| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ORIGINAL RESEARCH COMMUNICATION |
1 From the Jean Mayer US Department of Agriculture Human Nutrition Research Center on Aging at Tufts University, Boston (PKN); the Departments of Nutrition (FBH, EBR, SAS-W, LS, and WCW) and Epidemiology (EBR and WCW), Harvard School of Public Health, Boston; and the Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston (EBR, DF, and WCW).
2 Supported by research grants R01 HL60712, HL35464, and CA55075 from the National Institutes of Health. 3 Reprints not available. Address correspondence to PK Newby, Jean Mayer USDA Human Nutrition Research Center on Aging at Tufts University, 711 Washington Street, 9th Floor, Boston, MA 02111. E-mail: pknewby{at}post.harvard.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Objective: The objective of this study was to assess the reproducibility and validity of the DQI-R as measured by use of food-frequency questionnaires (FFQs).
Design: Diet was assessed separately by two FFQs at a 1-y interval and by two 1-wk diet records. DQI-R scores were computed from each method. Venous blood specimens were collected for measurement of dietary biomarkers. Participants (n = 127) were men aged 40-75 y in a validation study of the Health Professionals Follow-up Study.
Results: Mean DQI-R scores were 69.5 for FFQ-1, 67.2 for
FFQ-2, and 62.0 for the diet records out of a possible score of 100.
The reproducibility correlation for the 2 FFQ scores was 0.72.
Correlations between scores for each of the 2 FFQs and diet
records were 0.66 (FFQ-1) and 0.72 (FFQ-2). DQI-R scores from
FFQ-2 were directly correlated with plasma biochemical measurements of 
-carotene (r = 0.43, P < 0.0005), ß-carotene (r = 0.35,
P < 0.005), lutein (r = 0.31, P <0.005), and -tocopherol (r =
0.25, P < 0.05) and were inversely correlated with plasma total
cholesterol (r = -0.22, P < 0.05).
Conclusions: These data indicate reasonable reproducibility and validity of the DQI-R as assessed by an FFQ. Future studies are needed to examine whether this index and other instruments of diet quality can reliably predict disease outcomes.
Key Words: Diet index • diet quality • dietary pattern • bio-marker • reproducibility • validity
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
One methodologic approach to the measurement of total diet quality uses an index, in which separate nutritional elements or constructs are combined into a single score (1-4). The Diet Quality Index (DQI) is an instrument developed to measure overall diet quality that reflects a risk gradient for diet-related chronic disease (4). The original DQI is based on recommendations made in Diet and Health: Implications for Reducing Chronic Disease Risk (19) and consists of 8 dietary variables (total fat, saturated fat, cholesterol, fruit and vegetables, grains and legumes, protein, sodium, and calcium) that are summed into a composite diet quality score. Scores range from 0 to 16, where 0 reflects the highest quality diet and 16 the lowest.Comparing the original DQI with components not included in the index indicated strong relations between a low DQI score (excellent diet) and high fiber and vitamin C intakes (4).
The index was subsequently updated (20) to reflect additional aspects of diet quality not addressed in the original index, including variety, moderation, and proportionality, as reflected in the Food Guide Pyramid (21) and the Dietary Guidelines for Americans (5th edition) (22), as well as changes in nutritional recommendations and policy [eg, the score for the calcium component was changed from being based on the recommended dietary allowances (23) to being based on the dietary reference intakes (DRIs; 24)]. The Diet Quality Index Revised (DQI-R) includes 10 components, 4 of which are the same as in the original DQI (total fat, saturated fat, cholesterol, and calcium). The fruit and vegetable component is now 2 separate components, grains is its own category, and iron replaces protein. Dietary moderation and diversity are 2 new components. DQI-R scores range from 0 to 10 for each component, for a highest possible diet quality score of 100. Among a representative sample of 3202 adults participating in the 1994 Continuing Survey of Food Intakes by Individuals who had completed two 24-h recalls, higher DQI-R scores were related to lower fat consumption, higher fruit and vegetable intakes, and higher iron and calcium intakes (20).
