Quantification of the difference between 3D CT and plain radiograph for measurement of the position of medial unicompartmental knee replacements

Abstract The aim of this study was to quantify the differences in measurements obtained from 3D Computed Tomography and plain radiograph, for the positioning of the tibial component of the Oxford unicompartmental knee replacement. Post-operative 3D Computed Tomography data and plain radiographs (long antero-posterior (AP) and short lateral) were obtained for 28 knees of patients who had undergone medial unicompartmental knee replacement. Parameters of the orientation/positioning of the tibial component: Varus/valgus, posterior tibial slope and rotation were measured with both modalities. Bland–Altman plots were used to calculate the 1.96 standard deviation limits of agreement (LOA) between imaging modalities. Intra class correlation was used to assess inter-method and inter-rater reliability (> 0.81 = very good reliability). Radiographs were less reliable in all parameters, when compared with 3D CT (intra class correlation coefficients: tibial rotation 0.94 vs 0.96, varus/valgus 0.76 vs 0.94, and posterior tibial slope 0.82 vs 0.92). The LOA were −4.9° to 3.4° for varus/valgus (bias −0.7°, one third > 3° different); −4.9° to 0.1° for posterior tibial slope (bias −2.4°, one third > 3° different); and −20.6° to 16.1° for rotation (bias −2.2°, one third > 10° different). There was some disagreement between measurement by 3D Computed Tomography and plain radiograph for all three parameters of tibial component orientation, especially tibial rotation. This will be particularly relevant to research into the relationship between the accuracy of implant positioning/orientation and patient satisfaction/implant survival rates. This method offers a more reliable standard for the reporting of knee arthroplasty.