The Dietary Quality Score: validation and association with ... - Nature

European Journal of Clinical Nutrition (2007) 61, 270–278

& 2007 Nature Publishing Group All rights reserved 0954-3007/07 $30.00 www.nature.com/ejcn

ORIGINAL ARTICLE

The Dietary Quality Score: validation and association with cardiovascular risk factors: the Inter99 study U Toft1, LH Kristoffersen1, C Lau2, K Borch-Johnsen2 and T Jørgensen1 1 Research Centre for Prevention and Health, Glostrup University Hospital, Glostrup, Denmark and 2Steno Diabetes Center, Gentofte, Denmark

Objective: To develop and assess the validity of the Dietary Quality Score (DQS) and investigate the association with cardiovascular risk factors. Design: Cross-sectional population-based study. Setting: Copenhagen County, Glostrup, Denmark. Subjects: A random sample of 12 934 men and women aged 30–60 years were invited to a health examination. A total of 6542 participants were included in the statistical analysis. Methods: The DQS was developed using eight questions from a 48-item food frequency questionnaire (FFQ) and validated using a 198-item FFQ. Associations between the DQS and fasting serum lipids, homocysteine, blood pressure and the absolute risk of ischaemic heart disease (IHD) were explored. Results: A higher DQS was shown to be associated with higher dietary quality in general, including a low intake of fat, especially saturated fat; a high intake of fibre; various vitamins and minerals; and fruit, fish, vegetables and whole–grain products. A higher score according to the DQS was significantly negatively associated with total cholesterol (P ¼ 0.0031), triglyceride (P ¼ 0.0406), low-density lipoprotein-cholesterol (P ¼ 0.0071), homocysteine (Po0.0001) and the absolute risk of IHD (Po0.0001), adjusted for sex, age, smoking habits and physical activity level. Conclusions: The DQS is a simple, valid and quick tool to make a rough classification of individuals into groups with high, average and low dietary quality. The DQS is negatively associated with serum lipids, homocysteine and the absolute risk of IHD. Sponsorship: The Inter99 study is supported economically by The Danish Medical Research Council, The Danish Centre for Evaluation and Health Technology Assessment, Novo Nordisk, Copenhagen County, The Danish Heart Foundation, The Danish Pharmaceutical Association, Augustinus Foundation, Ib Henriksen Foundation and Becket Foundation, Copenhagen County.

European Journal of Clinical Nutrition (2007) 61, 270–278. doi:10.1038/sj.ejcn.1602503; published online 23 August 2006 Keywords: diet index; validation studies; diet quality; food frequency questionnaire

Introduction During the last decade, both observational and intervention studies have provided evidence of the effect of dietary Correspondence: U Toft, Research Centre for Prevention and Health, Glostrup University Hospital, Nordre Ringvej 57, Building 84/85, DK-2600 Glostrup, Denmark. E-mail: [email protected] Guarantor: U Toft.M.Sc. Contributors: TJ was responsible for the design and conduct of the Inter99 study. UT was responsible for the construction of the score, conducting the data analysis, interpretation of the data and writing of the manuscript. All the authors participated in editing the manuscript and provided advice regarding interpretation of the results. None of the contributing authors had any financial or personal interests in any of the bodies sponsoring this research. Received 6 September 2005; revised 30 March 2006; accepted 20 June 2006; published online 23 August 2006

patterns on especially all-cause mortality and cardiovascular disease (Kant, 2004). Thus, studies of elderly people have shown that dietary patterns were a stronger predictor for survival than single nutrients or food components (Huijbregts et al., 1995; Trichopoulou et al., 1995) and Tucker et al. (2005) recently demonstrated an additive protective effect of low intake of saturated fat and high intake of fruit and vegetables in relation to coronary heart disease mortality. Similar clinical trials have shown that changing dietary pattern rather than single dietary components is more effective in lowering blood pressure (Appel et al., 1997; Conlin et al., 2000; Sacks et al., 2001). Investigating the effect of the overall dietary habits is therefore a relevant alternative or supplement to the traditional approach of investigating the effect of single dietary components.

The Dietary Quality Score U Toft et al

271 Hence, a variety of dietary scores and indexes (Patterson et al., 1994; Huijbregts et al., 1995; Kennedy et al., 1995; Kant and Thompson, 1997; Haines et al., 1999; Trichopoulos et al., 2003) have been developed to measure the overall dietary habits, but no gold standard has emerged. Most of the dietary indexes that have been developed require extensive dietary assessment methods. However, when investigating a large study population and especially when the assessment of the nutrition intake of individuals is not the main focus of a study, it is necessary to have dietary assessment tools that are valid, short and easy to administer. Although a simple dietary assessment tool, such as a short food frequency questionnaire (FFQ), does not make it possible to compute the nutrition intake in detail, a rough classification of the participants into groups with low, medium or high intake still permits the examination of nutritional hypotheses and the assessment of dose–response relationships (Block, 1982). The challenge, however, is to correctly classify individuals despite the shortness of the instrument. A few indexes on overall dietary quality based on a short FFQ have been developed (Kant et al., 2000; Osler et al., 2001; Huot et al., 2004a, b; Massari et al., 2004), but the validations of most of these have been sparse. The objective of the present study was to develop a Dietary Quality Score (DQS) based on a short FFQ and to validate it against a 198-item FFQ. Furthermore, the objective was to investigate if the score was associated with biological risk factors for cardiovascular disease.

