Clin Interv Aging. 2026 Jun 18;21:600367. doi: 10.2147/CIA.S600367. eCollection 2026.
ABSTRACT
PURPOSE: Osteoporotic vertebral compression fractures (OVCFs) cause significant morbidity in aging populations. Hounsfield unit (HU) from CT and the vertebral bone quality (VBQ) from MRI show promise in assessing bone quality and fracture risk. This study aims to directly compare the predictive efficacy of HU and VBQ for OVCFs and develop a nomogram model integrating HU and VBQ.
PATIENTS AND METHODS: A retrospective study was conducted involving 385 patients (127 with OVCFs, 258 controls) who were hospitalized at our hospitals between September, 2020 and September, 2024. HU and VBQ were derived from picture archiving and communication system (PACS). Other variables included demographic, clinical, and radiological data. Statistical analyses included t-tests, chi-square tests, multivariable logistic regression, the least absolute shrinkage and selection operator method (LASSO) regression, and receiver operating characteristic (ROC) curve analysis. Then, a nomogram model was established. The calibration, discrimination and clinical practicability of the nomogram model were also evaluated.
RESULTS: The OVCF group had significantly higher VBQ and lower HU compared to controls. ROC analysis showed higher diagnostic accuracy for HU than VBQ.A nomogram model for predicting the risk of OVCF occurrence in patients has been developed based on three independent predictors, namely BMI, HU and VBQ. The AUC was 0.84 in the training set and 0.87 in the test set. The model has good practicability for clinics according to the decision curve analysis (DCA) and clinical impact curve (CIC).
CONCLUSION: Both HU and VBQ are effective predictors of OVCFs. The nomogram model showed good internal discrimination and calibration in our study. These findings suggest potential utility for opportunistic screening of OVCF risk in patients undergoing routine spinal CT and MRI. However, external validation in prospective and multi-center cohorts is needed before clinical implementation.
PMID:42339495 | PMC:PMC13285745 | DOI:10.2147/CIA.S600367
