Proximal humerus fractures are the third most common fragility fractures with treatment remaining challenging. Mechanical fixation failure rates of locked plating range up to 35%, with 80% of them being related to the screws perforating the glenohumeral joint. Secondary screw perforation is a complex and not yet fully understood process. Biomechanical testing and finite element (FE) analysis are expected to help understand the importance of various risk factors. Validated FE simulations could be used to predict perforation risk. This study aimed to (1) develop an experimental model for single screw perforation in the humeral head and (2) evaluate and compare the ability of bone density measures and FE simulations to predict the experimental findings.

Screw perforation was investigated experimentally via quasi-static ramped compression testing of 20 cuboidal bone specimens at 1 mm/min. They were harvested from four fresh-frozen human cadaveric proximal humeri of elderly donors (aged 85 ± 5 years, f/m: 2/2), surrounded with cylindrical embedding and implanted with a single 3.5 mm locking screw (DePuy Synthes, Switzerland) centrally. Specimen-specific linear µFE (ParOSol, ETH Zurich) and nonlinear explicit µFE (Abaqus, SIMULIA, USA) models were generated at 38 µm and 76 µm voxel sizes, respectively, from pre- and post-implantation micro-Computed Tomography (µCT) images (vivaCT40, Scanco Medical, Switzerland). Bone volume (BV) around the screw and in front of the screw tip, and tip-to-joint distance (TJD) were evaluated on the µCT images. The µFE models and BV were used to predict the experimental force at the initial screw loosening and the maximum force until perforation.

Initial screw loosening, indicated by the first peak of the load-displacement curve, occurred at a load of 64.7 ± 69.8 N (range: 10.2 – 298.8 N) and was best predicted by the linear µFE (R² = 0.90), followed by BV around the screw (R² = 0.87). Maximum load was 207.6 ± 107.7 N (range: 90.1 – 507.6 N) and the nonlinear µFE provided the best prediction (R² = 0.93), followed by BV in front of the screw tip (R² = 0.89). Further, the nonlinear µFE could better predict screw displacement at maximum force (R² = 0.77) than TJD (R² = 0.70). The predictions of non-linear µFE were quantitatively correct.

Our results indicate that while density-based measures strongly correlate with screw perforation force, the predictions by the nonlinear explicit µFE models were even better and, most importantly, quantitatively correct. These models have high potential to be utilized for simulation of more realistic fixations involving multiple screws under various loading cases. Towards clinical applications, future studies should investigate if explicit FE models based on clinically available CT images could provide similar prediction accuracies.

