ABC | Volume 113, Nº1, July 2019

Original Article Rocha et al Dietary patterns and adiposity in children Arq Bras Cardiol. 2019; 113(1):52-59 The assessed sociodemographic variables were gender and self-reported ethnicity of the child, maternal schooling, family and per capita income, and the type of school in which the child was enrolled (public or private). The behavioral variables were omission of breakfast and sedentary behavior. All questions were answered by parents or guardians. Breakfast consumption was assessed by the first food intake that the child consumed and/or drank within the first 2 hours after waking up. 24 Sedentary behavior was assessed as time spent by the child in activities that did not increase energy expenditure, such as watching television or engaging in other forms of screen-based entertainment. The used cutoff point was screen time ≥2 hours/day, according to the American Academy of Pediatrics (2013). 25 Maternal weight and height were measured using an electronic digital scale, with a capacity of 150 kg and sensitivity of 50 g (Tanita®, model BC 553, Arlington Heights, IL, USA) and a vertical stadiometer, divided in centimeters and subdivided in millimeters (Alturaexata®, Belo Horizonte, MG, Brazil), respectively. Using these data, it was possible to calculate the BMI and to classify it according to the WHO parameters (1998). 26 Statistical analysis Descriptive statistics were used to characterize the sample according to sociodemographic, behavioral characteristics, nutritional status and body composition. At this phase, each variable was assessed through the distribution of absolute and relative frequencies. The normality of the variables was evaluated by the Shapiro-Wilk test, in addition to the evaluation of graphical methods (histogram), kurtosis and asymmetry verification to classify variables regarding normality. Aiming to identify the food pattern, the foods covered by the R24h were measured in grams/day (g/d) or milliliters/ day (mL/d) and collected as isolated foods or food groups by nutritional similarity and their contribution to the hypothesis of diet-disease associations. Moreover, foods consumed by less than 10% of the assessed population were excluded or grouped. 27 To identify the patterns, all foods in milliliters/ day were transformed into grams/day according to the density table of the Food and Agriculture Organization (FAO, 2012). 28 Pattern identification was carried out using an a posteriori methodology, through the Principal Component Analysis (PCA). Before starting the analysis, the sample size was carefully evaluated in relation to the food groups formed in the PCA analyses. 29 For the PCA analysis, the results of the Kaiser-Meyer-Olkin test (KMO = 0.58) and the Bartlett sphericity test (p < 0.001) were estimated. They evaluate whether the data can be used in the PCA. 29 The varimax rotation was performed to facilitate the interpretation of the obtained results, in which factorial loads ≥0.25 (positive or negative) were retained. 24 The number of extracted factors was defined according to the eigenvalue criterion >1 followed by the scree plot graph of variance by the number of components, in which the points on the maximum slope indicated the number of components to be retained. The nomenclature of the found patterns was attributed according to the characteristics of the foods/ formed groups and extracted by PCA. Dietary patterns were presented as explanatory variables, and the schoolchildren’s dietary patterns scores were categorized according to the 75 th percentile of the sample. The Mann-Whitney test was performed to compare the medians of the anthropometric variables and body composition according to the classification of the dietary patterns. The crude analysis was performed using Poisson regression models with robust variance, having anthropometry and body composition as dependent variables. The variables considered important in the assessed association were used for the model adjustment, such as gender, maternal BMI, total energy consumption (kcal) and breakfast consumption. The prevalence ratio (PR) with 95% confidence interval (95% CI) was used as a measure of association. For all performed tests, the level of significance was set at 5%. Statistical analyses were performed using the Stata program version 13.0. Results It was observed that more than half of the children had a sedentary behavior (74.9%) and mothers with excess weight (56.9%). Breakfast omission was observed in almost 20.0% of the sample (Table 1). The PCA analysis identified five dietary patterns (DP): (i) unhealthy DP, consisting of foods/groups of simple sugars and chocolate, fat-rich snacks and whole dairy foods; (ii) snacks DP, consisting of bakery products/food groups and infusions; (iii) traditional DP, consisting of rice, beans, flours, tubers and cereals; (iv) industrialized DP, consisting mainly of ultra-processed products; (v) healthy DP, consisting of foods rich in complex carbohydrates and high biological value proteins (Table 2). Higher median BMI values (p = 0.001), body fat percentage (p = 0.002), waist circumference (p = 0.004), waist-to-height ratio (p = 0.030) and neck circumference (p = 0.001) were observed in children with higher consumption of the industrialized DP (Table 3). In the crude analysis of the regression model, a higher prevalence of increased neck circumference was found in children with higher consumption of snacks DP (PR: 1.79; 95% CI: 1.13 to 2.85). Children with excess weight (PR: 1.58; 95% CI: 1.18 to 2.10) and body fat (PR: 1.50; 95% CI: 1.23 to 1.82) showed higher adherence to the industrialized DP (Table 4). After the adjusted regression analysis, it was observed that children with excess weight (PR: 1.38, 95%CI: 1.02 to 1.87) and body fat (PR: 1.32, 95%CI: 1.07 to 1.64) showed greater adherence to the industrialized DP. Additionally, children with excess body fat (PR: 1.31; 95%CI: 1.01 to 1.74) showed lower adherence to the traditional DP (Figure 1). 54