Intraobserver and interobserver reliability of the modified Walch classification using radiographs and computed tomography.

Background: The Walch classification was introduced to classify glenoid morphology in primary glenohumeral osteoarthritis. A modified Walch classification was recently proposed, with 2 additional categories, B3 (monoconcave glenoid with posterior bone loss leading to retroversion > 15° or subluxation > 70%) and D (excessive anterior subluxation), as well as a more precise definition of subtypes A2 and C. The purpose of this study was to evaluate the intraobserver and interobserver agreement of the modified Walch classification system using both plain radiographs and computed tomography (CT).

Methods: Three fellowship-trained shoulder surgeons blindly and independently evaluated radiographs and CT scans of 100 consecutive shoulders (98 patients) with primary glenohumeral osteoarthritis and classified all shoulders according to the modified Walch classification in 4 separate sessions, each 4 weeks apart. Statistical analysis with the κ coefficient was used to evaluate reliability.

Results: The first reading by the most senior observer on the basis of CT scans was used as the gold standard (distribution: A1, 18; A2, 12; B1, 20; B2, 25; B3, 22; C, 1; and D, 2). The average intraobserver agreement for radiographs and CT scans was 0.73 (substantial; 0.72, 0.74, and 0.72) and 0.73 (substantial; 0.77, 0.69, and 0.72), respectively. The average interobserver agreement was 0.55 (moderate; 0.61, 0.51, and 0.53) for radiographs and 0.52 (moderate; 0.63, 0.50, and 0.43) for CT scans.

Conclusion: Intraobserver agreement of the modified Walch classification was substantial both for axillary radiographs and for CT scans. Interobserver agreement was fair. Although the modified Walch classification represents an improvement over the original classification, automated computer-based analysis of CT scans may be needed to further improve the value of this classification.