Download Evaluating the utility of nonechoplanar diffusionweighted

The Laryngoscope
C 2012 The American Laryngological,
V
Rhinological and Otological Society, Inc.
Evaluating the Utility of Non–Echo-Planar Diffusion-Weighted
Imaging in the Preoperative Evaluation of Cholesteatoma:
A Meta-analysis
Peter M. M. C. Li, MD; Eleni Linos, MD, DrPH; Richard K. Gurgel, MD; Nancy J. Fischbein, MD;
Nikolas H. Blevins, MD
Objectives/Hypothesis: To describe the accuracy of non–echo-planar diffusion-weighted magnetic resonance imaging
(DW MRI) in identifying middle ear cholesteatoma.
Study Design: A meta-analysis of the published literature.
Methods: A systematic review of the literature was performed to identify studies in which patients suspected of having
middle ear cholesteatoma underwent DW MRI scans prior to surgery. A meta-analysis of the included studies was performed.
Results: Ten published articles (342 patients) met inclusion criteria. Cholesteatoma was confirmed in 234 patients, of
which 204 were detected by DW MRI (true positives) and 30 were not (false negatives). One hundred eight patients did
not have cholesteatoma on surgical examination, and of these 100 were correctly identified by MRI (true negatives) whereas
eight were not (false positives). The overall sensitivity of DW MRI in detecting cholesteatoma was 0.94 (confidence interval,
0.80–0.98) and specificity 0.94 (confidence interval, 0.85–0.98). DW MRI sequences could not reliably detect cholesteatomas
under 3 mm in size.
Conclusions: Non–echo-planar DW MRI is highly sensitive and specific in identifying middle ear cholesteatoma. DW
MRI may help to stratify patients into groups of who would benefit from early second-look surgery and those who could be
closely observed.
Key Words: Cholesteatoma, magnetic resonance imaging, computed tomography, half-Fourier acquisition single-shot
turbo spin-echo imaging, periodically rotated overlapping parallel lines with enhanced reconstruction sequence, metaanalysis, non, echo-planar diffusion-weighted.
Level of Evidence: 2a.
Laryngoscope, 123:1247–1250, 2013
INTRODUCTION
Traditionally, computed tomography (CT) has been
considered the imaging modality of choice for cholesteatoma because of its excellent spatial resolution and
delineation of key anatomical landmarks. A major disadvantage of CT, however, is its low specificity in
differentiating cholesteatoma from other soft tissue (fibrosis, granulation tissue) or fluid whenever the middle
ear and mastoid cavity appear partially or completely
opacified.
Magnetic resonance imaging (MRI) is becoming
increasingly used as a tool for cholesteatoma diagnosis.
In particular, non–echo-planar–based diffusion-weighted
From the Department of Otolaryngology–Head and Neck Surgery
(P.M.M.C.L., R.K.G., N.J.F., N.H.B.), and Department of Radiology (N.J.F.),
Stanford University, Stanford, California; and the Department of
Dermatology ( E . L .), University of California San Francisco, San
Francisco, California, U.S.A.
Editor’s Note: This Manuscript was accepted for publication
September 5, 2012.
The authors have no funding, financial relationships, or conflicts
of interest to disclose.
Send correspondence to Peter M. M. C. Li, MD, Department of
Otolaryngology–Head and Neck Surgery, Stanford University, 801 Welch
Road, Stanford, CA, 94305-5739. E-mail: [email protected]
DOI: 10.1002/lary.23759
Laryngoscope 123: May 2013
magnetic resonance imaging (DW MRI) techniques have
been shown to have high reliability in detecting cholesteatoma with good interobserver agreement.1 These
techniques have the advantage of being rapid, convenient
and do not require the injection of contrast material. It
has been suggested that the use of DW MRI may reduce
or even obviate the need for second-look surgery.1
Detection of residual or recurrent cholesteatoma
can be a diagnostic challenge. At present, second look
surgery remains the only means of obtaining definitive
diagnosis. As a result, patients who do not have recurrent cholesteatoma may undergo unnecessary surgery.
Similarly, patients with residual disease may have their
diagnosis delayed in an attempt to avoid an additional
operative intervention. To address these issues, new
imaging techniques have been developed to accurately
and noninvasively detect cholesteatoma within the
middle ear and mastoid.
