Since this study was structured according to its three aims: (I) to conduct a meta-analysis on overweight/obesity prevalence in 9 to 11 year old Portuguese children; (II) to detect differences in behavioural characteristics among normal-weight and overweight/obese children, and (III) to investigate the importance of individual- and school-level correlates on children’s BMI variation, the discussion will follow accordingly.
4.1. Prevalence of Overweight/Obesity in 9–11 Year-Old Portuguese Children
There is compelling evidence that overweight/obesity in children is one of the most important public health problems worldwide. According to de Onis
et al. [
1], childhood obesity prevalence increased from 4.2% in 1990 to 6.7% in 2010, and it is expected to reach 9.1% in 2020. A recent Portuguese study [
15] among 7–9 year-old children, using data from 1970, 1990 and 2002, reported mean increases in BMI during these time periods, with greater increases between 1992 and 2002. However, as the 1970 and 1992 studies did not report overweight/obesity prevalence, it was not possible to determine a trend although they hypothesized that the increase observed from 1992 to 2002 could be linked to increases in the prevalence of overweight/obesity observed in the last years. Our meta-analysis results suggest that the previous increases in obesity may have slowed or plateaued in Portugal. Our findings are in agreement with recent trends of stability or decrease in the prevalence of overweight/obesity in youth. For example, among Dutch, Moroccan and Surinamese South Asian ethnic groups in the Netherlands, aged 3–16 years, de Wilde
et al. [
39] found a stabilization, or even a decrease, in the prevalence of overweight/obesity between 1999-2011. Moreover, similar results were reported by Moss
et al. [
40] studying German children, especially from the year 2004 onward; further, they reported a decline in the prevalence of overweight from 8.4%–11.9% to 3.3%–5.4%, varying according to regions, with an absolute decrease of prevalence rates up to 3% for overweight and 1.8% for obesity. Additionally, among Danish children aged 7–14 years [
41] (using self-reported values for height and weight), the prevalence of obesity did not increase significantly from 1995 to 2001 (2.3% to 2.4%), but the prevalence of overweight rose 10.9% to 14.4%). Among French children [
42], aged 3–14 years, no significant difference was found in the prevalence of overweight/obesity from 1999 to 2007. Notwithstanding World Health Organization data showing increases in overweight/obesity in young populations [
1], it seems that in some nations a levelling off in childhood overweight/obesity has been observed [
43,
44,
45], which calls for a more reliable analysis about variations among countries. In our meta-analysis, the prevalence of overweight/obesity ranged across studies from 19% to 35%, with an overall estimate (from fixed and random effects) of approximately 30% (for boys and girls, together), which is higher than those reported from previous studies. Although one study relied on self-reported height and weight [
36], its results were not different from the others.
The reason for the stability in youth overweight/obesity prevalence observed in some countries is not clearly understood, and the same occurs in the present study. Tambalis
et al. [
46] hypothesized that the obesity prevalence may have reached a race and/or country specific ceiling, implying that children with predisposition toward obesity are now obese and obesity prevalence will not increase systematically. Similarly, Olds
et al. [
44] suggested that the environment in developed countries may be saturated with unhealthy food and options for sedentariness that children with a predisposition to becoming overweight have become overweight, and the remaining children may be resilient to obesogenic environments. In the Portuguese context, there is no published evidence to support our results. However, available data on physical activity in this age range (9–11 years), showed that approximately 36% of children achieve the daily recommended levels of MVPA [
47]. Additionally, there has been an observed increase in organized sports participation during the last years among Portuguese youth (increases up to 149% from 1996 to 2010) [
48] which may also contribute to this stabilization, given the widely reported relationship between overweight/obesity and physical activity.
4.2. Biological, Behavioural and Socio-Demographic Differences between Normal-Weight and Overweight/Obese Children
Our results demonstrate that normal-weight and overweight/obese children are significantly different in their biological traits in that overweight/obese children are taller, heavier and ahead in their maturation. This is expected since previous research has shown that early maturing youth usually are taller, heavier and have higher BMI than their later maturing peers [
49]. There is also strong evidence showing familial aggregation in BMI, where children with parents with high BMI, tend to have high BMI values [
50,
51]. For example, among the Chinese Han population, Hu
et al. [
50] reported that children with overweight/obese parents had higher BMI, and Fuentes
et al. [
51], studying Finnish family’s aggregation in BMI showed that when one or both parents were obese, children were more likely to be in the highest quartile of BMI.
