Abstract
Background: A premise of the 2005 Dietary Guidelines for Americans (DGA) is chronic disease prevention.
Objective: The goal was to determine whether a diet meeting the DGA is associated with less atherosclerotic lesion progression.
Design: We used the data from 224 postmenopausal women with established coronary artery disease enrolled in the Estrogen Replacement and Atherosclerosis Study. Atherosclerosis progression was defined by repeated measures of quantitative angiography over a 3-y period. Adherence to the key DGA recommendations was measured by using the DGA Adherence Index (DGAI; possible range: 0–20), with each component weighted equally, and the modified DGAI score (wDGAI; possible range: −0.19–0.51), with each component weighted based on its relation to atherosclerosis progression. Mixed-model regression analyses were performed to assess the association between diet and atherosclerosis progression.
Results: No women consumed a diet meeting all of the DGA recommendations. The mean (range) of the DGAI score was 14.1 (8.0–19.0). DGAI was not associated with atherosclerosis progression (P = 0.44), whereas wDGAI was inversely associated; a 1-SD difference in wDGAI was related to 0.049-mm less narrowing of the coronary arteries (SE = 0.017, P = 0.004).
Conclusions: In postmenopausal women with established heart disease, under the assumption that all DGA recommendations are similarly effective, overall adherence was not associated with atherosclerosis progression. However, assigning differential weights to the DGA recommendations, the adherence was significantly associated with slower atherosclerosis progression. Assuming equity of associations between all dietary recommendations and disease outcomes is a limitation in accurately examining the effectiveness of the DGA.
See corresponding editorial on page 13.
INTRODUCTION
One of the underlying premises of the 2005 Dietary Guidelines for Americans (DGA) is chronic disease prevention (1, 2). Whether or not dietary guidelines can affect disease progression among individuals with existing disease has yet to be adequately studied (3–5). Single dietary components are associated with reduced heart disease progression (4–9), which is the major cause of death in developed and now developing countries (10, 11). However, evidence relating overall adherence to dietary recommendations and coronary artery atherosclerotic lesion progression is lacking (12).
To address the question of adherence to dietary recommendations and its association with health outcomes, Kennedy et al (13) proposed the use of a dietary index—the Healthy Eating Index—as a measure of adherence to the 1995 DGA. Fogli-Cawley et al (14) developed a dietary index—the 2005 DGA Adherence Index (DGAI)—to assess adherence to the 2005 DGA and showed cross-sectional associations with risk factors of chronic diseases in a population-based cohort (14, 15). The dietary index technique has proven to be useful at measuring adherence to dietary recommendations and at testing the hypothesis that diets concordant with them have a favorable effect on sustaining heart health in a general population (14–16).
Dietary indexes are constructed by summing scores based on adherence to individual dietary recommendations. Contributions of individual components to a total score are frequently considered of equal weight in calculating an index score. This construction leads to the biologically implausible assumption that adherence to each recommendation is associated to the same extent with all health outcomes considered (17). The use of prior knowledge to construct a summary measure of overall diet is a strength of this approach. The assumption of equal weights is a potential weakness of the index approach, limiting our ability to evaluate the effective of adherence to dietary recommendations in both clinical and general populations. In contrast with the a priori criterion–based approach, dietary pattern analyses have been widely used as data-driven approach to obtain summary variables of dietary factors (18, 19). Therefore, there are advantages to using a hybrid method consisting of a criteria-based approach and a data-driven approach, where a priori criteria are used to score adherence to dietary recommendations and a data-driven approach is used to derive weights of the components.
Using the dietary index approach, we tested the hypothesis that consuming a diet consistent with the current DGA recommendations is associated with a reduced progression of atherosclerosis in postmenopausal women with established coronary artery atherosclerosis. Given the aforementioned limitations of the dietary index approach, we took a novel hybrid approach with a priori–based criteria to score adherence to dietary recommendations and with data-driven weights on each component of a dietary index.
SUBJECTS AND METHODS
Subjects
The subjects in this study were participants in the Estrogen Replacement and Atherosclerosis (ERA) study—a randomized, double-blind, placebo-controlled trial of hormone replacement therapy in postmenopausal women with established heart disease. The study design and primary results have been reported previously (20). Briefly, 309 postmenopausal women aged <80 y who were not receiving estrogen-replacement treatments and had one or more segments with an epicardial coronary stenosis of ≥30% of the luminal diameter were eligible for the study. They were enrolled in 1995–1996 from 6 US sites: the University of Alabama, Birmingham, AL; Carolinas Medical Center, Charlotte, NC; Moses Cone Health System, Greensboro, NC; Hartford Hospital, Hartford, CT; Forsyth Memorial Hospital, Winston-Salem, NC; and Wake Forest University School of Medicine, Winston-Salem, NC. Women with one or more of the following conditions were excluded from the trial: known or suspected breast or endometrial carcinoma, previous or planned coronary bypass surgery, a history of deep vein thrombosis or pulmonary embolism, symptomatic gallstones, a serum aspartate aminotransferase concentration >1.5 times normal, a fasting triacylglycerol concentration >400 mg/dL, a serum creatinine concentration >2.0 mg/dL, left main stenosis >70%, uncontrolled hypertension, or uncontrolled diabetes. Eligible subjects were randomly assigned to receive 1) 0.625 mg conjugated equine estrogen, 2) 0.625 mg conjugated equine estrogen plus 2.5 mg medroxyprogesterone acetate, or 3) placebo. The subjects were followed for a mean (±SD) 3.3 ± 0.6 y. Neither treatment had a significant effect on the progression of coronary atherosclerosis (20). The institutional review board at each participating center approved all study procedures, and all participants in the ERA study gave informed consent.
At baseline, the participants completed standard interviews on drug use, diseases history, and physical activity frequency (Physical Activity Scale for Elderly; PASE) (21, 22); completed a semiquantitative food-frequency questionnaire (FFQ) to assess dietary intake (23, 24); and underwent clinical examinations and laboratory testing to measure cardiovascular disease risk factors, sex hormones, and inflammatory biomarkers.
For the current analyses, we excluded 61 women for whom follow-up coronary angiography was not available (including 12 women who died before follow-up angiography). Also excluded were 24 women who either did not complete the FFQ at baseline (n = 13), reported extreme energy intakes (<500 or >4500 kcal/d; n = 7), or had missing information of factors needed for calculation of a dietary index or for statistical adjustment (weight, height, physical activity, and education status; n = 4). For this study, complete data were available for 224 women. Total energy intake and dietary index components were not significantly different between the included and excluded subjects among those with dietary data (P > 0.2). No significant differences were observed for other baseline characteristics, except for aspirin use (54% and 73% for the included and excluded, respectively; P = 0.003).
