Professional Documents
Culture Documents
Strokeaha 107 509877 PDF
Strokeaha 107 509877 PDF
Background and Purpose—Accurate diagnosis of the degree of internal carotid artery (ICA) stenosis is needed for
decisions regarding optimal stroke prevention. Noninvasive magnetic resonance angiography (MRA) is being proposed
and used as a replacement for the gold standard, intra-arterial angiography. Our purpose was to perform a systematic
review and diagnostic meta-analysis to determine the sensitivity and specificity of time-of-flight (TOF) MRA and
contrast-enhanced (CE) MRA for the detection of (1) high-grade (ⱖ70% to 99%) ICA stenoses; (2) ICA occlusions; (3)
moderately severe (50% to 69%) ICA stenoses; and (4) compare the overall accuracy of the 2 MRA techniques.
Methods—The medical literature on MRA and the diagnosis of ICA steno-occlusive disease was reviewed through the
PubMed, EMBASE, and SCOPUS databases. All publication years were included through to November 2006. Studies
were eligible for inclusion if they compared the accuracy of TOF or CE MRA for the detection of ICA disease against
intra-arterial angiography and reported sufficient data.
Results—The overall sensitivity of TOF MRA for the detection of ⱖ70% to 99% ICA stenoses was 91.2% (95% CI: 88.9%
to 93.1%) with a specificity of 88.3% (86.7% to 89.7%), whereas the sensitivity of CE MRA was 94.6% (92.4% to
96.4%) with a specificity of 91.9% (90.3% to 93.4%). For the detection of ICA occlusions, the sensitivity of TOF MRA
was 94.5% (91.2% to 96.8%) and the specificity was 99.3% (98.9% to 99.6%), whereas the sensitivity and specificity
values for CE MRA were 99.4% (96.8% to 100%) and 99.6% (99.2% to 99.9%), respectively. For moderately severe
(50% to 69%) stenoses, TOF MRA had a sensitivity of only 37.9% (29.3% to 47.1%) and a specificity of 92.1% (89.6%
to 94.1%); for CE MRA, the pooled sensitivity value was somewhat better at 65.9% (57.0% to 74.0%), whereas the
Downloaded from http://ahajournals.org by on July 29, 2020
Received November 12, 2007; final revision received December 27, 2007; accepted January 14, 2008.
From the Stroke Neuroscience Unit (S.M.D., V.L.W., H.Y., J.K.L., A.E.B.), National Institute of Neurological Diseases and Stroke, National Institutes
of Health, Bethesda, MD; and Boston University School of Dental Medicine (S.J.J.), Boston, Mass, USA and HUG Hôpital Cantonal, Geneva,
Switzerland.
Correspondence to Alison E. Baird, FRACP, PhD, Professor of Neurology, Physiology and Pharmacology, and, Director of the Division of
Cerebrovascular Disease and Stroke, SUNY Downstate Medical Center, 450 Clarkson Avenue, Box 1213, Brooklyn, NY 11203. E-mail alison.baird@
downstate.edu
© 2008 American Heart Association, Inc.
Stroke is available at http://stroke.ahajournals.org DOI: 10.1161/STROKEAHA.107.509877
2237
2238 Stroke August 2008
patients with moderate (50% to 69%) stenoses: symptomatic patients only.22 Our purpose was to provide an update and
patients with 50% to 69% ICA stenoses had an absolute risk determine the sensitivity and specificity of TOF MRA and
reduction of stroke of 4.6% at 5 years.3 Furthermore, in the CE MRA for the detection of (1) high-grade (70% to 99%)
Asymptomatic Carotid Atherosclerosis Study, asymptomatic ICA stenoses; (2) ICA occlusions; (3) moderately severe
patients with 60% to 99% stenoses showed a modest benefit (50% to 69%) ICA stenoses; and (4) to compare the overall
from CEA.6 Similar results were obtained in patients with accuracies of the 2 MRA techniques using IAA as the gold
asymptomatic 60% to 99% stenoses randomized to either standard.
immediate or deferred CEA.7 Based on these figures, CEA is
recommended for the secondary prevention of stroke in Materials and Methods
patients with high-grade stenoses and after consideration of Data Sources
other clinical factors for many with moderate-grade stenoses The medical literature on MRA and diagnosis of ICA stenosis was
and also some with asymptomatic disease.8,9 Carotid artery reviewed through a variety of databases. Searches were run by 2 of
stenting may be an alternative to CEA in patients who are at the reviewers (SMD and AEB) using PubMed, EMBASE, SCOPUS,
very high surgical risk.10 –13 and the Cochrane Controlled Trials Register. All publication years
were included with the search being performed for all papers published
Intra-arterial angiography (IAA), most commonly performed through to November 2006. PubMed was searched using medical
using digital subtraction angiography (DSA), is considered the subject headings terms and from the papers identified using these
gold standard technique for evaluating the presence and terms, reference lists were searched to identify any additional articles.
extent of ICAD. The stenosis grades used in the NASCET No initial limits were set in this search, although only papers
and ECST trials were determined by IAA. However, IAA is published in English were accepted. The Cochrane Controlled Trials
Register database, SCOPUS, and EMBASE were all searched using
invasive, costly, and has risks that include an approximate 4% particular keywords, described subsequently. Hand searches of
risk of transient ischemic attack and a 1% risk of disabling selected papers’ references as well as recent publications of several
stroke or death.14 Because of these risks, there have been key journals, including Stroke, Neurology, Circulation, and Radiol-
increasing calls for noninvasive screening methods for ICAD. ogy, were performed. Two independent searches were performed,
In many hospitals, noninvasive imaging methods such as one for TOF MRA and one for CE MRA. However, the key words
and medical subject headings terms searched were the same, because
ultrasound, magnetic resonance angiography (MRA) and com- both TOF and CE MRA can be searched with identical terms. By
puted tomography angiography, either alone or in combina- performing separate searches for forms of MRA, the researchers
tion, are already being used to screen and select patients for were able to accurately exclude papers according to specific criteria.
