Habitual Diets Rich in Dark-Green Vegetables Are Associated with an Increased Response to ω-3 Fatty Acid Supplementation in Americans of African Ancestry^,,

Abstract

Although substantial variation exists in individual responses to omega-3 (ω-3) (n–3) fatty acid supplementation, the causes for differences in response are largely unknown. Here we investigated the associations between the efficacy of ω-3 fatty acid supplementation and a broad range of nutritional and clinical factors collected during a double-blind, placebo-controlled trial in participants of African ancestry, randomly assigned to receive either 2 g eicosapentaenoic acid (EPA) + 1 g docosahexaenoic acid (n = 41) or corn/soybean oil placebo (n = 42) supplements for 6 wk. Food-frequency questionnaires were administered, and changes in erythrocyte lipids, lipoproteins, and monocyte 5-lipoxygenase–dependent metabolism were measured before and after supplementation. Mixed-mode linear regression modeling identified high (n = 28) and low (n = 13) ω-3 fatty acid response groups on the basis of changes in erythrocyte EPA abundance (P < 0.001). Compliance was equivalent (∼88%), whereas decreases in plasma triglycerides and VLDL particle sizes and reductions in stimulated monocyte leukotriene B₄ production were larger in the high-response group. Although total diet quality scores were similar, the low-response group showed lower estimated 2005 Healthy Eating Index subscores for dark-green and orange vegetables and legumes (P = 0.01) and a lower intake of vegetables (P = 0.02), particularly dark-green vegetables (P = 0.002). Because the findings reported here are associative in nature, prospective studies are needed to determine if dietary dark-green vegetables or nutrients contained in these foods can enhance the efficacy of ω-3 fatty acid supplements. This trial was registered at clinicaltrials.gov as NCT00536185.

Introduction

Cardiovascular disease (CVD)¹⁰ is a leading cause of mortality in the United States, where it is attributed to 1 in 3 deaths (1). Advances in science and technology have contributed to our understanding of the pathology of CVD and its associated risk factors, and have provided novel treatment strategies. Diet is a modifiable factor in CVD risk reduction (2). Accumulating evidence suggests that consuming the long-chain ω-3 FAs EPA (20:5ω-3) and DHA (22:6ω-3) can decrease CVD risk through a combination of mechanisms, including lowering of plasma TGs, decreasing heart rate and blood pressure, improving endothelial and autonomic function, and reducing inflammation (3–7). However, there is substantial heterogeneity in responses to ω-3 FA supplementation. For example, high variance has been reported in ω-3 FA supplementation–associated increases in RBC EPA+DHA (i.e., the Omega-3 Index) (8) and reductions in plasma TGs (9–11). Baseline ω-3 FA status, as well as innate factors such as gender, age, and genetics, may contribute to such variability in the benefits of dietary interventions. For instance, the TG-lowering effects of ω-3 FAs are more pronounced in individuals with higher circulating TGs at baseline (11). Therefore, external factors such as diet and other lifestyle variables may also play a significant role in defining the heterogeneity of responses to ω-3 FA supplementation observed in populations.

Black Americans appear to be at increased risk of CVD (12) and have a higher CVD prevalence than whites (1). A CVD risk–associated 5-lipoxygenase gene (Alox5) promoter region polymorphism in the specificity protein 1 transcription factor is abundant in populations of African ancestry (13, 14). In a recent study exploring ω-3 FA interactions with this polymorphism, we found that eicosanoid production in response to supplementation differed by genotype (15). Interestingly, we also observed substantial variability in various responses to the dietary ω-3 FA enrichment, including RBC membrane FA changes and TG reductions not explained by the Alox5 promoter polymorphism. Despite the fact that high variation has been reported previously, the etiology and individual characteristics associated with responsive phenotypes are largely unknown. In the present study, we performed a secondary analysis to examine the variation in response to ω-3 FA supplementation in our cohort of black Americans randomly assigned by Alox5 genotype and describe the phenotypic association with reported habitual diets in high and low responders.

Participants and Methods

This study constitutes a secondary analysis of a randomized, double-blind, placebo-controlled intervention trial originally designed to examine the effect of ω-3 FA supplementation on CVD risk factors in participants with different Alox5 gene variants (15). The project was approved by the institutional review boards of The University of California, Davis, and Alta Bates Summit Medical Center. Written informed consent was obtained from all study participants. The study is detailed at clinicaltrials.gov under NCT00536185.

Participants.

Healthy adults aged 20–59 y self-identified as African American, black, or of African ancestry, with low habitual ω-3 FA intake and who were not first-degree relatives of other participants, were recruited from 3 Californian cities (Davis, Sacramento, and Oakland), and a sub-Saharan African Ancestry of 72.7 ± 16.0% was confirmed as previously described (15). Health-related exclusion criteria included physician-diagnosed chronic inflammatory diseases, lipid disorders, regular anti-inflammatory or lipid-lowering medication use, and abnormal complete blood count, standard clinical chemistry, or lipid panels suggesting an undiagnosed disease. Other exclusion criteria included practices that could affect ω-3 FA status and immune function: i.e., antioxidant (e.g., vitamin E, vitamin C, carotenoid) consumption greater than the RDA, current fish-oil supplementation, >2 servings/wk of oily fish or shellfish, >1 serving/wk of flaxseed or flaxseed oil, >2 servings/wk of walnuts or walnut oil, >1 α-linolenic acid (18:3ω-3)–enriched egg/d, >14 cigarettes/wk, or >14 alcoholic drinks/wk. Individuals who reported a fish allergy, had a BMI <18.5 or >35 kg/m², were pregnant, or were a first-degree relative of another study participant were also excluded. A total of 79 women and 37 men were recruited with an age of 35.4 ± 11.5 y and a BMI of 27.6 ± 4.6 kg/m². Of the 116 participants enrolled, 83 completed the intervention, returned supplement containers, and provided useful FFQs as described below.

