Background: Simulated interventions using observational data have the potential to inform policy and public health interventions where randomised controlled trials are not feasible. National childhood obesity policy is one such area. Overweight and obesity are primarily caused by energy-rich and low-nutrient diets that contribute to a positive net energy imbalance. Using data from the Avon Longitudinal Study of Parents and Children (ALSPAC), we investigated whether causal modelling techniques could be applied to simulate the potential impact of policy-relevant calorie-reduction interventions on population prevalence and inequalities in obesity in childhood.
Methods: Predicted probabilities of obesity at age 11 (UK90 cut offs) were estimated from logistic marginal structural models (MSM) accounting for observed calorie consumption at age 7 and confounding, overall and by maternal occupational social class. A series of population intervention scenarios were modelled to simulate daily calorie-reduction interventions that differed in effectiveness, targeting mechanism and programme uptake level.
Results: The estimated effect of maternal social class on obesity after accounting for confounding and observed calorie intake was provided by the controlled direct effect (CDE), in which, 18.3% of children were living with obesity at age 11 years,. A universal simulation to lower median intake to the estimated average requirement (EAR) (a 6.1% reduction in daily calories) with 75% uptake reduced overall obesity prevalence by 0.6%; there was little impact on inequalities. A targeted intervention to limit consumption to the EAR for children with above average intake reduced population obesity prevalence at 11 years by 1.5% but inequalities remained broadly unchanged. A targeted intervention for children of low-income families reduced prevalence by 0.7% and was found to slightly reduce inequalities.
Conclusions: MSMs allow estimation of effects of simulated calorie-reduction interventions on childhood obesity prevalence and inequalities, although estimates are limited by the accuracy of reported calorie intake. Further work is needed to understand causal pathways and opportunities for intervention. Nevertheless, simulated intervention techniques have promise for informing national policy where experimental data are not available.