An evaluation of the Gilbert-McKern pubic symphyseal age estimation method using Bayesian statistics and principal component analysis: A computed tomographic exploration in an Indian population

Age of an individual constitutes a primary facet of human identification. In cases where skeletal remains present for examination, bony markers distributed throughout the skeletal framework are employed for age estimation. Amongst these markers, the pubic symphysis constitutes one of the more commonly utilized structures. Gilbert-McKern's pubic symphyseal age estimation method was derived to complement the original three component method, and permit accurate age estimation in females. However, subsequent investigations with the Gilbert-McKern method are limited, and completely lacking for an Indian population. In the present study, CT scans of 380 consenting individuals (190 males and 190 females) undergoing CT examinations for therapeutic purposes, aged 10 years and above, were scored according to Gilbert-McKern's three component method. A significant sexual dimorphism was observed with the scoring of the ventral rampart and symphyseal rim. An overall accuracy of 29.50% was obtained in females, indicating that the method lacks forensic utility in its primal form. Highest posterior density and highest posterior density region values were computed for each component using Bayesian analysis in both sexes, to enable age estimation from individual components, whilst overcoming issues of age mimicry. Amongst the three components, symphyseal rim furnished the most accurate and precise estimates of age, whereas the ventral rampart garnered highest error computations, in both sexes. Principal component analysis was utilized for multivariate age estimation by taking into consideration this differential contribution of individual components. Weighted summary age models, derived using principal component analysis, furnished inaccuracy values of 12.19 years and 12.30 years in females and males, respectively. Bayesian error computations obtained with the symphyseal rim in both sexes were even lower than those obtained with weighted summary age models, demonstrating its suitability as an independent age marker. Despite the use of statistical modalities of Bayesian inference and principal component analysis for age estimation, the method did not yield significantly reduced error rates in females, demonstrating its limited forensic applicability. While, statistically significant sex differences were observed with the scoring of Gilbert-McKern's components, concordant correlations, comparable accuracy and absolute error values were obtained for both sexes, indicating that the Gilbert-McKern method can be utilized to age either sex. However, inaccuracy and bias values obtained with different statistical modalities, as well as broad age intervals furnished with Bayesian analysis demonstrate the overall limited applicability of the Gilbert-McKern method in aging males and females of an Indian population.