Dietary patterns associated with metabolic syndrome, sociodemographic and lifestyle factors in young adults: the Bogalusa Heart Study

Abstract

Objective

To examine the association between dietary patterns (DP) and risk for metabolic syndrome (MetS); and to identify differences in DP by socio-economic, demographic and lifestyle factors.

Design

Dietary intake (from an FFQ), anthropometric/biochemical parameters and sociodemographic/lifestyle information (from a self-reported questionnaire) were evaluated, using a cross-sectional design. Statistical methods included principal component factor analysis, analysis of covariance and linear regression. All analyses were covariate-adjusted.

Setting

The Bogalusa Heart Study (1995–1996), USA.

Subjects

Young adults (19–39 years; n 995; 61 % females/39 % males; 80 % whites/20 % blacks) from a semi-rural southern US community were examined.

Results

The ‘Western Dietary Pattern’ (WDP) consisted of refined grains, French fries, high-fat dairy foods, cheese dishes, red meats, processed meats, eggs, snacks, sweets/desserts, sweetened beverages and condiments. The ‘Prudent Dietary Pattern’ (PDP) consisted of whole grains, legumes, vegetables, fruits, 100 % fruit juices, low-fat dairy products, poultry, clear soups and low-fat salad dressings. The DP explained 31 % of the dietary intake variance. Waist circumference (P = 0·02), triceps skinfold (P = 0·01), plasma insulin (P = 0·03), serum TAG (P = 0·05), and the occurrence of MetS (P = 0·03) were all inversely associated with PDP. Insulin sensitivity (P < 0·0005) was positively associated with PDP. Serum HDL cholesterol (P = 0·05) was inversely associated with WDP. Blacks consumed more servings from WDP than whites (P = 0·02). Females consumed more servings from PDP than males (P = 0·002). Those with >12 years of education consumed more servings from PDP than their counterparts (P < 0·0001). Current smokers consumed more servings from WDP than current non-smokers (P < 0·0001). Physically very active young adults consumed fewer servings from WDP than their sedentary counterparts (P = 0·02).

Conclusions

More studies are warranted to confirm these findings in other populations.

Studies involving dietary patterns (DP) and their association with diseases have several benefits over the conventional approach, which has focused largely on the effects of single nutrients or individual foods(1,2). As the measurement of diet is complex, and foods are typically consumed in combinations, the combined effect of nutrients and foods can be observed only when DP are examined(1,2). Moreover, results from DP analyses are more helpful in disseminating diet-related messages to consumers that they may be more likely to adhere to rather than those related to single foods or nutrients(3). DP have also been related to selected biomarkers of dietary exposure(1,2) and have been reported to contribute in the development or prevention of CHD and type 2 diabetes mellitus (T2DM)(4).

Recent focus has been on the occurrence of metabolic syndrome (MetS), a constellation of metabolic abnormalities including central obesity, elevated blood levels of CHD-promoting lipids, hypertension, insulin resistance and hyperglycaemia(5). In adults, MetS increases the risk of CHD by two-fold and the risk for T2DM by five-fold(5–7). The age-adjusted prevalence of MetS in US adults (≥20 years) participating in the 2003–2006(8) National Health and Nutrition Examination Survey (NHANES) was 34 % v. 29·2 % reported in the 1988–1994 NHANES(9). Among young adults (20–39 years), the prevalence of MetS has increased from 10·8 % (in 1988–1994)(9) to 15·6 % (in 2003–2006) in females(8), and from 15·7 % (in 1988–1994)(9) to 20·3 % (in 2003–2006)(8) in males. Young adulthood is an important period of transition from adolescence into adulthood, when individuals begin to live an independent life. Pressures of independence, hurried lifestyles and providing support for new families may lead to shifts in their dietary and lifestyle patterns. Consequently, unhealthy dietary habits such as skipping breakfast(10), relying on fast food(11) and eating outside home(12) are prevalent among young adults. Moreover, individuals from rural and semi-rural US communities tend to have poorer dietary and health habits because of their lower socio-economic status (SES)(13–15). It is therefore critical to examine the DP of young adults and their relation to risk factors for chronic diseases in order to administer effective dietary and lifestyle prevention and treatment programmes for metabolic disorders such as the MetS, in this age group.

Despite the rising prevalence of MetS, few recent studies have examined the role of DP and their relationship with MetS(16–20); their results, in general, showed that healthy DP were inversely associated with the occurrence of MetS in adults(16–20). However, to date only one Bogalusa Heart Study (BHS)(21) has examined the relationship of diet with MetS in young adults, and showed that lower fruit and vegetable consumption and higher sweetened beverage consumption were independently associated with one to two risk factors for MetS(21). Yet, the above-mentioned BHS explored the association of only single food groups rather than DP with MetS.

The present study, although an extension of the previous BHS(21), had two additional objectives. First, it aimed to identify various DP among young adults and to examine the association of these DP with the risk factors of MetS. Second, because food consumption differs by SES and demographic factors such as gender and ethnicity(13), and the occurrence of MetS is related to lifestyle factors such as physical inactivity, smoking and alcohol consumption(22), the present study also assessed SES, and demographic and lifestyle differences among the DP in these young adults.

