Study design and sample
This is an observational longitudinal study with participants from the Schoolchildren’s Health Study, which began in 2011. All children and adolescents enrolled in 25 randomly selected public (municipal and state) and private schools from Santa Cruz do Sul, Brazil, were invited to participate in the baseline assessment (1,687 children and adolescents). All individuals were invited to participate in the follow-up assessment in 2014, only 420 participants accepted to be followed-up (24.9% retention), however 13 participants were excluded due to missing information, totaling 407 participants at follow-up, aged from 8 to 17 years (Figure 1).
This study was approved by the University of Santa Cruz do Sul research ethics committee (nº 1.836.983) and it was conducted following the Resolution 466/2012 of the National Council of Health in Brazil. The schoolchildren’s parents or legal guardians signed free and informed consent forms.
**Insert figure 1
Measures
All measurements were taken at University of Santa Cruz do Sul at baseline (2011) and follow-up (2014) periods. Arterial blood pressure was measured with a sphygmomanometer with appropriate brachial perimeter and a stethoscope were placed on their arm. Then, SBP and DBP were determined by manual auscultation, with the student sitting, resting forfive minutesprior to measurement. Each device had three different sized cuffs so that researchers could select the most suitable for each arm circumference. Two measurements on the right arm were made, and the lowest blood pressure recorded. All procedures were adopted following the VI Brazilian Guidelines for Hypertension[9].
The BF% was determined through the measures of tricipital and subscapular skinfolds, evaluated using a Lange® caliper (Beta Technology Inc, Houston, TX) by the same evaluator at both the 2011 and 2014 time points. Each skinfold was evaluated twice, and if the difference between measurements was higher than 2 mm, a third evaluation was performed. The lowest value was used for analyses. The BF% was calculated using equations of Slaughter et al (1988)[10]according to sex.
The CRF was assessed by the 9-minute running and walking cardiorespiratory fitness test in 2011, described by Projeto Esporte Brasil[11], and by the 6-minute running and walking CRF test in 2014, described by Projeto Esporte Brasil[12]. The indirect submaximal exercise tests, assessed in meters, were used to estimate peak oxygen uptake (VO2peak) by the following equations: 9-minute test→VO2peak = 47.547 + 0.008 * (Test) – 0.805 * (BMI) + 4.236 * (Sex)[13], and 6-minute test→VO2peak = 41.946 + 0.022 * (Test) – 0.875 * (BMI) + 2.107 * (Sex) [14], where test is the value of meters performed by the student, and sex equals to 1 and 0 for boys and girls, respectively.
Covariates
Information about age, sex, and skin color were obtained through a self-reported questionnaire. Height was measured on the anthropometric scale with a coupled stadiometer. The pubertal status was evaluated at follow-up using Tanner´s criteria[15]. The participant should filled the image corresponding to their current pubertal status considering genital and pubic hair. Therefore, five stages of sexual maturation were considered and classified into pre-pubertal (stage 1), initial development (stage 2), continuous maturation (stages 3 and 4), and matured (stage 5). Socioeconomic status was assessed by the questionnaire of the Brazilian Association of Research Companies[16], considering the head of household’s educational level and the quantity of appliances the family has (car, washing machine, bathroom, among others). A score was obtained according to the answers, thus, the sum of these scores indicated the family’s social class: low (D-E), medium (C), and high (A-B).
Statistical analysis
Descriptive data are presented as means and standard deviations for continuous variables and absolute and relative frequencies for categorical variables. The independent student t-test was used to verify differences between sexes, whereas the t-test for paired samples was used to verify differences between baseline and follow-up scores. Effect size (Cohen’s d) was calculated. Values of d<0.39 indicated a small difference, 0.40<d<0.79 indicated a medium difference, and d>0.80 indicated a large difference[17]. Effect sizes (Phi [φ] and Cramer’s V) were also calculated for the chi-squared test, which verified the difference of frequencies between sexes for the categorical variables. Linear regression models were used to test the relationship between baseline values and changes in CRF (VO2peak)and adiposity with blood pressure at follow-up. Moderation analyses were tested using multiple linear regression models through PROCESS macro,which is a program extension for the Statistical Package for Social Sciences (SPSS) version 24.0 (IBM Corp, Armonk, NY, USA). The following models were tested: a) associations between changes in CRF (VO2peak) with SBP and DBP at follow-up, b) associations between BF% at baseline with SBP and DBP at follow-up, c) Interactions between changes in CRF (VO2peak) and BF% at baseline with SBP and DBP at follow-up.
The Johnson-Neyman technique was used to probe interactions by assessing whether changes in CRF (VO2peak) moderated the relationship between BF% at baseline with SBP and DBP at follow-up. This technique verifies the association between the independent and dependent variable across different levels of the moderator variable (we present the relationship at 16th, 50th, and 84th percentiles because of the skewness of the moderator variable). In the context of the current study, the technique highlights specific changes in CRF (VO2peak) cut point in which the significant relationship between BF% at baseline with SBP and DBP at follow-up appears or disappears. All analyses were adjusted for sex, age, pubertal status, height, socioeconomic level, skin color/ethnicity, and the dependent variable itself at baseline. The probability value p<0.05 was considered as significant for all analysis.
Multiple linear regression was used as a statistical test for sample calculation on G*Power 3.1 program (Heinrich- Heine-Universität), considering the following parameters: test power (1-β) = 0.95, a significance level of α = 0.05, and effect size of 0.05. The number of predictors considered was 10, and the minimum number of participants was established as 348. However, to avoid probably difficulties with sample loss, an increase of 15% was assumed, totaling 400 children and adolescents.