The Metha short stem (B. Braun, Aesculap, Tuttlingen, Germany) (Figure 1a) is a cementless partial collum sparing implant with a metaphyseal anchorage [16, 17]. It has a 20 µm thick Calciumphosphate layer in the proximal and middle part, and is polished distally. According to previous studies, a Metha stem size 3 with a 135° adapter was implanted [18, 19].
The CLS standard stem (Zimmer, Warsaw, Indiana, USA) (Figure 1b) is a cementless, straight and collarless implant with a proximal anchorage and proven good long-term results. It has a porous surface treatment (Ra = 4.4 µm) and a rectangular cross-section with sharp proximal, anterior and posterior ribs. According to previous studies, a CLS prostheses size 13.25 with a 135° neck-shaft-angle was used [20, 21].
In order to acquire the FE models, the specimens were derived from ongoing biomechanical experiments evaluating the micromotions in our Laboratory  (Figure 1). This study used synthetic composite bones (Model 3306, sawbones Pacific Research Laboratories, USA) to avoid geometric and mechanical variances as seen in cadaveric bones . Besides, the composite bones were found to have analog bone properties and mimic the structural properties of average healthy adult human bones .
The femora are positioned laterally by 9° in the frontal plane and dorsally by 16° in the sagittal plane to create physiological loading conditions . All implantations were performed by one senior surgeon (FS) according to the manufacturers’ instructions. A sinusoid dynamic load was applied downward vertically with an amplitude between 300 N to 1700 N and a frequency of 1 Hz, to simulate a post-operative patient with 70 kg body weight walking on level ground .
One intact native composite femur and two composite femora, one with implanted SHA (Metha) and one with THA (CLS), were used to create the FE models. The homogenized FE models were generated as outlined in Figure 2. Details about the model generation are provided in the following subsections in accordance to the four modelling steps: (1) 3-dimensional (3D) model generation of the femurs and implants in the clinical quantitative computerized tomography (QCT) scans of the prepared samples, (2) alignment of the implants, (3) material properties assignment, and (4) boundary condition assignment. The final subsection dealt with model solving and post-processing. Except stated otherwise, all these steps were conducted using custom-made programs in Python, C++ and Fortran as recently described by Chevalier Y. .
QCT scanning and 3D model generation
QCT scans of the implanted and intact specimens, as well as of the two selected femoral stems were conducted using a clinical computed tomography (CT) scanner (64-slice) (Siemens Somatom Emotion 6, Siemens AG, Germany). 3D images were reconstructed with a voxel size of approximately 0.17 × 0.17 × 0.6 mm3. Trabecular and cortical bone were segmented from the CT scans based on gray-scale transition values using in-house written code . Furthermore, 3D models of the two prostheses were also created after segmentation of the implant in the implant scan images.
Alignment of the implants
To assure the accurate implant position, the 3D models of the isolated SHA and THA stems were placed by aligning them to the positions as recorded in the CT scans of the femur with the implanted stem. The positioned implant models were then converted to digitized images with custom codes in Python and insight toolkit (ITK) . Then, the bone and implant images were combined into a binarized image with three distinct regions (compact bone, trabecular bone, and stem), and then meshed with 2-mm 4-noded tetrahedral with computational geometry algorithms library (CGAL)  to create 3D models of the femurs with stems as described previously . The merged models of SHA (Metha) and THA (CLS) femur contained between 52 and 58×103 nodes, and 21 and 24×104 elements, respectively. The merged model of the native femur contained approximately 66×103 nodes and 29×104 elements.
Material properties assignment
Synthetic composite femur was characterized by isotropic material properties of cortical bone and trabecular bone, which were assumed to be linearly elastic and homogeneous with Poisson’s ratio setting to 0.35. Cancellous stiffness modulus value was designed by 155 MPa, and cortical stiffness value was designed by 16.7 GPa. The stiffness modulus values of the SHA and THA stem were designed by 25 GPa.
Boundary condition assignment
Models were loaded to mimic the experimental conditions of the specimens as in the in vitro study . Loading vector was defined based on the anatomical orientation and corresponded to a 9 degree angle in the frontal plane and 16 degrees angle in the sagittal plane. A resultant load with 1400N was applied on the tip nodes of the prosthesis neck, while bottom nodes of the bone were fully constrained.
Solving and post-processing
Linear analyses were performed using Abaqus 6.13 (Simulia, Dassault Systèmes, Vélizy-Villacoublay, France). To analyze the cortical stress distribution patterns of the FE models, a custom code was written dividing the FE models into equal regions of 10 mm starting from proximal to distal in the z-axis direction. The mean and peak values of von Mises stress in each region were then computed. To further analyze the mean cortical stress distribution around the stems, the femora were divided into a proximal (region 1-6), metaphyseal (region 7-12) and distal (region 13-18) region. Visualization of the mean stresses for the FE models was done in Paraview v3.14.
The FEA results (Native, SHA and THA) are depicted and described comparatively for the mean cortical stress distributions and peak cortical stress distributions. To further analyze the mean cortical stress distributions in the proximal, metaphyseal and distal region of the three groups, one-way analysis of variance (with a Bonferroni post hoc test) was conducted. Data analysis and graphic representation were conducted using GraphPad Prism 5 (GraphPad Software, San Diego, USA). The level of significance was set at 0.05.