Objectives The present Bayesian network meta-analysis aimed to compare the various strategies for acute ischemic stroke: direct endovascular thrombectomy within the thrombolysis window in patients with no contraindications to thrombolysis (DEVT); (2) direct endovascular thrombectomy secondary to contraindications to thrombolysis (DEVTc); (3) endovascular thrombectomy in addition to thrombolysis (IVEVT); and (4) thrombolysis without thrombectomy (IVT).
Methods Six electronic databases were searched from their dates of inception to May 2017 to identify randomized controlled trials (RCTs) comparing IVT versus IVEVT, and prospective registry studies comparing IVEVT versus DEVT or IVEVT versus DEVTc. Network meta-analyses were performed using ORs and 95% CIs as the summary statistic.
Results We identified 12 studies (5 RCTs, 7 prospective cohort) with a total of 3161 patients for analysis. There was no significant difference in good functional outcome at 90 days (modified Rankin Scale score ≤2) between DEVT and IVEVT. There was no significant difference in mortality between all treatment groups. DEVT was associated with a 49% reduction in intracranial hemorrhage (ICH) compared with IVEVT (OR 0.51; 95% CI 0.33 to 0.79), due to reduction in rates of asymptomatic ICH (OR 0.47; 95% CI 0.29 to 0.76). Patients treated with DEVT had higher rates of reperfusion compared with IVEVT (OR 1.73; 95% CI 1.04 to 2.94).
Conclusions To our knowledge, this is the first network meta-analysis to be performed in the era of contemporary mechanical thrombectomy comparing DEVT and DEVTc. Our analysis suggests the addition of thrombolysis prior to thrombectomy for large vessel occlusions may not be associated with improved outcomes.
Statistics from Altmetric.com
Recent randomized controlled trials (RCTs) demonstrate that endovascular thrombectomy in addition to intravascular thrombolysis (IVEVT) is safe and markedly improves functional outcomes in patients with acute ischemic stroke in the anterior circulation.1–4 Since these landmark trials, there have been several non-randomized studies suggesting that intravascular thrombolysis (IVT) does not confer any additional benefit over direct endovascular thrombectomy.5–7 Investigators have raised questions about the role of IVT in this modern era of rapid, highly successful recanalization with mechanical thrombectomy.7 8 A recent meta-analysis contradicted these publications, suggesting improved functional outcomes and reduced death/severe dependency in patients receiving IVEVT compared with patients receiving only direct endovascular thrombectomy.9–11 However, the direct endovascular thrombectomy cohort in this meta-analysis included studies with patients contraindicated to receive IVT rather than patients receiving direct endovascular thrombectomy based on choice and with no contraindications. Differences in baseline may have confounded the outcomes seen in the analysis. This study is further limited by small sample sizes, wide confidence intervals, and lack of direct comparative analysis of baseline characteristics and outcomes of the direct endovascular thrombectomy within the IVT timeframe and no contraindications to IVT (DEVT) cohort versus other endovascular strategies.
In light of the disparate data and resultant uncertainty, we conducted a network meta-analysis comparing the various endovascular therapy strategies for acute ischemic stroke with large vessel occlusions. Patients undergoing direct endovascular thrombectomy were assessed in separate cohorts, the first was patients who underwent direct endovascular thrombectomy within the IVT timeframe and had no contraindications to IVT (DEVT cohort); the second cohort included patients undergoing direct endovascular thrombectomy secondary to contraindications to IVT (DEVTc).
Informed consent was not required for this study as no patients were involved.
Search strategy and selection criteria
The methodology and guidelines of the systematic review followed PRISMA guidelines.12 The search strategy is summarized in the online supplemental methods. To facilitate higher quality evidence, we used strict inclusion and exclusion criteria for each comparison. For studies comparing IVT versus IVEVT, we restricted included studies to RCTs where specifically all patients in the medical arm received IVT, and excluded studies in which only a fraction of patients received IVT. For studies comparing IVEVT with DEVT or DEVTc, we selected only those that analyzed prospectively collected institutional data or registry data. When propensity score matched data were available, this was used in the analysis. Studies that included patients who did not receive thrombolysis in the IVT arm were excluded. Studies involving exclusively first generation EVT devices were excluded. When institutions published duplicate studies with accumulating numbers of patients or increased lengths of follow-up, only the most complete reports were included for quantitative assessment at each time interval. The DEVT registries were chosen such that DEVT was done by operator choice, and not by contraindication to IVT.
