Aims The aim of the present work is to assess the main predictors of the most clinically relevant radio-induced effects after Gamma Knife stereotactic radiosurgery (GKRS) for uveal melanoma (UM). Materials and methods Medical records and three-dimensional dosimetry data of critical structures of 66 patients were retrospectively reviewed. Cox's proportional hazard model was used to identify clinical and dosimetric variables as independent risk factor for GKRS-related complications. Results The fraction of the posterior segment receiving more than 20Gy (V20), Bruch's membrane rupture and tumour thickness were significant prognostic factors for neovascular glaucoma. A clear relationship with the dose received by 1% of the optic nerve (D1%) was found for radiation retinopathy and papillopathy. Multivariables models resulted for visual acuity (VA) reduction >20% of the basal value and for complete VA loss, both including largest tumour diameter and D1% to the optic nerve. The predictive model for complete VA loss includes also Bruch's membrane rupture. An alternative model for complete visual acuity loss, including the optic nerve-prescription isodose minimum distance, was also suggested. Conclusions We found clinical and dosimetric variables to clearly predict the risk of the main side effects after GKRS for UM. These results may provide dose constraints to critical structures, potentially able to reduce side effects. Constraining D1% to the optic nerve below 12- 13Gy may result in a dramatic reduction of blindness risk, while reducing V20 of the posterior segment of the bulb could limit the neovascular glaucoma onset. © 2018 BMJ Publishing Group. All rights reserved.