点击显示 收起
1 From the Department of Nutrition, Harvard School of Public Health, Boston, MA (MKJ, MF, LS, and EBR); the Department of Clinical Epidemiology, Aalborg Hospital, Aarhus University Hospital, Aalborg, Denmark (MKJ); the Department of Preventive Medicine, The University of Tennessee Health Science Center, Memphis, TN (PK-B); the Centre for Alcohol Research, National Institute of Public Health, Copenhagen, Denmark (MG); the Department of Epidemiology, Harvard School of Public Health, Boston, MA (EBR); and the Channing Laboratory, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, MA (EBR)
2 Supported by research grants HL35464 and CA55075 from the National Institutes of Health (Bethesda, MD) and a scholarship from the Danish Research Foundation (to MKJ). The Kellogg Company provided unrestricted funding for the development of the whole-grain database. The funding organizations had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, and approval of the manuscript.
3 Address reprint requests to EB Rimm, Department of Nutrition, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115. E-mail: erimm{at}hsph.harvard.edu.
See corresponding CME exam on page 392.
ABSTRACT
Background: Intake of whole grains is inversely associated with risk of diabetes and ischemic heart disease in observational studies. The lower risk associated with high whole-grain intakes may be mediated through improvements in glycemic control, lipid profiles, or reduced inflammation.
Objective: The aim was to examine whether the intake of whole grains, bran, and germ is related to homocysteine, plasma markers of glycemic control (fasting insulin, hemoglobin A1c, C-peptide, and leptin), lipids (total cholesterol, triacylglycerol, HDL cholesterol, and LDL cholesterol), and inflammation (C-reactive protein, fibrinogen, and interleukin 6).
Design: This was a cross-sectional study of the relations of whole grains, bran, and germ intakes with homocysteine and markers of glycemic control, lipids, and inflammation in 938 healthy men and women.
Results: Whole-grain intake was inversely associated with homocysteine and markers of glycemic control. Compared with participants in the bottom quintile of whole-grain intake, participants in the highest quintile had 17%, 14%, 14%, and 11% lower concentrations of homocysteine (P < 0.01), insulin (P = 0.12), C-peptide (P = 0.03), and leptin (P = 0.03), respectively. Inverse associations were also observed with total cholesterol (P = 0.02), HDL cholesterol (P = 0.05), and LDL cholesterol (P = 0.10). Whole-grain intake was not associated with the markers of inflammation. Whole-grain intake was most strongly inversely associated with markers of glycemic control in this population.
Conclusion: The results suggest a lower risk of diabetes and heart disease in persons who consume diets high in whole grains.
Key Words: Whole grains • bran • germ • homocysteine • glycemic control • lipids • inflammation
INTRODUCTION
Consumption of grain products with a high content of whole-grain flour, milled from all edible components of grains, has been inversely associated with mortality from and incidence of diabetes and ischemic heart disease (IHD) in several prospective population studies (1–4). Conversely, intake of refined flour, which consists mainly of the starchy endosperm, was not associated with diabetes and IHD risk in these studies (2, 5, 6).
It has been hypothesized that the observed health benefits of whole-grain intake may be attributable to the synergistic effects of dietary fiber and micronutrients found in whole-grain foods (7). The bran and germ components of whole grains are rich in fiber, vitamins, minerals, and phytoestrogens (8). In a recent analysis within the Health Professionals Follow-Up Study (HPFS), we found a strong inverse association between bran intake and IHD risk but no association between IHD risk and the germ component. However, the intake of germ was very low in this population and may not have substantial variation (4).
The observed lower risk of diabetes and IHD among participants with high whole-grain and bran intakes may be mediated through effects on glycemic control, plasma lipids, or inflammation. Results from cross-sectional studies suggest that consumption of foods with a high content of whole grains is associated with improved insulin sensitivity and lower concentrations of serum triacylglycerol and total and LDL cholesterol (9–13). To our knowledge, no data have been presented on the association between whole-grain intake and inflammation.
Although these studies support associations between intake of whole grains and glycemic control and lipid profiles, the commonly used method for classifying whole-grain foods in these past studies has limitations. Traditionally, foods with a whole-grain or bran content >25% by weight are classified as whole-grain items, although the amount of whole grains in each serving can vary considerably (3). This may result in a large degree of misclassification. Furthermore, this classification does not allow a separate analysis of the bran or germ content of whole grains, for which the nutrient composition and health effects may differ.