All data were extracted from article texts, tables, and figures. Two investigators independently reviewed each retrieved article. Where required, an attempt was made to contact the authors for the provision of additional data. Discrepancies between the two reviewers were resolved by discussion and consensus. Endpoints for extraction are stated in the online supplemental methods.
Dichotomous outcome variables were compared with ORs and 95% CIs for the network meta-analysis. Hierarchical Bayesian statistics was used for the network meta-analysis due to greater flexibility and ability to rank treatments according to their comparative effectiveness. It allows for combining of all available comparisons between treatments with the advantage of greater power and precision for rare events. Analyses were performed using Bayesian Markov chain Monte Carlo modeling. To provide a comparative hierarchy of procedural efficacy and safety, ‘Rankograms’ with surface under the cumulative ranking curve (SUCRA) probabilities were reported. All analyses were performed with NetMetaXL 1.6.1 (Canadian Agency for Drugs and Technologies in Health, Ottawa, Canada) and WinBUGS 1.4.3 (MRC Biostatistics Unit, Cambridge, UK). Further details are provided in the online supplemental methods.
From six electronic databases, 2151 studies were identified, of which 36 were examined for full text evaluation (figure 1). After the inclusion and exclusion criteria were applied, 12 studies were included in our analysis. One study was excluded because most of the endovascular thrombectomy arm included intra-arterial thrombolysis.13 Three studies were excluded for inclusion of patients who did not receive thrombolysis in the IVT arm.1 2 14 The study characteristics are summarized in the online supplementary table 1 and quality appraisal in the online supplementary table 2. There were five RCTs comparing IVEVT and IVT,3 15–18 two prospective cohort studies with propensity score matching comparing IVEVT and DEVT,19 20 one study analysis of a prospective database comparing IVEVT, DEVT, and DEVTc,21 and four studies using prospective registries, databases, or trials comparing IVEVT and DEVTc.5–7 22 Risk of bias assessments for each of the RCTs are presented in the online supplementary table 3. In total, there were 3161 patients, of whom 519 were treated with IVT, 1516 with IVEVT, 248 with DEVT, and 878 with DEVTc. The complete evidence network for all outcomes is shown in figure 2.
The mean ages of the patients undergoing IVT, IVEVT, DEVT, and DEVTc were 68±0.8 years, 67.6±0.6 years, 71.3±3.0 years, and 68.7±0.6 years, respectively. Patients undergoing IVT were slightly older than patients undergoing IVEVT (68.0 years vs 67.6 years; weighted mean difference (WMD) 1.93; 95% CI 1.02 to 2.83). There were also slightly more women undergoing IVEVT compared with IVT (44.6% vs 45.2%; OR 1.35; 95% CI 1.02 to 1.77) (see online supplementary table 4).
Fewer patients undergoing IVEVT had hyperlipidemia compared with IVT (41.4% vs 55.4%; OR 1.58; 95% CI 1.10 to 2.27). Compared with IVEVT, a greater proportion of patients undergoing DEVTc had atrial fibrillation (42.3% vs 32.7%; OR 0.45; 95% CI 0.37 to 0.55), a history of prior strokes/transient ischemic attacks (19.4% vs 10.1%; OR 0.51; 95% CI 0.38 to 0.68), heart failure (8.5% vs 7.5%; OR 0.45; 95% CI 0.27 to 0.75), and were on anticoagulation therapy at the time of the stroke (33.8% vs 8.4%; OR 0.17; 95% CI 0.09 to 0.30).
