Clinical Data
This retrospective study included forty-four consecutive patients who had undergone ultrasonography prior to revision arthroscopic surgery performed to treat postoperative pain in the shoulder between January 1996 and October 2000. Inclusion criteria included previous surgery on the shoulder, persistent pain despite appropriate conservative measures, the need for revision surgery, and ultrasonographic examination of the rotator cuff prior to the revision surgery. Exclusion criteria included inflammatory arthropathy or a previous arthroplasty. The time interval between the ultrasound study and the revision arthroscopic surgery averaged eight weeks (range, one to fifty-six weeks). There were sixteen women and twenty-eight men, and their ages ranged from thirty-one to seventy-five years (mean, fifty-three years). The right shoulder was operated on in twenty-nine patients and the left, in fifteen. Thirty-four patients had previously undergone rotator cuff repairs (
Table I ), nine of which had been done with nonabsorbable suture anchors. Ten patients had not had previous repair of the cuff; seven of them had previously undergone subacromial decompression, two had had débridement of a small partial-thickness rotator cuff tear, and one had undergone an instability (Bristow) procedure.
The final clinical diagnosis after the revision arthroscopy was a full-thickness rotator cuff tear in twenty-one patients and a large partial-thickness rotator cuff tear (>75% of the cuff thickness) in one. The remaining twenty-two patients had an intact rotator cuff and, of those patients, sixteen had a subacromial decompression alone at the time of the arthroscopy, two underwent a decompression and a concomitant distal clavicular resection, two had débridement of a small partial-thickness rotator cuff tear, and two underwent an arthroscopic capsular release for adhesive capsulitis.
For the purposes of this study, partial-thickness tears were classified according to the gross size of the tear as it related to surgical indications. Hence, patients with a small partial-thickness tear that did not require any treatment other than simple débridement were considered to have an intact rotator cuff, whereas those with a large partial-thickness tear that required a rotator cuff repair were considered to have a full-thickness tear. A partial-thickness tear was considered large if it was >75% of the total cuff thickness. The gross extent of the partial-thickness tear was estimated by viewing the amount of detachment of the cuff lateral to the lateral extent of the articular surface. Given that a footprint of the supraspinatus insertion approximates 5 to 10 mm, tears involving >5 mm of insertional loss lateral to the articular surface were considered to be large partial-thickness tears requiring excision and repair
4,5 .
Ultrasound Technique
All ultrasonograms were obtained in real time with use of an ATL HDI 3000 scanner (Advanced Technology Laboratories, Bothell, Washington) or a Siemens Elegra scanner (Siemens Medical Systems, Issaquah, Washington) and a variable high-frequency linear-array transducer (7.5 to 10 MHz). All patients had standardized bilateral ultrasonography of the shoulder performed by one of two radiologists who were very experienced with the technique, having conducted more than 2500 examinations during a ten-year period.
The ultrasonographic examination was performed with the patient seated on a stool and the radiologist standing behind the patient. First, the biceps tendon was examined in the transverse plane from the level of the acromion inferiorly to the point where the tendon merged with the biceps muscle.
Images of the supraspinatus tendon were made with the shoulder extended, the elbow flexed, and the hand placed on the iliac wing. This position was necessary in order to expose as much of the supraspinatus tendon as possible from under the acromion. The transducer was oriented parallel to the tendon (approximately 45° between the coronal and sagittal planes) in order to visualize the fibers in a longitudinal plane, and it was moved anteriorly to posteriorly in order to visualize the supraspinatus and infraspinatus tendons. The transducer was rotated 90° in order to examine the tendons in the transverse plane.
Ultrasonographic Criteria
A finding of a full-thickness rotator cuff tear was recorded when the rotator cuff could not be visualized because of complete avulsion and retraction under the acromion, when there was a focal defect in the rotator cuff, or when the torn cuff was retracted a variable degree from the surgical trough. A finding of a partial-thickness tear was recorded when a distinct hypoechoic or mixed hyperechoic and hypoechoic defect was visualized in both the longitudinal and the transverse plane at the deep articular side of the rotator cuff (an articular-side partial-thickness tear). A thinned cuff or one with a subtle concave contour was considered to be intact in the absence of a focal defect.
