Preoperative Assessment of the Carotid Bifurcation
Can Magnetic Resonance Angiography and Duplex Ultrasonography Replace Contrast Arteriography?
Background and Purpose Noninvasive studies are used with increasing frequency to assess the carotid bifurcation before endarterectomy. Therefore, assessment of their diagnostic accuracies is essential for appropriate patient management. We prospectively evaluate two noninvasive tests, magnetic resonance angiography (MRA) and duplex ultrasonography (DU), as potential replacements for contrast arteriography (CA).
Methods A blinded comparison of three-dimensional time-of-flight (TOF) MRA, two-dimensional TOF MRA, and DU in 176 arteries was performed. CA was used as the standard of comparison.
Results Three-dimensional TOF MRA had a sensitivity of 94%, a specificity of 85%, and an accuracy of 88% for the identification of 70% to 99% stenosis; two-dimensional TOF MRA had a sensitivity and specificity that were approximately 10% lower than those of three-dimensional TOF MRA. DU resulted in a sensitivity of 94%, a specificity of 83%, and an accuracy of 86%. Combining data from three-dimensional TOF MRA and DU, allowing for CA only for disparate results, yielded a sensitivity of 100%, a specificity of 91%, and an accuracy of 94% among concordant noninvasive tests, with CA required in 16% of arteries. MRA accurately differentiated 17 carotid occlusions from 16 high-grade (90% to 99%) stenoses, whereas with DU two patent arteries were identified as occluded and one occluded artery was identified as patent.
Conclusions Three-dimensional TOF MRA is the most accurate noninvasive test. Combined use of MRA and DU results in a marked increase in accuracy to a level that obviates the need for CA in a majority of patients.
CA has generally been proposed as a necessary preoperative study for carotid endarterectomy. However, the cost of CA, as well as the risk of stroke associated with this procedure, has encouraged the development of alternative noninvasive imaging modalities, including MRA and DU, for the detection of significant carotid disease. Accurate evaluation of the carotid bifurcation has become particularly important in view of the NASCET, which confirmed a benefit of carotid endarterectomy in symptomatic patients with 70% to 99% carotid stenosis.1 Similar results were obtained from the MRC European Carotid Surgery Trial, although the method of stenosis measurement differed from that used in the NASCET.2 The more recent multicenter Asymptomatic Carotid Atherosclerosis Study suggests a benefit of carotid endarterectomy in asymptomatic patients with a stenosis greater than 60%.3 As the indications for carotid endarterectomy become increasingly precise, so does the need for accurate and safe assessment of the carotid bifurcation, so that appropriate referral for carotid endarterectomy can be made.
The authors of numerous comparative studies using a variety of noninvasive modalities report varied conclusions.4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Many investigators have assessed the utility of MRA as a screening examination4 5 6 7 8 9 10 11 rather than as a preoperative diagnostic study.12 13 14 15 16 17 18 19 Also, in a majority of these studies, two-dimensional TOF and phase-contrast MRA techniques were used. Both of these latter techniques produce more signal loss at the carotid bifurcation because of dephasing phenomena than do three-dimensional TOF MRA techniques. The previous studies often involved too small a number of patients with disease to address specific subsets of stenosis, ie, the ability of the imaging test to accurately and reliably differentiate between a high-grade stenosis and occlusion. Although MRA and DU have individually been evaluated as potential replacements for CA, the concept of combining the results of these two tests has been suggested but not thoroughly investigated.14 15 16 17 We sought to determine whether concordant three-dimensional TOF MRA and DU results could improve diagnostic accuracy to a level that obviates the need for CA.
Subjects and Methods
Over a 19-month period, 88 patients referred for CA because of suspected atherosclerotic disease involving the carotid bifurcation were entered into our study. There were 30 women and 58 men ranging in age from 48 to 87 years (average age, 70 years). Patient risk factors for atherosclerotic disease included hypercholesterolemia (47%), hypertension (73%), diabetes mellitus (35%), and smoking (64%). There were 74 symptomatic carotid bifurcations (42%). Symptoms included transient ischemic attack (24), stroke (21), and amaurosis fugax (29). Carotid endarterectomy was subsequently performed on 86 arteries in 84 patients.
