Purpose: The purpose of this study was to determine inter- and intraobserver variability of MR arthrography of the shoulder in the detection and classification of superior labral anterior posterior (SLAP) lesions.
Methods: MR arthrograms of 78 patients who underwent MR arthrography before arthroscopy were retrospectively analysed by three blinded readers for the presence and type of SLAP lesions. MR arthrograms were reviewed twice by each reader with a time interval of 4 months between the two readings. Inter- and intraobserver agreement for detection and classification of SLAP lesions were calculated using kappa coefficients.
Results: Arthroscopy confirmed 48 SLAP lesions: type I (n = 4), type II (n = 37), type III (n = 3), type IV (n = 4). Sensitivity and specificity for detecting SLAP lesions with MR arthrography for each reader were 88.6%/93.3%, 90.9%/80.0% and 86.4%/76.7%. MR arthrographic and arthroscopic grading were concurrent for 72.7%, 68.2% and 70.5% of SLAP lesions for readers 1-3, respectively. Interobserver agreement was excellent (kappa = 0.82) for detection and substantial (kappa = 0.63) for classification of SLAP lesions. For each reader intraobserver agreement was excellent for detection (kappa = 0.93, kappa = 0.97, kappa = 0.97) and classification (kappa = 0.94, kappa = 0.84, kappa = 0.93) of SLAP lesions.
Conclusion: MR arthrography allows reliable and accurate detection of SLAP lesions. In addition, SLAP lesions can be diagnosed and classified with substantial to excellent inter- and intraobserver agreement.