Estado nutricional de menores de 5 anos de idade no Brasil : evidências da polarização epidemiológica nutricional Nutritional status of children under 5 years of age in Brazil : evidence of nutritional epidemiological polarisation

1 Programa de PósGraduação em Saúde Coletiva, Universidade Federal do Rio Grande do Norte (UFRN). Av. Senador Salgado Filho 1787, Lagoa Nova. 59056-000 Natal RN Brasil. ingrid_frc@hotmail.com 2 Departamento de Ciências Atmosféricas e Climáticas, UFRN. Natal RN Brasil. 3 Departamento de Nutrição, UFRN. Natal RN Brasil. Estado nutricional de menores de 5 anos de idade no Brasil: evidências da polarização epidemiológica nutricional


Introduction
Since the second half of the twentieth century, Brazil has experienced significant demographic changes 1 and changes to its populational morbidity-mortality and disability profiles.However, in contrast to developed countries, the Brazilian epidemiological transition has been marked by the simultaneous existence of high rates of morbidity and mortality from chronic non-communicable diseases (CNCDs) and the permanence or resurgence of infectious and parasitic diseases 2 .
Changes in the demographic and epidemiological profiles of the population are reflected by a decline in the prevalence of malnutrition and the significant prevalence of overweight/obesity, characterising the progression known as the nutritional transition 3 .Between 1974 and 2003, the nutritional epidemiological panorama in Brazil changed significantly, with a notable 72% cumulative decline in stunting in children under 5 4 .Childhood obesity remained stable in the periods between 1974-1975 and 2006-2007, with percentages of approximately 6 to 7%.However, this trend does not apply to other populational age groups, such as adolescents and adults, in whom the prevalence of obesity increased substantially over the same period 5 .
Globally, in 2011, approximately 101 million children under 5 years of age were underweight, whereas in 2013, an estimated 42 million children worldwide (6.3%) in that same age group were overweight.Malnutrition is linked to more than 1/3 of all deaths in children worldwide, although it is rarely listed as a direct cause.Obesity, in turn, is considered 1 of the 4 principal risk factors for CNCDs, and it is more damaging the earlier it occurs.Therefore, it is important to study children's nutritional status, given the demonstrated association of nutritional conditions in the infant-juvenile phase with health levels in adulthood 6,7 .
Children's nutritional status is considered an important tool in gauging a population's health conditions and quality of life 8 .Considering its complex and multifactorial character 9 , a child's nutritional status is determined by the population's living conditions, particularly regarding social and economic aspects 10,11 .
By contrast, children's nutritional conditions are associated with short-and long-term individual and collective consequences, including lower height in adulthood, lower school performance, increased morbidity and mortality, reduced productivity in adulthood and risk of chronic dis-ease.Child growth patterns are, therefore, strong predictors of future human capital, social progress and the health of future generations 12,13 .
It is essential to know how malnutrition, obesity and their intervening factors affect the population so that health care models can be developed that are based on the individual as a whole and the environment in which she/he lives.Against this backdrop, the objective of this study was to evaluate the nutritional status of children under 5 years of age in Brazil in 2009 and to identify discrepancies and peculiarities regarding social and demographic aspects.

