Socioeconomic inequalities in the food environment and body composition among school-aged children: a fixed-effects analysis

Childhood obesity is a significant public health concern, with its prevalence rising rapidly in recent decades. Changes in the food environment, characterized by increased exposure to food outlets and high-energy, ultra-processed foods, are likely contributing factors. However, evidence for a causal link between changing food environments and childhood obesity is limited, with many studies relying on cross-sectional designs susceptible to confounding factors like residential self-selection. Studies employing fixed-effects models with repeated measurements can help mitigate this bias by accounting for time-invariant and time-varying unmeasured variables. While one US study using this approach showed a counterintuitive negative association between increased fast-food outlets and BMI, further research is needed, particularly regarding socioeconomic inequalities. Prior research indicates that access to fast-food is higher in deprived areas, and individuals from lower socioeconomic positions spend more time in their neighborhoods, potentially leading to differential impacts of the food environment on obesity across socioeconomic groups. This study leverages the Generation R Study, a large birth-cohort study, to examine if changes in the food environment are associated with changes in body composition, hypothesizing that children from lower socioeconomic backgrounds will reside in areas with more unhealthy food outlets, leading to unfavorable changes in their body composition.

Existing literature reveals inconsistent findings regarding the association between the food environment and childhood obesity. Cross-sectional studies have shown a correlation between fast-food access and obesity, particularly in lower socioeconomic groups. However, these studies are prone to selection bias. Longitudinal studies, which are crucial for establishing causality, are scarce. A systematic review of 21 studies highlighted the higher prevalence of fast-food outlets in more-deprived areas. Other studies have also shown that individuals from lower socioeconomic backgrounds spend more time in their neighborhoods, increasing exposure to the local food environment. Despite this, a recent systematic review primarily based on cross-sectional studies failed to confirm a differential impact of the food environment on obesity across socioeconomic groups. The existing evidence underscores the need for longitudinal studies focusing on socioeconomic inequalities in the food environment and their connection to changes in obesity.

This study utilized data from the Generation R Study, a prospective birth-cohort study in Rotterdam, Netherlands. The study included objective measures of BMI (at ages 4, 6, 10, and 14 years) and DXA-derived fat mass indices (at ages 6 and 10 years). Home addresses were linked with yearly updated food environment data from Locatus, a commercial company collecting information on food retailers. The analysis focused on the period from 2004 onwards due to data availability. A total of 9901 children were initially part of the Generation R Study, with 4235 children ultimately included in the main analysis after various exclusions (e.g., younger siblings, missing data, children not living at one address for a sufficient duration). The food environment was assessed using a 400m buffer around the child's home, considering the number of fast-food outlets (absolute and relative exposure) and an overall healthiness score of all food outlets. Maternal education level was used as a socioeconomic indicator, categorized into four levels. Other sociodemographic variables included age, sex, ethnicity, and net household income. Linear regression models were used to analyze changes in the food environment over time (standardized to 7.1 years), stratified by maternal education. Fixed-effects linear regression models with repeated measurements were applied to examine within-person changes in the food environment and their association with changes in body composition (BMI, FMI, FFMI), controlling for time between measurements and interactions with maternal education level. Two sensitivity analyses were conducted: one adjusting for net household income and another including changes due to residential moves. R version 3.4.1 and the plm package were used for statistical analysis, employing clustered sandwich estimators to account for within-child correlation. P-values <0.05 were considered statistically significant.

At all time points, children from lower-educated mothers were exposed to more fast-food outlets, both in absolute and relative terms, and experienced a less healthy food environment. Over a median follow-up of 7.1 years, the absolute and relative increases in fast-food outlets were significantly greater for children from lower-educated mothers. The healthiness of the food environment did not differ significantly across maternal education levels. Importantly, changes in the food environment (fast-food outlets, healthiness score) were not significantly associated with changes in BMI, FMI, and FFMI in the overall population. However, in a subgroup analysis focusing on children without initial fast-food exposure, the introduction of fast-food outlets was associated with small increases in BMI, particularly among children from lower-educated mothers. These increases were more pronounced in FMI (fat mass index) than FFMI (fat-free mass index), but the confidence intervals were wide. Sensitivity analyses adjusting for household income and including residential moves did not significantly alter the results.

This study's findings demonstrate widening socioeconomic inequalities in exposure to fast-food outlets. While children from lower-educated mothers experience more significant increases in fast-food exposure over time, there's limited evidence of an independent effect of changes in fast-food access on body composition in already saturated fast-food environments. The observed effect for children from lower-educated mothers with new fast-food exposure highlights the potential vulnerability of this group. The absence of a widespread impact on body composition measures might be attributed to the already high prevalence of fast-food outlets, suggesting that incremental changes may not have a substantial effect in such contexts. However, even small shifts in distribution can have a major population-level effect on disease risk and contribute to health disparities. The study's strength lies in its longitudinal design using fixed-effects models and objective body composition measures. Limitations include potential confounding from unmeasured time-varying variables related to dietary choices and body composition. The lack of data on perceived exposure, food consumption patterns, and food environments outside the home address also limit the scope of conclusions.

This study provides evidence of increasing socioeconomic inequalities in children's exposure to unhealthy food environments. Although there is some indication that the introduction of fast food outlets is associated with small increases in BMI and FMI for children from low educated mothers with no initial exposure, the changes in the food environment were not largely associated with changes in body composition in an area already characterized by high fast-food outlet density. These findings highlight the need for policy interventions aimed at creating equitable food environments. Future research should explore causal mechanisms using instruments like consumption patterns and examine the cumulative impact of food environments across various daily settings.

The study's reliance on maternal education as a sole socioeconomic indicator might not fully capture the complexity of socioeconomic factors. Unmeasured time-varying confounders could affect the results, although fixed-effects models address time-invariant confounders. Furthermore, the focus on home addresses may not encompass the full range of locations where children are exposed to food. The lack of information on dietary behaviors, and the limited evaluation of other environments, such as school environments or physical activity settings, prevented further causal assessments. Lastly, the study did not utilize instrumental variable approaches, potentially not fully capturing some of the unobserved confounders.