Whole grain intake has been related to reductions in total mortality (Jacobs et al. Reference Jacobs, Meyer, Kushi and Folsom1999, Reference Jacobs, Meyer and Solvoll2001), coronary artery disease mortality and morbidity (Morris et al. Reference Morris, Marr and Clayton1977; Fraser et al. Reference Fraser, Sabate, Beeson and Strahan1992; Pietinen et al. Reference Pietinen, Rimm and Korhonen1996; Jacobs et al. Reference Jacobs, Meyer, Kushi and Folsom1998; Liu et al. Reference Liu, Stampfer and Hu1999; Steffen et al. Reference Steffen, Jacobs, Stevens, Shahar, Carithers and Folsom2003b) and diabetes incidence (Liu et al. Reference Liu, Manson and Stampfer2000; Meyer et al. Reference Meyer, Kushi, Jacobs, Slavin, Sellers and Folsom2000; Fung et al. Reference Fung, Hu and Pereira2002; Montonen et al. Reference Montonen, Knekt, Jarvinen, Aromaa and Reunanen2003), independent of other health behaviours. In a review of whole grain intake, Jacobs & Gallaher (Reference Jacobs and Gallaher2004) found that habitual consumers of whole grain consistently had a 20–40 % reduction in long-term risk of coronary artery disease and type II diabetes as compared with those who rarely ate whole grains. This evidence contributed to an emphasis on the consumption of whole grains in the 2005 US Department of Agriculture Dietary Guidelines for Americans, which state: ‘Consume 3 or more ounce-equivalents of whole-grain products per day, with the rest of the recommended grains coming from enriched or whole-grain products. In general, at least half the grains should come from whole grains’ (http://www.healthierus.gov/dietaryguidelines) (US Department of Health & Human Services & the US Department of Agriculture (2005).
Whole grain food intake and dietary fibre intake, especially from cereal sources, have also been associated with favourable levels of insulin sensitivity (Lovejoy & DiGirolamo, Reference Lovejoy and DiGirolamo1992; Feskens et al. Reference Feskens, Loeber and Kromhout1994; Vitelli et al. Reference Vitelli, Folsom and Shahar1996; Marshall et al. Reference Marshall, Bessesen and Hamman1997; Pereira et al. Reference Pereira, Jacobs and Pins2002; Liese et al. Reference Liese, Roach, Sparks, Marquart, D'Agostino and Mayer-Davis2003; Steffen et al. Reference Steffen, Jacobs and Murtaugh2003a;), BMI (Pereira et al. Reference Pereira, Jacobs and Pins2002; Steffen et al. Reference Steffen, Jacobs and Murtaugh2003a) and 10-year weight gain (Ludwig et al. Reference Ludwig, Pereira and Kroenke1999). Despite the strong body of evidence relating high consumption of whole grain food intake to CVD risk factors and CVD morbidity and mortality, there have been no studies of whole grain food and subclinical atherosclerosis. Additionally, little research has assessed whether racial/ethnic heterogeneity exists in the relationship between whole grain intake and various CVD risk factors.
This paper focuses on the cross-sectional relationship between whole grain intake and selected CVD risk factors and measures of subclinical atherosclerosis using baseline data from the Multi-Ethnic Study of Atherosclerosis (MESA). We hypothesized that whole grain intake would be inversely associated with the following variables: BMI; serum insulin; C-reactive protein (CRP); IL-6; homocysteine; newly diagnosed diabetes and impaired fasting glucose; urine albumin:creatinine ratio (A/kC); carotid artery intima-media thickness; presence of coronary artery calcification (CAC).
Methods
Subjects
MESA is a prospective epidemiological cohort study initiated in July 2000 with the aim of exploring the prevalence, correlates and progression of subclinical and clinical CVD, with focus on assessing possible differences between non-Hispanic whites, Hispanics, African Americans and Chinese. A full description of the design and methods has been published elsewhere (Bild et al. Reference Bild, Bluemke and Burke2002). The MESA protocol was approved by local institutional review committees and all subjects gave informed consent. A total of 6814 men and women between the ages of 45 and 84 years, all of whom were free of clinical CVD at baseline, were selected from six US field centres.
