European Journal of Clinical Nutrition (2014) 68, 1088–1094 © 2014 Macmillan Publishers Limited All rights reserved 0954-3007/14 www.nature.com/ejcn
ORIGINAL ARTICLE
Baseline patterns of adipose tissue fatty acids and long-term risk of breast cancer: a case-cohort study in the Danish cohort Diet, Cancer and Health JA Schmidt1,2, A Gorst-Rasmussen3, PW Nyström4, JH Christensen5, EB Schmidt3, C Dethlefsen3, A Tjønneland6, K Overvad2,3 and CC Dahm2 BACKGROUND/OBJECTIVES: The evidence regarding fatty acids and breast cancer risk is inconclusive. Adipose tissue fatty acids can be used as biomarkers of fatty acid intake and of endogenous fatty acid exposure. Fatty acids in adipose tissue are correlated owing to common dietary sources and shared metabolic pathways, which group fatty acids into naturally occurring patterns. We aimed to prospectively investigate associations between adipose tissue fatty acid patterns and long-term risk of total breast cancer and breast cancer subtypes characterised by oestrogen and progesterone receptor status (ER and PR). SUBJECTS/METHODS: This case-cohort study was based on data from the Danish cohort Diet, Cancer and Health. At baseline, a fat biopsy and information on lifestyle and reproductive factors were collected. From the 31 original fatty acids measured, patterns of fatty acids were identified using the treelet transform. During a median follow-up of 5.3 years, 474 breast cancer cases were identified. Hazard ratios and 95% confidence intervals of risk of total breast cancer and of subtypes according to quintiles of factor score were determined by weighted Cox proportional hazards regression. RESULTS: After adjustment for potential confounders, factor scores for the seven patterns identified by the treelet transform were not associated with risk of total breast cancer, nor with risk of ER+, ER − , PR+ or PR − tumours. CONCLUSIONS: No clear associations between the patterns of fatty acids at baseline and long-term risk of total breast cancer or ER+, ER − , PR+ or PR − tumours were observed. European Journal of Clinical Nutrition (2014) 68, 1088–1094; doi:10.1038/ejcn.2014.28; published online 19 March 2014
INTRODUCTION With an annual incidence of 1.4 million new cases, corresponding to 23% of cancer diagnoses in women, breast cancer is the most frequent cancer among women worldwide.1 Several risk factors for breast cancer have been established, but the aetiology of breast cancer remains largely unknown.2 Furthermore, subtypes of breast cancer categorised according to oestrogen and progesterone receptor (ER and PR, respectively) status might have different aetiologies.3 Intake of fatty acids may be related to breast cancer risk,4 but the intake is difficult to measure accurately with dietary assessment instruments, and may not represent the endogenous exposure of the breast cells.5 In contrast, the fatty acid composition of adipose tissue and blood lipids reflects dietary intake of some fatty acids over the past month and up to 2 years, but also absorptive and metabolic processes, genetics and lifestyle.6,7 Levels of fatty acids in tissue stores may thus more clearly reflect the environment the breast cells are exposed to than estimated dietary intake,5 and may therefore display stronger associations with disease.8,9 Epidemiological studies on the fatty acid composition of tissue stores and risk of breast cancer have yielded inconclusive results.10 Most studies have investigated individual fatty acids or the traditional classes of fatty acids based on chemical structure such
as saturated fatty acids (SFA), monounsaturated fatty acids (MUFA), trans fatty acids and polyunsaturated fatty acids (PUFA). However, many fatty acids originate from the same dietary sources or follow the same elongation and desaturation pathways. Their levels within the adipose tissue are therefore highly correlated, giving rise to naturally occurring patterns in fatty acid composition.11,12 Studies of individual fatty acids do not capture these correlations and may overly simplify the complexity of fatty acid composition in the body. Methods describing patterns by taking correlations into account, for example, the treelet transform (TT),13 could potentially facilitate new insight into the potential associations of fatty acids with breast cancer risk. The purpose of the present study was to prospectively investigate associations between patterns of adipose tissue fatty acids determined by TT and long-term risk of total breast cancer, as well as its subtypes categorised by ER and PR status. SUBJECTS AND METHODS Subjects This case-cohort study was based on data from the prospective Danish cohort Diet, Cancer and Health, the Danish branch of the European Prospective Investigation into Cancer and Nutrition.14 During the period 1993–1997, a total of 57 053 men and women were recruited; they were
1 Cancer Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, UK; 2Section for Epidemiology, Department of Public Health, Aarhus University, Aarhus, Denmark; 3Department of Cardiology, Aalborg University Hospital, Aalborg, Denmark; 4Department of Oncology, University Hospital of Uppsala, Uppsala, Sweden; 5Department of Nephrology, Aalborg University Hospital, Aalborg, Denmark and 6Diet, Genes and Environment Unit, Danish Cancer Society Research Center, Copenhagen, Denmark. Correspondence: JA Schmidt, Cancer Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Richard Doll Building, Roosevelt Drive, Oxford OX3 7LF, UK. E-mail: [email protected] Received 4 July 2013; revised 23 December 2013; accepted 26 January 2014; published online 19 March 2014
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1089 between 50 and 64 years of age, born in Denmark and free of cancer at the time of enrolment.15 The participants were followed from the date of their baseline examination until cancer diagnosis, death, emigration or 27 April 2006, whichever occurred first. All participants gave written informed consent, and Diet, Cancer and Health was approved by the ethical committees in Aarhus and Copenhagen. Information on vital status and emigration was obtained by linking cohort members to the Civil Registration System using the unique civil registration number assigned to all Danish residents. Information on cancer diagnoses was found in three national registries: the Danish Cancer Registry, which holds information on all cancer cases in the Danish population owing to mandatory reporting since 1987,16 the Pathology Bank and the Danish Breast Cancer Cooperative Group. From this latter registry, information on receptor status was obtained, but information on PR status was not registered systematically.17 A standardized immunohistologic method was used to detect ER and PR, and receptor status was defined as positive if ⩾ 10% of tumour cells possessed the receptors. During follow-up, 1074 women were diagnosed with breast cancer5 (Figure 1). Only women with the two most common types of breast cancer—invasive ductal and lobular carcinoma—were included in the present study corresponding to 95% of breast cancer cases. Fat biopsies from 409 breast cancer cases diagnosed between baseline and 31 December 2000 had been used in an earlier study,18 leaving no fat for analysis of fatty acids. Thus 1 January 2001 was defined as the start date of follow-up for the present study. All participants who were diagnosed with cancer (breast cancer or other types), had died, or had emigrated before this date were excluded, leaving 29 207 women in the study cohort. In order to economize the use of fat biopsies, a case-cohort design was used. Accordingly, the study population consisted of all breast cancer cases identified between 1 January 2001 and 27 April 2006, alongside a subcohort of 1215 participants randomly sampled among participants who were still in the cohort after the change in the start of the follow-up. Breast cancer cases for whom the diagnosis could not be histologically validated were excluded. The validation was done by reviewing pathology reports from the Pathology Bank and comparing this information with information from the other registries. Furthermore, participants with an incomplete fatty acid profile were excluded. Before analysis of breast cancer risk, participants with missing data on one or more of the potential covariates were excluded.
Baseline examination 1993–1997 n = 29 875 women
End of follow-up, 27 April 2006 n = 29 875
Cases n = 1074
Censored before 1 January 2001
Cohort after change of start of follow-up n = 29 207
Cases n = 500
Participants not in the sub-cohort Sub-cohort n = 1215
n = 17 Cases n = 500 Non-validated cases n = 23 Incomplete fatty acid profiles n = 105
Study population n = 1570
Sub-cohort n = 1112
n = 16 Cases n = 474 Missing covariates n = 57
Fatty acid analysis An adipose tissue biopsy of 35–50 mg subcutaneous fat was taken from the upper, outer part of the buttock at the baseline examination using a luer-lock system (Terumo, Terumo Corporation, Tokyo, Japan), consisting of a needle, a connector (Venoject multisample Luer adaptor) and an evacuated blood tube.15 The biopsies were stored in the connector in liquid nitrogen at − 150 °C until analysis. Studies have shown that longterm storage of fat biopsies only caused minor changes in fatty acids composition.6,19 Before analysis, the biopsies were defrosted and 2–4 mg of fat was removed from the connector. The fat was dissolved in heptane and fatty acids were trans-esterified by potassium hydroxide in methanol consistent with standard methods.20 The composition of fatty acids was analysed by gas chromatography using a Varian 3900 GC with a CP-8400 auto sampler and CP-Sil 88, 60 m × 0.25 mm ID capillary column (Varian, Middelburg, Netherlands) and using helium as carrier gas. Identification of individual fatty acids was performed using commercial standards (Nu-Chek-Prep, Inc., Elysian, MN, USA) that enable quantification of fatty acid methyl esters with 12–24 carbon atoms, and 31 fatty acids were determined (Table 1). The combinations of 18:1 n-6 (Δ12t) and 18:1 n-8 (Δ10t), and 18:1 n-10 (Δ8t) and 18:1 n-12 (Δ6t) were analysed owing to unsuccessful separation of peaks for these fatty acids.12 The inter-assay coefficients of variation for the fatty acids ranged from 0.7 to 11.7%.
Assessment of covariates At the baseline examination, participants filled in a self-administered questionnaire on lifestyle, medical history and reproductive factors. The questionnaires were checked for missing information and errors by optical scanning immediately after completion. Any unclear information was corrected during an interview performed by a trained lab technician who also measured weight and height for calculation of body mass index.15 © 2014 Macmillan Publishers Limited
Analysis population n = 1513
Sub-cohort n = 1070
n = 16 Cases n = 459
Figure 1. Flowchart of participant selection from baseline examination to analysis population.
Statistical analyses Baseline characteristics were examined for cases and members of the subcohort separately, and the average percentages of adipose tissue fatty acids in sub-cohort members were examined. To derive patterns of fatty acids we used the TT. This is a dimensionreduction method, which uses the correlation matrix of multidimensional data like fatty acids to derive patterns, or factors, which are in reality vectors.13 The numeric size of the original fatty acid in a factor is called the loading. Because analysing the correlation matrix standardizes the fatty acid variables, a positive loading corresponds to a larger-than-average value of the fatty acid; a zero loading corresponds to an average value; and a negative loading corresponds to a smaller-than-average value of the fatty acid. Factors derived by TT are sparse so that not all fatty acids contribute to every factor. Participants are assigned a factor score using the individual-level weights associated with each factor. A high score for a particular factor corresponds to a participant having a fatty acid profile that matches the factor loadings well. TT was performed according to the method described by GorstRasmussen et al.13,21 and based on fatty acids from members of the subcohort (n = 1112). In brief, closely correlated fatty acids were joined into one variable (depicted by a cluster tree; Figure 2), which had to be cut to derive factors. Initially, the tree was cut at a range of cut-levels (from 22 to 26) to determine how many factors to retain; namely factors with variance ⩾ 1.25.22 European Journal of Clinical Nutrition (2014) 1088 – 1094
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1090 Table 1. Mean (s.d.) percentage of fatty acids in adipose tissue, their loadings on treelet transform (TT) factors and the explained variance of extracted factors, in the study population sub-cohort (n = 1112) Fatty acid
Common name
Mean (s.d.)
