European Journal of Clinical Nutrition (2010) 64, 231–238 & 2010 Macmillan Publishers Limited All rights reserved 0954-3007/10 $32.00 www.nature.com/ejcn ORIGINAL ARTICLE Dietary patterns and socioeconomic position P Mullie1,2, P Clarys3, M Hulens1,4 and G Vansant1 1 Department of Nutrition, Preventive Medicine and Leuven Food Science and Nutrition Research Centre (LFoRCe), Catholic University Leuven, Leuven, Belgium; 2Unit of Epidemiology and Biostatistics, Queen Astrid Military Hospital, Brussels, Belgium; 3Faculty of Physical Education and Physiotherapy, Laboratory for Human Biometrics and Biomechanics, Vrije Universiteit Brussel, Brussels, Belgium and 4Department of Rehabilitation Sciences, Research Centre for Musculoskeletal Rehabilitation, Catholic University Leuven, Heverlee, Belgium Background/Objectives: To test a socioeconomic hypothesis on three dietary patterns and to describe the relation between three commonly used methods to determine dietary patterns, namely Healthy Eating Index, Mediterranean Diet Score and principal component analysis. Subjects/Methods: Cross-sectional design in 1852 military men. Using mailed questionnaires, the food consumption frequency was recorded. Results: The correlation coefficients between the three dietary patterns varied between 0.43 and 0.62. The highest correlation was found between Healthy Eating Index and Healthy Dietary Pattern (principal components analysis). Cohen’s kappa coefficient of agreement varied between 0.10 and 0.20. After age-adjustment, education and income remained associated with the most healthy dietary pattern. Even when both socioeconomic indicators were used together in one model, higher income and education were associated with higher scores for Healthy Eating Index, Mediterranean Diet Score and Healthy Dietary Pattern. The least healthy quintiles of dietary pattern as measured by the three methods were associated with a clustering of unhealthy behaviors, that is, smoking, low physical activity, highest intake of total fat and saturated fatty acids, and low intakes of fruits and vegetables. Conclusions: The three dietary patterns used indicated that the most healthy patterns were associated with a higher socioeconomic position, while lower patterns were associated with several unhealthy behaviors. European Journal of Clinical Nutrition (2010) 64, 231–238; doi:10.1038/ejcn.2009.145; published online 20 January 2010 Keywords: dietary pattern; nutritional assessment; nutritional epidemiology; public health; socioeconomic status; prevention Introduction Dietary pattern analysis, based on the concept that foods eaten together are as important as a reductive methodology characterized by a single food or nutrient analysis, has emerged more than a decade ago as an alternative approach to study the relation between nutrition and disease (Schwerin et al., 1982; Randall et al., 1991). As reviewed by Hu (Hu, 2002), dietary pattern analysis is a better method to examine Correspondence: Drs P Mullie, Department of Nutrition, Preventive Medicine and Leuven Food Science and Nutrition Research Centre (LFoRCe), Catholic University Leuven, Kapucijnenvoer 35 bus 7001, B-3000 Leuven, Belgium. E-mail: [email protected] Contributors: PM conceived the original idea together with GV, performed the study, analyzed the data and wrote the first draft; PC and MH conceived and refined the original idea, advised on study design and data analysis and critically appraised the paper. Received 10 March 2009; revised 24 October 2009; accepted 6 November 2009; published online 20 January 2010 the effect of overall diet: food and nutrients are not eaten in isolation, and the ‘single food or nutrient’ approach will not take into account the complex interactions between food and nutrients. Two major methods are used to reduce complex dietary data: a hypothesis-oriented approach using previous information to stratify a dietary pattern and a statistical approach using study-specific data to rank individuals, that is, principal component analysis or reduced rank regression models (Hoffmann et al., 2004; Schulze and Hoffmann, 2006). The Healthy Eating Index and the Mediterranean Diet Score are two frequently used hypothesis-oriented approaches (Waijers et al., 2007; Arvaniti and Panagiotakos, 2008; Fransen and Ocke, 2008; Kourlaba and Panagiotakos, 2009). The Healthy Eating Index represents the degree to which a dietary pattern conforms to official guidelines summarized in the United States Department of Agriculture Food Guide Pyramid (Kennedy et al., 1995). The Mediterranean Diet Score, according with the Mediterranean dietary pattern, has Education, income and nutrition P Mullie et al 232 received a lot of attention because of the associated reduction in mortality (Sofi et al., 2008). An example of commonly used exploratory approach is principal component analysis identifying foods that are consumed together. This statistical technique may be able to detect correlations between foods or food groups contained in an array of nutritional data. Few publications have reviewed the different methods to determine a dietary pattern (Hu, 2002; Kourlaba and Panagiotakos, 2009). A first aim of this study was to compare the degree of accordance of Healthy Eating Index, Mediterranean Diet Score and principal component analysis on the ranking