CEA.15 Ultrasound is affordable and readily available, but the The medical subject headings terms were carotid arteries, carotid
stenosis, magnetic resonance angiography, and angiography. The
images produced by ultrasound are often lower in quality than
key words were carotid artery, carotid artery stenosis, magnetic
Downloaded from http://ahajournals.org by on July 29, 2020
that of computed tomography angiography or MRA and the resonance angiography, contrast-enhanced magnetic resonance an-
accuracy is highly operator-dependent. giography, and angiography. Although other search terms exist in
MRA has become widely used as a noninvasive diagnostic Embase and Scopus, it was found that the terms used were thorough
imaging modality for ICAD. It avoids the radiation and and sufficient, due to the large quantity of results, in addition to the
high overlap of papers found using different search terms. Hand
iodinated contrast exposure associated with computed tomog- searches of relevant journals resulted in complete concordance with
raphy angiography. The first MRA method, phase-contrast the online databases, thereby validating these database searches as
MRA, was developed in the 1980s and was quickly followed comprehensive and complete. During the initial search strategy, over
by 2-dimensional (2D) and 3-dimensional (3D) time-of-flight 11 000 papers (duplicates included) were identified using 3 online
(TOF) MRA. TOF MRA has been widely adopted for an databases (8231 with PubMed, 1209 with EMBASE, and 2119 with
SCOPUS). The titles of all articles obtained were screened, and all
array of clinical indications but is relatively insensitive to abstracts of possible articles were further screened. Papers with
slow flow and is associated with long scan times and signal abstracts that clearly fitted into a category for exclusion or diagnostic
voids, all of which can lead to poor-quality imaging and tests for medical conditions unrelated to cerebrovascular disease
overestimation of stenoses. More recently, contrast-enhanced were eliminated at this level of the search before a full text was
(CE) MRA has been introduced. CE MRA produces high- printed. After search completion, independent readers checked for
agreement. On identification of a possible abstract for inclusion, the
quality images in a very short period of time and may alleviate full text of the article was read and assessed for inclusion. In our
some of the drawbacks associated with TOF MRA.16 Yet, to search, there were no possible abstracts lacking full text versions.
be a useful method for studying ICAD, the accuracy of MRA Any disagreements over article inclusion were solved by discussion.
is imperative, because decisions regarding optimum stroke The search protocol is provided in Figure 1.
prevention depend on the results obtained. Of particular
importance are the potential overcall of stenosis grade and the
Study Selection
differentiation of high-grade stenosis from occlusion (occlu- Inclusion Criteria
sions not usually being amenable to surgery) and the accurate Studies were eligible for inclusion if they (1) evaluated the accuracy
of 2D and/or 3D TOF MRA and/or CE MRA for the diagnosis of
diagnosis of moderate-grade (50% to 69%) stenosis for the ICA stenosis (with or without data on occlusion) against the gold
patient to receive optimal management. standard of IAA; (2) reported sufficient data to construct a 2⫻2 table
Despite the rapid technological advances in MRA and a of test performance; (3) conformed to the Standards for Reporting
large radiological literature, there has been little systematic of Diagnostic Accuracy criteria23; and (4) had image evaluations
evaluation of its diagnostic accuracy.17–21 The one prior performed by blinded observers. Due to the limited number of
studies published on this topic, all study designs were included
meta-analysis included a select population of symptomatic (Tables 1 and 2). Papers that included only symptomatic patients
patients studied up to April 2004, and data for moderate- were included and evaluated separately in the subgroup analyses.
grade stenoses were available in a very small number of The QUADAS tool for evaluating the quality of included studies was
Debrey et al Diagnostic Accuracy of MRA for ICA Disease 2239
receiving the test in practice, which would potentially include only negative likelihood ratio less than 0.1 support high diagnostic
symptomatic or “high-risk” patients. However, it was necessary in accuracy for the respective test. Because likelihood ratios, diagnostic
this analysis to include studies that examined a range of healthy and ORs, and summary ROC curves take into account both the sensitivity
ill subjects to determine the accuracy of these imaging techniques and specificity data, and therefore illustrate any tradeoffs that may
across a large spectrum. Additionally, the clinical data available to occur, these values are considered to be more valuable in terms of
radiologists during image interpretation for these studies was not the diagnostic accuracy than a pooled, weighted value for sensitivity and
same as would be available when the test is used in practice, and specificity.
uninterpretable test results in many studies were often not discussed
in great detail.
Results
Exclusion Criteria
Exclusion criteria were lack of comparison with angiography as the Included Studies
gold standard or lack of data for a 2⫻2 table. Where data were After the initial screening period, 72 papers investigating TOF
insufficient, efforts were made to contact the study authors, but if MRA and 24 papers investigating CE MRA, some overlapping,
there was no response, the papers were excluded. Studies that used were selected for further screening. Of these papers, 37 TOF
duplicate data were excluded. MRA17,18,26 – 60 and 21 CE MRA papers16,26,27,33,35,37,61–75 ful-
filled the inclusion criteria (Tables 1 and 2). Of the TOF
Data Extraction
papers, 11 used 2D TOF MRA, 14 used 3D TOF MRA, and
Measurement of Stenosis 12 included both 2D and 3D TOF data (Table 1). For TOF
In the majority of the studies after 1995, the degree of ICA stenosis
MRA, there were 1651 subjects, whereas for CE MRA, there
was measured using either NASCET guidelines (in which the grade
of the stenotic lesion is determined relative to a segment of distal were 1047 subjects. Reasons for exclusion between primary
disease-free ICA) or ECST guidelines (where the grade of the stenotic and secondary screening included: duplicate data (5 TOF),
lesion is compared relative to what the original ICA arterial outline poor quality (4 TOF), failure to use conventional angiography
would have been at the point of maximum stenosis); a 70% ECST as the gold standard (4 TOF), data not provided/failed
stenosis is approximately equivalent to a 50% NASCET stenosis.
