Rotator cuff tear has a prevalence of approximately 20% in the general population and causes substantial disability, including shoulder pain and dysfunction [
1]. Given recent advances in imaging and surgical techniques, rotator cuff repair generally yields satisfactory clinical outcomes [
2,
3]. However, postoperative retear occurs in 34.2–70% of cases [
4,
5].
Over time after rotator cuff tear, the ruptured tendon gradually contracts, and the torn muscle exhibits fatty degeneration and atrophy with associated loss of elasticity [
6,
7]. As a result, passive tensile load must be applied during surgical repair to relocate the torn tendons to their original insertion on the greater tuberosity. However, as the required tensile force increases, the postoperative retear rate also increases [
8,
9]. The severity of the tendon tear substantially impacts the ability to achieve a sufficient tendon repair, with higher frequencies of retear reported for larger tears and tears with greater tendon retraction as well as for tears resulting in reduced muscle quality [
4,
5,
8–
12]. Studies have also shown a greater frequency of repair failure and retear as the severity of muscle degeneration increases [
10,
11,
13]. Therefore, accurate evaluation of muscle quality during preoperative planning of rotator cuff repair is important for guiding prognostic assessments [
7,
14].
MRI not only assesses tendon pathology but also provides information regarding muscle quality [
12]. For example, on MRI, the Goutallier grading system [
14,
15] can be used to assess the degree of muscular fatty infiltration and the occupation ratio [
15–
17] and muscle atrophy grade [
17] can be used to assess the degree of muscular atrophy [
14–
17]. However, MRI does not provide information regarding muscle tensile force or stiffness [
18].
Shear-wave elastography (SWE) is an ultrasound-based functional imaging method that quantifies tissue elasticity and has been applied for musculoskeletal imaging [
19,
20]. In particular, studies have applied SWE to evaluate various pathologies of the supraspinatus tendon of the rotator cuff [
21–
25]. However, a paucity of studies have investigated the relationship between SWE measurements and supraspinatus muscle quality [
26] or between SWE measurements and the ability to achieve successful repair of rotator cuff tears in patients [
7]. We thus conducted this study to explore the utility of preoperative SWE measurements of the supraspinatus muscle to predict successful rotator cuff repair, including comparisons of SWE measurements with MRI-based measures.
Methods
Study Population
This prospective observational study was approved by the institutional review board. Written informed consent was obtained from all participants. We recruited patients admitted to the hospital between May 2019 and January 2021 to undergo arthroscopic rotator cuff repair. At our institution, surgical repair is performed in patients with at least 6 months of atraumatic shoulder pain, and shoulder MRI is performed within 3 months before surgery. Enrolled patients underwent ultrasound of the shoulder, including SWE, performed for purposes of this investigation, within 2 days before the surgery. Both patient recruitment and ultrasound examinations were performed during the same hospital admission as the surgery: Recruitment occurred at least 1 day before the surgery, and the ultrasound examination was potentially performed on the same day as surgery.
Enrolled patients were subsequently excluded from the analysis after completion of the investigational ultrasound if the patient had undergone prior ipsilateral rotator cuff repair, if the patient had undergone any other prior ipsilateral orthopedic surgery, if the rotator cuff tear was massive (> 5 cm) based on the greatest dimension of the tendon tear as measured at surgery, if the supraspinatus tendon was not identified as the primary pathology at the time of surgery (e.g., isolated infraspinatus or subscapularis tendon tear), or if the SWE examination of the supraspinatus muscle was nondiagnostic (e.g., severe artifacts in the elastogram). Patients with massive tears were excluded because massive tears are generally considered to not be reparable and are managed by surgical alternatives to attempted rotator cuff repair (e.g., reversed total shield arthroplasty).
