The aim of this study was to determine the rate of increase in prevalence of overweight and obesity over a 3-year-period in pre-pubertal South African children. This is a first study in SA, a developing country, to provide prevalence statistics of childhood obesity obtained by follow-up measurements in pre-pubertal children, aged 6 to 9-years. 12.7% of the group were OW or OB in 2010, compared to 16.7% of the same group in 2013. The rate of increase in the group was 4.0% over the 3-year period. Increases in the group were similar in boys and girls, while white children had much higher increases compared to black children as a group, and higher SES (which included white and black children) also contributed to higher prevalences and rates of increase. A different picture of the extent of the problem emerged, when interactions of race and SES were considered, than when the participants were analysed as a group. White children had a prevalence increase of 7.0% (20.3% to 27.3%), which was double compared to black children where the increase was 3.0% (10.3% to 13.3%). However, in Q4 and Q5 schools black children showed higher increases in combined OW/OB compared to white children (8.2% vs 5.2% Q4, 9.4% vs 7.7% Q5, Table 1), which was also much higher than the increase for black children when analysed as a group (3%) or for Q1–Q3 schools were only black children were enrolled in these schools. Overall Q4 and Q5 schools which represent children from more affluent families and environments showed much bigger increases in prevalence compared to Q1 to Q3 schools which can be ascribed to improved living conditions. The main effects of SES in 2013 were significant for all variables, while sex and race were only significant for fat percentage. The Quintile*race interaction effect for BMI showed significantly higher BMI values among black children compared to white children in Q4 schools. The higher BMI of black children compared to white children in higher SES, can in part be ascribed to the economic transition of black families in South Africa, although a longitudinal study in America over a period of 17-years of racial differences [30] in 5–14 year old children, also confirmed contrasting patterns of increase in BMI between white and black children. Annual increases in BMI varied in this study from 0.60 kg/² per year in white girls to 0.78 kg/² per year in black girls and yearly increases in BMI before age 18 were 25% to 55% higher in black compared to white girls. Our results, however confirm the conclusions made by Rossouw et al. [8] indicating that the prevalence of childhood obesity in SA does not give a true reflection of the problem, as overweight and obesity differ markedly between age groups, boys and girls, ethnic groups and geographical areas. However, evidence of an overweight/obesity transition in school-aged children in Sub-Saharan Africa is substantiated by research [4], while the fastest growth rates of obesity among pre-school children are also found in Africa where the number of overweight and obese children in 2010, were more than double that found in 1990 [6].

In order to offer a perspective on the rate of increase of 4.0% in our group over a 3-year period, we compared this percentage increase with estimated increases that are reported by other studies. There is however, a lack of longitudinal studies, to enable direct comparisons with. Studies that were used are thus not necessarily based on the same age groups or time periods and findings obtained by trend analysis, were also incorporated. Researchers studied 450 obesity surveys of 144 countries to quantify the worldwide prevalence of OW and OB among pre-school children in Africa [6], and reported the estimated prevalence of childhood OW and OB in Africa, in 2010, as 8.5%, which they expect to reach 12.7% in 2020, indicating a predicted increase of 4.2% over a ten-year-period. A longitudinal study of 306 black children from low income families in Jamaica, reported that overweight increased by 6% (3.5% to 9.5%), from 7–8 years to 11–12%, while tracking of BMI was also high during follow-up [31]. An increase is also reported in OW and OB in first grade children in Chile [32], which is also a developing country, from 6.5% to 7.8% in boys and girls respectively in 1987, to 17% and 18.6% in 2000, which shows an increase of 12% over a 13-year period. Prevalence for developing countries is reported to have changed from 1980 to 2013 in child and adolescent boys and girls, from 8.1% to 12.9% in boys and from 8.4% to 13.4% in girls, indicating an estimate increase of 5% over this period [1].

