PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Am J Clin Nutr. Author manuscript; available in PMC 2009 March 26.
Published in final edited form as:
PMCID: PMC2661261
NIHMSID: NIHMS90582

Peroxisome proliferator–activated receptor α genetic variation interacts with n–6 and long-chain n–3 fatty acid intake to affect total cholesterol and LDL-cholesterol concentrations in the Atherosclerosis Risk in Communities Study1,2,3

Abstract

Background

Peroxisome proliferator–activated receptor-α (PPARA) regulates the expression of genes involved in lipid metabolism. The binding of polyunsaturated fatty acids (PUFAs) to PPARA results in rapid changes in the expression of genes involved in lipid oxidation, with long-chain n–3 fatty acids being potent activators of PPARA.

Objective

We evaluated the potential effect modification of PPARA genetic variation on the association between PUFA intake, specifically n–6 and long-chain n–3 fatty acid intakes, and multiple lipid measures in the large biethnic Atherosclerosis Risk in Communities (ARIC) Study.

Design

Study participants (10 134 whites and 3480 African Americans) were selected from the ARIC Study—aprospective investigation of atherosclerosis and its clinical sequelae. Multiple linear regression models were used to assess the relation between PPARA genotypes, as well as dietary fatty acid intake, and baseline lipid measures. PPARA-specific effects of variation were assessed by including genotype-by-fatty acid interaction terms in each statistical model.

Results

PPARA genotype frequencies were significantly different between whites and African Americans. No significant associations between lipid measures and PPARA genotype were observed in either whites or African Americans. Significant genotype-by-n–6 fatty acid intake interactions were observed only in whites for the 3'untranslated region (UTR) G→A single nucleotide polymorphism (SNP) and total cholesterol (P = 0.03) and LDL cholesterol (P = 0.03). Significant genotype-by-long-chain n–3 fatty acid intake interactions were observed only in African Americans for the 3'UTR C→T SNP and total cholesterol (P = 0.03) and LDL cholesterol (P = 0.02).

Conclusions

Findings from the current study suggest that PPARA 3'UTR SNPs modulate the association between lipid concentrations and dietary n–6 fatty acid intake (in whites) and long-chain n–3 fatty acid intake (in African Americans) such that persons with homozygous variant genotypes have significantly lower total cholesterol and LDL-cholesterol measures when consuming higher quantities of n–6 or long-chain n–3 fatty acids.

INTRODUCTION

Peroxisome proliferator–activated receptors (PPARs) are ligand-dependent nuclear transcription factors belonging to the nuclear receptor superfamily, with 3 subtypes expressed in humans and encoded by different genes (PPARA, PPARG, and PPARB/D) (13). PPARs regulate target gene expression by binding to specific peroxisome proliferator response elements in enhancer sites of regulated genes as a heterodimer with the retinoid X receptor (3). PPAR-α (PPARA) regulates the expression of genes involved in lipid metabolism, and polyunsaturated fatty acids (PUFAs) are natural ligands of PPARA (4, 5). Studies have shown that binding of PUFAs to PPARA results in rapid changes in expression of genes involved in lipid oxidation (68). PUFAs are composed of 2 families, n–3 fatty acids and n–6 fatty acids, which interact metabolically (9). Studies have shown that long-chain n–3 fatty acids are more potent activators of PPARA than are n–6 fatty acids (10, 11).

The most commonly studied variant of the PPARA gene is a missense mutation (L162V) that has functional consequences on PPARA activity (5, 12, 13). Previous studies have shown the L162V variant allele to be associated with higher concentrations of LDL cholesterol, total cholesterol, apolipoprotein B (apo B), apolipoprotein C-III (apo C-III), and triacylglycerol (2, 12, 14). A recent study by Tai et al found the effect of the L162V polymorphism on triacylglycerol and apoC-III concentrations to be dependent on PUFA intake, with high intake triggering lower apoC-III and triacylglycerol concentrations in carriers of the 162V allele (5). The study by Tai et al was limited to ≈2000 white individuals from a single geographic location. For the current study, we evaluated the potential effect modification of PPARA genetic variation on the association between dietary n–6 and long-chain n–3 fatty acid intake and multiple lipid measures in the large biethnic (≈10 000 whites and ≈3500 African Americans) ARIC Study.

