Testing the Walker method in a Southeast Asian sample

Accurate estimation of biological attributes in unknown skeletons is crucial for routine anthropological assessment. Many extant forensic standards are derived from European and United States populations. Contemporary research, however, has shown that accuracy and reliability can vary when such standards are applied to individuals outside the reference sample (e.g. non-US samples). The validation of forensic standards is required to avoid misclassification and ensure scientifically sound interpretation of findings. The present study represents a preliminary assessment of the Walker (2008) sex estimation method in a Southeast Asian (Malaysian and Chinese) population. The sample comprises 77 (38 males, 39 females) high-resolution adult cranial multi-slice computer tomographic (MSCT) scans (≤1.0 mm). The MSCT scans were visualised using OsiriX® and assessed based on the descriptions and illustrations given by Walker (2008) of five cranial traits: nuchal crest; mastoid process; supraorbital margin; glabella; and mental eminence. An intra-observer test was developed to calculate the level of agreement using Cohen’s Kappa (K). Accuracy was assessed by comparing the estimated sex with the documented sex. Univariate logistic regressions were performed to calculate the accuracy of individual traits. The intra-observer agreement revealed substantial to almost perfect agreement for all traits (K=0.62 to 0.83), except for nuchal crest (K=0.55). After applying Walker’s six American/English population equations to the Southeast Asian sample, the overall accuracy achieved was between 76.9% and 93.5%. The sex bias value was ≤5% in all equations, except functions one (-5.5%), four (-15.9%) and five (41.0%). Univariate analysis indicated that the traits achieving the overall highest accuracy were the glabella (85.7%), orbital margin (76.7%) and mastoid process (76.5%). Considering the sex bias values for functions one, four and five in this population, they differed by 7.5,12.2 and 40.8% respectively compared to those reported by Walker (2008); the present study thus reaffirms that population-specific methods are crucial to avoid misclassifications.