Detection and Measurement of Rotator Cuff Tears with Sonography: Analysis of Diagnostic Errors

Transcription

1 Musculoskeletal Imaging Teefey et al. Detection of Rotator Cuff Tears with Sonography Sharlene A. Teefey 1 William D. Middleton 1 William T. Payne 2 Ken Yamaguchi 3 Teefey SA, Middleton WD, Payne WT, Yamaguchi K Received July 30, 2004; accepted after revision September 15, Supported by a grant from Society of Radiologists in Ultrasound. 1 Mallinckrodt Institute of Radiology, Washington University School of Medicine, 510 S Kingshighway Blvd., St. Louis, MO Address correspondence to S. A. Teefey (teefeys@mir.wustl.edu). 2 Department of Orthopaedics and Rehabilitation, Southern Illinois University School of Medicine, Springfield, IL Department of Orthopaedic Surgery, One Barnes-Jewish Hospital Plaza, St. Louis, MO AJR 2005;184: X/05/ American Roentgen Ray Society Detection and Measurement of Rotator Cuff Tears with Sonography: Analysis of Diagnostic Errors OBJECTIVE. The purpose of this study was to analyze the causes of errors in the detection and measurement of rotator cuff tears in our patient population. SUBJECTS AND METHODS. Seventy-one consecutive patients with shoulder pain who were prospectively studied with sonography had subsequent arthroscopy that showed a full-thickness or partial-thickness tear or intact cuff. For sonography and arthroscopy, the length or degree of retraction and width of a tear, when present, was recorded. When there were discrepant findings, representative images were jointly evaluated by the radiologist and orthopedic surgeon to determine the cause of the error. RESULTS. Fifteen detection errors were found, including five misses (three < 5-mm subscapularis and two small partial-thickness tears), four errors inherent with the test (distinguishing large bursal side or extensive partial-thickness from full-thickness tears and tendinopathy from partial-thickness tears), three errors of an unknown cause, two due to misinterpretation, and one error inherent with the patient. Seventeen measurement errors occurred with full-thickness tears, 15 of those in patients with large or massive tears. Bursal thickening (n = 4), nonvisualization of the torn tendon end (n = 2), nonretracted tear (n = 2), and complex tear (n = 1) contributed to the errors. Eight measurement errors occurred with partial-thickness tears. Difficulty distinguishing tendinopathy from partial-thickness tears (n = 3) and complex tears (n = 3) accounted for six errors. CONCLUSION. Although infrequent, detection errors were due to limitations inherent with the test or misses. Limitations inherent with the patient and misinterpretation of the findings were rare. Most measurement errors occurred in patients with large or massive cuff tears. otator cuff disease is the most R common cause of shoulder pain. In addition to history and physical examination, evaluation of a patient with shoulder pain often involves assessment of the cuff with a diagnostic test such as sonography or MRI. A review of the sonography literature has shown sensitivities ranging from % and specificities from % for the detection of full-thickness rotator cuff tears in series that varied in size from 38 to 71 patients and used surgery as a gold standard [1 8]. These wide ranges can be explained in part by the use of sonographic criteria that have since been refined or changed, use of older equipment and lower frequency transducers, or use of a scan technique that has since been modified to improve cuff visualization [4, 7]. Only a few series of 15 to 27 patients have reported sensitivities and specificities for detecting partial-thickness tears, with ranges from 41 93% and 85 94%, respectively [1, 2, 9]. One of these latter studies may have experienced some of the limitations described above [2]; however, none of the studies explained the causes for their interpretation errors nor have reasons for errors been reported, to our knowledge. The purpose of this study was to analyze the causes of the errors for the detection and measurement of rotator cuff tears in our patient population. Subjects and Methods Patients The study comprised 71 consecutive patients with acute or chronic shoulder pain who had undergone preoperative sonography and subsequent arthroscopy between December 1998 and April Forty-one men and 30 women participated, with a mean age of 59 (range, years old). Our institutional review board approved the study protocol and informed consent was obtained from all patients. The arthroscopic or surgical diagnoses were a full-thickness tear of the rotator cuff in 46 (65%) 1768 AJR:184, June 2005

