Musculoskeletal Imaging Original Research

de Jesus et al. MRI, MR Arthrography, and Ultrasound of Rotator Cuff Tears Musculoskeletal Imaging Original Research

Accuracy of MRI, MR Arthrography, and Ultrasound in the Diagnosis of Rotator Cuff Tears: A Meta-Analysis
Joseph O. de Jesus1 Laurence Parker Andrea J. Frangos Levon N. Nazarian
de Jesus JO, Parker L, Frangos AJ, Nazarian LN

OBJECTIVE. The purpose of this study was to compare the diagnostic accuracy of MRI, MR arthrography, and ultrasound for the diagnosis of rotator cuff tears through a meta-analysis of the studies in the literature. MATERIALS AND METHODS. Articles reporting the sensitivities and specificities of MRI, MR arthrography, or ultrasound for the diagnosis of rotator cuff tears were identified. Surgical (open and arthroscopic) reference standard was an inclusion criterion. Summary statistics were generated using pooled data. Scatterplots of the data sets were plotted on a graph of sensitivity versus (1 specificity). Receiver operating characteristic (ROC) curves were generated. RESULTS. Sixty-five articles met the inclusion criteria for this meta-analysis. In diagnosing a full-thickness tear or a partial-thickness rotator cuff tear, MR arthrography is more sensitive and specific than either MRI or ultrasound ( p < 0.05). There are no significant differences in either sensitivity or specificity between MRI and ultrasound in the diagnosis of partial- or full-thickness rotator cuff tears ( p > 0.05). Summary ROC curves for MR arthrography, MRI, and ultrasound for all tears show the area under the ROC curve is greatest for MR arthrography (0.935), followed by ultrasound (0.889) and then MRI (0.878); however, pairwise comparisons of these curves show no significant differences between MRI and ultrasound ( p > 0.05). CONCLUSION. MR arthrography is the most sensitive and specific technique for diagnosing both full- and partial-thickness rotator cuff tears. Ultrasound and MRI are comparable in both sensitivity and specificity. n the workup of patients with shoulder pain, the role of imaging is to guide treatment decisions [1, 2]. The diagnosis of a rotator cuff tear and its extent, full or partial thickness, can determine whether the patient will be managed conservatively or will need surgery [3, 4]. Furthermore, the surgical approach, open versus arthroscopic, can be chosen once the correct diagnosis is made [3, 5]. Of the various imaging tests that have been used to evaluate the painful shoulder, unenhanced MRI, indirect and direct MR arthrography, and ultrasound have become the standards by which a rotator cuff tear is diagnos ed. In the medical literature, various sensitivities and specificities have been reported for these techniques. Although each technique has its inherent strengths and weaknesses, there seems to be no general consensus about which is the most accurate test, despite the large number of studies in the literature. To our knowledge only one other study has reviewed the existing literature and provid-

Keywords: MR arthrography, MRI, rotator cuff tear, shoulder, sports medicine, ultrasound DOI:10.2214/AJR.08.1241 Received May 15, 2008; accepted after revision December 30, 2008.
1

ed summary sensitivities and specificities of MRI and ultrasound for the diagnosis of rotator cuff tears [6]. That study included literature with both surgical and nonsurgical reference standards and was published in 2003, after which important advances in both MRI and ultrasound imaging were made. We report an updated meta-analysis comparing the diagnostic capabilities of MRI, MR arthrography, and ultrasound for rotator cuff tears using surgery as the only reference standard. Materials and Methods Data Sources and Literature Search
A comprehensive literature search of the MEDLINE database was performed using the following keywords: rotator cuff and rotator cuff tear; magnetic resonance imaging, magnetic resonance, MRI, and MR; magnetic resonance arthrography and MR arthrography; ultrasound, ultrasonography, sonography, and US. Articles published from 1966 to September 2007 were searched and included publications in

All authors: Department of Radiology, Thomas Jefferson University Hospital, Rm. 763-E, Main Bldg., 132 S 10th St., Philadelphia, PA 19107-5244. Address correspondence to L. N. Nazarian (levon.nazarian@jefferson.edu).

