Non-Cochrane · DTA · PM42449875
Adult Age Threshold Estimation Using Radiographic Evaluation of Wrist-Hand Skeletal Maturation: A Systematic Review and Meta-Analysis.
Bianchi Ilenia, Focardi Martina, Costantino Andrea, et al. · Diagnostics (Basel, Switzerland) · 2026 · PubMed · doi:10.3390/diagnostics16132093
Findings
37 findings
| Population | ||||
|---|---|---|---|---|
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Accuracy at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Sensitivity analysis: high risk of bias studies only | Accuracy (%) 86.9 [79.4, 94.5] GRADE not reported 2 studies |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Sensitivity analysis: low risk of bias studies only | Specificity (%) 89.9 [88.2, 91.6] GRADE not reported 3 studies |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Alternative pooled estimate shown in GRADE figure (model not stated) | Sensitivity (%) 85.4 [75.7, 91.6] Very low Downgraded: risk of bias, inconsistency, indirectness 5 studies · 4587 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Accuracy at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples | Accuracy (%) 85.9 [79.4, 92.3] Very low Downgraded: risk of bias, inconsistency, indirectness 5 studies · 4587 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Sensitivity analysis: high risk of bias studies only | Sensitivity (%) 87.5 [71.88, 100] GRADE not reported 2 studies |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Area under the ROC curve at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples | AUC (approximate, SROC) 0.775 Very low 5 studies · 4587 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples | Sensitivity (%) 69.5 [61.6, 77.3] Very low Downgraded: risk of bias, inconsistency, indirectness 5 studies · 4587 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Alternative pooled estimate shown in GRADE figure (model not stated) | Specificity (%) 82.2 [76.8, 86.5] Very low Downgraded: risk of bias, inconsistency, indirectness 5 studies · 4587 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples | Specificity (%) 85.6 [83.5, 87.6] Very low Downgraded: risk of bias, inconsistency, indirectness 5 studies · 4587 participants |
| Adolescents and young adults undergoing forensic age assessment (studies ages 4-25; threshold age 18 years) | Wrist-hand radiographic skeletal maturation methods (GP, TW2/TW3, AI/ML), all methods pooled | Verified chronological age (reference standard), 18-year threshold | Mean absolute error (MAE) of age estimation, years Not specified · Sensitivity analysis: low risk of bias studies only Unresolved: the two extractors disagree on this row. | Sensitivity analysis by risk of bias was not feasible because all MAE studies were high risk of bias No studies found 0 studies |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Accuracy at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Sensitivity analysis: low risk of bias studies only | Accuracy (%) 85.4 [77.8, 93.1] GRADE not reported 3 studies |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Sensitivity analysis: low risk of bias studies only | Sensitivity (%) 64.8 [55.1, 74.6] GRADE not reported 3 studies |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Sensitivity analysis: high risk of bias studies only | Specificity (%) 68.3 [42.4, 94.2] GRADE not reported 2 studies |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Greulich-Pyle (GP) atlas and GP-based methods (including BoneXpert and linear regression on GP 16 bones) at 18-year threshold | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: GP/GP-based | Sensitivity (%) 82.6 GRADE not reported 4 studies · 4496 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Greulich-Pyle (GP) atlas and GP-based methods (including BoneXpert and linear regression on GP 16 bones) at 18-year threshold | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: GP/GP-based | Specificity (%) 77.3 GRADE not reported 4 studies · 4496 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Greulich-Pyle (GP) atlas and GP-based methods (including BoneXpert and linear regression on GP 16 bones) at 18-year threshold | Chronological age from verified documentation (reference standard) | Accuracy at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: GP/GP-based | Accuracy (%) 86.9 GRADE not reported 4 studies · 4496 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Tanner-Whitehouse 3 (TW3) manual assessment at 18-year threshold | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: TW3 | Sensitivity (%) 41.0 GRADE not reported 1 study · 688 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Tanner-Whitehouse 3 (TW3) manual assessment at 18-year threshold | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: TW3 | Specificity (%) 90.3 GRADE not reported 1 study · 688 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Tanner-Whitehouse 3 (TW3) manual assessment at 18-year threshold | Chronological age from verified documentation (reference standard) | Accuracy at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: TW3 | Accuracy (%) 82.3 GRADE not reported 1 study · 688 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Automated TW3 using CNN-based machine learning at 18-year threshold | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: CNN | Sensitivity (%) 76.5 GRADE not reported 1 study · 688 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Automated TW3 using CNN-based machine learning at 18-year threshold | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: CNN | Specificity (%) 91.3 GRADE not reported 1 study · 688 participants |
