Key words
subscapularis tear - MRI - sensitivity - healthcare research
Introduction
In the German healthcare system, patients with shoulder problems often present initially
to a general practitioner or general orthopedic doctor with no special training regarding
the shoulder. Subscapularis (SSC) tendon tears are especially difficult to diagnose
for non-specialized doctors and are often missed early on, leading to a delay in therapy
[1]. Therefore, the written report from the radiologist is very important in terms of
referring patients to specialist shoulder surgeons. This is of even greater importance
since SSC tendon tears have been found to be present more often, with newer studies
reporting a prevalence between 27 % and 30 % in all shoulder arthroscopies and between
49 % and 59 % in arthroscopic rotator cuff surgery [2]
[3]
[4].
An overlooked SSC lesion shows a tendency for early retraction with progression to
muscle atrophy and fatty degeneration that makes a delayed refixation less likely
to be successful [5]
[6]. The loss of the strongest internal rotator at the front and the loss of the force
coupled with the infraspinatus (ISP) tendon are commonly believed to be catastrophic
for shoulder function and the long-term prognosis [5]
[7]. Salvage procedures, such as pectoralis major transfer (PMT) and anterior latissimus
dorsi (ALTD) transfer, are less likely to restore shoulder function and have a higher
risk for complication and failure of treatment. This underlines the importance of
a correct diagnosis in the first instance.
The aim of this study was to assess the usefulness of the written radiological report
provided to our center by a heterogeneous group of 39 general radiologic centers (37
without dedicated musculoskeletal-specialized radiologists (i. e., certificates indicating
a level of competence in musculoskeletal radiology) as a tool for patient guidance
and decision making regarding SSC tendon tears.
Materials and Methods
This retrospective study (Ethics Committee Approval is granted with number A 2013–0160)
includes 103 patients who underwent shoulder arthroscopy between April 2013 and January
2015 with intraoperative confirmation of SSC tendon tears in two centers by two experienced
consultant orthopedic surgeons (each with over 10 years of experience in shoulder
surgery). Tear size was intraoperatively classified according to Fox and Romeo ([Table 1]) [8] and additional lesions were noted. The intraoperative findings were then correlated
to the written radiological report. Exclusion criteria were time from magnetic resonance
imaging (MRI) to surgery of more than 180 days (five patients), missing written report
(one case), and previous surgery on the examined shoulder. A total of 97 patients
were included for analysis. The mean age was 62.4 years (range 39 to 81), with 63
men and 34 women. The right side was affected in 58 cases (56 %). All patients had
conventional MRI scans performed at a mean of 57.4 days (range 4–164) before arthroscopic
surgery. The written MRI reports by the radiologists were assessed with respect to
SSC, supraspinatus (SSP)/ISP tendon tears, and biceps tendon pathology and correlated
to the intraoperative findings. Calculation of the sensitivity for each type of lesion
was performed.
Table 1
Comparison of the sensitivity of MRI-positive SSC lesions depending on the size of
the defect in accordance with the Fox and Romeo classification. Type I 1: partial
thickness tear; type 2: complete tear of upper 25 % of tendon; type 3: complete tear
of upper 50 % of tendon; and type 4: complete rupture of tendon.
Tab. 1 Vergleich der Sensitivität von MRT-positiven SSC-Läsionen in Abhängigkeit von der
Größe des Defekts gemäß der Fox- und Romeo-Klassifikation. Typ 1: partieller Dickenriss;
Typ 2: vollständiger Riss der oberen 25 % der Sehne; Typ 3: vollständiger Riss der
oberen 50 % der Sehne; Typ 4: vollständiger Sehnenriss.
