Construction and clinical validation of risk model for predicting bone cement leakage after the surgical management of spinal metastases.

This study aimed to comprehensively analyze the risk factors associated with bone cement leakage (LCK) during the surgical management of spinal metastases, construct a joint risk model for predictive assessment, and validate the clinical applicability of the risk model in an independent patient cohort. A retrospective analysis was conducted on patients who underwent surgery for spinal metastases between February 2022 and June 2023. Patients were divided into a non-LCK group (n=134) and an LCK group (n=86) based on the presence or absence of bone cement leakage after surgery. Additionally, a validation group was established, consisting of 21 patients with LCK and 65 patients without. Analysis focused on patient demographics, intraoperative parameters, LCK location, complications, pain management, and improvements in activities of daily living (ADL). Logistic regression, calibration curve, clinical impact curve (CIC) analysis, decision curve analysis (DCA) and receiver operating characteristic (ROC) analysis were used to assess the risk factors and construct a joint risk model. There were significant differences between the two groups in pathologic fracture, Tomita classification, posterior wall destruction, injected laterality, injected bone cement volume, radicular pain, pulmonary embolism, and medullary compression. Pathologic fracture, radicular pain, pulmonary embolism, and medullary compression were positively correlated with the occurrence of LCK, while Tomita classification, posterior wall destruction, injection laterality, and injected bone cement volume were negatively correlated with the occurrence of LCK. Pathological fracture, Tomita classification, posterior wall destruction, injected laterality, injected bone cement volume, and specific postoperative complications were identified as significant risk factors associated with LCK. The constructed joint risk model, incorporating these risk factors, demonstrated substantial predictive value, with an Area Under the Curve (AUC) of 0.885. Clinical validation in an independent patient cohort further confirmed the predictive power of the joint risk model, with an AUC of 0.846. This study underscores the multifactorial nature of LCK in surgical management of spinal metastases, providing valuable insights for risk assessment and management.