In the present study, we examined the associations between daily intakes of whole grains, bran, and germ and plasma concentrations of homocysteine, markers of glycemic control [fasting plasma insulin, hemoglobin A1c (Hb A1c), C-peptide, and leptin], lipid status (total cholesterol, triacylglycerol, HDL cholesterol, and LDL cholesterol), and inflammation [C-reactive protein (CRP), fibrinogen, and interleukin 6 (IL-6)] within healthy subsamples of the HPFS and the Nurses' Health Study II (NHS II).
SUBJECTS AND METHODS
The Health Professionals Follow-Up Study
The HPFS was initiated in 1986 when 51 529 male dentists, veterinarians, pharmacists, optometrists, osteopathic physicians, and podiatrists between 40 and 75 y of age completed a detailed self-administered food-frequency questionnaire (FFQ) and medical-history questionnaire. The participants have been followed with repeated questionnaires on lifestyle and health every 2 y and detailed FFQs every 4 y. Between 1993 and 1995, a blood sample was requested from all subjects and returned by 18 225 participants. Participants underwent local phlebotomy, and whole-blood samples were sent to our laboratory on ice via overnight courier; the vast majority of samples were returned within 24 h of being drawn. The men who provided samples were somewhat younger but otherwise similar to those who did not provide samples (14). Men with a history of myocardial infarction, angina, stroke, diabetes, intermittent claudication, gastric or duodenal ulcers, gallbladder disease, liver disease, or cancer (except nonmelanoma skin cancer) before 1994 were excluded. Of the remaining healthy men who had provided blood, 468 participants were randomly selected within strata of different patterns of self-reported alcohol consumption (determined by frequency, amount, and time of alcohol consumption, eg, with meals). The original purpose for the selection of this subset was to investigate the association between alcohol drinking patterns and novel biomarkers of IHD as previously reported (15, 16). It was previously shown that, besides the exclusion criteria, the characteristics of the 468 men evaluated in the present analysis did not differ from those of the remaining 17 757 who had provided blood samples (14).
The Nurses' Health Study II
The NHS II is a cohort of 116 671 female registered nurses aged 25–42 y at baseline in 1989. Health and disease status have been assessed with methods similar to those described above, with diet assessed every 4 y starting in 1991. Blood samples, obtained between 1996 and 1998, were collected from > 29 000 premenopausal women, whose blood was collected during the luteal phase of their menstrual cycle and who were not taking any hormones. Women who gave blood were similar to those in the overall cohort; the only difference was a greater percentage of women with a family history of breast cancer in the subgroup (17). Similar to the HPFS substudy, women with any of the described preexisting conditions at the time blood was drawn were excluded, and a subset of 473 women was randomly selected from the remaining participants on the basis of drinking patterns. For the present analyses, 3 women were excluded because of missing information on whole-grain intake.
Measurement of the biomarkers
The whole-blood samples were centrifuged at 2500 x g for 30 min at room temperature and portioned into cryotubes as plasma, buffy coat, and red blood cells and then stored in the vapor phase of liquid nitrogen freezers at –130 °C or colder. The measurement of the different biomarkers in the samples was previously described. The markers generally showed good stability and reproducibility during simulated transport and storage (18).
Candidate biomarkers assessed in both studies included homocysteine, Hb A1c, C-peptide, leptin, total cholesterol, triacylglycerol, HDL cholesterol, LDL cholesterol, CRP, fibrinogen, and IL-6. Of the 468 men, a total of 199 had not been fasting for >8 h at the time blood was drawn. These nonfasting men were excluded in all analyses of biomarkers considered to be influenced by fasting status (homocysteine, triacylglycerol, insulin, and C-peptide).
