
| Non available||18 (7.5)|
Adipose tissue preparation and fatty acids analysis
A fragment of adipose tissue was removed from the lumpectomy or mastectomy specimen during initial surgery. This fragment was freed from epithelial breast or carcinoma tissue, and kept frozen in liquid nitrogen until analysis. Total lipids were extracted from adipose tissue with chloroform-methanol 2:1 (v:v).28 The extract was washed with NaCl and the mixture was allowed to separate into two phases. The chloroform layer was transferred to a tube and evaporated to dryness under nitrogen. The lipid extract was dissolved in 200 μl of chloroform-methanol 2:1 (v:v) and directly used for the column procedure.
Triglycerides were purified by adsorption chromatography on silica tubes (Supelco, France) as follows: the column was pre-treated with 2 ml chloroform-methanol 2:1 (v:v). Then, the lipid sample was applied to the column and the tube treated with 100 μl of this same system solvent added to the column. Triglycerides were eluted with 20 ml of chloroform. Chloroform fractions were collected in screw-cap tubes with a Teflon seal. Fractions obtained were evaporated to dryness. Three hundred microliters of sodium methoxyde 2 N and 500 μl of boron trifluoride were added to convert the fatty acids to their methyl esters. The mixture was incubated and shaken 15 min at room temperature. Fatty acid methyl esters (FAME) were extracted twice into hexane.
For chromatography procedures, samples were analyzed in batches of 12 that included 8 samples drawn from cases, 3 samples from controls and 1 standard (Supelco™, 37 component FAME Mix, France). Samples were identified solely by a code number and ordered randomly within a batch. Except for the standard, the laboratory was blind to the sample origin (case or control). FAME composition was determined by capillary gas chromatography. The system was composed of a GC 8000 SERIES chromatograph (Thermoquest, France) equipped with an AS 800 autosampler (Thermoquest), a cold on-column injector, and a flame ionization detector. A 50 m long × 0.32 mm internal diameter fused silica column (SGE, France) was used. Stationary phase was a 0.25 μm thick 70% Cyanopropyl polysilphenylene-siloxane bonded phase. Helium was used as a carrier gas at a flow rate of 1.5 ml/min. The detector temperature was 270°C. One microliter was injected through the cold on-column injector (60°C). The oven temperature was programmed to rise from 60–170°C at a rate of 15°C/min and kept constant for 20 min, from 170–190°C at a rate of 5°C/min and kept constant for 10 min, and from 190–215°C at a rate of 1°C/min and kept constant for 10 min. Identification of FAME was obtained by comparison of their relative retention times with those of pure standard mixtures. The relative amount of each fatty acid was quantified by integrating the peak at baseline and dividing the results by the total area for all fatty acids. All integrations were performed by the same laboratory worker. Peaks accounting for less than 1% of total area, such as alpha-linolenic acid (18:3 n-3) or docosahexaenoic acid (DHA, 22:6 n-3), were well detected and quantified.
Within-day coefficients of variation (CV) were based on the analysis of one sample, aliquoted into 10 subsamples, all extracted and analyzed during the same day. CV ranged from 0.5% for the largest peaks to about 10% for the smallest peaks.
Clinical characteristics were compared between the 2 populations by a χ2 test or a 2-sample t-test. Pearson correlation coefficients were calculated between fatty acid levels and some clinical characteristics of patients.
An unconditional logistic regression analysis was performed using SAS System.29 Odds ratios and 95% confidence intervals (CI) were calculated to estimate the relative risk for tertiles as well as for continuous variables. Tertile cutpoints were determined by the distribution of fatty acids among controls, and the lowest tertile was used as the reference category (tertile 1). Potential confounding effects of age, height, BMI, and menopausal status (pre- and post-menopausal) were adjusted for in multivariate logistic regression models. Age, height and BMI were included either as continuous variables or were categorized according to tertiles. Modeling age, height and BMI in continuous variables or categorized in tertiles did not change substantially the estimates. In our data set, there are few controls, and it might lead to some unstable parameter estimates due to sparse numbers in some categories. Thus, only the values adjusted for these exposure variables as continuous were kept. Because the association of BMI with breast cancer varies according to menopausal status, an additional analysis adjusting for age, height, BMI as continuous variables, menopausal status and taking into account the interaction between menopausal status and BMI was performed. The estimates obtained did not change. Because the estimates did not change between the different adjustments, only the crude values and the values adjusted for age, height, BMI (as continuous variables), menopausal status and BMI-menopausal status interaction were kept. Tests for trend were performed for the tertiles of fatty acids on the basis of the likelihood estimates. All p-values quoted are 2-sided.