Of these carotid bifurcations, 171 were adequately imaged by CA, 163 by both three-dimensional TOF MRA and CA, 156 by both DU and CA, and 148 by all three modalities. The comparison of two-dimensional TOF MRA and three-dimensional TOF MRA was performed on 87 bifurcations imaged by both techniques. CA images of 5 carotid bifurcations were technically inadequate because of poor vessel opacification resulting from the inability to achieve selective catheterization of the CCA. One minor stroke occurred after CA. Eight bifurcations imaged by MRA were inadequate for evaluation. Reasons for MRA failure included motion (1), failure to image a low-lying bifurcation (1), and marked background signal (6). DU was not performed in 6 patients, and in 1 patient marked calcification obscured the bifurcation. Seventy-one patients had MRA of the circle of Willis.
CA was performed with digital subtraction techniques. MRA of the carotid bifurcation was performed on a 1.5-T Magnetom SP (Siemens Medical Systems) by use of a Helmholtz coil. The multislab three-dimensional transverse acquisition TOF technique used imaging parameters 40/7/1 (TR [repetition time, in milliseconds]/TE [excitation time, in milliseconds]/excitations) at a 20° flip angle, with 52-mm thickness per slab with 62 partitions and 0.812-mm slice thickness.20 The two-dimensional transverse acquisition TOF technique imaging parameters were 31/9/1 (TR/TE/excitations) at a 30° flip angle, with a slab thickness of 124 mm, a 3-mm slice thickness, and a 0.75-mm slice overlap with 55 slices. The three-dimensional TOF MRA of the circle of Willis was performed by use of a circularly polarized head coil with either imaging parameters 29/6/1 (TR/TE/excitations), at a 20° flip angle, or 35/6/1 at a 20° flip angle with magnetization transfer contrast. Two slabs were obtained with 52-mm slab thickness and 64 partitions per slab with 0.812-mm slice thickness. The MRA studies were reviewed after postprocessing with an MIP algorithm, with targeted MIPs used to display projections of each carotid bifurcation separately. The total MRA imaging time was approximately 30 minutes.
MRA and CA images were each reviewed independently by two radiologists; the MRA reviewers were different from the CA reviewers. For determination of extracranial stenoses, the CA and MRA studies were uniformly projected and magnified, and the outline of the arterial lumen was traced and measured with digital calipers (Fowler). A percent stenosis was determined in accordance with the NASCET protocol.1 A signal void was defined as a discrete discontinuity in the MIP projection. Carotid occlusion was determined by absence of signal on both two- and three-dimensional TOF MRA. The values for percent stenosis for both reviewers were averaged, and the κ statistic was used to assess interobserver reliability.21 Visual estimates were used to determine stenoses of the intracranial vessels. CA and MRA images were also evaluated for the presence, depth, and morphology of ulceration.
DU was performed either at the Brigham and Women’s Hospital (60 carotid bifurcations), the Beth Israel Hospital (80 carotid bifurcations), or at an outside certified laboratory (16 carotid bifurcations). PSV and end-diastolic velocity for the CCA and ICA were measured. We used two criteria to determine the sensitivity and specificity of DU for predicting a 70% to 99% stenosis: (1) ICA PSV/CCA PSV >4.022 and (2) ICA PSV ≥2300 mm/s.23
We calculated the sensitivity and specificity for DU, MRA, and their combination for determining various degrees of stenosis, with CA as the standard of comparison. Overall test performance was assessed with ROC analysis.24
Assessment of Stenosis by CA
Evaluation of CA revealed 99 carotid bifurcations (58%) with a 0% to 69% stenosis, 37 bifurcations with a 70% to 89% stenosis (22%), 17 bifurcations with a 90% to 99% stenosis (10%), and 18 carotid occlusions (10%). Interobserver reliability between readers was excellent (κ=0.93).
Assessment of Stenosis by MRA
Three-dimensional TOF MRA yielded a sensitivity of 94%, a specificity of 85%, and an accuracy of 88% for the identification of a 70% to 99% CA stenosis. The interobserver reliability between readers for three-dimensional TOF MRA was very good (κ=0.88). There were 20 carotid bifurcations (12%) for which three-dimensional TOF MRA and the CA stenosis were discrepant: 17 false-positive (Fig 1⇓) and 3 false-negative bifurcations. The mean difference between the degree of stenoses found by MRA and CA, excluding signal voids, was 5.0±14.3. In 2 false-positive bifurcations, a spurious stenosis found on MRA appeared to be a result of tortuosity or kinking of a vessel. There was no obvious explanation for the discrepancy in the remaining arteries. In 2 of the 3 false-negative bifurcations, the field imaged by MRA did not include the more severe stenosis, which in 1 case was in the mid-CCA and in the second was in the distal ICA. In the third case, artifact from dental fillings on CA partially obscured the area of stenosis, raising the possibility that the MRA may have actually been more accurate.
