Magnetic resonance imaging reproducibility for rotator cuff partial tears in patients up to 60 years.

Background: Magnetic resonance imaging (MRI) is the gold standard in diagnosing rotator cuff pathology; however, there is a lack of studies investigating the reliability agreement for supraspinatus partial-thickness tears among orthopaedic surgeons and musculoskeletal (MSK) radiologists.
Methods: Sixty digital MRI scans (1.5 Tesla) were reviewed by two orthopaedic shoulder surgeons, two MSK radiologists, two fellowship-trained shoulder surgeons, and two fellowship-trained orthopaedic surgeons at two distinct times. Thirty-two scans of partial-thickness tears and twenty-eight scans of the supraspinatus tendon with no tears were included. Supraspinatus tendonosis and tears, long head of the biceps pathology, acromial morphology, acromioclavicular joint pathology and muscle fatty infiltration were assessed and interpreted according to the Goutallier system. After a four-week interval, the evaluators were asked to review the same scans in a different random order. The statistical analyses for the intra- and interobserver agreement results were calculated using the kappa value and 95% confidence intervals.
Results: The intraobserver agreement for supraspinatus tears was moderate among the MSK radiologists (k = 0.589; 95% CI, 0.446-0.732) and the orthopaedic shoulder surgeons (k = 0.509; 95% CI, 0.324-0.694) and was fair among the fellowship-trained shoulder surgeons (k = 0.27; 95% CI, 0.048-0.492) and the fellowship-trained orthopaedic surgeons (k = 0.372; 95% CI, 0.152-0.592). The overall intraobserver agreement was good (k = 0.627; 95% CI, 0.576-0.678). The intraobserver agreement was moderate for biceps tendonosis (k = 0.491), acromial morphology (k = 0.526), acromioclavicular joint arthrosis (k = 0.491) and muscle fatty infiltration (k = 0.505). The interobserver agreement results for supraspinatus tears were fair and poor among the evaluators: the MSK radiologists and the orthopaedic shoulder surgeons had the highest agreement (k = 0.245; 95% CI, 0.055-0.435).
Conclusions: In this sample of digital MRI scans, there was an overall good intraobserver agreement for supraspinatus partial tears; however, there were also poor and fair interobserver agreement results. The evaluators with higher levels of experience (the orthopaedic shoulder surgeons and the MSK radiologists) demonstrated better results than evaluators with lower levels of experience.