Human dental age estimation using third molar developmental stages: Accuracy of age predictions not using country specific information P.W. Thevissen a , A. Alqerban b , J. Asaumi c , F. Kahveci d , J. Kaur a , Y.K. Kim e , P. Pittayapat a , M. Van Vlierberghe a , Y. Zhang f , S. Fieuws g , G. Willems a, * a Katholieke Universiteit Leuven, Faculty of Medicine, School of Dentistry, Forensic Odontology Department, Kapucijnenvoer 7, Leuven B-3000, Belgium b Riiyadh Military Hospital (RKH), Dental Department, Building 131, Riyadh, Saudi Arabia c Department of Oral and Maxillofacial Radiology, Field of Tumor Biology, Okayama University Graduate School of Medicine and Dentistry, 5-1, Shikata-cho, 2-Chome, Okayama 700- 8525, Japan d Gulumse Agiz Ve Dis Sagligi Klinigi, Perpa t.m. b blok kat:8, Okmeydani, Sisli, Turkey e Department of Oral Medicine and Oral Diagnosis, School of Dentistry & Dental Research Institute, Seoul National University, Yunkeon-Dong 28, Chongro-Ku, Seoul 110-749, Republic of Korea f Dalian Stomatological Hospital, No. 959 Changjiang Road, Shahekou District 116021, PR China g I-BioStat, Katholieke Universiteit Leuven and Universiteit Hasselt. Kapucijnenvoer 35, Blok D, bus 7001, B-3000 Leuven, Belgium 1. Introduction Forensic age estimations on living individuals are requisi- tioned to verify the chronological age of individuals who are, willfully or not, unable to deliver documentary evidence about Forensic Science International 201 (2010) 106–111 ARTICLE INFO Article history: Received 16 March 2010 Received in revised form 20 April 2010 Accepted 25 April 2010 Available online 18 May 2010 Keywords: Forensic odontology Age determination by teeth Third molars Country specific populations Bayesian Accuracy ABSTRACT Unquestionable forensic age investigations are based on statistical models constructed on a sample containing subjects of identical origin as the examined individual. In cases where corresponding models are unavailable, the established report has to describe the possible effects of this unrelated information on the predicted age outcome. The aim of this study is to collect country specific databases of third molar development and to verify how the related dental age estimations are influenced if we were to use dental developmental information only from Belgium or from all collected countries together. Data containing third molar developmental stages scored following Gleiser and Hunt (modified by Ko ¨ hler) were collected from 9 country specific populations (Belgium, China, Japan, Korea, Poland, Thailand, Turkey, Saudi-Arabia and South-India). Age predictions were obtained from a training dataset and validated on a test dataset. Bayes rule using the repeated third molar scores is applied to get age predictions and prediction intervals. Three age predictions were compared for males and females separately. For the first prediction, the training dataset contains only Belgian subjects. For the second prediction, the training dataset for each country consists only of subjects of the country itself. For the final prediction, subjects from all countries are pooled into one common training dataset. Besides the (absolute) difference between the chronological age and the predicted age, specific interest lies in the juvenile–adult distinction. In the age range from 16 to 22 years 6982 subjects (3189 male and 3793 female) were analyzed. Using information on third molar development from Belgium compared to information from the country specific databases hardly increased the mean absolute differences (MAD) and mean squared errors (MSE): the MAD and MSE increased on average with 0.5 and 2.5 months with maximal increases of, respectively 1.6 and 7.3 months. Using information from all countries pooled compared to country specific information provided even on average negligible increases (0.05 and 0.2 months for MAD and MSE, respectively). For the juvenile– adult discrimination, using information from all countries instead of country specific information yielded comparable performances. Using Belgium instead of country specific information increased the percentage of correctly identified juveniles, but decreased the percentage of correctly identified adults. The adult–juvenile discrimination based on information used from Belgium provides judicially the best applied reference. ß 2010 Elsevier Ireland Ltd. All rights reserved. * Corresponding author. Tel.: +32 16 33 24 59; fax: +32 16 33 75 78. E-mail address: guy.willems@med.kuleuven.be (G. Willems). Contents lists available at ScienceDirect Forensic Science International journal homepage: www.elsevier.com/locate/forsciint 0379-0738/$ – see front matter ß 2010 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.forsciint.2010.04.040