Intervention.

Briefly, participants were randomly assigned to ω-3 FA (5.0 g fish-oil concentrate containing 3.0 g ω-3 FAs: 2.0 g EPA and 1.0 g DHA) or placebo (5.0 g corn/soybean) groups and supplemented for 6 wk with 5 capsules/d. To randomly distribute the 6 Alox5 genotypes investigated in the primary study, 6 randomization lists and a randomized 2-block design were used. 40/20 Ethyl Ester ω-3 FA and Corn/Soybean Placebo capsules (1.0 g/capsule) were provided in bulk by Ocean Nutrition Canada and were indistinguishable by either color or size. FA analysis was performed quarterly on stored capsules to monitor the stability of ω-3 FA amounts. No significant differences were recorded.

Assessment of compliance.

Compliance was monitored with biweekly phone calls and daily diaries, which were assessed in concert with returned pill counts. Each participant was provided with 252 capsules (50.4 d equivalent) and scheduled to return to the clinic after 42 d. Time between visits was 44.3 ± 4.6 (37–58) d. Participants failing to return supplement containers were excluded from this analysis. Compliance was calculated as follows: (252 – returned pills)/d between dispensing and return. Of the 83 participants retained, 4 (1 ω-3 FA– and 3 placebo-supplemented) returned empty containers, and 6 (1 ω-3 FA– and 5 placebo-supplemented) had calculated overcompliance equivalent to >1 pill/wk over the course of the intervention. Participants consumed 87.6 ± 17.0% of provided pills, and compliance was equivalent between ω-3 FA and placebo groups (Table 1).

TABLE 1.

Baseline physical characteristics and RBC FA composition and selected changes in Americans of African ancestry who received placebo or ω-3 FA supplements for 6 wk, by RBC response to supplementation¹

Variable	ω-3 High (n = 28)	ω-3 Low (n = 13)	Placebo (n = 42)	P
Genotype (dd:d5:55),2 n	6:13:9	6:5:2	9:21:12	0.43
Gender (male:female), n	9:19	1:12	12:30	0.23
Age, y	37.2 ± 12	38.0 ± 9.6	34.1 ± 12	0.43
Systolic BP, mm Hg	116 ± 13	110 ± 13	112 ± 14	0.42
Diastolic BP, mm Hg	71.1 ± 11	69.7 ± 9.7	71.1 ± 8.9	0.90
Body weight, kg	77.2 ± 14	80.2 ± 11	79.6 ± 13	0.71
Δ Body weight, % initial	0.157 ± 0.37	0.79 ± 0.77	0.0327 ± 0.34	0.56
BMI, kg/m2	27.0 ± 4.3	30.2 ± 4.3	27.7 ± 4.6	0.11
Compliance, %	86.1 ± 18	87.1 ± 18	89.9 ± 19	0.67
RBCs, mol%
16:0Pre	27.9 ± 3.3a	32.1 ± 6.4b	29.3 ± 4.8a,b	0.043
18:1ω-9Pre	12.4 ± 2.2a	14.4 ± 2.4b	13.6 ± 2.4a,b	0.023
20:4ω-6Pre	15.1 ± 2.5a	11.1 ± 5.3b	14.0 ± 4.3a,b	0.038
EPAPre	0.460 ± 0.22a	0.320 ± 0.19b	0.490 ± 0.24a	0.034
DHAPre	3.95 ± 0.12a	2.25 ± 0.15b	3.54 ± 1.6a	0.003
Omega-3 Index, mol%	4.41 ± 1.3a	2.57 ± 1.6b	4.03 ± 1.7a	0.003
Δ RBC EPA, mol%	1.82 ± 0.76a	0.180 ± 0.23b	−0.020 ± 0.26b	<0.001
Δ RBC DHA, mol%	1.43 ± 1.1a	0.520 ± 0.70a,b	−0.120 ± 0.95b	<0.0001
Δ Omega-3 Index, mol%	3.20 ± 1.7a	0.689 ± 0.08b	−0.140 ± 1.1b	<0.001

¹All values are means ± SDs unless otherwise indicated. Gender and genotypes were compared by χ² distribution (α < 0.05). Means without a common letter differ by ANOVA (or Kruskal-Wallis for Δ RBC EPA) using the Holm-Sidak multiple-comparison post hoc test (α < 0.05). BP, blood pressure; Pre, presupplementation; SP1, specificity protein 1 transcription factor; ω-3 High, ω-3 high responders; ω-3 Low, ω-3 low responders.

²Participants had 1 of 6 allelic variants in 5-lipoxygenase gene-promoter SP1 binding site number and were grouped into 3 categories: 55 = wild-type with 2 alleles containing 5 SP1 sites, d5 = single allelic deletions, and dd = double allelic deletions.

Anthropometric measurements, blood pressure, and heart rate.

Weight, height, and blood pressure were determined before and after intervention and BMI was calculated. Blood pressure was measured in participants at rest for 5 min by using an automated instrument.

Dietary analysis.