Methods
This was a cross-sectional population-based study, in which the population of interest was children under 5 years of age who participated in the 2008-2009 Household Budget Survey (Pesquisa de Orçamento Familiar -POF) of the Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística -IBGE); in total, 14,569 children were included.
In planning the POF sample, complex sampling procedures were employed involving geographical stratification and statistics derived from the collection of national census tracts, the random selection of groups of sectors within strata and the random selection of households within sectors.Following the selection of sectors and households, the sectors were distributed over the 4 survey quarters, ensuring that in all quarters, geographical and socioeconomic strata were represented by the selected households 14 .
Each household belonging to the POF sample represents a certain number of permanent private households of the population (universe) from which this sample was selected.In this manner, each household in the sample is associated with a sample weight or expansion factor, which, when attributed to the characteristics investigated in the study, provides estimates for the values of interest in the study universe.To calculate estimates for the variables of interest, the adjusted expansion factors provided with the survey data by the IBGE were used in this study.
Weight and height (length and height) data were subjected by the IBGE itself, to a critique and imputation treatment, in which the imputation method was used to treat non-responses and response errors associated with values rejected at the critical stage.Pre-imputed weight and height variable data were used in the present study.
The weight, height, age and gender variables were used to classify nutritional status.These were processed using the WHO Anthro 15 software to obtain the z-scores for each child, adopting the growth curves proposed by the World Health Organisation (WHO) 16 as a reference.Nutritional diagnosis was determined from 3 anthropometric indices: height-for-age (H/A), weight-for-age (W/A) and weight-for-height (W/H).
Using the z-score statistical criterion and the classification recommended by the Food and Nutrition Surveillance System (Sistema de Vigilância Alimentar e Nutricional -SISVAN) 17 , children whose H/A, W/A and W/H indices had z-scores lower than -2.0 were classified as stunted, underweight or thin, respectively; children with W/H index z-scores greater than or equal to -2.0 and less than or equal to +2.0 were classified as eutrophic.The W/H ratio was also used to rate children whose index z-score value was greater than +2.0 and less than or equal to +3.0 as overweight and children with z-scores greater than +3.0 as obese.
If at least 1 of the 3 indices used in this study had values considered biologically implausible (outliers) for a child, this was considered an exclusion criterion according to the method proposed by the WHO Anthro 15 software.The following cut-off points were used for exclusion: z < -6 or > 5 for the W/A index; z < -5 or > 5 for the W/H index; and z < -6 or > 6 for the H/A index.A total of 14,013 of the 14,569 children were thus included in this study, representing a loss of 3.8%.
To classify the monthly per capita family income, the value expressed in the consumer unit per capita income variable was divided by the value corresponding to 1 minimum salary on the survey reference date (15 January 2009), which was R$ 415.00.
The Statistical Package for Social Science (SPSS), version 20, was used for data analysis.Pearson's chi-square test and logistic regression models were used in the first stage of the study to evaluate associations between the dependent variables W/A, H/A and W/H and the explanatory variables represented by social and demographic information (age, gender, colour/ race, home location, monthly family per capita income and macro-region).A binary logistic regression was used for the W/A and H/A variables, and a multinomial logistic regression was used for the W/H variable, with eutrophic as a reference category.
First, the chi-square test was applied to identify which explanatory variables were significantly associated with the response variables.The variables that were statistically significant at 20% were then included in the logistic regression model, with those that were significant after the stepwise selection method remaining in the final model, when the odds ratio values were estimated with their confidence intervals.The level of statistical significance adopted in the logistic regression analysis was 5% (p < 0.05).
In the second stage of the study, sets of relationships were explored by applying a correspondence analysis between socioeconomic and demographic factors and the child's nutritional status, based on the categorised W/H variable.The software XLSTAT was used for this analysis.