of individuals according to their dietary patterns. The second aim of this study was to describe the relation between Healthy Eating Index, Mediterranean Diet Score and principal component analysis versus education and income as indicators for socioeconomic position. A healthy dietary pattern has been consistently associated with a higher socioeconomic position (Dynesen et al., 2003; Huot et al., 2004; Robinson et al., 2004; Park et al., 2005; Kant and Graubard, 2007). To our knowledge, the relation between socioeconomic position and different methods to describe a dietary pattern has seldom been studied. Materials and methods In February 2007, air and terrestrial components of the Belgian army totaled 33 053 men. After stratification for military rank and age, 5000 men were selected representative for the total army structure. The selection consisted of 598 officers; 2103 non-commissioned officers and 2299 soldiers. This population has the advantage to limit the influence of occupation as socioeconomic determinants allowing us to focus on the influence of income and education. A semi-quantitative food frequency questionnaire with 150 food items was mailed to the participants. The following categories of consumption frequency were used: never, one to three times a month, once a week, two to four times a week, five to six times a week, once a day, two to three times a day, four to six times a day and more than six times a day. Portion sizes were predefined using familiar measuring devices (teaspoon, glass, cup...). The validity of the questionnaire was tested on a sample of 100 men representative for the participants to the cross-sectional study (Mullie et al., 2009). A second questionnaire was used to register health-related and lifestyle characteristics. This questionnaire was selfreporting regarding smoking, marital status, main occupation, age, weight, height, number of children and knowledge of cardiovascular risk factors. This questionnaire was used in previous research (Autier et al., 2003). Yearly gross salary was obtained from administrative services, taking into account the rank and years of active duty. European Journal of Clinical Nutrition The individual characteristics of the responders were categorized in age-category (20 to 29 years, 30 to 39 years, 40 to 49 years and 50 to 59 years); Body mass index (BMI) classified according to the World Health Organization in normal weight, 18.5pBMI o25.0 kg/m2, overweight, 25.0p BMI o30.0 kg/m2 and obesity, BMIX30.0 kg/m2 (World Health Organisation, 2003); physical activity (stratified in low, moderate and high according to the International Physical Activity Questionnaire) (Hallal and Victora, 2004); use of vitamin supplements (yes or no); actual smoking (yes or no); educational level (low for vocational education, moderate for secondary level and high for bachelor or master level); and income (low for lowest tertile of yearly gross income of the group, moderate for intermediate tertile of income and high for highest tertile of yearly income). Participation was on a voluntary basis and without incentives. An informed consent was signed by all participants. Statistical analysis Using a w2 test, we assessed the differences in the proportion of officers, non-commissioned officers and soldiers that responded. Using data from military records, that is, age and rank, the differences between responders and nonresponders were tested with the same test. For descriptive statistics, mean and s.d. were calculated for the individual characteristics, according to quintiles of dietary patterns. Differences between quintiles were tested with w2 and analysis of variance. The Healthy Eating Index and the Mediterranean Diet Score were computed as described earlier (Basiotis et al., 2002; Sofi et al., 2008). The possible scores for Healthy Eating Index ranged from 0 to 100 and for Mediterranean Diet Score from 0 to 9, with a high score for the most healthy pattern. Principal component analysis was applied to the data of the semi-quantitative food frequency questionnaire. First, 150 food items were classified into 34 predefined food groups with similar nutrient profile, according to Hu et al. (2000). Principal components analysis was used to derive dietary patterns based on the 34 food groups. Varimax transformation was effectuated to achieve uncorrelated factors with a greater interpretability. Components with eigenvalues more than 1.5, interpretability of the factors and Scree plot were used to determine the number of selected factors. The eigenvalues of the factors dropped after the second factor (from 2.44 to 1.77) and after the third factor (from 1.77 to 1.44). The remaining factors were more similar after the fourth factor (ranging from 1.38 for the fifth factor to 1.10 for the tenth factor). Three major dietary patterns were clearly identified for further analysis. The factor scores for each pattern were constructed by summing up the observed intakes of the component food items, weighted by the individual factor loadings. Those factor scores rank individuals according to their agreement with each dietary pattern. The most healthy dietary pattern was selected, that Education, income and nutrition P Mullie et al 233 Table 1 Characteristics of the participants Characteristics Total Age (in years) Military rank Body mass