SUBJECTS AND METHODS

ARIC Study

Study participants were selected from the ARIC Study—a prospective investigation of atherosclerosis and its clinical sequelae involving 15 792 individuals aged 45–64 y at recruitment (1987–1989). Institutional review boards approved the ARIC Study, and all participants provided their written informed consent. A detailed description of the ARIC Study design and methods was published elsewhere (1517). Briefly, subjects were selected by probability sampling from 4 communities: Forsyth County, NC; Jackson, MS; northwestern suburbs of Minneapolis, MN; and Washington County, MD. Participants were excluded from the analyses (n = 2178) if they 1) prohibited the use of their DNA for research purposes, 2) had an ethnic background other than white or African American, 3) took cholesterol-lowering medications within 2 wk of the baseline clinical examination, 4) were missing data for any of the lipid measures or covariates included in the analysis, or 5) were missing genotype information for the PPARA polymorphisms. After the exclusions, a total of 10 134 whites and 3480 African Americans were available for analysis.

Baseline examination and laboratory measures

Home and clinic interviews and questionnaires included assessment of dietary intake, use of cholesterol-lowering medications, and cigarette smoking. Cigarette smoking status was analyzed by comparing current smokers with individuals who had formerly or never smoked. Usual dietary intake was assessed with the use of a 66-item, interviewer-administered, semiquantitative food-frequency questionnaire (FFQ) administered at the baseline examination. The questionnaire was a modified version of the 61-item FFQ designed and validated by Willett et al (18). Participants were asked to report the frequency of consumption of each food or beverage on the basis of 9 categories, ranging from never or <1 time/mo to ≥6 times/d. Interviewers obtained additional information, including the type of fat typically used in frying and baking (butter, margarine, vegetable oil, vegetable shortening, and lard), as well as the brand name of breakfast cereal usually consumed (open-ended response).

Fat intake data were calculated in grams per day for both n–6 and long-chain n–3 fatty acid intake. Dietary n–6 (linoleic acid) and long-chain n–3 fatty acid [eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA)] intakes were included in the analyses as categorical variables. Dietary n–6 and long-chain n–3 fatty acid intakes were classified into 2 groups (low and high; separately by race) according to the race-specific mean value. For whites, dietary intakes were classified as follows: n–6 fatty acids (low, ≤7.99 g/d; high, >7.99 g/d) and long-chain n–3 fatty acids (low, ≤0.22 g/d; high, >0.22 g/d). For African Americans, dietary intakes were classified as follows: n–6 fatty acids (low, ≤7.05 g/d; high, >7.05 g/d) and long-chain n–3 fatty acids (low, ≤0.32 g/d; high, >0.32 g/d).

Body mass index (BMI; in kg/m2) was calculated from height and weight measurements. Plasma concentrations of total cholesterol and triacylglycerol were measured enzymatically, and the concentration of LDL cholesterol was calculated (19). HDL cholesterol was measured after dextran-magnesium precipitation of non-HDL lipoproteins (20).

Genotype determination

Single nucleotide polymorphisms (SNPs) were selected on the basis of being non-synonymous variants or variants located within the untranslated regions (UTRs) of the PPARA gene. The SNPs included the following: L162V (rs1800206), R127Q (rs1800204), V227A (rs1800234), D304N (rs1800242), A268V (rs1042311), Q413L (rs9615759), R409T (rs1800243), 3'UTR G→A (rs6008259), and 3'UTR C→T (rs3892755). Genotyping was carried out using the SNPlex or TaqMan Assays (Applied Biosystems, Inc, Foster City, CA). Primers and probes are available from the authors on request.

Statistical analysis

Allele frequencies were estimated by gene counting. Hardy-Weinberg equilibrium expectations were tested with a chi-square goodness-of-fit test. Proportions, means, and SEMs of the base-line measurements were calculated. All statistical analyses were conducted with the use of STATA software (version 9.2; College Station, TX). Because allele frequencies were different between whites and African Americans for all SNPs studied, all analyses were performed separately by race. Variant alleles were identified as the low frequency allele in whites, and homozygous nonvariant genotypes were designated as the referent group in the statistical analyses. Six of the 9 SNPs had a variant allele frequency <0.01 in both whites and African Americans and were not investigated further. The 3 SNPs investigated in the current study were L162V (rs1800206), 3'UTR G→A (rs6008259), and 3'UTR C→T (rs3892755).

