The complex nature of diet, with specific combinations of foods and nutrients, impedes nutrition epidemiology when investigating diet–disease associations; therefore a dietary pattern approach may be more appropriate when investigating disease outcomes(Reference Jacobs and Tapsell1–Reference Michels and Schulze7). Diet quality indices consist of a combination of foods and/or nutrient components that together represent dietary guidelines(Reference Newby and Tucker8–Reference Waijers, Feskens and Ocké10). The use of index-based dietary patterns is one way to address the complexity of diet, while at the same time evaluating the effectiveness of current dietary guidelines to prevent chronic diseases(Reference Arvaniti and Panagiotakos11). Recent reviews have reported more than twenty indices currently in existence(Reference Waijers, Feskens and Ocké10–Reference Kourlaba and Panagiotakos14). Studies using dietary indices have found associations with nutrient adequacy(Reference Kim, Haines and Siega-Riz15–Reference Neuhouser, Patterson and King18), biomarkers of disease(Reference Kant and Graubard19–Reference Boynton, Neuhouser and Wener21), premature mortality(Reference Huijbregts, Feskens and Räsänen22–Reference Kant, Schatzkin and Graubard26), CVD(Reference McCullough, Feskanich and Stampfer27, Reference Alberti, Fruttini and Fidanza28) and certain forms of cancer(Reference Fung, Hu and McCullough29, Reference Reedy, Mitrou and Krebs-Smith30). However, few previously constructed indices are suitable for direct application in a Swedish population since most include foods not typically consumed in Swedish diets or are based on dietary guidelines different from those in Sweden.
The official Swedish dietary guidelines consist of the Swedish nutrition recommendations (SNR) 2005 and the Swedish dietary guidelines (SDG)(31, 32). The SNR 2005 are based on the Nordic Nutrition Recommendations 2004 (NNR 2004) and include recommendations for the macronutrient composition of the diet, the daily micronutrient intake, and reference values for energy intake(33). The NNR 2004 are based on available scientific evidence and provide a basis for evaluating the intake of nutrients in groups of healthy individuals(33). The SDG are food-based dietary guidelines published by the Swedish National Food Administration which communicate the concept of a diet that fulfils the SNR 2005(32). In order to illustrate the practical application of the SNR 2005, the Swedish Nutrition Recommendations Objectified (SNO) project developed a food intake pattern for adults based on seventy-one foods considered to be representative of Swedish eating habits(Reference Enghardt-Barbieri and Lindvall34). A comparison between SNO and the most recent national dietary survey, Riksmaten 1997/98, formed the basis for the main SDG, which aim at increasing the population's intakes of fruit and vegetables, wholegrain products, fish and vegetable oils, and limiting intakes of animal fat, sugar and salt(32, Reference Becker and Pearson35).
Despite increased public awareness of the importance of diet in decreasing the risk of chronic disease, large gaps remain between current recommendations and actual dietary practices in Sweden(Reference Becker and Pearson35, Reference Becker36). The nutritional problems in Sweden relate mainly to a macronutrient distribution with too much fat (especially saturated fat), refined sugar and alcohol, and too little complex carbohydrates and dietary fibre. Micronutrient intakes are generally sufficient apart from deficient intakes of vitamin D, folate and Fe in certain subgroups of the population, and an overconsumption of salt(33, Reference Becker and Pearson35, Reference Becker36).
The aim of the present study was to develop and evaluate a diet quality index (DQI) that may be used as a tool to distinguish high- and low-quality diets by assessing adherence to the SNR 2005 and SDG.
Materials and methods
Study population
The Malmö Diet and Cancer (MDC) study is a population-based prospective cohort (n 28 098). Baseline examinations were conducted between 1991 and 1996, and eligible participants were men born between 1923 and 1945 and women born between 1923 and 1950, living in the city of Malmö and with Swedish reading and writing skills. The MDC study was approved by the Ethical Committee at the Medical Faculty, Lund University (LU 51-90). Details of the recruitment procedures and the cohort are given elsewhere(Reference Berglund, Elmståhl and Janzon37, Reference Manjer, Carlsson and Elmstahl38). The data collection included dietary habits, socio-economics, medical history and lifestyle habits using questionnaires and interview. Halfway through the baseline data collection, a change of coding routines was implemented in order to reduce interview time. For descriptive studies it has therefore been recommended to use the subset of the population that completed the baseline examinations after this revision(Reference Wirfält, Mattisson and Johansson39). Thus the present study is based on participants, men (n 4525) and women (n 8466), enrolled in the MDC study after 1 September 1994 (total n 12 991).