|
Fox/Romeo I
|
Fox/Romeo II
|
Fox/Romeo III
|
Fox/Romeo IV
|
arthroscopy (n = 97)
|
17
|
35
|
30
|
15
|
MRI-positive SSC lesion (n = 37)
|
5
|
7
|
14
|
11
|
sensitivity
|
5/17 (29.4 %)
|
7/35 (20 %)
|
14/30 (46.7 %)
|
11/15 (73.3 %)
|
MRI
Every included patient had a high-field strength (i. e., ≥ 1.5 T) standard MRI scan
that fulfilled the criteria established by the German Radiological Society (https://www.ag-bvb.drg.de/de-DE/3644/protokollempfehlungen/). No patient received contrast-enhanced arthrography, but in 14 MRIs, intravenous
contrast medium was applied. The examinations were performed in an outpatient setting
in different locations (39 radiological institutions) in northern Germany. Three of
them were associated with a communal hospital. The rest were either private centers
or part of a radiology network. Only two of them had certificates of musculoskeletal
specialization (DGMSR) listed in their reports or on their homepages. We analyzed
the written radiologic report of each patient and evaluated the interpretation of
the SSC and SSP tendons by the radiologist (i. e., complete and partial tears versus
intact tendon). There was no influence on the particular MRI protocols of each radiologist.
All of the MRI scans where a protocol was transmitted included T2-weighted coronal
oblique, axial, and sagittal oblique films and T1-weighted sagittal oblique films.
Arthroscopy
The arthroscopic examination was standardized by an inspection of the glenohumeral
joint via the posterior approach. The structural integrity and stability of the long
biceps tendon from its origin to the intertubercular course was checked. Inspection
of the joint cartilage and the labrum was followed by examination of the rotator cuff
tendons. The SSC tendon was tested by watching its movement during rotation. Visualization
was enhanced with a posterior lever push maneuver [7]
[8]. A (partially) bare lesser tuberosity is indicative of tearing. In cases of disturbing
tissue (i. e., scar or synovitis), careful debridement was performed in order not
to miss hidden lesions. SSC tendon tears were measured and then classified by the
same protocol using the Fox and Romeo classification [8].
The biceps pulley was described as damaged or intact. Thus, the co-incidence of pulley
lesions and SSC tendon tears could be determined. In addition, supraspinatus, ISP,
and biceps tendon lesions (partial and complete tears), as well as labrum lesions,
were recorded during arthroscopy ([Fig. 1]).
Fig. 1 Corresponding MRI (unenhanced proton density fat saturated sequences in axial orientation)
and intraoperative findings of SSC tears with different defect sizes. A Axial MRI view of a right shoulder with a luxation of the long head of the biceps
tendon and a partial tear of the subscapularis tendon (arrow) without relevant detachment
of the lesser tuberosity (LT). B Standard posterior viewing portal with 30° arthroscope showing a partial lesion (Fox/Romeo
I). C Axial MRI view of a right shoulder demonstrating a focal area of fluid signal intensity
of the superior subscapularis tendon attachment at the lower tuberosity (LT) (arrow)
consistent with a partial thickness tear. D Standard posterior viewing portal with 30° arthroscope showing a partial lesion (Fox/Romeo
II). E Axial MRI view of a right shoulder with a subtotal tear of the subscapularis tendon.
F Standard posterior viewing portal with 30° arthroscope showing a subtotal lesion
(Fox/Romeo III). G Axial MRI view of a right shoulder with a full-thickness tear of the subscapularis
tendon (arrow) and ruptured long head of biceps tendon. H Standard posterior viewing portal with 30° arthroscope confirming the complete lesion
of the SSC (Fox/Romeo IV). (MR images were mirrored for better orientation), (SSC = subscapularis
tendon; LHBT = long head of the biceps tendon; HH = humeral head; G = glenoid).
