Machine Learning Algorithms Identify Optimal Sagittal Component Position in Total Knee Arthroplasty

Abstract

BackgroundAdvanced technologies, like robotics, provide enhanced precision for implanting total knee arthroplasty (TKA) components; however, the optimal targets for implant position specifically in the sagittal plane do not exist. This study identified sagittal implant position which may predict improved outcomes using machine learning algorithms. MethodsA retrospective review of 1091 consecutive TKAs was performed. All TKAs were posterior cruciate ligament retaining or sacrificing with an anterior-lip (49.4%) or conforming bearing (50.6%) and performed with modern perioperative protocols. Preoperative and postoperative tibial slope and postoperative femoral component flexion were measured with standardized radiographic protocols. Analysis groups were categorized by satisfaction scores and the Knee Society Score question ‘does this knee feel normal to you?’ Machine learning algorithms were used to identify optimal sagittal alignment zones that predict superior satisfaction and knees “always feeling normal” scores. ResultsMean age and median body mass index were 66 years and 34 kg/m2, respectively, with 67% being female. The machine learning model predicted an increased likelihood of being “satisfied or very satisfied” and a knee “always feeling normal” with a change in tibial slope closer to native (−2 to +2°) and femoral component flexion 0 to +7°. Worse outcomes were predicted with any femoral component extension, femoral component flexion beyond +10°, and adding or removing >5° of native tibial slope. ConclusionSuperior patient-reported outcomes were predicted with approximating native tibial slope and incorporating some femoral component flexion. Deviation from native tibial slope and excessive femoral flexion or any femoral component extension were predictive of worse outcomes. Level of EvidenceTherapeutic level III.