Meta-Analytic Review of Ultrasound and Magnetic Resonance Imaging
for Diagnosing Tendon Tears of the Foot and Ankle
Adam Fleischer, DPM, MPH, Sarah Dickey, DPM, Maureen Allanson, BS, Jessica Minder, BS,
Daniel Evans, DPM, Martin Yorath, DPM
Department of Surgery, Advocate Illinois Masonic Medical Center
Abstract:
Introduction: Recent evidence suggests that musculoskeletal ultrasound (US) may
offer similar, if not greater, accuracy for detecting tendon tears when compared
to magnetic resonance imaging (MRI). Our aim was to better characterize how
US and MRI perform in patients with clinically suspected tears of the foot or
ankle by deriving summary estimates of sensitivity, specificity, and diagnostic
accuracy from the published literature.
Methods: We undertook meta-analysis of studies comparing US and/or MRI to the
reference standard of intra-operative inspection in patients suspected of having
a foot or ankle tendon tear. Medline, EMBASE and citation lists from 1980 to
July 2010 were searched. Random effects meta-analytic techniques were used
to derive pooled estimates of sensitivity, specificity and diagnostic odds ratio
(DOR) for both MRI and US.
Results: We identified 28 studies (824 tendons) comparing ultrasound (423
tendons) and/or high-field MRI (401 tendons) to operative inspection in patients
with suspected foot or ankle tendon injuries. Overall sensitivity, specificity and
DOR for US (95% confidence interval) was 93.2% (88.6 to 96.0), 89.8% (82.8 to
94.1), and 71.9 (27.2 to 189), respectively. Overall sensitivity, specificity and
DOR for MRI was 91.0% (80.6 to 96.1), 81.3% (59.2 to 92.8), and 33.8 (12.0 to
95.0), respectively. However, MRI studies lacked homogeneity (Chi-square test
for heterogeneity p < 0.05 for sensitivity and specificity calculations), suggesting
that some pooled estimates should be interpreted with caution.
Conclusion: US is a reliable test for tendon tears of the foot and ankle with greater
reported accuracy than MRI. The high accuracy and lower cost supports using
US as the first-line imaging modality in suspected, isolated tendon tears in this
region of the body.
Table 1. Studies Utilizing Magnetic Resonance Imaging to Diagnose Foot and
Ankle Tendon Tears
Source
n
Bonnin M
1997
8
Brandes CB
2000
12
Conti S 1992 20
DiGiovanni
BF 2000 122
Karjalainen
PT 2000
25
Level of
Evidence
Study
Type
Consecutive
Enrollment
Mean Age of Pts
(range)
M/F
Results:
Table 2. Studies Utilizing Sonography to Diagnose
Foot and Ankle Tendon Tears
Dedicated
Prevalence Magnet Extremity
Tendon of Interest of Tears, % Strength, T Coil Used?
Blinded
Level of Study
Evidence Type
Consecutive
Enrollment
Mean Age
of Pts
(range)
M/F
Tendon of
Interest
Prevalence of
Tears, %
Probe
Frequency
Transducer Type
Blinded
Source
n
Astrom M
1996
26
II
P
Yes
44 (21-68)
22:5
Achilles
15.4
5 or 5-7
convex or linear
array
Yes
II
P
NS
32 (18-63)
10:8
PB
100.0
NS
NS
NS
II
NS
NS
55 (30-75)
5:7
PL
58.3
NS
NS
NS
Chen Y 1997 14
II
P
Yes
52 (24-68)
3:11
PTT
100.0
10
NS
No
II
R
Yes
53.3 (26-70)
6:13
PTT
15.0
1.5
NS
Yes
Chiou H 1999 22
II
P
Yes
NS
NS
Achilles
100.0
10
linear
NS
II
R
Yes
35 (NS)
NS
PB, PL
13.9
NS
NS
NS
66
II
P
Yes
45.2 (18-72)
36:22
PB, PL
37.9
11, 12, or 15
NS
NS
III
P
No
33.0 (15-28)
75:25
Achilles
84.0
1.5
100%
NS
Hartgerink P
2001
26
II
R
Yes
40 (14-61)
17:9
Achilles
53.8
7.5 - 12
linear array
Yes
10
linear array
NS
Grant TH
2005
Khoury NJ
1996 AJR
11
III
R
No
53.5 (NS)
1:16
PTT
54.5
1.5
NS
NS
Hsu T 1997
16
II
P
NS
55 (NS)
5:11
PTT
43.8
Khoury NJ
1996 Rad
24
III
R
No
35.1 (17-68)
5:7
PB, PL
54.2
1.5
NS
NS
Kainberger
FM 1990
18
I
P
Yes
38.0 (13-60)
54:19
Achilles
100.0
Kalebo P
1992
37
III
P
No
35.0 (NS)
19:11
Achilles
100.0
7.5 MHz
linear array
NS
Mathieson JR
1988
6
III
P
No
30.0 (13-60)
13:60
Achilles
50.0
5.0 MHz
linear array
Yes
Miller SD
1996
17
II