| Adolescents and young adults (target 16 years and older) undergoing forensic age assessment; included samples aged roughly 6 to 25 years | Automated TW3 using CNN-based machine learning at 18-year threshold | Chronological age from verified documentation (reference standard) | Accuracy at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by method: CNN | Accuracy (%) 88.6 GRADE not reported 1 study · 688 participants |
| Individuals from clinical samples undergoing wrist-hand radiography | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Accuracy at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by population type: clinical | Accuracy (%) 85.9 [79.4, 92.3] GRADE not reported 4 studies |
| Individuals from clinical samples undergoing wrist-hand radiography | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by population type: clinical | Sensitivity (%) 71.3 [63.1, 79.6] GRADE not reported 4 studies |
| Individuals from clinical samples undergoing wrist-hand radiography | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by population type: clinical | Specificity (%) 85.9 [83.8, 88.0] GRADE not reported 4 studies |
| Individuals from forensic samples (one study, Moroccan males) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Specificity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by population type: forensic | Specificity (%) 82.3 [74.5, 90.1] GRADE not reported 1 study · 114 participants |
| Adolescents and young adults undergoing forensic age assessment (studies ages 4-25; threshold age 18 years) | Wrist-hand skeletal maturation methods in forensic samples | Verified chronological age (reference standard), 18-year threshold | Mean absolute error (MAE) of age estimation, years Not specified · Forensic · Subgroup by population type: forensic Unresolved: the two extractors disagree on this row. | Subgroup MAE meta-analysis by population type was not feasible because all three included studies were clinical Italian samples No studies found 0 studies |
| Individuals from forensic samples (one study, Moroccan males) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, AI/ML) classifying individuals as under or over 18 years | Chronological age from verified documentation (reference standard) | Sensitivity at 18-year threshold 18-year age threshold · Forensic age estimation; clinical and forensic radiograph samples · Subgroup by population type: forensic | Sensitivity (%) 56.6 [34.0, 79.1] GRADE not reported 1 study · 114 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, ML-based) for age estimation | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Univariate meta-regression: population type (forensic vs clinical) | Meta-regression coefficient (years) 0.47 GRADE not reported 3 studies · 1237 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, ML-based) for age estimation | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Univariate meta-regression: method type (GP vs TW3/ML-based) | Meta-regression coefficient (years) 0.16 GRADE not reported 3 studies · 1237 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, ML-based) for age estimation | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Univariate meta-regression: sex (males vs females; Tise only) | Meta-regression coefficient (years) -0.12 GRADE not reported 1 study |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, ML-based) for age estimation | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy | MAE (years) 0.537 [0.387, 0.686] Very low Downgraded: risk of bias, inconsistency, indirectness 3 studies · 1237 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, ML-based) for age estimation | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Multivariate meta-regression: full model (population, method, sample size) | R2 variance explained (%) 36.8 GRADE not reported 3 studies · 1237 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Wrist-hand radiographic skeletal maturation methods (GP, TW3, ML-based) for age estimation | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Univariate meta-regression: sample size (log-transformed) | Meta-regression coefficient (years per log-unit) -0.21 GRADE not reported 3 studies · 1237 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Greulich-Pyle (GP) atlas | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Subgroup by method: GP | MAE (years) 0.503 [0.311, 0.695] GRADE not reported 3 studies · 2494 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Tanner-Whitehouse 3 (TW3) manual assessment | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Subgroup by method: TW3 | MAE (years) 0.700 [0.641, 0.759] GRADE not reported 1 study · 453 participants |
| Children, adolescents and young adults from Italian clinical samples (aged 6 to 20 years) | Automated GP/TW3 machine learning (ML-based) | Chronological age from verified documentation (reference standard) | Mean absolute error of skeletal age estimation Clinical radiograph samples in Italy · Subgroup by method: ML-based | MAE (years) 0.620 [0.573, 0.667] GRADE not reported 1 study · 453 participants |
Included primary studies 23
Open in Studies| Study | Year | Also included by | Outcomes it informs here | Trial data |
|---|---|---|---|---|
| Alcina 2018 Reliability of the Greulich and Pyle method for chronological age estimation and age majority prediction in a Spanish sample. PMID:29248958 | 2018 | this review only | Sensitivity at 18-year threshold | not extracted yet |
| Behera 2023 Correlation of skeletal age by Greulich-Pyle atlas, physiological age by body development index, and dental age by London Atlas and… DOI:10.1186/s41935-023-00359-w | 2023 | this review only | not stated | not extracted yet |
