An analytical and numerical study of the stability of bone remodelling theories: dependence on microstructural...

The origin of unstable bone remodelling simulations using strain-energy-based remodelling rules was studied mathematically in order to assess whether the unstable behavior was due to the mathematical rules proposed to characterize the processes, or to the numerical approximations used to exercise the mathematical predictions. A condition which is necessary for the stability of a strain-energy-based remodelling theory was derived analytically using the calculus of variation. The analytical result was derived using a simple elastic model which consists of a long beam loaded by an axial force and a bending moment. This loading situation mimics the coupling between local density and global density distributions seen in vivo. A condition necessary for a stable remodelling scheme is arrived at, but the conditions necessary to guarantee a stable remodelling scheme are not. In this remodelling scheme, the elastic modulus is proportional to volumetric density raised to an exponent n, and the microstructural stimulus is taken as the strain energy density divided by volumetric density raised to an exponent m. In order for a remodelling scheme to be stable in this loading situation, m must be greater than n. Finite-difference time-stepping is used to verify the predictions of the analytical study. These numerical studies appear to confirm the analytical studies. Physiologic interpretation of the behavior found with n greater than m indicates that this type of unstable behavior is unlikely to be observed in vivo. Since numerical approximations are not made in deriving this stability condition, we conclude that the mathematical rules proposed to characterize bone remodelling based on strain energy density should meet this condition to be relevant to physiologic bone remodelling.