Patient-specific finite element models of human vertebrae with simulated osteolytic defects: Predictive accuracy of...

Approximately 40% of cancer patients develop spinal metastases, often leading to vertebral fragility fractures. Osteolytic metastases are the most severe, as they significantly alter strain distribution in bone tissue. This pilot study extends a previously validated experimental-computational framework for intact vertebrae to assess the accuracy of personalized finite element models (FEM) in the presence of controlled artificial lytic defects. Three human lumbar vertebrae (L3-L5) were tested in intact and defect conditions. Progressively enlarging hemispherical defects (10-25 mm diameter) were drilled within the trabecular bone through the caudal surface after endplate removal, without involving the cortical wall. A compressive load of 2 kN was homogeneously distributed on the cranial endplate. Surface displacements and strains were measured using Digital Image Correlation (DIC), and cranial endplate deflection using a linear displacement transducer (LVDT). Corresponding FEM were developed from CT data and validated against experimental measurements. Predicted and measured endplate deflection (R² = 0.79) and vertebral wall displacements (R² = 0.87-0.92) were significantly correlated across all defect sizes. Longitudinal strain predictions demonstrated good agreement with experimental data, with median relative errors across defect sizes ranging from -4% to +13%, comparably to the intact condition (+10%). Circumferential strains in defect conditions were consistently overestimated, similarly to the intact condition as well. In summary, the proposed FEM retained high accuracy in longitudinal strain prediction and captured vertebral endplate deflection even in the presence of large trabecular defects, although experimental limitations suggest modifying the caudal constraint in future studies.