Childhood Transitions Between Weight Status Categories: Evidence from the UK Millennium Cohort Study

Abstract

Background Assessing the cost-effectiveness of interventions targeting childhood excess weight requires estimates of the hazards of transitioning between weight status categories. Current estimates are based on studies characterized by insufficient sample sizes, a lack of national representativeness, and untested assumptions. Objectives We sought to (1) estimate transition probabilities and hazard ratios for transitioning between childhood weight status categories, (2) test the validity of the underlying assumption in the literature that transitions between childhood bodyweight categories are time-homogeneous, (3) account for complex sampling procedures when deriving nationally representative transition estimates, and (4) explore the impact of child, maternal, and sociodemographic characteristics. Methods We applied a multistate transition modeling approach accounting for complex survey design to UK Millennium Cohort Study (MCS) data to predict transition probabilities and hazard ratios for weight status movements for children aged 3–17. Surveys were conducted at ages 3 (wave 2 in 2004), 5 (wave 3 in 2006), 7 (wave 4 in 2008), 11 (wave 5 in 2012), 14 (wave 6 in 2015), and 17 (wave 7 in 2018) years. We derived datasets that included repeated body mass index measurements across waves after excluding multiple births and children with missing or implausible bodyweight records. To account for the stratified cluster sample design of the MCS, we incorporated survey weights and jackknife replicates of survey weights. Using a validation dataset from the MCS, we tested the validity of our models. Finally, we estimated the relationships between state transitions and child, maternal, and sociodemographic factors. Results The datasets for our primary analysis consisted of 10,399 children for waves 2–3, 10,729 for waves 3–4, 9685 for waves 4–5, 8593 for waves 5–6, and 7085 for waves 6–7. All datasets consisted of roughly equal splits of boys and girls. Under the assumption of time-heterogeneous transition rates (our base-case model), younger children (ages 3–5 and 5–7 years) had significantly higher annual transition probabilities of moving from healthy weight to overweight (0.033, 95% confidence interval [CI] 0.026–0.041, and 0.027, 95% CI 0.021–0.033, respectively) compared to older children (0.015, 95% CI 0.012–0.018, at ages 7–11; 0.018, 95% CI 0.013–0.023, at ages 11–14; and 0.018, 95% CI 0.013–0.025 at ages 14–17 years). However, the resolution of unhealthy weight was more strongly age-dependent than transitions from healthy weight to non-healthy weight states. Transition hazards differed by child, maternal, and sociodemographic factors. Conclusions Our models generated estimates of bodyweight status transitions in a representative UK childhood population. Compared to our scenario models (i.e., time-homogeneous transition rates), our base-case model fits the observed data best, indicating a non-time-homogeneous pattern in transitions between bodyweight categories during childhood. Transition hazards varied significantly by age and across subpopulations, suggesting that conducting subgroup-specific cost-effectiveness analyses of childhood weight management interventions will optimize decision-making.