Ultrasound Examinations Including Shear-Wave Elastography
The investigational shoulder ultrasound examinations were performed using an ultrasound system (Aplio i800, Canon Medial System) with an 18-MHz linear transducer. The examinations were performed by a single musculoskeletal radiologist with 6 years of posttraining experience (E.K.K, reader 1). First, B-mode ultrasound was performed with the patient in the sitting position. Then, SWE was performed with the patient relaxing in the supine position, the neck gently stretched, and the head turned slightly to the contra-lateral side. The supine position was used instead of the sitting position to help maintain a fixed position of the probe during SWE acquisition and to reduce the amount of pressure required on the probe. A transverse view of both the supraspinatus and trapezius muscles was first obtained, corresponding to the expected location of the Y view on shoulder MRI (defined as the most lateral parasagittal image on which the scapular spine is in contact with the scapular body [
15]) (
Fig. 1A). Then, the probe was turned to obtain the longitudinal view while maintaining the alignment of the supraspinatus and trapezius muscles, and SWE was used to evaluate the fibers of the trapezius and supraspinatus muscles in the longitudinal plane (
Fig. 1B). SWE was acquired in continuous mode at a rate of 0.4 frames per second using the shear modulus with the elasticity range set to 0–200 kPa. Corresponding SWE boxes were placed on the propagation map and color elastogram.
After completion of the examinations, two individuals (reader 1 and reader 2, a 4th-year radiology resident [A.Y.K.]) independently extracted SWE measurements from the obtained images. Measurements were obtained in regions where the propagation map appeared smooth and without obvious artifact. Three circular ROIs with a diameter of 0.3 cm were placed on the supraspinatus muscle on the propagation map on each of three longitudinal views, for a total of nine ROIs (
Fig. 2A). The ROIs were automatically propagated to the elastogram, and the mean elasticity of each ROI was recorded. Anterior and posterior portions of the deep aspect of the muscle were selected for ROI placement [
21]. A similar process was used to obtain three ROIs from the trapezius muscle on each of three longitudinal images (
Fig. 2B).
The mean and median elasticity from the nine supraspinatus ROIs, as well as the mean elasticity of the nine trapezius ROIs, were determined for each patient. In addition, the elasticity ratio was calculated using the trapezius muscle as reference as the ratio of the mean elasticity of the supraspinatus muscle to the mean elasticity of the trapezius muscle. The two readers also assessed the degree of fatty infiltration of the supraspinatus muscle on the gray-scale ultrasound images using a 3-point scale with the echogenicity of the trapezius muscle serving as reference [
27]: grade 1, isoechogenic; grade 2, mildly hyperechogenic; grade 3, markedly hyperechogenic. Twelve weeks after completion of these interpretations, reader 1 repeated the SWE measurements. The SWE measurements by reader 2 and the second set of SWE measurements by reader 1 were used solely to assess interobserver and intraobserver reproducibility; all analyses otherwise used solely the first set of measurements by reader 1.
MRI Examinations
Patients underwent clinical shoulder MRI using a 3-T system (Skyra, Siemens Healthineers). Examinations included axial, coronal, and sagittal proton-density fat-saturation sequences; axial, sagittal, and coronal T2-weighted sequences; and a sagittal T1-weighted sequence. The sagittal T1-weighted sequence, which was used for assessing muscle quality, was acquired using the following parameters: TR/TE, 632–689/12; matrix size, 448 × 269; FOV, 16 × 16 cm; section thickness, 3 mm; intersection gap, 0 mm; echo-train length, 4; number of signals acquired, 1; and scan time, 104–114 seconds. The MRI parameters for all sequences are shown in Table S1, which can be viewed in the electronic supplement to this article at
https://doi.org/10.2214/AJR.21.27129.
At least 4 weeks after completion of all ultrasound interpretations, readers 1 and 2 independently evaluated the MRI examinations blinded to the SWE findings. The readers evaluated muscle quality on MRI using the following three methods during a single session for a given examination: Goutallier grade, occupation ratio, and muscle atrophy grade. The readers first identified the Y view on the sagittal T1-weighted sequence. The Goutallier grade was assessed using a subjective 5-point (0–4) scale: grade 0, no fatty infiltration; grade 1, small fatty streaks; grade 2, less fatty infiltration than muscle; grade 3, equal fatty infiltration to muscle; and grade 4, more fatty infiltration than muscle [
28]. To determine the occupation ratio, the readers traced the outer boundaries of the supraspinatus muscle and supraspinatus fossa and recorded the cross-sectional surface area of both regions; the ratio of the two areas was computed. The muscle atrophy grade was assessed using a subjective 4-point (0–3) scale, which was proposed by Warner et al. [
29], that is based on a reference line (the line defining the “tangent” sign [
17]) drawn along the upper margin of the scapular spine and body and was classified as follows: grade 0, muscle extending above the line; grade 1, muscle at the line; grade 2, muscle restricted to below the line by less than 50% of the supraspinatus fossa depth; grade 3, muscle restricted to below the line by 50% or greater of the supraspinatus fossa depth [
29,
30].