AJR 2009; 192:17011707 0361803X/09/19261701 American Roentgen Ray Society

AJR:192, June 2009

1701

de Jesus et al.
all languages and involving human and animal subjects. In one article [16] that assessed partialthickness tears, three full-thickness tears were identified as partial-thickness tears; however, because a tear was identified, we considered those studies to be TPs. Another article [21] evaluated the sensitivity and specificity of six MRI findings in detecting a full-thickness rotator cuff tear; TPs, TNs, FPs, and FNs were recorded for each finding. Because we were concerned with techniques and not with specific findings, we picked one finding to represent MRI. For each finding, we calculated the [(sensitivity + specificity) / 2] and picked the finding with the highest value. This finding turned out to be thinning of the tendinous cuff. In one article [22], some MRI diagnoses by re port were equivocal (e.g., partial- versus fullthickness tear). The authors of this study used the more severe diagnosis in the final tally of no disease versus disease with the rationale that a more significant diagnosis is more likely to affect management. One article [27] compared the accuracy of two MR pulse sequences. Because we were not interested in individual sequences but more in the actual technique, we omitted one of the sequences. The sequence we omitted was trivial because the authors of that article found no diagnostically significant difference between the two sequences, and the reported sensitivities and specificities were identical. Similarly, another article [28] looked at the accuracy of T2-weighted sequences with and without fat saturation in diagnosing full- and partial-thickness rotator cuff tears. For each of the four sets of data, we took [(sensitivity + specificity) / 2] and omitted the sequence with the lowest value (i.e., the fat-saturated fast spin-echo sequence). In one article [30], the authors divided patients into two groups on the basis of who performed ultrasound: Group 1 patients underwent ultrasound performed by a sonographer with 5 years of experience and group 2, by a radiologist with 10 years of experience. We omitted the data from the group examined by the sonographer. Another study [31] examined the interobserver agreement of five readers, each of whom inter preted MR images twice, first as a blinded review and second with knowledge of the surgical outcome. Sensitivities and specificities were calculated for full- and partial-thickness tears for each reader, generating five sets of data. For our purposes, we averaged the TPs, TNs, FPs, and FNs for full- and partial-thickness tears for both the blinded and retrospective readings. In one study [32], investigators also reported TPs, TNs, FPs, and FNs for two readers and in another study [33], for four readers. In a similar fashion, we took the average TP, average TN, average FP, and average FN and calculated sensitivities and specificities. One article [36] compared MR arthrography performed on a low-field magnet (0.2-T) and on a high-field magnet (1.5-T); two sets of data were reported. The reported sensitivities and specificities were identical, so we dropped one of the data sets. Similar to other studies, one study [38] had two sets of data for two independent readers. We averaged the TP, TN, FP, and FN values for those readers. In one article [45], the diagnosis of a rotator cuff tear was established on the basis of findings from transverse, parasagittal, or both transverse and parasagittal MR images of the shoulder. Two radiologists independently evaluated the planar images for each shoulder. We used the reported data for diagnosis based on both the transverse and parasagittal images. As in prior studies, we averaged the data for the two readers. Another article [46] compared T2-weighted fast spin-echo with and without fat suppression. Of the data for the two sequences, we retained the data set with the higher sensitivity and specificity. One study [49] compared the results of MR arthrography performed using three different solutions for intraarticular injection: Ringer solution and two different concentrations of gadoteridol. All three solutions were reported as equivalent in diagnostic accuracy. Furthermore, two independent readers evaluated all the studies, and interobserver agreement was calculated as a kappa value for each of the three solutions. We took the data set for the contrast solution with the highest kappa value and averaged the data for the two readers. In studies with multiple observers and multiple data sets [3133, 38, 49], we took averages of the contingency data (TPs, TNs, FPs, FNs) to calcu late sensitivities and specificities, as described earlier. In one article [52] that looked at ultrasound of the rotator cuff, the authors established the diagnosis of a tear using published diagnostic criteria and again using a subset of the published criteria. We used the data set that yielded the higher accuracynamely, the one generated from the more restricted subset of the published criteria. In references 4551, multiple reference standards were used in each study. For each article, we used only the data sets with surgically proven findings that is, either open surgery or arthroscopy. Meister et al. [61] reported nine full-thickness tears, 28 partial-thickness tears, and 39 intact tendons by arthroscopy. Using MR arthrography, the authors recorded nine full-thickness tears that were classified as true-negatives in this analysis of partial-thickness tears [61]. In the calculation

Study Selection
Our query of the MEDLINE database return ed 1,195 hits. The articles were analyzed for concordance with the inclusion criteria. These criteria are English language; absolute (raw) data on rotator cuff tears (full or partial thickness or both) in the form of true-positives (TPs), truenegatives (TNs), false-positives (FPs), and falsenegatives (FNs) either provided or extractable; surgical reference standard (arthroscopy or open surgery); and diagnostic imaging studies interpreted by radiologists. In addition, data must not have been published in a prior study. To prevent this possibility, we included only the article with the earlier publication date if two articles with common authors or from the same institution had overlapping dates of subject inclusion. Non-English-language (n = 160) and animal (n = 1) studies were excluded. The abstracts of the remaining studies were evaluated for relevance to our study. Of these, 270 relevant articles were retrieved. One hundred sixty-two of the 270 were excluded because either raw data were not provid ed or the data could not be extracted into discrete TPs, TNs, FPs, and FNs. Of the remaining 108 articles with data, 43 were excluded for the following reasons: 18 studies had a nonsurgical standard of reference, 15 had ultrasound read by nonradiologists, one had MRI read by non radiologists, and nine had overlapping dates of subject inclusion with other studies by common authors from the same institution. Therefore, 65 of 270 (24.1%) of the English-language articles met the inclusion criteria [771]. Twenty-five studies analyzed MRI only; five, MRI and MR arthro graphy; nine, MR arthrography only; five, MRI and ultrasound; one, MR arthrography and ultrasound; and 20, ultrasound only. The most recently published article was in July 2007. The oldest was published in May 1985. From the 65 articles that fulfilled the inclusion criteria, we retrieved a total of 140 data sets: 48 ultrasound, 67 MRI, and 25 MR arthrography. The breakdown of the data sets by diagnostic end point was as follows: 56 evaluated for the presence of a rotator cuff tear (including full or partial thickness) versus no tear; 49 evaluated full-thickness tear versus non-full-thickness tear (including partial thickness tear or no tear); and 35 evaluated partial-thickness tear versus nonpartial-thickness tear (including full-thickness tear or no tear). The heterogeneous nature of these studies required judgments regarding which data to include and how to avoid including duplicate data. These judgments are detailed here.

