Catheter Angiography Is Still Necessary for the Measurement of Carotid Stenosis

What is the role of noninvasive imaging tests such as MR angiography (MRA), Doppler sonography (DUS), or computed tomographic angiography in the diagnostic evaluation of patients with suspected carotid stenosis? Can they be used instead of conventional angiography to identify candidates for surgery? Should they be used as screening tests to limit conventional angiography to those with a high likelihood of significant stenosis or an uncertain degree of stenosis, as suggested by Hatout et al in this issue of the AJNR? Is a combination of noninvasive tests better than one alone? The answers to these questions are not entirely clear and will vary from one practice to another.

For historical and scientific reasons, the measurement of luminal diameter narrowing by conventional angiography remains the single validated method for the identification of candidates for surgical carotid endarterectomy (CEA). All three pivotal multicenter randomized trials of CEA—the North American Symptomatic Carotid Endarterectomy Trial (NASCET [1]), the European Carotid Stenosis Trial (2), and the Asymptomatic Carotid Atherosclerotic Study—used luminal diameter narrowing by conventional angiography as enrollment criteria (3). Furthermore, in NASCET, the degree of stenosis by angiography correlated with increased risk of stroke with medical therapy (1). Cross-sectional area reduction, increased blood velocity, intraplaque hemorrhage, and the presence of plaque ulceration may all be related to stroke risk as well, but the use of these features to select patients for surgery has not been established in a randomized clinical trial.

Consequently, the use of noninvasive modalities in lieu of conventional angiography must be guided by two kinds of data. First, validation studies of the accuracy of these methods for the detection of luminal diameter narrowing as compared with conventional angiography must be performed. It is critical to note that these studies must be done on an individual institutional or machine basis: one cannot assume that published data reflect the performance of any individual Doppler laboratory or MR imager (4). In addition, for any given method, particularly DUS, the accuracy of different threshold values should be tested for different applications. For example, one threshold velocity value may be optimal for screening out patients with <70% stenosis and another for the detection of patients with >70% stenosis. Second, because none of these tests will have 100% accuracy, cost-effectiveness studies must be performed (5). These studies examine the trade-off between the reduction in costs and risks of stroke with angiography versus the added costs and risks of surgery in patients who have false-positive noninvasive studies and the costs and risks of stroke in the patients with false-negative studies.

The literature is full of validation and cost-effectiveness studies in this area, yet no clear consensus has been developed. A problem with many validation studies has been bias—including publication bias (only good results are published) and verification bias (only positive studies will be confirmed by angiography). The problem with many cost-effectiveness studies is their critical dependence on the assumptions and data that are fed into their mathematical models—accuracy of the noninvasive methods, risks of angiography and surgery. These data will be locally variable, and therefore the extent to which the data from any published study is applicable to any given institution or practice will be variable.

The decision on what combination of imaging methods should be used for the diagnosis of carotid stenosis must be based on locally generated data, including the risk of stroke with angiography, the accuracy of locally available noninvasive methods, and, to some extent, the risks of surgical or endovascular treatment. Another important factor in this decision is whether the patient is symptomatic. The penalty for a false-negative study in a symptomatic patient is much greater, in terms of stroke risk, than for an asymptomatic patient. For example, the noninvasive diagnosis of occlusion in an asymptomatic patient does not require conventional angiography for confirmation, given the very low risk of stroke with medical therapy even if this was actually a high-grade stenosis. The possibility of a high-grade stenosis in a symptomatic patients, on the other hand, should be pursued with conventional angiography, because the risk of stroke at 2 years with medical therapy may be as high as 30% (1).

In many practices, it may be reasonable to use DUS, MRA, or both to screen patients before angiography, provided that the local accuracy of these tools has been established. For DUS, this will require the identification of the optimal velocity screening thresholds. We adopted the use of DUS for this application at our institution after an extensive validation study (6). In addition, as shown Hatout et al in this issue of the AJNR, a well-validated noninvasive method may also allow the accurate diagnosis of severe stenosis. The degree of stenosis by gadolinium-enhanced MRA had a 95% confidence level of ±13.6%. Consequently it would appear to be reasonable to proceed with intervention at their institution with a gadolinium-MRA measurement of 80% or greater. At their institution, the accurate diagnosis of carotid stenosis for patients with >50% but <80% stenosis by MRA requires conventional angiography.

Despite the lack of strong data validating a completely noninvasive method to the selection of patients for CEA, many institutions have adopted this approach. There may be other factors driving this development, beyond the intent of reducing the costs and risks associated with angiography. MRA and DUS tend to overestimate stenosis. A complete reliance on these tests will likely increase the volume of CEA performed at a given institution, with a reduction in per-patient costs. This will be profitable to the hospital and vascular surgeon. Furthermore, conventional angiography is time and physician intensive. In many busy practices, it may be more efficient and profitable to keep the radiologist in the reading room interpreting noninvasive studies than tied up in the angiography suite.

In conclusion, the use of noninvasive carotid imaging tools to limit the use of conventional angiography in patients with possible carotid stenosis can be justified in some situations. These applications include both screening out patients with minimal stenoses or complete occlusion from further evaluation and identifying patients with high-grade stenosis as candidates for intervention. The appropriateness of these two applications, however, requires rigorous validation of the local accuracy of the noninvasive approach. In addition, whether the patient is symptomatic must be taken into consideration. The patients in the middle range still require conventional angiography for the accurate measurement of stenosis, to make appropriate treatment decisions. Angiographic complication rates, however, must be within acceptable limits.