Thrombectomy for anterior circulation stroke beyond 6 h from time last known well (AURORA): a systematic review and individual patient data meta-analysis

Summary

Background

Trials examining the benefit of thrombectomy in anterior circulation proximal large vessel occlusion stroke have enrolled patients considered to have salvageable brain tissue, who were randomly assigned beyond 6 h and (depending on study protocol) up to 24 h from time last seen well. We aimed to estimate the benefit of thrombectomy overall and in prespecified subgroups through individual patient data meta-analysis.

Methods

We did a systematic review and individual patient data meta-analysis between Jan 1, 2010, and March 1, 2021, of randomised controlled trials of endovascular stroke therapy. In the Analysis Of Pooled Data From Randomized Studies Of Thrombectomy More Than 6 Hours After Last Known Well (AURORA) collaboration, the primary outcome was disability on the modified Rankin Scale (mRS) at 90 days, analysed by ordinal logistic regression. Key safety outcomes were symptomatic intracerebral haemorrhage and mortality within 90 days.

Findings

Patient level data from 505 individuals (n=266 intervention, n=239 control; mean age 68·6 years [SD 13·7], 259 [51·3%] women) were included from six trials that met inclusion criteria of 17 screened published randomised trials. Primary outcome analysis showed a benefit of thrombectomy with an unadjusted common odds ratio (OR) of 2·42 (95% CI 1·76–3·33; p<0·0001) and an adjusted common OR (for age, gender, baseline stroke severity, extent of infarction on baseline head CT, and time from onset to random assignment) of 2·54 (1·83–3·54; p<0·0001). Thrombectomy was associated with higher rates of independence in activities of daily living (mRS 0–2) than best medical therapy alone (122 [45·9%] of 266 vs 46 [19·3%] of 238; p<0·0001). No significant difference between intervention and control groups was found when analysing either 90-day mortality (44 [16·5%] of 266 vs 46 [19·3%] of 238) or symptomatic intracerebral haemorrhage (14 [5·3%] of 266 vs eight [3·3%] of 239). No heterogeneity of treatment effect was noted across subgroups defined by age, gender, baseline stroke severity, vessel occlusion site, baseline Alberta Stroke Program Early CT Score, and mode of presentation; treatment effect was stronger in patients randomly assigned within 12–24 h (common OR 5·86 [95% CI 3·14–10·94]) than those randomly assigned within 6–12 h (1·76 [1·18–2·62]; p_interaction=0·0087).

Interpretation

These findings strengthen the evidence for benefit of endovascular thrombectomy in patients with evidence of reversible cerebral ischaemia across the 6–24 h time window and are relevant to clinical practice. Our findings suggest that in these patients, thrombectomy should not be withheld on the basis of mode of presentation or of the point in time of presentation within the 6–24 h time window.

Funding

Stryker Neurovascular.

Introduction

The first generation of prospective randomised trials that showed clinical benefit of mechanical thrombectomy in acute stroke due to anterior circulation large vessel occlusion (LVO) enrolled most patients within 6 h of time last seen well and selected patients primarily based on imaging studies aiming to exclude haemorrhage, confirm the presence of LVO, and exclude a large baseline infarct (core).1, 2, 3, 4, 5, 6, 7 In these trials, CT perfusion or MRI, considered more accurate imaging methods than non-contrast CT for assessment of salvageable brain tissue were not required for patient enrolment. Furthermore, benefit of thrombectomy was shown to be time dependent as the lower boundary of the 95% CI for demonstrable benefit was found to cross the line representing the common odds ratio (OR) of 1 at 7 h from time last seen well to procedure initiation.⁸ Two subsequent trials (DAWN⁹ and DEFUSE 3¹⁰) have shown that in selected patients with acute stroke due to anterior circulation LVO randomly assigned within 6–24 h and 6–16 h of time last seen well respectively, the benefit of thrombectomy is no less than that shown in the early time window. DAWN and DEFUSE 3 only included patients with evidence of salvageable brain tissue identified through the presence of substantial mismatch, a pathophysiological term characterising a condition in which the volume of brain tissue that is viable but imminently threatened to undergo infarction due to ongoing vessel occlusion far exceeds the volume of brain tissue that is already infarcted. Mismatch was assessed through computer generated volumetric imaging criteria, alone (DEFUSE 3) or in conjunction with clinical criteria (DAWN). Four other thrombectomy trials4, 5, 11, 12 allowed randomisation between 0–8 h or 0–12 h from time last seen well depending on each individual study protocol, and enrolled patients randomly assigned beyond 6 h from time last seen well (table 1). These trials used various imaging selection criteria (table 2) to confirm the presence of salvageable brain but did not require volumetric mismatch measurement as a prerequisite for patient enrolment. In total, six thrombectomy trials have included patients randomly assigned beyond 6 h of time last seen well with late enrolment limit ranging from 8–24 h.4, 5, 9, 10, 11, 12

