MR arthrography of traumatic anterior shoulder lesions showed modest reproducibility and accuracy when evaluated under clinical circumstances

We examined the reproducibility and accuracy of high-field MRA in traumatic anterior shoulder instability under conditions resembling clinical practice and assessed the influence of observer experience. Two radiologists with different experience levels evaluated 61 MRAs. Assessment was independent, blinded and non-sequential. For 40 MRAs, surgical reports were available to assess diagnosis accuracy and influence of observer experience. The assessed lesions were cuff lesions, Hill-Sachs lesions, bony and classic Bankart lesions, greater humeral tuberosity fractures, SLAP lesions and joint capsule lesions. Reproducibility was quantified using kappa coefficients. Accuracy was evaluated with sensitivity and specificity rates, positive and negative predictive values. Differences in the percentage of correctly diagnosed MRAs between the radiologists were tested using McNemar's test for paired proportions. Inter-observer k-values ranged from 0.03 for joint capsule lesions to 0.45 for humeral head lesions. The overall kappa was 0.21 (95% CI; 0.12-0.30). We also observed markedly lower sensitivity and specificity rates than those reported in the literature for most lesions. The more experienced radiologist correctly diagnosed 78.9% of all lesions compared to 65.4% for the less experienced radiologist (P <0.001; McNemar's test). MRA-interpretations of traumatic anterior shoulder instability should be regarded with caution in clinical practice. The experience level of radiologists can affect reproducibility and accuracy