The Block FFQ version 2005 (Block Dietary Data Systems) was administered at baseline to estimate dietary patterns and the intake of ω-3 FAs (i.e., EPA, DHA, α-linolenic acid, and flaxseed) and fats over the preceding year. The Block FFQ was self-administered after instruction by trained clinical staff, with serving-size pictures and food models provided to help estimate portion sizes. A registered dietitian reviewed the questionnaires, and participants were contacted about missing information, unusual responses, or discrepancies before data analysis. Participants below the 5th percentile and above the 95th percentile of calculated energy intake (kcal) per kg body weight were considered under- and overreporters, respectively, and excluded from dietary analyses. Six participants (3 placebo, 3 ω-3 FA) were identified as under- or over-reporters. FFQ information was converted to daily average consumption of USDA MyPyramid food groups and nutrients for each participant. To assess diet in terms of overall quality, FFQ analyses were further determined by using an adaptation of the Healthy Eating Index (HEI)–2005 (16, 17). The HEI-2005 was designed to transform 24-h dietary recall data into a standardized comparative index of dietary patterns. A cumulative score of 0 to 100 points accounts for 12 dietary categories: fruit, whole fruit, total vegetables, dark-green and orange vegetables and legumes, grains, whole grains, dairy, meat and beans, sodium, calories from oils, calories from SFAs, and calories from solid fats, alcohol, and added sugars. Here, because data for whole fruit was not available, all participants were assigned 50% of the maximum score for this dietary component. Similarly, percentage of calories from solid fats, alcohol, and added sugars was estimated by adding the percentage of calories from saturated fat, the percentage of calories from sweet desserts, and the percentage of calories from alcohols, and thus likely overrepresents this class. All other HEI scores were calculated as described (16, 18).

Sample collection and processing.

Blood (80 mL) was collected on sodium heparin at baseline and at week 6 and processed as described (15). Platelet-depleted plasma was stored at −80°C, buffy coats were removed and used for monocyte isolation, and buffer-washed RBC pellets were washed and stored under nitrogen at −80°C.

Lipid analyses.

Plasma lipid and lipoproteins were measured before and after intervention. RBC FAs were quantified as methyl esters by GC/MS and internal standard methodologies (19). Briefly, FAs were extracted with methanol containing 20% toluene and trans-esterified with methanolic sodium hydroxide in the presence of the analytical surrogate triheptadeca-(17-Z)-eneoylglyceride. FA methyl esters were extracted into hexane, enriched with tricosanoic acid (23:0) as an internal standard, separated by GC on a 30 m × 0.25 mm × 0.25 μm DB-225ms (Agilent Technologies), and detected on an Agilent Technologies 5973N mass spectral detector operated in simultaneous selected ion-monitoring/full-scan mode. Concentrations were calculated against authentic standards, and results are reported as the mol% of the total measured residues. NMR analysis (LipoScience, Inc.) was used to measure mean VLDL particles and particle concentration from frozen plasma samples as previously described (20).

Monocyte cultures and oxylipin analyses.

Monocytes isolated from peripheral blood mononuclear cells were stimulated with the calcium ionophore A23187 (Sigma), and generated oxylipins were quantified by LC with tandem MS (15). Monocytes were isolated by positive selection using anti-CD14 antibody on magnetic beads (Miltenyi Biotec) as previously described (15). Purified monocytes were cultured with 10% heat-inactivated autologous plasma. Monocyte 5-lipoxygenase activity was stimulated with 10 μmol/L of A23187 in 0.1% DMSO and compared with a vehicle control response at 60 min. Oxylipins from culture supernatants were extracted and quantified by LC with tandem MS by using internal standard methodologies (15). Briefly, samples were enriched with antioxidants and deuterated oxylipin surrogates, and analytes were trapped and eluted from 60-mg Oasis HLB solid-phase extraction cartridges (Waters Corp.). Solvents were evaporated and residues were reconstituted in methanol containing the internal standard cyclohexyl-urido-docecanoic acid (Cayman Chemical). Analytes were separated on a 2.1 mm × 150 mm, 1.7 μm Acquity BEH C18 column and quantified by negative-mode electrospray ionization on a Quattro Micro tandem mass spectrometer (Waters).

Response phenotype identification.

Participants randomly assigned to the ω-3 FA treatment were classified into high (ω-3 High) and low (ω-3 Low) responders by analyzing the change in RBC EPA content as a function of the ω-3 FA dose consumed by using either simplified or complex expressions of this behavior (Fig. 1) with the finite mixtures of regression protocol FlexMix in the R statistical computing environment version 2.9.1 (21, 22). FlexMix uses an expectation-maximization algorithm to iteratively assess and find the most probable finite regression solution to a given data set. In the simplest form, the mol% change in RBC EPA was expressed as a function of the reported mean percentage of the daily supplements consumed (Fig. 1A). However, because the change in RBC EPA is reduced as baseline status increases (23), and the mg/d provide was constant, while participant body mass varied, we also analyzed the relation between the log(RBC EPA_Pre) and log(RBC EPA_{Fold Change}) divided by the mean dose consumed by each individual: i.e., mg/(kg ⋅ d) dose = [(mg EPA + mg DHA) (% Compliance/100)]/kg body mass. Regression analyses were performed 10 times while allowing for up to 5 components. Aikaike’s information criterion identified 2 distinct regressions representing response subgroups.

FIGURE 1.

