Diet quality and weight gain among black and white young adults: the Coronary Artery Risk Development in Young Adults (CARDIA) Study (1985–2005)^,

Abstract

Background: Little is known about the long-term health consequences of following the 2005 Dietary Guidelines for Americans (DGA; Washington, DC: US Government Printing Office, 2005).

Objective: The objective was to examine the longitudinal association between diets consistent with the 2005 DGA and subsequent weight gain.

Design: We used data from the Coronary Artery Risk Development in Young Adults (CARDIA) Study, a cohort of black and white men and women aged 18–30 y at baseline who attended ≤7 examinations from 1985–1986 to 2005–2006 (n = 4913). We created a 100-point Diet Quality Index (2005 DQI) to rate participants’ diets based on meeting the 2005 DGA key recommendations. Longitudinal models of weight gain were adjusted for physical activity, smoking, energy intake, age, education, sex, and initial body mass index (BMI) and included interaction terms of DQI by race and initial BMI (if statistically significant).

Results: We found effect modification by race (likelihood ratio test, P < 0.03 in all models). The mean adjusted 20-y weight change was +19.4 kg for blacks and +11.2 kg for whites with high diet quality (DQI >70) and +17.8 for blacks and +13.9 for whites with a DQI <50 (P < 0.05). In race-specific Cox models (with interaction terms for DQI × initial BMI, P < 0.05), a 10-point increase in DQI score was associated with a 10% lower risk of gaining 10 kg in whites with an initial BMI (in kg/m²) <25 but with a 15% higher risk in blacks with baseline obesity (P < 0.001).

Conclusions: Our findings do not support the hypothesis that a diet consistent with the 2005 DGA benefits long-term weight maintenance in American young adults. Greater need for attention to obesity prevention in future DGAs is warranted.

INTRODUCTION

A healthy diet has been characterized in many ways; however, there is no consensus about what the best definition is. One commonly used definition of a healthy diet is adherence to the Dietary Guidelines for Americans (DGA) (1). Because the DGA is intended to promote health and reduce risk of chronic disease (1), it is often assumed that they can help prevent weight gain. Indeed, people who adhere to the 2005 version of the guidelines may have lower energy intakes (2). Yet, there is little evidence that people who have followed the DGA (or similar dietary patterns) actually gained less weight over a long period of time. One reason may be that past dietary guidelines were not intended to prevent weight gain in the population. Another reason may be that the knowledge base used in creating the guidelines was limited. For example, the 2005 DGA is mostly based on studies that reduce diets to individual components (eg, grams of fiber, percentage of energy from fat) (3). This poses a challenge because, due to the complex interactions among known and unknown food components (4), the relation between a single dietary component and disease may differ when all aspects of the diet are considered (5).

For the most part, the studies that found inverse associations between adherence to the DGA and body weight or obesity have been short-term or cross-sectional (4, 6–9). We found 7 longitudinal studies of the association between diets consistent with the DGA and subsequent changes in body weight. Whereas 2 studies found that diets consistent with the 2005 DGA were associated with lower weight gain in whites (10, 11), other studies produced inconsistent results whereby DGA-like dietary patterns were not clearly better at preventing weight gain than were other patterns (12–16). An important limitation of this literature is that most of these studies were performed in subjects who were white; thus, the findings may not be generalizable to African Americans, who are at highest risk of obesity (17). Furthermore, findings from studies performed in subjects who are older (past the age of highest risk of weight gain) or who are already overweight or obese may not be applicable to young and normal-weight adults (10, 18–22).

We used longitudinal data from a cohort of black and white young adults to 1) create an index of overall diet quality based on the key messages conveyed by the 2005 DGA 2), examine the association between diet quality and 20-y risk of weight gain, and 3) determine whether diet quality had the same association with body weight regardless of the participants’ race or initial body mass index (BMI; in kg/m²). We hypothesized that higher diet quality would be associated with less weight gain and that this association would be stronger in whites and in those with normal weight (BMI <25) at baseline.

SUBJECTS AND METHODS

Participants

We used data from the Coronary Artery Risk Development in Young Adults (CARDIA) Study, a prospective epidemiologic study of the determinants and evolution of cardiovascular disease risk factors among young adults. The baseline examination was conducted in 1985–1986, and follow-up exams were conducted 2, 5, 7, 10, 15, and 20 y later. The initial cohort consisted of 5115 young adults recruited from Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA; and was balanced as to age (18–30), sex, race (black and white), and educational status (high-school graduate or less, more than a high-school education). In addition, eligibility criteria included freedom from chronic disease or disability that would interfere with any part of the examination. The retention rate at year 7 was 81% and at year 20 was 72%. Details of the study design and participants were reported elsewhere (23, 24). In the present analysis, we excluded subjects missing data for key variables or pregnant at time of interview (n = 74). We also excluded subjects who had unusually high or low average daily caloric intakes (<800 or >8000 kcal for men and <600 or >6000 kcal for women; n = 128). Baseline sample size was 4913 after exclusions. All analyses were in accordance with the ethical standards of the University of North Carolina at Chapel Hill; the study was approved by the Institutional Review Board at the University of North Carolina at Chapel Hill.

