Participants for this study were recruited from the Tehran Lipid and Glucose Study (TLGS), a large-scale population-based cohort study performed to determine risk factors for non-communicable diseases in a representative sample of residents of Tehran. At first phase of the study (1999-2001), 15005 individuals aged ≥3 years were selected using multistage stratified cluster random sampling and follow-up examinations were conducted in five consecutive phases: Phase 2 (2002-2005), Phase 3 (2005-2008), Phase 4 (2008-2011), Phase 5 (2012-2015) and Phase 6 (2015-2018). The details of the study have been published elsewhere (15). Of 8843individuals aged ≥18 years who participated in Phase 4,a total of 6791 subjects (3016 men) completed the dietary assessment. We selected these subjects as baseline population in this study and followed them at next Phases (Phases 5 and 6). We compared characteristics of adult participants who had dietary data (respondents, n=6791) and those who did not have (non-respondents, n=2052). Among respondents, 44.4% were male, 22.6% were current smoker and 5.5% had family history of CVD (FH-CVD),compared with 44.8%, 20.6% and 4.7%, respectively, in non-respondents (P>0.05). The mean (SD) of age and body mass index (BMI) were 40.8 (14.1) and 27.3 (4.9), respectively, in respondents vs. 44.8 (17.1) and 27.7 (5.2) in non-respondents (P<0.001). The mean (SD) of SBP and DBP blood pressure among respondents were 114 (16.7) and 75.5 (11.1), respectively, compared with 118 (19.7) and 77.2 (11.5) mmHg in non-respondents (P<0.001). The level of physical activity did not differ in two groups.
Of 6791 participants, we excluded under- or over-reporters of energy intake (<800 or ≥4200 kcal/day, n=457), those with hypertension at baseline (n=1116) and subjects with missing data on hypertension status at baseline (n=19). Finally, after excluding participants without any follow up data (n=406), 4793 subjects (1986 men) were remained and entered in the analysis.
At baseline and next phases, information on age, sex, smoking status, medical history and medication use was obtained through a personal interview using a standardized questionnaire(15).
Anthropometric and blood pressure Measurements
Bodyweight was measured using a calibrated digital scale (Seca 707). Height was measured using a portable stadiometer. A qualified physician measured BP using a standard mercury sphygmomanometer after the participants remained seated for 15 minutes. The calibration of sphygmomanometer was done by Iranian Institute of Standards and Industrial Researches. The cuff was placed on the right arm, which was at the heart level and bloated as high rate as possible increments until the cuff pressure was 30 mmHg, above the level at which the radial pulse disappeared. Two separate measurements were performed with an at least 30-secondinterval time; the mean of two measurements was recorded as the participant's BP. The SBP was determined as the appearance of the first sound [Korotkoff phase 1], and the DBP was determined as the disappearance of the sound [Korotkoff phase 5] during deflating the cuff at a 2–3 mm per second decrement rate(15).
Physical activity Measurements
The physical activity level (PAL) was assessed using the Persian-translated modifiable activity questionnaire (MAQ) with high reliability and relative validity (16).
Dietary intake assessment
Dietary data were collected at baseline through face-to-face personal interviews with the use of a valid and reliable 168-items semi-quantitative food frequency questionnaire (FFQ). Participants used the standard serving sizes to report the usual frequency of consumption of individual food items on a daily, weekly, or monthly basis during the last year. The usual food intakes were then converted to daily intake (in grams) and were calculated in energy-adjusted terms (serving per 1000 kcal/day). Because the Iranian food composition table (FCT) is incomplete, the United States Department of Agriculture (USDA) FCT was used to analysis of food composition(17). Foods listed in the FFQ were collapsed into 20 mutually exclusive food categories based on the similarity of type of food and nutrient composition.
The DASH diet is a dietary pattern originally developed to prevent and control hypertension. DASH score is an index for measuring diet quality(18).We computed a 40-points DASH score which includes 8 dietary components(19).All components were computed per 1000 kcal and were then divided into quintiles. Each quintile intake received 1 point. For fruits, vegetables, whole grains, low-fat dairy, nuts and legumes, a score 5 was given to those in the top quintile. For sodium, red and processed meats, and sweetened beverages, the lowest quintile was given a score of 5, and the top quintile was given a score of one. The overall DASH score was then obtained by adding the component scores rangingfrom 8 to 40. A higher DASH score indicates better adherence.
HEI is based on key recommendations of the 2015-2020dietary guidelines for Americans (DGA). It is comprised of 13 dietary components. Nine adequacy components include total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and fatty acids. Four moderation components (those that should be limited) include refined grains, sodium, added sugars, and saturated fats (20). The HEI scoring is based on density (amount per 1000 kcal, ratio of fatty acids) and recommendations are in the range of 1200-2400 kcal dietary intake. To compute the score of HEI, six components from nine adequacy components (total fruit, whole fruit, total vegetables, greens and beans, total protein foods and seafood and plant proteins) each received a score of 0 and 5 respectively for the lowest and highest consumption. The other three adequacy components (whole grains, dairy and fatty acids) were scored from 0 to 10 for the lowest and highest consumption, respectively.