The ECST measurements in this meta-analysis were not corrected to
attempts to locate data (11 TOF MRA, 3 CE MRA), and other
NASCET measurements; therefore, because this contributes to het- reasons, including papers only investigating intracranial ste-
erogeneity in the study, a subgroup analysis was performed in which noses or unrelated subject matter (12 TOF). The NASCET
the ECST papers were excluded. After final selection of included criteria for the grading of stenoses were used in the majority
papers, data were extracted to form 2⫻2 tables of diseased and of studies; ECST criteria were used in 2 of 37 TOF MRA
nondiseased vessels as seen on MRA using conventional angiogra-
phy as the gold standard. The following comparisons of stenosis studies and one of 21 CE MRA study; in 11 early TOF MRA
grades were performed: (1) to evaluate high-grade stenoses, patients studies (before 1995), the method of stenosis measurement
with ⱖ70% to 99% ICA stenosis (ie, operative candidates) were was not described. In one CE MRA study, the method of
2240 Stroke August 2008
Nicholas 1995 40 2D Prescreened with ultrasound No Not stated Not stated ⬎70%
and DSA
Dadachanji 1995 20 2D Clinical signs of CVI Not stated Yes Not stated ⬎70%‡
Vanninen 1995 65 3D 55 of 65 referred for DSA Not stated Yes 1 day to 2 months ⬎70%
Young 1994 70 2D and 3D Prescreened with ultrasound Yes Yes 2 to 3 weeks ⬎70%‡
and symptomatic
DeMarco 1994 20 2D and 3D Referred for DSA Yes Yes 3 days ⬎70%‡
White 1994 60 2D Symptomatic Yes Yes Within days ⬎60%‡
Mittl 1994 38 2D Clinical signs of CVI No Not stated Not stated ⬎70%
Sitzer 1993 50 2D Prescreened with ultrasound Yes Yes 2 to 43 days ⬎70%‡
and symptomatic
Turnipseed 1993 30 2D Symptomatic Yes Yes Not stated ⬎70%
Chiesa 1993 63 3D Prescreened with ultrasound Not stated Not stated Not stated ⬎60%‡
and symptomatic
Huston 1993 50 2D Symptomatic, referred for DSA Yes Yes Same day ⬎80%
Anson 1993 20 2D Symptomatic No Not stated Less than 2 weeks ⬎75%‡
Pan 1992 34 2D and 3D Symptomatic Not stated Not stated Not stated ⬎60%‡
Furuya 1992 22 3D Suspected ICAD No Not stated Not stated ⬎60%‡
Mattle 1991 20 2D or 3D 13 of 20 symptomatic No Not stated Not stated ⬎70%‡
Kido 1991 31 2D Suspected ICAD Not stated Not stated 24 hours ⬎80%‡
*NASCET measurement method used to determine stenosis grade except where stated.
†ECST method.
‡Method used to determine stenosis grade measurement not stated.
CVI indicates cerebrovascular insufficiency.
measurement was not reported. Less than half of these studies TOF papers, 26 defined high-grade stenosis as 70% to 99%,
were prospectively carried out or were conducted in un- 5 as 60% to 99%, 2 as 75% to 99%, and 4 as 80% to 99%
selected populations. Only 12 of 37 TOF MRA and 5 of 21 (Table 1). Of the 21 CE MRA papers included in this
CE MRA papers had sample sizes ⱖ50 subjects. Of the 37 analysis, 18 defined high-grade stenosis as 70% to 99% (Table
Debrey et al Diagnostic Accuracy of MRA for ICA Disease 2241
2), whereas one measured stenosis of 80% to 99%, one of In addition to pooled sensitivity and specificity plots, positive
66% to 99%, and the last 50% to 99%. and negative likelihood ratios and diagnostic ORs were also
calculated. For TOF MRA, the positive and negative likeli-
Detection of >70% to 99% Internal Carotid hood ratios were 7.53 (5.75 to 9.87) and 0.13 (0.10 to 0.16),
Artery Stenoses respectively, whereas for CE MRA, the values were 12.26
The results for 70% to 99% stenosis include the 32 of 37 TOF
(8.00 to 18.8) and 0.07 (0.05 to 0.11). Although the diagnos-
MRA and 19 of 21 CE MRA papers that reported results for
tic accuracy for both MRA types was high, a difference
ⱖ70% to 99% stenoses, because these patients are operative
between TOF MRA and CE MRA does exist, CE MRA being
candidates (Table 3). The results for pooled sensitivities and
specificities for ⱖ70% to 99% ICA stenosis are depicted in the more accurate and proven to be more robust in the likelihood
Figure 2A–D. The pooled sensitivities for TOF MRA and CE ratio data than in the pooled sensitivity and specificity data.
MRA for the detection of ⱖ70% to 99% stenosis were 91.2%
(95% CI: 88.9% to 93.1%) and 94.6% (92.4% to 96.5%), respec- Detection of Internal Carotid Artery Occlusions
tively. The results for specificity were 88.3% (86.7% to 89.7%) Occlusion data were available in 29 of 37 TOF MRA studies
for TOF MRA and 91.9% (90.3% to 93.4%) for CE MRA and 18 of 21 CE MRA studies (Table 3). The sensitivity of
(Table 3). Summary ROC results are presented in Figure 3A–B. TOF MRA for the detection of ICA occlusions was 94.5%
Table 3. Diagnostic Accuracy for >70% to 99% ICA Stenosis, ICA Occlusion, and 50% to 69% ICA Stenosis
TOF MRA CE MRA
Pooled Sensitivity = 0.91 (0.89 to 0.93) Pooled Specificity = 0.88 (0.87 to 0.90)
Chi-square = 51.00; df = 31 (p = 0.0133) Chi-square = 146.80; df = 31 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 39.2 % 0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 78.9 %
Sensitivity Specificity
B CE MRA D CE MRA
Specificity (95% CI)
Sensitivity (95% CI)
Wright 2005 0.86 (0.80 - 0.92)
Wright 2005 0.64 (0.35 - 0.87)
Fellner 2005 0.84 (0.66 - 0.95)
Downloaded from http://ahajournals.org by on July 29, 2020
Figure 2. A–D, Pooled sensitivity and specificity values for the detection of ⱖ70% to 99% stenosis by TOF MRA and CE MRA.
