Using synthetic growth trajectories to predict childhood obesity trends at the individual and population level

Introduction: Nationally-representative longitudinal height and weight data spanning childhood are surprisingly scarce. Pooling data covering shorter periods, we constructed BMI (kg/m2) trajectories across childhood to predict obesity risk during young adulthood from early childhood obesity.

Methods: Within-person height and weight trajectory segments were pooled from National Longitudinal Survey of Children and Young Adults (NSLY) (1986-2010, n=9,402) and National Longitudinal Study of Adolescent Health (1996-2009, n=4,972). Overlapping height and weight trajectories from early childhood and adolescence were jointly matched using Bayesian methods.  A cohort of one million 5 year olds (5yo) was matched to anthropometric and sociodemographic measures from National Health and Nutrition Examination Survey (NHANES) (2005-2010,n=565 5yo) and simulated through age 19 years.  Projected mean BMI and obesity prevalence was validated against NHANES 2005-2010 data.

Results: At baseline 15.57% of 5yo were obese [mean BMI:16.42].  At age 19 years, the predicted mean BMI was 25.49 (95% CI:25.48-25.50) and 17.38% were obese (95% CI:17.31%-17.44%),  similar to the observed mean BMI [25.27 (95% CI: 24.60-25.96)] and obesity rate [17.43% (95% CI: 13.04%-21.82%)] in NHANES.  Obese 5yo were more likely to be obese at age 19 compared to non-obese (38.4% vs. 11.5%) (relative risk (RR) 3.37 (95% CI:3.34-3.40). 

Conclusion: We created realistic long-term trajectories from shorter span, observed data points.  We found that childhood obesity is a strong predictor of early adult obesity, but that substantial numbers of non-obese children became obese by early adulthood as well, strengthening the case for population-based obesity prevention efforts as well as targeted interventions for obese children.