| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Family, Nutrition, and Exercise Sciences, Queens College of the City University of New York, Flushing, NY (A.K.K.)
Division of Cancer Epidemiology and Genetics, Biostatistics Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD (B.I.G.)
Address correspondence to: Ashima K. Kant, PhD, Professor, Dept. of Family, Nutrition, and Exercise Sciences, Remsen Hall, Room 306E, Queens College of the City University of New York, Flushing, NY 11367. E-mail: akant{at}qc.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
Methods: We used dietary and laboratory data from the third National Health and Nutrition Examination Survey to study these associations in 8719 disease-free adults aged 
20 y. The HEI, developed by the USDA, was a sum of scores on consideration of ten individual components; the RFS was a sum of all recommended foods (lean meat, poultry and fish, whole grains, fruits and juices, low-fat dairy, and vegetables) mentioned in the recall; the DDS-R examined whether or not a recommended food was mentioned from each of the five major food groups. The independent association of the dietary pattern indexes with body mass index (BMI), blood pressure, and serum concentrations of several biomarkers were examined using regression methods to adjust for multiple covariates.
Results: All indexes were strong independent positive predictors of serum concentrations of vitamin C, E, folate, and all carotenoids (p 
0.00001), except lycopene, and were negative predictors of BMI, serum homocysteine, C-reactive protein, plasma glucose, and hemoglobin A1C (p < 0.05). The RFS and DDS-R were inversely associated with blood pressure and serum cholesterol (p 0.03).
Conclusions: The RFS and DDS-R performed as well or better than the HEI for predicting serum concentration of nutrients and biomarkers of disease risk.
Key words: diet quality, dietary indexes, biomarkers, BMI, blood pressure, NHANES III, components of variance
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
Dietary Assessment Method
A 24-hour dietary recall was collected by a trained dietary interviewer in a MEC interview using an automated, micro-computer based interview and coding system [10]. The type and amount of foods consumed were recalled using aids such as abstract food models, special charts, measuring cups, and rulers to help in quantifying the amounts consumed. Special probes were used to help recall commonly forgotten items such as condiments, accompaniments, fast foods, and alcoholic beverages, etc.
As part of a sub-study within NHANES III, a nonrandom subsample of approximately 5% of those who completed a visit to the MEC were invited back for a second visit [11]. During this repeat visit, another dietary recall, using methods similar to the first one, was also obtained.
Analytic Sample
All adults aged 20 y and over with a complete and reliable dietary recall (as determined by NCHS) were eligible for inclusion (n = 15,979). We excluded respondents stating that food intake on recalled day was "much less" or "much more" than usual (n = 2245), women who were pregnant (n = 282) or nursing (n = 91), and those missing information on body weight (n = 33) or height (n = 16). Finally, respondents who reported that they were informed by a doctor that they had diabetes, hypertension, or history of heart attack (or reported use of medications for hypertension, hyperglycemia, or hyperlipidemia) were excluded (n = 4681)—leaving an analytic sample of 8719. Some respondents were in more than one exclusion category.
From the second exam subsample, a reliable second recall was available for 617 respondents in our analytic sample who also provided a reliable first recall.
Dietary Indexes
From the 24-hour recall data, we derived three dietary indexes: the Healthy Eating Index (HEI), the Recommended Foods Score (RFS), and a Dietary Diversity Score (DDS-R).
Healthy Eating Index (HEI).
The HEI was developed by the USDA to assess concordance with dietary guidance and to monitor changes in diet quality of the US population [6]. The HEI comprises ten components (saturated fat intake, total fat intake, cholesterol intake, sodium intake, grain intake, fruit intake, dairy intake, meat intake, vegetable intake, and dietary variety) that contribute 10 points each to the maximum possible score of 100. The food group serving criteria are gender and age specific. The HEI available on the NHANES public release data file was used for examination of association with biomarkers [12].
