Abstract
Background/Aims There is increasing interest in the study of preventive and therapeutic interventions for childhood obesity. Without accounting for baseline variability in children’s weight measures, however, it is difficult to accurately explore the true impact of interventions. We wanted to identify a) what percentage of children and adolescents have large longitudinal within-individual variability in their standardized weight measures and b) groups of individuals with differential growth trajectories over time. Our hypothesis is that a substantial fraction of children have large longitudinal within-individual variability and/or a non-constant growth trajectory.
Methods We used a cohort of ~100,000 relatively healthy children and adolescents (2–20 years), seen in large ambulatory care organizations between 2000–2013, who had at least 3 weight measurements recorded longitudinally. The standardized weight measures we used were the weight-for-age z-score (WAZ) and the weight-for-age percentile (WAPCT). We quantified the within-individual variability by a) the slope and b) the root-mean-square-error (RMSE) of the regression of longitudinal standardized weight measures vs. age. Clusters of growth trajectories were identified using Growth Mixture Models (GMM). The number of clusters was determined by Akaike information criterion and relative cluster size.
Results The mean duration of longitudinal follow-up of individuals in our cohort was 4.7 yrs (median 4.2 yrs, IQR 3.9 yrs). Approximately 19% of all children and adolescents had substantial longitudinal within-individual variability (|slope| ≥ 0.02 and/or RMSE ≥ 0.35). We identified two clusters within this group: 1) those with an initial relatively constant growth-trajectory (between 2–9 years) and then a slightly upward trend (between 10–20 years), 61% of the total sample and 2) those with an initial upward trend (between 2–9 years) and then downward trend (between 10–20 years) in their growth-trajectory, 39% of the total. Large within-individual variability was identified in 19.5% of children in cluster 1 and 18.7% of children in cluster 2.
Conclusions Relatively healthy children and adolescents have large within-individual variability in their standardized weight measures that needs to be considered in study design when weight changes are used as study endpoints. Two subgroups exhibit potentially important growth patterns that warrant investigation.
