Empirically derived dietary patterns and obesity among Iranian Adults: Yazd Health Study‐TAMYZ and Shahedieh cohort study

Abstract The aim was to determine the relationship between dietary patterns derived by principal component analysis (PCA) in association with obesity from a large group of Iranian adults in the urban and suburb areas. A cross‐sectional study was conducted on 10,693 Iranian adults. The data were collected from two cohort studies: Shahedieh city annexed to Yazd area as well as Yazd Health Study (YaHS)‐TAMYZ (Yazd Nutrition Survey in Persian) in urban area. Dietary intakes were assessed using a validated semi‐quantitative food frequency questionnaire. The PCA was applied to identify the dietary patterns. Multiple logistic regressions were run to assess the relationship between dietary patterns and obesity. In Shahedieh cohort study, three major dietary patterns were identified traditional, unhealthy, and prudent pattern. Prudent pattern was associated with lower odds of obesity (OR: 0.68; 95% CI: 0.53, 0.88). Higher adherence to the unhealthy (OR: 1.24; 95% CI: 1.02, 1.50) and traditional (OR: 1.38; 95% CI: 1.11, 1.72) patterns was related to greater odds of obesity. Moreover, we identified traditional and unhealthy dietary patterns in YaHS study. Higher adherence to the unhealthy dietary pattern was associated with greater odds of obesity (OR: 1.21 95% CI: 1.02, 1.44). Greater adherence to unhealthy dietary patterns was associated with higher odds of obesity in participants. Greater adherence to traditional and prudent dietary patterns increased and decreased the obesity odds, respectively. Further prospective studies are needed to find out the causal relationship between the variables.

& Ghodrati, 2017; Shu et al., 2015;Slagter et al., 2018;Zhang et al., 2015). However, odds of obesity increased in dietary patterns with greater intakes of red meat, refined grains, and carbohydrates (Huybrechts et al., 2017;Livingstone & McNaughton, 2017;Papavagelis et al., 2018;Shaker-Hosseini & Ghodrati, 2017;Shu et al., 2015;Slagter et al., 2018;Zhang et al., 2015). In this regard, a cross-sectional study investigated the association between traditional dietary pattern and obesity among Chinese adults (Shu et al., 2015). According to the results, traditional dietary pattern, consisting of high amounts of fruits, vegetables, seeds, and tubers was associated with lower odds of obesity (Shu et al., 2015). On the contrary, animal dietary pattern that included high amounts of rice, red meat, and fat was associated with higher odds of obesity (Shu et al., 2015).
According to a research over 52 countries from eight different geographical regions, mean body mass index (BMI) of people from the Middle East countries (27.4) was higher than individuals from Europe (26.7), South America (26.7), Africa (26.7), China (24.4), Australia (27.0), and other parts of Asia (24.9; Yusuf et al., 2005).
However, mean BMI of adults from the Middle Eastern countries (27.4) was lower than North America (27.7;Yusuf et al., 2005).
This study was conducted considering the high prevalence of obesity (Asfaw, 2007) as well as different cultural and socioeconomic levels among the Middle Eastern populations (Monteiro, Moura, Conde, & Popkin, 2004). Our aim was to investigate the association between dietary patterns and obesity in Iran, where the existing various dietary patterns may provide some novel insights into the dietdisease relationships. To the best of our knowledge, the association between principal component analysis (PCA)-derived dietary patterns and obesity was studied in few researches with low sample size in Middle East countries (Jomaa et al., 2016;Mirzababaei et al., 2019;Naja et al., 2011;Rezazadeh & Rashidkhani, 2010;Shaker-Hosseini & Ghodrati, 2017). Moreover, most of these studies were conducted in urban areas, where the culture and dietary habits are different from the suburb regions (Hooper, Calvert, Thompson, Deetlefs, & Burney, 2008;Maruapula & Chapman-Novakofski, 2007). This study targeted at evaluating the relationship between major dietary patterns derived from posteriori methods and obesity among a large sample of Iranian adults living in urban and suburb areas.

