This was a cross-sectional study performed in the city of Florianópolis, in the State of Santa Catarina, South of Brazil, during the months from April 2007 to October 2007.
The sampling procedure is described in detail elsewhere [13–15]. Briefly, for sample size calculation, the following parameters were assumed: excess weight prevalence of 22.1% for school children aged 7–10 years  and 12.6% for those aged 11–14 years ; acceptable error of 3 percentage points; two-tailed test; confidence level of 95%; design effect of 1.5; and an addition of 10% for losses. Subjects were excluded if the parents or primary caregiver did not sign the written informed consent. The sampling procedure was probabilistic, stratified by clusters and performed in two stages (school and children). Schools were grouped in four strata, according to the geographical area and school type: center/continent public schools, center/continent private schools, beaches public and beaches private schools, considering a probability proportional to the size (number of students) of each stratum. In the first stage, schools were randomly selected from each stratum. From a total of 87 schools (33 private and 54 public), 17 schools were selected (6 private and 11 public). In the second stage, school children were randomly selected according to age. For the present study, 7 year old children (n = 421) were excluded, as there was no information on sexual maturation. The final sample consisted of 2412 school children aged 8 to 14 years (1144 males, 47.4%).
Anthropometric measures were taken following the protocol of the World Health Organization , as proposed by Lohman . Body weight was measured to the nearest of 50g using an electronic scale (model PP 180, Marte®, Minas Gerais, Brazil), capacity of 180 kg; height was measured to the nearest of 1 mm by Alturexata® stadiometer.
The team responsible for data collection was composed of 10 people previously trained in a workshop in the period of 2006 September to 2007 March. This workshop consisted of a theoretical and practical course in anthropometric measures, as well as of a pilot study in two schools (one private and one public), in which the intra- and inter- observer errors in anthropometric measures were assessed . Both schools were excluded from the sampling of the study.
Sex-specific body mass index (BMI)-for-age growth charts were used to assess excess body weight, according to the World Health Organization criteria [4
]. The variable excess body weight comprised the sum of overweight and obesity derived from BMI, in which 0 represents no excess of body weight and 1 represents having excess body weight. Z-scores for height-for-age, weight-for-age and BMI-for-age were also calculated for each sex separately, using LMS values. The LMS sums up the data in smoothed curves that are specific to each stratum, which in this case are the ages and genders. Parameter M is the median value of the index observed inside each stratum; parameter S is the coefficient of variation for each stratum; and parameter L is the Box-Cox coefficient employed for the mathematical transformation of the values of the variable in question in order to obtain a normal distribution in each stratum [21
]. The cutoff point for biologically implausible values was 5 z-scores [4
]. To calculate the z-score, LMS values were used, by age and gender, according to the following formula:
Sexual maturation was determined based on the stages of development proposed by Tanner, comprising 5 stages for genital (males) and breast (females) development . School children were asked to perform a self-evaluation; the instructions were given individually.
For each sex, the age distribution (in decimals) of each sexual maturation stage was used to define tertiles of age. Based on this, individuals were classified in three groups: early sexual maturation (below first tertile of age), late sexual maturation (second tertile or more) and normal sexual maturation or reference group (between the first and the second tertile) . This criterium considers that a younger age at a sexual maturation stage indicates precocious sexual maturation.
The birth weight of school children was reported by the parent or primary caregiver, who was asked to check the child’s health record. The subjects were classified into low (< 2.500 g), normal (≥ 2.500 g - 3.999 g) and high birth weight (≥ 4.000 g).
The weight status of the mother was evaluated by BMI, using self-reported weight and height. Excess body weight classification (≥ 25 kg/m2) followed the recommendation of the World Health Organization .
Information on the mode of commuting to school was collected by an illustrated questionnaire and subjects were classified as active (walking and biking) or inactive (car, bus, passenger in motorcycle or bicycle).
Further information (name, date of birth, school grade and type of school) were obtained in documents provided by the school. The type of school refers to public or private.
Given the non-normal distribution of anthropometric data (Shapiro-Wilk test, p < 0.05), quantitative variables (age, BMI Z score, height Z score, weight Z score) were described as median, 25th and 75th percentiles. Qualitative variables (excess body weight, sexual maturation classification, maternal excess body weight, birth weight, active commuting to school, school type) were described by relative frequencies (%). To compare quantitative variables between groups, Mann–Whitney (female vs. male) and Kruskal-Wallis (early, normal or late sexual maturation) tests were used. The associations among qualitative variables were assessed by Rao-Scott test. The relation between the independent variables and excess body weight (0-absence; 1-presence) was tested with univariate and multivariate Poisson regression models, using robust variance and stepwise forward strategy (multivariate model). The association between the independent variables and height was evaluated using a linear regression model (homoscedasticity checked by Breusch-Pagan and Cook-Weisberg test). Stata 11.0® was used in all statistical analysis, including the command svy to allow for sampling weights and stratification [25, 26].
This study was approved by the Ethics Committee on Human Research of the Santa Catarina University (UFSC) (protocol number 028/06).