Abstract
Background: Imaging of the carotid arteries is important for the evaluation of patients with ischemic stroke or TIA. CT angiography (CTA) of the head and neck is readily available and can be part of the routine imaging of stroke patients. To evaluate the accuracy of CTA, the authors compared the degree of stenosis found using CTA with digital subtraction angiography (DSA) in consecutive patients during a 3-year period.
Methods: The authors included all patients with interpretable CTA and DSA of the cervical carotid arteries from April 2000 to November 2002 at a single academic medical center. This yielded a total of 81 vessels. Stenosis on CTA of the internal carotid artery was measured in the axial plane at the point of maximum stenosis and referenced to the distal cervical internal carotid by two blinded readers. Two blinded readers measured stenosis from the DSA using the North American Symptomatic Carotid Endarterectomy Trial method.
Results: Using a 70% cutoff value for stenosis, CTA and DSA were in agreement in 78 of 81 (96%; 95% CI, 90 to 99%) vessels. CTA was 100% sensitive (n = 5) and 63% specific (95% CI, 25 to 88%), and the negative predictive value of a CTA demonstrating <70% stenosis was 100% (n = 73).
Conclusions: In this consecutive series of patients with CT angiography of the neck and digital subtraction angiography, the authors found that CT angiography has a high sensitivity and high negative predictive value for carotid disease. CT angiography appears to be an excellent screening test for internal carotid artery stenosis, and the authors advocate its use for the initial imaging of patients with suspected stroke or TIA.
Imaging of the carotid arteries is recommended as part of the diagnostic evaluation of a patient with ischemic stroke or TIA because patients with high-grade carotid stenosis will benefit from carotid endarterectomy if used properly and may benefit from stenting for secondary prevention of ischemic stroke.1–3⇓⇓ The accepted gold standard for evaluation of carotid artery stenosis is catheter angiography; however, this technique is expensive and invasive and has substantial risks.4 The techniques of duplex ultrasonography, MR angiography (MRA), and CT angiography (CTA) have the advantage of being noninvasive, and MRA and CTA provide multiple viewing planes and therefore may not suffer from limitations of the finite viewing planes of conventional angiography. Because CT is the recommended evaluation tool for ischemic stroke, coupling CTA with CT imaging during acute evaluation of stroke could expedite diagnostic evaluation.5 Thus, symptomatic carotid disease could be identified immediately, and treatment decisions could proceed more rapidly (e.g., urgent endarterectomy for TIA); alternatively, elimination of the carotid arteries as a source of emboli could direct the workup toward other sources, including the heart and the intracranial vasculature.
We use a stroke CT protocol for all acute stroke and TIA patients evaluated at our institution consisting of a noncontrast CT of the head, followed by CTA from the aortic arch through the circle of Willis, and two-level CT perfusion of the brain.6 We tested the ability to identify carotid stenosis by simple evaluation of the CTA cross-sectional images (partition images) alone without use of advanced reconstruction software. Such a method has the advantage of being performed at the scanner console or remote reader station and can be implemented with no additional software or hardware to existing multidetector CT scanners. The goal of this study is to determine the accuracy of CTA for evaluation of carotid stenosis compared with conventional angiography using this simple method to determine carotid stenosis.
Methods.
The radiology database at our institution was searched to identify all patients who received a CTA of the head and neck from April 2000 through November 2002. The brain and neck were imaged using a multidetector CT scanner (GE Lightspeed, Waukesha, Wisconsin) with the following protocol: noncontrast axial CT of the brain (3.5-mm thickness), CTA from vertex to aortic arch (70 to 110 mL contrast using 18 to 20 gauge IV at 3 to 4 mL/s; scan delay, 20 seconds; 3:1 pitch; slice thickness, 1.25 mm; 120 KV; and 170 MA), and dynamic CT perfusion using the toggling-table technique7 with 40 mL of IV contrast or at two locations with a fixed table location using 40 mL contrast each.