RBC EPA response phenotype identification in Americans of African ancestry who received ω-3 FA supplementation for 6 wk. (A) Finite mixture regression analysis identified 2 distinct response phenotypes in the ω-3 FA–supplemented group, one showing a significant correlation between postintervention change in RBC EPA (i.e., ω-3 High; n = 29; r² = 0.59, P < 0.001) and one that did not (ω-3 Low; n = 12; r² = 0.07, P = 0.4). The placebo control also showed no correlation with compliance (n = 42; r² = 0.02, P = 0.4). (B) Correcting changes in RBC EPA for baseline status and the mg/kg-consumed dose reassigned 1 high-response member to the low-response group and showed that the magnitude of response was greater in participants with the lowest baseline RBC EPA concentrations in both groups (ω-3 High: n = 28; slope = −0.02, y-intercept = 0.0129, r² = 0.52, P < 0.0001; ω-3 Low: n = 13; slope = −0.02, y-intercept = -0.0038, r² = 0.52, P = 0.005) as well as in the placebo controls (n = 42; slope = −0.008, r² = 0.17, P = 0.006). Regression results ± 95% CIs are shown. Placebo data and regressions with nonsignificant slopes are not shown. Pre, presupplementation; ω-3 High, ω-3 high responders; ω-3 Low, ω-3 low responders.

Statistical analysis.

Statistical analyses were performed by using PASW Statistics version 18.0 for Windows (IBM SPSS, Inc.). When not normally distributed, normal transformations of continuous variables were attempted using an iterative subroutine in the multivariate analysis software utility imDEV (version 1.4.2; available from: https://sourceforge.net/projects/imdev/) (24). If transformation failed to achieve normality, nonparametric comparisons were made. Untransformed data are presented in tables and figures. Values reported in all tables and text are presented as means ± SDs, and differences were considered significant at P < 0.05. Initial ω-3 FA and placebo group comparisons were made with t tests and χ² analyses. After separating the ω-3 FA–supplemented group into responsive groups (see section entitled “Response phenotype identification”), χ² statistics and ANOVA procedures were used to compare the baseline characteristics and RBC EPA mean change between ω-3 High, ω-3 Low, and placebo groups. Continuous variables not normally distributed were transformed to ranks and the Mann-Whitney U or Kruskal-Wallis tests were used where appropriate. When P values were <0.05, pairwise comparisons were performed to determine significant differences between groups.

Treatment effects on plasma lipoproteins were assessed after adjusting data for gender and BMI by using general linear model (GLM) ANOVA using baseline values as covariates and response to treatment (ω-3 High, ω-3 Low, and placebo) as a factor. Differences in dietary intake were assessed by using GLM ANOVA adjusting for gender and total energy intake (kcal). When GLM ANOVA overall P values were <0.05, post hoc multiple comparisons were made by using the Holm-Sidak test. Within-group repeated-measures analyses were performed on untransformed data by using the nonparametric Wilcoxon's signed-rank test. Linear regression analyses were used to examine the relation between RBC EPA concentrations and variables identified as having significantly different concentrations between groups. Where significant relations were identified, the ability of these variables to accurately predict response to ω-3 FA supplementation was assessed by receiver operating characteristic curve analysis. A cutoff providing optimal sensitivity and specificity for the identification of responders was calculated.

Multivariate analyses were performed using imDEV version 1.4.2 to highlight differences in dietary patterns between ω-3 High and ω-3 Low groups, after truncating the data set to contain variables with a 2-tailed t test P < 0.1. Gender- and energy-adjusted dietary data were subjected to a hierarchal cluster analysis by using the complete linkage (farthest neighbor) agglomeration and Minkowski distances to highlight covariant foods and nutrients. A partial least-squares discriminant analysis was used to highlight differences in dietary patterns between ω-3 High and ω-3 Low groups.

Results

Participant characteristics.

A total of 83 participants completed the 6-wk intervention, returned their capsules, and provided a postintervention plasma sample. This subset included 61 women and 22 men with a mean age of 35.8 ± 12 y and BMI of 27.9 ± 4.5 kg/m². The ω-3 FA (n = 41) and placebo (n = 42) groups did not differ with respect to age (ω-3 FA: 37.5 ± 11 y; placebo: 34.2 ± 12 y; P = 0.20), BMI (ω-3 FA: 28.0 ± 4.5 kg/m²; placebo: 27.7 ± 4.6 kg/m²), gender (ω-3 FA: 10 men and 31 women; placebo: 12 men and 30 women), baseline RBC ω-3 FA concentrations (e.g., RBC EPA: ω-3 FA, 0.42 ± 0.22; placebo, 0.49 ± 0.24), or percentage of capsules consumed (ω-3 FA: 85.9 ± 17%; placebo: 88.0 ± 17%).

RBC ω-3 FA supplementation responses.

As previously reported, ω-3 FA supplementation increased RBC long-chain ω-3 FAs including EPA (ω-3 FA: 1.30 ± 1.00 mol; placebo: −0.02 ± 0.26 mol; P < 0.001) and DHA, at the expense of ω-6 FAs (15). However, in the ω-3 FA–supplemented group, a large variation in response was evident, most notably in RBC EPA change, which ranged from −0.15% to +3.25%, an ∼4-fold increased variance compared with the placebo group (F-test, P < 0.001). Changes in RBC EPA were analyzed as functions of either the ω-3 FA consumed dose (i.e., % compliance; Fig. 1A) or the consumed dose adjusted for body mass and analyzed as a function of baseline EPA status (Fig. 1B). Mixed-model regressions indicated that both expressions of EPA change were best solved by a combination of 2 distinct regression equations, indicating a discontinuous variance. Considering that both dose and baseline status provided the best group discrimination, indicating an ∼70/30 split in high and low responders and a mean change in RBC EPA that was different across ω-3 High, ω-3 Low, and placebo groups (Table 1).

