Superior labral anteroposterior tears: accuracy and interobserver reliability of multidetector CT arthrography for diagnosis.

Abstract

To evaluate the accuracy and interobserver reliability of multidetector computed tomographic (CT) arthrography for the diagnosis and classification of superior labral anteroposterior (SLAP) lesions. Institutional review board approval and informed consent were obtained. Retrospective review of images from 161 multidetector CT arthrographic examinations was performed by two radiologists independently for detection and classification of SLAP lesions (type I-X), and sensitivity, specificity, accuracy, and interobserver agreement were evaluated. The SLAP group included 94 patients, and the no-SLAP group included 67 patients with normal labrum. At arthroscopy, a total of 88 SLAP lesions (excluding type I) were found. For detection of SLAP lesions excluding SLAP type I lesions, sensitivity, specificity, and accuracy were 94.3%, 76.7%, and 86.3% for reader 1 and 97%, 72.6%, and 86.3% for reader 2, respectively, and the interobserver agreement was very good (κ = 0.87). The distribution of SLAP lesions was as follows: six type I, 58 type II, one type III, one type IV, 16 type V, one type VI, five type VII, three type VIII, one type IX, one type V and VI, and one type V and VII. Percentages of correct classification of SLAP lesions were variable according to the types, but the overall percentage was noted to be 69.2% for reader 1 and 68.1% for reader 2. The interobserver agreement of classification of SLAP lesions was good (κ = 0.72). Multidetector CT arthrography shows high accuracy and good interobserver reliability for diagnosis of SLAP lesions in spite of its limitations in specific classification.