Modified Linac radiosurgery was utilized at our institution between 1990 and 1995 in 54 patients with cerebral arteriovenous malformations (AVMs) as part of a multimodality therapeutic approach. Most patients also underwent surgery and embolization of the AVMs prior to radiosurgery. The goal of the adjunctive radiosurgical treatment was the complete angiographic obliteration of the deep residual AVM after subtotal embolization and resection. Radiosurgery was used as the primary treatment of some small deep AVMs which were judged to have a high risk of morbidity if approached surgically. Of the 54 patients, 37 have had follow-up angiograms available for review. The median time interval between radiosurgery and the most recent angiogram was 22 months (range 5-66 months). This is a retrospective report on the angiographic follow-up available on this selected group of patients with difficult AVMs. All angiograms were evaluated for the presence of residual AVM. If any residual AVM was present, a decrease in size of more than 50% was noted as significant. The hospital and office charts of all patients were reviewed and the patients and families were contacted for up-to-date clinical follow-up. Prior to treatment, 11 patients had Spetzler-Martin grade 5 AVMs, 12 patients had grade 4 AVMs, and 14 patients had grade 3 AVMs. There were no grade 1 or 2 AVMs treated with radiosurgery. Of these 37 patients, 13 (35%) were completely obliterated on follow-up angiography. Three of the 11 (27%) grade 5 AVMs were completely obliterated. Seven of the 12 (58%) grade 4 AVMs were obliterated. Three of the 14 (21%) grade 3 AVMs have documented radiosurgical obliteration, however 8 others have had a significant decrease in size on the first posttreatment angiogram (median follow-up 14 months) and have further angiographic follow-up pending. There were 5 hemorrhages after radiosurgery, 1 was fatal and 1 resulted in permanent morbidity. Six patients have been reembolized since radiosurgery and 2 of those had their AVMs surgically resected thereafter. Only 1 patient of 6 (17%) in this series who was treated with radiosurgery alone has had documented AVM obliteration. None of 3 treated with embolization alone prior to radiosurgery have had complete obliteration. Three of the 5 patients (60%) who had only surgery prior to radiosurgery had their AVMs completely obliterated. Of the 23 patients who had all three treatments (embolization, surgery, and radiosurgery), 10 (43%) had complete obliteration of their AVMs. One patient (2%) developed moderate permanent neurological disability as a result of radiosurgery-induced necrosis. This patient did have complete obliteration of her AVM. These data support the use of radiosurgery for treatment of cerebral AVMs as part of a multimodality approach if the surgical risk of any residual AVM after embolization and microsurgical resection is deemed excessive. The adjunctive use of radiosurgery in this series resulted in the safe complete obliteration of many very difficult grade 4 and 5 AVMs. These data do not support the use of radiosurgery as the primary treatment of surgically resectable AVMs since there is a risk of interval hemorrhage and the obliteration rate is far inferior to that of microsurgical resection.