Operator variation in applying a knee loading device for evaluation of tibial component loosening in total knee arthroplasty

Background: A CT-based method has been developed to aid diagnosis of aseptic loosening of the tibial component by scanning the knee in a loading device, and subsequent 3D-image analysis to quantify component displacement. This study evaluated the effect of operator differences in applying the loading device on the component displacement variables, using two image analysis protocols. Methods: Sixteen subjects underwent repeated CT examinations with valgus and varus loading. Two operators applied the loading device to each patient. With each load, a CT scan was made, and tibial component displacement relative to the tibia was quantified as rotation about the screw-axis, Maximum Total Point Motion and mean Target Registration Error. Two protocols were used: (1) analyzing the entire tibia (100 %) and (2) the proximal tibia (20 %) to mitigate tibia deformation. Inter-operator reliability and measurement error were assessed using intraclass correlation coefficient (ICCagreement), standard error of measurement, standard error of operator and smallest detectable change. Findings: The 100 % tibia protocol showed moderate-to-good ICCagreement(0.64– 0.84 for the different displacement variables) with standard error of measurement around 0.15 mm or degree. The 20 % tibia protocol showed poor-to-moderate ICCagreement, ranging from 0.17 to 0.31, with the standard error of measurement around 0.10 mm or degree. This protocol showed smaller measurement errors but poorer ICCagreement due to reduced subject variance explained by smaller apparent implant displacements. Operator related error was statistically and clinically negligible. The smallest detectable change values ranged between 0.27 and 0.44 mm or degree. Interpretation: The 100 % tibia protocol showed moderate-to-good reliability, whereas the 20 % protocol reduced reliability but lower measurement error. Level of evidence: Level II.