| Study design and population
The present cross-sectional study was carried out based on the data collected from two cohort studies (Shahedieh and YaHS). Dietary foods and supplements have been addressed in our substudy, called Yazd Nutrition Survey (YNS), locally known as TAMYZ in Persian.
Detailed information about the design and baseline population of YaHS study was published previously (Mirzaei et al., 2017). This research included a total of 8,966 individuals from the suburb region (Shahedieh) aged 35-70 years and 10,038 people from the urban and rural areas (YaHS study) aged 20-70 years in Iran. Considering the above-mentioned substudies, the exclusion criteria were being on a weight loss or specific diet and having a history of diseases such as diabetes, cardiovascular diseases, stroke, fatty liver, hypertension, cancer, and thyroid, since such diseases may change the participants' diet. Furthermore, participants with a total daily energy intake of less than 800 or higher than 6,500 kcal were excluded.

| Dietary assessment
The semi-quantitative FFQ was administered to assess the dietary foods and supplements. The original semi-quantitative FFQ contains 168 items, but 10 more questions were added on consumption of Yazd-specific frequently consumed food items, which made a total of 178 food items. The semi-quantitative FFQ was previously validated for the Iranian population , so the questionnaire was completed by trained interviewers. Participants were supposed to report the amount and frequency of consuming each food item per month, week, or day in the past year. Moreover, a food photograph book was used for all participants as a reference, so that they could estimate the portion size of foods as a unit accurately.
Participants were also asked to report their intake frequency with regard to all food items based on 10 multiple-choice frequency response categories ranging from "never or less than once a month" to "10 or more times per day." Later, the amount of food consumed at each intake was estimated using questions with five predefined answers.

| Anthropometric assessment
Participants' body weight was measured in standing position with light clothing. All anthropometric indices were measured three times: before the interview, after completing one-third of the questionnaire, and after completing two-thirds of the questions. Participants' height was also measured to the nearest centimeter with barefoot while their heads, shoulder blades, buttocks, and heels were rested against the wall. BMI (kg/m 2 ) was calculated using weight and height measurements according to the following formula: weight (kg)/height squared (m 2 ). Waist circumference was recorded to the nearest 0.5 cm using nonstretch tape placed midway between iliac crest and lowest rib while participants were in the standing position (Edwards, Williams-Roberts, & Sahely, 2008).

| Assessment of covariates
The demographic and medical history questionnaires were also administered, and the related information was collected from all participants: age, gender, marital status, tobacco smoking, SES, and diseases. The SES score was calculated to determine the individuals' SES based on the infrastructure facilities (source of drinking water and sanitation facility), housing condition (e.g., the number of rooms, type of home ownership), durable assets' ownership (e.g., dishwasher, car, television), and education level (Karyani et al., 2019). Later, the total SES score, ranging from 0 to 3, was calculated by summing up the assigned scores; a score of 3 showed high SES. In addition, the Iranian version of International Physical Activity Questionnaire was applied to calculate the participants' physical activity (Moghaddam et al., 2012) and individuals with more than 1 hr of activity per week were considered as physically active.

| Statistical analysis
To determine the major dietary patterns based on the food groups (N = 22), the PCA was applied and the factors were rotated using the varimax rotation. Furthermore, the study factors were naturally interpreted in conjunction with eigenvalues >1.5 and the scree plot was determined. The derived dietary patterns were labeled according to data interpretation and similar studies. To calculate the factor score of each identified pattern, the food group intakes weighted by their factor loadings were summed for each participant. Later, the participants were categorized based on the dietary pattern scores' quartiles (quartile 1: low consumption, quartile 4: high consumption of a given food pattern). Next, the participants' characteristics were measured across quartiles of each dietary pattern and the data were calculated by mean ± standard deviation for continuous variables and percentage for categorical variables. Analysis of variance was run to describe the mean differences of the continuous variables, and the chi-squared test was applied to determine the difference between categorical variables. Multivariable logistic regression analysis was also used to study the association of dietary patterns with obesity in different models. Initially, the confounder variables were adjusted: age, energy intake (kcal/d), gender, smoking status (nonsmoker, ex-smoker, current smoker), SES (weak, moderate, high), marital status (married, single, widowed, divorced), physical activity level (never, <1, >1 hr/week), and diseases. With regard to all analyses, we considered the first quartiles of dietary pattern scores as the reference. The quartile categories were also considered as ordinal variables in the analyses to calculate the overall trend of odds ratios (OR) across increasing quartiles of dietary pattern scores. The IBM SPSS version 20.0 was run to analyze the data, and the significant p value was set at <.05.

