Effect of Glenoid Prosthesis Design Variables on Glenoid Bone Remodeling: a Finite Element Based Simulation and Validation Study

Introduction Glenoid prosthesis loosening is a major complication of shoulder arthoplasty, thus highlighting the importance of improving glenoid prostheses longevity [1]. Finite element (FE) stress analyses have been used to assist joint prosthesis designs. Prior FE analyses have employed static or non-linear solutions to compare design variations with respect to stresses in the glenoid bone and prosthesis [2, 3]. However, bone is known to modify its shape and structure based on Wolff’s Law. To design longer-lasting glenoid prostheses, bone’s remodeling due to prosthesis fixation geometries and materials needs to be considered. Previous numerical models have not incorporated Wolff’s Law to simulate normal glenoid bone remodeling or remodeling in response to implanted glenoid prostheses. In this study, the internal bone structure was remodeled using strain-energy, while the shape of the outside bone remained unchanged. The aims of this study were to (1) create glenoid prostheses with varying fixation geometries and material combinations, and (2) perform subject-specific 2-D FE-based numerical simulation of glenoid bone remodeling in response to various prosthesis design variables. Materials and Methods An intact normal scapula cadaver specimen (male 55y) was computed tomography (CT) imaged. The axial section at the glenoid center was selected for bone segmentation. The bone contours were imported into computer-aided design software SolidWorks®. 12 glenoid prosthesis designs were created having 3 different fixation geometries and 4 different material combinations. Simulated surgical implantation of these glenoid prostheses were performed in SolidWorks®. 2-D FE models for all designs were generated using the finite element analysis software Ansys®. Six-node solid triangle elements with average edge length of 0.3mm were created (Figure 1 (left)). The models had on average 12,473 elements and 25,332 nodes. Bone was modeled as linearly elastic, isotropic, and nonhomogeneous material. A custom program assigned location-specific properties to each bone element based on the CT value at that element. Poisson’s ratio for bone was assigned 0.3. The glenoid prosthesis was modeled as linearly elastic, isotropic, and homogeneous material [metal: Ti-6Al-4V alloy (E=110GPa, ν=0.33) and polyethylene (PE): (E=1.2GPa, ν=0.46)]. 3 loads were applied. Center load of 1 bodyweight (force=800N) simulated the arm at 90° abduction, while posteriorand anterior-offset loads of half bodyweight (force=400N) simulated extremes of the range of motion. The medial border was fixed to avoid rigid body motion. All interfaces were considered to be fully bonded. The glenoid internal bone remodeling simulations were based on element strain-energy density adapted from Jacobs et al and Weinans et al [4, 5] and described in Sharma et al [6]. Prior to execution of the simulations, each FE model was assigned the actual specimen’s location-specific material properties. Each load was applied by itself and the FE problem was solved for the element reference stimulus values. For simulations, the bone was assigned an initial homogeneous density of 0.6g/cc [4]. The 3 loads were applied consecutively for a total of 300 iterations. After each iteration, the remodeling stimulus was compared with the reference stimulus for all elements and the material properties were modified. For validation, the simulation was performed for the intact bone FE model. Upon completion, the predicted intact bone apparent density was compared with the actual specimen on a location-specific basis. For analysis, six regions of interest were considered, one in distal glenoid, two in mid-glenoid, and three in proximal glenoid. The mean predicted bone apparent density and the mean von Mises stress values were compared between all glenoid prosthesis FE models. Results The glenoid bone remodeling process results are shown in Figure 1 (right). The process was validated for the intact glenoid FE model. The predicted intact glenoid bone density (mean±sd:0.79±0.54g/cc) was similar to that of the actual specimen (0.8±0.46g/cc, p=0.36). The predicted distal glenoid bone density was lower in glenoid prosthesis FE models compared to intact. Increased reduction was seen with metal fixations (0.82±0.7g/cc). PE fixations (0.92±0.6g/cc) resulted in predicted density closer to intact (0.94±0.6g/cc). The predicted proximal glenoid bone density increased with PE fixations compared to intact. With metal fixations, the bone density in proximal anterior (0.57±0.1g/cc) was reduced, whereas it increased in proximal center (0.63±0.2g/cc) and posterior (0.7±0.3g/cc) regions compared to intact (anterior:0.68±0.3g/cc, center:0.55±0.2g/cc, posterior:0.66±0.2g/cc). The posterior fixation geometry showed greater reduction in the predicted distal glenoid bone density (0.88±0.8g/cc) compared to intact glenoid (0.94±0.6g/cc), and the center (0.92±0.6g/cc) and anterior (0.92±0.6g/cc) fixation geometries. The mean von Mises stress decreased for metal fixations in distal (0.263±0.3MPa) and proximal anterior (0.02±0.03MPa) glenoid compared to the intact bone (distal:0.274±0.3MPa, proximal anterior:0.06±0.05MPa) and PE fixations (distal:0.278±0.3MPa, proximal anterior:0.06±0.03MPa), respectively. Discussion The FE-based subject-specific adaptive bone remodeling simulation was successfully executed for the intact and glenoid prostheses models. The simulation process was validated for the normal intact bone model. The predicted bone density distribution obtained from an initial homogeneous density was similar to the actual specimen location specific bone density. No validation for the glenoid prosthesis FE models was performed in this study. The metal fixations resulted in lower predicted distal glenoid bone density compared to the intact and PE fixation FE models. This may be attributed to the increased stiffness of the metal fixation compared to the PE. This observation is in coherence with the “stress-shielding” phenomenon reported in glenoid stress analysis literature [3]. The anatomically angled PE fixation designs resulted in relatively close to intact predicted bone density and von Mises stress values. This is comparable to a prior stress analysis study comparing center keel and off-set keel designs [7]. The technique developed in this preliminary study may be a useful tool to examine the effects of glenoid prosthesis design on glenoid bone remodeling. Improved understanding of prosthesis induced remodeling should lead to improved designs for longer lasting glenoid prostheses.