It is generally accepted that inadequate or excessive nutrient intake can have important health consequences, such as nutritional deficiencies, increased risk of type 2 diabetes, CVD and obesity. Food-based dietary guidelines (FBDG) have been developed to prevent such dietary-related health problems(1, Reference Sandstrom2). These guidelines are targeted at the general population and contain messages that give an indication of what a person should be eating in terms of foods rather than nutrients. These FBDG can be broad and non-specific, such as ‘eat a variety of foods each day’ or more targeted such as ‘eat five portions of fruits and vegetables a day’. Messages may also specify the type of food, such as ‘at least half of the grains consumed should be whole grains’, or be meal specific such as ‘eat a breakfast every day’(Reference Sandstrom2, 3).
Over the last decades, a number of diet quality indices, measuring adherence to such dietary guidelines, have been developed(Reference Waijers, Feskens and Ocke4–Reference Kant7). The advantage of such indices is that they capture the complexity of human diets in a single value, taking into account the interactions between nutrients, food preparation methods and eating patterns(Reference Waijers, Feskens and Ocke4, Reference Kant8, Reference Hu9). In general, indices representing overall diet quality showed stronger correlations with health outcomes than individual nutrients or foods(Reference Kant7).
The majority of existing indices are, however, unsuitable for children and adolescents, because their development is based on dietary recommendations for adults(Reference Manios, Kourlaba and Grammatikaki10). Nevertheless, appropriate indices for children and adolescents have been developed, based on recommendations specific for these age groups, e.g. the Youth Healthy Eating Index(Reference Feskanich, Rockett and Colditz11), the Revised Children's Diet Quality Index (DQI)(Reference Kranz, Findeis and Shrestha12), the DQI for Preschoolers(Reference Huybrechts, Vereecken and De Bacquer13), the Preschoolers Diet-Lifestyle Index(Reference Manios, Kourlaba and Grammatikaki10) and the Healthy Lifestyle-Diet Index(Reference Manios, Kourlaba and Grammatikaki14). However, most of these indices are calculated based on a combination of food and nutrient (e.g. cholesterol, Na) intake. This implies the need for detailed dietary information and use of food composition tables.
The purpose of the present study was to develop a DQI for Adolescents (DQI-A) calculated on food intake data only for assessing adherence to FBDG. Furthermore, the aim of the present study was to investigate whether adherence to FBDG (using the DQI-A as a proxy measure) is associated with better adherence to nutrient dietary recommendations and a better nutritional biomarker blood profile in European adolescents.
Subjects and methods
Study design
The ‘Healthy Lifestyle in Europe by Nutrition in Adolescence (HELENA) – Cross Sectional Study’ is a population-based multi-centre investigation of the nutritional and lifestyle status of adolescents, carried out in ten European cities (Vienna in Austria, Ghent in Belgium, Lille in France, Dortmund in Germany, Athens and Heraklion in Greece, Pécs in Hungary, Rome in Italy, Zaragoza in Spain and Stockholm in Sweden). Data were collected from October 2006 to December 2007. The purpose of the study was to obtain standardised, reliable and comparable data from a random sample of European adolescents on a broad battery of relevant nutrition and health-related parameters like dietary intake, food choices and preferences, serum indicators of lipid metabolism and glucose metabolism, vitamin and mineral status, anthropometry, physical activity, fitness and genetic markers. A detailed description of the HELENA study design and sampling procedure has been published elsewhere(Reference De Henauw, Gottrand and De Bourdeaudhuij15–Reference Moreno, De Henauw and González-Gross17).
