Higher Plasma Phospholipid n–3 PUFAs, but Lower n–6 PUFAs, Are Associated with Lower Pulse Wave Velocity among Older Adults^,,

Abstract

Background: Higher intake of polyunsaturated fatty acids (PUFAs) and higher circulating PUFAs are associated with lower cardiovascular disease (CVD) risk. The positive influence of PUFAs might be via lowering arterial stiffness, resulting in a better CVD risk profile; however, studies investigating circulating PUFAs in relation to arterial stiffness in a general population are limited.

Objective: We investigated the associations of plasma phospholipid n–3 (ω-3) and n–6 PUFAs and fish oil intake with arterial stiffness.

Methods: We used data from a subgroup of the Age, Gene/Environment Susceptibility–Reykjavik (AGES-Reykjavik) Study (n = 501, 75.0 ± 4.96 y, 46% men), a population-based study of community-dwelling older adults. Plasma phospholipid PUFAs were measured by GC at baseline, and fish oil intake was assessed at 3 time points: early life (ages 14–19 y), midlife (ages 40–50 y), and late life (ages 66–96 y, AGES-Reykjavik baseline) with the use of a validated food-frequency questionnaire. Arterial stiffness was determined as carotid–femoral pulse wave velocity (cf-PWV) with the use of an electrocardiogram after a mean follow-up of 5.2 ± 0.3 y. Regression coefficients (95% CIs), adjusted for demographics, follow-up time, risk factors, cholesterol, triglycerides, and serum vitamin D, were calculated by linear regression per SD increment in PUFAs.

Results: Plasma total n–3 PUFAs, eicosapentaenoic acid, and docosahexaenoic acid were associated with lower cf-PWV [β (95% CI): −0.036 (−0.064, −0.008); −0.031 (−0.059, −0.003); −0.036 (−0.064, −0.009), respectively]. In contrast, plasma total n–6 PUFAs and linoleic acid were associated with higher cf-PWV [0.035 (0.009, 0.061) and 0.034 (0.008, 0.059)]. Regular fish oil consumption at early-, mid-, and late-life was not associated with cf-PWV.

Conclusions: Our results show a positive association between plasma n–6 PUFAs and arterial stiffness, and suggest that higher concentrations of plasma long-chain n–3 PUFAs are associated with less arterial stiffness and therein may be one of the mechanisms underlying the association between plasma n–3 PUFAs and lower CVD risk.

Introduction

Higher dietary and higher circulating n–3 PUFAs are largely associated with better cardiovascular health and lower risk of fatal ischemic heart disease (1, 2). However, there is also evidence that n–3 PUFAs may have limited effects on cardiovascular health (3). Higher plasma n–6 PUFAs have been associated with a decreased risk of coronary heart disease (4) and cardiovascular disease (CVD)¹² mortality (5), but the opposite relation also has been reported (6). The uncertainty regarding the role of PUFAs on CVD risk is important to clarify. Exploration of potential mediators of the relation may provide insight.

Arterial stiffness is a risk factor for CVD with a pathogenesis that differs from atherosclerosis. Compared with atherosclerosis, much less is known regarding relations between pulse wave velocity (PWV) and PUFAs, and the role of PUFAs may be specific to each pathologic process. Arterial stiffness, i.e., decreased arterial wall elasticity, leads to a loss of the cushioning function of the arterial tree. This results in increased systolic pressure and increased transmission of potentially harmful pulsatile energy into the periphery, where it may damage microcirculation. Thereby, arterial stiffness may contribute to risk of stroke, coronary heart disease, heart failure, cognitive impairment, kidney disease, and all-cause mortality (7–9). Arterial stiffness can be determined by carotid–femoral pulse wave velocity (cf-PWV) which is considered to be the noninvasive reference standard (8, 10). Lower cf-PWV indicates less stiffness of the arteries.

Previous research has shown inverse associations between n–3 PUFAs and arterial stiffness (11–17). Long chain n–3 PUFAs may reduce arterial stiffness by antithrombotic, hypotensive, anti-inflammatory, endothelial NO–stimulating, and atherosclerotic plaque growth–inhibiting mechanisms (1, 2, 18–20). However, little is known about n–6 PUFAs and arterial function. To date, one cross-sectional study in Japanese healthy middle-aged men has been published. It found a positive association between higher concentrations of serum n–6 PUFAs and markers indicative of greater arterial stiffness, such as higher brachial–ankle PWV and elevated serum C-reactive protein concentrations (21).

