Change in overweight and obesity over a decade according to sociodemographic factors in Brazilian adolescents

Resumo Os objetivos do estudo foram examinar as mudanças na prevalência de sobrepeso e obesidade entre os estudantes de 2001 a 2011 e verificar se essas alterações diferem de acordo com a faixa etária, a área de residência e a renda familiar. Analisamos dois levantamentos transversais realizados em 2001 e 2011 com adolescentes brasileiros. As características sociodemográficas e antropométricas foram autorrelatadas por meio de um questionário. Regressão logística multinomial foi utilizada para analisar as mudanças no sobrepeso e obesidade entre os dois inquéritos. O estado nutricional foi considerado o desfecho do presente estudo e os inquéritos (2001 e 2011) foram as exposições. As razões de chance de sobrepeso e obesidade foram maiores entre os meninos e meninas em áreas urbanas no inquérito de 2011 em comparação com 2001. Meninos e meninas com maior renda também apresentaram maior chance de sobrepeso e obesidade no segundo inquérito. Sugerimos que futuras intervenções considerem a área de residência e a renda familiar para fortalecer a eficácia das ações desenvolvidas para prevenir e controlar esses indicadores entre os adolescentes. Palavras-chave Obesidade, Adolescente, Renda, Área urbana, Área rural Abstract The aims were to examine changes in the prevalence of overweight and obesity among students from 2001 to 2011, and to verify if these changes differ according to age group, area of residence, and family income. We analyzed two cross-sectional surveys conducted in 2001 and 2011 with brazilian adolescents. Sociodemographic and anthropometric characteristics were self-reported using a questionnaire. Multinomial logistic regression was used to analyze changes in overweight and obesity between the two surveys. The nutritional status was the outcome of the present study and the surveys (2001 and 2011) were the exposure. The odds of being overweight and obese were higher among boys and girls in urban areas in the 2011 survey compared to the 2001. Boys and girls with higher income were also more likely to present overweight and obesity in the second survey compared to the first. An increase of overweight and obesity was observed over a decade. We suggest that future interventions consider the area of residence and the family income to strengthen the effectiveness of actions developed to prevent and control these indicators among adolescents.

introduction An increasing trend in the prevalence of overweight and obesity in adolescents has been observed in many countries in the last decades [1][2][3][4][5][6] . Likewise, in Brazil, the National Household Budget Survey 7 showed that obesity increased from 0.4% to 5.9% among boys and from 0.7% to 4.0% among girls in a period of 34 years. This scenario has become a public health concern worldwide 8,9 , since higher body mass index is associated with risk of obesity in adulthood [10][11][12] , and the presence of these outcomes in any of the life stages may result in negative health consequences, increasing public costs 8 .
Although a global increase in the prevalence of overweight and obesity in adolescents is reported 8,9 , there are large variations in the behavior of these outcomes over time for specific subgroups. Wang and Lim 8 mention that prevalence may vary according to area of residence, age, gender, race/ethnicity, and socio-economic status. However, there is lack of information regarding these trends in other sociodemographic subgroups 13 such as family income, which has been associated with overweight, especially in middle income countries like Brazil 14 .
Therefore, monitoring of overweight and obesity trends is extremely relevant for the comparison over time 15 , as well as to explore which subgroups have been more or less affected. These results can assist in the comprehension of these phenomena and allow the development of more specific strategies and targeted at the different needs of each context. In this sense, the aims of the study were to examine changes in the prevalence of overweight and obesity among students in Santa Catarina State from 2001 to 2011, as well as to verify if these changes differ according to age group, area of residence, and gross monthly household income.

