Accuracy of age estimation in children using radiograph of developing teeth.

The aims of this study were: first, to determine the accuracy of the Cameriere method for assessing chronological age in children based on the relationship between age and measurement of open apices in teeth and, second, to compare the accuracy of this method with the widely used Demirjian et al. method and with the method proposed by Willems et al. Orthopantomographs taken from white Italian, Spain and Croatian children (401 girls, 355 boys) aged between 5 and 15 years were analysed following the Cameriere, Demirjian and Willems methods. The difference between chronological and dental age was calculated for each individual and each method (residual). The accuracy of each method was assessed using the mean of the absolute values of the residuals (mean prediction error). Results showed that the Cameriere method slightly underestimated the real age of children. The median of the residuals was 0.081 years (interquartile range, IQR=0.668 years) for girls and 0.036 years for boys (interquartile range, IQR=0.732 years). The Willems method showed an overestimation of the real age of boys, with a median residual error of -0.247 years and an underestimation of the real age of girls (median residual error=0.073 years). Lastly, the Demirjian method overestimated the real age of both boys and girls, with a median residual error of -0.750 years for girls and -0.611 years for boys. The Cameriere method yielded a mean prediction error of 0.407 years for girls and 0.380 years for boys. Although the accuracy of this method was better for boys than for girls, the difference between the two mean prediction errors was not statistically significant (p=0.19). The Demirjian method was found to overestimate age for both boys and girls but the mean prediction error for girls was significantly greater than that for boys (p=0.024), and was significantly less accurate than the Cameriere method (p<0.001). The Willems method was better than that of Demirjian (p=0.0032), but was significantly less accurate than that of Cameriere (p<0.001).