Factor loading TT1
12:0 14:0 15:0 16:0 17:0 18:0 19:0 20:0 14:1 n-5 16:1 n-7 18:1 n-7 18:1 n-9 20:1 n-9 20:1 n-11 22:1 n-9 18:1 n-6+8 (Δ12t+Δ10t)a 18:1 n-9 (Δ9t) 18:1 n-10+12 (Δ8t+Δ6t)a 18:1 n-7 (Δ11t) 18:2 n-7 (Δ9c11t) 18:2 n-6 18:3 n-3 18:3 n-6 20:2 n-6 20:3 n-6 20:4 n-3 20:4 n-6 20:5 n-3 22:4 n-6 22:5 n-3 22:6 n-3 Explained variance (%)
Lauric acid Myristic acid Pentadecanoic acid Palmitic acid Heptadecanoic acid Stearic acid Nonadecanoic acid Arachidic acid Myristoleic acid Palmitoleic acid Asclepic acid Oleic acid n-9 eicosenoic acid n-11 eicosenoic acid Eurcic acid Elaidic acid Vaccenic acid Rumenic acid Linoleic acid α-linolenic acid γ-linolenic acid Eicosadienoic acid Dihomo-γ-linolenic acid Arachidonic acid Arachidonic acid Eicosapentaenoic acid Docosatetraenoic acid Docosapentaenoic acid Docosahexaenoic acid
0.4 2.7 0.3 19.3 0.21 3.2 0.1 0.2 0.5 7.1 2.2 44.3 0.8 0.2 0.05 0.3 0.5 0.3 0.3 0.5 10.9 0.8 0.1 0.2 0.2 0.1 0.4 0.1 0.1 0.3 0.3
(0.1) (0.5) (0.1) (1.9) (o 0.1) (0.9) (o 0.1) (0.1) (0.1) (1.5) (0.4) (1.9) (0.1) (0.1) (o 0.1) (0.1) (0.2) (0.1) (0.1) (0.1) (2.0) (0.2) (o 0.1) (o 0.1) (0.1) (0.3) (0.1) (0.1) (o 0.1) (0.1) (0.1)
a
TT2
European Journal of Clinical Nutrition (2014) 1088 – 1094
TT4
TT5
TT6
TT7
0.3 0.4 0.4 0.2 0.4 0.4 0.4 0.5 0.6 0.5 0.4 0.5 0.5 0.5 0.6 0.6 0.6 0.4 0.6 0.6 0.5 0.6 0.5 0.5 0.5 0.5 0.5 15.1
0.5 0.5 10.5
Peaks for 18:1n-10 (Δ8t) and 18:1n-12 (Δ6t), and 18:1n-6 (Δ12t) and 18:1n-8 (Δ10t) could not be separated.
The optimal cut-level for this number of factors was then determined by use of a 10-fold cross-validation.13,21 The TT tree was cut at the optimal level, the factors extracted and participants were assigned factor scores for each factor according to their own fatty acid profiles. To examine the association between patterns of fatty acids and total breast cancer risk, factor scores were divided into quintiles based on members of the study population sub-cohort. Hazard ratios (HR) and 95% confidence intervals (CI) for each quintile were calculated relative to the first quintile by weighted Cox proportional hazards regression using age as the time axis and robust variance estimates.23 Cases were assigned weight 1, whereas non-cases were assigned weight N/n; N and n being the number of non-cases in the study cohort and the sub-cohort, respectively. A P-value for trend (P(trend)) was calculated by entering quintiles of factor scores as continuous ordinal variables in the regressions. Two-sided P-values o0.05 were considered statistically significant in all analyses. An unadjusted and one adjusted regression models were explored. The adjusted model included: body mass index (o 20; 20–o25; 25–o 30; ⩾30 kg/m2), smoking (never; former; current o15; current 15–25; current >25 g tobacco/day), alcohol consumption (g/day, as a continuous variable), hormone replacement therapy (never; former; current), number of childbirths (0; 1; 2; 3; ⩾4), age at first childbirth (years, as a continuous variable), age at menarche (o 12; 12–o14; ⩾14 years; unknown), benign breast tumour surgery (yes; no; unknown), physical activity (⩽3.5; >3.5 h/week) and education (⩽7; 8–10; >10 years). The proportional hazard assumption was visually examined by using plots of log(− log(probability of being free of breast cancer)) and no appreciable violations were found. The association between patterns of fatty acids and risk of breast cancer subtypes, ER+, ER − , PR+ and PR − , was analysed as described for total breast cancer risk. TT factor scores can be correlated, potentially leading to confounding.13 Thus, a sensitivity analysis was performed: the factor-specific adjusted
TT3
9.3
9.2
9.0
7.4
6.0
12
regressions for total breast cancer risk were further adjusted (using quintiles) for all factor scores that had an absolute correlation > 0.2 with the score of the factor under investigation. All analyses were performed in STATA, version 11 (StataCorp LP, College Station, TX, USA).
RESULTS After exclusion of non-validated cases (n = 23) and participants with incomplete fatty acid profiles (n = 105), the study population consisted of 1570 women (Figure 1). Of these, 474 were breast cancer cases, 1112 were members of the sub-cohort and 16 were both. Information on covariates was missing for 57 of these women who were excluded before analyses of breast cancer risk. Thus, these analyses were based on 459 cases and 1070 members of the sub-cohort including 16 cases. The median duration of follow-up was 5 years and 4 months. The mean age at start of the follow-up was 61 years (Table 2). Cases were more often current users of hormone replacement therapy, had higher alcohol consumption, had more often had benign breast tumour surgery and had given birth fewer times than members of the sub-cohort. Body mass index, smoking, age at first childbirth and menarche, physical activity and level of education were more equally distributed between the groups. The most abundant fatty acids in adipose tissue were 18:1 n-9, 16:0 and 18:2 n-6 constituting 44.3% (s.d. 1.9), 19.3% (s.d. 1.9) and 10.9% (s.d. 2.0) of total fatty acids, respectively (Table 1). © 2014 Macmillan Publishers Limited
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1091 Cut–level 0 12:0 14:0 15:0 17:0 18:0 19:0 18:1 n–7 (Δ11t) 16:0 20:0 22:1 n–9 20:1 n–9 20:1 n–11 18:1 n–10+12 (Δ6t+Δ8t) 18:1 n−9 (Δ9t) 18:1 n–6+8 (Δ12t+Δ10t) 18:3 n–6 14:1 n–5 18:2 n–7 (Δ9c11t) 16:1 n–7 18:1 n–9 18:1 n–7 20:3 n–6 20:4 n–6 22:4 n–6 20:4 n–3 20:5 n–3 22:6 n–3 22:5 n–3 18:2 n–6 20:2 n–6 18:3 n–3
5
10
15
20
25
30
35
Table 2.
Baseline characteristics, measured in 1993 − 1997, of members of the sub-cohort (n = 1112) and the breast cancer cases (n = 474) in the study population n (%)/median (5th; 95th percentile)a
Variables TT1
Sub-cohort (n = 1112)b Cases (n = 474)c Age at start of the follow-up (years) Median (5th; 95th percentile)
TT3
60.9 (55.0; 69.0)
61.5 (55.1; 69.1)
Body mass index (kg/m ) o20 20–o25 25–o30 ⩾ 30
58 541 376 137
(5.2) (48.7) (33.8) (12.3)
18 247 145 64
(3.8) (52.1) (30.6) (13.5)
Smoking (g tobacco/day) Never Former Current o15 Current 15–25 Current >25
478 262 172 175 25
(43.0) (23.6) (15.5) (15.7) (2.3)
215 109 76 60 14
(45.4) (23.0) (16.0) (12.7) (3.0)
2
TT5
TT6
TT4
TT2
Alcohol consumption (g/day) Median (5th; 95th percentile)
TT7
Figure 2. Cluster tree produced by the treelet transform (TT) on adipose tissue fatty acid variables among women in the study population sub-cohort, n = 1112, cut at level 22 (dotted line), retaining 7 factors.
The TT identified seven factors, or patterns, with variances of ⩾ 1.25. These factors explained 66.5% of the total variance, and fatty acid loadings for each factor were positive (Table 1). Fatty acids joined at the optimal cut-level 22 were grouped in the same pattern and all except 18:1 n-9 and 18:3 n-6 were contained in exactly one pattern (Figure 2). TT2 and TT5 were comprised solely of n-3 PUFA and trans fatty acids, respectively. TT1, TT3 and TT4 were dominated by SFA, long-chain MUFA and n-6 PUFA, respectively, whereas MUFA and n-6 PUFA were the main fatty acids in TT6 and TT7, respectively. No clear associations between patterns of fatty acids and risk of total breast cancer were observed (Table 3). A slight tendency towards higher risk of breast cancer with higher factor score for TT3, TT4 and TT5 was observed, but no clear dose-response associations were found. Of the 459 cases in the analysis population, 384 (84%) had ER+ tumours, 59 (13%) had ER − tumours, 162 (35%) had PR+ tumours and 91 (20%) had PR − tumours. ER and PR status was unknown for 16 (3%) and 206 (45%) cases, respectively. Similar to the analysis of total breast cancer risk, no clear associations between patterns of fatty acids and risk of ER+, ER − , PR+ or PR − tumours were found (Table 4). Only weak tendencies towards a higher risk of ER+ tumours for TT5 and a higher risk of ER − tumours for TT4 and TT7 were observed. In contrast, a slight tendency towards a lower risk of PR+ tumours for TT2 was observed. In general, the correlations between factor scores were substantial; approximately half of the correlations (11 out of 21) were more extreme than the chosen cutoff at ± 0.2 (data not shown). When these factor scores were simultaneously added in quintiles to the adjusted model for total breast cancer risk, the association for both TT3 and TT4 with risk of breast cancer were strengthened, and both reached statistical significance (HR (95% CI) for TT3 adjusted for TT1, TT4 and TT5 in the 5th quintile was 1.42 (0.91–2.21), P(trend) = 0.04; for TT4 adjusted for TT1–TT3 and TT5 the corresponding numbers were 1.54 (0.92–2.58), P(trend) = 0.03). However, no clear dose-response relationships were found. © 2014 Macmillan Publishers Limited
9.2 (1.2; 40.9)
10.7 (0.6; 42.4)
Hormone replacement therapy Never Former Current
578 (53.0) 169 (15.5) 344 (31.5)
186 (39.8) 61 (13.1) 220 (47.1)
Number of childbirths 0 1 2 3 ⩾4
129 171 518 225 61
76 85 210 78 25
Age at first childbirthd (years) Median (5th; 95th percentile)
(11.7) (15.5) (46.9) (20.4) (5.5)
23 (18; 32)
(16.0) (17.9) (44.3) (16.5) (5.3)
24 (18; 32)
Age at menarche (years) o12 12–o14 ⩾ 14 years Unknown
97 393 581 41
Benign breast tumour surgery Yes No Unknown
142 (12.8) 966 (87.1) 1 (0.1)
88 (18.6) 383 (80.8) 3 (0.6)
Physical activity (h/week) o3.5 ⩾ 3.5
718 (64.6) 394 (35.4)
309 (65.2) 165 (34.8)
Education (years) ⩽7 8–10 >10
363 (32.7) 555 (50.0) 192 (17.3)
136 (28.7) 247 (52.1) 91 (19.2)
(8.7) (35.4) (52.3) (3.7)
43 186 228 17
(9.1) (39.2) (48.1) (3.6)
Data are n (%) if not specified as median (5th; 95th percentile). bn = 1091 for hormone replacement therapy, 1104 for number of childbirths, 1109 for benign breast surgery and 1110 for education, owing to missing data. c n = 467 for hormone replacement therapy, owing to missing data. dOnly for women who had given birth; cases, n = 394; sub-cohort, n = 967. a
DISCUSSION In the present study, no clear associations between patterns of fatty acids and long-term risk of breast cancer or risk of ER+, ER − , PR+ or PR − tumours were observed. After adjusting for correlated factor European Journal of Clinical Nutrition (2014) 1088 – 1094
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1092 Table 3.