No difference in the Alox5 promotor specificity protein 1 transcription factor genotype distribution, gender, age, blood pressure, body weight, or compliance was observed between the 3 groups (Table 1), and the mean ω-3 FA dose consumed was 34.3 ± 9.4 mg/(kg ⋅ d). Whereas there were more women in the ω-3 Low group, the gender distribution in the ω-3 High group was not significantly different (Fig. 1B). However, when groups were evaluated for gender effects, baseline RBC EPA content was found to increase with age in women (n = 56; r = 0.27, P = 0.041) but not in men (n = 22; r = −0.17, P = 0.45). Significant age-associated increases were also detected in women from both the low-response (n = 12; r = 0.58, P = 0.047) and high-response (n = 19; r = 0.49, P = 0.033) groups at baseline. Similar age-associated elevations were observed in women postsupplementation (ω-3 Low: r = 0.76, P = 0.004; ω-3 High: r = 0.56, P = 0.012). Age did not significantly correlate with the magnitude of the postintervention change in RBC EPA in any group (P > 0.05).

Post hoc comparisons showed that the relative abundance of palmitate (i.e., 16:0) and oleate (i.e., 18:1ω-9) were higher, whereas arachidonate (i.e., 20:4ω-6) was lower in the RBCs of the ω-3 Low versus the ω-3 High groups at baseline. Similarly, the RBC EPA, DHA, and EPA+DHA (i.e., the Omega-3 Index) contents of the ω-3 Low group were lower than that in both the ω-3 High and placebo groups at baseline, whereas ω-3 High and placebo groups were not different. In addition, the ω-3 High group had greater EPA increases (∼5-fold) than both the ω-3 Low (∼0.5-fold; P < 0.001) and placebo groups (P < 0.001), and the ω-3 Low and placebo groups also differed (P = 0.003). Whereas results were similar for DHA and the Omega-3 Index, the differential changes in ω-3 status between the ω-3 High and ω-3 Low groups were greater for RBC EPA (∼10-fold) than either DHA (∼3-fold) or the Omega-3 Index (∼5-fold). The fold-changes from baseline for these variables are included in Table 1.

Monocyte oxylipin production responses.

The 10-fold relative increase in RBC EPA was also reflected by a 3-fold increase in the production of 5-hydroxy EPA (5-HEPE) by monocytes stimulated with the calcium ionophore A23187. As seen in Fig. 2A, the changes in RBC EPA and 5-HEPE were positively correlated in both the ω-3 High and ω-3 Low groups. However, changes in RBC EPA and the proinflammatory arachidonic acid metabolite leukotriene B₄ were only negatively correlated in the ω-3 High group (Fig. 2B; n = 29; r = −0.50, P < 0.01).

FIGURE 2.

RBC EPA enrichment in African Americans receiving fish oil supplementation for 6 wk correlates with monocyte capacity to produce immunomodulatory metabolites. Changes in RBC EPA correlated with A23187-stimulated monocyte EPA-derived 5-HEPE increases in both responsive groups (ω-3 High: n = 28; slope = 11, r² = 0.55, P < 0.0001; ω-3 Low: n = 11; slope = 34, r² = 0.57, P = 0.007) (A) and with arachidonic-derived LTB4 decreases in the high-response group only (ω-3 High: n = 29; slope = −2.0, r² = 0.25, P = 0.005; ω-3 Low: n = 12; P = 0.98) (B). Placebo controls showed no changes in either variable. Regression results ± 95% CIs are shown. Placebo data and regressions with nonsignificant slopes are not shown. LTB4, leukotriene B₄; 5-HEPE, 5-hydroxyeicosapentaenoic acid; ω-3 High, ω-3 high responders; ω-3 Low, ω-3 low responders.

Plasma TG and VLDL responses.

A significant group effect was evident in plasma lipids and lipoproteins (Table 2). Overall the ω-3 High group showed greater beneficial changes in lipid and lipoprotein concentrations compared with the ω-3 Low group, with no major changes evident in the placebo group. For example, ω-3 FA supplementation decreased TG concentrations relative to the placebo group (P < 0.001), with declines being greater in the ω-3 High group (−28.9%) compared with the ω-3 Low group (−13.6%). As expected, these findings were reflected in VLDL TG concentrations in which the ω-3 High group showed a 27.1% greater decrease in total VLDL TGs than did the ω-3 Low group (P < 0.001). These effects were also observed for large, medium, and small VLDL TG concentrations (Table 2). The number of VLDL particles showed similar patterns in which the ω-3 High group exhibited a greater reduction than did the ω-3 Low group, whereas the placebo group did not change. Within-group repeated-measures analysis confirmed that reductions in total TG, VLDL TG, and VLDL particles were significant in the ω-3 High group (all P < 0.001). Although less responsive, the ω-3 Low group reductions in total TGs (P = 0.009), total VLDL TGs (P = 0.02), and large VLDL TGs (P < 0.05) were all significantly different from the placebo group. The placebo group showed only a slight increase in small VLDL TG (P < 0.05) and small VLDL particles (P < 0.05).

TABLE 2.