A history of prior myocardial infarction and/or coronary artery disease was more prevalent in the DEVT group compared with IVEVT (28.4% vs 15.8%; OR 0.57; 95% CI 0.38 to 0.87). More patients treated with IVEVT had an M1 occlusion compared with DEVT (63.5% vs 42.8%; OR 1.57; 95% CI 1.06 to 2.33). Cardioembolic strokes were more common in patients treated with IVEVT than those treated with IVT (53.3% vs 40.0%; OR 0.35; 95% CI 0.13 to 0.92). Patients undergoing IVEVT also had higher serum glucose levels than those patients treated with DEVT (6.9 mmol/L vs. 6.6 mmol/L; WMD 0.41; 95% CI 0.28 to 0.54). The IVEVT group had a higher baseline National Institutes of Health Stroke Scale score than the IVT group (17.0 vs 16.5; WMD −0.91; 95% CI −1.77 to −0.05).
The time from symptom onset to hospital arrival, needle puncture, and groin puncture are summarized in the online supplementary table 5. Compared with patients treated with DEVT, patients treated with IVEVT had a longer delay from symptom onset to groin puncture time (241 min vs 180 min; WMD 62.0; 95% CI 40.4 to 83.6), hospital arrival to groin puncture time (117 min vs 101 min; WMD 53.0; 95% CI 38.8 to 67.2), and time from symptom onset to reperfusion (286 min vs 232 min; WMD 62.3; 95% CI 32.4 to 92.2). Patients treated DEVTc had significantly longer delays from symptom onset to groin puncture time than those treated with IVEVT (280 min vs 241 min; WMD −56.6; 95% CI −108.4 to −4.9) (see online supplementary table 5).
In the network meta-analysis, there was no difference in the number of patients who achieved a modified Rankin Scale score of ≤2 at 90 days for DEVT versus IVEVT (OR 1.11; 95% CI 0.75 to 1.66), DEVTc versus IVEVT (OR 1.30; 95% CI 1.00 to 1.79), and DEVT versus DEVTc (OR 1.45; 95% CI 0.93 to 2.35) (figure 3A). IVEVT was associated with a 97% increase in the modified Rankin Scale score of ≤2 compared with IVT (OR 1.97; 95% CI 1.42 to 2.78). Similarly, DEVT was associated with significantly improved functional outcome compared with IVT (OR 2.19; 95% CI 1.32 to 3.75). There was no significant difference in rates of achieving a modified Rankin Scale score of ≤2 at 90 days between DEVTc and IVT (OR 1.52; 95% CI 0.96 to 2.33). Heterogeneity was low (τ2=0.14). Bayesian Markov chain Monte Carlo modeling demonstrated that DEVT had a 70% probability of being ranked as the best treatment for achieving a modified Rankin Scale score of ≤2, higher than IVEVT (29%), DEVTc (1%), and IVT (0%) (figure 4A).
For comparison, pairwise analysis was also performed (see online supplementary figures 4–6). IVT had significantly fewer numbers of patients with a modified Rankin Scale score of ≤2 compared with IVEVT (OR 0.53; 95% CI 0.40 to 0.69). However, there was no significant difference found between IVEVT versus DEVT (OR 0.92; 95% CI 0.65 to 1.30; P=0.65), or IVEVT versus DEVTc (OR 1.19; 95% CI 0.99 to 1.44; P=0.06). DEVT was superior to DEVTc although this was based on a single study (OR 2.01; 95% CI 1.02 to 3.94) (see online supplementary figure 4).
There was no significant difference in 90 day mortality between all treatments, including DEVT versus IVEVT (OR 0.77; 95% CI 0.53 to 1.05), IVEVT versus DEVTc (OR 0.78; 95% CI 0.50 to 1.20), and IVEVT versus IVT (OR 0.76; 95% CI 0.48 to 1.18) (figure 3B). There was no significant heterogeneity (τ2=0.14). Bayesian Markov chain Monte Carlo modeling demonstrated that DEVT had an 86% chance of being ranked as the treatment associated with the lowest mortality at 90 days, higher than IVEVT (10%), IVT (3%), and DEVTc (1%) (figure 4B). Pairwise analyses also showed no significant difference between treatment groups (see online supplementary figure 5).