Risk of breast cancer expressed as hazard ratios (HR) according to quintiles of treelet transform (TT) factor scores for patterns of adipose tissue fatty acids (n = 1513) HR (95% CI)a
Fatty acids Factor score pattern quintile
Unadjusted TT1
TT2
TT3
TT4
TT5
TT6
TT7
1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend)
1.00 0.82 0.93 1.02 1.04 1.00 0.81 0.85 1.03 0.88 1.00 0.80 1.22 1.06 1.19 1.00 0.96 1.04 1.21 1.04 1.00 0.93 0.98 1.13 0.99 1.00 1.30 1.02 1.39 1.13 1.00 1.19 0.90 1.13 0.99
(reference) (0.58–1.18) (0.66–1.31) (0.72–1.43) (0.74–1.46) 0.47 (reference) (0.57–1.15) (0.60–1.21) (0.73–1.44) (0.62–1.25) 0.99 (reference) (0.55–1.15) (0.86–1.71) (0.75–1.51) (0.84–1.68) 0.12 (reference) (0.67–1.37) (0.73–1.48) (0.86–1.69) (0.73–1.48) 0.42 (reference) (0.65–1.32) (0.70–1.39) (0.81–1.59) (0.70–1.40) 0.65 (reference) (0.92–1.85) (0.71–1.48) (0.98–1.89) (0.79–1.63) 0.44 (reference) (0.85–1.67) (0.63–1.28) (0.81–1.59) (0.70–1.41) 0.87
Adjustedb 1.00 0.74 0.86 0.98 1.01 1.00 0.80 0.90 1.10 0.98 1.00 0.78 1.17 1.04 1.27 1.00 0.97 1.05 1.38 1.12 1.00 1.11 1.13 1.45 1.20 1.00 1.24 0.96 1.32 1.00 1.00 1.29 1.00 1.31 1.09
(reference) (0.51–1.09) (0.60–1.25) (0.68–1.41) (0.69–1.47) 0.51 (reference) (0.55–1.17) (0.62–1.30) (0.76–1.60) (0.66–1.47) 0.53 (reference) (0.52–1.16) (0.81–1.69) (0.69–1.55) (0.85–1.91) 0.10 (reference) (0.66–1.42) (0.72–1.55) (0.95–2.01) (0.75–1.68) 0.20 (reference) (0.76–1.62) (0.77–1.64) (0.99–2.11) (0.80–1.81) 0.17 (reference) (0.85–1.80) (0.65–1.41) (0.92–1.90) (0.68–1.48) 0.90 (reference) (0.90–1.85) (0.68–1.46) (0.91–1.89) (0.74–1.60) 0.65
Abbreviation: CI, confidence interval. aHazard ratios were calculated by use of the weighted Cox proportional hazard regression. bAdjusted for body mass index ( o20; 20–o25; 25–o30; ⩾30 kg/m2), smoking (never; former; current o15; current 15–25; current >25 g tobacco/day), alcohol consumption (g/day, as a continuous variable), hormone replacement therapy (never; former; current), number of childbirths (0; 1; 2; 3; ⩾ 4), age at first childbirth (years, as a continuous variable), age at menarche (o 12; 12–o 14; ⩾ 14 years; unknown), benign breast tumour surgery (yes; no; unknown), physical activity (⩽3.5; >3.5 h/week) and education (⩽ 7; 8–10; >10 years).
scores, positive associations were observed between the risk of breast cancer and TT3 and TT4 (dominated by MUFA and n-6 PUFA, respectively). However, interpretation of the fully adjusted model is complex. Common dietary sources partly explain the correlations and thus it is not possible to change the factor score for one pattern without affecting ones factor score in other patterns. The fatty acids grouped in patterns were only to some extent comparable to the traditional classes of fatty acids based on chemical structure. This could be expected when the correlations between fatty acids induced by dietary sources, metabolic pathways and the complex European Journal of Clinical Nutrition (2014) 1088 – 1094
interplay of product inhibition, genetics and fatty acids competing for enzymes7 are taken into consideration. The results of this study must be interpreted in light of the potential for bias and confounding. First, selection bias owing to misclassification of breast cancer cases is unlikely to have affected the results, as cases were identified via three national registers, which contain information on all cancer diagnoses in Danish residence, and diagnoses were subsequently validated. However, the lack of information on the fatty acid composition from women developing breast cancer close to baseline is a limitation. This precluded assessment of short-term effects of fatty acid patterns on breast cancer, and it might have resulted in selection bias.24 Breast cancer occurring shortly after baseline is likely to be in women who are susceptible to a potential effect of fatty acid composition on breast cancer, but these are overlooked in the present analysis. Thus, the study cohort is likely to comprise less susceptible women compared with all women recruited in the original Diet, Cancer and Health cohort. Second, adipose tissue fatty acids do not possess the weaknesses related to collection of dietary information, such as erroneous reporting,25 reducing the potential for information bias. Furthermore, the investigation of patterns of fatty acids identified by TT allowed us to investigate the naturally occurring patterns in fatty acid composition and did not limit us to investigation of individual fatty acids, where information on these correlations would be overlooked. However, the patterns identified by TT explain variation of fatty acids in the population, and the biological relevance of mathematically derived patterns may be questioned.26,27 Just one measurement of fatty acids composition was available, and as the composition might change over time owing to changes in diet or metabolism, our results may underestimate effect sizes. Finally, to address confounding, several factors related to breast cancer were included in the multivariate regression models. In addition, a sensitivity analysis further adjusting for correlated fatty acid patterns was conducted. However, residual confounding may still have affected our results. Studies have indicated that the association between tissue fatty acids and risk of breast cancer might differ by menopausal status.28,29 However, owing to the small number of women still premenopausal at the start of follow-up in the study, we were unable to investigate this association stratified by menopausal status. Only one previous study has investigated the relationship between breast cancer risk and patterns of fatty acids (determined by principal component analysis).11 This case-control study found a protective effect of a pattern characterised by a low n-6/n-3 PUFA ratio, low levels of 18:2 n-6 and high levels of MUFA. A pattern characterised by low levels SFA and high levels of n-6 PUFA was not associated with risk of breast cancer. Other studies of fat composition in tissue stores all investigated individual fatty acids or the traditional classes of fatty acids. The majority of cohort studies investigating SFA,10,30–33 n-6 PUFA,28,30,33,34 trans fatty acids28,31,34 or MUFA 28,30,33,34 in relation to breast cancer risk have not found any associations. However, other studies have reported positive associations between breast cancer risk and SFA,28 trans fatty acids30 or MUFA,31 whereas others again have found inverse associations with SFA34 or n-6 PUFA10,32 in cohort studies. A meta-analysis published in 2004 of cohort and case-control studies is suggestive of a protective effect of n-3 PUFA.10 Some studies published later have supported this finding, over all35,36 or for premenopausal women only,29 whereas others have not.5,30,33,37,38 The discrepancy might be attributed to the proportions of n-3 PUFA in tissue stores. In most studies suggestive of a protective effect, the women had a high proportion of n-3 PUFA,31,35,36 but in studies not documenting an association, the women had low proportions of n-3 PUFA in the tissue.5,28,37–40 This could indicate a threshold for a protective effect of n-3 PUFA on breast cancer risk. Our findings of no clear © 2014 Macmillan Publishers Limited
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1093 Table 4. Risk of breast cancer subtypes characterised by receptor status expressed as hazard ratios (HR) according to quintiles of treelet transform (TT) factor scores for patterns of adipose tissue fatty acids Fatty acids pattern
HR (95% CI)a
Factor score quintile ER+b n (case) = 384
TT1
TT2
TT3
TT4
TT5
TT6
TT7
1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend)
1.00 0.71 0.78 0.99 0.98 1.00 0.82 0.85 1.13 1.01 1.00 0.68 1.02 1.01 1.20 1.00 0.89 0.92 1.35 1.11 1.00 1.19 1.18 1.47 1.25 1.00 1.20 0.93 1.31 0.96 1.00 1.33 1.01 1.38 0.04
(reference) (0.48–1.07) (0.52–1.16) (0.67–1.46) (0.65–1.46) 0.55 (reference) (0.55–1.23) (0.57–1.26) (0.76–1.67) (0.66–1.54) 0.51 (reference) (0.45–1.04) (0.69–1.51) (0.67–1.55) (0.78–1.84) 0.14 (reference) (0.59–1.34) (0.61–1.39) (0.91–2.02) (0.72–1.71) 0.21 (reference) (0.80–1.78) (0.78–1.77) (0.98–2.20) (0.81–1.93) 0.17 (reference) (0.80–1.78) (0.61–1.40) (0.89–1.93) (0.64–1.46) 0.95 (reference) (0.90–1.97) (0.67–1.52) (0.94–2.05) (0.69–1.57) 0.76
ER − b n (case) = 59 1.00 0.99 1.34 0.94 0.83 1.00 0.93 1.31 1.21 1.11 1.00 1.11 2.18 0.80 2.12 1.00 1.74 2.53 1.78 1.69 1.00 0.62 0.57 1.26 0.97 1.00 1.56 1.24 1.32 1.15 1.00 0.71 1.25 1.29 1.77
(reference) (0.36–2.71) (0.54–3.35) (0.37–2.39) (0.30–2.30) 0.66 (reference) (0.35–2.47) (0.51–3.39) (0.45–3.28) (0.40–3.09) 0.70 (reference) (0.40–3.11) (0.88–5.42) (0.23–2.73) (0.79–5.67) 0.22 (reference) (0.60–5.00) (0.94–6.86) (0.64–4.98) (0.58–4.93) 0.35 (reference) (0.24–1.62) (0.23–1.40) (0.55–2.87) (0.37–2.53) 0.62 (reference) (0.60–4.06) (0.46–3.31) (0.50–3.52) (0.41–3.24) 0.98 (reference) (0.27–1.88) (0.53–2.98) (0.53–3.15) (0.74–4.26) 0.12
PR+b n (case) = 162 1.00 0.64 0.66 0.86 1.11 1.00 0.61 0.65 0.76 0.59 1.00 0.51 0.79 0.90 0.80 1.00 0.73 1.14 1.40 1.12 1.00 1.38 1.13 1.56 1.07 1.00 1.45 1.01 1.20 0.98 1.00 1.10 0.69 1.02 0.90
(reference) (0.36–1.15) (0.37–1.18) (0.48–1.55) (0.62–1.98) 0.49 (reference) (0.34–1.08) (0.37–1.12) (0.44–1.33) (0.32–1.09) 0.22 (reference) (0.27–0.95) (0.45–1.38) (0.49–1.65) (0.43–1.51) 0.87 (reference) (0.40–1.35) (0.64–2.01) (0.80–2.46) (0.60–2.07) 0.21 (reference) (0.79–2.39) (0.62–2.07) (0.89–2.74) (0.56–2.06) 0.62 (reference) (0.83–2.55) (0.56–1.84) (0.68–2.11) (0.54–1.80) 0.69 (reference) (0.65–1.86) (0.38–1.25) (0.59–1.77) (0.51–1.58) 0.66
PR − b n (case) = 91 1.00 0.79 0.97 0.86 1.04 1.00 0.82 1.10 1.45 1.09 1.00 0.90 1.79 0.50 1.70 1.00 0.87 1.58 1.63 1.11 1.00 0.77 0.70 1.00 0.97 1.00 1.54 1.15 1.89 0.95 1.00 1.11 1.22 1.61 1.11
(reference) (0.35–1.79) (0.46–2.04) (0.42–1.76) (0.50–2.18) 0.84 (reference) (0.38–1.78) (0.51–2.37) (0.68–3.06) (0.48–2.45) 0.41 (reference) (0.41–1.98) (0.90–3.57) (0.18–1.36) (0.79–3.65) 0.29 (reference) (0.39–1.96) (0.73–3.43) (0.75–3.55) (0.48–2.52) 0.39 (reference) (0.36–1.64) (0.33–1.46) (0.50–2.00) (0.45–2.11) 0.88 (reference) (0.71–3.34) (0.50–2.62) (0.89–4.01) (0.41–2.24) 0.92 (reference) (0.54–2.31) (0.59–2.55) (0.78–3.32) (0.50–2.45) 0.47
Abbreviations: CI, confidence interval; ER ± , oestrogen receptor positive/negative; PR ± , progesterone receptor positive/negative. aHazard ratios were calculated by use of the weighted Cox proportional hazard regression and were adjusted for body mass index ( o 20; 20–o 25; 25–o 30; ⩾30 kg/m2), smoking (never; former; current o 15; current 15–25; current >25 g tobacco/day), alcohol consumption (g/day, as a continuous variable), hormone replacement therapy (never; former; current), number of childbirths (0; 1; 2; 3; ⩾ 4), age at first childbirth (years, as a continuous variable), age at menarche (o 12; 12–o14; ⩾ 14 years; unknown), benign breast tumour surgery (yes; no; unknown), physical activity (⩽ 3.5; >3.5 h/week) and education (⩽ 7; 8–10; >10 years). bNumbers of sub-cohort members: ER, n = 1069; PR, n = 1063.