Dietary assessment of whole-grain, bran, and germ intakes
In both studies, dietary information was collected with a validated semiquantitative, 131-item FFQ, which was described in detail elsewhere (19, 20). Briefly, the questionnaire was designed to assess average food intake over the previous year and included questions regarding the consumption of grain foods such as cooked and cold breakfast cereals, white and dark bread, white and brown rice, and pasta. For each food, a commonly used unit or portion size was specified along with 9 possible response categories for frequency of intake ranging from "never" to "6 times/d." Besides these questions, the participants were asked to specify the brand and type of cold breakfast cereal usually eaten. Finally, the FFQ included open-ended questions regarding the usual serving size and frequency of consumption of foods not listed on the FFQ. For all foods, the portions were converted to grams per serving, and nutrient intakes were computed by multiplying the frequency of consumption of each unit of food by the nutrient content in grams.
Details on the estimation of whole-grain intake in grams per day in these cohorts was published previously (21). Briefly, daily intakes of whole grains, bran, and germ were calculated by determining the whole-grain content of all grain foods (rice, bread, pasta, and breakfast cereals) according to the dry weight of whole-grain ingredients. For the question on "dark bread," we used a composite of recipes to represent commercial wheat (not whole wheat) and wheat berry breads selected on the basis of marketplace shelf spacing of breads. The nutrient profile of 4 commercial wheat breads was derived from US Department of Agriculture nutrient data and product labels. Cookbooks were used to create recipes for home-prepared bakery items. Whole grains were considered in their intact and pulverized forms, where, by definition, they must contain the expected proportion of bran, endosperm, and germ for the specific grain type. The naturally occurring bran or germ corresponded to the amount that would typically be found in the type of grain in the food. We further estimated the amount of bran or germ that was added to foods either during processing or by the participants while cooking. Total bran and germ consumption was estimated by summing the naturally occurring and the added components.
Whole-grain intake (g/d) was estimated from all grain foods, both with the exclusion and inclusion of added bran and germ. Because previously performed analyses of whole grains, bran, and germ in relation to IHD risk in the entire HPFS cohort suggested that the added versus naturally occurring bran and germ components did not differ on a gram-to-gram basis (4), we used the measure of whole-grain intake including added bran and germ in the analyses for the present study.
Statistical analysis
Dietary factors, except fruit and vegetable intakes, were reported in grams per day, adjusted for total energy intake by the residual method (22), and categorized into quintiles. Alcohol intake was not adjusted for energy and categorized as follows: 0, 0.1–4.9, 5–14.9, 15–29.9, and 30 g/d. To reduce within-person variation and best represent long-term dietary intake, we calculated cumulative averages of food and nutrient intakes from the first 3 administrations of the FFQ. The nondietary covariates [age (5-y intervals), sex, smoking (never, ever, and current), BMI (in kg/m2: <20, 20–24.9, 25–29.9, 30–34.9, and 35), physical activity (quintiles of metabolic equivalent hours/wk), and hypercholesterolemia at baseline (yes or no)] were collected biennially in both cohorts, and in the present analysis the most recent measure was used.
We used multivariate linear regression with robust variance (PROC MIXED, version 8.2; SAS Institute, Cary, NC) to ensure validity without the need for normal distribution assumptions (23). However, leptin, CRP, and IL-6 concentrations were transformed by the natural logarithm because their distributions were heavily skewed. Mean concentrations of the biomarkers were estimated in quintiles of whole-grain intake for an average person (least-squares means). The means of the log-transformed biomarkers are presented as geometric means in all tables. Relative differences in mean biomarker concentrations between whole-grain quintiles 1 and 5 were calculated. Tests for linear trends were performed by treating the median intake of whole grains across quintiles as a continuous variable.
Primarily, analyses were conducted in the 2 separate study populations and only combined when the study-specific associations were similar. All dietary-adjusted models included adjustment for total energy, alcohol, fruit, and vegetable intakes (servings/d), but the choice of other dietary confounders was different between the 3 groups of biomarkers. We also considered protein, sucrose, fructose, and saturated, trans, monounsaturated, and polyunsaturated fat intakes and whether or not the participants reported taking multivitamin supplements. Intakes of folate, choline, betaine, riboflavin, vitamin B-6, niacin, and vitamin E from food and supplements were also considered as predictors in the analyses of homocysteine concentrations. Final models of markers related to glycemic control included adjustment for dietary glucose and fructose, whereas models of lipids, homocysteine, and inflammatory-related markers included adjustment for saturated, trans, monounsaturated, and polyunsaturated fat intakes. Protein and use of vitamin supplements did not have an appreciable effect on the associations and were not included in these models.
