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

The study addresses whether socioeconomic position (SEP) is associated with differential exposure to the local food environment during childhood and whether changes in that environment contribute to changes in body composition. Childhood obesity has risen substantially alongside shifts toward greater availability of energy-dense and ultra-processed foods and more food outlets in residential areas. Evidence for causal effects of changing food environments on child obesity is limited, largely cross-sectional, and potentially confounded by residential self-selection. Prior fixed-effects research in U.S. children paradoxically suggested that increases in fast-food outlets were linked to small BMI reductions, warranting replication in other contexts. The authors hypothesized that children of lower SEP (proxied by maternal education) live in neighborhoods with more unhealthy food outlets and that over time these exposures would lead to more unfavorable changes in body composition. They further examined whether effects of changes in the food environment on BMI, FMI, and FFMI differ by maternal education.

Prior work indicates socioeconomic disparities in food environments: a review of 21 studies found higher fast-food access in more-deprived areas. Time-use studies suggest individuals of lower SEP may spend more time in their neighborhoods, increasing exposure to local food outlets. However, a systematic review (mostly cross-sectional) did not confirm consistent SEP modification of food environment–obesity associations. Longitudinal evidence is scarce; one U.S. fixed-effects study in children found that increases in fast-food outlets were associated with small BMI reductions. Other longitudinal studies in adults showed that food environment changes can affect diet or BMI, with stronger effects in low-income groups or among those with obesity. Reviews also highlight potential bias from unmeasured confounding and the importance of differentiating absolute versus relative food outlet measures. These gaps motivated a longitudinal, fixed-effects approach with repeated objective measures in children to assess SEP inequalities and potential impacts on body composition.

Design and population: Prospective analysis within the Generation R birth cohort (Rotterdam, The Netherlands). Eligible participants had at least two outcome measures (BMI or DXA-derived indices) that could be linked to food environment data between ages 4 and 14 years. Exclusions: younger siblings (n=397) and missing maternal education (n=441). The main analytic sample included children with at least two observations while living at the same address, yielding 4235 children with 11,277 BMI person-observations and 6240 FMI/FFMI person-observations. Median within-address follow-up: 7.1 years. Ethics approval and written informed consent were obtained. Body composition: Height and weight at ages 4 (child health centers), 6, 10, and 14 years (research center) with standardized protocols. BMI calculated and converted to sex- and age-specific SDS using Dutch reference charts. DXA (iDXA, GE-Lunar; enCORE v12.6) at ages 6 and 10 years provided fat mass and fat-free mass; FMI = fat mass (kg)/height (m³), FFMI = fat-free mass (kg)/height (m³); SDS for FMI/FFMI computed from cohort distributions. Food environment exposure: Food retailer data (location/type) from Locatus annual field audits (validated). For each outcome time-point, exposures were linked to the prior year’s food environment. Using ArcGIS, Euclidean 400 m buffers around home addresses counted outlets. Exposures: (1) absolute fast-food exposure = count of fast-food outlets (fast-food, grillroom/kebab, takeaway, ice cream shops); (2) relative fast-food exposure = fast-food outlets / total food outlets (%); (3) healthiness score = mean of outlet healthiness ratings (range −5 very unhealthy to +5 very healthy) from a Dutch Delphi-based index (fast-food −4.9; green-grocer +4.8). Addresses without any food outlet were excluded from the healthiness score analyses. Socioeconomic stratifier and covariates: Maternal education (measured at child age 6) categorized as high, mid-high, mid-low, low; for interaction tests, dichotomized into higher (mid-high, high) vs lower (low, mid-low). Other variables included sex, ethnicity (Dutch, other-Western, non-Western), and net household income (low, intermediate, high) at ages 6, 10, 14 years (imputation for occasional missing assuming stability when appropriate). Statistical analysis: Descriptive characteristics at first and last within-address measurements; standardized changes over 7.1 years to account for varying follow-up durations. Kernel density plots described within-person exposure changes. Linear regression tested change over 7.1 years in exposures across four maternal education levels. Associations between within-person changes in exposures and within-person changes in outcomes were estimated using individual-level fixed-effects linear models with a first-difference specification, adjusting for time between measurements. Interactions by maternal education were tested; for power, education dichotomized (lower vs higher). Effect units: +10%-points in relative fast-food, +1 outlet absolute fast-food, +0.5 point in healthiness score (healthier). Subgroup analysis restricted to children with no fast-food outlets at baseline assessed the impact of fast-food introduction. Sensitivity analyses: (1) additional adjustment for time-varying household income (ages 6, 10, 14); (2) inclusion of observations spanning residential moves (expanded sample: 4594 children; 13,528 BMI and 7856 fat-mass person-observations). Fixed-effects implemented in R (plm), clustered sandwich SEs; two-sided P<0.05.