1702

AJR:192, June 2009

MRI, MR Arthrography, and Ultrasound of Rotator Cuff Tears

of specificity of MR arthrography for the diagnosis of partial-thickness tears, a TN is not a partial tearthat is, composed of full-thickness tears and intact tendons. However, the authors reported a specificity of 96% (43/45), which appears instead to correspond to the negative predictive value. Specificity should actually be 90% (43/48) if the nine full-thickness tears are counted as TNs. In their study, Milosavljevic et al. [64] noted that in the calculation of the sensitivity of ultrasound for partial-thickness tears, in seven additional shoulders, ultrasound identified a full-thickness tear instead of a partial-thickness tear. Because a tear was identified, these studies were considered to be true-positive. For the purpose of calculating sensitivity for partial-thickness tears in this metaanalysis, we counted only the 17 partial-thickness tears as TPs. Herold et al. [65] recorded data for two readers. These data were averaged in the aforementioned fashion. Additionally the authors noted that each patient underwent MR arthrography twice: once with the patients shoulder in a neutral position and a second time with it abducted and externally rotated (ABER). Sensitivity, specificity, and accuracy were compared. As done previously, we calculated [(sensitivity + specificity) / 2] and took the data set with the higher values, which turned out to be the ABER data set. Fritz et al. [67] evaluated the association of cystic changes at tendon insertion sites with rotator cuff disorders. Overall performance for tears was reported. Data for full- or partial-thickness tears could not be gleaned. Ferrari et al. [71] classified full-thickness supraspinatus tears as focal, subtotal, or total and classified partial-thickness tears as intratendinous, articular, or bursal-sided. For the meta-analysis, these subcategories of full- and partial-thickness tears were not considered separately. compare the Q* points of the curvethat is, the points on an ROC curve where sensitivity equals specificity [72]. We analyzed the data for this meta-analysis in both ways. It should be noted that an article with a small sample size has a small impact on the summary analysis because it is aggregated into a pool of data. However, such a study may appear as an outlier on the scatterplots. Additionally, in some instances, a value for sensi tivity or specificity could not be calculated from the contingency data. For instance, Toyoda et al. [63] report a data set with 41 shoulders and a sensitivity of 100% for 41 TPs and 0 FNs; specificity could not be calculated because there were 0 TNs and 0 FPs. These data sets were not included in either the distribution plots or the ROC curves because an actual value, 0 or a positive value, is needed to generate a point on the plot or curve.

Statistical Analysis
Our central interest was to compare the diagnostic accuracies of MRI, MR arthrography, and ultrasound in the evaluation of rotator cuff tears using a surgical reference standard. Because investigators from different studies reported us ing different criteria for the diagnosis of rotator cuff tears (full-thickness, partial-thickness, or both), we compared techniques within and across criteria. Two common approaches appear in the metaanalysis literature for the analysis of diagnostic tests. One approach is to pool data from a number of studies to obtain overall sensitivities and specificities and compare them using the chisquare test. A second approach involves using regression to construct summary ROC curves for each technique and then computing a z test to

Results Figure 1 shows distribution plots for fullthickness tears, full- and partial-thickness tears, and partial-thickness tears. These scatterplots show sensitivity and (1 specificity) on the y- and x-axes, respectively. Each point on the diagrams represents a published study. The meta-analysis of an ideal imaging techniquethat is, one that is highly sensitive with a low FP ratewould be depicted with a cluster of data points in the upper left area of the plot. The data points for MR arthrography in all three plots are tightly clustered in the upper left, although there are only 25 points: eight for fullthickness tears, 11 for full-thickness and partial tears, and six for partial-thickness tears. There is considerable overlap in the distribution of data points for MRI and ultrasound. For full-thickness tears (Fig. 1A), MRI and ultrasound have a grouping of data points in the upper left. However, for full- and partial-thickness tears (Fig. 1B), MRI and ultrasound have data points somewhat loosely dispersed across the upper half of the graph, corresponding to high sensitivity and varying FP rates. Both have prominent outliers. For partial-thickness tears (Fig. 1C), MRI and ultrasound have a low FP rate but a wide variation in sensitivities, with data points dispersed loosely in the left one third of the plot. Again, there are outliers for both. Table 1 illustrates the pooled sensitivities and specificities of each imaging technique with respect to full-thickness tears; partialthickness tears; full- or partial-thickness tears; and total tears, which is the sum of the first three groups.

