Socioeconomic inequalities in overweight and obesity among 6‐ to 9‐year‐old children in 24 countries from the World Health Organization European region

Childhood overweight and obesity have significant short‐ and long‐term negative impacts on children's health and well‐being. These challenges are unequally distributed according to socioeconomic status (SES); however, previous studies have often lacked standardized and objectively measured data across national contexts to assess these differences. This study provides a cross‐sectional picture of the association between SES and childhood overweight and obesity, based on data from 123,487 children aged 6–9 years in 24 countries in the World Health Organization (WHO) European region. Overall, associations were found between overweight/obesity and the three SES indicators used (parental education, parental employment status, and family‐perceived wealth). Our results showed an inverse relationship between the prevalence of childhood overweight/obesity and parental education in high‐income countries, whereas the opposite relationship was observed in most of the middle‐income countries. The same applied to family‐perceived wealth, although parental employment status appeared to be less associated with overweight and obesity or not associated at all. This paper highlights the need for close attention to context when designing interventions, as the association between SES and childhood overweight and obesity varies by country economic development. Population‐based interventions have an important role to play, but policies that target specific SES groups are also needed to address inequalities.

overweight and obesity varies by country economic development. Population-based interventions have an important role to play, but policies that target specific SES groups are also needed to address inequalities. hood obesity, as well as inequalities in exposure to these factors within and across populations. 2,3 Poverty is known to have significant impacts on child health and well-being, 4 and the resulting food insecurity can be associated with both undernutrition and overnutrition in countries at different stages of economic development. 5 Overweight and obesity in children are important health problems due to their association with a range of serious short-and long-term complications. 6 Research shows that children with overweight and obesity are very likely to maintain their weight status in adult life, 7,8 leading to an increased risk of morbidity and premature mortality from noncommunicable diseases (NCDs) in childhood and adulthood. [9][10][11][12] In addition to these delayed consequences of excess body weight, children also suffer from immediate consequences, such as stigmatization, bullying in school, social exclusion, low selfesteem, and body image dissatisfaction. [13][14][15] Additionally, childhood and adolescent obesity has been repeatedly linked to depression and depressive symptoms. 16 A pooled analysis of the NCD risk factor collaboration estimated that the number of children and adolescents aged 5-19 years with overweight worldwide to be approximately 75 (95% credible interval  17 In the same analysis, the authors mention that the previously observed trend of increasing childhood overweight and obesity in many high-income countries (HICs) appears to have plateaued. At the same time, numerous studies indicate that children in lower socioeconomic groups within these countries have not benefited from the stabilization of the trend, suggesting growing disparities in prevalence between socioeconomic groups. [18][19][20][21] Worldwide, most studies that have detected inequalities in childhood obesity prevalence due to socioeconomic status (SES) show an increase in inequalities after 2000. 22 The World Health Organization (WHO) describes the social determinants of health as "The conditions in which people are born, grow, live, work and age." 23 These conditions affect health across the lifecourse via three related mechanisms. 24 Firstly, a lower SES is thought to expose an individual to long periods of psychosocial stress, which is known to be damaging to health through hormonal and nervous system reactions. Secondly, people with lower SES are more prone to exposure to risk factors such as tobacco use, air pollution, poor nutrition, and (in certain contexts) sedentary behaviors. 25,26 Finally, children born into families with lower SES are more likely to be exposed to factors such as poor maternal health in utero, as well as early-life influences such as early cessation of breastfeeding and earlier introduction of foods and drinks high in fat, salt, and/or sugar. 27,28 These associations are evidently at play in the case of childhood overweight and obesity, which research has repeatedly linked to parental body weight, parental education, family income, and sociodemographic factors. [29][30][31][32][33][34] Globally, SES is negatively associated with childhood overweight and obesity in the majority of HICs. Hence, the lower a child's SES, the more likely they are to suffer from obesity. 22 However, in lowincome countries (LICs) and low-middle-income countries (LMICs), it is still more likely to observe overweight and obesity among children in families with higher SES. 35,36 LICs and LMICs may also bear a "double burden" of increasing childhood overweight and obesity, coupled with the persistence of childhood undernutrition. 37 From a regional perspective, data analyzed by Knai et al. from 22 European countries between 1990 and 2005 suggested that greater inequality in household income was positively associated with self-reported and measured child overweight prevalence. 38 Studies from the United States have also suggested that childhood obesity prevalence is inversely associated with income and education, although patterns might differ between children and adolescents. [39][40][41] However, in previous studies, SES indicators were defined in different ways including measures of parental education, parental occupation, family income, composite SES measures, and neighborhood-level SES indicators. 33,42,43 These studies often relied on self-reported anthropometric measures rather than standardized measurements.
Pooled results from the WHO European Childhood Obesity Surveillance Initiative (COSI) fourth round (2015-2017) indicated that 28.7% of boys and 26.5% of girls aged 7-9 years had overweight (including obesity) and 12.5% of boys and 9.0% of girls had obesity according to the WHO growth reference curves. 44 Data collected in COSI first round (2008)(2009)