Plasma lipid and lipoprotein concentrations in Americans of African ancestry who received placebo or ω-3 FA supplements for 6 wk, by RBC response to supplementation¹

	ω-3 High (n = 28)		ω-3 Low (n = 13)		Placebo (n = 42)
Variable	Pre	Post	Pre	Post	Pre	Post	P
Plasma TGs, mg/dL	81.7 ± 58	58.1 ± 35a	84.6 ± 32	73.1 ± 26b	72.7 ± 29	71.1 ± 30b	<0.001
VLDL TGs, mg/dL
Total	51.5 ± 55	27.9 ± 33a	49.9 ± 31	40.6 ± 24b	41.3 ± 28	39.2 ± 29b	<0.001
Large	14.3 ± 23	5.95 ± 8.7	13.3 ± 9.5	8.55 ± 5.5	10.7 ± 9.0	8.99 ± 8.3	0.08
Medium	25.2 ± 33	14.0 ± 23a	21.8 ± 19	18.7 ± 19a,b	19.8 ± 17	18.1 ± 22b	0.008
Small	12.0 ± 4.9	8.04 ± 6.5a	14.8 ± 9.9	13.3 ± 9.0a,b	10.8 ± 6.8	12.1 ± 6.0b	<0.001
VLDL particles, nmol/L
Total	46.3 ± 30	28.4 ± 26a	50.0 ± 32	44.0 ± 29b	39.4 ± 24	41.0 ± 24b	<0.001
Large	1.15 ± 3.1	0.330 ± 0.82a	0.910 ± 1.1	0.390 ± 0.47b	0.690 ± 1.1	0.490 ± 1.0b	<0.001
Medium	18.9 ± 22	10.9 ± 15a	16.4 ± 13	14.4 ± 14a,b	15.3 ± 12	14.1 ± 14b	0.013
Small	26.2 ± 11	17.2 ± 14a	32.8 ± 22	29.3 ± 20a,b	23.4 ± 15	26.4 ± 14b	<0.001

¹Values are means ± SDs adjusted for gender and BMI (kg/m²). Postintervention means without a common letter differ by ANOVA using the Holm-Sidak multiple-comparison post hoc test (α < 0.05). Post, postsupplementation; Pre, presupplementation; ω-3 High, ω-3 high responders; ω-3 Low, ω-3 low responders.

Dietary pattern analysis.

Analysis of the FFQ data revealed differences in the self-reported habitual diets of low and high ω-3 response groups as shown in Table 3. Neither energy intake nor the percentage of energy from protein or carbohydrate differed between groups, and no differences in dietary fat composition were detected (Table 3). Whereas a direct comparison of the ω-3 Low and ω-3 High groups showed an elevation in the percentage of energy from sweets in the ω-3 Low group that approached significance (P = 0.10), this was influenced by a single individual who could not be excluded as a statistical outlier. Conversely, 3 of the 6 measures of vegetable consumption were low in the ω-3 Low group. Specifically, the ω-3 Low group consumed less of “not legumes/potatoes,” which includes a wide variety of vegetables such as green beans, broccoli, avocado, carrots, corn, tomatoes, squash, sweet potato, and spinach, as well as vegetables from sources such as stew, spaghetti, and soup (P = 0.02). The ω-3 Low group also consumed less of the “other” vegetables group compared with the ω-3 High group (P = 0.04), and more clearly, significantly fewer “dark-green” vegetables, which include greens, broccoli, green salad, and spinach (P = 0.002), than either the ω-3 High or placebo groups, who showed similar intakes. Transformation of the FFQ data into HEI-2005 scores allowed an alternative assessment of cohort diet quality. The overall diet quality as reported by cumulative HEI scores were similar between groups and were comparable to those reported for the non-Hispanic black population in 2005 (Fig. 3). Of the HEI-2005 categories, only the HEI category of dark-green and orange vegetables and legumes significantly differed between the low and high ω-3 groups.

TABLE 3.

Habitual dietary component intake estimated from Block FFQ analysis in Americans of African ancestry who received placebo or ω-3 FA supplements for 6 wk, by RBC response to supplementation¹

Dietary component2	ω-3 High (n = 25)	ω-3 Low (n = 13)	Placebo (n = 39)	P (ANCOVA)
Energy, kcal/d	1860 ± 730	1610 ± 640	1780 ± 790	0.79
Fat, % of daily energy	33.9 ± 5.3	36.8 ± 8.4	36.0 ± 7.7	0.53
Protein, % of daily energy	15.8 ± 2.5	14.1 ± 2.4	15.6 ± 3.1	0.30
Carbohydrate, % of daily energy	51.5 ± 7.5	49.8 ± 8.7	49.2 ± 10	0.68
Sweets, % of daily energy	15.5 ± 6.1	21.2 ± 10	15.9 ± 10	0.16
Alcohol, % of daily energy	1.65 ± 1.9	1.53 ± 1.9	1.76 ± 2.4	0.99
Fats consumed
Total fat, g/d	69.9 ± 30	65.1 ± 30	69.2 ± 30	0.75
SFAs, g/d	20.8 ± 10	19.3 ± 9.0	20.7 ± 9.2	0.70
MUFAs, g/d	26.4 ± 11	24.4 ± 11	26.7 ± 11	0.63
PUFAs, g/d	16.8 ± 8.7	16.2 ± 8.0	16.0 ± 7.4	0.52
ω-3 PUFAs, g/d	1.77 ± 1.2	1.42 ± 0.73	1.79 ± 1.0	0.40
Total ω-3 PUFAs,3 g/d	1.78 ± 1.2	1.42 ± 0.72	1.84 ± 1.0	0.32
trans-Fat, g/d	2.13 ± 0.96	2.10 ± 1.3	2.29 ± 1.1	0.39
Cholesterol, mg/d	231 ± 130	187 ± 71	258 ± 150	0.28
Fiber, g/d	20.2 ± 11a	12.7 ± 6.1b	17.8 ± 9.9a,b	0.038
Fruit servings, cups/d	1.33 ± 0.94	0.990 ± 0.75	1.36 ± 1.0	0.58
Vegetable servings
Not legumes/potatoes, cups/d	1.63 ± 1.3a	0.920 ± 0.69b	1.60 ± 1.2a	0.016
Dark-green, cups/d	0.570 ± 0.57a	0.230 ± 0.21b	0.590 ± 0.59a	0.002
Orange, cups/d	0.120 ± 0.10	0.100 ± 0.11	0.140 ± 0.15	0.34
Legumes/soy, cups equiv/d	0.340 ± 0.44	0.240 ± 0.40	0.230 ± 0.35	0.39
Potatoes, cups/d	0.230 ± 0.17	0.240 ± 0.23	0.270 ± 0.20	0.70
Other, cups/d	0.940 ± 0.68	0.580 ± 0.43*	0.850 ± 0.62	0.06
Total grain servings, oz equiv/d	5.57 ± 2.1	4.47 ± 2.6	4.93 ± 2.9	0.33
Whole-grain servings, oz equiv/d	1.57 ± 1.1	0.960 ± 0.47	1.26 ± 0.91	0.31
Meat servings, oz/d	4.18 ± 3.3	3.32 ± 2.0	4.12 ± 2.8	0.70
Nuts/seeds servings, oz equiv/d	0.550 ± 0.58	0.460 ± 0.84	0.430 ± 0.39	0.78
Dairy servings, cup equiv/d	1.13 ± 0.92	0.690 ± 0.55	0.900 ± 0.71	0.26
Oils servings, tsp/d	2.61 ± 1.8	2.58 ± 1.9	2.40 ± 1.4	0.95

