Sex prediction from human tooth dimension by ROC curve analysis: a preliminary study.

Though mathematical approaches are effective in developing reliable odontometric equations with high predictive accuracy for sex estimation in forensic odontology, these methods are often complex to apply, making them impractical in certain scenarios. Consequently, efforts continue to refine these models for greater accessibility. This study aims to develop a sex prediction method based on the horizontal width of anterior teeth using ROC analysis and evaluate its accuracy. Maximum horizontal measurements were taken from all anterior teeth on 100 dental casts from male and 100 from female patients. ROC curve analysis, along with Youden's index and Euclidean index, was used to determine cut-off points for sex estimation. An additional set of 50 male and 50 female dental casts was used to assess the method's accuracy. Results showed that only the canines were statistically significant for sex identification. AUC values were 0.634 and 0.626 for maxillary canines (p = .001) and 0.724 and 0.727 for mandibular canines (p < .001), with accuracy ranging from 69 to 77%. In conclusion, a sex prediction method based on the horizontal width of the canine, using ROC analysis, was developed, demonstrating reasonable accuracy. This methodological improvement is expected to enhance its practical application in forensic investigations.