The study population comprised adolescents, aged 12·5–17·5 years. Adolescents were excluded from the database a posteriori if they met one of the exclusion criteria, namely age < 12·5 or >17·5 years, no measurement of weight and/or height, completion of < 75 % of the tests, participating simultaneously in another clinical trial or an acute infection during the week prior to the examination(Reference Moreno, González-Gross and Kersting16). The total HELENA population consisted of 3528 eligible adolescents (52·3 % females). For the purpose of the present study, adolescents who provided data on two non-consecutive 24 h dietary recalls were included in the analysis, resulting in 2330 subjects. Participants from Heraklion and Pécs were excluded for these analyses, as no nutrient intake information was calculated for these two cities due to logistical problems. Underreporters were excluded for the analyses, as previous reports, using concentration biomarkers and the Triads method(Reference Kaaks18), have shown that the validity of food and nutrient intakes compared to ‘true’ intake was better when excluding underreporters(Reference Vandevijvere, Geelen and Gonzalez-Gross19). Exclusion of underreporters resulted in a final sample of 1804 adolescents (52·6 % females) for statistical analysis. Underreporting was considered when the individual ratio of energy intake divided by the estimated BMR was lower than 0·96(Reference Black20). The group of underreporters consisted of a slightly higher percentage of females (57·8 % compared to 52·6 % in the plausible reports, P =0·036) and had higher median (minimum, maximum) BMI values (22·5 (14·99, 45·63) kg/m2 compared to 20·1 (14·08, 40·77) kg/m2 in the plausible reports; P <0·001). No differences in age and DQI-A score were observed.
Blood samples were collected in a randomly selected subset of the total HELENA study population (1089 adolescents), of whom 697 provided two 24 h recalls. Exclusion of underreporters resulted in a final sample of 552 adolescents (52·3 % females) with biomarkers in the present study. Characteristics of adolescents for whom no biochemical parameters were obtained were compared to those included in the sub-analysis. No significant differences in age, sex and BMI were observed between these two groups; however, the adolescents in the sub-analysis had higher mean values for the DQI-A compared to the total study population (52·6 (sd 15·6) and 51·3 (sd 16·5) %, respectively, P =0·018).
The study was performed following the ethical guidelines of the Declaration of Helsinki, the Good Clinical Practice rules and the legislation regarding clinical research in human subjects in each of the participating countries. All study participants and their parents provided a signed informed consent form. The protocol was approved by the Human Research Review Committees of the institutions involved(Reference Beghin, Castera and Manios21).
Dietary intake assessment
Dietary intake was assessed by two non-consecutive 24 h recalls(Reference Biro, Hulshof and Ovesen22), comprising weekdays and weekend-days (except from Fridays and Saturdays), though not necessarily including a week and weekend-day for each individual. The 24 h recalls were collected by use of a computer-based self-administered tool, the HELENA-Dietary Intake Assessment Tool (DIAT). This tool was adapted from a previous version developed and validated for Flemish adolescents(Reference Vereecken, Covents and Matthys23). This assessment tool is based on six meal occasions (breakfast, morning snacks, lunch, afternoon snacks, evening meal, evening snacks) referring to the previous day. Trained dietitians assisted the adolescents to complete the 24 h recalls when needed. Adolescents selected autonomously all the consumed foods and beverages from a standardised food list in the HELENA-DIAT(Reference Vereecken, Covents and Sichert-Hellert24). Items not available in the list could be added by the participant at any moment. Consumed foods were translated to nutrients by use of the German Food Code and Nutrient Data Base (Bundeslebensmittelschlüssel, BLS, version II.3.1)(Reference Dehne, Klemm and Henseler25). The Multiple Source Method was used to estimate the usual dietary intake of nutrients and foods(Reference Haubrock, Hartigg and Souverein26, 27). This statistical modelling technique takes into account within-person variability and calculates usual intakes corrected for age, sex and study centre.
Diet Quality Index for Adolescents
A previously validated DQI, originally developed for pre-school-aged children(Reference Huybrechts, Vereecken and De Bacquer13), was adapted for use in adolescents to measure their compliance to the Flemish FBDG(28). These FBDG put forward three basic principles for a healthy and balanced diet, namely dietary quality, dietary diversity and dietary equilibrium. Furthermore, the daily diet was divided into nine recommended food groups, namely (1) water, (2) bread and cereals, (3) grains and potatoes, (4) vegetables, (5) fruit, (6) milk products (7), cheese, (8) meat, fish, eggs and substitutes and (9) fat and oils. Milk products and cheese were allocated to different food groups because of the important difference in fat content. Meat and fish are considered in the same food group; because of the differences in nutrient content, the FBDG additionally recommended the consumption of fish preferably two times per week. However, as only two 24 h recalls were assessed, the frequency of fish consumption could not be considered separately in the DQI-A calculation. For each of the food groups, a range of recommended daily intakes, specifically for adolescents, was provided by the FBDG. The ranges in these FBDG were based upon the nutrient recommendations of the Belgian Health Council(29) and the WHO, combined with data on habitual dietary intake in the Belgian population. These FBDG were very similar to dietary guidelines in other countries and to the CINDI pyramid (Countrywide Integrated Non-communicable Disease Intervention program) developed by the WHO(30), making the index applicable for a European population.