To our knowledge, the reproducibility of the DQI-R and its comparability across dietary assessment methods have not been assessed. Establishing the validity of this index will help to demonstrate its utility in assessing diet quality and hence its potential use in assessing diet and disease associations. The objective of our study was to assess the reproducibility and validity of the DQI-R as measured by use of food-frequency questionnaires (FFQs) among male health professionals.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Dietary assessment
The 127 men in the diet validation study completed FFQs at
baseline in 1986 (FFQ-1) and again in 1987 (FFQ-2), roughly
1 y apart, as well as two 1-wk diet records <7 mo apart during
the year between the 2 FFQs. To complete the diet records, the
participants were contacted by a research dietitian and provided
detailed instruction on how to record diet intake by using
specially designed booklets and scales to weigh foods. Records were individually reviewed and the participants were
re-contacted to provide further detail when necessary. The
reproducibility and validity of nutrient intakes (26), food intakes (27), and dietary patterns as measured by factor analysis
(28) have been described elsewhere.
The FFQ included 131 food items that were selected to
describe usual dietary intake over the past year. Participants
were asked to describe their average intake of each food by
using 9 frequency of consumption categories ranging from
"almost never" to "
6 times/d." For FFQ-1, 35% of subjects
had complete FFQs, with no missing food items; 76% of
subjects had no more than 3 missing food items; and 8 subjects
(6%) were missing > 10% of items. For FFQ-2, 39% of
subjects had complete FFQs, with no missing food items; 91%
of subjects had no more than 3 missing food items, and 4
subjects (3%) were missing > 10% of items. Food items that
were not answered on the FFQ were considered to reflect
nonconsumption and were recoded as "almost never."
Dietary data from the FFQ were converted to average daily intake values (eg, 1 serving/wk = 0.14 serving/d). Serving size definitions for the FFQ were based on "natural" portions (eg, 1 slice of bread) or typical serving sizes (21, 29); as aforementioned, scales were provided to participants to weigh and record food for the diet records. On the 2 wk of diet records, there were 1565 unique foods reported; mixed dishes were converted to recipes to obtain food ingredient data (27). The average daily intake from the two 1-wk diet records was calculated and used in all analyses to reduce the effect of within-person variation in daily food consumption.
Daily food intake data were grouped into the food-based DQI-R dietary components, and the DQI-R nutrient components were assessed directly (29). According to the method of Haines et al (20), the total fat, saturated fat, and cholesterol components were calculated as a percentage of total energy and were categorically scored as 0, 5, or 10, and the remaining components were scored as continuous variables from 0 to 10, proportional to the recommended range of intake. Scores were summed across the 10 components for a highest possible score of 100 points.
The goals for the fruit, vegetable, grain, and added sugar components as defined by the food guide pyramid (21) depend on daily energy intake. In our study population, most of the participants (> 80%) reported energy intakes between 7531 and 10 878 kJ/d (1800-2600 kcal/d), with a mean close to 9205 kJ/d (2200 kcal/d) from the average of the 2 wk of diet records and the FFQs, so we used the recommendations for this energy range. The fruit and vegetable components included fresh, canned, and dried fruit and vegetables and juices. The grains component included breads, grains, cereals, rice, pasta, popcorn, and crackers. According to the method of Haines et al (20), sweets such as pies, cakes, cookies, and pastries were excluded from the grains score, although these foods are considered part of the grains food group in the food guide pyramid.
The DQI-R components for cholesterol, calcium, and iron
were not adjusted for total energy because the cutoffs for
intakes used in the DQI-R are not energy dependent. The
adequate intake (23) for calcium and the recommended dietary
allowance for iron (23) are age dependent, and age-specific
cutoffs are indicated in the footnotes to Table 1
.
|
Because Haines et al (20) used a diversity cutoff of 1/2
serving/d over a 2-d period, we thought it most comparable if
we halved the cutoff in our study to 1/4 serving/d, because the
FFQ (and the mean of the diet records) provided us with daily
estimates (ie, 1/2 serving per 2-d period = 1/4 serving per 1-d
period). Participants received 1 point if they consumed 1/4
serving/d of the foods within each subgroup (alone or in
combination) and 0 points if they consumed < 1/4 serving/d.
For each food group, points were summed across the subgroups
and divided by the total number of subgroups and then multiplied by 2.5 to receive a top score of 2.5 points per food group,or 10 points in total for the diversity component across the 4
food groups.