The HEI is not available for the second recall subsample. Because of our interest in exploring components of variance in the HEI using the second recall data, we computed the HEI following the methods described in the NHANES III Healthy Eating Index documentation [12] for both days 1 and 2. To enable this computation, we used the NHANES III pyramid servings data base (PSDB) [13]. All components except the dietary variety score could be replicated and were highly correlated with those available on the data file (with Pearson’s correlation coefficients ranging from >0.91–0.99). According to the NHANES III documentation, the HEI variety score is commodity based, ingredients of foods reported contribute to the variety score, and different forms of a food item are collapsed, but excludes ingredients in "fats, sweets, seasonings, and similar foods". We modified the variety component to consist of all unique food codes within the energy-dense food groups (fruit, vegetable, meat/and alternates, grains, and dairy) if mentioned in amounts that were at least half serving of a food as defined in PSDB. Mixed dishes contributed as many points as component food groups. The variety component score computed by us and that available in the NHANES III data file were highly correlated (r = 0.72). The overall HEI score computed by us (labeled HEI-C) and that available in the data file were highly correlated (r = 0.96).
Recommended Foods Score (RFS).
We previously developed the recommended foods score using food frequency data in the Breast Cancer Detection and Demonstration Project cohort [7]. Because the HEI in the NHANES III was computed from the 24-hour recall data, to ensure comparability, we adapted the original RFS computed from the food frequency questionnaire to 24-hour recall. Individual foods reported in the recall were examined and grouped into the recommended category. Out of a total of 4263 foods reported by the analytic sample, (1439) or 38% met the criteria for recommended foods. Foods considered to be recommended (low fat dairy; lean meats, poultry, fish, and alternatives; low fat whole grains such as whole wheat breads and cereals; all fruits and juices; all vegetables that were not fried, pickled, or creamed) in each recall were summed. Mixed dishes were considered recommended if their fat content was <35% of energy, and contributed one point to the overall score. The foods were counted only if they were consumed in at least a minimum amount—the minimum amount threshold was 15 g for non-beverages, and 30 gm for beverages. For recommended foods reported more than once, the criterion for minimum amount was applied after summing all mentions of a food. Finally, several mentions of a recommended food in a recall contributed only one point to the score.
Dietary Diversity Score (DDS-R).
The DDS as previously reported considers whether or not any food from each of the five major food groups was reported in the recall, with each food group contributing one point to the overall score, for a total of 5 [9]. For this study, we modified the previously published DDS by consideration of only those foods that are currently recommended in dietary guidance from each of the five major food groups. Accordingly, we evaluated each recall for whether a recommended food was mentioned from each of the five major food groups. Any mention of a recommended food from the fruit, vegetable, meat, dairy and grain in amounts meeting or exceeding the minimum amounts threshold mentioned above was considered to contribute one point to the score, for a maximum of five. Mixed dishes meeting the recommended criteria contributed component food group mentions to the score.
Biomarkers
The biomarkers available in the NHANES III public release data file included established, and emerging biomarkers of risk of cardiovascular disease, diabetes, and obesity: body mass index, systolic and diastolic blood pressure, serum total cholesterol, low density lipoprotein (LDL) cholesterol, high density lipoprotein (HDL) cholesterol, serum triglycerides, plasma glucose, serum insulin, glycated hemoglobin, serum C-peptide, lipoprotein (a), apolipoprotein AI and B, homocysteine, plasma fibrinogen, C-reactive protein, and leptin [14, 15]. Finally, we examined biomarkers of dietary exposure: serum vitamin C, serum vitamin E, serum folate, and the carotenoids—beta carotene, alpha carotene, lutein/zeaxanthin, beta cryptoxanthin, and lycopene. For serum triglycerides, plasma glucose, serum c-peptide, and insulin, respondents reporting <9 hours of fasting before phlebotomy were excluded.
Statistical Analyses
All statistical analyses were performed using SAS [16], and software designed for analysis of survey data (SUDAAN) [17]. This software generates variance estimates that are corrected for multi-stage stratified cluster probability design of complex surveys. Sample weights provided by the NCHS to correct for differential probabilities of selection, non-coverage, and non-response were used in all analyses to obtain point estimates.
