Age estimation of living adolescents and young adults of unknown age has been of considerable interest in forensic practice and research, especially for purposes of age identification of refugees. A variety of methods that use the skeletal maturation stages for age estimation have been proposed and practiced worldwide. However, studies focusing on the methodological quality of this researche is scarce.
In order to appraise the quality of age estimation studies, we conducted a systematic review of the various methods used for forensic age assessments.
In April 2017 we searched major databases for age estimation studies using the left hand, clavicle, knee and ankle, and third molar teeth, in living individuals under 25 years of age. We used a revised QUADAS-2 checklist to address the quality and applicability issues, including patient selection, index-test interpretation, reference standard, patient flow and timing. Notably, we added an additional domain to address a neglectable issue in age estimation studies -- age-mimicry bias, which would cause significant inaccuracy on the results.
We found 10,722 references and eventually included 52 studies for further evaluation. Most of the studies were published after 2010, and involved 42,125 participants from all continents. Analysis showed that 59.6% (n = 31) of included studies had a low risk of bias in the patient selection process. Most of the included studies (n = 43, 82.7%) avoided bias of index test interpretation by using medical records for chronological age determination. All the included studies showed a low risk of bias in the reference standard domain, and 78.8% of studies (n = 41) had a low risk of bias in the patient flow and timing domain. However, only 28.8% of included studies (n = 15) showed a low risk of age-mimicry bias, leading to considerable heterogeneity of chronological age estimation across studies, even for studies conducted in the same region.
Our systematic review highlights an easily neglected but prevalent selection bias, namely age mimicry, which would distort the reference sample distribution, leading to inaccurate and inconsistent chronological age estimates. Future studies with even number of participants in each age group, wide age range and sufficient sample size are warranted to account for age mimicry bias.
Patient or healthcare consumer involvement: