Objectives
The aim of the present study was to investigate the risk factors for external root resorption (ERR) of the second molars (M2) associated with impacted third molars (M3) and to develop a prediction model for clinical assessing the risk of ERR based on the individual characteristics of M3.
Materials and methods
A total of 798 patients with 2156 impacted third molars were collected from three centers between December 1, 2018 and December 15, 2018. ERR was identified by cone beam computed tomography (CBCT)examinations. The effects of different risk factors on the presence/absence of ERR and its severity were analyzed using Chi-Square or Fisher test. Multivariate logistic regressive analysis with stepwise variable selection methods was performed to identify factors which were significant predictors for ERR and its severity. Subsequently, a prediction model was developed, and the model performance was validated internally and externally.
Results
The overall incidence of ERR of second molars was found to be 16.05%. Risk factors of ERR included age, position (upper or lower jaw), impact depth, impact type, contact position, M2 situation in opposing jaw and root number of M3. The prediction model was established using six factors including position, impact type, impact depth (including PG:A-B-C and PG:I-II-III),contact position and root number of M2. In terms of internal validation, the model achieved an AUC of 0.959 and a prediction accuracy of 0.896 (95% CI, 0.867, 0.921). The Kappa value was 0.572, with a sensitivity of 0.956 and a specificity of 0.571. Additionally, for external validation, the model demonstrated an AUC of 0.975 and a prediction accuracy of 0.910 (95% CI, 0.870–0.941). The Kappa value was calculated as 0.653, with a sensitivity of 0.941 and a specificity of 0.732.
Conclusion
A risk prediction model for ERR was established in the present study. Position (upper or lower jaw), impact type, impact depth, contact position and root number of M2 were identified as influencing variables which were significant predictors in the development of this predictive model. The prediction model showed great discrimination and calibration.
Clinical relevance:
This prediction model has the potential to aid dentists and patients in making clinical decisions regarding the necessity of M3 extraction.