indexa Physical activitya Regularly use of vitamin supplementation Actual smoking Educational levela Incomea Categorization Responders 20–29 30–39 40–49 50–59 Officers Non-commissioned officers Soldiers Normal (o25.0) Overweight (X25.0–o30.0) Obesity (X30.0) Missing Low Moderate High Missing Low Moderate High Low Moderate High Non-responders n % n % 1852 119 358 1064 311 217 936 699 744 836 244 28 365 383 1016 88 283 447 789 811 252 618 616 618 100.0 6.4* 19.3* 57.5* 16.8* 11.7* 50.5* 37.7* 40.2 45.1 13.2 1.5 19.7 20.7 54.9 4.8 15.3 24.1 42.6 43.8 13.6 33.4 33.3 33.4 3148 461 753 1439 495 381 1167 1600 100.0 14.6* 23.9* 45.7* 15.7* 12.1* 37.1* 50.8* a Body mass index (in kg/m2) was classified according to the World Health Organization (World Health Organisation, 2003); physical activity was stratified in low, moderate and high according to the International Physical Activity Questionnaire (Hallal and Victora, 2004); educational level is low for vocational education, moderate for secondary level and high for bachelor or master level; income is low for lowest tertile of yearly gross income, moderate for intermediate tertile of income and high for highest tertile of yearly income. *Po0.001. is, the Healthy Dietary Pattern (principal components analysis), because a high factor score is associated with the most healthy pattern, which is also the case for Healthy Eating Index and Mediterranean Diet Score. This Healthy Dietary Pattern was associated with a high intake of fruits, vegetables, nuts, fish, whole grain and low-fat diary products. Participants were divided in quintiles according to the scores for Healthy Eating Index, Mediterranean Diet Score and factor scores for the Healthy Dietary Pattern (principal components analysis). Spearman correlation coefficients, percentages of participants classified into the same and opposite quintiles of intake, and Cohen’s kappa coefficient of agreement were calculated for the three dietary patterns. Age-adjusted and BMI stratified linear regression was executed to separately estimate the independent effect of education and income categories on Healthy Eating Index, Mediterranean Diet Score and principal components analysis dietary pattern as continuous dependent variable. Tolerances were checked for all variables. Plots of the residuals versus the predicted values were examined to ascertain that basic model assumptions were met. Correlation between education and income was 0.4, which excluded multicollinearity problems. A two-sided level 0.05 level of significance was defined. SPSS 16.0 (SPSS Inc., Chicago, IL, USA) statistics software was used. The Bioethical Committee of the University of Leuven approved the complete research protocol. Results Table 1 presents the demographic and lifestyle characteristics of the participants. Out of the 5000 selected men, only 37% participated to the study. The most prevalent age-category was 40–49 years, 76% were non-smokers. Approximately 58% had a BMIX25.0 kg/m2 while 42.6% had a low level of education. Responders to the mailing tended to be older than non-responders (74.3% were older than 40 years compared with 61.4% for the non-responders) (Po0.001). Moreover, soldiers were less incline to participate to the study than officers and non-commissioned officers (Po0.001). Tables 2 and 3 present the factor groupings used in the principal component analysis and the factor-loading matrix for the three major factors identified by using food consumption data from the food frequency questionnaire. The greater the factor loading for a specific food or food item, the greater the effect of that food or food item to a specific factor. The first factor was heavily loaded with red meats, processed meats, beer, garlic, tomatoes, wine, eggs, poultry, liquor, organ meats and vegetables. This factor explained 7.4% of the total variance. This was labelled Meat European Journal of Clinical Nutrition Education, income and nutrition P Mullie et al 234 Table 2 Factor groupings used in the dietary pattern analysis Food or food groups Food items Processed meats Processed meats, bacon, hot dogs, salami, sausage, ham Beef, pork, lamb, hamburger Liver Fish Chicken or turkey Eggs and all types of preparations with eggs Butter Margarine Skim or low-fat milk or yoghurt or chocolate milk, buttermilk, low-fat cheese Whole milk or yoghurt or chocolate milk, half-and-half milk, cream, ice cream, all types of cheese Liquor Red and white wine Beer Tea Coffee Oranges, grapefruit, raisins, grapes, bananas, fresh apples or pears, strawberries, apricots, nectarines, cherries, kiwi, pineapple, peaches, plums Orange juice, other fruit juice Broccoli, coleslaw and uncooked cabbage, cooked cabbage, cauliflower, Brussels sprouts, kale, sauerkraut, carrots, yams, spinach, iceberg or other lettuce, celery, mushroom, green pepper, eggplant, all other vegetables Beans, peas, lentils, soybeans Tomatoes, tomato juice Garlic Potatoes French fries Cooked breakfast cereals, dark bread, brown rice, other grains, wheat germ Cold breakfast cereals White bread, biscuits, white rice, pasta, sandwiches Potato chips, pancakes Nuts Cola or other beverages with sugar Cola or other beverages without sugar Oil, vinegar Mayonnaise, dressings Home-made or ready-made soup Chocolate, candy bars, cookies, cake, pie, pastry, sugar, jam, waffles Red