(91.2% to 96.8%) with a specificity of 99.3% (98.9% to papers published data on this level of stenosis. Only 7 of 37
99.6%), whereas the sensitivity of CE MRA was 99.4% (95% TOF MRA and 8 of 21 CE MRA papers presented these data
CI: 96.8% to 100%) with a specificity of 99.6% (99.2% to (Table 3; Figure 4). The 7 TOF papers resulted in a pooled
99.9%). The positive and negative likelihood ratios of TOF sensitivity of only 37.9% (29.3% to 47.1%) and a pooled
MRA were 56.24 (38.87 to 81.36) and 0.12 (0.07 to 0.19). For specificity of 92.1% (89.7% to 94.1%). The 8 CE MRA
CE MRA, these values were 80.5 (48.16 to 134.70) and 0.07 papers had a pooled sensitivity of 65.9% (57.0% to 74.0%)
(0.04 to 0.13), respectively, again supporting a superior accu- and a pooled specificity of 93.5% (91.3% to 95.3%).
racy of CE MRA over TOF MRA for the detection of ICA
occlusion. The corresponding summary ROC curves are shown Studies Comparing Time-of-Flight Magnetic
in Figure 3C–D. Resonance Angiography and Contrast-Enhanced
Magnetic Resonance Angiography in the Same
Detection of Moderate-Grade Internal Carotid Patients for the Detection of High-Grade Internal
Artery Stenoses 50% to 69% Carotid Artery Stenoses
Although it has been shown that CEA can be effective in Only 5 studies assessed both TOF MRA and CE MRA in the
symptomatic patients with 50% to 69% stenosis, very few same subjects (n⫽172) for the detection of high-grade steno-
Debrey et al Diagnostic Accuracy of MRA for ICA Disease 2243
Symmetric SROC
0.9 AUC = 0.9597 Symmetric SROC
SE(AUC) = 0.0063 0.9 AUC = 0.9880
SE(AUC) = 0.0047
Q* = 0.9040
SE(Q*) = 0.0091 Q* = 0.9537
0.8 SE(Q*) = 0.0107
0.8
0.7
0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
1-specificity 1-specificity
D CE MRA
B CE MRA
Sensitivity SROC Curve
Sensitiv ity SROC Curv e 1
1
Symmetric SR
Symmetric SROC 0.9 AUC = 0.9922
0.9 AUC = 0.9812 SE(AUC) = 0.0
SE(AUC) = 0.0055 Q* = 0.9640
Q* = 0.9395 SE(Q*) = 0.010
SE(Q*) = 0.0106 0.8
0.8
0.7
0.7
Downloaded from http://ahajournals.org by on July 29, 2020
0.6
0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
1-specificity 1-specificity
Figure 3. A–D, Summary ROC curves for detection of 70% to 99% stenosis and occlusion by TOF MRA and CE MRA.
ses.26,27,33,35,37 CE MRA was more accurate than TOF MRA was performed in patients with angiographic stenoses of 70%
with a higher sensitivity and a similar specificity to TOF to 100%, but this gave essentially the same results (Table 4).
MRA (Table 3). TOF MRA produced a pooled sensitivity of The other sections of the sensitivity analyses include papers
89.6% (82.2% to 94.7%) and a pooled specificity of 92.7% that studied high-grade stenosis detection in only symptomatic
(88.5% to 95.7%), whereas CE MRA produced a pooled subjects as opposed to a population of symptomatic and asymp-
sensitivity of 96.2% (90.5% to 99.0%) and a pooled speci- tomatic patients and papers published after 2003 (Table 4). Both
ficity of 92.7% (88.6% to 95.7%). methods showed high accuracy for the detection of high-grade
stenosis in symptomatic patients. Results for both MRA methods
Subgroup Analyses for high-grade stenosis were highly accurate for papers pub-
To study the robustness of the results, a number of subgroups lished after 2003.
were analyzed in a sensitivity analysis. When the analyses As shown in Tables 1 and 2, some papers did not report the
were broadened to include the data from all of the included method of stenosis measurement. To decipher whether these
studies, according to each paper’s definition of “high-grade” studies, as well as the few studies that used ECST instead of
stenosis, the results were very similar (Table 4) to those for NASCET methods, contributed to heterogeneous results, a
ⱖ70% to 99% stenosis. In case overcall of occlusion by MRA sensitivity analysis was completed, including only those
was not being detected in these analyses, a separate analysis papers that explicitly stated the NASCET protocol was
2244 Stroke August 2008
B CE MRA D CE MRA
Sensitivity (95% CI) Specificity (95% CI)
Fellner 2005 0.00 (0.00 - 0.84) Fellner 2005 0.95 (0.83 - 0.99)
Anzalone 2005 0.92 (0.64 - 1.00) Anzalone 2005 0.92 (0.84 - 0.97)
U-King-Im 2004 0.57 (0.42 - 0.70) U-King-Im 2004 0.86 (0.80 - 0.91)
Alvarez-Linera 2003 0.80 (0.44 - 0.97) Alvarez-Linera 2003 0.99 (0.92 - 1.00)
Cosottini 2003 0.79 (0.61 - 0.91) Cosottini 2003 0.98 (0.94 - 1.00)
Wutke 2002 0.67 (0.22 - 0.96) Wutke 2002 1.00 (0.93 - 1.00)
Randoux 2001 0.80 (0.28 - 0.99) Randoux 2001 0.97 (0.87 - 1.00)
Serfaty 2000 0.14 (0.00 - 0.58) Serfaty 2000 0.86 (0.74 - 0.94)
Pooled Sensitivity = 0.66 (0.57 to 0.74) Pooled Specificity = 0.93 (0.91 to 0.95)
Chi-square = 23.48; df = 7 (p = 0.0014) Chi-square = 34.29; df = 7 (p = 0.0000)
0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 70.2 % 0 0.2 0.4 0.6 0.8 1 Inconsistency (I-square) = 79.6 %
Sensitivity Specificity
Figure 4. A–D, Pooled sensitivity and specificity values for detection of 50% to 69% stenosis by TOF MRA and CE MRA.
followed. Results for this subgroup were nearly identical to TOF MRA is that for every 30 MRA studies, one ⱖ70% to
the overall analysis results for high-grade stenosis. 99% stenosis would be detected on CE MRA that was missed
on TOF MRA. Similarly, for every 20 MRA studies, an ICA
Downloaded from http://ahajournals.org by on July 29, 2020
further major sources of heterogeneity were the variation in sion. As can be seen from both the forest plots of sensitivity
(a) the methods used to measure stenosis grade, and (b) the and specificity, as well as the ROC curves, more heteroge-
number and types of stenosis grades that were reported as neity exists in the TOF MRA data than in the CE MRA data
evidenced by the lack of data on moderate-grade stenoses. for high-grade stenosis. In particular, the CE MRA sensitivity
Evaluation of varying degrees of stenosis can result in plot shows very little heterogeneity. This could be due to a
suboptimal positive predictive values as evidenced in a recent variety of factors and is most likely the result of similar study
study.76 The use of varying stenosis grade may explain why designs across these papers.
