Obesity is associated with an increased risk for non-communicable diseases (NCD) such as CVD, type 2 diabetes, certain types of cancers and premature death. The most important variables that contribute to obesity include over-eating, sedentary lifestyle, genetic, environmental, neurological, physiological, biochemical, socio-cultural and psychological factors( Reference Hruby, Manson and Qi 1 ).
Diet is a major modifiable determinant of obesity, and diet quality has been defined as the degree to which a diet reduces the risk for NCD( Reference Hruby, Manson and Qi 1 – Reference Kant 3 ). To assess diet quality in a population, the a priori dietary pattern was introduced as a tool known as ‘diet quality/dietary quality indices’, which evaluates the level of adherence to a specified pattern or a set of recommendations( Reference Wirt and Collins 4 ) or reflects the risk gradient for major diet-related chronic diseases( Reference Kant 5 ). Assessing diet quality can provide information on dietary behaviours. Behaviours, as a holistic view, include several components that work synergistically on health and disease( Reference Kant 3 , Reference Kant 5 ). Thus, in recent years, diet quality indices have been developed to meet the requirement of this field of nutrition research( Reference Kennedy, Ohls and Carlson 6 ). To the best of our knowledge, there are no published comprehensive systematic reviews clarifying which of the developed indices for adherence to dietary recommendations can predict the risk for obesity among healthy populations.
The aim of the current review was to systematically summarise all available literature regarding the association of dietary indices with general obesity/abdominal obesity in observational studies.
Methods
Eligibility criteria
Study designs
Prospective (cohort), cross-sectional and nested case–control studies were included, whereas reviews, meta-analyses, commentaries, clinical trials, editorials or duplicate publications were excluded.
Participants
Studies were included if their participants had been healthy adult individuals (aged ≥18 years) of both sexes, and were excluded if they had been conducted on infants, children, adolescents, pregnant or lactating women, and patients.
Exposure
We included studies that investigated the Healthy Eating Index (HEI), the Diet Quality Index (DQI), the Dietary Guidelines for Americans Index (DGAI), the Dietary Guideline Index (DGI), the Framingham Nutritional Risk Score (FNRS), the Recommended Food Score (RFS), the Not Recommended Food Score (NRFS), the French Program National Nutrition Sante´-Guideline Score (PNNS-GS), the Dietary Quality Score (DQS), the Food Variety Score (FVS), the Elderly Dietary Index (EDI), the Australian Recommended Food Score (ARFS) and the Dietary Diversity Score (DDS) as exposures (Appendix). These studies included the DQI to score dietary guideline recommendations and the food guide pyramid. Studies were excluded if they had investigated dietary patterns or specific diets that were not based on dietary guidelines such as the Mediterranean dietary pattern or the Dietary Approaches to Stop Hypertension (DASH) diet.
Outcomes
The main outcomes in the studies reviewed were overweight, general obesity, overweight and/or obesity, and abdominal obesity. Weight status as defined by BMI (normal weight: 18·5–24·9 kg/m2; overweight: 25–29·9 kg/m2; obese: ≥30 kg/m2; overweight/obesity: ≥25 kg/m2) was based on World Health Organization classification( 7 ). Abdominal obesity was defined according to author-specified criteria such as the National Cholesterol Education Program Adult Treatment Panel Ш (waist circumference (WC) >88 cm for women and WC>102 cm for men)( Reference Grundy, Cleeman and Daniels 8 ), World Health Organization (WC≥94 cm for men and WC≥80 cm for women)( 7 ) or national cutoff points. Results of studies were reported using means for WC and BMI, weight or BMI changes, and OR or relative risk for overweight and obesity.
Search strategy
A search was performed of literature published between January 1990 and January 2016 using MEDLINE (http://www.nlm.nih.gov/bds/pmresources/html) and Embase (http://www.embase.com) by using diet(ary) quality, diet(ary) patterns, diet score, diet (quality) index, or diet indices in combination with BMI, WC, (general) obesity, or abdominal (central) obesity. Search results were downloaded and imported directly into EndNote (version X5).
Study selection
Study selection was a three-step process; first, we removed exact duplicate articles using EndNote tools. Second, to detect irrelevant studies, two authors (E. Y. and G. A.) screened all titles identified through the electronic database search. Third, E. Y. and G. A. read all abstracts and full texts and based on inclusion criteria, selected studies from the remaining articles. If the results of a study were reported in more than one publication, only the publication with the most complete results was retained. To avoid double-counting data from multiple publications, we juxtaposed author names, sample sizes and outcomes. Lastly, to identify other potentially relevant articles, G. A. examined the reference lists of selected articles; disagreements between E. Y. and G. A. were resolved by consensus, and when consensus could not be reached, a third author (P. M.) made the final decision.