Multiple linear regression models were used to assess the relation between PPARA genotypes and baseline lipid measures. Regression model coefficients were determined to be significant with the use of a standard t test. For all analyses, the covariates included age (continuous), sex (categorical), ARIC field center (categorical), BMI (continuous), smoking status (categorical), use of medications that secondarily lower cholesterol (categorical), and total energy intake (continuous). Evidence for a PPARA-specific effect of variation was assessed by including a genotype-by-fatty acid interaction term in the model; statistical significance was assessed in each gene model with a standard t test (n–6 and long-chain n–3 fatty acid intakes were considered in separate models).

RESULTS

Race-specific characteristics of the study population are presented in Table 1. PPARA genotype frequencies were significantly different between whites and African Americans, and genotype distributions for each racial group were in accordance with Hardy-Weinberg equilibrium expectations. The 162V allele was more commonly observed in whites than in African Americans, with the 162V allele being rare in African Americans [f(V) = 0.015]. The 3'UTR T allele was more commonly observed in African Americans than in whites, with the 3'UTR T allele being rare in whites [f(T) = 0.003]. The 3'UTR G→A SNP had disparate allele frequencies in whites [f(A) = 0.18] compared with African Americans [f(A) = 0.67].

TABLE 1
Baseline characteristics of the Atherosclerosis Risk in Communities Study population, by race-ethnic group1

Mean lipid measures for each racial group, according to PPARA genotype, are presented in Table 2. No significant associations between lipid measures and PPARA genotype were observed in either whites or African Americans. Mean lipid measures for each racial group are shown in Table 3 according to n–6 and long-chain n–3 fatty acid intakes. A significant association was observed for HDL cholesterol and long-chain n–3 fatty acid intake in African Americans, ie, persons with higher intakes of long-chain n–3 fatty acids (EPA + DHA) had higher HDL- cholesterol measures. No significant associations between lipid measures and long-chain n–3 fatty acid intakes were observed in whites. No significant associations between lipid measures and n–6 fatty acid intake were observed in African Americans. However, significant associations were observed between n–6 fatty acid intake and cholesterol measures in whites. Whites consuming higher amounts of n–6 fatty acids had significantly higher measures of total and LDL cholesterol and significantly lower measures of HDL cholesterol.

TABLE 2
Lipid concentration in whites and African Americans according to peroxisome proliferator–activated receptor-α genotype1
TABLE 3
Lipid concentrations according to long-chain n–3 and n–6 fatty acid intake, by race-ethnic group1

Significant genotype-by-long-chain n–3 fatty acid intake interactions were observed only in African Americans for the 3'UTR C→T SNP and total cholesterol (P = 0.03) and LDL cholesterol (P = 0.02). Significant genotype-by-n–6 fatty acid intake interactions were observed only in whites for the 3'UTR G→A SNP and total cholesterol (P = 0.03) and LDL cholesterol (P = 0.03). Mean measures of total cholesterol and LDL cholesterol, specific to African Americans, according to PPARA 3'UTR CT genotype and long-chain n–3 fatty acid intake are shown in Figure 1. Consistent results were observed for total cholesterol and LDL cholesterol such that African Americans with a high long-chain n–3 fatty acid intake (>0.32 g/d) had significantly lower total and LDL-cholesterol concentrations if they were homozygous for the PPARA 3'UTR TT genotype. In contrast, African Americans consuming higher amounts of long-chain n–3 fatty acids had higher total and LDL-cholesterol concentrations if they had either the PPARA 3'UTR CC or CT genotype. Mean measures of total cholesterol and LDL cholesterol, specific to whites, according to PPARA 3'UTR GA genotype and n–6 fatty acid intake are shown in Figure 2. Consistent results were observed for total cholesterol and LDL cholesterol such that white individuals with a high n–6 fatty acid intake (>7.99 g/d) had significantly lower total and LDL-cholesterol concentrations if they were homozygous for the PPARA 3'UTR AA genotype. In contrast, whites consuming higher amounts of n–6 fatty acids had higher total and LDL-cholesterol concentrations if they had either the PPARA 3'UTR GG or AG genotypes.