Sampling
This study analyzes secondary data from two cross-sectional surveys (2001 and 2011) of the "Health risk behaviors project (CompAC) in youth of the Santa Catarina State, Brazil", with a representative sample of high-school students (15-19-year-old) from public schools of Santa Catarina State, Brazil. This research used the procedures of sampling by conglomerates, considering as strata the geographic regions and its respective regional education management. The calculation of the number of schools required in each stratum was obtained applying the proportionality criterion. For the number of classes required, an average of 25 students per class was estimated in 2001. As this number of students per class did not reach the required sample, an average of 17 students were considered for the 2011 survey. Whit this information, the probabilities of participation in the sample were obtained for the sample weight calculation and correction of the imbalance introduced by the adopted sampling plan.
The same statistical parameters for sample size calculation (prevalence of the phenomenon of 50%, maximum error of 2%, design effect of two, and addition of 25% to account for possible losses or refusals) were adopted in both surveys (resulting in an estimated sample size of 5,932 adolescents), as well as the research tools used. After the exclusion of ineligible students (out of the age range of 15-19 years) in both surveys (2001: n = 380; 2011: n = 508), it was observed that 55 of the 5,083 participants in the 2001 survey filled-out the questionnaire incorrectly (final sample of 5,028) and, of the 6,569 students in the 2011 survey, 40 had their information lost due to the same reason (final sample of 6,529). Additional information on the study methodology can be found on a previously published paper 16 .
Both surveys (2001 and 2011) followed ethical principles and were approved by the Ethics Committee on Human Research of the Federal University of Santa Catarina, Brazil. For students under the age of 18, parents or legal guardians needed to give their consent for participation in the research, and passive informed consent was used 16 .

Data collection
A standardized questionnaire was applied. The reproducibility of the instrument ranged between 0.69 and 0.99 in 2001 and between 0.51 and 0.96 in 2011 16 . In the present study, we used self-reported measures of weight and height to calculate participants' body mass index (BMI). Nutritional status was calculated according to WHO cut-off points 17 : normal weight: <-2SD; overweight: >+1SD (equivalent to BMI 25 kg/m 2 at 19 years); obesity: >+2SD (equivalent to BMI 30 kg/m 2 at 19 years).
Sociodemographic strata were: age (15-16 and 17-19 years old); area of residence (rural and urban), and gross monthly household income (until five minimum wages and six or more minimum wages earned by all household members). In 2001, the minimum wage in Brazil was R$180.00 (US$71.71) and in 2011 it was R$545.00 (US$340.63).

Statistical analysis
Descriptive statistics was used to present the prevalence of overweight and obesity among adolescents, stratified by gender, in the two surveys (2001 and 2011) and their respective 95% confidence interval (95%CI). Multinomial logistic regression was used and nutritional status was the outcome (normal weight was the reference category) and the surveys (2001 and 2011) were the exposure. The adjusted analysis was performed in the following manner: a) age subgroup: adjustment for gross monthly household income and area of residence; b) area of residence subgroup: adjustment for age group and gross monthly household income; c) gross monthly household income subgroup: adjustment for age group and area of residence. Values were expressed as odds ratios (OR) and 95%CI to verify the association between the survey year (2001 vs. 2011) and the outcomes (overweight and obesity). This decision was justified by the trends investigated in other studies 8,9 . With the aim of verifying the average outcome variation over the years, we estimated annual change rates. Therefore, for each outcome, we calculated the difference (subtraction) of the prevalence rates from 2011 and 2001. Then, the result obtained was divided by ten (corresponding to the number of years between the two surveys: 2001 to 2011) 18,19 . It should be highlighted that these calculations were made for each of the subgroups considered in the present study (age group, area of residence, and gross monthly household income).
The value corresponding to the relative outcome change (percent change) over ten years was calculated as the difference between the prevalence of each outcome referent to 2011 and 2001. Next, the value found was divided by the prevalence of 2001 and multiplied by 100, being represented by the term "delta".
Data analyses were performed using Stata® Standard Edition, version 13.0 for Microsoft® Windows™ (StataCorp LP, USA) and in Microsoft Office Excel version 2013. We considered the methodology for analysis of complex data, adding the prefix "svyset" to incorporate strata, clusters, and sampling weights. For all tests, the level of significance was 5%.

