Reliability of the modified Walch classification for advanced glenohumeral osteoarthritis using 3-dimensional computed tomography analysis: a study of the ASES B2 Glenoid Multicenter Research Group.

Background: Variations in glenoid morphology affect surgical treatment and outcome of advanced glenohumeral osteoarthritis (OA). The purpose of this study was to assess the inter- and intraobserver reliability of the modified Walch classification using 3-dimensional (3D) computed tomography (CT) imaging in a multicenter research group.

Methods: Deidentified preoperative CTs of patients with primary glenohumeral OA undergoing anatomic or reverse total shoulder arthroplasty (TSA) were reviewed with 3D imaging software by 23 experienced shoulder surgeons across 19 institutions. CTs were separated into 2 groups for review: group 1 (96 cases involving all modified Walch classification categories evaluated by 12 readers) and group 2 (98 cases involving posterior glenoid deformity categories [B2, B3, C1, C2] evaluated by 11 readers other than the first 12). Each case group was reviewed by the same set of readers 4 different times (with and without the glenoid vault model present), blindly and in random order. Inter- and intraobserver reliabilities were calculated to assess agreement (slight, fair, moderate, substantial, almost perfect) within groups and by modified Walch classification categories.

Results: Interobserver reliability showed fair to moderate agreement for both groups. Group 1 had a kappa of 0.43 (95% confidence interval [CI]: 0.38, 0.48) with the glenoid vault model absent and 0.41 (95% CI: 0.37, 0.46) with it present. Group 2 had a kappa of 0.38 (95% CI: 0.33, 0.43) with the glenoid vault model absent and 0.37 (95% CI: 0.32, 0.43) with it present. Intraobserver reliability showed substantial agreement for group 1 with (0.63, range 0.47-0.71) and without (0.61, range 0.52-0.69) the glenoid vault model present. For group 2, intraobserver reliability showed moderate agreement with the glenoid vault model absent (0.51, range 0.30-0.72), which improved to substantial agreement with the glenoid vault model present (0.61, range 0.34-0.87).

Discussion: Inter- and intraobserver reliability of the modified Walch classification were fair to moderate and moderate to substantial, respectively, using standardized 3D CT imaging analysis in a large multicenter study. The findings potentially suggest that cases with a spectrum of posterior glenoid bone loss and/or dysplasia can be harder to distinguish by modified Walch type because of a lack of defined thresholds, and the glenoid vault model may be beneficial in determining Walch type in certain scenarios. The ability to reproducibly separate patients into groups based on preoperative pathology, including Walch type, is important for future studies to accurately evaluate postoperative outcomes in TSA patient cohorts.