Delayed Dental Development in Children With Non-Syndromic Hypodontia: A Cross-Sectional Study Using a Machine Learning Approach to Dental Age Estimation.

OBJECTIVES

To investigate the influence of non-syndromic hypodontia on radiographic dental development and to estimate dental age in children with bilateral mandibular agenesis using a machine learning approach.

MATERIALS AND

METHODS

This retrospective cross-sectional study included 626 children aged 6-15 years (311 with dental agenesis, 315 matched controls). Dental age (DA) was assessed using the original Demirjian method. In cases with bilateral congenitally missing teeth in the mandible, DA was estimated using supervised machine learning models based on specific random forests, following a secondary-diagnosis approach. Dental developmental delay was calculated as the difference between dental and chronological age (DA-CA) and compared across groups defined by agenesis status, severity and pattern. Multiple linear regression was applied to evaluate the effects of hypodontia, sex, chronological age and their interactions on DA-CA.

RESULTS

Eight random-forest models were trained, achieving good age-prediction accuracy (MAE = 0.08-0.28 years, R2 > 0.95). A 0.77-year difference in dental development (95% CI 0.61-0.94) separated children with hypodontia from controls (p < 0.001). The regression model confirmed that agenesis status, sex and CA were significant predictors of DA-CA, with an interaction between sex and agenesis. Unilateral or bilateral agenesis of mandibular second premolars was associated with a dental developmental delay compared with controls (padj < 0.05).

CONCLUSION

Our results were consistent with broader evidence linking hypodontia to altered developmental timing. Machine learning imputation offers a robust approach for missing teeth and can be implemented for age estimation in larger cohorts for orthodontic or forensic purposes.