The technical aspects of the calculation of the DQI-A are given in Table 1. Parallel to the FBDG, the DQI-A consisted of three components, namely quality, diversity and equilibrium.
Table 1 Overview of the calculation of the Diet Quality Index for Adolescents (DQI-A)*
FBDG, food-dased dietary guidelines; FG, food groups; DQ, dietary quality; DD, dietary diversity; DA, dietary adequacy; DEx, dietary excess; DE, dietary equilibrium
* Further details on ‘preference group’, ‘intermediate group’ and ‘low-nutrient, energy-dense group’ can be found in Table 2.
Dietary quality expressed whether the adolescent made the optimal food quality choices within a food group and was represented by a ‘preference group’ (e.g. cereal/brown bread, fresh fruit, fish), an ‘intermediate group’ (e.g. white bread, minced meat) and a ‘low-nutrient, energy-dense group’ (e.g. soft drinks, sweet snacks, chicken nuggets). A comprehensive description of the allocation of food items to the different quality groups is given in Table 2.
Table 2 Classification of food items to the different quality groups within each food group, as advised by the Flemish Food-Based Dietary Guidelines
Dietary diversity expressed the degree of variation in the diet. This diversity component was obtained by giving points ranging from 0 to 9 when at least one serving of food of a recommended food group was consumed.
Dietary equilibrium was calculated from the difference between the adequacy component (which was the percentage of the minimum recommended intake for each of the main food groups, truncated to 1) and the excess component (which was the percentage of intake exceeding the upper level of the recommendation, truncated to 1 if larger than 1 and truncated to 0 when below 0; see Table 1). In the food group of meat, fish, eggs and substitutes, the daily intake of the total group was considered. As such, a too high consumption of meat and fish was penalised in the excess component. However, the fish consumption was granted points in the quality component, as fish is always allocated to the preference group in contrast to semi-fat and fat meat products.
These three components of the DQI-A were presented in percentages. The dietary quality component ranged from − 100 to 100 %, while dietary diversity and dietary equilibrium ranged from 0 to 100 %. To compute the DQI-A, the mean of these components was calculated; as such, the DQI-A ranged from − 33 to 100 %, with higher scores reflecting a higher diet quality. The score was calculated for each day and a mean of the daily scores was taken as global index score of the individual.
Blood analyses
After an overnight fasting period, venous blood samples were drawn in the morning at school according to a standardised blood collection protocol. Details about the sampling, processing and transportation can be found elsewhere(Reference Gonzalez-Gross, Breidenassel and Gomez-Martinez31). The studied biomarkers were chosen in view of clinical relevance to evaluate nutritional status (vitamin D, vitamin B12, retinol and TAG) or as a dietary biomarker reflecting true intake (vitamin C, plasma folate, carotenoids, n-3 fatty acids (FA) and trans-FA). Although plasma vitamin D concentrations are influenced by several factors such as sunlight exposure and adiposity, evidence also showed weak correlations with dietary intakes(Reference Jacques, Sulsky and Sadowski32). Strong correlations of dietary intakes of vitamin C and serum ascorbic acid concentrations have been reported mainly when habitual dietary intakes of vitamin C are relatively modest(Reference Bates, Rutishauser and Black33, Reference Jenab, Slimani and Bictash34). Many factors influence serum folate concentrations and bioavailability of dietary folate; however, intakes correlate moderately with serum concentrations(Reference Jacques, Sulsky and Sadowski32). Weak but positive correlations were reported for males and females between dietary vitamin B12 intake and holo-transcobalamin status, being a marker of long-term vitamin B12 status(Reference Bor, von Castel-Roberts and Kauwell35, Reference Hvas, Gravholt and Nexo36). Weak correlations may be linked to the large size of liver vitamin B12 stores. Blood concentrations of carotenoids appear to be moderately correlated with fruit and vegetable intake(Reference Jenab, Slimani and Bictash34, Reference Rock, Swendseid and Jacob37, Reference Willett, Stampfer and Underwood38). Plasma retinol concentrations are only responsive to vitamin A intake in individuals with inadequate vitamin A status(Reference Willett, Stampfer and Underwood38). Plasma TAG have been shown to be positively correlated with total fat intake and negatively with fibre intake. Levels may, to some extent, be indicative of the level of dietary fibre intake, but the findings to date are conflicting(Reference Jenab, Slimani and Bictash34, Reference Sonnenberg, Quatromoni and Gagnon39). n-3 FA intake is moderately correlated with plasma phospholipid levels, reflecting intake in the short to medium term(Reference Hodge, Simpson and Gibson40, Reference Andersen, Solvoll and Drevon41). Correlations between the intake of specific types of trans-FA and their levels in blood are generally good; however, correlation between the total sum of trans-FA intake and the sum of serum trans-FA levels is only weak(Reference Hodson, Skeaff and Fielding42, Reference Baylin, Kim and Donovan-Palmer43).
