Introduction:

While survivorship of total knee arthroplasty (TKA) is excellent, up to 25% of patients remain dissatisfied with their outcome [1, 2]. Knee instability, which is common during high demand activities, contributes to patient dissatisfaction [3]. As younger patients undergo TKA, longevity requirements and functional demands will rise [4]. Design factors influence the functional outcome of the procedure [5, 6], although in clinical studies it can be difficult to distinguish joint mechanics differences between designs due to confounding variability in patient-related factors. The objective of the current study was to assess the stability and mechanics of several current TKA designs during high-demand dynamic activities using a computational model of the lower limb.

Introduction:

There is substantial range in kinematics and joint loading in the total knee arthroplasty (TKA) patient population. Prospective TKA designs should be evaluated across the spectrum of loading conditions observed in vivo. Recent research has implanted telemetric tibial trays into TKA patients and measured loads at the tibiofemoral (TF) joint [1]. However, the number of patients for which telemetric data is available is limited and restricts the variability in loading conditions to a small subset of those which may be encountered in vivo. However, there is a substantial amount of fluoroscopic data available from numerous TKA patients and component designs [2]. The purpose of this study was to develop computational simulations which incorporate population-based variability in loading conditions derived from in vivo fluoroscopy, for eventual use in computational as well as experimental activity models.

Complications of the patellofemoral (PF) joint remain a common cause for revision of total knee replacements. PF complications, such as patellar maltracking, subluxation, dislocation and implant failure, have been linked to femoral and patellar component alignment. Computational analyses represent an efficient method for investigating the effects of patellar and femoral component alignment and loading on output measures related to long term clinical success (i.e. kinematics, contact mechanics) and can be utilized to make direct comparisons between common patellar component design types. Prior PF alignment studies have generally involved perturbing a single alignment parameter independently, without accounting for interaction effects between multiple parameters. The objective of the current study was to determine critical alignment parameters, and combinations of parameters, in three patellar component designs, and assess whether the critical parameters were design specific.

A dynamic finite element (FE) model of an implanted PF joint was applied in conjunction with a 100-trial Monte Carlo probabilistic simulation to establish relationships between alignment and loading parameters and PF kinematics, contact mechanics and internal stresses (Figure 1). Seven parameters, including femoral internal-external (I-E) alignment, patellar I-E, flexion-extension (F∗∗∗∗∗E) and adduction-abduction (A-A) rotational alignment, and patellar medial-lateral (M-L) and superior-inferior (S-I) translational alignment, as well as percentage of the quadriceps load on the vastus medialis obliquus (VMO) tendon, were perturbed in the probabilistic analysis. Ten output parameters, including 6-DOF PF kinematics, peak PF contact pressure, contact area, peak von Mises stress and M-L force due to contact, were evaluated at 80 intervals during a simulated deep knee bend. Three types of patellar component designs were assessed; a dome-compatible patellar component (dome), a medialized dome-compatible patellar component (modified dome), and an anatomic component (anatomic). Model-predicted bounds at 5 and 95% confidence levels were determined for each output parameter throughout the range of femoral flexion (Figure 2). Traditional sensitivity analysis, in addition to a previously described coupled probabilistic and principal component analysis (probabilistic-PCA) approach, were applied to determine the relative importance of alignment and loading parameters to knee mechanics in each of the three designs.

The dome component demonstrated the least amount of variation in contact mechanics and internal stresses, particularly in the 30–100° flexion range, with respect to alignment and loading variability. The modified dome had substantially reduced M-L contact force when compared with the dome. The anatomic design, while wide bounds of variability were predicted, had consistently greater contact area and lowered contact pressure than the dome and modified dome designs. The anatomic design also reproduced more natural sagittal plane patellar tilt than the other components. All three designs were most sensitivity to femoral I-E alignment. Thereafter, sensitivity to component alignment was design specific; for the anatomic component, the main alignment parameter was F-E, while for the domed components it was a combination of F-E and translation (M-L and S-I) (Figure 3). Understanding the relationships and design-specific dependencies between alignment parameters can add value to surgical pre-operative planning, and may help focus instrumentation design on those alignment parameters of primary concern.

Experimental knee simulators for component evaluation or in vitro testing provide valuable insight into the mechanics of the implanted joint. The Kansas knee simulator (KKS) is an electro-hydraulic whole joint knee simulator, with five actuators at the hip, ankle and quadriceps muscle used to simulate a variety of dynamic activities in cadaveric specimens. However, the number and type of experimental tests which can feasibly be performed is limited by the need to make physical component parts, obtain cadaveric specimens and the substantial time required to carry out each test. Computational simulations provide a complementary toolset to experimental testing; experimental data can be used to validate the computational model which can subsequently be used for early evaluation and ranking of component designs. The objective of this study was to explore potential improvements to loading and boundary conditions in current computational/experimental models, specifically the KKS, in order to develop representations of several activities of daily living (ADLs) which reproduce in vivo knee joint loading measurements.

An existing finite element model of the KKS was modified to extend the capability, and improve the fidelity, of the computational model beyond the experimental setup. An actuator to allow anterior-posterior (A-P) motion of the hip was included and used to prescribe relative hip-ankle A-P kinematics during the simulations. The quadriceps muscle, which in the experimental simulator consisted of a single quadriceps bundle with a point-to-point line of action, was divided into four heads of the quadriceps with physiological muscle paths. The hamstrings muscle, which was not present in the experiment, was represented by point-to-point actuators in four bundles. A flexible control system was developed which allowed control of the quadriceps and hamstrings actuators to match a knee flexion profile, similar to actuation of the experimental KKS, but also allowed control of the compressive tibiofemoral (TF) joint force, medial-lateral (M-L) load distribution, internal-external (I-E) torque and A-P load at the joint. A series of sensors, measuring all six load components on the medial and lateral compartments of the tibial insert, as well as knee flexion angle, were incorporated into the simulation. Instantaneous measurements from the sensors were fed to a control system, implemented within an Abaqus/Explicit user subroutine (Figure 1). The controller was used to drive actuators in the FE model to match target in vivo joint loading profiles, measured from telemetric patient data. The control system was applied to recreate in vivo loading conditions at the knee joint during three ADLs for three different subjects (Figure 2), with excellent agreement between simulation joint loading conditions and the target profiles; RMS differences were less than 1°, 80N, 2.5%, and 0.8Nm for knee flexion angle, compressive joint load, M-L load split and I-E torque, respectively, throughout the cycle for all three activities (Figure 3). The flexible nature of the control system ensures that it can be used to evaluate an expansive variety of ‘effect of’ studies, as well as to determine advanced loading profiles for the experimental simulator.