Methods

Study design and participants

The BHS was conducted in the semi-rural community of Washington Parish, Bogalusa, which is 70 miles north of New Orleans, LA(23). The study began in 1973 as a long-term epidemiological investigation of cardiovascular risk factors and their environmental determinants in a bi-racial (black/white) paediatric population. Eventually, the study was expanded to include observations of young adults. Details of the BHS study design, participation rates and protocols are presented elsewhere(23). Data for the present study were collected during a follow-up post-high school cross-sectional survey conducted in 1995–1996 on young adults aged 19–39 years (mean age 30 (sd 5·1) years). Data on ninety-four subjects were excluded from an initial sample of 1089 subjects: i.e. females with energy intakes <2092 kJ (500 kcal) or >14644kJ (3500 kcal)(2) (n 47), males with energy intakes <3347kJ (800 kcal) or >16736kJ (4000 kcal)(2) (n 23) and pregnant and/or lactating females (n 24). The final sample (n 995) was 61 % females and 39 % males, with 80 % whites and 20 % blacks. The present study was approved by the Tulane Medical Center Institutional Review Board, and written informed consent was obtained from all the participants.

Measurements

Dietary measures

All young adults from the present study completed the youth/adolescent questionnaire (YAQ), a 131 food-item, self-administered, semi-quantitative food-frequency questionnaire(24,25). The YAQ is valid and reliable for use in epidemiological studies(24,25). Briefly, this questionnaire included specific foods/beverages (including alcohol) along with a commonly used unit or portion size. Each food/beverage item provided three to six possible responses, ranging from ‘never or less than once a month’ to ‘five or more times per day.’ Participants indicated how often, on average, they had consumed a given amount of the specified food/beverage during the past year. Usual portion sizes were calculated for each of the food/beverage items. The selected frequency choice indicated by the participants for each food/beverage was converted to daily intake, e.g. one serving/week was converted to 0·14 serving/d. Food/beverage items were then grouped into specific food categories (as reported earlier)(13), and were further categorised based on food characteristics, e.g. refined or whole grains, low-fat or high-fat dairy foods and so on. Thirty-six food groups from the YAQ were identified for the analyses, of which twenty-four food groups representing the DP from the current study are presented in Table 1.

Table 1.

Components of food groups^* included in the DP from the YAQ: The Bogalusa Heart Study

Food groups	Foods included
1. Whole grains	Hot breakfast cereal (e.g. oatmeal, grits), dark breads, other grains (e.g. bulgur, kasha,    couscous)
2. Refined grains	White bread, pita bread or toasts, muffins, cornbreads, bagels, biscuits or rolls, rice, noodles,    pasta, pancakes, waffles, tortillas
3. Low-fat dairy products	Skim or 1 % milk, non-fat or low-fat yoghurt and cheese, other non-fat dairy products
4. High-fat dairy products	Whole or 2 % milk/chocolate milk, whipped cream, regular yoghurt, cheese, cottage cheese,    cream cheese, pudding, frozen yoghurt, ice cream, milkshake or frappe
5. Fruits	Grapes, raisins, bananas, cantaloupes, melons, apples/apple sauce, pears, oranges,    strawberries, peaches, plums, apricots
6. 100 % fruit juices	100 % fruit juices
7. Tomatoes	Tomatoes, tomato sauce, spaghetti sauce, salsa
8. Legumes	Beans, lentils, soybeans, peas, lima beans
9. Cruciferous vegetables	Broccoli, greens, coleslaw, kale
10. Green leafy vegetables	Spinach, lettuce, tossed salad
11. Dark-yellow/orange vegetables	Carrots, yams, sweet potatoes
12. Other vegetables	String beans, beets, corn, peppers, eggplant, zucchini, mixed vegetables, summer squash
13. French fries	French fries
14. Red meats	Beef, steak, lamb, pork, meatballs, meatloaf, ham
15. Processed meats	Processed meats, bacon, hot dogs, salami, bologna
16. Poultry	Chicken, turkey, chicken nuggets
17. Eggs	Eggs
18. Main dishes
Dishes with cheese	Pizza, tacos/burritos, lasagna, baked ziti, macaroni and cheese, spaghetti, grilled cheese
Burgers and sandwiches	Cheese burger, hamburger, peanut butter sandwich, chicken/turkey sandwich, roast beef/ham    sandwich, deli meat sandwich, tuna sandwich, other fish sandwich
19. Snacks	Potato chips, corn chips, nachos, popcorn, pretzels, crackers, peanuts, fun fruit, graham    crackers, saltines, wheat thins
20. Sweets and desserts	Pop tarts, cakes, snack cakes, Twinkies, Danish pastries, pastries, donuts, cookies, brownies,    pie, chocolates, candy bars, other candy such as mints, flavoured gelatin, pudding, frozen    yoghurt, ice cream, milkshake, popsicles
21. Sweetened beverages	Soda, punch, lemonade, non-carbonated fruit drink, iced tea
22. Low-fat salad dressings	Low-fat salad dressing
23. Condiments	Brown gravy, ketchup, mayonnaise, added sugar
24. Low-fat soups	Clear soup, chicken noodle soup

DP, dietary patterns; YAQ, youth and adolescent food frequency questionnaire.

^*Only twenty-four food groups identified in the DP (from Table 2) are discussed above from a total of thirty-six food groups from the YAQ.