Dietary data

The CARDIA diet history (available for 1985–1986, 1992–1993, and 2005–2006) is an interviewer-administered instrument that consists of a questionnaire regarding usual dietary practices and a quantitative diet-history questionnaire that assessed consumption of foods over the past month. One hundred header questions, such as “Do you eat meat,” “How much do you eat,” and “How was it prepared” elicited foods eaten in an open-ended fashion. Questions were asked about brand names, preparation methods, and frequency of consumption. The open-ended aspect elicited information about special ethnic foods and unusual dietary preferences. Because of the diversity in socioeconomic status and literacy among the CARDIA population, the dietary history was designed to place the responsibility for documentation on the nutritionist, who was trained to probe for specific information in a standard way.

Nutrient and energy intakes were computed by using the nutrient table developed by the Nutrition Coordinating Center (NCC) at the University of Minnesota, based on the 1609 distinct NCC food codes referenced at either year 0 (NCC tape 10) or 7 (NCC tape 20) and several thousand codes at year 20 (NDS-R in 2005). Food groups developed by the NCC were used. Additional details about the quality control and validation of the CARDIA Diet History are available elsewhere (25–27).

Creation of the 2005 Diet Quality Index

The original Diet Quality Index (DQI) was designed in 1994 to evaluate the overall quality of the diet based on the 1989 recommendations by the National Academy of Sciences Food and Nutrition Board (28). It has since been revised in adaptation to different populations and to reflect changes in nutrition knowledge (29–33). Our dietary assessment tool is based on one of these revisions, the DQI-R, which quantifies adherence to the 1995 DGA (32). This new index, the 2005 DQI, reflects the key messages conveyed by the 2005 DGA (1, 34). Because there is no “gold standard” for measuring adherence to the DGA, our index, as well as others that are also based on the 2005 DGA (35), represents its authors’ interpretation of the dietary recommendations.

Each of the 10 DQI (2005) components and how they were scored are shown in Table 1. Scores were based on the percentage of dietary recommendations met, specific cutoffs for nutrient intake, or distribution of values in our sample. Consumption of vitamin supplements did not contribute to estimates of nutrient intakes. Three of the 10 DQI components include intake of key nutrients addressed by the DGA: fat (between 20% and 35% of total energy), saturated fat (≤10% of total energy), and cholesterol (≤300 mg). Four components quantify adequate intake of dairy (reduced fat), fruit, vegetables, and grains. The specific serving recommendations for different levels of energy intake were obtained from Appendix A-2 of the 2005 DGA (1), which lists sample eating patterns from the US Department of Agriculture Food Guide. As a general rule, foods were excluded only when they were specifically mentioned in the DGA (eg, whole-milk consumption was not counted for the dairy intake recommendation because the DGA specifies reduced-fat milk). Points were neither added nor subtracted for servings in excess of recommended intakes. The last 3 components relate to broader health messages, including an emphasis in 2005 on consumption of a variety of foods within and among the basic food groups to achieve the recommended nutrient intakes as well as reduced intake of “empty” calories and sodium. The diversity component reflects consumption of foods from 17 broad food-group categories. Eight of the groups include different types of fruit and vegetables (eg, dark-green vegetables, deep-yellow vegetables, tomatoes, and potatoes), 3 represent reduced-fat dairy products, and 6 represent meat and meat alternatives (eg, eggs, fish, legumes, nuts and seeds, and lean meats). The moderation component reflects “discretionary” behavior on the part of consumers and is based on limiting the consumption of alcohol and reducing sodium intake. The 2005 DGA gives different recommendations for sodium intake for blacks and whites (≤1500 mg for blacks, ≤2300 mg for whites). For blacks, we assigned 5 points for sodium intakes ≤1500 mg, 2.5 points for 1500–3200 mg, and 0 points for >3200 mg; for whites, the respective cutoffs were ≤2300, 2300–4000, and >4000 mg. Similarly, the 2005 DGA gives different recommendations for alcohol consumption for men and women (2 servings/d for men, 1 serving/d for women). We assigned 5 points for ≤1 (women) or ≤2 (men) alcohol servings, 2.5 points for 1–1.5 (women) or 2–3 (men) alcohol servings, and 0 points for >1.5 (women) or >3 (men) alcohol servings. The added sugars component was based on the 2005 DGA's recommendation to limit added sugars and caloric sweeteners. We ranked participants into quintiles based on their intake of added sugars from foods (sugars not naturally occurring in foods, eg, honey, table sugar, and candy) and sugar-sweetened beverages. Our scoring was based on the assumption that people in the lowest quintiles of intake were limiting their intake. Points were summed across the 10 components for a maximum score of 100, with low values reflecting a poor diet and high values reflecting a healthy diet. We categorized the continuous score into 3 categories based on the distributions of total score: low (DQI <50), middle (DQI 50–70), and high (DQI >70).

