Reproducibility for linear and nonlinear micro-finite element simulations with density derived material properties of the human radius

Finite element (FE) simulations based on high-resolution peripheral quantitative computed-tomography (HRpQCT) measurements provide an elegant and direct way to estimate bone strength. Parallel solvers for nonlinear FE simulations allow the assessment not only of the initial linear elastic behavior of the bone but also materially and geometrically nonlinear effects. The reproducibility of HRpQCT measurements, as well as their analysis of microarchitecture using linear-elastic FE simulations with a homogeneous elastic modulus has been investigated before. However, it is not clear to which extent density-derived and nonlinear FE simulations are reproducible. In this study, we introduced new mechanical indices derived from nonlinear FE simulations that describe the onset of yielding and the behavior at maximal load. Using 14 embalmed forearms that were imaged three times, we found that in general the in vitro reproducibility of the nonlinear FE simulations is as good as the reproducibility of linear FE. For the nonlinear simulations precision errors (PEs) ranged between 0.4 and 3.2% and intraclass correlation coefficients were above 0.9. In conclusion, nonlinear FE simulations with density derived material properties contain important additional information that is independent from the results of the linear simulations.