The probability of whole-bone fatigue fracture can be accurately predicted using specimen-specific finite element analysis incorporating a stochastic failure model

A better understanding of the mechanisms of mechanical fatigue in bone could help improve understanding of the etiology of stress fractures. Investigations of small material samples of bone have identified a nonlinear relationship between strain magnitude, strained volume, and fatigue life, but it is non-trivial to extend these principles to predict the fatigue-life of whole bones which experience complex loading and non-uniform strain distribution. The purpose of this investigation was to experimentally validate a specimen-specific finite element (FE) model that predicts whole-bone fatigue failure using a stochastic model based on strain magnitude and volume. Thirty-four rabbit tibiae were previously tested to failure under cyclic compression, torsion, or both. Strain distribution during the test was estimated from computed-tomography based specimen-specific FE models, and a stochastic failure model based on strain magnitude and volume was used to predict the probability of failure as a function of loading cycles. Model predicted fracture risk matched experimental observations. Respectively, for the 25%, 50%, 75%, and 95% probabilistic predictions, we observed experimental failure <= model predicted values in 41%, 53%, 76%, and 80% of the tested specimens. A Brier scoring rule further demonstrated that this model, using strain magnitude and volume, more accurately predicted failure probability compared to two reference models that considered strain magnitude only. In conclusion, the stochastic model may be a powerful tool in future studies to assess mechanical factors that influence stress fracture risk.