Plasma folate was measured by means of an immunoassay using the Immunolite 2000 analyser (DPC Biermann GmbH). Holo-transcobalamin (the biologically active form of vitamin B12) was determined by an automated microparticle enzyme immunoassay with the AxSYM analyser (Abbott Laboratories). Vitamin C, β-carotene and retinol were analysed by HPLC (Sykam) using UV detection (UV-Vis 205, Merck). Serum phospholipid FA composition was determined by capillary GC (GC-2010, FID detection, Shimadzu GmbH) after extraction performed by TLC. Serum TAG were measured enzymatically on the Dimension RxL clinical chemistry system (Dade Behring) using the manufacturer's reagents and instructions. Plasma 25-hydroxyvitamin D (25(OH)D) was analysed by ELISA using a kit (OCTEIA 25-Hydroxy Vitamin D) from Immunodiagnostic System and measured with a SunriseTM Photometer by TECAN.
Statistical analysis
Statistical analyses were performed using the statistical software PASW for Windows version 18 (SPSS, Inc.).
Descriptive characteristics were summarised by calculating means and standard deviations for continuous variables and percentages for categorical variables. Pearson χ2 and t tests were used to test differences between sexes in categorical and continuous variables, respectively.
Normality was evaluated visually and based on the skewness of the data distributions. Skewness of variables on intake ranged from 0·869 (meat intake) to 9·809 (alcohol intake – due to a high number of non-consumers). As these variables were studied in a large sample size (n 1804), it was considered that parametrical tests were allowed without transformations. However, variables with a skewness >3 were transformed (log-transformation and square root transformations were tested) and the multilevel analyses were repeated with the transformed data. For all the variables, except for Na intake, the results before and after transformation were similar. To facilitate the interpretation of the results, it was chosen to display the non-transformed data, except for Na intake which was log-transformed. The skewness of the variables on blood values ranged from − 0·210 to 2·017 and the skewness of holo-transcobalamin was 3·110. As these variables were only studied in a sample of 552 adolescents, it was decided to perform a log-transformation of holo-transcobalamin in order to achieve a more normal distribution. Multilevel linear regression analysis with inclusion of a random intercept for study centre was used to examine the relationship between the DQI-A and foods, nutrient intakes or blood biomarkers. Confounders (age and sex) were entered as covariates. The random intercept for centre ranged from 0·19 to 24·42 %, with the highest influence of centre observed for oils, butter and animal fats, and milk and yoghurt. Significant differences in mean DQI-A scores were observed between both sexes; however, results of validation were very similar. As such, results were not stratified. To adjust for multiple testing, a Bonferroni correction was applied to lower the significance level (α) taking into account the number of tests (0·05/number of tests). Pvalues of 0·0019, 0·0013 and 0·006 were used as thresholds of significance for the associations between DQI-A and foods, nutrients and biomarkers, respectively.
Results
The total study population consisted of 1804 participants (52·6 % females) and the mean age was 14·7 (sd 1·2) years. The DQI-A score ranged from − 11·1 to 84·1 %; the mean DQI-A scores were 49·0 (sd 17·0) and 53·3 (sd 15·9) % for males and females, respectively (P <0·001). No differences were observed in mean DQI-A between adolescents in different BMI classes or between adolescents complying with the recommendation of 60 min physical activity v. non-compliers.