Although previous studies provide promising evidence for n–3 PUFAs in the reduction of CVD risk, they were limited by lack of generalizability through use of specific population samples (healthy middle-aged men), use of self-reported dietary PUFA intake, lack of a distinction between individual PUFAs, or cross-sectional measurement of arterial function. In addition, although limited to one study, the potential harmful associations of high n–6 PUFAs and arterial function measures are important to clarify. Understanding associations between individual circulating PUFAs and subsequent measurements of vascular stiffness might provide insight into the mechanisms underlying CVD development.

Therefore, we investigated whether baseline concentrations of plasma phospholipid long chain n–3 PUFAs [EPA (20:5n–3), docosapentaenoic acid (DPA) (22:5n–3), and DHA (22:6n–3)], the intermediate chain n–3 PUFA α-linolenic acid (ALA) (18:3n–3), and plasma phospholipid long chain n–6 PUFAs [linoleic acid (LA) (18:2n–6) and arachidonic acid (AA) (20:4n–6)] were associated with arterial stiffness as measured by cf-PWV values obtained at a 5 y follow-up examination. Because FAs were measured at a single time point, we also inquired about dietary intake of fish oil in early life, midlife, and recently in old age. We hypothesized that higher concentrations of plasma phospholipid n–3 and n–6 PUFAs and daily intake of fish oil each were associated with lower values of cf-PWV.

Methods

Study population.

We used data from the Age, Gene/Environment Susceptibility–Reykjavik (AGES-Reykjavik) Study, an ongoing single-center, prospective population study of survivors from the Reykjavik Study. Details of the study design were previously published (22). A random sample of 5764 men and women participated in study’s baseline examination, which occurred between 2002 and 2006. During a mean follow-up of 5.2 ± 0.3 y 1039 participants died, 1198 were not willing to participate, and 211 were lost to follow-up. Follow-up measurements took place between 2007 and 2011 and included 3316 participants.

Baseline plasma phospholipid PUFAs were measured in 2 substudies of AGES-Reykjavik; ICELAND-MI (23, 24) and a case cohort study of fracture (25). A total of 1012 participants had PUFA data. Of these, 501 participants had data on cf-PWV at the follow-up examination and complete data on baseline covariates. The study design is depicted in Figure 1. At baseline, participants excluded were older and less physically active, had a larger waist circumference and lower HDL cholesterol concentrations, and were more likely to be hypertensive than those included in the current analyses (data not shown).

FIGURE 1.

Study design of the AGES-Reykjavik Study, including participants with complete data on plasma phospholipid PUFAs and cf-PWV. AGES-Reykjavik, Age, Gene/Environment Susceptibility–Reykjavik; cf-PWV, carotid–femoral pulse wave velocity.

All participants provided written informed consent. The study was approved by the National Bioethics Committee in Iceland (VSN 00–063), as well as the Institutional Review Board of the Intramural Research Program at the National Institute on Aging.

PUFAs.

Baseline blood samples were collected after an overnight fast and stored at –80°C. PUFAs were measured in plasma phospholipids, which reflect short-term dietary intake and FAs available to the periphery. PUFA determination was performed at the Biomarker Lab, Fred Hutchinson Cancer Research Center, Seattle, WA. In brief, phospholipids were separated from other lipids by one-dimensional TLC (26). FA methyl esters were prepared by direct transesterification (27) and separated with the use of GC. PUFAs are expressed as a relative percentage of the total phospholipid FAs analyzed. For this study, we focused on total and individual long-chain n–3 PUFAs (EPA + DPA + DHA), ALA, and total and individual long-chain n–6 PUFAs (LA + AA). The CV from pooled quality-control samples for EPA, DPA, DHA, ALA, LA, and AA were all <2.5%.

Fish oil intake.

