From 1980 through 1982, intravenous extracranial digital subtraction angiography (DSA) was performed in 6684 patients at the Cleveland Clinic. Of these, 211 previously unoperated patients had prior transient ischemic attacks (TIA) or strokes in conjunction with greater than 50% carotid stenosis on unequivocal DSA studies. Nonoperative management was employed in 126 patients, including 120 (84%) who received antiplatelet therapy or formal anticoagulation. The remaining 85 patients underwent carotid endarterectomy. During a mean follow-up interval of 36 months, there was no overall difference in survival, late TIA, or stroke between the nonoperated and surgical cohorts. However, parity was maintained in these groups only because of the prevalence of intermediate (50-69%) stenotic lesions and stable internal carotid occlusions in the nonoperated group. Carotid endarterectomy provided superior stroke prevention for patients with greater than 70% unilateral stenosis (p = 0.04), for those with greater than 50% bilateral stenosis (p = 0.004), and for those with internal carotid occlusions associated with greater than 50% contralateral stenosis (p = 0.03). The special risk of such discrete subsets should be recognized in order to plan appropriate treatment at the present time, as well as to conduct safe clinical trials in the future.