This study prospectively compared the accuracy of published duplex ultrasonographic criteria for 70%-99% internal carotid artery (ICA) stenosis according to the North American Symptomatic Carotid Endarterectomy Trial (NASCET) method to determine angiographic stenosis. From March 1, 1995 to December 1, 1995, all patients considered for carotid endarterectomy (CEA) were studied with carotid duplex ultrasound and carotid angiography within 1 month of the ultrasound study. Duplex measurements of ICA peak systolic velocity (PSV), end diastolic velocity (EDV), and ratio of the ICA to common carotid artery (CCA) PSVs were recorded. Degree of stenosis on angiography was determined using NASCET criteria. A MEDLINE search to identify duplex ultrasound criteria to predict NASCET defined 70%-99% ICA stenosis was carried out. In addition, the original University of Washington criteria for critical stenosis (> or = 80%) was also examined. The accuracy of these criteria was determined with angiographic results and the positive predictive value (PPV) of each criterion were compared. Ninety-nine patients with 185 carotid bifurcations were available for comparison. The different duplex criteria for determining NASCET defined 70%-99% ICA stenosis were: ICA PSV > 175 cm/sec or PSV < 40 cm/sec, PSV > 230 cm/sec, ratio of ICA to CCA PSVs > 4, PSV > 130 cm/sec plus EDV > 100 cm/sec, and PSV > 270 cm/sec plus EDV > 110 cm/sec. When compared with angiography, the calculated PPVs for these criteria were 71% (73/103), 81% (71/88), 86% (67/78), 88% (62/70), and 90% (57/63), respectively. The University of Washington criteria for critical stenosis (PSV > 125 cm/sec plus EDV > 135 cm/sec) had the highest PPV at 91.6% (55/60). The University of Washington criteria for critical stenosis had the highest PPV to predict a 70%-99% angiographic stenosis.