Spine - 2026-03-10 - Journal Article
Comparative Analysis of Three Lordosis Correction Parameters for Predicting Proximal Junctional Kyphosis in Adult Spinal Deformity Surgery.
Park SJ, Park JS, Kang DH, Lee CS, Kim HJ
Topics
Key Takeaway
L1PA offset outperformed PI-LL and age-adjusted PI-LL in calibration for predicting PJK (Brier score 0.196 vs 0.211 and 0.210), though AUCs were statistically similar across all three parameters (0.704 vs 0.638 and 0.648).
Summary Depth
Choose how much analysis to show on this article page.
Summary
This study compared the predictive performance of PI-LL offset, age-adjusted PI-LL offset, and L1PA offset for PJK in 323 ASD patients undergoing lower thoracic-to-pelvis fusion with minimum 2-year follow-up. Discriminative ability was similar across all three parameters (AUCs 0.638, 0.648, 0.704; P>0.05), but L1PA offset demonstrated superior calibration (Brier score 0.196 vs 0.210–0.211) and was the only parameter showing a clear dose-response relationship between overcorrection and PJK risk. L1PA had only moderate correlation with PI-based parameters, indicating it captures distinct alignment information.
Key Limitation
The retrospective design precludes standardization of correction intent, meaning measured postoperative offsets reflect achieved rather than planned alignment, conflating surgical decision-making with outcome prediction.
Original Abstract
STUDY DESIGN
Retrospective study.
OBJECTIVE
To compare the predictive performance of three lordosis correction parameters for proximal junctional kyphosis (PJK) in adult spinal deformity (ASD) surgery.
SUMMARY OF BACKGROUND DATA
Although the extent of lordosis correction, as measured by pelvic incidence minus lumbar lordosis (PI-LL), age-adjusted PI-LL, and L1PA, has been identified as a risk factor for PJK, direct comparison of their predictive ability has not been fully explored.
METHODS
A total of 323 patients who underwent lower thoracic spine to pelvis fusion with a 2-year follow-up were included. Three lordosis correction parameters were measured postoperatively: PI-LL, age-adjusted PI-LL offset, and L1PA offset. PJK was defined both radiographically and clinically. Logistic regression models with restricted cubic splines were used to assess the relationship between each parameter and the occurrence of PJK. Discriminative performance was evaluated using the area under the receiver operating characteristic curve (AUC), and calibration performance was assessed using Brier scores (lower scores indicate better performance). Odds ratios were calculated at multiple offset levels to evaluate effect sizes.
RESULTS
All three parameters demonstrated similar discriminative abilities (AUCs: 0.638 for PI-LL, 0.648 for age-adjusted PI-LL offset, and 0.704 for L1PA offset; P>0.05). However, the L1PA offset model showed significantly superior calibration performance (Brier score=0.196) compared to PI-LL (0.211) and age-adjusted PI-LL offset (0.210). Effect size analysis revealed that while overcorrection (negative offset values) increased PJK risk across all models, only the L1PA offset demonstrated a clear dose-response relationship. Correlation analysis showed that L1PA had only a moderate correlation with PI-based parameters, suggesting it captures distinct alignment features.
CONCLUSIONS
All three lordosis correction parameters were predictive of PJK, but the L1PA offset showed improved calibration and more consistent risk stratification. L1PA may serve as a more robust and individualized metric for guiding alignment in ASD surgery.