Anthropometric measures

Duplicate measures of all anthropometric parameters were collected by trained examiners using standardised protocols(23); least-square means and their standard errors are presented. Height was measured to the nearest 0·1 cm on a stadiometer and weight was measured to the nearest 0·1 kg on a balance-beam metric scale. The National Heart, Lung and Blood Institute reference standards(26) were used to classify participants’ BMI (weight (kg)/height² (m²)) into normal weight (BMI ≥ 18·5 and ≤24·9kg/m²) or overweight/obese (BMI ≥ 25 kg/m²). Young adults who were underweight (BMI < 18·5 kg/m²) were included with those in the normal weight category. Waist circumference was measured midway between the rib cage and superior border of the iliac crest. Hip circumference was measured at the greater trochanters. Waist-to-hip ratio was calculated. Triceps skinfold was measured to the nearest millimetre with Lange skinfold calipers (Cambridge Scientific Industries, Inc., Cambridge, MD, USA). A description of the reproducibility of these measures used in the BHS is discussed elsewhere(23).

Laboratory measures

Venous blood was collected following a 12h fast. Plasma glucose concentrations were measured by the glucose oxidase method using a Beckman glucose analyzer (Beckman Instruments, Fullerton, CA, USA); a commercial radioimmunoassay kit measured plasma immunoreactive insulin concentration (Padebas Pharmacia, Piscataway, NJ, USA). Indices of insulin sensitivity were calculated according to the Quantitative Insulin Sensitivity Check Index (QUICKI) formula (=1/(log fasting plasma glucose (mg/dl) + log fasting plasma insulin (µU/ml)); higher QUICKI values indicate greater insulin sensitivity)(27); and the Homeostasis Model Assessment of insulin resistance (HOMA-IR) formula (=(fasting plasma glucose (mg/dl) × fasting plasma insulin (µU/ml))/405; higher HOMA values indicate greater insulin resistance)(28). Serum total cholesterol and TAG concentrations were measured using enzymatic procedures on the Abbott VP instrument (Abbott Laboratories, North Chicago, IL, USA), and serum LDL cholesterol (LDL-C) and HDL cholesterol (HDL-C) were analysed using a combination of heparin–calcium precipitation and agar–agarose gel electrophoretic procedures(29). Blind duplicates were used for quality control for all analyses(23). Right arm systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured in triplicate by trained nurses(23), using the first and fifth Korotkoff phase readings with the participant sitting relaxed, and also using the automated instrument (the readings of which were used in the present study). Means of all replicate measures were used for statistical analyses.

Diagnostic criteria for metabolic syndrome

Several definitions exist for MetS(30). The International Diabetes Federation (IDF) criteria help to generate greater prevalence estimates for MetS (especially, in the European population) by using central obesity as a mandatory criterion. Conversely, the revised Adult Treatment Panel (ATP III)(5) criteria mandate the selection of a wider range of risk factors for identifying individuals with MetS (as discussed below), with no single mandatory criterion. Although, in the USA, both the IDF and revised ATP III criteria identify mostly the same people, the IDF criteria have a lower predictive power for coronary events(30). Therefore, in the present study, the young adults were classified as having MetS using the revised ATP III criteria(5,30). Data on medications prescribed to the young adults to increase serum HDL-C and/or lower serum TAG were not available in the present data set. Therefore, we slightly modified the criteria for serum TAG and HDL-C from the original revised ATP III criteria(5,30).

The criteria for classifying young adults with MetS for the present study were ≥3 of the following risk factors: (i) abdominal obesity (waist circumference ≥ 102 cm in males and ≥88 cm in females); (ii) high serum TAG (≥150 mg/dl); (iii) low serum HDL-C (<40 mg/dl in males or <50 mg/dl in females); (iv) high blood pressure (≥130 or ≥85 mmHg or those taking medications for hypertension); and (v) high fasting plasma glucose (≥100 mg/dl or those taking medications, i.e. oral hypoglycaemic agents/insulin).

Demographic, socio-economic status and lifestyle information

Participants completed a questionnaire eliciting information on their age, gender, ethnicity, smoking status (i.e. non-smoker, current smoker and ex-smoker) and alcohol intake (based on the frequency, type and length of alcohol use during the past 12 months). White/black males and females were classified into four ethnicity × gender groups. The SES of the young adults was determined using income (i.e. ≤$15 000, $15 001–30 000, $30 001–45 000, >$45 000) and education levels (i.e. ≤12 years or >12 years). To determine marital status, young adults were asked whether they were currently married and/or cohabiting or were single. Physical activity outside work was measured with a self-reported subjective rating on a 5-item Likert scale adapted from the Lipid Research Clinic’s questionnaire(31). Participants were considered sedentary if they classified themselves as 1 or 2; were considered moderately active if they classified themselves as 3; and were considered very active if they classified themselves as 4 or 5 on the Likert scale. The test–retest reliability of this questionnaire has been reported to be high (r = 0·85). Also, this questionnaire has been noted to be significantly associated with a 4-week physical activity history(31).

Data analysis

The Statistical Analysis Software (version 8·2; SAS Inc., Cary, NC, USA)(32) was used to conduct data analyses. To identify the DP, principal components factor analysis was conducted. Factor analysis helps to summarise and refine large data sets containing several variables, simultaneously, into a small number of orthogonal variables named as ‘patterns’. Factor analysis has earlier shown to have good reproducibility and validity with data from a food frequency questionnaire(33).

