PAPER ANTHROPOLOGY Stephanie E. Calce, 1 M.Sc. and Tracy L. Rogers, 2 Ph.D. Evaluation of Age Estimation Technique: Testing Traits of the Acetabulum To Estimate Age at Death in Adult Males* ABSTRACT: This study evaluates the accuracy and precision of a skeletal age estimation method, using the acetabulum of 100 male ossa coxae from the Grant Collection (GRO) at the University of Toronto, Canada. Age at death was obtained using Bayesian inference and a computational application (IDADE2) that requires a reference population, close in geographic and temporal distribution to the target case, to calibrate age ranges from scores generated by the technique. The inaccuracy of this method is 8 years. The direction of bias indicates the acetabulum technique tends to underestimate age. The categories 46–65 and 76–90 years exhibit the smallest inaccuracy (0.2), suggesting that this method may be appropriate for individuals over 40 years. Eighty-three percent of age estimates were €12 years of known age; 79% were €10 years of known age; and 62% were €5 years of known age. Identifying a suitable reference population is the most significant limitation of this technique for forensic applications. KEYWORDS: forensic science, forensic anthropology, skeletal age estimation, acetabulum, adult male, os coxae The ability to estimate age at death of human remains is an important tool in medico-legal investigations. Recently, it has been suggested that the acetabulum provides age-related changes that may aid in accurate age estimation of unknown individuals (1–4). Studies by Rissech et al. (1,4), describe seven age changes in the fused acetabulum and propose a method for estimating age with 89–100% accuracy based on these characteristics (Fig. 1). Rissech et al. (1) conclude that acetabular observations enable accurate age at death estimates of adults over 40 years of age. The results of Rissech et al. (1) are significant for a number of reasons: (i) standard techniques for age estimation using the pubic symphysis and auricular surface exhibit inaccuracies for individuals over 40 years of age (5–11), (ii) most techniques lump individuals over 60 years of age into one age class (5–16), (iii) the acetabulum is more likely to survive postdepositional processes than fragile areas of the skeleton, such as the pubic symphysis (17), and (iv) few age estimation techniques report such high accuracy for indi- viduals over 40 years of age. The purpose of this research is to (i) test the precision of Rissech et al.’s (1) method of scoring each trait, (ii) evaluate the accuracy of age at death estimates for individ- uals over 40 years, and (iii) compare the results achieved by using different reference populations to determine the impact of choosing an inappropriate reference sample. Methods and Materials The Grant Collection (GRO), housed at the University of Tor- onto, is composed of mostly male subjects (n = 147) over the age of 40 (n = 133). Age at death is known for each individual (8). It is therefore an appropriate collection to test the method developed by Rissech et al. (1,4). A sample of 100 male individuals was ran- domly selected from the GRO and, as in the original publications, the left os coxae of each individual was examined. The nondestruc- tive method of evaluation proposed by Rissech et al. (1,4), com- prises close morphological examination of seven traits of the acetabular region: (i) acetabular groove, (ii) acetabular rim shape, (iii) acetabular rim porosity, (iv) apex activity, (v) activity of ace- tabular fossa, (vi) activity of outer edge of acetabular fossa, and (vii) porosities of the acetabular fossa (Figs 2–8). Each variable was scored based on a series of states within each observed mor- phological condition (1,4), for example acetabular groove can be scored as no groove [0]; groove [1]; pronounced groove [2]; or very pronounced groove [3]. See Table 1 for a description of each trait, variable states, and characteristics as published by Rissech et al. (1). Individuals with noninflammatory osteoarthritis or diffuse idiopathic skeletal hyperostosis were not excluded, as was the case in the original study. Linear regression was used to associate the degree of trait expression with age at death. Standard error of the estimate versus known age was conducted using a paired t-test. To test for intra-observer error, every third individual (for a total of 34 individuals) was selected from the same sample and sub- jected to re-examination by the author (Calce). Intra-observer error was calculated from the difference in years between the first and second age estimates (n = 34). A comparison of percentage wrong estimates greater than €5 years and p-value <0.05 from paired t-test was considered significant. Standard error of the age estimate, that 1 Department of Anthropology, University of Victoria, P.O. Box 3050, STN CSC, Victoria, British Columbia, Canada, V8W 3P5. 2 Department of Anthropology, University of Toronto at Mississauga, 3359 Mississauga Rd. N., Mississauga, Ontario, Canada L5L 1C6. *Presented in part at the Canadian Association for Physical Anthropology (CAPA) Annual Meeting, November 17, 2007, University of Calgary Alberta, Banff, Alberta, Canada. Received 27 July 2009; and in revised form 19 Oct. 2009; accepted 10 Dec. 2009. J Forensic Sci, March 2011, Vol. 56, No. 2 doi: 10.1111/j.1556-4029.2011.01700.x Available online at: onlinelibrary.wiley.com 302 Ó 2011 American Academy of Forensic Sciences