classes: (1) no vocational training, (2) up to 1 year of vocational training, (3) 2–4 years of vocational training and (4) over 4 years of vocational training. Physical activity during leisure time was recorded by the participants scoring themselves into one of four categories, mainly sedentary, moderate activity, regular exercise and heavy training (Saltin and Grimby, 1968). On the day of attendance at the centre, all participants were asked to be fasting from midnight. Fasting blood samples were drawn for assessment of total cholesterol, high-density lipoprotein (HDL)-cholesterol, triglyceride and homocysteine. Homocysteine level was only determined on approximately half of the Inter99 population (2788 persons), selected a priori. Homocysteine was determined using a fluorescent polarization immunoassay. Cholesterol and triglyceride were determined with enzymatic techniques (Boeringer, Mannheim, Germany). Low-density lipoprotein (LDL)-cholesterol was calculated by Friedewalds’s equation (Friedewald et al., 1972). Blood pressure was measured twice after 5 min of rest in the lying position. Height was measured without shoes to the nearest 0.5 cm and weight was measured without shoes and overcoat to the nearest 0.1 kg. Waist circumference was measured midway between the lower rib margin and iliac crest. Body mass index (BMI) was calculated (kg/m2). For each participant, the 10-year absolute risk of ischaemic heart disease (IHD) at a fixed age (60 years) was estimated using The Copenhagen Risk Score (Thomsen et al., 1997). During their stay at the centre, the participants were additionally asked to complete a 198-item FFQ (long FFQ).

Study population and methods This study used baseline data from the Danish populationbased Inter99 study, which is an intervention study on diet, physical activity and smoking. The study aimed to decrease the incidence of cardiovascular diseases. The study population comprised all 61 301 individuals born in selected years (1939–1965) from a defined area of the suburb of Copenhagen. All individuals were drawn from the Civil Registration System. An age- and sex-stratified random sample of 13 016 individuals was drawn from the study population. Of the 13 016 people sampled, 82 individuals were non-eligible, as they had died or could not be traced. The remaining 12 934 subjects were invited for a healthscreening programme at The Research Centre for Prevention and Health. A total of 6906 (53.4%) subjects turned up for the investigation. Out of these, 122 subjects were excluded because of alcoholism or drug abuse (n ¼ 23) or linguistic barriers (n ¼ 99). The Inter99 study is described in detail elsewhere (Jorgensen et al., 2003). The invitation included a detailed questionnaire to be completed before attendance at the centre including information on sociodemographic and lifestyle factors. Dietary habits were assessed by a 48-item FFQ. Education was defined on the basis of questions regarding number of years of vocational training categorized into four

Data on dietary intake Data on dietary intake used for developing the DQS were assessed using the 48-item FFQ. It included questions about the type of bread, spread and fats used for cooking. The participants were further asked how often 27 food items (including hot meals, accompaniment to hot meals, vegetables, etc.) were consumed the last week choosing between four possible responses: 0, 1–2, 3–4 or 5–7 times a week. For fruit intake, there were eight possible responses ranging from none to more than six pieces a day. The long FFQ covered 198 food items and beverages. It was based on FFQs previously used and validated in the Diet, Cancer and Health Study and The Danish National Birth Cohort (Tjonneland et al., 1991; Olsen et al., 2001), but the questionnaire was modified to improve estimates of fatty acids, cholesterol and complex carbohydrates. The long FFQ has been described previously (Tjonneland et al., 1991; Olsen et al., 2001; Lau et al., 2004). Briefly, participants were asked to report their average intake of different foods and beverages the last month, choosing between 7 and 11 possible responses, ranging from never to eight or more times a day. The questionnaire also included questions about the types of bread, spread and fat used for cooking. The food consumption quantity was obtained by multiplying the frequency of consumption of each unit of food by standard European Journal of Clinical Nutrition