Dietary assessment and DGAI
Habitual food consumption and nutrient intakes were assessed at baseline with an interviewer-administered 126-item FFQ (25). Participants were asked to choose 1 of 9 categories to indicate how often, on average, they had consumed given amounts of various specified foods during the past year. Other questions asked about supplements, including multivitamin supplements, and specific foods, including breakfast cereals, margarine, and vegetable oil used for frying or baking. Nutrient intakes were calculated by multiplying the reported frequency of consumption for each food item with each prespecified portion size and the nutrient composition for that item. The reproducibility and validity of measuring dietary variables were described in detail previously (23, 24). For example, in our study population, the correlation between dietary and plasma n−3 long-chain polyunsaturated fatty acids was 0.41 (26). This value is similar to that observed previously in a study of questionnaire validation (27).
The DGAI is an index that measures adherence to 20 key dietary recommendations in the 2005 DGA (1). Components related to nondietary behaviors, such as food safety, or those for special populations, such as lactating or pregnant women, were not included in the DGAI. We modified the original DGAI developed by Fogli-Cawley et al (14) so that each component has a continuous score from 0 to 1, instead of discrete scores of 0, 0.5, or 1 (Table 1 for the algorithm).
TABLE 1.
Criteria of 20 Dietary Guidelines for Americans Adherence Index (DGAI) components for individuals with an 1800-kcal/d caloric requirement to maintain body weight1
| Criteria for diet with an 1800-kcal/d EER3 |
||
| DGAI components2 | 0 point | 1.0 point |
| DGAI score4 | ||
| Food Group score5 | ||
| Dark green vegetables (cups/wk) | 0 | ≥3.0 |
| Orange vegetables (cups/wk) | 0 | ≥2.0 |
| Legumes (cups/wk) | 0 | ≥3.0 |
| Other vegetables (cups/wk) | 0 | ≥6.5 |
| Starchy vegetables (cups/wk)6 | 0 | 2.5 |
| Fruit (cups/d) | 0 | ≥6.5 |
| Variety (no. of components)7 | 0 | 6 |
| Meat and beans (oz/d) 6 | 0 | 4.5 |
| Dairy products (cups/d)6 | 0 | 2.5 |
| All grains (oz/d)6 | 0 | 5.5 |
| Added sugar (% of energy) | ≥7.5% | ≤5.0% |
| Healthy Choice score8 | ||
| Whole grains (% of grains) | 0% | ≥50% |
| Fiber intake (g/1000 kcal/d) | 0 | ≥14 |
| Total fat (% of energy) | ≤10%, ≥45% | ≥20%, ≤35% |
| Saturated fat (% of energy) | ≥15% | ≤10% |
| trans Fat (% of energy) | ≥1.5% | ≤1% |
| Cholesterol intake (mg/d) | ≥450 | ≤300 |
| Low-fat products (%)9 | 0% | ≥75% |
| Sodium intake (mg/d) | ≥3450 | ≤2300 |
| Alcohol consumption (drinks/d) | ≥1.5 | ≤1.0 |
The DGAI was calculated as the sum of 20 components with a possible score of 0 to 20; EER, estimated energy requirement per day calculated on the basis of formula provided from National Research Council (28).
1 cup = 237 mL (US), 0.946 cup in a metric unit; 1 kcal = 4.184 kJ; 1 oz = 28.35 g; 1 drink = 355 mL regular beer, 118 mL wine, or 45 mL distilled spirits.
Criteria of the food group components were determined on the basis of EER; the DGA determines 10 different criteria for 10 daily EER levels of 1400–3200 kcal incremented by 200 kcal (5.86–13.39 MJ by 0.84 MJ). The criteria based on intermediate EER were proportionately determined between the 2 criteria based on the EERs provided by the DGA; for example, the criteria for individuals with an EER of 1900 kcal were set middle values of the criteria for EER of 1800 and 2000 kcal. The criteria for individuals with an EER <1400 or >3200 kcal were set equal to criteria for individuals with daily EERs of 1400 and 3200 kcal, respectively. Intermediate intakes between criteria for 0 and 1.0 points were scored proportionately between 0 and 1.0.
Total score of the 20 components.
Total score of the 11 components below.
Overconsumption more than recommendation was penalized. Consumption ≥1.25 times recommendations was scored as 0.5. Intermediate consumption was scored proportionally between 0.5 and 1.0.
Variety is determined as the sum of scores of 6 fruit and vegetable components.
Total score of the 9 components below.
Adherence to recommendations of lean-fat meat and low-fat dairy product choice was the sum of adherences, calculated separately. For each adherence to recommendations of low-fat dairy product choices (%) and lean fat (%) meat product choices: ≥75% assigns 0.5 point to an individual, 0% assigns 0 points to an individual; intermediate %, proportionate score between 0 and 0.5. Sum of the 2 scores had a possible range from 0 to 1 and is considered the component score (14).
Of the 20 DGAI components, each of the 11 “Food Group” items was assigned a score of 0–1.0 on the basis of adherence to each recommendation for food group consumption, depending on the estimated energy requirement of each individual. DGA defines patterns of the recommendations for 10 levels of daily energy requirements to maintain body weight. Therefore, first, the caloric need of each subject was estimated by using a published equation for Estimated Energy Requirement (EER), which is dependent on sex, height, weight, age, and physical activity level (28). Physical activity level was determined by responses to the PASE questionnaire. Second, based on the calculated EER and calorie-specific DGA diet pattern, criteria of adherence to food group recommendations were assigned to each subject. Finally, adherence to each recommendation of 11 items was scored proportionally with a value that ranged from 0 to 1. The sum of the 11 scores was considered a measure of adherence to the “Food Group” recommendations. For the food groups starchy vegetables, meat and beans, total grains, and dairy products, overconsumption was penalized by reducing the score proportionally for scores up to 1.25 times higher than recommended consumption. A maximum penalty of 0.5 was assigned to those consuming ≥1.25 times the recommendation. Nine “Healthy Choice” items assessed adherence to recommendations of nutrient intake. The recommended intake of sodium for elderly and hypertensive persons is <1500 mg/d The change in the criterion for sodium intake component to 1500 mg/d did not appreciably change the results of our analyses. Adherence to each item was scored proportionally by a value that ranged from 0 to 1 on the basis of predetermined cutoffs of the 9 recommendations that were not based on estimated energy requirements. The maximum possible DGAI score is 20.
Outcome measurement
At baseline and at the end of follow-up, each woman underwent coronary angiography using a standardized protocol to measure luminal diameters of up to 10 proximal epicardial segments (mean: 9.3 segments per women) and the degree of stenosis as a percentage of the reference diameter (29). Review and analyses of the paired films were performed by using a previously validated system of cine projectors (SME-3500; Sony, Park Ridge, NJ) and software (QCAPlus; Sanders Data Systems, Palo Alto, CA), giving a mean intraoperator difference equal to 0.02 mm between blinded duplicate measurements of minimum diameter for vessels with lesions. All measurements were performed by operators blinded to the temporal sequence of the films. Segments totally occluded at baseline or affected by intervening coronary interventions were excluded from the analyses. In total, repeated measures of 2054 segments were used for the analyses.