Multilevel regression analysis of the DQI-A scores with the usual consumption of different foods is shown in Table 3. A strong positive association between the DQI-A score and water intake (g/d) was observed (β = 19·529, P <0·0001). In contrast, soft drinks, fruit juices and alcoholic beverages showed a significant negative association with the DQI-A. Furthermore, the DQI-A score and bread/cereals had a positive association, but there was no significant association with potatoes and grains. Milk and cheese (g/d) were positively associated with the DQI-A score, and animal fat (g/d) and vegetable fat (g/d) showed a small, however, significant positive association with the DQI-A (β = 0·100, P <0·0001 and β = 0·118, P <0·0001, respectively). No significant relation was present with meat, fish, eggs and substitutes. All non-recommended (energy-dense and low-nutritious) foods showed a significant negative association with the DQI-A score.
Table 3 Association between Diet Quality Index for Adolescents (DQI-A) scores and food intake* (β-Coefficients and 95 % confidence intervals)
* Multilevel regression analyses with inclusion of a random intercept for centre and corrected for age and sex as independent variables. Bonferroni correction resulted in level of significance < 0·0019.
At the level of macronutrients (Table 4), a positive association was observed between the DQI-A and water (g/d) and fibre (g/d) intake, and a negative relationship was found with total energy intake (kJ/d) (β = − 2·893, P =0·0005). The usual intake of polysaccharides (g/d) was positively related to the dietary quality (β = 0·230, P =0·0004), whilst the intake of mono- and disaccharides (g/d) showed a negative relationship (β = − 0·853 and − 0·289, respectively; both P <0·0001). No significant association was seen between DQI-A and protein intake (g/d) or fat intake (g/d). All investigated minerals (Table 4), except Fe and Cu, were positively associated with the DQI-A score. Furthermore, intake of almost all vitamins, except niacin, vitamin C and vitamin E, showed a significant positive association with the calculated index.
Table 4 Associations between Diet Quality Index for Adolescents (DQI-A) score and usual intake of macro- and micronutrients* (β-Coefficients and 95 % confidence intervals)
* Multilevel regression analysis with inclusion of a random intercept for centre and corrected for age and sex as independent variables. Bonferroni correction resulted in level of significance < 0·0013.
† Variable was log-transformed to obtain a normal distribution.
Table 5 describes the results of the multilevel regression analysis between the DQI-A scores and nutritional biomarkers in a subgroup of 552 adolescents. Only for plasma 25(OH)D and holo-transcobalamin, a significant positive association was observed with the index score. The positive association with the n-3 FA status (μmol/l) was borderline significant (β = 0·376, P =0·007).
Table 5 Associations between the Diet Quality for Adolescents (DQI-A) scores and nutritional biomarkers* (β-Coefficients and 95 % confidence intervals)
* Multilevel regression analyses with inclusion of a random intercept for centre and corrected for age and sex as independent variables. Bonferroni correction resulted in level of significance < 0·006.
† Variable was log-transformed to obtain a normal distribution.
Discussion
Diet quality indices are valuable tools to obtain a global assessment of the dietary quality of a person or population. The present study aimed to investigate whether the developed DQI-A, calculated solely from food items, was an adequate proxy measure for adherence to the FBDG. This was done by comparing the DQI-A scores with the usual intake of different foods, of which some were not included in the calculation. The results showed that DQI-A scores were significantly associated with most food items in the expected direction. Nutrient-dense food items, such as fruits and vegetables, were positively associated, whilst non-recommended foods showed negative associations. No significant relation between the DQI-A score and the usual intake of meat, fish, eggs or their substitutes was found. This was due to the fact that this food group (calculated as the sum of meat, fish and eggs) was often consumed in excess, resulting in a lower score of the dietary equilibrium component. However, it is noteworthy that the overconsumption of this food group is mainly due to an excessive intake of meat products rather than fish or meat substitutes. Fish is only over-consumed in a minority of the adolescents (twenty-six of the 1804 adolescents had a habitual fish consumption of >100 g/d; range 0–328 g/d). Consumption of meat substitutes ranged from 0 to 143 g/d, with eight adolescents consuming more than 100 g/d. Furthermore, a positive relation was observed between the DQI-A score and the consumption of fat and oils. A moderate intake of fat and oil is recommended in the Flemish FBDG. Also, a high consumption of vegetable oils is in line with the Mediterranean diet and the FBDG of Greece and Spain(30). Moreover, adherence to the Mediterranean diet has been shown to have beneficial effects on cardiovascular risk factors(Reference Estruch, Martinez-Gonzalez and Corella44). In the past, it was generally assumed that saturated fats induced a higher risk of CVD; however, this has been questioned lately, as replacement of SFA with refined carbohydrates was suspected to increase the risk of CVD(Reference Mozaffarian45, Reference Willett46). Furthermore, children and adolescents have higher lipid intake needs, which is essential for growth. Besides energy delivery, lipids have an important function as structural components in all tissues, because they are indispensable for cell and plasma membrane synthesis(Reference Uauy, Mize and Castillo-Duran47).
