Aims: Bone mineral density (BMD) is the diagnostic standard for osteoporosis but reflects only part of bone mass change. Testing bone yield strength is a better measure of biomechanical competence, yet requires destructive testing. We developed two clinically feasible approaches for assessing yield strength to improve osteoporosis identification: an intraoperative model based on surgical mechanical data, and a preoperative model using demographic and noninvasive measures. Methods: Axial force and torque during reaming/tapping were integrated to compute process energy in synthetic bone analogues (SBA). This computation was calibrated with finite element analysis (FEA) by modelling the same processes and aligning calculated and simulated energies. A relationship between destruction energy and yield strength was then derived in SBA. The protocol was applied to ex vivo human cancellous bone, where the computed energies were converted to predicted yield strength via the SBA mapping, and calibrated against paired uniaxial compression. Intraoperative mechanical data were then collected across multiple centres to estimate vertebral yield strength and to train a preoperative model from demographic and noninvasive variables. Results: Data were acquired for six SBA densities. Computed energies correlated with finite element energies (R² = 0.95), and the SBA energy to yield strength relationship showed excellent fit (R² = 0.99). In ex vivo human cancellous bone (n = 14), predicted yield strengths were corrected against uniaxial compression with high agreement (R² = 0.96) and validated in an independent sample (n = 6). Intraoperative data from 200 procedures supported a high-performing preoperative model (R² = 0.99). SHapley Additive exPlanations (SHAP) indicated positive contributions from T-score and Hounsfield units (HU) and negative contributions from BMI, female sex, and age. Conclusion: This study proposes two clinically applicable models for bone yield strength prediction, circumventing destructive testing. The preoperative model enables risk stratification, while the intraoperative model provides accurate vertebral assessment.
Bai et al. (Fri,) studied this question.