The extent of the rotator cuff tear was determined with transverse measurements. According to empirical guidelines instituted prior to the inception of this study, if the tear extended posteriorly =1.5 cm from the intra-articular portion of the biceps tendon it was recorded as involving only the supraspinatus tendon, whereas if it extended >1.5 to 3.0 cm it was recorded as involving both the supraspinatus and the infraspinatus tendon. The teres minor tendon was not evaluated when the extent of the tear was determined.
Surgical Technique and Criteria
All arthroscopic examinations and operative procedures were performed by one surgeon who recorded all findings in a standardized manner. The surgeon was not blinded to the ultrasound results. The presence or absence of a full-thickness or partial-thickness (bursal or articular-side) rotator cuff tear was recorded. Representative arthroscopic images were made of all tears and other pathological findings, such as abnormalities of the biceps tendon. Data on the status of the rotator cuff tendons were derived from standardized operative reports based on operative observations.
When arthroscopic examination revealed a partial-thickness tear, a tagging suture was placed through the defect from the bursal side. This suture was then used as a guide during arthroscopic bursal imaging to inspect the corresponding superficial side of the rotator cuff.
Data Analysis
The ultrasonographic and arthroscopic findings with regard to the integrity of the rotator cuff were correlated. The subscapularis was classified only as torn or intact as demonstrated by the ultrasound or arthroscopy. When there was a disagreement between the interpretations of the ultrasonographic and arthroscopic results, representative images were reevaluated jointly by the radiologist and surgeon to explain the discrepancy; however, the previously reported results were used for statistical analysis. Reported results were not changed as a result of the discussion between the radiologist and surgeon.
Detection of Rotator Cuff Tears
True-Positive Findings
Ultrasound correctly identified twenty of the twenty-two rotator cuff tears (twenty-one full-thickness tears and one large partial-thickness tear) that were confirmed at the revision arthroscopic surgery (
Figs. 1-A and
1-B ). Ultrasound correctly identified all three partial-thickness rotator cuff tears, which included one large partial-thickness tear and two small partial-thickness tears. The large partial-thickness tear was repaired, and thus managed as a full-thickness tear, and the two small partial-thickness tears were débrided and thus considered to be intact rotator cuffs. Overall, the sensitivity of the ultrasound for detection of rotator cuff tears was 91%, the specificity was 86%, and the accuracy was 89%.
False-Negative Findings
There were two false-negative studies, both of small (less than 1 × 1 cm
2 ) full-thickness rotator cuff tears that had previously been treated with a mini-open, deltoid-splitting, rotator cuff repair (one with a nonabsorbable suture anchor). One of these tears was in a morbidly obese patient with a markedly limited range of shoulder motion, and the other was in a shoulder in which sutures were visualized at the rotator cuff insertion.
True-Negative Findings
Ultrasound correctly identified twenty of twenty-two intact rotator cuffs confirmed at the revision surgery (
Figs. 2-A and
2-B ).
False-Positive Findings
There were two false-positive studies. Both were of patients who had previously undergone a mini-open, deltoid-splitting, rotator cuff repair. One of the patients had a markedly limited range of motion due to adhesive capsulitis, and the other had an overreduced rotator cuff repair.
Overall Accuracy
If the index procedure involved treatment of a shoulder with an intact rotator cuff, the sensitivity, specificity, and accuracy of the postoperative ultrasound were all 100%. If the index procedure included a rotator cuff repair, the sensitivity, specificity, and accuracy of the postoperative ultrasound were 90%, 79%, and 85%, respectively.
