Development and internal-external validation of the ATHE Scale: predicting acute large vessel occlusion due to underlying intracranial atherosclerosis prior to endovascular treatment

Abstract

OBJECTIVE The diagnosis of intracranial atherosclerosis (ICAS) associated with large vessel occlusion (LVO) before endovascular treatment (EVT) remains a clinical challenge. This study was aimed at developing a predictive model for ICAS-LVO in the anterior circulation preceding EVT. METHODS Patients from two national stroke centers who had undergone EVT for acute ischemic stroke in the anterior circulation were evaluated. Those from one center served as the derivation cohort, whereas patients from another center functioned as the external validation cohort. ICAS-LVO was characterized as stenosis exceeding 70% or stenosis surpassing 50% accompanied by distal blood flow disruption or recurrent occlusion evidence during the intervention. A random forest algorithm helped to identify key predictors within the derivation cohort. Utilizing these predictors, the authors formulated a logistic regression model from the derivation cohort data, and the model was then internally validated through a bootstrapping method. Subsequently, a predictive score based on this model was constructed and evaluated in both cohorts. RESULTS Among all the patients, 470 from the derivation cohort and 147 from the external validation cohort met the inclusion criteria. After random forest regression, the key predictors of ICAS-LVO included the absence of atrial fibrillation, the presence of truncal-type occlusion, the absence of a hyperdense artery sign, and a lower baseline examination National Institutes of Health Stroke Scale (NIHSS) score (ATHE Scale). Incorporating these variables into the logistic regression model yielded an area under the curve (AUC) of 0.920 (95% CI 0.894–0.947) for ICAS-LVO prediction. After bootstrapping validation, the model produced a mean AUC of 0.915. Subsequently, the ATHE score, derived from these predictors, registered an AUC of 0.916 (95% CI 0.887–0.939, p < 0.001) in the derivation cohort and 0.890 (95% CI 0.828–0.936, p < 0.001) in the external validation cohort. CONCLUSIONS The ATHE Scale, incorporating atrial fibrillation, truncal-type occlusion, hyperdense artery sign, and baseline examination NIHSS score, is an accurate, objective tool for predicting ICAS-LVO prior to EVT.