Serum Phospholipids Fatty Acids and Breast Cancer Risk by Pathological Subtype

This study evaluates whether serum phospholipids fatty acids (PL-FAs) and markers of their endogenous metabolism are associated with breast cancer (BC) subtypes. EpiGEICAM is a Spanish multicenter matched case-control study. A lifestyle and food frequency questionnaire was completed by 1017 BC cases and healthy women pairs. Serum PL-FA percentages were measured by gas chromatography-mass spectrometry. Conditional and multinomial logistic regression models were used to quantify the association of PL-FA tertiles with BC risk, overall and by pathological subtype (luminal, HER2+ and triple negative). Stratified analyses by body mass index and menopausal status were also performed. Serum PL-FAs were measured in 795 (78%) pairs. Women with high serum levels of stearic acid (odds ratio (OR)T3vsT1 = 0.44; 95% confidence interval (CI) = 0.30–0.66), linoleic acid (ORT3vsT1 = 0.66; 95% CI = 0.49–0.90) and arachidonic to dihomo-γ-linolenic acid ratio (OR T3vsT1 = 0.64; 95% CI = 0.48–0.84) presented lower BC risk. Participants with high concentrations of palmitoleic acid (ORT3vsT1 = 1.65; 95% CI = 1.20–2.26), trans-ruminant palmitelaidic acid (ORT3vsT1 = 1.51; 95% CI = 1.12–2.02), trans-industrial elaidic acid (ORT3vsT1 = 1.52; 95% CI = 1.14–2.03), and high oleic to stearic acid ratio (ORT3vsT1 = 2.04; 95% CI = 1.45–2.87) showed higher risk. These associations were similar in all BC pathological subtypes. Our results emphasize the importance of analyzing fatty acids individually, as well as the desaturase activity indices.


Introduction
Breast cancer (BC) is the most frequently diagnosed cancer in women and the leading cause of cancer-related deaths worldwide [1]. In Spain, age-standardized incidence rates have increased in recent decades [2]. In 2018, it reached a rate of 101 cases per 100,000 [3], which represents 28.7% of all newly diagnosed cancers among women [1].
There is growing evidence for a plausible role of dietary factors in BC risk, but evidence for fatty acids intake is too limited to draw conclusions [4]. Several studies have shown a positive association with the intake of saturated fatty acids (SFAs) [5][6][7][8]. The relationship between monounsaturated fatty acids (MUFAs) and BC risk is more conflicting, and seems to depend on the contributing food source, such as olive oil and margarines [5,7]. On the other hand, higher consumption of marine omega-3 polyunsaturated fatty acids (n-3 PUFAs) as well as elevated intake of n-3/n-6 PUFA ratio seem to exert a protective effect on this tumor [5,7,[9][10][11][12]. To assess the usual intake of fatty acids, these studies were based on commonly-used dietary questionnaires. However, self-reported methods are limited by systematic and random measurement errors [13]. Moreover, only some fatty acids, particularly those not endogenously synthesized, are good biomarkers of the diet, and are estimated by food frequency questionnaires. Therefore, the fatty acid profile in biological samples is a more accurate measure that reflects the interaction between dietary intake, de novo lipogenesis and the activity of different enzymes involved in this metabolic process [14]. Although data from a meta-analysis of prospective studies found no association between serum levels of SFAs, MUFAs and PUFAs and risk of BC [15], more recent studies have described positive associations with circulating levels of several SFAs [16,17], palmitoleic acid [18], some trans fatty acids [17][18][19] and the ratio of palmitoleic acid to palmitic acid (DI 16 ) [18], as well as an inverse association with n-3 PUFAs among overweight/obese women [17].