Minor changes were made to the food subgroups because of differences in foods contained on the FFQs used in the Health Professionals Follow-up Study. Notably, whole-grain and non-whole-grain cereals were combined into one subgroup of cereals. We also added an "other grains" subgroup that included wheat germ, bran, and other grains not specified. In addition, the fruit category was expanded from 2 subgroups to 3 subgroups to reflect the large number of "other" fruit contained in our FFQ.
DQI-R dietary moderation component
The dietary moderation component is comprised of 4 subgroups (added sugar, discretionary fat, sodium, and alcohol)
that each contribute a maximum score of 2.5 points. The added
sugar component is defined by the US Department of Agriculture Food Surveys Research Group (30) and the food guide
pyramid (21) to reflect "1 teaspoon of added sugar, where 1
teaspoon is the quantity of a sweetener that contains the same
amount of carbohydrate as one teaspoon of table sugar." Products that contribute to added sugar include all sweeteners that
are eaten separately or used as ingredients in processed or
prepared foods (21). To quantify the added sugar component,
we included teaspoons of added sugar consumed per day (eg,
added to coffee or cereal), which was directly assessed from
the FFQs and diet records. We also derived added sugar intake
by summing the sucrose content per serving across major
foods, including muffins and biscuits, pancakes and waffles,
nondiet cola, chocolate and nonchocolate candy, cookies,
brownies, donuts, cake, pie, sweet roll and coffee cake, and
jam, jelly, syrup, and honey. Teaspoons of sugar were derived
from the total sucrose intake of the added sugar foods, where
3.8 g sucrose = 1 tsp sugar. Together, the direct and derived
sugar values were summed to total added sugar consumption.
Discretionary fat is defined in the food guide pyramid (21) as the difference in fat content between full-fat and low-fat products, specifically, "all excess fat...beyond amounts that would be consumed if only the lowest fat forms were eaten, and fats added to foods in preparation or at the table." To derive discretionary fat intake, we included the fat from foods including cream, butter, margarine, cream cheese, oils, salad dressings, chocolate, whole milk, sour cream, ice cream, mayonnaise, coffee whitener, and baked goods. The grams of fat contained in a serving (29) were then separately calculated (eg, 2 tablespoons cream at 2.9 g fat/tablespoon = 5.8 g discretionary fat) and summed across all foods to derive total discretionary fat consumption. Our method differed slightly from the food guide pyramid definition (21). In this study, we included the absolute number of fat grams contained in these products as discretionary, given that fat-free choices for most of the foods listed above are available and an individual can choose not to consume these foods at all.
Alcohol intake was counted directly for both dietary assessment methods, and sodium (and other nutrient components) was derived by using a nutrient database (29). For sodium, salt used in cooking and at the table was included for both the FFQs and the diet records. Specifically, on the FFQ, the participants were asked how much salt they added to staple foods (eg, rice or pasta), meats, vegetables, and soups during cooking (1/8, 1/4, or 1/5 tsp per serving); the frequency of the foods in the different groups were summed and then multiplied by the salt quantity selected. Participants were also asked on the FFQ to estimate how often they added salt to foods at the table, as well as how many shakes of salt they usually added, where 0.3 g salt was represented in each shake. For diet records, there was a column for participants to directly record both salt added during cooking (in teaspoons) and shakes of salt added to food at the table.
Laboratory analyses
Blood samples were obtained from 121 nonfasting participants shortly before they completed the second FFQ. Blood
specimens were collected into EDTA-treated tubes in the
morning and then covered with aluminum foil and stored in the
dark on ice until the plasma was separated; plasma was stored
at -70 °C until analyzed. Plasma carotenoids, tocopherols, and
retinol were measured by reversed-phased HPLC in the laboratory of Hoffmann-La Roche (Basel, Switzerland) (31).
Plasma cholesterol and triacylglycerol concentrations were
measured by using kits from Hoffmann-La Roche, according to
the methods of Richmond (32) and Bucolo and David (33),
respectively.
Statistical analyses
All analyses were performed by using the SAS statistical
software package (version 6; SAS Institute Inc, Cary, NC). For
individual DQI-R components, the proportion of men in each
scoring category and mean (±SD) intakes were calculated for
each dietary assessment method. Mean scores for individual
components and the total DQI-R score were also calculated.
