PREDICTING PLATE FAILURE USING SPECIMEN-SPECIFIC FINITE ELEMENT MODELS COMBINED WITH IMPLANTABLE SENSORS

Fixation failure via plastic deformation of implants in midshaft fracture treatment is a significant clinical concern, with failure rates up to 17% often requiring reoperation [1,2]. A recent pre-clinical study highlighted plastic plate bending in animal models, underscoring the need for better understanding and monitoring of biomechanical implant behavior under load [3]. Trustworthy finite element (FE) models could predict these failures, but validation with in-vivo data is challenging due to the lack of loading information. The AO Fracture Monitor (AOFM) enables continuous monitoring of implant deformation and fracture healing, providing unique novel insights into implant loading. To interpret the in-vivo sensor data accurately, it is essential to validate these predictions through controlled experiments. Therefore, the aim of this study was to biomechanically induce plastic deformation in cadaveric ovine plated constructs and predict sensor signals at yield using sample-specific FE models. The virtual AOFM was first calibrated using a four-point bending test and compared to the actual sensor output. Seven cadaveric ovine tibiae were then prepared; each was instrumented with either titanium or stainless-steel Locking Compression Plates (LCP), each mounted with an AOFM, and osteotomized to create varying gap sizes. These samples were subjected to axial compression until plastic deformation, and the sensor signal at yield point was recorded. Concurrently, the physical samples were digitally replicated via FE analysis using previously validated material properties [4] and the predicted sensor signals at yield, adjusted by the calibration factor, were compared against the experimental signals. The calibration showed a liner relationship with a high correlation between experimental and virtual sensor signals in the four-point bending test, with an R² value greater than 0.99. When comparing the FE-predicted and measured sensor data from the cadaveric experiments, the concordance correlation coefficient was 0.89, suggesting strong agreement. Additionally, the standard error of the estimate was 187.0 N, with a relative standard error of 11.9%. This study demonstrates that FE models can predict AOFM signals at the onset of plastic plate bending in a controlled environment, enabling retrospective model validation and aiding in patient-specific rehabilitation protocols. Future research should focus on expanding these findings to in-vivo data to further refine predictive models. Acknowledgements. This study was supported by the AO Foundation via AOTRAUMA (Grant:AR2021_03).