Metabolic syndrome (MetS) is a group of metabolic disorders including insulin resistance, dyslipidemia, abdominal obesity, and high blood pressure (BP), which together increase the risk of coronary heart disease, diabetes, stroke, and other serious health problems.¹ Nutritional recommendations to manage MetS have been proposed over the years, but the increasing prevalence of this syndrome in the past few decades has prompted health experts to reevaluate existing nutritional strategies.² In the past, nutritional recommendations were based on the consumption of one or more nutrients, while current recommendations are mostly based on consumption of whole foods due to food–nutrient interactions. In this regard, food quality indicators have been proposed as tools for predicting MetS.

The diet quality index is an indicator of the diversity of key food groups consumed based on national recommended guidelines.³ The evolution of methods to measure diet quality in recent years has led to the emergence of nearly 30 food quality indicators.⁴ Higher scores on these food quality indicators have been associated with a lower prevalence of MetS in previous studies. In the study of Saraf-Bank et al.⁵, the risk of developing MetS was 28% lower in women who were in the upper quintile of healthy eating indices (HEI). In other studies, higher Dietary Approach to Stop Hypertension (DASH) scores were associated with a lower prevalence of MetS.^6,7 However, these indicators have limitations that cannot be ignored; for example, many of these indicators need a food composition database which may not be available in countries with limited resources.⁸ Furthermore, use of many of these indicators is difficult and time-consuming, and finally, many of these indicators are based on data from developed countries; it is not clear to what extent they accurately predict nutritional adequacy in developing countries.^9,10 All these limitations have led to the development of an overall metric of diet quality called the global diet quality score (GDQS). This index uses a combination of healthy and unhealthy food groups and not only addresses all the above-mentioned limitations but also allows comparison of food quality between communities.¹⁰ Furthermore, GDQS can be used to independently compare the relationships between healthy and unhealthy food groups with MetS.⁹

The effectiveness of the GDQS in tracking and predicting non-communicable diseases (NCDs) related to nutrition has been investigated in only a handful of studies. Fung et al.¹¹ reported that a higher of GDQS was inversely related to the risk of diabetes in women; inverse associations between GDQS and cardiovascular disease or gestational diabetes have also been observed in other studies.^12,13 However, no prior study has investigated the effectiveness of GDQS in predicting MetS or its components. Therefore, our aim in this study was to investigate the relationship between the GDQS with MetS and its components in a group of Tehranian adults.

Materials and methods

The Tehran Lipid and Glucose Study (TLGS) was performed prospectively on district no. 13 Tehran residents to determine risk factors for NCDs.^14,15 Subjects of this cohort study were selected from among TLGS participants using a multistage stratified cluster random sampling technique. The first observation study was performed from 1999 to 2001 on 15,005 subjects aged ≥3 years, and follow-up observations were performed every 3 years; 2002–2005 (surveillance 2), 2005–2008 (surveillance 3), 2008–2011 (surveillance 4), 2012–2015 (surveillance 5), and 2015–2018 (surveillance 6) to identify recently developed diseases or risk factors. For our secondary analysis, among subjects participating in surveys 3 (n=3,665) and 4 (n=7,847; baseline), 8,091 subjects aged ≥18 years were randomly selected to complete a dietary assessment. Subjects with MetS diagnosis at baseline¹⁶ and pregnant or lactating women were excluded. Furthermore, individuals with over- or under-reporting of energy intake (≥4,200 or <800 kcal/day; n=780) were omitted.¹⁷ A total of 4,954 adult women and men with accessible biochemical, anthropometric, and dietary data were selected as the baseline population and they were tracked until survey 6. Subjects who did not provide follow-up data were removed from the analysis; as a result, 4,548 subjects remained and were included in the analysis (Fig. 1). Other independent lines of exclusion were done for the MetS components of abdominal obesity, high BP, low high-density lipoprotein cholesterol (HDL-C), high fasting blood glucose (FBG), and high triglycerides (TG) (Fig. 1).

All participants signed a written informed consent form before taking part in this study. The study was performed according to the tenants of the Declaration of Helsinki. The study proposal was approved by the ethics committee of the Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran (IR.SBMU.ENDOCRINE.REC.1401.122) (Grant no. 43003733).