The accuracy with which a treating physician can determine the integrity of the rotator cuff can have important implications for the management of a shoulder that is painful postoperatively. Authors of numerous clinical studies have reported variable success with nonsurgical management of painful rotator cuff tears
6-12 . In the setting of an intact rotator cuff, the surgeon and patient can be more confident of achieving success with nonoperative treatment, without the long-term risks of rotator cuff atrophy, progression of the rotator cuff tear, or development of rotator cuff-related degenerative changes.
Magnetic resonance imaging provides an anatomic picture, demonstrates the quality of rotator cuff muscles, and shows other intra-articular and extra-articular pathology. However, the diagnostic role of magnetic resonance imaging of a shoulder that has undergone surgical treatment is controversial, with reported accuracy rates ranging from 70% to 90%
13-16 . Additionally, in patients who have undergone shoulder surgery, sutures, suture anchors, and/or osseous changes may alter signal intensities within the acromion, humeral head, and rotator cuff tissue, and these altered signals may be indistinguishable from those of an acute rotator cuff tear
13-16 . This is of particular concern given the widespread use of suture anchors for rotator cuff repairs.
Prior to the introduction of magnetic resonance imaging, arthrography was the imaging method used mostfrequently to evaluate the integrity of the rotator cuff
17 . Disadvantages of this technique included its invasive nature and its inability to localize or quantitate a rotator cuff tear. In addition, arthrography yielded a substantial number of false-positive results following rotator cuff repair because of a non-watertight closure
18 .
There have been very few studies on the use of ultrasound to evaluate shoulders postoperatively
3 . Ultrasound may be especially helpful in this setting as the image should not be affected by the presence of intraosseous hardware. The reported sensitivities and specificities of ultrasonography before and after surgery on the shoulder have ranged from 50% to 100%
2,3,19,20 . However, investigators who reported that ultrasound had poor accuracy for the diagnosis of rotator cuff tears used ultrasound criteria that are no longer accepted, a scanning technique that has since been modified to improve visualization of the rotator cuff, and/or equipment that is now outdated. Recently, Teefey et al.
2 reported on 100 consecutive patients who had undergone shoulder ultrasound prior to arthroscopy. On the basis of modern ultrasound criteria, the imaging had 100% sensitivity, 85% specificity, and 96% accuracy for detecting full-thickness rotator cuff tears.
We also found ultrasound to be highly accurate for evaluating the integrity of the rotator cuff after operations on the shoulder. It is inexpensive, noninvasive, and well tolerated, and it can often be performed in a timely fashion with immediate results. Importantly, the image is not degraded by suture anchors. However, its usefulness is limited in patients with a severely restricted range of motion because the entire rotator cuff cannot be adequately visualized. Its accuracy in detecting small partial-thickness rotator cuff tears may also be limited.
One limitation of our study was its retrospective design. This limitation was reduced in part by the standardized documentation of the operative findings and the photographic record of the rotator cuff. Another limitation was the potential bias toward improved accuracy because the surgeon who performed the arthroscopy had knowledge of the findings of the preoperative physical examination and the ultrasound results; however, as the study was performed retrospectively, there was not necessarily a bias by the surgeon to verify the results of the ultrasound at the time of the surgery.
One other important consideration was our statistical treatment of partial-thickness rotator cuff tears. Given previous reports
2,13-16 of the relative inaccuracy of both magnetic resonance imaging and ultrasound in the detection of partial-thickness tears in shoulders without previous operative treatment, we thought that the emphasis in our study should be on full-thickness rotator cuff defects. Partial-thickness tears thus were classified on the basis of their clinical relevance. We thought that this would add more clinical practicality to our conclusions. For this reason, rotator cuffs with a small partial-thickness tear treated only with débridement were considered to be intact, and large partial-thickness tears that were thought clinically to require repair were considered to be the same as full-thickness tears.
In conclusion, our findings suggest that ultrasonography can be highly accurate in the evaluation of the rotator cuff in shoulders that have been treated operatively. The results are comparable with those reported for shoulders without prior operative treatment. The lack of image degradation in the presence of commonly used suture anchors suggests that this modality may compare favorably with magnetic resonance imaging for the evaluation of shoulders that have been treated operatively.