ABSTRACT NUMBER: ESOC2026A2417 POSTPROCEDURAL GFAP PREDICTS CLINICALLY RELEVANT HEMORRHAGIC TRANSFORMATION AFTER ENDOVASCULAR THROMBECTOMY Authors: Ricardo Varela, Alexandre Dias, Liliana Igreja, Ana Moutinho, Beatriz Ramos, Catarina Vaz, José Silva, José Fins, Francisca Costa, Daniela Dinis, Joel Simrén, José Rocha, Rui Magalhães, Manuel Correia, Henrik Zetterberg, Luís Maia Published: 2026 ABSTRACT AbstractBackground and aimsHemorrhagic transformation (HT) remains a major concern after endovascular therapy (EVT) for large-vessel occlusion stroke (LVO-AIS), and reliable predictors of HT are lacking. Glial fibrillary acidic protein (GFAP), a marker of astroglial injury, may reflect tissue vulnerability to reperfusion-related hemorrhage. We aimed to assess the predictive value of post-EVT plasma GFAP for HT.MethodsProspective single-centre study including two temporally distinct cohorts of anterior-circulation-LVO-AIS patients successfully recanalized after EVT with/without IVT (mTICI≥2B): derivation cohort (cohort-1, n=78) and evaluation cohort (cohort-2, n=110). Plasma GFAP was sampled immediately after EVT and measured using Simoa® in cohort-1 and Lumipulse® in cohort-2. The primary outcome was relevant HT, defined as parenchymal hematoma on 24-hour CT. ROC analyses compared the predictive performance of GFAP with clinical and imaging variables and across cohorts.ResultsCohort-1 (50% female; mean age 72.8 years) and cohort-2 (57% female; mean age 74.5 years) had 16% and 9.1% HT, respectively. HT-patients presented approximately 6-fold higher GFAP mean concentration across cohorts. Baseline ASPECTS was the only significant predictor of HT (0.743500pg/mL identified 20% of HT with 100% specificity, while <61.4pg/mL excluded HT with 100% specificity in 23% of patients.ConclusionsIn successfully recanalized LVO-AIS, post-EVT GFAP consistently predicted relevant HT across two independent cohorts. Early and rapid clinical-grade testing may enable more individualized post-EVT monitoring, especially in neurocritical patients with limited neurological assessment.Conflict of interestAll authors: nothing to disclose