Another interesting fact was the inverse relationship between the DQI-A score and energy intake. This suggested that adolescents with large, excessive food intake, and thus more likely to meet minimal intake recommendations, did not necessarily obtain a higher diet quality score. This was in contrast with other DQI validation studies where participants consuming more food, and thus more total energy, had higher quality scores compared with adolescents who ate less(Reference Feskanich, Rockett and Colditz11, Reference Lowik, Hulshof and Brussaard48).
In the development of the DQI-A, three other variants have been studied, namely, one with inclusion of a measure for meal frequency, one with inclusion of a measure of physical activity and both (data not shown). The DQI-A, as described in the present paper, however, showed the strongest associations with the different food items as well as the largest variance between individuals. Overall, based on these results, it can be concluded that the DQI-A is indeed a good surrogate marker for adherence to FBDG.
The second aim of the present study was to investigate whether adherence to these FBDG, using the DQI-A as a proxy measure, resulted in a better nutritional intake and blood biomarker profile. Indeed, the DQI-A was strongly related to higher intakes of water, fibre and most minerals and vitamins. The high fibre intake was a clear representation of the FBDG to consume sufficient vegetables and fruits, and to choose wholegrain products(28). Also, the DQI-A was positively associated with complex carbohydrates, whilst the usual intake of mono- and disaccharides decreased with better adherence to the guidelines, as would be expected(28, 49). Increased consumption of simple carbohydrate-rich foods has been associated with obesity, type 2 diabetes and the metabolic syndrome(49, Reference Novac, Matasaru and Tataru50). The absence of association with vitamin C intake might be attributed to the fact that vitamin C is prevalent in fruit juices, which are considered as non-recommended foods because of their high energy density. As such, high intakes of these items, and thus vitamin C, resulted in a lower DQI-A, whilst a high intake of fruits and vegetables, and thus also vitamin C, tended to increase the DQI-A. Furthermore, the lack of association of Fe and niacin with the adherence to FBDG could be due to the high concentration of these nutrients in meat products, whilst this food group showed no association with the DQI-A. In the present study population, no association could be found between the DQI-A score and the absolute intake of fat and FA. This might be due to the fact that meat was the largest contributor to fat intake, followed by non-recommended foods and dairy products(Reference Vyncke, Libuda and De51). As such, the ‘penalisation’ for the excessive intake of meat and non-recommended foods is counteracted by the recommended intake of dairy products. Also, associations might be attenuated due to faults that may arise through linkage with food composition tables. Of course, this finding might also indicate that the present FBDG are well tuned to the micronutrient recommendations, but that the guidelines are not efficient in transferring the recommendations for fats and FA.
Previously, it has been shown that biomarkers do not always perform better than food intake assessment methods to evaluate true nutrient intake(Reference Kabagambe, Baylin and Allan52). Moreover, not all nutrients have well-defined biological markers and many are influenced by other factors than intake. In the present study population, a positive association was found between the DQI-A and levels of 25(OH)D and holo-transcobalamin, representing the bioactive fraction of vitamin B12, and n-3 FA. Amongst others, plasma levels of 25(OH)D are related to a better bone mineralisation(Reference Holick53), while deficiency has been linked to the pathogenesis of several disorders, including cancer, hypertension, multiple sclerosis and diabetes(Reference Zhang and Naughton54). Adequate folate and vitamin B12 levels are essential for good growth and development of the central nervous system in fetal and infant life(Reference Selhub and Paul55). Both folate and vitamin B12 are also essential for the synthesis of nucleotide precursors, so if both are deficient, this can result in impaired cell division and anaemia(Reference Pepper and Black56). In addition, deficiencies in folate and vitamin B12 result in high values of homocysteine in blood and tissues, which in turn is associated with organ dysfunction in children that may lead to disease later in life, such as CVD(Reference Moreno, González-Gross and Kersting16, Reference Bjorke Monsen and Ueland57). Also, trans-FA levels showed an inverse relation with the DQI-A, which supports the recommendation of discouraging trans-FA intake in the human diet because of their association with an increased cardiovascular risk(Reference Remig, Franklin and Margolis58). This finding was, however, not significant at the corrected level of P <0·006.