Results
Using the above-mentioned search engines, 479 papers were identified, which described original English-language research studies conducted on humans. After reviewing titles, abstracts and full texts, 445 articles were excluded because they did not include a priori dietary patterns. Articles that assessed specific diets, for example, the Mediterranean or DASH diet scores were also excluded. In the final step, full texts were reviewed for cited references and one more article was included. Finally, thirty-four studies were selected for the current review (Fig. 1). Characteristics of the studies included are illustrated in Tables 1 and 2.
Summary of the findings for diet quality and obesity in cross-sectional studies
HEI, Healthy Eating Index; WC, waist circumference; WHR, waist:hip ratio; BF, body fat; F, female; M, male; SU.VI.MAX, SUpplementation en VItamines et Minéraux AntioXydants; AHEI, alternative HEI; DQI-R, Diet Quality Index Revised; RFS, Recommended Food Score; HEI-f, healthy eating index-food-frequency questionnaires; NRFS, Not Recommended Food Score; FVS, Food Variety Score; DQI-I, Diet Quality Index-International; DQS, Dietary Quality Score; DQI, Diet Quality Index; DGAI, Dietary Guidelines for Americans Index; DGI, Dietary Guideline Index; DDS, Dietary Diversity Score; DDSP, Dietary Diversity Score with Portions; EDI, Elderly Dietary Index; MEDIS, Mediterranean Islands Study.
For HEI, updated or modified versions such as alternative HEI (AHEI)( Reference McCullough, Feskanich and Stampfer 20 ), HEI-05( Reference Gao, Beresford and Frank 35 ) and HEI-2005( Reference Nicklas, O’Neil and Fulgoni 17 ) have been reported; however, the original HEI had the highest frequency in articles (n 13), followed by DQI (n 7), DGAI (n 5), RFS (n 4), DDS (n 3), and FNRS, DGI and FVS (n 2 each). However, EDI, DQS, ARFS and PNNS-GS scores, along with some of the scores mentioned above, were investigated in single studies only. Only one article had targeted a population of the elderly (≥65 years) and EDI was used as the DQI in this population( Reference Kourlaba, Polychronopoulos and Zampelas 34 ).
In all, ten studies had a prospective design, with only one of them focusing solely on women( Reference Wolongevicz, Zhu and Pencina 36 ) and the remaining on both sexes. Out of ten published articles, five had >10-year follow-ups, of which the study by Zamora et al.( Reference Zamora, Gordon-Larsen and Jacobs 37 ) had a 20-year follow-up. A total of five prospective studies were performed on the North American population( Reference Quatromoni, Pencina and Cobain 24 , Reference Gao, Beresford and Frank 35 – Reference Kimokoti, Newby and Gona 38 ) and the remaining were conducted in Iran( Reference Asghari, Mirmiran and Rashidkhani 18 ), Australia( Reference Arabshahi, van der Pols and Williams 39 ), France( Reference Lassale, Fezeu and Andreeva 40 ), Canada( Reference Sundararajan, Campbell and Choi 22 ) and Spain( Reference Funtikova, Baena-Diez and Koebnick 41 ). However, among cross-sectional studies, thirteen had been performed on North American( Reference Fung, McCullough and Newby 11 – Reference Guo, Warden and Paeratakul 13 , Reference McCullough, Feskanich and Rimm 15 – Reference Nicklas, O’Neil and Fulgoni 17 , Reference de Koning, Chiuve and Fung 19 , Reference McCullough, Feskanich and Stampfer 20 , Reference Quatromoni, Pencina and Cobain 24 , Reference Fogli-Cawley, Dwyer and Saltzman 25 , Reference Fogli-Cawley, Dwyer and Saltzman 27 , Reference Fogli-Cawley, Dwyer and Saltzman 28 , Reference Kourlaba, Polychronopoulos and Zampelas 34 ), three on Iranian( Reference Asghari, Mirmiran and Rashidkhani 18 , Reference Azadbakht, Mirmiran and Esmaillzadeh 31 , Reference Azadbakht and Esmaillzadeh 32 ) and one each on Brazilian( Reference Tardivo, Nahas-Neto and Nahas 9 ), French( Reference Drewnowski, Fiddler and Dauchet 10 ), Guatemalan( Reference Gregory, McCullough and Ramirez-Zea 23 ), Sri Lankan( Reference Jayawardena, Byrne and Soares 30 ), Danish( Reference Toft, Kristoffersen and Lau 33 ), Spanish( Reference Schröder, Marrugat and Covas 14 ) and Australian( Reference McNaughton, Dunstan and Ball 29 ) populations.
