A systematic review and meta‐analysis of randomized controlled trials of endovascular thrombectomy compared with best medical treatment for acute ischemic stroke

Background Acute ischemic strokes involving occlusion of large vessels usually recanalize poorly following treatment with intravenous thrombolysis. Recent studies have shown higher recanalization and higher good outcome rates with endovascular therapy compared with best medical management alone. A systematic review and meta‐analysis investigating the benefits of all randomized controlled trials of endovascular thrombectomy where at least 25% of patients were treated with a thrombectomy device for the treatment of acute ischemic stroke compared with best medical treatment have yet to be performed. Aim To perform a systematic review and a meta‐analysis evaluating the effectiveness of endovascular thrombectomy compared with best medical care for treatment of acute ischemic stroke. Summary of review Our search identified 437 publications, from which eight studies (totaling 2423 patients) matched the inclusion criteria. Overall, endovascular thrombectomy was associated with improved functional outcomes (modified Rankin Scale 0–2) [odds ratio 1·56 (1·32–1·85), P < 0·00001]. There was a tendency toward decreased mortality [odds ratio 0·84 (0·67–1·05), P = 0·12], and symptomatic intracerebral hemorrhage was not increased [odds ratio 1·03 (0·71–1·49), P = 0·88] compared with best medical management alone. The odds ratio for a favorable functional outcome increased to 2·23 (1·77–2·81, P < 0·00001) when newer generation thrombectomy devices were used in greater than 50% of the cases in each trial. Conclusions There is clear evidence for improvement in functional independence with endovascular thrombectomy compared with standard medical care, suggesting that endovascular thrombectomy should be considered the standard effective treatment alongside thombolysis in eligible patients.


Introduction
Intravenous (IV) thrombolysis with recombinant tissue-type plasminogen activator (rt-PA) is currently the standard medical treatment for patients with acute ischemic stroke (AIS). However, recanalization rates with IV rt-PA are usually low, particularly for major vessel occlusion [basilar artery, internal carotid artery, and middle cerebral artery (1,2)]. An earlier meta-analysis (MA) with IV rt-PA has demonstrated recanalization rates as low as 14% for internal carotid artery and 55% for middle cerebral artery occlusions (3). Similarly, a systematic review reported a recanalization rate of only 10-15% for major artery occlusions and a 20-40% success rate in a cohort of AIS patients treated with IV thrombolysis (4). While recanalization with IV rt-PA is correlated with improved functional outcome, the low recanalization rates mean that a significant proportion of treated patients are not benefitting from this treatment.
The low rate of recanalization for large vessel occlusion instigated a search for improved recanalization strategies such as endovascular therapy (EVT). EVT can include rt-PA delivered directly into the artery or retracting a clot using aspiration or a stent retriever. EVT is thought to produce higher recanalization rates, improve functional outcome and reduce mortality in patients with AIS. Despite higher recanalization rates, early randomized controlled trials (RCTs) of EVT failed to show any benefits of endovascular treatment over IV thrombolysis alone (5)(6)(7). The use of older generation devices, lack of documented vessel occlusion before treatment, and time delay were some of the factors responsible for the failure of these EVT trials. However, recent clinical data from multiple RCTs that evaluated endovascular thrombectomy over best medical treatment (which may or may not include IV rt-PA) for patients with major vessel occlusion in AIS have shown overwhelming evidence in favor of mechanical thrombectomy. The first convincing evidence of the benefit of endovascular thrombectomy for the management of stroke came from the Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands (MR CLEAN) (8). This was subsequently supported by other trials: the Endovascular Treatment for Small Core and Anterior Circulation Proximal Occlusion With Emphasis on Minimizing CT to Recanalization Times (ESCAPE) (9), Extending the Time for Thrombolysis in Emergency Neurological Deficits -Intra-Arterial (EXTEND-IA) (10), Randomized Trial of Revascularization With the Solitaire FR Device Versus Best Medical Therapy in the Treatment of Acute Stroke Due to Anterior Circulation Large Vessel Occlusion (REVASCAT) (11), and Solitaire With the Intention for Thrombectomy as Primary Endovascular Treatment Trial (SWIFT PRIME) (12) studies.
The findings in these recent trials suggest a superior outcome following treatment with IV thrombolysis and thrombectomy using modern thrombectomy devices compared with best medical treatment alone. The successes from the new RCTs have been attributed mostly to improved thrombectomy devices with faster and higher rates of recanalization and better study protocols with documentation of vessel occlusion before randomization. This is in addition to improved time to revascularization.
While previous systematic reviews (SRs) and MA failed to show any superiority or benefit of EVT over IV thrombolysis for AIS (13)(14)(15), a recent SR and MA involving all the previous prospective studies, including MR CLEAN (but excluding ESCAPE, EXTEND-IA, REVASCAT, and SWIFT PRIME), demonstrated superiority of EVT over best medical treatment alone for AIS (16). The recent publication of these additional four RCTs has prompted the need for an updated SR and MA to amalgamate the latest evidence comparing EVT using mechanical thrombectomy devices over medical treatment alone. The aim of this study is to perform a comprehensive SR and MA evaluating RCTs of AIS management comparing thrombolysis to EVT wherein at least 25% of EVT cases were treated with thrombectomy devices.

