|Home | About | Journals | Submit | Contact Us | Français|
To examine the association between breast cancer risk and the fatty acid composition of phospholipids in prediagnostic serum samples.
We analyzed the fatty acid composition in 130 incident postmenopausal breast cancer cases and 257 matched controls nested within the β-Carotene and Retinol Efficacy Trial Cohort. The fatty acid composition was measured by gas chromatography. Multivariate-adjusted odds ratios and corresponding 95% confidence intervals for the risk of breast cancer were estimated using logistic regression. Stratified analysis was conducted by smoking status.
There were no associations with breast cancer risk for total saturated, monounsaturated, n-3, n-6, or trans fatty acids among all women. For individual fatty acids, we observed an inverse association with the trans linoleic acid, 18:2n6tt (p trend = 0.0002). Among current smokers, long-chain saturated fatty acids (22:0 and 24:0) and total 16:1 trans fatty acids were positively associated with the risk of breast cancer, whereas these fatty acids showed no association among former smokers.
Overall, we observed no significant association between serum phospholipid fatty acids and breast cancer risk, except for the trans linoleic acid isomer 18:2n6tt, which was unexpected. Our finding of a positive association of long-chain saturated fatty acids (22:0 and 24:0) and total 16:1 trans fatty acids with the risk of breast cancer only in current smokers may suggest an effect modification by smoking status. Our findings need to be replicated in future epidemiologic studies.
A long-standing hypothesis on the effect of fat on postmenopausal breast cancer remains inconclusive (1). A meta-analysis of observational studies measuring circulating fatty acid composition showed that α-linolenic acid was inversely associated with breast cancer risk in case-control studies (2), but was positively associated in cohort studies. Moreover, in cohort studies, most saturated fatty acids (SFAs) and monounsaturated fatty acids (MUFAs) also tend to be positively associated with postmenopausal breast cancer risk, whereas n-6 and n-3 fatty acids, except α-linolenic acid, were inversely associated. However, overall, the number of prospective studies investigating the association between circulating fatty acids and breast cancer risk is small. Furthermore, the current evidence suggests different isomers of trans fatty acids may have different biological effects (3,4); however, epidemiologic studies investigating the effect of each isomer on breast cancer risk are sparse. Therefore, we conducted a case-control study nested within the β-Carotene and Retinol Efficacy Trial (CARET) (5) to investigate the association between breast cancer risk and the fatty acid composition in prediagnostic serum.
Participants in this study were selected from CARET (5). In brief, CARET was a multicenter, double-blind randomized trial that investigated the effect of β-carotene and retinol supplementation on lung cancer and cardiovascular disease risk among heavy cigarette smokers and asbestos-exposed workers. The participants were continued to be followed as a cohort after the termination of the trial in 1996 due to increased risk of lung cancer observed in the supplementation arm (6).
Postmenopausal women included in this study were between 50 and 69 years of age at enrollment in 1985-94, had smoked at least 20 pack-years, were active smokers or had quit within 6 years prior to enrollment, were not taking any β-carotene or > 5,500 IU/day of vitamin A supplement, and were free of cancer other than non-melanoma skin cancer for at least five years prior to enrollment. Cases eligible for the present study were women who were diagnosed with breast cancer, and had a baseline fasting blood sample available at least 3 years before the diagnosis. Potential controls had an available blood sample collected at least three years prior to their last follow-up contact. Two controls were matched to each case by age at enrollment (5-year age group), race, study center (Irvine, Seattle or Portland), and year of enrollment (two-year intervals). The follow-up rate in this study was about 96%. The final analysis consisted of 130 cases and 257 controls, after excluding three controls for which laboratory analyses were unsuccessful.
Health status was updated routinely during the trial with three contacts (two phone calls and one clinic visit) per year targeted for most participants through 1997 and annual contacts (phone or mail) thereafter. Initial reports of breast cancer diagnoses as well as other cancers were primarily based on self-report and secondarily through cancer registry, state boards of health, and private health insurance plan data base (7). Once a breast cancer case was reported, the CARET Endpoints Review Committee reviewed medical records and pathology reports and confirmed the cancer case (5).