In the present study, thirty-six food groups from the YAQ were subjected to principal component factor analysis with varimax rotation to identify the DP. Specific food items were aggregated based on the degree to which the food items were correlated with one another in the data set. Eigenvalues (>1), the scree test (a graph from which the number of factors were chosen where the plot levelled off to a linear decreasing pattern) and interpretability of derived factors were used to derive the DP. Linear regression examined the association between DP and MetS risk factors (dependent variables). Analysis of covariance with Tukey–Kramer’s post-hoc test was used to examine: (i) ethnicity × gender differences in the occurrence of metabolic risk factors (dependent variables); and (ii) differences in mean servings of foods from the DP (dependent variables) by SES, demographic and lifestyle characteristics. The mean number of servings of foods consumed from the DP was used in the latter analyses because factor scores by themselves have no biological meaning. The covariates varied for each analyses and included age, energy intake, gender, ethnicity, ethnicity × gender, SES, marital status, physical activity, smoking, alcohol consumption and BMI. Statistical significance was set at P ≤ 0·05.

Results

Identification of dietary patterns (Table 2)

Table 2.

Identification of DP from factor loadings^* for foods from the YAQ: The Bogalusa Heart Study

	Factor loadings
Food items	WDP	PDP
1. Whole grains	–	0·46
2. Legumes	–	0·61
3. Cruciferous vegetables	–	0·70
4. Other vegetables	–	0·74
5. Green leafy vegetables	–	0·69
6. Dark-yellow vegetables	–	0·70
7. Tomatoes	–	0·58
8. Fruits	–	0·64
9. 100 % fruit juices	–	0·43
10. Low-fat dairy products	–	0·36
11. Poultry	–	0·40
12. Clear soups	–	0·36
13. Low-fat salad dressings	–	0·49
14. Refined grains	0·43	–
15. French fries	0·53	–
16. High-fat dairy products	0·53	–
17. Dishes with cheese	0·58	–
18. Red meats	0·50	–
19. Processed meats	0·59	–
20. Eggs	0·39	–
21. Snacks	0·53	–
22. Sweets and desserts	0·54	–
23. Sweetened beverages	0·44	–
24. Condiments	0·40	–
Variability explained	19 %	12 %

DP, dietary patterns; YAQ, youth and adolescent food frequency questionnaire; WDP, Western dietary pattern; PDP, prudent dietary pattern.

^*Data (1–24) are factor loadings (correlation coefficients between the variables and factors) derived from principal component factor analysis. Absolute values of factor loadings <0·30 are indicated by ‘–’ for simplicity.

Factor analysis retained two DP, which contained twenty-four of the original thirty-six food groups from the YAQ. The DP were labelled as: the ‘Western Dietary Pattern’ (WDP; consisting of refined grains, French fries, high-fat dairy products, dishes with cheese, red meats, processed meats, eggs, snacks, sweets and desserts, sweetened beverages and condiments) and the ‘Prudent Dietary Pattern’ (PDP; consisting of whole grains, legumes, vegetables (i.e. cruciferous, other leafy and dark-yellow vegetables), tomatoes, fruits, 100 % fruit juices, low-fat dairy products, poultry, clear soups and low-fat salad dressings). The WDP and the PDP explained 19 % and 12 % of the dietary intake variance, respectively.

Covariate-adjusted mean metabolic profiles of young adults (Table 3)

Table 3.

Covariate-adjusted metabolic profiles of young adults^* (19–39 years) by ethnicity and gender: The Bogalusa Heart Study

	White males (n 311)		White females (n 486)		Black males (n 78)		Black females (n 120)		P-values†
	Least-square mean	se	Least-square mean	se	Least-square mean	se	Least-square mean	se	Unadjusted model	Adjusted model
Age (years)	30·2a	0·3	29·6a,b	0·2	30·3a,b	0·6	28·5b	0·5	<0·01
Total energy intake (kJ)‡	920a	182·0	8155b	156·1	9866a	366·3	8966a	294·0		<0·0001¶
Total energy intake (kcal)	2201a	43·5	1949b	37·3	2358a	87·5	2143a	70·3		<0·0001¶
Obesity measurements
BMI (kg/m2)§	28·2a	0·4	25·8b	0·3	27·8a,b	0·9	29·8a	0·7		<0·0001**
Waist circumference (cm)	94·8a	0·9	79·4b	0·8	89·3c	1·9	88·4c	1·5		<0·0001**
Waist-to-hip ratio	0·9a	0·0	0·76b	0·0	0·84c	0·0	0·8d	0·0		<0·0001**
Triceps skinfold (mm)	18·5a	0·7	27·2b	0·6	16·7a	1·4	30·0b	1·1		<0·0001**
Physical activity	3·4a	0·1	3·1b	0·5	3·6a	0·1	3·2a,b	0·1		<0·0001††
Blood pressure (mmHg)
Systolic	113·3a	0·7	106·2b	0·6	117·0c	1·3	109·6d	1·1		<0·0001‡‡
Diastolic	75·8a	0·6	71·6b	0·5	77·4a	1·1	72·3b	0·9		<0·0001‡‡
Plasma glucose (mg/dl)	81·7a	0·8	77·5b	0·6	81·8a	1·5	78·9a,b	1·2		<0·0001‡‡
Plasma insulin (µU/ml)	12·3a,b	0·5	11·0a	0·5	13·9a,b	1·1	13·9b	0·9		<0·0001‡‡
Serum lipids
Total cholesterol (mg/dl)	200·1a	2·7	191·2b	2·3	199·3a,b	5·4	190·0a,b	4·3		<0·02‡‡
HDL-C (mg/dl)	41·1a	0·8	50·3b	0·7	48·4b	1·6	54·6c	1·3		<0·0001‡‡
LDL-C (mg/dl)	133·3a	2·3	120·2b	2·0	131·0a,b	4·6	122·2b	3·7		<0·0001‡‡
TAG (mg/dl)	139·9a	7·6	116·8b	6·5	108·5a,b,c	15·3	77·7c	12·3		<0·0001‡‡
	n	%	n	%	n	%	n	%
MetS∥	46	14·8	56	11·5	12	15·4a	7	5·8b		0·03