Surgical Technique and Determination of Sufficient Repair
All surgeries were performed by a single orthopedic surgeon (J.Y.J., with 10 years of posttraining experience) who was aware of the clinical MRI interpretations but was blinded to the SWE results and to the MRI assessments of muscle quality. All surgical procedures were performed with patients in the lateral decubitus position under general anesthesia with an interscalene block. The glenohumeral joint was assessed using a 30° arthroscope via a posterior approach. The arthroscope was inserted into the subacromial space, and the size of the rotator cuff tear was recorded. The surgery also included subacromial space débridement and, as needed, any additional bone intervention (acromioplasty) and treatment of biceps tendon tear. Repairs generally involved mobilizing the rotator cuff tendon to the far lateral end of the greater tuberosity using a suture-bridge technique. However, when this could not be performed due to severe retraction of the supraspinatus tendon, partial repair was performed for the infra-spinatus and subscapularis tendons using a single-row modified Mason-Allen technique.
On completion of the surgery, the surgeon prospectively classified all patients based on the amount of footprint coverage achieved by the rotator cuff origins as having a sufficient repair or an insufficient repair, using the approach described by Yoo et al. [
31]. For patients who underwent repair using the suture-bridge technique, insufficient repair was defined as when the repaired rotator cuff tendon did not cover at least half of the original footprint, typically due to poor mobility of the torn cuff. In patients who underwent repair using the modified Mason-Allen technique, insufficient repair was defined as when more than 10 mm of the humeral head remained exposed [
31]. Based on the size of rotator cuff tears measured at surgery, rotator cuff tears were classified as follows: small (< 1 cm), medium (1 to < 3 cm), large (3–5 cm), and massive (> 5 cm) [
32].
Statistical Analyses
Patient characteristics were summarized and were stratified by the classification of the repair as sufficient or insufficient. Categoric variables were compared using chi-square tests or Fisher exact tests; continuous variables were compared using t tests or Wilcoxon rank-sum tests. Interobserver agreement of Goutallier grade, muscle atrophy grade, and gray-scale ultrasound fatty infiltration grade was assessed by weighted kappa coefficients, and interobserver agreement of occupation ratio was assessed by intraclass correlation coefficients (ICCs). Interobserver and intraob-server agreement of the three SWE measures were assessed by ICCs. Spearman correlation coefficients were computed among each of the three MRI measures (Goutallier grade, occupation ratio, and muscle atrophy grade) and of the gray-scale fatty infiltration grade with each of the three SWE measures (mean elasticity, median elasticity, and elasticity ratio). Trend tests using the generalized linear model were performed between each of the three SWE measures with Goutallier grade and muscle atrophy grade.
The diagnostic performance for predicting insufficient repair was assessed by ROC analysis for the three MRI measures, the three SWE measures, and the gray-scale fatty infiltration grade, and AUC values, sensitivity, and specificity were calculated at cutoff values determined by the Youden index. Multivariable stepwise logistic regression analysis for predicting insufficient repair was performed using, as independent predictors, tendon tear size, the Goutallier grade, occupation ratio, muscle atrophy grade, gray-scale fatty infiltration grade, and the highest-performing SWE measure. The regression analysis used the penalized maximum likelihood estimation method; p values < .05 indicated statistically significant differences. SPSS Statistics software (version 20.0, IBM) and SAS software (version 9.4, SAS Institute) were used for statistical analyses.