For full-thickness tears, chi-square analysis shows no significant difference in sensitivity among the three techniques. Furthermore, ultrasound and MRI are not significantly different in sensitivity or specificity. MR arthrography is more specific than either MRI or ultrasound (specificity vs MRI, 2 = 40.142, p < 0.0001; specificity vs ultrasound, 2 = 25.836, p < 0.0001). For partial-thickness tears, MR arthrography is more sensitive and more specific than either MRI or ultrasound (sensitivity and specificity of MR arthrography vs MRI: 2= 27.358 and 14.134, p < 0.0001 and 0.0002, respectively; vs ultrasound: 2 = 21.635 and 5.111, p < 0.0001 and 0.02). Although there is no statistically significant difference between MRI and ultrasound for the diagnosis of partial-thickness tears, ultrasound tends to be more sensitive and more specific (sensitivity: 2 = 2.057, p = 0.15; specificity: 2 = 3.347, p = 0.067). For the diagnosis of full- or partial-thickness tears, MR arthrography is more sensitive and more specific than either MRI or ultrasound (sensitivity and specificity of MR arthrography vs MRI: 2 = 8.130 and 54.990, p < 0.0004 and 0.0001, respectively; vs ultrasound: 2 = 14.843 and 28.073, p < 0.0001 and 0.0001). However, there is no statistically significant difference between MRI and ultrasound ( p > 0.05). Figure 2 shows the summary ROC curves for MR arthrography, MRI, and ultrasound for all tears. We followed the method described by Langlotz and Sonnad [73], which consists of fitting a regression line to the sum and difference of the logit transforms of the sensitivities and specificities. MR arthrography has the greatest area under the ROC curve (0.935), followed by ultrasound (0.889) and then MRI (0.878). The curves for ultrasound and MRI cross at the location where sensitivity is approximately 88% and 1 specificity is 15%, after which MRI is shown to be superior to ultrasound. Q* is the point on an ROC curve where sensitivity and specificity are equal and has been proposed as a more relevant measure of test efficacy than the entire ROC curve. After Moses et al. [72], we calculated Q* and its SD for each technique: For MR arthrography, these values were 0.92 and 0.057, respectively; MRI, 0.86 and 0.032; and ultrasound, 0.86 and 0.049. The results of z tests show that pairwise comparisons are not significantly different ( p > 0.05).

AJR:192, June 2009

1703

de Jesus et al.

100 90 80

100 90 80

Sensitivity (%)

60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100

Sensitivity (%)

70

70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100

1 Specificity

1 Specificity

A
100 90 80

B
Fig. 1 Distribution plots of sensitivities and false-positive rates (1 specificity) of MR arthrography (), MRI (), and ultrasound () studies for analyzing rotator cuff tears. Each symbol corresponds to data set from published study. AC, Plots show data for detection of full-thickness tears ( A ), full- and partialthickness tears (B), and partial-thickness tears (C).

Sensitivity (%)

70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100

1 Specificity

C For the construction of the distribution plots, studies in which TP, TN, FP, FN, or a combination of these values equal 0 are not included because 0 values cannot be plotted. Likewise, for the composite ROC curve, these studies are excluded. For this reason, two of the MR arthrography studies [24, 63] are not plotted in the distribution plots and ROC curve. Therefore, MR arthrography is shown to be significantly better than MRI or ultrasound in the chi-square values but not the ROC curves. Discussion For the patient with shoulder pain, a host of therapeutic options, ranging from medical management to physiotherapy to open surgery, are available [6, 74]. The role of diagnostic imaging is to help guide surgical or nonsurgical management. The ideal imaging technique should have a high rate of TPs and an acceptable rate of FPs to limit unnecessary surgical intervention. In the 65 studies included in our analysis, most of the reported sensitivities and specificities fall in the range of 60100% for ultrasound. A similarly wide variation is observed for MRI and MR arthrography. There are a few possible reasons for such a variation in numbers including small sample sizes per study, differing study designs, varying quality of imaging equipment (e.g., a wide range of ultrasound probe frequencies and MRI field strengths), and differing imaging criteria for diagnosis. We sought to draw on the strengths of a meta-analysis to evaluate the large body of literature to overcome the small sample sizes and heterogeneous designs of individual trials by pooling the data and obtaining summary sensitivities and specificities. Our findings show that MR arthrography is more sensitive and more specific than ei-

