ABSTRACT NUMBER: ESOC2026A116 PREDICTION OF CLINICAL OUTCOME AFTER ENDOVASCULAR TREATMENT: COMPARISON BETWEEN POST-PROCEDURAL SCORES

Abstract Background and aims Less than half of patients treated with endovascular treatment (EVT) for acute ischemic stroke (AIS) achieve a good outcome (modified Rankin [mRS] Scale 0-2) at three months. This study aimed to compare the performance of several prediction tools for good outcomes at three months after EVT. Methods From our center Safe Implementation of Thrombolysis in Stroke (SITS) registry, we retrospectively identified patients treated with EVT for anterior circulation LVO. We evaluated the diagnostic accuracy for mRS 0–2 at three months of NIHSS at 24 hours, MRPREDICTS@24H, Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials (HERMES)-24 score, Bronx Endovascular Thrombectomy (BET), and SNARL (Symptomatic haemorrhage, baseline NIHSS, Age, Reperfusion, Location of clot) scores. ROC curves were generated to determine optimal cut-off values and to quantify the predictive performance. Calibration plots, Brier scores, and Hosmer-Lemeshow tests were applied to evaluate calibration and overall performance. Results At three months, 204 patients (43%) achieved good outcome, while 274 (57%) had poor outcome (mRS 3-6). In descending order, the best-performing tools for predicting good outcome were MRPREDICTS@24H, HERMES-24, NIHSS at 24 hours, BET, and SNARL scores (AUCs 0.91, 0.88, 0.87, 0.79, and 0.75, respectively). Conclusions MRPREDICTS@24H, HERMES-24, and NIHSS are robust models for predicting good outcomes after EVT, showing reliable discrimination and calibration. BET and SNARL scores remain useful but demonstrate lower accuracy and should be interpreted cautiously. Conflict of interest A. El Fahdi : nothing to disclose, A. Peeters : nothing to disclose, A. Sellimi : nothing to disclose, JM. Raymackers : nothing to disclose, S. Jacobs : nothing to disclose, B. Chemin : nothing to disclose, F. Hammer : nothing to disclose, A. Loubet : nothing to disclose, P. Scoppetuolo : nothing to disclose.