Potential Ultrasound Facilitated Overuse of Carotid Interventions
Ultrasound has clearly emerged as the primary method of stratification of atherosclerosis induced internal carotid artery (ICA) stenosis. It has replaced angiography as the test of choice prior to carotid endarterectomy. For such a widely accepted diagnostic test, carotid ultrasound is subject to a surprising number of factors that contribute to its variability. Technologists, laboratory protocols, physicians, the ultrasound machines themselves, the method of measuring angiographic stenosis, and individual patient populations can all influence proposed criteria for ultrasound stratification of ICA stenosis. Given these many potential sources of variability it is not surprising there exists great discrepancy in proposed criteria for stratification of ICA stenosis. This well acknowledged variability of carotid ultrasound led to the largest accrediting organization for vascular laboratories, the ICAVL, to demand individual laboratories adhere to laboratory specific protocols, and perform validation of diagnostic criteria to receive accreditation in the performance of ultrasound studies of the cervical carotid artery. It was one of the prime motivating factors for the Society of Radiology in Ultrasound (SRU) Carotid Consensus Conference. Participants in the conference were charged with reviewing recent literature and developing a reasonable set of criteria for performance and interpretation of carotid ultrasound studies and for stratification of ICA stenosis.
Conference participants easily agreed that adherence to protocols with consistently applied diagnostic criteria by all individuals within a laboratory is highly desirably. Participants also agreed that any proposed diagnostic criteria deriving from the conference potentially should have maximum clinical relevance. This meant proposing criteria relevant to incorporating threshold values of stenosis pertinent to the results of the randomized trials of carotid endarterectomy. Thus the criteria agreed to incorporate 50%, 60% and 70% levels of stenosis and minimize the importance of stratifying minor (<50%) degrees of ICA stenosis.
It is certain the opinions deriving from the conference will not please everyone. In particular, the relatively low flow velocities that the conference participants agreed upon to suggest >50%, >60% and >70% ICA stenosis have profound implications in selection of patients for prophylactic carotid endarterectomy. Given the marginal therapeutic benefit of prophylactic carotid endarterectomy documented by ACAS and the recently published European trial of carotid endarterectomy for asymptomatic carotid stenosis, strict adherence to the SRU consensus conference guidelines may contribute to an even greater increase in carotid endarectomy for asymptomatic disease.
The conference emphasized sensitivity for predicting ICA stenosis. However, while sensitivity is important, for the current clinical care of patients, positive and negative predictive values are perhaps of more value. This derives from the fact that the randomized trials of carotid endarectomy have established threshold levels of angiographic ICA stenosis beyond which symptomatic and asymptomatic patients benefit from carotid endarterectomy.
When the therapeutic index for a carotid intervention is relatively broad, i.e., the risk of a procedure is relatively small in comparison to the natural history of the disease, a noninvasive test that has a high negative predictive value to exclude the presence of disease is desirable. Lower positive predictive values may be acceptable. In such cases it is likely patients with lesions just below the "optimal" threshold will also statistically derive some benefit from the procedure. Consider a patient with symptomatic ICA stenosis. High-grade,>70% symptomatic ICA stenosis, has a relatively high therapeutic index for carotid endarterectomy. NASCET data also indicates that patients with < 70% to 50% symptomatic ICA stenosis also benefit from carotid endarterectomy, although to a lesser extent. It follows that it is undesirable to fail to identify a patient with symptoms referable to the carotid artery who also has a >70% ICA stenosis. A high negative predictive value is desirable to be sure the patient does not have an appropriate lesion for treatment, as treatment in such cases is very likely to be effective in preventing stroke. A high positive predictive value is not as important. Even if the lesion is somewhat "overcalled" it likely also will benefit from endarterectomy.
The opposite is true for patients with asymptomatic carotid stenosis. The data suggest a relatively narrow therapeutic index for prophylactic carotid endarterectomy in patients with asymptomatic ICA stenosis. In such cases, given the marginal proven benefit of carotid endarterectomy, one does not mind avoiding operation on some patients with a >60% (ACAS threshold) ICA asymptomatic stenosis. Since only about 1 in 20 will derive benefit from the operation in five years, failure to operate for an ACAS threshold lesion is less of a problem than failure to operate for a high-grade symptomatic ICA stenosis. For asymptomatic disease therefore negative predictive value is not so important. Conversely, one does not wish to subject asymptomatic patients with less than a threshold lesion to prophylactic carotid endarterectomy. For preoperative noninvasive evaluation of ICA stenosis in asymptomatic patients, one wishes the noninvasive test to have a high positive predictive value to avoid subjecting patients to a potentially dangerous procedure that carries only a small proven benefit.
Data derived from a prospective NIH funded study of progression of atherosclerosis conducted at our institution suggests using even the SRU ctiteria for >70% ICA stenosis to select patients for a prophylactic carotid intervention would result in significant overuse of prophylactic carotid intervention. The risk of an unheralded stroke in asymptomatic patients with ICA stenosis and an ICA peak systolic velocity >230cm/s but less than 290cm/s is sufficiently small that the risk of intervention likely exceeds the risk associated with the natural history of ICA stenoses with PSV between 230 and 290 cm/s.
In the future, stratification of diagnostic criteria of ICA stenosis should focus on sensitivity and specificity but also on positive and negative predictive values. Sensitivity and specificity data are very important for research studies and for assessing the accuracy of a diagnostic modality. However, if clinicians wish to utilize the SRU consensus guidelines for threshold levels of stenosis as a means to determine who is a candidate for carotid intervention, negative and positive predictive values for identifying threshold lesions are more important than sensitivity and specificity.
The consensus conference proposed diagnostic criteria will therefore not be acceptable to all laboratories or all physicians. They should not be utilized by all laboratories. The SRU criteria would clearly result in over utilization of carotid intervention at our institution. Vascular laboratories with sufficient volume of material to perform onsite validation studies should continue to do so. A locally validated set of diagnostic criteria in an individual laboratory currently remains the standard to which all laboratories should strive to achieve.
The criteria deriving from the conference should be tested with respect to their implications for clinical practice especially with regard to asymptomatic carotid stenosis. The results of such testing should influence the desire of laboratories to adopt the criteria deriving from the conference.