Population effects and cross-population generalisability of the third molar maturity index for 18-year age classification: a multi-centre study with machine learning validation.

Accurate forensic age estimation is critical in legal contexts where documentation is unavailable. The Third Molar Maturity Index (I3M) is a quantitative, reproducible method based on the ratio of the open apical width to the length of the mandibular third molar. This multicentre, cross-sectional study evaluated I3M performance for age-threshold classification at 18 years across five populations. A total of 3000 orthopantomograms from individuals aged 13-22 years were collected in Brazil, China, Croatia, Portugal and Sweden. Teeth 38 and 48 were measured, and binary classification (≥ 18 vs. <18 years) was assessed using ROC analysis, hierarchical logistic regression with likelihood ratio testing, and machine learning with leave-one-population-out (LOPO) validation. I3M showed excellent discriminatory performance in all five populations (AUC: 0.917-0.942 for tooth 38). Inter-population differences in I3M distributions were significant; population was an independent predictor of 18-year-old classification. Intra-population sex differences were significant in three populations, with females showing higher I3M values. Optimal cut-off values varied from 0.126 (Croatia) to 0.223 (China). LOPO validation confirmed robust cross-population generalisability (AUC: 0.917-0.958). I3M is a reliable age classification tool across diverse ethnic groups. Population-specific thresholds are recommended when ethnic background is known; when unknown, a conservative threshold of 0.10-0.15 is advised to minimise misclassification of minors.