We calculated Pearson correlation coefficients to assess the
reproducibility from the repeated FFQs and the validity comparing the FFQs to the diet records for individual DQI-R
components and total DQI-R scores. We calculated deattentuated correlation coefficients to reduce the effect of week-to-week variation in diet record intake, as suggested by Rosner
and Willett (34).
Our validation analysis compared total DQI-R scores from
the FFQs and diet records with plasma biochemical measurements. Additional validation compared DQI-R scores from
both methods with nutrients derived from the diet records. Both
analyses used Pearson correlations. Nutrients were energy-adjusted by using the residual method (35) and log-transformed
to improve normality. Plasma measurements of retinol, - and
ß-carotene, and -tocopherol were adjusted for age, plasma
cholesterol, plasma triacylglycerols, and body mass index.
Smokers and users of multivitamin or ß-carotene supplements
were excluded from retinol and carotenoid analyses, whereas
users of multivitamin or single vitamin E supplements were
excluded from -tocopherol analyses.
The distribution of scores for total DQI-R and individual index components and the reproducibility and validity correlations are presented for both FFQ-1 and FFQ-2. All other data are presented for FFQ-2 only because of the similarity of results between the 2 FFQs.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
). The weak validity correlation for iron (r = 0.06) may have
been due to limited variation in iron intake, because most
subjects met the goal for iron according to both FFQ-2 and the
diet records. Intakes of fat, saturated fat, and cholesterol were
underreported on the FFQs compared with the diet records. For
example, 23% of men met the goal for total fat consumption
(
30% of total energy intake) as assessed by the diet records,
whereas 43% of men met the goal as assessed by FFQ-2 (r =
0.45). Similarly, 43% of men met the goal for cholesterol
consumption ( 300 mg/d) as assessed by the diet records,
whereas 65% met the goal according to FFQ-2 (r = 0.47). The validity correlations between each FFQ and the diet records, which were statistically adjusted to reduce the effect of week-to-week variation in diet records, were 0.66 (FFQ-1) and 0.72 (FFQ-2). Validity varied among food group components of the DQI-R. Even though fruit consumption was overreported on the FFQs compared with the diet records, the fruit score was the most highly correlated component between FFQ-2 and the diet records (r = 0.71). A lower correlation was observed for the vegetable score (r = 0.19). Very few men met the goal for grain consumption according to both methods (r = 0.39).
Mean DQI-R scores were 69.5 for FFQ-1, 67.2 for FFQ-2,
and 62.0 for the diet records out of a possible 100 points (Table
2). Mean scores were higher for the fat, saturated fat, cholesterol, fruit, vegetable, diversity, and moderation components on
FFQ-2 than on the diet records. Calcium and iron scores were
similar, although the mean intake of iron was almost 4 mg/d
higher on the diet records.
|
. Most of the subjects met the goal for sugar consumption according to both FFQ-2 (90%) and the diet records (94%)
(r = 0.21), whereas almost 3 times as many subjects met the
goal for discretionary fat as estimated by FFQ-2 (43%) compared with the diet record (16%) (r = 0.50). Alcohol intake
was strongly correlated between assessment methods (r =
0.65) (correlations not shown).
|
are the correlation coefficients and P
values comparing DQI-R scores from FFQ-2 and the diet
records with plasma biochemical measurements and nutrients
from the diet records. DQI-R scores from both methods were
directly correlated with plasma biochemical measurements of
-carotene, ß-carotene, lutein, and -tocopherol, but inversely
correlated with cholesterol (P < 0.05). DQI-R scores calculated from both diet assessment methods were also directly
related to vitamins B-6 and C, fiber, folate, magnesium, calcium, and carotene intakes and inversely related to fat, saturated fat, monounsaturated fat, and cholesterol intakes from the
diet records (P < 0.05).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
We also compared the DQI-R scores from the FFQs and diet records with nutrient intakes estimated by the diet records because diet records have the fewest correlated errors with FFQs and are therefore the most widely used dietary assessment method for validating an FFQ. The major sources of error associated with FFQs are limited food items, memory of food consumed, assessment of portion size, and interpretation of questions. These sources of error are minimally shared with the diet record method, which is open-ended, involves recording of foods as they are consumed, and involves direct weighing of food portions (37). Diet quality assessed from the FFQ showed reasonable validity when compared with nutrient intakes from diet records. Our findings were generally as expected, given that many of the nutrients were specifically incorporated into the DQI-R. However, several nutrients significantly related to the total DQI-R score were not measured by the diet records (eg, fiber, vitamin C, and folate), suggesting that the DQI-R captures additional aspects of diet quality.