Statistical analysis
All statistical analyses were performed by using SAS software (version 9.13; SAS Inc, Cary, NC). For descriptive purposes, baseline characteristics of the study population are presented as means or frequencies by DGAI tertile categories. Associations between the total DGAI score and subjects’ characteristics were tested by age-adjusted regression analyses, in which the DGAI score and each variable was modeled as a dependent variable and independent variable, respectively; each categorical variable was converted to dummy variables as appropriate. Dietary factors including total energy intake, each dietary component of DGAI, and scores of food group and healthy choice components were described by means and SDs in strata of the 3 DGAI tertile categories. Associations between the total DGAI score and each of the dietary factors were described by partial Spearman correlation coefficients adjusted for age.
To examine the association between the total DGAI score and the change in minimal diameter over the period of follow-up, repeated-measures regression analyses were performed with a mixed-model procedure using PROC MIXED; the technique accounted for multiple correlated observations of each individual (10 coronary segments and 2 time points) (30). The variance-covariance matrix of the 20 correlated observations was estimated from the sum-of-squares cross-products matrix under the unstructured assumption, which provided the best goodness-of-fit compared with other common assumptions, according to Akaike Information Criteria. The main independent variable was the standardized DGAI score. Standardization allows interpretation of the regression slope as the strength of association based on a 1-SD difference in DGAI score. The DGAI variable was included in the model as a single variable (DGAI) and cross-product term with time (time × DGAI). The regression coefficient of time × DGAI allows testing associations between the DGAI score and change of minimal diameters over time. The analyses were adjusted for each coronary segment and selected covariates. The selection of covariates was determined by examining a change in the regression coefficient of time × DGAI (31). Briefly, we excluded covariates one-by-one in a backward manner from the fully adjusted model. The covariate with the least influence on the regression coefficient was excluded from the model. The reduced model was considered the next adjusted model. This repetitive procedure was stopped when exclusion of any one of covariates yielded change in the regression coefficient by >0.002. The model with the remaining covariates was considered the final model. The selected covariates were as follows: age categories ( <60, 60–64.9, 65–69.9, or ≥70 y), study site [Charlotte, NC (referent); Winston-Salem, NC; Hartford, CT; Greensboro, NC), reported use of a cholesterol-lowering drug (yes or no), education status (less than high school, high school level, or at least college level), frequency of walking (days per week, continuous), current smoking status (yes or no), reported history of chest pain (yes or no), total energy intake (quartiles), systolic blood pressure (continuous), and casual glucose concentration (continuous). Other covariates, which did not meet our selection criteria, included the following: experimental arms (hormone replacement therapies), race-ethnicity, physical activity level determined by PASE, cumulative exposure of smoking (pack-years, continuous), other reports of disease diagnoses and gynecological or cardiovascular surgeries, other reports of dietary supplement or medication use for cardiovascular health, age of menopause, body mass index (in kg/m2), waist circumference, duration of follow-up, blood pressure, fasting glucose, fasting insulin, serum triglyceride, serum HDL cholesterol, and sex hormones including estradiol, estrogen, and progestin. We examined statistical interactions involving selected factors, such as coronary segments, medications and clinical conditions, including a cross-product term in the same regression model. Because no significant interactions were observed, the results are not presented.
Weighted DGAI
To address the limitations of the DGAI associated with assigning an equal weight to each of 20 components, we created a modified index by assigning a different weight based on the relation of each component to atherosclerosis progression. To develop the wDGAI (Step I), we first performed a regression analysis including all the DGAI components, except for a variety component, which was created from a score of the sum of fruit and vegetables components. The model included the 19 DGAI component scores (DGAIi, i = 1–19) and each time interaction (time × DGAIi), in addition to the selected covariates. Second, 19 regression coefficients of the time × DGAIi were saved as weights; a positive weight indicates that better adherence to the component was associated with slower atherosclerosis progression, with control for adherence to the other components; a negative weight indicates that better adherence to the component resulted in more atherosclerosis progression. Third, the weights were multiplied by the DGAI component scores of each individual. Finally, the products were summed to create the wDGAI. To allow for a direct comparison of the wDGAI and the DGAI, both of these indexes were standardized to let the mean and SD equal 0 and 1, respectively.
To examine the association between the modified index and atherosclerosis progression (Step II), we repeated the analyses described above using the wDGAI in place of the original DGAI. The regression analysis was repeated by including both the DGAI and wDGAI to examine which one had a higher predictive capability of minimal diameter change. SEs of the regression coefficients should incorporate the uncertainty of the first step (development of the wDGAI) and the second step (application of the wDGAI). This need was addressed by a nested bootstrap technique to obtain pooled SEs from the 2 steps as described elsewhere (see Supplemental Tables 1 and 2 under “Supplemental data” in the online issue) (32–36). Cross-validation analyses were repeated for the 2-step approaches, where the 2 steps were set independent by iteratively splitting our study population (37). Although the association was attenuated by holding independency between the 2 steps, our results were not affected substantially (see Supplemental Table 3 under “Supplemental data” in the online issue).
Our study had a statistical power of 0.67, 0.80, and 0.89 to detect regression coefficients of 0.030, 0.035, and 0.040, respectively, based on a 1-SD difference in DGAI score, with a pooled SD of 0.20 obtained from a residual matrix of the repeated-measures regression model. The regression coefficient of ≥0.030 for the interaction term (time × DGAI or time × wDGAI) would be clinically meaningful based on past studies using quantitative coronary angiography (38). A possible increase in type 1 error due to a small sample size was unlikely in the results of our regression analyses according to Hotelling-Lawley-McKeon approximation of F statistics in a study with small sample size (30). Influence analyses were performed by calculating Cook’s distance (30, 39). The exclusion of those with a high Cook’s distance did not affect our conclusion (see Supplemental Table 4 under “Supplemental data” in the online issue). According to bias estimates by bootstrapping analyses, the bias of our estimates of regression coefficients was negligible (see Supplemental Tables 1 and 2 under “Supplemental data” in the online issue).
RESULTS
Age-adjusted descriptive subject characteristics are presented according to tertile categories of DGAI score in Table 2. None of the women reported adherence to all of the recommendations. The mean (±SD) DGAI was 14.1 ± 2.1, which indicated that, on average, our study population was adherent to about two-thirds of the 20 DGA key dietary recommendations. Compared with those with lower DGAI score, those with higher DGAI scores were more likely to be older (P < 0.03), have a college education (P = 0.002), use cholesterol lowering drugs (P = 0.01), and walk more frequently (P = 0.03). They were less likely to be current smokers (P < 0.001).
