Multiple myeloma (MM) is associated with vertebral fracture. Finite element (FE) modelling simulates mechanical loading to directly estimate strength; however, the thin cortical wall, with a thickness on the order of the pixel size of clinical CT scanners, cannot be fully resolved. We developed an FE model with thin cortical elements and applied it to a dataset with MM patients. Vertebral quantitative computed tomography (qCT) scans were evaluated in 101 MM patients. An FE mesher was developed (Matlab) to generate nonlinear specialized-cortex (SC) FE models from the images. The spongiosa region was meshed with tetrahedral elements, while the cortical mesh was created by applying pentahedral elements (triangular prisms) to the spongiosa surface. The element-level elastic modulus, failure stress and post-yield behavior were set from the local BMD according to relations from Keyak et. al., 1998. The patient-specific thickness of the cortical elements was determined from the density-weighted thickness and the elastic modulus set as 10GPa. A tetrahedral full-vertebral (FV) mesh was generated with no special consideration for the cortex for comparison. Uniaxial compression was simulated (Calculix v.2.7) and the apparent level stiffness, yield load and work to yield as well as stress distribution were determined. Comparisons of the two models were made using linear regression and t-tests. Strong correlations were observed between the two models for stiffness (R2=0.66), and yield force (R2=0.61), with more moderate correlations observed in work to yield (R2=0.39). The FV mesh showed significantly higher extrinsic stiffness (p=0.006) and yield force (p=0.02), while the SC model predicted a higher work to yield (p=0.002). In the SC model, the mean von Mises (VM) stress within the spongiosa and cortex were both associated with yield load (R2=0.51 and 0.36, respectively). The distribution of the VM stress in the spongiosa (R2=0.17) but not in the cortex (R2=0.007) was associated with yield load. All parameters correlated significantly with BMD, with the correlation coefficents consistently higher in the FV model. Similar predictions of mechanics were obtained with both models; however stiffness and strength may be overestimated and energy absorption underestimated when the cortical geometry is not taken into account. In MM patients, the strength is affected by the stress distribution in the spongiosa but not in the cortex, indicating that stress raisers in the spongiosa may be the leading cause of fracture in these patients.
Disclosure: The authors declared no competing interests.