¹Values are means ± SDs. Participants identified as over- and underreporters (3 ω-3 High, 3 placebo) were removed before analysis. After adjusting for gender and total energy intake, means without a common letter differed by ANCOVA with a Holm-Sidak multiple-comparison test (a, b) or differed from the ω-3 High group by 2-tailed t tests (*) (α < 0.05). equiv, equivalents; oz, ounce; tsp, teaspoon; ω-3 High, ω-3 high responders; ω-3 Low, ω-3 low responders.

²Servings represent USDA My Pyramid servings: 1 cup = 236.6 mL, 1 oz = 29.6 mL, 1 tsp = 4.92 mL.

³Total ω-3 PUFAs represent the sum of ω-3 PUFAs from food and supplements.

FIGURE 3.

Assessment of diet quality in Americans of African ancestry who received placebo or ω-3 FA supplements for 6 wk using modified HEI–2005 scoring. (A) The HEI categories are shown for the study groups and for the U.S. 2005 national average in non-Hispanic blacks. Although the study cohort differed significantly from the national population in multiple categories, calories from SOFAAs are not directly comparable (see text). The HEI–DGOL subscores without common letters differed by ANOVA (P < 0.05). (B) Study HEI-DGOL subscores ranged from 0.22 to 5.0, with 62% of the ω-3 Low group and 64% of the ω-3 High group found in the second (1.4–2.6) and fourth (3.8–5.0) quartiles, respectively. (C) The HEI-Total score did not differ between groups (ω-3 Low: 56 ± 4; ω-3 High: 64 ± 2; Placebo: 61 ± 1; P = 0.2) and was comparable to that in the 2005 non-Hispanic black population (HEI = 54). The ω-3 High group HEI-Total scores were skewed to the observed range (36–78) upper quartiles. DGOL, dark-green and orange vegetable plus legume; HEI, Healthy Eating Index; Non-Hisp, non-Hispanic; Q, quartile; SOFAA, solid fats, added sugars, and alcohols; Veg, vegetables; ω-3 High, ω-3 high responders; ω-3 Low, ω-3 low responders.

Postintervention RBC EPA in the ω-3 FA–supplemented group (n = 41) was significantly correlated with intake measurement of both dark-green vegetables (r = 0.58, P < 0.001) and “not legumes/potatoes” (r = 0.52, P = 0.001), categories with overlapping compositions. To examine the utility of dark-green vegetable intake as a predictor of RBC EPA response to ω-3 FA supplementation in the current study, a receiver operating characteristics curve was constructed by using data from the ω-3 low responders and those who achieved an EPA concentration >2% after supplementation (AUC = 0.85; 95% CI: 0.698, 1.0; P = 0.002). A cutoff of 0.29 cup/d of dark-green vegetables enabled high- and low-response group membership assignment with a sensitivity of 86% and a specificity of 77%. Furthermore, the low vegetable intake in the ω-3 Low group coincided with lower intake of other nutrients. A hierarchical cluster analysis of the FFQ-derived data comparing the ω-3 Low and ω-3 High groups identified 3 distinct nutrient clusters containing the following: 1) percentage of calories from sweets and BMI (ω-3 Low > ω-3 High), 2) vegetable-derived nutrients and associated measures (ω-3 Low < ω-3 High), and 3) protein- and dairy-derived nutrients and associated measures (ω-3 Low < ω-3 High) (data not shown). The strongest associations between calculated dark-green vegetable and nutrient intake were for vitamin K and lutein/zeaxanthin > folate from food > vitamin A and β-carotene > vitamin B-3 (niacin), vitamin B-6, fiber, iron, total folate, magnesium, and potassium. Other nutrients present in lower quantities in the diets of ω-3 low responders that clustered with baseline RBC EPA content included dietary protein, selenium, zinc, phosphorus, vitamin B-2 (riboflavin), and vitamin D.

