Downloaded from www.ajronline.org by 60.48.26.21 on 10/16/15 from IP address 60.48.26.21. Copyright ARRS. For personal use only; all rights reserved
i 21
Rotator
Cuff
Tears:
Diagnosis
with
Sonography
C. Whitley Sidney
Vick1
The
purpose of this study was to determine the accuracy of sonography in the of rotator cuff tears. Eighty-one patients were referred by orthopedic specialists because of a clinical suspicion of rotator cuff tear. The standards of comparison were arthrography in 79 cases and surgery in two cases. The sonographic technique used was based on a review of the literature and experience gained by scanning normal
A. Bell2
detection
subjects.
The sonographic
criteria
for the diagnosis
of a complete
rotator
cuff tear were
a focal defect in the cuff or complete absence or nonvisualization of the cuff. All sonograms were interpreted prospectively without knowledge of arthrographic or surgical results. With arthrography as the standard of comparison for the diagnosis of rotator cuff tear, sonographic results included 15 true-positives, 52 true-negatives, eight false-negatives, and four false-positives. With surgery as the standard of comparison, the results were one true-positive and one true-negative sonogram. The sensitivity of sonography in detecting rotator cuff tear was 0.67, the specificity was 0.93, and the accuracy was 0.85. Our study found lower sensitivity and accuracy results for shoulder sonography than have been previously reported. 154:121-123,
AJR
Authors
January
of several
1990
recent
the use of sonography [1 2] have found the accuracy of shoulder sonography to be 0.87-0.94, the sensitivity 0.93, and the specificity 0.83-0.93. The procedure is appealing because, unlike shoulder arthrography, there is no discomfort to the patient, radiation exposure, or potential for adverse reactions. We describe our findings in using sonography to detect rotator
for the diagnosis
of tears
cuff
patients
tears
in 81
studies
of the
[i -4]
rotator
in whom
available
for use as the diagnostic
Patients
and Methods
have
cuff.
the
advocated
Prior
results
investigations
,
of arthrography
or
surgery
were
standard.
One hundred fifteen consecutive patients were referred for shoulder sonography. Of these, 79 who underwent arthrography and two who had surgery without arthrography were included in the study. All patients were referred by a group of six orthopedic surgeons because of a clinical suspicion of rotator cuff tear. Sonograms were obtained by using a 5-MHz linear phased-array transducer (Acuson 128, Received June 6, 1989; accepted after revision September 1
8, 1989.
Rome Radiology
Blvd., Rome, GA 30161 to C. W. Vick.
0361-803x/90/1541-0121 © American
Roentgen
View, used
,
humerus
adducted
to view
based
were
on a review
the fibers
in neutral
groove was located the rotator
Ray Society
CA). The procedures was
performed
of the
and interpreted
literature
and
by one radiologist.
experience
gained
normal subjects [1 , 2, 5-7]. Verbal consent was obtained in all cases. Both shoulders were scanned. The examination was done with the patient
Group, 1104 Martha Berry . Address reprint requests
2 The Harbin Clinic, 1825 Martha Berry Blvd. Rome, GA 30161 .
Mountain technique
supraspinatus
cuff.
The
examination
began
with
the
seated
transducer
of the rotator cuff in a transverse or cross-sectional plane. and used as a landmark for identification of the component
The transducer and
position.
infraspinatus
was tendons
then
moved from
their
laterally insertion
and posteriorly on
the
humerus
The
by scanning
with the oriented
The bicipital tendons of
to visualize to
the
the point
122
VICK
AND
where they became obscured by the acromion. The transducer was then brought back to the bicipital groove and then moved medially and superiorly to view the subscapularis tendon. The same areas were then scanned in the longitudinal plane. The sequence was repeated both with the shoulder elevated and with the arm pulled behind the patient’s back in a position of hyperextended internal
Downloaded from www.ajronline.org by 60.48.26.21 on 10/16/15 from IP address 60.48.26.21. Copyright ARRS. For personal use only; all rights reserved
rotation.