Dietary assessment
Information on dietary habits was obtained through a modified diet history method combining a 7 d menu book (collecting descriptions of prepared meals, nutrient supplements and cold beverages) and a 168-item quantitative diet questionnaire using both exact frequencies and a picture booklet to assess portion sizes of regularly eaten foods other than prepared meals during the past year. During a 1 h interview, the questionnaire and the menu book were checked so that reported food consumption did not overlap and detailed information was collected on cooking practices and recipes. Energy and nutrient intakes were computed from the reported food intake using the MDC Food and Nutrient Database, originating mainly from PC Kost2-93 of the National Food Administration in Uppsala, Sweden(Reference Callmer, Riboli and Saracci40). Data on the validity(Reference Elmståhl, Riboli and Lindegärde41, Reference Riboli, Elmståhl and Saracci42) and reproducibility(Reference Elmståhl, Gullberg and Riboli43) of the method have been published. The relative validity of the MDC method is high compared with other dietary assessment methods in similar populations(Reference Thiébaut, Kipnis and Schatzkin44, Reference Willett45).
Socio-economic and lifestyle variables
Information on socio-economic and lifestyle factors was collected from the extensive lifestyle and socio-economic questionnaire. A high educational level was defined in the present study as upper secondary school or higher (more than 11 years). The smoking habits of the participants were defined as current smokers (including irregular smokers), former smokers or never smokers. Leisure-time physical activity was assessed by a list of activities in the questionnaire, modified from the Minnesota Leisure Time Physical Activity Questionnaire(Reference Taylor, Jacobs and Schucker46–Reference Li, Aronsson and Hedblad48). Participants were asked to report how many minutes per week on average, and for each of the four seasons, they spend on a specific activity. A physical activity score was obtained by multiplying the number of minutes for each activity with an activity-specific factor. A high leisure-time physical activity was defined as individuals in the highest gender-specific tertile of activity score. Physical activity at work self-rated by participants as very light or light was defined as sedentary work. Alcohol habits were classified as zero, low, moderate or high consumption. Participants reporting no alcohol intake in the 7 d menu book and reporting no alcohol intake during the preceding year in the questionnaire were classified as zero alcohol consumers. For all other participants low, medium and high alcohol consumption level was set at alcohol intakes of <15, 15–30 and >30 g/d for women, and <20, 20–40 and >40 g/d for men. Mattisson et al.(Reference Mattisson, Wirfält and Aronsson49) have previously defined low, adequate and high energy reporters in the MDC cohort using the approach described by Goldberg et al.(Reference Goldberg, Black and Jebb50) and later refined by Black(Reference Black51). Energy misreporting was defined as having a ratio of energy intake to BMR outside the 95 % CI limits of the calculated physical activity level. Past food habit change was based on the questionnaire item ‘Have you substantially changed your dietary habits because of illness or another reason?’ Previous cardiovascular event (including coronary event or stroke) was determined though local registers and diabetes diagnosis was self-reported by the participants.
Anthropometric variables
Nurses measured height (m), weight (kg), waist and hip circumferences (cm). BMI was calculated as weight divided by the square of height (kg/m2). Blood pressure (mmHg) was measured after 10 min of rest.
Development of the diet quality index (DQI-SNR)
Index component selection
The SNR and SDG are overlapping and complementary, and therefore both were considered for identification of suitable index components (Table 1). Three main aspects were considered during the selection process. First, information on dietary intake of candidate components had to be available within the MDC cohort. Initial assessment therefore excluded trans fatty acids and salt intake, since information on these dietary factors was either lacking or had poor validity. Second, dietary components considered most important in assuring overall diet quality (i.e. food groups and macronutrients) in relation to chronic disease were primarily considered. Third, inter-correlation between components was investigated to assess the mutual independence between components(Reference Waijers, Feskens and Ocké10, Reference Kourlaba and Panagiotakos14).
