Trends in body mass index values of Brazilian enlisted men , 1980 to 2005

This study was to investigate the temporal evolution of body mass index (BMI) in young enlisted men of 18 years in Brazil between 1980 and 2005, to identify specific points of greatest variance in time series and compare specific points in time, and the temporal evolution of BMI with socioeconomic changes in Brazil. The present study explores a temporal series of twenty-six national surveys of Brazilian men enlisted between 1980 and 2005. Heteroscedasticity in BMI time series was tested using Engle’s Lagrange-multiplier (LM) test, and analyses were performed using the autoregressive conditional heteroscedasticity (ARCH) model. As possible explanations for these increases in mean BMI, changes in economic indicators were considered. Particularly in 1985 to 1986 and 1994 to 1995, there was a sharp and significant increase in BMI. These two points occurred after two major economic policy changes that increased the purchasing power of the population related to economics factors such as: reducing the level of social inequality, increased family income, poverty reduction, inflation control, and increased consumption of foods. The present study showed a sharp increase of obesity in the population of young men in Brazil on two occasions during this series (years 1985 to 1986 and 1994 to 19955), when a possible reduction in caloric expenditure and increased food consumption population were observed.


INTRODUCTION
Trends of increasing overweight and obesity are now well documented in both developed countries for example, United States (Komlos and Brabec, 2010;Burkhauser et al., 2009), France (de Saint Pol, 2009), Japan (Yoshiike et al., 2002) and developing countries for example, Brazil (Wang et al., 2002) and Mexico (Rivera et al., 2004) but few studies have used time series with more frequent data collection points on weight change, which would allow testing, whether specific points in time are associated with more substantial increases in the rates of overweight and obesity.
Recent data from the Brazilian Institute of Geography and Statistics (IBGE, 2010) reported that overweight has been increasing steadily in adults since the mid-1970s and it is currently found in approximately half of all Brazilians.Data from enlisted men have been used to describe the prevalence of overweight in Israel (Gross et al., 2009) and the United States (Hsu et al., 2007).These data are usually collected annually therefore they are a good resource for testing temporal changes, mainly when measured by professionals.
The objectives of this study are: (1) To evaluate the temporal change in BMI of enlisted men in Brazil from 1980 to 2005; (2) to test for specific points in time that display greater variance and; (3) to further compare specific points in time in the BMI temporal evolution with economic changes in Brazil.

Study population
The present study explores a temporal series of twenty-six national surveys of Brazilian men who enlisted between 1980 and 2005.Each survey comprises a 35 to 40% of all Brazilian men aged 18 years at the time of examination.The article 143 in the Constitution of the Federative Republic of Brazil (1998) provides that a one year military service is compulsory.Women and clergymen are exempted from compulsory military service.Weight and height were measured to the nearest 0.1 kg and 0.5 cm, respectively using a scale with incorporated stadiometer (Filizola ® ).All measurements were performed by previously trained examiners.Participants were weighed and measured wearing light clothing and no shoes.

Body mass index data analysis
The Brazilian candidates for compulsory military service are representative of the male population in general thus allowing one to assess the increase in the prevalence of obesity during the period of analysis and identify periods of change.Prevalence of overweight and obese men was calculated with 95% confidence intervals, and the BMI values were plotted to make a descriptive analysis of the time series.This analysis showed that the variance of the error term was not equal over time that is, heteroscedasticity.The cutoff points of the World Health Organization (WHO, 2007) were used to classify overweight (BMI ≥ 25 kg/m 2 and < 30 kg/m 2 ) and obese (BMI ≥ 30 kg/m 2 ) men.
Heteroscedasticity in BMI time series was tested using Engle's Lagrange-multiplier (LM) test, and analyses were performed using the autoregressive conditional heteroscedasticity (ARCH (1)) model (Engle, 1982;Bollerlev et al., 1994) with a level of significance set at p < 0.05.For those points in time with higher oscillations of the mean of BMI (1985BMI ( , 1994BMI ( and 2000)), dummy variables were included under the assumption that the growth rate of mean BMI was not the same throughout the period.
When heteroscedasticity is present, the regression coefficients for an ordinary least squares regression (OLS) are unbiased, but the variance and confidence intervals estimated by conventional procedures will be too narrow, giving a false sense of precision.In this context, an ARCH model, introduced by Engle (1982), is an appropriate framework for analyzing the data.All analyses were performed using the STATA.

Economic factors
As possible explanations for these increases in mean BMI, changes in economic indicators were considered (Brazilian Institute of Geography and Statistics and Institute of Applied Economic Research).The economics factors which have been analyzed were: annual inflation rate, food production, poverty (%), soft drinks consumption and average annual real income.

Annual inflation rate
The annual inflation rate was measured by the General Price Index, an index that attempts to reflect the monthly variations in prices (%).The index calculates the price variations of agricultural raw materials and wholesale industrial and final goods and services for consumption.Annual inflation was measured by the average of the 12 months of the year (IPEA, 2010).

Food production
Food production was measured by indicators for aggregated products that represent the sectors that provide supplies directly to livestock or delivering the first industrial processing of goods resulting from activities in the primary sector.The series consists of products and aggregates defined class-based agribusiness industry (year 1989 = 100) (ABIA, 2010).

Poverty
Poverty was measured by the percentage (%) of people in total population with per capita household income below poverty line.The poverty line is twice the extreme poverty line, an estimate of the value of a basic food basket with a minimum of calories needed to adequately sustain life, based on the recommendations of Food and Agriculture Organization and WHO (IPEA, 2009).