Characteristics of patients
No significant difference in the distribution of age at menarche and age at first pregnancy was found between patients with breast cancer and patients with benign breast disease. There were significant differences in age at diagnosis, menopausal status, height and BMI distributions of case and control patients. There was a higher proportion of case patients with age above 49 years or a BMI above 22.6 kg/m2 or a height ≤160 cm and a higher proportion of case patients than of controls were post-menopausal.
Fatty acid composition of adipose tissue
Individual fatty acid levels and summed classes of adipose breast tissue in case and control patients are given on Table III. The major fatty acids in breast adipose tissue of patients (cases and controls) were oleic acid (18:1 n-9c), palmitic acid (16:0), linoleic acid (18:2 n-6) and stearic acid (18:0). These fatty acids accounted for about 80% of total area under the chromatographic curve. For each individual fatty acid, a large variability between subjects was observed.
Table III. Fatty Acid Composition of Breast Adipose Tissue in Cases and Controls
| 16:1 n-7c||3.66||0.153||3.38||0.070|
| 18:1 n-9c||38.43||0.289||37.23||0.209|
| 18:2 n-6||13.42||0.394||15.07||0.256|
| 18:3 n-6||0.13||0.004||0.15||0.004|
| 20:4 n-6||0.33||0.017||0.39||0.010|
| Total n-63||14.67||0.420||16.54||0.265|
| 18:3 n-3||0.51||0.016||0.46||0.007|
| 22:6 n-3||0.20||0.010||0.21||0.006|
| Total n-34||0.98||0.027||0.99||0.014|
| Long chain n-35/total n-6||0.034||0.002||0.034||0.001|
Correlation between adipose tissue fatty acid levels and characteristics of patients
In the whole population of patients (n = 329), age at diagnosis was inversely associated with saturated fatty acids, 14:0 (r = −0.37; p = 0.0001), 16:0 (r = −0.31; p = 0.0001) and 18:0 (r = −0.39; p = 0.0001) and positively with monounsaturated fatty acid 16:1 n-7 (r = 0.45; p = 0.0001), and with 18:2 n-6 (r = 0.17; p = 0.002), total long-chain n-6 PUFA (r = 0.52; p = 0.0001) and total long-chain n-3 PUFA (r = 0.58; p = 0.0001). A weak inverse association was observed between age and 18:3 n-3 (r = −0.13; p = 0.02). The same tendency was observed between BMI and fatty acid levels, with an inverse link with 14:0 (r = −0.37; p = 0.0001), 18:0 (r = −0.43; p = 0.0001) and 18:3 n-3 (r = −0.17; r = 0.003) and a positive link with 16:1 n-7 (r = 0.45; p = 0.0001), total long-chain n-6 (r = 0.57; p = 0.0001) and total long-chain n-3 (r = 0.40; p = 0.0001).
Calculations were done also for subgroups of control (n = 88) and case patients (n = 241). In each subgroup of patients, the same tendency remained as in the whole population (data not shown).
Relative risk (odds ratio) of breast cancer by adipose tissue fatty acid levels
The crude odds ratio of breast cancer according to levels of saturated and monounsaturated fatty acids showed that 14:0, 16:0, 18:0 and 18:1 n-9c were inversely linked to breast cancer-risk (Table IV). When adjusting for age at diagnosis, height, BMI, menopausal status and the interaction BMI-menopausal status, no significant association was found between saturated fatty acids and breast cancer risk, except for a positive association between 17:0 levels in adipose tissue and breast cancer risk. Women in the highest tertile of 17:0 had an odds ratio of 1.41 (95% CI = 0.72–2.78) compared to women in the lowest tertile (trend p = 0.01). Women in the highest tertile of 18:1 n-9c had an odds ratio of 0.41 (95% CI = 0.21–0.82) compared to women in the lowest tertile (trend p = 0.0018). Among n-6 PUFA, linoleic acid, the primary essential fatty acid of the n-6 series, 20:4 n-6 and total n-6 PUFA were positively associated with breast cancer-risk. When adjusting for the same exposure variables, women in the highest tertile of 18:2 n-6 had an odds ratio of 2.31 (95% CI = 1.15–4.67) compared to women in the lowest tertile (trend p = 0.06). This link accounts for the positive association observed between total n-6 PUFA and breast cancer: women in the highest tertile of total n-6 PUFA had an odds ratio of 2.29 (95% CI = 0.12–4.69) compared to women in the lowest tertile (trend p = 0.07). For 20:4 n-6, adjusted odds ratio did not show a significant association.