Abb. 1 Korrespondierende MRT (nichtverstärkte protonendichte fettsaturierte Sequenzen in
axialer Orientierung) und intraoperative Befunde von SSC-Rissen mit unterschiedlichen
Defektgrößen. A Axiale MRT-Ansicht einer rechten Schulter mit einer Luxation der langen Bizepssehne
und einer Teilruptur der Subscapularissehne (Pfeil) ohne relevante Ablösung des Tuberculum
minus (LT). B Standard-Betrachtungsportal von posterior mit 30°-Arthroskop, das eine partielle
SSC-Läsion zeigt (Fox/Romeo I). C Axiale MRT-Ansicht einer rechten Schulter, die ein fokales Flüssigkeitssignal des
oberen Subscapularissehnenansatzes am unteren Tubeculum minus (LT) (Pfeil) zeigt,
übereinstimmend mit einer Partialruptur. D Standard-Portal von posterior mit 30°-Arthroskop, das eine partielle SSC-Läsion zeigt
(Fox/Romeo II). E Axiale MRT-Ansicht einer rechten Schulter mit einem subtotalen Riss der Subscapularissehne.
F Standard-Betrachtungsportal von posterior mit 30°-Arthroskop, das eine subtotale
Läsion zeigt (Fox/Romeo III). G Axiale MRT-Ansicht einer rechten Schulter mit einer kompletten Ruptur der Subscapularissehne
(Pfeil) und rupturierter langer Bizepssehne. H Standard-Portal von posterior mit 30°-Arthroskop zur Bestätigung der vollständigen
Läsion des SSC (Fox/Romeo IV). Die MRT-Bilder wurden zur besseren Orientierung gespiegelt.
SSC = Subscapularissehne; LHBT = lange Bizepssehne; HH = Humeruskopf; G = Glenoid.
A tenotomy or tenodesis of the long biceps tendon was usually performed prior to SSC
therapy, depending on the age and functional demands of the patient. Afterwards, arthroscopic
repair of the SSC tendon was performed with one or two suture anchors, depending on
rupture size and preferred surgical technique. Additional rotator cuff tears were
addressed depending on their size and tear configuration.
Statistical analysis
Microsoft Excel (Microsoft, Redmond, Washington, USA) was used for data collection.
Continuous variables are presented as means, standard deviations (SDs), maximums and
minimums. Categorical variables are presented as percentages. The ANOVA test for linear
regression was used to calculate correlations. Statistical significance was set to
a p-value of < 0.05. Data was analyzed using SPSS statistics software version 23.0
(IBM, New York, USA).
Results
Arthroscopic findings
SSC tendon tears (n = 97) were classified according to the classification by Fox and
Romeo ([Table 1]): I, 17 patients (17.5 %); II, 35 patients (36.1 %); III, 30 patients (30.1 %);
and IV, 15 patients (15.5 %). Isolated SSC tendon tears were present in 24 cases (24.7 %).
Additional tears of the ISP or SSP tendon were observed in 73 cases (75.3 %). SSP
tendon partial or full-thickness tears were present in 68/97 patients (70.1 %). ISP
partial or full-thickness tendon tears were present in 22/97 patients (22.7 %) (no
isolated ISP tendon tears). The long head of the biceps (LHB) and pulley systems showed
a normal structure in only 6/97 cases (6.2 %). We detected isolated pulley lesions
in 39/97 cases (40.2 %) and combined with dislocation of the LHB tendon (LHBT) in
13/97 cases (13.4 %). Other LHBT lesions (54.6 %) included complete or partially torn
LHB tendons or tendinitis of the tendon. Among the sizes of the torn SSC tendons,
there was a rising co-incidence of pathologies of the LHBT ([Table 2]).
Table 2
Correlation of LHBT pathology with the size of the subscapularis tears in accordance
with the Fox and Romeo classification.