Eight studies had <1000( Reference Tardivo, Nahas-Neto and Nahas 9 , Reference Fung, McCullough and Newby 11 , Reference Asghari, Mirmiran and Rashidkhani 18 , Reference Jayawardena, Byrne and Soares 30 – Reference Azadbakht and Esmaillzadeh 32 , Reference Kourlaba, Polychronopoulos and Zampelas 34 , Reference Wolongevicz, Zhu and Pencina 36 ); fifteen had between 1000 and 10 000( Reference Tande, Magel and Strand 12 , Reference Schröder, Marrugat and Covas 14 , Reference Gregory, McCullough and Ramirez-Zea 23 – Reference McNaughton, Dunstan and Ball 29 , Reference Gao, Beresford and Frank 35 , Reference Zamora, Gordon-Larsen and Jacobs 37 – Reference Drewnowski, Henderson and Driscoll 42 ) and six studies had>10 000( Reference Tande, Magel and Strand 12 , Reference Guo, Warden and Paeratakul 13 , Reference Nicklas, O’Neil and Fulgoni 17 , Reference McCullough, Feskanich and Stampfer 20 , Reference Toft, Kristoffersen and Lau 33 , Reference Cade, Upmeier and Calvert 43 ) participants. FFQ (n 19) and 24-h recalls (n 9) were the most common dietary methods; however, some studies used multiple dietary methods. A total of seven studies investigated BMI, not as the main outcome, rather, as the baseline characteristic in an unadjusted or in age- and sex-adjusted models( Reference McCullough, Feskanich and Rimm 15 , Reference McCullough, Feskanich and Stampfer 16 , Reference de Koning, Chiuve and Fung 19 , Reference McCullough, Feskanich and Stampfer 20 , Reference Jayawardena, Byrne and Soares 30 , Reference Gao, Beresford and Frank 35 , Reference Cade, Upmeier and Calvert 43 ).
The relation of the Healthy Eating Index and its modified versions with obesity-related outcomes
Associations of weight status with the original HEI were examined in eight cross-sectional studies, seven of which reported a significant relation( Reference Drewnowski, Fiddler and Dauchet 10 – Reference McCullough, Feskanich and Stampfer 16 , Reference Asghari, Mirmiran and Rashidkhani 18 ); two studies found no correlation in women( Reference Tardivo, Nahas-Neto and Nahas 9 , Reference Drewnowski, Fiddler and Dauchet 10 ), whereas one of them found a correlation in men( Reference Drewnowski, Fiddler and Dauchet 10 ). Other versions of the HEI include AHEI, HEI-2005, HEI-2010 and HEI-05, which were assessed in six cross-sectional and two prospective studies. Significant inverse associations with BMI and WC was observed in all cross-sectional studies( Reference Fung, McCullough and Newby 11 , Reference Nicklas, O’Neil and Fulgoni 17 , Reference de Koning, Chiuve and Fung 19 – Reference Sundararajan, Campbell and Choi 22 ) and one prospective study( Reference Gao, Beresford and Frank 35 ); however, one study of HEI-2005 documented no significant association( Reference Asghari, Mirmiran and Rashidkhani 18 ).
The average mean score of the original HEI was 63 in French( Reference Drewnowski, Fiddler and Dauchet 10 ), 67·8 in Spanish( Reference Schröder, Marrugat and Covas 14 ), 64·9 in Iranian( Reference Mirmiran, Azadbakht and Azizi 44 ), 56·6 in Brazilian( Reference Tardivo, Nahas-Neto and Nahas 9 ) and 63·8 in American( Reference Kennedy, Ohls and Carlson 6 ) studies.
All studies, whether conducted on population-based subjects or on specific groups such as nurses or health professionals, showed an inverse relation between HEI and its modified versions and obesity status. It needs to be mentioned that in non-significant association studies( Reference Tardivo, Nahas-Neto and Nahas 9 , Reference Drewnowski, Fiddler and Dauchet 10 , Reference Asghari, Mirmiran and Rashidkhani 18 ), the 24-h recall was used for dietary assessment, indicating that one or two 24-h recalls cannot extract the dietary patterns and usual food intakes. The sample populations of the above-mentioned studies were small, and the studies were conducted outside the USA – namely, in Brazil, Iran and France.