Results
A total of 103 patients satisfied the initial screening criteria and were recruited for potential study participation. Of those patients, 11 declined to participate, and the remaining 92 patients who agreed to participate were enrolled. The investigational pre-operative ultrasound examination including SWE was performed in all 92 patients. A total of 18 patients were then excluded for the following reasons: prior ipsilateral rotator cuff repair (
n = 2), prior ipsilateral orthopedic surgery (
n = 1), massive rotator cuff tear identified at surgery (
n = 9), supraspinatus tear not identified as primary pathology at surgery (
n = 4), and nondiagnostic SWE evaluation of the supraspinatus muscle (
n = 2). These exclusions resulted in a final study sample of 74 patients (37 men, 37 women; mean age, 63.9 ± 10.0 [SD] years; age range, 37–83 years). Rotator cuff repair was sufficient in 60 patients and insufficient in 14 patients.
Figure 3 shows the flow of patient inclusions and exclusions, and
Table 1 summarizes patient characteristics.
The mean interval between preoperative MRI and surgery was 15.2 ± 18.9 days (range, 0–69 days), and the mean interval between preoperative ultrasound and surgery was 0.0 ± 0.4 days (range, 0–2 days). At surgery, all patients showed a full-thickness rotator cuff tendon tear. A total of 70 patients underwent repair using the suture-bridge technique, and four patients underwent repair using the modified Mason-Allen technique. Patients with sufficient repair and those with insufficient repair were not significantly different in terms of sex, age, body mass index, and laterality of surgery (all, p > .05).
Intraobserver and Interobserver Agreement
Interobserver agreement between the two readers was 0.900 for Goutallier grade, 0.960 for occupation ratio, 0.874 for muscle atrophy grade, 0.934 for mean elasticity, 0.925 for median elasticity, 0.949 for elasticity ratio, and 0.856 for gray-scale fatty infiltration grade. Intraobserver agreement for reader 1 was 0.925 for mean elasticity, 0.899 for median elasticity, and 0.965 for elasticity ratio.
Table 2 provides the 95% CIs for these values.
Comparison of MRI and Shear-Wave Elastography Measures Between Patients With Sufficient Repair and Patients With Insufficient Repair
Tables 1 and
3 show comparisons of MRI and SWE measures between patients with sufficient repair and those with insufficient repair. Patients with insufficient repair, compared with those with sufficient repair, more commonly exhibited a large tear (100.0% vs 50.0%), a Goutallier grade on MRI of 4 (78.6% vs 5.0%), a muscle atrophy grade on MRI of 2 (64.3% vs 10.0%) or 3 (21.4% vs 0.0%), and a gray-scale fatty infiltration grade of 3 (85.7% vs 8.1%) (all,
p < .05). In addition, patients with insufficient repair, compared with sufficient repair, exhibited significantly higher mean Goutallier grade on MRI (3.8 ± 0.4 vs 1.9 ± 1.1), mean muscle atrophy grade on MRI (2.0 ± 0.8 vs 0.5 ± 0.7), mean elasticity (44.15 ± 8.06 vs 30.84 ± 7.89 kPa), median elasticity (43.51 ± 7.73 vs 29.73 ± 7.61 kPa), mean elasticity ratio (3.66 ± 0.66 vs 1.83 ± 0.58), and mean gray-scale fatty infiltration grade (2.86 ± 0.36 vs 1.63 ± 0.66) and exhibited significantly lower occupation ratio on MRI (0.3 ± 0.1 vs 0.6 ± 0.1) (all,
p < .001).
Correlations Among Measures
Table 4 summarizes correlations among pairwise combinations of measures. Correlation coefficients with the three SWE measures ranged from 0.59 to 0.66 for Goutallier grade, –0.63 to –0.69 for occupation ratio, from 0.59 to 0.64 for muscle atrophy grade, and from 0.51 to 0.70 for gray-scale fatty infiltration grade. Absolute values of the correlation coefficients with the MRI measures ranged from 0.59 to 0.63 for mean elasticity, 0.59–0.63 for median elasticity, and 0.64–0.69 for elasticity ratio. The three SWE measures showed progressive stepwise increases with increasing Goutallier grade and increasing muscle atrophy grade (all,
p < .001, based on trend tests) (Table S2, which can be viewed in the electronic supplement to this article at
https://doi.org/10.2214/AJR.21.27129).
