Participants
The current cross-sectional study was carried out within the framework of the Study on the Epidemiology of Psychological- Alimentary Health and Nutrition (SEPAHAN). The SEPAHAN project was conducted in two phases in a large population of Iranian adults in Isfahan province in the central part of Iran. Details about SEPAHAN have been reported previously (23). SEPAHAN was conducted in two phases. In the first phase, questionnaires were distributed among 10,087 people aged 19–65 years, and 8691 subjects returned the completed questionnaires (response rate 86.16 %). At the second phase, questionnaires with focus on common gastrointestinal, psychological, and personality traits were distributed and 6239 questionnaires were completed (response rate: 64.64%). After merging data from these two phases, complete information was available for 4763 people. We excluded participants with missing data (outcome and covariate variables), women with pregnancy and lactation, and those with under- or over-estimation of energy intake (<800 or >4200 kcal/day). Finally, data on 2818 subjects used in the current analysis. Written informed consent forms were completed by all participants before enrolment. The Bioethics Committee of Isfahan University of Medical Sciences, Isfahan, Iran, approved this study (#189069, #189082, and #189086) (23).
Dietary assessment
Dietary data were collected using a Willett-format (24) Dish-based 106-item Semi-quantitative Food Frequency Questionnaire (DS-FFQ), which was designed and validated specifically for Iranian adults (25). Briefly, the DS-FFQ contained five categories of foods and dishes: (1) mixed dishes (2) carbohydrate-based foods (3) dairy products (4) fruits and vegetables; and (5) miscellaneous food items and beverages (including sweets, fast foods, nuts, desserts and beverages). Common portion sizes and the average reported frequency were used to determine the amount of food consumed. Daily intakes of all food items were computed and converted to g/d using household measures (26). In addition, we calculated daily intakes of 38 nutrients (and bioactive substances) for each participant using the US Department of Agriculture’s (USDA) national nutrient databank (27).
Dietary Diversity Score Estimation
DDS was calculated using Kant et al. method (28). Briefly, DDS consisted of five food groups which divided into twenty-three subgroups: bread-grain (refined bread, biscuits, macaroni, whole grain bread, corn flakes, rice and refined flour), fruit (fruit and fruit juice, berries and citrus), vegetables (vegetables, potato, tomato, other starchy vegetables, legumes, yellow vegetables and green vegetables), meat (red meat, poultry, fish and eggs) and dairy (milk, yoghurt and cheese). Based on scoring criteria, each food group receives a diversity score between 0-2, and total DDS score would range from 0 to 10. To be counted as a 'consumer' for any of the food group categories, a respondent had to consume at least one-half of the serving of any subgroup in a day as defined by the Food Pyramid quantity criteria. For example, in the vegetables group, if a person consumed legumes and green vegetables, his or her score was calculated as (2 ÷ 7) x 2 = 0.57. Therefore, the diversity score of vegetables group would be 0.57.
Assessment of Somatoform Symptoms
There was not a separate questionnaire to examine somatoform symptoms in the SEPAHAN study. However, we settled a validated questionnaire based on 31 items in SEPAHAN's questionnaires common with the 47-items questionnaire used in the Lacourt et al.’s study (29), and “the patient health questionnaire (PHQ)” (30) as valid and standard tools for the assessment of somatoform symptoms. The reliability of this instrument has been evaluated previously (31) and showed a strong internal reliability with a Cronbach's alpha score of 0.903. The 31-items questionnaire was used to examine the frequency of somatoform symptoms. Based on a four points Likert scale (never, sometimes, often, and always) each subject determined the frequency of experiencing each symptom in the past three months. For one item (Dry mouth), the rating scale was as: never, low and high.
Assessment of other variables
Self-administered questionnaires were distributed to collect information on age, sex, weight, height, marital status (married, single), education levels (≤12 yrs, 12-16 yrs, >16 yrs), smoking habits (non-smoker/former smoker/current smoker), disease history (hyperlipidemia, hypertension, diabetes mellitus, asthma, stroke, myocardial infarction, gastrointestinal bleeding, gallstone, cancer, Crohn’s disease and ulcerative colitis), current use of anti-psychothropic medications and dietary supplements. Body mass index (BMI) was computed as weight (kg) divided by height2 (m2). Overweight and obesity were defined as BMI=25-30 and BMI>30, respectively. General Practice Physical Activity Questionnaire (GPPAQ) was used to assess the physical activity status of participants (32). According to GPPAQ, participants were divided into inactive or moderately inactive and moderately active or active categories. Stressful life events questionnaire was used to assess the stressful life events over the past 6 months (33). Total stress score ranged from 0 to 83, with higher scores indicating more sever stressful life events.
Statistical analysis
Exploratory factor analysis using principal component extraction approach was used to identify profiles of somatoform complaints based on the 31 individual somatoform symptoms. The orthogonal varimax rotation procedure was used to find the interpretable factors. Factors were retained for further analysis based on their natural interpretation and eigenvalues on the Scree plot. In this study, we retained factors with eigenvalues > 2 as this cut-off could resulted in more interpretable somatoform complaints profiles and explain sufficient amounts of total variance. Four main somatoform complaints profiles were labelled based on the loaded somatoform symptoms in each factor. The factor score for each profile was computed by summing up items of somatoform complaints weighted by their factor loadings and assigned into each participant (Table 1) (31), and then based on the median value were categorized into high or low score of each factor. We also classified the participants based on tertiles of DDS. Continuous and categorical demographic variables were compared across tertiles of DDS using analysis of variance and Chi-square tests, respectively. Analysis of covariance was used for assessment of age–, sex- and energy-adjusted intakes of foods and nutrients across quartiles of DDS. The association of DDS with being in the higher than median of somatoform complaints profiles scores was assessed using univariate and multivariable logistic regression in the crude and different adjusted models. First, adjustments were done for age, sex and energy intake. In the second model, we further controlled for marital status (single/married), education (≤12 yrs, 12-16 yrs, >16 yrs), smoking status (non-smoker/former smoker/current smoker), physical activity (inactive/ moderately inactive vs. moderately active/active) and BMI (continuous). Ultimately, the effects of stressful life event (continuous), anti-psychotic medicine (yes/no) and medical condition were additionally adjusted in model 3. All these analyses were done using analysis of covariance with Bonferroni correction. To calculate the trend of OR across increasing tertiles of DDS, we considered DDS tertiles as an ordinal variable. In all analyses, the third tertile of the DDS was considered as the reference category and p for linear trends was determined using Mantel-Haenszel extension of chi-square test. Stratified analyses by sex, applying the above-mentioned adjusted models, was run to examine potential modifying effect of sex in relation to the association of DDS and somatoform complaints profiles. All statistical analyses were done using the Statistical Package for Social Sciences (SPSS, version 16.0 for Windows, 2006, SPSS, Inc, Chicago, IL). P value <0.05 was considered statistically significant.