Bias and accuracy of age estimation using developing teeth in 946 children.

Developing teeth are used to assess maturity and estimate age in several disciplines. The aim of the study was to determine which of the most well known dental age estimation methods was best at estimating age. The target sample of dental radiographs (N = 946, ages 3–16) was described by Maber et al. (Forensic Sci Int 159 (2006) S68–S73). Seven mandibular permanent teeth (I1–M2) were assessed, and dental age was calculated using four dental maturity scales and fifteen methods that use data for individual teeth. The mean difference between dental age and real age was calculated (bias) as well as several other measures of accuracy (mean/median absolute difference, percentage aged to within six months and to within 10% of real age). Most methods estimated age with significant bias and standard deviation of bias ranged from 0.86 to 1.03 years. Analysis by age group showed most methods over-aged younger children, and considerably under-aged older children. The method that performed best was the dental maturity scale of Willems et al. (J Forensic Sci 46 (2001) 893–895) with bias of −0.14 ± 0.86 years (N = 827), mean absolute difference of 0.66 years, 71% aged to 10% or less of age, and 49% aged to within six months. Two individual teeth, P2 and M2, estimated age with bias not significantly different to zero for most formation stages using methods based on a large reference sample (L9a Demirjian stages) and a uniform age distribution (N25a Moorrees stages). Standard deviation of bias was least for early crown stages and most for late root stages. Methods that average ages for individual teeth improve if schedules for ‘mean age entering a stage’ are adjusted for prediction. Methods that directly calculate ‘mean age within stage’ can be improved by drawing from a uniform age distribution.