Purpose

Trochlear dysplasia is one of the anatomical risk factors for patellar dislocation, and the Dejour classification is commonly used for morphological assessment of trochlear dysplasia. However, it has been reported that the interobserver reliability of the Dejour classification is not high. The assessment using a three-dimensional (3D) bone CT model may improve the interobserver reliability of the Dejour classification, but the effectiveness of 3D-CT bone model remains unclear. Thus, the purpose of the present study was to investigate the effect of high-resolution 3D-CT bone model on the interobserver reliability of the Dejour classification. It was hypothesized that interobserver reliability of Dejour classification would be higher when using 3D-CT bone model than using an only plain radiograph.

Methods

Thirty patients (age: 24.6 years, M/F: 9/21) who underwent surgical treatment for patellar dislocation were included in the present study. A plain lateral radiograph and high-resolution CT images, which were taken prior to the surgery, were used for the evaluation. 3D-CT model of distal femur was created, and several anatomical landmarks were identified using the 3D processing software. Briefly, highest points of medial/lateral facet, and lowest points of the groove of trochlea with reference to the femoral posterior condyle axis were marked using axial images. If there were lowest points on the anterior part of distal femur other than trochlea groove, those points were marked. These points were connected and projected onto the 3D-CT model. Four orthopedic surgeons classified the cases into four groups (A, B, C, D) according to the Dejour classification three times. The first round used only lateral plain radiograph, the second round added 3D-CT models, and the third round used only axial CT images. Interobserver reliability was assessed using Fleiss’ kappa. In addition, interobserver reliability was assessed when simplifying the Dejour classification into two grades (A, mild; B-D,
severe). The categorization of Fleiss’ kappa was determined a priori as follows: 0.21-0.40, fair agreement; 0.41-0.60, moderate; 0.61-0.80, good; and 0.81-1.00, very good agreement.

Results

Interobserver reliability for the Dejour classification into four groups was 0.40 for the first round, 0.51 for the second round, and 0.37 for the third round. In the two-grade classification, interobserver reliability was 0.38 for the first round, 0.63 for the second round, and 0.57 for the third round. In 9 out of 30 cases (30%), two or more observers changed their classification from mild to severe or vice versa between the first and second rounds.

Conclusion

3D-CT bone model of distal femur with marked bony landmarks appears to improve interobserver reliability of the Dejour classification compared to the assessment using only plain radiograph, suggesting the usefulness of 3D-CT model in evaluating trochlear dysplasia