Longitudinal Prediction Of Childhood Caries Risk Using Salivary Microbiota

The objective of this study was to assess whether dynamic changes in salivary microbiota can predict caries risk in children. Using longitudinal data from 120 children, unstimulated saliva samples were collected at baseline, 6 months, and 12 months and analyzed by 16S rRNA gene sequencing. Differential taxa were identified through LEfSe and random forest analyses, and an optimal microbial feature set was selected via an incremental modeling strategy. Predictive models—including logistic regression, XGBoost, random forest, and support vector machine—were constructed and evaluated with time series cross-validation. Key taxa, such as Neisseria, Actinomyces, Capnocytophaga, and Fusobacterium, were significantly associated with caries risk. The optimal feature combination achieved an AUC of 0.74 in 6-month prediction and similar performance was observed in the 12-month prediction, with stable sensitivity and specificity across models. Overall, the dynamic microbial features demonstrated robust predictive capability. Dynamic monitoring of salivary microbiota effectively predicts caries risk in children and provides a promising basis for early screening.