There were no significant differences in rates of symptomatic intracranial hemorrhage (ICH) between all treatment groups (figure 3C). Heterogeneity was low (τ2=0.14). However, DEVT was associated with a 53% reduction in asymptomatic ICH compared with IVEVT (OR 0.47; 95% CI 0.29 to 0.76). The Bayesian Markov chain Monte Carlo modeling Rankograms are shown in figure 4C. Similarly, pairwise meta-analyses demonstrated no significant difference in the rates of symptomatic ICH between all treatment groups (see online appendix supplementary figure 5).
There was no difference in asymptomatic ICH between DEVTc and IVEVT (OR 0.83; 95% CI 0.50 to 1.33) (figure 3D). There was a trend for lower rates of asymptomatic ICH in DEVT compared with DEVTc (OR 0.57; 95% CI 0.30 to 1.09). Heterogeneity was low (τ2=0.13). Compared with IVEVT (0%), DEVTc (2%), and IVT (9%), DEVT had the highest probability (89%) of being the treatment with the lowest rates of asymptomatic ICH (figure 4D). In pairwise meta-analysis, IVT was associated with lower rates of asymptomatic ICH compared with IVEVT (OR 0.68; 95% CI 0.49 to 0.96), and IVEVT was associated with higher rates of asymptomatic ICH compared with DEVT (OR 2.03; 95% CI 1.34 to 3.07) (see online supplementary figure 7).
Compared with IVEVT, DEVT was associated with a 49% reduction in any ICH (OR 0.51; 95% CI 0.33 to 0.79) (see online supplementary figure 1A). There was no significant difference in rates of ICH between IVEVT and DEVTc (OR 0.88; 95% CI 0.56 to 1.36) and between DEVT and DEVTc (OR 0.58; 95% CI 0.32 to 1.04). There was no significant difference between all other treatments. Heterogeneity was low (τ2=0.12).
The league table is shown in the online supplementary figure 2A. Bayesian Markov chain Monte Carlo modeling revealed that DEVT had a 87% probability of being associated with the lowest rates of all ICH, higher than IVT (11%), DEVTc (2%), and IVEVT (0%) (see online supplementary figure 3A). Pairwise meta-analysis revealed that IVT was associated with lower rates of any ICH compared with IVEVT (OR 0.74; 95% CI 0.53 to 1.03), and IVEVT was associated with higher rates of any ICH compared with DEVT (OR 1.93; 95% CI 1.31 to 2.85) (see online supplementary figure 8).
Patients treated with DEVT had higher rates of reperfusion compared with IVEVT (OR 1.73; 95% CI 1.04 to 2.94) (see online supplementary figure 1B). DEVT was also associated with improved reperfusion rates compared with DEVTc (OR 2.03; 95% CI 1.16 to 3.59). There was no significant difference in reperfusion rates between IVEVT and DEVTc (OR 1.17; 95% CI 0.86 to 1.60). All three endovascular thrombectomy strategies had greater rates of reperfusion compared with IVT; IVEVT (OR 9.7; 95% CI 4.7 to 21.0), DEVT (OR 16.8; 95% CI 6.9 to 43.4), and DEVTc (OR 8.28; 95% CI 3.8 to 19.2). The league table is shown in the online supplementary figure 2B. Bayesian Markov chain Monte Carlo modeling demonstrated that DEVT had a 98% probability of having the highest reperfusion rates compared with IVEVT (2%), DEVTc (0.1%), and IVT (0%) (see online supplementary figure 3B). The individual pairwise meta-analyses are shown in the online supplementary figure 9. Patients treated with IVEVT had higher rates of reperfusion compared with those treated with IVT (OR 0.11; 95% CI 0.05 to 0.22). Patients treated with DEVT had greater rates of reperfusion than those treated with IVEVT (OR 0.56; 95% CI 0.35 to 0.90).
Individual pairwise meta-analyses were performed for each of the outcomes and similar trends to the network meta-analyses were identified (see online supplementary figures 4–9). Sensitivity analyses were performed for network meta-analyses after exclusion of each trial, with no major differences in trends (see online supplementary figures 10–15). Bayesian network meta-analyses were also separately completed for RCTs and prospective studies subgroups. These analyses yielded similar results to those that were mentioned above.
