Background and PurposeArteriovenous malformations (AVMs) have an overall 2% to 4% annual risk of hemorrhage. The purpose of this study was to determine whether specific clinical and radiographic factors predispose AVMs to bleed and to predict the bleeding risk for individual AVM patients.MethodsWe reviewed the clinical histories and cerebral angiograms of 315 AVM patients who underwent stereotactic radiosurgery at our center. One half of the patient data (analysis cohort) was used to determine risk factors for bleeding and to construct AVM hemorrhage risk groups. These risk groups were then tested with the second half of the patient data (test cohort).ResultsThe mean AVM volume was 4.0 +/- 3.4 mL (approximate maximum diameter of 2 cm). One hundred ninety-six initial hemorrhages occurred in 10 348 patient-years for an annual initial bleed rate of 1.89%; 44 of these 196 patients had a repeat bleed in 591 patient-years for an annual rebleed rate of 7.45%. The overall crude annual hemorrhage rate was 2.40%. Multivariate analysis revealed three factors associated with hemorrhage: history of a prior bleed (relative risk [RR], 9.09; 95% confidence interval [CI], 5.44 to 15.19; P < .001), a single draining vein (RR, 1.66; 95% CI, 1.13 to 2.38; P < .01), and a diffuse AVM morphology (RR, 1.64; 95% CI, 1.12 to 2.46; P < .01). Four AVM hemorrhage risk groups were constructed on the basis of the significant factors. The annual rate of bleeding was 0.99% for low-risk AVMs, 2.22% for intermediate-low-risk AVMs, 3.72% for intermediate-high-risk AVMs, and 8.94% for high-risk AVMs.ConclusionsAnalysis of a large group of AVM patients who underwent stereotactic radiosurgery demonstrated that small AVMs have an annual hemorrhage risk similar to that of the general AVM population. AVM patients have a wide variability of bleeding risk that can be predicted from their clinical presentation and the angiographic characteristics of the AVM. The management of AVM patients should be based not only on the morbidity of the proposed treatment but also those factors that predispose individual patients to either a low or high hemorrhage risk.(Stroke. 1996;27:1-16.)