Several studies to date have explored the diagnostic
accuracy of DW MRI, most from single centers with a
small number of patients. The purpose of this study was
to perform a systematic review of the established literature in this field and to perform a meta-analysis of the
studies evaluating the accuracy of non–echo-planar diffusion- weighted MRI techniques in correctly identifying
cholesteatoma.
Li et al.: DW MRI in cholesteatoma diagnosis
1247
Laryngoscope 123: May 2013
1248
The size limit represents the smallest size of cholesteatoma detectable by DW MRI. All studies utilized non–echo-planar diffusion-weighted MRI, but two studies did not describe the specific sequence used.
DW ¼ diffusion-weighted; EP ¼ echo-planar; HASTE ¼ half-Fourier acquisition single-shot turbo spin-echo imaging; MRI ¼ magnetic resonance imaging; NA ¼ not available; NPV ¼ negative predictive
value; PPV ¼ positive predictive value; PROPELLER ¼ periodically rotated overlapping parallel lines with enhanced reconstruction sequence.
93
100
91
100
5
Revision
Adults and children
24
2
France
2006
1.5 T, Non-EP DW MRI
100
100
100
100
100
3
Revision
Adults and children
23
1
Australia
2009
Dephnorrarat
et al.
Dubrulle et al.
1.5 T, HASTE, 3 mm
100
100
100
100
100
100
100
3
Revision
3
Children
Adults and children
10
15
1
1.5 T, HASTE, 3 mm
3
France
2009
Lehmann et al.
Australia
2010
Rajan et al.
3 T, PROPELLER, 3 mm
100
100
75
100
1
100
92
100
3
2
Revision
Primary and
revision
Revision
Adults and children
Adults
32
26
1
2
1.5 T, HASTE, 3 mm
3 T, HASTE, 3 mm
2010
2010
Huins et al.
Pizzini et al.
France
2010
Plouin-Gaudon
et al.
United Kingdom
Italy
58
89
88
62
3
Revision
Children
21
1
Adults and children
120
4
1.5 T, Non-EP DW MRI,
2 mm
Belgium
2010
De Foer et al.
1.5 T, HASTE, 3 mm
57
96
87
83
NA
96
83
64
93
90
71
82
96
2
3
Revision
Primary and
revision
Primary and
revision
NA
Adults
38
33
1
1
NPV
%
Specificity
%
Sensitivity
%
Size Limit,
mm
Type of
Procedure
Population
Patient
No.
Radiologists
No.
MRI, Sequence,
Slice Thickness
1.5 T, HASTE, 2 mm
1.5 T, HASTE, 3 mm
United Kingdom
Slovak Republic
2011
2011
Khemani et al.
Profant et al.
Fourteen studies met inclusion criteria. Of these,
one study was excluded due to insufficient reported
data.5 Three further studies were excluded to restrict the
analysis to original data.6,7 The methodological quality of
all remaining studies was appropriate for inclusion. Studies used prospective designs and enrolled consecutive
patients with suspected cholesteatoma. All studies used
surgery as the gold standard for diagnostic confirmation.
The final analysis included 10 studies and 342
patients (Table I).8–17 Cholesteatoma was confirmed
intraoperatively in 234 patients, in whom 204
Country
RESULTS
Publication
Year
Traditional meta-analysis methods are insufficient for
summarizing studies of diagnostic accuracy because these
studies typically include two measures, such as sensitivity and
specificity, which are inversely correlated. In addition, betweenstudy heterogeneity is common in studies of diagnostic accuracy,
and models must account for this. Therefore, new methods for
meta-analysis of diagnostic studies have been developed.2,3 We
used bivariate meta-analysis and the hierarchical summary receiver operating characteristic model.4 We calculated the total
TPs, FPs, TNs, and FNs of all patients included in each study.
Using the metandi command (STATA version 11; StataCorp LP,
College Station, TX), the summary sensitivity and specificity of
the test, including 95% CIs for each value were calculated. In
addition, we calculated and plotted the 95% confidence regions
and 95% prediction regions around the summary accuracy estimate. To test for heterogeneity, we also performed separate
forest plots for sensitivity and specificity, and calculated the I2
statistic and P value. Overall, there was no significant evidence
of heterogeneity (I2 ¼ 11.2%, P ¼ .34 for sensitivity and I2 ¼
26.6%, P ¼ .18 for specificity). To assess for publication bias,
funnel plots were created by plotting the effect size found by
each study against the inverse of its standard error. These were
symmetric on visual inspection and on formal testing with
Begg’s test (P ¼ .11), suggesting no significant publication bias.