Results
Among the 5,028 students that filled out the questionnaire, a total of 4,924 answered the questions relative to weight and height in 2001 (response rate = 97.9%), and the 6,529 total students, 6,366 answered these questions in 2011 (response rate = 97.5%). Most of the study participants were females (59,4% in 2001; 55,5% in 2011), aged between 15 Table 1 presents the trend of overweight and obesity among boys over a decade, according to sociodemographic strata. Delta values, representing percent change in outcomes over the years, ranged from -7.0% to 45.4% for overweight and 97.6% to 363% for obesity. Annual rates, showing the percent change per year, were of -0.01% to 0.62% for overweight and of 0.08% to 0.27% for obesity. In the 2011 survey, compared to the 2001 survey, the odds of being overweight and obese were higher among younger (ORobesity: 4.

Discussion
The present study showed that among both boys Studies shows evidences about global increase in the prevalence of overweight and obesity in adolescents 2,4,20 . In Brazil, the National Household Budget Survey also identified an increased prevalence of obesity among boys and girls in a period of 34 years, especially in the South Region (boys: 0.6% to 7.7% and girls: 1.0% to 5.4%) 7 .
In Guangzhou (China), between 2007 to 2011, the prevalence of overweight increased in boys from 10.5% to 12.9% and in girls from 5.9% to 7.0%. In the same period, obesity increased from 4.2% to 5.7% and from 2.1% to 2.7% among boys and girls, respectively 20 . There was also an increase in overweight and obesity among Mexican adolescents, from 11.1% in 1998 to 35.8% in 2012 21 .
In fact, a systematic review conducted by Wang and Lim 8 describing the prevalence of overweight and obesity worldwide states that, over the last two decades, the increase in the prevalence of these outcomes has been occurring especially in countries that experienced rapid economic and social transitions, like China, Mexico, and Brazil. The authors of the systematic review affirm that a considerable number of studies point to the important contribution to the current scenario of globalization and its resulting change in society, environments, and individual behavioral patterns. Adding to this point, there are other factors such as economic growth, modernization, urbanization, and food market globalization 4 . This social and cultural transition results in an increase of fast-food consumption and a reduction in traditional and healthier meals. Also, low levels of physical activity and high time spent in sedentary behavior are factors that might be related to the high prevalence of overweight and obesity among adolescents 2,22 .
On the other hand, it has been observed in Australia 23 , Canada 24 , and some European countries a trend of stability or even decrease in the prevalence of overweight and obesity among both boys and girls 22,25 , Authors' arguments to explain this scenario are based on increased media attention 26 , intervention programmes 22 and increased parental awareness 25 of the consequences of obesity. However, it is necessary more attention to the fact that children from lower socio-economic strata have not benefited from this trend 25 .
The literature shows trends regarding the overweight and obesity in the overall population, however, our findings, highlights the importance of examining trends in overweight and obesity taking into account some characteristics. In general, we have observed that in 2011 there was an increase in the odds of being overweight and obese in most of the subgroups, however, we emphasize that, the odds of the outcomes differ according to the subgroup investigated. Among the both younger and older age groups, the odds of being overweight were higher from 2011, compared to 2001. In relation to obesity, odds were higher only for girls. It is observed that, even considering subgroups with different ages, the new generation has significantly lower physical activity levels, significantly higher screen time and changes in dietary intake, a trend that occur concomitantly with the increase in obesity 3,27 .
The prevalence of overweight and obesity presents a distinct behavior in rural and urban areas, being influenced by the culture and the level of country development 28,29 . Over a decade, we verified that the odds of being overweight and obese were higher in adolescents living in urban areas only, which was verified in previously studies, especially in developing countries 20,30 . According to a meta-analysis conducted by Yu et al. 31 , from 1991 to 2010 the prevalence of overweight and obesity among the Chinese increased more in urban areas (8.2%; CI: 6.4%, 9.9%) than in rural areas (5.3%; CI: 4.2%, 6.3%).
These findings might be explained by the increased access to ultra-processed foods and energy-dense foods and reduced intake of fruits and vegetables, as well as the lower levels of physical activity among adolescents living in urban areas. Moreover, families living in urban areas usually have more televisions and computers and prefer to use the car instead of walking or using a bicycle 31,32 , increasing the risk of hypokinetic diseases such as obesity. Conversely, families living in rural areas still tend to engage in manual labor, use more active transportation and eat more fruits and vegetables compared to those living in urban areas 20,30,33 .