association between TT2 (comprised solely of n-3 PUFA) and risk of breast cancer would be consistent with a threshold effect, as proportions of n-3 PUFA in the present study were low. In conclusion, no clear associations between the baseline patterns of fatty acids identified by TT and long-term risk of breast cancer or of ER+, ER − , PR+ or PR − tumours were observed. These findings indicate that the dominating patterns of adipose tissue fatty acids, explaining the most variance in the population, do not affect the long-term risk of breast cancer.
ACKNOWLEDGEMENTS We thank the participants of Diet, Cancer and Health for their invaluable contribution to the study. We also acknowledge the Danish Cancer Society for supporting this study financially.
REFERENCES
The authors declare no conflict of interest.
1 Ferlay J, Shin H-R, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer 2010; 127: 2893–2917. 2 Tibor T. The theory of the sick lobe. In: Tibor T (ed). Breast Cancer—A Lobular Disease. 1st edn. Springer-Verlag: London, UK, 2011, pp 1–17. 3 Yang XR, Chang-Claude J, Goode EL, Couch FJ, Nevanlinna H, Milne RL et al. Associations of breast cancer risk factors with tumor subtypes: a pooled analysis
© 2014 Macmillan Publishers Limited
European Journal of Clinical Nutrition (2014) 1088 – 1094
CONFLICT OF INTEREST
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1094 4 5
6
7 8
9
10
11 12
13
14
15
16 17
18
19
20
21
from the Breast Cancer Association Consortium studies. J Natl Cancer Inst 2011; 103: 250–263. Turner LB. A meta-analysis of fat intake, reproduction, and breast cancer risk: an evolutionary perspective. Am J Hum Biol 2011; 23: 601–608. Witt PM, Christensen JH, Schmidt EB, Dethlefsen C, Tjønneland A, Overvad K et al. Marine n-3 polyunsaturated fatty acids in adipose tissue and breast cancer risk: a case-cohort study from Denmark. Cancer Causes Control 2009; 20: 1715–1721. Hodson L, Skeaff CM, Fielding BA. Fatty acid composition of adipose tissue and blood in humans and its use as a biomarker of dietary intake. Prog Lipid Res 2008; 47: 348–380. Vessby B, Gustafsson IB, Tengblad S, Boberg M, Andersson A. Desaturation and elongation of fatty acids and insulin action. Ann N Y Acad Sci 2002; 967: 183–195. Simonsen NR, Fernandez-Crehuet Navajas J, Martin-Moreno JM, Strain JJ, Huttunen JK, Martin BC et al. Tissue stores of individual monounsaturated fatty acids and breast cancer: the EURAMIC study. European Community Multicenter Study on Antioxidants, Myocardial Infarction, and Breast Cancer. Am J Clin Nutr 1998; 68: 134–141. Tjønneland A, Overvad K, Thorling E, Ewertz M. Adipose tissue fatty acids as biomarkers of dietary exposure in Danish men and women. Am J Clin Nutr 1993; 57: 629–633. Saadatian-Elahi M, Norat T, Goudable J, Riboli E. Biomarkers of dietary fatty acid intake and the risk of breast cancer: a meta-analysis. Int J Cancer 2004; 111: 584–591. Bougnoux P, Giraudeau B, Couet C. Diet, cancer, and the lipidome. Cancer Epidemiol Biomarkers Prev 2006; 15: 416–421. Dahm CC, Gorst-Rasmussen A, Jakobsen MU, Schmidt EB, Tjønneland A, Sørensen TI et al. Adipose tissue fatty acid patterns and changes in anthropometry: a cohort study. PLoS One 2011; 6: e22587. Gorst-Rasmussen A, Dahm CC, Dethlefsen C, Scheike T, Overvad K. Exploring dietary patterns by using the treelet transform. Am J Epidemiol 2011; 173: 1097–1104. Riboli E, Kaaks R. The EPIC Project: rationale and study design. European Prospective Investigation into Cancer and Nutrition. Int J Epidemiol 1997; 26: S6–14. Tjønneland A, Olsen A, Boll K, Stripp C, Christensen J, Engholm G et al. Study design, exposure variables, and socioeconomic determinants of participation in Diet, Cancer and Health: a population-based prospective cohort study of 57 053 men and women in Denmark. Scand J Public Health 2007; 35: 432–441. Storm HH, Michelsen EV, Clemmensen IH, Pihl J. The Danish Cancer Registry – history, content, quality and use. Dan Med Bull 1997; 44: 535–539. Olsen A, Knudsen KE, Thomsen BL, Loft S, Stripp C, Overvad K et al. Plasma enterolactone and breast cancer incidence by estrogen receptor status. Cancer Epidemiol Biomarkers Prev 2004; 13: 2084–2089. Raaschou-Nielsen O, Pavuk M, Leblanc A, Dumas P, Philippe Weber J, Olsen A et al. Adipose organochlorine concentrations and risk of breast cancer among postmenopausal Danish women. Cancer Epidemiol Biomarker s Prev 2005; 14: 67–74. Beynen AC, Katan MB. Rapid sampling and long-term storage of subcutaneous adipose-tissue biopsies for determination of fatty acid composition. Am J Clin Nutr 1985; 42: 317–322. IUPAC Methods 2.301. IUPAC Standard Methods of Oil, Fats and Derivatives, Report of IUPAC Working Group WG 2/87. 7th edn. Blackwell Scientific Publications: Oxford, UK, 1987. Gorst-Rasmussen A. tt: Treelet transform with Stata. Stata J 2012; 1: 130–146.
European Journal of Clinical Nutrition (2014) 1088 – 1094
22 Slattery ML, Boucher KM, Caan BJ, Potter JD, Ma KN. Eating patterns and risk of colon cancer. Am J Epidemiol 1998; 148: 4–16. 23 Kalbfleisch JD, Lawless JF. Likelihood analysis of multi-state models for disease incidence and mortality. Stat Med 1988; 7: 149–160. 24 Hernan MA. The hazards of hazard ratios. Epidemiology 2010; 21: 13–15. 25 Bates CJ, Nelson M, Ulijaszek SJ. Nutritional assessment methods. In: Geissler CA, Powers HJ (eds) Human Nutrition. 11th edn. Churchill Livingstone: Edinburgh, UK, 2005, pp 573–595. 26 Jacques PF, Tucker KL. Are dietary patterns useful for understanding the role of diet in chronic disease? Am J Clin Nutr 2001; 73: 1–2. 27 Martinez ME, Marshall JR, Sechrest L. Invited commentary: factor analysis and the search for objectivity. Am J Epidemiol 1998; 148: 17–19. 28 Saadatian-Elahi M, Toniolo P, Ferrari P, Goudable J, Akhmedkhanov A, Zeleniuch-Jacquotte A et al. Serum fatty acids and risk of breast cancer in a nested case-control study of the New York University Women's Health Study. Cancer Epidemiol Biomarkers Prev 2002; 11: 1353–1360. 29 De Bree E, Mamalakis G, Sanidas E, Hatzis C, Askoxylakis I, Daskalakis M et al. Adipose tissue fatty acid composition in Greek patients with breast cancer versus those with benign breast tumors. Anticancer Res 2013; 33: 1667–1672. 30 Chajes V, Thiebaut AC, Rotival M, Gauthier E, Maillard V, Boutron-Ruault MC et al. Association between serum trans-monounsaturated fatty acids and breast cancer risk in the E3N-EPIC Study. Am J Epidemiol 2008; 167: 1312–1320. 31 Pala V, Krogh V, Muti P, Chajes V, Riboli E, Micheli A et al. Erythrocyte membrane fatty acids and subsequent breast cancer: a prospective Italian study. J Natl Cancer Inst 2001; 93: 1088–1095. 32 Rissanen H, Knekt P, Jarvinen R, Salminen I, Hakulinen T. Serum fatty acids and breast cancer incidence. Nutr Cancer 2003; 45: 168–175. 33 Wirfalt E, Vessby B, Mattisson I, Gullberg B, Olsson H, Berglund G. No relations between breast cancer risk and fatty acids of erythrocyte membranes in postmenopausal women of the Malmo Diet Cancer cohort (Sweden). Eur J Clin Nutr 2004; 58: 761–770. 34 Chajes V, Hulten K, Van Kappel AL, Winkvist A, Kaaks R, Hallmans G et al. Fatty acid composition in serum phospholipids and risk of breast cancer: a prospective cohort study in Northern Sweden. Int J Cancer 1999; 26: 585–590. 35 Kuriki K, Hirose K, Wakai K, Matsuo K, Ito H, Suzuki T et al. Breast cancer risk and erythrocyte compositions of n-3 highly unsaturated fatty acids in Japanese. Int J Cancer 2007; 121: 377–385. 36 Shannon J, King IB, Moshofsky R, Lampe JW, Gao DL, Ray RM et al. Erythrocyte fatty acids and breast cancer risk: a case-control study in Shanghai, China. Am J Clin Nutr 2007; 85: 1090–1097. 37 Bagga D, Anders KH, Wang HJ, Glaspy JA. Long-chain n-3-to-n-6 polyunsaturated fatty acid ratios in breast adipose tissue from women with and without breast cancer. Nutr Cancer 2002; 42: 180–185. 38 Mamalakis G, Hatzis C, de Bree E, Sanidas E, Tsiftsis DD, Askoxylakis J et al. Adipose tissue fatty acids in breast cancer patients versus healthy control women from Crete. Ann Nutr Metab 2009; 54: 275–282. 39 London SJ, Sacks FM, Stampfer MJ, Henderson IC, Maclure M, Tomita A et al. Fatty acid composition of the subcutaneous adipose tissue and risk of proliferative benign breast disease and breast cancer. J Natl Cancer Inst 1993; 85: 785–793. 40 Simonsen N, van't Veer P, Strain JJ, Martin-Moreno JM, Huttunen JK, Navajas JF et al. Adipose tissue omega-3 and omega-6 fatty acid content and breast cancer in the EURAMIC study. European Community Multicenter Study on Antioxidants, Myocardial Infarction, and Breast Cancer. Am J Epidemiol 1998; 147: 342–352.