All analyses were repeated for intakes of total bran and germ, which were mutually adjusted in all models. Potential interactions between whole-grain intake and overweight (BMI > 25) or hypercholesterolemia were addressed by stratified analyses. We conducted tests for statistical interaction by treating whole-grain intake as a continuous variable and testing the significance of the added product term between whole grains and the stratification variable.
RESULTS
The median intake of whole grains was 22.3 g/d. In men, whole-grain intake ranged from 0.4 to 124 g/d (median: 23.4 g/d) and in women it ranged from 3.1 to 77 g/d (median: 21.9 g/d). In the combined study population, the median intake of bran was 5.0 g/d (range: 0–74 g/d) and the median intake of germ was 0.93 g/d (range: 0–23 g/d) (data not shown). The Pearson's correlation coefficient (r) between bran and germ intakes was 0.32 (P < 0.0001).
As shown in Table 1, a higher intake of whole grains was associated with higher levels of physical activity, greater fruit and vegetable intakes, less smoking, and lower intakes of alcohol and saturated and monounsaturated fats. Among men and women with the highest intake of whole grains, carbohydrate intake contributed a higher percentage of total calories. High intakes of bran and germ were also related to overall healthier participant characteristics (data not shown).
View this table:
TABLE 1. Lifestyle and dietary characteristics according to quintile (Q) of whole-grain intake in US men in the Health Professionals Follow-Up Study (HPFS) and women in the Nurses' Health Study II (NHS II)1
The estimated relative difference in mean values of the biomarkers by quintiles of whole-grain intake was not different in the 2 substudies, with the exception of triacylglycerol (data not shown); thus, all other analyses were conducted in the combined study population adjusted for sex.
Among these healthy men and women, mean Hb A1c was stable across whole-grain intakes, whereas mean concentrations of insulin, C-peptide, and leptin were inversely related to intake of whole grains. (Table 2) Adjustment for dietary factors only slightly attenuated these associations. In a comparison of extreme quintiles, participants with the highest whole-grain intake had 11–14% lower concentrations of insulin (P for trend = 0.13), C-peptide (P for trend = 0.03) and leptin (P for trend = 0.03).
View this table:
TABLE 2. Estimated markers of glycemic control according to quintile (Q) of whole-grain intake in the total study population1
Whole-grain intake was inversely related to all assessed plasma lipids, although the association with LDL cholesterol was not statistically significant. (Table 3) Adjustment for lifestyle and dietary factors strengthened the inverse association between whole-grain intake and total and LDL cholesterol and attenuated the inverse relation between whole grains and HDL cholesterol. Compared with the bottom quintile of whole-grain intake, the highest quintile had 3%, 5%, and 2% lower mean values of total cholesterol, HDL cholesterol, and LDL cholesterol, respectively. Attempts to adjust for more refined carbohydrates by including sugars and total carbohydrates in the model of HDL cholesterol attenuated the estimated P for trend but the observed difference between extreme quintiles remained. The association between whole-grain intake and triacylglycerol concentrations differed between the 2 study populations (p interaction: 0.05). Among the young women, triacylglycerol concentrations were 22% higher among participants in the highest quintile of whole-grain intake, compared with lowest (P for trend = 0.02). Among the middle-aged men, the estimated difference between extreme quintiles was –36% (P for trend = 0.02) (data not shown).
View this table:
TABLE 3. Estimated lipid marker concentrations according to quintile (Q) of whole-grain intake in the total study population1
Modestly inverse, but not statistically significant, associations were observed between whole-grain intake and concentrations of CRP and IL-6, whereas a strong inverse association was observed with homocysteine, which remained highly significant even after multivariate adjustments. (Table 4) The mean concentration of homocysteine was 17% lower in the participants with the highest category of whole-grain consumption than in the participants with the lowest intake. Because this result may have been due to the folate fortification of all grain products after 1996, the analysis was repeated after adjustment for folate intake. The difference in mean homocysteine concentrations across extreme quintiles was modestly attenuated (mean concentrations of homocysteine across whole-grain quintiles: 13.0, 12.5, 11.9, 11.8, and 11.0 µmol/L); however, the trend remained highly significant (P for trend = 0.001). Additional adjustment for other B vitamins and vitamin E intake (found in whole-grain foods) did not alter the association (data not shown).
