Age-specific profiles of tissue-level composition and mechanical properties in murine cortical bone

There is growing evidence that bone composition and tissue-level mechanical properties are significant determinants of skeletal integrity. In the current study, Raman spectroscopy and nanoindentation testing were co-localized to analyze tissue-level compositional and mechanical properties in skeletally mature young (4 or 5 months) and old (19 months) murine femora at similar spatial scales. Standard multivariate linear regression analysis revealed age-dependent patterns in the relationships between mechanical and compositional properties at the tissue scale. However, changes in bone material properties with-age are often complex and nonlinear, and can be missed with linear regression and correlation-based methods. A retrospective data mining approach was implemented using non-linear multidimensional visualization and classification to identify spectroscopic and nanoindentation metrics that best discriminated bone specimens of different age-classes. The ability to classify the specimens into the correct age group increased by using combinations of Raman and nanoindentation variables (86-96% accuracy) as compared to using individual measures (59-79% accuracy). Metrics that best classified 4 or 5 month and 19 month specimens (2-age classes) were mineral to matrix ratio, crystallinity, modulus and plasticity index. Metrics that best distinguished between 4,5 and 19 month specimens (3-age classes) were mineral to matrix ratio, crystallinity, modulus, hardness, cross-linking, carbonate to phosphate ratio, creep displacement and creep viscosity. These findings attest to the complexity of mechanisms underlying bone tissue properties and draw attention to the importance of considering non-linear interactions between tissue-level composition and mechanics that may work together to influence material properties with age. The results demonstrate that a few non-linearly combined compositional and mechanical metrics provide better discriminatory information than a single metric or a single technique. (C) 2012 Elsevier Inc. All rights reserved.