Methods
This study was guided by established review protocols of the Cochrane Collaboration (17) and followed the PRISMA standard (18) to report findings.

Eligibility criteria
Studies were included if they were RCTs meeting the following criteria: • Completed studies were published in peer-reviewed journals, • Study populations were greater than 20 cases, • Patients had AIS due to major vessel occlusion and had received treatment with endovascular intervention, IV thrombolysis or best medical care, or endovascular treatment with IV thrombolysis, • Major vessel occlusion was confirmed by CT angiography or MR angiography, • In studies with endovascular thrombectomy, both old (MERCI) and new generation (Stent retrievers and Penumbra aspiration 5 MAX, ACE) devices were utilized in at least 25% of cases (pure manipulation of the clot with a guide wire, without use of a thrombectomy device, was not considered endovascular thrombectomy), • The studies reported the following outcomes: functional outcome [measured by the modified Rankin Scale (mRS)], allcause mortality, and symptomatic intracerebral hemorrhage (sICH), and • The studies reported a risk estimate [relative risk or odds ratio (OR)] or had available data for the calculation of a risk estimate.

Study selection and analysis
The titles and abstracts of studies were examined by four reviewers (J. S. B., B. A. S., G. H., A. A. N.), and then full texts were scrutinized if necessary for inclusion. Eligibility assessment and data extraction were performed independently by five reviewers (J. S. B., B. A. S., L. D. E., A. A. N., G. H.) and entered into a standard format. Researchers then met to discuss areas of agreement and completeness, and disagreements were resolved by consensus. Figure 1 shows the PRISMA flow chart of study selection. Study quality was assessed using the CASP Randomized Controlled Trial checklist (19).

Data extraction
The following information was extracted from the included primary studies as well as published supplementary materials: • background characteristics of the included studies (trial and first author, trial period, location, number of patients, number of centers, and devices used) ( • secondary characteristics [onset to groin puncture, CT to groin • proportion of EVT patients that underwent mechanical thrombectomy using a thrombectomy device (Table 4), and • outcome measure data as described below.

Outcome measures
The prespecified primary outcome measure was clinical functional independence as determined by an mRS score of 0-2 at 90 days. The prespecified secondary outcomes were sICH as defined by the trials and all-cause mortality at 90 days. Other outcomes of interest included: the number of patients who were able to walk unassisted but with increasing levels of disability from none to moderate (mRS 0-3) at 90 days; and the number of patients who had mRS 0-2 at 90 days with a baseline ASPECTS 8-10 (minimal evidence of underlying ischemic change), ASPECTS 5-7 (moderate evidence of underlying ischemic change), or a baseline ASPECTS 0-4 (substantial evidence of underlying ischemic change).