© 2014 Macmillan Publishers Limited
ORIGINAL ARTICLE
Baseline patterns of adipose tissue fatty acids and long-term risk of breast cancer: a case-cohort study in the Danish cohort Diet, Cancer and Health JA Schmidt1,2, A Gorst-Rasmussen3, PW Nyström4, JH Christensen5, EB Schmidt3, C Dethlefsen3, A Tjønneland6, K Overvad2,3 and CC Dahm2 BACKGROUND/OBJECTIVES: The evidence regarding fatty acids and breast cancer risk is inconclusive. Adipose tissue fatty acids can be used as biomarkers of fatty acid intake and of endogenous fatty acid exposure. Fatty acids in adipose tissue are correlated owing to common dietary sources and shared metabolic pathways, which group fatty acids into naturally occurring patterns. We aimed to prospectively investigate associations between adipose tissue fatty acid patterns and long-term risk of total breast cancer and breast cancer subtypes characterised by oestrogen and progesterone receptor status (ER and PR). SUBJECTS/METHODS: This case-cohort study was based on data from the Danish cohort Diet, Cancer and Health. At baseline, a fat biopsy and information on lifestyle and reproductive factors were collected. From the 31 original fatty acids measured, patterns of fatty acids were identified using the treelet transform. During a median follow-up of 5.3 years, 474 breast cancer cases were identified. Hazard ratios and 95% confidence intervals of risk of total breast cancer and of subtypes according to quintiles of factor score were determined by weighted Cox proportional hazards regression. RESULTS: After adjustment for potential confounders, factor scores for the seven patterns identified by the treelet transform were not associated with risk of total breast cancer, nor with risk of ER+, ER − , PR+ or PR − tumours. CONCLUSIONS: No clear associations between the patterns of fatty acids at baseline and long-term risk of total breast cancer or ER+, ER − , PR+ or PR − tumours were observed. European Journal of Clinical Nutrition (2014) 68, 1088–1094; doi:10.1038/ejcn.2014.28; published online 19 March 2014
INTRODUCTION With an annual incidence of 1.4 million new cases, corresponding to 23% of cancer diagnoses in women, breast cancer is the most frequent cancer among women worldwide.1 Several risk factors for breast cancer have been established, but the aetiology of breast cancer remains largely unknown.2 Furthermore, subtypes of breast cancer categorised according to oestrogen and progesterone receptor (ER and PR, respectively) status might have different aetiologies.3 Intake of fatty acids may be related to breast cancer risk,4 but the intake is difficult to measure accurately with dietary assessment instruments, and may not represent the endogenous exposure of the breast cells.5 In contrast, the fatty acid composition of adipose tissue and blood lipids reflects dietary intake of some fatty acids over the past month and up to 2 years, but also absorptive and metabolic processes, genetics and lifestyle.6,7 Levels of fatty acids in tissue stores may thus more clearly reflect the environment the breast cells are exposed to than estimated dietary intake,5 and may therefore display stronger associations with disease.8,9 Epidemiological studies on the fatty acid composition of tissue stores and risk of breast cancer have yielded inconclusive results.10 Most studies have investigated individual fatty acids or the traditional classes of fatty acids based on chemical structure such
as saturated fatty acids (SFA), monounsaturated fatty acids (MUFA), trans fatty acids and polyunsaturated fatty acids (PUFA). However, many fatty acids originate from the same dietary sources or follow the same elongation and desaturation pathways. Their levels within the adipose tissue are therefore highly correlated, giving rise to naturally occurring patterns in fatty acid composition.11,12 Studies of individual fatty acids do not capture these correlations and may overly simplify the complexity of fatty acid composition in the body. Methods describing patterns by taking correlations into account, for example, the treelet transform (TT),13 could potentially facilitate new insight into the potential associations of fatty acids with breast cancer risk. The purpose of the present study was to prospectively investigate associations between patterns of adipose tissue fatty acids determined by TT and long-term risk of total breast cancer, as well as its subtypes categorised by ER and PR status. SUBJECTS AND METHODS Subjects This case-cohort study was based on data from the prospective Danish cohort Diet, Cancer and Health, the Danish branch of the European Prospective Investigation into Cancer and Nutrition.14 During the period 1993–1997, a total of 57 053 men and women were recruited; they were
1 Cancer Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, UK; 2Section for Epidemiology, Department of Public Health, Aarhus University, Aarhus, Denmark; 3Department of Cardiology, Aalborg University Hospital, Aalborg, Denmark; 4Department of Oncology, University Hospital of Uppsala, Uppsala, Sweden; 5Department of Nephrology, Aalborg University Hospital, Aalborg, Denmark and 6Diet, Genes and Environment Unit, Danish Cancer Society Research Center, Copenhagen, Denmark. Correspondence: JA Schmidt, Cancer Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Richard Doll Building, Roosevelt Drive, Oxford OX3 7LF, UK. E-mail: [email protected] Received 4 July 2013; revised 23 December 2013; accepted 26 January 2014; published online 19 March 2014
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1089 between 50 and 64 years of age, born in Denmark and free of cancer at the time of enrolment.15 The participants were followed from the date of their baseline examination until cancer diagnosis, death, emigration or 27 April 2006, whichever occurred first. All participants gave written informed consent, and Diet, Cancer and Health was approved by the ethical committees in Aarhus and Copenhagen. Information on vital status and emigration was obtained by linking cohort members to the Civil Registration System using the unique civil registration number assigned to all Danish residents. Information on cancer diagnoses was found in three national registries: the Danish Cancer Registry, which holds information on all cancer cases in the Danish population owing to mandatory reporting since 1987,16 the Pathology Bank and the Danish Breast Cancer Cooperative Group. From this latter registry, information on receptor status was obtained, but information on PR status was not registered systematically.17 A standardized immunohistologic method was used to detect ER and PR, and receptor status was defined as positive if ⩾ 10% of tumour cells possessed the receptors. During follow-up, 1074 women were diagnosed with breast cancer5 (Figure 1). Only women with the two most common types of breast cancer—invasive ductal and lobular carcinoma—were included in the present study corresponding to 95% of breast cancer cases. Fat biopsies from 409 breast cancer cases diagnosed between baseline and 31 December 2000 had been used in an earlier study,18 leaving no fat for analysis of fatty acids. Thus 1 January 2001 was defined as the start date of follow-up for the present study. All participants who were diagnosed with cancer (breast cancer or other types), had died, or had emigrated before this date were excluded, leaving 29 207 women in the study cohort. In order to economize the use of fat biopsies, a case-cohort design was used. Accordingly, the study population consisted of all breast cancer cases identified between 1 January 2001 and 27 April 2006, alongside a subcohort of 1215 participants randomly sampled among participants who were still in the cohort after the change in the start of the follow-up. Breast cancer cases for whom the diagnosis could not be histologically validated were excluded. The validation was done by reviewing pathology reports from the Pathology Bank and comparing this information with information from the other registries. Furthermore, participants with an incomplete fatty acid profile were excluded. Before analysis of breast cancer risk, participants with missing data on one or more of the potential covariates were excluded.
Baseline examination 1993–1997 n = 29 875 women
End of follow-up, 27 April 2006 n = 29 875
Cases n = 1074
Censored before 1 January 2001
Cohort after change of start of follow-up n = 29 207
Cases n = 500
Participants not in the sub-cohort Sub-cohort n = 1215
n = 17 Cases n = 500 Non-validated cases n = 23 Incomplete fatty acid profiles n = 105
Study population n = 1570
Sub-cohort n = 1112
n = 16 Cases n = 474 Missing covariates n = 57
Fatty acid analysis An adipose tissue biopsy of 35–50 mg subcutaneous fat was taken from the upper, outer part of the buttock at the baseline examination using a luer-lock system (Terumo, Terumo Corporation, Tokyo, Japan), consisting of a needle, a connector (Venoject multisample Luer adaptor) and an evacuated blood tube.15 The biopsies were stored in the connector in liquid nitrogen at − 150 °C until analysis. Studies have shown that longterm storage of fat biopsies only caused minor changes in fatty acids composition.6,19 Before analysis, the biopsies were defrosted and 2–4 mg of fat was removed from the connector. The fat was dissolved in heptane and fatty acids were trans-esterified by potassium hydroxide in methanol consistent with standard methods.20 The composition of fatty acids was analysed by gas chromatography using a Varian 3900 GC with a CP-8400 auto sampler and CP-Sil 88, 60 m × 0.25 mm ID capillary column (Varian, Middelburg, Netherlands) and using helium as carrier gas. Identification of individual fatty acids was performed using commercial standards (Nu-Chek-Prep, Inc., Elysian, MN, USA) that enable quantification of fatty acid methyl esters with 12–24 carbon atoms, and 31 fatty acids were determined (Table 1). The combinations of 18:1 n-6 (Δ12t) and 18:1 n-8 (Δ10t), and 18:1 n-10 (Δ8t) and 18:1 n-12 (Δ6t) were analysed owing to unsuccessful separation of peaks for these fatty acids.12 The inter-assay coefficients of variation for the fatty acids ranged from 0.7 to 11.7%.
Assessment of covariates At the baseline examination, participants filled in a self-administered questionnaire on lifestyle, medical history and reproductive factors. The questionnaires were checked for missing information and errors by optical scanning immediately after completion. Any unclear information was corrected during an interview performed by a trained lab technician who also measured weight and height for calculation of body mass index.15 © 2014 Macmillan Publishers Limited
Analysis population n = 1513
Sub-cohort n = 1070
n = 16 Cases n = 459
Figure 1. Flowchart of participant selection from baseline examination to analysis population.
Statistical analyses Baseline characteristics were examined for cases and members of the subcohort separately, and the average percentages of adipose tissue fatty acids in sub-cohort members were examined. To derive patterns of fatty acids we used the TT. This is a dimensionreduction method, which uses the correlation matrix of multidimensional data like fatty acids to derive patterns, or factors, which are in reality vectors.13 The numeric size of the original fatty acid in a factor is called the loading. Because analysing the correlation matrix standardizes the fatty acid variables, a positive loading corresponds to a larger-than-average value of the fatty acid; a zero loading corresponds to an average value; and a negative loading corresponds to a smaller-than-average value of the fatty acid. Factors derived by TT are sparse so that not all fatty acids contribute to every factor. Participants are assigned a factor score using the individual-level weights associated with each factor. A high score for a particular factor corresponds to a participant having a fatty acid profile that matches the factor loadings well. TT was performed according to the method described by GorstRasmussen et al.13,21 and based on fatty acids from members of the subcohort (n = 1112). In brief, closely correlated fatty acids were joined into one variable (depicted by a cluster tree; Figure 2), which had to be cut to derive factors. Initially, the tree was cut at a range of cut-levels (from 22 to 26) to determine how many factors to retain; namely factors with variance ⩾ 1.25.22 European Journal of Clinical Nutrition (2014) 1088 – 1094
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1090 Table 1. Mean (s.d.) percentage of fatty acids in adipose tissue, their loadings on treelet transform (TT) factors and the explained variance of extracted factors, in the study population sub-cohort (n = 1112) Fatty acid
Common name
Mean (s.d.)
