Intervention versus observation for brain arteriovenous malformations: long-term hemorrhagic outcomes in a single-center cohort study

OBJECTIVE: To compare long-term hemorrhagic and functional outcomes between intervention and observation for brain arteriovenous malformations (bAVMs). METHODS: We performed a retrospective, single-center cohort study of angiographically confirmed bAVMs, ruptured and unruptured, evaluated between 2000 and 2025. Initial management was observation or intervention (embolization, microsurgery, stereotactic radiosurgery or multimodal therapy). The primary endpoint was any bAVM-related intracranial hemorrhage from diagnosis, including perioperative and follow-up events; secondary endpoints included follow-up hemorrhage, perioperative hemorrhage and a favorable functional outcome at last follow-up (modified Rankin Scale (mRS) 0-2). Kaplan-Meier and multivariable Cox regression were used, with inverse probability of treatment weighting and time-dependent analyses to assess robustness. RESULTS: Of the 337 screened patients, 220 were eligible (108 ruptured, 112 unruptured); 138 (63%) underwent initial intervention and 82 (37%) were observed. Over 1943 patient-years (median, 7.4 (IQR, 1.9-14) years), 28 follow-up hemorrhages and 13 perioperative hemorrhages occurred. For the primary endpoint, 10-year hemorrhage-free survival was 66% (observation) versus 93% (intervention; log-rank p<0.001). The annual follow-up hemorrhage rates were 3.2% and 0.4% in the observation and intervention groups, respectively. Intervention was independently associated with a lower risk of follow-up hemorrhage, with similar estimates across models and in ruptured and unruptured bAVMs. At the last follow-up, 74% of patients in the intervention group and 68% in the observation group had an mRS score of 0-2. CONCLUSIONS: Compared with observation, intervention for bAVMs was associated with lower long-term hemorrhagic risk, without worse functional outcomes. These findings support individualized, multidisciplinary treatment decisions.