TABLE 1.

Scoring of the 2005 Diet Quality Index (DQI) components and distribution by Coronary Artery Risk Development in Young Adults (CARDIA) Study year¹

			Subjects meeting the respective scoring criteria
			Baseline (1985)			Year 7 (1992)
Recommendation	Scoring criteria	Points	Black(n = 2786)	White(n = 2427)	Total(n = 4913)	Black(n = 1725)	White(n = 2014)	Total(n = 3739)
			%			%
Keep total fat intake between 20% and 35% of total energy	>40% or <15%	0	40	31	36	30	18	24
	36–40% or 15–19%	5	34	36	35	34	33	33
	≤35% and ≥20%	10	26	33	29	36	49	43
Reduce saturated fat intake to <10% of total energy	>13%	0	68	64	66	42	35	39
	11–13%	5	26	28	27	40	39	39
	≤10%	10	7	8	7	18	26	22
Reduce cholesterol intake to <300 mg/d	>400 mg	0	57	42	50	41	21	30
	300–400 mg	5	16	20	18	19	18	19
	<300 mg	10	27	38	33	40	61	51
Choose foods and beverages that limit intake of added sugars2	Based on population distribution of intake of added sugars and sweetened beverages	0–5	73	51	62	72	55	63
		6–8	22	32	27	20	31	26
		9–10	5	17	11	8	14	11
2–3 Servings of reduced-fat milk or milk alternatives3	≤50% of recommended servings	0	88	64	76	86	61	72
	60–80% of recommended servings	5	7	18	13	9	23	17
	≥90% of recommended servings	10	5	18	11	5	16	11
2–5 Servings of fruit3	≤50% of recommended servings	0–5	37	38	38	41	38	39
	60–80% of recommended servings	6–8	26	30	28	26	32	30
	≥90% of recommended servings	9–10	37	32	34	33	30	31
2–8 Servings of vegetables3	≤50% of recommended servings	0–5	53	33	43	43	23	32
	60–80% of recommended servings	6–8	34	41	38	38	46	42
	≥90% of recommended servings	9–10	13	26	19	19	31	26
1.5–5 Servings of whole grains;3 at least half the grains should come from whole grains4	Operationalized as 2 separate recommendations (5 points each) and then summed	0–5	66	50	58	58	42	50
		6–8	25	32	29	29	35	32
		9–10	9	18	13	13	23	18
Consume a variety of nutrient-dense foods5	Dietary diversity score6	0–10	5.6 ± 1.8	6.5 ± 1.8	6.0 ± 1.9	5.9 ± 1.9	7.1 ± 1.8	6.5 ± 1.9
Limit sodium and alcohol consumption7	Dietary moderation score6	0–10	5.6 ± 2.1	6.1 ± 2.5	5.9 ± 2.3	5.5 ± 2.2	6.0 ± 2.3	5.7 ± 2.3

¹The 2005 DQI is based on the 2005 Dietary Guidelines for Americans (1).

²Calculated quintiles of intake of added sugars from foods and quintiles of intake of calorically sweetened beverages. We assigned 5 points to the first quintile; 3 and 2 points to the second and third quintiles, respectively; and 0 points to the fourth and fifth quintiles.

³The specific serving recommendations for different amounts of energy intake were obtained from Appendix A-2 of the 2005 Dietary Guidelines for Americans (1).

⁴Five points were assigned for meeting 100% of the recommended whole-grain servings (lower scores were prorated) and 5 points for consuming ≥50% of all grain servings as whole grains (whole-grain servings divided by all-grain servings, multiplied by 10, prorated).

⁵For each of 4 food groups, participants received ≤4 points if they were consumers (≥1.75 servings/wk) of each of the respective subgroups. Vegetable subgroups (4 points): dark-green, deep-yellow, orange, starchy vegetables; legumes; and other vegetables. Fruit subgroups (2 points): citrus fruit and noncitrus fruit. Meat subgroups (2 points): lean beef, pork, and poultry; eggs; fish; and nuts and seeds. Dairy subgroups (2 points): reduced-fat milk, cheese, and yogurt.

⁶Values for dietary diversity and dietary moderation are provided as mean ± SD scores.