Factor loading TT1
12:0 14:0 15:0 16:0 17:0 18:0 19:0 20:0 14:1 n-5 16:1 n-7 18:1 n-7 18:1 n-9 20:1 n-9 20:1 n-11 22:1 n-9 18:1 n-6+8 (Δ12t+Δ10t)a 18:1 n-9 (Δ9t) 18:1 n-10+12 (Δ8t+Δ6t)a 18:1 n-7 (Δ11t) 18:2 n-7 (Δ9c11t) 18:2 n-6 18:3 n-3 18:3 n-6 20:2 n-6 20:3 n-6 20:4 n-3 20:4 n-6 20:5 n-3 22:4 n-6 22:5 n-3 22:6 n-3 Explained variance (%)
Lauric acid Myristic acid Pentadecanoic acid Palmitic acid Heptadecanoic acid Stearic acid Nonadecanoic acid Arachidic acid Myristoleic acid Palmitoleic acid Asclepic acid Oleic acid n-9 eicosenoic acid n-11 eicosenoic acid Eurcic acid Elaidic acid Vaccenic acid Rumenic acid Linoleic acid α-linolenic acid γ-linolenic acid Eicosadienoic acid Dihomo-γ-linolenic acid Arachidonic acid Arachidonic acid Eicosapentaenoic acid Docosatetraenoic acid Docosapentaenoic acid Docosahexaenoic acid
0.4 2.7 0.3 19.3 0.21 3.2 0.1 0.2 0.5 7.1 2.2 44.3 0.8 0.2 0.05 0.3 0.5 0.3 0.3 0.5 10.9 0.8 0.1 0.2 0.2 0.1 0.4 0.1 0.1 0.3 0.3
(0.1) (0.5) (0.1) (1.9) (o 0.1) (0.9) (o 0.1) (0.1) (0.1) (1.5) (0.4) (1.9) (0.1) (0.1) (o 0.1) (0.1) (0.2) (0.1) (0.1) (0.1) (2.0) (0.2) (o 0.1) (o 0.1) (0.1) (0.3) (0.1) (0.1) (o 0.1) (0.1) (0.1)
a
TT2
European Journal of Clinical Nutrition (2014) 1088 – 1094
TT4
TT5
TT6
TT7
0.3 0.4 0.4 0.2 0.4 0.4 0.4 0.5 0.6 0.5 0.4 0.5 0.5 0.5 0.6 0.6 0.6 0.4 0.6 0.6 0.5 0.6 0.5 0.5 0.5 0.5 0.5 15.1
0.5 0.5 10.5
Peaks for 18:1n-10 (Δ8t) and 18:1n-12 (Δ6t), and 18:1n-6 (Δ12t) and 18:1n-8 (Δ10t) could not be separated.
The optimal cut-level for this number of factors was then determined by use of a 10-fold cross-validation.13,21 The TT tree was cut at the optimal level, the factors extracted and participants were assigned factor scores for each factor according to their own fatty acid profiles. To examine the association between patterns of fatty acids and total breast cancer risk, factor scores were divided into quintiles based on members of the study population sub-cohort. Hazard ratios (HR) and 95% confidence intervals (CI) for each quintile were calculated relative to the first quintile by weighted Cox proportional hazards regression using age as the time axis and robust variance estimates.23 Cases were assigned weight 1, whereas non-cases were assigned weight N/n; N and n being the number of non-cases in the study cohort and the sub-cohort, respectively. A P-value for trend (P(trend)) was calculated by entering quintiles of factor scores as continuous ordinal variables in the regressions. Two-sided P-values o0.05 were considered statistically significant in all analyses. An unadjusted and one adjusted regression models were explored. The adjusted model included: body mass index (o 20; 20–o25; 25–o 30; ⩾30 kg/m2), smoking (never; former; current o15; current 15–25; current >25 g tobacco/day), alcohol consumption (g/day, as a continuous variable), hormone replacement therapy (never; former; current), number of childbirths (0; 1; 2; 3; ⩾4), age at first childbirth (years, as a continuous variable), age at menarche (o 12; 12–o14; ⩾14 years; unknown), benign breast tumour surgery (yes; no; unknown), physical activity (⩽3.5; >3.5 h/week) and education (⩽7; 8–10; >10 years). The proportional hazard assumption was visually examined by using plots of log(− log(probability of being free of breast cancer)) and no appreciable violations were found. The association between patterns of fatty acids and risk of breast cancer subtypes, ER+, ER − , PR+ and PR − , was analysed as described for total breast cancer risk. TT factor scores can be correlated, potentially leading to confounding.13 Thus, a sensitivity analysis was performed: the factor-specific adjusted
TT3
9.3
9.2
9.0
7.4
6.0
12
regressions for total breast cancer risk were further adjusted (using quintiles) for all factor scores that had an absolute correlation > 0.2 with the score of the factor under investigation. All analyses were performed in STATA, version 11 (StataCorp LP, College Station, TX, USA).
RESULTS After exclusion of non-validated cases (n = 23) and participants with incomplete fatty acid profiles (n = 105), the study population consisted of 1570 women (Figure 1). Of these, 474 were breast cancer cases, 1112 were members of the sub-cohort and 16 were both. Information on covariates was missing for 57 of these women who were excluded before analyses of breast cancer risk. Thus, these analyses were based on 459 cases and 1070 members of the sub-cohort including 16 cases. The median duration of follow-up was 5 years and 4 months. The mean age at start of the follow-up was 61 years (Table 2). Cases were more often current users of hormone replacement therapy, had higher alcohol consumption, had more often had benign breast tumour surgery and had given birth fewer times than members of the sub-cohort. Body mass index, smoking, age at first childbirth and menarche, physical activity and level of education were more equally distributed between the groups. The most abundant fatty acids in adipose tissue were 18:1 n-9, 16:0 and 18:2 n-6 constituting 44.3% (s.d. 1.9), 19.3% (s.d. 1.9) and 10.9% (s.d. 2.0) of total fatty acids, respectively (Table 1). © 2014 Macmillan Publishers Limited
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1091 Cut–level 0 12:0 14:0 15:0 17:0 18:0 19:0 18:1 n–7 (Δ11t) 16:0 20:0 22:1 n–9 20:1 n–9 20:1 n–11 18:1 n–10+12 (Δ6t+Δ8t) 18:1 n−9 (Δ9t) 18:1 n–6+8 (Δ12t+Δ10t) 18:3 n–6 14:1 n–5 18:2 n–7 (Δ9c11t) 16:1 n–7 18:1 n–9 18:1 n–7 20:3 n–6 20:4 n–6 22:4 n–6 20:4 n–3 20:5 n–3 22:6 n–3 22:5 n–3 18:2 n–6 20:2 n–6 18:3 n–3
5
10
15
20
25
30
35
Table 2.
Baseline characteristics, measured in 1993 − 1997, of members of the sub-cohort (n = 1112) and the breast cancer cases (n = 474) in the study population n (%)/median (5th; 95th percentile)a
Variables TT1
Sub-cohort (n = 1112)b Cases (n = 474)c Age at start of the follow-up (years) Median (5th; 95th percentile)
TT3
60.9 (55.0; 69.0)
61.5 (55.1; 69.1)
Body mass index (kg/m ) o20 20–o25 25–o30 ⩾ 30
58 541 376 137
(5.2) (48.7) (33.8) (12.3)
18 247 145 64
(3.8) (52.1) (30.6) (13.5)
Smoking (g tobacco/day) Never Former Current o15 Current 15–25 Current >25
478 262 172 175 25
(43.0) (23.6) (15.5) (15.7) (2.3)
215 109 76 60 14
(45.4) (23.0) (16.0) (12.7) (3.0)
2
TT5
TT6
TT4
TT2
Alcohol consumption (g/day) Median (5th; 95th percentile)
TT7
Figure 2. Cluster tree produced by the treelet transform (TT) on adipose tissue fatty acid variables among women in the study population sub-cohort, n = 1112, cut at level 22 (dotted line), retaining 7 factors.
The TT identified seven factors, or patterns, with variances of ⩾ 1.25. These factors explained 66.5% of the total variance, and fatty acid loadings for each factor were positive (Table 1). Fatty acids joined at the optimal cut-level 22 were grouped in the same pattern and all except 18:1 n-9 and 18:3 n-6 were contained in exactly one pattern (Figure 2). TT2 and TT5 were comprised solely of n-3 PUFA and trans fatty acids, respectively. TT1, TT3 and TT4 were dominated by SFA, long-chain MUFA and n-6 PUFA, respectively, whereas MUFA and n-6 PUFA were the main fatty acids in TT6 and TT7, respectively. No clear associations between patterns of fatty acids and risk of total breast cancer were observed (Table 3). A slight tendency towards higher risk of breast cancer with higher factor score for TT3, TT4 and TT5 was observed, but no clear dose-response associations were found. Of the 459 cases in the analysis population, 384 (84%) had ER+ tumours, 59 (13%) had ER − tumours, 162 (35%) had PR+ tumours and 91 (20%) had PR − tumours. ER and PR status was unknown for 16 (3%) and 206 (45%) cases, respectively. Similar to the analysis of total breast cancer risk, no clear associations between patterns of fatty acids and risk of ER+, ER − , PR+ or PR − tumours were found (Table 4). Only weak tendencies towards a higher risk of ER+ tumours for TT5 and a higher risk of ER − tumours for TT4 and TT7 were observed. In contrast, a slight tendency towards a lower risk of PR+ tumours for TT2 was observed. In general, the correlations between factor scores were substantial; approximately half of the correlations (11 out of 21) were more extreme than the chosen cutoff at ± 0.2 (data not shown). When these factor scores were simultaneously added in quintiles to the adjusted model for total breast cancer risk, the association for both TT3 and TT4 with risk of breast cancer were strengthened, and both reached statistical significance (HR (95% CI) for TT3 adjusted for TT1, TT4 and TT5 in the 5th quintile was 1.42 (0.91–2.21), P(trend) = 0.04; for TT4 adjusted for TT1–TT3 and TT5 the corresponding numbers were 1.54 (0.92–2.58), P(trend) = 0.03). However, no clear dose-response relationships were found. © 2014 Macmillan Publishers Limited
9.2 (1.2; 40.9)
10.7 (0.6; 42.4)
Hormone replacement therapy Never Former Current
578 (53.0) 169 (15.5) 344 (31.5)
186 (39.8) 61 (13.1) 220 (47.1)
Number of childbirths 0 1 2 3 ⩾4
129 171 518 225 61
76 85 210 78 25
Age at first childbirthd (years) Median (5th; 95th percentile)
(11.7) (15.5) (46.9) (20.4) (5.5)
23 (18; 32)
(16.0) (17.9) (44.3) (16.5) (5.3)
24 (18; 32)
Age at menarche (years) o12 12–o14 ⩾ 14 years Unknown
97 393 581 41
Benign breast tumour surgery Yes No Unknown
142 (12.8) 966 (87.1) 1 (0.1)
88 (18.6) 383 (80.8) 3 (0.6)
Physical activity (h/week) o3.5 ⩾ 3.5
718 (64.6) 394 (35.4)
309 (65.2) 165 (34.8)
Education (years) ⩽7 8–10 >10
363 (32.7) 555 (50.0) 192 (17.3)
136 (28.7) 247 (52.1) 91 (19.2)
(8.7) (35.4) (52.3) (3.7)
43 186 228 17
(9.1) (39.2) (48.1) (3.6)
Data are n (%) if not specified as median (5th; 95th percentile). bn = 1091 for hormone replacement therapy, 1104 for number of childbirths, 1109 for benign breast surgery and 1110 for education, owing to missing data. c n = 467 for hormone replacement therapy, owing to missing data. dOnly for women who had given birth; cases, n = 394; sub-cohort, n = 967. a
DISCUSSION In the present study, no clear associations between patterns of fatty acids and long-term risk of breast cancer or risk of ER+, ER − , PR+ or PR − tumours were observed. After adjusting for correlated factor European Journal of Clinical Nutrition (2014) 1088 – 1094
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1092 Table 3.