Although not directly comparable to other studies, the rate of increase of 4.0% that were found among our 6–9-year-old group of SA children, thus displayed a more rapid increase over a shorter period of time, providing evidence of an expanding epidemic in pre-pubertal children in our study. In Japan [11], the largest odds ratio was also observed in the 6–8year-old children in whom the prevalence of obesity more than doubled from 4.2% (1976–1980) to 9.7% (1996–2000). The increase in our group was however influenced by SES and the interaction between SES and race, indicating that white children as a group and children from higher SES which included white and black children showed the most rapid increases. Statistics reported on this prevalence among white children in other SA studies [2,7,33] are also consistent with the higher prevalence found among white children. Differences reported in another SA study between ethnic groups also indicated that the results may be confounded by differences in SES [7]. However, the higher prevalence among white children compared to black children as a group, still differs from the statistics reported in developed countries such as the USA, Canada and Norway [34].

The rate of increase in overweight showed a slighter upward trend, compared to obesity prevalence which increased nearly twofold during the 3-year period. A 20 year trend analysis in Slovenia also reported a higher increase in obesity compared to overweight [12]. A trend analysis on 1–9 year old South African children, furthermore found that overweight decreased over time, while obesity increased [9]. The prevalence of obesity and severe obesity were studied over a period of 14 years in the USA in children, ages 2 to 19 years [15]. From 2011 to 2012, 17.3% of the children were obese, 5.9% met criteria for class 2 obesity (BMI ≥ 120% of the 95^th percentile and 2.1% met criteria for class 3 obesity (BMI ≥ 140% of the 95^th percentile). The researchers concluded that these rates were not significantly different from 2009 to 2010, but that more severe forms of obesity have increased over the last 14 years. Although percentages of Class 2 and 3 obesity were not determined in our study, obesity prevalence was the most severe form of OB in our study, and increased more during the 3-year-period than OW, which seems to follow the same pattern as described in the USA, a developed country.

The prevalence of obesity increased more in boys compared to girls over the 3-year period although girls still had a higher prevalence of obesity during follow-up. In addition more girls moved back to a healthier BMI compared to boys over the 3-year period. This trend of a higher increase in boys is consistent with other studies worldwide. In Canadian children, an increase is reported for OW (11 to 33% in boys, 13 to 27% in girls) and OB (2% to 10% in boys and 2 to 9% in girls) between 1981 and 1996 [16]. In 6-14-year old Japanese children, age adjusted BMI increased with 0.32 kg/^m per 10 years in boys and 0.24 kg/^m per 10 years in girls over 25-years, 6–14 years as derived from a national nutrition survey [11]. Time trends in the UK (1995–2010) based on the Health Survey of England (HSE) indicate an increase in prevalence of boys 2–15 from 11% to 17% while prevalence in girls increased from 12% to 15%. Trend analysis in the USA [14] of two large representative federal health surveys and data systems show a four-fold increase in obesity prevalence among 6-17-year-old male (5.5% to 21.6%) and a three-fold increase among female children (5.8% to 17.7%) between 1976 and 2008. The average annual rate of increase in obesity prevalence was furthermore 4.5% for male children and 3.8% for females in this study.