TABLE 2.
Baseline characteristics for postmenopausal women with established coronary artery atherosclerosis enrolled in the Estrogen Replacement and Atherosclerosis Trial (n = 224) by tertile of the Dietary Guidelines for Americans Adherence Index (DGAI)
| DGAI tertile group |
||||
| First | Second | Third | P1 | |
| Median DGAI | 12.2 (8.0–13.5) | 14.3 (13.5–15.1) | 16.1 (15.1–19.0) | |
| Number of subjects | 73 | 73 | 73 | |
| Number of coronary segments | 704 | 716 | 715 | |
| Age (y)2 | 64.1 ± 6.9 | 66.3 ± 6.8 | 66.6 ± 6.9 | 0.026 |
| Follow-up duration (y)2 | 3.3 ± 0.01 | 3.4 ± 0.01 | 3.4 ± 0.01 | 0.805 |
| Exam sites [n (%)] | ||||
| Charlotte | 23 (32.0) | 31 (43.1) | 18 (25.0) | |
| Winston-Salem | 31 (40.8) | 20 (26.3) | 25 (32.9) | |
| Hartford | 5 (16.7) | 9 (30.0) | 16 (53.3) | |
| Greensboro | 15 (32.6) | 15 (32.6) | 16 (34.8) | 0.06 |
| Experimental arms [n (%)] | ||||
| Placebo | 23 (30.7) | 29 (38.7) | 23 (30.7) | |
| Equine estrogen | 30 (39.0) | 20 (26.0) | 27 (35.1) | |
| Progesterone and equine estrogen | 21 (29.2) | 26 (36.1) | 25 (34.7) | 0.47 |
| Education [n (%)] | ||||
| Less than high school | 40 (47.1) | 19 (22.4) | 26 (30.6) | |
| High school level education | 26 (29.6) | 34 (38.6) | 28 (31.8) | |
| College level education | 8 (15.7) | 22 (43.1) | 21 (41.2) | <0.01 |
| Current smoking [n (%)] | ||||
| Yes | 29 (61.7) | 12 (25.5) | 6 (12.8) | |
| No | 45 (25.4) | 63 (35.6) | 69 (39.0) | <0.01 |
| Cholesterol-lowering drug [n (%)] | ||||
| Yes | 23 (29.5) | 19 (24.4) | 36 (46.2) | |
| No | 51 (34.9) | 56 (38.4) | 39 (26.7) | 0.01 |
| History of chest pain [n (%)] | ||||
| Yes | 43 (36.8) | 39 (33.3) | 35 (29.9) | |
| No | 31 (29.0) | 36 (33.6) | 40 (37.4) | 0.376 |
| Walking frequency (d/wk)2 | 2.4 ± 0.3 | 3.3 ± 0.3 | 3.2 ± 0.3 | 0.026 |
| BMI (kg/m2)2 | 29.2 ± 0.8 | 30.9 ± 0.8 | 29.4 ± 0.8 | 0.860 |
| Waist circumference (cm)2 | 93.9 ± 1.8 | 93.6 ± 1.8 | 92.9 ± 1.8 | 0.696 |
| Systolic blood pressure (mm Hg)2 | 135.4 ± 2.0 | 132.4 ± 2.0 | 133.4 ± 2.0 | 0.469 |
| Diastolic blood pressure (mm Hg)2 | 74.5 ± 1.0 | 74.6 ± 1.0 | 73.6 ± 1.0 | 0.527 |
| Total cholesterol (mg/dL)23 | 217.0 ± 4.9 | 214.8 ± 4.8 | 219.3 ± 4.9 | 0.749 |
| Triglyceride (mg/dL)23 | 180.8 ± 12.5 | 192.3 ± 12.4 | 209.1 ± 12.5 | 0.113 |
| HDL cholesterol (mg/dL)23 | 45.4 ± 1.4 | 45.4 ± 1.4 | 42.4 ± 1.4 | 0.132 |
| LDL cholesterol (mg/dL)3 | 138.7 ± 4.5 | 131.9 ± 4.5 | 135.6 ± 4.5 | 0.617 |
| Glucose concentration (mg/dL)2 | 117.3 ± 5.1 | 125.0 ± 5.1 | 116.0 ± 5.1 | 0.874 |
P for associations with the DGAI score tested by age-adjusted linear regression for continuous variables except for age and age-adjusted multivariate analysis of variance for categorical variables.
Continuous variables are presented as mean ± SE and are adjusted for age categories (<60, ≥60 to <65, ≥65 to <70, and ≥70 y).
Three, 2, 3, and 17 individuals were missing measures of total cholesterol, triglyceride, HDL-cholesterol, and LDL-cholesterol concentrations, respectively.
At least 70% of the study subjects adhered to the recommendations (score > 0.9) for “other vegetables” (which included iceberg lettuce and tomatoes), dietary cholesterol and sodium, percentage of energy from total fat, percentage energy of from saturated fat, and alcohol consumption (Table 3). Few subjects adhered to the recommendations for starchy vegetables and dairy products, with <10% of subjects receiving scores >0.9. The low scores for starchy vegetables were due to overconsumption, whereas the low scores for dairy products were due to underconsumption. Weak correlations were observed between the overall DGAI score and intakes of many of the component nutrients and foods, including starchy vegetables, percentage of energy from total fat, sodium intake, and alcohol consumption. This is likely attributable to the limited variability for the component or to less influence on total score relative to the other components.
TABLE 3.
