Interobserver variation in reporting CT arthrograms of the shoulder

Computed tomography (CT) arthrography of the shoulder is an imaging modality of great diagnostic accuracy with regard to glenohumeral instability and in particular labral lesions. Interpretation of the scans is made difficult by the frequent occurrence of normal anatomic variants and the complexity of injuries to the bone and soft tissues. We selected a continuous sample of 50 CT arthrograms of the shoulder and they were reported by two consultant musculoskeletal radiologists. The results were collated and analysed for the level of agreement. Hill-Sachs showed Kappa (K) statistic to be 0.37 (fair agreement), soft tissue Bankart 0.32 (fair agreement), bony Bankart 0.61 (substantial agreement), anterior capsular laxity 0.41 (moderate agreement) and glenohumeral osteoarthritis 0.20 (slight agreement). All the results were significant with a p value of <0.05. Nine (18%) of the 50 scans were in complete agreement. The results demonstrate that there can be considerable interobserver variation (IOV) in the reports of a CT arthrogram of a shoulder. They highlight the potential difficulties in reporting such images and suggests ways in which the report could be more focussed to provide a clinically reliable report and one which matches the surgical findings accurately.