Fatty acid concentrations were measured by gas chromatography. Details of the procedure have been described elsewhere (8,9). In short, total lipids were extracted by the Folch method (10) and the phospholipid fraction was separated by one-dimentional thin-layer chromatography and transesterified using Lepage method (11), yielding fatty acid methyl esters (FAMEs). The prepared FAMEs in hexane were injected onto gas chromatography, the peaks for the corresponding fatty acids were identified using model mixtures (GLC-87, NIH-D, NIH-F, Nu Chek, Elysian, MN; trans 18:2 standard cat #47791, Supelco-Sigma-Aldrich Inc., St. Louis, MO) with serum sample confirmation from USDA lipid laboratory (Peoria, IL). The coefficients of variation (CVs) of the quality control samples ranged from 0.9 to 1.3%, for major fatty acids (≥5%) and from 1.3 to 9.9% for the remaining fatty acids (<5%). The CVs for the trans fatty acids were as follows: 16:1n-7t (17.7%), 18:1n-10-12t (15.6%), 18:1n-9t (8%), 18:1n-8t (18.6%) 18:1n-7t (8.9%), 18:1n-6t (3.5%), 18:2tt (17.1%), 18:2ct (17.3%), 18:tc (6.7%). Laboratory analyses were conducted by technicians blinded to case control status.
The fatty acid composition was expressed as the weighted percentage of the total fatty acids. Given that n-3 and n-6 fatty acids compete with the same desturases to be converted into more bioavailable forms, we also used ratios and saturation index (SI) such as the ratio of total n-3 and n-6 fatty acids and SI n-9 (the ratio of stearic to oleic acids).
To compare characteristics between cases and controls, we performed t-test for continuous variables and χ2-test for categorical variables. In order to estimate the odds ratios (OR) and the corresponding 95% confidence intervals (CIs) for risk of breast cancer, logistic regression models were applied while adjusting for all matching criteria [i.e., age, study center, and year of the enrollment] as well as intervention arm, smoking status at baseline and at blood draw (current vs. former smokers), body mass index (BMI) and alcohol use. Participants were categorized into quartiles by applying the quartile cut-off values of each fatty acid composition among controls and assigned the median value of the quartile to test for a linear trend. The analyses were repeated for subgroups by smoking status at the time of blood draw considering potential differential composition of fatty acids between current and past female smokers (12). An interaction between smoking status (current vs. former) and fatty acid composition, indicated by the median value of the quartile, was also tested by assessing the p-value of an interaction term included in logistic regression models. All statistical analyses were conducted by SAS version 9.1 (SAS Institute, Inc., Carey, NC).
At the time of blood draw, about 60% of women in this study were current smokers, while the others were former smokers (Table 1). Alcohol consumption and the proportion of obesity were higher among cases than controls, while smoking status and educational attainment was similar between cases and controls. On average, women were diagnosed 4.4 years after the blood draw. Except for two controls and one case who were Hispanic, all participants were non-Hispanic Caucasian.
Among controls, about 44% of the total fatty acids were composed of SFAs, followed by n-6 PUFAs (35%), MUFAs (13%), n-3 PUFAs (4%) and trans fatty acids (2%) (Table 2). Of the individual fatty acids, the highest fractions were contributed by palmitic acid (26%) and linoleic acid (20%). There was no difference in the fatty acid composition between the supplementation and the placebo arms among controls (data not shown).
Individual and total SFAs and MUFAs compositions were not significantly associated with risk of breast cancer (Table 3). Among the n-3 fatty acids, α-linolenic acid exhibited a significant 50% lower risk of breast cancer at the second quartile than the first quartile; however, point estimates of the third and fourth quartiles were not significant. Total n-3 fatty acids and the other n-3 fatty acids, individually or when the ratio was taken (i.e., EPA: DHA), were not associated with breast cancer risk. For the n-6 fatty acids, we observed a suggestive non-significant inverse association with γ-linolenic acid (p trend = 0.09). None of the other n-6 fatty acids was associated with breast cancer risk, when assessed individually, as total n-6 fatty acids or as a ratio.
We observed no association between risk of breast cancer and total trans fatty acids of 16:1, 18:1, 18:2, and all combined (p trend = 0.26, 0.62, 0.94, and 0.55, respectively). When major isomers of 18:1 and 18:2 trans fatty acids were examined separately, the trans linoleic acid, 18:2n6tt was significantly inversely associated with breast cancer risk; the risk difference between the highest and the lowest quartiles was 68% (OR = 0.32; 95% CI = 0.17-0.61; p trend = 0.0002). In contrast, no significant trend was observed for 18:2n6ct (p for trend = 0.44) and for 18:2n6tc isomer (p trend = 0.44), although those in the third quartile for 18:2n6tc isomer had a significant 2.3 times greater risk than those at the first quartile.
In the stratified analyses by smoking status long-chain SFAs (22:0 and 24:0) and total 16:1 trans fatty acids were positively associated with the risk (p trend = 0.03, 0.04, and 0.03, respectively) in current smokers (Table 4). However, the interaction effect with smoking status was borderline significant only for 16:1 trans fatty acids (p interaction = 0.05). Moreover, vaccenic and α-linolenic acids were suggestively inversely associated with breast cancer risk (p trend = 0.06 and 0.07, respectively). In contrast, no association with fatty acids was observed among former smokers.