R
Yes
53.1 (31-78)
3:14
PTT
82.4
7.5 MHz
linear array
NS
Neuhold A
1992
13
III
P
No
41.7 (NS)
21:7
Achilles
92.3
7.5 or 10.0 MHz
linear
NS
Neustadter J
2004
24
II
P
Yes
30.4 (16-66)
3:10
PB, PL
25.0
12 MHz or 13
MHz
linear array
NS
Lamm BM
2004
32
III
R
No
40.4 (13-64)
10:22
PB, PL
75.0
1.5
100%
Yes
Mengiardi B
2005
11
Neuhold A
1992
13
II
R
Yes
63.2 (30-82)
8:20
ATT
91.0
1.5
100%
NS
III
P
No
41.7 (NS)
21:7
Achilles
100.0
0.5
0%
NS
Rockett MS
1998
38
III
P
No
46.5 (21-78)
8:20
FDL, PB, PL,PTT,
44.7
1.5
100%
NS
5.0-10.0MHz linear-phased array
Rosenberg
ZS 1988
22
III
R
No
NS (43-70)
8:19
PTT
91.0
1.5
90%
No
Rosenberg
ZS 1997
7
III
R
No
36 (21-77)
14:13
PB
100.0
1.5
100%
No
Sammarco
GL 1995
20
II
P
No
46.5 (31-63)
10/4
PB, PL
80.0
NS
NS
No
Paavola M
1998
66
III
R
No
42.0 (11.088.0)
60:19
Achilles
56.1
5-10 MHz
linear array
NS
Schweitzer
ME 1997
36
II
R
NS
40 (10-70)
15:27
PL, PB
40.0
1.5
100%
Yes
Raikin SM
2008
28
II
P
Yes
34.0 (15-53)
0.0
PB, PL
17.9
"high
frequency"
linear array
Yes
Rockett MS
1998
54
III
P
No
46.5 (21-78)
8:20
FDL, PB,
PL,PTT,
44.4
7.5 Mhz,
10MHz
linear array, broad
band linear array
NS
Abbreviations: ATT, Anterior Tibial Tendon, PB, Peroneus Brevis, PL, Peroneus Longus, PTT, Posterior Tibial Tendon, FDL, Flexor
Digitorum Longus, n, number of tendons included, NS, Not Specified, P, Prospective, R, Retrospective
Introduction:
Current opinion generally regards MRI as the gold standard for assessing soft tissue
pathology of the foot and ankle. However, recent evidence suggests that
musculoskeletal US affords comparable, if not greater, accuracy than MRI for
detecting and characterizing tendon tears within the foot and ankle. With only a
handful of studies directly comparing ultrasound performance with the criterion gold
standard (i.e., operative inspection), and even fewer studies examining the
diagnostic performance of MRI and US head-to-head, third party payers in the United
States remain reluctant to reimburse for in-office use of musculoskeletal US despite
its growing popularity and dramatic potential for cost-savings.
In this study, we undertook a systematic review and meta-analysis of Englishlanguage studies evaluating US and MRI findings with respect to operative
inspection for patients with suspected foot or ankle tendon injury. The aim of the
project was to better compare how US and MRI perform in diagnosing tears by
deriving definitive estimates of sensitivity, specificity, and diagnostic accuracy.
Methods:
We sought to identify all studies of patients with a clinically suspected tendon tear or
rupture who underwent testing with US and/or MRI followed by the reference
standard of intra-operative inspection (Figure 1). We searched Medline and
EMBASE (1980 to July 2010) for original articles containing the following words
“tendon tear” and “MRI” or “ultrasound” or “sonogram”. Three reviewers (SD, MA,
JM) screened the titles and abstracts of all articles to identify a list of potentially
relevant articles. We excluded studies that used a reference standard other than
intra-operative examination; case series and studies with less than five patients;
articles whose data could not be abstracted into 2x2 tables; non-original studies; and
studies examining low-field extremity MRI units (Figure 2). The references of
included articles were also scanned for additional articles. For each included article,
data was then extracted into 2x2 tables and the study characteristics recorded by two
independent raters (Tables 1 and 2). Discrepancies were checked and resolved by a
third independent reviewer (AF).
Random effects models were used to estimate overall sensitivity, specificity, and
DOR for both MRI and US studies. Summary ROC curves were also examined for
US and MRI. Chi-square test or Cochran-Q test was used to test for heterogeneity.