A signal void was identified by three-dimensional TOF MRA in 28 bifurcations and corresponded to CA stenoses ranging from 40% to 99% (Fig 2⇓). Four signal voids (confirmed on review of the source images) were associated with a CA stenosis of less than 70%. Weak correlation was found between degree of stenosis and length of signal void (Spearman’s rank correlation of .25).
In 87 bifurcations both three- and two-dimensional TOF MRAs were available for comparison. For the identification of a 70% to 99% stenosis, three-dimensional TOF MRA resulted in a sensitivity of 92%, a specificity of 83%, and an accuracy of 85%, whereas the two-dimensional TOF MRA resulted in a sensitivity of 83%, a specificity of 75%, and an accuracy of 77%. There were 14 bifurcations for which a signal void found by two-dimensional TOF MRA corresponded to a measurable lumen on three-dimensional TOF MRA (Fig 3⇓), whereas in 8 bifurcations a signal void was seen with both techniques. There were no instances in which an artery with a signal void on three-dimensional TOF MRA had a measurable lumen on two-dimensional TOF MRA.
The sensitivity, specificity, and accuracy of three-dimensional TOF MRA for depicting a 60% to 99% stenosis were 93%, 83%, and 87%, respectively. Diagnostic accuracies of MRA for four different levels of CA stenosis are summarized in Table 1⇓. Overall diagnostic performance of three-dimensional TOF MRA, as defined by the area under the ROC curve, ranged from .94 for detection of a 60% to 99% stenosis to .96 for detection of a 50% to 99% stenosis. This value was .95 for detection of a 70% to 99% stenosis.
Assessment of Stenosis by DU
With a ratio of ICA/CCA PSV >4.0 used to define a positive test, the sensitivity, specificity, and accuracy of DU for 70% to 99% stenosis were 79%, 86%, and 84%, respectively. Alternatively, with ICA PSV ≥2300 mm/s used to define a positive test, these values were 94%, 83%, and 86%, respectively (Table 2⇓). We found the ICA PSV parameter to have better diagnostic performance than the ICA/CCA PSV ratio on the basis of the areas under the ROC curves for the ranges of angiographic stenoses evaluated. Specifically, for detection of a 70% to 99% stenosis, the areas under the ROC curve were .93 and .90 for ICA PSV and the ICA/CCA PSV ratio, respectively.
Assessment of Stenosis by Both MRA and DU
We evaluated whether concordant data from three-dimensional TOF MRA and DU could result in accuracies that are sufficient to obviate the need for CA. With a ratio of ICA/CCA PSV >4.0, MRA and DU were concordant in 122 of 148 arteries (82%). For these 122 arteries, the sensitivity, specificity, and accuracy for a 70% to 99% stenosis were 97%, 92%, and 93%, respectively. Twenty-six arteries (18%) with disparate findings on MRA and DU would have needed to have been imaged by CA (Table 2⇑). With ICA PSV ≥2300 mm/s, there was agreement between the noninvasive tests in 124 of 148 arteries (84%). For these 124 arteries, a sensitivity of 100%, a specificity of 91%, and an accuracy of 94% resulted. Twenty-four (16%) of these arteries would have required imaging with CA. Therefore, combining concordant data from both noninvasive tests markedly increased overall accuracy (94%) compared with three-dimensional TOF MRA (88%) or DU alone (86%).
CA identified 16 bifurcations with 90% to 99% stenosis and 17 occlusions (14 ICA and 3 CCA) in vessels imaged by MRA and CA. MRA was able to differentiate between high-grade stenoses and occlusions in all cases. Two occluded arteries were identified as patent by DU, and one patent ICA was determined to have an occlusion by DU.
Ulcerations Involving the Carotid Bifurcation
We found multiple examples in which three-dimensional TOF MRA was not able to depict ulcers that were clearly visible on CA. We graded ulcerations by quantitating their depth and width. An ulcer was considered present if the depth was greater than 0.4 mm. There was tremendous variability between readers’ results with regard to the presence and quantification of ulceration. For the detection of ulceration, three-dimensional TOF MRA yielded a sensitivity of 22% compared with CA.
With the MRA techniques used in this study, the vertical extent of coverage was limited to 120 mm on the two-dimensional TOF MRA and 70 mm on the three-dimensional TOF MRA. Thus, the entire length of the extracranial ICA was not visualized. This led to errors in two cases. In one vessel, a significant lesion in the distal ICA was not visualized, and in the second artery a lesion in the mid-CCA was not visualized.