Tab. 2 Korrelation der Pathologie der langen Bicepssehne mit der Größe der Subskapularissehnenrisse
nach der Fox- und Romeo-Klassifikation.
|
Fox 1
|
Fox 2
|
Fox 3
|
Fox 4
|
pulley lesion
|
2/17 (11.8 %)
|
18/35 (51.4 %)
|
12/30 (40 %)
|
7/15 (46.7 %)
|
LHBT dislocation
|
1/17 (5.9 %)
|
0/35 (0 %)
|
9/30 (30 %)
|
3/15 (20 %)
|
LHBT partial lesion
|
5/17 (29.4 %)
|
8/35 (22.8 %)
|
10/30 (33.3 %)
|
3/15 (20 %)
|
LHBT complete lesion
|
4/17 (23.5 %)
|
2/35 (5.7 %)
|
6/30 (20 %)
|
5/15 (33.3 %)
|
LHBT tendinitis
|
2/17 (11.7 %)
|
6/35 (17.1 %)
|
1/30 (3.3 %)
|
1/15 (6.6 %)
|
MRI
We collected MRI findings from 39 different radiological centers. On average, 2.3
patients were examined by one center (range 1–21). For two centers, a certificate
concerning subspecialization in musculoskeletal imaging could be identified. In 89
cases, the MRI protocols were transmitted. Intravenous administration of contrast
medium was performed 14 times. Arthrographies were not used.
In order to obtain an indication of the precision of the examination results depending
on the clinical question submitted by the requester, the clinical information was
divided into the five following groups (double counting possible): suspected SSC tendon
tear (3/97), suspected SSP tendon tear 14(97), general rotator cuff lesion (46/97),
disorders of the biceps tendon (8/49), and trauma (14/97). The tear was identified
correctly by the radiologist in only 1 of 3 received suspected lesions of the SSC
tendon. Due to the very heterogeneous information provided by the referring physician,
we did not see any statistical significance with regard to a correlation with the
accuracy of the findings.
Preoperatively written MRI reports showed a frequency of SSC lesions of 37 of 97 for
SSP and of 61 of 97 for SSP/ISP. The correlation with the intraoperative findings
resulted in an overall low sensitivity of 38.1 % for the correct diagnosis of an SSC
tendon tear confirmed by arthroscopy. With increasing size of the SSC tendon tear,
the sensitivity increased (Fox and Romeo I 29.4 %, II 20 %, III 46.7 % and IV 73.3 %).
In contrast, the radiologists classified SSP and SSP/ISP tendon tears more often correctly.
From 68 arthroscopically proven tears, 60 tears were correctly identified in the written
report, which creates an overall sensitivity of 88.2 %. Only 1 of 29 MRI reports was
false-positive, which results in a specificity of 96.5 % ([Table 3]).
Table 3
Comparison of MRI and arthroscopic findings in diagnosing supraspinatus tears.
Tab. 3 Vergleich von MRT und arthroskopischen Befunden hinsichtlich der Diagnose von Supraspinatussehnenrissen.
|
ASK
|
|
|
supraspinatus
|
tear
|
no tear
|
total
|
MRI
|
tear
|
60 (61.8 %)
|
1 (1 %)
|
61 (62.8 %)
|
|
no tear
|
8 (8.2 %)
|
28 (28.7 %)
|
36 (37.1 %)
|
|
total
|
68 (70.1 %)
|
29 (29.9 %)
|
97 (100 %)
|
When comparing the mean intervals between MRI and surgical treatment, a small difference
was found depending on the size of the defect but without statistical significance.
(Fox and Romeo I: 73.16 d vs. 48.25 d, II: 60.75 d vs. 49.14 d, III: 60.9 d vs. 45.8 d,
and IV: 52.8 d vs. 51.2 d (not detected vs. detected)).
Discussion
The main finding of our study is the overall low sensitivity of the written MRI reports
for detecting SSC lesions, with a strong tendency to underdiagnose these lesions.
Delayed or missed treatment, especially of larger lesions, has detrimental effects
on the patient’s chances to restore normal shoulder function (as explained above),
as well as an additional impact in terms of sick leave and treatment costs for the
social system [5]. Clinical diagnosis of an SSC lesion is also difficult [9]
[10], highlighting the need for experienced MRI examinations of patients with SSC tendon
tears to get the correct diagnosis and timely further therapy.