Table 1 Summary of the Swedish nutrition recommendations (SNR 2005) and the Swedish dietary guidelines (SDG)
% E, percentage of energy.
Considering these three aspects, the following components were selected: SFA, PUFA, fish and shellfish, dietary fibre, fruit and vegetables, and sucrose. SFA and PUFA were selected to reflect the fat quality of the diet as well as reflecting the recommendation to use liquid margarine or vegetable oils in food preparation. There is a strong correlation between the subtypes of fat in the MDC cohort. The strong correlation (r = 0·65) between MUFA and SFA in the MDC cohort suggests that the dietary sources of these two fats are mainly the same (i.e. from meat and dairy products). This has recently been highlighted by a pooled analysis of dietary fat and CHD(Reference Jakobsen, Eilis and Heitmann52). Therefore, MUFA was not included as a component. In addition, SFA and PUFA jointly reflect total fat intake and low SFA intake is also likely to reflect a preference for low-fat dairy and low-fat meat products. The different subtypes of PUFA were similarly highly correlated (r = 0·97) and total PUFA was therefore selected as an index component, as it is likely to reflect intakes of both α-linolenic acid and linoleic acid. Dietary fibre reflects intake of high-fibre cereals (i.e. whole grains; r = 0·39) and fibre from fruit and vegetables (r = 0·63). The food-based dietary guideline for increasing dietary fibre intake by choosing fibre-rich foods was recently changed to selecting primarily wholegrain cereals. However, there is no specified amount of recommended whole grains intake; therefore dietary fibre was included as a component to reflect this aspect of diet. Sucrose was included to reflect intake of refined sugars and consumption of sugar-rich foods. Table 2 shows a description of the DQI-SNR.
Table 2 DQI-SNR: componentsFootnote †, cut-offsFootnote ‡ and differences in component adherence (%) between men (n 4525) and women (n 8491) of the Malmö Diet and Cancer cohort
DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005.
† Expressed as non-alcohol energy percentage (%E) unless otherwise noted.
‡ Adherence provides 1 point to total score, non-adherence 0 points.
§ P from χ 2 test. Values for non-adherence are not shown.
Cut-offs and scoring
A cut-off for each index component was assigned based on the recommended level of intake in the SNR 2005 or the SDG (see Table 1). A score of 1 was assigned to individuals complying with the recommendation and a score of 0 was assigned to those not complying with the recommendation. A total score was created by summing the score of each component, thus total score ranged from 0 to 6. Using predefined cut-offs can cause problems when the population has intake levels that are not close to the recommendation. This was the case for SFA, where only 2 % of the MDC population complied with the recommendation (≤10 % of energy (%E)). Therefore, the cut-off for SFA was defined by adding one standard deviation of the mean population intake to the recommended level, resulting in a new cut-off (≤14 %E). The recommendation for dietary fibre intake is 25–35 g/d (approximately 3 g/MJ). The energy relative recommendation has no defined upper or lower limit, and therefore a similar approach was applied for the dietary fibre component (i.e. ±1 sd of the population mean). The fruit and vegetable component excludes fruit juices and the cut-off was therefore set at ≥400 g/d, instead of the recommended ≥500 g/d (including a maximum of 100 g fruit juices).
Evaluation of the diet quality index (DQI-SNR)
It is important that the index assesses diet quality independent of diet quantity. Because most nutrient intakes are positively correlated with energy intake, a diet quality index could overrate high-energy diets if intakes are measured in terms of absolute amounts. In nutrition epidemiology, energy-adjusted variables ensure the isoenergetic principle and in addition reduce the impact of measurement errors associated with self-reported data(Reference Willett, Howe and Kushi53). Recommendations for most of the index components are expressed in energy relative terms (%E or g/MJ). However, fish intake and fruit and vegetable intake are expressed as gram per week or d. Energy independence was therefore investigated by examining correlations between components, total score and total energy intake. In addition, since alcohol intake tends to dilute diet composition and has established health effects, we used non-alcohol energy intake for calculation of energy percentage (%E).