Several diet quality indexes have been associated with
plasma biomarkers and specific nutrient intakes. A diet quality
index similar to the DQI-R was positively correlated with
plasma concentrations of eicosapentaenoic acid and docosahexaenoic acid and inversely related to cholesterol in a representative sample in southern France (38), whereas an earlier
dietary scoring method based on the (then) basic 4 food groups
(1) was positively correlated with intakes of calcium and magnesium. More recently, Hann et al (39) observed associations
between the Healthy Eating Index and plasma -carotene (r =
0.40, P < 0.05) and ß-carotene (r = 0.28, P < 0.05) that were
of a similar magnitude to those obtained in our study.
In our study, the correlations between DQI-R scores calculated from both methods and the plasma biomarkers were similar for many of the nutrients, indicating that the FFQ is a reasonable estimate of diet quality when compared with 2 wk of diet records. In general, although FFQs are not as reliable in assessing absolute intakes as are diet records, the reasonable correlations observed indicate that individuals can be ranked with sufficient accuracy with respect to diet quality, as has been shown for intakes of nutrients (26) and foods (27). The underreporting of dietary components perceived as relatively less healthy (eg, saturated fat) and the overreporting of dietary components perceived as relatively more healthy (eg, fruit) on the FFQ contribute to the higher total DQI-R scores seen on the FFQs than on the diet records. These biases of FFQs compared with diet records have been previously discussed (26, 27). Assessment of diet quality, then, must consider the limitations of the primary dietary assessment method when interpreting results generated from different methods.
A limitation of our study was our ability to exactly reproduce the method used in the DQI-R, specifically for discretionary fat and added sugar. Because our FFQ and diet records did not contain direct measures of discretionary fat or added sugar, these components were derived from relevant foods for both methods and our choices of foods to include was instructed by the definitions used in the food guide pyramid (21). It is arguable that additional fat-containing foods should have been counted as discretionary, such as fried foods, which could be baked or broiled rather than fried, or high-fat cuts of meat, which could be replaced by low-fat cuts of meat. A similar problem was encountered for the added sugar component, in which we created a list of foods containing added sugars; additional foods could also have been included, such as sweetened juice drinks or dairy desserts. Omission of these foods with discretionary fat (ie, fried foods or high-fat cuts of meat) or added sugar (ie, sweetened juice drinks or dairy desserts) would, if many participants consumed them in large amounts, lead to a lower score on the moderation component and an overall lower DQI-R score than reported here.
Counting food ingredients from mixed dishes always poses a challenge when assessing dietary intake, and a limitation of this study is the potential loss of dietary information from mixed dishes. For example, a consistent loss of flour from mixed dishes would lead to an underestimation of grain consumption. This may have contributed to the low mean grain consumption observed in our study, thus artificially decreasing the DQI-R score. Another possible reason for our low grain score could be because, in replicating the method of Haines et al (20), we did not count grains consumed from baked goods in our estimate. This method differs from the food guide pyramid (21), which does count these foods toward grain consumption. Therefore, the decision to omit grains from baked goods from the DQI-R score reflects the authors' preferences (20) and not current nutrition policy.
An additional limitation of this study is the homogeneity of our study population, all of whom were highly educated white males. Diet quality varies among populations, and validity of the DQI-R method may vary for women, who generally have healthier diets (20), and for groups of a lower socioeconomic status, who generally have less healthy diets (39).
Assessment of reproducibility and validity of an instrument
such as the DQI-R is only one step in the evaluation of a dietary
assessment method. Whether the index or other measures of
diet quality can predict disease across diverse populations is the
ultimate test of validity. A review of diet quality indexes found
that diet quality was related to the risk of disease more strongly
than were individual nutrients or foods (40), but recent studies
examining the relation have led to inconsistent results (3, 12,
36, 41). The inconsistencies could be attributable to specific
components included in the indexes that may not be clearly
associated with disease risk. For the DQI-R, consuming 30%
of total energy from fat is one of the components, but total fat
intake may not be associated with either coronary heart disease
(42) or cancer (43), although there is considerable controversy
in this area (19, 44). In addition, whole grains may be protective against coronary heart disease, whereas refined grains may
increase risk (45), although the DQI-R recommendation for
grain consumption, as based on the food guide pyramid (21),
does not distinguish between whole and refined grains. As
such, diet quality indexes based on current nutrition policy may
be limited in utility if nutrition policy itself does not reflect
current nutrition knowledge. In addition, an index that is based
on current policy may become outdated as nutrition science
evolves. Diet quality indexes, then, are only as good as the
components on which they are based; hence, they inevitably
must be revised if they are truly to reflect the latest nutrition
science and policy.
