Optimizing Outcome Prediction Scores in Patients Undergoing Endovascular Thrombectomy for Large Vessel Occlusions Using Collateral Grade on Computed Tomography Angiography

Abstract

BACKGROUND: Although several outcome prediction scores incorporated with pretreatment variables have been developed for acute ischemic stroke (AIS) patients, there is not currently a prediction score that includes pretreatment imaging that can show salvageable brain tissue.

OBJECTIVE: To evaluate whether addition of the collateral grade on computed tomography angiography to previously published prediction scores could increase accuracy of clinical outcome prediction in endovascular thrombectomy (EVT) for AIS.

METHODS: This study used a retrospective multicenter registry for patients undergoing EVT for anterior circulation large vessel occlusion. Three previously published outcome prediction scores (Houston intra-arterial therapy 2, HIAT2; totaled health risks in vascular events, THRIVE; and Pittsburgh response to endovascular therapy, PRE scores) were tested in this study. Using 482 deprivation cohorts, areas under the receiver operating characteristic curves (AUC-ROCs) were compared between prediction scores with/without collateral grades in predicting the poor outcomes (modified Rankin Scale 4-6 at 3-mo follow-up) after EVT. We developed modified prediction scores by adding the collateral grade, and their advancement of outcome prediction was validated using 208 independent validation cohorts.

RESULTS: AUC-ROCs of HIAT2, THRIVE, and PRE scores that incorporated with collateral grade were superior in predicting poor outcomes when compared to that of the unmodified scores (P < 0.001). In modified prediction models, 3, 3, and 10 points were added for poor collateral grade to HIAT2, THRIVE, and PRE score. Modified models outperformed unmodified models in testing of the validation cohorts (P < 0.001).

CONCLUSION: The addition of the collateral grade to outcome prediction scores resulted in better prediction of poor outcome after EVT for AIS compared to the prediction scores alone.