BC represents a heterogeneous disease, characterized by the presence versus absence of specific receptors on the tumor cells: hormonal receptors (HR) (estrogen receptors (ER) and progesterone receptors (PR)) and human epidermal growth factor receptor 2 (HER2) [20]. It has been described that the expression of fatty acid metabolism proteins differs with respect to intrinsic molecular BC subtype [21]. To date, very few studies have analyzed the association between circulating fatty acids and BC risk defined by hormonal receptor status [16,18], and, to our knowledge, none of them have studied this relationship taking into account the HER2 status. The objective of the present study was to investigate the association between fatty acids assessed in serum phospholipids and BC risk, as well as to evaluate whether this association differed according to specific pathological BC subtype.

Study Population
Data came from a multicenter case-control study on female BC, including incident cases and individually matched healthy controls. Between 2006 and 2011, 1017 incident cases of BC were recruited in the Oncology departments of 23 hospitals belonging to the Spanish Breast Cancer Research Group (GEICAM) (https://www.geicam.org/). These hospitals are located in nine of the 17 Spanish Autonomous Regions, where 78% of the Spanish population resides. The oncologists invited the patients to participate at the time of diagnosis. Participants were between 18 and 70 years old, with no history of BC. They should reside in the area attended by the hospital, and they should be able to answer the epidemiological questionnaire. Each patient was matched with a healthy control of a similar age (±5 years), generally selected by the patient. This could be a friend, a neighbor or a work colleague residing in the same city. Of the 2494 women who were initially invited to participate in the EpiGEICAM study (1362 patients and 1132 controls), 1017 case-control pairs participated; thus, the overall participation rate was 82% (75% in BC cases and 90% in controls).
During the first three months after diagnosis, patients and controls filled a structured questionnaire collecting anthropometric and demographic data, personal and family background, gynecological, obstetric and occupational history, smoking status, physical activity and diet. Postmenopausal status was based on self-reported information and was defined as absence of menstruation in the past 12 months. Dietary intake during the preceding five years was estimated using a 117-item semi-quantitative food frequency questionnaire (FFQ), similar to the Harvard questionnaire [25], and adapted to and validated in different Spanish adult populations [26]. The responses for each food item were converted to mean daily intake (in grams) and total energy intake (in kcals/day). Verification and data entry were carried out at GEICAM headquarters. The accuracy of the information recorded in the database was verified randomly by selecting and reviewing 10% of the questionnaires. Additionally, a basal serum sample was collected from patients and healthy controls for phospholipids fatty acid (PL-FA) analysis.
All subjects gave their informed consent for inclusion before participating in the study. This was conducted in accordance with the Declaration of Helsinki, and the protocol was approved on 25 May 2006 by the Ethics Committee of Hospital Clínico San Carlos (EpiGEICAM-01, Committee internal code E-06/141). Patient information was anonymized and de-identified prior to analysis. Further details regarding the study design have been previously published [27,28].

Analysis of Serum PL-FAs
PL-FAs were determined by using the protocol proposed by Criado-Navarro et al. [29], which is based on the isolation of PLs using 30 mg HybridSPE ® cartridges from Supelco (Bellefonte, PA, USA), derivatization of the resulting extract to convert PL-FAs into their more volatile fatty acid methyl esters of phospholipids (PL-FAMEs) and gas chromatography coupled to mass spectrometry (GC-MS) analysis. The NIST Mass Spectral Search Program v.11.0 (NIST, Washington, DC, USA) was used for spectral search (Mainlib and Replib libraries). Tentative identification was reported when the correlation between experimental and database spectra was above 0.75 in normal search mode. Confirmatory analysis was carried out by analysis of a FAMEs multistandard from Sigma-Aldrich (Steinheim, Germany).