- Socioeconomic inequalities in exposure: At all time-points (ages 4–14), children of lower educated mothers had higher absolute and relative exposure to fast-food outlets and unhealthier overall food environments; they also had higher BMI and FMI on average. - Widening disparities over time: Over a median 7.1 years, increases in fast-food exposure were larger for children of lower educated mothers than for higher educated mothers: absolute +0.6 outlets (95% CI: 0.4–0.8) and relative +2.0%-points (95% CI: 0.7–3.4). The average healthiness of the food environment decreased over time overall, but trends did not differ by maternal education. - High baseline exposure: At first measurement, 39.7% of children of low educated mothers vs 20.7% of children of high educated mothers had ≥4 fast-food outlets within 400 m of home. - No overall association with body composition: In fixed-effects models, within-person changes in relative fast-food exposure (+10%-points), absolute fast-food exposure (+1 outlet), or healthiness score (+0.5 point) were not significantly associated with within-person changes in BMI SDS, FMI SDS, or FFMI SDS in either lower or higher education groups (e.g., BMI SDS per +1 outlet: 0.01 [−0.01; 0.02] in both groups; BMI SDS per +0.5 healthiness point: 0.00 [−0.02; 0.02] lower education, 0.01 [−0.01; 0.02] higher education). - Subgroup with no initial fast-food exposure: Among children of lower educated mothers without fast-food outlets at baseline, increases in fast-food exposure were associated with small BMI increases (e.g., BMI SDS per +10%-points relative fast-food: 0.04 [0.00; 0.07]); effects were driven primarily by fat mass (FMI) but with wide confidence intervals. No such associations were observed among children of higher educated mothers. - Sensitivity analyses: Additional adjustment for time-varying household income did not alter findings. When including exposure changes due to residential moves, the associations observed in the main analyses were not present.

The study demonstrates persistent and widening socioeconomic inequalities in children’s exposure to fast-food outlets within residential areas. Children of lower educated mothers not only faced more fast-food outlets and a less healthy food environment at baseline but also experienced faster increases in fast-food exposure over time. Despite these disparities, changes in fast-food access or overall food environment healthiness were not associated with concurrent changes in BMI, FMI, or FFMI in the overall cohort. A plausible explanation is saturation: in a context with already high density of food outlets (“fast-food paradise”), marginal additions may not measurably affect body composition over 2–4 year intervals. However, among lower-SEP children initially unexposed to fast-food, the introduction of fast-food outlets coincided with small BMI increases, suggesting that effects may be more detectable at the low end of exposure or among vulnerable groups. Findings align with literature indicating unequal distribution of unhealthy food environments and potential SEP-specific sensitivity to environmental changes. Policy implications include prioritizing structural, equity-focused strategies to improve food environments, as educational approaches alone may be insufficient. The absence of strong short-term effects does not preclude meaningful cumulative impacts across time and settings; thus, population-level shifts in the distribution of exposures remain relevant for reducing health inequalities.

This longitudinal fixed-effects study in a large urban Dutch cohort provides evidence that socioeconomic inequalities in the residential food environment widen during childhood: children of lower educated mothers are increasingly exposed to fast-food outlets and an overall less healthy foodscape. In an environment with ubiquitous fast-food availability, additional outlets did not translate into measurable changes in BMI, FMI, or FFMI overall over typical 2–4 year intervals. However, among lower-SEP children with no prior fast-food exposure, introducing fast-food outlets was associated with small BMI increases. These findings support equity-focused, structural interventions to improve local food environments. Future research should clarify causal mechanisms linking outlet exposure to purchasing and dietary behaviors, incorporate perceived access and exposure across multiple daily settings (e.g., schools), examine longer-term cumulative effects, and explore advanced causal designs (e.g., valid instrumental variables, natural experiments) to address time-varying confounding.

- Potential residual confounding from unmeasured time-varying factors (e.g., dietary preferences, financial stress) despite fixed-effects modeling. - Lack of data on perceived exposure/access and on actual purchasing or dietary behaviors to elucidate mechanisms. - Food environment assessed only around home; exposures at schools or other activity spaces and changes in physical activity environments were not captured. - Analyses of healthiness score excluded addresses with zero outlets; exposure metrics may not fully reflect use. - Instrumental variable approaches that could address time-varying confounding were not implemented due to challenges in identifying valid instruments. - Associations observed in main analyses were not present when including residential moves, suggesting other factors accompanying moves (e.g., school changes) may influence body composition. - Differential follow-up duration by SEP necessitated standardization and may influence precision.