meats Organ meats Fish Poultry Eggs Butter Margarine Low-fat diary products High-fat diary products Liquor Wine Beer Tea Coffee Fruit Fruit juice Vegetables Legumes Tomatoes Garlic Potatoes French fries Whole grain Cold breakfast cereals Refined grains Snacks Nuts High-energy drinks Low-energy drinks Oil, vinegar Mayonnaise Soup Sweets and desserts Dietary Pattern. The second factor, explaining 7.2% of the total variance, was more loaded for tomatoes, fruit, low-fat diary products, whole grain, vegetables, cold breakfast cereals, fruit juice, fish, tea and nuts. This was labelled Healthy Dietary Pattern. The last factor, explaining 6.2% of the total variance, was heavily loaded with red meats, processed meats, sweets, desserts, snacks, high-energy drinks, high-fat diary products, refined grains, mayonnaise and potatoes. This was labelled Sweet Dietary Pattern. Table 4 presents the distribution of lifestyle and nutritional exposure in function of the quintiles of Healthy Eating Index, European Journal of Clinical Nutrition Table 3 Factor-loading matrix for the major factors identified by using food consumption data from the food frequency questionnairea Factor 1 (meat dietary pattern) Red meats Processed meats Beer Garlic Tomatoes Wine Eggs Poultry Liquor Organ meats Fruit Low-fat diary products Whole grain Vegetables Cold breakfast cereals Fruit juice Fish Tea Nuts Sweets and desserts Snacks High-energy drinks High-fat diary products Refined grains Mayonnaise Potatoes Factor 2 (healthy dietary pattern) Factor 3 (sweet dietary pattern) 0.60 0.58 0.47 0.43 0.43 0.40 0.38 0.37 0.37 0.33 0.58 0.47 0.43 0.39 0.38 0.37 0.36 0.32 0.30 0.53 0.45 0.42 0.40 0.36 0.30 0.30 a Absolute values o0.30 were excluded from the table for simplicity. Foods or food groups with factor loadings o0.30 for three factors were excluded; see Table 2 for food groupings. The percentage of explained variance was 7.4% for factor 1, 7.2% for factor 2 and 6.2% for factor 3. Mediterranean Diet Score and Healthy Dietary Pattern (principal components analysis). The range of the scores was for the lowest and the highest quintiles of Healthy Eating Index, Mediterranean Diet Score and Healthy Dietary Pattern, respectively 22–45 and 68–95; 0–2 and 6–9; 3.6 to 0.8 and 0.7–4.8. There was no relation between the quintiles of dietary patterns and age or BMI. The highest quintiles of Healthy Eating Index, Mediterranean Diet Score and Healthy Dietary Pattern were systematically associated with higher physical activity (all Po0.001), general use of vitamin supplements (all Po0.001), non-smoking (all Po0.001), high education (all Po0.05), high income (Po0.001, except for the Healthy Eating Index). The highest quintiles of the three dietary patterns were associated with the lowest daily intake of total fat (Po0.001), saturated fatty acids (Po0.001), mono-unsaturated fatty acids (Po0.001), poly-unsaturated fatty acids (Po0.001 except for the Mediterranean Diet Score) and with the highest intake of carbohydrates (Po0.001), all expressed in energy-percent. In Table 5, the Spearman correlation coefficients, percentages of participants classified into the same and opposite quintiles of intake, and Cohen’s kappa coefficient are presented. The highest correlation was found between Education, income and nutrition P Mullie et al 235 Table 4 Baseline characteristics according to Healthy Eating Index, Mediterranean Diet Score and Healthy Dietary Pattern (principal components analysis) (n ¼ 1852)a Healthy eating index a Q1 n Range scores Age (mean s.d.) Body mass indexa (mean s.d.) Mediterranean diet score a Q3 Q5 362 362 362 22–45 53–60 68–95 43.5 (7.2) 43.2 (6.9) 42.8 (7.2) 26.3 (3.9) 25.9 (3.4) 26.4 (3.5) Q1 Q3 Healthy dietary pattern (principal components analysis) a Q5 344 379 406 0–2 4–4 6–9 43.0 (7.2) 42.6 (7.3) 43.4 (7.2) 26.4 (4.0) 26.0 (3.4) 26.0 (3.3) Q1 Q3 Q5 362 362 362 3.6– 0.8 0.3–0.1 0.7–4.8 43.0 (7.4) 43.5 (6.6) 43.3 (6.8) 25.8 (3.7) 26.3 (3.4) 26.5 (3.6) Subject characteristics (%) a Physical activity Low Moderate High Use of vitamin supplements Smokinga Educational categoriesa Low Moderate High Income categoriesa Low Moderate High 30.1 22.5 47.4 11.0 34.8 22.2 21.3 56.5 11.3 24.0 10.4*** 21.6*** 68.0*** 20.4*** 10.8*** 25.2 23.4 51.4 7.8 32.6 20.9 19.8 59.3 14.2 21.4 14.9** 21.3** 63.8** 25.6*** 14.5*** 30.8 26.9 42.3 8.0 37.6 22.8 20.5 56.7 15.7 21.3 12.3*** 17.9*** 69.8*** 24.6*** 11.9*** 47.8 43.6 8.6 40.3 43.6 16.0 37.0* 47.2* 15.7* 48.8 42.4 8.7 41.4 44.1 14.5 33.7*** 46.6*** 19.7*** 45.0 46.1 8.8 45.9 42.3 11.9 40.3** 45.6** 14.1** 32.9 31.5 35.6 24.6 36.5 39.0 24.0 35.9 40.1 32.8 36.6 30.5 28.8 33.8 37.5 19.0*** 33.0*** 48.0*** 32.6 34.5 32.9 22.1 40.9 37.0 23.2** 35.1** 41.7** Nutrient intakes-mean (s.d.) in energy_percent Protein Total fat Saturated fat Monounsaturated fat Polyunsaturated fat Carbohydrates 15.4 46.0 18.8 17.4 8.5 35.0 (3.6) (5.9) (3.8) (2.8) (3.1) (5.9) 16.7 36.8 14.2 14.3 7.0 42.5 (3.2) (4.9) (3.1) (2.6) (2.6) (5.7) 17.3 28.7 10.2 11.5 5.8 49.8 (3.1)*** (4.9)*** (2.0)*** (2.5)*** (1.9)*** (6.3)*** 16.4 40.7 17.2 15.0 7.3 39.9 (3.3) (7.7) (4.1) (3.1) (2.6) (7.5) 16.2 37.0 14.1 14.5 7.2 43.0 (3.4) (7.8) (4.0) (3.3) (2.9) (7.9) 16.9 33.8 11.9 13.7 7.0 44.8 (3.1)* (7.1)*** (3.4)*** (3.1)*** (2.4) (7.4)*** 15.0 42.8 17.3 16.2 8.4 37.5 (3.4) (7.6) (4.2) (3.0) (3.2) (6.9) 16.7 37.1 