Dietary assessment

Skilled nutritionists gathered dietary data by applying a valid and reliable semi-quantitative food frequency questionnaire through face-to-face personal interviews. Usual dietary intake was determined based on household portion sizes during the last year.^18-20

The portion amounts of foods consumed were converted from household measurements to grams per day. Due to the incompleteness of and limited data on the nutrient content of cooked food items in the Iranian Food Composition Table (FCT), United States Department of Agriculture (USDA) FCT data were used to determine the nutrient composition of food items (e.g., breads, legumes, nuts, and white or red meat). The Iranian FCT was applied to national foods not present in the USDA FCT.

Globally, the GDQS is a validated index that reflects nutritional adequacy and predicts major NCDs.¹⁰ Scores are based on 3 or 4 categories of eaten amounts (g/day) of specific food groups. It comprises 16 healthy, seven unhealthy, and two unhealthy food groups when consumed in excessive amounts (Supplementary Table 1). The achievable range score is 0–49 points. In the healthy food group, higher scores are given for higher intake, while in the unhealthy group, higher scores are given for lower intakes. In the unhealthy-when-consumed-in-excessive-amounts group, higher scores are given until a certain amount has been eaten, after which no additional score is added. Nutrient contents of food items and food groups were estimated cumulatively during follow-up visits from the first survey to the time of diagnosis of MetS or its components or the last follow-up survey.

Physical activity measurements

A Persian-translated modifiable activity questionnaire was applied to estimate physical activity levels by an expert interviewer.²¹ A previous study reported moderate validity and high reliability of this questionnaire.²¹ Based on routine daily activities, the time, frequency, and intensity of light, middle, high, and challenging activities were recorded during the last year. Activity data were transformed into metabolic equivalent/minutes/week.

Blood pressure and anthropometric measurements

Weight (kg) and height (cm) were determined using a digital scale (accuracy 100 g; Seca 707) with light clothes and no shoes in the standing position and a non-flexible tape measure (accuracy 0.5 cm), respectively. We measured waist circumference (WC) after exhaling without pressure on the surface of the body with light clothing. Our measurements were accurate to 0.1 cm. A standardized mercury sphygmomanometer was applied to determine systolic and diastolic BP (SBP and DBP; mmHg) according to accepted protocols described previously.

Biochemical analysis

Blood samples were collected between 7:00 AM and 9:00 AM after 12–14 hours of overnight fasting. FBG and 2 hours postprandial glucose levels were measured. Blood samples were analyzed on the day of blood collection using an Selectra 2 auto-analyzer at the TLGS research laboratory. FBG concentration was determined using an enzymatic colorimetric method as well as the glucose oxidase technique (Vital Scientific).

At follow-up appointments, all participants who did not take glucose-lowering medications were given 82.5 g glucose monohydrate solution (equivalent to 75 g anhydrous glucose) orally. TG and total cholesterol (TC) concentrations were estimated by an enzymatic colorimetric method using glycerol phosphate oxidase and cholesterol esterase and cholesterol oxidase, respectively (Pars Azmoon Inc.). HDL-C concentration was evaluated after precipitation of apolipoprotein B-containing lipoproteins with phosphotungstic acid. Inter- and intra-assay coefficients of variations of glucose were 2.2%. Inter- and intra-assay coefficients of variations of TG were 1.6% and 0.6%, respectively.

Outcome definition

Subjects with three or more of the following criteria were diagnosed with the MetS phenotype according to the Iranian modified National Cholesterol Education Program/Adult^1,16: (1) abdominal obesity (WC ≥95 cm in men and women); (2) BP ≥130/85 mmHg or antihypertensive drug treatment; (3) HDL-C <1.30 mmol/L (<50 mg/dL) in women and <1.04 mmol/L (<40 mg/dL) in men or receiving drug treatment; (4) FBG ≥6.11 mmol/L (≥110 mg/dL) or drug treatment for hyperglycemia; and (5) TG ≥1.70 mmol/L (≥150 mg/dL) or drug treatment.

Statistical analysis

IBM SPSS software version 26 (IBM Co.), and STATA version 12 (StataCorp.) were used to analyze data. A two-sided P-value <0.05 was recognized as statistically significant. To compare the means and frequencies of participants’ baseline characteristics across quartiles of the GDQS, χ² tests and one-way analysis of variance were used for categorical and continuous variables, respectively. P-values for trends across the GDQS subgroup categories were calculated by designating continuous variables in a linear regression model. Multivariable Cox proportional hazard regression analyses were used to determine hazard ratios (HRs) and 95% confidence intervals (CIs) for MetS and its components’ incidences. There were no interactions between GDQS and age or sex concerning MetS incidence. The first quartile was considered the reference quartile. Confounders were chosen based on the literature^15,22,23 and were age, sex, education level (elementary, diploma, and higher diploma), total energy intake (kcal/day), saturated fat intake (percentage of energy), fiber (g/1,000 kcal), smoking (never smoked, past smoker, or current smoker), physical activity, and body mass index (BMI) at baseline. Moreover, each confounder was entered in the univariable Cox regression model; a two-tailed P<0.20 was applied to determine inclusion in the model.