Statistical Methods
Serum PL-FA levels were analyzed for 795 case-control participant pairs. Given the matching nature of our data, missing values were imputed using Multiple Imputation by Chained Equation [32]. This methodology was used to fill incomplete information for height (9.2%), weight (2.8%), waist circumference (2.6%), age at menarche (0.8%), age at first birth (4.6%), hormone replacement therapy (HRT) use (4.9%), smoking (0.4%), alcohol consumption (2.9%), overall physical activity during the previous year (8.5%) and caloric intake (4.4%). Imputation models also included the following potential explanatory variables with complete data: case-control status, age, parity, menopausal status, educational level, hypercholesterolemia, previous benign breast problems, family history of BC and all serum PL-FAs. Five imputed data sets were generated using the multiple imputation procedure implemented in STATA [33]. A complete BC case analysis was also performed to check the validity of the imputation.
Descriptive characteristics of participants were summarized for patients and controls. We calculated absolute figures and percentages for categorical variables and means and standard deviations for continuous variables. Significant differences between patients and controls were tested using Pearson chi-square for categorical and Student's t-test for continuous variables.
To analyze the association between BC and relative percentage of serum PL-FAs, the latter were divided into tertiles, based on the distribution of serum levels in controls. Second and third tertiles were compared with the first tertile (reference) using conditional logistic regression models adjusted for educational level (no formal school or first grade; second grade or vocational training; university graduate), body mass index (BMI) one year prior to the interview, menopausal status (these two variables were included in the models with the corresponding interaction term), age at menarche (continuous), age at first birth (<20, 20-24; 25-29; ≥30; nulliparous), HRT use (never; ever), alcohol consumption (g/day), self-assessed physical activity during the previous year (sedentary or slightly active; moderately active; active or very active), history of benign breast disease (no; yes), family history of BC (none; second degree; first degree) and caloric intake (Kcal/day). The linear trend across tertiles was also tested with the Wald test. In addition to categorical analyses, each group of serum PL-FAs was also modelled through a restricted quadratic spline with knots at the 5th, 50th and 95th percentiles [34]. These restricted quadratic splines allowed two different quadratic trends on either side of the median PL-FAs that were restricted to be linear below the 5th percentile and above the 95th percentile; thus, they could reproduce a large variety of smooth dose-response curves while avoiding implausible shapes at extreme PL-FA levels. As sensitivity analyses, we conducted stratified analyses by BMI and menopausal status, comparing the third tertile of each PL-FA with the first tertile. The potential effect modification was evaluated including in the models the corresponding interaction terms. To take into account the problem of multiple testing, p-values were suitably adjusted using the false discovery rate as proposed by Benjamini and Hochberg [35].
Finally, multinomial logistic regression models were used to evaluate the association of serum PL-FAs with each of the aforementioned intrinsic BC subtypes. These models were adjusted for age (continuous), recruiting hospital and the same set of potential confounders described above. Heterogeneity of effects was tested using the Wald test, comparing the coefficients obtained for the different cancer subtypes. All statistical analyses were performed with STATA/MP 15.1 software (StataCorp LLC, College Station, TX, USA). Table 1 shows the main characteristics of the 795 case-control pairs with information on the relative concentrations of serum PL-FAs. Their mean age was 50 years. Compared with controls, patients had lower educational level, higher proportion of premenopausal women and postmenopausal overweight women, higher caloric intake, higher frequency of previous benign breast diseases and more relatives with BC. Abbreviations: HR+ = hormone receptor positive tumors (ER+ and/or PR+, with HER2-); HER2+ = human epidermal growth factor receptor 2 positive tumors; TN = triple negative tumors. a p-value resulting from Pearson Chi-Square test (completed