Paraspinal muscle parameters’ predictive value for new vertebral compression fractures post-vertebral augmentation: Nomogram development and validation

TL;DR

This study shows that paraspinal muscle parameters can predict new vertebral fractures after treatment, using a validated nomogram model.

Contribution

A novel nomogram model using paraspinal muscle parameters to predict new vertebral fractures after PVP/PKP is developed and validated.

Findings

Multifidus skeletal muscle index is an independent predictor of new vertebral fractures.

The nomogram model demonstrated good accuracy with AUC values of 0.801 to 0.832 in validation.

The model showed greater negative predictive value compared to spinal CT values.

Abstract

Prior research underscores the significance of paraspinal muscles in maintaining spinal stability. This study aims to investigate the predictive value of paraspinal muscle parameters for the occurrence of new vertebral compression fractures (NVCF) following percutaneous vertebroplasty (PVP) or percutaneous kyphoplasty (PKP) in patients with osteoporotic vertebral compression fractures (OVCF).

Retrospectively collected data from October 2019 to February 2021 (internal validation, n = 235) and March 2021 to November 2021 (external validation, n = 105) for patients with OVCF treated with PVP/PKP at our institution. They were randomly divided into training (188 cases) and validation groups (47 cases) at an 8:2 ratio. Lasso regression and multivariable logistic regression identified independent risk factors in the training set, and a Nomogram model was developed. Accuracy was assessed using receiver operating characteristic curves (ROC), calibration was evaluated with calibration curves and the Hosmer-Lemeshow test, and clinical utility was analyzed using decision curve analysis (DCA) and clinical impact curve (CIC).

Surgical approach, spinal computed tomography (CT) values, and multifidus skeletal muscle index (SMI) are independent predictors of postoperative NVCF in OVCF patients. A Nomogram model, based on the identified predictors, was developed and uploaded online. Internal validation results showed area under the curve (AUC) values of 0.801, 0.664, and 0.832 for the training set, validation set, and external validation, respectively. Hosmer-Lemeshow goodness-of-fit tests (χ2 = 7.311–14.474, p = 0.070–0.504) and calibration curves indicated good consistency between observed and predicted values. DCA and CIC demonstrated clinical net benefit within risk thresholds of 0.06–0.84, 0.12–0.23, and 0.01–0.27. At specificity 1.00–0.80, the partial AUC (0.106) exceeded that at sensitivity 1.00–0.80 (0.062).

Compared to the spinal CT value, the multifidus SMI has certain potential in predicting the occurrence of NVCF. Additionally, the Nomogram model of this study has a greater negative predictive value.

Figures