Limited sensitivity of ultrasound for the detection of rotator cuff tears

Bilateral rotator cuff sonography was performed on 56 patients referred for shoulder arthrography to detect rotator cuff tears. Since one patient had bilateral arthrography there were 57 arthrograms. Twenty-seven shoulders had rotator cuff tears shown on arthrography, 15 of which were detected by sonography. Eleven were false negative by sonography, and one was indeterminate. Of the 30 shoulders with negative arthrograms, 28 had normal sonograms and two were false positive. Sensitivity was 58%, specificity was 93% and overall accuracy was 77%. The positive and negative predictive values were 88% and 72%. These results suggest that sonography has limited value in screening patients for rotator cuff tears and that a positive sonographic reading is more reliable than a negative one. Theories relating to the variable anatomic configuration of a tear are postulated to explain these results, which differ from those previously reported.