Wright et al16 was an outlier; this study also used a head and It should also be noted that publication bias may be a
neck coil and hence evaluated ICA stenoses in a much wider source of heterogeneity. It is possible that studies demonstrat-
field of view. (3) Although it was necessary to compare data ing poor accuracy of MRA as compared with IAA may not
from both CE MRA and TOF studies, the majority of the have been published. The accuracy of imaging results de-
included TOF papers were published earlier than those pends on the instruments and technology used as well as the
investigating the accuracy of CE MRA. Approximately 76% experience of the operator and radiologist. Studies showing
of the CE MRA papers were published in the last 4 years excellent concordance between MRA and IAA are most often
compared with only 16% of the TOF papers. MRA is a performed in academic radiology departments. Therefore,
relatively new technology with first reports of its clinical use results from centers with decreased quality control may be
being less than 25 years old and the rate of improvement in less optimistic and perhaps less likely to be published.
technological advancements and user reliability has been Differences in diagnostic threshold could potentially have
marked. Although it was believed that the dates of the studies contributed to heterogeneity in this analysis. It has been stated
would be a serious source of heterogeneity, the data proved previously that differences in study design, including NASCET/
otherwise with surprisingly high rates of sensitivity and speci- ECST measurements and varying definitions of “high-grade”
ficity for TOF MRA over 10 years ago. (4) This apparent high stenosis contribute to heterogeneity. However, it is possible
degree of accuracy raises the question as to whether some of for differences in diagnostic threshold to exist despite the use
the earlier TOF MRA images were dropped from analysis due of the same diagnostic threshold such as defining 70% to 99%
to poor imaging quality, thus giving the potential for selection stenosis as severe. If a threshold effect is present, it can be
bias in the TOF MRA studies, resulting in the overprojection visually interpreted through the ROC plots. Sensitivity will
of its accuracy. This issue was frequently discussed in the increase with decreasing specificity or vice versa. This is
early MRA literature, because poor imaging quality, due to a illustrated on the corresponding ROC plots, which, in this
variety of factors including patient movement, artifact, and case, show weak evidence of diagnostic threshold, because
Downloaded from http://ahajournals.org by on July 29, 2020
slow flow, were common concerns with TOF MRA. In later both sensitivity and specificity increase simultaneously.
years, technology and imaging technique improvements led Only limited data were available for 50% to 69% stenoses;
to a decrease in these difficulties. The raw data in the 2⫻2 sometimes 30% to 69% stenosis data were reported, although
tables do not show this dilemma, because images graded as such a large range of data is not useful for CEA decision-
“nondiagnostic” quality are simply removed from the analysis making.8 At present, many, but not all, patients with moderate-
before being compared with the corresponding IAA images. grade stenosis are referred for surgery after consideration of a
This was not as common an occurrence in the CE MRA studies; number of clinical factors; patients with the best operative
in the cases of blurred images due to patient motion or artifact, outcomes are male, those with hemispheric stroke or transient
these images were more often included in the analyses. Other ischemic attack (as opposed to retinal ischemia), and those
papers stated that the precise definition of stenosis was not who do not have contralateral ICA occlusion. In cases of
always feasible with the TOF images due to flow voids or endovascular angioplasty or stenting, the results of noninva-
signal dropout. In these cases, a signal dropout was usually sive imaging will be de facto confirmed or rejected by IAA.
identified in these papers as a severe stenosis, which would Even with the limited data available, both CE MRA and
alter the data dramatically on the topic of differentiating especially TOF MRA appear to be poor diagnostic tools for
high-grade stenosis from occlusion. Signal dropout can be the moderate ICA stenosis. Their higher specificities only show
result of many factors, including turbulent flow, patient that MRA is somewhat more accurate in ruling out the presence
movement, and other technical difficulties and may actually of a moderate stenosis. The cause of the reduced sensitivity
not be due to a high-grade stenosis. Regardless, imaging may relate to image resolution. It is quite possible that some
failure and the reporting of signal voids as a high-grade patients assessed as having high-grade stenosis on MRA, in
stenosis are limitations in early TOF studies and must be fact, only have moderate-grade disease and vice versa,
taken into account in the comparison of TOF MRA and CE patients with moderate-grade disease may have high-grade
MRA data in this meta-analysis. (5) The majority of TOF disease. The potential clinical impact of these findings is most
MRA studies used 3D TOF MRA; however, there are several important for good clinical practice and will need to be
studies that incorporated both 2D and 3D techniques. Al- considered and investigated as a matter of urgency.
though 3D is often considered a stronger diagnostic technique This new information updates results obtained in the prior
due to improved signal-to-noise ratio, in areas of slow flow it systematic reviews. MRA techniques have shown a continued
may be advantageous to use 2D instead of 3D TOF. Studies evolution and improvement in their accuracy for the assess-
using both forms may be more complete, because transverse ment of ICA stenoses. It is clear that many research centers
2D images are more sensitive to slow flow and can therefore and hospitals have already made the transition to noninvasive
discriminate between critical stenosis (ie, ⬎95%) and occlu- MRA as a sole imaging modality and determinant for CEA
2246 Stroke August 2008
despite some reports that may demonstrate that MRA does 3. Barnett HJ, Taylor DW, Eliasziw M. Benefit of carotid endarterectomy in
not meet the threshold of diagnostic quality that would allow patients with symptomatic moderate or severe stenosis. North American
Symptomatic Carotid Endarterectomy Trial Collaborators. N Engl J Med.
its independent use in all clinical settings.77 MRA has the 1998;339:1415–1425.
advantages of being less expensive, less time-consuming, and 4. North American Symptomatic Carotid Endarterectomy Trial Collabo-
noninvasive as compared with DSA. It is also capable of rators. Beneficial effect of carotid endarterectomy in symptomatic
patients with high-grade carotid stenosis. North American Symptomatic
imaging vessels with unusual pathologies and produces stron-
Carotid Endarterectomy Trial Collaborators. N Engl J Med. 1991;325:
ger images of vessels not normally imaged with traditional 445– 453.
noninvasive imaging, including the vertebral and subclavian 5. MRC European Carotid Surgery Trial: interim results for symptomatic
arteries. Although the data for TOF MRA and CE MRA are patients with severe (70 –99%) or with mild (0 –29%) carotid stenosis.
