We studied the intra- and interobserver reliability of measurements of the position of the components after total knee replacement (TKR) using a combination of radiographs and axial two-dimensional (2D) and three-dimensional (3D) reconstructed CT images to identify which method is best for this purpose.

A total of 30 knees after primary TKR were assessed by two independent observers (an orthopaedic surgeon and a radiologist) using radiographs and CT scans. Plain radiographs were highly reliable at measuring the tibial slope, but showed wide variability for all other measurements; 2D-CT also showed wide variability. 3D-CT was highly reliable, even when measuring rotation of the femoral components, and significantly better than 2D-CT. Interobserver variability in the measurements on radiographs were good (intraclass correlation coefficient (ICC) 0.65 to 0.82), but rotational measurements on 2D-CT were poor (ICC 0.29). On 3D-CT they were near perfect (ICC 0.89 to 0.99), and significantly more reliable than 2D-CT (p < 0.001).

3D-reconstructed images are sufficiently reliable to enable reporting of the position and orientation of the components. Rotational measurements in particular should be performed on 3D-reconstructed CT images. When faced with a poorly functioning TKR with concerns over component positioning, we recommend 3D-CT as the investigation of choice.

The rotational alignment of the tibia is an unresolved issue in knee replacement. A poor functional outcome may be due to malrotation of the tibial component. Our aim was to find a reliable method for positioning the tibial component in knee replacement.

CT scans of 19 knees were reconstructed in three dimensions and orientated vertically. An axial plane was identified 20 mm below the tibial spines. The centre of each tibial condyle was calculated from ten points taken round the condylar cortex. The tibial tubercle centre was also generated as the centre of the circle which best fitted eight points on the outside of the tubercle in an axial plane at the level of its most prominent point.

The derived points were identified by three observers with errors of 0.6 mm to 1 mm. The medial and lateral tibial centres were constant features (radius 24 mm (sd 3), and 22 mm (sd 3), respectively). An anatomical axis was created perpendicular to the line joining these two points. The tubercle centre was found to be 20 mm (sd 7) lateral to the centre of the medial tibial condyle. Compared with this axis, an axis perpendicular to the posterior condylar axis was internally rotated by 6° (sd 3). An axis based on the tibial tubercle and the tibial spines was also internally rotated by 5° (sd 10).

Alignment of the knee when based on this anatomical axis was more reliable than either the posterior surfaces or any axis involving the tubercle which was the least reliable landmark in the region.