Study design and subjects
Detailed information about study design, inclusion and exclusion criteria of this hospital-based case-control study have been reported previously [29]. Briefly, among 235 newly diagnosed pathologically confirmed glioma patients, the following cases were not included: 25 cases due to not meeting the inclusion criteria, 30 cases due to having severe form of glioma and disability, 22 cases due to their avoidance to cooperate and 30 cases due to defective medical information. Finally, 128 patients, including 75 men and 53 women aged between 20 and 75, were included. Among outpatients or admitted patients to the orthopedic and reconstructive surgery wards, 256 subjects, including 150 men and 106 women aged between 20 and 75, met our inclusion criteria. Included subjects were younger, leaner and also had higher education compared with excluded volunteers. Finally, we enrolled 128 cases and 256 matched (in the term of age and sex) controls between November 2009 and September 2011 in Tehran, Iran. The hospitals affiliated with Shahid Beheshti University of Medical Sciences were chosen as sampling sites using the convenience sampling method. Patients referring to the Neurosurgery department of the hospital, who had met our inclusion criteria, were regarded as cases. Individuals with pathologically confirmed glioma in the first month following detection (ICD-O-2, morphology codes 9380e9481), who were aged within the range of 20–75 years, were recruited. Controls were chosen from apparently healthy individuals (aged within the range of 20–75) who had been referred to other wards (orthopedic or surgery wards) of the same hospital. All cases and controls completed an informed consent form before data collection initiation. The main project on glioma was first ethically approved by the Iran National Nutrition and Food Technology Research Institute in 2009, before the study (Ethical code: 39414). Then, based on the main dataset, other projects were defined, each with a different exposure. Mostly, each of these projects that were later written based on that dataset has its own approvals. For the current study, the Medical Ethics Committee of Tehran University of Medical Sciences ethically approved the study (2020/06/22, Ethical code, IR.TUMS.MEDICINE.REC.1399.162).
Inclusion and exclusion criteria
Qualified cases for taking part in our project were those who met the following inclusion criteria (1) individuals with pathologically confirmed glioma with a maximum one-month interval after diagnosis of glioma, (2) aged between 20 and 75 years old. Controls were healthy subjects that had the same age and sex as cases.
The following items were regarded as exclusion criteria: (1) being pregnant or lactating, (2) having a history of some disorders including cancer (only in control groups), neurological, gastrointestinal, hepatic, endocrine, immune, kidney and cardiovascular diseases in medical records, (3) being on special diets, which might result in changes in routine dietary intakes, (4) having any history of chemotherapy or radiation therapy, and (5) use of nitrosamine-enhancing drugs (Fig. 1).
Dietary intake assessment
In this study, trained interviewers administered a Block-format-validated 123-item semi-quantitative food frequency questionnaire (FFQ) to evaluate dietary intakes of subjects over the past year [30]. Each participant reported his/her average intakes of different food items (per day, week or month) in a face-to-face interview. Considering the U.S. Department of Agriculture’s food composition database (modified for Iranian foods) [31], daily nutrients and energy intakes were estimated using Nutritionist IV software (First Databank, Hearst Corp., SanBruno, CA, USA). A validation study [17] revealed reasonable estimates of long-term dietary intakes for this questionnaire because good correlations were seen between dietary intakes obtained from this questionnaire and those from the average of 24-h dietary recalls (two recalls in each month of a year) as the gold standard. For example, energy-adjusted correlation coefficients for vitamin C, vitamin E and b-carotene were estimated as 0.65, 0.65, and 0.68, respectively [16, 18].
Calculation of dietary phytochemical index
We estimated DPI using McCarty equation [10]:
$$ (DPI)=\frac{dietary\kern0.34em energy\kern0.34em derived\kern0.34em from\kern0.34em phytochemical- rich\kern0.34em foods\kern0.28em (kcal)}{total\kern0.34em daily\kern0.34em energy\kern0.17em \mathrm{intake}\kern0.17em (kcal)}\mathrm{X}100 $$
The phytochemical-rich foods we considered in the current study were as follows: Whole grains (Sangak and Barbari bread, which are traditional Iranian breads); fruits (red, yellow and orange fruits); vegetables (dark green vegetables, red, orange vegetables, starchy vegetables and other vegetables); soy products (soybean); nuts (peanut, almond, walnut, pistachio and hazelnut); legumes (lentil, beans, chickpea); olives; olive oil; natural fruit and vegetable juices (carrot juice, orange juice, Limon juice). Potato, as a food item in the vegetable group, was not considered in DPI calculation due to its low content of phytochemicals.
Assessment of glioma
Detection of glioma was performed based on the pathological test ICD-O-2 and morphology codes 9380–948 [29]. Glioma patients who had passed a maximum of one month of the disease confirmation were included in our study.
Assessment of other variables
A pretested questionnaire including several variables of sociodemographic status such as age (years), gender (male/female), the status of marriage (married/unmarried), residence place (urban/rural), occupation (farmer/non-farmer), education (university graduated/non-university graduated) and family history of glioma and any other cancer (yes/no), a history of trauma, hypertension and allergy (yes/no), dealing with chemicals during the past ten years (yes/no), methods of cooking (barbecue/microwave/canned foods/fried foods), drug use (yes/no), use of hair dye (yes/no), cell phone use duration (years), exposure to the radiographic x-ray (yes/no); was applied to collect general information of participants. Measurement of participants’ physical activity was done using a short form of the International Physical Activity Questionnaire (IPAQ). Data from IPAQ were stated as Metabolic Equivalent per week (METs/week). Anthropometric measurements were quantified via standard methods. Considering weight and height, body mass index (BMI) was calculated for each participant.
Statistical analysis
All subjects were classified based on the DPI score into tertile ranges. The distribution of study participants in terms of general characteristics across tertiles of DPI was assessed using the Chi-square test. Differences in continuous variables across DPI tertiles were determined using one-way analysis of variance (ANOVA), followed by pairwise post hoc tests with Bonferroni correction. Binary logistic regression was used to evaluate the association of DPI with glioma. Age (continuous), sex (male/female), energy intake (kcal/day), physical activity (continues), family history of cancers (yes/no), family history of glioma (yes/no), marital status (yes/no), education (university graduated/ non-university graduated), high-risk occupation (farmer/non-farmer), high-risk residential area (yes/no), duration of cell phone use (continues), supplement use (yes/no), history of exposure to the radiographic X-ray (yes/no), history of head trauma (yes/no), history of allergy (yes/no), history of hypertension (yes/no), smoking status (smoker/non-smoker), exposure to chemicals (yes/no), drug use (yes/no), personal hair dye (yes/no), frequent fried food intake (yes/no), frequent use of barbecue (yes/no), canned foods and microwave (yes/no), dietary intakes of red and processed meat, fish, tea, coffee, sugar-sweetened beverages, egg, total fat, dietary fiber, cholesterol, calcium, SFA, folate, and selenium were adjusted in the multivariable-adjusted model. The selection of these confounders was made based on previous publications [11, 32,33,34]. The model goodness-of-fit was examined using Hosmer–Lemeshow test. By considering tertiles of DPI as ordinal variables, the overall trend of ORs across increasing tertiles of DPI was examined. All the statistical analyses were performed using SPSS (SPSS Inc., version 19). The significance of P-values was considered at < 0.05.