Abstract
Background and Purpose—Whether all acute stroke patients with large vessel occlusion need to undergo intravenous thrombolysis before mechanical thrombectomy (MT) is debated as (1) the incidence of post-thrombolysis early recanalization (ER) is still unclear; (2) thrombolysis may be harmful in patients unlikely to recanalize; and, conversely, (3) transfer for MT may be unnecessary in patients highly likely to recanalize. Here, we determined the incidence and predictors of post-thrombolysis ER in patients referred for MT and derive ER prediction scores for trial design. Methods—Registries from 4 MT-capable centers gathering patients referred for MT and thrombolyzed either on site (mothership) or in a non MT-capable center (drip-and-ship) after magnetic resonance– or computed tomography–based imaging between 2015 and 2017. ER was identified on either first angiographic run or noninvasive imaging. In the magnetic resonance imaging subsample, thrombus length was determined on T2*-based susceptibility vessel sign. Independent predictors of no- ER were identified using multivariable logistic regression models, and scores were developed according to the magnitude of regression coefficients. Similar registries from 4 additional MT-capable centers were used as validation cohort. Results—In the derivation cohort (N=633), ER incidence was ≈20%. In patients with susceptibility vessel sign (n=498), no-ER was independently predicted by long thrombus, proximal occlusion, and mothership paradigm. A 6-point score derived from these variables showed strong discriminative power for no-ER (C statistic, 0.854) and was replicated in the validation cohort (n=353; C statistic, 0.888). A second score derived from the whole sample (including negative T2* or computed tomography–based imaging) also showed good discriminative power and was similarly validated. Highest grades on both scores predicted no-ER with >90% specificity, whereas low grades did not reliably predict ER. Conclusions—The substantial ER rate underlines the benefits derived from thrombolysis in bridging populations. Both prediction scores afforded high specificity for no-ER, but not for ER, which has implications for trial design.