Dietary consumption was assessed by validated FFQ in early life (ages 14–19), midlife (ages 40–50), and later life (ages 66–96, AGES-Reykjavik baseline) (28). The FFQ assessed frequency of intake of 10 common foods and food groups, including fish and fish oil, with the use of the same questions for all 3 time periods (29). The most commonly consumed fish in Iceland are cod and haddock (30), both of which contain low concentrations of n–3 PUFAs. Therefore, we focused on fish liver oil intake by supplements in liquid or capsules (referred to as fish oil hereafter) which is rich in n–3 PUFAs, as well as vitamin D, and has been common in the Icelandic diet for many decades. Fish oil intake was categorized as never, occasionally (<1 time/wk, 1–2 times/wk, or 3–4 times/wk), or regularly (5–6 times/wk or daily).

cf-PWV.

Arterial stiffness, determined after a mean of 5.2 ± 0.3 y of follow-up, was noninvasively assessed by cf-PWV, as previously described (31). Briefly, after 15–20 min of supine posture, brachial blood pressure and heart rate were assessed and arterial tonometry with electrocardiogram was obtained from the carotid and femoral arteries with the use of a custom transducer (Cardiovascular Engineering). Transit distance of the pulse wave from the carotid to the femoral artery was assessed by body surface measurements from the suprasternal notch to the carotid and femoral pulse recording sites. cf-PWV was computed as the pulse wave transit distance divided by the transit time of the pulse wave from the carotid to the femoral artery, with adjustment for parallel transmission of the arterial pulse wave in the brachiocephalic artery and aortic arch. During tonometry, mean arterial pressure and heart rate were determined as well. Higher values of cf-PWV represent increased stiffening of the arteries (8).

Covariates.

All covariates were assessed at baseline, except for mean arterial pressure and heart rate, which were assessed during the cf-PWV measurement. BMI (kilograms per meter squared) was calculated from measured weight and height, and waist circumference (centimeters) was measured with the use of standardized protocols (22). Education (less than high school, high school, and postsecondary), smoking status (never, former, and current), and physical activity (frequency of moderate-to-vigorous activity) were assessed by questionnaire. Fasting serum concentrations of total cholesterol, HDL cholesterol, and TGs were determined on a chemistry analyzer with the use of comparable enzymatic procedures (Hitachi 912; Roche Diagnostics) (32). Serum LDL cholesterol was calculated with the use of the Friedewald equation. Serum 25-hydroxyvitamin D [25(OH)D]was measured with the use of Liaison chemiluminescence immunoassay (DiaSorin). Use of antihypertensive medications was ascertained from medication vials brought to the research center. Blood pressure was assessed from the mean value of 2 measurements taken in a supine position with a mercury sphygmomanometer. Hypertension was defined as measured systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg, or self-reported doctor’s diagnosis of hypertension, or using antihypertensive medications. Mean arterial pressure (millimeters of mercury) and heart rate (beats per minute) were determined during the tonometry test. Diabetes mellitus was defined as a self-reported doctor’s diagnosis of diabetes, use of blood glucose–lowering drugs, or fasting blood glucose concentration ≥7.0 mmol/L. Other medical conditions (e.g., congestive heart failure and coronary heart disease) were determined from self-report, medications, and clinical assessments.

Statistical analysis.

Baseline characteristics were compared for men and women separately with 2-sided t tests for continuous variables or a chi-square test for categorical variables. There was no interaction between sex and PUFAs in relation to cf-PWV. Therefore, analyses were performed for men and women combined. Concentrations of baseline plasma phospholipid PUFAs according to fish oil intake at early life, midlife, and late life were presented as means ± SDs, and trends across fish oil intake groups were tested with the use of linear regression analyses. The distribution of cf-PWV, measured after 5.2 ± 0.3 y after the baseline visit, was skewed to the right and was therefore log-transformed to meet normality assumptions. Multivariable linear regression analyses were performed to assess associations between plasma phospholipid PUFAs in relation to cf-PWV. Unstandardized regression coefficients (β) and corresponding 95% CIs were expressed per SDs of PUFAs. The regression coefficient can be interpreted as percentage difference in cf-PWV per SD of PUFAs. Model 1 was adjusted for age, sex, mean arterial pressure, heart rate, and follow-up time. Model 2 included Model 1 plus established CVD risk factors, including waist circumference, education, smoking status, physical activity, use of antihypertensive medication, diabetes, and coronary heart disease. Model 3 further adjusted for HDL cholesterol, LDL cholesterol, TGs, and 25(OH)D. Serum 25(OH)D has been included, because it is inversely associated with CVD events and CVD risk factors (33, 34), and daily intake of fish oil, rich in vitamin D and a main source of n–3 PUFAs, was high (62%) in this population. Multivariable linear regression analyses for fish oil were performed, adjusting for the same covariates mentioned in the models described above. P- trend values across intake categories were calculated, with “never” as the reference group. At the time of the cf-PWV measurement, mean arterial pressure and heart rate were also determined. Mean arterial pressure was included as a covariate to determine vascular properties, independent of distending pressure, which can alter aortic stiffness. Adjustment for heart rate was performed because heart rate influences the intraindividual variation of cf-PWV, meaning that when heart increases, PWV also increases (35).