European Carotid Surgery Trialists’ Collaborative Group. Lancet. 1991;
fairly similar in this meta-analysis, especially for high-grade 337:1235–1243.
ICA stenoses and occlusions, there are other factors that must 6. Executive Committee for the Asymptomatic Carotid Atherosclerosis
be taken into account. For instance, CE MRA unquestionably Study. Endarterectomy for asymptomatic carotid artery stenosis. JAMA.
provides clearer, more defined images than TOF MRA. Also, 1995;273:1421–1428.
7. Halliday A, Mansfield A, Marro J, Peto C, Peto R, Potter J, Thomas D;
because CE MRA is flow-independent, there is not a risk of MRC Asymptomatic Carotid Surgery Trial (ACST) Collaborative Group.
signal void due to turbulent flow. Because TOF MRA appears Prevention of disabling and fatal strokes by successful carotid endarter-
to be in the process of being phased out in some centers, some ectomy in patients without recent neurological symptoms: randomised
controlled trial. Lancet. 2004;363:1491–1502. Erratum in Lancet. 2004;
studies have only looked at issues such as patient preference 364:416.
with CE MRA versus DSA. It has been found in a recent study 8. Chaturvedi S, Bruno A, Feasby T, Holloway R, Benavente O, Cohen SN,
that patients appear to prefer CE MRA over DSA. According to Cote R, Hess D, Saver J, Spence JD, Stern B, Wilterdink J. Carotid
U-King-Im et al, 67% of subjects were willing to have a repeat endarterectomy- an evidence-based review: report of the therapeutics and
technology assessment subcommittee of the American Academy of Neu-
MRA versus 41% for DSA.78 rology. Neurology. 2005;65:794 – 801.
In conclusion, from this meta-analysis, the current change 9. Mohammed N, Anand SS. Prevention of disabling and fatal strokes by
to the use of MRA to detect high-grade ICA stenoses and successful carotid endarterectomy in patients without recent neurological
occlusions appears justified with CE MRA having the edge symptoms: randomized controlled trial. MRC Asymptomatic Carotid
Surgery Trial (ACST) collaborative group. Vasc Med. 2004;10:77–78.
over TOF MRA. However, the accuracy of MRA has yet to 10. Kastrup A, Gröschel K, Krapf H, Brehm BR, Dichgans J, Schulz JB.
be proven for the detection of moderate-grade stenoses, and it Early outcome of carotid angioplasty and stenting with and without
is recommended that a second noninvasive study be obtained cerebral protection devices: a systematic review of the literature. Stroke.
2003;34:813– 819.
to confirm the grade of ICA disease at this time before 11. Yadav JS, Wholey MH, Kuntz RE, Fayad P, Katzen BT, Mishkel GJ,
treatment decisions are made. A complimentary diagnostic
Downloaded from http://ahajournals.org by on July 29, 2020
Bajwa TK, Whitlow P, Strickman NE, Jaff MR, Popma JJ, Snead DB,
tool such as ultrasound or computed tomography angiography Cutlip DE, Firth BG, Ouriel K. Stenting and Angioplasty with Protection
or DSA is necessary in these cases for decision-making in Patients at High Risk for Endarterectomy Investigators. Protected
carotid-artery stenting versus endarterectomy in high-risk patients. N Engl
regarding potential carotid endarterectomy or stenting. With J Med. 2004;351:1493–1501.
over 660 000 CEAs performed in the United States over the 12. Mas JL, Chatellier G, Beyssen B, Branchereau A, Moulin T, Becquemin
past 5 years, there is a need for larger studies including both JP, Larrue V, Lievre M, Leys D, Bonneville JF, Watelet J, Pruvo JP,
Albucher JF, Viguier A, Piquet P, Garnier P, Viader F, Touze E, Giroud
symptomatic and asymptomatic subjects with tighter study
M, Hosseini H, Pillet JC, Favrole P, Neau JP, Ducrocq X; EVA-3S
guidelines to address the serious problem of heterogeneity Investigators. Endarterectomy versus stenting in patients with symptom-
and reporting of age data. Future studies should include patients atic severe carotid stenosis. N Engl J Med. 2006;355:1660 –1671.
with varying degrees of ICA stenosis severity, including mild 13. Furlan AJ. Carotid-artery stenting— case open or closed? N Engl J Med.
2006;355:1726 –1729. Erratum in N Engl J Med. 2006;355:2712.
stenoses. This work is also needed to confirm that the more 14. Willinsky RA, Taylor SM, TerBrugge K, Farb RI, Tomlinson G, Mon-
expensive CE MRA, which the literature has shown to be tanera W. Neurologic complications of cerebral angiography: prospective
largely replacing TOF MRA in recent studies, is truly a better analysis of 2,899 procedures and review of the literature. Radiology.
diagnostic tool. It is essential that these studies be completed 2003;227:522–528.
15. Barth A, Arnold M, Mattle HP, Schroth G, Remonda L. Contrast-
in a prospective and blinded fashion to fully answer the enhanced 3-D MRA in decision making for carotid endarterectomy: a
questions arising from this meta-analysis and systematic 6-year experience. Cerebrovasc Dis. 2006;21:393– 400.
review and to ensure that each patient receives optimal care. 16. Wright VL, Olan W, Dick B, Yu H, Alberts-Grill N, Latour LL, Baird
AE. Assessment of CE-MRA for the rapid detection of supra-aortic
vascular disease. Neurology. 2005;65:27–32.
Sources of Funding 17. Patel MR, Kuntz KM, Klufas RA, Kim D, Kramer K, Polak JF, Skillman
This research was supported in part by the Intramural Research JJ, Whittemore AD, Edelman RR, Kent KC. Preoperative assessment of
Program of the NIH, NINDS. S.D.J. is supported by the National the carotid bifurcation. Stroke. 1995;26:1753–1758.
Scientist Development Grant from the American Heart Association. 18. Elgersma OEH, Wust AFJ, Buijs PC, van der Graaf Y, Eikelboom BC,
Mali WPTM. Multidirectional depiction of the internal carotid arterial
Disclosures stenosis: three-dimensional time-of-flight MR angiography versus rota-
tional and conventional digital subtraction angiography. Radiology. 2000;
None.
216:511–516.