In models to estimate the HRs of high BP, high TG, high FBG, and low HDL-C, the continuous amount of each risk factor at the baseline of study corresponding to each model was added to the adjustment factors. Furthermore, we applied each confounder in the univariable Cox regression model with a two-tailed P<0.2. The quantitative score of the GDQS was used as a continuous variable to estimate the P for trends in the Cox proportional hazard regression models. The definition of time to event was based on the mid-time between baseline and the event date (for incidence cases) or the time between baseline and the last follow-up (for censored subjects), whichever occurred first.

The proportional hazard assumption was confirmed by the Schoenfeld residuals test and plot of log [–log (survival)] versus log (time) to see if they were parallel.

Four thousand five hundred forty-eight subjects were included in the analysis of the relationship between GDQS and MetS (1,762 men and 2,786 women). Mean ages of men and women were 38.6±14.3 and 35.9±11.8 years, respectively. Among these subjects, there were 1,279 incident cases of MetS (incidence rate, 47.7; 95% CI, 45.1 to 50.4) with a median follow-up of 6.23 years. There were 1,002 (incidence rate, 66.0; 95% CI, 62.0 to 70.2), 1,391 (incidence rate, 44.8; 95% CI, 42.5 to 47.3), 1,373 (incidence rate, 57.5; 95% CI, 54.6 to 60.6), 1,009 (incidence rate, 45.8; 95% CI, 43.1 to 48.6), and 1,385 (incidence rate, 47.4; 95% CI, 44.9 to 49.9) incident cases of low HDL-C, high FBG, high TG, high WC, and high BP, respectively. The mean±standard deviation (SD) of the GDQS for the total population was 29.0±18.5.

In Table 1, we show the baseline attributes of subjects according to quartiles of GDQS. Subjects in the higher quartiles of the GDQS were older than subjects in the lower quartiles. There was a statistically significant association between sex and GDQS quartile. Percentage of current smokers was lower in the upper quartiles of the GDQS than the lower quartiles. In addition, individuals with a higher level of education and physical activity, BMI, and WC were in the higher quartiles of GDQS than the lower quartiles, but there were associations between TG, HDL-C, SBP, DBP, TC, and FBG with GDQS quartiles.

Participants in the higher quartiles of the GDQS had higher energy intake, carbohydrates, total protein, and fiber intake than lower quartiles (Table 2). Moreover, participants in the higher quartiles of the GDQS had lower total fat, poly-unsaturated fatty acid, and mono-unsaturated fatty acid intakes than those in lower quartiles. Individuals with higher total GDQS, healthy, unhealthy, and unhealthy-in-excessive-amount food group scores were in the upper quartiles of the GDQS (in the unhealthy and unhealthy-in-excessive amounts groups, scores are reversed, so a higher score indicates a lower intake of these food groups).

HRs (95% CI) of MetS and its components across quartiles of the GDQS and its subgroups are shown in Table 3. The incidence of MetS was inversely associated with GDQS quartiles (HR, 1.00; 0.90 [95% CI, 0.82 to 0.98]; 0.84 [95% CI, 0.76 to 0.91]; 0.80 [95% CI, 0.73 to 0.89]; P for trend <0.001) after adjusting for confounding variables. After adjusting for confounding variables, the healthy food component subgroup of the GDQS was conversely related to MetS incidence (HR, 1.00; 0.91 [95% CI, 0.83 to 0.99]; 0.87 [95% CI, 0.79 to 0.95]; 0.81 [95% CI, 0.72 to 0.89]; P for trend <0.001). The unhealthy food component of the GDQS showed no relationship with MetS incidence after adjusting for confounding factors.

GDQS was related to high FBG incidence after adjusting for confounding variables (HR, 1.00; 0.91 [95% CI, 0.84 to 0.98]; 0.86 [95% CI, 0.80 to 0.94]; 0.83 [95% CI, 0.76 to 0.92]; P for trend <0.001). Healthy and unhealthy components of the GDQS decreased the incidence of high FBG by 13% and 10%, respectively, in the fourth quartile of the adjusted models.