ther ultrasound or MRI in diagnosing both full- and partial-thickness rotator cuff tears. Additionally, there is no statistically significant difference between the sensitivities and specificities of MRI versus ultrasound in diagnosing either full- or partial-thickness tears. To our knowledge, only one other metaanalysis of rotator cuff tears has been published [6]. In that systematic review published in 2003, Dinnes et al. [6] evaluated the diagnostic effectiveness of MRI, MR arthrography, ultrasound, and clinical examination in the evaluation of a painful shoulder, with rotator cuff tears as the disease end point. They concluded that either MRI or ultrasound could be used for equal detection of full-thickness rotator cuff tears but that ultrasound is the more cost-effective test. They also stated that MR arthrography appeared to perform better than either MRI or ultrasound but that any such benefit must be set against the invasiveness and potential discomfort to patients of the procedure [6]. Our inclusion criteria differed from those used for the systematic review by Dinnes et al. [6] because we used data sets with only surgically proven findings. In the study by Dinnes and colleagues, a large number of studies used nonsurgical techniques, such as arthrography, as the reference standard for disease. In one article, investigators had even used MRI as the reference standard for disease [75]. Also, Dinnes et al. included studies

1704

AJR:192, June 2009

MRI, MR Arthrography, and Ultrasound of Rotator Cuff Tears TABLE 1: Summary Sensitivities and Specificities for MR Arthrography, MRI, and Ultrasound
Technique Full-thickness tear MR arthrography MRI Ultrasound Partial-thickness tear MR arthrography MRI Ultrasound Full- or partial-thickness tear MR arthrography MRI Ultrasound All tearsa MR arthrography MRI Ultrasound 907 1,528 1,803 91.7 85.5 85.1 1.8 1.8 1.6 2,623 2,464 1,945 96.5 90.4 92.0 0.7 1.2 1.2 485 667 915 92.3 87.0 85.1 2.4 2.6 2.3 869 463 481 94.5 81.7 86.1 1.5 3.5 3.1 195 236 249 85.9 63.6 66.7 4.9 6.2 5.9 875 916 790 96.0 91.7 93.5 1.3 1.8 1.7 227 625 639 95.4 92.1 92.3 2.7 2.1 2.1 879 1,085 674 98.9 92.9 94.4 0.7 1.5 1.7 No. of Positive Cases Sensitivity (%) CI (%) No. of Negative Cases Specificity (%) CI (%)

NoteThe results are arranged by the type of tear reported on in the individual study. CI = confidence interval. aSum of data listed under full-thickness tear, partial-thickness tear, and full- or partial-thickness tear.
100 90 80

Sensitivity (%)

70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100

Fig. 2 Overall summary receiver operating characteristic curves for detection of rotator cuff tears using MR arthrography (solid line ), MRI (dotted line ), and ultrasound (dashed line ).

1 Specificity

in which ultrasound had been performed by nonradiologists, usually orthopedic surgeons or rheumatologists. We included only those studies in which radiologists performed and interpreted the imaging studies. Furthermore, we directly compared MR arthrography, MRI, and ultrasound using ROC curves and single-statistic summaries (Q*). Despite all of these differences, our conclusions are similar to those of Dinnes and colleagues. Our study has limitations. First, although the criteria for diagnosing a full- or partial-thickness tear of the rotator cuff have been published, the criteria used to make the diagnosis in each study varied. This variability reflects

the wide range in time over which the included studies were published and is a manifestation of the maturation process of each technique for which criteria for diagnosis must be refined. The selection bias and workup bias inherent to each individual study also may play an important role in our pooled study. The design of most studies was retrospective to identify patients who underwent the index test. Of this set of patients, a subset who subsequently underwent surgery was identified as the sample population. Because of workup bias, most patients who underwent the index test and subsequent surgery had a high suspicion and probability of, indeed, having a tear; effectively, those

studies assessed the accuracy of the index test in detecting a tear in a patient with a high pretest clinical suspicion of a rotator cuff tear. Additional forms of bias encountered by Dinnes et al. [6] were present in our study. In partial verification bias, only patients who underwent the reference test were included in a sample, with the results of the remaining patients who underwent the index test (MRI, ultrasound, MR arthrography) unreported. Publication bias is a particular drawback of meta-analyses. Studies with favorable results have a higher likelihood of being published, creating an inherent selection bias during a literature review [6, 76, 77]. In summary, MR arthrography is more accurate than MRI and ultrasound in diagnosing rotator cuff tears. Ultrasound is as accurate as MRI for both full-thickness tears and partialthickness tears. These results, combined with the lower cost for ultrasound, suggest that ultrasound may be the most cost-effective imaging method for screening for rotator cuff tears provided that the examiner has been properly trained in this operator-dependent technique. For practitioners without ultrasound expertise, MRI can be used. MR arthrography can be performed in cases in which ultrasound and MRI are not definitive. References
1. Post M, Silver R, Singh M. Rotator cuff tear: diagnosis and treatment. Clin Orthop Relat Res 1983;

