Prediction models for subsequent vertebral fractures after percutaneous vertebroplasty or kyphoplasty: a systematic review and critical appraisal

OBJECTIVE: To systematically review and critically appraise existing risk prediction models for SVFs following PVP or PKP, with emphasis on methodological quality, model performance, and clinical applicability. METHODS: Databases were conducted from inception to December 2025 for relevant studies that developed or validated multivariable models predicting subsequent vertebral fractures after PVP or PKP. Study selection and data extraction were performed in duplicate. Methodological quality and applicability were assessed using the PROBAST framework. RESULTS: A total of 30 studies met the inclusion criteria. The identified models were primarily derived using logistic regression or various machine learning. Frequently incorporated predictors comprised bone mineral density, patient age, cement leakage, cement dispersion pattern, and use of anti-osteoporosis medication. Reported discrimination performance ranged from 0.664 to 0.998 (AUC or C-index). However, calibration measures and decision curve analysis were insufficiently described in several studies. And every model was considered to have a high overall risk of bias. CONCLUSIONS: Although existing models demonstrate promising discriminatory performance, their high risk of bias and limited external validation substantially restrict clinical applicability. Future research should prioritize prospective, multicenter studies with adequate sample sizes and external validation to enhance robustness and transportability.