AUTHOR=Rong Yi , Zhu Yihua , Yin Heng , Yu Hao , Wang Jianwei , Shao Yang , Li Shaoshuo , Ye Jiapeng , Guo Yang , Ma Yong , Wang Lining , Hua Zhen 

TITLE=Establishment of a risk prediction model for residual back pain after percutaneous kyphoplasty in osteoporotic vertebral compression fractures

JOURNAL=Frontiers in Surgery

VOLUME=Volume 12 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/surgery/articles/10.3389/fsurg.2025.1625518

DOI=10.3389/fsurg.2025.1625518

ISSN=2296-875X

ABSTRACT=PurposeSevere residual back pain (RBP) after percutaneous kyphoplasty (PKP) significantly impacts postoperative prognosis and quality of life in patients. The aim of this study was to identify the risk factors for RBP in osteoporotic vertebral compression fracture (OVCF) patients after PKP, to establish a risk prediction model, and to validate its effectiveness.MethodsA case-control study was carried out among OVCF patients, who were assigned to either the training set (these patients were recruited from January 2018 and June 2020) or the validation set (these patients were recruited from July 2020 and December 2020). Risk factors were identified by univariate analysis and multifactor logistic regression analysis. The performance of the prediction model was determined by using the area under the receiver operating characteristic (ROC) curve (AUC) to assess discrimination. A nomogram for risk prediction was constructed, the Hosmer-Lemeshow test and calibration curves were used to assess calibration, and decision curve analysis was used to assess the clinical use of the model.ResultsA total of 647 patients were included, 569 cases were used to train the model and 78 cases were used for external validation. Based on the data of model training set, age, bone mineral density, trauma history, posterior fascial edema, platelet distribution width, serum chloride, and middle vertebral height were independent risk factors for RBP after PKP (P ≤ 0.05). The AUC of the risk prediction model constructed thus was 0.788 (95% CI, 0.740–0.836), cut off (0.710, 0.761), with good discrimination. Calibration curves of the model training and validation sets were between the standard curve and the acceptable line, and the Hosmer-Lemeshow test indicated that the model training and validation sets were χ2 = 6.354 and χ2 = 7.240, (P = 0.608 and 0.511), respectively, which have good calibration. The decision curve analysis showed that the threshold probability interval of the net benefit value of the model was 6.3%–82.3% for the training set, 8.7%–55.6% and 72.5%–81.3% for the validation set.ConclusionThe constructed model showed good predictive ability in the occurrence of residual back pain after PKP, which can provide a scientific basis and guidance for clinical prevention and treatment.