Dental age estimation and assessment of the 18-year threshold using the third molar maturity index and dental development stages in a Korean population

Third molar development is widely used for forensic age estimation and for classifying individuals as adults or minors based on the legal age threshold. This study aimed to evaluate the applicability of Cameriere’s third molar maturity index and three stage-based indicators, Demirjian, Köhler, and Liversidge, for assessment at the 18-year threshold and age estimation in a Korean population. A total of 2,800 panoramic radiographs from individuals aged 15.00 to 21.99 years (1,400 males and 1,400 females) were analyzed. I 3M was measured for the left and right mandibular third molars, and dental development stages were assessed using the Demirjian, Köhler, and Liversidge methods. Youden’s index from the receiver operating characteristic curves was used to determine Korean-specific I 3M cut-off values and cut-off values for the predicted probabilities of being an adult derived from logistic regression models. The I 3M cut-off value approach and logistic regression models were developed for classification at the 18-year threshold, while linear regression models were constructed for age estimation. For the LMM3, the Korean-specific cut-off values were 0.18 in males, 0.17 in females, and 0.17 in the combined-sex sample. For the RMM3, the corresponding values were 0.16, 0.15, and 0.16, respectively. These cut-off values generally improved sensitivity and accuracy compared with the Cameriere cut-off value of 0.08, which showed greater specificity. In the logistic regression models, depending on sex and side, I 3M typically yielded higher sensitivity, whereas stage-based indicators yielded higher specificity, with both maintaining comparable accuracy. For age estimation, prediction errors were similar among the four evaluated dental maturity indicators, with only minute differences, typically within approximately 0.2 years. The I 3M cut-off value approach and logistic regression models may be applicable for classification at the 18-year threshold. For age estimation, stage-based indicators showed practical feasibility in linear regression-based models, while I 3M may also have potential for linear regression-based age estimation. The cut-off values and regression models established in this study may provide useful reference data for forensic age estimation and classification in Korean adolescents and young adults, although further external validation is required before broader forensic application.