variables). b p-value resulting from conditional logistic regression models (imputed variables). c Two breast cancer cases could not be classified. Table 2 shows mean percent of each serum PL-FA (percent of total) in patients and controls, as well as BC risk by tertiles of PL-FAs. When we compared women in the third tertile with women in the first tertile, those with high relative concentrations of stearic acid (odds ratio (OR) T3vsT1 = 0.44; 95% confidence interval (CI) = 0.30-0.66) and linoleic acid (OR T3vsT1 = 0.66; 95% CI = 0.49-0.90) presented lower BC risk, while women with high levels of palmitoleic acid showed higher risk (OR T3vsT1 = 1.65; 95% CI = 1.20-2.26). Associations were also observed with other PL-FAs whose relative serum concentrations were much lower. In this way, we found a protective association with behenic acid, as well as a positive association linked to pentadecanoic, gondoic and γ-linolenic acids. On the other hand, women with high relative concentrations of trans fatty acids showed a significant increase in BC risk, both of animal origin (OR T3vsT1 = 1.44; 95% CI = 1.07-1.93)-mainly due to the action of palmitelaidic acid-and of industrial origin (OR T3vsT1 = 1.38; 95% CI = 1.04-1.85)-mainly due to the action of elaidic acid. Finally, regarding desaturation indices, the ratio of oleic acid to stearic acid (SCD-18) (OR T3vsT1 = 2.04; 95% CI = 1.45-2.87) and the ratio between dihomo-γ-linolenic and linoleic acids (OR T3vsT1 = 1.35; 95% CI = 1.02-1.78) were positively associated with BC risk, while the ratio between arachidonic and dihomo-γ-linolenic acids showed and inverse relationship (OR T3vsT1 = 0.64; 95% CI = 0.48-0.84).

Results
When these analyses were repeated, stratifying by menopausal status and BMI (Table 3), no significant heterogeneity was observed in the associations. However, we observed that the association of trans fatty acids of animal origin was particularly intense in postmenopausal women (OR T3vsT1 = 1.87; 95% CI = 1.20-2.92), while the BC risk associated with those of industrial origin was higher in premenopausal women (OR T3vsT1 = 1.67; 95% CI = 1.14-2.47). An excess risk of BC associated with elevated serum cis-MUFAs levels was observed only in postmenopausal women and in overweight/obese women. In the latter women, there was also an increased risk associated with trans ruminant fatty acids (OR T3vsT1 = 1.73; 95% CI = 1.14-2.64), a stronger inverse association linked to the FADS1 desaturation index (OR T3vsT1 = 0.52; 95% CI = 0.35-0.77), and a protective relationship that was not observed in women with BMI < 25 Kg/m 2 , linked to docosahexaenoic acid (DHA) (OR T3vsT1 = 0.71; 95%CI = 0.47-1.06).   Abbreviations: BMI, body mass index; CI, confidence interval; FADS, fatty acid desaturase; MUFAs, monounsaturated fatty acids; n-3 PUFAs, omega-3 polyunsaturated fatty acids; n-6 PUFAs, omega-6 polyunsaturated fatty acids; OR, odds ratio; SCD, stearoyl-CoA desaturase; SFAs, saturated fatty acids. a Adjusted for educational level, body mass index, menopausal status (these two variables were included in the models with the corresponding interaction term), age at menarche, age at first birth (with a category of history of nulliparous), hormone replacement therapy use, alcohol consumption, physical activity, previous benign breast problems, family history of breast cancer and energy intake. Adjusted for age and hospital by design. b p-value for heterogeneity. c Includes 16:1 n-7t and 18:1 n-7t. d Includes 18:1 n-9t and 18:2t.
In general, the association between PL-FAs and BC risk by pathological subtype was similar to that observed for total BC risk (Table 4), although the associations with HER2+ tumors and TN tumors failed to attain statistical significance, probably due to the smaller number of cases. In particular, high relative concentrations of the saturated pentadecanoic fatty acid were associated with increased risk of HR+ and HER2+ tumors, but with a decreased risk of TN tumors (p for heterogeneity = 0.098). On the other hand, serum erucic acid concentrations were inversely associated with HR + tumors and, to a lesser extent, with TN tumors, while a positive association with HER2 + tumors was observed (p for heterogeneity = 0.030). Figure 1 depicts the dose-response curve for the different groups of fatty acids. As can be seen, BC risk increases as the relative concentrations of trans fatty acids increase. A slight increasing trend is also observed with cis-MUFA levels and a slight decreasing trend is associated with n-6 PUFAs.
