[A nomogram for predicting residual back pain after percutaneous vertebroplasty for osteoporotic vertebral compression fractures]
Jun-Ke Li,Xu-Bin Ma,Liang Li,Qing Ma,Xu-Dong Wang
DOI: https://doi.org/10.12200/j.issn.1003-0034.20230652
2024-06-25
Abstract:Objective: To construct percutaneous vertebroplasty for predicting osteoporotic vertebral compression fractures (OVCFs) nomogram of residual back pain (RBP) after percutaneous vertebroplasty(PVP). Methods: Clinical data of 245 OVCFs patients who were performed PVP from January 2020 to December 2022 were retrospectively analyzed, including 47 males and 198 females, aged from 65 to 77 years old with an average of (71.47±9.03) years old, and were divided into RBP group and non-RBP group according to whether RBP occurred. Gender, age, comorbidities, fracture stage, body mass index (BMI), bone mineral density (BMD), visual analogue scale (VAS), Oswestry disability index (ODI) and other general information were collected; anterior vertebral height (AVH), anterior vertebral height ratio (AVH), anterior vertebral height ratio(AVHR), Cobb angle, intravertebral vacuum cleft (IVC), thoracolumbar fascia (TLF) injury, paravertebral muscle steatosis, injection volume and leakage of bone cement, bone cement dispersion pattern, anterior vertebral height recovery ratio (AVHRR), Cobb angle changes, etc. imaging parameters before operation and 24 h after operation were collected. Univariate analysis was performed to analysis above factors, and multivariate Logistic regression model was used to investigate independent risk factors for postoperative RBP, and Nomogram model was constructed and verified;receiver operating characteristic(ROC) curve and calibration curve were used to determine predictive performance and accuracy of the model, and Hosmer-Lemeshow (H-L) test was used for evaluation. The area under curve (AUC) of ROC was calculated, and Harrell consistency index (C index) was used to evaluate the predictive efficiency of model;decision curve analysis (DCA) was used to evaluate clinical practicability of model. Results: There were 34 patients in RBP group and 211 patients in non-RBP group. There were no significant differences in gender, age, comorbidities, fracture stage, BMI, BMD, VAS, ODI, AVH, AVHR and Cobb angle between two groups (P>0.05). Univariate analysis showed 6 patients occurred IVC in RBP group and 13 patients in non-RBP, the number of IVC in RBP group was higher than that in non-RBP group (χ2=5.400, P=0.020);6 patients occuured TLF injury in RBP group and 11 patients in non-RBP group, the number of TLF injury in RBP group was higher than that in non-RBP group (χ2=7.011, P=0.008);In RBP group, 18 patients with grade 3 to 4 paraptebral steatosis and 41 patients in non-RBP group, RBP group was higher than non-RBP group (χ2=21.618, P<0.001), and the proportion of bone cement mass in RBP group was higher than non-RBP group (χ2=6.836, P=0.009). Multivariate Logistic regression analysis showed IVC (χ2=4.974, P=0.025), TLF injury (χ2=5.231, P=0.023), Goutallier grade of paravertebral steatosis >2 (χ2=15.124, P<0.001) and proportion of bone cement (χ2=4.168, P=0.038) were independent risk factors for RBP after PVP. ROC curve of model showed AUC of original model was 0.816[OR=2.862, 95%CI (0.776, 0.894), P<0.001]. The internal verification of model through 200 bootstrap samples showed the value of C index was 0.936, and calibration curve showed predicted probability curve was close to actual probability curve. H-L goodness of fit test results were χ2=5.796, P=0.670. DCA analysis results showed the decision curve was above None line and All line when the threshold value ranged from 6% to 71%. Conclusion: IVC, TLF combined injury, paravertebral muscle steatosis with Goutallier grade> 2, and bone cement diffusion with mass type are independent risk factors for RBP after PVP. The risk prediction model for RBP after PVP established has good predictive performance and good clinical practicability.