The aim of the DQI-A was to obtain a measure for overall dietary quality of an individual. In the present study, statistically significant associations were only found with biomarkers representing long-term dietary intake (25(OH)D and holo-transcobalamin), whilst no statistically significant relationship was seen with biomarkers representing short- to medium-term intake. This might indicate that the DQI-A is a valid measure for long-term dietary habits. Also, the associations with the other biomarkers might be attenuated, as supplement use was not taken into account in the present study.
Several diet quality indices have been associated with specific nutrient intakes and plasma biomarkers. Comparison is difficult as different statistical approaches have been applied. Similar to the present results, Hann et al. (Reference Hann, Rock and King59) found significant positive associations between the Healthy Eating Index and vitamin C, folate and fibre intake in a sample of adult women. In contrast to the present results, carbohydrates and total energy were also positively associated. Also Newby et al. (Reference Newby, Hu and Rimm60) found significant associations between the DQI-Revised and intakes of several vitamins and minerals. In contrast to the present results, however, total fat and saturated fat were negatively correlated with the index. However, in this index, low total fat intake and low saturated fat intake were incorporated as two of the ten separate components in the calculation.
Both the Healthy Eating Index and the DQI-Revised showed significant correlations with β-carotene serum values (r 0·12–0·42)(Reference Hann, Rock and King59–Reference Weinstein, Vogt and Gerrior61). Furthermore, Weinstein et al. (Reference Weinstein, Vogt and Gerrior61) found significant, but generally weak, correlations between the Healthy Eating Index and serum vitamin C (r 0·21), serum folate (r 0·15), serum vitamin B12 (r 0·01), serum retinol (r 0·05) and serum TAG (r 0·06) in a large study population (n >16 000). Moreover, Neuhouser et al. (Reference Neuhouser, Patterson and King62) could not observe any correlation between the DQI and long-chain n-3 phospholipid FA in a sample of 102 women, whilst Gerber et al. (Reference Gerber, Scali and Michaud63) found a significant association with n-3 erythrocyte FA. Both studies could not observe a relationship with serum β-carotene.
Strengths and limitations
This was the first study to investigate the use and validity of a DQI in a European adolescent population. In the HELENA study, all data were collected according to standardised protocols and strict procedures. Furthermore, in contrast to many other diet quality indices, the DQI-A in the present study was not based on nutrient intakes, avoiding the limitations that coincide with the use of food composition data (such as the use of various tables in different countries with different methods of analysis used, unavailability of food items, loss of dietary information from mixed dishes, etc.).
A possible limitation of the present study is the observed significant differences in sex distribution and BMI between underreporters and non-underreporters. This differential underreporting could most probably attenuate the present results, as generally it would be expected that adolescents with a higher BMI would have a less healthy dietary pattern. Underreporting by these adolescents might, however, result in better DQI-A scores than that in reality. This could attenuate the present results, especially the associations with the biomarkers. Therefore, it was chosen to exclude the underreporters. This decision was supported by a previous evaluation of food and nutrient intake assessment in the HELENA study population(Reference Vandevijvere, Geelen and Gonzalez-Gross19), showing better correlations with the ‘true intake’ (calculated with the Triads method(Reference Kaaks18)) after exclusion of underreporters. However, underreporting cannot be fully precluded, as exclusion was only applied for adolescents indicating a negative energy balance. As such, attenuation of results might still be present.