GDQS was related to low HDL-C incidence after adjusting for confounding variables (HR, 1.00; 0.90 [95% CI, 0.80 to 1.01]; 0.87 [95% CI, 0.77 to 0.97]; 0.88 [95% CI, 0.78 to 1.0]; P for trend=0.02). The unhealthy component of the GDQS was related to the incidence of low HDL-C in the adjusted model.

GDQS was related to abdominal obesity incidence after adjusting for confounding variables (HR, 1.00; 0.96 [95% CI, 0.87 to 1.05]; 0.90 [95% CI, 0.81 to 0.99]; 0.88 [95% CI, 0.79 to 0.98]; P trend <0.001). Healthy and unhealthy components of the GDQS decreased the incidence of high WC by 6% and 15%, respectively, in the fourth quartile of the adjusted models.

After controlling for confounders, a significant reverse association was observed between GDQS and the HR of high BP across quartiles of the GDQS (HR in fourth quartile, 0.83; 95% CI, 0.75 to 0.91; P trend <0.001). This significant finding was also found for the healthy and unhealthy components of the GDQS, indicating that healthy and unhealthy components of the GDQS decreased the incidence of high BP by 15% and 10%, respectively, in the fourth quartile of the adjusted models.

A significant reverse association was observed between GDQS and the HR of high TG across quartiles of the GDQS (HR in the fourth quartile, 0.84; 95% CI, 0.76 to 0.94; P for trend <0.001) after controlling for confounders. Healthy and unhealthy components of the GDQS decreased the incidence of high TG by 9% and 6%, respectively, in the fourth quartile of adjusted models.

Participants in the fourth quartile received the highest scores for consumption of healthy food group foods such as whole grains, other fruits, other vegetables, and citrus fruits while participants in the unhealthy food group received the highest scores from consumption of sugar-sweetened beverages, high fat dairy, and red meat (Supplementary Table 2).

We found an inverse relationship between GDQS and the risk of MetS incidence among Tehranian adults based on 6.23 years of follow-up. This association appears to be due intake of healthy food group items and limiting unhealthy foods. We also found inverse associations between the GDQS and all five MetS components in adjusted models and a decrease in the incidences of MetS components in the GDQS range of 12% to 17%. However, among the MetS components, significant reverse relationships between the healthy food group component of the GDQS and risk of high FBG, central obesity, high BP, and hypertriglyceridemia were found after adjusting for multiple confounding factors. The unhealthy food group component of the GDQS was inversely related to the incidence of all five MetS components.

GDQS is a food group-based score of diet quality that has found to be associated with health outcomes and nutrient quality in diverse regions of the world.¹⁰ The mean GDQS score in our study was higher than that reported in the 2010–2012 China National Nutrition and Health Survey of 19.8 (standard error, 0.03).²⁴ Average scores of other studies are not available; however, a study of non-pregnant, non-lactating women of reproductive age in 10 African countries, as well as China, India, Mexico, and the United States reported that a GDQS ≥23 was associated with a low risk of both NCD risk and nutrient adequacy, while scores ≥15 and <23 were associated with moderate NCD risk and scores <15 were associated with a risk of NCD.¹⁰

GDQS significantly predicted MetS in urban China (odds ratio, 0.58; 95% CI, 0.45 to 0.75) and decreased odds of high WC, hypertension, and low HDL-C in rural China in cross-sectional studies.^10,24 Unlike our study, GDQS was not predictive of MetS in rural or urban Mexico in cross-sectional studies^9,10; however, GDQS was inversely related to BMI, WC, and low-density lipoprotein cholesterol in this population.⁹ In Mexican teachers’ and Nurses’ Health studies, a 1-SD increase in GDQS was related to less weight and WC gains.^25,26 Fung et al.¹¹ reported that a higher GDQS was associated with lower type 2 diabetes mellitus risk, mostly due to a lower dietary intake of unhealthy foods, in the United States women. Several studies that used different indices of diet quality including the DASH diet and dietary guidelines for adherence index (DGA-2005) reduced the risk of MetS in our population²⁷; however, they all consumed more whole grains, fruits and vegetables, legumes, lean protein, and less sugar, red and processed meats, and refined grains in general.²⁷ Like other diet quality indices, the GDQS index takes into account intake of healthy and unhealthy food groups, but in the GDQS, food groups are classified in a more detailed manner to form 25 groups, and nutrients are more extensively examined, so GDQS better predicts cardiovascular risk factors than other diet quality indices.¹⁰ Furthermore, calculating the GDQS is easier than calculating other diet quality metrics because GDQS relies only on food groups, not FCTs.¹⁰