Our nested case-control study among postmenopausal smokers did not provide evidence for associations with breast cancer risk for total of saturated, monounsaturated, n-3, n-6, or trans fatty acids. By individual fatty acids, we unexpectedly observed an inverse association with the trans linoleic acid isomer, 18:2n6tt. In stratified analyses by smoking status, long-chain SFAs (22:0 and 24:0) and total 16:1 trans fatty acids were positively associated with breast cancer risk among current smokers, whereas none of the fatty acids was associated among former smokers.
Previous epidemiologic studies on the association between total trans fatty acids and the risk of breast cancer yielded mixed results. Two observational studies reported a positive association (13,14). In contrast, a favorable breast cancer prognosis was associated with a higher trans fatty acids in adipose tissue (15). Consistent with our finding, four observational studies reported no association (16-19). Furthermore, separate analysis by major isomers of trans fatty acids unexpectedly showed an inverse association with one trans linoleic acid isomer (18:2n6tt), while the other two isomers (18:2n6ct and 18:2n6tc) were not associated or, if any, possibly positively for 18:2n6tc. This finding does not corroborate with a recent case-control study that reported no association with 18:2n6tt isomer (19), nor with putative negative effects of trans fatty acids derived from partially hydrogenated vegetable oils on cancer (3). Partially hydrogenated vegetable oils are considered as the major dietary source of the trans linoleic acid. To explore if specific foods contribute to 18:2n6tc compared with the other trans fatty acids, we calculated correlations for all individual and combined trans fatty acids isomers with several foods and food groups. However, we did not observe any moderate or strong correlations for any trans fatty acids isomer; the strongest correlation was 0.18 for margarine and total trans 18:1 fatty acids. Accordingly, and given that the 18:2n6tt isomer contributes only 0.03% to the total fatty acid composition, this unexpected inverse association might be a chance finding.
Our finding of no association between the activity of stearoyl-CoA desaturase, indicated by SI n-9, and breast cancer risk is consistent with several studies (20-23) while two studies reported an inverse association (24,25). Since it was postulated that factors other than fatty acids (e.g., insulin sensitivity and estrogen levels) may affect the stearoyl-CoA desaturase activity (24), the observed association may have been modified by these factors. However, due to lack of measurement on these factors, we were unable to further investigate this potential interaction.
The underlying biological mechanism for a potential interaction of long-chain SFAs (22:0 and 24:0) and 16:1 trans fatty acids with smoking status is unknown. However, we previously showed that a moderate-fat diet (34% of energy from fat) resulted in higher fractions of 22:0 and 24:0 fatty acids as well as 16:1 trans fatty acids in plasma phospholipids than a low-fat diet (17% of energy from fat) in an 6-week intervention study (9). Hence, it is interesting to speculate if dietary habits of former smokers (e.g., by lowering fat intake) could have changed when they stopped smoking. These findings need to be replicated.
Strengths of our study include our prospective design with up to 14 years of follow-up and the use of an objective measure for the fatty acid composition. Furthermore, this is the first study among cigarette smokers that examined the effect of current and past smoking on the association between fatty acid composition and breast cancer risk. Conversely, the generalizability of our findings to a healthy population may be limited. This is also one of the very few studies that examined the effect of the different trans fatty acid isomers on breast cancer risk.
A limitation of our study is the small sample size with limited power, especially for subgroup analysis. Another potential limitation is our inability to adjust for some of the established risk factors for breast cancer (i.e., age at menarche and hormone regimen) in the analyses due to lack of information. Two early studies using biomarker of fatty acids (13,25) reported comparable risk estimates with and without adjustment for breast cancer risk factors. Thus, this limitation is probably less likely to affect our risk estimates.
Overall, we observed no significant association between fatty acids in serum phospholipids and breast cancer risk, except for the trans linoleic acid, 18:2n6tt. Our finding of a positive association with long-chain SFAs (22:0 and 24:0) and total 16:1 trans fatty acids only in current smokers is intriguing and may suggest an effect modification by smoking status. Our findings need to be confirmed in larger epidemiologic studies in order to further elucidate the effect of specific fatty acids, particularly isomers of trans fatty acids, on the risk of breast cancer.
We thank the CARET participants and research staff for their effort and time. All the authors contributed to the interpretation of findings and commented on the manuscript. YT and UP guided the statistical analysis and drafted the manuscript, IBK oversaw the laboratory measurement, IBK and MLN provided expertise on fatty acid, YT, SS and MB conducted data analysis, SS and MB managed the data set, MT and GEG designed the study, and GEG obtained the funding for this study. This study was funded by National Cancer Institute (NCI CA063673).
Financial support: This study was funded by National Cancer Institute (NCI CA063673).