Studies in males and females revealed different correlations. Tande et al.( Reference Tande, Magel and Strand 12 ) showed that the correlation between the HEI score and abdominal obesity was stronger in men than in women (1·4 v. 0·8 % risk reduction), and Drewnowski et al.( Reference Drewnowski, Fiddler and Dauchet 10 ) found a significant correlation between higher the HEI score and lower BMI (β-coefficient=–0·08) in males but not in females. Moreover, Drenowatz et al.( Reference Drenowatz, Shook and Hand 21 ) revealed that a higher HEI-2010 score was correlated with a decreased risk for general obesity and overweight among men but not among women. In addition, Guo et al.( Reference Guo, Warden and Paeratakul 13 ) showed that, in men, lower scores of HEI increased the risk for general obesity and overweight by 90 and 50 %, respectively; however, in women, lower scores of HEI increased the risk for general obesity by 70 % but not of overweight. A general overview may imply that the HEI score showed a better correlation with general obesity or abdominal obesity in males and a weak correlation in females.
The relation of the Diet Quality Index and its modified versions with obesity-related outcomes
Associations of weight status with the original DQI, Diet Quality Index-International (DQI-I), and the Diet Quality Index Revised (DQI-R) were observed in four( Reference Sundararajan, Campbell and Choi 22 , Reference Quatromoni, Pencina and Cobain 24 , Reference Zamora, Gordon-Larsen and Jacobs 37 , Reference Funtikova, Baena-Diez and Koebnick 41 ), three( Reference Asghari, Mirmiran and Rashidkhani 18 , Reference Gregory, McCullough and Ramirez-Zea 23 , Reference Lassale, Fezeu and Andreeva 40 ) and one( Reference Fung, McCullough and Newby 11 ) study, respectively; of these, three had prospective( Reference Zamora, Gordon-Larsen and Jacobs 37 , Reference Lassale, Fezeu and Andreeva 40 , Reference Funtikova, Baena-Diez and Koebnick 41 ) and three had cross-sectional designs( Reference Fung, McCullough and Newby 11 , Reference Sundararajan, Campbell and Choi 22 , Reference Gregory, McCullough and Ramirez-Zea 23 ), whereas two had both cross-sectional and longitudinal designs( Reference Asghari, Mirmiran and Rashidkhani 18 , Reference Quatromoni, Pencina and Cobain 24 ). Of three studies investigating the original DQI, one study had significant findings on WC but not on BMI( Reference Funtikova, Baena-Diez and Koebnick 41 ); the second revealed ethnicity-specific findings, with an inverse association among whites and a direct one among blacks( Reference Zamora, Gordon-Larsen and Jacobs 37 ); and the third study, with a cross-sectional design, found correlations in both sexes. However, no significant prediction was found in its prospective design for men( Reference Quatromoni, Pencina and Cobain 24 ). Of the studies conducted on DQI-I, conflicting results were found for both sexes: Gregory et al.( Reference Gregory, McCullough and Ramirez-Zea 23 ) found a positive correlation, whereas one study found an inverse prediction in men and no prediction in women( Reference Lassale, Fezeu and Andreeva 40 ) and Asghari et al.( Reference Asghari, Mirmiran and Rashidkhani 18 ) observed no significant associations. In the only study on DQI-R that revealed a significant inverse correlation( Reference Fung, McCullough and Newby 11 ), mean scores of the DQI-I were 63, 67 and 59 in Iran, Guatemala and USA, respectively.
