Background and purpose: The accuracy of routine ultrasonography in detecting severe carotid artery stenosis was evaluated in comparison with cerebral angiography. The precision of ultrasonographic criteria in predicting the risk of stroke was also assessed.
Methods: A total of 1011 symptomatic carotid bifurcations were studied in patients from the North American Symptomatic Carotid Endarterectomy Trial (NASCET). Given that all patients were considered for entry into the trial, the chance of a verification bias affecting the analyses was minimized. The ultrasonographic data consisted of peak systolic velocities and frequency changes from both the internal and common carotid arteries. Angiographic stenosis was calculated as in NASCET. Receiver operating characteristic (ROC) curves were constructed from the ultrasonographic data for the detection of 70% or greater stenosis on the basis of an angiographic assessment. Kaplan-Meier stroke-free survival curves were used to predict the risk of stroke.
Results: The areas under the ROC curves ranged from 0.74 to 0.75 (95% confidence interval [CI], 0.69 to 0.79). The sensitivities and specificities ranged from 0.65 to 0.71. The risk of stroke at 18 months declined sharply as the degree of angiographically defined stenosis declined from 99% to 70%. No pattern of decline was apparent on the basis of the ultrasonographic data.
Conclusions: The results indicate that the accuracy of ultrasonography is moderate when flow parameters are used to assess the degree of stenosis. Ultrasonography should be used as a screening tool to exclude patients with no carotid artery disease from further testing. Conventional angiography remains an essential investigation before assigning the risk of stroke and deciding appropriate treatment for extracranial carotid artery disease.