Ninety-seven patients were referred to the Toronto Hospital (Western Division) between January 1984 and January 1992 for selective angiographic embolization (108 embolizations, including repeat procedures) to control intractable or recurrent severe epistaxis. Eighty-one patients (comprising 94 embolizations) were referred on an emergent basis because of failure of conventional conservative therapy, consisting of anterior and posterior packing. The remaining 16 patients (14 embolizations) were referred electively for recurrent epistaxis. A retrospective review of these cases was performed, with long-term telephone follow-up achieved in over 95% of cases. Embolization safely controlled active hemorrhage in 88% of the emergent cases. The success rate increased to 90% when two cases in which the source of epistaxis was found to be from the internal carotid artery were excluded (because these vessels could not be safely embolized). Of the patients whose epistaxis was initially controlled by embolization, 82% had no further nosebleeds (follow-up time ranged from 2 to 82 months; average, 26.8 months). More than half of the long-term failures were seen in patients with Osler-Weber-Rendu disease. Overall, the mortality rate was 0% and the long-term morbidity rate was 2% (one cerebral vascular accident and one case of skin slough in the territory of the superficial temporal artery.