Background: The coronal plane alignment of the knee (CPAK) classification is increasingly used in daily practice and in research. The reliability of the angles measured has been estimated, but not that of the classification itself. This study aimed to assess the inter- and intra-observer reliability of the CPAK classification using different imaging modalities: full-length weight-bearing radiographs, two-dimensional (2D) electrons-optiques systemes (EOS), and three-dimensional (3D) EOS in osteoarthritic knees.

Methods: In this monocentric repeat cross-sectional study, 39 patients (78 knees) with all three types of images were included. There were two examiners who performed each planning run twice, with an interval of at least two weeks between the two planning runs for full weight-bearing radiographs, 2D EOS, and 3D EOS. The intraclass correlation coefficient (ICC) was used to evaluate the reliability of angles, and the Cohen's Kappa coefficient was used to evaluate the reliability of the CPAK classification.

Results: For CPAK classification, the intraobserver reliability (Kappa _1st = 0.39; Kappa _2nd = 0.48) and the interobserver reliability (Kappa = 0.39) were poor. The intraobserver reliability of the lateral distal femoral angle (LDFA) was good (ICC_1st = 0.85) and excellent (ICC_2nd = 0.92). For medial proximal tibial angle (MPTA), the intraobserver reliability was moderate (ICC_1st = 0.72) and good (ICC_2nd = 0.75). The interobserver reliability of the LDFA was excellent (0.91) and good (0.80) for MPTA. Comparing the three different long leg radiographs, the intraobserver and interobserver reliability were better with full weight-bearing radiographs (Kappa_1st = 0.55, Kappa _2nd = 0.61, Kappa = 0.57) than with 2D EOS (Kappa _1st = 0.35, Kappa _2nd = 0.35, Kappa _{inter-reliability} = 0.25) and 3D EOS (Kappa _1st = 0.40, Kappa _2nd = 0.34, Cohen's Kappa _{inter-reliability} = 0.22).

Conclusions: The reliability of the CPAK classification is poor to moderate, and full-length weight-bearing radiographs should be preferred over EOS.

Level of evidence: IV, retrospective cohort.