Inter-rater reliability of an arthritic glenoid morphology classification system

To our knowledge, no independent analysis of the inter-rater agreement of the widely used Walch classification for osteoarthritic glenoid morphology has been performed. The computed tomography scans of 24 shoulders with primary osteoarthritis were used by 4 experienced shoulder surgeons to classify the glenoids independently according to Walch et al. The weighted κ statistic was calculated to determine the inter-rater and intrarater agreement among observers. The overall inter-rater agreement for the Walch classification was fair (κ = 0.37) when classified into the 5 types (A1, A2, B1, B2, and C). Agreement for the various subclassifications was as follows: A1, κ = 0.22; A2, κ = 0.33; B1, κ = 0.17; B2, κ = 0.32; and C, κ = 0.86. When the classification system was simplified to just the 3 major types (A, B, and C), overall agreement was moderate (κ = 0.44). Agreement for each type was moderate for A (κ = 0.59) and B (κ = 0.59) and almost perfect for C (κ = 0.89). Overall intrarater agreement was fair (κ = 0.37). We conclude that only fair agreement was found among experienced shoulder surgeons when classifying arthritic shoulders using the classification system of Walch et al. A glenoid classification scheme that relies more upon glenoid morphology and less upon humeral head position may demonstrate greater observer agreement and, therefore, may offer greater value.