Vertebral fractures are likely to occur if a daily task-related load exceeds the vertebra's capacity to carry such loading, i.e exceeds its structural strength. Most of those loads originate from the trunk muscles, which hold the body upright and balance the trunk for any loading held in the hands. The magnitude of such loads depends on body size, spinal curvature, vertebral level, and the task performed. Measuring muscle-generated forces and resulting vertebral loading faces significant technical difficulties and requires a unique laboratory setting, both impractical when evaluating infirm cohorts of cancer patients with spinal metastasis.
Derived from clinical imaging, spine musculoskeletal models incorporate personalized anatomy (location and cross-sectional areas) of the flexor, extensor and stabilizing thoracic and abdominal muscles and the anatomy and strength of the vertebral column and can be used to evaluate the loading carried by each vertebra under a daily task. This data is used to derive the level's load-to-strength ratio (LSR), a metric for the patient's risk of fracture.
In patients with metastatic spine disease, tumors change vertebral strength, with the systemic effects of cancer affecting sarcopenia and cachexia, both hallmarks of cancer. Yet, clinical prediction tools used to judge patients' risk of spinal instability and, indirectly, risk of vertebral failure, such as the Spinal Instability Neoplastic Score (SINS), base their prediction on how a lesion appears on imaging but do not consider either the lesion's vertebral mechanics or the loading applied to this vertebra during daily life.
Our group, working with Dennis Anderson, uses individualized musculoskeletal models of the spine to estimate the loading and resulting LSR in cancer patients during everyday tasks as a metric of their vertebral fracture. We first applied this approach to cadaveric spines from donors with cancer [1] and then to routine clinical CT from 135 patients planned for radiotherapy at Dana-Farber Cancer Institute (DFCI) and Brigham and Women's Hospital (BWH) [2]. In both studies, the cancer spines were compared with a population-based reference group from the Framingham Heart Study. The models draw on our finite element work for vertebral strength and on our image analysis and AI tools
Figure 1. Workflow for establishing musculoskeletal spine models for personalized assessment of cancer patients' estimated task- and vertebral-specific loading. The patient/subject CT data is used to extract the subject-specific information (A: sex, height/weight, muscle morphology, and spinal curvature). This information is used to adjust and personalize our generic spine models (B). Simulating different postures and loading, we apply static optimization to predict muscle forces and calculate spinal joint loading using the joint analysis tool in OpenSim ™. The outcome of this analysis is individual vertebral loading (D). Combined with CT estimation of vertebral strength (E - F), we compute vertebral-specific load-to-strength ratio (LSR) (G).
