This study confirms a social gradient of overweight in children according to the mother’s educational level. However, the gradient was not as clear for maternal occupation and household income, given that it was rather a threshold effect contrasting the lowest SEP categories to the others, namely, “manual worker, unemployed and student” for occupation and the lowest quintile of income. Additionally, the mother’s migration status seemed to play an important role, without interacting with other SEP variables. Hence, this study challenges the common assumption that a gradient applies to all SEP indicators. We believe our analysis adds value to the existing literature by taking a holistic approach because it incorporates the perspective of migration history, thus better disentangling the multifaceted socio-cultural determinants contributing to this health condition.
The overall prevalence of overweight in our sample (8.3%, obesity included) was relatively low as compared with national estimates reported in three other studies. A joint survey, the ELFE-PMI, conducted by the ELFE cohort team and volunteers from French maternal and child health services from 2014 to 2016, found an overall prevalence of overweight of 10.4% at ages 3 to 4 years.22 However, the comparability of this estimate with our results is limited because the study covered only 30 French departments, and we know that regional variations in prevalence exist. Meanwhile, the INCA3 surveys, conducted in 2014 in metropolitan France and using IOTF cut-offs, estimated a prevalence of 13.7% in children 4 to 6 years old.23 Notably, children participating in the ELFE cohort study were younger than these children, precisely 3.5 years old in 2014. Likewise, a nationwide representative school survey conducted in 2013 reported an overweight rate in kindergarten (mean age 5 years) of 11.9% (90% CI 11.5–12.5);24 these latter figures apply to all regions of France, including the overseas departments (except Mayotte), where the prevalence is much higher.25 Both studies share a cross-sectional design, which could explain the difference in prevalence with the ELFE cohort, more likely to be affected by selection and attrition biases (as is the case in most cohorts) that can only partly be corrected by weighting. Nevertheless, our findings are consistent with the inverse socio-economic gradient reported from pre-school age in these two other national studies, whether the definition of SEP was based on the highest occupational category among the two parents24 or the educational level of the respondent parent in INCA3.23 The latter study found no association with respondent occupational status. Although the territorial divide of childhood overweight has not yet been described in France, in adults, the Obepi survey reported a downward North-South gradient of obesity.26 Here, we show that the geographical distribution is slightly different in early childhood, with the Northern, Southern and Western regions most affected.
The overall inverse socio-economic gradient of overweight observed in the 3.5-year-old children of the ELFE cohort did not apply uniformly to all SEP dimensions. This consideration is crucial because each indicator, although interdependent, reflects different facets of the SEP. For example, parents’ educational level is indicative of their knowledge and, to some extent, their health literacy and skills. Parents’ occupational category provides insight into the social environment in which the child is raised. Household income reflects the family’s purchasing capacity and ability to access goods and services.6 Our findings show that for household income levels, children in the lowest quintile but not the intermediate quintiles were more likely to have overweight than those in the highest income quintile. Likewise, children from workers and clerk mothers were more likely to have overweight than children born to mothers in other occupational categories. These nuances are not anecdotal and refer to different paradigms and specific perspectives regarding social inequalities in health.27 The socio-economic gradient refers to the graded association between SEP and health outcomes. However, using household income and mother’s occupational category as indicators of socioeconomic position, we observed that beyond a certain point, SEP is no longer associated with an individual’s health outcomes. These two different paradigms have implications for interventions aimed at reducing health inequalities in terms of components and sub-population groups to be targeted in priority. In particular, the paradigms can influence the adoption of proportionate universalism, a perspective that seeks to provide universal access to health services and interventions while directing resources to those who are most in need. This approach recognises that everyone should have access to the same level of basic care but acknowledges that some individuals and communities require additional support to achieve optimal health outcomes.28,29
Furthermore, our results show that the association between maternal education and child overweight was substantially attenuated after adjusting for household income. This observation suggests that the effect of maternal education on child overweight in the EFE cohort is partly explained by household income, as was further confirmed by a mediation analysis (results not shown, but available on request). However, there is some remaining direct effect of education, as confirmed by other studies:6,30 Van Rossem et al.30 also reported a persistent but attenuated association between maternal educational level and child overweight after adjusting for material hardship in the Dutch Generation R cohort, a multi-ethnic birth cohort of 4,581 children born between 2002 and 2006 in Rotterdam. Nevertheless, the magnitude of the association is not directly comparable with our findings because material hardship represents a different concept from household income6,31 and the baseline categories in the two studies were not identical.