Differences between weight groups in nutritional habits, physical activity levels, sedentariness, sleep time and SES were not statistically significant. There is no clear evidence that children with healthy weight differ from those with excess weight in varied sets of behaviours and demographic characteristics [
12,
13]. For example, with regards to nutritional habits, Yannakoulia
et al. [
12] using a 3-day food record to study food patterns between normal-weight and overweight children, reported no significant differences between groups. Additionally, Garaulet
et al. [
9], investigating the association between energy and nutrient intake with the prevalence of overweight and obesity, found that overweight boys derived a greater percentage of their energy from fat and less from carbohydrates as compared to normal weight boys, whereas overweight girls consumed less carbohydrates than normal-weight ones. Furthermore, Storey
et al. [
52], investigating non-overweight, overweight and obese adolescent diets, showed that non-overweight students consumed significantly more carbohydrate and fiber, significantly less fat and high calorie beverages, and had a higher frequency of consuming breakfast and snacks compared to their overweight or obese peers. Among Portuguese children, no significant differences between weight groups in healthy and unhealthy diet consumption were found, and children from both groups had equal access to healthy and unhealthy foods. One possible explanation for these results may be related to characteristics of Portuguese schools, which offer a nutrient balanced lunch for children. Moreover, they have food policies and controls regarding snacking and fast food, allowing children from both groups to have equal access to healthy and unhealthy food. In addition, schools also have a national program called “education for health” that aims to teach children about healthy choices and healthy living.
Physical activity and sedentary time/behaviour are usually correlated with weight status where normal-weight children are more active and less sedentary [
53,
54]. However, in our data no differences were found in physical activity levels or time spent in sedentary behaviour. These results are in line with data reported by Wang
et al. [
14] in Chinese children, as well as with those from Maier
et al. [
13]. This last investigation showed that all overweight or obese children reached the recommendation of spending one hour per day in sportive activities which was similar to normal-weight children. In Portugal, physical education is mandatory (twice a week); in addition, all schools offer free sports club participation, and most children (independent of their SES) have access to private sports club which may explain our results.
There is evidence that short sleep duration is consistently associated with concurrent and future obesity [
55], where overweight/obese children tend to spend less time sleeping than children of normal weight [
56,
57,
58]. This relationship may be explained by alterations in glucose metabolism, up-regulation of appetite, and decreased energy expenditure [
59]. However, our results differ from previous studies in that no differences were found in sleep time between normal-weight and overweight/obese children meaning that sleep deprivation may not be a major risk factor for development of overweight/obesity among Portuguese children. However, our sample included both overweight and obese children, and different results may be obtained by focusing on obesity alone.
In our study SES was determined by annual household income, and it was not found to be different between children of different weight status. Available data do not consistently show a clear effect size and direction in the association between weight status and SES. Previous research suggests that in developed countries, children of low and medium SES are more likely to be obese than those of high SES [
60], while others report that higher SES is positively associated with overweight and obesity in Chinese children [
61]. Given the distribution of SES in our sample, without any noticeable income inequality, the null results were expected.
4.3. Individual- and School-Level Correlates of BMI Variation
The third purpose of the present study was to investigate the importance of individual- and school-level correlates on variation in children’s BMI. At the child level, most variables were significant predictors. Similar to our results, sex differences in BMI have been consistently reported, with boys having higher BMI. For example, Meigen
et al. [
62] studied secular trends in German children’s and adolescents’ BMI and showed a greater increase in boys. Moreover, among Chinese children, Song
et al. [
63] identified sex disparities in BMI-for-age z-score during a 15-year period in which girls were stable whereas a linear increase was observed in boys. Additionally, Ogden
et al. [
64], found a significant increase in obesity prevalence between 1999–2000 and 2009–2010 in American boys as contrasted with girls. Similarly, Skinner and Skelton [
65] also reported a stabilization, with a non-significant increase, in the prevalence of obesity among US children from 1999 to 2012. In the Portuguese context, results differ between studies [
66], and there is no clear trend for boys having higher BMI than girls.
