Willems method measures the developmental stages of the seven left permanent mandibular teeth and is frequently used for dental age estimation. The aim of this study was to test its accuracy on a large sample of the Croatian population and to develop new models for estimation based on one to seven mandibular teeth. The developmental stages were evaluated on the digital, standardized orthopantomograms of 1868 Croatian children aged 5 to 16. Univariate regression was used for age estimation based on one tooth and regression with forward and backward elimination for selection of the best combination on two to six teeth. Tested in parallel with Willems method, the accuracy of the new models was assessed within intervals of ± 0.5, ± 1, ± 1.5 and ± 2 years and shows the percentage of correct estimations. Using Willems method, the average overestimation was 0.41 years for boys and 0.22 years for girls. Newly developed models that use two to seven teeth proved to be significantly more accurate (p < 0.001). The accuracy of age estimation increases significantly with the number of teeth in a model. Predictably, within the interval ± 0.5 years, a model with three teeth has 3% fewer accurate age estimations than a model including seven teeth. In both theory and practice, Willems method has been considered suitable for age estimation among Croatian children. However, these newly developed models significantly surpass its accuracy. Models using two to seven teeth represent a simple, reliable, and accurate method for age estimation, even in cases with missing mandibular teeth.
Keywords: Croatia; Dental age estimation; Forensic odontology; Newly developed models; Willems method.