An individual who had an amount of zero per 1000 kcal receive a minimum score of zero; therefore, densities of exactly half the standard for maximum points would interpret to a score of half the possible points (e.g., 5 out of 10 score).
The four moderation components (refined grains, sodium, added sugars, and saturated fats) received a score of 10 and 0 for the lowest and highest intakes, respectively. Intermediate intakes between the minimum and maximum were prorated. The scores from the 13 components were added for a total HEI score ranging from 0 to 100. Higher total HEI scores indicate greater adherence to DGA recommendations(20).
Blood samples were collected after a 12- to 14-hour overnight fasting. Fasting plasma glucose (FPG) was measured using an enzymatic colorimetric method with glucose oxidase. Triglyceride (TG) concentrations were assayed with glycerol phosphate oxidase using the enzymatic colorimetric method. Analyses were performed using Pars Azmoon kits (Pars Azmoon Inc., Tehran, Iran) and a Selectra 2 auto-analyser (Vital Scientific, Spankeren, theNetherlands). Inter- and intra-assay coefficients of variation(CVs) were both <2.2% for FPG, 0.6% and 1.6% for TG, respectively (15).
Definition of terms and outcome
Smoking status was categorized as smoker (current smokers) versus non-smoker (including past and never-smokers). A current smoker was defined as a person who smokes cigarettes or other smoking implements daily or occasionally. A positive FH-CVD was defined as diagnosis of CVD in a male first degree relative <55 or in a female first degree relative <65 years. Individuals were considered physically active when they achieved a minimum of at least 600 MET (metabolic equivalent task)- minutes per week (21). Type 2 diabetes mellitus (DM) was defined as FPG ≥7 mmol/L or 2 h-PLPG ≥11.1 mmol/L (22) or using glucose-lowering treatment. Hypertension was defined as a SBP ≥140 mmHg or a DBP ≥90 mmHg or taking antihypertensive medications (23).
Missing data among total population (after applying the exclusion criteria) were 1.1, 1.1, 0.1, 2.4 and %9.6 for baseline covariates including smoking status, BMI, TC, DM status and PAL, respectively. Thus, multivariate imputations by chained equations (MICE) (mice package in R software) (24)were used to impute missing values at baseline.The PCA was used as a posteriori method with orthogonal rotation to identify dietary patterns on 20 food groups (as servings per 1000 kcal/day). Eigenvalues>1 derived from the correlation matrix, scree plots, factor interpretability and variance explained >5% were used to extract key dietary patterns. Food groups with absolute factor loadings values >0.2 were considered as contributing highly to the extracted pattern. Each person received a factor score for each dietary pattern by summing intakes of food groups weighted by the loadings generated by the PCA. The posteriori and priori dietary patterns (DASH and HEI) scores were then stratified into quartiles. The baseline characteristics of the study population were compared across quartile categories of each dietary pattern using descriptive analysis. To test linear trend for categorical and continuous variables across quartiles of dietary pattern scores, logistic and linear regression tests were performed respectively; with the use of quartiles of dietary pattern scores as a continuous variable in those models. The incidence density rate of hypertension was calculated by dividing the number of events by the person-years at risk.
The association between different dietary patterns and incidence of hypertension was analyzed using time-dependent Cox proportional hazard (Cox PH) regression. The interaction of dietary patterns and age/sex in relation to hypertension were not significant. All covariates (excluding sex and dietary patterns) were included in the models as time-dependent variables. Missing data on time-dependent variables was imputed by the last observation carried forward (LOCF) approach. For these analyses, the lowest quartiles of the different dietary patterns were considered as the reference category. Time to event was defined as the time between baseline and the event date (for event cases) or the last follow-up (for censored cases), whichever occurred first. The event date was defined as the mid-time between the date of the follow-up visit at which hypertension was detected for the first time, and the most recent follow-up visit prior to the diagnosis. Study participants were censored due to death, loss to follow-up or non-occurrence of hypertension before the end of the follow-up (18th April 2018). Two models were developed; model 1 was adjusted for the age and sex. Model 2 was further adjusted for time dependent BMI, smoking, DM status, PAL, TG, FH-CVD, total energy and salt intakes as the most important confounders. The PH assumption was verified using Schoenfeld residuals test and plot of log [−log (survival)] versus log (time) to see if they are parallel. We conducted tests for linear trends with the use of quartiles of dietary patterns as a continuous variable and modeled this variable in separate Cox PH models. Analyses were conducted with R software (version 3.6.2) and the Statistical Package for Social Sciences (SPSS; version 21.0), and a two sided P values <0.05 were considered statistically significant.