Comparison of Two Probabilistic Methods for Finite Element Analysis of Total Knee Replacement

Probabilistic Finite Element (FE) models have recently been developed to assess the impact of experimental variability present in knee wear simulator on predicted Total Knee Replacement (TKR) mechanics by determining the performance envelope of joint kinematics and contact mechanics. The gold standard for this type of analysis is currently the Monte Carlo method, however, this requires a larger number of trials and is therefore computationally expensive. Alternatively, probabilistic methods exist, such as response surface methods that can offer considerable savings in computational cost. The aim of the current study was to compare the performance envelopes obtained for three metrics (Anterior-Posterior (AP) translation, Internal-External (IE) rotation and peak Contact Pressure (CP)) for a FE model of TKR mechanics using two different probabilistic methods: the Monte Carlo technique and the Response Surface Method (RSM), implemented with PamCrash FE solver and PamOpt optimization/probabilistic software. The influence of implant alignment was considered, based on a study from the literature. The results of a 1000 trial Monte Carlo analysis were compared to predictions from 25, 50 and 100 trial response surface calculations. Overall, the Response Surface Method (RSM) was capable of predicting similar results to the Monte Carlo method, but with a substantially reduced computational cost (RSM-50 4 hours as compared to 4 days with the Monte Carlo method).