HDL-C, HDL cholesterol; LDL.C, LDL cholesterol; MetS, metabolic syndrome.

^a,b,c,dMean (percentage) values within a row with unlike superscript letters were significantly different (P < 0·01).

^*All young adults (n 995) were fasting before the blood draw.

^†Least-square means for dependent variables compared by analysis of covariance and Tukey–Kramer’s post hoc test.

^‡1 kcal54194 kJ.

^§Normal weight, BMI ≥ 18·5 and ≤24·9 kg/m² ; overweight/obese, BMI ≥ 25·0 kg/m².

^∥Diagnosis of MetS based on ≥3 of the following risk factors: waist circumference ≥102 cm in males and ≥88 cm in females; serum TAG ≥ 150 mg/dl; serum HDL-C < 40 mg/dl in males and <50 mg/dl in females; blood pressure ≥130 or ≥85 mmHg or taking medications for hypertension; and fasting plasma glucose ≥100 mg/dl or taking medications (oral hypoglycaemic agents/insulin).

^¶Energy intake model adjusted for age, socio-economic status (SES), smoking status, alcohol intake and physical activity.

^**Obesity measurement model(s) adjusted for energy intake, age, SES, smoking status, alcohol intake and physical activity.

^††Physical activity model (1 = sedentary, 3 = moderately active and 5 = very active) adjusted for energy intake, age and SES.

^‡‡Blood/plasma/serum parameters model(s) adjusted for energy intake, age, SES, smoking status, alcohol intake and physical activity.

Among the four ethnicity × gender groups, white females had the lowest energy intake, waist circumference, waist-to-hip ratio and SBP; white males had the highest waist circumference and waist-to-hip ratio but the lowest serum HDL-C; and black males had the highest SBP. Compared to white males, white females had lower energy intake, BMI, waist circumference, waist-to-hip ratio, SBP, DBP, plasma glucose, serum total cholesterol, LDL-C and TAG and physical activity, but higher triceps skinfold and serum HDL-C. Compared to black males, black females had higher triceps skinfold and serum HDL-C, but lower waist-to-hip ratio, SBP and DBP. Compared to black females, white females had lower energy intake, BMI, waist circumference, waist-to-hip ratio, SBP, plasma insulin and serum HDL-C, but higher serum TAG. Compared to black males, white males had higher waist circumference and waist-to-hip ratio, but lower SBP and serum HDL-C.

The overall occurrence of MetS in young adults was 12·2 %, with 14·9 % in males v. 10·4 % in females (P = 0·03, data not shown). No ethnic differences in the occurrence of MetS were observed (12·8 % in whites v. 9·6 % in blacks (P = 0·22, data not shown)). However, black males had a higher occurrence of Mets than black females (15·4 % v. 5·8 %; P = 0·03).

Covariate-adjusted associations between dietary patterns and components of metabolic syndrome (Table 4)

Table 4.

Covariate-adjusted associations between risk factors for MetS^* and DP in young adults (19–39 years)

	Model 1†				Model 2‡
	PDP (n 995)		WDP (n 995)		PDP (n 995)		WDP (n 995)
Risk factors (dependent variables)	Std β	P value	Std β	P value	Std β	P value	Std β	P value
Obesity measurements
BMI§	−0·07	0·15	−0·05	0·42	–	–	–	–
Waist circumference	−0·10	0·02	−0·05	0·46	−0·04	0·009	0·00	1·00
Triceps skinfold	−0·11	0·01	−0·02	0·75	−0·06	0·17	0·02	0·62
Blood pressure
Systolic	−0·05	0·29	−0·05	0·42	−0·03	0·51	−0·04	0·60
Diastolic	−0·01	0·80	0·00	0·99	0·01	0·86	0·17	0·80
Diabetes mellitus
Plasma glucose	−0·03	0·52	−0·05	0·49	−0·01	0·83	−0·03	0·64
Plasma insulin	−0·10	0·03	−0·04	0·53	−0·07	0·09	−0·01	0·83
Insulin resistance: HOMA-IR	−0·07	0·11	−0·04	0·58	−0·04	0·33	−0·01	0·89
Insulin sensitivity: QUICKI	0·16	<0·0005	0·02	0·78	0·13	0·001	−0·01	0·82
Lipid profiles
Total serum cholesterol	−0·01	0·88	−0·08	0·22	0·00	0·94	−0·08	0·27
Serum LDL-C	−0·01	0·91	−0·02	0·77	0·01	0·92	−0·01	0·93
Serum HDL-C	0·05	0·26	−0·11	0·10	0·03	0·45	−0·12	0·05
Serum TAG	−0·10	0·05	−0·13	0·05	−0·07	0·10	−0·12	0·07
MetS*
≥3 risk factors	−0·10	0·03	−0·08	0·22	−0·07	0·07	−0·05	0·35
	OR	CI	OR	CI	OR	CI	OR	CI
MetS*
(Reference = no MetS)	0·93	0·80, 1·07	0·93	0·80, 1·07	0·93	0·80, 1·07	0·93	0·80, 1·07

MetS, metabolic syndrome; DP, dietary patterns; PDP, prudent dietary pattern; WDP, Western dietary pattern; Std β, standardised β; HOMA-IR, Homeostasis

Model Assessment of insulin resistance; QUICKI, Quantitative Insulin Sensitivity Check Index; LDL-C, LDH cholesterol; HDL-C, HDL cholesterol.