| De Donno 2013 Age assessment using the Greulich and Pyle method on a heterogeneous sample of 300 Italian healthy and pathologic subjects. PMID:23570965 | 2013 | this review only | Mean absolute error (MAE) of age estimation, years; Mean absolute error of skeletal age estimation | not extracted yet |
| Garamendi 2003 matched by author and year (probable) | NR | this review only | Sensitivity at 18-year threshold | not extracted yet |
| Gelbrich 2015 Combining wrist age and third molars in forensic age estimation: how to calculate the joint age estimate and its error rate in age… PMID:26079219 | 2015 | this review only | not stated | not extracted yet |
| Hackman 2012 Does mirror imaging a radiograph affect reliability of age assessment using the Greulich and Pyle atlas? PMID:22509947 | 2012 | this review only | not stated | not extracted yet |
| Kadam 2012 P02.50. Naturopathy and yoga based life style intervention for cardiovascular risk reduction in patients with cardiovascular risk factors:… DOI:10.1186/1472-6882-12-s1-p106 | 2012 | this review only | not stated | not extracted yet |
| Kumar 2016 A model of quality assurance and quality improvement for post-graduate medical education in Europe. PMID:20163217 | 2010 | this review only | not stated | not extracted yet |
| Maggio 2016 Skeletal age estimation in a contemporary Western Australian population using the Tanner-Whitehouse method. PMID:27080619 | 2016 | this review only | not stated | not extracted yet |
| Maggio 2018 Assessment of the accuracy of the Greulich and Pyle hand-wrist atlas for age estimation in a contemporary Australian population DOI:10.1080/00450618.2016.1251970 | 2016 | this review only | not stated | not extracted yet |
| Malkocoglu 2025 12-18 yaşlarındaki bireylerin el ve el bileği radyografilerinin kemik yaşı tayini açısından Gilsanz-Ratib ve Kahn-Gaskin atlaslarıyla… DOI:10.17986/blm.1748 | 2025 | this review only | not stated | not extracted yet |
| Mohammed 2015 Is Greulich and Pyle standards of skeletal maturation applicable for age estimation in South Indian Andhra children? PMID:26229357 | 2015 | this review only | not stated | not extracted yet |
| Pereira 2025 Bone age and dental age to assess criminal responsibility: Part II. PMID:40014080 | 2025 | this review only | Sensitivity at 18-year threshold | not extracted yet |
| Pinchi 2014 Skeletal age estimation for forensic purposes: A comparison of GP, TW2 and TW3 methods on an Italian sample. PMID:24681971 | 2014 | this review only | not stated | not extracted yet |
| Pinchi 2026 AI-assisted age estimation in children based on a combination of bone and tooth maturity. PMID:41129910 | 2026 | this review only | Mean absolute error (MAE) of age estimation, years; Mean absolute error of skeletal age estimation | not extracted yet |
| Santoro 2019 A Comparison of 3 Established Skeletal Age Estimation Methods in an African Group From Benin and an Italian Group From Southern Italy. PMID:30864958 | 2019 | this review only | not stated | not extracted yet |
| Schmeling 2006 Age estimation of unaccompanied minors. Part I. General considerations. PMID:16529895 | 2006 | this review only | not stated | not extracted yet |
| Schmidt 2007 Comparative analysis of the applicability of the skeletal age determination methods of Greulich-Pyle and Thiemann-Nitz for forensic age… PMID:17401574 | 2007 | this review only | not stated | not extracted yet |
| Schmidt 2008 Studies in use of the Greulich-Pyle skeletal age method to assess criminal liability. PMID:18331810 | 2008 | this review only | not stated | not extracted yet |
| Thodberg 2017 Automated determination of bone age from hand X-rays at the end of puberty and its applicability for age estimation. PMID:27757577 | 2017 | this review only | Sensitivity at 18-year threshold | not extracted yet |
| Tise 2011 Applicability of Greulich and Pyle method for age assessment in forensic practice on an Italian sample. PMID:21221985 | 2011 | this review only | Mean absolute error (MAE) of age estimation, years; Mean absolute error of skeletal age estimation | not extracted yet |
| Wang 2025 An X-ray bone age assessment method for hands and wrists of adolescents in Western China based on feature fusion deep learning models. PMID:40402226 | 2025 | this review only | Sensitivity at 18-year threshold | not extracted yet |
| Yarimoglu 2005 matched by author and year (probable) | NR | this review only | not stated | not extracted yet |
Review question
| Element | Planned (methods) | Found (results) |
|---|---|---|
| Population | Adolescents and young adults (>=16 years) undergoing forensic age assessment for legal, immigration, asylum or criminal justice purposes | Individuals aged 4-25 years (23 studies, n=11,425; 56.5% clinical samples; 15 countries) |
| Intervention | Wrist-hand radiographic skeletal maturation methods: Greulich-Pyle, Tanner-Whitehouse (TW2, TW3), AI/ML (BoneXpert, deep learning), epiphyseal fusion, combined approaches | GP atlas (17 studies), TW2/TW3 (5), AI/ML (3), epiphyseal fusion |
| Comparison | Chronological age from verified documentation (reference standard) | Documented chronological age (n=23) |
| Outcomes | Sensitivity, specificity, PPV, NPV, accuracy, AUC, LR+, LR-, DOR at thresholds 16, 18, 21 years; MAE, MSE, RMSE, correlation; inter-rater reliability | Pooled sensitivity 69.5%, specificity 85.6%, accuracy 85.9%, AUC 0.775 (5 studies, n=4587); pooled MAE 0.537 years (3 studies) |
| Study designs | Observational cohort (prospective or retrospective), cross-sectional, validation, method comparison and diagnostic accuracy studies | Cross-sectional and retrospective diagnostic accuracy studies (n=23) |
| Timing | Not specified | Threshold age 18 years |