A cf-PWV measurement was also obtained from a subsample of individuals at baseline (n = 940). We performed sensitivity analyses among those who had data on PUFAs and cf-PWV measured at both study baseline and follow-up (n = 86). Linear regression analyses with additional adjustment for baseline cf-PWV were performed to determine associations between PUFAs and cf-PWV at follow-up.

All P values were 2-tailed (α = 0.05) and data were analyzed with STATA version 12.1.

Results

Baseline data were available for 501 participants with a mean age of 75.0 ± 4.96 y, of whom 46% were men. Compared with women, men had a larger waist circumference, had completed more years of education, were more physically active and more likely to be former smokers, and had poorer lipoprotein profiles, higher serum 25(OH)D concentrations, lower heart rate, and more comorbidities. In addition, men had higher concentrations of plasma n–3 PUFAs than women, and higher values of cf-PWV (P < 0.05) (Table 1).

TABLE 1.

Characteristics of men and women with data on baseline plasma phospholipid PUFAs and cf-PWV at 5 y follow-up in a subset of the Age, Gene/Environment Susceptibility–Reykjavik Study¹

	Men	Women	P
Participants, n	228	273
Age, y	74.9 ± 4.60	75.1 ± 5.24	0.58
BMI, kg/m2	26.6 ± 3.22	27.0 ± 3.96	0.23
Waist circumference, cm	101.4 ± 9.03	97.0 ± 11.2	<0.001
Moderate to vigorous physical activity, h/wk	1.99 ± 2.65	1.33 ± 2.27	0.003
Education			<0.001
Less than high school	153 (67)	218 (80)
High school	33 (14)	35 (13)
Postsecondary school	42 (18)	20 (7)
Smoking status			<0.001
Never	72 (32)	143 (52)
Former	135 (59)	96 (35)
Current	21 (9)	34 (12)
Serum HDL cholesterol, mmol/L	1.45 ± 0.40	1.75 ± 0.42	<0.001
Serum LDL cholesterol, mmol/L	3.26 ± 0.95	3.66 ± 1.03	<0.001
Serum TGs, mmol/L	1.15 ± 0.59	1.18 ± 0.58	0.55
Serum 25-hydroxyvitamin D, nmol/L	59.3 ± 26.3	53.8 ± 26.8	0.022
Heart rate,2 bpm	59.6 ± 9.9	62.2 ± 9.9	0.004
Mean arterial pressure,2 mm Hg	93.5 ± 12.5	95.3 ± 13.4	0.13
Systolic blood pressure, mm Hg	144 ± 19.6	140 ± 18.6	0.06
Antihypertensive medication	101 (44)	132 (48)	0.37
Hypertension	183 (80)	205 (75)	0.17
Type 2 diabetes mellitus	24 (11)	14 (5)	0.023
Coronary heart disease	63 (28)	34 (12)	<0.001
Current fish intake			0.82
Never	56 (25)	69 (25)
Occasionally (<1 time/wk, 1–2 times/wk, or 3–4 times/wk)	22 (10)	22 (8)
Regularly (5–6 times/wk or daily)	150 (66)	182 (67)
Plasma PUFAs, %
Long-chain n–3 PUFAs	10.9 ± 3.17	9.94 ± 2.94	<0.001
EPA	3.16 ± 1.70	2.72 ± 1.55	0.003
DPA	1.20 ± 0.22	1.11 ± 0.19	<0.001
DHA	6.58 ± 1.53	6.11 ± 1.46	<0.001
α-Linolenic acid	0.22 ± 0.07	0.22 ± 0.07	0.52
Long-chain n–6 PUFAs	24.5 ± 3.26	25.0 ± 2.78	0.10
Linoleic acid	17.6 ± 3.09	18.0 ± 2.60	0.12
Arachidonic acid	6.96 ± 1.49	7.01 ± 1.76	0.73
Carotid–femoral pulse wave velocity,2 m/s	15.0 ± 5.25	13.4 ± 4.91	<0.001

