Trabecular bone microarchitecture and tissue heterogeneity strongly influence mechanical behaviour, yet homogeneous micro-finite element (µFE) models neglect tissue heterogeneity, limiting accuracy in local stress–strain predictions. This study evaluated the effects of mesh type (voxel-based linear hexahedral, HEX, vs geometry-based quadratic tetrahedral, TET), mesh resolution, and material property assignment (homogeneous vs heterogeneous) on µFE predictions of apparent elastic modulus, von Mises stress, and principal strains in human femoral trabecular bone cores. Fourteen cores were micro-computed tomography scanned and meshed into HEX and TET µFE models at three resolutions, with tissue modulus assigned either homogeneously or heterogeneously using trilinear interpolation from voxel-level density. Models were calibrated to experimentally measured apparent elastic modulus. All models accurately predicted apparent elastic modulus. However, mesh type and resolution substantially influenced local mechanical outcomes. Tetrahedral models produced higher local principal strains and von Mises stresses, indicating improved sensitivity to localized strain concentrations, whereas hexahedral models yielded smoother stress–strain distributions with lower peak values but greater computational efficiency. Finer meshes enhanced the resolution of local strain concentrations, while coarser meshes underestimated peak responses. Incorporating tissue heterogeneity increased local principal strains and reduced peak von Mises stresses, although differences from homogeneous models were modest. These findings demonstrate that mesh type has a stronger influence on local mechanical predictions than material heterogeneity when models are calibrated to bulk stiffness. Geometrybased tetrahedral meshes are preferable when accurate estimation of local strain concentrations is required, whereas voxel-based hexahedral meshes provide a computationally efficient alternative for bulk property prediction. This work provides practical guidance for optimizing µFE modelling strategies in trabecular bone research.