Conversely, a different picture emerges when it comes to overweight, a condition associated with excess nutrient and energy intake relative to expenditure. Overweight and obesity increase the risk of a number of chronic diseases, which can lead to reduced quality of life and premature death [8,9]. In addition, there is evidence of tracking of overweight, with several studies showing that overweight children tend to remain so during adolescence [10,11].

The association between socioeconomic position (SEP) and nutritional status is well established, but it can go in different directions according to the level of socioeconomic development of the population. It also varies by gender. McLaren [12] analyzed the socioeconomic patterning of weight across societies in various stages of socioeconomic development. In general, both men and women from the lowest SEP have a higher likelihood of obesity in developed countries while in low and middle-income countries higher SEP is frequently associated with excessive weight [12,13]. Similar associations have been reported for children [14]. In high-income populations, child obesity prevalence is directly related to poverty [15,16]. The association between SEP and overweight/obesity among children from low and middle-income countries is less well studied. Wang [17] showed that while in China high socioeconomic children were at high risk of obesity, in Russia, both low-income and high-income groups were at an increased risk of obesity compared to the medium-income group. However, recent findings from Hong Kongs population-representative “Children of 1997” birth cohort found no consistent association between family or neighborhood SEP and childhood adiposity [18].

These three studies offer a unique opportunity for documenting the nutrition transition in a population from the same location and which belong to the same ethnic background. The comparisons rely on uniform modes of data collection (prospective information obtained among large population-based samples) combined with the use of similar questionnaires, standardised anthropometric measurements performed by highly trained fieldworkers, high follow-up rates and low frequencies of missing data (below 5%) for nutritional variables in the three birth cohort studies. Another advantage of our study over those using cross-sectional data is that it was possible to examine both cohort effects – i.e. changes in stunting and overweight over time comparing the same age group across cohorts, and age effects – i.e. comparing across ages within each cohort.

The one-year follow-up in the 1982 cohort, and the one and four-year follow-ups in 1993 cohort were based on subsamples rather than the whole cohorts. The 1993 subsamples were similar to the rest of the cohort in terms of socioeconomic position and maternal characteristics. Nevertheless, due to over-representation of low birthweight babies in the lower income groups we needed to weight all analyses of the 1993 cohort subsamples. In the 1982 cohort, children not included in the one-year subsample were born to poorer and less educated mothers than those followed up. It is unlikely that attrition or use of subsamples have introduced important biases in the present results.

There is debate about the accuracy of BMI as a measure for assessing adiposity in individual children [29]; however, it is widely accepted as a population estimate of obesity risk [29,30]. We opted to use BMI-for-age indicator to estimate child overweight in our population, following World Health Organizations (WHO) recommendations [23].

Levels and trends in stunting and overweight in Pelotas have to be placed in the international context. In South America, Guatemala, Peru, Honduras and Bolivia have the highest prevalence of stunting (54%, 30%, 29% and 22%, respectively). Brazils prevalence of stunting (7%) is almost the same as found in Argentina (8%), but higher than that of Chile (1%) [7]. In the 2004 cohort, stunting prevalence ranged from 4-6% depending on age. Most developing countries showed declines in the last decades [7]. In South America, overall stunting levels fell from 20.9% in 1990 to 13.1% in 2007 [31]. The decline in Brazil has been particularly rapid (from a prevalence of 37.1% in 1974–75 to 7.1% in 2006–7) [32], and our current findings are consistent with those of Brazil as whole, although the local prevalence remained below the national average.

Childhood overweight and obesity are assessed through different definitions, by different authors. However, studies are consistent in showing that their prevalence is increasing both in high-income [33] and low and middle-income countries [1,7]. Popkin et al[34]. compared trends in overweight prevalence (International Obesity Task Force, IOTF, criteria) among adults and children above five years of age, showing that in many countries - including the US, the UK, Australia, China and Brazil - child overweight is increasing at a faster rate than adult obesity. Wang & Lobstein [35] estimated that over 46% of school-age children will be overweight (IOTF criteria), and about one in seven children will be obese (IOTF criteria) in the Americas by 2010. In contrast, an analysis of trends in Brazil [32] did not find an increase in overweight among all under-five children over a 33-year period (proportions of under-five children whose weight-for-height was ≥2 z scores were 8.4% in 1989, 6.6% in 1996 and 7.3% in 2006–7). Because these national studies were cross-sectional and included children in broad age ranges (under three or under five years of age) they were unable to describe age patterns in the nutrition transition process. For example, we showed that overweight (percent of children with more than 2 z-scores of BMI-for-age) increased primarily among four year olds, but not among younger children. The national, cross-sectional analyses from Brazil may have missed this increase, because the increase in overweight is only happening for older under-fives, and these children only account for a small fraction of all under-fives.