Both CA and three-dimensional TOF MRA were used to evaluate 126 arteries for the presence of 50% to 99% stenosis of the intracranial ICA. There were 11 vessels in which CA indicated a 50% to 99% stenosis of the ICA. There were false-negative results in 2 arteries (sensitivity, 82%) in which severe occlusive disease in the more proximal extracranial ICA resulted in marked distal signal loss with inaccurate depiction of the lesions. There were false-positive results in 6 vessels (specificity, 95%); all involved areas in which the ICA was tortuous, such as the carotid siphon.
Evaluation for branch stenoses involving the anterior, middle, and posterior cerebral arteries revealed numerous stenoses on MRA that were either not visualized or much less severe when the same distribution was appropriately imaged on CA (Fig 4⇓).
Recent studies indicate the appropriateness of carotid endarterectomy for both symptomatic and asymptomatic patients with significant carotid stenoses. For carotid endarterectomy to be of maximal benefit, a low risk of stroke must be associated with the surgical procedure as well as the preoperative imaging study. Although CA has been the traditional preoperative test for carotid endarterectomy, there is a small but definite probability of stroke, ranging from 0.5% to 4%, associated with this procedure.25
Noninvasive imaging modalities including MRA and DU have been proposed as replacements for CA, because both studies have negligible immediate risks. The direct costs of these tests and patient discomfort are also less than those of CA. However, if these tests are not accurate, inappropriate treatment could result.
This study was designed to compare MRA and DU with CA. We felt information derived from both noninvasive tests could be combined in a way that would obviate the need for CA when the results of these tests were in agreement. The sensitivities and specificities that we report for MRA and DU alone are similar to those in previous studies.8 13 22 23 Three-dimensional TOF MRA alone resulted in a sensitivity of 94%, a specificity of 85%, and an accuracy of 88%. DU alone, with ICA PSV ≥2300 mm/s used as the parameter, yielded a sensitivity of 94%, a specificity of 83%, and an accuracy of 86%. For those arteries in which both noninvasive tests were in agreement, a sensitivity of 100%, a specificity of 91%, and an accuracy of 94% were achieved. Such an approach would require that all arteries with discordant MRA and DU results be imaged by CA. If this strategy were used, CA would have been necessary for only 16% of the arteries in our series.
Although our group has previously introduced the concept that overall accuracy in the diagnosis of carotid artery stenosis is significantly increased by combining the results of MRA and DU,14 15 Mittl et al9 did not find these tests to be complimentary. However, in their study two-dimensional TOF MRA was used, a technique that we have found to be less accurate than three-dimensional TOF MRA.
Because of our findings, we no longer schedule routine CA in symptomatic patients with carotid stenosis. Rather, we reserve CA for those circumstances in which MRA and DU are not in agreement. Because DU is often used as an initial screening test, the only additional cost is for MRA, which is significantly less expensive than CA.
It is possible that the NASCET and European Carotid Surgery trials will eventually demonstrate that patients with a stenosis of 50% to 70% may benefit from carotid endarterectomy. Recent data from the Asymptomatic Carotid Atherosclerosis Study support performing carotid endarterectomy in asymptomatic patients with a stenosis of greater than 60%.3 We therefore evaluated the ability of MRA to evaluate cutoff points other than 70%. We found that MRA was consistently accurate in identifying CA stenoses ranging from 50% to 80% (Table 1⇑).
It has been proposed that the identification and the morphological appearance of ulceration should influence the decision to perform carotid endarterectomy.26 We found MRA to be an extremely poor predictor of ulceration. In recent prospective studies from which current indications for carotid endarterectomy are derived, ulceration alone is not considered a criterion for surgery. Consequently, for most surgeons the insensitivity of MRA to the detection of ulcers does not preclude use of this test in the preoperative management of carotid disease.