¹Values are means ± SDs for continuous variables and n (%) for categorical variables unless otherwise indicated. Plasma PUFAs are expressed as a relative percentage of the total plasma phospholipid FAs analyzed. DPA, docosapentaenoic acid.

²Assessed at 5 y follow-up; all other variables reflect baseline measurements.

PUFAs in relation to cf-PWV.

Associations between plasma phospholipid PUFAs and cf-PWV are shown in Table 2. Plasma total long-chain n–3 PUFAs were associated with lower cf-PWV after adjustments for age, sex, mean arterial pressure, heart rate, and follow-up time (model 1) (P = 0.007); further adjustments for waist circumference, education, smoking status, physical activity, antihypertensive medication, diabetes, and coronary heart disease (model 2) (P = 0.003); and more adjustments for HDL cholesterol, LDL cholesterol, TGs, and 25(OH)D (model 3) (P = 0.012). This association reflects the inverse association of plasma EPA and plasma DHA with cf-PWV. In contrast, each SD increment in plasma total long-chain n–6 PUFAs and plasma LA were associated with higher cf-PWV in all models (P = 0.009 for both) (model 3). No other associations between plasma PUFAs and cf-PWV were observed (P > 0.05).

TABLE 2.

Associations between baseline plasma phospholipid PUFAs (per SD increment) and cf-PWV in 501 participants of the Age, Gene/Environment Susceptibility–Reykjavik Study¹

	cf-PWV
	Model 1	Model 2	Model 3
Long-chain n–3 PUFAs	−0.035 (−0.061, −0.010)	−0.039 (−0.064, −0.013)	−0.036 (−0.064, −0.008)
P	0.007	0.003	0.012
EPA	−0.036 (−0.061, −0.010)	−0.036 (−0.062, −0.011)	−0.031 (−0.059, −0.003)
P	0.006	0.005	0.029
DPA	−0.011 (−0.037, 0.015)	−0.016 (−0.042, 0.009)	−0.012 (−0.039, 0.014)
P	0.40	0.21	0.37
DHA	−0.031 (−0.057, −0.006)	−0.037 (−0.063, −0.011)	−0.036 (−0.064, −0.009)
P	0.017	0.005	0.010
α-Linolenic acid	0.009 (−0.017, 0.034)	0.019 (−0.006, 0.044)	0.020 (−0.006, 0.045)
P	0.50	0.15	0.12
Long-chain n–6 PUFAs	0.036 (0.011, 0.062)	0.036 (0.011, 0.061)	0.035 (0.009, 0.061)
P	0.006	0.005	0.009
Linoleic acid	0.023 (−0.003, 0.048)	0.034 (0.009, 0.059)	0.034 (0.008, 0.059)
P	0.08	0.009	0.009
Arachidonic acid	0.026 (0.001, 0.052)	0.009 (−0.017, 0.035)	0.003 (−0.026, 0.031)
P	0.042	0.50	0.86

¹Values are unstandardized regression coefficients (β) and 95% CIs. Model 1 was adjusted for age, sex, mean arterial pressure, heart rate, and follow-up time. Model 2 was adjusted for model 1 plus waist circumference, education, smoking status, physical activity, use of antihypertensive medication, diabetes mellitus, and heart disease. Model 3 was adjusted for model 2 plus HDL cholesterol, LDL cholesterol, TGs, and serum vitamin D. cf-PWV, carotid femoral pulse wave velocity; DPA, docosapentaenoic acid.

Fish oil intake in relation to cf-PWV.