We computed the least square mean and standard error of each score variable by selected socio-demographic and lifehabit characteristics from multivariate regression models. The association of each serum or plasma analyte with each dietary index was examined using linear regression models to adjusted for a number of covariates. All covariates included in the various multiple regression models were decided apriori based on known relationships of biomarkers with socio-demographic, and lifestyle factors, and included: gender, age, race (non-Hispanic White, non-Hispanic Black, Hispanic, other), years of education (<12, 12, >12), smoking status (never, former, current), level of weekly recreational physical activity (none, 1–2 times/week, >2 times/week), body mass index (<25, 25–30, >30), alcohol intake (none, 1, >1–2, >2 drinks/day), hours of fasting, supplement use in the 24 hours before phlebotomy, and supplement use in the past month. The multiple regression models for examining the independent association of dietary indexes with body mass index and blood pressure included all of the above variables except hours of fasting, supplement use in the past 24 hours before phlebotomy, supplement use in the past month. The and mean of each biomarker was obtained from covariate adjusted models by approximate quartiles of each dietary pattern index. To examine the trends in biomarker concentration, the dietary pattern index variables were also modeled as continuous variables. The Pearson’s correlation method was used to examine the association of each diet index variable with energy, macro and micronutrient intake, using the NCHS recommended sample weights.
From the second dietary recall obtained from the second exam subsample for NHANES III, we computed within- and between-person components of variance for each type of dietary pattern index, using the varcomp procedure available in SAS.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
50 y, with >12 y of education, and <25 BMI were in the highest approximate quartile of each index variable (Table 2). Approximately 7% of the population mentioned no recommended food in the 24-hour recall, 38% mentioned a low-fat or non-fat dairy product, 49% mentioned a fruit or fruit juice, 38% mentioned a whole grain, 42% mentioned a lean meat, poultry, fish or alternative, and 77% mentioned a vegetable that was not fried, creamed, or pickled. Only 5% mentioned at least one recommended food from all five major food groups on the recall day (data not shown).
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
In accord with other observational studies that have examined the association of dietary indexes or scores with BMI, we found all scores in the present study to be negative predictors of BMI [18]. We had also expected the three scores to be associated with blood pressure because they approximate the DASH dietary patterns [19]. The results for the association of RFS and DDS-R were consistent with this expectation. Surprisingly, however, the HEI was not a predictor of blood pressure. It is possible that the observed associations may reflect a somewhat stronger association of RFS and DDS-R with dietary protein and most micronutrients (except sodium). It is noteworthy that few observational studies have examined the association of diet pattern indexes with blood pressure. In a comparison of dietary patterns for predicting the risk of incident hypertension in women in the EPIC-POTSDAM study, few associations were found between hypertension risk and dietary patterns. Although an association with the third quartile of a DASH diet based score was reported, the trend of association of risk of hypertension and the DASH diet score was not significant [20].
Given the single measurement of dietary intake in the present study, we did not expect to find associations of serum lipids with the dietary pattern indexes. We had further expected that due to the stronger association of HEI (relative to the other two indexes) with dietary saturated fat, HEI would perform better for predicting serum lipid /lipoprotein profiles. A stronger association of HEI with dietary fat and saturated fat was expected since these two nutrients are components of the overall score. Although HEI was a negative predictor of LDL concentration, it is notable that HEI also had a negative association with HDL concentration. On the other hand, the RFS and DDS-R were far weaker predictors of dietary intake of total and saturated fat, but were independent negative predictors of serum total cholesterol. The reasons for these disparate associations are not clear. We note that Weinstein and colleagues have reported similar findings in a much larger sample of NHANES III which had no exclusions for known chronic disease or its pharmacologic treatment [21].
Few studies have examined the independent association of index based dietary patterns with the range of biomarkers that were examined in the present study. Generally, the associations of serum biomarkers of fruit and vegetable intake with index based patterns have been reported to be consistently positive [22–25]. Hann et al found HEI determined from 3-day diet records to be a positive predictor of serum carotenoids and vitamin C, but not serum folate and lycopene in 340 women [22]. Three studies have examined different versions of another complex quantitative index—the diet Quality Index or DQI in relation to dietary biomarkers [23–25]. Neuhouser derived DQI from a 122-item FFQ and found it to be related positively with serum vitamins C and E, and beta cryptoxanthin but not serum folate [23]. Gerber et al found a DQI adapted for the French diet to be a predictor of serum vitamin E, DHA, EPA, and beta carotene [24]. Newby et al also reported that plasma carotenoids and vitamin E were positive correlates, and serum cholesterol a negative correlate, of a revised DQI [25]. It is noteworthy that although the RFS and DDS-R are relatively simple indexes that do not require quantitative estimation of servings of foods beyond a minimum threshold amount, but were positive predictors of dietary biomarkers.