FIGURE 1
Total cholesterol and LDL-cholesterol concentrations according to long-chain n–3 fatty acid (FA) intake (eicosapentaenoic acid + docosahexaenoic acid) and peroxisome proliferator–activated receptor-α (PPARA) 3'untranslated region ...
FIGURE 2
Total cholesterol and LDL-cholesterol concentrations according to n–6 fatty acid (FA) intake (linoleic acid) and peroxisome proliferator– activated receptor-α (PPARA) 3'untranslated region (UTR) GA genotype in whites. ...

DISCUSSION

Previous research has shown that long-chain n–3 fatty acids (EPA + DHA), essential fatty acids that cannot be synthesized by mammals and therefore must be obtained from dietary sources, are potent activators of PPARA, resulting in the altered expression of genes involved in lipid metabolism (68, 10, 11, 21). Previous studies have shown the PPARA L162V SNP to be associated with multiple lipid and lipoprotein measures, with one study finding the effect of the L162V SNP on triacylglycerol and apoC-III concentrations to be dependent on PUFA intake (2, 5, 12, 14). In the current study, we did not observe significant associations between the PPARA L162V SNP and multiple lipid measures. However, we did observe a significant interaction between the PPARA 3'UTR C→T SNP and long-chain n–3 fatty acid intake with regards to total cholesterol and LDL-cholesterol concentrations in African American participants of the ARIC Study and a significant interaction between the PPARA 3'UTR G→A SNP and n–6 fatty acid intake with regards to total cholesterol and LDL-cholesterol concentrations in white participants of the ARIC Study. This finding suggests that genetic variation within the PPARA 3'UTR region can modulate the association of dietary n–6 and long-chain n–3 fatty acid intakes with total and LDL-cholesterol concentrations in whites and African Americans.

Our study confirms the differing genetic structure of the PPARA locus in various ethnic groups (25). The L162V SNP was first identified in a white population, but is very rare in Asian and African American populations (12, 22, 23). Another PPARA SNP (V227A) was identified in Chinese and Japanese populations, but has not been observed in non-Asian populations (22, 24). In the current study, each PPARA SNP investigated had disparate allele frequencies between whites and African Americans and therefore may explain why significant effect modification in the 2 racial groups was observed for different SNPs.

It has been reported that the effect of PPARA genetic variation on the transcriptional activation associated with ligand (eg, PUFAs) binding to PPARA depends on the concentration of the ligand to which it is exposed (5). Although the functional consequences of the PPARA 3'UTR C→T and 3'UTR G→A SNPs are unknown, we hypothesize that higher long-chain n–3 fatty acid intakes (>0.32 g/d) in African Americans leads to increased activation of PPARA, which results in lower lipid concentrations in persons with the 3'UTR TT genotype. Whereas lower long-chain n–3 fatty acid intakes (≤0.32 g/d) in African Americans lead to decreased activation of PPARA, which results in higher lipid concentrations in persons with the 3'UTR TT genotype. Similarly, we hypothesize that higher n–6 fatty acid intakes (>7.99 g/d) in whites lead to increased activation of PPARA, which results in lower lipid concentrations in persons with the 3'UTR AA genotype. However, lower n–6 fatty acid intakes (≤7.99 g/d) in whites lead to decreased activation of PPARA, which results in higher lipid concentrations in persons with the 3'UTR AA genotype.

Our finding of a significant PPARA genotype-by-long-chain n–3 fatty acid intake specific to African Americans and a significant PPARA genotype-by-n–6 fatty acid intake in whites is intriguing. There are, however, a number of potential explanations. With regard to the significant findings observed with different PPARA SNPs, owing to the fact that linkage disequilibrium patterns are different between whites and African Americans, the SNPs may be in linkage disequilibrium with the true functional variant. Therefore, we may have detected an association with one marker polymorphism in African Americans and not in whites and with another marker polymorphism in whites and not in African Americans. We must also note that the 3'UTR T allele is very rare in whites [f(T) = 0.003], which may have reduced our power to detect a significant interaction effect in this racial group. In addition, PUFA intakes were categorized as low and high based on the race-specific mean, which was significantly different between whites and African Americans for both long-chain n–3 fatty acid intake (0.22 and 0.32, respectively; P < 0.0001) and n–6 fatty acid intake (7.99 and 7.05, respectively; P < 0.0001). It is possible that the level of long-chain n–3 fatty acid intake in whites in the ARIC Study is not high enough to affect the association between PPARA genotype and lipid measures; similarly, the level of n–6 fatty acid intake in African Americans in the ARIC Study may not be high enough to affect the association between PPARA genotype and lipid measures. Typical recommendations for long-chain n–3 fatty acid intakes (EPA + DHA) are 0.3–0.5 g/d (25), for which whites from the ARIC Study have mean intakes below these recommendations. Ethnic differences in dietary patterns may also contribute to these discordant results (26, 27). Within each racial group there may be a unique set of potential food and nutrient confounders that are difficult to adequately disentangle. A limitation of the ARIC Study is that it used a 66-item FFQ; a more comprehensive dietary questionnaire may be needed to adequately characterize intake and evaluate how dietary PUFA and long-chain fatty acid intakes interact with the PPARA gene.