14.1 14.7 7.0 42.5 (3.2) (6.3) (3.5) (3.1) (2.6) (6.4) 17.7 30.7 11.3 12.1 5.9 48.3 (3.3)*** (6.3)*** (3.2)*** (2.9)*** (1.9)*** (7.0)*** a Healthy Eating Index, Healthy Dietary Pattern (principal components analysis) (both expressed in 0 to 100 scale of agreement) and Mediterranean Diet Score (expressed in 0 to 9 scale of agreement); Body mass index measured in kg/m2; physical activity was measured in metabolic equivalent task (as calculated by the International Physical Activity Questionnaire); smoking expressed as actual smokers versus non-smokers or ex-smokers; educational categories were low (for vocational level), medium (for secondary level) and high (for bachelor or master level); income categories were low (for lowest tertile of yearly gross income), medium (for middle tertile of yearly gross income) and high (for highest tertile of yearly gross income). Physical activity, use of vitamin supplements, smoking, educational and income categories are expressed in percents. *Po0.05; **Po0.01; ***Po0.001. Healthy Eating Index and Healthy Dietary Pattern, the lowest between Healthy Eating Index and Mediterranean Diet Score. Surprisingly, the kappa coefficients expressed only slight agreements between the three dietary patterns. Table 6 expresses the relation between the socioeconomic indicators education and income and the three dietary patterns, stratified in normal weight, overweight and obesity. All the age-adjusted linear regressions showed the same relation, that is, a higher education or income level was associated with the most healthy dietary pattern, independently of the weigth-category. When the models were adjusted for age and for both indicators, the socioeconomic relation is attenuated but still present. In summary, a higher socioeconomic position was associated with an increasing score for Healthy Eating Index, Mediterranean Diet Score and Healthy Dietary Pattern: in the category normal weight, the score increased respectively with 1.59, 0.25 and 0.10 with increasing education; and with 1.38, 0.28 and 0.11 for increasing income (Table 6). Discussion The first aim of this work was to study the relation between three commonly used methods to determine dietary patterns, namely Healthy Eating Index, Mediterranean Diet Score and Healthy Dietary Pattern (principal components analysis). The low correlation and Cohen’s kappa coefficient of agreement did not influence the hypothesis, that is, the relation between socioeconomic indicators and dietary patterns. After age-adjustment, education and income remained associated with the most healthy dietary pattern. Even when both socioeconomic indicators were used together in a model, a higher income and education were associated European Journal of Clinical Nutrition Education, income and nutrition P Mullie et al 236 with a higher score for Healthy Eating Index, Mediterranean Diet Score and Healthy Dietary Pattern. Stratification in normal weight, overweight and obesity did not influence Table 5 Correlation coefficients, percentages of participants classified into the same and opposite quintiles of intake, and Cohen’s kappa coefficient (K) for Healthy Eating Indexb, Mediterranean Diet Scoreb and Healthy Dietary Pattern (principal components analysis) (n ¼ 1852)b Dietary pattern Percentage classified by in Spearman correlation coefficients Same Adjacent quintile quintile Healthy eating index— Mediterranean diet score Healthy eating index— Healthy dietary pattern (principal components analysis) Mediterranean diet score— Healthy dietary pattern (principal components analysis) K Non-adjacent quintile 0.43a 28.5 39.1 32.4 0.10 0.62a 36.0 40.1 23.9 0.20 0.49a 33.5 36.9 29.6 0.17 a Correlation is significant at the 0.01 level (two-tailed). Healthy Eating Index and Healthy Dietary Pattern (principal components analysis) (both expressed in 0 to 100 scale of agreement) and Mediterranean Diet Score (expressed in 0 to 9 scale of agreement). b the relation between socioeconomic indicator and dietary pattern. Although socioeconomic differences in prevalence of obesity have been described, that is, higher prevalence for the lowest socioeconomic positions (Mullie et al., 2008), dietary pattern analysis seems not to be able to detect a specific dietary pattern explaining this socioeconomic occurrence of obesity. The positive association between socioeconomic position and dietary pattern has been confirmed by research carried out on different populations, using different indexes or statistical techniques to determine dietary patterns. As reviewed by Darmon et al. (2008), higher values of Healthy Eating Index (Loughley et al., 2004; Angelopoulos et al., 2009; Manios et al., 2009), Diet Quality Index (Patterson et al., 1994; Lallukka et al., 2006), dietary diversity scores (Kant and Graubard, 2007), and other diet-quality measures (Groth et al., 2001; Dynesen et al., 2003; Robinson et al., 2004) have all been associated with a higher socioeconomic position, usually estimated by education level. Using principal components analysis to determine dietary patterns, Robinson et al. (2004) found that educational attainment was the most important determinant of a healthy eating pattern. The general observed socioeconomic nutritional gradient can be mediated by food costs, meaning that lowest cost diets mainly consumed by the lowest