MRI is the imaging modality of choice for the diagnosis of rotator cuff lesions. Nevertheless,
MRI does not always provide appropriate preoperative information, especially in the
case of tears involving less than half the cephalad-to-caudal width of the tendon
originating in the articulating face [11]. Garavaglia et al. also showed a clearly elevated frequency of arthroscopically
determined SSC lesions compared with the radiological report [12]. They showed a sensitivity of 37 % for MRI-documented SSC tendon tears, compared
to 38 % in our study. However, all 37 patients with preoperative MRI scans that were
interpreted by the radiologists as positive for SSC tendon tears were confirmed to
be positive by arthroscopy. This resulted in a sensitivity of 38.1 % and a specificity
of 100 %. This is comparable with other studies concerning the value of preoperative
MRI in the diagnosis of SSC tendon tears. For example, Adams et al. conducted a retrospective
study of patients undergoing shoulder arthroscopy in correlation with preoperative
MRI examinations (90 patients). They defined an SSC tendon tear when at least 20 %
of the craniocaudal length of the tendon insertion was involved. Their results showed
both 100 % specificity and positive predictive value, as well as a sensitivity of
36 %, negative predictive value of 62 %, and an accuracy of 69 %. Larger tears (at
least 50 % of the craniocaudal length) were more likely to be seen on MRI than smaller
tears (< 50 %) [9]. Foad and Wijdicks evaluated SSC tendon tears using MRI and MR arthrography and
reported relatively low sensitivities of 40 % and 36 %, respectively, which were not
different between these two techniques, and concluded that there was no advantage
for arthrography in the diagnosis of SSC tendon tears [13]. Other authors showed an accuracy of up to 84 % with the use of MRI arthrography.
However, it is more expensive and requires intraarticular injection [14].
Our data support the assumption that cranial partial ruptures are especially overlooked
with the standard MRI protocols. This low sensitivity arises because lesions of the
superior part of the SSC tendon insertion are visualized obliquely on transverse MR
images and in parallel on oblique sagittal MR images, which leads to distortion from
the partial-volume effect. Most SSC tendon tears start as disinsertion of the superior
border of the tendon and extend inferiorly [15]. This might be the reason for the higher rates for accuracy reported by Ryu et al.
using a sagittal oblique technique, with a sensitivity, specificity, and accuracy
of SSC tendon tear detection of 0.72, 0.77, and 0.75, respectively, for the radiologists
[16]. Additionally, radial slice magnetic resonance images reached sensitivity values
of 94.7 % and a specificity of 82.4 % and were promoted as useful for diagnosing these
lesions. In particular, the sensitivity for tears in the superior part of the SSC
tendon is higher than that of conventional methods [17]. In our study collective, none of the mentioned special sequences were used according
to the transmission of the study protocols in the written report. Concerning the tear
size, we could demonstrate an increasing sensitivity with increasing tear size (type
III 46.7 % and type IV 73.3 %). This underlines the fact that larger and full-thickness
tears of the SSC tendon were more frequently detected, which was comparable to other
studies [9]
[18]
[19]. However, still some Fox and Romeo IV lesions were missed. Therefore, we also looked
specifically at the time from MRI to arthroscopic diagnosis, and we found no difference
between correctly diagnosed and missed Fox and Romeo IV SSC lesions. Moreover, the
time from MRI to clinical confirmation and scheduling of an operation was generally
short (< 2 months) for these lesions.
Additionally, concurrent pathology of the LHBT can be assessed on axial MRI views.