The face validity of the index, i.e. quantitative evaluation of how well the index measures what it is supposed to measure(Reference Trochim54), was assessed in several ways. First, the correlation between the total score and the index components was investigated to assess the relative weight of components. Second, we examined how total score was associated with nutrient intakes and selected food groups. The nutrient variables investigated in the study included total energy intake and relative intakes of fat (total fat, SFA, MUFA, PUFA, n-6 and n-3 fatty acids), carbohydrates (including %E from total carbohydrates and dietary fibre intake), protein and alcohol. The intakes of selected vitamins and minerals (including retinol equivalents, vitamin D, vitamin E, thiamin, riboflavin, niacin equivalents, vitamin B6, folic acid, vitamin B12, ascorbic acid, Ca, P, K, Mg, Se, Zn and Fe) were assessed by investigating the proportion of the study sample reaching the recommended intake levels in the NNR 2004 (g/MJ)(33). In addition, intakes of fruit and berries, vegetables, fish and shellfish, butter and margarines, cereals, high-fibre cereals, potatoes, rice and pasta, high- and low-fat dairy, high- and low-fat meat, soft drinks, cakes and biscuits, and sweets and chocolate were examined. Third, we examined the concurrent-criterion validity; whether the index can distinguish between groups of people with different lifestyle and socio-economic characteristics that have previously been observed to report diets of diverging quality(Reference Trochim54).
Statistical analyses
The proportion of participants meeting each component cut-off was calculated and the χ 2 test was used to compare differences between men and women. Pearson partial correlations were estimated to evaluate inter-correlation among the index components as well as the correlation between index components, energy intake and total score; analysis was adjusted for age, sex and season (autumn, winter, spring, summer). Mean intakes of the DQI-SNR components by total index score were assessed while adjusting for age, season and total energy intake. Because there were few individuals in the low and high DQI-SNR score subgroups, subsequent analyses investigated categories of DQI-SNR score, i.e. low score (0 or 1 point), medium score (2 or 3 points) and high score (4–6 points). Mean nutrient and food intakes were investigated across categories of DQI-SNR score and linear trends across continuous DQI-SNR score were calculated adjusting for age, season and total energy intake. Associations between DQI-SNR score and participant characteristics were investigated using the linear regression coefficient for the DQI-SNR score for continuous variables and the logistic regression coefficient for the DQI-SNR score for dichotomous variables, while adjusting for age and total energy intake. Additional adjustments were made for leisure-time physical activity for anthropometric characteristics and for leisure-time physical activity and BMI for disease history characteristics. The SPSS statistical software package version 17 (SPSS Inc., Chicago, IL, USA) was used for all statistical analyses; P < 0·05 was considered significant.
Results
Table 2 describes the construction of the DQI-SNR and shows the percentage of men and women adhering to the selected cut-offs. Table 3 shows the correlation coefficients for components, total score and total energy intake. Most components showed low to intermediate inter-correlation. Dietary fibre was negatively correlated with SFA (r = −0·52). There was also a positive correlation between fruit and vegetables and dietary fibre (r = 0·63). Total score showed energy independence (r = −0·04); however, there was some correlation between the individual components and total energy intake. Correlations between individual components and total score ranged from |r = 0·21| to |r = 0·53|.
Table 3 Partial correlationsFootnote † between DQI-SNR componentsFootnote ‡, total score and total energy intake among men and women (n 12 991) of the Malmö Diet and Cancer cohort
DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005.
* P < 0·01.
† Adjusted for gender, age and season.
‡ Dietary variables are expressed as percentage of energy (%E) from non-alcohol energy intake (SFA; PUFA; sucrose), g/MJ (dietary fibre), g/week (fish and shellfish) or g/d (fruit and vegetables).