AJR:192, June 2009

1705

de Jesus et al.
173:7891 2. Boenisch U, Lembcke O, Naumann T. Classification, clinical findings and operative treatment of degenerative and posttraumatic shoulder disease: what do we really need to know from an imaging report to establish a treatment strategy? Eur J Radiol 2000; 35:103118 3. Ruotolo C, Nottage WM. Surgical and nonsurgical management of rotator cuff tears. Arthroscopy 2002; 18:527531 4. Mantone JK, Burkhead WZ Jr, Noonan J Jr. Nonoperative treatment of rotator cuff tears. Orthop Clin North Am 2000; 31:295311 5. Gartsman GM, Khan M, Hammerman SM. Arthroscopic repair of full-thickness tears of the rotator cuff. J Bone Joint Surg Am 1998; 80:832 840 6. 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:iii, 1166 7. Tuite MJ, Yandow DR, DeSmet AA, Orwin JF, Quintana FA. Diagnosis of partial and complete rotator cuff tears using combined gradient echo and spin echo imaging. Skeletal Radiol 1994; 23:541545 8. van Holsbeeck MT, Kolowich PA, Eyler WR, et al. US depiction of partial-thickness tear of the rotator cuff. Radiology 1995; 197:443446 9. Iannotti JP, Zlatkin MB, Esterhai JL, Kressel HY, Dalinka MK, Spindler KP. Magnetic resonance imaging of the shoulder: sensitivity, specificity, and predictive value. J Bone Joint Surg Am 1991; 73:1729 10. Kurol M, Rahme H, Hilding S. Sonography for diagnosis of rotator cuff tear: comparison with observations at surgery in 58 shoulders. Acta Orthop Scand 1991; 62:465467 11. Wiener SN, Seitz WH. Sonography of the shoulder in patients with tears of the rotator cuff: accuracy and value for selecting surgical options. AJR 1993; 160:103107 12. Paavolainen P, Ahovuo J. Ultrasonography and arthrography in the diagnosis of tears of the rotator cuff. J Bone Joint Surg Am 1994; 76:335340 13. Misamore GW, Woodward C. Evaluation of degenerative lesions of the rotator cuff: a comparison of arthrography and ultrasonography. J Bone Joint Surg Am 1991; 73:704706 14. Hodler J, Fretz CJ, Terrier F, Gerber C. Rotator cuff tears: correlation of sonographic and surgical findings. Radiology 1988; 169:791794 15. Bretzke CA, Crass JR, Craig EV, Feinberg SB. Ultrasonography of the rotator cuff: normal and pathologic anatomy. Invest Radiol 1985; 20:311 315 16. 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; 82:498504 17. Farin PU, Jaroma H. Acute traumatic tears of the rotator cuff: value of sonography. Radiology 1995; 197:269273 18. Brenneke SL, Morgan CJ. Evaluation of ultrasonography as a diagnostic technique in the assessment of rotator cuff tendon tears. Am J Sports Med 1992; 20:287289 19. Karzel RP, Snyder SJ. Magnetic resonance arthrography of the shoulder: a new technique of shoulder imaging. Clin Sports Med 1993; 12:123 136 20. Traughber PD, Goodwin TE. Shoulder MRI: arthroscopic correlation with emphasis on partial tears. J Comput Assist Tomogr 1992; 16:129133 21. Farley TE, Neumann CH, Steinbach LS, Jahnke AJ, Petersen SS. Full-thickness tears of the rotator cuff of the shoulder: diagnosis with MR imaging. AJR 1992; 158:347351 22. Wnorowski DC, Levinsohn EM, Chamberlain BC, McAndrew DL. Magnetic resonance imaging assessment of the rotator cuff: is it really accurate? Arthroscopy 1997; 13:710719 23. Nelson MC, Leather GP, Nirschl RP, Pettrone FA, Freedman MT. Evaluation of the painful shoulder: a prospective comparison of magnetic resonance imaging, computerized tomographic arthrography, ultrasonography, and operative findings. J Bone Joint Surg Am 1991; 73:707716 24. Wagner SC, Schweitzer ME, Morrison WB, Fenlin JM Jr, Bartolozzi AR. Shoulder