Age-adjusted Dietary Guidelines for Americans Adherence Index (DGAI) component scores > 0.9 and Spearman correlations between the total DGAI score and dietary variables in postmenopausal women with established coronary artery atherosclerosis enrolled in the Estrogen Replacement and Atherosclerosis Trial (n = 224)1
| DGAI tertile group |
|||||
| DGAI component2 | First | Second | Third | >0.9 points3 | r4 |
| DGAI score median | 12.2 (8.0–13.5) | 14.3 (13.5–15.1) | 16.1 (15.1–19.0) | 1.0 | |
| Energy intake (kcal/d) | 1525.8 ± 67.2 | 1711.4 ± 66.7 | 1707.1 ± 66.8 | 0.17 | |
| Estimated energy requirement (kcal/d) | 1849.5 ± 28.6 | 1849.6 ± 28.4 | 1833.9 ± 28.4 | 0.02 | |
| Food Group point | 5.8 ± 0.1 | 7.3 ± 0.1 | 8.3 ± 0.1 | 0.80 | |
| Dark green vegetables (cups/wk) | 2.2 ± 0.4 | 3.3 ± 0.4 | 5.4 ± 0.4 | 115 (51.3) | 0.44 |
| Orange vegetables (cups/wk) | 1.2 ± 0.3 | 2.2 ± 0.3 | 4.5 ± 0.3 | 111 (49.6) | 0.49 |
| Legumes (cups/wk) | 1.4 ± 0.2 | 2.0 ± 0.2 | 2.9 ± 0.2 | 68 (30.4) | 0.36 |
| Other vegetables (cups/wk) | 11.3 ± 1.0 | 15.6 ± 1.0 | 20.9 ± 1.0 | 202 (90.2) | 0.30 |
| Starchy vegetables (cups/wk) | 4.6 ± 0.3 | 4.8 ± 0.3 | 4.5 ± 0.3 | 15 (6.7) | 0.09 |
| Fruit (cups/d) | 1.3 ± 0.2 | 2.6 ± 0.2 | 3.0 ± 0.2 | 142 (63.4) | 0.57 |
| Variety (no. of components) | 3.6 ± 0.1 | 4.5 ± 0.1 | 5.0 ± 0.1 | 26 (11.6) | 0.70 |
| Meat and beans (oz/d) | 4.4 ± 0.3 | 4.8 ± 0.3 | 5.0 ± 0.3 | 38 (17.0) | 0.34 |
| Dairy products (cups/d) | 0.9 ± 0.1 | 1.2 ± 0.1 | 1.3 ± 0.1 | 11 (4.9) | 0.31 |
| All grains (oz/d) | 3.5 ± 0.2 | 3.9 ± 0.2 | 4.4 ± 0.2 | 30 (13.4) | 0.24 |
| Added sugar (% of energy) | 13.5 ± 1.0 | 10.2 ± 0.9 | 7.9 ± 0.9 | 70 (31.3) | 0.34 |
| Healthy Choice points | 6.0 ± 0.1 | 7.0 ± 0.1 | 8.0 ± 0.1 | 0.75 | |
| Whole grains (% of grains) | 19.9 ± 2.2 | 25.8 ± 2.2 | 45.1 ± 2.2 | 52 (23.2) | 0.52 |
| Fiber intake (g/1000 kcal/d) | 10.0 ± 0.4 | 12.9 ± 0.4 | 15.0 ± 0.4 | 100 (44.6) | 0.62 |
| Total fat (% of energy) | 31.0 ± 0.7 | 27.1 ± 0.7 | 23.8 ± 0.7 | 176 (78.6) | 0.11 |
| Saturated fat (% of energy) | 10.7 ± 0.3 | 8.7 ± 0.3 | 7.4 ± 0.3 | 166 (74.1) | 0.50 |
| trans Fat (% of energy) | 2.1 ± 0.1 | 1.6 ± 0.1 | 1.1 ± 0.1 | 54 (24.1) | 0.61 |
| Cholesterol intake (mg/d) | 223.4 ± 13.0 | 221.2 ± 12.9 | 194.6 ± 12.9 | 199 (88.8) | 0.23 |
| Low-fat products (%) | 40.6 ± 2.3 | 56.1 ± 2.3 | 69.8 ± 2.3 | 38 (17.0) | 0.59 |
| Sodium intake (mg/d) | 1761.9 ± 83.8 | 1995.1 ± 83.2 | 1956.1 ± 83.3 | 179 (79.9) | 0.00 |
| Alcohol (drinks/d) | 0.1 ± 0.0 | 0.1 ± 0.0 | 0.1 ± 0.0 | 218 (97.3) | −0.06 |
A brief description of DGAI, Food Group, and Healthy Choice are given in Subjects and Methods and provided by Fogli-Cawley et al (14).
1 cup = 237 mL (US), 0.946 cup in a metric unit; 1 kcal = 4.184 kJ; 1 oz = 28.35 g; 1 drink = 355 mL regular beer, 118 mL wine, or 45 mL distilled spirits.
Values are the number of subjects with a DGAI component score > 0.9 points; percentages in parentheses.
Age-adjusted Spearman correlations with the total DGAI score; >|0.14| was statistically significant (P > 0.05).
The weights assigned to the index components are shown in Table 4. By definition, 1 is considered an equal weight for each component to calculate the original DGAI. The regression coefficients between the DGAI components and atherosclerosis progression are also presented as weights assigned to each component to calculate the wDGAI. None of the regression coefficients was statistically significantly different from 0 when entered simultaneously in the regression model, and they varied among the DGAI components, ranging from −0.068 (fruit) to 0.153 (% of energy from total fat). Adherence to recommendations of whole grains, total fat intake, and cholesterol intake showed associations with slower atherosclerosis progression; a difference in the component score of 1 was related to less narrowing of the coronary arteries by 0.076 mm (SE = 0.042, P = 0.08), 0.153 mm (SE = 0.079, P = 0.05), and 0.0125 mm (SE = 0.069, P = 0.07).
TABLE 4.
Weights assigned to each component of the original Dietary Guidelines for Americans Adherence Index (DGAI) and the weighted DGAI (wDGAI) in postmenopausal women with established coronary artery atherosclerosis enrolled in the Estrogen Replacement and Atherosclerosis Trial (n = 224)
| DGAI component | Original DGAI | wDGAI1 |
| Dark green vegetables | 1 | 0.003 (0.047) |
| Orange vegetables | 1 | −0.027 (0.041) |
| Beans and legumes | 1 | 0.002 (0.046) |
| Starchy vegetables | 1 | −0.019 (0.100) |
| Other vegetables | 1 | 0.028 (0.074) |
| Fruit | 1 | −0.068 (0.062) |
| Variety | 1 | 02 |
| Meat | 1 | −0.004 (0.065) |
| Dairy products | 1 | 0.008 (0.060) |
| All grain | 1 | 0.033 (0.064) |
| Added sugar (% of energy) | 1 | 0.019 (0.032) |
| Whole grains (% of grains) | 1 | 0.076 (0.042) |
| Fiber intake (per 1000 kcal) | 1 | 0.014 (0.113) |
| Total fat (% of energy) | 1 | 0.153 (0.079) |
| Saturated fat (% of energy) | 1 | 0.019 (0.075) |
| trans Fat (% of energy) | 1 | −0.029 (0.036) |
| Cholesterol | 1 | 0.125 (0.069) |
| Low-fat products (%) | 1 | −0.032 (0.066) |
| Sodium intake | 1 | −0.014 (0.065) |
| Alcohol | 1 | 0.031 (0.110) |
| Possible total score3 | 0–20 | −0.19–0.51 |
| Observed value4 | 14.1 (8.0–19.0) | 0.21 (0.12–0.38) |
Weights are regression coefficients (SE) estimated by repeated-measures regression analysis including minimal diameter (outcome), all components together, and selected covariates of age categories (<60, 60–64.9, 65–69.9, or ≥70 y), study site (3 levels), education status (3 levels), frequency of walking (days per week, continuous), cigarette smoking status (yes or no), energy intake (quartiles), systolic blood pressure (continuous), casual glucose concentration (continuous), self-report of cholesterol-lowering drug use (yes or no), and self-report of chest pain (yes or no).