272 portion sizes (Haraldsdottir et al., 1998; Biltoft-Jensen, unpublished work). To translate food consumption into energy intake and daily nutrient intake, all food items were linked to food items in the Danish Food Composition Databank version 6 (in this paper we used an updated version compared to the earlier study by Lau et al., 2004) (Saxholt et al., 2005). The software program FoodCalc version 1.3 (Lauritsen, 1998) was used for the calculations. Developing the DQS Using the 48-item FFQ, we developed the DQS as a crude index of the overall quality of the dietary habits. The calculations were based on indices of overall diet quality (Kant, 1996) and the Danish Dietary Guidelines (The National Consumer Agency, 1995; Trolle et al., 1998; Becker et al., 2004) that emphasize consumption of at least 600 g of fruit and vegetables each day, at least 200 g of fish a week and a low intake of total fat (o30% of the energy intake) and saturated fat (o10% of the energy intake). The score was based on the following questions: fruits, boiled vegetables, raw vegetables, vegetable or vegetarian dishes, fish for lunch, fish for dinner, fat spread and fat used for preparation. A three-point scoring system was developed for each of the four food groups: fish, fruits, vegetables and fats. The recommended intake of fish, fruits and vegetables (The National Consumer Agency, 1995; Trolle et al., 1998), respectively, was used to approximate the upper cutoff point for the corresponding food groups. The cutoff point for a high score in the fat group was no use of spread and no use of fat or use of olive oil for cooking. The lower cutoff point was defined as only using saturated fats for cooking and spread; no intake of fish; or a very low intake of fruits (p2 pieces/ week) and vegetables (p2 servings/week), respectively. Thereby, the participants were categorized into three groups in each food group: two smaller groups with a very healthy (three points) and a very unhealthy (one point) intake, respectively, according to the recommendations; and a larger group with an average intake (two points). Summation of the four food variables resulted in a score ranging from 1 to 12 points. The construction of the DQS is shown in Table 1. A total of 242 participants did not complete the 48-item FFQ, leaving 6542 for analyses. Additionally, 206 subjects did not complete the 198-item FFQ or clearly misunderstood the questionnaire (by repeatedly putting two or more marks for each questions), leaving 6336 for analyses including the long FFQ. Subjects treated with lipid-lowering medication (n ¼ 90) or antihypertensive medication (n ¼ 470) were excluded from the analyses of the association between the DQS and cardiovascular risk factors.

Statistical analysis Statistical analyses were carried out using the SAS statistical software package (version 9.1, SAS Institute Inc., Cary, NC, USA). European Journal of Clinical Nutrition

Dietary intake according to the 198-item FFQ was used to evaluate whether the DQS adequately reflects dietary quality. Linear trend across the categories was evaluated by modelling the score as a continuous variable in the model and testing for model reduction. Because the data on dietary intake were not normally distributed, correlation coefficients between the DQS and various food and nutrient intakes were assessed using Spearman’s correlation coefficient. Furthermore, gamma coefficients were calculated. The association between the DQS and socioeconomic and lifestyle factors was examined in an ordinal logistic regression analysis, with the DQS as the dependent variable and education, sex, age, smoking and physical activity as independent or confounding variables. The association between the DQS and various cardiovascular risk factors (serum lipids, homocysteine, systolic blood pressure, waist circumference, BMI and The Copenhagen Risk Score) was evaluated through two sets of linear regression models: linear regression models adjusted for sex and age; multiple linear regression models adjusted for sex, age, physical activity and smoking. To investigate the confounding effect of physical activity and smoking, these variables were eliminated to check for changes in beta coefficients. Finally, BMI was included in the multiple regression models. In the analysis with BMI as the dependent variables, waist circumference was included instead of BMI. In the analysis of the association between the DQS and The Copenhagen Risk Score, there was only adjustment for physical activity, because BMI and smoking are incorporated in the score. Linear trend across the DQS was evaluated as described above. F-tests were used to test for significance, defined as Po0.05. The standardized residuals of the final regression models were examined to verify assumptions of linearity, variance homogeneity and normality. To normalize the distribution and to improve the variance homogeneity of the residuals, homocysteine, triglyceride and The Copenhagen Risk Score were natural logarithmically transformed.

Results Characteristics of the participants by DQS score are shown in Table 2. The DQS was significantly, positively associated with female sex, higher social position, no smoking, age and physical activity level at leisure time. A higher DQS was associated significantly with a lower intake of total and saturated fat and a higher intake of carbohydrate, fibre, various vitamins and minerals, fish, fruit, vegetables and whole–grain products. Spearman’s correlation coefficients between the DQS and the 198-item FFQ varied much between dietary components and were highest for fruits (r ¼ 0.55) and lowest for vitamin A (r ¼ 0.05) (Table 3). Gamma coefficients ranged from 0.46 (fruits) to 0.04 (vitamin A). The DQS showed a significantly positive association with HDL-cholesterol and a significant negative association with


273 Table 1 The construction of the DQS Food

Frequency

Score

Vegetables (cooked or raw) and/or vegetarian dishes

X5–7 servings/week The answers in between p2 servings/week

3 points 2 points 1 point

Fruit

X3 pieces/day X3 pieces/week and p 2 pieces/day p2 pieces/week


Fish

X 200 g/week The answers in between No intake


Fat Fat, bread

3 points – none 2 points – minarine, vegetable margarine 1point – butter, blended spread, lard 3 points – none/olive oil 2 points – vegetable margarine, oil 1 point – margarine/butter/blended spread/lard 6 points, summarized 3–5 points, summarized 2 points, summarized


Fat, cooking

Fat, summarized

Abbreviation: DQS, DietaryQuality Score.