socioeconomic positions are generally unhealthy. People who have less Table 6 Age-adjusted linear regression with dietary pattern as continuous dependent variable: effects of educational categories and income categories unadjusted for each other (Model 1), and simultaneous adjustment for both socioeconomic indicators (Model 2) (n ¼ 1852) Healthy eating index—unstandardized linear regression coefficient (95% confidence interval) Model 1 b Body mass index Normal (n ¼ 744) Educational 1.59 (0.29–2.89)* categoriesa Income categoriesa 1.38 (0.15–2.61)* Model 2 Overweight (n ¼ 836) 1.50 (0.26–2.75)* Obesity (n ¼ 244) Normal (n ¼ 744) Overweight (n ¼ 836) 2.27 (0.26 to 4.79) 1.19 (0.26 to 2.65) 1.04 (0.36 to 2.44) Obesity (n ¼ 244) 1.90 (0.89 to 4.64) 1.37 (0.23–2.52)* 1.58 (0.74 to 3.90) 0.83 (0.55 to 2.20) 0.94 (0.34 to 2.20) 0.89 (1.64 to 3.41) Mediterranean Diet Score—unstandardized linear regression coefficient (95% confidence interval) Model 1 b Body mass index Normal (n ¼ 744) Educational 0.25 (0.06–0.44)* categoriesa Income categoriesa 0.28 (0.10–0.46)** Model 2 Overweight (n ¼ 836) 0.43 (0.26–0.60)*** Obesity (n ¼ 244) Normal (n ¼ 744) Overweight (n ¼ 836) 0.18 (0.15 to 0.51) 0.15 (0.06 to 0.36) 0.33 (0.14–0.52)*** Obesity (n ¼ 244) 0.08 (0.28 to 0.44) 0.34 (0.18–0.49)*** 0.24 (0.06 to 0.54) 0.22 (0.01–0.42)* 0.21 (0.12 to 0.54) 0.20 (0.03–0.37)* Healthy Dietary Pattern (principal components analysis)—unstandardized linear regression coefficient (95% confidence interval) Model 1 Model 2 Body mass indexb Normal (n ¼ 744) Overweight (n ¼ 836) Obesity (n ¼ 244) Normal (n ¼ 744) Overweight (n ¼ 836) Obesity (n ¼ 244) Educational 0.10 (0.01 to 0.20) 0.06 (0.04 to 0.15) 0.05 (0.13 to 0.24) 0.06 (0.06 to 0.18) 0.01 (0.09 to 0.12) 0.07 (0.13 to 0.27) categoriesa Income categoriesa 0.11 (0.01–0.21)* 0.09 (0.01–0.18)* 0.01 (0.18 to 0.16) 0.08 (0.03 to 0.19) 0.08 (0.02 to 0.18) 0.03 (0.22 to 0.15) a All analyses were adjusted for age as continuous variable; operationalization of variables: the continuous dependent variable Healthy Eating Index, Healthy Dietary Pattern (principal components analysis) (both expressed in 0 to 100 scale of agreement) and Mediterranean Diet Score (expressed in 0 to 9 scale of agreement); educational categories were low (for vocational level), medium (for secondary level) and high (for bachelor or master level); income categories were low (for lowest tertile of yearly gross income), medium (for middle tertile of yearly gross income) and high (for highest tertile of yearly gross income). b Body mass index (in kg/m2) was classified according to the World Health Organization (World Health Organisation, 2003) in normal (o25.0 kg/m2), overweight (X25.0–o30.0 kg/m2) and obesity(X30.0 kg/m2). *Po0.05; **Po0.01; ***Po0.001. European Journal of Clinical Nutrition Education, income and nutrition P Mullie et al 237 money, choose to buy cheaper foods, and these cheaper foods are less healthy (Drewnowski, 2003; Drewnowski and Darmon, 2005; Darmon and Drewnowski, 2008). In this study, the least healthy quintiles of dietary pattern measured by the three methods were associated with a clustering of unhealthy behaviors, that is, smoking, low physical activity, highest intake of total fat and saturated fatty acids, and low intakes of fruits and vegetables. The correlation between Healthy Eating Index and Mediterranean Diet Score was 0.43, a rather moderate correlation coefficient. This may be explained by the fact that the two hypothesis-based indexes have different foundations: the United States Department of Agriculture Food Guide Pyramid for the Healthy Eating Index and the Mediterranean dietary pattern for the second. Both indexes had a rather comparable score system for vegetables, fruits, milk and meat but the other components will differ. The Mediterranean Diet Score focused on the ratio monounsaturated fatty acids on saturated fatty acids, legumes and alcohol; the Healthy Eating index on total fat, saturated fat, cholesterol, sodium and diet variety. A major drawback of the Healthy Eating Index is that it is unable to distinguish between whole grains and refined grains, which will limit the capacity to assess dietary fibers (Arvaniti and Panagiotakos, 2008). Second, in the Mediterranean Diet Score the median intake of each component serves as cut-off value. This approach does not automatically means that a high score is associated with a healthy level of intake. The consequence of the binary method used to categorize intakes in two groups is that people with moderate high or low intake of a component are classified in the same category as people with a very high or very low intake. The advantage of using the median is that half of the participants will score positively and half negatively, creating enough contrast for further research (Waijers et al., 2007). The Healthy Eating Index is a scoring system based on current views of healthy eating, which allows more comparisons between populations because the scoring system remains the same. A major disadvantage of the Healthy Eating Index is the low discriminative power of the components if all the participants have a low score for a component. In Western diets with a high intake of saturated fatty acids, the discriminative power of those fatty acids could be very low. Moreover, the fact that energy