The sample size was relatively low, ranging from 467 to 4913, compared with studies on HEI. The DQI was evaluated in various countries including France( Reference Lassale, Fezeu and Andreeva 40 ), USA( Reference Fung, McCullough and Newby 11 , Reference Quatromoni, Pencina and Cobain 24 , Reference Zamora, Gordon-Larsen and Jacobs 37 ), Iran( Reference Asghari, Mirmiran and Rashidkhani 18 ), Guatemala( Reference Gregory, McCullough and Ramirez-Zea 23 ) and Spain( Reference Funtikova, Baena-Diez and Koebnick 41 ). FFQ and 24-h recalls were the most common dietary methods( Reference Asghari, Mirmiran and Rashidkhani 18 , Reference Gregory, McCullough and Ramirez-Zea 23 , Reference Quatromoni, Pencina and Cobain 24 , Reference Lassale, Fezeu and Andreeva 40 , Reference Funtikova, Baena-Diez and Koebnick 41 ). Four studies reported BMI( Reference Fung, McCullough and Newby 11 , Reference Asghari, Mirmiran and Rashidkhani 18 , Reference Gregory, McCullough and Ramirez-Zea 23 , Reference Funtikova, Baena-Diez and Koebnick 41 ) and two studies measured weight gain during follow-up( Reference Zamora, Gordon-Larsen and Jacobs 37 , Reference Lassale, Fezeu and Andreeva 40 ), and in one study, the combination of weight gain and BMI was calculated( Reference Quatromoni, Pencina and Cobain 24 ).
Funtikova et al.( Reference Funtikova, Baena-Diez and Koebnick 41 ) found that after adjustment for confounders, a ten-point increment in the DQI predicted a 3·2 cm reduction in WC, but there was no significant prediction of BMI on the basis of DQI over a 10-year follow-up period among Spanish men and women. Zamora et al.( Reference Zamora, Gordon-Larsen and Jacobs 37 ) observed that a ten-point increment in the DQI score predicted 15 % more weight gain in blacks (particularly obese ones), and 10 % less weight gain in whites, after adjustment for confounders. In addition, Quatromoni et al.( Reference Quatromoni, Pencina and Cobain 24 ) found an inverse, linear association between better adherence to DQI and lower weight gain over an 8-year follow up in the Framingham Offspring Study. Fung et al.( Reference Fung, McCullough and Newby 11 ) showed that higher adherence to the DQI-R was inversely correlated with BMI in the Nurses’ Health Study. Lassale et al.( Reference Lassale, Fezeu and Andreeva 40 ) observed a 32 % increment in the risk for general obesity after 13 years of follow up in French men with higher adherence to the DQI; these predictions were not statistically significant in women. In a cross-sectional study from Guatemala, DQI-I was positively correlated with BMI and WC in both men and women( Reference Gregory, McCullough and Ramirez-Zea 23 ). Asghari et al.( Reference Asghari, Mirmiran and Rashidkhani 18 ) reported no significant association in the Tehran Lipid and Glucose Study for both longitudinal and cross-sectional analyses. Overall, studies regarding DQI or its modified versions indicate controversial and sex-specific findings.
The relation of variety scores with obesity-related outcomes
Among four cross-sectional studies, all in developing countries (two in Iran( Reference Azadbakht, Mirmiran and Esmaillzadeh 31 , Reference Azadbakht and Esmaillzadeh 32 ), one in Sri Lanka( Reference Jayawardena, Byrne and Soares 30 ) and one in Guatemala( Reference Gregory, McCullough and Ramirez-Zea 23 )), DDS had the highest frequency (n 3). In the Tehran Lipid and Glucose Study, the highest quartile of DDS, compared with the lowest, increased the risk for general obesity by 39 %( Reference Azadbakht, Mirmiran and Esmaillzadeh 31 ); in contrast, healthy females with higher DDS had an approximately 80 % decrement in the risk for general obesity and abdominal obesity( Reference Azadbakht and Esmaillzadeh 32 ). In Sri Lanka, using a 24-h dietary recall, it was found that BMI and WC in the lowest category compared with the highest category of DDS were 22·16 v. 23·82 kg/m2 and 77 v. 80 cm, respectively( Reference Jayawardena, Byrne and Soares 30 ). Regarding FVS, Jayawardena et al.( Reference Jayawardena, Byrne and Soares 30 ) observed a positive correlation and Gregory et al.( Reference Gregory, McCullough and Ramirez-Zea 23 ) reported a non-significant correlation with general obesity risk. Also, the Dietary Diversity Score with Portions score was positively correlated with general obesity( Reference Jayawardena, Byrne and Soares 30 ). Considering different effects of higher scores of food variety on general obesity, two pathways were determined: First, greater food variety is associated with higher energy intake, and therefore may be associated with general obesity; second, greater variety can be increased by consumption of a diversity of healthy and low-energy-dense food groups (vegetables, whole grains and fruits), with a simultaneous decrease in the risk for general obesity.