⁷Higher scores represent lower intakes of sodium (for blacks, we assigned 5 points for sodium intakes ≤1500 mg, 2.5 points for 1500–3200 mg, and 0 points for >3200 mg; for whites, the respective cutoffs were ≤2300, 2300–4000, and >4000 mg) and moderation in consumption of alcohol. We assigned 5 points for alcohol servings ≤1 (women) or ≤2 (men), 2.5 points for 1–1.5 (women) or 2–3 (men), and 0 points for >1.5 (women) or >3 (men).

Weight gain

Body weight and height were measured at each examination by trained staff. Body weight was measured on a calibrated balance scale while participants were dressed in light clothing and without shoes and was recorded to the nearest 0.2 kg. Height (without shoes) was assessed by using a vertical ruler and recorded to the nearest 0.5 cm. Body mass index (BMI) was calculated by dividing weight (in kg) by height (in meters squared). Participants were categorized as normal (BMI <25), overweight (25 ≤ BMI < 30), or obese (BMI ≥30) (36). We calculated a weight gain of ≥10 kg by subtracting baseline weight from weight at each subsequent examination. Weight gain was chosen as the outcome instead of obesity, because all people are at risk of weight gain, regardless of current weight, whereas the likelihood of becoming obese depends importantly on the starting BMI and excludes those who are already obese. Furthermore, a 10-kg weight gain has been linked to adverse changes in serum cholesterol, triglycerides, fasting insulin, and blood pressure (37).

Additional covariates

Physical activity was assessed by using the CARDIA Physical Activity History questionnaire—a self-report of frequency of participation in leisure, occupational, and household physical activities over the past 12 mo (38). Physical activity level was expressed in units of total activity based on the frequency and intensity of each activity and is available from each examination. Standard questionnaires were used at each visit to assess sociodemographic variables. Participants were classified as smokers or nonsmokers at each examination.

Statistical analysis

We used survival analysis methodology (39) in Stata version 10 (Stata Corp, College Station, TX) to examine the associations between diet quality and risk of major weight gain from 1985 to 2005. To avoid the issue of reverse causality (weight change affecting diet, rather than diet affecting weight change), our models were set up so that diet at year 0 predicted weight gain from baseline to years 2, 5, and 7; diet at year 7 predicted weight gain from baseline to years 10, 15, and 20. Hazard ratios (HRs) were estimated by using Cox proportional hazards regression models (40). Models were adjusted for baseline BMI, age, sex, race, education, clinic of recruitment, and time-varying physical activity score, energy intake, and smoking status. Effect modification by race and initial BMI classification was assessed in separate models through the inclusion of interaction terms and significance was determined by using likelihood ratio tests with α = 0.05. We used race-specific models if both race and BMI interaction sets were significant, then tested for BMI × DQI interactions within each race-specific model. Interactions with time were included for variables that did not meet the proportional hazards assumption. We compared models with and without energy intake as well as models coding DQI as a continuous or categorical variable. (DQI <50 was used as the reference for the categorical exposure.)

The longitudinal association between participants’ continuous weight change from baseline and diet quality was examined by using generalized estimating equation (GEE) models adjusted for baseline BMI, age, sex, race, education, clinic of recruitment, and time-varying physical activity score, energy intake, and smoking status. Effect modification by race and initial BMI was assessed as described above. To allow flexibility in the shape of the distribution of weight at each follow-up year, indicator variables for exam year and interaction terms for DQI × exam year were added to the statistical models. Multivariate analyses were repeated by using different cutoffs for excluding “implausible” energy intakes; the most stringent cutoffs excluded men with <1000 or >4000 kcal and women with <800 or >3500 kcal.

RESULTS

Overall DQI scores increased over time, although not all DQI components increased (Table 1). The largest increases were driven by reduced intakes of fat, saturated fat, and cholesterol. In both 1985 and 1992, a greater proportion of whites met the 2005 DGA recommendations for all DQI components except for fruit intake, for which more blacks reported eating ≥90% of the recommended servings. Relatively few participants met the recommended intakes of dairy products or whole grains or were in the lowest quintiles for both added sugars from foods and beverages.

At baseline, most participants were classified as having a low DQI score (Table 2). Blacks with high DQI scores had a higher BMI (P < 0.01), and a higher proportion was obese compared with blacks with low scores (P < 0.01). In contrast, whites with higher DQI scores had lower BMIs (P < 0.01). There was a trend for increased physical activity and education with higher DQIs (P < 0.01) in both blacks and whites.

TABLE 2.