All analyses were performed using SAS version 9.2 for Windows.
NS
b
a
MRI
US studies were generally of higher quality than those looking at MRI performance.
Blinding of the observers to the presence of tendon pathology occurred in just 64%
(7/11) of all studies that reported blinding, and the majority of studies did not indicate
blinding status (19/28, 68%) at all. The high overall prevalence of tendon tears
suggests a slight selection bias for both MRI and US studies.
Figures 3 and 4 show the Forest plots of sensitivity and specificity, respectively, for
US and MRI. Point estimates of sensitivity and specificity are plotted, with 95%
confidence intervals, for each study. Pooled sensitivity (95% CI) for US in detecting
tendon tears was 93.2% (88.6 to 96.0), whereas pooled sensitivity for MRI was
91.0% (80.6 to 96.1). Pooled specificity for US was 89.8% (82.8 to 94.1), while
pooled specificity for MRI was 81.3% (59.2 to 92.8). Lower than expected estimates
and wider confidence intervals were observed for studies with multiple zero cells
where a 0.5 correction was needed. Chi-square test for heterogeneity for sensitivity
and specificity calculations among MRI studies was statistically significant (p<0.05),
indicating that these estimates should be interpreted with caution.
Figure 5 shows the Forest plot of summary DOR (95% CI) for US (71.9 [27.2 to 189])
and MRI (33.8 [12.0-95.0]) in the diagnosis of tendon tears of the foot and ankle. US
clearly performed better in the diagnosis of tendon tears, with a DOR almost twice
that of MRI. The Cochran Q test was also insignificant for both groups, suggesting
appropriate enough homogeneity to report our summary estimates. Figure 6 shows
the summary receiver operating characteristic (SROC) curves of US and MRI. The
area under the curve (AUC) for US studies is greater than that for MRI studies, again
suggesting a better overall accuracy for US.
Figure 3.
Forest Plot
Sensitivity.
c
The flow of articles is outlined in Figure 1. We scanned 1453 titles/abstracts and
selected 99 potentially relevant articles for retrieval. Review of the full articles
identified 17 that met the inclusion criteria. Review of the bibliographies of the
selected articles identified 11 additional articles for inclusion. Two articles duplicated
data published elsewhere and were excluded. One study was excluded where the
radiologists were aware of the operative diagnosis of a tendon tear. We were unable
to extract or analyze appropriate data from 5 articles. Two articles directly compared
MRI to US and could be used in both analyses. The meta-analysis therefore
included a total of 28 studies—423 tendons evaluated with US and 401 tendons
evaluated with MRI.
US
Figure 6. SROC Curves.
Figure 1. Longitudinal “split tear” of the peroneus brevis tendon
(PB) as seen on (a) MRI, (b) US, and (c) intra-operative inspection.
Figure 4.
Forest Plot
Specificity.
Search for articles on
MEDLINE and EMBASE
and via Hand Search
(n=1453)
Excluded (n=1354)
Lack of relevance (n=1354)
Full text articles
reviewed by authors
(n=99)
MRI
Excluded (n=82) Exclusion Criteria:
1. Does not use MRI/US to evaluate tendon
tear (n= 4) .
2. Lacks intra-op gold standard (n=26)
3. Outcome of interest not evaluated (n=28)
4. Data cannot be extracted (n=5).
5. Case study n<5 (n=14)
6. Observer Not Blinded (n=1)
7. Not original study (n=2)
8. Low field MRI (n=2)
Eligible articles (n=17)
Additional artlces
obtained by reference
search (n=327)
Excluded (n=316): Lack of relevance
Total eligible reference
search articles (n=11)
Total eligible included
in analysis (n=28)
Figure 2. Search results.
Figure 5.
Forest Plot of
Diagnostic
Odds Ratios
(DOR).
US
MRI
US
AUC = 0.925
AUC = 0.958
Conclusions:
MRI
US
• Our study confirms that US is a strong test for both including and excluding the
diagnosis of underlying tendon tear in this region of the body—with extremely high
pooled sensitivity (0.93) and pooled specificity (0.90) values.
• Our study also suggests that US may be more accurate than MRI in diagnosing
tendon tears of the foot and ankle (DOR 71.9 vs. 33.8).
Pooled DOR = 33.8
Pooled DOR = 71.9
10
100
100
10
• A higher accuracy and lower cost would clearly favor using US as the first-line
imaging modality for suspected isolated tendon tears of the foot or ankle.
• Formal decision modeling will be needed to fully characterize the place of MRI and
US in the diagnostic algorithm of tendon tears of the foot and ankle.