OBJECT Large arteriovenous malformations (AVMs) remain difficult to treat, and ideal treatment parameters for volume-staged stereotactic radiosurgery (VS-SRS) are still unknown. The object of this study was to compare VS-SRS treatment outcomes for AVMs larger than 10 ml during 2 eras; Era 1 was 1992-March 2004, and Era 2 was May 2004-2008. In Era 2 the authors prospectively decreased the AVM treatment volume, increased the radiation dose per stage, and shortened the interval between stages. METHODS All cases of VS-SRS treatment for AVM performed at a single institution were retrospectively reviewed. RESULTS Of 69 patients intended for VS-SRS, 63 completed all stages. The median patient age at the first stage of VS-SRS was 34 years (range 9-68 years). The median modified radiosurgery-based AVM score (mRBAS), total AVM volume, and volume per stage in Era 1 versus Era 2 were 3.6 versus 2.7, 27.3 ml versus 18.9 ml, and 15.0 ml versus 6.8 ml, respectively. The median radiation dose per stage was 15.5 Gy in Era 1 and 17.0 Gy in Era 2, and the median clinical follow-up period in living patients was 8.6 years in Era 1 and 4.8 years in Era 2. All outcomes were measured from the first stage of VS-SRS. Near or complete obliteration was more common in Era 2 (log-rank test, p = 0.0003), with 3- and 5-year probabilities of 5% and 21%, respectively, in Era 1 compared with 24% and 68% in Era 2. Radiosurgical dose, AVM volume per stage, total AVM volume, era, compact nidus, Spetzler-Martin grade, and mRBAS were significantly associated with near or complete obliteration on univariate analysis. Dose was a strong predictor of response (Cox proportional hazards, p < 0.001, HR 6.99), with 3- and 5-year probabilities of near or complete obliteration of 5% and 16%, respectively, at a dose < 17 Gy versus 23% and 74% at a dose ≥ 17 Gy. Dose per stage, compact nidus, and total AVM volume remained significant predictors of near or complete obliteration on multivariate analysis. Seventeen patients (25%) had salvage surgery, SRS, and/or embolization. Allowing for salvage therapy, the probability of cure was more common in Era 2 (log-rank test, p = 0.0007) with 5-year probabilities of 0% in Era 1 versus 41% in Era 2. The strong trend toward improved cure in Era 2 persisted on multivariate analysis even when considering mRBAS (Cox proportional hazards, p = 0.055, HR 4.01, 95% CI 0.97-16.59). The complication rate was 29% in Era 1 compared with 13% in Era 2 (Cox proportional hazards, not significant). CONCLUSIONS VS-SRS is an option to obliterate or downsize large AVMs. Decreasing the AVM treatment volume per stage to ≤ 8 ml with this technique allowed a higher dose per fraction and decreased time to response, as well as improved rates of near obliteration and cure without increasing complications. Reducing the volume of these very large lesions can facilitate a surgical approach for cure.