Baseline characteristics of Coronary Artery Risk Development in Young Adults (CARDIA) Study participants presented by race and Diet Quality Index (DQI) score, 1985–1986¹

	Blacks				Whites
Characteristic	DQI < 50(n = 1859)	DQI 50–70(n = 553)	DQI > 70(n = 74)	Total(n = 2486)	DQI < 50(n = 1130)	DQI 50–70(n = 938)	DQI > 70(n = 359)	Total(n = 2427)
DQI score	35.9 ± 8.22	57.6 ± 5.3	76.1 ± 4.7	41.9 ± 13.23	38.9 ± 7.6	58.8 ± 5.7	77.4 ± 5.6	52.3 ± 15.43
Age (y)	24.1 ± 3.7	24.9 ± 3.8	25.3 ± 3.7	24.3 ± 3.83	25.2 ± 3.4	25.5 ± 3.2	25.7 ± 3.2	25.4 ± 3.43
Education (y)	12.9 ± 1.8	13.4 ± 1.9	14.1 ± 1.8	13.1 ± 1.83	14.2 ± 2.4	15.0 ± 2.3	15.3 ± 2.2	14.6 ± 2.43
Physical activity score	383 ± 303	358 ± 256	450 ± 319	380 ± 3013	428 ± 277	448 ± 281	522 ± 298	450 ± 2833
Weight (kg)	72.4 ± 17.7	74.9 ± 18.1	70.7 ± 15.8	72.9 ± 17.83	71.4 ± 15.1	69.0 ± 14.1	65.9 ± 13.1	65.0.9 ± 13.13
BMI (kg/m2)	25.1 ± 5.7	26.5 ± 5.8	25.9 ± 6.1	25.3 ± 5.83	23.8 ± 4.2	23.7 ± 4.0	23.1 ± 3.7	23.7 ± 4.13
Overweight (%)	22.2	32.5	23.0	24.53	22.9	21.0	17.5	21.4
Obese (%)	15.6	20.2	20.3	16.83	7.1	7.0	5.0	6.8
Male (%)	47.9	29.8	21.6	43.23	59.7	40.7	24.8	47.23
Daily intake4
Total energy (kcal)	3205 ± 1426	2368 ± 1289	1788 ± 1010	2977 ± 14443	3076 ± 1214	2383 ± 1053	1981 ± 846	2646 ± 11823
Total fat (% of energy)	39.8 ± 5.1	34.2 ± 5.4	28.9 ± 4.7	38.2 ± 5.93	40.1 ± 4.9	36.1 ± 4.8	30.0 ± 5.2	37.0 ± 6.03
Saturated fat (% of energy)	15.0 ± 2.6	12.2 ± 2.4	9.6 ± 2.1	14.0 ± 2.93	15.6 ± 2.6	13.5 ± 2.4	10.6 ± 2.55	14.0 ± 3.03
Total carbohydrates (% of energy)	44.4 ± 6.5	51.4 ± 7.7	57.6 ± 6.8	46.3 ± 7.63	42.2 ± 6.1	47.0 ± 5.8	53.5 ± 6.95	45.7 ± 7.33
Cholesterol (mg/1000 kcal)	189.6 ± 70.3	144.7 ± 56.0	130.3 ± 45.2	177.8 ± 69.83	172.5 ± 57.1	147.7 ± 52.0	119.6 ± 50.9	155.1 ± 57.53
Sodium (g/1000 kcal)	1.5 ± 0.3	1.3 ± 0.3	1.3 ± 0.3	1.4 ± 0.33	1.6 ± 0.3	1.5 ± 0.3	1.5 ± 0.35	1.5 ± 0.3
Fiber (g/1000 kcal)	1.6 ± 0.6	2.3 ± 1.0	3.9 ± 1.5	1.8 ± 0.93	1.8 ± 0.6	2.5 ± 1.0	3.7 ± 1.5	2.4 ± 1.63
Dairy, total (servings/1000 kcal)	1.0 ± 0.6	1.0 ± 0.9	1.0 ± 0.7	1.0 ± 0.7	1.3 ± 0.9	1.4 ± 0.7	1.4 ± 0.85	1.4 ± 0.8
Dairy, reduced-fat (servings/1000 kcal)	0.2 ± 0.3	0.4 ± 0.5	0.6 ± 0.5	0.2 ± 0.43	0.4 ± 0.5	0.7 ± 0.6	0.9 ± 0.75	0.6 ± 0.63
Vegetables, total (servings/1000 kcal)	1.1 ± 0.6	1.5 ± 1.0	2.5 ± 1.7	1.2 ± 0.83	1.3 ± 0.6	1.8 ± 1.2	2.8 ± 2.0	1.7 ± 1.33
Dark-green	0.1 ± 0.1	0.2 ± 0.2	0.4 ± 0.5	0.1 ± 0.23	0.1 ± 0.2	0.2 ± 0.4	0.5 ± 0.9	0.2 ± 0.53
Deep-yellow	0.0 ± 0.1	0.1 ± 0.2	0.4 ± 1.0	0.1 ± 0.23	0.1 ± 0.1	0.2 ± 0.3	0.4 ± 0.6	0.1 ± 0.33
Tomatoes	0.1 ± 0.1	0.2 ± 0.3	0.3 ± 0.3	0.1 ± 0.23	0.2 ± 0.1	0.2 ± 0.2	0.3 ± 0.3	0.2 ± 0.23
Starchy	0.2 ± 0.2	0.2 ± 0.2	0.2 ± 0.2	0.2 ± 0.2	0.2 ± 0.2	0.2 ± 0.2	0.2 ± 0.2	0.2 ± 0.2
Other	0.3 ± 0.3	0.6 ± 0.5	0.9 ± 0.8	0.4 ± 0.43	0.5 ± 0.4	0.7 ± 0.7	1.2 ± 0.9	0.7 ± 0.73
Fried	0.3 ± 0.2	0.2 ± 0.2	0.2 ± 0.3	0.3 ± 0.23	0.2 ± 0.2	0.2 ± 0.2	0.1 ± 0.2	0.2 ± 0.23
Whole grains (servings/1000 kcal)	0.4 ± 0.4	0.6 ± 0.5	1.1 ± 0.7	0.5 ± 0.53	0.5 ± 0.4	0.8 ± 0.6	1.2 ± 0.7	0.7 ± 0.63
Alcohol (servings/1000 kcal)	0.2 ± 0.4	0.2 ± 0.4	0.1 ± 0.3	0.2 ± 0.43	0.4 ± 0.4	0.3 ± 0.4	0.3 ± 0.45	0.3 ± 0.43
Fruit, total (servings/1000 kcal)	1.0 ± 0.7	2.0 ± 1.2	3.5 ± 1.5	1.3 ± 1.13	0.7 ± 0.6	1.5 ± 0.9	2.1 ± 1.15	1.2 ± 1.03
100% Fruit juice	0.6 ± 0.6	1.2 ± 1.0	1.8 ± 1.4	0.8 ± 0.83	0.4 ± 0.4	0.7 ± 0.7	0.9 ± 0.95	0.6 ± 0.63
Sweetened drinks (servings/1000 kcal)	0.7 ± 0.6	0.8 ± 0.9	0.4 ± 0.6	0.7 ± 0.73	0.5 ± 0.6	0.4 ± 0.6	0.2 ± 0.35	0.4 ± 0.63

