Establishing a nomogram to predict refracture after percutaneous kyphoplasty by logistic regression

Area of Science:

• Orthopedics
• Medical Imaging
• Biostatistics

Background:

• Percutaneous kyphoplasty (PKP) is a common procedure for vertebral compression fractures.
• Identifying patients at risk for new vertebral compression fractures (NVCFs) post-PKP is crucial for effective management.
• Existing prognostic models for post-PKP refractures lack the application of Random Forest (RF) modeling.

Purpose of the Study:

• To investigate risk factors for NVCFs after PKP.
• To compare the predictive performance of logistic regression and RF models for post-PKP fractures.
• To develop a visual predictive tool for post-PKP fractures.

Main Methods:

• Collected clinical data from 349 patients undergoing PKP.
• Utilized Lasso regression to identify significant risk factors for NVCFs.
• Developed and compared logistic regression and RF models, including a nomogram visualization.

Main Results:

• Key risk factors identified include bone density, cement distribution, fracture location, preoperative vertebral height, and height restoration rate.
• The logistic regression model demonstrated superior predictive performance (AUC: 0.868 training, 0.786 validation) compared to the RF model (AUC: 0.786 training, 0.599 validation).
• The logistic regression model's calibration curve also outperformed the RF model's.

Conclusions:

• Logistic regression models offer superior predictive accuracy for post-PKP vertebral fractures compared to RF models.
• The identified risk factors and developed logistic regression model can aid in predicting NVCFs after PKP.