instability: accuracy of MR imaging performed after surgery in depicting recurrent injuryinitial findings. Radiology 2002; 222:196203 25. Quinn SF, Sheley RC, Demlow TA, Szumowski J. Rotator cuff tendon tears: evaluation with fatsuppressed MR imaging with arthroscopic correlation in 100 patients. Radiology 1995; 195:497 500 26. Shellock FG, Bert JM, Fritts HM, Gundry CR, Easton R, Crues JV. Evaluation of the rotator cuff and glenoid labrum using a 0.2-Tesla extremity magnetic resonance (MR) system: MR results compared to surgical findings. J Magn Reson Imaging 2001; 14:763770 27. Sonin AH, Peduto AJ, Fitzgerald SW, Callahan CM, Bresler ME. MR imaging of the rotator cuff mechanism: comparison of spin-echo and turbo spin-echo sequences. AJR 1996; 167:333338 28. Needell SD, Zlatkin MB. Comparison of fat-saturation fast spin echo versus conventional spinecho MRI in the detection of rotator cuff pathology. J Magn Reson Imaging 1997; 7:674677 29. Hodler J, Kursunoglu-Brahme S, Snyder SJ, et al. Rotator cuff disease: assessment with MR arthrography versus standard MR imaging in 36 patients with arthroscopic confirmation. Radiology 1992; 182:431436 30. Chang CY, Wang SF, Chiou HJ, Ma HL, Sun YC, Wu HD. Comparison of shoulder ultrasound and MR imaging in diagnosing full-thickness rotator cuff tears. Clin Imaging 2002; 26:5054 31. Balich SM, Sheley RC, Brown TR, Sauser DD, Quinn SF. MR imaging of the rotator cuff tendon: interobserver agreement and analysis of interpretive errors. Radiology 1997; 204:191194 32. Tuite MJ, Asinger D, Orwin JF. Angled oblique sagittal MR imaging of rotator cuff tears: comparison with standard oblique sagittal images. Skeletal Radiol 2001; 30:262269 33. Robertson PL, Schweitzer ME, Mitchell DG, et al. Rotator cuff disorders: interobserver and intraobserver variation in diagnosis with MR imaging. Radiology 1995; 194:831835 34. Prickett WD, Teefey SA, Galatz LM, Calfee RP, Middleton WD, Yamaguchi K. Accuracy of ultrasound imaging of the rotator cuff in shoulders that are painful postoperatively. J Bone Joint Surg Am 2003; 85:10841089 35. Blanchard TK, Bearcroft PW, Constant CR, Griffin DR, Dixon AK. Diagnostic and therapeutic impact of MRI and arthrography in the investigation of full-thickness rotator cuff tears. Eur Radiol 1999; 9:638642 36. Loew R, Kreitner KF, Runkel M, Zoellner J, Thelen M. MR arthrography of the shoulder: comparison of low-field (0.2 T) vs. high-field (1.5 T) imaging. Eur Radiol 2000; 10:989996 37. Wang YM, Shih TT, Jiang CC, et al. Magnetic resonance imaging of rotator cuff lesions. J Formos Med Assoc 1994; 93:234239 38. Yagci B, Manisali M, Yilmaz E, et al. Indirect MR arthrography of the shoulder in detection of rotator cuff ruptures. Eur Radiol 2001; 11:258262 39. Read JW, Perko M. Shoulder ultrasound: diagnostic accuracy for impingement syndrome, rotator cuff tear, and biceps tendon pathology. J Shoulder Elbow Surg 1998; 7:264271 40. Sonnabend DH, Hughes JS, Giuffre BM, Farrell R. The clinical role of shoulder ultrasound. Aust N Z J Surg 1997; 67:630633 41. Torstensen ET, Hollinshead RM. Comparison of magnetic resonance imaging and arthroscopy in the evaluation of shoulder pathology. J Shoulder Elbow Surg 1999; 8:4245 42. Zlatkin MB, Iannotti JP, Roberts MC, et al. Rotator cuff tears: diagnostic performance of MR imaging. Radiology 1989; 172:223229 43. Pattee GA, Snyder SJ. Sonographic evaluation of the rotator cuff: correlation with arthroscopy. Arthroscopy 1988; 4:1520 44. Teefey SA, Rubin DA, Middleton WD, Hildebolt CF, Leibold RA, Yamaguchi K. Detection and