Table 2 Characteristics of individuals according to the DQS category, n ¼ 6112 Healthy dietary habits (7–9 points) n (%)

Average dietary habits (4–6 points) n (%)

Unhealthy dietary habits (7–9 points) n (%)

Sex Men Women

296 (34) 569 (66)

2058 (48) 2232 (52)

608 (64) 349 (36)

Age (years) 30–35 40–50 55–60

88 (10) 518 (60) 259 (30)

700 (16) 2649 (62) 941 (22)

172 (18) 588 (61) 197 (21)

Education No vocational training p1 year of vocational training 2–4 years of vocational training 44 years of vocational training

98 42 562 163

713 211 2836 530

244 33 605 75

Daily smoker Yes No

189 (22) 672 (78)

1481 (35) 2796 (65)

501 (52) 455 (48)

Physical activity level, leisure time Low Moderate High

123 (14) 554 (65) 180 (21)

837 (20) 2681 (63) 710 (17)

322 (34) 498 (53) 128 (13)

P-values a

Pp0.0001

Pp0.0001

Pp0.0001 (11) (4.9) (65.0) (18.8)

(17) (5) (66) (12)

(26) (3) (63) (8)

Pp0.0001

Pp0.0001

Abbreviation: DQS, Dietary Quality Score. a The association with the DQS, estimated in a multiple ordinal logistic regression analysis with the dietary score as response and including sex, age, education, smoking status and physical activity level in the model. Owing to missing values for education, smoking status and physical activity level, the number of participants in this analysis is reduced to 6112.

total cholesterol, triglyceride, LDL-cholesterol, homocysteine, waist circumference and the absolute risk of IHD (The Copenhagen Risk Score) when adjustments for sex and age were made (Table 4). Taking physical activity level,

smoking habits and education into account, the described associations remained, except for waist circumference and HDL-cholesterol. The association between the DQS and waist circumference was confounded by physical activity level, European Journal of Clinical Nutrition


274 Table 3 Differences in daily food and nutrient intakes by DQS category Unhealthy dietary habits (1–3 points) Median (range) n Total energy (kJ) Carbohydrate (E%) Fibre (g) Fat (E%) Saturated fat (E%) Vitamin A (RE) Vitamin B6 (mg) Vitamin B12 (mg) Vitamin E (mg) Vitamin C (mg) Vitamin D (mg) Vitamin K (mg) Folate (g) Calcium (mg) Magnesium (mg) Iron (mg) Fruits (g) Vegetables (g) Fish (g) Whole-grain products (g)

8912 43.1 18.4 36.6 14.9 936 1.2 4.7 6.5 43.4 2.4 78.0 289 679 318 9.2 26.8 59.6 11.9 123.5

987 (1470–34 012) (23–69) (2–80) (15–56) (4–28) (87–7707) (0.2–3.4) (0.3–27) (1–132) (7–235) (0.3–21) (5–594) (49–1158) (121–2669) (89–838) (1–28) (0–525) (1–548) (0–218) (0–606)

Average dietary habits (4–6 points) Median (range)

9317 47.4 23.4 32.5 12.4 898 1.4 5.3 7.5 66.0 3.1 104.8 350 789 348 10.3 121.0 104.3 23.9 140.8

4447 (1355–39 155) (19–84) (2–104) (8–59) (2–26) (94–14 973) (0.1–4.9) (0.3–70) (1–74) (4–290) (0.2–4.2) (7–771) (71–2405) (88–3873) (65–1189) (1–39) (0–1106) (3–937) (0–398) (0–919)

Healthy dietary habits (7–9 points) Median (range)

9716 50.9 29.0 28.7 10.1 839 1.6 5.6 8.5 108.3 3.9 147.4 440 853 381 11.5 326.4 166.4 34.3 153.4

902 (1323–25 281) (29–80) (3–86) (10–48) (3–26) (105–12 536) (0.4–3.3) (1–57) (1–49) (15–321) (0.3–19) (23–838) (132–2274) (177–2563) (58–1016) (2–32) (0–1050) (7–957) (0–197) (0–676)

P-values for trend a

Spearman’s correlation coefficient a,b

Gamma coefficienta,b

o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 NS o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001 o0.0001

0.08 0.32 0.34 0.35 0.41 0.05 0.27 0.09 0.16 0.48 0.23 0.36 0.34 0.13 0.16 0.19 0.55 0.48 0.33 0.12

0.07 0.26 0.28 0.29 0.34 0.04 0.22 0.07 0.13 0.41 0.19 0.30 0.28 0.10 0.13 0.16 0.46 0.40 0.27 0.10

Abbreviation: DQS, Dietary Quality Score. a P-values for trend, gamma coefficient and Spearman’s correlation coefficient analyses are made using the 9-classed score. b All Spearman’s correlation coefficients and gamma coefficients were significant, Po0.001.

whereas both physical activity and smoking habits were confounding factors in the association with HDLcholesterol. The DQS showed no association with BMI. Neither did BMI seem to be a confounding factor in the associations with cardiovascular risk factors, as none of the estimates and Pvalues changed markedly by adjusting for BMI in the analyses (data not shown).

