Multidimensional Analysis of Early Adiposity Rebound Determinants in Children: Two Key “1000 Days”
Ling Luo, Xing Wang, Juan Tong, Yongkang Liu, Ping Lv, Qizheng Huang, Jie Wang, Cong Geng, Hong Gan, Menglong Geng, Beibei Zhu, Shuman Tao, Hui Gao, Xiulong Wu, Kun Huang, Shuangqin Yan, Fangbiao TaoABSTRACT
Objective
Adiposity rebound (AR), the second rise in BMI during growth, increases the risk of obesity and metabolic diseases when it occurs early. We aimed to identify key factors influencing early AR (EAR) within two critical 1000‐day periods and develop a predictive model.
Methods
Based on the Ma'anshan Birth Cohort (MABC), we collected serial BMI from birth through school age and applied linear mixed models to construct individual growth curves for determining AR. EAR was defined as AR timing at or below the sex‐specific 25th percentile (≤ 55.46 months in boys and ≤ 60.54 months in girls) in MABC. Among the 2072 parent‐offspring trios included, 65 potential factors were collected from maternal, pregnancy, paternal, and offspring sources. Ten machine learning techniques and SHAP were used to evaluate feature importance.
Results
Multimethod screening identified EAR determinants, including paternal and maternal pre‐pregnancy overweight/obesity, pregnancy organic phosphate esters (OPE) exposures, offspring high genetic risk (the predominant contributor), catch‐up BMI trajectory, and high waist circumference trajectory. The categorical boosting model demonstrated the best performance among 10 machine learning models for predicting EAR (AUC training = 0.764, 95% CI = 0.733–0.795; AUC testing = 0.721, 95% CI = 0.669–0.773).
Conclusions
EAR determinants may inform future interventions, including weight management, prenatal OPE exposure reduction, and early screening for genetically high‐risk individuals.