Nutrients 2020, 12, x FOR PEER REVIEW 12 of 18 Figure 1 depicts the dose-response curve for the different groups of fatty acids. As can be seen, BC risk increases as the relative concentrations of trans fatty acids increase. A slight increasing trend is also observed with cis-MUFA levels and a slight decreasing trend is associated with n-6 PUFAs. polyunsaturated fatty acids, with knots at their 5th, 50th and 95th percentiles. The reference value for each group of fatty acids was set at the median of the first tertile (56.4%, 10.9%, 1.3%, 0.1%, 26.6% and 2.4% respectively). Odds ratios were obtained from conditional logistic regression models adjusted for educational level, body mass index one year prior to the interview, menopausal status, age at menarche, age at first birth, hormone replacement therapy use, alcohol consumption, self-assessed physical activity during the previous year, history of benign breast disease, family history of breast cancer and caloric intake. Bars represent the histogram of each group of fatty acids.

Discussion
In the present study, we found evidence that higher relative concentrations of phospholipid monounsaturated palmitoleic and gondoic acids, trans fatty acids and a high SCD-18 desaturation index were associated with increased BC risk, while increased concentrations of stearic acid, linoleic acid and FADS1 index were inversely associated. This relationship was similar in all BC pathological subtypes.
Stearic acid is a long-chain saturated fatty acid found primarily in foods typical of a Western diet, including meat, poultry, fish, grain products, chocolate and milk fats [36,37]. The protective association provided by this fatty acid has been previously reported in premenopausal women by Saadatian-Elahi et al. in a meta-analysis of fatty acids in biological samples and BC risk [38]. The influence of stearic acid on the inhibition of tumor cells in vitro and tumor development in vivo has been demonstrated for many years [39]. More recently, stearate has been shown to induce apoptosis of human BC cells involving protein kinase C in the signaling cascade [40] and, in addition, to reduce visceral adipose tissue by apoptosis of preadipocytes [37]. polyunsaturated fatty acids, with knots at their 5th, 50th and 95th percentiles. The reference value for each group of fatty acids was set at the median of the first tertile (56.4%, 10.9%, 1.3%, 0.1%, 26.6% and 2.4% respectively). Odds ratios were obtained from conditional logistic regression models adjusted for educational level, body mass index one year prior to the interview, menopausal status, age at menarche, age at first birth, hormone replacement therapy use, alcohol consumption, self-assessed physical activity during the previous year, history of benign breast disease, family history of breast cancer and caloric intake. Bars represent the histogram of each group of fatty acids.

Discussion
In the present study, we found evidence that higher relative concentrations of phospholipid monounsaturated palmitoleic and gondoic acids, trans fatty acids and a high SCD-18 desaturation index were associated with increased BC risk, while increased concentrations of stearic acid, linoleic acid and FADS1 index were inversely associated. This relationship was similar in all BC pathological subtypes.
Stearic acid is a long-chain saturated fatty acid found primarily in foods typical of a Western diet, including meat, poultry, fish, grain products, chocolate and milk fats [36,37]. The protective association provided by this fatty acid has been previously reported in premenopausal women by Saadatian-Elahi et al. in a meta-analysis of fatty acids in biological samples and BC risk [38]. The influence of stearic acid on the inhibition of tumor cells in vitro and tumor development in vivo has been demonstrated for many years [39]. More recently, stearate has been shown to induce apoptosis of human BC cells involving protein kinase C in the signaling cascade [40] and, in addition, to reduce visceral adipose tissue by apoptosis of preadipocytes [37].