| Study population characteristic
In Shahedieh cohort study, 73.3% of the participants were in the age range of 35-49 years and 26.6% were above 50 years. Prevalence of obesity was calculated as 26.7% (men, 10.1%; women, 16.6%). Food groups and their corresponding food items used in PCA to derive dietary patterns are shown in Table 1. The results of factor analysis showed three dietary patterns: "Traditional" (highly loaded by vegetable, red and processed meats, fish, soft drinks, fruits, nuts, pickles, eggs, legumes, dairy, mayonnaise, potatoes, refined grains, snacks, poultry, vegetable oils, sweet and sugars, olive group and tea and coffee), "Unhealthy" (highly loaded by sweet and sugars, tea and coffee, eggs, potatoes, and snack but low intake of vegetables, fruits, olive group, and dairy), and "Prudent" (highly loaded by fruits, vegetables, whole grains, dairy, but low intake of pizza, snacks, soft drinks, refined grains, and vegetable oils). The factor loading matrixes for these dietary patterns are shown in Table 2. These dietary patterns explained 28.1% of the whole variance. The participants' characteristics and dietary intakes across dietary patterns' quartiles are represented in Tables S1 and S3, respectively. Individuals in the highest quartiles of all dietary patterns were significantly more likely to be males and married, but they had a higher intake of energy, red and processed meat, vegetable, fruits, egg, potato, refined grains, mayonnaise, and nuts compared to those in the lowest quartiles.
Participants in the top quartile of traditional pattern were younger and had higher weight and height compared with those in the lowest quartile. Among the people with unhealthy dietary pattern, the highest quartile members were likely to be older and had higher weight, height, and SES but, they had lower physical activity compared with individuals in the lowest quartile. Furthermore, members of the highest quartile of the prudent pattern had a lower prevalence of central obesity, BMI, and SES but they are more likely to be younger compared with those in the lowest quartile.
In YaHS and TAMYZ study, 74.8% of the participants were in the age range of 20-49 years and 25.1% were above 50 years old. Prevalence of obesity was 21.2% (men, 8.6%; women, 12.6%).
Two dietary patterns with 33.7% of the whole variance were recognized: "Traditional" (highly loaded by red and processed meats, vegetables, dairy, fruits, legumes, poultry, whole grains, fish, pickles, refined grains, eggs, and sweet and sugars), and "Unhealthy" (highly loaded by sweet and sugars, condiments, snack, soft drink, nuts, mayonnaise, tea, and coffee).  (p-trend = .005) and after full adjustments (p-trend = .04). Also, this result remained significant for males before (p-trend = .008) and after full adjustments (p-trend = .008). In addition, there was a significant increasing trend in the odds of obesity across increasing quartiles of the unhealthy dietary pattern among whole population after adjustment for age and energy intake (p-trend = .01). This result remained significant among women (p-trend = .001). However, there was a significant decreasing trend in the odds of obesity across increasing quartiles of the prudent dietary pattern among whole population before (p-trend ˂ .001) and after full adjustments (p-trend = .001).

| Dietary patterns and obesity in Shahedieh study (suburb area)
This result remained significant for men before (p-trend ˂ .001) and after full adjustments (p-trend ˂ .001). Model II: in addition to age and total energy intake additionally adjusted for age; gender; smoking status; socioeconomic status; marital status; physical activity level; diseases and total energy intake. Statistically significant p-values (p ˂ .05) should be indicated in bold.

