Predicting Clinical Outcomes in TKR

Background:  Computational modeling is increasingly useful for analyzing medical implants used in orthopedic medicine, such as joint replacements implanted into more than 3 million people worldwide each year. However, it is rarely used for predicting clinical outcomes due to lack of available clinical data suitable for use as model inputs and the difficulty of achieving correspondence between model predictions and clinical performance. During ongoing collaborations with surgeons specializing in total knee replacement (TKR), we obtained TKR alignment data and in vivo joint motions (i.e. kinematics) collected from patients, some of whom experienced poor clinical outcomes. The involved surgeons suspected TKR malalignment and inadequate soft tissue tension may be contributing to the poor outcomes, as those factors are known to contribute to half of all TKR failures within five years of the initial surgery. However, adequate methods for measuring soft tissue tension in vivo do not exist. Finite element (FE) modeling with ABAQUS provides a tool to utilize this clinical dataset and predict unknown soft tissue tensions for comparing TKR patients with good and poor clinical outcomes. This project develops a FE knee model that uses patient-specific clinical inputs (TKR alignment, kinematics) to drive patient-specific simulations and discriminate clinical outcomes based on model predictions. Modeling:  A generalized FE model of a knee joint implanted with a TKR was developed in ABAQUS/EXPLICIT from an open source dataset (http://digitalcommons.du.edu/natural_knee_data/7/). That dataset included images used to generate specimen-specific 3D anatomical geometry reconstructions and ligament attachment sites; data unavailable for the TKR patients in the clinical dataset.  3D rigid elements (R3D3) available in ABAQUS allowed modeling of bone geometry and implant components as rigid bodies and a computationally efficient approach for simulating joint motion by defining contact using a penalty-based method. ABAQUS also provided advantages in modeling the physiological mechanical behavior of soft tissues as bundled 1D tension-only springs using connector elements (CONN3D2), which allowed for the implementation of nonlinear force-displacement relationships using specimen-specific parameters calibrated in Isight. The versatility of ABAQUS connector elements was further utilized to model kinematic linkages as three cylindrical elements for applying patient-specific model inputs, including (i) 6 degrees of freedom (DOF) alignment of the TKR components relative to the bone, and (ii) 6 DOF kinematics of the TKR components to run the simulations.  Since soft tissues were modeled with connector elements, tension was output for each soft tissue bundle during the simulations and were compared between patients. Impact:  This study demonstrates the potential of the FE model to discriminate clinical outcomes after TKR when simulating patient-specific conditions. Initially, two TKR patients were simulated and model outputs showed the poor outcome patient exhibited higher tensions in medial soft tissues while the good outcome patient exhibited higher tensions in posterior soft tissues. Future work will include 15 additional TKR patients to further exercise the model across a broader patient population, informing surgical decisions and intraoperative technology.  Modeling this physiological system required patient-specific inputs into ABAQUS and successfully generated clinically useful model predictions.