¹The DQI is based on the 2005 Dietary Guidelines for Americans (1). Scores range from 0 to 100; higher scores indicate a diet more consistent with the guidelines. DQI categories: low = DQI <50, middle = DQI 50–70, high = DQI >70.

²Mean ± SD (all such values).

³Significant difference (chi-square or ANOVA, α = 0.05) across DQI categories within each race group.

⁴Serving sizes based on the 2005 Dietary Guidelines for Americans (1). Total mean daily intake was significantly different between blacks and whites (ANOVA, P < 0.03) for all nutrients or foods shown except for saturated fat and starchy vegetables.

⁵Significantly different from blacks with a high DQI score, P < 0.03 (ANOVA).

For the most part, baseline mean total food and nutrient intakes differed significantly between blacks and whites. For example, compared with blacks, whites reported lower intakes of total calories, fat, cholesterol, and sweetened beverages, but higher intakes of dairy, whole grains, and vegetables. Among participants with a high diet quality, blacks reported consuming a higher percentage of energy from carbohydrates; a higher intake of sweetened beverages, fruit, and 100% fruit juice; and lower intakes of reduced-fat dairy products, sodium, and saturated fat compared with whites.

Most participants gained weight over the 20-y study period, regardless of diet quality, with a mean (±SD) weight gain of 17.9 ± 14.5 kg in blacks and 12.5 ± 11.7 kg in whites. In multivariable-adjusted GEE models of continuous weight change (see Figure 1 and Supplemental Table 1 under “Supplemental data” in the online issue)) we observed significant effect modification by race (likelihood ratio test, P < 0.01) but not by baseline BMI. On average, blacks with a high diet quality (DQI >70) gained significantly more weight than did blacks with a low diet quality (DQI <50) over 20 y (19.4 compared with 17.8 kg). In contrast, whites with a high diet quality gained significantly less weight than did whites with a low diet quality (11.2 compared with 13.9 kg).

FIGURE 1.

Adjusted 20-y mean weight change in CARDIA (Coronary Artery Risk Development in Young Adults) Study participants with low (<50) or high (>70) Diet Quality Index (DQI) scores. Weight change estimates were based on a generalized estimating equation model that includes DQI score interactions with year and race and adjusted for baseline BMI, age, sex, education, clinic of recruitment, and time-varying physical activity score, energy intake, and smoking status. Baseline n = 4913 (average of 5 observations per person). At years 15 and 20, the mean weight change was significantly different (P < 0.05) between all subgroups shown.