Table IV. Estimated Relative Risk of Breast Cancer and 95% CI by Adipose Tissue Fatty Acid Levels from the Whole Population
| 14:0||1.00||0.62 (0.34–1.11)||0.30 (0.16–0.56)||0.0002|
|0.99 (0.51–1.19)1||0.67 (0.31–1.43)||0.28|
| 15:0||1.00||0.96 (0.52–1.74)||0.77 (0.42–1.40)||0.38|
|1.14 (0.58–2.24)1||0.86 (0.43–1.71)||0.50|
| 17:0||1.00||0.57 (0.30–1.07)||1.02 (0.57–1.83)||0.86|
|0.68 (0.33–1.41)1||1.41 (0.72–2.78)||0.01|
| 16:0||1.00||0.49 (0.27–0.89)||0.32 (0.17–0.60)||0.0002|
|0.72 (0.36–1.41)1||0.44 (0.22–0.88)||0.34|
| 18:0||1.00||0.39 (0.21–0.72)||0.48 (0.26–0.86)||0.008|
|0.76 (0.37–1.53)1||1.17 (0.57–2.40)||0.10|
| 18:1 n-9c||1.00||0.66 (0.36–1.19)||0.50 (0.27–0.91)||0.02|
|0.70 (0.36–1.36)1||0.41 (0.21–0.82)||0.0018|
| 18:2 n-6c||1.00||1.67 (0.88–3.17)||2.97 (1.60–5.51)||0.0005|
|1.60 (0.76–3.36)1||2.31 (1.15–4.67)||0.06|
| 18:3 n-6||1.00||0.84 (0.42–1.68)||1.21 (0.62–2.35)||0.39|
|0.53 (0.24–1.19)1||0.56 (0.26–1.21)||0.74|
| 20:4 n-6||1.00||0.95 (0.49–1.85)||2.94 (1.55–5.58)||0.0002|
|0.87 (0.41–1.84)1||0.98 (0.42–2.29)||0.32|
| Total n-6||1.00||1.65 (0.86–3.18)||3.15 (1.69–5.88)||0.0002|
|1.50 (0.71–3.21)1||2.29 (0.12–4.69)||0.07|
| 18:3 n-3||1.00||1.06 (0.58–1.94)||0.41 (0.22–0.74)||0.005|
|0.97 (0.50–1.90)1||0.39 (0.19–0.78)||0.01|
| 22:6 n-3||1.00||1.53 (0.82–2.85)||1.40 (0.74–2.67)||0.36|
|0.84 (0.40–1.75)1||0.31 (0.13–0.75)||0.016|
| Total n-3||1.00||1.73 (0.94–3.19)||1.43 (0.77–2.64)||0.31|
|0.91 (0.45–1.87)1||0.40 (0.17–0.94)||0.001|
| 18:3n-3/18:2n-6||1.00||0.62 (0.34–1.12)||0.31 (0.17–0.57)||0.0002|
|0.89 (0.46–1.75)1||0.41 (0.20–0.81)||0.0004|
| Long chain n-3/total n-6||1.00||1.14 (0.62–2.10)||1.34 (0.74–2.45)||0.33|
|0.48 (0.23–0.97)1||0.33 (0.17–0.66)||0.0002|
Considered n-3 fatty acids in isolation, alpha-linolenic acid was inversely associated with breast cancer-risk. When adjusting for age, height, BMI, menopause and menopausal status-BMI interaction, women in the highest tertile of 18:3 n-3 had an odds ratio of 0.39 (95% CI = 0.19–0.78) compared to women in the lowest tertile (trend p = 0.01). Moreover, women in the highest tertile of 22:6 n-3 had an odds ratio of 0.31 (95% CI = 0.13–0.75) compared to women in the lowest tertile (trend p = 0.016). When considering the ratio of n-3 to n-6 PUFA, women in the highest tertile of 18:3 n-3/18:2 n-6 had an odds ratio of 0.41 (95% CI = 0.20–0.81) compared to women in the lowest tertile (trend p = 0.0004), and women in the highest tertile of long chain n-3/ total n-6 PUFA had an odds ratio of 0.33 (95% CI = 0.17–0.66) compared to women in the lowest tertile (trend p = 0.0002).
In our study, we examined whether n-3 PUFA, including both alpha-linolenic acid and DHA, may protect against the risk of breast cancer. For this purpose, individual n-3 PUFA levels of breast adipose tissue were used as a biomarker of past dietary intake of n-3 PUFA, to investigate the relation between exposure to n-3 fatty acids and breast cancer. We compared the fatty acid composition of the adipose tissue from 241 women with breast cancer, and from 88 controls with benign breast tumors. We found that the level of alpha-linolenic acid in adipose tissue displayed a moderate inverse association with breast cancer. The level of DHA, and the ratios 18:3n-3/18:2 n-6 and long-chain n-3/total n-6 PUFA were inversely associated with breast cancer.