2 Detection of Rotator Cuff Tears with Sonography patients, a partial-thickness tear in 19 (27%), and no tear in six (8%). Of the six patients with no tear, chronic tendinitis was diagnosed in four and adhesive capsulitis in one; one patient had a normal cuff. Eight patients (11%) had subscapularis tendon tears, seven of whom had associated full-thickness tears involving at least the supraspinatus tendon. Imaging and Surgical Techniques All sonograms were obtained with an Elegra scanner (Siemens Medical Solutions) using a highfrequency, MHz linear array transducer. Real-time scanning of the shoulder was performed in a standardized fashion as described by Teefey et al. [1]. In nearly all cases, harmonic imaging was performed with a transmit frequency of 4.5 MHz. A single subspecialty-trained shoulder surgeon performed all arthroscopic examinations and surgeries. Representative arthroscopic images and videotapes were made of all tears. Imaging and Surgical Interpretation One of two radiologists similarly experienced with musculoskeletal sonography prospectively obtained and interpreted each sonogram. Established criteria were used for the diagnosis of a fullor partial-thickness tear [1, 3, 9]. For each sonogram, the absence or presence of a full- or partialthickness rotator cuff tear was recorded. The length of a partial-thickness tear or degree of retraction of a full-thickness tear (measured on longitudinal views oriented parallel to the long axis of the cuff) and width (measured on transverse views oriented perpendicular to the long axis of the cuff) were also recorded. When the width was greater than 30 mm, an attempt was often made to measure it with extended field-of-view imaging. The presence or absence of a subscapularis tendon tear was recorded separately without measurements. Each radiologist also recorded whether the sonogram was diagnostic, suboptimal but diagnostic, or nondiagnostic. At arthroscopy, the orthopedic surgeon, who was not blinded to the sonographic findings, recorded the absence or presence of a full- or partialthickness rotator cuff tear. The degree of retraction or length and the width of each tear were measured with a linear tool notched with 1-mm marks. The presence or absence of a subscapularis tendon tear was also recorded. Representative arthroscopic images were taken of all tears. Data Analysis Diagnoses and measurements made with sonography were compared with the arthroscopic findings. Findings were compared for the presence or absence of a full- or partial-thickness tear, tear length or retraction, and tear width. Sonographic measurements within 5 mm of the measurement recorded at arthroscopy were considered correct. Tears 30 mm or greater in width and those retracted beneath the acromion were assigned a value of 30 mm. When discrepant findings between sonography and arthroscopy were found, sonographic and representative arthroscopic images were retrospectively reviewed jointly by one radiologist and the orthopedic surgeon to determine the cause of the discrepancy. In one discrepant case, arthroscopic images were not available for review. Afterward, the causes for the detection errors (including subscapularis tears) were categorized as follows: inherent with the test (i.e., the study would have been interpreted the same way in retrospect based on the standard sonographic criteria for diagnosing a tear), inherent with the patient (i.e., an inability to perform the test because of patient limitations), misinterpretation of a finding (the abnormality was identified but misinterpreted), a miss (no abnormality detected), and an unknown cause. The causes for measurement errors were descriptive. Results Detection of Rotator Cuff Tears Of the 46 full-thickness rotator cuff tears, there was only one diagnostic error and it was due to misinterpretation of a finding. In this case, an extensive partial-thickness tear rather than a full-thickness tear was diagnosed. A review of the images showed a contour deformity of the cuff. All 46 sonograms were considered diagnostic. Of the 19 partial-thickness rotator cuff tears, six diagnostic errors occurred. In two cases, the finding was missed. In these two cases, a review of the representative sonographic images showed a normal-appearing cuff (Fig. 1). The articular-side partial-thickness tears measured 5 15 mm and 15 5 mm (length width) on arthroscopy. In the other four cases, full-thickness tears were diagnosed. In two of these cases, large bursal-side, partial-thickness tears were found on arthroscopy involving most of the cuff (in one case, more than 90% of the cuff fibers were torn) (Fig. 2) and in another case, a markedly thinned cuff with an extensive partial-thickness tear was found (Fig. 3). These three errors were categorized as inherent with the test. In the fourth case, the cause for the error was unknown. A review of the representative arthroscopic images showed no full-thickness tear; however, the sonographic images showed a small contour defor- A B Fig year-old woman with an articular-side, partial-thickness tear. A, Longitudinal sonographic image shows a normal, intact cuff. B, Transverse sonographic image shows a normal, intact cuff. The small tear was missed. It measured 5 (length) 15 (width) mm at surgery. BT = biceps tendon. AJR:184, June