To the best of our knowledge, this is the first network meta-analysis to be performed in the current era following major contemporary stroke trials with a specific focus on DEVT and DEVTc. When comparing IVEVT with DEVT, our network meta-analysis suggests there are no differences in good functional outcome and mortality at 90 days. DEVT was associated with a 49% reduction in ICH compared with IVEVT, mostly driven by a lower rate of asymptomatic ICH. Patients treated with DEVT also had significantly higher rates of reperfusion compared with bridging therapy with IVEVT. Our analysis also demonstrated that despite DEVTc being performed in sicker patients who presented later, comparable 90 day good functional outcome and mortality rates can be achieved compared with endovascular strategies performed in healthier patients without contraindications to tissue plasminogen activator (tPA). At this time, there are no RCTs which have examined DEVT, and thus registry data constitutes the ‘next best thing’. Since the crux of our network meta-analysis is the comparison of DEVT with currently accepted methods, this was a necessary choice. The challenge that faces the meta-analyst on this scale is that patient level data are not equally available between studies, which might bias results due to this potentially large inequality. It is standard statistical practice to use summary data for a meta-analysis when individual patient data are lacking.
There remains a lack of randomized evidence to support or refute the role of DEVT compared with IVEVT in the context of acute ischemic stroke within the tPA delivery timeframe and without contraindications. Although pooled analyses of randomized studies have attempted to address this question, these studies are limited by the lack of direct comparative evidence, as well as conclusions based on patients with contraindications to tPA.9–11 Patients who undergo DEVTc (ie, with contraindications to tPA) are a distinct population who are likely to present later (>4.5 hours), have a prior history of strokes or transient ischemic attacks, be on anticoagulation, and have a higher number of comorbidities, and as such their results cannot be directly applied to DEVT. Although prior studies have supported the feasibility of a direct endovascular thrombectomy approach in patients who are ineligible for IVT, these prior pooled studies do not address the key clinical questions relating to whether bridging IVT should be withheld in patients undergoing mechanical thrombectomy who otherwise have no contraindications to IVT.
The recent series of RCTs clearly and consistently demonstrated that endovascular thrombectomy results in improved neurological outcomes when used in addition to tPA.1–3 These results reaffirm this finding, as patients treated with IVEVT bridging therapy had a 97% increase in the number of patients who achieved a modified Rankin Scale score of ≤2 compared with those receiving IVT only.
In the present study, the rate of ICH was 49% lower in the DEVT group compared with the IVEVT group, largely due to a reduction in asymptomatic ICH. Although the prognostic outcomes associated with asymptomatic ICH are currently unclear, there has been some suggestion that they may lead to cognitive and functional neurological impairment over time.23
Interestingly, reperfusion rates for DEVT were superior to IVEVT. One possible explanation may be the incomplete lysis of the clot by tPA, resulting in embolization of smaller fragments into the distal arteries, which become irretrievable by DEVT. It is important to note that the cerebral vascular territories may remain at risk after the initial thrombolysis, depending on the pattern of vascular branching, primary thrombus fragmentation, and subsequent distal impaction. Alternatively, reperfusion may be time dependent and this analysis, in addition to others, has shown that the administration of tPA increases the time delay to endovascular therapy19 24 (see online supplementary table 5).
It is striking that the higher recanalization rate with DEVT does not in fact eclipse the functional outcomes of IVEVT. We suppose that is explained by the fact that brain which is lost cannot be recovered. Operators use CT and angiography for selection to save time, but also treat large lesions. In the studies with adequate selection, outcomes are good for both IVTEVT and DEVT. However, in IVEVT, patients would have to have received intravenus tPA in the 3 hour window and thus more viable brain is available for salvage. Stroke outcomes are dependent on many factors, some of the key ones being age, premorbid status, workflow and time, baseline imaging, and quality and speed of reperfusion. In addition, Thrombolysis in Cerebral Infarction (TICI) 2b is significantly different from TICI 3, and increasingly there is use of TICI 2c due to this. Unless we can match the rest of these factors precisely (and given the relatively small sample size), we cannot expect a direct correlation between TICI and outcomes.