^*Diagnosis of MetS based on ≥3 of the following risk factors: waist circumference ≥ 102 cm in males and ≥88 cm in females; serum triacylglycerol ≥ 150 mg/dl; HDL-C < 40 mg/dl in males and <50 mg/dl in females; blood pressure ≥ 130 or ≥85 mmHg or taking medications for hypertension; and fasting plasma glucose ≥ 100 mg/dl or taking medications (oral hypoglycaemic agents/insulin).

^†Model 1 adjusted for age, energy intake, ethnicity, gender, ethnicity × gender, socio-economic status (SES), physical activity, alcohol intake and smoking status.

^‡Model 2 adjusted for BMI in addition to age, energy intake, ethnicity, gender, ethnicity × gender, SES, physical activity, alcohol intake and smoking status.

^§BMI calculated as weight (kg)/height² (m²) with normal weight defined as BMI ≥18·5 and ≤24·9 kg/m² ; and overweight/obese defined as BMI ≥ 25·0 kg/m².

Using the covariate-adjusted model (excluding BMI), waist circumference, triceps skinfold, plasma insulin and the occurrence of MetS were all inversely associated with the PDP. Insulin sensitivity was positively associated with the PDP. Serum TAG was negatively associated with both PDP and WDP. After adjusting for BMI in addition to other covariates, serum HDL-C was inversely associated with the WDP. The overall occurrence of MetS did not differ by the two DP.

Covariate-adjusted demographic, socio-economic status and lifestyle differences in dietary patterns (Table 5)

Table 5.

Covariate-adjusted demographic, SES and lifestyle differences in servings of foods from DP among young adults (19–39 years): The Bogalusa Heart Study

		Fully adjusted model†		Unadjusted‡/partially adjusted model§
		PDP		WDP		PDP		WDP
Characteristics of young adults	Sample size (n)*	Mean	se	Mean	se	Mean	se	Mean	se
Demographics
Ethnicity
Whites	797	4·2	0·1	10·0	0·1	4·7	2·8	9·8	3·8
Blacks	198	4·0	0·2	10·5	0·2	4·6	2·4	11·1	4·2
		P = 0·46†		P = 0·02†		P = 0·74‡		P < 0·0001‡
Gender
Males	389	3·8	0·2	10·1	0·2	4·6	2·8	10·6	4·3
Females	606	4·5	0·2	10·4	0·2	4·7	2·7	9·4	3·6
		P = 0·002†		P = 0·18†		P = 0·45‡		P < 0·0001‡
Ethnicity × gender
White males	311	3·8a	0·2	10·1a	0·2	4·6	2·9	10·3a	4·1
White females	486	4·6b	0·1	10·0a	0·1	4·7	2·7	9·0b	3·5
Black males	78	3·7a,b	0·3	10·2a,b	0·3	4·5	2·0	11·5a	4·9
Black females	120	4·3a,b	0·3	10·9b	0·3	4·6	2·7	10·9a	3·8
		P = 0·8†		P = 0·07†		P = 0·87‡		P < 0·0001‡
		a v. b, P = 0·01†		a v. b, P = 0·01†
Age groups (years)
19–24	208	3·5a	0·2	10·5	0·2	4·2	2·8	10·2	4·2
25–29	273	4·0a,b	0·2	10·3	0·2	4·7	2·8	10·0	4·0
30–34	302	4·4b	0·2	10·1	0·2	4·7	2·6	9·6	3·6
35–39	212	4·5b	0·2	10·2	0·2	4·9	2·6	9·6	3·9
		P = 0·001†		P = 0·63†		P = 0·08‡		P = 0·30‡
SES*
Income ($)
≤ 15000	262	3·8	0·2	10·2	0·2	4·2a	2·2	10·6a	4·1
15 001–30000	259	4·0	0·2	10·4	0·2	4·5a,b	2·9	10·0a,b	3·9
30001–45000	195	4·2	0·2	10·3	0·2	4·8a,b	3·1	9·6a,b	3·7
≥45001	227	4·5	0·2	10·2	0·2	5·1b	2·6	9·0b	3·8
		P = 0·10†		P = 0·88†		P < 0·001§		P < 0·05§
Education (years)
≤12	272	3·6	0·2	10·7	0·2	3·9	1·9	10·7	4·1
>12	641	4·6	0·2	9·8	0·2	5·0	2·9	9·4	3·8
		P < 0·0001†		P < 0·0001†		P < 0·0001§		P < 0·001§
Lifestyle factors
Current smokers
No	665	4·07	0·2	9·9	0·1	4·9	2·7	9·5	3·9
Yes	330	4·14	0·1	10·6	0·2	4·1	2·5	10·6	4·0
		P < 0·0001†		P < 0·0001†		P < 0·0001§		P < 0·0001§
Current drinkers
No	261	4·5	0·2	10·3	0·2	4·6	2·7	9·7	3·9
Yes	734	3·7	0·2	10·2	0·1	4·7	2·7	9·9	4·0
		P = 0·74†		P = 0·57†		P = 0·56§		P = 0·38§
Physical activity (outside of work)*
1 (sedentary)	61	3·8	0·3	10·3a,b	0·3	4·5	2·5	10·1	3·8
2	156	3·9	0·2	10·8a	0·2	4·6	2·5	10·3	40
3 (moderately active)	457	4·0	0·2	10·3a,b	0·2	4·5	2·5	9·7	3·8
4	155	4·4	0·2	10·2a,b	0·2	4·8	3·1	9·5	3·8
5 (very active)	166	4·5	0·2	9·8b	0·2	50	30	10·0	4·6
		P = 0·08†		P = 0·02†		P = 0·31§		P = 0·15§
Marital status*
Unmarried	452	4·3	0·2	10·2	0·2	4·6	2·5	10·1	4·2
Married	533	4·0	0·2	10·3	0·2	4·6	2·9	9·6	3·7
		P = 0·09†		P = 0·55†		P = 0·73§		P = 0·66§