The relation of the Dietary Guidelines for Americans Index with obesity-related outcomes
Only one prospective and four cross-sectional studies provided data on adherence to DGAI scores and obesity( Reference Fogli-Cawley, Dwyer and Saltzman 25 – Reference Fogli-Cawley, Dwyer and Saltzman 28 , Reference Lassale, Fezeu and Andreeva 40 ); of these five studies, three studies conducted in US populations indicated that a higher DGAI score could imply a 50 % reduction in enlarged WC and two units of BMI( Reference Fogli-Cawley, Dwyer and Saltzman 25 , Reference Fogli-Cawley, Dwyer and Saltzman 27 , Reference Fogli-Cawley, Dwyer and Saltzman 28 ). Hosseini-Esfahani et al.( Reference Hosseini-Esfahani, Jessri and Mirmiran 26 ) reported no correlation between enlarged WC and BMI and DGAI in the Tehran Lipid and Glucose Study population. However, in a prospective study conducted by Lassale et al.( Reference Lassale, Fezeu and Andreeva 40 ) on French participants, higher scores of DGAI after adjustment for confounding variables predicted a 40 % decreased risk for general obesity in men, but not in women, after 13 years of follow-up. It seems that DGAI has a good correlation with central and abdominal obesity.
The relation of the Dietary Guideline Index with obesity-related outcomes
In Australian populations, two prospective and cross-sectional studies provided evidence for DGI scores and anthropometric status( Reference McNaughton, Dunstan and Ball 29 , Reference Arabshahi, van der Pols and Williams 39 ). In a cross-sectional study by McNaughton et al.( Reference McNaughton, Dunstan and Ball 29 ), a significant decrease in the risk for enlarged WC with increasing compliance with DGI, and insignificant findings for women, was reported. Also, in the Nambour Skin Cancer Study, men in the highest quartile had the lowest gain in BMI (0·05 v. 0·11 kg/m2 per year), and those in the third quartile had the smallest increase in WC (0·04 v. 0·26 cm/year) during a 15-year follow-up. However, in women, the DGI score was not associated with change in any of the anthropometric measures( Reference Arabshahi, van der Pols and Williams 39 ). In addition, a study of Australian women indicated no relation of dietary scores with general obesity, probably because overweight individuals adhered to healthier diets to manage their weight and dieting issues were more popular in women than in men( Reference Arabshahi, van der Pols and Williams 39 ).
Other dietary indices with obesity-related outcomes
Two longitudinal studies investigated the FNRS and the risk for general obesity after a 16-year follow-up in the Framingham study offspring and spouses and found a positive association( Reference Wolongevicz, Zhu and Pencina 36 , Reference Kimokoti, Newby and Gona 38 ). Wolongevicz et al.( Reference Wolongevicz, Zhu and Pencina 36 ) found that weight gain was about 1·4 kg for those with higher quality diets compared with 2·3–3·6 kg for those with poorer quality diets. In addition, Kimokoti et al.( Reference Kimokoti, Newby and Gona 38 ) showed that women with a lower diet quality gained an additional 5·2 kg, compared with those with higher diet quality( Reference Kimokoti, Newby and Gona 38 ).
All of the studies on RFS used FFQ for dietary intakes collection( Reference Fung, McCullough and Newby 11 , Reference de Koning, Chiuve and Fung 19 , Reference McCullough, Feskanich and Stampfer 20 , Reference Gregory, McCullough and Ramirez-Zea 23 ); two had significant findings( Reference de Koning, Chiuve and Fung 19 , Reference McCullough, Feskanich and Stampfer 20 ), whereas others had unrelated findings( Reference Fung, McCullough and Newby 11 , Reference Gregory, McCullough and Ramirez-Zea 23 ). Three studies were conducted in USA and one in Guatemala( Reference Gregory, McCullough and Ramirez-Zea 23 ). McCullough et al.( Reference McCullough, Feskanich and Stampfer 20 ) indicated that a higher RFS score was correlated with a 0·2 unit decrement and a 0·3 unit increment of BMI in men and women, respectively.
Toft et al.( Reference Toft, Kristoffersen and Lau 33 ), in a Danish population, found no correlation of DQS with general obesity. A cross-sectional study from the Greek islands, conducted by Kourlaba et al.( Reference Kourlaba, Polychronopoulos and Zampelas 34 ) showed that a higher EDI was correlated with a 60 % lower risk for general obesity in the elderly. A study conducted on 1220 Guatemalan young adults by Gregory et al.( Reference Gregory, McCullough and Ramirez-Zea 23 ) showed no significant prediction of NRFS with BMI and abdominal obesity. Other dietary indices have been investigated on sporadic studies. The PNNS-GS predicted lower risk for general obesity after a 13-year follow-up in men, but not in women( Reference Lassale, Fezeu and Andreeva 40 ).
