Spine - 2026-10-15 - Journal Article
Estimated Three-Dimensional Thoracic Kyphosis Supplements Radiographic Bone Age in Predicting Adolescent Idiopathic Scoliosis Curve Progression.
McCoy L, Tang N, Kelly BA, Brouillet K, Luhmann SJ
Topics
Key Takeaway
e3DTK alone achieves AUC 0.89 for predicting AIS curve progression at a threshold of 22.7°, but adds no statistically significant improvement over Sanders score alone (AUC 0.80, p=0.15 for combined model).
Summary Depth
Choose how much analysis to show on this article page.
Summary
This study asked whether estimated 3D thoracic kyphosis derived from standard radiographs adds predictive value to Sanders skeletal maturity staging for AIS curve progression (≥6° Cobb increase or progression to ≥50°). In 126 braced AIS patients, Sanders score AUC was 0.80 and e3DTK alone AUC was 0.89 (threshold 22.7°), but combining both did not improve the model (AUC 0.69, p=0.15). Stratified analysis showed e3DTK differentiates progression risk within Sanders strata (p=0.022), suggesting a complementary rather than additive role.
Key Limitation
The combined model AUC (0.69) was paradoxically lower than either predictor alone, suggesting possible overfitting or collinearity in a small single-center cohort of only 126 patients.
Original Abstract
STUDY DESIGN
Retrospective cohort study.
OBJECTIVE
To evaluate whether estimated three-dimensional thoracic kyphosis (e3DTK) adds predictive value to radiographic bone age in identifying patients with adolescent idiopathic scoliosis (AIS) at risk for curve progression.
SUMMARY OF BACKGROUND DATA
Accurate prediction of curve progression in AIS remains challenging. Radiographic bone age is commonly used for risk stratification; however, three-dimensional spinal morphology may better reflect the biomechanical factors influencing progression.
MATERIALS AND METHODS
Patients with AIS were retrospectively identified at a single tertiary pediatric center. Progression was defined as a ≥6° increase in Cobb angle or progression to ≥50°. e3DTK was estimated from standard radiographs using a validated equation. Predictive performance was assessed using ROC analysis, DeLong testing, and likelihood ratio comparisons. In addition, patients were stratified by Sanders score and e3DTK risk category to evaluate subgroup progression rates.
RESULTS
A total of 126 patients were included (79.4% female; 109 braced). e3DTK ( r =0.25, P =0.005) and Sanders score ( r =-0.33, P <0.001) were both correlated with curve magnitude change. Sanders score demonstrated strong predictive performance (AUC=0.80). Adding e3DTK to Sanders did not improve model performance (AUC=0.69; P =0.15). e3DTK alone showed good discrimination for progression (AUC=0.89) with an optimal threshold of 22.7°. However, this did not outperform Sanders-based prediction or improve combined models. Stratified analyses showed higher progression rates in high-risk e3DTK groups across Sanders stages ( P =0.022), suggesting within-stratum risk differentiation.
CONCLUSION
e3DTK is associated with AIS curve progression and demonstrates discriminatory ability in univariate analysis; however, it does not improve predictive performance beyond Sanders skeletal maturity staging. While it may help characterize risk heterogeneity within maturity strata, its role appears complementary rather than independently additive in AIS progression prediction.