View this table:
TABLE 4. Estimated homocysteine concentrations and markers of inflammation according to quintile (Q) of whole-grain intake in the total study population1
The fully adjusted analyses (including lifestyle and dietary factors) of total bran and germ are shown in Table 5 and Table 6. Compared with participants with the lowest bran intake, participants with the highest bran intake had lower concentrations of C-peptide (13%), cholesterol (3%), HDL cholesterol (4%), CRP (22%), and homocysteine (11%). Except for the strong trend with homocysteine concentrations, these associations were all borderline significant at the 0.05 level of statistical significance (Table 5). Although the intake of germ only ranged from a mean concentration of 0.33 g/d in the lowest quintile to 2.23 g/d in the highest quintile, germ intake was related to 12–16% lower concentrations of insulin (P for trend = 0.10), leptin (P for trend = 0.03), and IL-6 (P for trend = 0.04) (Table 6). Overall, the associations between whole grains and the biomarkers did not differ substantially by BMI or by hypercholesterolemia. (data not shown).
View this table:
TABLE 5. Estimated biomarker concentrations according to quintile (Q) of bran intake in the total study population1
View this table:
TABLE 6. Estimated biomarker concentrations according to quintile (Q) of germ intake in the total study population1
DISCUSSION
In combined analyses of healthy men from the HPFS and women from the NHS II, we found intakes of whole grains, and of the bran and germ components, to be inversely related to general markers of diabetes and IHD. Except for a strong association with homocysteine, the associations were modest, but if causal they could have important implications for prevention of diabetes and IHD.
Although the HPFS and NHS II cannot be considered representative of the general American population, the biological mechanisms behind the observed associations are unlikely to differ across populations. Furthermore, the range in intake of whole grains in these cohorts is wide and may represent a more meaningful comparison for public health recommendations than would be possible in a similar size sample of the general American population.
The participants in the present study are likely to be relatively homogeneous with regard to education, work, and ability to purchase whole-grain foods. This may reduce potential confounding by characteristics that may otherwise be difficult to assess. Nevertheless, confounding by factors related to both whole-grain intake and the biomarkers remain an important concern. However, confounding beyond that of the included covariates is unlikely to explain the remaining associations. Both the relatively small study sample and random measurement error in the exposure assessments contribute to reduced statistical power in our study. Furthermore, only one measure of the biomarkers was available. Because the biomarkers may be susceptible to short-term variation, this random individual variation may also result in attenuation of the estimated associations with diet.
The results of the present analyses are in line with those of 4 recent cross-sectional studies of whole-grain foods and markers of insulin sensitivity and lipid status. In 2 reports from the Framingham Offspring Cohort Study, the number of daily servings of whole-grain foods was inversely related to fasting insulin concentrations and insulin resistance, whereas Hb A1c was unrelated to whole-grain intake (9, 10). In the Insulin Resistance Atherosclerosis Study, intake of dark bread and cereals was significantly related to improved insulin sensitivity (11), and a similar relation between whole-grain intake and insulin sensitivity was observed in a cross-sectional study of 285 adolescents from Minnesota (12). The associations were more pronounced inoverweight participants (10, 12). We found no important differences by underlying level of obesity in our study; however, we had limited power to test for this interaction.
In the Framingham Offspring Cohort Study, servings of whole-grain foods were related to lower LDL-cholesterol and total cholesterol concentrations and not to HDL-cholesterol or triacylglycerol concentrations (10). Intake of traditional Iranian whole-grain foods was associated with lower triacylglycerol concentrations in 827 men and women from the Tehran Lipid and Glucose Study but not with any other lipid variables (13). In these studies, intake of whole grain was estimated on the basis of servings of whole-grain foods and not quantitatively, which makes a comparison of past and current results difficult. Furthermore, grains consumed in the traditional Iranian diet are different from those consumed in the United States and may have a higher soluble fiber content, which may have a stronger effect on triacylglycerol concentrations. We found that whole-grain intake was associated with lower triacylglycerol concentrations in the HPFS, whereas triacylglycerol concentrations were highest among NHS II participants with the highest whole-grain intake. This discrepancy was not explained by the adjustment for potential confounders and may have been due to chance, unaccounted for study-specific differences, or to sex differences in triacylglycerol metabolism. Additionally, men and women may differ in their choice of whole-grain products, but we could not discern this from our questionnaire.