To our knowledge, our study is the first to investigate the association between maternal migration status and childhood overweight in France. The results are consistent with a systematic review of 19 studies conducted in six European countries (not France) published from 1999 to 2009 that overall found a higher risk of overweight and obesity in immigrant children than their non-immigrant counterparts.10 However, most of these studies did not control for socio-economic factors in their analyses and the definition used to classify a child as an “immigrant” was not consistent across all studies. Other research from the Generation R study showed that children whose parents were born abroad were more likely to be overweight at age 4 years than children whose parents were both born in The Netherlands. However, after adjusting for maternal education, parental BMI, and infant weight change, the associations were attenuated, with the strongest attenuation observed after adjustment for maternal education.30
The “healthy migrant effect” describes migrants having better health outcomes than non-migrants in both their source and host countries. Evidence for the healthy migrant effect is inconsistent across regions and contexts. Previous studies14,32 have documented the healthy migrant effect in Canada, the United States, Australia and the United Kingdom for the prevalence of chronic conditions, self-assessed health, and obesity, but the situation in other European countries is complex and heterogeneous. Moullan et al.13 found evidence of a healthy migrant effect for self-assessed health in Italy and Spain but not Belgium and France. Some other studies have reported poorer health status among immigrants than non-immigrants in terms of chronic conditions and self-perceived health in France15,16,33 and Europe.34 Differences between countries can be explained by factors such as the country of origin, the host country’s legal framework for immigration, the length of stay, and the healthcare system; the healthy migrant effect also depends on the reasons for migration.14 Whether this effect extends to the offspring of immigrant mothers with regard to obesity remains unclear. A US study did not find any “healthy foreign-born effect” for childhood obesity.35
A recent study based on the ELFE cohort showed that immigrant parents’ pre-migration education had a positive impact on their children’s birth outcomes including birth weight,36 but this advantage declined with length of residence in the new country. The acculturation process, which refers to the change and adaptation that occurs when individuals or groups encounter a different culture, may partly account for why this benefit diminishes with length of residence. This process entails changes in attitudes, values, beliefs, behaviors and language and can take place at both individual and group levels. A systematic review concluded that acculturation was associated with obesity in adult immigrants from low/middle-income to high-income countries.37 Another Swedish study found that children of immigrants had greater risk of low physical activity and overweight than children of Swedish parents, despite a better-quality diet.14 However, acculturation is a complex and nuanced phenomenon that requires careful examination.38 Another study of ELFE data found that the diet of immigrant mothers was better than that of descendants of immigrants, who in turn had a better diet than non-immigrant mothers.39 Therefore, women who were less acculturated had both a healthier and less processed diet than non-immigrant mothers.40 However, this relationship must be tested in children to determine the extent to which it applies to them. In light of the literature,39,41 we were expecting a more pronounced socio-economic gradient of childhood overweight in the non-immigrant than immigrant population, which was not confirmed given the lack of interaction between migration status and all three SEP indicators.