Summary. Time-lapsed CT offers new opportunities to predict the risk of cement leakage and to evaluate the mechanical effects on a vertebral body by monitoring each incremental injection step in an in-vitro vertebroplasty procedure. Introduction. Vertebroplasty has been shown to reinforce weak vertebral bodies and to prophylactically reduce fracture risks. However, bone cement leakage is a major vertebroplasty related problem which can cause severe complications. Leakage risk can be minimised by injecting less cement into the vertebral body, inevitably compromising the mechanical properties of the augmented bone, as a proper endplate-to-endplate connection of the injected cement is needed to obtain a mechanical benefit. Thus the cement flow in a vertebroplasty procedure requires a better understanding. This study aimed at developing a method to monitor the cement flow in a vertebral body and its mechanical effect. Materials and Methods. Eight fresh frozen human cadaveric vertebrae were prepared for augmentation by performing a bitrans- or bipedicular approach. Following they were XTremeCT-scanned (Scanco, Switzerland) at a nominal resolution of 82µm. A custom made setup enabled to fix the vertebrae in the CT bore (Siemens Emotion6) centrically. Bone cement (Vertecem V+, Synthes GmbH, Switzerland) was injected monopedicularly via a syringe driver (Harvard Apparatus, USA). Injection forces were recorded through a load cell (Type 9211, Kistler Instrumente AG, Switzerland) placed on the driver. Either a custom PEEK cannula or a trocar was inserted into each pedicle of a vertebra to allow artifact-free CT scanning. After each milliliter of injection a CT scan of the vertebra was performed at a nominal resolution of 0.63mm. Subsequently, the CT images were resampled to the original XTremeCT image and the cement cloud was segmented. The image data were then further processed for micro finite element (microFE) modeling (FAIM, Numerics88, Canada). The models were then solved for axial stiffness and Von Mises Stress (VMS) distribution. Finally, the vertebrae underwent a biomechanical quasistatic axial compression test (Mini Bionix II 858, MTS Systems Corp., USA). Results. Endplate-to-endplate connection of the cement was reached in 4 vertebrae. The average volume needed to reach the connection was 5.0±1.2 ml. Cement leakage occurred in all vertebrae, whereby in 4 cases the cement leaked into the spine channel. Each successive cement injection step was characterised with an increase of peak injection forces (16.5±12.7N at 1ml to 70.82±21.14N at 6ml). With respect to axial stiffness the mechanical tests and the microFE models correlated well (R. 2. = 0.778). Analyzing the top 100 VMS an elevated stress concentration between the endplate and the cement was observed unless the endplate was in direct contact with the cement. Conclusion. Cement flow can be monitored precisely at each injection step using the time-lapsed CT approach. Combined with microFE modeling the mechanical properties of the augmented bone can be evaluated for different incremental cement volumes injected. Our results suggest augmenting the bone until an endplate-to-endplate connection is established as otherwise partial filling would increase the risk of failure in the trabecular bone structure. This is in close agreement to other studies

Introduction. Glenoid loosening, still a main complication for shoulder arthroplasty, was suggested to be related implant design, surgical aspects, and also bone quality. However, typical studies of fixation do not account for heterogeneity in bone morphology and density which were suggested to affect fixation failure. In this study, a combination of cyclic rocking horse tests on cadaver specimens and microCT-based finite element (microFE) analysis of specimens of a wide range of bone density were used to evaluate the effects of periprosthetic bone quality on the risks of loosening of anatomical keeled or pegged glenoid implants. Methods. Six pairs of cadaveric scapulae, scanned with a quantitative computer tomography (QCT) scanner to calculate bone mineral density (BMD), were implanted with either cemented anatomical pegged or keeled glenoid components and tested under constant glenohumeral load while a humeral head component was moved cyclically in the inferior and superior directions. Edge displacements were measured after 1000, 4000 and 23000 test cycles, and tested for statistical differences with regards to changes or implant design. Relationships were established between edge displacements and QCT-based BMD below the implant. Four other specimens were scanned with high-resolution peripheral QCT (82µm) and implanted with the same 2 implants to generate virtual models. These were loaded with constant glenohumeral force, varying glenohumeral conformity and superior or inferior load shifts while internal stresses at the cement-bone and implant-cement interfaces were calculated and related to apparent bone density in the periprosthetic zone. Results. Mean displacements at the inferior and superior edges showed no statistical difference between keeled and pegged designs (p>0.05). Compression and distraction were however statistically different from the initial reference measurement at even 1000 and 4000 cycles for both implant designs (p<0.05). For both implant designs, superior and inferior distractions were generally highest at each measurement time in specimens where BMD below the lifting edge was lower, showing a trend of increased distraction with decreased BMD. Moreover, the microFE models predicted higher bone and cement stresses for specimens of lower apparent bone density. Finally, highest peak stresses were located at the cement-bone interface, which seemed the weaker part of the fixation. Discussion. With this combined experimental and numerical study, it was shown that implant distraction and stresses in the cement layer are greater in glenoids of lower bone density for both implant designs. This indicates that fixation failure will most likely occur in bone of lower density, and that fixation design itself may play a secondary role. These results have important impact for understanding the mechanisms of glenoid component failure, a common complication of total shoulder arthroplasty