|
nif:isString
|
-
The Adolescent Nutritional Assessment Longitudinal Study (ELANA) followed two cohorts (middle and high school) at two public and four private schools from the metropolitan area of Rio de Janeiro, Brazil. The metropolitan area of Rio de Janeiro is formed by 18 municipalities representing the second largest centre of national wealth in the country [23]. In contrast, the geographic area of ELANA’s schools faces many social disparities which results in a lack of security and vulnerable neighbourhoods and poverty [24]. The main aim of the ELANA was to examine changes in anthropometric indicators and body composition, and to verify the influence of health, socioeconomic and psychosocial factors on inadequate development and nutritional status. This study included data from the middle school cohort from 2010 to 2012. At baseline, a self-reported questionnaire was administered to investigate physical activity, sedentary behaviour, socioeconomic and demographic variables among the adolescents. Questionnaire of physical activity was repeated in 2011, and sedentary behaviours in 2011 and 2012. This research was conducted according to the guidelines established in the Declaration of Helsinki. The Ethics Committee of Research of the Institute of Social Medicine of the State University of Rio de Janeiro (certificate number 0020.0.259.000–09) approved all procedures involving human subjects before the start of the study. Adolescents’ legal guardians provided a written informed consent. All the data we used were confidential and the researchers did not have access to any personal data that could identify individuals included in the study. Following the internal regulations of the ELANA Study committee, data are made available for specific research projects. Thus, we are not allowed to share the data we used for this study with other researchers. However, we are glad to answer questions about the data used in this study and to share unpublished results.
Out of the 946 adolescents available, 888 met the eligibility criteria for this study, which included not having a physical or mental condition preventing the completion of questionnaires and/or not being pregnant or lactating at the time of data collection. Of the 888 eligible adolescents, 32 (3.6%) refused to participate, 46 (5.2%) did not have parental consent, and 4 (0.45%) had no birth date information. The study sample comprised 810 students at baseline, of which 786 adolescents completed information of physical activity and sedentary behaviours at baseline, corresponding to a response rate of 88.5%. In 2011, 526 students repeated the evaluation of physical activity and sedentary behaviours (33% missing cases), and 435 of sedentary behaviours in 2012 (46% missing cases). Imputation of missing cases was carried out by means of maximum likelihood estimation, allowing the study to include 810 adolescents in all time points.
Physical activity (PA) was assessed by the short version of the self-reported ‘International Physical Activity Questionnaire–IPAQ’ which has eight open questions estimating the weekly time of vary intensities of PA during the last 7 days (light PA: walking; moderate PA: carrying light loads, bicycling at a regular pace, or doubles tennis; and vigorous PA: lifting, digging, aerobics, or fast bicycling). This questionnaire was validated by Guedes et al. (2005) [25] for Brazilian adolescents older than 14 years old. The Brazilian version of short IPAQ has been validated against a 24-hour recall instrument of the daily activities, identifying correlation coefficients varying between 0.49 to 0.83. When detecting blank answers and unusual values for responses (e.g. the adolescent said they performed eight hours of MVPA per day) the research staff returned to the student in order to check the coherence of the answer, avoiding loss of information and classification errors. The time (in minutes) of light PA and MVPA per day was used, and it was based on the guidelines for data processing and analyses of the IPAQ [26].
The television viewing (TV) and video game/computer use (VG) times were assessed separately by two questions from a self-reported questionnaire [27]. The first question was ‘How many days do you watch TV and use VG per week? (1) Almost never or never, (2) 1 to 2 times per week, (3) 3 to 4 times per week, (4) 5 to 6 times per week, and (5) every day’. The responses were categorised in a five-point scale: category 1 = 0 days (almost never or never), 2 = 1.5 days, 3 = 3.5 days, 4 = 5.5 days, 5 = 7 days. The second question was ‘In general, how many hours do you usually spend watching TV and using VG per day?’. Average daily time in minutes was calculated multiplying ‘hours per day’ by ‘days per week’ for TV and VG variables, applying the formula: [(days per week)*(hours per day)]*60/7.
