Background: Advances in shoulder magnetic resonance imaging (MRI) and arthrography (MRA) have revolutionised musculoskeletal diagnosis and surgical planning. Despite this, the overall accuracy of MRI, with or without intra-articular contrast, can be variable.
Methods: In this prospective non-randomised analysis, 200 participants (74.5% males) with suspected shoulder injuries underwent MRI (41.0%) or MRA followed by arthroscopy. A study specific proforma was developed to ensure consistency of reporting by radiologists and surgeons. The reports were compared to assess the predictive power of MRI/MRA. Specific assessment of rotator cuff tendon appearance, long head of biceps (LHB) tendon appearance, position and anchor, subacromial space, glenoid labrum and humeral cartilage grade were included.
Results: Shoulder MRA demonstrated a higher agreement with arthroscopy than MRI for supraspinatus, infraspinatus and subscapularis tendon appearance (κ = 0.77 vs. κ = 0.61, κ = 0.55 vs. κ = 0.53 and κ = 0.58 vs. κ = 0.46 respectively). There were also superior agreement rates with MRA compared to MRI for LHB tendon appearance (κ = 0.70 vs. κ =0.54) and position (κ = 0.89 vs. κ = 0.72). As an overall assessor of shoulder pathology we found significantly higher total agreement scores when MRA was used (p = 0.002).
Discussion: Whilst magnetic resonance imaging with arthrography is an extremely useful tool to assess underlying pathological shoulder states it does not confer 100% accuracy. In cases whereby this modality is inconclusive, an examination under anaesthesia and diagnostic arthroscopic assessment for the detection of intra-articular shoulder pathology may be considered.
Keywords: arthroscopy; magnetic resonance; sensitivity; shoulder; specificity.
© 2021 Royal Australasian College of Surgeons.