Finite element (FE) models of bone derived from clinical quantitative computed tomography (QCT) rely on realistic material properties to accurately predict patient-specific bone strength in vivo. QCT cannot resolve microarchitecture, therefore QCT-based FE models lack the directionality apparent within trabecular bone. Maps of anisotropy were constructed from high-resolution peripheral QCT (HR-pQCT) images of seven femur specimens using a „direct mechanics‟ method to measure local anisotropy. The resulting directionality reflected all the major structural patterns visible within the microarchitecture of the proximal femur. Principal stiffness directions were interpolated into QCT-based femur models, and whole bone stiffness was calculated for orthotropic and isotropic models in a sideways fall configuration. Comparing model stiffness to experimental data revealed no difference in correlation (R²_ORTH=0.780, R²_ISO=0.788). These results suggest that the variability in stiffness explained by anisotropy at the microarchitecture level does not scale to whole bone models for this specific loading configuration.