Doppler sonographic parameters for detection of carotid stenosis: is there an optimum method for their selection?

A wide range of Doppler threshold values for carotid stenosis is found in the literature. We undertook this study to compare methods of derivation and to determine if an optimum strategy of threshold selection exists for a high-risk population.

From the sonograms of all patent internal carotid arteries, peak systolic velocity in the internal carotid artery (ICA(PSV)) and the ratio of peak systolic velocity in the internal carotid artery to that of the common carotid artery (ICA(PSV)/ CCA(PSV)) were compared with the percentage of angiographically determined stenosis. Receiver operating characteristic curves were generated for levels of stenosis > or =60% and > or =70%. Doppler thresholds were chosen on the basis of maximum accuracy and on the basis of > or =90% sensitivity and specificity. Patients were then segregated into symptomatic and asymptomatic cohorts, and the above process was repeated. An effectiveness analysis was also conducted using various Doppler thresholds. Thresholds derived using these three methods were compared and optimal values chosen. RESULTS. Of 333 carotid arteries that fit inclusion criteria, 132 were found in asymptomatic patients and 201 in symptomatic patients. Maximum accuracy, > or =90% sensitivity and specificity, and effectiveness analysis each produced different ranges of thresholds. We chose final thresholds that maintained patient outcome profiles. For asymptomatic patients at the > or =60% stenosis level, thresholds were ICA(PSV) = 200 cm/sec and ICA(PSV)/CCA(PSV) = 3.0. For symptomatic patients with stenosis > or =70%, thresholds were ICA(PSV) = 175 cm/sec and ICA(PSV)/CCA(PSV) = 2.5.

Considerable latitude exists in the choice of carotid Doppler thresholds. We propose a rational strategy for threshold selection based on a combination of three commonly used methods. Our observations indicate that it appears advisable to consider symptomatic and asymptomatic patients separately and to apply appropriately derived thresholds.