The unhealthy food group component of the GDQS, driven by lower intake of unhealthy food groups and unhealthy-in-excessive amount food groups, contributed significantly to the association between GDQS score and MetS risk factors. Moderate consumption of red meat and full-fat dairy is scored in the GDQS to highlight the value of these foods as sources of protein, iron, and calcium in low-income countries.^9,10,25

The unhealthy food group score displayed minimal variation across quartile groups, whereas the healthy food group score exhibited much larger variation across quartile groups. Additionally, it is worth noting that the effect size of HRs was also larger in the healthy component of the GDQS than in the unhealthy component of the GDQS.

Unlike in our study, the ATTICA study (a population-based health and nutrition survey in Greece) found that healthy food group intake was more important than unhealthy food group intake concerning protection against cardio-metabolic outcomes; this study also reported that lower consumption of unhealthy food groups was not associated with cardio-metabolic outcomes. However, this study was conducted in a population with moderate to high adherence to the Mediterranean diet.²⁸

We found that lower intake of unhealthy food groups was strongly related to a lower risk of MetS components. Previous studies reported that high intake of fast foods, red and processed meats, refined grains (white rice), energy-dense nutrient-poor snacks (both salty and sweet), and sugar-sweetened beverages was directly associated with MetS or its components.^15,27 The unhealthy food group component of the GDQS was inversely associated with MetS, low HDL-C, high WC, and high TG in China, and this component was related to a lower odds of high WC in India.^10,24 The unhealthy food group component of the GDQS was not related to MetS in Mexico in a cross-sectional study.⁹ We found that a higher unhealthy food group component score was associated with a 17% reduced incidence of MetS in the third quartile of this score; however, this association disappeared in the fourth quartile. Overall unhealthy dietary pattern extracted from principal component analysis was associated with MetS in a previous study.¹⁸

In our study, the healthy food group component of the GDQS was inversely associated with MetS, high FBG, high WC, high BP, and high TG. Healthy dietary patterns with consumption of healthy components of the GDQS have been found in previous studies to be inversely associated with MetS and other dietary indices like HEI or DASH diet have been successful at predicting MetS or its components.^5,7

HEI was established based on key recommendations of the 2015–2020 dietary guidelines for Americans, including 13 (nine healthy and four unhealthy) dietary components. DASH score includes eight (five healthy and three unhealthy) dietary components; however, the GDQS includes an extensive series of local food groups which are categorized into healthy and unhealthy food groups. This categorization allows further insight into the association between healthy and unhealthy food groups and outcomes of interest.

Fung et al.¹¹ did not observe an association between the healthy food group components of the GDQS with diabetes, whereas in our study, the healthy food group component of the GDQS reduced the incidence of high FBG. This may be due to the higher score of healthy components in our population than in Fung et al.’ study.¹¹

Strengths of this study include its large sample size study and cohort design in a Middle Eastern country. Lifestyle and dietary information were not self-reported. Multiple confounders were adjusted for in the models; however, other possible confounders like tea or coffee consumption were not adjusted for. We applied the GDQS, which has been validated and used in different regions of the world, so the result of our study are comparable with those conducted in different regions around the world with a wide range of nutritional threats and economic resources.

Previous analyses of the associations of the GDQS with cardiovascular risk factors have been performed in Western populations.^24-26 Regional differences in food consumption and the effect of this on the high incidence of cardiovascular risk factors in developing countries warrants further exploration.¹⁵

In conclusion, we found that higher GDQS scores were associated with a lower risk of the incidence of MetS or its components among Tehranian adults. Our results showed that higher intake of healthy food group components and lower consumption of unhealthy food group components of the GDQS predicted a lower risk of MetS and its component factors. This score can be used to improve dietary strategies to reduce the increasing trend of MetS risk factors in the Iranian population.

Supplementary materials can be found online at https://doi.org/10.7570/jomes24001.

The authors declare no conflict of interest.

The authors would like to thank all participants and Tehran Lipid and Glucose Study (TLGS) personnel for their participation.

Study concept and design: FHE and SD; acquisition of data: FHE, SD, and GK; analysis and interpretation of data: FHE, SD, and AI; drafting of the manuscript: FHE, SD, AI, and GK; critical revision of the manuscript: FHE, PM, and FA; statistical analysis: FHE, SD, and AI; administrative, technical, or material support: FHE, SD, and GK; and study supervision: PM and FA.