A variety component was not included because it is the sum of the 6 components of vegetables and fruit.
The ranges are based on the assignment of 0 or 1 points to all components.
Values are means; ranges in parentheses.
The regression analyses yielded no significant association between the original DGAI and minimal diameter change over the period of the study follow-up (P = 0.44) (Table 5). In contrast, the wDGAI calculated from the sum of products of the weights and component score was significantly associated with slower progression of atherosclerosis. A 1-SD difference in the wDGAI score was related to less narrowing of the coronary arteries by 0.049 mm (bootstrap SE = 0.017, P = 0.004). The similar estimates of regression coefficients for wDGAI were obtained, even when both the original DGAI and wDGAI scores were included in the same regression model.
TABLE 5.
Results of regression coefficients (SE) from repeated-measures regression analyses to examine the association between each of the standardized Dietary Guidelines for Americans Adherence Index (DGAI) and weighted DGAI (wDGAI) values and changes in minimal diameters of coronary arteries in the postmenopausal women with established coronary artery atherosclerosis enrolled in the Estrogen Replacement and Atherosclerosis Trial (n = 224)
| Original DGAI |
wDGAI |
|||
| β (SE)1 | P | β (SE)1 | P | |
| Crude | −0.002 (0.013) | 0.897 | −0.029 (0.015) | 0.059 |
| Multivariate model 12 | −0.007 (0.014) | 0.636 | −0.045 (0.017) | 0.010 |
| Multivariate model 23 | −0.011 (0.014) | 0.441 | −0.049 (0.017) | 0.004 |
SEs of the coefficients for wDGAI and DGAI were estimated by 2-stage (nested) bootstrap analyses.
Multivariate model 1 included age categories (<60, 60–64.9, 65–69.9, or ≥70 y), study site (3 levels), education status (3 levels), frequency of walking (days per week, continuous), cigarette smoking status (yes or no), energy intake (quartiles).
Multivariate model 2 included the variables in model 1 plus systolic blood pressure (continuous), casual glucose concentration (continuous), self-report of cholesterol-lowering drug use (yes or no), and self-report of chest pain (yes or no).
DISCUSSION
No significant association was identified between the DGAI as a measure of diet consistent with the 2005 DGA and narrowing of coronary arteries after a mean 3.3-y follow-up period in postmenopausal women with established coronary artery atherosclerosis. However, adherence to an index that was weighted by dietary recommendations based on their association with heart disease risk (wDGAI) was significantly associated with reduced progression of atherosclerosis. These findings suggest that certain dietary recommendations may help limit atherosclerotic lesion progression, but assuming equity of associations between all dietary recommendations may limit our ability to identify the relation between the DGA and chronic disease prevention accurately.
The differential weighting is a novel approach to address the limitation of dietary indexes to assess healthy dietary patterns. Past studies taking the dietary index approach used indexes defined on the basis of predetermined criteria and applied equal weights for each of the individual components (17). Equality in weights is biologically implausible, because adherence to each dietary recommendation is not necessarily of equal importance for the reduction of specific disease progression or overall chronic disease risk. In our data set, this apparent weakness with respect to atherosclerotic lesion progression was addressed by assigning data-driven weights to the index components on the basis of results from regression analyses. Therefore, because we combined a priori criterion–based and data-driven approaches to calculate a dietary index, our approach is a novel hybrid of these 2 types of dietary pattern analysis. The value of such an approach will ultimately be determined by the consistency of findings among independent cohorts.
Of the components of the overall index, adherence to individual recommendations for total fat, whole grain, and cholesterol recommendations were found to be inversely associated with atherosclerosis progression. The observation of whole grain was expected and consistent with past studies using diseased populations (40–42), whereas past observations relating total fat and cholesterol intake to progression are inconsistent and controversial (5). From these data, we cannot determine whether the putative factors are these dietary components or if these dietary components are surrogate markers for other dietary components and are more amenable to detection with an FFQ. The collinearity of dietary components is well established and potentially contributes to this uncertainty (43, 44).
In our study, no women reported complete adherence to all dietary recommendations. This is consistent with past studies suggesting a less than robust compliance with the diet and lifestyle recommendations and individual variability (45–49). Selected dietary recommendations have been shown to improve heart disease risk factors, such as achieving and maintaining recommended plasma lipid and lipoprotein concentrations, BMI, and blood pressure, in clinical trials of dietary counseling or other types of extensive support to ensure compliance (50, 51). Nonetheless, despite established efficacy, nationwide studies have shown that adherence to these recommendations is poor. National Health and Nutrition Examination Survey data indicated that only 20% of hypertensive adults committed to a diet concordant with the Dietary Approaches to Stop Hypertension (DASH) eating pattern highlighted in the 2005 DGA (52). Similar observations were made among the Women’s Health Initiative participants (45–47). Reports from the Nurses’ Health Study and the 2000 National Behavioral Risk Factor Surveillance observed that only 3% of each study population was categorized to a low-risk or a healthy lifestyle group according to criteria based on both dietary and nondietary (eg, smoking and physical activity) (48, 49) factors. These observations and ours indicate the need for clinical practice and public health policy to foster improved adherence to dietary recommendations and other healthy lifestyle behaviors.
The main strength of our study was the use of longitudinal data from angiography as a measure of atherosclerosis progression rather than surrogate measures of heart disease, such as conventional risk factors. An additional strength of this work was the use of the wDGAI. This comprehensive measure of adherence to current dietary recommendations was designed to be independent of actual energy intake and was previously shown to favorably predict chronic disease risk factors in healthy adults (14, 15). Last, by generating and using individual weight factors for each component of the index, we were able to directly address the major limitation of the criterion-based indexes that rely on equal weights for each component (17).
The major limitation of our study was the relatively small sample size, which may have resulted in marginal statistical power to detect modest associations, were they to exist, with diet. We were also unable to assess potential residual confounding in the data set because of complex medical conditions. However, our sensitivity analyses and confounder selection indicated that this limitation was not substantial. The accuracy and precision of FFQ measures could influence the results, because FFQ use is not necessarily suitable for estimating absolute intakes. As with all FFQ data, the results should be interpreted cautiously. Another limitation was the potential lack of generalizability of our findings to other clinical populations with diseases and general populations without diseases because of the highly selective nature of the study cohort. Our study cohort was derived from a randomized trial of estrogen replacement therapy; therefore, the study subjects were highly selected based on a variety of inclusion and exclusion criteria and included only those participants who had established coronary artery atherosclerosis and completed the repeated measures of angiography (20). Finally, although not unique to different cohorts, our study population had an unbalanced distribution of some DGAI components. For example, 97% of subjects received > 0.9 points for the recommendation of alcohol consumption. Thus, adherence to this component could not be evaluated for an association with atherosclerosis progression in our population because of insufficient variability among study subjects. On the basis of these limitations, it will be of interest to see this approach applied to other independent data sets.
