BMI-for-age Z-score trajectories in early childhood: a comparison of GMM and KML and associated antenatal and postnatal factors

BACKGROUND: Early-life growth trajectories, especially during the first two years, are important for future health. However, in the context of rapidly rising childhood obesity, it is essential to characterize BMI-for-age Z-score (BAZ) trajectories in early childhood, compare optimal modeling approaches, and identify associated risk factors. METHODS: This prospective birth cohort study enrolled participants in Jinan, Shandong from 2019 to 2022. World Health Organization (WHO)-standardized BAZ scores were calculated for children at eight timepoints (0-24 months). Sex-specific BAZ trajectories were derived using growth mixture modeling (GMM) and k-means for longitudinal data (KML). The possible risk of being overweight (BAZ > 1) at age three years was predicted using logistic regression with five folds cross-validation, comparing model performance via the area under the curve (AUC). Associations between trajectory classification and antenatal and postnatal factors were examined. RESULTS: A total of 4629 mother-infant dyads were enrolled in this study. The KML-derived four-class trajectory model demonstrated optimal predictive performance for possible risk of being overweight at age three years, achieving an AUC of 0.749 and significantly outperforming the GMM model (AUC = 0.571). Male sex [odds ratio (OR) = 1.314; 95% confidence interval (95% CI): 1.098, 1.572], high birth weight (OR = 4.739; 95% CI: 3.594, 6.250), multiparity (OR = 1.265; 95% CI: 1.003, 1.594), and gestational hypertension (OR = 2.581; 95% CI: 1.387, 4.801) were identified as risk factors for high growth trajectory. CONCLUSIONS: The KML model was identified as the more effective approach for delineating BAZ trajectories and predicting the risk of being overweight in our study population. Furthermore, male sex, high birth weight, multiparity, and gestational hypertension were identified as key antenatal and postnatal factors associated with higher growth trajectories.