Résumé
Age at death estimation is a major part of forensic anthropology, but is often poor in the latter decades of life and should benefit from specific population standards. The aim of this study was to test a virtual reference sample with a uniform age distribution in order to improve the accuracy of age estimation for individuals over 40 years of age. We retrospectively built a random virtual reference sample of pubic symphyses from 1100 clinical cases using computed tomography at two French hospitals; this was compared with a test sample (pubic symphyses from 75 corpses undergoing post-mortem computed tomography at a French forensic department) and with the sample originally used in the Suchey-Brooks (SB) system. Inaccuracy and bias were calculated and the proportions of cases in which the real age fell within the estimated age ± 1 standard deviation were calculated. Compared to using the SB sample, using our French sample resulted in a lower inaccuracy for males over 55 years and bias showed that that for males aged 56 to 70 years were less underestimated. Compared to using the SB sample, using our French sample resulted in a lower inaccuracy for females over 70 years and bias showed that that for females aged 56 and older were less underestimated. This study presents a large data set of pubic symphyseal phases from a French virtual sample that allows for improving age estimation accuracy at death, particularly for individuals over 40 years. This kind of material can be useful to improve the age estimation accuracy in a specific region. However, the reliability remains poor and the variability of pubic symphysis morphology related to bone degeneration seems to be an unavoidable limit of the method.