3 Teefey et al. Fig year-old man with an extensive bursal-side tear involving greater than 90% of the cuff. Longitudinal image shows a 1-cm tear (cursors) surrounded by fluid with a few intact articular-side fibers that was interpreted as full-thickness tear on the sonogram. Tear was detected but misinterpreted due to inherent limitations of the test. mity of the cuff. All 19 sonograms were considered diagnostic. Of the eight subscapularis tendon tears, four diagnostic errors were found. A review of the sonographic images in three cases showed no tear. All three of these errors were categorized as misses. In all three cases, the subscapularis tears were < 5-mm partialthickness tears of the superior aspect of the tendon as noted on arthroscopy. Sonographic visualization of the subscapularis tendon was hindered by limited external rotation in the fourth case and was the only case considered nondiagnostic. This error was categorized as inherent with the patient. Two subscapularis tendon tears were diagnosed with sonography in patients with intact subscapularis tendons on arthroscopy (Fig. 4). In one of these cases, a tear involving the supraspinatus and infraspinatus tendons was noted to extend over the rotator interval on arthroscopy although it did not involve the subscapularis tendon. The arthroscopic images were not available for review in the second case. Review of the sonographic images showed a small defect in the subscapularis tendon. The first diagnostic error was categorized as a misinterpretation of the findings and the second as from an unknown cause. Of the six intact cuffs, two diagnostic errors were found. A small full-thickness tear was diagnosed in one case that was found to have adhesive capsulitis without a tear on arthroscopy. Review of the sonographic images showed a small contour deformity of the cuff (Fig. 5) but no tear was identified on review of the arthroscopic images. The cause for this error was categorized as unknown. In the second case, a tiny partial-thickness tear was reported; chronic tendinitis was found on arthroscopy. Review of the sonographic images showed a tiny hypoechoic region at the articular side of the cuff. This error was categorized as inherent with the test. All six sonographic studies were considered diagnostic. Size and Extent of the Tears Of the 45 full-thickness rotator cuff tears correctly diagnosed on sonography, 11 errors were in tear retraction and six were errors in tear-width measurements. Retraction was overestimated in eight cases and underestimated in three. In all but one case, large or massive tears (20 45 mm in width) were found on arthroscopy. In two of the three cases in which retraction was underestimated, the medial torn ends were retracted beneath the acromion. In three of the cases in which tear retraction was overestimated and in one case in which it was underestimated, a review of the sonograms showed bursal thickening. In another case, review of the arthroscopic images showed a complex, irregularly shaped tear. In the remaining four cases, a cause for the errors could not be determined. Tear width was underestimated in three cases and overestimated in three. In five of these cases, tear widths measured mm on arthroscopy. In two of the three cases in which tear width Fig year-old woman with a markedly thinned, extensive partial-thickness tear involving almost the entire rotator cuff. Longitudinal image shows flattening of the cuff and loss of cuff substance (cursors) that was interpreted as a full-thickness tear on the sonogram. The tear was detected but misinterpreted due to the inherent limitations of the test AJR:184, June 2005

