Dental age estimation in Somali children using the Willems et al. model

The aim of the current study was to retrospectively collect dental panoramic radiographs from Somali children living in Finland, to use the radiographic data to develop a new age estimation model based on the model established by Willems et al. (J Forensic Sci 46(4):893-895, 2001), and to compare the age prediction performances of the Willems et al. model (WM) and the newly developed model.

Dental panoramic radiographs from 808 healthy Somalis born in Finland were selected. The development of the seven left mandibular permanent teeth, from the central incisor to the second molar, was staged according to Demirjian et al. (Hum Biol 45(2):211-227, 1973). Radiographs with all listed permanent teeth completely developed were excluded. The studied sample consisted of 635 subjects (311 females, 324 males) ranging in age from 4 to 18 years. Kappa and weighted Kappa statistics were used to quantify intra- and inter-observer agreement in stage allocation. The collected dataset was used to validate the WM, constructed on a Belgian Caucasian reference sample, and to establish a Somali-specific age estimation model (SM) based on the WM. Both models were validated and their age prediction performances quantified using mean error (ME), mean absolute error (MAE) and root mean squared error (RMSE).

The SM resulted in a slight underestimation of age when the sex groups were analysed separately or combined, with ME varying between 0.04 (standard deviation (SD) 1.01) and 0.05 (SD 1.04) years, MAE between 0.77 and 0.80 years and RMSE between 1.01 and 1.04 years. The WM statistically significantly underestimated the age of females, with an ME of 0.20 (SD 1.01) years (p = 0.0006). For males, and for females and males combined, no statistically significant ME was observed.

The WM and SM were similar in their age prediction performances, and the use of the WM in dental age assessment in the Somali population is justified.