We must also acknowledge that our results may have been due to chance findings when one considers multiple testing and that significant findings were not consistent between racial groups. However, previous studies of genetic variation in the PPARA gene have found gene-by-diet interactions specific to different polymorphisms in different racial-ethnic groups (ie, L162V in whites and V227A in Asians). We initially evaluated 9 PPARA SNPs (7 nonsynonymous and 2 UTR SNPs), but were restricted to further evaluation of only 3 because of rare allele frequencies. Two of the SNPs presented in the current study were rare in one of the studies (L162V in African Americans and 3'UTR C→T in whites). Additional investigation of genetic variation within the PPARA gene is needed to determine whether there is an SNP common to both whites and African Americans that is significantly associated with lipid measures and gene-diet interaction effects in both racial groups. There may also be a gene in close proximity to the PPARA gene that contains the functional SNP, in linkage disequilibrium with one or more PPARA SNPs, responsible for the significant findings observed in the different studies of different racial groups.

Acknowledgments

We thank the staff and participants of the ARIC Study for their important contributions.

The authors’ responsibilities were as follows—KAV: performed the statistical analysis and prepared the first draft of the manuscript. All authors contributed to the data interpretation and critically reviewed the manuscript.

Footnotes

2Supported by National Heart, Lung, and Blood Institute contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022.

None of the authors had a personal or financial conflict of interest.

