Evaluating postcranial macromorphoscopic traits to estimate ancestry among modern South Africans

Abstract

When decomposed human remains are recovered, the expertise of a forensic anthropologist is required to assist in the identification of the decedent. Identification involves establishing a biological profile, which includes the estimation of age, sex, ancestry, and stature of the individual. Robust methods are needed to assist in creating an accurate biological profile. While osteometric methods are currently preferred for ancestry estimation for forensic analyses in South Africa, non-metric methods can provide valuable information and need to be further explored. The current study aimed to assess postcranial macromorphoscopic traits as a tool to estimate ancestry among modern South Africans. A sample of 271 postcranial skeletons belonging to black, white, and coloured South Africans were used to score a series of eleven macromorphoscopic traits. The skeletal material was sourced from the Pretoria Bone Collection (University of Pretoria) and the Kirsten Skeletal Collection (Stellenbosch University). The intra- and inter-observer agreement ranged from fair to almost perfect for all but one trait (accessory transverse foramen of C1). The traits varied in frequency and rarity among the populations, with only seven traits demonstrating significant differences between at least two of the groups. Univariate and multivariate random forest models were created to test the positive predictive performance of the traits to classify ancestry. The univariate models performed poorly, with accuracies that ranged from 33.0% to 53.0%. The overall classification accuracy for the multivariate model incorporating all traits was not much better at 54.6% The results of the current study indicate that the postcranial macromorphoscopic approach does not outperform current methods employed to estimate ancestry. Furthermore, the low accuracies and Kappa values obtained with the random forest models suggest that the traits are not reliable classifiers, and as such, the method does not currently have practical applicability for medicolegal casework. However, with significant differences observed, more research needs to be conducted to potentially improve the method for use in South Africa.