A neurologist and a neuroradiologist, blinded to patient data and radiology interpretation, interpreted the CTAs. Each artery was evaluated using magnified axial sections at a computer workstation. No three-dimensional reconstructions were used in evaluation. Using magnification, a measure (dmin) was taken of the diameter of the narrowest portion of the cervical internal carotid in the axial plane. This was compared with the maximal diameter (dnorm) of the cervical internal carotid artery distal to the carotid bulb at a location in which the imaging plane was orthogonal to the artery, the arterial walls were parallel, and where there was no arterial disease (figure 1, A and B). The degree of cross-sectional stenosis was calculated in percent as: percent stenosis = (1 − dmin/dnorm) × 100%, analogous to the method used in the North American Symptomatic Carotid Endarterectomy Trial (NASCET).1 Examiners also reported whether calcification was present in the artery, whether a radiographic “string sign” was observed, and whether a total occlusion was seen. The final stenosis was calculated as the average between that calculated by each examiner.
Figure 1. Method of CT angiographic (CTA) stenosis measurement and example of CTA “overcall” of stenosis compared with digital subtraction angiography (DSA). (A) The cross-section at maximum stenosis of the internal carotid artery in the axial plane was measured (dmin) and referenced to the diameter of the distal cervical carotid (B) imaged in the orthogonal plane (dnorm) to calculate stenosis of 76% (see Methods section). (C) DSA of carotid artery imaged in A and B, revealing a 43% stenosis.
Digital subtraction angiography (DSA) was performed, and a neuroradiologist selected the viewing angle revealing the maximum degree of internal carotid stenosis for blinded interpretation. A second neuroradiologist and neurologist blinded to other patient data measured the degree of stenosis on printed film using digital calipers (Mitutoyo Corporation, Aurora, IL). The percentage stenosis was calculated according to the NASCET trial, as above.
Interobserver agreement was calculated for cutoff points of 50% and 70% stenosis between observers of CTAs and observers of DSA.8
Sensitivity, specificity, and negative predictive value were calculated for CTA using DSA as the gold standard with 50% and 70% cutoff points for stenosis. Exact binomial 95% CIs were calculated for proportions between 1 and 99%. Receiver operator curve areas were calculated. All statistical analyses were performed with Stata, version 8 (College Station, TX) or Excel, version 2002 (Microsoft Corporation, Redmond, WA).
Results.
Search of the radiology database during the study period revealed 238 patients who had CTA of the head and neck. Of these, 66 patients (132 vessels) underwent DSA within 2 months of the CTA examination. In 26 vessels, the angiogram was unavailable for review. In 24 vessels, the CTA was of poor quality either because of inadequate contrast bolus or excessive streak artifact from dental hardware. In one vessel, the conventional angiogram was of poor quality. This yielded a total of 81 vessels for our study (figure 2).
Figure 2. Flow diagram of study subjects.
The relationship between CTA and conventional angiographic stenosis is shown for all arteries in figure 3. Using a 70% cutoff value, CTA and DSA were in agreement in 78 of 81 (96%; 95% CI, 90 to 99%) vessels (table 1). Three of 81 (4%) vessels were evaluated by CTA as >70% stenosis but were found to be <70% by DSA. As an example, CTA was interpreted to read 76% stenosis, and DSA showed 43% in a highly calcified vessel as shown in figure 1. CTA was 100% sensitive (n = 5) and 63% specific (95% CI, 25 to 88%), and the negative predictive value of a CTA demonstrating <70% stenosis was 100% (n = 73).
Figure 3. Scatter plot of conventional angiographic-measured stenosis (ordinate) and CT angiographic-measured stenosis (abscissa). Cross-bars are at 50 and 70% stenosis; some data points overlap.
Table 1 Accuracy of CTA compared with DSA at 70% stenosis
Because there is marginal benefit of carotid endarterectomy for 50 to 69% carotid stenosis,9 we also chose to explore this cutoff point. Using a 50% cutoff value, CTA and DSA were in agreement in 72 of 81 (89%; 95% CI, 81 to 95%) vessels (table 2). CTA was 86% sensitive (95% CI, 57 to 100%) and 43% specific (95% CI, 14 to 71%), and the negative predictive value of a CTA demonstrating <50% stenosis was 99% (95% CI, 94 to 100%).
Table 2 Accuracy of CTA compared with DSA at 50% stenosis
Receiver operator curves were calculated for the 70% and 50% cutoff points. The area under the curve was 0.99 for the 70% cutoff point and 0.96 for the 50% cutoff point.