REFERENCES

1. Newaz M, Blanton A, Fidelis P, Oyekan A. NAD(P)H oxidase/nitric oxide interactions in peroxisome proliferator activated receptor (PPAR)α-mediated cardiovascular effects. Mut Res. 2005;579:163–171. [PubMed]
2. Tai ES, Semissie S, Cupples LA, et al. Association between the PPARA L162V polymorphism and plasma lipid levels: the Framingham Off-spring Study. Arterioscler Thromb Vasc Biol. 2002;22:805–810. [PubMed]
3. Berger J, Moller DE. The mechanisms of action of PPARs. Annu Rev Med. 2002;53:409–435. [PubMed]
4. Zak I, Balcerzyk A, Sarecka B, Niemiec P, Ciemniewski Z, Dylag S. Contemporaneous carrier-state of two or three “proatherosclerotic” variants of APOE, ICAM1, PPARA and PAI-1 genes differentiate CAD patients from healthy individuals. Clin Chim Acta. 2005;362:110–118. [PubMed]
5. Tai ES, Corella D, Demissie S, et al. Polyunsaturated fatty acids interact with the PPARA-L162V polymorphism to affect plasma triglyceride and apolipoprotein C-III concentrations in the Framingham Heart Study. J Nutr. 2005;135:397–403. [PubMed]
6. Sampath H, Ntambi JM. Polyunsaturated fatty acid regulation of gene expression. Nutr Rev. 2004;62:333–339. [PubMed]
7. Jump DB, Clarke SD, MacDougald OA, Thelen A. Polyunsaturated fatty acids inhibit S14 gene transcription in rat liver and cultured hepatocytes. Proc Natl Acad Sci U S A. 1993;90:8454–8458. [PubMed]
8. Jump DB, Clarke SD, Thelen AT, Liimata M. Coordinate regulation of glycolytic and lipogenic gene expression by polyunsaturated fatty acids. J Lipid Res. 1994;35:1076–1084. [PubMed]
9. Ordovas JM. Genetic interactions with diet influence the risk of cardiovascular disease. Am J Clin Nutr. 2006;83 suppl:443S–446S. [PubMed]
10. Clarke SD. Polyunsaturated fatty acid regulation of gene transcription: a molecular mechanism to improve the metabolic syndrome. J Nutr. 2001;31:1129–1132. [PubMed]
11. Zheng X, Rivabene R, Cavallari C, et al. The effects of chylomicron remnants enriched in n–3 or n–6 polyunsaturated fatty acids on the transcription of genes regulating their uptake and metabolism by the liver: influence of cellular oxidative state. Free Radic Biol Med. 2002;32:1123–1131. [PubMed]
12. Vohl MC, Lepage P, Gaudet D, et al. Molecular scanning of the human PPARA gene: association of the L162V mutation with hyperapobetali-poproteinemia. J Lipid Res. 2000;41:945–952. [PubMed]
13. Sapone A, Peters JM, Sakai S, et al. The human peroxisome proliferator-activated receptor α gene: identification and functional characterization of two natural allelic variants. Pharmacogenetics. 2000;10(4):321–333. [PubMed]
14. Robitaille J, Brouillette C, Houde A, et al. Association between the PPAR –L162V polymorphism and components of the metabolic syndrome. J Hum Genet. 2004;49:482–489. [PubMed]
15. ARIC Investigators. The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. Am J Epidemiol. 1989;129:687–702. [PubMed]
16. White A, Folsom A, Chambless L, et al. Community surveillance of coronary heart disease in the Atherosclerosis Risk in Communities (ARIC) Study: methods and initial 2 years’ experience. J Clin Epidemiol. 1996;49:223–233. [PubMed]
17. Rosamond W, Folsom A, Chambless L, et al. Stroke incidence and survival among middle-aged adults: 9-year follow-up of the Atherosclerosis Risk in Communities (ARIC) cohort. Stroke. 1999;30:736–743. [PubMed]
18. Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol. 1985;122:51–65. [PubMed]
19. Siedel J, Hagele EO, Ziegenhorn J, Wahlefeld AW. Reagent for the enzymatic determination of serum total cholesterol with improved lipolytic efficiency. Clin Chem. 1983;29:1075–1085. [PubMed]
20. Warnick GR, Benderson JM, Albers JJ. Dextran sulfate-Mg2+ precipitation for quantitation of high-density lipoprotein cholesterol. Clin Chem. 1982;28:1379–1388. [PubMed]
21. Sessler AM, Ntambi JM. Polyunsaturated fatty acid regulation of gene expression. J Nutr. 1998;128:923–926. [PubMed]
22. Chan E, Tan CS, Deurenberg-Yap M, Chia KS, Chew SK, Tai ES. The V227A polymorphism at the PPARA locus is associated with serum lipid concentrations and modulates the association between dietary polyunsaturated fatty acid intake and serum high density lipoprotein concentrations in Chinese women. Atherosclerosis. 2006;187:309–315. [PubMed]
23. Flavell DM, Pineda TI, Jamshidi Y, et al. Variation in the PPARA gene is associated with altered function in vitro and plasma lipid concentrations in type II diabetic subjects. Diabetologia. 2000;43:673–680. [PubMed]
24. Yamakawa-Kobayashi K, Ishiguro H, Arinami T, Miyazaki R, Hamaguchi H. A Val227Ala polymorphism in the peroxisome proliferator activated receptor alpha (PPARA) gene is associated with variations in serum lipid levels. J Med Genet. 2002;39:189–191. [PMC free article] [PubMed]
25. Kris-Etherton PM, Harris WS, Appel LJ. Fish consumption, fish oil, omega-3 fatty acids, and cardiovascular disease. Circulation. 2002;106:2747–2757. [PubMed]
26. Houston DK, Stevens J, Cai J, Haines PS. Dairy, fruit, and vegetable intakes and functional limitations and disability in a biracial cohort: the ARIC study. Am J Clin Nutr. 2005;81(2):515–522. [PubMed]
27. Steffen LM, Jacobs DR, Jr, Stevens J, Shahar E, Carithers T, Folsom AR. Associations of whole-grain, refined-grain, and fruit and vegetable consumption with risks of all-cause mortality and incident coronary artery disease and ischemic stroke: the ARIC study. Am J Clin Nutr. 2003;78(3):383–390. [PubMed]