Recent studies have suggested that carotid endarterectomy can be performed safely based solely on the noninvasive duplex ultrasound evaluation in selected patients. We have prospectively evaluated 60 consecutive patients who underwent 65 carotid endarterectomies, 48 patients without preoperative angiography and 12 with angiography. Forty-two patients were operated on for symptomatic disease, and 23 procedures were done for critical, asymptomatic stenoses. Long term followup consisted of physical examination and serial duplex scans every 3-6 months postoperatively over a mean followup period of 2.4 years. Clinical management indicated by duplex ultrasound was altered in only one of the 12 patients who had preoperative angiography, a change in the timing of the endarterectomy in a symptomatic patient with an ulcerated lesion seen at angiography. At operation the severity of disease predicted by duplex ultrasound was confirmed in all cases (100 per cent sensitivity), including one > 80% diameter stenosis interpreted by angiography as occluded; no unsuspected anatomic anomalies were found at surgery. The duplex scan also correlated well with intraoperative findings of surface ulceration and gross intraplaque hemorrhage. There was one intraoperative stroke with good recovery in a patient with preoperative angiography; and there were no deaths, for a combined morbidity and mortality of 1.6 per cent. During long term followup, 97 per cent of patients have remained symptom-free. We conclude that clinical assessment with a preoperative duplex ultrasound scan of good technical quality and interpreted in collaboration with the vascular surgeon provides appropriate information on which to base carotid endarterectomy and allows a safe alternative to the routine use of preoperative angiography.(ABSTRACT TRUNCATED AT 250 WORDS)