Research in context

Evidence before this study

Time is an essential determinant of patient eligibility for thrombectomy in acute stroke. In 2015, several randomised trials established that thrombectomy—when performed predominantly within 6 h of time last seen well with newer generation devices and more efficient workflow— significantly reduces disability rates after acute stroke due to proximal large vessel occlusion in the anterior circulation. Some of these studies allowed inclusion of patients who could be randomly assigned beyond 6 h from time last seen well but few such patients were enrolled. As such, data supporting thrombectomy in patients treated beyond 6 h from time last seen well were scarce until the publication in 2018 of two trials that enrolled only patients who could be randomly assigned between 6 h and up to 24 h from time last seen well. These studies required proof of substantial mismatch between extent of infarcted brain tissue and extent of at-risk brain tissue on the basis of advanced imaging studies (CT perfusion or MRI) for patient inclusion. They were stopped for efficacy based on prespecified criteria after enrolment of a relatively small number of patients. As such, all individual trials were underpowered to provide convincing evidence of efficacy across some of the subgroups of great relevance to clinical practice. Between Jan 1, 2018, and March 1, 2021, on several occasions we searched PubMed, Embase, MEDLINE, and ClinicalTrials.gov databases without language restrictions for any patient-level systematic reviews and meta-analyses reporting on data from randomised controlled trials testing the efficacy of endovascular treatment using modern thrombectomy devices in patients presenting beyond 6 h from time last seen well. The search focused on articles published between Jan 1, 2000, to March 1, 2021, using the search terms “stroke”, “endovascular”, “large vessel occlusion” and “6 hours”. Although meta-analyses of observational studies and study level meta-analyses of randomised controlled trials have been reported, we did not identify any patient-level meta-analyses of such data in our search.

Added value of this study

Our study is an individual patient meta-analysis of all individuals enrolled in published randomised endovascular stroke trials in which randomisation occurred beyond 6 h of time last seen well and thrombectomy for large vessel anterior circulation occlusion was conducted according to current clinical practice. Treatment was with modern generation thrombectomy devices leading to highly effective reperfusion rates, proof of vessel occlusion before enrolment, exclusion of patients with large infarcts at baseline, and evidence of reversible cerebral ischaemia especially when ascertained by advanced imaging (MRI or CT perfusion). These data will enable a better understanding of the degree of precision of effect size estimates and safety outcome estimates in the overall population and of effect size estimates in some clinically relevant subgroups. We have shown that the clinical benefit derived from thrombectomy extends across a wide range of ages and initial stroke severity or extent of baseline infarct. Furthermore, our analysis shows that thrombectomy is beneficial regardless of presentation mode (stroke symptoms discovered upon awakening, witnessed onset, or unwitnessed onset) and that the treatment effect is stronger in the later (12–24 h) time window than the earlier (12–24 h) time window.

Implications of all the available evidence

Our study has several implications for health policy considerations and for clinical practice, which might lead to changes in current guidelines. First, by showing consistent results across different patient populations with respect to geography, our study suggests that benefit from thrombectomy in patients presenting within 6–24 h from time last seen well is generalisable to a broad range of patients with large vessel ischaemic stroke with imaging evidence of reversible cerebral ischaemia. Second, by providing a more precise treatment effect estimate than each individual trial, the cost effectiveness of this intervention at a societal level can be assessed with higher precision. Third, our study provides evidence that, in clinical practice, endovascular therapy for stroke in patients with proximal large vessel anterior circulation occlusion should not be withheld on the basis of advanced age, moderate size infarcts on baseline CT, moderate or severe clinical deficit, mode of presentation, or the point in time of presentation within the 6–24 h time window.

Although time represents an important determinant of patient eligibility for reperfusion therapies, frequently the time of symptom onset cannot be confirmed with precision because in many patients (including those who wake up with neurological deficits), the point in time at which neurological deficits first occur is unknown. Although the time of symptom discovery is typically when medical attention is sought, for the purposes of establishing eligibility for thrombectomy, the universally accepted reference point is the time last seen well. There are three distinct modes by which patient stroke symptoms first come to attention: wake-up stroke, unwitnessed onset stroke, and witnessed onset stroke (appendix p 7).

The distinction between the 0–6 h and 6–24 h time periods is important in clinical practice; current guidelines15, 16 recommend demonstration of mismatch as defined in DAWN⁹ or DEFUSE 3¹⁰ as a prerequisite for selection for endovascular therapy in patients who are late-presenting but not for those who are early-presenting.

Uncertainties remain for the benefits of endovascular therapy in patient subpopulations that were under-represented in individual trials, including those who present in the latest time period, who are not selected on the basis of volumetric mismatch assessment, and those whose symptoms are witnessed to have occurred for longer than 6 h. Moreover, since these trials were either individually moderate in size or their contributing dataset entailed a small number of patients, data pooling can provide more precise treatment effect estimation. Herein, we seek to address questions about the risks and benefits of endovascular therapy with second-generation thrombectomy devices in patients with LVO stroke presenting beyond 6 h from time last seen well by performing an individual patient data pooled meta-analysis of randomised controlled trials of thrombectomy for acute ischaemic stroke involving this patient population.