Sonographic criteria for the diagnosis of a complete tear were a focal defect in the cuff or complete absence or nonvisualization of the rotator cuff. The sonograms were interpreted prospectively before arthrography or surgery as showing either the presence or the absence of a complete tear. A decision regarding tear vs no tear was made in each case, with no attempt to exclude or tabulate indeterminate cases. Sensitivity, specificity, and accuracy were determined. Subsequently, the sonograms were reviewed retrospectively to as-
sess possible causes of errors. Single-contrast arthrography was performed after the sonography on the same day. Knowledge of the results of the sonograms was usually available to the radiologist before arthrography was performed. The contrast agent used was either methylglucamine diatri-
zoate (Renografin,
Squibb, New Brunswick,
NJ) or iopamidol
(Isovue,
Squibb). Approximately 10-15 ml of contrast agent was injected into the joint under fluoroscopic guidance. Fluoroscopic spot films were made at the discretion of the radiologist, usually if contrast material was observed entering the subacromial bursa. After injection of contrast material, anteroposterior radiographs were made in internal and external rotation and an axillary view was made. If no gross abnormality was seen on these films, a second set of films was made after exercise. A full-thickness rotator cuff tear was diagnosed when
contrast
material entered the subacromial
bursa.
An arthrography, 23 patients had tears of the rotator cuff and 56 patients had no tear. At surgery, one patient had a rotator cuff tear and one patient had an intact cuff.
Results Using rotator
arthrography cuff
tears,
as the we found
1 5 true-positives, 52 and four false-positives. arthrography. true-positive;
the
standard
for the diagnosis
sonographic
true-negatives, Two patients
In one of these in the other, the
results
eight had
patients, sonogram
of
included
false-negatives, surgery without
the sonogram was was true-negative.
The sensitivity of sonography in detecting rotator cuff tear was 0.67, the specificity was 0.93, and the accuracy was 0.85. Retrospective analysis of all 1 6 of the true-positive cases showed that in 1 0 (63%), the sonographic finding was a focal defect in the rotator cuff, and in six (38%), the cuff was not visualized. Review of the eight false-negative cases showed that
four
thinning
had of the
no
sonographic
posterior
abnormality,
aspect
of the
one cuff
not
had
focal
Discussion The ences
sensitivity than
in study
dependent,
and
previously
accuracy
of sonography
reported.
This
the
that
design,
and technical
fact
limitations.
may
in our
be related
sonography
study to differ-
is operator
is
AJR:154,
January
1990
We believe the design of our study was comparable with those of studies reported previously [1 2]. Scanning technique and sonographic equipment used were the same or comparable. The sonographic criteria for rotator cuff tear were the same as those used by Mack et al. [2]. In both studies, visualization of a defect in the cuff was required to diagnose a tear. However, Middleton et al. [1] and Crass et al. [3] also include the finding of a central echogenic band or focus in the cuff as a sign of tear. Use of more liberal criteria such as these will increase sensitivity and lower specificity. In retrospect, three false-negative cases in our study had echogenic foci in the cuff, but 1 4 true-negative cases also had a similar-appearing echogenic area in the cuff. If we had included the criterion of a focal echogenic area in the cuff as a sign of tear, our sensitivity would have improved to 0.79, but specificity and accuracy would have dropped to 0.68 and 0.72, respectively. These figures are still well below those of earlier studies. Arthrography was used as the standard of comparison in ,
the majority
of our
cases
and
in previous
comparable
reports
[1 2]. It is recognized that errors may occur with arthrography, most notably failure to detect partial rotator cuff tears that do not communicate with the joint space. However, the majority of our sonographic errors involved failure to detect tears shown by arthrography. In this situation, contrast agent enters the subacromial bursa, and the likelihood that arthrography is in error is exceedingly low. Therefore we believe that using arthrography as the standard of comparison does not substantially affect our results or comparison with previous studies. Shoulder sonography is highly operator dependent. We cannot rule out the possibility that individual differences in technique or skill account for our less favorable results cornpared with those of previous reports. We believe that technical limitations of shoulder sonography contribute to its inability to identify some rotator cuff tears. Tears occurring in that portion of the cuff behind the acromion cannot be visualized [1 -4, 7, 8]. However, most tears occur near the insertion of the supraspinatus tendon, an area that is routinely seen with sonography. In this location, the size of a tear may affect sonographic visualization. Mack et al. [2] showed that some small (