Nonetheless, special attention should be given to girls living in rural areas, since they presented higher odds of being overweight and obesity in 2011. A possible hypothesis is that girls, compared to boys, might be starting to become more susceptible to lifestyle changes accompanying modernization and rural development, resulting in social, cultural, and economic transformations, including changes in eating patterns 34 .
For girls, income does not seem to exert as much influence as there was a greater chance of being overweight and obese in 2011 for the low-and high-income group. We highlight that, for both genders in 2011, adolescents with higher gross monthly household income had higher odds of being overweight and obese compared to 2001. Similar results for the family income pattern were observed in other studies 27,30 . In Brazil, between 1974 and 2009, obesity prevalence increased (from 0.1% to 2.1%) among adolescents of both genders across all income strata, but with greater magnitude among those with higher income (from 1.2% to 9.2%) 35 .
With regards to family income, the level of development of the country has to be considered. In Canada 24 , United States 36 , Australia 37 and Europe 38 for instance, the prevalence and the odds of obesity are increasing more among low-income adolescents over the years. In California, United States, for instance, the income disparities showed that the prevalence in those from low income raised more than double, when compared with those from higher income (2001: 17% vs 10% and 2007: 23% vs8%, respectively) 39 . In these countries, low-income families have had their purchasing power increased, having more access to video games, computer, and ultra-processed foods 40,41 , in addition, high-income adolescents have more options for caring for and preventing obesity 39 . However, in emerging countries like India, the prevalence of overweight and obesity increased significantly (2006-2009) among high-income adolescents 30 . These differences point to a possible reduction in economic disparity associated with the development and increase in obesity prevalence.
Obesity is complex because it is influenced by physical, socio-cultural and environmental issues. Therefore, it is fundamental to understand the characteristics of the population that present the greatest risks to be able to intervene in the best way. Thus, from the trend analysis of overweight and obesity in children and adolescents, in different strata, it is possible to direct future public policies for the subpopulations that present greater chances of developing this unfavorable aspect to health, such as people that live in urban areas and with high gross monthly household income.
Our study has some limitations, such as the minimum wage difference between the two surveys; also, it is important to emphasize that the variable gross monthly household income was reported by the students, and the variability between those values and the real values is unknown. The area of residence was also reported by the adolescents, resulting in possible mistakes due to the lack of knowledge about the criteria that should be used to classify area of residence. The use of self-reported weight and height to cal-culate the BMI of the adolescents can be considered another limitation, because it may result in response bias, since girls tend to underestimate their body mass while boys tend to overestimate it, due to cultural patterns. However, this measure shows validity when used with children and adolescents 42,43 . Besides that, self-reported family income also can be considered a limitation, because it is possible to be difficult for the adolescent to know the exact information. On the other hand, the high response rate in both surveys (2001 and 2011) and the representativeness of the sample are positive aspects, as well as the fact that the study has shown differences between overweight and obesity stratified by gender and discussing other sociodemographic factors such as area of residence and gross monthly household income.

Conclusions
We conclude that there was an increase in the prevalence of overweight and obesity among students in Santa Catarina State over a decade. Among the sociodemographic factors investigated, boys and girls living in urban areas and with high gross monthly household income presented, in the 2011 survey, higher odds of being overweight and obese compared to those in the 2001 survey. This information may foster the development of current policies, as well as the development of future intervention programs to prevent and decrease overweight and obesity in adolescents. Since the trends were different in the subgroups studied, specific interventions for each context are essential to increase the effectiveness of actions.  Overweight Obesity