Combining medial clavicular epiphysis ossification and clavicle length on conventional radiography for forensic age estimation: A multivariable regression approach

Abstract

Accurate age estimation is crucial in forensic contexts, yet current methods have limitations. This study evaluates whether integrating clavicle length with medial clavicular epiphysis (MCE) ossification improves age estimation on conventional chest radiographs. A retrospective analysis was performed on 743 chest radiographs from Italian individuals aged 10–30 years. MCE ossification was staged according to Schmeling's classification, and clavicle length was measured. Three multivariable linear regression models were developed, and post-hoc subgroup analyses were conducted to investigate age-dependent performance. Results showed that while clavicle length correlated significantly with age in the overall sample (p < 0.001), its predictive value was limited to early adolescence (10–15 years) and became non-significant in the critical forensic window of 16–20 years (p = 0.342). Nevertheless, the combined regression model (MCE stage + clavicle length + sex) achieved the best overall performance (R2 = 0.685, RMSE = 3.20 years, MAE = 2.51 years; AUC = 0.94, OA = 87.2%). In classifying individuals around the 18-year threshold, this model significantly outperformed traditional binary cut-offs and single-variable models (p < 0.001). While the “Stage 4” rule yielded high specificity (98.1%) but low sensitivity (62.5%), the combined regression model achieved a sensitivity of 91.4% with acceptable specificity (80.1%). These findings suggest that transitioning from rigid cut-offs to statistical modeling provides a more balanced risk assessment, reducing false negatives, although the specific added value of clavicle length diminishes after the pubertal growth spurt.