Participants who had no diet data (n 577) or implausible energy intakes as defined by consuming >25 081 kJ/d (6000 kcal/d) or < 2508 kJ/d (600 kcal/d) (n 157) were excluded. Furthermore, participants were excluded if they had been previously diagnosed with diabetes (n 610), as these individuals may have changed their diets in response to disease. These criteria were not mutually exclusive, thus the present report includes baseline data on 5496 participants.
Data collection
Dietary assessment
At baseline, diet was assessed using a staff-assisted self-administered 127-item FFQ and dietary supplement form in Block format (Block et al. Reference Block, Woods, Potosky and Clifford1990). For each questionnaire item, participants were asked to report their frequency of consumption of various foods from among nine categories, ranging from rarely or never to two or more servings/d (six or more servings/d for beverages) and also their serving size as either small, medium or large. Servings per d were calculated from these categories. The FFQ was patterned after the FFQ used in the Insulin Resistance Atherosclerosis Study, which has been validated in non-Hispanic white, African-American and Hispanic persons (Mayer-Davis et al. Reference Mayer-Davis, Vitolins and Carmichael1999). Concerning validity, the mean correlation coefficients between nutrient intake estimated from the FFQ and intake from the average of eight 24-h recalls were 0·62 for non-Hispanic whites, 0·50 for African Americans and 0·41 for Hispanics. For total carbohydrates the correlation coefficient was 0·39. Among non-Hispanic whites, however, carbohydrate intake estimated from the FFQ tended to be lower than carbohydrate intake estimated from the dietary recalls. Concerning reproducibility, the mean correlation coefficient for nutrients across two administrations of the FFQ was 0·62 and did not differ by ethnic subgroup. In order to accommodate the MESA subject population, the Insulin Resistance Atherosclerosis Study FFQ was modified to include Chinese foods and culinary practices.
Whole grain intake
Servings per d of the following foods were summed to calculate total whole grain intake: whole grain breakfast cereal; oatmeal; dark bread; bran muffins; brown or wild rice. Further descriptions of whole grain food items, including verbatim FFQ wordings, are provided in Table 1.
Table 1 Descriptions of mean intake of whole grain food groups among Multi-Ethnic Study of Atherosclerosis participants†
* Cold cereals were classified as either whole grain or refined grain. The classification criteria are described in Methods.
† For details of subjects and procedures, see Methods.
If participants reported eating cold cereal, they were asked to name the breakfast cereal that they usually ate. Breakfast cereals mentioned were then evaluated for dietary fibre and whole grain content as determined by package labels, dietary databases, such as the Nutrition Data System and the US Department of Agriculture Food Composition Data, or by records shared in 1996 by General Mills, Inc (Minneapolis, MN, USA). Of the 144 breakfast cereals mentioned, 121 were classified as whole grain cereals (most mentioned by very few participants) as they contained ≥ 3 g dietary fibre per 100 g dry weight. Given that 12 g dietary fibre corresponds to 100 g whole wheat, all wheat cereals classified as whole grain contained at least 25 % of a serving of whole grain/100 g. Of breakfast cereal consumers, 54·5 % consumed whole grain varieties, while 45·6 % consumed refined grain varieties. No nutrient available in the MESA database was closely correlated with whole grain food intake; the correlation with dietary fibre was 0·36 (P < 0·0001), while with total carbohydrates it was 0·29 (P < 0·0001).
BMI, serum insulin, newly diagnosed diabetes and impaired fasting glucose
BMI was calculated as weight over height squared (kg/m2). Participants were asked to fast for at least 8 h. Serum insulin was measured by the Linco Human Insulin Specific RIA Kit (Linco Research, Inc., St. Charles, MO, USA), and serum glucose by rate reflectance spectrophotometry using thin film adaptation of the glucose oxidase method on the Vitros analyzer (Johnson & Johnson Clinical Diagnostics, Inc., Rochester, NY, USA) at the Collaborative Studies Clinical Laboratory at Fairview University Medical Center (Minneapolis, MN, USA). Non-medicated participants with fasting glucose ≥ 7·0 mmol/l (126 mg/dl) who did not self-report pre-existing diabetes were classified as newly diagnosed diabetics and those with fasting glucose levels between 5·6 mmol/l (100 mg/dl) and 6·9 mmol/l (125 mg/dl) were classified as having impaired fasting glucose. The homeostasis model assessment (Matthews et al. Reference Matthews, Hosker, Rudenski, Naylor, Treacher and Turner1985) estimate of insulin resistance was calculated as insulin*glucose/22·5 (mU/l*mmol/l).