Authors
Statistical Methods
TABLE I.
Description of Studies Included in the Meta-analysis.
Two independent investigators (P.L., R.G.) searched MEDLINE and Web of Science through October 2010. MeSH and
free-text terms included the following: ‘‘cholesteatoma’’ and
‘‘magnetic resonance imaging,’’ ‘‘MR,’’ ‘‘MRI,’’ ‘‘diffusionweighted imaging,’’ and ‘‘diffusion MRI.’’ After selecting a study,
a search of its reference list was conducted to identify additional sources. The full article of each potentially eligible study
was obtained and assessed for inclusion.
We included prospective studies in which patients suspected of having cholesteatoma underwent both a preoperative
non–echo-planar DW MRI scan and a subsequent otologic procedure for diagnosis and treatment. For each study, the following
variables were extracted: total number of patients, DW MRI
sequences and imaging parameters, the number of radiologists
who reviewed the MRI scans, the range of cholesteatoma size
seen at surgery, the type of surgical procedure performed (primary or revision), the number of true positives (TP), false
positives (FP), true negatives (TN), false negatives (FN), or sensitivity and specificity with confidence intervals (CIs).
Studies were excluded if original data were not presented.
For multiple articles from the same group of authors or institution, we included only the most recent larger patient series
from any given group of authors, and we excluded earlier publications of a smaller subset of patients if there was known
duplication of patient data between articles.
PPV
%
MATERIALS AND METHODS
Li et al.: DW MRI in cholesteatoma diagnosis
level (for example, if the radiologists’ threshold were to
change based on cholesteatoma size).
DISCUSSION
Fig. 1. Hierarchical summary receiver operating characteristic
(HSROC) curve for included study data. This figure shows both the
pooled summary sensitivity and specificity values or accuracy (red
box), as well as the individual study values as circles, with circle size
proportional to study weight. The 95% confidence region is shown in
yellow, and the 95% prediction region is shown in gray. [Color figure
can be viewed in the online issue, which is available at
wileyonlinelibrary.com.]
cholesteatomas were detected by DW MRI (true positives) and 30 cholesteatomas were not detected (false
negatives). One hundred eight patients did not have cholesteatoma found intraoperatively, and of these patients,
100 were correctly identified by DW MRI (true negatives) whereas eight were not (false positives).
The majority of false negative results were due to
cholesteatoma pearls measuring <3 mm in size, thus
exceeding the lower limit of reliability of current imaging technology. The occasional false positive result was
attributed to susceptibility artifacts, presence of cholesterol granuloma, abscess, or bone powder.17
The overall pooled sensitivity was 0.94 (95% CI,
0.80–0.98) and specificity 0.94 (95% CI, 0.85–0.98).
Figure 1 presents these measures of study accuracy in
graphical form. This figure shows both the pooled summary sensitivity and specificity values or accuracy (red
box) as well as the individual study values as circles
with circle size proportional to study weight.
The 95% confidence region, shown in yellow, is the
two-dimensional analogy of the 95% CI for the pooled
study accuracy based on the data included. The 95% prediction region, shown in gray, is a larger area because
this reflects the area in which a future study could fall
with 95% probability. The green curved line represents
the receiver operator curve, which reflects the curve
on which different diagnostic thresholds would affect
sensitivity and specificity, given a constant accuracy
Laryngoscope 123: May 2013
Cholesteatomas arise from aberrantly located squamous epithelium in the middle ear or mastoid. Left
untreated, keratin debris from the squamous epithelium
accumulates, causing local bony erosion and acute or
chronic infections. The diagnosis of cholesteatoma has
historically been a clinical one. Imaging has not traditionally played a significant role in the diagnosis or
postoperative surveillance of cholesteatoma due to the inherent limitations of the various imaging modalities.18
Although CT imaging can define the bony anatomy of
the temporal bone, it cannot differentiate among types of
soft tissue (i.e., scar, granulation tissue, fluid, inflamed
mucosa, or recurrent cholesteatoma). Traditional MRI
sequences have also lacked the ability to differentiate
various forms of middle ear soft tissue and do not provide
good anatomic resolution of fine structures.