We reported previously an inverse association between adipose tissue alpha-linolenic acid level and the risk of breast cancer in a case-control study conducted among women treated for breast tumors in Tours.27 This finding, however, remained to be confirmed, and we had to rule out the effect of the potential methodological biases derived from the comparison of a population of control patients to the population of breast cancer patients previously used for a prognostic study and not selected during the same period. For that purpose, we conducted an independent study, and not an extension of the previous one,27 among a population of new case and control patients selected at the University Hospital during a subsequent period. This new study allowed confirming the inverse link between alpha-linolenic acid and breast cancer risk. Moreover, we also provided new data on long-chain n-3 PUFA, which showed an inverse association between DHA and breast cancer risk.
The possibility that a low level of n-3 PUFA in adipose breast tissue in breast cancer patients reflected a reduced dietary intake of n-3 PUFA rather than metabolic interactions between breast adipose tissue and breast epithelium deserves consideration. In a previous study, we reported in breast cancer patients that n-3 PUFA levels were highly correlated between breast and subcutaneous adipose tissue, suggesting that their levels in breast reflect the body stores of these fatty acids.30 In a case-control study, the relationship between dietary fatty acids and breast cancer has been evaluated on the basis of fatty acid levels in both subcutaneous abdomen and the breast sites. Our study reported that the difference in the fatty acid composition between sites had no relationship to breast cancer.31 Finally, in an animal system of chemically-induced mammary tumors, we reported a lack of an association of 18:3 n-3 levels in mammary adipose tissue and tumor growth.32 This set of data suggests that a low level of n-3 PUFA as a consequence of interactions between carcinoma and breast adipose tissue is very unlikely.
Because samples of subcutaneous fat taken from different parts of the body did not show substantial differences within the same subject, the choice of where to take the sample may be based on practical consideration.24 In breast cancer patients, samples of fat are easily taken from the breast during surgery, allowing to avoid practical and ethical difficulties to obtain aspirates of subcutaneous fat from the buttocks during the hospitalization. The necessity of obtaining breast fat requires that the control group be chosen from women undergoing breast surgery, and the more practical solution is to consider patients with benign breast disease as hospital controls. The interpretation of our results might be hampered by the difficulty of considering women with benign breast disease as representative of the general population. Circumstantially, one study conducted in Finland reported that the composition of the diet in subjects with benign breast disease was similar to that reported for the general female population in some other Finnish studies.33
Our case-control study has several limitations. As in any case-control study, the possibility of patients selection bias should be considered. We believe that the choice of hospitalized control patients referred within a healthcare network identical to that of case patients should minimize this bias. The main limitation is represented by the difference of age between cases and controls, because long-chain n-3 PUFA levels increase with age. This bias should not influence the results because it was taken into account as a confounder in the regression models used. In contrast, the potential confounding effect of several covariates, including education, contraceptive use, alcohol consumption and tobacco smoking, was not controlled for in our study population and cannot be excluded. Finally, the possibility that cases display adipose tissue fatty acid changes as a consequence of breast cancer is very unlikely as discussed above.
Considered alpha-linolenic acid in isolation, we interpret our results as only a weak protective effect on the risk of breast cancer, which was limited to the highest levels in adipose tissue. The strongest associations that we found were observed for DHA, and for the ratios 18:3 n-3/18:2 n-6 or long-chain n-3/total n-6 PUFA. In contrast to our findings, 2 case-control studies conducted in North America found no consistent association between adipose tissue n-3 fatty acid levels and breast cancer risk.31, 34 In 1 case-control study conducted in Finland, DHA level in breast adipose tissue, along with its dietary intake, were significantly lower in breast cancer patients compared to control patients, suggesting a protective effect of DHA in the risk of breast cancer.33 In agreement with our findings, a case-control study conducted across 5 European countries found a decrease risk of breast cancer with an elevated ratio of long-chain n-3 PUFA to total n-6 PUFA in 4 of 5 centers.35 In our study, the ratio of alpha-linolenic acid to linoleic acid showed an inverse association with disease in one center and a positive association in three centers. These associations were weak (and not statistically significant) in all but one center. This set of data supports the idea that the protective effect of n-3 fatty acids depends on background levels of n-6 PUFA and that inhibition of eicosanoid production from n-6 PUFA precursors may be involved in the inhibitory effect of n-3 PUFA.