In conclusion, although our findings of reasonable reliability and validity of the DQI-R are positive, they do not necessarily mean that this index is useful in predicting disease outcomes among persons who conform to the recommendations. Studying whether the DQI-R can reliably predict disease risk is the next step in the validation of its utility as a dietary assessment method.
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. Reedy, P. N. Mitrou, S. M. Krebs-Smith, E. Wirfalt, A. Flood, V. Kipnis, M. Leitzmann, T. Mouw, A. Hollenbeck, A. Schatzkin, et al. Index-based Dietary Patterns and Risk of Colorectal Cancer: The NIH-AARP Diet and Health Study Am. J. Epidemiol., July 1, 2008; 168(1): 38 - 48. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. M. Tomey, M. R. Sowers, C. Crandall, J. Johnston, M. Jannausch, and M. Yosef Dietary Intake Related to Prevalent Functional Limitations in Midlife Women Am. J. Epidemiol., April 15, 2008; 167(8): 935 - 943. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. N. Healy, D. W. Dunstan, J. Salmon, E. Cerin, J. E. Shaw, P. Z. Zimmet, and N. Owen Breaks in Sedentary Time: Beneficial associations with metabolic risk Diabetes Care, April 1, 2008; 31(4): 661 - 666. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. N. Healy, K. Wijndaele, D. W. Dunstan, J. E. Shaw, J. Salmon, P. Z. Zimmet, and N. Owen Objectively Measured Sedentary Time, Physical Activity, and Metabolic Risk: The Australian Diabetes, Obesity and Lifestyle Study (AusDiab) Diabetes Care, February 1, 2008; 31(2): 369 - 371. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Godwin, S. Streight, E. Dyachuk, E. C. van den Hooven, J. Ploemacher, R. Seguin, and S. Cuthbertson Testing the Simple Lifestyle Indicator Questionnaire: Initial psychometric study Can Fam Physician, January 1, 2008; 54(1): 76 - 77. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Koushik, D. J. Hunter, D. Spiegelman, W. L. Beeson, P. A. van den Brandt, J. E. Buring, E. E. Calle, E. Cho, G. E. Fraser, J. L. Freudenheim, et al. Fruits, Vegetables, and Colon Cancer Risk in a Pooled Analysis of 14 Cohort Studies J Natl Cancer Inst, October 3, 2007; 99(19): 1471 - 1483. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. T. Fung, F. B. Hu, M. L. McCullough, P. K. Newby, W. C. Willett, and M. D. Holmes Diet Quality Is Associated with the Risk of Estrogen Receptor-Negative Breast Cancer in Postmenopausal Women J. Nutr., February 1, 2006; 136(2): 466 - 472. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. A. Meyerhardt, D. Heseltine, H. Campos, M. D. Holmes, W. C. Willett, E. P. Winer, P. C. Enzinger, C. A. Bunnell, M. H. Kulke, and C. S. Fuchs Assessment of a Dietary Questionnaire in Cancer Patients Receiving Cytotoxic Chemotherapy J. Clin. Oncol., November 20, 2005; 23(33): 8453 - 8460. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. K. Kant and B. I. Graubard A Comparison of Three Dietary Pattern Indexes for Predicting Biomarkers of Diet and Disease J. Am. Coll. Nutr., August 1, 2005; 24(4): 294 - 303. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. T Fung, M. L McCullough, P. Newby, J. E Manson, J. B Meigs, N. Rifai, W. C Willett, and F. B Hu Diet-quality scores and plasma concentrations of markers of inflammation and endothelial dysfunction Am. J. Clinical Nutrition, July 1, 2005; 82(1): 163 - 173. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