Interobserver agreement between CTA and DSA measurements of stenosis were calculated for both pairs of blinded examiners (table 3). Although agreement was high with both techniques of carotid imaging, the kappa scores for CTA were higher than those of DSA, suggesting that a single interpretation of CTA stenosis would be more reliable than a single interpretation of DSA stenosis.
Table 3 Interobserver agreement, kappa value,Z andp values between pairs of blinded readers for CTA and DSA measurements of stenosis
Discussion.
In this consecutive series of patients who had CTA of the neck as part of our routine stroke imaging protocol and who also had conventional angiography, we found that CTA had a high sensitivity and negative predictive value for internal carotid artery stenosis. The high negative predictive value (100% for 70% stenosis and 99% for 50% stenosis) can assure the clinician that a negative result is highly predictive of minimal carotid disease, allowing one to avoid further imaging of the artery. The high sensitivity (100% for 70% stenosis and 86% for 50% stenosis) means that the test is unlikely to miss important carotid disease. Additionally, the method we used to measure the degree of carotid stenosis can be performed without additional software, making it readily and widely available. Previous studies have compared CTA evaluation of carotid artery stenosis with that of MRA, DSA, ultrasound, or some combination of these modalities and have demonstrated good agreement with conventional angiography.10–13⇓⇓⇓ Some of these studies used software-driven reconstruction techniques in CTA evaluation,10,11⇓ whereas others used patients selected based on known ultrasound or DSA stenoses.11,14⇓ Our results confirm these studies and use a simpler method of measure with a consecutive series of unselected patients.
The finding that interobserver agreement was higher for CTA-measured stenosis than with DSA-measured stenosis has important clinical implications. Although these results must be considered preliminary, this finding implies that the clinician can be more confident in the single interpretation of stenosis—as typically occurs in clinical practice—measured by CTA compared with conventional angiography. There is debate that conventional angiography is the gold standard for documenting carotid disease.15 Our results support the notion that cross-sectional techniques may provide more accurate imaging of carotid anatomy.13,14⇓
There were a small number of arteries in which CTA overcalled stenosis (three in the 70% cutoff group). Retrospective examination of these studies showed that two were secondary to calcification artifact on the CTA (see figure 1). The third false-positive CTA was secondary to a distal supraclinoid complete occlusion that resulted in decreased flow and thus caliber of the cervical internal carotid. This “string sign” also causes an under-reporting of stenosis on conventional angiography because the internal carotid artery collapses under a low-flow state. Because of these false-positive findings, alternate confirmatory studies of stenosis should be considered, especially if there is carotid calcification.
Our study has several limitations. Selection bias may have made this sample of patients unrepresentative. Not every patient who had a CTA had DSA during the study period. Some patients had DSA because of concerns of intracranial atherosclerosis, aneurysm, or arteriovenous malformation, for example, and some received DSA for confirmation of carotid stenosis. However, the number of arteries with severe carotid disease was only 10% of the vessels sampled. This prevalence of stenosis reasonably agrees with the pretest probability of carotid disease in a population of stroke patients; therefore, the performance characteristics of CTA found in our study are likely relevant to clinical practice. We did not perform measurements of stenosis on three-dimensional reconstructions of the carotid arteries; therefore, the method could miss stenosis within a vessel segment that is parallel to the image plane. We did not observe this potential false-negative finding in our study, but this could be observed in a larger series of patients.
In the routine evaluation of a patient with anterior circulation stroke or TIA, imaging of the carotid artery is indicated, even if there is a likely source of cardioembolism.16 Clinicians often order duplex ultrasonography or MRA, which require, at minimum, patient transport and, at most, a delay of several days to weeks. Because second stroke from a symptomatic carotid lesion may occur shortly after first stroke or TIA, this potentially long waiting time before carotid disease is identified places patients at risk.17 A previous study has shown that the administration of IV contrast at the doses used for this protocol is safe.6 Because this imaging sequence is also feasible in acute stroke patients,6 we advocate the routine addition of CTA of the neck to current society recommendations of CT imaging of the brain for patients with stroke and TIA.5
Footnotes
-
See also page 412
- Received December 29, 2003.
- Accepted May 25, 2004.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.