A correspondence analysis is an exploratory statistical technique used to examine associations or similarities between qualitative or categorical variables.Using a graphical representation, the spatial positions of the categories of each variable on a multi-dimensional map can be interpreted as associations, in which variables perceived as similar or associated are allocated to closer points on the map, and those perceived as not similar are represented by distant points 18 .
The application of a correspondence analysis is very appropriate in population data studies and cross-sectional designs because it allows the association between variables to be explored without assuming a causal relationship between them and without assuming a probability distribution 19 .
Weight deficits, measured by the W/A ratio, and thinness, defined by the W/H index, were more prevalent in children under 1 year of age, at 3.6% and 8.9%, respectively.Stunting was higher among children 3 of years of age (12.0%).For all analysed indices, the monthly per capita income variable showed deficit prevalence levels that were inversely proportional to the income increase (Table 1).
The binary logistic regression model (Table 2) revealed that all colour/race categories, except for black in the case of stunting and Asian for both deficits (weight and height), had greater odds of developing nutritional deficits than Caucasian children.Those most likely to present some deficit were indigenous children, whose re-sult for stunting was almost double that of Caucasians (OR = 1.754, 95% CI 1.714 to 1.794).
Children of 4 years of age had a reduced chance of having deficits compared to children of all other ages, particularly those younger than 1 year of age for a weight deficit and those 3 years of age for a height deficit.The income variable continued to demonstrate an inverse association with weight and height deficits, indicating that the lower the income level, the greater the chance of the child developing a nutritional deficiency.This association was almost 10 times higher for the group of up to ¼ minimum salary regarding stunting (OR = 9.568, 95% CI 9.106 to 10.052) and almost 4 times higher regarding low weight (OR = 3.493, 95% CI 3.437 to 3.549) (Table 2).
Children of any other Brazilian region were more likely to have deficits than those living in the South region, with special emphasis on the North, where the chances are almost double compared to the South (low weight: OR = 1.908; 95% CI 1.876 to 1.941; stunting: OR = 1.901, 95% CI 1.886 to 1.917) (Table 2).
The multinomial logistic regression model (Table 3) revealed that males of all age groups had higher odds of developing overweight and obesity than 4-year-old children, except those 3 years old for obesity.Those from urban areas also had higher odds of developing overweight and obesity than those from rural areas.All other geographical regions had lower odds of becoming overweight and higher odds of being obese than those from the South.
Because the Asian race category did not have children with obesity, it was grouped with the mixed-race category in the multinomial logistic regression model, given the similarity between the 2 regarding children's nutritional status patterns.After this data grouping adjustment, all other categories had lower odds of becoming overweight and obese than Caucasians, except obesity for indigenous children (OR = 1.893, 95% CI 1.836 to 1.952) (Table 3).
The results of the correspondence analysis (Figure 1) were in line with those found when the logistic regression models were applied.For the thinness classification (W/H deficit), there was a stronger relationship with children from the North and Northeast regions, those belonging to families with lower levels of per capita income (up to ¼ minimum salary and ¼ to ½ minimum salary) and those of black colour/race.
In the upper left quadrant of Figure 1, overweight appears to be more closely related to the children from the South, Southeast and Central-West regions, to males, to those residing in the country's urban strata, to Caucasians, to those aged 3 and to those belonging to families with intermediate per capita income levels (½ to 1 minimum salary and from 1 to 5 minimum salaries).
The positions of children under 1 year of age and indigenous children should also be noted.They are located between obesity and thinness, showing a high percentage of these 2 deficits in children belonging to these social groups.
Figure 2 shows the correlation analysis between the W/H index and Federal Units.The states that were most closely associated with thinness were Roraima, Maranhão, Amapá, Amazonas, Alagoas, Pernambuco and the Federal District.These associations corroborate the above findings; notably, most of these states are located in the North and Northeast regions.The states most closely related to excess weight were Santa Catarina, Paraná, São Paulo, Rio Grande do Sul, Espírito Santo, Mato Grosso do Sul, Paraíba, Minas Gerais, Goiás and Ceará.Notably, most of the Federal Units that were associated with excess weight are located in the Southeast, South and Central-West, with the exception of 2 states in the Northeast.The states associated most closely with obesity were Rio de Janeiro, Mato Grosso, Ceará, Alagoas, Amazonas and Amapá.