PUFA concentrations according to fish oil intake groups are presented in Supplemental Table 1. Associations between fish oil intake at early, mid-, and late life and cf-PWV are shown in Table 3. There were no significant associations observed for early life intake in relation to cf-PWV in all models (P-trend = 0.85) (model 3). Occasional and regular intake of fish oil in midlife was significantly associated with lower cf-PWV in model 1 (P-trend = 0.038). However, this association was attenuated after adjustments for covariates and became nonsignificant (P-trend = 0.20) (model 3). In late life, occasional intake was associated with lower cf-PWV in all models; however, the P-trend across intake categories was nonsignificant (P-trend = 0.96).

TABLE 3.

Associations between fish oil intake across lifetime and cf-PWV in 501 participants of the Age, Gene/Environment Susceptibility–Reykjavik Study¹

	n	cf-PWV, m/s	Model 1	Model 2	Model 3
Early life (ages 14–19 y)
Never	190	14.1 ± 4.91	Reference	Reference	Reference
Occasionally	102	14.0 ± 5.10	−0.022 (−0.092, 0.048)	−0.023 (−0.091, 0.046)	−0.025 (−0.094, 0.043)
Regularly	209	14.3 ± 5.34	−0.013 (−0.070, 0.044)	−0.013 (−0.070, 0.043)	−0.006 (−0.062, 0.051)
P-trend		0.66	0.66	0.65	0.85
Midlife (ages 40–50 y)
Never	126	14.9 ± 5.61	Reference	Reference	Reference
Occasionally	111	13.8 ± 5.20	−0.086 (−0.160, −0.012)	−0.073 (−0.146, 0.000)	−0.072 (−0.145, 0.001)
Regularly	264	13.9 ± 4.82	−0.073 (−0.134, −0.012)	−0.056 (−0.117, 0.006)	−0.049 (−0.112, 0.014)
P-trend		0.11	0.038	0.12	0.20
Late life (ages 66–96 y)
Never	125	14.6 ± 5.35	Reference	Reference	Reference
Occasionally	44	12.8 ± 4.43	−0.127 (−0.226, −0.028)	−0.110 (−0.207, −0.013)	−0.108 (−0.205, −0.011)
Regularly	332	14.1 ± 5.10	−0.035 (−0.094, 0.025)	−0.027 (−0.086, 0.033)	−0.011 (−0.074, 0.052)
P-trend		0.53	0.43	0.56	0.96

¹Values are means ± SDs or unstandardized regression coefficients (β) and 95% CIs. Fish oil intake was defined as never, occasionally (<1 time/wk, 1–2 times/wk, or 3–4 times/wk), or regularly (5–6 times/wk or daily). Model 1 was adjusted for age, sex, mean arterial pressure, heart rate, and follow-up time. Model 2 was adjusted for model 1 plus waist circumference, education, smoking status, physical activity, use of antihypertensive medication, diabetes mellitus, and heart disease. Model 3 was adjusted for model 2 plus HDL cholesterol, LDL cholesterol, TGs, and serum vitamin D. cf-PWV; carotid femoral pulse wave velocity.

Sensitivity analyses.

We performed sensitivity analyses to investigate whether associations between plasma PUFAs and cf-PWV were driven by baseline values of cf-PWV. At AGES-Reykjavik baseline, cf-PWV was determined in 940 participants. Of those, 143 had plasma PUFA measurements, and after excluding participants with missing data on cf-PWV at follow-up (n = 55) or covariates (n = 2), 86 participants were included for sensitivity analyses. cf-PWV at follow-up was 9.28 ± 23.8% higher than baseline cf-PWV values. After additional adjustment for baseline cf-PWV, linear regression showed significant associations between long-chain n–3 PUFAs, EPA, DPA, long-chain n–6 PUFAs, and LA and cf-PWV (Supplemental Table 2). In the fully adjusted model, each SD increment in plasma total long-chain n–3 PUFAs (P = 0.049), EPA (P = 0.036), and DPA (P = 0.047) was associated with lower cf-PWV. Also in line with our main analyses, higher cf-PWV values were observed for each SD increment in plasma total long-chain n–6 PUFAs (P = 0.019) and LA (P = 0.023).