1706

AJR:192, June 2009

MRI, MR Arthrography, and Ultrasound of Rotator Cuff Tears

quantification of rotator cuff tears: comparison of ultrasonographic, magnetic resonance imaging, and arthroscopic findings in seventy-one consecutive cases. J Bone Joint Surg Am 2004; 86:708716 45. Pfirrmann CW, Zanetti M, Weishaupt D, Gerber C, Hodler J. Subscapularis tendon tears: detection and grading at MR arthrography. Radiology 1999; 213:709714 46. Singson RD, Hoang T, Dan S, Friedman M. MR evaluation of rotator cuff pathology using T2weighted fast spin-echo technique with and without fat suppression. AJR 1996; 166:10611065 47. Martn-Hervs C, Romero J, Navas-Acin A, Reboiras JJ, Munuera L. Ultrasonographic and magnetic resonance images of rotator cuff lesions compared with arthroscopy or open surgery findings. J Shoulder Elbow Surg 2001; 10:410415 48. Motamedi AR, Urrea LH, Hancock RE, Hawkins RJ, Ho C. Accuracy of magnetic resonance imaging in determining the presence and size of recurrent rotator cuff tears. J Shoulder Elbow Surg 2002; 11:610 49. Binkert CA, Zanetti M, Gerber C, Hodler J. MR arthrography of the glenohumeral joint: two concentrations of gadoteridol versus Ringer solution as the intraarticular contrast material. Radiology 2001; 220:219224 50. Magee TH, Gaenslen ES, Seitz R, Hinson GA, Wetzel LH. MR imaging of the shoulder after surgery. AJR 1997; 168:925928 51. Gaenslen ES, Satterlee CC, Hinson GW. Magnetic resonance imaging for evaluation of failed repairs of the rotator cuff. J Bone Joint Surg Am 1996; 78:13911396 52. Brandt TD, Cardone BW, Grant TH, Post M, Weiss CA. Rotator cuff sonography: a reassessment. Radiology 1989; 173:323327 53. Palmer WE, Brown JH, Rosenthal DI. Rotator cuff: evaluation with fat-suppressed MR arthrography. Radiology 1993; 188:683687 54. Soble MG, Kaye AD, Guay RC. Rotator cuff tear: clinical experience with sonographic detection. Radiology 1989; 173:319321 55. Mack LA, Gannon MK, Kilcoyne RF, Matsen RA. Sonographic evaluation of the rotator cuff: accuracy in patients without prior surgery. Clin Orthop Relat Res 1988; 234:2127 56. Burk DL Jr, Karasick D, Kurtz AB, et al. Rotator cuff tears: prospective comparison of MR imaging with arthrography, sonography, and surgery. AJR 1989; 153:8792 57. Vick CW, Bell SA. Rotator cuff tears: diagnosis with sonography. AJR 1990; 154:121123 58. Jaovisidha S, Jacobson JA, Lenchik L, Resnick D. MR imaging of rotator cuff tears: is there a diagnostic benefit of shoulder exercise prior to imaging? Clin Imaging 1999; 23:249253 59. Burk DL Jr, Torres JL, Marone PJ, Mitchell DG, Rifkin MD, Karasick D. MR imaging of shoulder injuries in professional baseball players. J Magn Reson Imaging 1991; 1:385389 60. Sperling JW, Potter HG, Craig EV, Flatow E, Warren RF. Magnetic resonance imaging of painful shoulder arthroplasty. J Shoulder Elbow Surg 2002; 11:315321 61. Meister K, Thesing J, Montgomery WJ, Indelicato PA, Walczak S, Fontenot W. MR arthrography of partial thickness tears of the undersurface of the rotator cuff: an arthroscopic correlation. Skeletal Radiol 2004; 33:136141 62. Zlatkin MB, Hoffman C, Shellock FG. Assessment of the rotator cuff and glenoid labrum using an extremity MR system: MR results compared to surgical findings from a multi-center study. J Magn Reson Imaging 2004; 19:623631 63. Toyoda H, Ito Y, Tomo H, Nakao Y, Koike T, Takaoka K. Evaluation of rotator cuff tears with magnetic resonance arthrography. Clin Orthop Relat Res 2005; 439:109115 64. Milosavljevic J, Elvin A, Rahme H. Ultrasonography of the rotator cuff: a comparison with arthroscopy in one-hundred-and-ninety consecutive cases. Acta Radiol 2005; 46:858865 65. Herold T, Bachthaler M, Hamer OW, et al. Indirect MR arthrography of the shoulder: use of abduction and external rotation to detect full- and partial-thickness tears of the supraspinatus tendon. Radiology 2006; 240:152160 66. Cullen DM, Breidahl WH, Janes GC. Diagnostic accuracy of shoulder ultrasound performed by a single operator. Australas Radiol 2007; 51:226229 67. Fritz LB, Ouellette HA, OHanley TA, Kassarjian A, Palmer WE. Cystic changes at supraspinatus and infraspinatus tendon insertion sites: association with age and rotator cuff disorders in 238 patients. Radiology 2007; 244:239248 68. Waldt S, Bruegel M, Mueller D, et al. Rotator cuff tears: assessment with MR arthrography in 275 patients with arthroscopic correlation. Eur Radiol 2007; 17:491498 69. Stetson WB, Phillips T, Deutsch A. The use of magnetic resonance arthrography to detect partial-thickness rotator cuff tears. J Bone Joint Surg Am 2005; 87:8188 70. Magee T, Williams D. 3.0-T MRI of the supraspinatus tendon. AJR 2006; 187:881886 71. Ferrari FS, Governi S, Burresi F, Vigni F, Stefani P. Supraspinatus tendon tears: comparison of US and MR arthrography with surgical correlation. Eur Radiol 2002; 12:12111217 72. Moses L, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med 1993; 12:12931316 73. Langlotz CP, Sonnad SS. Meta-analysis of diagnostic procedures: a brief overview. Acad Radiol 1998; 5[suppl 2]:S269S273 74. Green S, Buchbinder R, Hetrick S. Physiotherapy interventions for shoulder pain. Cochrane Database Syst Rev 2003; 2:CD004258 75. Naredo AE, Aguado P, Padron M, et al. A comparative study of ultrasonography with magnetic resonance imaging in patients with painful shoulder. J Clin Rheumatol 1999; 5:184192 76. Duval SJ, Tweedie RL. Practical estimates of the effect of publication bias in meta-analysis. Australasian Epidemiologist 1998; 5:1417 77. Sutton AJ, Duval SJ, Tweedie RL, Abrams KR, Jones DR. Empirical assessment of effect of publication bias on meta-analyses. BMJ 2000; 320: 15741577

AJR:192, June 2009

1707