4 Detection of Rotator Cuff Tears with Sonography Fig year-old woman with a full-thickness rotator cuff tear involving the supraspinatus and infraspinatus tendons extending over the rotator interval but not involving the subscapularis tendon. The longitudinal image of the subscapularis tendon shows defect (cursors) that was misinterpreted as a full-thickness tear on the sonogram. was underestimated, portions of the tears were nonretracted. A review of the arthroscopic and sonographic images showed no apparent reasons for the remaining four errors. Of the 13 partial-thickness tears that were correctly diagnosed with sonography, two errors had occurred in tear length and six in tearwidth measurements. In one of the two cases in which tear-length measurements were underestimated, a review of the arthroscopic images showed a complex tear with multiple, separate components and in the other case, a review of the sonograms showed a hypoechoic region that was larger than what was measured. In two of the six cases in which tear widths were underestimated, a review of the arthroscopic showed irregularly shaped tears. In two other cases, a review of the sonograms showed apparent hypoechoic regions on transverse images that were larger than what was measured (Fig. 6). In the remaining two cases, a cause for the measurement errors could not be determined. Discussion Although infrequent, we found a wide variety of causes for errors when diagnosing and measuring rotator cuff tears (Table 1). All of our misses occurred in patients with small partialthickness tears; three were in patients with < 5- mm partial-thickness subscapularis tears. Given the size and partial-thickness nature of these tears, it is not surprising that they were missed. Although the clinical significance of these tiny partial-thickness subscapularis tears is unknown, full-thickness subscapularis tears, which are clinically important, can accurately be diagnosed with sonography [10]. The only other misses occurred in patients with small articular-side partial-thickness tears. It has been reported that sonography is less accurate for diagnosing partial-thickness tears compared with full-thickness tears [8]. This may be due to the difficulty in distinguishing tendinopathy from some partial-thickness tears. Although little has been written on the sonographic appearance of tendinopathy apart from brief comments in musculoskeletal sonography textbooks [11], which have described a focal hypoechoic region, our experience suggests that it can be difficult to differentiate these two entities in some cases. This distinction, however, may not be clinically relevant because both are initially managed nonoperatively. We found that one third of our diagnostic errors were inherent with the test or patient and beyond the control of the radiologist. One of these errors occurred because of difficulty distinguishing an extensive partial-thickness tear from a full-thickness tear on sonography (Fig. 3). The error occurred because of the substantial loss of cuff substance and compressibility of the few remaining fibers with the transducer. One study reported that an extensive partial-thickness tear involving greater than 50% of the cuff substance can mimic a full-thickness tear by virtue of its compressibility [1]. Large partialthickness bursal-side tears can also be mistaken for full-thickness tears (Fig. 2). A large bursalside cuff tear, by virtue of its size and location, Fig year-old man with adhesive capsulitis and an intact cuff. The longitudinal image shows contour deformity (cursors) of the supraspinatus tendon that was interpreted as a full-thickness tear on the sonogram. The reason for the error could not be determined. AJR:184, June

5 Teefey et al. Fig year-old woman with an articular-side, partial-thickness supraspinatus tear. Transverse image shows the original measurement of the tear width (cursors), which was underestimated. The more likely tear width (arrows) correlates with the arthroscopic measurement of 15 mm. BT = biceps tendon. will appear on sonography as a contour deformity and simulate a full-thickness tear, as in two of our cases. This difficulty in distinguishing partial- and full-thickness tears has recently been shown as the primary cause of interobserver variability between two experienced observers [12]. Clinically, a failure to distinguish extensive partial-thickness from full-thickness tears may not be important because orthopedic surgeons generally treat extensive partial-thickness tears or large bursal-side tears as if they were full-thickness ones. Difficulty distinguishing tendinopathy from an articular-side partial-thickness tear was another source of error inherent with the test but occurred only once. As stated earlier, differentiating these two entities can be problematic. Only one error was inherent with the patient; this error occurred when sonographic visualization of the subscapularis tendon was hindered by a limited range of motion, as the patient was unable to externally rotate his arm to evaluate the tendon. This was the only case in