C-reactive protein, IL-6 and homocysteine
CRP was measured using the BNII nephelometer (N High Sensitivity CRP; Dade Behring Inc., Deerfield, IL, USA) and IL-6 by an ultra-sensitive ELISA (Quantikine HS Human IL-6 Immunoassay; R&D Systems, Minneapolis, MN, USA), both at the Laboratory for Clinical Biochemistry Research (University of Vermont, Burlington, VT, USA). Plasma total homocysteine was measured by a fluorescence polarization immunoassay (IMx Homocysteine Assay; Axis Biochemicals ASA, Oslo, Norway) using the IMx Analyzer (Abbott Diagnostics, Abbott Park, IL, USA) at the Biochemical Genetics Clinical Laboratory at Fairview University Medical Center (Minneapolis, MN, USA).
Urine albumin excretion
Urine albumin and creatinine concentrations were assayed in a single untimed urine sample at the Fletcher Allen Health Care Clinical Chemistry Laboratory (Burlington, VT, USA) Urine albumin was measured by the Array 360 CE Protein Analyzer (Beckman Instruments, Inc., Drea, CA, USA). Serum creatinine was measured by rate reflectance spectrophotometry using thin film adaptation of the creatine amidinohydrolase method on the Vitros analyzer (Johnson & Johnson Clinical Diagnostics, Inc.). To estimate albumin excretion rate, sex-standardized A/kC (where albumin is expressed as μg/ml and creatinine is expressed as mg/ml) were calculated after multiplying men's urine creatinine concentrations by k 17/25, based on the higher rate of creatinine excretion typical of men compared with women (Warram et al. Reference Warram, Gearin, Laffel and Krolewski1996; Jacobs et al. Reference Jacobs, Murtaugh, Steffes, Yu, Roseman and Goetz2002). The sex-standardized A/kC is represented both linearly and dichotomously, with participants having values ≥ 25 and < 250 defined as having microalbuminuria (Jacobs et al. Reference Jacobs, Murtaugh, Steffes, Yu, Roseman and Goetz2002). Participants with macroalbuminuria (A/kC ≥ 250 mg/g) (n 127) were excluded in analyses of the urine albumin data.
Carotid artery intima-media thickness
Images of bilateral common carotid and internal carotid arteries were obtained via high-resolution B-mode ultrasonography using a Logiq 700 ultrasound machine (GE Medical Systems, Waukesha, WI, USA). Images of the near and far walls were obtained, on the basis of a previous study (O'Leary et al. Reference O'Leary, Polak, Kronmal, Manolio, Burke and Wolfson1999). Central reading of the intima-media thickness was done at Tufts-New England Medical Center (Boston, MA, USA) (Espeland et al. Reference Espeland, Evans and Wagenknecht2003); maximal intima-media thickness at any site was used in analysis. Additionally, a dichotomous variable indicated the presence of atherosclerotic plaque (any stenosis in either the right or left carotid artery).
Coronary artery calcification
Computed tomography of the coronary arteries was performed, as has been previously described (Carr et al. Reference Carr, Nelson and Wong2005), with electron beam scanners (Imatron C-150; Imatron, Inc., San Francisco, CA, USA) cardiac-gated at 80 % of the R-R interval at three centres and with a prospective electrocardiogram-triggered scan acquisition at 50 % of the R-R interval with multidetector scanners at the remaining three centres. The scanners are comparable in their ability to measure Ca (Carr et al. Reference Carr, Crouse, Goff, D'Agostino, Peterson and Burke2000). Scans were read centrally at Harbor University of California Medical Center (Los Angeles, CA, USA) and Agatston coronary artery Ca scores were quantified by blinded computer tomography (CT) image analysts. Participants with CAC scores >0 were considered to have CAC in the dichotomous variable representation.