Discussion

In this current study, we investigated associations between baseline plasma phospholipid n–3 and n–6 PUFAs with cf-PWV measured 5 y later. This work expands upon previous cross-sectional studies to provide insight into whether plasma long-chain PUFAs are associated prospectively with arterial stiffness as measured by cf-PWV. Higher concentrations of plasma total long-chain n–3 PUFAs, EPA, and DHA were associated with lower cf-PWV. In contrast, higher plasma total long-chain n–6 PUFAs and LA were associated with higher cf-PWV. We also investigated whether fish oil intake (measured at 3 time points across the lifetime) was associated with PWV, and we showed that, in our study populations, regular fish oil intake was not associated with cf-PWV.

Our results align with previous studies of associations between PUFA concentrations and arterial stiffness. Cross-sectional results from the Framingham Offspring Study and Omni cohort showed that higher red blood cell n–3 PUFA (EPA + DHA) content was associated with lower cf-PWV (16). In a small cross-sectional study in multiethnic middle-aged adults in Europe, Anderson et al. (12) showed that plasma FA profiles characterized by higher proportions of EPA, DHA, and AA and lower proportions of oleic, palmitic, and LA concentrations were associated with lower PWV. These observations are comparable with our results. Our results showing an inverse association between plasma EPA and cf-PWV are also in line with results from another cross-sectional study performed in Korean men (13). Our results for plasma phospholipid n–3 PUFAs are confirmed by a recent randomized, controlled trial showing that n–3 supplementation (4g/d for 12 wk) resulted in lower cf-PWV in healthy adults aged 66 ± 2 y (17). The effect of n–3 PUFA supplementation on cf-PWV might be explained by improved endothelial function (36) and a decrease in inflammatory markers (1).

Previous studies have shown that LA, the most abundant n–6 PUFA in the diet, reduces CVD risk by reducing total and low-density lipoprotein (37, 38). In addition, a meta-analysis of prospective studies showed that higher plasma LA was associated with lower coronary heart disease risk (39), and a meta-analysis by Harris et al. (40) showed an inverse relation between LA and nonfatal CVD endpoints. Knowing of these beneficial associations between n–6 PUFAs and CVD-related outcomes (4, 41, 42), we hypothesized that higher plasma n–6 PUFAs would be associated with lower cf-PWV. However, in this study, we found that higher concentrations of plasma long-chain n–6 PUFAs and LA were associated with higher cf-PWV. A similar result was seen in a cross-sectional Japanese study in healthy middle-aged men investigating serum PUFAs in relation to measures of arterial stiffness (21). In yet another cross-sectional Japanese study, higher concentrations of n–6 PUFAs (LA + AA) were weakly associated with the presence of carotid plaques—an intermediate for atherosclerosis and decreased arterial compliance—although neither n–3 nor n–6 PUFAs were correlated with PWV in that study (43). Our results thereby add to the existing uncertainty about the role of individual plasma n–6 PUFAs in CVD risk factors and outcomes. Not all studies show a positive influence from LA on CVD risk factors or incidence (44). AA, the other main n–6 PUFA, might be associated with proinflammatory function by raising eicosanoids (45), although higher AA concentrations were not associated with coronary heart disease risk (39, 40). In addition, a trial investigating the effects of a fish-based diet on plasma EPA and AA concentrations and, in turn, brachial–ankle PWV showed that a higher plasma EPA-to-AA ratio was accompanied by a significant reduction in brachial–ankle PWV in high-risk CVD outpatients compared with low-risk CVD outpatients (46). However, caution should be taken in interpreting our results to be evidence of n–6 PUFA harm. Further longitudinal studies are warranted to investigate and clarify the associations and possible underlying biological mechanisms through which individual n–6 PUFAs, alone and in combination, influence risk of CVD.

PUFA concentrations are dependent on both dietary intake and endogenous synthesis. As expected, participants who reported regular fish oil intake had higher total and individual n–3 PUFA concentrations throughout life, but lower concentrations of n–6 PUFAs. We observed inverse associations between fish oil intake in midlife in relation to lower cf-PWV in a minimally adjusted model only. We can partly draw parallels with one large community-based cohort study among men performed by Livingstone et al. (15). This study observed an inverse association between total PUFA intake and aortic PWV after 17.8 y of follow-up. However, our finding, based on fish oil intake in midlife relative to lower cf-PWV in old age, did not withstand adjustments for risk factors and 25(OH)D. In addition, Livingstone et al. (15) analyzed total PUFAs (n–3 + n–6), but did not analyze them separately. This is of interest, because we observed opposing associations between n–3 and n–6 PUFAs and cf-PWV. To facilitate comparison, we analyzed plasma total PUFA concentrations and found no association between cf-PWV and every SD increment in plasma total PUFAs [−0.006 (−0.022, −0.033) (P = 0.68)]. A possible explanation for the discrepancy in the results is that Livingstone et al. determined intake of PUFAs based on FFQ instead of plasma phospholipid concentrations, which is more prone to misreporting. Another explanation might be the difference in age. The population of the study by Livingstone et al. was younger at the time of PWV measurements (73.8 vs. 80.2 y).