| Classification of children's weight status
The classification of children's weight status was based on the 2007 WHO recommended growth reference for school-aged children and adolescents. 48,49 The WHO 2007 cut-offs were used to compute BMI-for-age Z-scores and to estimate prevalence of overweight/ obesity. According to the WHO definitions, overweight and obesity are defined as a BMI-for-age value >+1 Z-score and >+2 Z-scores, respectively. Moreover, the estimated prevalence of overweight includes children with obesity. 48 Children with a biologically implausible (or extreme) BMI-for-age value were excluded from the analysis (values below −5 or above +5 Z-scores relative to the 2007 WHO growth reference median). 50

| Family SES variables
We assessed family SES according to three variables: parental education, parental employment, and family-perceived wealth.
The COSI optional family record form included an item on the education and employment of the responding caregiver and his/her partner/spouse. Therefore, the information about parental education and employment was available only if the family form was completed by the mother or the father. In Bulgaria, Czechia, Italy, Malta, San Marino, Spain, and Turkey, data on education level and employment status specifically of the parents were gathered, regardless of which caregiver completed the questionnaire. Because information on family composition was not gathered in the fourth round of COSI, it was not possible to identify children living in a single-parent family nor to properly classify the educational attainment and the employment status of their parent. These children were thus excluded from the analysis, which focused on children living in a traditional two-parent family structure.
Three categories of parental education were then created: (1) low parental education (both parents with lower education); (2) medium parental education (one parent with lower education, one parent with higher education); (3) high parental education (both parents with higher education). We described parents as having "lower education" if they reported their educational attainment as "primary school or less," "secondary or high school," or "vocational school." We described parents as having "higher education" if they reported their educational attainment as "undergraduate or bachelor's degree" and "master's degree or higher." We created two categories for parental employment status: (1) low parental employment (one or more parent(s) unemployed or economically inactive, i.e., not working at all and neither available nor looking for work); (2) high parental employment (both parents employed). Parents were classified as "employed," "unemployed," or "inactive" based on the following answer options from the optional family record form: "employed" comprises the answers "government employed," "nongovernment employed," and "self-employed"; "unemployed" is indicated by the answer "unemployedable to work"; and "inactive" comprises the answers "unemployed unable to work," "student," "homemaker," and "retired." We generated three categories to describe family-perceived wealth: (1) low family-perceived wealth (those who had trouble meeting the end of the month with their own earnings); (2) medium familyperceived wealth (those who met the end of the month with their own earnings without serious problems); (3) high family-perceived wealth (those who easily met the end of the month with their own earnings).
For the purpose of this paper, the following inclusion criteria were applied: (i) children aged between 6 and 9 years; (ii) children with available information on body weight, height, sex, and age; (iii) children with available information on education or employment status of both parents.