Age, Asset ownership, body mass index (BMI) and sexual maturation were treat as potential baseline confounders of the investigated association. The economic characteristics of the families of the adolescents were assessed through the Brazilian Socioeconomic Classification Criteria (CCEB) [28], which takes into account the purchasing power of urban households based on a score obtained by the sum of the household assets, the presence of domestic workers in the household and the schooling level of the head of household. An indicator constructed from information on ownership of durable assets in the home was used to calculate the presence or absence of, respectively: TV; VCR or DVD player; radio; bathroom; automobile; washing machine; refrigerator; and freezer (independent appliance or part of duplex refrigerator). A weighted score was attributed considering presence and relative frequency of each item [29]. The Body mass index (BMI in kg/m2) was calculated based on direct measurements of height and weight [30]. Sexual maturation stage was investigated using the self-evaluation technique validated by Saito (1984) [31], focusing on the development of genitalia for boys and breast for girls, according to Tanner’s (1962) [32] criteria. This information was coded as a categorical variable (pre-pubescent, early spurt, maximum speed peak, and slowing growth).
Firstly, the distribution of all variables in the sample was tested by the Kolmogorov Smirnov test. Comparisons between the genders were performed through the non-parametric Wilcoxon Test, and Wilcoxon Rank Compared Test to evaluate differences between means ranks over time from T1 to T2. In order to facilitate the interpretation of the descriptive results, we presented means, standard deviations, minimum and maximum, previously tested and that showed similar statistical results with non-parametric tests (p values) (details available upon request). Due to the non-normal distribution of the sample, an asymptotic estimation (more appropriate for non-parametric distributions, without missing value imputation) was performed instead of the maximum likelihood missing values estimation. All coefficient values remained similar in this analysis, but lost their statistical significance, most likely due to the decreased sample size. Hence, it was decided to continue the analysis with the maximum likelihood for missing values estimation. The maximum likelihood estimation was used to deal with missing values, which is probably the most pragmatic missing data estimation approach for structural equation modelling (SEM), and has been shown to produce unbiased parameter estimates and standard errors for data missing at random and data missing completely at random [33]. School changes by students are very common in Brazil, making it difficult to follow them, and these are probably not related to the outcomes or the other variables being investigated. Further sensitivity analysis was performed to support the assumption of missing at random or completely random for the imputation process (available upon request). Autoregressive cross-lagged panel models were estimated by means of SEM to simultaneously address bidirectional association of PA and screen activities, with maximum likelihood missing values estimation. Two different models were performed, considering the intensities of PA [Model A: light PA; Model B: MVPA, both stratified by gender. A baseline model was constructed with autoregressive paths (measuring stability over time) from light PA and MVPA (Models A and B, respectively) at Time 1 (T1-2010) to Time 2 (T2-2011), and from screen activities (TV and VG) at T1 to T2, and from T2 to Time 3 (T3-2012). The autoregressive model also included correlations between PA (light PA: model A and MVPA: model B) and screen activities at T1 and T2; and within screen activities (TV and VG) at T3. Apart from the baseline model, three additional models were assessed which included one or more cross-lagged estimate (I. Forward causation model: PA at T1 predicts screen activities at T2; II. Reversed causation model: screen activities at T1 predict PA at T2; III. Reciprocal model: PA and screen activities have reciprocal effects). Baseline confounders were included in all models but not shown in illustrations. The models were illustrated in Fig 1 (1a [Model A] and 1b [Model B]).
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0211470.g001 Logic models. (1a) Model A for light physical activity (PA); (1b) Model B for moderate and vigorous physical activity (MVPA). The Adolescent Nutritional Assessment Longitudinal Study, 2010–2012, Brazil.
For each of the models (A and B), a set of model fit statistics was derived: the Standardised Root Mean Square Residual (RMSEA), which should be below or close to 0.06, as well as the Comparative Fit Index (CFI) and the Tucker-Lewis Index (TLI), which should both be close to or above 0.95. The p value for chi-square test was also presented for the model fit statistics, which should be above 0.05. Age, BMI, asset ownership and sexual maturation were used in the models as confounding variables assessed at baseline (suppressed on the Fig 1a and 1b [methods], and Figs 2 and 3 [results]. All analyses were performed in Stata version 13.1 (Stata Corp, Texas, USA).
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0211470.g002 Associations between light activity, TV and VG (Males n = 436, Females n = 374).Results from structural equation modelling. Estimates (standardized) are displayed as males/females. *** p<0.001, ** p<0.01, * p<0.05(baseline confounders were omitted) The Adolescent Nutritional Assessment Longitudinal Study, 2010–2012, Brazil.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0211470.g003 Associations between MVPA, TV and VG (Males n = 436, Females n = 374).Results from structural equation modelling. Estimates (standardized) are displayed as males/females. *** p<0.001, ** p<0.01, * p<0.05 (baseline confounders were omitted). The Adolescent Nutritional Assessment Longitudinal Study, 2010–2012, Brazil.
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}