Discussion The DQS was developed to create a simple and quick instrument to assess the overall quality of the dietary habits. A higher DQS was shown to be associated with higher dietary quality in general, including a low intake of fat, especially saturated fat; a high intake of fibre; various vitamins and minerals; and fruit, fish, vegetables and whole–grain products. In accordance with the results of other studies, a higher DQS was significantly negatively associated with the serum level of total cholesterol (Keys and Pralin, 1966; Verschuren et al., 1995; Brunner et al., 1997), triglyceride (Singh et al., 1992; Volozh et al., 2002), LDL-cholesterol (Keys and Pralin, 1966; Brown, 1983), homocysteine (Fung et al., 2001a, b; Villegas et al., 2004) and the absolute risk of IHD (Burr et al., 1989; Law and Morris, 1998; De Lorgeril et al., 1999; Hu et al., 2000; Fung et al., 2001a, b). European Journal of Clinical Nutrition

A few other scores and indexes have been developed; similar to the DQS as these were constructed to capture the overall dietary quality based on dietary recommendations, requiring only a short FFQ. The Food Pyramid Index (Massari et al., 2004) and the Global Dietary Index of Food Quality (Huot et al., 2004a, b) focused on food items rich in fat, whereas the Healthy Food Index (HFI) included questions on butter/margarine, vegetables, fruit and coarse bread (Osler et al., 2001), and the Healthy Diet Index (HDI) included questions on vegetables, fruits, meat and fish (Dynesen et al., 2003). The Food Pyramid Index (FPI), HFI and HDI were found to be associated with risk factors, mortality and socio– demographic factors, respectively, but none of the scores has, to our knowledge, been validated against other dietary measures. The Global Dietary Index (GDI) was validated against seven 24-h diet recalls, but the score was only validated against energy and fat intake, which does not necessarily reflect the overall dietary quality (Huot et al., 2004a, b). The Recommended Food Score (RFS) (Kant et al., 2000) is probably the index, in the category of simple indexes, that has been validated and used the most. The weight in this index is on the intake of fruit and vegetables (65% of the 23 questions) and the score further includes whole grains, lean meats or meat alternatives, and low-fat dairy products. The RFS has been shown to be associated with mortality, various risk factors and biomarkers of dietary intake and has been validated against a 24-h diet recall.


275 Table 4 Associations between the DQS and cardiovascular risk factors Categories of the DQS Unhealthy (1–3 points)

Average (4–6 points)

Healthy (7–9 points)

P for trend

Cholesterol (mmol/l) Mean (s.d.) Simple modela, b (CI) Multivariate modelb, b (95% CI)

5.21 (1.01) 0.21 (0.12–0.30) 0.14 (0.05–0.23)

5.13 (0.97) 0.12 (0.05–0.19) 0.08 (0.01–0.16)

5.07 (0.96) 0 0

o0.0001 0.0031

LDL-cholesterol (mmol/l) Mean (s.d.) Simple modela, b (95% CI) Multivariate modelb, b (95% CI)

3.60 (0.98) 0.21 (0.12–0.30) 0.13 (0.03–0.22)

3.50 (0.96) 012 (0.05–0.18) 0.07 (0.00–0.14)

3.41 (0.91) 0 0

o0.0001 0.0071

HDL-cholesterol (mmol/l) Mean (s.d.) Simple modela, b (95% CI) Multivariate modelb, b (95% CI)

1.37 (0.40) 0.14 (0.09–0.19) 0.05 (0.01 to 0.10)

1.51 (0.41) 0 0

0.0203 0.4429

Triglycerided (log mmol/l) Mean (s.d.) Simple modela, b (95% CI) Multivariate modelb, b (95% CI)

1.44 (0.98) 0.14 (0.09–0.19)d 0.05 (0.01 to 0.10)d

1.26 (0.84) 0.05 (0.01–0.09)d 0.01 (0.03 to 0.05)d

1.15 (0.67) 0 0

o0.0001 0.0406

Homeocysteind (log mmol/l) Mean (s.d.) Simple modela, b (95% CI) Multivariate modelb, b (95% CI)

3.59 (0.98) 0.16 (0.12–0.21)d 0.13 (0.08–0.18)d

3.48 (0.96) 0.06 (0.02–0.10)d 0.04 (0.01–0.08)d

3.38 (0.91) 0 0

o0.0001 o0.0001

1.44 (0.39) 0.02 (0.05 to 0.01) 0.01 (0.03 to 0.05)

Systolic blood pressure (mm Hg) Mean (s.d.) Simple modela, b (95% CI) Multivariate modelb, b (95% CI)

129.3 (16.5) 0.31 (1.15 to 1.78) 1.20 (0.42 to 2.00)

128.7 (16.7) 0.48 (0.05–0.18) 0.79 (0.42 to 0.56)

128.3 (16.3) 0 0

0.7111 0.1366

Waist circumference (cm) Mean (s.d.) Simple modela, b (95% CI) Multivariate modelb, b (95% CI)

88.4 (13.4) 1.62 (0.55–2.70) 0.65 (0.48 to 1.79)

85.9 (12.4) 0.84 (0.01 to 0.70) 0.55 (0.33 to 1.42)

83.5 (12.4) 0 0

0.0031 0.2716

Body mass index (kg/m2) Mean (s.d.) Simple modela, b (95% CI) Multivariateb, b (95% CI)

26.2 (4.6) 0.20 (0.23 to 0.63) 0.10 (0.34 to 0.55)

26.1 (4.5) 0.19 (0.15 to 0.53) 0.21 (0.14 to 0.56)

25.9 (4.4) 0 0

0.3798 0.6946

The Copenhagen Risk Scorec,d Mean (s.d.) Simple modela, b (95% CI) Multivariate modelb, b (95% CI)

6.97 (4.98) 0.22 (0.18–0.26)d 0.17 (0.12–0.21)d

5.67 (4.63) 0.11 (0.07–0.14)d 0.09 (0.05–0.12)d

4.50 (3.55) 0 0

o0.0001 o0.0001

Abbreviations: DQS, Dietary Quality Score; HDL, high-density lipoprotein; LDL, low-density lipoprotein; s.d., standard deviation. a Simple linear regression models adjusted for sex and age. b Multivariate linear regression models adjusted for sex, age, physical activity level at leisure time, smoking status and education. c The model including The Copenhagen Risk Score was not adjusted for smoking, as smoking is included in the score. d Data are natural logarithmically transformed.