Additional variables
Sex, race, age, educational level ( < high school, high school, some college, bachelor's degree, graduate or professional degree), current cigarette smoking (Yes/No), current alcohol use (Yes/No) and current hormone replacement therapy use (Yes/No) were self-reported. Physical activity was assessed using a detailed, semi-quantitative questionnaire adapted from the Cross-Cultural Activity Participation Study (B. Ainsworth, personal communication, San Diego State University). Leisure physical activity was computed as the sum of metabolic equivalent min/week of walking, conditioning, sports and dance, while a sedentariness score was the sum metabolic equivalent min/week of sitting or reclining, reading, knitting, sewing, etc, driving a car or watching television; metabolic equivalent activity intensity codes were based on a published table (Ainsworth et al. Reference Ainsworth, Haskell and Whitt2000). Neither separating former and never smokers nor treating alcohol as a continuous variable altered findings noticeably (data not shown).
HDL-cholesterol was measured in EDTA plasma using the cholesterol oxidase method (Roche Diagnostics, Indianapolis, IN, USA) after precipitation of non-HDL-cholesterol with Mg/dextran, and LDL-cholesterol was calculated in plasma specimens having a TAG value < 400 mg/dl using the Friedewald formula, at the Collaborative Studies Clinical Laboratory at Fairview University Medical Center. Resting blood pressure was measured three times in the seated position using a Dinamap model Pro 100 automated oscillometric sphygmomanometer (Critikon, Tampa, FL, USA). The average of the last two measurements was used in analyses.
Statistical analysis
SAS was used for all analyses (version 9.1; SAS Institute, Inc., Cary, NC, USA). Mean levels of demographics, behaviours and physiological variables were provided by quintile of whole grain intake. Regression analyses were used to evaluate the association of each variable with whole grain intake, providing a P value for trend over the continuous whole grain variable. Linear regression was used for continuous dependent variables (PROC GLM). Logistic regression (PROC GENMOD) was used for dichotomous dependent variables and provided the P for trend. However, as logistic regression is a nonlinear procedure and therefore gives biased estimates of probabilities, which are on the arithmetic scale, linear regression was used to compute the adjusted percentages within whole grain intake quintiles for dichotomous dependent variables. The natural logarithm transformation was utilized because of skewness in serum insulin, CRP, A/kC, the common carotid intimal-medial thickness, the internal carotid intimal-medial thickness and the Agatston score. Geometric means of these variables were reported. To account for Agatston scores of zero, one was added to all Agatston score values prior to transformation, then subtracted after exponentiation when estimating the geometric means.
Three models were developed to evaluate relationships with whole grain intake. Model 1 was adjusted for age, sex, race, education, survey centre and energy intake (base adjustment). Model 2, our primary model of interest, further adjusted for behavioural factors including current smoking (yes if one or more cigarettes/week or no), current alcohol use (Yes/No) and dietary intake of the following food groups: fruit; vegetables; refined grains; dairy; fish and poultry; meat. Model 3 (mechanistic model) was adjusted for model 2 factors as well as for BMI and serum insulin, two variables thought to be in the causal pathway between whole grain intake and CVD; these models were intended to assess whether observed relationships with whole grain intake were mediated by BMI or insulin. Furthermore, race/ethnicity interaction with whole grain food intake was assessed in each model for each dependent variable by adding the product of the continuous whole grain variable with the categorical race/ethnicity variable. Race/ethnicity interaction was insubstantial except for the dependent variable albumin excretion rate.
Results
The mean daily intake of whole-grain foods was 0·54 servings/d (Table 1), while the median ranged from 0·02 servings/d for the lowest quintile of whole grain intake to 1·39 servings/d for the highest (Table 2). Whole grain intake varied by race, with whites having the highest mean intake (0·60 servings/d), followed by blacks (0·53 servings/d), Hispanics (0·52 servings/d) and Chinese (0·32 servings/d).
Table 2 Means and percentages* of demographics, behaviours, diet, blood lipids and blood pressure by category of whole grain intake in 5496 participants, MESA 2000–2002‖
* All means and percentages were adjusted for sex, age, race, education, site and energy intake; except for age, sex, education and energy intake, each of which is adjusted for the relevant combination of these variables only; and race, for which crude values are presented.
† √(mean squared error/mean n per quintile) is included to facilitate statistical comparison of pairs of whole grain quintiles.