| Data analysis
The prevalence values of overweight and obesity were estimated at the country level. Differences across SES categories were tested using the Pearson's χ 2 test corrected using the Rao-Scott method. 51 A multivariate multilevel logistic regression analysis was carried out to estimate the odds ratios (ORs) and their 95% confidence intervals (CIs) of having overweight (compared with having normal weight or thinness) for parental education (reference category: high parental education), family-perceived wealth (reference category: high perceived wealth), and parental employment (reference category: high employment). The ORs were estimated adjusting for the child's sex and age, the degree of urbanization of the child's residence or school (urban versus rural), and the region/administrative division of the family's place of residence. The categorization of urbanization has been described elsewhere. 52 A model was estimated for each country included in the analysis. All models included random effects for primary schools attended by children to consider the clustered structure of the data. For Czechia's models, random effects for paediatric clinics where children were enrolled were used instead of primary schools. In the multivariate regression analysis, children with a missing value for any of the covariates were excluded. Sex-stratified models were also estimated for all countries.
The same regression analysis was carried out for obesity. More specifically, country-specific models for having obesity compared with having normal weight or thinness were estimated for all countries but Czechia, Denmark, Ireland, San Marino, and Tajikistan where the limited number of children with obesity did not allow reliable results.
Sex-stratified models were estimated as well.
Both descriptive and regression analyses showed a high level of heterogeneity among countries in terms of the direction and magnitude of the association between overweight/obesity and SES variables. For this reason, we did not carry out any analysis pooling together data from different countries.
Sampling weights to adjust for the sampling design, oversampling, and nonresponse at the child level 47  The results are presented in the tables by grouping countries in six macroregions according to United Nations "Standard Country or Area Codes for Statistical Use" 53 : Northern Europe, Western Europe, Eastern Europe, Southern Europe, Central Asia, and Western Asia.
The World Bank classification of countries by income was also used to report and discuss results. 54 Table 1.
In Table 2, the percentage of boys, children's age (mean and standard deviation [SD]) and percentage of children having overweight and obesity according to the WHO growth reference are reported by country. The table also shows how children were distributed by parental education level, parental employment status, and familyperceived wealth at the country level.

| Prevalence of overweight and obesity in children with low and high level of family SES
As shown in Figure 1A, in European HICs, the prevalence of overweight was higher among children whose parents had lower education status relative to children with high parental education The patterns between SES and obesity were similar to those observed with SES and overweight. However, the association between parental education and obesity was stronger than the association between parental education and overweight. In seven European HICs, the obesity prevalence among children with low parental education was around twice of which was observed among children with high parental education.