Data analyses
All MAs were conducted using Review Manager version 5.3 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark). MAs were performed for the prespecified primary and secondary outcomes with the results stratified by the administered treatment type. The ORs and 95% confidence intervals were calculated from the extracted data using the Mantel-Haenszel fixed-effects model. The heterogeneity between studies was tested using the inconsistency index (I 2 ): values less than 25% were considered low; between 25 and 70% were considered moderate; and greater than 70% were considered highly heterogeneous (20).
A DerSimonian and Laird's random effect model was only used where there was a high level of heterogeneity between studies. A P value of <0·05 was considered statistically significant.

Study characteristics
The search yielded 437 studies, eight of which were retained for MA (Fig. 1). The eight RCTs comparing endovascular therapies with IV thrombolysis meeting the eligibility criteria included ESCAPE (9), EXTEND-IA (10), IMS III (5), MR CLEAN (8), MR RESCUE (7), REVASCAT (11), SWIFT PRIME (12), and SYN-THESIS (6). THERAPY (21) and THRACE (22), trials still awaiting publication, were not incorporated in this analysis. 1 Seven of the studies were multicenter trials based in several countries with REVASCAT the only trial to be conducted in one country (Spain). Patient populations ranged from n = 70 to n = 656, with a total of 2423 patients; 1313 patients were randomized to EVT, with mean age ranging from 65 to 71 years, and 1110 patients to IV thrombolysis with mean age ranging from 66 to 70 years. The median baseline NIHSS ranged from 13 to 21 for the IV thrombolysis group and 13-19 for the EVT group. Different thrombectomy devices were used in the RCTs, with newer generation devices being used in more recent trials ( Table 1). The characteristics of the selected studies are summarized in Tables 1-4.

MA of outcome measures
Two sets of MAs were conducted. The first analysis was performed on the eight identified studies ( Table 1). The second analysis was conducted on the six published studies that had >50% of patients in the EVT group treated by mechanical thrombectomy with MERCI and new generation devices (ESCAPE, EXTEND, MR CLEAN, MR RESCUE, REVASCAT, and SWIFT PRIME) ( Table 4).

Secondary outcome
No significant difference was found in the prespecified secondary outcomes mortality and sICH when EVT was compared with IV 1 The authors are aware of THERAPY and THRACE, randomized controlled trials on EVT for AIS that are undergoing final analysis before publication. The trials' results were presented at the European Stroke Organisation Conference (Glasgow, United Kingdom, April 17, 2015), and while the trials meet the eligibility criteria for MA, the results were not included here.  rt-PA. EVT had a tendency toward decreased mortality at 90 days [OR 0·84 (0·67-1·05), P = 0·12] compared with best medical management alone (Fig. 3a). EVT did not increase the incidence of sICH [OR 1·03 (0·71-1·49), P = 0·88; Fig. 3b]. Unlike the primary outcome, no heterogeneity was noted between studies for the secondary outcomes (I 2 = 0% for both mortality and sICH).

Subgroup analysis with ASPECTS
This subgroup analysis divided each treatment arm into three subgroups: baseline ASPECTS 8-10 (minimal evidence of underlying ischemic change), baseline ASPECTS 5-7 (moderate evidence of underlying ischemic change), and baseline ASPECTS 0-4 (substantial evidence of underlying ischemic change) (23). Four studies were included in this analysis as these were the only studies that presented stratified ASPECTS data for mRS [ESCAPE (9), MR CLEAN (8), REVASCAT (11), and SWIFT PRIME (12)]. IMS III (5) presented stratified data for ASPECTS 8-10 and 0-7, but due to the difference in stratification and having <50% of EVT patients undergoing thrombectomy, this study was excluded from this subgroup MA. EVT improved functional independence compared with best medical treatment in patients with high baseline ASPECTS [OR 2·10 (1·61-2·73), P < 0·00001; Fig. 6]. EVT also improved functional independence with moderate baseline ASPECTS [OR 2·04 (1·25-3·32), P = 0·004]. There was no evidence of benefit of EVT in patients with low baseline ASPECTS [OR 1·09 (0·14-8·46), P = 0·93], but only 28 patients were included in this analysis due to MR CLEAN being the only study that incorporated this group of patients in their trial. Overall, these results suggest that patients with baseline ASPECTS >4 benefit from EVT.