| D ISCUSS I ON
According to the results, unhealthy dietary pattern increased the odds of obesity in residents of the suburb and urban areas. In addition, traditional dietary pattern increased the odds of obesity while the prudent dietary pattern was inversely related to obesity in suburb region. In men with greater adherence to the traditional dietary pattern, the odds of obesity increased. However, the odds of obesity were lower for higher adherence to the prudent dietary pattern in the suburb region. Moreover, the odds of obesity were higher in women with higher adherence to unhealthy pattern. In YaHS and TAMYZ studies, odds of obesity were lower in women with higher adherence to traditional pattern.
To the best of our knowledge, this was the first study conducted Unhealthy patterns contain high amounts of sweet and sugar and provide carbohydrate energy, so they differ from energy provided by fat (Hill & Prentice, 1995). High intake of carbohydrates leads to weight gain caused by consumption of high concentration of dietary fat with sugar, which increases the odds of obesity (Hill & Prentice, 1995). In unhealthy dietary pattern, consumption of snacks contributes to high caloric intake of saturated fat, cholesterol, and salt (Heald, 1992).
Greater adherence to prudent dietary pattern was associated with lower odds of obesity in the participants of suburb region. Prudent/ healthy dietary patterns decreased the odds of obesity in a study over a small sample of Iranian women Mirzababaei et al., 2019;Rezazadeh & Rashidkhani, 2010).
Studies from Canada, and Mexico, Argentina, and Mexico reported similar findings. These results were also confirmed in a meta-analysis (Rezagholizadeh et al., 2017).
The odds of obesity are reduced in the prudent dietary pattern, since it is rich in fruits and vegetables, which contain fiber (Burton-Freeman, 2000;Glore, Van Treeck, Knehans, & Guild, 1994;Zank & Kemp, 2012). In addition, chewing fiber requires more time than normal foods (Heaton, 1973) and it absorbs more water, creates a viscous gel that increases stomach distention (Howarth, Saltzman, & Roberts, 2001), provides low energy, and slows down the gastric emptying (Ruhee & Suzuki, 2018). Fiber also creates a feeling TA B L E 5 Odds ratio (95% CI) for obesity according to quartiles (Q) of dietary pattern in a sample of Iranian adults (n = 6,750); and also stratified by gender in YaHS cohort study (urban area) 1 Model II: in addition to age and total energy intake additionally adjusted for age; gender; smoking status; socioeconomic status; marital status; physical activity level; diseases and total energy intake.
Adherence to the traditional dietary pattern increased the odds of obesity. Our findings are consistent with the results of previous data collected from 141 Iranian adults. (Sherafat-Kazemzadeh et al., 2010). This result was also in the same line with the research carried out among Chinese adults (Shu et al., 2015;Yu et al., 2015;Zhang et al., 2015), young Japanese women (Okubo et al., 2008), and the Mexican American population (Carrera, Gao, & Tucker, 2007).
Conversely, in a case-control study including 147 Iranian adults, participants in the highest quartile of traditional dietary pattern had significantly lower odds of obesity (Yosaee et al., 2016). This result was also confirmed by a study on old (Xu, Byles, Shi, McElduff, & Hall, 2016) and young  Chinese population. In the same vein, a cross-sectional study over 486 women showed no significant association between Iranian dietary pattern (high in refined grains, potato, tea, whole grains, hydrogenated fats, legumes, and broth) and general obesity .
Our results showed that traditional dietary pattern increased the obesity prevalence in Shahedieh (suburb) study, but not in YaHS (85% urban /15% suburb area) study. Considering that Shahedieh is a suburb region, its residents have different dietary quality and lifestyle than the urban region. cities. In addition, the researchers had to make several subjective decisions to conduct this study; for example, categorizing food items into food groups, extracting the number of factors, applying the rotation method, and labeling the factors. Moreover, we cannot reject the possibility of residual confounding bias, since unknown or unmeasured confounders may exist that affected our results. Finally, our participants with odds of obesity might have been advised to reduce their fat intake, which led them to alter their dietary habits.

| Strengths and limitations
However, such possibility cannot be resolved in a cross-sectional study.

| CON CLUS ION
Greater adherence to unhealthy dietary patterns was associated with higher odds of obesity in residents of suburb and urban areas in Yazd Greater Area-Iran. In suburb area, greater adherence to traditional dietary pattern was associated with higher odds of obesity, while adherence to prudent dietary pattern was associated with lower odds of obesity. In order to reflect on the causal relationship between the studies variables, further prospective studies are needed.

ACK N OWLED G M ENTS
The authors would like to thank all those who have helped in carrying out the research. This study was extracted from a MSc dissertation that was approved by the School of Health Shahid Sadoughi University of Medical Sciences, Yazd, Iran.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest to report regarding this study.

AUTH O R CO NTR I B UTI O N
SS-KH and MH made substantial contributions to the conception and design of the manuscript, preparation manuscript, as well as performing statistical analysis and data interpretation. They also approved the final manuscript for submission and critical revision.
HM-KH, MM, and AN contributed to data interpretation and also critically revised the manuscript for important intellectual content and approved the final manuscript for submission.

E TH I C A L A PPROVA L
The study's protocols and procedures were ethically reviewed and approved by a recognized ethical body (Ethics Committee of Shahid Sadoughi University of Medical Science with ethics code of (IR.SSU. SPH.REC.1397.123)). This study does not involve any human or animal testing. Also, this study conforms to the Declaration of Helsinki, US, and/or European Medicines Agency Guidelines for human subjects.

I N FO R M E D CO N S E NT
Written informed consent was obtained from all study participants.