Risk of breast cancer expressed as hazard ratios (HR) according to quintiles of treelet transform (TT) factor scores for patterns of adipose tissue fatty acids (n = 1513) HR (95% CI)a
Fatty acids Factor score pattern quintile
Unadjusted TT1
TT2
TT3
TT4
TT5
TT6
TT7
1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend)
1.00 0.82 0.93 1.02 1.04 1.00 0.81 0.85 1.03 0.88 1.00 0.80 1.22 1.06 1.19 1.00 0.96 1.04 1.21 1.04 1.00 0.93 0.98 1.13 0.99 1.00 1.30 1.02 1.39 1.13 1.00 1.19 0.90 1.13 0.99
(reference) (0.58–1.18) (0.66–1.31) (0.72–1.43) (0.74–1.46) 0.47 (reference) (0.57–1.15) (0.60–1.21) (0.73–1.44) (0.62–1.25) 0.99 (reference) (0.55–1.15) (0.86–1.71) (0.75–1.51) (0.84–1.68) 0.12 (reference) (0.67–1.37) (0.73–1.48) (0.86–1.69) (0.73–1.48) 0.42 (reference) (0.65–1.32) (0.70–1.39) (0.81–1.59) (0.70–1.40) 0.65 (reference) (0.92–1.85) (0.71–1.48) (0.98–1.89) (0.79–1.63) 0.44 (reference) (0.85–1.67) (0.63–1.28) (0.81–1.59) (0.70–1.41) 0.87
Adjustedb 1.00 0.74 0.86 0.98 1.01 1.00 0.80 0.90 1.10 0.98 1.00 0.78 1.17 1.04 1.27 1.00 0.97 1.05 1.38 1.12 1.00 1.11 1.13 1.45 1.20 1.00 1.24 0.96 1.32 1.00 1.00 1.29 1.00 1.31 1.09
(reference) (0.51–1.09) (0.60–1.25) (0.68–1.41) (0.69–1.47) 0.51 (reference) (0.55–1.17) (0.62–1.30) (0.76–1.60) (0.66–1.47) 0.53 (reference) (0.52–1.16) (0.81–1.69) (0.69–1.55) (0.85–1.91) 0.10 (reference) (0.66–1.42) (0.72–1.55) (0.95–2.01) (0.75–1.68) 0.20 (reference) (0.76–1.62) (0.77–1.64) (0.99–2.11) (0.80–1.81) 0.17 (reference) (0.85–1.80) (0.65–1.41) (0.92–1.90) (0.68–1.48) 0.90 (reference) (0.90–1.85) (0.68–1.46) (0.91–1.89) (0.74–1.60) 0.65
Abbreviation: CI, confidence interval. aHazard ratios were calculated by use of the weighted Cox proportional hazard regression. bAdjusted for body mass index ( o20; 20–o25; 25–o30; ⩾30 kg/m2), smoking (never; former; current o15; current 15–25; current >25 g tobacco/day), alcohol consumption (g/day, as a continuous variable), hormone replacement therapy (never; former; current), number of childbirths (0; 1; 2; 3; ⩾ 4), age at first childbirth (years, as a continuous variable), age at menarche (o 12; 12–o 14; ⩾ 14 years; unknown), benign breast tumour surgery (yes; no; unknown), physical activity (⩽3.5; >3.5 h/week) and education (⩽ 7; 8–10; >10 years).
scores, positive associations were observed between the risk of breast cancer and TT3 and TT4 (dominated by MUFA and n-6 PUFA, respectively). However, interpretation of the fully adjusted model is complex. Common dietary sources partly explain the correlations and thus it is not possible to change the factor score for one pattern without affecting ones factor score in other patterns. The fatty acids grouped in patterns were only to some extent comparable to the traditional classes of fatty acids based on chemical structure. This could be expected when the correlations between fatty acids induced by dietary sources, metabolic pathways and the complex European Journal of Clinical Nutrition (2014) 1088 – 1094
interplay of product inhibition, genetics and fatty acids competing for enzymes7 are taken into consideration. The results of this study must be interpreted in light of the potential for bias and confounding. First, selection bias owing to misclassification of breast cancer cases is unlikely to have affected the results, as cases were identified via three national registers, which contain information on all cancer diagnoses in Danish residence, and diagnoses were subsequently validated. However, the lack of information on the fatty acid composition from women developing breast cancer close to baseline is a limitation. This precluded assessment of short-term effects of fatty acid patterns on breast cancer, and it might have resulted in selection bias.24 Breast cancer occurring shortly after baseline is likely to be in women who are susceptible to a potential effect of fatty acid composition on breast cancer, but these are overlooked in the present analysis. Thus, the study cohort is likely to comprise less susceptible women compared with all women recruited in the original Diet, Cancer and Health cohort. Second, adipose tissue fatty acids do not possess the weaknesses related to collection of dietary information, such as erroneous reporting,25 reducing the potential for information bias. Furthermore, the investigation of patterns of fatty acids identified by TT allowed us to investigate the naturally occurring patterns in fatty acid composition and did not limit us to investigation of individual fatty acids, where information on these correlations would be overlooked. However, the patterns identified by TT explain variation of fatty acids in the population, and the biological relevance of mathematically derived patterns may be questioned.26,27 Just one measurement of fatty acids composition was available, and as the composition might change over time owing to changes in diet or metabolism, our results may underestimate effect sizes. Finally, to address confounding, several factors related to breast cancer were included in the multivariate regression models. In addition, a sensitivity analysis further adjusting for correlated fatty acid patterns was conducted. However, residual confounding may still have affected our results. Studies have indicated that the association between tissue fatty acids and risk of breast cancer might differ by menopausal status.28,29 However, owing to the small number of women still premenopausal at the start of follow-up in the study, we were unable to investigate this association stratified by menopausal status. Only one previous study has investigated the relationship between breast cancer risk and patterns of fatty acids (determined by principal component analysis).11 This case-control study found a protective effect of a pattern characterised by a low n-6/n-3 PUFA ratio, low levels of 18:2 n-6 and high levels of MUFA. A pattern characterised by low levels SFA and high levels of n-6 PUFA was not associated with risk of breast cancer. Other studies of fat composition in tissue stores all investigated individual fatty acids or the traditional classes of fatty acids. The majority of cohort studies investigating SFA,10,30–33 n-6 PUFA,28,30,33,34 trans fatty acids28,31,34 or MUFA 28,30,33,34 in relation to breast cancer risk have not found any associations. However, other studies have reported positive associations between breast cancer risk and SFA,28 trans fatty acids30 or MUFA,31 whereas others again have found inverse associations with SFA34 or n-6 PUFA10,32 in cohort studies. A meta-analysis published in 2004 of cohort and case-control studies is suggestive of a protective effect of n-3 PUFA.10 Some studies published later have supported this finding, over all35,36 or for premenopausal women only,29 whereas others have not.5,30,33,37,38 The discrepancy might be attributed to the proportions of n-3 PUFA in tissue stores. In most studies suggestive of a protective effect, the women had a high proportion of n-3 PUFA,31,35,36 but in studies not documenting an association, the women had low proportions of n-3 PUFA in the tissue.5,28,37–40 This could indicate a threshold for a protective effect of n-3 PUFA on breast cancer risk. Our findings of no clear © 2014 Macmillan Publishers Limited
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1093 Table 4. Risk of breast cancer subtypes characterised by receptor status expressed as hazard ratios (HR) according to quintiles of treelet transform (TT) factor scores for patterns of adipose tissue fatty acids Fatty acids pattern
HR (95% CI)a
Factor score quintile ER+b n (case) = 384
TT1
TT2
TT3
TT4
TT5
TT6
TT7
1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend) 1st 2nd 3rd 4th 5th P(trend)
1.00 0.71 0.78 0.99 0.98 1.00 0.82 0.85 1.13 1.01 1.00 0.68 1.02 1.01 1.20 1.00 0.89 0.92 1.35 1.11 1.00 1.19 1.18 1.47 1.25 1.00 1.20 0.93 1.31 0.96 1.00 1.33 1.01 1.38 0.04
(reference) (0.48–1.07) (0.52–1.16) (0.67–1.46) (0.65–1.46) 0.55 (reference) (0.55–1.23) (0.57–1.26) (0.76–1.67) (0.66–1.54) 0.51 (reference) (0.45–1.04) (0.69–1.51) (0.67–1.55) (0.78–1.84) 0.14 (reference) (0.59–1.34) (0.61–1.39) (0.91–2.02) (0.72–1.71) 0.21 (reference) (0.80–1.78) (0.78–1.77) (0.98–2.20) (0.81–1.93) 0.17 (reference) (0.80–1.78) (0.61–1.40) (0.89–1.93) (0.64–1.46) 0.95 (reference) (0.90–1.97) (0.67–1.52) (0.94–2.05) (0.69–1.57) 0.76
ER − b n (case) = 59 1.00 0.99 1.34 0.94 0.83 1.00 0.93 1.31 1.21 1.11 1.00 1.11 2.18 0.80 2.12 1.00 1.74 2.53 1.78 1.69 1.00 0.62 0.57 1.26 0.97 1.00 1.56 1.24 1.32 1.15 1.00 0.71 1.25 1.29 1.77
(reference) (0.36–2.71) (0.54–3.35) (0.37–2.39) (0.30–2.30) 0.66 (reference) (0.35–2.47) (0.51–3.39) (0.45–3.28) (0.40–3.09) 0.70 (reference) (0.40–3.11) (0.88–5.42) (0.23–2.73) (0.79–5.67) 0.22 (reference) (0.60–5.00) (0.94–6.86) (0.64–4.98) (0.58–4.93) 0.35 (reference) (0.24–1.62) (0.23–1.40) (0.55–2.87) (0.37–2.53) 0.62 (reference) (0.60–4.06) (0.46–3.31) (0.50–3.52) (0.41–3.24) 0.98 (reference) (0.27–1.88) (0.53–2.98) (0.53–3.15) (0.74–4.26) 0.12
PR+b n (case) = 162 1.00 0.64 0.66 0.86 1.11 1.00 0.61 0.65 0.76 0.59 1.00 0.51 0.79 0.90 0.80 1.00 0.73 1.14 1.40 1.12 1.00 1.38 1.13 1.56 1.07 1.00 1.45 1.01 1.20 0.98 1.00 1.10 0.69 1.02 0.90
(reference) (0.36–1.15) (0.37–1.18) (0.48–1.55) (0.62–1.98) 0.49 (reference) (0.34–1.08) (0.37–1.12) (0.44–1.33) (0.32–1.09) 0.22 (reference) (0.27–0.95) (0.45–1.38) (0.49–1.65) (0.43–1.51) 0.87 (reference) (0.40–1.35) (0.64–2.01) (0.80–2.46) (0.60–2.07) 0.21 (reference) (0.79–2.39) (0.62–2.07) (0.89–2.74) (0.56–2.06) 0.62 (reference) (0.83–2.55) (0.56–1.84) (0.68–2.11) (0.54–1.80) 0.69 (reference) (0.65–1.86) (0.38–1.25) (0.59–1.77) (0.51–1.58) 0.66
PR − b n (case) = 91 1.00 0.79 0.97 0.86 1.04 1.00 0.82 1.10 1.45 1.09 1.00 0.90 1.79 0.50 1.70 1.00 0.87 1.58 1.63 1.11 1.00 0.77 0.70 1.00 0.97 1.00 1.54 1.15 1.89 0.95 1.00 1.11 1.22 1.61 1.11
(reference) (0.35–1.79) (0.46–2.04) (0.42–1.76) (0.50–2.18) 0.84 (reference) (0.38–1.78) (0.51–2.37) (0.68–3.06) (0.48–2.45) 0.41 (reference) (0.41–1.98) (0.90–3.57) (0.18–1.36) (0.79–3.65) 0.29 (reference) (0.39–1.96) (0.73–3.43) (0.75–3.55) (0.48–2.52) 0.39 (reference) (0.36–1.64) (0.33–1.46) (0.50–2.00) (0.45–2.11) 0.88 (reference) (0.71–3.34) (0.50–2.62) (0.89–4.01) (0.41–2.24) 0.92 (reference) (0.54–2.31) (0.59–2.55) (0.78–3.32) (0.50–2.45) 0.47
Abbreviations: CI, confidence interval; ER ± , oestrogen receptor positive/negative; PR ± , progesterone receptor positive/negative. aHazard ratios were calculated by use of the weighted Cox proportional hazard regression and were adjusted for body mass index ( o 20; 20–o 25; 25–o 30; ⩾30 kg/m2), smoking (never; former; current o 15; current 15–25; current >25 g tobacco/day), alcohol consumption (g/day, as a continuous variable), hormone replacement therapy (never; former; current), number of childbirths (0; 1; 2; 3; ⩾ 4), age at first childbirth (years, as a continuous variable), age at menarche (o 12; 12–o14; ⩾ 14 years; unknown), benign breast tumour surgery (yes; no; unknown), physical activity (⩽ 3.5; >3.5 h/week) and education (⩽ 7; 8–10; >10 years). bNumbers of sub-cohort members: ER, n = 1069; PR, n = 1063.
