Abstract Background and aims Predicting clinical outcomes before endovascular treatment (EVT) remains challenging. Several prognostic tools estimate the likelihood of a good (modified Rankin Scale mRS 0–2) or poor outcome (mRS 3–6) at 3 months. This study validated the MR PREDICTS and Totaled Health Risks in Vascular Events–EVT (THRIVE-EVT) models for good outcomes, and the Risk strAtification for eNdovascular treatment in acute anterior circulation occlusive stroKe (RANK), Stroke Prognostication using Age and NIH Stroke Scale (SPAN), and Houston Intra-Arterial Therapy 2 (HIAT-2) scores for poor outcomes. Methods From the Safe Implementation of Thrombolysis in Stroke (SITS) registry, we retrospectively included patients treated with EVT for anterior circulation large-vessel occlusion within 6 hours of last-seen-well time. Receiver Operating Characteristic curves determined optimal cut-offs and diagnostic accuracy. Results Among 296 patients, 145 (49%) had a good outcome and 151 (51%) a poor outcome. AUCs for MR PREDICTS and THRIVE-EVT were 0.67 and 0.68, with no significant difference. Optimal cut-offs of treatment benefit of 9.7% (MR PREDICTS) and 16% (THRIVE-EVT) yielded 22 (7.4%) and 49 (16.5%) false negatives (FN), respectively. For poor outcome prediction, AUCs were 0.71 (RANK), 0.76 (SPAN),and 0.74 (HIAT-2), with a significant difference between SPAN and RANK (ΔAUC = 0.056, p = 0.026). FN rates were 21.88% for RANK, 15.8% for SPAN and 14.8% for HIAT-2. Conclusions Accuracy for good outcome prediction was lower than in larger cohorts, and the FN question clinical usefulness. In contrast, FN rates for poor outcome prediction were lower than previously reported. Conflict of interest The authors declare no conflicts of interest.
Merschaert et al. (2026) studied this question.