Discussion

The aim of this investigation was to examine the variation in response to ω-3 FA supplementation in a cohort of healthy black Americans and to identify factors that modulated these effects. Reported variability in response to ω-3 FA supplementation is common, and numerous diet × gene interactions have been reported (13–15, 25, 26). Participants in the current study were recruited on the basis of specific Alox5 gene promoter variants that have been suggested to influence the cardiovascular benefit of dietary ω-3 FAs (13, 14). Whereas mean ω-3 FA responses were lower in the Alox5“dd” (double allelic deletion) phenotype (15), these gene variants did not segregate the identified ω-3 Low from the ω-3 High phenotypes. In the Fish Oil Intervention and Genotype (FINGEN) study, 31% of ω-3 FA–supplemented participants showed no changes in plasma TGs despite compliance of ∼95% (10). Similarly, ∼32% of our ω-3 FA participants were classified as low responders, showing a nonsignificant ∼13% reduction in TG concentration (P = 0.099), as opposed to a ∼30% decrease seen in the ω-3 High group (P = 0.002). Considering the known dose-response association with ω-3 FA supplementation (27–29), the 3-g/d dose used here versus 1.8-g/d dose in the FINGEN study may explain our ability to observe marginal changes in the low-response group.

Although variation in response to treatment with ω-3 FAs has been reported, few factors associated with responsiveness have been described in detail. Gender and basal ω-3 FA status are notable exceptions. Because women are, on average, smaller than men, studies providing a constant ω-3 FA dose per participant must account for body mass to avoid the appearance of an elevated response in women (10, 23). Unlike the FINGEN study (10) where men were found to have a dose-dependent hypotriglyceridemic response after adjusting for the mass-specific dose, such gender-specific trends were not observed here. However, we found that women but not men showed a significant increase in RBC EPA content with age and that this gender × age effect was evident presupplementation. Although this suggests the possibility of hormonal effects on ω-3 FA status, 2 separate reports based on U.S. study populations of 160 and 160,000 participants demonstrated that the Omega-3 Index increased by 0.5 units for every decade of age in both genders (30, 31). Therefore, whereas the age-associated changes may be linked to either lifestyle or metabolic changes, the lack of this effect in the small number of male participants (n = 10) in our study is likely due to sample size limitations.

We recently reported that individuals with low EPA concentration at baseline had greater dose-dependent responses to ω-3 FA supplementation (23). This relation was also observed in the current study; however, basal status did not account for the differential response between groups. In contrast, these response groups were characterized by distinct dietary patterns based on Block 2005 FFQ surveys of habitual diet. Although such dietary surveys have limitations in accurately determining dietary composition and/or intake, they are considered sufficiently accurate for evaluating dietary associations with disease phenotypes in epidemiologic studies (32, 33). In addition, caloric intake adjustment can result in excellent diet composition agreement between instruments with different levels of complexity (32, 34). Participants consuming ω-3 FAs showed positive correlations between postsupplementation RBC EPA concentration and dietary intake of dark-green vegetables and their associated nutrients, including vitamin K, folate, lutein/zeaxanthin, and vitamin B-6. Moreover, pairing ω-3 FA supplementation with a dark-green vegetable intake of more than ∼0.3 cup/d was associated with achieving an RBC EPA abundance of >2% and an Omega-3 Index of ∼8%, the target for optimal health (35).

Although elucidating the biologic mechanisms linking vegetable intake to ω-3 FA supplementation response is beyond the scope of this study, it is possible to hypothesize such mechanisms. The benefits of a diet rich in fruit and vegetables are well established (36–41). For instance, fruit and vegetable consumption can reduce oxidative stress (42, 43), including in vivo lipid peroxidation (44). Thus, it is conceivable that lower vegetable intake induces a prooxidative state that may diminish ω-3 FA availability, thereby reducing their availability as substrates for developing RBC membranes in the ω-3 Low group (45). Alternatively, digestive and absorptive processes may regulate the availability and incorporation of ω-3 FAs into the body, and diet-associated changes in the gut microbiota may mediate these processes. Importantly, foods rich in plant fiber and polyphenolics are known to modulate the composition and activity of the gut microbiota (46). Furthermore, ω-3 FAs can influence the bacterial composition of the gut (47, 48), and intestinal bacteria can regulate mammalian lipid metabolism (49, 50). The question therefore arises as to whether the gut microbiota might influence FA composition of host tissues. Such a case was recently reported in which changing the gut microbiota composition of mice significantly increased EPA and DHA concentrations in adipose tissues (51). Taken together, these findings suggest that diets rich in dark-green vegetable may modulate gut microbiota and in turn influence FA availability and cellular composition in humans. Additional studies will be needed to determine whether an integrative network linking habitual dietary patterns, gut microbiota, and ω-3 FA metabolism is a plausible biologic mechanism.

In conclusion, our findings provide an indication that habitual consumption of dark-green vegetables may influence the efficacy of ω-3 FA supplementation. Because this study was a secondary analysis, the experimental design and cohort selection were not powered to address the hypotheses regarding vegetable intake and ω-3 FA supplementation response that emerged. In addition, limitations associated with compliance assessment and FFQ data collection must be acknowledged when interpreting these findings. Therefore, to assess the broader implications of these findings, it will be critical to determine how robust low dark-green vegetable intake is in predicting a low ω-3 FA supplementation response phenotype and to further determine if a person’s response phenotype can be altered by a change in dietary patterns. If supported, one could envisage a situation in which particular doses of ω-3 FAs could be recommended along with complementary guidelines (e.g., portions and types of vegetables to be consumed) on the basis of an individual’s basal RBC EPA concentrations and an assessment of habitual dietary patterns. Although further work will be required to confirm these findings and investigate their biologic mechanisms, this study has generated an intriguing and unexpected hypothesis for diet × diet interactions in the context of ω-3 FA supplementation.