19. Kallmes DF, Omary RA, Dix JE, Evans AJ, Hillman BJ. Specificity of
References MR angiography as a confirmatory test of carotid artery stenosis. AJNR
1. Mohr JP, Caplan LR, Melski JW, Goldstein RJ, Duncan GW, Kistler JP, Am J Neuroradiol. 1996;17:1501–1506.
Pessin MS, Bleich HL. The Harvard Cooperative Stroke Registry: a 20. Nederkoorn PJ, van der Graaf Y, Hunink MG. Duplex ultrasound and
prospective registry. Neurology. 1978;28:754 –762. magnetic resonance angiography compared with digital subtraction
2. National Hospital Discharge Survey. Available at: www.cdc.gov/nchs/ angiography in carotid artery stenosis: a systematic review. Stroke. 2003;
about/major/hdasd/nhds.htm. Accessed November 12, 2007. 34:1324 –1332.
Debrey et al Diagnostic Accuracy of MRA for ICA Disease 2247
21. Westwood ME, Kelly S, Berry E, Bamford JM, Gough MJ, Airey CM, 39. Scarabino T, Carriero A, Magarelli N, Florio F, Giannatempo GM,
Meaney JF, Davies LM, Cullingworth J, Smith MA. Use of magnetic Bonomo L, Salvolini U. MR angiography in carotid stenosis: a com-
resonance angiography to select candidates with recently symptomatic parison of three techniques. Eur J Radiol. 1998;28:117–125.
carotid stenosis for surgery: systematic review. BMJ. 2002;324:198 –203. 40. Magarelli N, Scarabino T, Simeone AL, Florio F, Carriero A, Salvolini U,
22. Wardlaw JM, Chappell FM, Best JJ, Wartolowska K, Berry E. Non- Bonomo L. Carotid stenosis: a comparison between MR and spiral CT
invasive imaging compared with intra-arterial angiography in the angiography. Neuroradiology. 1998;40:367–373.
diagnosis of symptomatic carotid stenosis: a meta-analysis. Lancet. 2006; 41. Jackson MR, Chang AS, Robles HA, Gillespie DL, Olsen SB, Kaiser WJ,
367:1503–1512. Goff JM, O-Donnell SD, Rich NM. Determination of 60% or greater
23. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig carotid stenosis: a prospective comparison of magnetic resonance
LM, Lijmer JG, Moher D, Rennie D, de Vet HC. Standards for Reporting angiography and duplex ultrasound with conventional angiography. Ann
of Diagnostic Accuracy. Towards complete and accurate reporting of Vasc Surg. 1998;12:236 –243.
studies of diagnostic accuracy: the STARD initiative. Clin Chem. 2003; 42. Strotzer M, Fellner F, Fraunhofer S, Gmeinwieser J, Albrich H, Seitz J,
49:1– 6. Feuerbach S. Dedicated head-neck coil in MR angiography of the supra-
24. Whiting P, Rutjes AWS, Reitsma JB, Bossuyt PMM, Kleijnen J. The aoritc arteries from the aortic arch to the circle of Willis. Acta Radiol.
development of QUADAS: a tool for the quality assessment of studies of 1998;39:249 –256.
diagnostic accuracy included in systematic reviews. BMC Med Res 43. Huston J, Nichols DA, Luetmer PH, Rydberg CH, Lewis BD, Meyer FB,
Methodol. 2003;3:25. Brown RD, Schleck CD. MR angiographic and sonographic indications
25. Zamora J, Abraira V, Muriel A, Khan KS, Coomarasamy A. Meta-DiSc: for endarterectomy. AJNR Am J Neuroradiol. 1998;19:309 –315.
a software for meta-analysis of test accuracy data. BMC Med Res Methodol. 44. Fellner C, Strotzer M, Fraunhofer S, Held P, Spies V, Seitz J, Fellner F.
2006;6:31. MR angiography of the supra-aortic arteries using a dedicated head and
26. Anzalone N, Scomazzoni F, Castellano R, Strada L, Righi C, Politi LS, neck coil: image quality and assessment of stenoses. Neuroradiology.
Kirchin MA, Chiesa R, Scotti G. Carotid artery stenosis: intraindividual 1997;39:763–771.
correlations of 3D time-of-flight MR angiography, contrast-enhanced MR 45. Nicholas GG, Osborne MA, Jaffe JW, Reed JF. Carotid artery stenosis:
angiography, and conventional DSA, and rotational angiography for preoperative noninvasive evaluation in a community hospital. J Vasc
detection and grading. Radiology. 2005;236:204 –213. Surg. 1995;22:9 –16.
27. Fellner C, Lang W, Janka R, Wutke R, Bautz W, Felner FA. Magnetic 46. Dadachanji MC, Shroff MM, Modi D, Jankharia BG. Comparison of MR
resonance angiography of the carotid arteries using three different tech- angiography with contrast angiography for the diagnosis of carotid artery
niques: accuracy compared with intraarterial x-ray angiography and end- stenosis. J Assoc Physicians India. 1995;43:92–95.
arterectomy specimens. J Magn Reson Imaging. 2005;21:424 – 431. 47. Vanninen RL, Manninen HI, Partanen PLK, Vainio PA, Soimakallio S.
28. DeMarco JK, Huston J, Bernstein MA. Evaluation of classic 2D time-of- Carotid artery stenosis: clinical efficacy of MR phase-contrast flow quan-
tification as an adjunct to MR angiography. Radiology. 1995;194:
flight MR angiography in the depiction of severe carotid stenosis. AJR
459 – 467.
Am J Roentgenol. 2004;183:787–793.
48. Young GR, Humphrey PRD, Shaw MDM, Nixon TE, Smith ETS. Com-
29. Scarabino T, Fossaceca R, Carra L, Florio F, Marano G, Carriero A.
parison of magnetic resonance angiography, duplex ultrasound, and
Actual role of unenhanced magnetic resonance angiography (MRA TOF
digital subtraction angiography in assessment of extracranial internal
3D) in the study of stenosis and occlusion of extracranial carotid artery.
carotid stenosis. J Neurol Neurosurg Psychiatry. 1994;57:1466 –1478.
Radiol Med. 2003;106:497–503.