In conclusion, as assessed by using the wDGAI score, but not the DGAI score, postmenopausal women with established coronary artery atherosclerosis consuming diets more, rather than less, consistent with the 2005 DGA showed a slower rate of atherosclerosis progression over a mean 3.3-y follow-up period. From a clinical and public health perspective, modifying the approach to analyze dietary data using the assumption that not all components have an equal weight in describing diet-disease relations may improve the predictive ability. Although the generalizability of the approach is an important limitation, our study supports the concept that not all dietary recommendations are equally related to disease progression. Our findings highlight the need for the development of more sophisticated approaches to the assessment of dietary recommendations on disease progression and other chronic disease outcomes.
Supplementary Material
Acknowledgments
We thank Georgia Saylor for data management.
The authors’ responsibilities were as follows—FI: conducted the statistical analyses and drafted the manuscript; and DMH: contributed to the data collection and preparation of the data for analyses. All authors designed the project, prepared the statistical strategies and discussion materials, and participated in editing the manuscript. None of the authors had a conflict of interest
REFERENCES
- 1.US Department of Agriculture and Department of Health and Human Services Dietary Guidelines for Americans. 6th ed Washington, DC: US Department of Agriculture and Department of Health and Human Services, 2005 [Google Scholar]
- 2.Lichtenstein AH, Appel LJ, Brands M, et al. Diet and lifestyle recommendations revision 2006: a scientific statement from the American Heart Association Nutrition Committee. Circulation 2006;114:82–96 [DOI] [PubMed] [Google Scholar]
- 3.Colquhoun D, Kostner K, Ferreira-Jardim A. Letter by Colquhoun et al regarding article “Effect size estimates of lifestyle and dietary changes on all-cause mortality in coronary artery disease patients: a systematic review.” Circulation 2006;114:e40.(author reply e41) [DOI] [PubMed] [Google Scholar]
- 4.Iestra JA, Kromhout D, van der Schouw YT, Grobbee DE, Boshuizen HC, van Staveren WA. Effect size estimates of lifestyle and dietary changes on all-cause mortality in coronary artery disease patients: a systematic review. Circulation 2005;112:924–34 [DOI] [PubMed] [Google Scholar]
- 5.Knopp RH, Retzlaff BM. Saturated fat prevents coronary artery disease? An American paradox. Am J Clin Nutr 2004;80:1102–3 [DOI] [PubMed] [Google Scholar]
- 6.Balady GJ, Williams MA, Ades PA, et al. Core components of cardiac rehabilitation/secondary prevention programs: 2007 update: a scientific statement from the American Heart Association Exercise, Cardiac Rehabilitation, and Prevention Committee, the Council on Clinical Cardiology; the Councils on Cardiovascular Nursing, Epidemiology and Prevention, and Nutrition, Physical Activity, and Metabolism; and the American Association of Cardiovascular and Pulmonary Rehabilitation. Circulation 2007;115:2675–82 [DOI] [PubMed] [Google Scholar]
- 7.Parikh P, McDaniel MC, Ashen MD, et al. Diets and cardiovascular disease, an evidence-based assessment. J Am Coll Cardiol 2005;45:1379–87 [DOI] [PubMed] [Google Scholar]
- 8.Zarraga IGE, Schwarz ER. Impact of dietary patterns and interventions on cardiovascular health. Circulation 2006;114:961–73 [DOI] [PubMed] [Google Scholar]
- 9.Hamer M, Steptoe A. Influence of specific nutrients on progression of atherosclerosis, vascular function, haemostasis and inflammation in coronary heart disease patients: a systematic review. Br J Nutr 2006;95:849–59 [DOI] [PubMed] [Google Scholar]
- 10.Rosamond W, Flegal K, Friday G, et al. Heart disease and stroke statistics—2007 update: a report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation 2007;115:e69–171 [DOI] [PubMed] [Google Scholar]
- 11.World Health Organization Avoiding heart attacks and strokes: don't be a victim—protect yourself. Geneva, Switzerland: WHO Press, 2005 [Google Scholar]
- 12.King JC. An evidence-based approach for establishing dietary guidelines. J Nutr 2007;137:480–3 [DOI] [PubMed] [Google Scholar]
- 13.Kennedy ET, Ohls J, Carlson S, Fleming K. The Healthy Eating Index: design and applications. J Am Diet Assoc 1995;95:1103–8 [DOI] [PubMed] [Google Scholar]
- 14.Fogli-Cawley JJ, Dwyer JT, Saltzman E, McCullough ML, Troy LM, Jacques PF. The 2005 Dietary Guidelines for Americans Adherence Index: development and application. J Nutr 2006;136:2908–15 [DOI] [PubMed] [Google Scholar]
- 15.Fogli-Cawley JJ, Dwyer JT, Saltzman E, et al. The 2005 Dietary Guidelines for Americans and insulin resistance in the Framingham Offspring Cohort. Diabetes Care 2007;30:817–22 [DOI] [PubMed] [Google Scholar]
- 16.Fung TT, Chiuve SE, McCullough ML, Rexrode KM, Logroscino G, Hu FB. Adherence to a DASH-style diet and risk of coronary heart disease and stroke in women. Arch Intern Med 2008;168:713–20 [DOI] [PubMed] [Google Scholar]
- 17.Waijers PM, Feskens EJ, Ocke MC. A critical review of predefined diet quality scores. Br J Nutr 2007;97:219–31 [DOI] [PubMed] [Google Scholar]
- 18.Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 2002;13:3–9 [DOI] [PubMed] [Google Scholar]
- 19.Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev 2004;62:177–203 [DOI] [PubMed] [Google Scholar]
- 20.Herrington DM, Reboussin DM, Brosnihan KB, et al. Effects of estrogen replacement on the progression of coronary-artery atherosclerosis. N Engl J Med 2000;343:522–9 [DOI] [PubMed] [Google Scholar]
- 21.Washburn RA, McAuley E, Katula J, Mihalko SL, Boileau RA. The Physical Activity Scale for the Elderly (PASE): evidence for validity. J Clin Epidemiol 1999;52:643–51 [DOI] [PubMed] [Google Scholar]
- 22.Schuit AJ, Schouten EG, Westerterp KR, Saris WHM. Validity of the physical activity scale for the elderly (PASE): according to energy expenditure assessed by the doubly labeled water method. J Clin Epidemiol 1997;50:541–6 [DOI] [PubMed] [Google Scholar]
