This is the first study carried out with adolescents from all geographic regions of Brazil, which has characteristics of a country with continental dimensions, which different ethnicities, customs and traditions. Furthermore, this study evaluated if the factors associated (sex, age and geographic region) with overweight and obesity are the same for the different classification criteria used. It is noteworthy that the three criteria used have in common the fact of using arbitrary cutoff points and adjusting BMI values that characterize underweight, overweight and obesity in adults for the population of children and adolescents .
The three references used in this study showed marked differences between each other. Both for boys and girls, increased overweight prevalence was observed by the Brazilian reference of Conde & Monteiro , followed by WHO  and IOTF .
Some studies carried out in Brazil [24, 25] and elsewhere [26, 27] also compared different criteria for nutritional status classification in children and adolescents. Most of these studies demonstrated that the criteria for nutritional status classification differ, but a study conducted in adolescents of Florianópolis, Santa Catarina, Brazil, Farias Jr et al.  found no significant differences in the overweight prevalence determined by the different BMI classification criteria, except for male adolescents, in which the BMI values recommended by Conde & Monteiro  showed higher prevalences compared to the other criteria.
Contrasting these findings, a study carried out in Rio Grande, Rio Grande do Sul, Brazil  with students aged 10 to 15 years revealed that the overweight prevalence was significantly higher using the Conde & Monteiro  criteria than using IOTF . In Canadian preschoolers , the overweight prevalence using the WHO criteria was higher than that using the IOTF and the Centers for Disease Control and Prevention (CDC) criteria .
One possible explanation for the higher overweight prevalence found using the criteria proposed by Conde & Monteiro  is due to the fact that the critical BMI values are lower than those of the other criteria, as already observed by Farias Jr et al. . Moreover, it could be speculated that unlike criteria of IOTF  and WHO , the cutoff points of Conde and Monteiro  were defined with the specific population of Brazil. Thus, one might think that the prevalence detected by the criteria proposed by Conde and Monteiro  may represent the real problem of the high overweight and obesity prevalence in Brazilian adolescents.
When the Kappa test was applied, high concordance between criteria of Conde & Monteiro  and WHO  was found for all age groups and for the data set. Moreover, strong concordance (kappa> 0.75) was observed for the most overweight classifications analyzed. These results corroborate the findings of Dumith and Farias Jr , who also found that the concordance between criteria of Conde & Monteiro and WHO was higher than the concordance obtained between the other criteria.
Regarding associated factors, similar associations were found for the three criteria for BMI classification examined: being male, age 11–12 and 13–14 years, living in the Midwestern, Southeastern and Southern regions of Brazil had greater chances of showing overweight. A study conducted among adolescents in southern Brazil revealed that the direction and magnitude of the association measures were similar between criteria for BMI classification . Corroborating these findings, Abrantes et al.  also observed higher overweight prevalence among adolescents in southeastern Brazil when compared to those of northeastern Brazil.
The results of this study make some hypotheses to be raised: the lower overweight prevalence in girls may be related to a greater concern with body image . The decline in overweight and obesity with increasing age is expected, since moderate excess weight might be compensated by growth, and may even represent a positive aspect of the obesity treatment ; however, excess weight in adolescence should not be underestimated, since there is an increased risk of persisting into adulthood . It is assumed that differences between regions are related to different socioeconomic development between regions , resulting in a greater prevalence in the most developed regions. The higher overweight prevalences found in the Midwestern, Southeastern and Southern regions can be explained due to the difference socioeconomic development between regions, resulting in a greater prevalence in more developed regions, where there are better conditions of purchase and choice of food . Nevertheless, there is an increasing trend in the prevalence of overweight in the North and Northeast and the lower socioeconomic levels .
The present study has the following limitations: 1) the sample selection was performed only in schools that joined the PROESP project, and thereby the participation of more adolescents with obesity characteristic was limited, 2) data are restricted to adolescent students who were in school during the period of data collection and could not be extrapolated to adolescents who were not enrolled in schools.