Inter-laboratories comparison of different numerical models to four experimental datasets of human vertebrae with or without metastases.

Bone metastases, remarkably common in the vertebral bodies of advanced cancer patients, are associated with an increased risk of vertebral fracture. To assess vertebral strength, subject-specific biomechanical models have been proposed based on quantitative computed tomography. Such models have been evaluated on a single dedicated dataset often acquired by the same team. The goal of this paper is to compare different biomechanical models with four experimental datasets to strengthen their validation. Finite element models developed by three groups at the University of Bern, University of Lyon, and "Institut de Biomécanique Humaine Georges Charpak" of Paris, were used to simulate the strength of four different experiments that included 1) intact non-metastatic vertebrae with endplates embedded in endcaps during experimental tests, vertebrae with simulated defects representing osteolytic bone metastasis foci with endplates removed, and metastatic vertebrae from cancer donors with the endplates removed. The simulation results show the determination coefficient to range between 0.28 and 0.96 with a slope of the correlation between experimental and numerical failure strengths from 0.32 to 1.69. The bias (difference between experimental and numerical failure strengths) varied from -3749 and + 1368 N, corresponding to -48.6% to +77%. The experimental removal of the endplates influenced FE model predictions. This study illustrates the need for extensive validation of models before their potential use in clinical practice. To encourage such validation, the experimental datasets were made available in open data.