SES, socio-economic status; DP, dietary patterns; PDP, prudent dietary pattern; WDP, Western dietary pattern.

^a,bMean values within a column with unlike superscript letters were significantly different by analysis of covariance and Tukey–Kramer’s post hoc test (P < 0·05).

^*Sample size differs from the original sample (n 995) due to missing data for some subjects for SES, physical activity and marital status.

^†Fully adjusted model controlled for age, energy intake, gender, ethnicity, ethnicity × gender, SES, physical activity, smoking status, alcohol intake and marital status.

^‡Unadjusted model.

^§Partially adjusted model controlled for gender, ethnicity and ethnicity × gender.

Overall, young adults consumed more servings from the WDP than the PDP (mean 9·8 (sd 0·2) v. 4·5 (sd 0·2); P < 0·0001, data not shown). Blacks consumed more servings from the WDP than whites, and females consumed more servings from the PDP than males. Whites (males and females) consumed fewer servings from the WDP than black females. White females consumed more servings from the PDP than white males. Older young adults (30–39 years) consumed more servings from the PDP than their younger age group counterparts (19–24 years).

A higher percentage of young adults reported to be in the income category of <$15 000 (27·8 %) compared to $30 001–45 000 (20·6 %; P = 0·006, data not shown). Young adults reporting an income level of >$45 000 consumed more servings of the PDP than those reporting lower income, who consumed more servings from the WDP (income model showed significance when adjusted for only gender, ethnicity and ethnicity × gender, but not other covariates). Young adults with >12 years of education consumed more servings from the PDP than those with an education ≤12 years, who consumed more servings from the WDP. Current smokers consumed more servings from the WDP than current non-smokers, who consumed more servings from the PDP. Those who were physically very active (level 5) consumed fewer servings from the WDP than those who were sedentary (level 2).

Discussion

Factor analysis discerned two prominent DP in the present study, the ‘WDP’ mainly characterised by high-fat and high-refined carbohydrate foods, and the ‘PDP’ mainly characterised by low-fat and low-refined carbohydrate foods. A growing body of evidence suggests that increased consumption of healthier foods, including fruits and vegetables(34–37), whole grains/cereals(38), dairy products(39) and other low-fat foods(40), may prevent chronic nutrition-related diseases mainly by their vitamin/mineral(39,41), phytochemical(41) and fibre content(42,43). For example, whole grains have lower glycaemic index and higher fibre content than refined grains, and their consumption may increase insulin sensitivity(43) and plasma levels of anti-inflammatory cytokines (e.g. adiponectin)(44) and reduce serum markers of systemic inflammation (e.g. C-reactive protein and tumour necrosis factor alpha-receptor 2)(38). The calcium content in dairy foods has been hypothesised to lower central obesity and insulin resistance(39,45). Increased consumption of fruits and vegetables has been associated with lower incidence of stroke(34), ischaemic heart disease(34), hypertension(35), T2DM(36) and increased satiety(37), that may help to reduce body weight. Conversely, consumption of energy-dense (i.e. high-fat and/or high-refined carbohydrate foods) may contribute to a surplus intake of ‘discretionary calories’ in the diet (46) and may contribute to the prevalence of overweight/obesity and related chronic diseases over time.

The present study found that several risk factors for CHD, T2DM and MetS were associated with the DP (especially, the PDP). The occurrence of MetS (i.e. more than or equal to three MetS risk factors) was inversely associated with the PDP; however, no association was noted in the occurrence of MetS with the WDP in this study. In a recent longitudinal study(16), participants who were in the highest quintile of the WDP scores (comprising of refined grains, processed meat, fried foods and red meat) had an 18 % greater risk of MetS than those in the lowest quintile for the WDP scores; however, in the same study, consumption of the PDP was not associated with MetS(16). In another cross-sectional study(19), a dietary pattern characterised by a healthy balanced diet (with a frequent intake of raw and salad vegetables, fruits, fish, pasta and rice, and low intake of fried foods, sausages, fried fish and potatoes) was inversely correlated with central obesity, plasma glucose and TAG, and positively correlated with plasma HDL-C. The above dietary pattern(19) was also negatively associated with the risk of having undiagnosed diabetes, and this association was independent of age, gender, smoking and obesity.