Soft drink consumption
The Soft drink consumption was measured by the annual wholesale value of soft drinks and juices in liters per year (l/y) (ABIR, 2009).

Average annual income
The series were calculated from all jobs of employed persons with employment income (IPEA, 2008).The annual income was measured by the average of the 12 months of the year and in national currency: Real (R$).The study was approved by the Research Ethics Committee of Rio de Janeiro State University, Institute of Social Medicine (protocol number 397523).

RESULTS
The study population of 11,090,230 males showed an increase in mean BMI from 21.4 kg/m 2 in 1980 to 22.2 kg/m 2 in 2005.The prevalence of overweight men changed from 4.5% in 1980 to 12.5% in 2005 (2.6 times larger).The prevalence of obesity increased from 0.5% in 1980 to 1.9% in 2005, an increase of almost 300% during the period; although these values are still at low level in international comparison.Table 1 shows increasing variance of mean BMI over time, indicating heteroscedasticity.
The LM test of heteroscedasticity on BMI showed a p-value  2 is related to the fit of the model.The coefficient L1 = 0.49 indicates that the variance of the current period error depends on information that is revealed in the preceding period.The years of change in BMI were accompanied by the usual changes in Brazilian´s lifestyle (1985 to 1986 and 1994 to 1995) but by international comparison they are below average (Figure 1).

DISCUSSION
Non-genetic factors appear to have a major role in rapidly increasing rates of obesity (Ogden et al., 2007;Saarloos et al., 2009) but there is less consensus on the specific environmental factors that may contribute to such dramatic shifts (Jeffrey and Utter, 2003).Although young males are not the group with the greatest prevalence of overweight in Brazil, overweight changed from 4.5% in 1980 to 12.5% in 2005, the prevalence in the overall population according to nationwide surveys in Brazil has increased substantially.Overweight and obesity combined Log annual inflation (%), food production, poverty (%), soft drink consumption (10 5 liters/year, body mass index (kg/m 2 ) and average annual income (R$) in Brazil, 1980 to 2005.The shaded zones illustrate that the periods of economic growth (1985 to 1986 and 1994 to 1995) and are corresponding in time with increased sharp of BMI.
among adults has increased from 18.5 to 50% in the last 30 years indicating that the obesity epidemic is happen also in Brazil (IBGE, 2010).
The data of enlisted men provides a practical and large convenience sample of the Brazilian young male population and possibly reflects general population trends.We identified turning points in the BMI curve in 1985 and 1994 and also identified changes in the Brazilian economy of possible factors are corresponding in time with increased BMI.Beginning in 1985, a marked change in general consumption was noted with an increasing amount of available calories (Silveira and Almeida, 2008) and also the launching of many new food products (Mendonça and Anjos, 2004).Beginning in 1994, a trend was observed towards increased consumption of fast foods (Mendonça and Anjos, 2004) and soft drinks (ABIR, 2009).
The decline in inflation rate and consequent reduction in food prices, mainly for industrially-produced food products, was observed in both periods and many authors have concluded that low food prices induce people to consume more calories (Mitra, 2001;Cutler et al., 2003;Chou et al., 2004).Glanz et al. (1998) found that cost is one of the most important factors in food-purchasing decisions and that these trends toward increased consumption and reduced industrialized food prices (for example, soft drinks and vegetable oils) have also been observed in the past three decades in other developing countries (Popkin, 1993).
An increase in wages mainly among the poor (Lavinas, 2001;IBGE, 2007) and the reduction of inequality (Hoffman, 2002) were also a result of the 1994 real plan in Brazil, a major economic intervention by the government, which increased food consumption in res-taurants and fast food establishments (Mendonça and Anjos, 2004).Eating outside the home has been associated with obesity in studies conducted in the United States (Guthrie et al., 2002) and Europe (Orfanos et al., 2007) and in Brazil eating outside the home has been associated with obesity among men (Bezerra and Sichieri, 2009).
In the United States, increased rates of obesity have coincided with a large increase in average caloric intake between 1985 and 2000 when more fats were added to homemade foods and many processed foods (Chastenet, 2011) and in Brazil, an increase in processed foods and sodas has been well documented (Levy-Costa et al., 2005).Watching television has also received great attention for its contribution to a sedentary lifestyle (Finkeltein et al., 2005).Time spent on computers, video games and other electronic resources have increased substantially since the 1980's in Brazil.Furthermore, an increase in TV sales was observed in the period 1994 to 1995 (Mendonça and Anjos, 2004).

Conclusion
Finally, the results of this study show that between 1980 Much has yet to be investigated on the influence of economic factors on the development of obesity and although time series studies are not considered the best framework for causal analysis, this methodology clearly allows the identification of major changes over time, as shown in our analyses.Because of their public health importance, trends in overweight and obesity need to be monitored continuously and factors associated with these trends studied, in order to develop sound health-policy strategies to improve health and prevent obesity.
Figure1.Log annual inflation (%), food production, poverty (%), soft drink consumption (10 5 liters/year, body mass index (kg/m 2 ) and average annual income (R$) in Brazil, 1980 to 2005.The shaded zones illustrate that the periods of economic growth(1985 to 1986 and 1994 to 1995)  and are corresponding in time with increased sharp of BMI.

Table 1 .
Mean, standard deviation (SD) and variance of body mass index (BMI) and the prevalence of overweight and obesity (%) and the 95% confidence interval (CI) in Brazilian enlisted men, 1980 to 2005.

Table 2 .
ARCH model for BMI.Brazilian enlisted men, 1980 to 2005.we identified that two points of rate of change in BMI are corresponding in time with economic changes that altered Brazilians' lifestyles.