Palmitoleic acid is an n-7 MUFA biosynthesized from palmitic acid by the action of the SCD-1 enzyme in the liver. Although not commonly found in food, it is present in macadamia nuts, blue-green algae and marine oils [41,42]. Elevated plasma palmitoleic acid levels have been associated with increased risk of general [18,43,44] and post-menopausal BC [38]. The SCD-1 (or ∆9-desaturase) is an enzyme that converts SFAs into MUFAs in the liver. Several studies have involved the expression and activity of this enzyme in the development and progression of various types of cancer, including BC [30,45]. Thus, some epidemiological studies have shown that high serum concentrations of oleic acid together with low levels of stearic acid (high SCD-18 desaturation index) [46,47], as well as high levels of palmitoleic acid to the detriment of palmitic acid concentrations (high SCD-16 index) [18,43,44]-both indicators of high SCD-1 activity-are associated with an increased BC risk. Although we have not detected differences by histological type, it has been described that SCD-1 expression is greater in HR+ and HER2+ BC subtypes than in TN breast tumors [48].
Linoleic acid, the essential fatty acid of the n-6 PUFA family, is found in typical foods of the Western diet, such as vegetable oils, nuts, seeds, meat and eggs [49]. The role of linoleic acid in cancer development remains unclear. A systematic review of randomized controlled trials concluded that there is no evidence to show that the addition of linoleic acid to the diet increases the concentration of inflammatory markers [50]. In relation to BC, Zhou et al., in a meta-analysis of 12 prospective studies, showed that both linoleic acid intake and serum levels of this fatty acid were associated with decreased BC risk, although none of the associations was statistically significant [49]. This inverse relationship, and in line with what we observed in our results, was more evident among premenopausal women than among postmenopausal women.
Since our results show an inverse association with linoleic acid, a borderline positive association with serum levels of dihomo-γ-linolenic acid, and no relationship with arachidonic acid, the FADS2 desaturation index (indicator of ∆6-desaturase activity) was positively related to BC, while the FADS1 index (indicator of ∆5-desaturase expression) showed an inverse association. Although an increased ∆6-desaturase activity has been found in human BC tissue [51], these enzymes do not appear to influence the development of BC [31]. Only two previous studies have observed an association with this tumor in the opposite direction to our results [52,53].
Regarding trans fatty acids, a recent meta-analysis indicated that while dietary intake of these fatty acids was not associated with BC risk, a significant positive association was observed with serum trans fats in postmenopausal women [19]. We have detected an increased BC risk associated with high serum concentrations of trans-palmitelaidic and elaidic acids. These are precisely the two trans fatty acids that Chajes et al. found to be positively associated with this tumor in the E3N-EPIC cohort [43]. Palmitelaidic acid is a natural trans fatty acid produced by microbial hydrogenation of palmitoleic acid in the rumen of ruminants and, consequently, is present in all fats of these animals [54], although it can also be produced endogenously by the partial β oxidation of dietary vaccenic acid [55]. Although studies evaluating the relationship between BC and levels of this trans fatty acid in biomarkers are scarce, positive associations have been described in the serum of current postmenopausal smokers [56], in the erythrocytes membrane of overweight/obese women from the Nurses' Health Study II [17], and in the plasma of a cohort of Australian women, although in the latter case the association was only found with ER-/PR-tumors [16]. Elaidic acid, the trans form of oleic acid, is the main industrial trans fatty acid that is produced during the partial hydrogenation of vegetable oils, and is a reliable biomarker of highly processed foods among European populations [57]. Chajes et al., in a study that evaluated the variation in the percent of weight change at five years, found that doubling the plasma concentration of elaidic acid was associated with a decreased risk of weight loss and an increased risk of weight gain during the five-year follow-up in women [58]. Other studies, although they have not been able to separate trans fatty acids of industrial origin into their different individual isomers, have described a positive association between this type of trans fatty acids and BC risk in general [17], and ER-tumors in particular [18]. In contrast, elaidic acid was not associated with BC risk in another prospective Italian study [47].