Discussion
The current systematic review provides a comprehensive summary of the diet quality indices that have been developed to assess the overall healthfulness of dietary intakes and overweight, obesity, and weight gain in adults. Among thirteen studies on HEI, ten had inverse and three had no associations; however, of seven studies on DQI, only two reported inverse associations and others had conflicting associations based on race, sex and design of the study. Furthermore, studies on variety scores mostly had significant positive associations with general obesity. Of five studies providing data on adherence to DGAI, three were inversely associated with a lower risk for general obesity. Diet quality assessed by two population-specified indices in Australia and Framingham, DGI and FNRS, had consistent inverse associations with general obesity in a sex-specified manner.
The use of diet quality indices is increasing worldwide in various populations. Several diet quality scores reflect a common dietary pattern characterised by high intakes of plant-based foods such as whole grains, and moderate intakes of alcohol, and low intakes of red and processed meat, Na, sweetened beverages and trans-fatty acids.
An in-depth investigation of diet quality studies indicated that, in most cases, subjects who adhered to diet quality indices had favourable health behaviours associated with being older, married, higher education levels, higher physical activity levels and lower smoking( Reference Zamora, Gordon-Larsen and Jacobs 37 , Reference Parveen, Ali and Ali 45 – Reference Zamora, Gordon-Larsen and He 47 ).
Controversial findings on both men and women may be explained by some factors. It is particularly challenging for men to adhere to the cholesterol and Na recommendation because they consume more total energy content( Reference Morreale and Schwartz 48 ); despite this apparent sex difference in diet quality, menopausal women tend to gain more weight over time than men, resulting from the potentially confounding effect of hormonal changes( Reference Lassale, Fezeu and Andreeva 40 ). Despite the fact that people can meet the recommendations for fruits by eating either fresh–raw or processed fruits, the differences in nutritional quality and glycaemic effects can be huge. Hence, people may achieve higher scores by choosing many different food options.
Several studies have found dietary diversity to be directly associated with overweight and general obesity at the individual level( Reference Jayawardena, Byrne and Soares 30 , Reference Azadbakht, Mirmiran and Esmaillzadeh 31 , Reference Toft, Kristoffersen and Lau 33 ). Dietary variety and diversity may reflect consumption of both high- and low-quality foods( Reference Azadbakht, Mirmiran and Esmaillzadeh 31 ). Data shows dietary diversity to be directly associated with energy availability( Reference Jayawardena, Byrne and Soares 30 , Reference Azadbakht, Mirmiran and Esmaillzadeh 31 ).
One of the explanations for the conflicting findings on the association between general obesity and diet quality indices is that the range of higher categories, compared with lower categories was relatively narrow, indicating that subjects in the higher and lower categories could not clearly distinguish differences in dietary patterns, and could attenuate the association of diet with obesity. In addition, some of the scores do not show the extent to which a person deviates from the recommended values because persons consuming over the recommended amounts for food groups receive full points. Furthermore, diet scores assessed the adherence to dietary guidelines (e.g. HEI, DQI and DGAI), most of which were designed for US populations; however, populations may be incapable of assessing overall diet quality until more is known about patterns of consumption. The major issues in developing countries are both under- and over-nutrition. Measures of diet quality in developing countries are more complex to design and interpret; it is also complex in such countries to assess diet quality in terms of both micronutrient adequacy and prevention of overweight, indicating the need for better measures of diet quality specifically for these populations. In addition, there are differences in the scoring models of indices, based on dietary guidelines, for example, the HEI-2005 was designed to differentiate diet quality from diet quantity; however, the original HEI considered only quantity( Reference Kennedy, Ohls and Carlson 6 , Reference Guenther, Reedy and Krebs-Smith 49 ). Another point to consider is the critical role of energy intakes, which determines what kinds of dietary indices lead to loss or gain of weight. Adherence to dietary guidelines has a favourable effect on weight status when the proportion of energy requirements of individuals is taken into account. It is important that the energy density (e.g. intakes per 4184 kJ (1000 kcal)) be applied in the scoring systems. Moreover, in nutritional epidemiology, by considering energy intake as a confounder in the multivariable-adjusted models, the role of nutrients or foods is determined independent of body size and physical activity in terms of energy intake. However, studies evaluated the association of dietary indices with obesity both before and after energy adjustment. Interestingly, findings revealed that energy-adjusted-OR or energy-adjusted means were not markedly different from unadjusted OR or unadjusted-means in baseline models. Indeed, adjustment of energy did not change the results of the studies. Conflicting results may also be due to the fact that overweight individuals adopt a healthier diet to manage their weight, and the effect of a healthy diet as assessed by scores on their obesity status hence could not be detected. In addition, the population that the index is developed for is important; for example, FNRS and DGI were developed for the Framingham and Australian populations, respectively; hence, it is clear that specific indices can be evaluated only in specific populations.