Diagnostic Performance for Insufficient Repair
Table 5 presents the diagnostic performance of the various methods for predicting insufficient repair, and
Figure 4 shows the ROC curves of the three SWE measures. The AUCs for MRI measures were 0.945 for Goutallier grade, 0.961 for occupation ratio, and 0.900 for muscle atrophy grade. The AUCs for SWE measures were 0.874 for mean elasticity, 0.888 for median elasticity, and 0.971 for elasticity ratio. The AUC for gray-scale fatty infiltration grade was 0.912.
Goutallier grade at a cutoff of 4 achieved sensitivity of 78.6% and specificity of 95.0%. Occupation ratio at a cutoff of 0.38 or less achieved sensitivity of 85.7% and specificity of 96.7%. Muscle atrophy grade at a cutoff of 2 or greater achieved sensitivity of 85.7% and specificity of 90.0%. Mean elasticity at a cutoff of 34.85 kPa or greater achieved sensitivity of 92.9% and specificity of 70.0%. Median elasticity at cutoff of 38.60 kPa or greater achieved sensitivity of 78.6% and specificity of 85.0%. Elasticity ratio at cutoff of 2.51 or greater achieved sensitivity of 100.0% and specificity of 90.0%. Gray-scale fatty infiltration grade at cutoff of 3 achieved sensitivity of 85.7% and specificity of 90.0%. Figures S1 and S2, which can be viewed in the electronic supplement to this article at
https://doi.org/10.2214/AJR.21.27129, show representative images with associated MRI and SWE measurements in a patient with sufficient repair and in a patient with insufficient repair, respectively.
Among the three SWE measures, elasticity ratio was selected for inclusion in the stepwise multivariable logistic regression analysis for predicting insufficient repair given its highest AUC. Significant independent predictors of insufficient repair were muscle atrophy grade on MRI of 2 or 3 (odds ratio [OR] = 9.3 [95% CI, 1.1–76.4]; p = .04) and elasticity ratio (OR = 15.7 [95% CI, 2.8–87.7]; p = .002). Large tear size, Goutallier grade on MRI of 3 or 4, occupation ratio, and gray-scale fatty infiltration grade of 3 were removed from the model by the stepwise selection process.
Discussion
In this study of patients who underwent preoperative SWE before rotator cuff repair, those with insufficient repair exhibited stiffer supraspinatus muscles than those with sufficient repairs. The SWE measures of the supraspinatus muscle were, in general, moderately correlated with various MRI-based measurements of muscle properties, indicating complementary information provided by the SWE measures. Further, the elasticity ratio on SWE significantly predicted an insufficient repair independent of the MRI measures and of gray-scale assessment of fatty infiltration. These findings support a potential role for SWE in prognostic assessment before rotator cuff repair.
The emergence of ultrasound elastography has enabled functional tissue evaluation by elasticity measurements. SWE has been applied for musculoskeletal evaluation, showing fair interobserver and intraobserver reliability [
19]. SWE evaluation provides both shear-wave velocity (SWV) and shear modulus. The SWV measures shear speed in meters per second, and shear modulus measures elasticity in kilopascals, reflecting both speed and tissue density. As tissue stiffness increases, the propagation speed increases, and all of these values increase [
20]. Nonetheless, a paucity of studies have investigated the utility of the shear modulus from SWE for supraspinatus muscle evaluation.
