OBJECTIVE Arteriovenous malformations (AVMs) present no pathologic tissue, and radiation dose is confined in a clear targeted volume. The authors retrospectively evaluated the role of the biologically effective dose (BED) after Gamma Knife radiosurgery (GKRS) for brain AVMs. METHODS A total of 149 consecutive cases of unruptured AVMs treated by upfront GKRS in Lille University Hospital, France, were included. The mean length of follow-up was 52.9 months (median 48, range 12-154 months). The primary outcome was obliteration, and the secondary outcome was complication appearance. The marginal dose was 24 Gy in a vast majority of cases (n = 115, 77.2%; range 18-25 Gy). The mean BED was 220.1 Gy2.47 (median 229.9, range 106.7-246.8 Gy2.47). The mean beam- on time was 32.3 minutes (median 30.8, range 9-138.7 minutes). In the present series, the mean radiation dose rate was 2.259 Gy/min (median 2.176, range 1.313-3.665 Gy/min). The Virginia score was 0 in 29 (19.5%), 1 in 61 (40.9%), 2 in 41 (27.5%), 3 in 18 (12.1%), and 4 in 0 (0%) patients, respectively. The mean Pollock-Flickinger score was 1.11 (median 1.52, range 0.4-2.9). Univariate (for obliteration and complication appearance) and multivariate (for obliteration only) analyses were performed. RESULTS A total of 104 AVMs (69.8%) were obliterated at the last follow-up. The strongest predictor for obliteration was BED (p = 0.03). A radiosurgical obliteration score is proposed, derived from a fitted multivariable model: (0.018 x BED) + (1.58 x V12) + (-0.013689 x beam- on time) + (0.021 x age) - 4.38. The area under the receiver operating characteristic curve was 0.7438; after internal validation using bootstrap methods, it was 0.7088. No statistically significant relationship between radiation dose rate and obliteration was found (p = 0.29). Twenty-eight (18.8%) patients developed complications after GKRS; 20 (13.4%) of these patients had transient adverse radiological effects (perilesional edema developed). Predictors for complication appearance were higher prescription isodose volume (p = 0.005) and 12-Gy isodose line volume (V12; p = 0.001), higher Pollock-Flickinger (p = 0.02) and Virginia scores (p = 0.003), and lower beam- on time (p = 0.03). CONCLUSIONS The BED was the strongest predictor of obliteration of unruptured AVMs after upfront GKRS. A radiosurgical score comprising the BED is proposed. The V12 appears as a predictor for both efficacy and toxicity. Beam-on time was illustrated as statistically significant for both obliteration and complication appearance. The radiation dose rate did not influence obliteration in the current analysis. The exact BED threshold remains to be established by further studies.