Shape is only a weak predictor of deep knee flexion kinematics in healthy and osteoarthritic knees

Tibiofemoral shape influences knee kinematics but little is known about the effect of shape on deep knee flexion kinematics. The aim of this study was to examine the association between tibiofemoral joint shape and kinematics during deep kneeling in patients with and without osteoarthritis (OA). Sixty-one healthy participants and 58 patients with end-stage knee OA received a computed tomography (CT) of their knee. Participants completed full flexion kneeling while being imaged using single-plane fluoroscopy. Six-degree-of-freedom kinematics were measured by registering a three-dimensional (3D)-static CT onto 2D-dynamic fluoroscopic images. Statistical shape modeling and bivariate functional principal component analysis (bfPCA) were used to describe variability in knee shape and kinematics, respectively. Random-forest-regression models were created to test the ability of shape to predict kinematics controlling for body mass index, sex, and group. The first seven modes of the shape model up to three modes of the bfPCAs captured more than 90% of the variation. The ability of the random forest models to predict kinematics from shape was low, with no more than 50% of the variation being explained in any model. Furthermore, prediction errors were high, ranging between 24.2% and 29.4% of the data. Variations in the bony morphology of the tibiofemoral joint were weakly associated with the kinematics of deep knee flexion. The models only explained a small amount of variation in the data with high error rates indicating that additional predictors need to be identified. These results contribute to the clinical understanding of knee kinematics and potentially the expectations placed on high-flexion total knee replacement design.