Subscapularis tendon tears: detection and grading at MR arthrography.

PURPOSE To assess diagnostic accuracy in the detection and grading of subscapularis tendon lesions at magnetic resonance (MR) arthrography. MATERIALS AND METHODS MR arthrograms in 50 consecutive patients (29 with normal subscapularis tendons, 11 with a lesion in the cranial quarter, seven with a major tear but not complete detachment, three with complete detachment) with arthroscopic or surgical confirmation were evaluated independently by two radiologists. Diagnosis was established on findings from transverse and/or parasagittal images. RESULTS With transverse images alone, sensitivity was 95%/100% (reader 1/reader 2); specificity was 55%/62%. With parasagittal images alone, sensitivity was 91%/91%; specificity was 76%/90%. With combined images, sensitivity was 91%/91%; specificity was 86%/79%. Interobserver agreement was substantial (kappa = 0.67). Forty-one of 50 (82%) grades for subscapularis abnormalities matched at MR imaging and surgery; nine mismatches differed by only one degree. Several signs were specific (90%-100%) but insensitive (29%-62%); these included leakage of contrast material onto the lesser tuberosity, fatty degeneration of the subscapularis muscle, and abnormality in the course of the long biceps tendon (luxation, subluxation). CONCLUSION MR arthrography is accurate in the detection and grading of subscapularis tendon lesions. Specificity of findings on transverse images for this diagnosis can be improved by including indirect signs and findings on parasagittal images.