Sensor-validated simulations predict fracture healing outcomes in an ovine model

• Prognostic fracture healing simulations correctly identify delayed union and nonunion cases within the preclinical cohort • Further development of the healing simulations is needed to accurately predict the precise time of healing • Prognostic fracture healing simulations have the potential to correctly identify impaired healing at early time points, improving patient care Fracture management remains a significant clinical challenge, with nonunions occurring in approximately 5 % of fractures. In tibial shaft fractures 1 in 4 patients are considered unhealed after 5 months and a reoperation rate of 12 % is reported. By predicting healing outcomes, prognostic fracture healing simulations could assist in treatment selection and enable early intervention in the event of complications but have lacked comprehensive ground truth validation. This study validated a mechanoregulatory fracture healing simulation algorithm against continuous in vivo sensor data from an ovine preclinical study with eight animals, including a delayed union and a nonunion case. Tibia osteotomies were fixed with locking plates and a strain sensor was attached to the plate to monitor healing in vivo . Subject-specific finite element models, derived from CT post-operative scans, were used with the simulation algorithm, which updated material properties through iterative strain-based tissue differentiation. Specimens were grouped into different healing categories using an unsupervised clustering algorithm, distinguishing cases towards union, delayed union, and nonunion. Significant differences between the predicted healing trajectories were identified, enabling the correct classification of the healing outcomes between union, delayed union and non-union cases. Overall, this study demonstrated successful application of the healing algorithm with validation of the simulation results to healing outcomes. The potential of the simulation to predict healing patterns and to be used as a tool for non-union risk assessment was illustrated. Further refinements are necessary before the healing simulation can be implemented in clinical settings, requiring future research.