| Parental employment status
Multivariate regression analyses showed that, in case of association, children with low parental employment compared with high were less likely to have overweight, especially in UMICs and LMICs, such as Albania, Bulgaria, Kyrgyzstan, Georgia, Romania, and Turkey.
The country specific adjusted ORs for having obesity compared with having normal weight or thinness confirmed the associations observed for having overweight (Figure 3). In general, the association tended to be stronger for obesity, regardless of the direction, especially for parental education.  Table 1 boys (see Figure S1). In Czechia, parental education tended to be more strongly associated with overweight among boys, whereas The fact that we were only able to analyze children living in twoparent households is likely to have impacted our results. There is evidence that single parent families are at greater risk of food insecurity in HICs and that food insecurity is associated with obesity in adulthood. 58  Our findings are also consistent with a study by Pampel et al., which found that obesity (in adulthood) rose with a nation's economic development, and the relation between SES and obesity changed. 36 In lower income countries, people with higher SES were more likely to have obesity, whereas in HICs, those with higher SES were less likely to have obesity. A recent systematic review confirmed a similar situation for childhood and adolescent overweight and obesity in HICs.
The authors found an inverse relationship between SES and overweight/obesity in 72% of the included studies. 22 Meanwhile, a review by Dinsa et al. explored the relationship between SES and obesity in low-and middle-income countries. Of the 11 included studies that looked at children, all of them found a positive relationship; that is, obesity was more common among children with higher SES. 35 In our results, the sex-stratified multivariate regression analysis suggested only minor differences between boys and girls when comparing ORs for having overweight/obesity related to SES variables in most of the countries. However, some countries did show significant differences. This demonstrates the importance of sex-disaggregated data to identify disparities; for example, a 2016 Korean study found that childhood overweight was positively related to household income in boys but inversely related to household income in girls. 59 In 2020, evidence is beginning to emerge that lockdowns in response to the COVID-19 pandemic have had unintended consequences such as increasing less healthy behaviors (like reduced physical activity, the loss of healthy and nutritious meals in schools, and increased sedentary behaviors) among children, 60 and increased prevalence of overweight and obesity among young people. 61 It is likely that these impacts will be unequally distributed by SES and will consequently increase poverty and inequalities. Future research will be needed to expose the extent of these consequences in different contexts.
Our results identify key areas for intervention at the national level, as well as providing an opportunity for knowledge exchange between countries facing similar challenges. They add to the body of evidence linking increased risk of childhood overweight and obesity with SES and are particularly reliable given the standardized nature of data collection across countries. In general, they show a need for a blend of structural and individual approaches for effective prevention of childhood obesity. SES is multifaceted; therefore, interventions to address socioeconomic inequalities in childhood obesity should be comprehensive and address multiple risk factors.
Structural measures that reduce risk at the population level have been implemented in many countries around the world: a recent example is the proliferation of taxes on sugar-sweetened beverages.
At a basic level, poverty is associated with obesity in HICs due to the lower cost of energy-dense foods 62 ; therefore, fiscal interventions like taxes and subsidies can help reduce the accessibility of less healthy options while making healthier options more accessible. For example, the Mexican sugar tax was accompanied by efforts to increase the availability of fresh, potable water in schools. 63 Schools and day-care centers/kindergartens are a pragmatic site of intervention, given that they provide opportunities to deliver healthy school meals, in addition to age-congruent physical activity. There is evidence that targeted school food policies can increase consumption of specific foods such as fruits and vegetables; however, broader impacts on metabolic risk factors are limited. 64 As the onset of childhood obesity generally occurs in the early years, 65 and because we must assume that the obesogenic environment as a whole fosters obesity-related unhealthy lifestyle behaviors, 66 school interventions must be one component of more comprehensive action.
Taking a life course perspective, we know that maternal nutrition is vital to ensure optimal fetal development and reduce the risk of NCDs in later life. Improving access to quality antenatal care for all social groups can improve maternal and child health outcomes. 67 Another key area for population-level intervention is the exposure of children (especially in the early years) to digital marketing of high in fat, salt, and sugar (HFSS) foods. Given the rapid increase in access to digital devices at young ages, legislation has been slow to keep pace and protect children from subtle and manipulative marketing techniques that have been shown to increase requests for nutrient-poor and calorie-dense foods. 68

| Strengths and weaknesses
To the best of our knowledge, this is the largest study of its kind, using standardized methods to collect data on children's anthropometrics from nationally representative samples and using standardized indicators of SES to examine the association between SES and weight status among children. It therefore provides robust evidence to support national and regional policy decisions.
However, there are also some weaknesses that should be taken into consideration when interpreting the results. Firstly, measures of SES were self-reported and subjective, which may have introduced reporting bias, particularly in the case of family-perceived wealth.

ACKNOWLEDGMENTS
We gratefully acknowledge all participating, children, their parents, and the schoolteachers, principals, and examiners who have participated in COSI. We also thank the teams who have collected the data in each country. We gratefully acknowledge Jelena Jakovljevic, Liza Villas and Gerben Rienk for their support in preparing and carrying out COSI round 4 data collection. We also thank Natalia Fedkina (WHO Regional Office for Europe, Moscow, Russian Federation) and Karen McColl for their support in the production of this article. and data cleaning. All authors contributed to and approved the paper.

CONFLICTS OF INTEREST
The authors declare no conflict of interest. The funders played no role in the design of the COSI protocol, the decision to write this paper or its content.

ETHICS STATEMENT
The COSI study follows the International Ethical Guidelines for Biomedical Research Involving Human Subjects. Local ethics approval was also granted.

FUNDING INFORMATION
The authors gratefully acknowledge support through a grant from the