In our present study, the finding of serum HDL-C being inversely associated with the WDP (after controlling for BMI and other covariates) is not in agreement with earlier theories(47,48). In general, diets high in saturated fatty acids tend to increase the cardio-protective serum HDL-C levels along with increasing other CHD-causing lipids (e.g. serum total cholesterol)(47). Conversely, low-fat, high-carbohydrate diets tend to decrease serum HDL-C, but also decrease serum total cholesterol and LDL-C(48).Whether the consumption of a mixture of high-fat and high-refined carbohydrate foods from the WDP led to the inverse association of serum HDL-C with the WDP, or whether the adjustment of BMI as a covariate in the model led to this finding, is not fully understood.

We also found an inverse association of serum TAG with both the DP. Diets high in refined carbohydrates and low in fat, tend to increase serum TAG owing to increased VLDL cholesterol TAG secretion, which is a result of increased hepatic fatty acid availability due to lower fatty acid oxidation(48). The WDP included a mixture of high-fat and high-refined carbohydrate foods that may have resulted in the inverse association of serum TAG with the WDP. Conversely, the inverse association of serum TAG with the PDP could be because of the consumption of low-refined carbohydrate foods (e.g. whole grains, legumes, vegetables and so on) as well as consumption of some low-fat foods (e.g. low-fat dairy products, poultry and low-fat salad dressings). Nevertheless, further investigation in this area is warranted.

Blacks and white males had more risk factors for MetS than white females in the present study. Further, blacks consumed more servings of the WDP, and females consumed more servings from the PDP. Earlier research has reported that blacks were less likely to modify meats to make them lower in fat and ate more fried foods than whites(49). Ethnic disparities in dietary intakes could be attributed to the fact that a larger number of blacks are in the lower SES group than whites(50) and hence may consume poorer diets(13,14) owing to either their inability to afford healthier foods(51), or may have a decreased accessibility to healthier foods(52). White females had the lowest BMI, waist circumference and waist-to-hip ratio than other ethnicity × gender groups, and consumed fewer servings from the WDP than black females. Black females (especially, from low SES groups) tend to be less health conscious than white females, and hence may be less likely to choose healthy dietary and lifestyle patterns(53,54).

In the present study, young adults in the higher SES group (especially, those with higher education) consumed more servings from the PDP. Similar results with respect to SES and food group consumption were suggested earlier(13,19). The relationship between higher SES and the consumption of a PDP could be because of increased knowledge and health awareness, or increased pressures of social acceptability that occur with the increasing SES, which may influence their food consumption habits(53). Among the lower SES groups, the increased cost of healthier foods or decreased access to healthier foods may be important factors influencing their food choices(51,52).

Current smokers in the present study consumed more servings of the WDP than current non-smokers who consumed more servings from the PDP. Those young adults who reported to be in the highest level of physical activity consumed lesser number of servings from the WDP than those who reported to be less physically active. Often, healthy dietary and lifestyle habits tend to cluster together among individuals. In a recent study(55), relative to non-smokers, current smokers reported higher overall energy intake, higher percentages of energy from fat, sweets and alcohol, and a lower percentage of energy from protein among low-income women. Yet, our study failed to find differences in DP by alcohol consumption and marital status.

The use of factor analysis to identify DP is a major strength of the present study. Although factor analysis takes into account the issue of high inter-correlations of foods within the diet, decisions based on factor loadings may be subjective or arbitrary and can affect the study results and interpretation(56). Nevertheless, a recent study depicted that the young adult age-group frequently consumed less healthier foods(11) resembling the WDP from the current study, suggesting that the factor loadings from our study are robust and the DP are meaningful. Also, the similarity in the results on DP and MetS from our study to those reported in longitudinal(16) and cross-sectional studies(17–20) strengthens the present findings.

The present study has some limitations. Owing to its cross-sectional design, causal inferences cannot be made(57). Despite the large sample size, blacks were under-represented. The findings from this study may be specific to young adults of Bogalusa, and are not representative of national findings. The YAQ, which was originally developed for the dietary assessment of adolescents, was used for the dietary assessment of young adults in the current study. In comparison with a 24 h dietary recall, the YAQ has been more helpful in characterising snack food consumption among the young adults(58). Further, the energy intake when measured by YAQ or the 24 h dietary recall in young adults has been similar(58). Lastly, the dietary data used in our study were collected over 10 years ago; yet, it is not uncommon to publish results from long-term epidemiological studies with data that were collected earlier (e.g. the Framingham offspring cohort study)(20,59). The present BHS findings are thus still noteworthy, as they provide valuable information on the role of DP, MetS and its association with socio-demographic and lifestyle factors. These findings may also help to generate new hypotheses for future research.

Conclusions

Overall, DP are important in identifying relationships with occurrence of diseases such as the MetS. Specifically, a prudent balanced dietary pattern may be helpful in preventing MetS in this sample of BHS young adults. More studies are warranted to confirm these findings in other populations. Nonetheless, nutrition intervention programmes for young adults to promote healthy dietary and lifestyle habits tailored-based on their SES, demographic and lifestyle characteristics, may be beneficial.