Connectome-sparing stereotactic radiosurgery: Results from the phase II COG-SRS trial.

2002 Background: Cognitive decline remains common after radiosurgery. The cognition-sparing stereotactic radiosurgery trial (COG-SRS, NCT04343157) evaluated a novel connectome-sparing SRS strategy leveraging advanced imaging to autosegment the cognitive connectome for individualized functional sparing. Methods: COG-SRS is a prospective, single-institution, phase II trial that enrolled adults (KPS≥70) treated with SRS for benign or malignant intracranial lesions from 06/2018-10/2025. Connectome structures (10 discrete eloquent bilateral white matter tracts [WMTs], corpus callosum, and hippocampi [HC]) were auto-segmented from pretreatment diffusion tensor imaging and structural MRI and assigned evidence-based dose constraints (single fraction Dmax WMT=12Gy, HC=8.4Gy, with 3-/5-fraction equivalent doses) during SRS planning. Neuropsychological testing assessed cognitive domains of memory, language, executive function, attention/processing speed, and fine motor control at baseline and 3 months post-SRS. Outcomes were summarized using normed T scores and analyzed with mixed effects models adjusting for baseline score and prespecified clinical covariates. In the brain metastases cohort, cognitive performance was also compared with historical controls treated with standard brain SRS on Alliance N0574/N107C using domain Z scores and baseline adjusted mixed effects models. Questionnaires assessed mood/QOL, and disease time to event endpoints were estimated by Kaplan Meier. Results: Ninety-two patients enrolled (mean age 57.8 years), including 82 (89%) with brain metastases. All plans met standard target coverage and organ-at-risk constraints while optimizing dose to the cognitive connectome, with 76% meeting predefined constraints for all 23 connectome structures. At 3 months, 74% of patients demonstrated stable-to-improved performance across all 5 cognitive domains, with no significant mean decline in any domain. Mean performance improved in executive function (+4.30 T; p=0.043) and language (+4.91 T; p=0.011). Compared with historical controls, COG-SRS demonstrated superior 3 month cognitive outcomes (global composite Z difference +0.38; p=0.001), with the most pronounced domain-level benefit in executive function (+0.71; p=0.009) and language (+0.35; p=0.010). Mood and QOL remained stable. Among patients with brain metastases (median followup 35.6 months), 1 year local control was 98.5%, freedom from distant intracranial failure 57.4%, and overall survival 66.7%. Conclusions: Connectome-sparing brain SRS is feasible, leveraging advanced MRI-based autosegmentation of patient-specific cognitive networks, and is associated with preserved early cognition at the cohort level with superior 3 month cognitive outcomes versus historical controls. Cognitive benefits were achieved without compromising oncologic control, supporting prospective randomized study. Clinical trial information: NCT04343157 .