association between TT2 (comprised solely of n-3 PUFA) and risk of breast cancer would be consistent with a threshold effect, as proportions of n-3 PUFA in the present study were low. In conclusion, no clear associations between the baseline patterns of fatty acids identified by TT and long-term risk of breast cancer or of ER+, ER − , PR+ or PR − tumours were observed. These findings indicate that the dominating patterns of adipose tissue fatty acids, explaining the most variance in the population, do not affect the long-term risk of breast cancer.
ACKNOWLEDGEMENTS We thank the participants of Diet, Cancer and Health for their invaluable contribution to the study. We also acknowledge the Danish Cancer Society for supporting this study financially.
REFERENCES
The authors declare no conflict of interest.
1 Ferlay J, Shin H-R, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer 2010; 127: 2893–2917. 2 Tibor T. The theory of the sick lobe. In: Tibor T (ed). Breast Cancer—A Lobular Disease. 1st edn. Springer-Verlag: London, UK, 2011, pp 1–17. 3 Yang XR, Chang-Claude J, Goode EL, Couch FJ, Nevanlinna H, Milne RL et al. Associations of breast cancer risk factors with tumor subtypes: a pooled analysis
© 2014 Macmillan Publishers Limited
European Journal of Clinical Nutrition (2014) 1088 – 1094
CONFLICT OF INTEREST
Adipose tissue fatty acid patterns and breast cancer JA Schmidt et al
1094 4 5
6
7 8
9
10
11 12
13
14
15
16 17
18
19
20
21
from the Breast Cancer Association Consortium studies. J Natl Cancer Inst 2011; 103: 250–263. Turner LB. A meta-analysis of fat intake, reproduction, and breast cancer risk: an evolutionary perspective. Am J Hum Biol 2011; 23: 601–608. Witt PM, Christensen JH, Schmidt EB, Dethlefsen C, Tjønneland A, Overvad K et al. Marine n-3 polyunsaturated fatty acids in adipose tissue and breast cancer risk: a case-cohort study from Denmark. Cancer Causes Control 2009; 20: 1715–1721. Hodson L, Skeaff CM, Fielding BA. Fatty acid composition of adipose tissue and blood in humans and its use as a biomarker of dietary intake. Prog Lipid Res 2008; 47: 348–380. Vessby B, Gustafsson IB, Tengblad S, Boberg M, Andersson A. Desaturation and elongation of fatty acids and insulin action. Ann N Y Acad Sci 2002; 967: 183–195. Simonsen NR, Fernandez-Crehuet Navajas J, Martin-Moreno JM, Strain JJ, Huttunen JK, Martin BC et al. Tissue stores of individual monounsaturated fatty acids and breast cancer: the EURAMIC study. European Community Multicenter Study on Antioxidants, Myocardial Infarction, and Breast Cancer. Am J Clin Nutr 1998; 68: 134–141. Tjønneland A, Overvad K, Thorling E, Ewertz M. Adipose tissue fatty acids as biomarkers of dietary exposure in Danish men and women. Am J Clin Nutr 1993; 57: 629–633. Saadatian-Elahi M, Norat T, Goudable J, Riboli E. Biomarkers of dietary fatty acid intake and the risk of breast cancer: a meta-analysis. Int J Cancer 2004; 111: 584–591. Bougnoux P, Giraudeau B, Couet C. Diet, cancer, and the lipidome. Cancer Epidemiol Biomarkers Prev 2006; 15: 416–421. Dahm CC, Gorst-Rasmussen A, Jakobsen MU, Schmidt EB, Tjønneland A, Sørensen TI et al. Adipose tissue fatty acid patterns and changes in anthropometry: a cohort study. PLoS One 2011; 6: e22587. Gorst-Rasmussen A, Dahm CC, Dethlefsen C, Scheike T, Overvad K. Exploring dietary patterns by using the treelet transform. Am J Epidemiol 2011; 173: 1097–1104. Riboli E, Kaaks R. The EPIC Project: rationale and study design. European Prospective Investigation into Cancer and Nutrition. Int J Epidemiol 1997; 26: S6–14. Tjønneland A, Olsen A, Boll K, Stripp C, Christensen J, Engholm G et al. Study design, exposure variables, and socioeconomic determinants of participation in Diet, Cancer and Health: a population-based prospective cohort study of 57 053 men and women in Denmark. Scand J Public Health 2007; 35: 432–441. Storm HH, Michelsen EV, Clemmensen IH, Pihl J. The Danish Cancer Registry – history, content, quality and use. Dan Med Bull 1997; 44: 535–539. Olsen A, Knudsen KE, Thomsen BL, Loft S, Stripp C, Overvad K et al. Plasma enterolactone and breast cancer incidence by estrogen receptor status. Cancer Epidemiol Biomarkers Prev 2004; 13: 2084–2089. Raaschou-Nielsen O, Pavuk M, Leblanc A, Dumas P, Philippe Weber J, Olsen A et al. Adipose organochlorine concentrations and risk of breast cancer among postmenopausal Danish women. Cancer Epidemiol Biomarker s Prev 2005; 14: 67–74. Beynen AC, Katan MB. Rapid sampling and long-term storage of subcutaneous adipose-tissue biopsies for determination of fatty acid composition. Am J Clin Nutr 1985; 42: 317–322. IUPAC Methods 2.301. IUPAC Standard Methods of Oil, Fats and Derivatives, Report of IUPAC Working Group WG 2/87. 7th edn. Blackwell Scientific Publications: Oxford, UK, 1987. Gorst-Rasmussen A. tt: Treelet transform with Stata. Stata J 2012; 1: 130–146.
European Journal of Clinical Nutrition (2014) 1088 – 1094
22 Slattery ML, Boucher KM, Caan BJ, Potter JD, Ma KN. Eating patterns and risk of colon cancer. Am J Epidemiol 1998; 148: 4–16. 23 Kalbfleisch JD, Lawless JF. Likelihood analysis of multi-state models for disease incidence and mortality. Stat Med 1988; 7: 149–160. 24 Hernan MA. The hazards of hazard ratios. Epidemiology 2010; 21: 13–15. 25 Bates CJ, Nelson M, Ulijaszek SJ. Nutritional assessment methods. In: Geissler CA, Powers HJ (eds) Human Nutrition. 11th edn. Churchill Livingstone: Edinburgh, UK, 2005, pp 573–595. 26 Jacques PF, Tucker KL. Are dietary patterns useful for understanding the role of diet in chronic disease? Am J Clin Nutr 2001; 73: 1–2. 27 Martinez ME, Marshall JR, Sechrest L. Invited commentary: factor analysis and the search for objectivity. Am J Epidemiol 1998; 148: 17–19. 28 Saadatian-Elahi M, Toniolo P, Ferrari P, Goudable J, Akhmedkhanov A, Zeleniuch-Jacquotte A et al. Serum fatty acids and risk of breast cancer in a nested case-control study of the New York University Women's Health Study. Cancer Epidemiol Biomarkers Prev 2002; 11: 1353–1360. 29 De Bree E, Mamalakis G, Sanidas E, Hatzis C, Askoxylakis I, Daskalakis M et al. Adipose tissue fatty acid composition in Greek patients with breast cancer versus those with benign breast tumors. Anticancer Res 2013; 33: 1667–1672. 30 Chajes V, Thiebaut AC, Rotival M, Gauthier E, Maillard V, Boutron-Ruault MC et al. Association between serum trans-monounsaturated fatty acids and breast cancer risk in the E3N-EPIC Study. Am J Epidemiol 2008; 167: 1312–1320. 31 Pala V, Krogh V, Muti P, Chajes V, Riboli E, Micheli A et al. Erythrocyte membrane fatty acids and subsequent breast cancer: a prospective Italian study. J Natl Cancer Inst 2001; 93: 1088–1095. 32 Rissanen H, Knekt P, Jarvinen R, Salminen I, Hakulinen T. Serum fatty acids and breast cancer incidence. Nutr Cancer 2003; 45: 168–175. 33 Wirfalt E, Vessby B, Mattisson I, Gullberg B, Olsson H, Berglund G. No relations between breast cancer risk and fatty acids of erythrocyte membranes in postmenopausal women of the Malmo Diet Cancer cohort (Sweden). Eur J Clin Nutr 2004; 58: 761–770. 34 Chajes V, Hulten K, Van Kappel AL, Winkvist A, Kaaks R, Hallmans G et al. Fatty acid composition in serum phospholipids and risk of breast cancer: a prospective cohort study in Northern Sweden. Int J Cancer 1999; 26: 585–590. 35 Kuriki K, Hirose K, Wakai K, Matsuo K, Ito H, Suzuki T et al. Breast cancer risk and erythrocyte compositions of n-3 highly unsaturated fatty acids in Japanese. Int J Cancer 2007; 121: 377–385. 36 Shannon J, King IB, Moshofsky R, Lampe JW, Gao DL, Ray RM et al. Erythrocyte fatty acids and breast cancer risk: a case-control study in Shanghai, China. Am J Clin Nutr 2007; 85: 1090–1097. 37 Bagga D, Anders KH, Wang HJ, Glaspy JA. Long-chain n-3-to-n-6 polyunsaturated fatty acid ratios in breast adipose tissue from women with and without breast cancer. Nutr Cancer 2002; 42: 180–185. 38 Mamalakis G, Hatzis C, de Bree E, Sanidas E, Tsiftsis DD, Askoxylakis J et al. Adipose tissue fatty acids in breast cancer patients versus healthy control women from Crete. Ann Nutr Metab 2009; 54: 275–282. 39 London SJ, Sacks FM, Stampfer MJ, Henderson IC, Maclure M, Tomita A et al. Fatty acid composition of the subcutaneous adipose tissue and risk of proliferative benign breast disease and breast cancer. J Natl Cancer Inst 1993; 85: 785–793. 40 Simonsen N, van't Veer P, Strain JJ, Martin-Moreno JM, Huttunen JK, Navajas JF et al. Adipose tissue omega-3 and omega-6 fatty acid content and breast cancer in the EURAMIC study. European Community Multicenter Study on Antioxidants, Myocardial Infarction, and Breast Cancer. Am J Epidemiol 1998; 147: 342–352.
© 2014 Macmillan Publishers Limited