The DW MRI technique includes a period before the
image acquisition that enhances the signal intensity
attenuation in response to diffusion and other spin
motion.19,20 Cholesteatomas and histologically similar
epidermoid cysts are hyperintense on DW MRI images
compared with cerebrospinal fluid and brain parenchyma. Two major classes of diffusion weighting
techniques are currently in use; echo-planar (EP) and
non–echo-planar (non-EP) DW MRI. Non-EP–based
imaging techniques in the form of single-shot turbo spinecho sequences (HASTE DW MRI) and multishot turbo
spin-echo sequences (BLADE DW MRI and PROPELLER DW MRI) are preferred for imaging the temporal
bone, as they are less sensitive to magnetic susceptibility
artifacts and show less anatomical distortion than
images produced with EPI-based methods.18 An example
of these imaging modalities is given in Figures 2 and 3.
Fig. 2. Diffusion-weighted magnetic resonance imaging of a 55year-old male with a middle ear cholesteatoma. The cholesteatoma is clearly identified by the hyperintense signal.
Li et al.: DW MRI in cholesteatoma diagnosis
1249
to stratify patients into groups of those who would benefit from early second-look surgery and those who could
be closely observed. In the latter group, serial DW MRI
would either show patients to be free from cholesteatoma or allow a small below-threshold cholesteatoma to
reach a size allowing for detection and subsequent
treatment.
BIBLIOGRAPHY
Fig. 3. The corresponding location on a fine-cut, temporal bone,
computed tomography scan of the patient in Figure 2.
Despite the low anatomic resolution of MRI, DW
MRI results are useful to either rule in or out the
presence of cholesteatoma. Typically, this information is
adequate to plan surgery, because most anatomic landmarks and the cholesteatoma itself will be visualized
directly intraoperatively. Although a coregistered CT and
DW MRI is technologically possible, the additional cost
and radiation exposure likely do not justify the routine
use of fusion images to increase anatomic localization.21
Our results suggest that non–echo-planar DW MRI is
a highly sensitive and specific way to detect cholesteatoma
in the temporal bone. Moreover, this technology can detect
even very small (3 mm) cholesteatoma recurrences. Based
on these findings, non–echo-planar DW MRI is a reasonable alternative to second-look surgery for detecting
cholesteatoma in the properly selected patient population.
There are some limitations of this study that are inherent to all meta-analyses. Bias, including publication
bias, confounding, and quality issues can occur in the
original studies resulting from flaws in the study design
that tend to distort the magnitude or direction of associations in the data or from the way in which studies are
selected for inclusion.22 Publication bias, defined as the
selective publication of studies based on the magnitude
and direction of their findings, represents a particular
threat to the validity of meta-analysis of observational
studies.23 In this particular meta-analysis, there were
relatively few patients for each study included. The consistent results of each included study, however, support
the overall conclusion that non–echo-planar DW MRI
have very high sensitivity and specificity for detecting
even small cholesteatomas. Further clinical experience
with this imaging modality will be helpful in establishing its utility in cholesteatoma detection.
CONCLUSION
Non–echo-planar DW MRI is highly sensitive and
specific in identifying cholesteatoma. DW MRI may help
Laryngoscope 123: May 2013
1250
1. Jindal M, Riskalla A, Jiang D, Connor S, O’Connor AF. A systematic
review of diffusion-weighted magnetic resonance imaging in the assessment of postoperative cholesteatoma. Otol Neurotol 2011;32:1243–1249.
2. Harbord RM, Deeks JJ, Egger M, Whiting P, Sterne JA. A unification of
models for meta-analysis of diagnostic accuracy studies. Biostatistics
2007;8:239–251.
3. Harbord RM, Whiting P. Metandi: meta-analysis of diagnostic accuracy
using hierarchical logistic regression. Stata J 2010;9:211.
4. Rutter CM, Gatsonis CA. A hierarchical regression approach to meta-analysis of diagnostic test accuracy evaluations. Stat Med 2001;20:
2865–2884.
5. Kasbekar AV, Scoffings DJ, Kenway B, et al. Non echo planar, diffusionweighted magnetic resonance imaging (periodically rotated overlapping
parallel lines with enhanced reconstruction sequence) compared with
echo planar imaging for the detection of middle-ear cholesteatoma.
J Laryngol Otol 2011;125:376–380.