Prevalence in both white (4.2%) and black children (2.0%) increased significantly over the 3-year period, although white children displayed a much bigger increase and had almost double the prevalence of obesity (27.3%) than black children (13.3%) during follow-up. The combined OW/OB prevalence increase in white children was also bigger (20.3% to 27.3%) over the 3-year period than among black children (10.3% to 13.3%), although not in Q4 and Q5 schools where the increases among black children in Q4 schools were 8.2% (26.5%–34.7%) and 10.3% (20.7%–31.0%) in Q5 schools compared to those of white children (Q4, 5.6% (13.3%–18.9% and 7.7% (Q5, 24.5%–32.2%, Table 1). It further seems that South African children from higher SES have the highest prevalence and rates of increase in OW and OB. White children were all enrolled into the Q4 and Q5 schools that were part of the study, and these school quintiles represent more affluent schools, families and environments and also showed the highest combined prevalence. This differs from the findings of studies conducted in other countries such as the USA and UK, indicating the highest rates of obesity and severe obesity among children from minority groups or who are underserved by the health care system [34,35]. These studies are however conducted in developed countries while SA is considered a developing country in transition with high socio-economic disparities. It can thus be deduced that higher SES is currently associated with higher increases in overall prevalence in predominantly pre-pubertal white but also among black children in economic transition (Q4-19.4%–25.9%; Q5-23.3%–31.7%) as the increase in Q1–Q3 schools (Q1-9.1%–10%; Q2-8.3%–8.5%, Q3-3.9%–7.7%), based mainly on statistics of black children, were much less over the same period. Q1–Q3 schools enrol children from areas with high levels of food insecurity, [24,36,37], thus levels of underweight might be high in these schools. Black children in Q4 schools had significantly higher BMI and mass during follow-up compared to white children in these schools and their combined OW/OB prevalence increases were also higher. From this, the conclusion can be drawn that it can quite probably be the result of westernization and urbanization of more affluent black families. The high prevalence found among white children might also still be a result of the post-apartheid regime which exposed these children to circumstances equal to those in developed countries, such as sedentary lifestyles. Although many interrelated behaviour patterns can be contributing factors, decreased physical activity levels, among black girls, and higher food security which can contribute to higher availability and intake of processed foods, can be offered as reasons for these major differences between black children in low and high SES schools. Cultural beliefs regarding ideal body mass might be furthermore a possible contributing factor to the black and white differences that were found [36]. Spending money at school tuck shops on unhealthy foods might also play a role in the increased prevalence found in children attending Q4 and Q5 schools [38]. The prevalence of obesity that was established among 6–9-year old children by means of the first South African National Health and Nutrition Examination Survey (SANHANES-1), compared with the NFCS-2005, indicates a prevalence of 11.8% (OW 8.4% OB 3.4, mean BMI 16.2) (2013) and 10.3% (OW 7.8% OB 2.24, 2005, mean BMI 16.0) respectively [39]. This survey, however, has shortcomings in the sense that it essentially refers to African and coloured children residing in SA. When compared to our prevalence of 10.3% that were established in 2010 for 6-year-old black children, and 13.3% for 9-year old black children in 2013, a similar prevalence is confirmed in this ethnic group (11.8%). A longitudinal study of 306 black children from low income families in Jamaica, also reported low prevalence’s with an increase from baseline at 7–8 years (3.5%) to 9.5% at 11–12% [31].

Lastly, our results established that although a considerable percentage of the group transitioned over the 3-year period to an unhealthier BMI classification, a small percentage also moved to a healthier BMI. Boys and girls showed similar transition tendencies, while white children and children in schools representing higher SES (which included black children in Q4 and Q5 schools) showed higher shifts towards more unhealthy BMI’s in comparison to children in lower SES school types. Decreasing tendencies were also observed in BMI levels, although to a much smaller extent compared to the increasing tendencies that were found, resulting in a significant increase in combined OW/OB between 6 and 9 years. Although our changes in BMI were over a shorter follow-up period of 3 years, it agrees with a longitudinal study on tracking of BMI in Chinese children which was done on children who were aged between 6 and 13 years at baseline, reporting that over a 6-year period (1991–1997), that BMI remained unchanged in 40% of the group, while 30% moved to a lower or higher quintile and that overweight children were 2.8 times as likely as other children to become overweight adolescents [40]. This study, however, includes a high percentage of underweight children, and the researchers found that a smaller proportion of Chinese children in a rapidly changing society, continue to be overweight than what is reported in higher income countries.

Our study had limitations that need to be taken into consideration. This was not a nationally representative study but based on regional data of only 1 of the 9 provinces in South Africa. Research incorporating prevalence’s of all the provinces in SA are thus recommended. The strong points of the study are, however the stratified and longitudinal design, and the fact that the findings are based on real measurements and not self-reported height and weight data. This is also an on-going study with follow-up measurements due in 2016, which will provide an even more accurate picture of this growing problem among children over a period of 6 years as they move from early childhood into adolescence.