intake may be a confounder, that is, participants with a high intake will more easily meet the guidelines, can not be excluded for the Mediterranean Diet Score. The scoring system of the Healthy Eating Index depends on the recommended energy intakes: the adequate number of servings is expressed according to energy intake level, based on sex and age. The high correlation coefficient between Healthy Eating Index and Healthy Dietary Pattern (principal components analysis) can be the consequence of overlapping components: high scores for Healthy Eating Index and for Healthy Dietary Pattern (principal components analysis) are characterized by high intake of fruits, vegetables, cereals and low intake of meat and diary products, with low intake of total fat, cholesterol and saturated fatty acids as a consequence. The use of principal component analysis involves that several arbitrary decisions must be taken, such as the number of retained factors and the labels of the factors. The value of the labelling can be judged from the presented factor loadings. Moreover, the percentage of the variance explained by the factors in this study (20.8%) is comparable to other studies using comparable statistical methods (Hu et al., 2000; van Dam et al., 2003; Slattery, 2008). Three distinct dietary patterns were identified; similar factor loadings were extracted in other studies when two or three major patterns were selected (Hu et al., 2000; Schulze et al., 2001; Kim et al., 2004; Park et al., 2005). Some limitations of this study are worth noting. The response in this study was only 37%, but information could be gathered regarding non-responders. The responders were older than non-responders. A military population was selected for this study. This population has the advantage of limiting the influence of occupation as socioeconomic determinant, which allowed us to restrain our investigations to the influence of income and education as socioeconomic indicators. A second advantage is that we could have exact figures regarding income from administrative services. The sample can be observed as representative for Belgian army men. However, because of the different manual and non-manual tasks, occupations and education levels present in an army, our sample can be observed as a representative sample for men with an occupation. Moreover, our nutritional and lifestyle results match with the results of a recent cross-sectional study on a representative Belgian male population (Devriese et al., 2006). The low correlation between education and income in this study (r ¼ 0.40) indicate that each indicator involve different components of exposition variability. The correlation was comparable to other publications (Liberatos et al., 1988; Winkleby et al., 1992; Galobardes et al., 2001; Braveman et al., 2005). Colinearity and unstable models did not influence, because the correlation between education and income is below 0.50 (Winkleby et al., 1992; Turrell et al., 2003; Braveman et al., 2005). In conclusion, in this study, a higher socioeconomic position as measured by education and income was systematically associated with a more healthy dietary pattern, independently of the method to determine the dietary pattern. Healthy Eating Index, Mediterranean Diet score and Healthy Dietary Pattern (principal components analysis) obtained a comparable ranking. From a practical point of view, the choice of the Healthy Eating Index or the Mediterranean Diet Score seems to be obvious, because those two methods are less time consuming than principal component analysis. Moreover, principal component analysis involves that several arbitrary decisions must be taken. Conflict of interest The authors declare no conflict of interest. European Journal of Clinical Nutrition Education, income and nutrition P Mullie et al 238 Acknowledgements The authors are indebted to the participants of this study. The authors thank Ms Jeanine De Leeuw for her valuable assistance in realizing this study. References Angelopoulos P, Kourlaba G, Kondaki G, Fraqiadakis G, Manios Y (2009). Assessing children’s diet quality in Crete based on Healthy Eating Index: the children study. Eur J Clin Nutr 63, 964–969. Arvaniti F, Panagiotakos DB (2008). Healthy indexes in public health practice and research: a review. Crit Rev Food Sci Nutr 48, 317–327. Autier P, Creplet J, Vansant G, Brohet C, Paquot N, Muls E et al. (2003). The impact of reimbursement criteria on the appropriateness of ‘statin’ prescribing. Eur J Cardiovasc Prev Rehabil 10, 456–462. Basiotis PP, Carlson A, Gerrior SA, Juan WY, Lino M (2002). The Healthy Eating Index: 1999–2000. US Department of Agriculture, Center for Nutrition Policy Promotion. CNPP-12. Braveman PA, Cubbin C, Egerter S, Chideya S, Marchi KS, Metzler M et al. (2005). Socioeconomic status in health research: one size does not fit all. JAMA 294, 2879–2888. Darmon N, Drewnowski A (2008). Does social class predict diet quality? Am J Clin Nutr 87, 1107–1117. Devriese S, Huybrechts I, Moreau M, Van Oyen H (2006). The Belgian Food Consumption Survey 1–2004. IPH/EPI Reports No 2006–016 Epidemiology Unit, Scientific Institute of Public Health: Brussels, Belgium. Drewnowski A (2003). Fat and sugar: an economic analysis. J Nutr 133, 838S–840S. Drewnowski A, Darmon N (2005). The economics of obesity: dietary energy density