6. De Foer B, Vercruysse JP, Bernaerts A, et al. The value of single-shot
turbo spin-echo diffusion-weighted MR imaging in the detection of middle ear cholesteatoma. Neuroradiology 2007;49:841–848.
7. De Foer B, Vercruysse JP, Bernaerts A, et al. Detection of postoperative
residual cholesteatoma with non-echo-planar diffusion-weighted magnetic resonance imaging. Otol Neurotol 2008;29:513–517.
8. Dubrulle F, Souillard R, Chechin D, Vaneecloo FM, Desaulty A, Vincent C.
Diffusion-weighted MR imaging sequence in the detection of postoperative recurrent cholesteatoma. Radiology 2006;238:604–610.
9. Dhepnorrarat RC, Wood B, Rajan GP. Postoperative non-echo-planar diffusion-weighted magnetic resonance imaging changes after cholesteatoma
surgery: implications for cholesteatoma screening. Otol Neurotol 2009;
30:54–58.
10. Lehmann P, Saliou G, Brochart C, et al. 3T MR imaging of postoperative
recurrent middle ear cholesteatomas: value of periodically rotated overlapping parallel lines with enhanced reconstruction diffusion-weighted
MR imaging. AJNR Am J Neuroradiol 2009;30:423–427.
11. De Foer B, Vercruysse JP, Bernaerts A, et al. Middle ear cholesteatoma:
non-echo-planar diffusion-weighted MR imaging versus delayed gadolinium-enhanced T1-weighted MR imaging—value in detection. Radiology
2010;255:866–872.
12. Huins CT, Singh A, Lingam RK, Kalan A. Detecting cholesteatoma with
non-echo planar (HASTE) diffusion-weighted magnetic resonance imaging. Otolaryngol Head Neck Surg 2010;143:141–146.
13. Pizzini FB, Barbieri F, Beltramello A, Alessandrini F, Fiorino F. HASTE
diffusion-weighted 3-Tesla magnetic resonance imaging in the diagnosis
of primary and relapsing cholesteatoma. Otol Neurotol 2010;31:596–602.
14. Plouin-Gaudon I, Bossard D, Fuchsmann C, Ayari-Khalfallah S, Froehlich
P. Diffusion-weighted MR imaging for evaluation of pediatric recurrent
cholesteatomas. Int J Pediatr Otorhinolaryngol 2010;74:22–26.
15. Rajan GP, Ambett R, Wun L, et al. Preliminary outcomes of cholesteatoma
screening in children using non-echo-planar diffusion-weighted magnetic resonance imaging. Int J Pediatr Otorhinolaryngol 2010;74:
297–301.
16. Khemani S, Lingam RK, Kalan A, Singh A. The value of non-echo planar
HASTE diffusion-weighted MR imaging in the detection, localisation
and prediction of extent of postoperative cholesteatoma. Clin Otolaryngol 2011;36:306–312.
17. Profant M, Slavikova K, Kabatova Z, Slezak P, Waczulikova I. Predictive
validity of MRI in detecting and following cholesteatoma. Eur Arch
Otorhinolaryngol 2012;269:757–765.
18. Khemani S, Singh A, Lingam RK, Kalan A. Imaging of postoperative
middle ear cholesteatoma. Clin Radiol 2011;66:760–767.
19. Schwartz KM, Lane JI, Bolster BD Jr, Neff BA. The utility of diffusionweighted imaging for cholesteatoma evaluation. AJNR Am J Neuroradiol
2011;32:430–436.
20. Le Bihan D, Breton E, Lallemand D, Grenier P, Cabanis E, Laval-Jeantet
M. MR imaging of intravoxel incoherent motions: application to diffusion and perfusion in neurologic disorders. Radiology 1986;161:401–407.
21. Plouin-Gaudon I, Bossard D, Ayari-Khalfallah S, Froehlich P. Fusion of
MRIs and CT scans for surgical treatment of cholesteatoma of the
middle ear in children. Arch Otolaryngol Head Neck Surg 2010;136:
878–883.
22. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational
studies in epidemiology: a proposal for reporting. Meta-analysis Of
Observational Studies in Epidemiology (MOOSE) group. JAMA 2000;
283:2008–2012.
23. Easterbrook PJ, Berlin JA, Gopalan R, Matthews DR. Publication bias in
clinical research. Lancet 1991;337:867–872.
Li et al.: DW MRI in cholesteatoma diagnosis