‡ Based on χ2 statistic with 12 df.
§ Based on χ2 statistic with 16 df.
‖ For details of subjects and procedures, see Methods.
MESA, Multi-Ethnic Study of Atherosclerosis; HRT, hormone replacement therapy; MET, metabolic equivalent.
After adjustment for age, sex, race, education, survey centre and energy intake, higher whole grain intake was strongly associated with race and being older, female, more educated, a non-smoker, more leisure physical activity, a lower sedentariness score and with consuming more energy, fruits, vegetables and dairy and less refined grains, meat and alcohol. Whole grain intake was not related to hormone replacement therapy, HDL-cholesterol, LDL-cholesterol, systolic blood pressure or diastolic blood pressure.
Inverse associations were found between whole grain intake and BMI, insulin, homeostasis model assessment insulin resistance, CRP and homocysteine (Table 3). In the case of CRP, P for trend was not significant after possible mechanistic adjustment for BMI and insulin; however, the estimated mean CRP was lower in whole grain quintile 5 than 1 (P = 0·009). IL-6 was inversely associated with whole grain consumption in model 1; however, this was attenuated with further adjustments. Whole grain intake was inversely related to glucose and to impaired fasting glucose or newly diagnosed diabetes (glucose ≥ 100 mg/dl), but showed little relation to newly diagnosed diabetes when analysed separately.
Table 3 Means and percentages of body mass index, insulin resistance, inflammation, diabetes and subclinical CVD by category of whole grain intake in 5496 participants, MESA 2000–2002§
Model 1 (base model), age, sex, race, education, survey centre and energy intake.
Model 2 (behavioural model), model 1 plus current smoking, current alcohol use and dietary intake of fruit, vegetables, refined grains, dairy, fish and poultry, meat, leisure physical activity, and sedentariness score.
Model 3 (mechanistic model), model 2 plus BMI and insulin.
* Calculated as root MSE/(mean n per quintile) is included to facilitate statistical comparison of pairs of whole grain quintiles.
† Geometric mean.
‡ Zero values were included.
§ For details of subjects and procedures, see Methods.
MESA, Multi-Ethnic Study of Atherosclerosis; HOMA, homeostasis model assessment; CRP, C-reactive protein; IFG, impaired fasting glucose; A/kC, sex-standardized albumin:creatinine ratio where k = 1 for women and 0·68 for men; IMT, intima-media thickness.
Urine albumin excretion was inversely associated with whole grain intake after adjustment for age, sex, race, education, survey centre and energy intake. These relationships were attenuated with additional adjustments. The proportion of participants with microalbuminuria paralleled trends observed in urine albumin excretion levels. Associations of urine albumin excretion rate and whole grain intake (adjusted as in model 3) varied with race/ethnicity (P for interaction 0·03). The A/kC was 12 % and 19 % lower per whole grain food serving per d among Hispanics (P = 0·03) and Chinese (P = 0·02), respectively. These associations were null in whites and blacks.
Whole grain intake was inversely associated with probability of having any CAC in the base model, but this association was attenuated with further adjustment. Whole grain intake was not associated with carotid artery intima-media thickness or presence of plaque. These associations are presented in light of relatively low correlations among the different subclinical markers, which may suggest that each assesses a different aspect of subclinical CVD. The correlation between ln(Agatston score+1) and ln(A/kC) was r 0·19; ln(common carotid artery intima-media thickness) and ln(A/kC) was r 0·19; and between ln(common carotid artery intima-media thickness) and ln(Agatston score+1) was r 0·32.
Discussion
In this multi-ethnic sample of 5496 men and women, mean whole grain consumption of about 0·5 servings per d was slightly less than that estimated for the entire US population (Cleveland et al. Reference Cleveland, Moshfegh, Albertson and Goldman2000), and is well below the recommended intake (US Department of Health & Human Services & the US Department of Agriculture, 2005) of three or more servings of whole grain foods per d. In fact, less than 1 % of participants met the official recommendation of three or more servings per d. About 20 % of white, black and Hispanic participants reported eating one or more servings per d, but less than 10 % of Chinese participants ate whole grain foods that often. As in previous papers (Jacobs et al. Reference Jacobs, Meyer, Kushi and Folsom1998; Steffen et al. Reference Steffen, Jacobs, Stevens, Shahar, Carithers and Folsom2003b), we found that whole grain food intake was a good indicator of other healthful behaviours, with higher consumption associated with being a non-smoker, drinking less alcohol, more leisure physical activity, less sedentary behaviour and greater consumption of fruit, vegetables and dairy and less of meat and refined grains.
