Magnetic resonance imaging identification of rotator cuff retears after repair: Interobserver and intraobserver agreement

Background: Magnetic resonance imaging (MRI) is the most commonly used imaging modality to assess the rotator cuff. Currently, there are a limited number of studies assessing the interobserver and intraobserver reliability of MRI after rotator cuff repair. Hypothesis: Fellowship-trained orthopaedic shoulder surgeons will have good inter- and intraobserver agreement with regard to features of the repaired rotator cuff (repair integrity, fat content, muscle volume, number of tendons involved, tear size, and retract) on MRI. Study Design: Cohort study (diagnosis); Level of evidence, 3. Methods: Seven fellowship-trained orthopaedic shoulder surgeons reviewed 31 MRI scans from 31 shoulders from patients who had previous rotator cuff repair. The scans were evaluated for the following characteristics: rotator cuff repair status (fullthickness retear vs intact repair), tear location, tendon thickness, fatty infiltration, atrophy, number of tendons involved in retear, tendon retraction, status of the long head of the biceps tendon, and bone marrow edema in the humeral head. Surgeons were asked to review images at 2 separate time points approximately 9 months apart and complete an evaluation form for each scan at each time point. Multirater kappa (k) statistics were used to assess inter- and intraobserver reliability. Results: The interobserver agreement was highest (80%, k = 0.60) for identifying full-thickness retears, tendon retear retraction (64%, k = 0.45), and cysts in the greater tuberosity (72%, k = 0.43). All other variables were found to have fair to poor agreement. The worst interobserver agreement was associated with identifying rotator cuff footprint coverage (47%, k = 0.21) and tendon signal intensity (29%, k = 0.01). The mean intraobserver reproducibility was also highest (77%-90%, k = 0.71) for full-thickness retears, quality of the supraspinatus (47%-83%, k = 0.52), tears of the long head of the biceps tendon (58%-94%, k = 0.49), presence of bone marrow edema in the humeral head (63%-87%, k = 0.48), cysts in the greater tuberosity (70%-83%, k = 0.47), signal in the long head of the biceps tendon (60%-80%, k = 0.43), and quality of the infraspinatus (37-90%, k = 0.43). The worst intraobserver reproducibility was found in identification of the location of bone marrow edema (22%-83%, k = 0.03). Conclusion: The results of this study indicate that there is substantial variability when evaluating MRI scans after rotator cuff repair. Intact rotator cuff repairs or full-thickness retears can be identified with moderate reliability. These findings indicate that additional imaging modalities may be needed for accurate assessment of the repaired rotator cuff.