Discussion
These SR and MA combine the results from all published RCTs comparing EVT (where at least 25% of patients were treated with a thrombectomy device) to best medical management alone for the treatment of AIS. Eight trials (2423 patients) of EVT compared with best medical treatment alone were identified and included in this analysis. In the combined MA of the eight trials, EVT demonstrated improved clinical functional outcome (number of patients with mRS 0-2 at 90 days poststroke), but there was high heterogeneity between the studies. The heterogeneity could be accounted for by different study designs, times to treatment and differing imaging methodologies, as well as the use of newer generation thrombectomy devices with higher recanalization rates in the more recent trials. There was a tendency toward decreased mortality with EVT compared with best medical treatment alone, indicating the benefit of this therapeutic strategy. In addition, sICH incidence was not increased with EVT suggesting that the use of thrombectomy devices alongside IV rt-PA does not increase the risk of hemorrhage compared with IV rt-PA alone.
The eight primary trials included in the MA contained EVT groups where patients underwent either intra-arterial rt-PA therapy and/or mechanical manipulation of the clot (often not  involving a thrombectomy device but wire manipulation or ultrasound without clot extraction). While earlier trials [SYNTHESIS (6) and IMS III (5)] predominantly used intra-arterial rt-PA for EVT, the remaining six trials used a thrombectomy device in >50% of cases in the EVT group (Table 4). Therefore, we analyzed these trials with more frequent use of mechanical thrombectomy devices in a separate MA. The results showed that in these trials, the chances of a better outcome were superior to the primary MA (where only at least 25% of EVT patients needed to be treated with thrombectomy). There was also greater homogeneity between studies, which could be attributed to similar study design between these RCTs. Similar to the primary analysis, there was no difference in mortality or sICH rate between EVT and IV thrombolysis in the trials with substantial use of endovascular thrombectomy. Five trials [ESCAPE (9), IMS III (5), MR CLEAN (8), REVAS-CAT (11), and SWIFT PRIME (12)] reported baseline ASPECTS, a score that divides the brain into 10 regions and assesses each brain region for ischemic changes based on a CT image (23). A higher ASPECTS indicates minimal ischemic change, while a lower ASPECTS signifies substantial ischemic change. Subgroup analysis was only possible using data from ESCAPE (9), MR CLEAN (8), REVASCAT (11), and SWIFT PRIME (12) as these were the only studies that reported mRS scores stratified by high ASPECTS (8-10), moderate ASPECTS (5-7), or low ASPECTS (0-4). In fact, ESCAPE, REVASCAT, and SWIFT PRIME had baseline ASPECTS <6 as an exclusion criterion for their trials, and so the majority of ASPECTS 5-7 patients included in the MA would have been patients with scores of 6 or 7. The ASPECTS sub-group MA showed that with both high and moderate baseline ASPECTS, the odds of mRS 0-2 at 90 days improved with EVT compared to IV rt-PA, while this was not the case in patients with low (<5) baseline ASPECTS. The low baseline ASPECTS analysis only included 28 patients from MR CLEAN, and further analysis will be undertaken once more data have been published that includes patients stratified by ASPECTS. Until these data are available, we cannot confirm whether EVT is beneficial or not in this patient group.
The key aspect of EVT is the restoration of cerebral blood flow. In clinical trials, this is usually measured through the thrombolysis in cerebral infarction (TICI) scale, which assesses recanalization by digital subtraction angiography. In the majority of the eight trials identified, the recanalization score was reported, but unfortunately, MA was not possible due to variations in the comparisons made (EVT vs. IV thrombolysis or EVT vs. baseline) and the timing of measurements. These variations in the reporting of data from the TICI scale are evident throughout a number of other revascularization studies, making the interpretation and comparison of these scores across studies difficult (24). Time-torecanalization could also be a critical contributor to outcome, where it was noted that MR RESCUE had the longest time-torecanalization (>5 h) and did not show benefit, whereas the more recent trials with positive effects had a time-to-recanalization within three-to five-hours. In addition to assessment of recanalization, some studies also examined the extent to which the brain reperfused following recanalization. In EXTEND-IA (10) and SWIFT PRIME (12), EVT showed that reperfusion at 24 h poststroke (assessed by CT and MR perfusion imaging) was substantially improved and associated with better functional outcome compared with IV rt-PA alone.