Population median intake and interquartile range of the DQI-SNR components are shown in Table 4, along with mean intakes of the components across total index score. All trends across score were significant (P < 0·0001). Mean relative intakes of macronutrients across categories of DQI-SNR score (low, medium and high) are shown in Table 5. All trends were significant with P < 0·0001; however, the trends were not as clear as for all macronutrients (i.e. total fat and total carbohydrate). Figures 1(a) and 1(b) respectively show the mean intakes of different food groups in the categories of DQI-SNR score among men and women. Participants with a high DQI-SNR score had significantly higher intakes of fruit and berries, vegetables, fish and shellfish, margarines, cereals, high-fibre cereals, low-fat dairy products and meat (all P for intakes across continuous total index score <0·0001) and lower intakes of butter, high-fat dairy and meat, soft drinks, cakes and biscuits, and sweets and chocolate (all P < 0·0001). Men and women with a high DQI-SNR score were also more likely to reach the recommended intake levels for most micronutrients (Figs 2(a) and 2(b), respectively). However, although significantly higher intakes of vitamin D and folate (P for trend <0.0001) were found for participants with a high score, most of the study population was below the recommended intake level for both vitamin D (82 %) and folate (98 %). In addition, most participants (close to 100 %) reached the recommended intake level for retinol equivalents, riboflavin, niacin equivalents, vitamin B6, vitamin B12 and P.
Table 4 Population median intake and interquartile range of the specific DQI-SNR componentsFootnote †, total energy intakeFootnote ‡ and adjusted mean intakesFootnote § of the DQI-SNR components by index score, among men (n 4525) and women (n 8491) of the Malmö Diet and Cancer cohort
DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005; IQR, interquartile range.
† Expressed as non-alcohol energy percentage (%E) unless otherwise noted.
‡ Adjusted for age and season.
§ Adjusted for age, season and total energy intake.
Table 5 AdjustedFootnote † mean macronutrient intakes (%E) by categories of DQI-SNR score among men (n 4525) and women (n 8491) of the Malmö Diet and Cancer cohort
%E, percentage of energy; DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005.
† Adjusted for age, total energy intake and season.
‡ P for trend was determined using the DQI-SNR score in its continuous form and represents the P value associated with the linear regression coefficient for the DQI-SNR score for continuous nutrient variables.
Fig. 1 Mean intake (g/d) of different food groups adjusted for energy intake, age and season in categories of DQI-SNR score (- -○- -, low score (0 or 1 point); · · ● · ·, medium score (2 or 3 points); —▵—, high score (4–6 points)) among (a) men (n 4525) and (b) women (n 8491) of the Malmö Diet and Cancer cohort. All P for the linear trends of mean intake across total index score (continuous) were <0·0001 (DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005)
Fig. 2 Adherence (%) to recommended intake level (g/MJ) of selected micronutrients in categories of DQI-SNR score (- -○- -, low score (0 or 1 point); · · ● · ·, medium score (2 or 3 points); —▵—, high score (4–6 points)) among (a) men (n 4525) and (b) women (n 8491) of the Malmö Diet and Cancer cohort (DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005)
Tables 6a and 6b show how selected characteristics of participating men and women are related to DQI-SNR score. Individuals with higher DQI-SNR score were more likely to be older, more physically active during leisure-time and at work, non-smokers, have a higher educational level and socio-economic status, and have changed their food habits in the past. Among women, energy misreporting was also associated with a higher score. In addition, a high score was associated with diabetes diagnosis and a previous cardiovascular event.
Table 6a Anthropometric, lifestyle and socio-economic factors, and disease history by categories of DQI-SNR score among men (n 4525) of the Malmö Diet and Cancer cohort
DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005.
†P for trend was determined using the DQI-SNR score in its continuous form and represents the P value associated with the linear regression coefficient for the DQI-SNR score for continuous variables (age, BMI, waist, waist-to-hip ratio) and the logistic regression coefficient for the DQI-SNR score for dichotomous variables (high leisure-time physical activity, sedentary work, current smokers, moderate/high alcohol consumption, past food habit change, energy misreporting, high educational level, high socio-economic status, blood pressure, previous cardiovascular event and diabetes diagnosis). Analysis adjusted for age (except for mean age across score) and total energy intake, unless otherwise noted.
