Related Experiment Video
Updated: Feb 23, 2026
Minimally Invasive Treatment for Thoracolumbar Burst Fracture Using Sagittal Alignment Screws and A Trauma Reduction Device
Published on: November 8, 2024
Fine-tuning the Predictive Model for Proximal Junctional Failure in Surgically Treated Patients With Adult Spinal
Mitsuru Yagi1,2,3, Nobuyuki Fujita1,3, Eijiro Okada1,3
1Department of Orthopedic Surgery, Keio University School of Medicine, Tokyo, Japan.
Study Design:
Multicenter retrospective study.
Objective:
To validate and improve the predictive model for proximal junctional failure (PJF) with or without the bone mineral density (BMD) score.
Summary Of Background Data:
PJF is a serious complication of surgery for adult spinal deformity (ASD). A predictive model for PJF was recently reported that has good accuracy, but does not include BMD, a known PJF risk factor, as a variable.
Methods:
We included 145 surgically treated ASD patients who were older than 50 at the time of surgery and had been followed up for at least 2 years. Variables included age, sex, body mass index (BMI), fusion level, upper and lower instrumented vertebral (UIV and LIV) level, primary or revision surgery, pedicle subtraction osteotomy (PSO), Schwab-SRS type, and BMD. PJF was defined as a ≥ 20° increase from baseline (immediately postoperative) of the proximal junctional angle with concomitant deterioration of at least 1 SRS-Schwab sagittal modifier grade, or any proximal junctional kyphosis requiring revision. Decision-making trees were constructed using the C5.0 algorithm with 10 different bootstrapped models, and validated by a 7:3 data split for training and testing; 112 patients were categorized as training and 33 as testing samples.
Results:
PJF incidence was 20% in the training samples. Univariate analyses showed that BMD, BMI, pelvic tilt (PT), UIV level, and LIV level were PJF risk factors. Our predictive model was 100% accurate in the testing samples with an AUC of 1.0, indicating excellent fit. The best predictors were (strongest to weakest): PT, BMD, LIV level (pelvis), UIV level (lower thoracic), PSO, global alignment, BMI, pelvic incidence minus lumbar lordosis, and age.
Conclusion:
A successful model was developed for predicting PJF that included BMD. Our model could inform physicians about patients with a high risk of developing PJF in the perioperative period.
Level Of Evidence:
4.