The secondary outcomes analyzed as part of the MA showed that EVT was safe (EVT does not increase sICH beyond that which is already produced by rt-PA, and does not increase mortality) while producing a clear beneficial effect on clinical outcome (mRS score). The lack of adverse effects of endovascular thrombectomy highlights the applicability of this procedure as a therapeutic strategy for AIS.
There are a number of limitations of this MA. Like any MA, data were pooled from trials with differences in design and methodology, particularly with regard to patient selection, time to treatment, and the use of different mechanical devices which could all be sources of bias. Patients in the endovascular arm of the studies evaluated mainly received IV thrombolysis followed by mechanical clot extraction with or without the additional use of thrombolytics. There were also patients that did not undergo IV thrombolysis and received direct mechanical thrombectomy [e.g. 10% of patients in the MR CLEAN study (8)]. As a result, the MA population was not perfectly homogeneous which was only evident when assessing all eight trials for the primary outcome. In addition, our inclusion criteria required that at least 25% of patients in the EVT arm were treated by mechanical thrombectomy with a device. Most studies met this inclusion criterion, but IMS III (5) and SYNTHESIS (6) had a much greater proportion of patients treated endovascularly by intra-arterial rt-PA or wire manipulation alone than mechanical thrombectomy. IMS III had several issues regarding the trial design and execution with lack of evidence of vessel occlusion, the use of older devices, and resultant lower recanalization rates. Despite the limitations of IMS III, it is repeatedly quoted as a comparator and also used in MA. As the time course of restoration of cerebral blood flow is disparate between rt-PA (gradual) and thrombectomy (instant), these differences in endovascular treatment could have an effect on outcome. Also, with wire manipulation alone, vessel re-opening rates are low, and no clot is extracted. Hence, our sub-group analysis of trials that used mechanical thrombectomy in greater than 50% of patients in the EVT group showed a stronger positive effect on clinical outcome, with greater homogeneity between studies. A patient-level MA would provide greater scope for further analysis, particularly when assessing contributing factors to outcome, but with these included studies, this was not possible.   Despite these limitations, this MA was able to evaluate the positive impact of mechanical thrombectomy on clinical outcome in patients with AIS secondary to major vessel occlusion. Given the rigor associated with angiographic and clinical assessments in all eight trials, this allowed data pooling for a number of primary and secondary outcomes and some sub-group analyses where consistencies between studies were reached.

Conclusions
Our study has shown overwhelming evidence of improved functional outcome in favor of EVT over medical treatment alone. EVT did not increase the incidence of sICH and showed marginal evidence for reduced mortality at 90 days, confirming the safety of the procedure. The benefit of EVT on functional independence was even more marked in the subgroup analysis of RCT trials where >50% of EVT patients underwent mechanical thrombectomy. Overall, the substantial evidence of improvement in functional outcome without an increased incidence of adverse effects means that EVT should now be considered as a primary treatment option for eligible AIS patients, alongside IV thrombolysis.

Contributors
J. S. B. searched the literature, extracted the data, reviewed available studies, guided the review process, and wrote the paper. B. A. S. reviewed available studies, extracted the data, analyzed the results, and co-wrote the paper. L. D. E. guided the literature search and data extraction process, searched the literature, extracted the data, contributed to and reviewed the paper. I. Q. G.

Supporting information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Figure S1. Funnel plot of included studies for the primary outcome (mRS 0-2) showing symmetry of studies suggestive of lack of publication bias.