‡Adjusted for age, total energy intake and leisure-time physical activity.
§Adjusted for age, total energy intake, leisure-time physical activity and BMI.
Table 6b Anthropometric, lifestyle and socio-economic factors, and disease history by categories of DQI-SNR score among women (n 8491) of the Malmö Diet and Cancer cohort
DQI-SNR, diet quality index based on the Swedish nutrition recommendations 2005.
†P for trend was determined using the DQI-SNR score in its continuous form and represents the P value associated with the linear regression coefficient for the DQI-SNR score for continuous variables (age, BMI, waist, waist-to-hip ratio) and the logistic regression coefficient for the DQI-SNR score for dichotomous variables (high leisure-time physical activity, sedentary work, current smokers, moderate/high alcohol consumption, past food habit change, energy misreporting, high educational level, high socio-economic status, blood pressure, previous cardiovascular event and diabetes diagnosis). Analysis adjusted for age (except for mean age across score) and total energy intake, unless otherwise noted.
‡Adjusted for age, total energy intake and leisure-time physical activity.
§Adjusted for age, total energy intake, leisure-time physical activity and BMI.
Discussion
The present study describes the development of a diet quality index based on the SNR 2005 and the SDG. Such an index could potentially assume several different designs. We designed the DQI-SNR to assess the key dietary factors that are thought to be related to development of chronic disease, under the prerequisite that these would jointly reflect overall diet quality.
One drawback of dietary indices is that it is possible only to compare individuals with high and low scores(Reference Waijers, Feskens and Ocké10). A large proportion of the study population was given an intermediate score (i.e. 2 or 3 points) and these individuals could have very different diets. However, intakes of the DQI-SNR components showed significant trends across total score demonstrating the index's capability of separating individuals based on intakes of the selected index components. Results also showed that index score was associated with intakes of several foods groups. Notably, a high score was associated with a diet in accordance with the SDG: choosing margarine instead of butter, high-fibre cereals, low-fat dairy and meat products, and consuming larger amounts of fruit and vegetables. Individuals with high scores also seemed to avoid or consume less high-fat dairy, soft drinks, cakes, biscuits, sweets and chocolate. In addition, index score was associated with intake of several macronutrients other than the components. Individuals with high scores seemed to meet the recommendations for vitamins and minerals to a greater extent, notably for vitamin D, vitamin E, ascorbic acid, K, Mg, Fe, Zn and Se. Only a small proportion of the total study population managed to reach recommended intake of vitamin D (18 %) and folate (2 %), which is in accordance with what is previously known about the nutritional status of Swedish populations(33, Reference Becker and Pearson35, Reference Becker36). We have also shown that the DQI-SNR score relates as expected to participant characteristics. For example, one would expect that individuals with higher age, non-smokers, high physical activity level and high educational and socio-economic status would consume a higher-quality diet, as seen with the DQI-SNR. In addition, individuals reporting to have changed their food habits in the past (due to illness or other reason), having diabetes or having a previous cardiovascular event were more likely to have a higher total score. This is most likely a reflection of the fact that individuals who change their diets due to illness rely on the current dietary recommendations(Reference Sonestedt, Wirfält and Gullberg55). The associations seen with past food habit change and energy misreporting highlight the importance of excluding these individuals from analysis when investigating associations with disease endpoints.
Overall, the results of our study are consistent with the findings of previous studies using other indices in other populations(Reference Kim, Haines and Siega-Riz15, Reference Kennedy, Ohls and Carlson56–Reference Haines, Siega-Riz and Popkin58), as well as with other Swedish studies on dietary habits and nutritional status(33, Reference Becker36).
Recently published reviews on diet quality indices highlighted several unresolved methodological issues of the index approach including the selection of components, inter-correlation between components, cut-off points and scoring. The subjective decisions made by the researcher regarding these issues may affect the diagnostic capacity of indices(Reference Waijers, Feskens and Ocké10–Reference Wirt and Collins12, Reference Kourlaba and Panagiotakos14).