The correlation coefficients found in this study were at similar levels as those found for the RFS (Kant and Graubard, 2005) and also for other, more complex indexes (Kennedy et al., 1995; Tur et al., 2005). Recently, Kant and Graubard (2005) compared the association of the RFS, the Healthy Eating Index (HEI) and the Dietary Diversity Score (DDS) with cardiovascular risk factors. In accordance with the findings of this study, all three scores were associated with homocysteine whereas, in contrast with the

DQS, the three scores did not associate with triglyceride. The association with total cholesterol was also found for the RFS and the DDS, but not for the HEI, whereas only the HEI was associated with LDL-cholesterol. The RFS and the DDS were associated with systolic blood pressure, in contrast with the DQS and the HEI. Contradictory to the DQS, all three scores were associated with BMI. Similarly, the FPI (Massari et al., 2004) was found to be associated with BMI in men but not in women. The reason for European Journal of Clinical Nutrition


276 the missing association between the DQS and BMI is unclear. The missing association with HDL-cholesterol, when smoking and physical activity level were taken into account, was in contrast with our expectations, especially because the score includes fish intake. However, this finding is in accordance with the results of the RFS and the DDS (Kant and Graubard, 2005) and the Lyon Diet Heart Study (De Lorgeril et al., 1999). An explanation for this could be that some studies have shown that when substituting fat with carbohydrate both LDL- and HDL-cholesterol decrease (Martin et al., 1986). Hence, HDL-cholesterol concentration might be more closely related to, for example, physical activity level and smoking than a low–fat diet. When constructing the score, the creation of two smaller groups with healthy and unhealthy dietary habits, respectively, and a larger group of individuals with an average dietary quality for each food group was emphasized. This strategy was chosen to optimize the information on variability compared to if the score had been scaled as dichotomous measures according to the median for the dietary intake or in tertiles of equal size. The Danish Dietary Guidelines emphasize, in addition to the dietary recommendations included in the DQS, a high intake of fibre (43 g/MJ). In addition to the intake of fruit and vegetables, whole–grain products, including bread, are an important source of fibres. The questionnaire did not include sufficient information on bread intake, and therefore this food group was not included in the score. In spite of this, the results from the validation study showed a reasonable association between the score and the intake of fibre. The similarities in the designs of the long and the short FFQ may cause an overestimation of the validity of the score. Optimally, the score should be validated against the results of multiple-day food records. However, Coates et al., (1995) and Kristal et al., (1990) investigated this issue and did not find higher correlation coefficients between a long FFQ and a short FFQ than between a short FFQ and a food record. The fact that the two FFQs measure the dietary intake of different periods (the last month versus the last week) might on the other hand cause an underestimation of the validity of the score. Despite the extensive use of correlation analyses to validate dietary assessment methods, correlation coefficients provide only a limited measure of the level of agreement between two measurements (Bland and Altman, 1986) and correlation coefficients should therefore not be used alone. The correlation coefficients are actually not very useful when comparing results from different populations and the results are in general difficult to interpret. Correlation coefficients depend, for example, on the range of the true quantity in the sample (Bland and Altman, 1986). Thus, in this study, it would mean that the correlation coefficients would be greater if the score was categorized into more categories. Furthermore, correlation analyses ignore any systematic bias between the two variables. A better measurement is the European Journal of Clinical Nutrition

gamma coefficient. Gamma coefficients are appropriate for measuring the relation between two ordinally scaled variables and the interpretation of these is easier. Gamma coefficients are based on the number of concordant and discordant pairs of observations and measure the degree to which individuals are ranked equal by the two methods (Goodman and Kruskal, 1979). Although the DQS seems to be able to classify individuals reasonably into rough categories of high, average and low dietary quality, misclassification must be considered. Nevertheless, if overall dietary habits are of interest and nutrient intake is not necessary to estimate, or if dietary habits might be a confounding factor of the association explored, the DQS is very useful in large studies. In fact, to construct the DQS, it is only necessary to include eight questions about dietary intake to get a rough classification of the individuals in a population, according to their dietary habits. Because the score is such a simple, quick and inexpensive dietary assessment method, it is possible to include it in most studies. Furthermore, the DQS might be a good tool to assess dietary changes in a population. Future research will explore if the score is sensitive enough for this purpose. A limitation of this study is the cross-sectional design. A relevant future study would therefore be to explore the associations between changes in the DQS and changes in the described risk factors. Furthermore, it is relevant to investigate if the DQS can predict disease.

Conclusion The DQS is a simple, valid and quick tool to make a rough classification of individuals into groups with overall healthy, average or unhealthy dietary habits. The score is significantly negatively associated with total and LDL-cholesterol, triglyceride, homocysteine and the absolute risk of IHD.