Downloaded from http://ahajournals.org by on July 29, 2020
59. Mattle HP, Kent KC, Edelman RR, Atkinson DJ, Skillman JJ. Evaluation 68. Remonda L, Senn P, Barth A, Arnold M, Lovblad KO, Schroth G. Contrast-
of the extracranial carotid arteries: correlation of magnetic resonance enhanced 3D MR angiography of the carotid artery: comparison with con-
angiography, duplex ultrasonography, and conventional angiography. ventional digital subtraction angiography. AJNR Am J Neuroradiol. 2002;
J Vasc Surg. 1991;13:838 – 845. 23:213–219.
60. Kido DK, Panzer RJ, Szumowski J, Hollander J, Ketonen LM, Monajati 69. Remonda L, Heid O, Schroth G. Carotid artery stenosis, occlusion, and
A, Ouriel K, Manzione JV, Dumoulin CL, Souza SP, Katz JL, pseudo-occlusion: first-pass, gadolinium-enhanced, three-dimensional
Rothenberg BM. Clinical evaluation of stenosis of the carotid bifurcation MR angiography—preliminary study. Radiology. 1998;209:95–102.
with magnetic resonance angiographic techniques. Arch Neurol. 1991;48: 70. Borisch I, Horn M, Butz B, Zorger N, Draganski B, Hoelscher T,
484 – 489. Bogdahn U, Link J. Preoperative evaluation of carotid artery stenosis:
comparison of contrast-enhanced MR angiography and duplex sonography
61. Randoux B, Marro B, Koskas F, Chiras J, Dormont D, Marsault C.
with digital subtraction angiography. AJNR Am J Neuroradiol. 2003;24:
Proximal great vessels of aortic arch: comparison of three-dimensional
1117–1122.
gadolinium-enhanced MR angiography and digital subtraction
71. U-King-Im JM, Trivedi RA, Graves MJ, Higgins NJ, Cross JJ, Tom BD,
angiography. Radiology. 2003;229:697–702. Hollingworth W, Eales H, Warburton EA, Kirkpatrick PJ, Antoun NM,
62. Alvarez-Linera J, Benito-Leon J, Escribano J, Campollo J, Gesto R. Gillard JH. Contrast-enhanced MR angiography for carotid disease.
Prospective evaluation of carotid artery stenosis: elliptic centric contrast- Neurology. 2004;62:1282–1290.
enhanced MR angiography and spiral CT angiography compared with 72. Randoux B, Marro B, Koskas F, Duyme M, Sahel M, Zouaoui A,
digital subtraction angiography. AJNR Am J Neuroradiol. 2003;24: Marsault C. Carotid artery stenosis: prospective comparison of CT, three-
1012–1019. dimensional gadolinium-enhanced MR, and conventional angiography.
63. Cosottini M, Pingitore A, Puglioli M, Michelassi MC, Lupi G, Abbruzzese Radiology. 2001;220:179 –185.
A, Calabrese R, Lombardi M, Parenti G, Bartolozzi C. Contrast-enhanced 73. Scarabino T, Carriero A, Giannatempo GM, Marano R, De Matthaeis P,
three-dimensional magnetic resonance angiography of atherosclerotic Bonomo L, Salvolini U. Contrast-enhanced MR angiography (CE MRA)
internal carotid stenosis as the noninvasive imaging modality in revascu- in the study of the carotid stenosis: comparison with digital subtraction
larization decision making. Stroke. 2003;34:660 – 664. angiography (DSA). J Neuroradiol. 1999;26:87–91.
64. Willinek WA, von Falkenhausen M, Born M, Gieseke J, Holler T, 74. Yang CW, Carr JC, Futterer SF, Morasch MD, Yang BP, Shors SM, Finn
Klockgether T, Textor HJ, Schild HH, Urbach H. Noninvasive detection JP. Contrast-enhanced MR angiography of the carotid and vertebrobasilar
of steno-occlusive disease of the supra-aortic arteries with three- circulations. AJNR Am J Neuroradiol. 2005;26:2095–2101.
dimensional contrast-enhanced magnetic resonance angiography: a pro- 75. Butz B, Dorenbeck U, Borisch I, Zorger N, Lenhart M, Feuerbach S, Link
spective, intra-individual comparative analysis with digital subtraction J. High-resolution contrast-enhanced magnetic resonance angiography of
angiography. Stroke. 2005;36:38 – 43. the carotid arteries using fluoroscopic monitoring of contrast arrival:
diagnostic accuracy and interobserver variability. Acta Radiol. 2004;45:
65. Lenhart M, Framme N, Volk M, Strotzer M, Manke C, Nitz WR, Finken-
164 –170.
zeller T, Feuerbach S, Link J. Time-resolved contrast-enhanced magnetic
76. Feldmann E, Wilterdink JL, Kosinski A, Lynn M, Chimowitz MI, Sarafin
resonance angiography of the carotid arteries: diagnostic accuracy and
J, Smith HH, Nichols F, Rogg J, Cloft HJ, Wechsler L, Saver J, Levine
inter-observer variability compared with selective catheter angiography.
SR, Tegeler C, Adams R, Sloan M. The Stroke Outcomes and Neuro-
Invest Radiol. 2002;37:535–541. imaging of Intracranial Atherosclerosis (SONIA) Trial. Neurology. 2007;
66. Sundgren PC, Sunden P, Lindgren A, Lanke J, Holtas S, Larsson EM. 68:2099 –2106.
Carotid artery stenosis: contrast-enhanced MR angiography with two 77. Gluf WM, O’Neill B, Couldwell WT. False-negative magnetic resonance
Downloaded from http://ahajournals.org by on July 29, 2020
different scan times compared with digital subtraction angiography. angiography with extracranial internal carotid artery stenosis: a report of
Neuroradiology. 2002;44:592–599. two cases and review of the literature. Neurosurg Rev. 2005;28:154 –158.
67. Wutke R, Lang W, Fellner C, Janka R, Denzel C, Lell M, Bautz W, 78. U-King-Im JM, Trivedi R, Cross J, Higgins N, Graves M, Kirkpatrick P,
Fellner FA. High-resolution, contrast-enhanced magnetic resonance Antoun N, Gillard JH. Conventional digital subtraction x-ray angiography
angiography with elliptical centric k-space ordering of supra-aortic arteries versus magnetic resonance angiography in the evaluation of carotid disease:
compared with selective x-ray angiography. Stroke. 2002;33:1522–1529. patient satisfaction and preferences. Clin Radiol. 2004;59:358–363.