An important limitation of dietary indices is when some of the index items show high inter-correlation, which can result in some aspects of diet contributing greater weight to the overall index score(Reference Waijers, Feskens and Ocké10, Reference Arvaniti and Panagiotakos11). What level of correlation is acceptable, however, is highly subjective and there are no methodological papers discussing this issue. The inter-correlation between components of the DQI-SNR was low, apart from a negative correlation between dietary fibre intake and SFA (r = −0·54) and a positive correlation between dietary fibre and fruit and vegetables (r = 0·63). The correlation between dietary fibre and SFA is an example of an overlap in dietary behaviours that is inherent in the SNR 2005 (i.e. individuals consuming low-fat and high-fibre diets). Since there is no specified recommendation for intake of wholegrain cereals, dietary fibre was included to reflect intake of wholegrain/fibre-rich cereals. However, the high correlation between fibre and fruit and vegetables suggests some degree of overlap which will have to be investigated in future studies using this index. Despite using energy-adjusted components to a great extent, correlation with energy intake remained for some components as well as a very low correlation with total score. As other researchers have noted previously(Reference Kushi, Sellers and Potter59, Reference Mattisson, Wirfält and Gullberg60), this suggests that the energy density method is not sufficient for complete energy adjustment. Therefore, it will be important to adjust for total energy intake in diet–disease analyses using the DQI-SNR.
Another potential limitation is the categorization of food and nutrient intake into dichotomies of ‘compliance’ and ‘non-compliance’. Such categorization can conceal the true variability in the intake data and diminish the range of scores(Reference Waijers, Feskens and Ocké10, Reference Kourlaba and Panagiotakos14). In the present study we opted for simplicity and used simple cut-offs for all components in order to evaluate adherence to the recommended intake levels. However, in order to provide sufficient discriminatory power for the SFA component, we had to add 1 sd of the population mean to the cut-off value. In other study populations researchers should investigate the possibility of using the recommended intake as a cut-off (i.e. <10 %E). In addition, future studies need to examine the influence of each component as a continuous variable on disease outcome separately in the search for potential cut-offs with improved predictability.
Finally, the dietary assessment method used in the MDC is a diet history method, reflecting usual intake. Although the high relative validity of the method is a great advantage, it is not ideal for evaluating absolute intakes(Reference Wirfält, Mattisson and Johansson39). We therefore used relative intakes of most components, as well as for evaluation of nutrient intakes across total score. However, the fruit and vegetable component as well as the fish and shellfish component were based on absolute intakes (g/d or g/week) and thus the cut-offs for these components may not be appropriate. Previous studies within the MDC cohort have found that fruit and vegetable intake tends to be over-reported(Reference Elmståhl, Riboli and Lindegärde41).
Several dietary indices have previously been constructed. However, the Mediterranean Diet Score and the Healthy Diet Indicator for instance include food groups (e.g. olive oil and legumes) not consumed in large amounts in Swedish diets(Reference Huijbregts, Feskens and Räsänen22). Also, no previous score has specifically addressed the Swedish dietary recommendations. In addition, few studies consider inter-correlation between components. For example, the Recommended Food Score developed by Kant et al.(Reference Kant, Schatzkin and Graubard26), which has previously been successfully applied in a Swedish cohort(Reference Lauza, Håkansson and Brozozowska61), includes thirty-six recommended foods whereof a majority is fruit and vegetables, suggesting the possibility that this index is merely a reflection of fruit and vegetable intake rather than overall diet quality.
In conclusion, the results suggest that the DQI-SNR is a useful tool for assessing adherence to the SNR 2005 and the SDG. Future studies need to validate the DQI-SNR in relation to biochemical and clinical indicators of nutritional status, as well as assess its utility in assessing risk of mortality and morbidity. Also, the utility of the index needs to be examined in other population and age groups.
Acknowledgements
The study was supported by the Ernhold Lundström Foundation and Region Skåne. There are no conflicts of interest. I.D. is primary author of the paper and responsible for design, analysis and interpretation of data. All authors (I.D., B.G., U.E., E.S., J.N., P.W., B.H. and E.W.) contributed to the interpretation of data and finalisation of the manuscript. The authors are especially grateful to the men and women who participated in the Malmö Diet and Cancer Study.
