Background Inequalities in childhood obesity in the United Kingdom are marked and have widened in recent years. Early years risk factors have been linked to obesity, and the social patterning of risk factors (with higher prevalence in disadvantaged groups) is likely to be an important driver of childhood obesity inequalities. However, the extent to which policy interventions addressing these modifiable risk factors would decrease obesity inequalities is unknown. As part of an NIHR-funded Obesity PRU project, we used a causal mediation approach to investigate whether the socioeconomic gradient in obesity at 7 years would reduce following a hypothetical intervention to change the prevalence of exemplar early risk factors (maternal smoking in pregnancy and breastfeeding behaviour) across the population to that found in the most advantaged families.

Methods Millennium Cohort Study data were used to model the relationship between household income in infancy (quintiles), any smoking in pregnancy (yes/no), breastfeeding initiation (yes/no) and obesity at 7 years (UK90 cut-off), accounting for baseline and intermediate confounders. The adjusted total causal effect (TCE) between income and obesity was compared to the interventional disparity measure-direct effect (IDM-DE), which represents the association between income and obesity that would remain if an intervention had changed the distributions of both early years risk factor mediators in all income groups to match those found in the highest income group.

Results Socioeconomic inequalities were present for smoking in pregnancy and breastfeeding and obesity at 7 years, and both mediators were associated with subsequent obesity. The population average prevalence of obesity was 8.9% in the highest income group and 13.9% in the lowest income group (TCE expressed as a risk difference[RD]: 4.9% (confidence interval[CI]:1.5–8.4). Following the hypothetical intervention, the RD between highest and lowest income groups fell by 2.6% (IDM-DE: 2.3%, CI: -1.1–5.8); a 52.9% relative change in the RD for obesity prevalence between these income groups.

Conclusion Interventions targeting early years risk factors have the potential to appreciably reduce population inequalities in childhood obesity, and further results will differentiate the separate effects of maternal smoking and breastfeeding. Analyses are based on the assumption of no unmeasured confounding and thus may suffer from some bias, although we controlled for several confounders. While these findings hold promise, the evidence points to inequalities remaining even if risk factors were equalised across social groups.