our study in which such a limitation occurred, indicating that it is rare. We were unable to determine the causes of the diagnostic errors in three cases. One occurred in a patient with a partial-thickness tear; MRI and sonography showed a contour deformity in the cuff, suggestive of a fullthickness tear. The second error occurred in a patient thought to have a partial-thickness subscapularis tear; unfortunately, arthroscopic images were not available for review. In the third case, a patient with adhesive capsulitis was diagnosed with a small full-thickness tear, also reported in an MRI study (Fig. 5). Although it is difficult to explain these errors, it is possible that small tears may have been missed on arthroscopy. Sonographic findings were rarely misinterpreted. In one case, a contour deformity was TABLE 1 present on sonography, but an extensive partial-thickness tear was diagnosed rather than a full-thickness tear due to the perceived presence of intact bursal-side fibers during realtime scanning. One study reported that extensive partial-thickness tears can produce a cuff defect [1]. Nevertheless, had the cuff defect, an important criterion for the diagnosis of a fullthickness tear, been interpreted as such, the error would not have occurred. The only other error occurred in a patient thought to have a subscapularis tear (Fig. 4). The arthroscopic images showed a supraspinatus tear that extended over the rotator interval but did not involve the tendon. This subtle distinction was almost certainly the cause for the error. MRI also diagnosed a subscapularis tear in this case. We found that most of the retraction-measurement errors in patients with full-thickness tears occurred with large or massive tears. Although it was difficult to determine the exact cause of the discrepant measurements in every case, we speculate that several factors may have contributed. First, because large and massive tears can be irregularly shaped, it is doubtful that both sets of measurements were obtained at the same location in these cases. In addition, when bursal thickening was present, it was more difficult to determine the precise location of the torn tendon ends. Shoulders were also positioned differently during sonography and surgery. During sonography, the shoulder was extended and the elbow flexed and directed medially; this may have increased the distance between the torn tendon ends and may have caused local muscle groups to contract, resulting in further medial retraction. At surgery, the humeral head was reduced to an anatomic position, allowing the torn ends of the cuff to return to their relaxed positions. An inability to visualize the medial torn tendon end beneath the acromion was another cause for discrepant measurements. In such cases, an accurate measurement could not be obtained. Nevertheless, reporting the location of the torn tendon end as retracted beneath the acromion provided useful information to the orthopedic surgeon. Although far fewer tear-width measurement errors for full-thickness tears occurred, the tears were again large or massive in the majority of cases. We found it difficult to measure large or massive tear widths along a curved surface. In some cases, we used extended field-of-view imaging to better show the width of a large or massive tear and overcome the usual limitations of a small field of view. Nevertheless, an accurate Detection of Rotator Cuff Tears with Sonography Diagnostic Errors Sonographic Findings Inherent with Total Misinterpretation Miss Unknown Test Patient Full-thickness tear Partial-thickness tear Subscapularis tear Intact cuff Total AJR:184, June 2005

6 Detection of Rotator Cuff Tears with Sonography measurement could not be obtained because the resultant image was often foreshortened. In those cases in which tear widths were underestimated, arthroscopic images showed that a portion of the tears were nonretracted; these portions were most likely not included in the measurements accounting for the errors. Few errors occurred when measuring partial-thickness tear lengths; however, tear widths were underestimated on sonography in nearly one half of the cases. The irregular shape of the cuff tear was thought to be responsible for the incorrect sonographic measurements in two cases due to difficulty measuring such tears at the same location as arthroscopy. In two other cases, review of the sonographic images showed hypoechoic regions on transverse views that were larger than what was measured (Fig. 6). Difficulty distinguishing tendinopathy from the tear most likely accounted for these errors. All but one of the sonographic studies was considered diagnostic. Body