Prediction of Metastatic Spine Fracture Using CT Based Structural Analysis

Purpose: This prospective, in-vivo study evaluates whether metastatic spine fracture is better predicted by measuring the structural properties of affected vertebra on transaxial CT scans than by analysis of defect size and location only.

Background: One in nine women develop breast cancer; 60% develop skeletal metastases, the spine being the most frequent site of involvement. Pathologic fractures occur in ~30% of patients with skeletal metastases. Current radiographic criteria for estimating fracture risk have low specificity. Accurate prediction of pathologic fracture requires an assessment of whether the tumor has diminished the load bearing capacity of the vertebra relative to the loads applied to the spine such that fracture will occur. We have previously demonstrated that the load bearing capacity of vertebra with simulated lytic defects depends on the material and geometric properties of the host bone in addition to the size and location of the defect [1].

Methods: 94 women with metastatic breast cancer to the spine were CT scanned and prospectively evaluated for pathologic fracture over the ensuing four months. The load bearing capacity of each vertebra from T8-L5 was calculated using an algorithm that accounted for the material and geometric properties of the vertebra and the size and location of the defect by measuring the cross-sectional bending and axial rigidity of the vertebra on serial transaxial CT images and modeling the vertebra as a composite beam loaded in combined flexion and compression. The loads applied to each vertebra when lifting a 10 kg mass were estimated using an analytic model of the spine and torso. A fracture risk index (FRI) was calculated by dividing the load applied to the vertebra by the load bearing capacity. Fracture occurs if FRI >1. To derive an empiric threshold for predicting fracture a multivariate logistic regression analysis was performed including as independent variables the load applied to the vertebra, the calculated load bearing capacity and the patient's body mass index. Fracture risk was further evaluated from CT images through affected vertebrae using criteria developed by Taneichi [2] for predicting fracture based on defect size, defect location and vertebral level. All fracture predictions were compared to actual fracture occurrence during the four months of surveillance.

Results: 11 vertebral fractures occurred in 10 patients over 4 months. All methods were 100% sensitive for predicting fracture, however the CT based structural analysis was significantly more specific (49%) than the Taneichi criteria (20%). The fracture threshold determined by multivariate logistic regression best discriminated vertebral fracture (69% specific). When this threshold was exceeded, the relative fracture risk was 7.9.

Conclusion: CT based structural rigidity analysis was as sensitive and more specific than current radiographic guidelines for predicting vertebral fracture. Fracture risk assessment based on defect size alone is insufficient since the analysis fails to account for the structural properties of the host bone that may be altered by coexistent pathology such as osteoporosis.

References: [1] JBJS(2000) 82: 1240; [2] Spine(1997) 22: 239.