Discussion
The main results of this study demonstrate that children's nutritional deficits are highly linked to the socioeconomic conditions in which they live.By contrast, the overweight and obesity distributions had a more complex character, demonstrating heterogeneous prevalences in widely different population groups.
A study evaluating the trend of linear growth retardation in children younger than 5 years of age in Brazil, Batista Filho and Rissin 5 demonstrated a sharp decline in stunting in 1975, 1989 and 1996, although there was a higher prevalence in the rural stratum than in the urban stratum and in the North and Northeast regions than in the Central-South.In this study, the North presented much greater weight and height deficit levels than other regions.However, regarding stunting, the Northeast region had a percentage similar to those recorded in the Southeast and Central-West, demonstrating that the Northeast has been able to improve child nutrition levels, albeit more slowly than the macro-regions situated in the Central-South of the country.
Monteiro et al. 20 corroborated the findings of this study when identifying the high risk of exposure to malnutrition in the under 5 population in the North.In other macro-regions of the country, the value was moderate and very homogenous.They also noted the substantial reduction in the risk of child malnutrition, particularly in the Northeast, between 1996 and 2006.
While acknowledging that many factors affect nutritional status, and more specifically infant growth, the effect of conditional cash transfer programmes has been discussed, given that a considerable proportion of the funds transferred is destined for the purchase of food and to meet health care demands, due to conditions existing in these programs 21 .
Monteiro et al. 22 evaluated the effects of the Bolsa Família cash transfer programme on child malnutrition using data collected in the "Chamada Nutricional 2005" survey conducted in the Brazilian semi-arid region.They used adjusted prevalence estimates of H/A deficits and found that for children under 5 years of age, as a whole, participation in the programme resulted in a reduction of almost 30% in the frequency of malnutrition.
Specifically, in low-income families, the effects of the Bolsa Família Program, for example, have translated into higher household spending on food, greater availability of fresh or minimally processed foods and a greater availability of foods that usually diversify and improve the nutritional quality of a diet, such as meat, tubers and vegetables 23 .The repercussions of these programmes may have positively affected the nutritional status profile of Northeastern children, given that this region has the largest number of beneficiary families in the programme.
However, this greater availability of food may affect children's weight gain in an inappropriate manner.The tendency observed in Brazil to replace traditional meals based on fresh or minimally processed foods by ultraprocessed foods is causing harm to health.Ultraprocessed foods have higher energy density and higher sugar, to-tal fat, saturated fat and trans-fat contents and lower fibre and potassium contents 24 .This discussion highlights an emerging challenge for public health: to promote educational interventions in health, more specifically in Food

