This cross-sectional study was approved by the Ethics Committee on Human Research of the “Extremo Sul Catarinense” University under protocol 225/2009. Adolescents who participated in the survey and parents/guardians signed the consent form authorizing their participation in the study.

The study was conducted at the city of Criciúma, southern Brazil, located in the Brazilian state of Santa Catarina, with human development index of 0.788, considered high, and life expectancy at birth of 75.8 years. 12

The target population for this study consisted of 13 508 elementary school students from Criciúma, Santa Catarina, Brazil. The sample size was calculated using an unknown prevalence for the outcome (50%), confidence level of 95%, estimated error of 5%, design effect of 2.0 and increment of 20% for losses and refusals. Given these parameters, a sample of 897 adolescents was estimated.

To ensure that the sample represented the target population, distribution considered school system (public and private), and school size (small, with fewer than 200 students; intermediate, with 200–499 students and large, with 500 students or more). The sample selection used the sampling procedure of clusters in two stages: the school was the sampling unit of the first stage and the classes were the sampling unit of the second stage. All schools in Criciúma with basic education were eligible for inclusion in the study. In the first stage, school density was adopted as stratification criterion, according to system and size, so proportionately more schools in the school system that had more schools and higher density were selected. In the second stage, the density of classes in the selected schools was considered as a criterion to draw those in which the questionnaires would be applied. All students in the selected classes were invited to participate in the study.

Data collection was conducted in the first half of 2010. Questionnaires were applied in classroom, without the presence of teachers. The evaluation team attended a previous training program for the standardization of data collection procedures. Physical fitness tests were performed in school gyms. For these tests, the students received oral explanation and demonstration of their execution.

The following inclusion criteria were used in the study: (a) students aged 10–14 years, (b) absence of health problems that prevent them from performing physical tests and anthropometric measurements. Adolescents who refused to participate and those who did not have the consent form signed by parents/guardians were considered refusals.

The dependent variable of this study was aerobic fitness. The aerobic fitness was assessed by the 9-min running test of the American Alliance for Health, Physical Education, Recreation and Dance (AAHPERD), 13 which has been validated for children 14 and adolescents. 15 Students were instructed to run the longest distance as possible in 9 min. The test took place in school gymnasiums that were prepared for this test. Walking during the test was allowed, but the goal was to cover the longest distance as possible in 9 min.

The distance was measured in meters, so that it was later dichotomized as “high” and “low” aerobic fitness levels. For this classification, the cutoff points described in the AAHPERD Manual 13 were used, which establish the 50th percentile as the minimum level for adequate cardiorespiratory function. The cutoff points of AAHPERD 13 are in yards and for the present study, they were transformed into meters (1 yard ≅ 0.91 m).

The independent variables of this study were age in years, which was dichotomized into (10–12 years/13–14 years) and socioeconomic level, which was collected through instrument of the Brazilian Association of Research Companies. 16 In this instrument, the population was divided into eight classes in decreasing order of purchasing power (A1, A2, B1, B2, C1, C2, D and E). In this study, A1 and A2 were grouped into “High”, B1 and B2 into “Intermediate” and C1, C2, D and E into “Low”. Another independent variable was school system (public/private).

Physical activity was analyzed by estimating the energy expenditure. 17 The diary was applied in two days, one day of the week and one day of the weekend. In the diary, the adolescent recorded the physical activities carried out on that day every 15 min. Daily activities are classified on a continuum involving nine categories, according to estimates of the average energy expenditure of activities performed by humans: (1) resting in bed; (2) activities performed in a sitting position; (3) light activities in a standing position; (4) activities that require light walking (<4 km/h); (5) light manual work; (6) active leisure activities and recreational sports; (7) manual labor at a moderate pace; (8) active leisure activities and sports of moderate intensity; and (9) and heavy manual labor in competitive sports. This instrument provides reproducibility of r = 0.91 in subjects from 10 years of age. 17 In addition, the diary showed a correlation coefficient of 0.87 with the TriTrac accelerometer. 18 From the recorded activities, the physical activity level was determined by estimating the daily energy demand in kilocalories per kilogram of body weight (kcal/kg/day), considering the energy equivalents of activities involving humans. There are no cutoff points for the classification of the physical activity level from the daily energy demand in literature. Therefore, the sample was classified according to the division into energy expenditure percentiles. Subjects who were below the 25th percentile were classified as insufficiently active. Students showing energy expenditure above the 25th percentile were considered physically active. Similar strategy to classify adolescents as for the physical activity level was adopted by Eisenmann et al. 19 The goal of dividing into percentile is to compare the most active with the least active. The study does not aim to analyze adolescents who reach the recommended physical activity levels, because the instrument used does not have this feature.

Sedentary behavior was verified by time spent watching television in a common day of the week and weekend. The results of the week were added of a weight of five, equivalent to the days of the week and the results of the weekend were added of weight of two, dividing the total result by seven, thus obtaining the average per day. Sedentary risk behavior was considered as time spent watching TV more than 2 h/day. 20

The nutritional status was assessed by collecting weight and height 21 and calculating the body mass index (BMI). International cutoff points for BMI, 22 according to age and sex were used. This variable was dichotomized into normal weight and overweight (overweight + obesity).

Triceps and subscapular skinfolds were measured according to procedures described in literature, 21 which were measured using scientific caliper brand Cescorf®, a Brazilian model that features design and mechanics similar to English Harpenden model, with supposed constant pressure on any opening of its rods of approximately 10 g/mm2, measurement unit of 0.1 mm and contact area (surface) of 90 mm2. Excess peripheral (triceps skinfold) and central adiposity (subscapular skinfold) was rated based on a value corresponding to the 90th percentile of the reference distribution for sex and age according to the Centers for Disease Control and Prevention curve (CDC). 23 This study examined the variable considering with excess general body adiposity adolescents who had both triceps and subscapular skinfold thickness values above the 90th percentile. 23

Analyses were stratified by sex using the Stata 11.0 software. Descriptive and inferential statistics was used. The “t” test for independent samples was used to compare continuous variables between sexes. In crude and adjusted association analysis, Wald test and binary logistic regression were used to estimate odds ratio (OR) and confidence intervals of 95% (95%). Regression analysis followed a hierarchical model to determine low aerobic fitness level, hypothetically temporal, according to recommendations of Victora et al. 24 Statistical analysis followed the division into three groups of variables: (a) distal, which included sociodemographic variables (age, socioeconomic status, school system), (b) intermediate, which included health-related behaviors (physical activity and sedentary behavior); (c) proximal, which included nutritional status and body adiposity. All variables were adjusted for analysis, regardless of p-value in the crude analysis.

The adjusted analysis adopted hierarchical approach, following the backward procedure. Initially, the adjustment was made for variables of the distal block to the other factors of the same level. Then, variables of the intermediate block were controlled for variables of the same level and for those of the distal level that remained in the model. Finally, variables of the proximal block were adjusted for the other factors of intermediate and distal levels that remained in the model. The criterion used for the maintenance of the factor in the regression analyses was p-value in the Wald test below 0.20. 25 In addition, factors whose p-value was less than 0.05 were considered significantly associated with the outcome.