These findings in the MESA database are consistent with several studies that have observed more favourable values of BMI, insulin and diabetes among whole grain eaters (Lovejoy & DiGirolamo, Reference Lovejoy and DiGirolamo1992; Feskens et al. Reference Feskens, Loeber and Kromhout1994; Marshall et al. 1997; Ludwig et al. Reference Ludwig, Pereira and Kroenke1999; Liu et al. Reference Liu, Manson and Stampfer2000, Reference Liu, Willett, Manson, Hu, Rosner and Colditz2003; Meyer et al. Reference Meyer, Kushi, Jacobs, Slavin, Sellers and Folsom2000; Fung et al. Reference Fung, Hu and Pereira2002; Pereira et al. Reference Pereira, Jacobs and Pins2002; Liese et al. Reference Liese, Roach, Sparks, Marquart, D'Agostino and Mayer-Davis2003; Montonen et al. Reference Montonen, Knekt, Jarvinen, Aromaa and Reunanen2003; Steffen et al. Reference Steffen, Jacobs and Murtaugh2003a). Graded inverse relationships were observed between whole grain intake and BMI, serum insulin, homeostasis model assessment insulin resistance, glucose, and newly diagnosed impaired fasting glucose or diabetes.
As observed in other studies (Jensen et al. Reference Jensen, Koh-Banerjee, Franz, Sampson, Gronbaek and Rimm2006; Lutsey et al. Reference Lutsey, Steffen and Feldman2006), there was a strong inverse association between homocysteine and whole grain intake. This is as expected, since whole grains are a rich source of folate, which is inversely related to homocysteine (Wardlaw & Kessel, Reference Wardlaw and Kessel2002). Several food-based feeding trials have also shown reductions in homocysteine resulting from increased consumption of whole grains (Jang et al. Reference Jang, Lee, Kim, Park and Lee2001), fortified cereals (Malinow et al. Reference Malinow, Duell and Hess1998; Riddell et al. Reference Riddell, Chisholm, Williams and Mann2000) and cereals prior to folic acid fortification of refined grain food starting in 1998 (Tucker et al. Reference Tucker, Bermudez and Castaneda2000).
In recent literature, inverse associations were observed between whole grain intake and inflammatory markers in a subset of men from the Health Professionals Follow-Up Study and women from the Nurses' Health Study II (Jensen et al. Reference Jensen, Koh-Banerjee, Franz, Sampson, Gronbaek and Rimm2006); however, all became non-significant after accounting for lifestyle factors. In our analysis, CRP remained inversely associated with whole grain intake after adjustment for other behavioural characteristics, although the relationship was partially explained by adjustment for BMI and insulin, two factors that are believed to be in the causal pathway. IL-6 was inversely associated with whole grain intake after minimal adjustment; however, significance was not retained with further adjustment. No association between whole grain intake and IL-6 was observed in a previous study (Jensen et al. Reference Jensen, Koh-Banerjee, Franz, Sampson, Gronbaek and Rimm2006). Another recent study among female nurses with type 2 diabetes observed inverse associations (Qi et al. Reference Qi, van Dam, Liu, Franz, Mantzoros and Hu2006) of whole grain intake and both CRP and TNF receptor 2. Dietary fibre has also been inversely associated with serum CRP concentrations (King et al. Reference King, Egan and Geesey2003; Ajani et al. Reference Ajani, Ford and Mokdad2004).
As hypothesized, urine albumin excretion and microalbuminuria prevalence had inverse associations with whole grain intake in the base model. However, these associations became non-significant after additional adjustments. Urine albumin excretion was the only variable studied in which the present data suggested a race/ethnicity interaction; even in the fully adjusted models, Chinese and Hispanic participants showed an inverse association between whole grain intake and urine albumin excretion rate, whereas whites and blacks showed no relationship even in the base model. However, given the large number of variables assessed, this heterogeneity by ethnic group may have been a chance finding. To our knowledge no previous studies have examined the relationship between whole grain intake and urine albumin excretion or microalbuminuria. Total dietary fibre has, however, been evaluated, with one study showing reduced albuminuria among those consuming >26 g/d (Metcalf et al. Reference Metcalf, Baker, Scragg, Dryson, Scott and Wild1993) and the other showing no effect (Watts et al. Reference Watts, Gregory, Naoumova, Kubal and Shaw1988).