We could find no studies where the independent association of index based patterns with prothrombotic, proinflammatory, or glycemic biomarkers has been reported. Finally, we included known confounders of the association of dietary variables and biomarkers in our regression models, however, the possibility of residual confounding due to poorly measured or unknown confounders in an observational study such as this one, can not be excluded.
A number of studies where patterns were derived from factor or cluster analysis have also examined biomarkers and present a mixed picture. In an analysis of data from a subsample (n = 466 men) of the Health Professionals Study cohort, Fung et al found the pattern labeled as prudent (from factor analysis) to be an inverse predictor of serum homocysteine and insulin and a positive predictor of serum folate; however, fifteen other biomarkers of obesity and cardiovascular disease examined were unrelated with this pattern [26]. The pattern labeled western in this study was a positive predictor of HDL cholesterol, LDL cholesterol, homocysteine, insulin, and c-peptide. Kerver et al derived two patterns (western and American healthy) from factor analysis of the FFQ from the NHANES III. American healthy-showed no association with any examined biomarker, the western pattern showed a positive association with three glycemic biomarkers, C-peptide, insulin and glycated hemoglobin [27]. Recently, Newby et al found a healthy cluster and factor to be predictors of serum triacylglycerol concentration, but not total, LDL or HDL cholesterol [28]. This brief review of studies that have examined biomarker and pattern associations suggests that the dietary pattern indexes examined in this study perform somewhat better at predicting biomarkers of disease than several other patterns reported in the literature. However, because index-based patterns (as opposed to patterns derived from exploratory factor or cluster analyses) reflect known diet and disease associations, their stronger associations with biomarkers are not surprising.
We note that the three indexes reported here were derived from a single 24-hour recall with its well known limitations for measuring usual dietary intake [29]. Although the association of dietary pattern indexes with dietary biomarkers provides an objective validation of the self-reported intake, considerable dietary reporting errors are known to be present in the NHANES III data and probably contribute to misclassification of respondents [30]. Apart from random or systematic errors in self-reported intake, given the ratios of within to between-person components of variance for the three indexes (especially the HEI and the DDS-R), it is likely that with-in person variability in intake contributed to attenuation of estimates of association.
To our knowledge, this is the first study to present estimates of within-person and between-person variance in the three dietary pattern indexes reported here. Generally, the estimates of the ratio of within-person to between-person variance are smaller than other published variance estimates for most nutrients and food groups [31]. We note however, that our estimates are derived from only two dietary measurements and our methods are sensitive to variability in only the total score of each index. Variability in food selection which may not result in a change in the overall score is not captured by our methods.
An assumption inherent in all studies that examine the association of index based patterns with an outcome is that a given less than perfect score should relate similarly to an outcome. However, an identical numerical score on an index based pattern may reflect many different dietary profiles. For example, a score of 70 on the HEI can result from many different combinations of the ten contributing components, a score of four on the RFS can result from mention of four recommended foods from one or more food groups in several different combinations, and a score of 3 on the DDS-R can be due to omission of any two of the five food groups. Thus, it is likely that subjects with similar scores may not show a similar association with an outcome further contributing to attenuation of the association.
In conclusion, the two relatively simple dietary pattern indexes—the RFS and DDS-R—performed as well if not better than the HEI for predicting serum nutrient concentrations and biomarkers of disease risk.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
Received December 7, 2004. Accepted April 23, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. Boynton, M. L Neuhouser, M. H Wener, B. Wood, B. Sorensen, Z. Chen-Levy, E. A Kirk, Y. Yasui, K. LaCroix, A. McTiernan, et al. Associations between healthy eating patterns and immune function or inflammation in overweight or obese postmenopausal women Am. J. Clinical Nutrition, November 1, 2007; 86(5): 1445 - 1455. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. J. Fogli-Cawley, J. T. Dwyer, E. Saltzman, M. L. McCullough, L. M. Troy, J. B. Meigs, and P. F. Jacques The 2005 Dietary Guidelines for Americans and Insulin Resistance in the Framingham Offspring Cohort Diabetes Care, April 1, 2007; 30(4): 817 - 822. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. J. Fogli-Cawley, J. T. Dwyer, E. Saltzman, M. L. McCullough, L. M. Troy, and P. F. Jacques The 2005 Dietary Guidelines for Americans Adherence Index: Development and Application J. Nutr., November 1, 2006; 136(11): 2908 - 2915. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