and energy cost. Am J Clin Nutr 82, 265S–273S. Dynesen AW, Haraldsdottir J, Holm L, Astrup A (2003). Sociodemographic differences in dietary habits described by food frequency questions—results from Denmark. Eur J Clin Nutr 57, 1586–1597. Fransen HP, Ocke MC (2008). Indices of diet quality. Curr Opin Clin Nutr Metab Care 11, 559–565. Galobardes B, Morabia A, Bernstein MS (2001). Diet and socioeconomic position: does the use of different indicators matter? Int J Epidemiol 30, 334–340. Groth MV, Fagt S, Brondsted L (2001). Social determinants of dietary habits in Denmark. Eur J Clin Nutr 55, 959–966. Hallal PC, Victora CG (2004). Reliability and validity of the International Physical Activity Questionnaire (IPAQ). Med Sci Sports Exerc 36, 556. Hoffmann K, Schulze MB, Schienkiewitz A, Nothlings U, Boeing H (2004). Application of a new statistical method to derive dietary patterns in nutritional epidemiology. Am J Epidemiol 159, 935–944. Hu FB (2002). Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 13, 3–9. Hu FB, Rimm EB, Stampfer MJ, Ascherio A, Spiegelman D, Willett WC (2000). Prospective study of major dietary patterns and risk of coronary heart disease in men. Am J Clin Nutr 72, 912–921. Huot I, Paradis G, Receveur O, Ledoux M (2004). Correlates of diet quality in the Quebec population. Public Health Nutr 7, 1009–1016. Kant AK, Graubard BI (2007). Secular trends in the association of socio-economic position with self-reported dietary attributes and biomarkers in the US population: National Health and Nutrition Examination Survey (NHANES) 1971–1975 to NHANES 1999–2002. Public Health Nutr 10, 158–167. European Journal of Clinical Nutrition Kennedy ET, Ohls J, Carlson S, Fleming K (1995). The healthy eating index: design and applications. J Am Diet Assoc 95, 1103–1108. Kim MK, Sasaki S, Sasazuki S, Tsugane S (2004). Prospective study of three major dietary patterns and risk of gastric cancer in Japan. Int J Cancer 110, 435–442. Kourlaba G, Panagiotakos DB (2009). Dietary quality indices and human health: a review. Maturitas 62, 1–8. Lallukka T, Laaksonen M, Rahkonen O, Roos E, Lahelma E (2006). Multiple socio-economic circumstances and healthy food habits. Eur J Clin Nutr 61, 701–710. Liberatos P, Link BG, Kelsey JL (1988). The measurement of social class in epidemiology. Epidemiol Rev 10, 87–121. Loughley K, Basiotis P, Zizza C, Dinkins J (2004). Profiles of selected target audiences: promoting the dietary guidelines for Americans. Fam Econ Nutr Rev 13, 3–14. Manios Y, Kourlaba G, Kondaki G, Grammatikaki E, Birbilis M, Oikonomou E et al. (2009). Diet quality of preschoolers in Greece based on the Healthy Eating Index: the GENESIS study. J Am Diet Assoc 109, 616–623. Mullie P, Clarys P, Hulens M, Vansant G (2009). Reproducibility and validity of a semiquantitative food frequency questionnaire among military men. Mil Med 174, 852–856. Mullie P, Vansant G, Guelinckx I, Hulens M, Clarys P, Degrave E (2008). Trends in the evolution of BMI in Belgian army men. Public Health Nutr 18, 1–5. Park SY, Murphy SP, Wilkens LR, Yamamoto JF, Sharma S, Hankin JH et al. (2005). Dietary patterns using the Food Guide Pyramid groups are associated with sociodemographic and lifestyle factors: the multiethnic cohort study. J Nutr 135, 843–849. Patterson RE, Haines PS, Popkin BM (1994). Diet quality index: capturing a multidimensional behavior. J Am Diet Assoc 94, 57–64. Randall E, Marshall J, Graham S, Brasure J (1991). High-risk health behaviors associated with various dietary patterns. Nutr Cancer 16, 135–151. Robinson SM, Crozier SR, Borland SE, Hammond J, Barker DJ, Inskip HM (2004). Impact of educational attainment on the quality of young women0 s diets. Eur J Clin Nutr 58, 1174–1180. Schulze MB, Hoffmann K (2006). Methodological approaches to study dietary patterns in relation to risk of coronary heart disease and stroke. Br J Nutr 95, 860–869. Schulze MB, Hoffmann K, Kroke A, Boeing H (2001). Dietary patterns and their association with food and nutrient intake in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam study. Br J Nutr 85, 363–373. Schwerin H, Stanton J, Smith J, Riley A, Brett B (1982). Food, eating habits, and health: a further examination of the relationship between food eating patterns and nutritional health. Am J Clin Nutr 35, 1319–1325. Slattery ML (2008). Defining dietary consumption: is the sum greater than its parts? Am J Clin Nutr 88, 14–15. Sofi F, Cesari F, Abbate R, Gensini GF, Casini A (2008). Adherence to Mediterranean diet and health status: meta-analysis. BMJ 337, a1344. Turrell G, Hewitt B, Patterson C, Oldenburg B (2003). Measuring socio-economic position in dietary research: is choice of socioeconomic indicator important? Public Health Nutr 6, 191–200. van Dam RM, Grievink L, Ocke MC, Feskens EJ (2003). Patterns of food consumption and risk factors for cardiovascular disease in the general Dutch population. Am J Clin Nutr 77, 1156–1163. Waijers PM, Feskens EJ, Ocke MC (2007). A critical review of predefined diet quality scores. Br J Nutr 97, 219–231. Winkleby MA, Jatulis DE, Frank E, Fortmann SP (1992). Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease. Am J Public Health 82, 816–820. World Health Organisation (2003). Process for a WHO Global Strategy on Diet, Physical Activity and Health. WHA55.23.
© Copyright 2026 Paperzz