In our analyses, comparing DEVTc and IVEVT, there were no difference in the rates of ICH. There were also no differences in reperfusion rates between the two groups. DEVT was associated with significantly less ICH and higher rates of reperfusion compared with DEVTc. This likely reflects the differences between the study populations. Since DEVTc includes patients who were contraindicated for tPA, some of these patients may have been on oral anticoagulants or presented with significant delays from symptom onset to hospital presentation.
Wang et al reported a significant difference in procedure time between IVEVT and DEVT in their cohort but we did not find this on meta-analysis. Number of passes was not reported reliably enough for comparisons to be performed.20
There were no differences in mortality rates between all treatment groups. This is consistent with the individual RCTs comparing IVEVT and IVT, which demonstrated improvements in functional outcomes but not survival. However, there was a trend favoring DEVT over IVEVT, providing reassurance that DEVT alone is unlikely to have major deleterious impacts on mortality.
In a recent analysis by Bellwald et al, they authors noted no difference in outcome with IVEVT and DEVT, as in our study.25 Unlike in the present analysis, the authors found a significant difference in mortality favoring DEVT, likely explained by the fact that their analysis was constructed with matched pairs which we sought to avoid.
Whether or not bridging tPA is required is a provocative question. The key issue is that, prior to reaching the hospital, it is not known whether or not the patient has an emergent large vessel occlusion (ELVO) (most strokes will not have an ELVO). For patients that do not have an ELVO, going to a peripheral center makes sense, as these should generally be closer. However in patients with ELVO, one could argue that the mothership approach is reasonable based on our analysis. The RACECAT trial (Direct Transfer to an Endovascular Center Compared to Transfer to the Closest Stroke Center in Acute Stroke Patients With Suspected Large Vessel Occlusion) is currently investigating this matter. Future RCTs are needed to properly assess and compare the two treatment options, with our study suggesting equipoise exists. Currently, two RCTs (Solitaire with the Intention for Thrombectomy Plus Intravenous t-PA Versus DIRECT Solitaire Stent-retriever Thrombectomy in Acute Anterior Circulation Stroke (SWIFT DIRECT) and Multicenter Randomized CLinical trial of Endovascular treatment for Acute ischemic stroke in The Netherlands investigating the added benefit of intravenous alteplase prior to intra-arterial thrombectomy in stroke patients with an intracranial occlusion of the anterior circulation (MR CLEAN-NO IV))26 27 are aiming to compare endovascular therapy with and without bridging thrombolysis. The management of acute ischemic strokes may potentially undergo a paradigm shift, mirroring the changes seen in the treatment of ST elevation myocardial infarctions.28 29
The present study has several limitations. One of the consequences of the strict inclusion criteria was the paucity of studies available for analysis. There were only three studies which reported the outcomes of patients undergoing DEVT alone that did not include patients that were contraindicated for tPA. Second, the inclusion of non-randomized evidence means there is a risk of selection bias, however, this was mitigated using propensity score matched data where available and the inclusion of only prospectively collected data. Furthermore, the comparison of baseline characteristics demonstrated that the IVT, IVEVT, and DEVT groups were generally well matched. Network meta-analysis offers a method with greater power and precision for rare events while controlling for publication bias and small study effects. The network model was tested for consistency and heterogeneity, with low heterogeneity across outcomes reported and, in general, the analyses from random and fixed effects models were consistent.
In the present Bayesian network meta-analysis, DEVT was not associated with poorer neurological outcomes compared with IVEVT. Compared with IVEVT, DEVT in patients without contraindications to IVT was associated with decreased ICH, particularly asymptomatic ICH, and an increase in reperfusion rates. These results can be confirmed by future RCTs, which should seek to reassess the role of bridging IVT in patients with acute ischemic stroke undergoing endovascular thrombectomy.
KP and AAD contributed equally.
Contributors All authors made substantial contributions to the conception or design of the work, drafting the work or revising it critically for important intellectual content, final approval of the version to be published, and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial, or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.