The reason for the discrepancy between the North American and European studies remains unresolved. One hypothesis is that the inconsistencies may be due in part to interactions between n-3 fatty acids and antioxidant compounds in diet, affecting their role in breast cancer risk.36, 37 Several experimental studies showed that the anti-oxidant vitamin E suppressed the inhibitory effect of n-3 PUFA (alpha-linolenic and long-chain n-3 PUFA) on tumor growth in different models of mammary carcinogenesis in rats38, 39 or on the proliferation of mammary tumor cells in vitro.40 In contrast, administration of oxidative compounds to diets high in alpha-linolenic acid led to an inhibition of tumor growth in chemically-induced mammary carcinogenesis.39 These data suggest that the inhibitory effect of n-3 PUFA on tumor growth is mediated, at least in part, by increased formation of lipid peroxidation products and emphasize on the potential importance of the interaction of anti-and pro-oxidants compounds with n-3 PUFA. These recent findings may explain the lack of an association between n-3 PUFA and breast cancer risk in the studies conducted in the United States,14, 15, 31, 34 a country in which supplementation with anti-oxidant vitamins is a common practice.41
Although we focused on the relation of n-3 fatty acids to estimated breast cancer risk, we also found inverse associations between the monounsaturated oleic acid and breast cancer. Adipose tissue levels of non-essential saturated and monounsaturated fatty acids are determined both by their intake from numerous dietary sources and by de novo endogenous synthesis and metabolism through desaturation and elongation pathways. Because metabolic factors influence levels of non-essential fatty acids in tissues, no association, or only weak association, has been reported between levels of saturated or monounsaturated fatty acids in adipose tissue and dietary intake levels.25 Epidemiological data on the association between estimated dietary intakes of saturated and monounsaturated fatty acids and the risk of breast cancer are inconsistent. Results from a combined analysis of case-control studies reported an increased risk of breast cancer with increased dietary intake of saturated and monounsaturated fat in post-menopausal women42 whereas a pooled analysis of cohort studies showed no significant associations between saturated or monounsaturated fatty acids and the risk of breast cancer.43 The EURAMIC case-control study found a strong inverse association for adipose tissue oleic acid level, which was largely confined to the Spanish population.35 Thus, the significance of the relation of non-essential fatty acids in adipose fat to breast cancer risk remains difficult to evaluate.
We examined the association of the saturated fatty acids 15:0 and 17:0 to breast cancer risk, because recent data indicated that the content of these fatty acids in adipose tissue might be a valid biological marker of long-term milk fat intake in free-living individuals in populations with high consumption of dairy products.44 Epidemiological data suggested a protective effect of a high consumption of milk in breast cancer risk,45 although other studies reported inconsistent results.46, 47 In our population study, we found only a significant positive association between 17:0 level in breast adipose tissue and breast cancer, which is limited to the highest levels in adipose tissue.
We found a positive association (but not significant) between linoleic acid level in adipose tissue and breast cancer, suggesting an increased risk of breast cancer with increased dietary intake of linoleic acid. This finding, in line with experimental data obtained among animal models of mammary carcinogenesis,11, 48, 49 appears to be somewhat inconsistent with recent conclusions on the role of linoleic acid in breast cancer risk. Although it has been suggested that long-term consumption of large amounts of linoleic acid might increase breast cancer risk, results from a meta-analysis of risk estimates from case-control and prospective cohort studies indicated no increased risk of breast cancer with high intakes of linoleic acid.50
In conclusion, our data based on fatty acid levels in breast adipose tissue suggest a protective effect of n-3 PUFA, somewhat moderate for alpha-linolenic acid, on breast cancer risk and support the hypothesis that the balance between n-3 and n-6 fatty acids plays a role in breast cancer. Further epidemiological and experimental studies are needed to evaluate the potential role of the interaction of n-3 fatty acids with anti- and pro-oxidative vitamins and n-6 PUFA in the development of breast cancer.
We warmly thank N. Simonsen for criticism and helpful comments; we acknowledge the participation of J. Lansac and surgical gynecologists by providing adipose tissue samples. We thank Dr. A. Reynaud-Bougnoux and Dr. G. Calais for patients' information, Dr. F. Fetissof and Dr. T. Lefrancq for reviewing pathology. This work was supported in part by grants from the French ministry of research (Nutrialis), La Ligue Nationale contre le Cancer (Comités d'Indre et Loire, Indre, Loir et Cher and Charente) and from CERIN. V. Chajès is the recipient of a grant from the Agence Régionale d'Hospitalisation, Région Centre, France. V. Maillard is a recipient of a grant from the French ministry (MENRT).