- 23.Salvini S, Hunter DJ, Sampson L, et al. Food-based validation of a dietary questionnaire: the effects of week-to-week variation in food consumption. Int J Epidemiol 1989;18:858–67 [DOI] [PubMed] [Google Scholar]
- 24.Hu FB, Rimm E, Smith-Warner SA, et al. Reproducibility and validity of dietary patterns assessed with a food-frequency questionnaire. Am J Clin Nutr 1999;69:243–9 [DOI] [PubMed] [Google Scholar]
- 25.Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Litin LB, Willett WC. Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals. Am J Epidemiol 1992;135:1114–26 [DOI] [PubMed] [Google Scholar]
- 26.Erkkilä AT, Matthan NR, Herrington DM, Lichtenstein AH. Higher plasma docosahexaenoic acid is associated with reduced progression of coronary atherosclerosis in women with CAD. J Lipid Res 2006;47:2814–9 [DOI] [PubMed] [Google Scholar]
- 27.Hunter DJ, Rimm EB, Sacks FM, et al. Comparison of measures of fatty acid intake by subcutaneous fat aspirate, food frequency questionnaire, and diet records in a free-living population of US men. Am J Epidemiol 1992;135:418–27 [DOI] [PubMed] [Google Scholar]
- 28.Institute of Medicine Dietary Reference Intakes for energy, carbohydrate, fiber, fat, fatty acids, cholesterol, protein and amino acids (macronutrients). Washington, DC: National Academies Press, 2005 [Google Scholar]
- 29.Herrington DM, Reboussin DM, Klein KP, et al. The Estrogen Replacement and Atherosclerosis (ERA) Study: study design and baseline characteristics of the cohort. Control Clin Trials 2000;21:257–85 [DOI] [PubMed] [Google Scholar]
- 30.Schabenberger O, Wolfinger RD. The MIXED procedure. In: Jones A, Huddleston E, eds SAS/STAT 9.2 user's guide. Cary, NC: SAS Institute Inc, 2008:3885–4086 [Google Scholar]
- 31.Maldonado G, Greenland S. Simulation study of confounder-selection strategies. Am J Epidemiol 1993;138:923–36 [DOI] [PubMed] [Google Scholar]
- 32.Efron B. Bootstrap methods: another look at the Jackknife. Ann Statist 1979;7:1–26 [Google Scholar]
- 33.Henderson AR. The bootstrap: a technique for data-driven statistics. Using computer-intensive analyses to explore experimental data. Clin Chim Acta 2005;359:1–26 [DOI] [PubMed] [Google Scholar]
- 34.Ogden RT, Tarpey T. Estimation in regression models with externally estimated parameters. Biostatistics 2006;7:115–29 [DOI] [PubMed] [Google Scholar]
- 35.Maas CJM, Hox JJ. Robustness issues in multilevel regression analysis. Statistica Neerlandica 2004;58:127–37 [Google Scholar]
- 36.SAS Institute Inc Jackboot macro documentation. Cary, NC: SAS Institute, 1995 [Google Scholar]
- 37.Molinaro AM, Simon R, Pfeiffer RM. Prediction error estimation: a comparison of resampling methods. Bioinformatics 2005;21:3301–7 [DOI] [PubMed] [Google Scholar]
- 38.Balk EM, Karas RH, Jordan HS, Kupelnick B, Chew P, Lau J. Effects of statins on vascular structure and function: a systematic review. Am J Med 2004;117:775–90 [DOI] [PubMed] [Google Scholar]
- 39.Cook RD. Influential observations in linear regression. J Am Stat Assoc 1979;74:169–74 [Google Scholar]
- 40.Erkkilä AT, Herrington DM, Mozaffarian D, Lichtenstein AH. Cereal fiber and whole-grain intake are associated with reduced progression of coronary-artery atherosclerosis in postmenopausal women with coronary artery disease. Am Heart J 2005;150:94–101 [DOI] [PubMed] [Google Scholar]
- 41.Malik VS, Hu FB. Dietary prevention of atherosclerosis: go with whole grains. Am J Clin Nutr 2007;85:1444–5 [DOI] [PubMed] [Google Scholar]
- 42.Mellen PB, Liese AD, Tooze JA, Vitolins MZ, Wagenknecht LE, Herrington DM. Whole-grain intake and carotid artery atherosclerosis in a multiethnic cohort: the Insulin Resistance Atherosclerosis Study. Am J Clin Nutr 2007;85:1495–502 [DOI] [PubMed] [Google Scholar]
- 43.McGee D, Reed D, Yano K. The results of logistic analyses when the variables are highly correlated: an empirical example using diet and CHD incidence. J Chronic Dis 1984;37:713–9 [DOI] [PubMed] [Google Scholar]
- 44.Wax Y. Collinearity diagnosis for a relative risk regression analysis: an application to assessment of diet-cancer relationship in epidemiological studies. Stat Med 1992;11:1273–87 [DOI] [PubMed] [Google Scholar]
- 45.Howard BV, Van Horn L, Hsia J, et al. Low-fat dietary pattern and risk of cardiovascular disease: the Women's Health Initiative Randomized Controlled Dietary Modification Trial. JAMA 2006;295:655–66 [DOI] [PubMed] [Google Scholar]
- 46.Women's Health Initiative Study Group Dietary adherence in the Women's Health Initiative Dietary Modification Trial. J Am Diet Assoc 2004;104:654–8 [DOI] [PubMed] [Google Scholar]
- 47.Mozaffarian D, Rimm EB. Fish intake, contaminants, and human health: evaluating the risks and the benefits. JAMA 2006;296:1885–99 [DOI] [PubMed] [Google Scholar]
- 48.Stampfer MJ, Hu FB, Manson JE, Rimm EB, Willett WC. Primary prevention of coronary heart disease in women through diet and lifestyle. N Engl J Med 2000;343:16–22 [DOI] [PubMed] [Google Scholar]
- 49.Reeves MJ, Rafferty AP. Healthy lifestyle characteristics among adults in the United States, 2000. Arch Intern Med 2005;165:854–7 [DOI] [PubMed] [Google Scholar]
- 50.Champagne CM. Dietary interventions on blood pressure: the Dietary Approaches to Stop Hypertension (DASH) trials. Nutr Rev 2006;64:S53–6 [DOI] [PubMed] [Google Scholar]
- 51.Azadbakht L, Mirmiran P, Esmaillzadeh A, Azizi T, Azizi F. Beneficial effects of a Dietary Approaches to Stop Hypertension eating plan on features of the metabolic syndrome. Diabetes Care 2005;28:2823–31 [DOI] [PubMed] [Google Scholar]
- 52.Mellen PB, Gao SK, Vitolins MZ, Goff DC., Jr Deteriorating dietary habits among adults with hypertension: DASH dietary accordance, NHANES 1988-1994 and 1999-2004. Arch Intern Med 2008;168:308–14 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.