In our study, intake of whole grains was strongly inversely associated with homocysteine concentrations. Adjustment for folate intake along with other micronutrients found in whole grains that are involved in one-carbon metabolism did not appreciably alter this strong association. Although the present analyses only allowed us to speculate, we cannot exclude the possibility that synergistic interactions among the wide variety of micronutrients may have been involved.
Although other observational studies of whole grains have not included measures of inflammation, several studies have estimated measures of carbohydrate quality in relation to inflammatory markers. In analyses from the CARDIA study (24), the National Health and Nutrition Examination Surveys (25, 26), and the Women's Health Study (27), related measures such as dietary fiber and glycemic load were associated with lower inflammatory status.
Many experimental studies have been conducted to examine the metabolic effects of fiber-rich diets. In a series of short-term Finnish crossover trials, regular consumption of rye bread was favorably associated with a diminished insulin response to a controlled glucose feeding compared with refined-wheat bread (28–30). Many other trials with interventions such as diets with decreased glycemic load, supplementation with fiber from different sources (psyllium or bran), or substitution with different grain sources, have also been conducted (31–34). However, the interpretation of these studies with respect to our results is complicated by the varied ranges of exposure and outcome measures, the comparison food, the study populations (which are often hypercholesterolemic, hyperglycemic, or obese), and the duration of the trials.
The mechanisms by which whole grains contribute to health benefits remain to be elucidated. It is known that whole grains are a rich source of fiber, minerals (magnesium, potassium, phosphorous, selenium, manganese, zinc, and iron), vitamins (especially high in vitamins B and E), phenolic compounds, phytoestrogens (lignans), and related antioxidants (8). These compounds all may have important biological functions, which as a whole could make an important contribution to reductions in diabetes and IHD risk.
The high fiber content of whole-grain foods may lead to a slower or reduced digestion of macronutrients, giving rise to a reduced blood glucose burden and lower overall insulin concentrations. Recently, we reported that high intakes of whole grains and dietary fiber were associated with less weight gain among men from the HPFS (21). In the present analyses we also found that a high whole-grain intake was associated with lower plasma leptin concentrations. High leptin concentrations were previously shown to predict increased weight gain among overweight men from our cohort (14). Other micronutrients found in whole grains, such as magnesium and vitamin E, may also have a role in insulin sensitivity (35, 36). The soluble fiber content of whole grains may be responsible for the lower cholesterol concentrations (37, 38). However, whole wheat contains little soluble fiber but is the major contributor to the consumption of whole-grain products in the United States (39). This source of insoluble fiber does not affect plasma cholesterol concentrations (37). Plant estrogens and sterols are other components of grain foods that may have promising effects on lipid profile (40, 41). However, data on the relation of lignans (the main group of phytoestrogens found in grain products) and IHD risk factors in humans are limited. In our male cohort, a high dietary intake of lignans was not associated with improvements in lipid profile (42).
In conclusion, we found that higher intakes of whole grains in healthy men and women were associated with homocysteine concentrations and markers of improved glycemic control. Weak and somewhat inconsistent associations were observed for the lipid-related measures, and there were no associations between whole grains and the available markers of inflammation in our study. Furthermore, the associations between the whole-grain components—bran and germ—and the biomarkers were not appreciably different. Although a diet high in whole grains, bran, and germ may be associated with better glycemic control, limitations pertinent to a cross-sectional study should be recognized. Future investigations may consider specific sources of whole grains, which may have implications for the grain-particle size, the type of grain, the solubility of dietary fiber, and the amounts of bran, germ, and other micronutrients. Furthermore, investigations of intact compared with milled kernels deserve attention because the bioavailability of the many micronutrients in intact kernels is unexplored.
ACKNOWLEDGMENTS
All authors were involved in the critical revision of the manuscript and gave final approval of the submitted manuscript. MKJ, PK-B, MF, LS, MG, and EBR contributed to the design of the study and the analysis and interpretation of the data. MKJ drafted the manuscript. None of the authors had any conflicts of interest.
REFERENCES
Related articles in AJCN: