INTER-OBSERVER VARIABILITY IN THE ASSESSMENT OF PLAIN RADIOGRAPHS FOLLOWING UNICOMPARTMENTAL KNEE ARTHROPLASTY

Abstract

Radiological measurements are an essential component of the assessment of outcome following knee arthroplasty. However, plain radiographic techniques can be associated with significant projectional errors because they are a two-dimensional (2D) representation of a three-dimensional (3D) structure. Angles that are considered within the target zone on one film may be outside that zone on other films. Moreover, these parameters can be subject to significant inter-observer differences when measured. The aim of our study therefore was to quantify the variability between observers evaluating plain radiographs following Unicompartmental knee arthroplasty.

Twenty-three observers, made up of Orthopaedic Consultants and trainees, were asked to measure the coronal and sagittal alignment of the tibial and femoral components from the post-operative long-leg plain radiograph of a Unicompartmental knee arthroplasty. A post-operative CT scan using the low dose Imperial knee protocol was obtained as well and analysed with 3D reconstruction software to measure the true values of these parameters. The accuracy and spread of the pain radiographic measurements were then compared with the values obtained on the CT.

On the femoral side, the mean angle in coronal alignment was 1.5° varus (Range 3.8, SD 1, min 0.1, max 3.9), whereas the mean angle in sagittal alignment was 8.6° of flexion (Range 7.5, SD 1.5, Min 3.7, Max 11.2). The true values measured with CT were 2.4° and 11.0° respectively. As for the tibial component, the mean coronal alignment angle was 89.7° (Range 11.6, SD 3.3, Min 83.8, Max 95.4), and the mean posterior slope was 2.4° (Range 8.7, SD 1.6, Min -2, Max 6.7). The CT values for these were 87.6° and 2.7° respectively.

We conclude that the plain radiographic measurements had a large scatter evidenced by the wide ranges in the values obtained by the different observers. If only the means are compared, the plain radiographic values were comparable with the true values obtained with CT (that is; accuracy was good) with differences ranging from 0.3° to 2.4°. The lack of precision can be avoided with the use of CT, particularly with the advent of low-dose scanning protocols.