Although we did not observe associations between fish intake in later life and lower cf-PWV values, a healthy diet containing high fruit, vegetable, and fish intake and low meat and saturated fat intake are associated with an improved CVD risk profile, with lower arterial stiffness; this is based on results of several randomized controlled trials (47–49).

This study has several strengths. First, we measured cf-PWV, which is considered to be the gold-standard measurement of arterial stiffness (8, 10). Second, determination of plasma phospholipid n–3 and n–6 PUFAs has the advantage of identifying and quantifying individual PUFAs and therefore allows for determination of their individual contribution to cf-PWV. It also eliminates the potential biases associated with self-reported dietary information, whereby it is likely that participants underreport their energy intake (50, 51), which might lead to an underestimation of the true association. Furthermore, the detailed and standardized collection of demographic, lifestyle, and other covariates allowed adjustment for several potential confounders. However, our study was not without limitations. The cross-sectional design limits inferring causal relations from these results. In addition, plasma phospholipid PUFAs were determined at baseline only. Because of the single assessment, we were not able to determine changes in PUFA concentrations over time, and thus investigate influences of changes in plasma phospholipid PUFA concentration in relation to cf-PWV. Furthermore, we looked at a single nutrient, whereas dietary patterns could show the complexity of metabolism and interaction between food components (52). Determination of cf-PWV at both baseline and follow-up was only performed in a subsample of individuals at AGES-Reykjavik baseline, resulting in a relatively small number of participants with complete data on PUFAs and cf-PWV. However, our sensitivity analyses in a sample with cf-PWV at both time points showed comparable results to our main analyses. Another limitation is that among these aged participants, those who participated in the 5 y follow-up examination were younger, had a smaller waist circumference at baseline, and were less likely to have comorbidities and therefore may not be representative of younger individuals. To investigate potential effects of selective drop-out, we performed multiple imputation with the use of the chained equation function with predictive mean-matching in Stata. We imputed cf-PWV values for the participants with baseline plasma phospholipid PUFA data. The results of the pooled analyses over 50 imputed datasets indicated that every SD increment increase in plasma EPA remained associated with lower cf-PWV (P = 0.031). The associations between plasma total long-chain n–3 PUFAs, DHA, total long-chain n–6 PUFAs, and LA and cf-PWV lost their significance, but directions remained. It is possible that selective drop-out of participants because of poorer baseline health may have attenuated the observed associations between plasma PUFAs and cf-PWV; therefore, it is likely that survival bias resulted in an underestimation of the observed association. The AGES-Reykjavik Study consists of older Icelandic men and women (mean age 75 y) in whom deterioration of the arteries might already have occurred. In addition, the external validity may be limited to younger populations in whom pathophysiology is likely to differ compared with older adults. A total of 62% of the AGES-Reykjavik Study population consumed fish oil daily, which may also limit external validity. Furthermore, FA metabolism is different for individuals with different genetic backgrounds (53), and, therefore, results from this study may not be generalizable to other populations. In addition, differences in, e.g., heterogeneity between studies, such as in participant characteristics (race, age, and disease status), study design, or arterial stiffness, determination make it difficult to compare study results.

In conclusion, this study showed that higher plasma concentrations of total long chain n–3 PUFAs, specifically EPA and DHA, were associated with lower cf-PWV and higher concentrations of total long-chain n–6 PUFAs, specifically LA, were associated with higher cf-PWV. Future larger prospective studies are warranted to clarify the conflicting associations between plasma n–3 and n–6 PUFAs and cf-PWV and to provide insight into mechanisms by which PUFAs influence vascular health.