Selection Paradigms for Large Vessel Occlusion Acute Ischemic Stroke Endovascular Therapy

Background: Optimal patient selection methods for thrombectomy in large vessel occlusion stroke (LVOS) are yet to be established. We sought to evaluate the ability of different selection paradigms to predict favorable outcomes. Methods: Review of a prospectively collected database of endovascular patients with anterior circulation LVOS, adequate CT perfusion (CTP), National Institutes of Health Stroke Scale (NIHSS) ≥10 from September 2010 to March 2016. Patients were retrospectively assessed for thrombectomy eligibility by 4 mismatch criteria: Perfusion-Imaging Mismatch (PIM): between CTP-derived perfusion defect and ischemic core volumes; Clinical-Core Mismatch (CCM): between age-adjusted NIHSS and CTP core; Clinical-ASPECTS Mismatch (CAM-1): between age-adjusted NIHSS and ASPECTS; Clinical-ASPECTS Mismatch (CAM-2): between NIHSS and ASPECTS. Outcome measures were inclusion rates for each paradigm and their ability to predict good outcomes (90-day modified Rankin Scale 0-2). Results: Three hundred eighty-four patients qualified. CAM-2 and CCM had higher inclusion (89.3 and 82.3%) vs. CAM-1 (67.7%) and PIM (63.3%). Proportions of selected patients were statistically different except for PIM and CAM-1 (p = 0.19), with PIM having the highest disagreement. There were no differences in good outcome rates between PIM(+)/PIM(-) (52.2 vs. 48.5%; p = 0.51) and CAM-2(+)/CAM-2(-) (52.4 vs. 38.5%; p = 0.12). CCM(+) and CAM-1(+) had higher rates compared to nonselected counterparts (53.4 vs. 38.7%, p = 0.03; 56.6 vs. 38.6%; p = 0.002). The abilities of PIM, CCM, CAM-1, and CAM-2 to predict outcomes were similar according to the c-statistic, Akaike and Bayesian information criterion. Conclusions: For patients with NIHSS ≥10, PIM appears to disqualify more patients without improving outcomes. CCM may improve selection, combining a high inclusion rate with optimal outcome discrimination across (+) and (-) patients. Future studies are warranted.