In the present analysis, the association between maternal migration status and childhood overweight persisted even after adjusting for other SEP variables. This finding suggests that migration status has a direct effect, independent of SEP. In particular, socio-cultural norms, values and representations could be involved. For example, perception of weight and overweight may differ according to the migration status and induce less favorable energy balance-related behaviors, regardless of SEP: a British study found that parental perceptions of healthy body size and concerns about overweight in childhood varied by ethnicity.42 Disparities observed between different racial and ethnic groups could also indicate discrimination at an institutional or structural level, shaped by personal experiences, or a combination of both.6 However, we cannot rule out the possibility of residual confounding due to other unmeasured dimensions of SEP. For instance, a study31 of the ELFE cohort used a multidimensional approach to measure child poverty, combining income and deprivation measures. Income poverty did not perfectly overlap with deprivation: some low-income children were not considered poor and some higher-income children were considered deprived.
There are a number of limitations to our study. As mentioned before, the prevalence estimates reported in this cohort are lower than those from other cross-sectional studies. The specific weights provided by the ELFE team to adjust the sample and mitigate this selection and attrition biases were deemed effective but not sufficient to correct the selection bias due to attrition. However, our primary objective was not to measure the prevalence itself but rather to comprehend the social patterning of childhood overweight with a holistic approach, accounting for various social determinants, in particular migration. Of note, the ELFE cohort does not include children from the overseas departments. A study published in 2012 on children aged 5 to 9 years in Guadeloupe, Martinique, French Guiana and French Polynesia reported an overweight prevalence of 15.2%, 25.0%, 16.8% and 31.6%, respectively.25 Hence, this exclusion could potentially bias the results by underestimating the absolute prevalence and the prevalence in the non-immigrant population. In addition, several studies have reported ethnic variations in body composition.43–45 For the purpose of consistency, we decided to use the most commonly accepted definition of overweight: BMI greater than or equal to the IOTF-25 cut-off. However, by extension, this choice may slightly overestimate the real prevalence of overweight among some ethnic minorities.6
Our study also has notable strengths. The use of a hierarchical approach in the design of our conceptual model avoids a wrong interpretation of over-adjustment commonly encountered in other studies that incorporate all variables simultaneously in the same models. In addition to confirming that “one size does not fit all”,6 our approach demonstrated both social gradient and threshold effects: we cannot exclude that these differences would be more pronounced in the target population as compared with our selected population. Finally, the use of multiple imputation techniques mitigated potential bias introduced by missing data, which provided a significant advantage.
This study demonstrates the existence of social inequalities in the ELFE birth cohort. These findings, based on SEP and migration status, agree with those of other studies conducted in similar European cohorts,10 such as Generation R.30 However, more nuanced aspects of vulnerability31 need to be examined to gain a deeper understanding of the underlying mechanisms that structure and mediate the effect of these factors on child overweight. This examination would enable the development of concrete solutions to be applied in public health interventions. Nonetheless, if the latter have the overarching goal to reduce social inequalities in childhood overweight, their design requires an appropriate balance between individual versus structural components. Indeed, a systematic review revealed that interventions across different socio-economic groups differed in efficacy.46 Those that relied on information provision for individual behavior modification were ineffective in participants with lower SEP. Only community or policy interventions aimed at bringing structural changes to the environment were successful for them. This is also corroborated by a recent systematic review which highlighted that increasing the availability of healthier food options enhanced the likelihood of healthy choices and reduced the energy content of the diet similarly among individuals with higher and lower SEP.47 There are existing examples of this type of intervention that have demonstrated positive outcomes in encouraging healthier choices among consumers.48 Hence, policies that increase the availability of healthier food could have potential as equitable strategies to reduce obesity and improve population health. In addition to availability, accessibility is a crucial factor in food choices. Research suggests a significant positive effect of higher occupational social class on food expenditure, which in turn influences the healthiness of food purchases.49 Within the framework of proportionate universalism, only such structural interventions can empower the segment of the population facing social adversity and promote social equity in health initiatives. Another recent systematic review found that interventions delivered by lay agents among ethnic/racial minorities had some effect on lifestyle behaviors and obesity risk; additional factors for creating more effective, pragmatic, inclusive, and non-judgmental programs were to engage stakeholders, including users, in their development and adhering to theoretical frameworks.50