Study design
For the current cross-sectional study, we used data from the sixth wave of the NESDA study, which was collected across the Netherlands between 2013 and 2016. Recruitment for this study took place in the general population, in general practices (through a three-stage screening procedure), and in mental health organizations to recruit persons reflecting various settings and developmental stages of psychopathology. A total of 2069 individuals participated in wave six, of whom 396 were healthy controls, 568 had a current (6-month recency) diagnosis of depression (major depressive disorder or dysthymia) and/or anxiety (panic, social phobia, generalized anxiety disorder, agoraphobia) disorder, and 1105 individuals were at risk due to a (family) history of depressive and/or anxiety symptoms. Further details of this study can be found elsewhere [16].
Exposure to fast-food restaurants
The primary measure of exposure to the food environment was density of FFR in a 1600-m Euclidean buffer around the home. We used data (collected in 2013) on the location of FFR from Locatus [17], a commercial company that performs yearly audits on all retailers in the Netherlands. We considered all fast food and take away outlets available in the Locatus database as FFR. In accordance with the classification of Lake et al. [18], fast food and take away outlets were defined as outlets serving hot food ordered and paid for at the till, foods cooked in bulk in advance and providing minimal table service (chain and non-chain FFR, take away and delivery outlets, grillrooms and kebab shops). We calculated the density of FFR in a 1600-m, 800-m and 400-m Euclidean buffer around the centroid of each individuals’ six-digit postcode. In the Netherlands, six-digit postcodes consist of approximately 17 residential addresses, as such comprising a relatively small area. In sensitivity analyses, we used (1) density of FFR as quintiles, rather than a continuous variable, and (2) the density of all types of food outlets in a 400-m, 800-m and 1600-m buffer. The latter was done to test whether the overabundant availability of food in general, rather than the density of FFR, is related to an unhealthier diet.
Mastery
Individual sense of mastery was measured at wave six with an abbreviated version of the Pearlin Mastery Scale [11]. This abbreviated mastery scale has previously shown reasonable reliability (α = 0.67) in a non-institutionalized sample [19] and good reliability in individuals with a subclinical depression in the current study (α = 0.81) [20]. In the current study, internal consistency was excellent with α = 0.91. The respondents were asked to rate how much they agreed with five different statements on a five-point scale ranging from 1 (strongly disagree) to 5 (strongly agree). Negatively worded items were reversely coded prior to scoring, resulting in a score range 5–25, with higher scores indicating greater levels of mastery.
Diet
Dietary intake was measured at wave six using a 238-item semi-quantitative Food Frequency Questionnaire (FFQ) with a reference period of 4 weeks [21]. This FFQ was based on an existing validated Dutch FFQ [22]. Overall diet quality was evaluated using an index of dietary adherence to the DASH diet, adapted from that of Fung et al [23]. The index consists of eight dietary components (grains/grain products; vegetables; fruits; low-fat/fat-free dairy; red and processed meat; nuts/seeds/dry beans; dietary sodium; and foods high in added sugar). Unfortunately, the FFQ did not allow for the assessment of dietary sodium intake. As such, the seven remaining components were divided into quintiles (separately for men and women), and the quintiles were summed to create an overall DASH scores ranging between 7 (minimal adherence) and 35 (maximal adherence).
Blood pressure
Systolic and diastolic blood pressure was measured twice during supine rest, on the right arm, using an OMRON M4 IntelliSense digital blood pressure monitor (HEM-752A, Omron Healthcare, Inc., Bannockburn, Illinois, USA). We added 10 mmHg to the average of the two systolic blood pressure (SBP) measurements for those using antihypertensive medication (N = 364). These values represent the average decline in blood pressure in antihypertensive medication trials [24], and have been used in this manner before [25]. Analyses excluding those using antihypertensive medication generated similar results (data not shown).
Covariates
Analyses were adjusted for socio-demographics, total energy intake, depression status and presence of food outlets other than FFR. The latter was done to correct for the fact that food outlets tend to co-locate. Given the high correlation between density of FFR and other food outlets, we derived the residuals of density of fast-food restaurants and all other restaurants and used this as predictor in the models. Models with SBP were additionally adjusted for current smoking status, body mass index, and alcohol consumption. As such, we present Model 1, which is adjusted for age, gender, marital status, years of education, household income, depression status and presence of food outlets other than FFR (for both DASH score and systolic blood pressure as outcome), and Model 2, which is adjusted for total energy intake (for DASH score as outcome) or for total energy intake, current smoking status, body mass index, and alcohol consumption (for systolic blood pressure as outcome). Because 95% of the NESDA participants had a North European ethnicity, we did not adjust for ethnic background. Given the mix of healthy, at-risk and ill individuals, we examined the difference in mastery between individuals with and without current depression using an independent t test. We further tested the interaction between depression status and the densities of FFR in 1600-m, 800-m and 400-m buffers in relation to DASH adherence and blood pressure, as well as the three-way interaction between FFR, mastery and depression status, which were non-significant in all models.
Statistical analysis
From the 2069 individuals that participated in wave six, we excluded individuals who did not complete the FFQ (N = 433), who had extreme energy intake values [26] (N = 35) and those with missing data on their six-digit postcode (N = 58). The final analytical sample consisted of 1543 participants. Descriptives were presented across quintiles of DASH adherence and for the total sample. ANOVAs and Chi-square tests were conducted to examine differences in covariates across the DASH quintiles.
Missing values were handled using multiple imputations (M = 10) with the Predictive Mean Matching method, using all available variables [27].
The independent associations of density of FFR and mastery with adherence to the DASH diet and SBP were examined using linear regression analyses (quadratic and cubic transformations of the determinants were non-significant). For all statistical tests, a probability level of 95% was regarded as significant. The interaction between the independent variables was assessed by adding a cross-product term of the two continuous variables to the model. Following significant interaction terms, analyses were stratified by tertiles of mastery.