Results of multivariable Cox regressions for risk of major weight gain (≥10 kg) are presented in Table 3. Overall, having a high (compared with a low) diet quality was associated with a 25% lower risk of major weight gain (HR: 0.75; 95% CI: 0.65, 0.87). However, we observed significant (P < 0.05) effect modification by race and baseline BMI and therefore present effect estimates for each subgroup. After adjustment for potential confounders, a10-point increase in DQI score was associated with a 10% risk reduction in normal-weight (BMI < 25) whites. In contrast, blacks who were obese at baseline had 15% higher risk of gaining 10 kg for each 10-point increase in DQI score. Further adjustment for energy intake caused almost no change to estimates. Similar results were obtained from models using DQI score as a categorical variable.

TABLE 3.

Associations between Diet Quality Index (DQI) scores and risk of 10-kg weight gain from 1985 to 2005¹

Model	Blacks	Whites
DQI score as continuous variable (per 10-point increment)
Model 12
Overall	0.98 (0.95, 1.01)
Model 23
Normal weight	1.00 (0.95, 1.06)	0.90 (0.86, 0.95)4
Overweight	1.02 (0.96, 1.10)	1.04 (0.96, 1.12)
Obese	1.15 (1.05, 1.23)4	0.92 (0.81, 1.04)
DQI score as categorical variable (reference: DQI <50)
Model 1
Overall
DQI 50–70	0.98 (0.89, 1.08)
DQI >70	0.75 (0.65, 0.87)4
Model 23
Normal weight
DQI 50–70	0.97 (0.81, 1.16)	0.77 (0.65, 0.91)4
DQI >70	0.83 (0.59, 1.17)	0.61 (0.49, 0.75)4
Overweight
DQI 50–70	1.09 (0.87, 1.38)	1.11 (0.85, 1.44)
DQI >70	0.78 (0.48, 1.28)	1.10 (0.77, 1.57)
Obese
DQI 50–70	1.36 (1.03, 1.79)4	1.12 (0.75, 1.68)
DQI >70	1.68 (1.01, 2.82)4	0.54 (0.27, 1.09)

¹All values are hazard ratios; 95% CIs in parentheses. Cox proportional hazards regression models were adjusted for baseline age, education, BMI, clinic of recruitment, race, sex, and time-varying physical activity, energy intake, and smoking status (baseline n = 4913).

²Model without interaction terms.

³Race-specific models include DQI interactions with baseline BMI.

⁴α = 0.05.

In blacks, even though the sample was reduced by 549 participants, results were robust to more stringent exclusion criteria for implausible energy intakes (<1000 or >4000 kcal for men and <800 or >3500 kcal for women, compared with our original exclusion of men with <800 or >8000 kcal and women with <600 or >6000 kcal). In whites, the only appreciable difference seen with more stringent exclusion cutoffs (sample reduced by 239 participants) was that the effect modification by initial BMI became attenuated (data not shown). In models with continuous DQI as the exposure, effect modification by initial BMI was no longer significant (likelihood ratio test, P = 0.13; overall HR for 10-point increase in DQI score among whites: 0.92, 95% CI: 0.88, 0.97). In models with categorical DQI, the HRs for obese whites with high (compared with low) DQI scores were significant (HR: 0.48; 95% CI: 0.23, 0.98), but there was still an overall significant effect modification by initial BMI (likelihood ratio test, P = 0.05).

DISCUSSION

Despite a higher risk of obesity and weight gain among blacks, few longitudinal studies of diet and weight change have examined racial differences. The intent of this study was to examine the association between having a diet consistent with the 2005 DGA (as operationalized by the 2005 DQI) and subsequent weight gain in black and white young adults. Our findings suggest that a diet consistent with the DGA was associated with more weight gain in blacks (particularly if obese), but with less weight gain in whites after adjustment for participants’ physical activity, caloric intake, smoking, and other sociodemographic characteristics. However, even whites with high DQI scores experienced an average weight gain of >10 kg over a 20-y period.

Both blacks and whites with higher DQI scores ate more whole grains, fruit, low-fat dairy products, and nonfried vegetables and had lower intakes of total fat, saturated fat, cholesterol, and sugar-sweetened beverages. However, because of ambiguity in the 2005 DGA, a high DQI score does not represent a single dietary pattern. The reason is that many different food options can be used interchangeably to meet a dietary recommendation. For example, people can meet the DGA recommendations for fruit by eating either fresh raw or processed fruit; however, the differences in nutritional quality and glycemic effects can be large (41–44). Thus, it is possible that the observed differential associations for diet quality by race could be partly explained by differences in the nutritional quality of foods consumed. A few differences between the diets of blacks and whites in the United States have been found across studies, including a higher intake of refined grains and sweet beverages (45–49). Our own results show that, in relation to whites with a high diet quality, blacks with a high diet quality report a higher percentage of energy from carbohydrates and a higher intake of sugars, fruit, and 100% fruit juice. Such differences are indicative of a higher glycemic load in blacks than in whites (43, 50–52). However, the diets of blacks with high DQI scores are still closer to the DGA than are the diets of blacks with low DQI scores. Moreover, our analyses are race-specific; thus, blacks with a high diet quality are compared with blacks with a low diet quality and not to whites with a high diet quality.