The developed DQI-A score was based on the Flemish FBDG, whilst large variation in dietary habits was observed in the studied population. These guidelines were selected as the basis of the index score because of the great similarities with the CINDI pyramid developed by the WHO. The Flemish guidelines, however, are more specific concerning quality of different food items and recommended quantities. Compared to dietary guidelines of the other European countries, rather minor differences were found(30). The French and Austrian guidelines put greater emphasis on vegetables v. cereal intake compared to the Flemish. Furthermore, daily olive oil consumption is a specific recommendation in the Greek and Spanish guidelines. Given the large similarities on the most important aspects of the dietary guidelines, the authors considered it appropriate to apply the Flemish guidelines to a European population.
The DQI-A was calculated based on two self-administered, computer-assisted, non-consecutive 24 h recalls. Following recommendations of the ‘European Food Consumption Survey method’, 24 h recalls were preferred as these are open-ended questionnaires from which detailed information can be obtained. Furthermore, they are applicable in large populations of different ethnicity, and standardisation is possible by self-administered computer-assisted recall methods with pictures for portion sizes(Reference Biro, Hulshof and Ovesen22, Reference Vereecken, Dohogne and Covents64). According to Biro et al. (Reference Biro, Hulshof and Ovesen22), the 24 h recall method is appropriate to assess both acute and usual intake on the individual level by repeated short-term measurements and modelling.
A limitation of the method used is, however, that only information of 2 d was obtained. Although this allows inclusion of exceptional intakes at the individual level, this effect is neutralised by the large number of observations. The 24 h dietary recall method does not allow quantifying proportions of non-consumers for particular food items, especially for infrequently consumed foods. In order to decrease this influence, nutrient intakes were corrected for within-person variability by applying the Multiple Source Method. Moreover, accuracy of collected data relies on the individual's ability to remember foods and beverages consumed in the past 24 h, and might, therefore, be biased towards underreporting. In this respect, the 24 h dietary recalls were performed through the computer-assisted HELENA-DIAT program(Reference Vereecken, Covents and Sichert-Hellert24) to standardise the recall procedures as much as possible. Another limitation of the use of 24 h recall interviews is the potential loss of dietary information from mixed dishes, as food ingredients were sometimes counted from mixed dishes.
The same food composition table for conversion of food intake data to estimated nutrient intakes was used for all survey centres. In this way, differences in definitions, analytical methods, units and modes of expression were overcome. In this regard, the German food composition tables (Bundeslebensmittelschlüssel, BLS) were chosen. The BLS is based on German, American, English, Swedish, Danish and Dutch food composition tables, on analytical values of food producing firms, publications and research results of the Federal Research Centres and Universities(65). The BLS includes about 11 000 raw and cooked foods and recipes, and has been widely used in epidemiological studies(Reference Deharveng, Charrondiere and Slimani66).
Conclusion
The present study has shown good validity of the DQI-A by confirming the expected associations with food and nutrient intakes and some biomarkers in blood. However, further investigation is necessary to explore why the present guidelines do not reach their goal of obtaining a more favourable lipid intake with increasing DQI-A scores.
Acknowledgements
The HELENA study has taken place with the financial support of the European Community Sixth RTD Framework Programme (Contract FOOD-CT-2005-007034). There was additional support from the Spanish Ministry of Education (AGL2007-29784-E/ALI), Axis-Shield Diagnostics Limited (Oslo, Norway), Abbot Científica S.A. (Spain) and Cognis GmbH (Germany). K. V., T. D. V. and C. V. are financially supported by the Research Foundation Flanders. U. A. is financially supported by the Universidad Politécnica de Madrid. The content of the present article reflects only the authors' views and the European Community is not liable for any use that may be made of the information contained therein. Many thanks to Christel Bierschbach, Adelheid Schuch, Anke Berchtold, Petra Pickert, Andre Spinneker and Anke Carstensen for their contribution to the laboratory work. L. A. M., M. K., A. M., Y. M., M. G.-G., J. D., L. M., F. G., M. S., A. K., D. M., K. W., C. L. and S. D. H. designed the research. I. H. performed the data cleaning, K. V. and E. C. F. did the statistical analysis. K. V., M. F.-P., M. C.-G., W. D. K., T. M., T. D. V., C. B., K. B., C. V. and I. H. wrote the paper and L. B., C. B., U. A., K. D., E. G., W. G., M. V. W., L. E. D., A. G. and L. H. edited the manuscript. K. V. had primary responsibility for the final content. All authors read and approved the final manuscript. None of the authors had a personal or financial conflict of interest. The authors declare not having any conflicts of interest.