Previous research in the social patterning of overweight in the 1982 and 1993 Pelotas cohort studies showed that among 11-year-old boys and girls there was a strong direct association between SEP and overweight. In the 1982 cohort, overweight at 18years of age showed a positive association with SEP for males, and an inverse association among females [36]. This is in agreement with studies of Brazilian adults, where wealthy men tend to be fatter, and wealthy women thinner, than the rest of the population.

Few studies from low and middle-income countries have analyzed socioeconomic inequalities in childhood stunting or overweight status over time, and all of these have relied upon repeated cross-sectional designs. Analyses of national surveys in Mexico [37] and Brazil [32] showed a steady decrease in the rates of stunting for under-five children followed by a decrease in both absolute and relative socioeconomic inequalities. In India, even though in the 1992–2005 period there was a decrease in the rates of stunting among children less than three years of age, relative social disparities either widened or remained stable over time [38]. An analysis of 47 DHSs (Demographic and Health Surveys) that contained information on the nutritional status of children aged up to five years of four regions (sub-Saharan Africa, eastern Mediterranean, south and south-east Asia and Latin America and the Caribbean) showed that in almost all countries stunting disproportionately affected the poor, however, Latin America and the Caribbean region showed the largest socioeconomic inequalities [39]. Investigators identified three patterns of socioeconomic inequalities in stunting: “mass deprivation”, where stunting is highly prevalent within the majority of population while a small privileged class is much better off; “exclusion”, where the prevalence of stunting is relatively low in the majority of the population, but was much higher among the poor, and a third pattern, called “queuing” which shows an intermediate situation. Our results showed that socioeconomic inequalities in stunting in the Pelotas cohort studies have been showing the “exclusion” pattern since 1982. Hence programmes targeted at the poorest people are needed to continue reducing inequalities.

Less has been written about overweight/obesity trend inequalities among children from low and middle-income countries. A recent review showed that the prevalence of childhood overweight (percent of children with more than 2 z-scores of weight-for-height) tended to be higher among the rich than among the poor in Ghana, Sierra Leone, Tajikistan and the Dominican Republic [40]. Similar results were found among Mexican school-aged children [41]. In the same way, our results showed higher prevalence of overweight among the highest SEP group in almost every follow-up in the three cohort studies.

In the 22-year period between 1982 and 2004, there were major political and economic changes in Brazil. A remarkable change, in the field of politics, was the end of the military dictatorship in 1985, which was followed by periods of financial chaos until inflation was finally controlled in the early 1990s. The country experienced economic growth since 2000 and several programs targeting the poorest population groups were implemented. Inequalities in income distribution in the country persisted between 1982 and 1993, with the Gini income distribution index remaining at 0.60. However, a reduction to 0.57 was recorded in 2004, with a further drop to 0.55 by 2008 [42]. Brazil dropped from being the country with the highest income concentration in the world during the 1980s to a ranking of 14. All of these changes must have affected the health and nutrition situation of young children [43].

In spite of economic improvement in the country as a whole, the Pelotas region had slower growth than the rest of the country. In 1980, the per capita gross domestic product (GDP) of the city was equivalent to 81.4% of the average value for the state of Rio Grande do Sul, where it is located, and 93.6% of the average value for Brazil as a whole. In 2002, the per capita GDP of the city fell to 58% and 73.5% of the average value for the state of Rio Grande do Sul and Brazil, respectively [44].

Nevertheless, substantial improvements in maternal health and education - including decreased fertility, increased birth intervals, declining prevalence of smoking during pregnancy and higher levels of schooling – were observed in Pelotas during the course of the study period [45]. In contrast, there were negative changes in birth outcomes (the mean birthweight decreased by 37g and preterm births increased, from 6.3% in 1982 to 14.7% in 2004) [46] which may explain the higher prevalence of stunting at birth in 2004 relative to 1993. Possible explanations for the negative trends in birth outcomes, such as inadequate quality of antenatal care and increased medicalisation of pregnancy and childbirth, including labour induction, caesarean sections and inaccurate ultrasound scans, are provided elsewhere [45].

The 1982, 1993 and 2004 Pelotas birth cohort studies are currently supported by the Wellcome Trust Initiative entitled Major Awards for Latin America on Health Consequences of Population Change. Previous phases of these studies were supported by the International Development Research Center, The World Health Organization, Overseas Development Administration, European Union, National Support Program for Centers of Excellence (PRONEX), the Brazilian National Research Council (CNPq), the Brazilian Ministry of Health and the “Pastoral da Criança” (Catholic non-governmental organization, Curitiba, Brazil).