There was no single study investigating the differences of abdominal obesity definitions in relation to the dietary indices; therefore, we were unable to determine the effect/impact of different definitions of abdominal obesity on results.
Nutritional behaviour and food choice, as well as economic and cultural factors and ethnicity, play major roles( Reference Schröder, Marrugat and Covas 14 , Reference Gao, Beresford and Frank 35 ). In a Spanish population, diet cost increases with higher adherence to the HEI and higher scores of the HEI were inversely associated with general obesity; subjects who strongly adhere to HEI had to pay 42€ (52·5$) more per month than those with low adherence to this dietary score( Reference Schröder, Marrugat and Covas 14 ). Ethnicity, as a culturally relevant factor, is proposed to influence the HEI adherence and its relation to general obesity( Reference Gao, Beresford and Frank 35 ). Adherence to HEI scores for white populations was better than the other ethnicities. HEI predictability for outcome of obesity was different among ethnicities, efficient in whites, only fair in Chinese and Hispanic, and poor in African-Americans. Lack of understanding of nutrition guidelines and misconceptions about ‘good’ v. ‘bad’ foods are two of the major obstacles to a healthy diet, particularly in whites and African-Americans. Dietary habits are changing in the Chinese and Hispanics, and they are adopting new dietary behaviours. In addition, people who have migrated recently were less able to adapt their traditional foods to US nutritional guidelines( Reference Gao, Beresford and Frank 35 ).
Some limitations of this review need to be mentioned. First, it is possible that this review did not identify all relevant publications; although using wide search terms, repeating our search in numerous relevant databases, and hand-searching reference lists were attempted to minimise this possibility. Second, there are several factors that may have introduced bias in our findings; specifically, the selection of English-language publications. The third limitation is that twenty-four of the identified articles had a cross-sectional design. These studies cannot show causal effects of adhering to dietary indices on weight status, and only explained an association. The fourth limitation of the literature is that most of the studies were conducted in developed countries. This is important because other populations may differ with respect to weight reduction and acceptability of food items from dietary guidelines.
In conclusion, the review findings suggest that overall diet quality seems to be an important component of the diet–obesity relationship, and also provides potential new insights for use in future research on developing preventive nutrition strategies. Diet quality indices provide important information on updating food guidelines. We found that diet quality indices based on dietary guidelines were inversely associated with parameters of weight status in most studies. However, the difference in scores observed in different populations indicated that future dietary guidelines should be developed and updated to address the dietary needs of different specific population groups. Scoring on the basis of dietary diversity was directly associated with weight gain. Further research using longitudinal studies and field trials to confirm these findings are recommended.
Review flow chart for the association of diet quality indices with obesity.
Summary of findings for diet quality and obesity in prospective studies
DQI, Diet Quality Index; FNRS, Framingham Nutritional Risk Score; F, female; M, male; DGI, Dietary Guideline Index; HRT, hormone replacement therapy; WC, waist circumference; PNNS-GS, Program National Nutrition Sante´-Guideline Score; SU.VI.MAX, SUpplementation en VItamines et Minéraux AntioXydants; DGAI, Dietary Guidelines for Americans Index; DQI-I, Diet Quality Index-International; HEI, Healthy Eating Index; ARFS, Australian Recommended Food Score.
Acknowledgements
The authors would like to acknowledge Ms Niloofar Shiva for critical edition of English grammar and syntax of the manuscript.
This work was funded by a grant from the Research Institute for Endocrine Sciences, Shadid Beheshti University of Medical Sciences, Tehran, Iran.
G. A. and E. Y. Conceptualised and designed the study, and drafted the initial manuscript. F. A. and P. M. supervised the project and revised the final version of the manuscript. All authors read and approved the final manuscript.
None of the authors has any conflicts of interest to declare.
Appendix. Description for diet quality indices