habitus was not a limiting factor in our study, nor was it a cause of subdiagnostic or nondiagnostic studies. The one nondiagnostic study occurred in a patient with limited external rotation of his arm and only hindered evaluation of the subscapularis tendon. Our study was limited by the fact that the joint review was done months after the arthroscopic and sonographic examinations were completed rather than immediately after surgery. Although the sonographic and arthroscopic studies were performed and measurements obtained prospectively, we were nevertheless unable to determine the exact cause of some of the measurement errors and can only speculate as to the reasons. Had the joint review been done earlier, we may have been able to determine the exact causes of the measurement errors in more cases. Although all of these patients had MRI as part of another study [8], we did not determine if the diagnostic errors made with sonography were the same errors made with MRI. However, we did note that many of the cases that were problematic for sonography were also problematic for MRI [8]. This primarily reflects the difficulty that both tests have in distinguishing extensive partialthickness or large bursal-side tears from fullthickness tears and tendinopathy from partialthickness tears. Given this overlap in diagnostic capabilities, the choice as to which test to order depends on several factors. Based solely on diagnostic accuracy, there is ample evidence to support the use of either test in a patient with a suspected rotator cuff tear [8, 13]. Thus, when determining which test to order, other factors besides accuracy should be considered, including regional expertise with a particular test, the importance of ancillary clinical information (labral, capsular, ligamentous, or bone pathology), patient tolerance, and cost. We found that the patients perception of the test was also an important consideration; in our experience, sonography is preferred to MRI by most patients [14]. In summary, we have shown that detection errors, although infrequent, will occur when obtaining a shoulder sonogram, primarily due to limitations inherent with the test or misses in patients with small partial-thickness tears. From a clinical perspective, these errors should not have a significant impact on patient care. Most measurement errors for full-thickness tears occurred in patients with large or massive tears, errors that may not be rectifiable. References 1. Teefey SA, Hasan SA, Middleton WD, Patel M, Wright RW, Yamaguchi K. Ultrasonography of the rotator cuff: a comparison of ultrasonographic and arthroscopic findings in one hundred consecutive cases. J Bone Joint Surg Am 2000;82: Brenneke SL, Morgan CJ. Evaluation of ultrasonography as a diagnostic technique in the assessment of rotator cuff tendon tears. Am J Sports Med 1992;20: Wiener SN, Seitz WH Jr. Sonography of the shoulder in patients with tears of the rotator cuff: accuracy and value for selecting surgical options. AJR 1993;160: Paavolainen P, Ahovuo J. Ultrasonography and arthrography in the diagnosis of tears of the rotator cuff. J Bone Joint Surg Am 1994;17: Hodler J, Fretz CJ, Terrier F, Gerber C. Rotator cuff tears: correlation of sonographic and surgical findings. Radiology 1988;169: Mack LA, Matsen FA III, Kilcoyne RF, Davies PK, Sickler ME. US evaluation of the rotator cuff. Radiology 1985;157: Brandt TD, Cardone BW, Grant TH, Post M, Weiss CA. Rotator cuff sonography: a reassessment. Radiology 1989;173: Teefey S, Rubin D, Middleton W, Hildebolt C, Leibold R, Yamaguchi K. Detection and quantification of rotator cuff tears: a comparison of ultrasonographic, magnetic resonance imaging and arthroscopic findings in seventy-one consecutive cases. J Bone Joint Surg Am 2004;86: van Holsbeeck MT, Kolowich PA, Eyler WR, et al. US depiction of partial-thickness tear of the rotator cuff. Radiology 1995;197: Farin P, Jaroma H. Sonographic detection of tears of the anterior portion of the rotator cuff (subscapularis tendon tears). J Ultrasound Med 1996;16: Ptaszink R. Sonography of the shoulder. In: van Holsbeeck MT, Introcaso JH, eds. Musculoskeletal ultrasound, 2nd ed. St. Louis, MO: Mosby, 2001: Middleton WD, Teefey SA, Yamaguchi K. Sonography of the rotator cuff: analysis of interobserver variability. AJR 2004;183: Dinnes J, Loveman E, McIntyre L, Waugh N. The effectiveness of diagnostic tests for the assessment of shoulder pain due to soft tissue disorders: a systematic review. Health Technol Assess 2003;7: Middleton WD, Payne WT, Teefey SA, Hildebolt CF, Yamaguchi K. Sonography and MRI of the shoulder: comparison of patient satisfaction. AJR 2004;183: AJR:184, June