It is evident that more mature children have higher BMIs [
49], as early maturing children are taller and heavier than on time or late maturing peers, where early maturity affects weight relatively more than height [
49]. As such, our results were expected since those closer to their PHV (higher maturity offset positive values) had higher BMI values. Further, the relationship between parental and offspring BMI was significant,
i.e., children whose parents report high BMI tend also to have high BMI values, which is a consistent finding in previous twin and family studies [
50,
51]. For example, two Portuguese family studies by Souza
et al. [
67] and de Chaves
et al. [
68] reported the presence of genetic factors explaining from 30% to 50% of the total variance in different body composition phenotypes, which agrees with our findings that parental BMI is positively correlated with their child’s BMI.
Nutritional habits and physical activity levels are two behavioural phenotypes usually associated with BMI [
54]. Our results firstly showed that a healthy diet was negatively associated with BMI, meaning that children with a higher healthy diet score had lower BMI values. Among Mexican children [
69], food patterns characterized by a high intake of sugary cereals, sweetened beverages, industrial snacks, cakes, whole milk, and sweets were associated with higher risk of overweight/obesity. Moreover, accordingly to Maffeis [
70], there is some evidence that obese children show a certain preference for a fatty diet. On the other hand, non-overweight students tend to consume significantly more carbohydrate and fibre, and significantly less fat and high calorie beverages [
52]. With regard to physical activity levels, our results are also in line with previous research. For example, Bingham
et al. [
3] found that performing at least 1 hour of moderate physical activity every day is a protective factor against childhood overweight/obesity in Portuguese children. Furthermore, Janssen
et al. [
71], in a review paper examining associations between overweight, dietary and physical activity patterns in youth concluded that increasing physical activity participation was a relevant strategy to prevent and treat overweight and obesity.
School predictors were not significantly associated with children’s BMI. It is possible that the low number of schools (23 schools) and the similarity observed across Portuguese school environments may explain the null results, since only 2% of the BMI variation was attributed to school-level differences. Pallan
et al. [
72] investigated interschool variation in BMI z-scores and also found low intracluster correlations—between 0.9% and 4.2%. Further, they reported that the only school-level variable associated with BMI z-score was time spent in physical education classes (minutes/week).
The present study has several limitations. Firstly, the national studies included in the meta-analysis did not include information from the Autonomous Regions of Madeira and Azores, and this may bias the estimates to an unknown degree. However, one study reported a 29.1% prevalence of overweight/obesity in 11 year-old Azorean children [
73], and another [
74] showed a prevalence of 31.3% and 25.2% among Azorean 10 year-old girls and boys, respectively, which falls within the range of the confidence interval of the overall prevalence across the decade. Additionally, among 9–11 year-old Madeiran children [
75], the prevalence of overweight/obesity ranged from 14.7% to 17.1% for boys and 13.3% to 16.9% for girls, which are lower than the prevalences reported in studies involving mainland samples. Secondly, although we used subjective methods to determine nutritional habits and sedentary behaviour, this is current practice in previous studies cited in the present article. Thirdly, we use self-reported parental height and weight which are suitable proxies for their actual values, and this is usual procedure in epidemiological research [
76]. Fourthly, although ISCOLE utilized a validated questionnaire to obtain information concerning TV watching, it is possible that children underestimated their actual TV time. Fifthly, the present study combined overweight and obese children for analyses which may have attenuated effect sizes. Note that our total sample size is 686 children, and the distribution of BMI groups is 60.8% (n = 417) normal-weight, 28.3% (n = 194) overweight, and 10.9% (n = 75) obese. Separate analyses were not performed due to the lack of statistical power; further, our parameter estimates would be less precise, and our conclusions less reliable. Notwithstanding these limitations, the study has several merits: (1) the presentation of a meta-analysis of the prevalence of overweight/obesity in the last decade in Portuguese children within an important developmental transition period; (2) the use of an objective method to assess physical activity; (3) inclusion of objective information regarding sleep time; (4) use of standard methods and highly reliable data; and (5) the use of multilevel modelling to capture the complexity of nested information at children and school levels.