SSC tendon tears are associated with partial rupture or displacement of the biceps
from the bicipital groove due to the frequent disruption of the coracohumeral ligament
attachment on the humerus at the medial aspect of the bicipital groove (“Pulley lesion”)
[20]
[21]. Therefore, newer studies promote the position of the LHBT in the intertubercular
sulcus or lesions of the pulley system as signs associated with SSC tendon tears,
which probably can increase the accuracy of detecting SSC tendon tears [22]
[23]. The integrity or disintegrity of the LHBT as a marker for severity of a torn SSC
tendon is emphasized by our results. However, as Shi et al. mentioned, the diagnostic
value of a subluxated LHBT in axial MRI scans lies primarily in its negative predictive
value. They pronounced, if there is no subluxation, it is unlikely that a full-thickness
tear of the SSC tendon is present. In our study, there were also high rates of LHBT
lesions. A subluxation often could not be determined due to older tears with complete
retraction of the biceps tendon. Thus, surgeons should be cautious about relying on
static biceps subluxation as a primary diagnostic tool for predicting SSC tendon tears
[22]. Here, clinical data regarding positive pain symptoms of the LHBT could direct the
focus accordingly. In our study, the referring physician noted an indication of biceps
pain with a comparatively low frequency (n = 8) compared to the long biceps tendons
treated during arthroscopy (n = 79). Thus, arthroscopy remains the gold standard for
identifying SSC tendon tears and LHBT co-pathologies [24].
In contrast to the low sensitivity of MRI reports for SSC lesions, the correct diagnosis
of SSP lesions in 60 cases was much better and acceptable. A meta-analysis of de Jesus
et al. emphasizes the relatively high sensitivity (88 %) of MRI in diagnosing SSP
tendon tears without any significant differences between MR arthrography, conventional
MRI, and ultrasound regarding full- and partial-thickness tears [25]. There are several studies concerning the accuracy for the detection of rotator
cuff lesions, with little attention to abnormalities of the SSC tendon [25]
[26]
[27].
For the future, further improvement of MRI techniques is needed. New protocols or
supporting MR series may be helpful to accentuate the ability to also detect small
SSC tendon tears and to be more concise with intraoperative findings. It appears important
to develop better diagnostic tools and agreement on classification systems for SSC
pathology that are universal, as well as easy to manage and reproducible. The main
goal of this study is to highlight the importance of specialized shoulder expertise
in the diagnosis of rotator cuff lesions, especially SSC lesions, which is often underrepresented
in a changing medical landscape. In addition, one may conclude that, with the initiatives
of several radiological societies to certify the expertise in musculoskeletal radiology,
the diagnostic performance may be better when MRI reports are obtained by a group
of radiologists with certified expertise in musculoskeletal radiology. However, since
this was not the initial goal of our study and only two radiological centers were
identified with musculoskeletal specialization, we cannot prove this with our data.
This should be demonstrated in another study.
Study limitations
We are aware that some variables may affect the results since MRI investigations were
performed by different outpatient radiologists with varying levels of experience in
musculoskeletal imaging and differing imaging protocols. The time interval (in days)
from MRI scan to arthroscopic surgery was limited to 180 days. Progression of the
tendon lesion is conceivable with a longer interval between MRI and surgical treatment.
However, a comparison of the median intervals between detected and undetected ruptures
shows a statistically insignificant difference. Thus, the time elapsed between MRI
and arthroscopic treatment is not sufficient to explain the low sensitivity. Further
limitations are the retrospective cohort study design, with a potential for overreporting
bias. Including a large number of different reporting outpatient radiologists (39
radiologic institutions) is a further limitation, because of varying experience and
skill levels in the interpretation of MRIs. However, this reflects the current practice
in the German health care system and other systems worldwide and thus represents the
real clinical quality of care for our patients outside a dedicated musculoskeletal
center where every patient received imaging and therapy within this center.
Conclusion
Preoperative MRI and interpretation by a heterogeneous group of general (presumably
non-MSK-specialized) radiologic centers do not reliably detect SSC tendon tears and
are not sufficient for guiding patients to specialist centers, especially in the case
of partial-thickness tears. However, larger and full-thickness SSC tendon tears were
also overlooked. Often technical difficulties are the explanation. Therefore, the
radiologist’s report alone cannot be used as the main tool in terms of patient referral.
This study punctuates the need for the experienced examination of patients with SSC
tendon tears by radiologists to get the correct diagnosis and timely further therapy.
Future studies must show the extent to which specialization in musculoskeletal imaging
leads to improved findings.