There are several possible biological hypotheses by which trans fatty acids could influence the development of BC. These fatty acids are positively associated with markers of inflammation [59], which may be related to breast carcinogenesis [4]. Other possible mechanisms proposed are the elevation of circulating estrogen levels, induced oxidative stress, and body weight regulation [19].
Despite these results, the mean content of trans fatty acids in food products in Spain is low, and has been declining since 2010 [60]. Although Spanish legislation does not yet regulate its content, in May 2019 a new regulation of the European Commission came into force limiting the content of trans fats of industrial origin in food to a maximum of 2 g per 100 g of fat [61], regulation that has a transitional period of two years so that industries can adapt. However, this regulation only regulates trans fats of industrial origin and, precisely, the group of foods with the highest content of trans fatty acids in Spain are dairy products, which contain fatty acids of natural origin [60].
One of the main strengths of this study is that histologically confirmed patients and controls were recruited in nine Spanish regions located throughout the Spanish geography, which allowed us to have a broad representation of fatty acid intake in Spain. To our knowledge, this is the first study exploring the association between serum PL-FAs and BC risk considering HER2 status separately. The large number of premenopausal participants allowed us to adequately explore the association of serum PL-FAs with BC risk in these women. Another important strength is the wide range of fatty acids measured in serum phospholipids. In addition, this biomarker, compared to traditional self-reported assessment methods, constitutes a more objective measure of the bioavailable amount of fatty acids. Finally, The GC-MS analysis represents a powerful platform for the analysis of fatty acids in complex samples with exceptional sensitivity [62].
Several limitations should also be addressed. First, the study relied on self-reported data, and recall bias may be present in some exposures. However, we believe that this misclassification is most likely non-differential. Second, the fact of having selected friends as controls of BC cases could imply greater similarity in lifestyles, including eating habits and fat intake. However, this possible bias would imply an underestimation of the effects studied and, therefore, the associations reported here would underestimate the real ones. Third, although most of the established risk factors were taken into account, residual unmeasured confounders associated with serum PL-FA levels, such as cholesterol, triglycerides, insulin or other dietary factors, may have interfered with the detected associations. Fourth, chance findings could be occurred due to the large number of tests performed. However, to control for multiple comparisons, the false discovery rate (q values) was calculated using the Benjamini and Hochberg correction, a powerful test but at the cost of a higher number of Type I errors. Fifth, PL-FA levels are expressed as a percentage of total fatty acids, reflecting relative concentrations. Therefore, changes in one PL-FA affect the relative concentration of the others, particularly those representing lower proportions. Sixth, statistical power was limited in the analyses by menopausal status, by BMI and by pathological subtypes, since TN tumors are relatively uncommon in Spanish population [63]. Seventh, it is possible that metabolic alterations prior to BC diagnosis could have contributed to altering serum PL-FA levels in our patients. However, we believe that this is unlikely, since this effect would be observed mainly in the more advanced stages of the disease, and we have not found differences by pathological subtype. Eighth, our study is based on a single baseline measurement for blood levels, thus, we have not been able to evaluate changes over time. Finally, although there are other biological markers, such as adipose tissue, more appropriate to reflect long-term dietary intake [64], the fatty acid composition of serum phospholipids can be a useful tool in epidemiological studies [65], and it is considered a convenient alternative for the medium-to long-term assessment of habitual availability and metabolism of fatty acids in large-scale epidemiological studies [66].

Conclusions
In conclusion, our results suggest that SFAs, MUFAs and PUFAs cannot be considered as homogeneous groups when studying their association with BC risk, but should be analyzed individually. Furthermore, serum PL-FA levels are not only a reflection of what we eat, but are also the result of the activity of different enzymes involved in a variety of metabolic processes. Our results suggest that higher intake of stearic and linoleic acids, as well as a reduction in the consumption of palmitoleic acid and trans fatty acids, could be a good strategy for BC prevention. Additional studies are needed to better understand the influence of PL-FAs on BC development.