In a study by Itoigawa et al. [
7], preoperative SWE measures, preoperative Goutallier grade on MRI, and the stiffness of the supraspinatus musculotendinous unit measured during surgery were significantly higher in patients with incomplete repair than in those with compete repair. Correlations with surgically determined stiffness were higher for SWE measures than for the Goutallier grade. However, that study had a very small sample size of only 12 patients with complete repair and nine patients with incomplete repair and did not assess other measures, such as occupation ratio or muscle atrophy grade. In a cadaveric study by Giambini et al. [
33], the SWE elastic modulus exhibited a negative correlation with the experimentally measured extensibility of the supraspinatus muscle and a positive correlation with the muscle's fat fraction on MRI. The low extensibility of stiff muscles may contribute to an inability to cover the footprint in insufficient repairs, consistent with the association between stiffness and insufficient repair observed in the current study. Rosskopf et al. [
26] reported lower SWV of the supraspinatus muscle in patients with rotator cuff tear than in asymptomatic volunteers as well as lower SWV in patients with a positive tangent sign (i.e., in patients with muscular atrophy). However, SWV decreased with an increase in muscular fatty infiltration, indicating that SWV reflects tissue composition rather than stiffness. As muscular degeneration progresses with increasing chronicity of rotator cuff tears, fatty infiltration, fibrosis, and fibrosis increase. The fibrotic change includes collagen deposition, which is associated with an increase in extracellular matrix and in turn with increases in the elastic modulus and tissue stiffness [
34–
36]. Thus, although chronic muscular degeneration is associated with both fatty infiltration and fibrosis, the latter is believed to be responsible for the increased elasticity that is observed in torn muscles.
Among the SWE measures, the elasticity ratio exhibited the highest AUC, the strongest correlation with the MRI measures, and the highest interobserver and intraobserver agreement. Normal absolute reference ranges for measurements of the elasticity of musculoskeletal tissues have not yet been determined. However, the elasticity ratio provides a standardized SWE measurement, expressed as a ratio to a reference tissue in the given patient. The trapezius is located in the superficial layer of the supraspinatus muscle and can be readily evaluated by SWE [
37,
38], supporting its role as the reference tissue during supraspinatus assessment.
We also performed gray-scale ultrasound evaluation of the supraspinatus muscle. Although a prior study showed a positive correlation between SWV and the gray-scale fatty infiltration grade after rotator tendon tear [
35], we are unaware of prior studies comparing the gray-scale grade with SWE shear modulus measures. In this study, the gray-scale fatty infiltration grade was only moderately correlated with the SWE measures, indicating the distinct nature of the two approaches.
Among the MRI measures, the occupation ratio exhibited the highest interobserver agreement and the highest correlation coefficients with SWE measures. In comparison with the subjective Goutallier grade and muscle atrophy grade, the occupation ratio is an objective measure of muscle size. Specifically, with increasing chronicity of the injury and of associated muscular fibrosis, the muscle fibers shorten, leading to a lower occupation ratio [
6].
This study had limitations. First, the sample size was small, particularly in terms of patients with insufficient repair. It is possible that patients more likely to have insufficient repair (e.g., patients with large tears) were less likely to be scheduled for surgical repair. Second, all insufficient repairs were observed in patients with large tears. However, the size of the tear was determined surgically and thus was not known at the time of preoperative assessment; further, the tear size was not an independent predictor of insufficient repair in the multivariable model. Third, the attainment of successful repair may be affected by factors related to the surgeon. In this study, all repairs were performed by a single surgeon using a standard surgical method. Fourth, we measured SWE values with the patient in the supine position rather than in the sitting position as in some prior studies [
26,
34]. The potential decreased gravitational load on the musculature in the supine position is supported by a prior study reporting decreased tone, stiffness, and gravitational load of the trapezius muscle in the supine position compared with the sitting position [
37]. Future studies are warranted to assess for differences in SWE values between the two positions. Fifth, we assessed interobserver and intraobserver agreement solely in terms of ROI placement on the acquired images. We did not assess variability resulting from the actual performance of the SWE examinations, which may be a larger source of variability than the ROI placements. Finally, although SWE provides clinically relevant information about rotator cuff muscles and tendons, the technique does not provide information about bone, cartilage, or the labrum, which are also important to evaluate in patients undergoing arthroscopic rotator cuff repair. Therefore, SWE should be used as a complementary method in preoperative planning in patients with rotator cuff tear.
In conclusion, in patients with rotator cuff tears, SWE- and MRI-based measures of the supraspinatus muscle were moderately correlated. The SWE values were significantly higher in patients with insufficient rotator cuff repair than in those with sufficient rotator cuff repair. The elasticity ratio showed the highest AUC among the SWE measures in predicting insufficient repair and predicted insufficient repair independent of the MRI measures and of the gray-scale fatty infiltration grade. These findings indicate a complementary role for SWE as a prognostic marker during preoperative evaluation in patients with rotator cuff tear.