References Appel LJ, Moore TJ, Obarzanek E, Vollmer WM, Svetkey LP, Sacks FM et al. (1997). A clinical trial of the effects of dietary patterns on blood pressure. DASH Collaborative Research Group. N Engl J Med 336, 1117–1124. ˚ B, Ohlander EM, Lyhne N et al. (2004). 4th Edition Becker W, Konde A of the Nordic Nutrition Recommendations – Preface and Chapter 1 and 2. Nordic Counsil of Ministers: Copenhagen, Denmark. 1–19. Biltoft-Jensen A. The National Dietary Survey of Denmark 2000–2002. Technical Report (unpublished work). Bland JM, Altman DG (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, 307–310. Block G (1982). A review of validations of dietary assessment methods. Am J Epidemiol 115, 492–505. Brown HB (1983). Diet and serum lipids: controlled studies in the United States. Prev Med 12, 103–109. Brunner E, White I, Thorogood M, Bristow A, Curle D, Marmot M (1997). Can dietary interventions change diet and cardiovascular risk factors? A meta-analysis of randomized controlled trials. Am J Public Health 87, 1415–1422.


277 Burr ML, Fehily AM, Gilbert JF, Rogers S, Holliday RM, Sweetnam PM et al. (1989). Effects of changes in fat, fish, and fibre intakes on death and myocardial reinfarction: diet and reinfarction trial (DART). Lancet 2, 757–761. Coates RJ, Serdula MK, Byers T, Mokdad A, Jewell S, Leonard SB et al. (1995). A brief, telephone-administered food frequency questionnaire can be useful for surveillance of dietary fat intakes. J Nutr 125, 1473–1483. Conlin PR, Chow D, Miller III ER, , Svetkey LP, Lin PH, Harsha DW et al. (2000). The effect of dietary patterns on blood pressure control in hypertensive patients: results from the Dietary Approaches to Stop Hypertension (DASH) trial. Am J Hypertens 13, 949–955. De Lorgeril M, Salen P, Martin JL, Monjaud I, Delaye J, Mamelle N (1999). Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction: final report of the Lyon Diet Heart Study. Circulation 99, 779–785. 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. Friedewald WT, Levy RI, Fredrickson DS (1972). Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 18, 499–502. Fung TT, Rimm EB, Spiegelman D, Rifai N, Tofler GH, Willett WC et al. (2001a). Association between dietary patterns and plasma biomarkers of obesity and cardiovascular disease risk. Am J Clin Nutr 73, 61–67. Fung TT, Willett WC, Stampfer MJ, Manson JE, Hu FB (2001b). Dietary patterns and the risk of coronary heart disease in women. Arch Intern Med 161, 1857–1862. Goodman LA, Kruskal WH (1979). Measures of Association for Cross Classification. Springer-Verlag: New York, NY. Haines PS, Siega-Riz AM, Popkin BM (1999). The Diet Quality Index revised: a measurement instrument for populations. J Am Diet Assoc 99, 697–704. Haraldsdottir J, Seppa¨nen R, Steinrı´msdo´ttir L, Trygg K, Hagman U (1998). Portion Sizes: Nordic Standard Portions of Food and Foodstuffs. Nordic Council and Council of Ministers: Copenhagen, Denmark. 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. Huijbregts P, Feskens E, Rasanen L (1995). Dietary pattern and 20 year mortality in elderly men in Finland, Italy, and The Netherlands: longitudinal cohort study. BMJ 311, 1457–1460. Huot I, Paradis G, Ledoux M (2004a). Effects of the Quebec Heart Health Demonstration Project on adult dietary behaviours. Prev Med 38, 137–148. Huot I, Paradis G, Receveur O, Ledoux M (2004b). Correlates of diet quality in the Quebec population. Public Health Nutr 7, 1009–1016. Jorgensen T, Borch-Johnsen K, Thomsen TF, Ibsen H, Glumer C, Pisinger C (2003). A randomized non-pharmacological intervention study for prevention of ischaemic heart disease: baseline results Inter99 (1). J Cardiovasc Risk 10, 377–386. Kant AK (1996). Indexes of overall diet quality: a review. J Am Diet Assoc 96, 785–791. Kant AK (2004). Dietary patterns and health outcomes. J Am Diet Assoc 104, 615–635. Kant AK, Graubard BI (2005). A comparison of three dietary pattern indexes for predicting biomarkers of diet and disease. J Am Coll Nutr 24, 294–303. Kant AK, Schatzkin A, Graubard BI, Schairer C (2000). A prospective study of diet quality and mortality in women. JAMA 283, 2109– 2115. Kant AK, Thompson FE (1997). Measures of overall diet quality from a food frequency questionnaire: National Health Interview Survey, 1992. Nutr Res 17, 1443–1456. Kennedy ET, Ohls J, Carlson S, Fleming K (1995). The Healthy Eating Index: design and applications. J Am Diet Assoc 95, 1103–1108.