Mixed race
Symmetrical plot (axes F1 and F2: 93.27%) and Nutrition Education, through viewing eating behaviour as the result of social and historical relationships and thus overcoming its biomedical roots 25 and promoting the autonomous and voluntary practice of healthy eating habits 26 .
The effect of socioeconomic status on infant growth has been observed in different contexts in the Brazilian population and is already significantly represented in the literature, which confirms this relationship 11,[27][28][29][30] .Family income has a frequent significant inverse association with nutritional deficit situations.Using data from the National Demographic and Health Survey (Pesquisa Nacional de Demografia e Saúde -PNDS) from 1996 and 2006, Monteiro et al. 28 identified improvements in household purchasing power as a determinant responsible for 21.7% of the total decline in the prevalence of malnutrition in children under 5 years of age observed in Brazil between 1996 and 2006.
The data from this study revealed a higher prevalence of nutritional deficits in male children than in female children.Other studies with more specific aggregation levels obtained similar results 27,31 .One study suggests that boys appear to be more vulnerable to malnutrition in an environment unfavourable to growth 31 .
Regarding the high prevalence of weight and particularly height deficits in indigenous children, as in this study, research conducted in various indigenous villages and communities across the country has reported results that reveal prevalences of stunting, in particular, that are well above the national average 32,33 .However, although reports of malnutrition, in general, characterise children's anthropometric profiles among indigenous peoples, the increase in children with obesity in this population merits particular attention.Evaluating the nutritional status of indigenous children from Alto Xingu, and in a study of the nutritional profile of indigenous Kaingáng children, Mondini et al. 34 and Kuhl et al. 35 , respectively, also found a marked percentage of overweight, similar to that observed in this study.The authors also noted that these findings are not isolated cases in the indigenous context and note that recent studies have identified cases of overweight amongst children from different indigenous groups.
Westernisation of the diet, marked by the replacement of traditional eating habits, and the incorporation of practices from the urban population, particularly in relation to processed foods, along with a reduction in physical activity levels, may represent changes in the lifestyle of indige-nous peoples indicative of a food and nutritional transition, possibly resulting in significant repercussions for the epidemiological profile of this population in the future 36,37 .
This study also shows that with regard to age group, the prevalence of a deficit was higher in children under 1 compared to almost all other age groups except for children aged between 1 and 3 for the (H/A) index.The onset of malnutrition usually occurs between the fourth and sixth month of a child's life, when the transition to complementary foods may be inadequate regarding quality and quantity, and also exposes the child to infections, particularly diarrhoea, making this age group increasingly vulnerable 38,39 .
The greatest weight and height growth rates occur at 2 stages of an individual's life: in the first 2 years of age and in adolescence; the former is the period most vulnerable to growth disorders 40 .When a child of this age group develops a health problem, it can compromise both weight and height.Children can recover from a weight deficit fully, with no harm to the weight expected according to the individual's genetic potential.The full recovery of a height deficit is more complex and may even lead to significant consequences in later life, such as the onset of chronic diseases in adulthood 41,42 .
An important finding of this study was the manner in which overweight and obesity appeared to be distributed across diverse population groups, demonstrating their heterogeneous and complex character.A study based on PNDS -2006 data also revealed higher prevalences of overweight in children under 5 in all Brazilian regions, with the highest percentages in the South (9.3%) and the lowest percentages in the North region (5.1%) 43 .
Additionally, regarding spatial distribution, the urban stratum demonstrated a significantly higher prevalence of overweight than the rural stratum.This pattern may be particularly attributed to differences in access to food, health services, physical activity patterns and social norms between these strata 44 .
The heterogeneous pattern of overweight and obesity among the various regions and population groups in Brazil reflects the enormous physical, socioeconomic and cultural diversity of the country.This heterogeneous pattern makes deconstructing the idea of obesity as a disease of developed countries or socially advantaged groups important 45 .
In this study, obesity was associated with Brazilian states in which child malnutrition was once the only nutritional disorder.States in the North and Northeast known to have less economic power, such as Alagoas, Ceará, Amazonas and Amapá, were associated with this condition.
The increased obesity in groups with lower income levels appears to reflect a lack of knowledge regarding obesity-related damage, as well as unequal access to physical activity and proper nutrition regarding both quality and quantity.Such inequalities impose a pattern of poor nutrition on low income groups.It is increasingly common to observe adherence to a healthier diet and the practice of physical activity in more privileged social groups 45.Furthermore, although a decline in malnutrition is evident, it remains prevalent in Brazil in the most severe forms, particularly regarding H/A deficits.These are more serious in the North but also appear in pockets of poverty in other regions, characterising malnutrition as a result of social inequality and poverty in the country 46 .
Furthermore, regarding stunting, which reflects the chronicity of malnutrition, a higher percentage was found than that expected in a healthy reference population, i.e., 2.3%, in all sub-categories analysed in this study 17 .
The persistence of this deficit in the Brazilian population, coexisting with the growing and widespread distribution of obesity, speaks to the phenomenon of epidemiological polarisation and indicates that, although occurring at an accelerated pace, the nutritional transition process in Brazil is not yet complete.It coexists with the reduction in malnutrition associated with a decrease in the occurrence of communicable diseases, characterising the third phase of nutritional transition, known as "hunger reduction."However, concomitantly, there is an increase in obesity and related CNCDs, both in low-income populations and in those with greater purchasing power, characterising the fourth phase of the nutritional transition 47 .
A possible limitation of this study is the percentage (3.8%) of children excluded from the sample because they had biologically implausible z-score values (outliers).The WHO recognises anthropometric surveys in which the proportion of biologically implausible values does not exceed 1% of individuals studied to be good quality 48 .However, despite this limitation, the importance of this study must be stressed, given the reduced number of studies using population databases as comprehensive as the POF, which allows the investigation of anthropometric and nutritional aspects at a territorial level and that are not restricted to local realities or specific groups.Moreover, there was no significant selective loss related to the explanatory variables, which allows making statistical inferences without losing test power.
The results of this study show that in parallel to the phenomenon of epidemiological polarisation, Brazil is experiencing a nutritional epidemiological polarisation, marked by the persistence of malnutrition in certain regions and population groups and the diffuse distribution of overweight, which is indiscriminately reaching the most diverse population groups.In this sense, this study presents a dualised agenda for public health, which must seek to reduce nutritional deficiencies and thus malnutrition, but which concentrates on food and nutrition education as a strategy to promote healthy eating habits from childhood, thus changing the nation's nutritional and epidemiological profile and mortality.
Furthermore, surveys to investigate nutritional status at the population level should be conducted to ensure regular monitoring and to identify the most vulnerable population groups, both regarding nutritional deficit or excess.The use of innovative methodological treatments also enables new perspectives to be developed regarding factors associated with the nutritional profiles found.The systematisation of this more frequent and specific monitoring will provide a stronger foundation for creating health policies regarding food and nutrition.

Figure 2 .Figure 1 .F1
Figure 2. Correspondence analysis applied to the weight-for-height index and Federal Units in children under 5 years of age.Brazil, 2009.

Table 1 .
Prevalences of weight and height deficits, thinness, overweight and obesity in children under 5 years of age.Brazil, 2009.

Table 2 .
Odds ratios (ORs) and confidence intervals (95% CIs) of weight/height deficits in the binary logistic regression model with weight-for-age and height-for-age indices as dependent variables.Brazil, 2009.

Table 3 .
Odds ratios (ORs) and confidence intervals (95% CIs) of the multinomial logistic regression model with weight-for-height index as the dependent variable (reference category = eutrophic).Brazil, 2009.