In contrast and contrary to our initial hypotheses, however, whole grain intake showed little cross-sectional relationship with subclinical markers of vascular disease. There was limited evidence that whole grain intake was inversely associated with CAC, but no evidence that whole grain intake was predictive of carotid artery intima-media thickness. Associations of whole grain intake with these markers have not previously been reported. Numerous prospective studies have found graded and continuous relationships between subclinical disease and CVD outcomes (Levy et al. Reference Levy, Garrison, Savage, Kannel and Castelli1989; Chambless et al. Reference Chambless, Heiss and Folsom1997; O'Leary et al. Reference O'Leary, Polak, Kronmal, Manolio, Burke and Wolfson1999; Raggi et al. Reference Raggi, Callister and Cooil2000), yet they occur early in the atherosclerotic process and do not reflect impending clinical disease. The subclinical disease markers used here are relatively weakly correlated and therefore heterogeneous. Thus, another possible explanation is that although whole grain may have an influence on atherosclerosis, measurement of CAC and intima-media thickness may be too loosely connected to the whole body burden of atherosclerosis to detect this influence. Furthermore, it is possible that whole grain intake reduces risk of CHD in ways other than through direct reduction of atherosclerosis. The failure to find associations between whole grain intake and these subclinical markers may reflect the cross-sectional design. A cause–effect relationship cannot be inferred from these data and reverse causality may be an issue. For example, despite participants being free of clinical CVD at baseline, it is possible that participants at greater risk for CVD may have begun taking behavioural precautions to reduce their risk of having a CVD event, such as increasing their whole grain intake. Given the lack of follow-up, whether whole grain intake is associated with progression of subclinical disease or incidence of clinical events was not studied. Ongoing MESA follow-up will help to overcome this limitation.
Another limitation of the study is that the dietary measure may have limited accuracy as it was based on a single FFQ. Further, whole grain consumption was low and there was little variation. It is possible that there could be a threshold effect in which the impact of whole grains on subclinical markers is only evident at higher levels of consumption than reported in this study. Error in the measurement of potential confounders or failure to measure and adjust for potential confounders could have resulted in residual confounding. Ruling out the possibility of residual confounding is particularly difficult in this analysis, as whole grain consumers tend to report healthier lifestyle habits than non-consumers (Jacobs et al. Reference Jacobs, Meyer, Kushi and Folsom1998; Steffen et al. Reference Steffen, Jacobs, Stevens, Shahar, Carithers and Folsom2003b).
Strengths of this study are that MESA collected an extensive set of subclinical CVD measures in a large sample of participants using standardized procedures to increase measurement validity and that whole grain intake (including specific cereal brand) was reported by participants using a FFQ, which accounted for both frequency of consumption and serving size. Additionally, in light of recent discussions concerning both the benefits of whole food approaches (Jacobs & Steffen, Reference Jacobs and Steffen2003) and possible limitations of single nutrient approaches in assessing relationships between diet and complex disease (Lichtenstein & Russell, Reference Lichtenstein and Russell2005), the fact that we assessed diet in terms of whole grain intake may be considered a strength of the present study.
In summary, in this multi-ethnic sample of 5496 men and women, we found ethnic differences in whole grain intake, but few ethnic differences in the associations of whole grain foods with several dependent variables. There were strong cross-sectional associations between whole grain consumption and healthful behaviour, BMI, insulin, homocysteine, CRP and fasting glucose and possible associations with measures of urine albumin excretion rate, but no associations with measures of carotid artery intima-medial thickness or CAC.
Acknowledgements
This research was supported by contracts N01-HC-95 159 through N01-HC-95 165 and N01-HC-95 169 from the National Heart, Lung, and Blood Institute. The authors thank the other investigators, the staff, and the participants of the MESA study for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org
