The number of total knee joint replacements has increased dramatically, from 28,000 in 2004 to over 73,000 in 2008 in the UK. This increase in procedures means that there is a need to assess the performance of an implant design in the general population. For younger, more active patients, cementless tibial fixation is an attractive alternative means of fixation and has been used for over 30 years. However, the clinical results with cementless fixation have been variable, with reports of extensive radiolucent lines, rapid early migration and aseptic loosening [1]. This study investigates the inter-patient variability of bone strain at the implant-bone interface of 130 implanted tibias over a full gait cycle. A large scale FE study of a full gait cycle was performed on 130 tibias implanted with a cementless tibia tray (PFC Sigma, DePuy Inc, USA). A population of tibias was generated from a statistical shape and intensity (SSI) model [2]. The tibia tray was automatically positioned and implanted using ZIBAmira (Zuse Institute Berlin, Germany). Cutting and implanting were performed using Boolean operations on the meshed surfaces of the tibia and implant. After generation of a volume mesh from the resulting surface, the bone modulus was mapped onto the new mesh. The FE models were processed in Abaqus (SIMULIA, RI, USA). Associated force data (axial, anterior-posterior and medial-lateral forces and flexion-extension, varus-valgus and internal-external moments) was sampled from a statistical model of the gait cycle derived from musculoskeletal modelling of 20 elderly healthy subjects. Patient weight was estimated using the length of the tibia and a BMI sampled from NHANES data. Loads were applied to four groups of nodes on the tibia tray (anterior, posterior, medial and, lateral) for 51 steps in the gait cycle. The bone and implant were assumed to be bonded, simulating the osseointegrated condition.Introduction
Methods
Implant positioning is a critical factor in assuring the primary stability of cementless Total Hip Replacements (THRs). Although it is under the direct control of surgeons, finding the optimal implant position and achieving a perfect fit remain a challenge even with the advent of computer navigation. Placement of the femoral stem in an excessive ante/retroversion or varus/valgus orientation can be detrimental to the performance of THR. To determine the effect of such malalignment, finite element (FE) computer modelling is often used. However, this can be time consuming since FE meshes must be repeatedly generated and solved each time for a range of defined implant positions. In the present study, a mesh morphing technique is developed for the automatic generation of FE models of the implanted femur; in this way, many implant orientations can be investigated in a single analysis. An average femur geometry generated from a CT scan population of 13 male and 8 female patients aged between 43 and 84 years was considered. The femur was virtually implanted with the Furlong HAC titanium alloy stem (JRI Ltd, Sheffield, UK) and placed in the medullary canal in a baseline neutral nominal position. The head of the femur was then removed and both femur and implant volumes were joined together to form a single piece that was exported into ANSYS11 ICEM CFD (ANSYS Inc., 2008) for meshing. To adequately replicate implant ante/retroversion, varus/valgus or anterior/posterior orientations, the rigid body displacement of the implant was controlled by three rotations with respect to a local coordinate system. One hundred different implant positions were analysed and the quality of the morphed meshes analysed for consistency. To check the morphed meshes, corresponding models were generated individually by re-positioning the implant in the femur. Selected models were solved to predict the strain distribution in the bone and the boneimplant relative micromovements under joint and muscle loading. A good agreement was found for bone strains and implant micromotions between the morphed models and their individually run counterparts. In the postprocessing stage further metrics were analysed to corroborate the findings of the morphed and individually run models. These included: average and maximum strains in bone interface area and its entire volume, percentage of bone interface area and its volume strained up to and beyond 0.7%; implant average and maximum micromotions and finally percentages of implant area undergoing reported critical micromotions of 50 μm, 100 μm and 150 μm for bone in growth. Excellent correlation was observed in all cases. In conclusion, the proposed technique allowed an automatic generation of FE meshes of the implanted femur as the implant position varies; the required computational resources were considerably reduced and the biomechanical response was evaluated. This model forms a good basis for the development of a tool for multiple statistical analyses of the effects of implant orientation in pre-clinical studies.
Excessive implant migration and micromotion have been related to eventual implant loosening. The aim of this project is to develop a computational tool that will be able to predict the mechanical performance of a cementless implant in the presence of uncertainty, for example through variations in implant alignment or bone quality. To achieve this aim, a computational model has to be developed and implemented. However, to gain confidence in the model, it should be verified experimentally. To this end, the present work investigated the behavior of a cementless implant experimentally, and compared the results with a computational model of the same test setup. A synthetic bone (item 3406, Sawbones Europe AB, Sweden) was surgically implanted with a Furlong cementless stem (JRI, Sheffield, UK) in a neutral position and subjected to a compression fatigue test of −200 N to −1.6 kN at a frequency of 0.5 Hz for 50000 cycles. Measurements of the micromotion and migration were carried out using two linear variable differential transducers and the strain on the cortex of the femur was measured by a digital image correlation system (Limess Messtechnik &
Software Gmbh). A three-dimensional model was generated from computed tomography scans of the implanted Sawbone and converted to a finite element (FE) model using Simple-ware software (Simpleware Ltd, Exeter, UK). Face-to-face elements were used to generate a contact pair between the Sawbone and the implant. A contact stiffness of 6000 N/m and a friction coefficient of 0.3 were assigned. The analysis simulated a load of −1.6 kN applied to the head of the implant shortly post implantation. The motions and strains recorded in the experiment were compared with the predictions from the computational model. The micromotion (the vertical movement of the implant during a single load cycle), was measured at the proximal shoulder, at the distal tip of the implant and at the bone-implant interface. The maximum value calculated proximally using FE was 61.3 μm compared to the experimental value of 59.6 μm. At the distal end, the maximum micromotion from FE was 168.9 μm compared to 170 μm experimentally. As a point of reference, some authors have suggested that in vivo, fibrous tissue formation may take place at the bone-implant interface when the micromotion is above 150 μm. The maximum micromotion found computationally at this interface was 99 μm which is below the threshold value defined. The longitudinal strain over the surface of the bone was variable and reached values of up to 0.15% computationally and 0.4% experimentally; this may be related to the coordinate systems used. However, it was noted that digital image correlation identified qualitatively similar strain patterns, and has great potential for measuring low level surface strains on bone. In conclusion, the good correlation between the computational modelling and experimental tests provides confidence in the model for further investigations using probabilistic analyses where more complex configurations (for example change in implant alignment) can be analyzed.