Another possibility is that metabolic or physiologic differences between blacks and whites underlie differential responses to diet. Racial differences in several metabolic aspects of weight regulation have been observed, including fuel oxidation, resting energy expenditure, and the metabolic response to weight loss (53–61). This might help explain why in weight-loss trials black participants tend to lose less weight and regain weight faster than do their white counterparts (10, 62, 63). Furthermore, blacks of all ages tend to have higher plasma insulin concentrations and lower insulin sensitivity than do whites, independently of adiposity (64–69). Because insulin resistance and insulin secretion play a role in body weight regulation (70–74) and diet composition can affect these variables (51, 75), a person's insulin response may modify the association between diet and body weight (52). Results from trials comparing weight change among individuals of varying levels of insulin secretion (73, 76–78) suggest that the glycemic load of meals is more relevant for people who have higher insulin secretion. The 2005 DGA emphasizes a high-carbohydrate dietary pattern, but was not designed to have a low glycemic load. Hence, whether due to genetic or environmental factors, the higher insulin secretion documented in African Americans (67) may make them more susceptible to the glycemic effects of a high-carbohydrate dietary pattern.

Adjustment for energy intake did not attenuate the association between DQI scores and weight gain. Thus, the possibility that the observed associations are the result of misreporting of dietary intake cannot be ruled out. A common concern with self-reported dietary data is that heavier participants may underreport food intake to a greater extent (79). Moreover, differential reporting of diet by race is likely. Despite a concerted effort to include foods relevant to dietary preferences of both blacks and whites, the CARDIA dietary-history questionnaire did not yield estimates of caloric intakes at baseline that were as reasonable for blacks as for other populations (26). Also, results from a validation study suggest that there was more random variation in the diet histories of blacks than of whites (25). However, sensitivity analysis limiting the range of allowable energy intakes did not attenuate effect estimates in blacks, and the effect modification by initial BMI was attenuated only in whites (although effect estimates remained largely unchanged).

A comprehensive assessment of the whole diet is less subject to measurement error than is the assessment of energy intake alone (80). That is, because even when people under- or overreport the total amount they consume, the ratios of the foods that they do report is still likely to be reflective of actual consumption. Several components of the 2005 DQI are designed to account for misreporting by scoring based on each subject's reported intake, rather than their predicted intake based on weight, sex, age, etc. For example, a person with a predicted energy requirement of 2000 kcal who actually eats (or reports eating) 6000 kcal will not get a high DQI score simply by meeting the food and nutrient recommendations for a 2000-kcal intake. To get a high score, subjects have to meet the dietary recommendations for the amount and type of food they reported eating.

Limitations of this study relate mainly to its observational nature. Although we adjusted analyses for several characteristics of participants, the possibility of residual confounding precludes definitive conclusions about causality. Also, our statistical models relied on the assumption that dietary patterns were applicable several years (≤13 y) after they were reported. Although we cannot know for sure how those assumptions affected our results, some research suggests that the dietary patterns of adults are relatively stable over time (81, 82). Moreover, because of the collinearity among foods and nutrients, we are not able to accurately determine which specific components of the DQI score are driving the observed associations.

The 2005 DQI was designed to assess how well the participant's usual diets agreed with the dietary recommendations provided by the 2005 DGA (1), and great care was taken not to allow our preconceptions about optimal nutrition influence decisions regarding the scoring of the index. Hence, our results do not suggest that a healthy diet is ineffective in the prevention of weight gain, but rather that a different definition of a healthy diet may be needed to achieve better results (83). It is important to note that creating the DQI entailed making some subjective decisions based on our interpretation of the 2005 DGA (eg, which components to include). Thus, the possibility exists that a different interpretation of the DGA would yield different results. Nevertheless, the 2005 DQI is similar to the 2005 Healthy Eating Index (HEI-2005)—an index that has also been revised to meet the changing dietary guidelines (35). For example, both our DQI and the HEI-2005 were designed to uncouple diet quality from diet quantity, and they both include components for fruit, vegetables, grains, dairy products, fats, sugars, alcohol, and sodium and account for dietary variety. The main difference between the 2 indexes is that the HEI-2005 includes a component for “meat and beans” but not for cholesterol, whereas the reverse is true for our DQI.

Our findings do not support the hypothesis that a diet consistent with the 2005 DGA benefits long-term weight maintenance among American young adults. A greater need for attention to obesity prevention in future DGAs is warranted. More research is needed to determine whether the observed differential associations by race are due to differences in diet, dietary reporting, or physiologic mechanisms.