Predicting Recanalization Failure With Conventional Devices During Endovascular Treatment Related to Vessel Occlusion.

Background: Among patients with stroke eligible for endovascular treatment, preprocedure identification of those with low chances of successful recanalization with conventional devices (stent-retrievers and/or direct aspiration) may allow anticipating procedural rescue strategies. We aimed to develop a preprocedural algorithm able to predict recanalization failure with conventional devices (RFCD). Methods: Observational study. Data from consecutive patients with stroke who received endovascular treatment between 2019 and 2022 in 10 centers were collected from the Catalan Stroke Registry (Codi Ictus Catalunya Registry, CICAT). RFCD was defined as final thrombolysis in cerebral infarction ≤2a or the use of rescue therapy defined as balloon angioplasty±stent deployment. Univariate and multivariate analysis to identify variables associated with RFCD were performed. A gradient boosted decision tree machine learning model to predict RFCD was developed utilizing preprocedure variables previously selected. Clinical improvement at 24 hours was defined as a drop of ≥4 points from baseline National Institutes of Health Stroke Scale score or 0-1 at 24 hours. Results: In total, 984 patients were included; RFCD was observed in 14.3% (n:141) of the cases. Of these, 47.5% (n = 67) received balloon angioplasty±stent deployment as rescue therapy. Among patients receiving balloon angioplasty±stent deployment, clinical improvement was associated with lower number of attempts with conventional devices (median number of passes 2 versus 3; Conclusion: In RFCD patients, early balloon angioplasty±stent deployment rescue was associated with improved outcomes. A predictive model using affordable preprocedure clinical variables could be useful to identify these patients before intervention.