Keys A, Pralin RW (1966). Serum cholesterol response to changes in dietary lipids. Am J Clin Nutr 19, 175–181. Kristal AR, Schattuck AL, Henry HL, Fowler AS (1990). Rapid assessment of dietary intake of fat, fiber and saturated fat: validity of an instrument suitable for community intervention research and nutritional surveillance. Am J Health Promot 4, 288–295. ¨ mer C, Toft U, Tetens I, Borch-Johnsen K, Lau C, Færch K, Glu Jorgensen T (2004). Evaluation of dietary intake in a Danish population: the Inter99 study. Scand J Nutr 48, 136–143. Lauritsen J (1998). FoodCalc 1.3 1998. Available at:www:http:/ foodcalc.dk(accessed October 2002). Law MR, Morris JK (1998). By how much does fruit and vegetable consumption reduce the risk of ischaemic heart disease? Eur J Clin Nutr 52, 549–556. Martin MJ, Hulley SB, Browner WS, Kuller LH, Wentworth D (1986). Serum cholesterol, blood pressure, and mortality: implications from a cohort of 361,662 men. Lancet 2, 933–936. Massari M, Freeman KM, Seccareccia F, Menotti A, Farchi G (2004). An index to measure the association between dietary patterns and coronary heart disease risk factors: findings from two Italian studies. Prev Med 39, 841–847. Olsen J, Melbye M, Olsen SF, Sorensen TI, Aaby P, Andersen AM et al. (2001). The Danish National Birth Cohort – its background, structure and aim. Scand J Public Health 29, 300–307. Osler M, Heitmann BL, Gerdes LU, Jorgensen LM, Schroll M (2001). Dietary patterns and mortality in Danish men and women: a prospective observational study. Br J Nutr 85, 219–225. Patterson RE, Haines PS, Popkin BM (1994). Diet quality index: capturing a multidimensional behavior. J Am Diet Assoc 94, 57–64. Sacks FM, Svetkey LP, Vollmer WM, Appel LJ, Bray GA, Harsha D et al. (2001). Effects on blood pressure of reduced dietary sodium and the Dietary Approaches to Stop Hypertension (DASH) diet. DASH – Sodium Collaborative Research Group. N Engl J Med 344, 3–10. Saltin B, Grimby G (1968). Physiological analysis of middle-aged and old former athletes: comparison with still active athletes of the same ages. Circulation 38, 1104–1115. Saxholt E, Møller A, Ygil K (2005). . Danish Food Composition Databank Version 6.0 [Fødevaredatabanken Version 6.0]. Department of Nutrition, The Danish Institute for Food and Veterinary Research: Søborg, Denmark http://www.foodcomp.dk. Singh RB, Rastogi SS, Verma R, Laxmi B, Singh R, Ghosh S et al. (1992). Randomised controlled trial of cardioprotective diet in patients with recent acute myocardial infarction: results of one year follow up. BMJ 304, 1015–1019. The National Consumer Agency (1995). The Seven Food Guidelines. The National Consumer Agency: Copenhagen, Denmark. Thomsen T, Borch-Johnsen K, Davidsen M, Ibsen H (1997). The ‘PRECARD’ study: identification and management of individuals at risk of developing cardiovascular disease. Can J Cardiol 13, 286B–287B. Tjonneland A, Overvad K, Haraldsdottir J, Bang S, Ewertz M, Jensen OM (1991). Validation of a semiquantitative food frequency questionnaire developed in Denmark. Int J Epidemiol 20, 906–912. Trichopoulos A, Costacou T, Bamia C, Trichopoulos D (2003). Adherence to a Mediterranean diet and survival in a Greek population. N Engl J Med 348, 2599–2608. Trichopoulou A, Kouris-Blazos A, Wahlqvist ML, Gnardellis C, Lagiou P, Polychronopoulos E et al. (1995). Diet and overall survival in elderly people. BMJ 311, 1457–1460. Trolle E, Fagt S, Ovesen L (1998). Fruit and vegetables – Guidelines for intake [Frugt og grønt – Anbefalinger for indtagelse]. 244. The Danish Institute for Food and Veterinary Research: Copenhagen, Denmark. Tucker KL, Hallfrisch J, Qiao N, Muller D, Andres R, Fleg JL (2005). The combination of high fruit and vegetable and low saturated fat intakes is more protective against mortality in aging men than is either alone: the Baltimore Longitudinal Study of Aging. J Nutr 135, 556–561.

European Journal of Clinical Nutrition


278 Tur JA, Romaguera D, Pons A (2005). The Diet Quality IndexInternational (DQI-I): is it a useful tool to evaluate the quality of the Mediterranean diet? Br J Nutr 93, 369–376. Verschuren WM, Jacobs DR, Bloemberg BP, Kromhout D, Menotti A, Aravanis C et al. (1995). Serum total cholesterol and long-term coronary heart disease mortality in different cultures. Twenty-fiveyear follow-up of the seven countries study. JAMA 274, 131–136.

European Journal of Clinical Nutrition

Villegas R, Salim A, Collins MM, Flynn A, Perry IJ (2004). Dietary patterns in middle-aged Irish men and women defined by cluster analysis. Public Health Nutr 7, 1017–1024. Volozh O, Solodkaya E, Abina J, Kaup R, Goldsteine G, Olferiev A et al. (2002). Some biological cardiovascular risk factors and diet in samples of the male population of Tallinn, Estonia in 1984/1985 and 1992/1993. Eur J Public Health 12, 16–21.