Data from our study indicate that MRA has limited ability to accurately image the intracranial circulation. In the evaluation of branch arteries, stenoses were frequently found on MRA that could not be identified or that corresponded to minimal areas of narrowing by CA. We also evaluated the potential for MRA to depict the intracranial ICA. Two patients were identified who had high-grade lesions in the extracranial ICA in tandem with high-grade stenoses of the carotid siphon by CA. In neither patient was the carotid siphon disease depicted by MRA, presumably because of marked signal loss caused by the extracranial lesion. However, despite these two exceptions, MRA was able to assess the intracranial ICA with a sensitivity of 82% and a specificity of 95%. Therefore, MRA can provide useful information about the intracranial circulation with minimal additional cost. In addition, many surgeons will proceed with endarterectomy of an extracranial stenosis despite the existence of a tandem intracranial lesion.27 28 29
A nearly occluded carotid artery may be amenable to carotid endarterectomy, whereas a carotid occlusion, in most instances, is not. We found the combination of two- and three-dimensional TOF MRA to be an excellent means of differentiating between these two conditions. The increased sensitivity of two-dimensional TOF MRA for the slow flow in this situation has been demonstrated.30 The inability to evaluate carotid occlusion continues to be a source of inaccuracy for DU.31 In our series, two occluded carotid arteries were found to be patent by DU. In a third artery, DU predicted an occlusion; however, the artery was found to be patent by CA.
For evaluation of the carotid bifurcation, we have found that the accuracy of three-dimensional TOF MRA is significantly greater than that of two-dimensional TOF MRA. Similar findings have been reported by De Marco et al.12 However, for differentiation between high-grade stenoses and occlusions, two-dimensional TOF MRA provided additional information that was useful in making a final determination. Therefore, we recommend that both studies be routinely obtained in patients referred for evaluation of disease.
We propose the combined use of MRA and DU in the preoperative assessment of carotid bifurcation disease, with CA being used only for those circumstances in which the noninvasive tests are not in agreement. For such an approach to be implemented, MRA and DU need to be of high quality and require state-of-the-art equipment and interpretive expertise. Combining data from both tests can produce an accuracy as high as 94% with no false-negative results. With such a strategy, contrast angiography would be required in only a minority of patients. On the basis of our results, we advocate the combined use of these two noninvasive studies for the preoperative evaluation of patients with carotid artery disease.
Selected Abbreviations and Acronyms
|CCA||=||common carotid artery|
|ICA||=||internal carotid artery|
|MIP||=||Maximum Intensity Projection|
|MRA||=||magnetic resonance angiography|
|NASCET||=||North American Symptomatic Carotid Endarterectomy Trial|
|PSV||=||peak systolic velocity|
|ROC||=||receiver operating characteristic|
Reprint requests to K. Craig Kent, MD, Beth Israel Hospital, 330 Brookline Ave, Boston, MA 02215.
- Received June 1, 1995.
- Revision received July 11, 1995.
- Accepted July 11, 1995.
- Copyright © 1995 by American Heart Association
Laster RL Jr, Acker JD, Halford HH III, Nauert TC. Assessment of MR angiography versus arteriography for evaluation of cervical carotid bifurcation disease. AJNR Am J Neuroradiol. 1993;14:681-688.
Litt AW, Eidelman EM, Pinto RS, Riles TS, McLachlin SJ, Schwartzenberg ST, Weinreb JC, Kricheff IL. Diagnosis of carotid artery stenosis: comparison of 2DFT time-of-flight MR angiography with contrast angiography in 50 patients. AJNR Am J Neuroradiol. 1991;12:149-154.
Mittl RL, Broderick M, Carpenter JP, Goldberg HI, Listerud J, Mishkin MM, Berkowitz HD, Atlas SW. Blinded-reader comparison of magnetic resonance angiography and duplex ultrasonography for carotid artery bifurcation stenosis. Stroke. 1994;25:4-10.
Anderson CM, Saloner D, Lee RE, Griswold VJ, Shapeero LG, Rapp JH, Nagarkar S, Pan X, Gooding GAW. Assessment of carotid artery stenosis by MR angiography: comparison with X-ray angiography and color-coded Doppler ultrasound. AJNR Am J Neuroradiol. 1992;13:989-1003.
Sitzer M, Furst G, Fischer H, Siebler M, Fehlings T, Kleinschmidt A, Kahn T, Steinmetz H. Between-method correlation in quantifying internal carotid stenosis. Stroke. 1993;24:1513-1518.
Li W, Kramer J, Kleefield J, Edelman RR. MRA of the extracranial carotid arteries using a two-slab oblique 3-D acquisition. AJNR Am J Neuroradiol. 1992;13:1423-1428.
Hankey GJ, Warlow CP, Sellar RJ. Cerebral angiographic risk in mild cerebrovascular disease. Stroke. 1990;21:209-222.
Keller P. Time-of-flight magnetic resonance angiography. Neuroimag Clin N Am. 1992;4:639-656.