Identifying Key Predictors of Mid-Childhood Obesity in a Population-Based Cohort Study: An Evidence Synthesis and Predictive Modeling Study.

Abstract

A wide spectrum of predictors of childhood overweight and obesity has been identified over past decades, yet a quantitative overview of multidisciplinary predictors is missing, and the relative importance and their collective contribution to childhood obesity remains unclear. We synthesized evidence from 93 published studies from the Generation R Study, a population-based prospective cohort from early-pregnancy onwards in the Netherlands, to provide a quantitative overview of 210 predictors across preconception and childhood associated with body mass index (BMI) in mid-childhood and selected 59 candidate predictors. Then, we further identified 32 key predictors for either zBMI or weight status using model search algorithms and built prediction models with data from the same cohort. Associations of note were identified among predictors in preconception and prenatal parental health, early-life weight, and child behavior domains. An interquartile range increase in parental prepregnancy BMI (maternal β = 0.17, 95% CI: 0.14, 0.20; paternal β = 0.17, 95% CI: 0.14, 0.21), early-life weight (e.g., zBMI at age 2y β = 0.28, 95% CI: 0.23, 0.33), and food responsiveness (β = 0.25, 95% CI: 0.21, 0.28) were positively associated with mid-childhood BMI, and satiety responsiveness was negatively associated (β = -0.18, 95% CI: -0.21, -0.15). Together, identified key predictors accounted for 49.2% of the variance of BMI and 33.7% of the variance of the odds of overweight and obesity at the age of 10 years. Multimodal interventions targeting parental BMI before pregnancy, early adiposity rebound, and management of appetite in early childhood may be most effective in maintaining a healthy weight during childhood.