Intravenous tissue plasminogen activator (IV-TPA) administered within 4.5 hours after stroke onset is the only proven therapy for improving outcomes in patients with acute ischemic stroke (AIS) [1,2]. However, acute large artery occlusions are usually resistant to IV-TPA [3,4]. Endovascular recanalization therapy (ERT) has an advantage of a higher recanalization rate, but disadvantages of longer time required for treatment initiation, the need for advanced and resource-intensive stroke care system, and potential increase in symptomatic intracranial hemorrhage (SICH).

In an earlier meta-analysis of randomized controlled trials (RCTs), ERT substantially increased recanalization rates and improved clinical outcomes in AIS compared to no thrombolytic therapy. Despite an increased risk of SICH, ERT did not increase mortality [5]. However, two RCTs failed to demonstrate any benefit of ERT (used alone or in addition to IV-TPA) compared to IV-TPA alone [4,6]. However, recent RCTs using stent-retriever thrombectomy have consecutively demonstrated the benefits of ERT plus IV-TPA over IV-TPA alone [7-11].

We conducted a meta-analysis of all published RCTs to generate an overall estimate of the benefits of ERT when added to standard therapy.

We systematically searched PubMed and EMBASE databases (January 1998 to May 2015) with the search terms “ischemic stroke” AND “intra-arterial” AND “thrombolysis or thrombectomy.” We restricted our search to human and clinical trials and included no language restrictions. In addition, we performed manual searches of the bibliographies of all trials. Abstracts and articles were reviewed and identified by two investigators (Ko SB and Hong KS). Inclusion criteria were 1) RCT, 2) active arm receiving ERT, 3) control arm receiving standard therapy including IV-TPA but not ERT, and 4) modified Rankin Scale (mRS) score reported at 90 days or at the end of the trial. We assessed study quality using the Cochrane risk-of-bias algorithm (www.cochrane.org/training/cochrane-handbook) [12].

All data from selected RCTs were independently abstracted by two investigators (Ko SB and Hong KS). The quality of each study was also reviewed by another two investigators (Yu KH and Rha JH). Discrepancies were resolved by consensus and by referencing the original reports. In most outcomes, analyses were based on the intention-to-treat population (SICH rate on perprotocol population in one trial) [13]. The Cochrane Collaboration’s Review Manager Software Package (RevMan 5) was used for this meta-analysis. Using a random-effect model, we generated a pooled estimate as an odds ratio (OR) with a 95% confidence interval (CI). All statistical tests were 2-sided, and a P value of < 0.05 was considered statistically significant.

Primary efficacy outcome was good outcome defined by an mRS score of 0-2. Secondary efficacy outcomes were mRS 0-1, mRS 0-3, good neurological outcome (variable criteria on the National Institutes of Health Stroke Scale [NIHSS] score across trials), good activity of daily living (Barthel Index 85, 90, or 95 to 100); partial or complete recanalization modified arterial occlusive lesion (mAOL) recanalization score, Thrombolysis in Myocardial Infarction (TIMI) grade, Thrombolysis in Cerebral Infarction (TICI) 2-3, and complete recanalization (mAOL, TIMI, or TICI 3). Safety outcomes were SICH, any ICH including asymptomatic ICH and SICH, bed-ridden status or death (mRS 5-6), and mortality. We compared outcomes between the ERT and control groups for all RCTs, stent-retriever trials (stent-retriever thrombectomy >75% of ERT), and RCTs of control arms with IV-TPA therapy (IV-TPA rate >75%). Using data from five stent-retriever trials, shift analysis (shift in disability levels on mRS) was conducted to analyze the effect of ERT on the distribution of mRS outcomes. Heterogeneity was assessed by P values of χ² statistics and by I² statistics. Heterogeneity was considered significant if the P values of χ² statistics were <0.10. We regarded an I² of < 40% as minimal heterogeneity, 40%-75% as modest heterogeneity, and > 75% as substantial heterogeneity [12]. Publication bias was assessed graphically with a funnel plot, and statistically with the Begg’s test.

This updated meta-analysis, including all relevant published trials to date, clearly demonstrates the benefits of ERT in patients with AIS caused by large artery occlusion. ERT significantly increased good outcome (mRS 0-2) and excellent outcome (mRS 0-1) in parallel with increasing good neurological and activity of daily living outcomes. For all trials including patients with no IV-TPA and IV-TPA in the control arms, ERT increased these outcomes by1.8- to 3.1-fold. When separately analyzing trials in which the majority of cases were treated with IV-TPA in the control arms, the magnitudes of ERT’s benefits were similar. In particular, the benefits were pronounced in recent stent-retriever trials, showing that the chances of achieving these outcomes more than doubled or tripled with ERT. In addition, patients treated with stent-retriever thrombectomy had about a 2-fold odds of a favorable shift in the mRS disability score. Beyond statistical significance, the treatment effects were substantial. The NNTs for achieving one additional outcome of mRS 0-2 and mRS 0-1 were 9 and 11 for all trials. The NNT values were even lower with stent-retriever trials, in which 88% of patients assigned to control arms received IV-TPA therapy: for every five patients who were treated with stent-retriever in addition to IV-TPA therapy, one additional patient achieved an mRS 0-2 outcome; for every seven patients, one additional patient had no or minimal disability (mRS 0-1).

ERT substantially increased reperfusion rates, which likely accounted for the clinical benefits observed. The absolute increase in partial or complete recanalization rate with ERT compared to control was 29.1%. In stent-retriever trials, the rate of partial or complete recanalization with ERT was 80.2%, and the absolute increase compared to IV-TPA (88% treated with IV-TPA) was 33.6%. Therefore, for every three patients, stent-retriever ERT in addition to IV-TPA would achieve one additional recanalization.

Although ERT was associated with an increased risk of any ICH, ERT did not increase SICH and mortality in all trials, stent-retriever trials, and RCTs including a control arm with IV-TPA. Therefore, most ICHs were radiological and asymptomatic. In contrast, ERT significantly decreased the rates of extreme disability or death (mRS 5-6) in all analyses. For stroke outcomes, combining extreme disability and mortality seems to be more clinically meaningful than mortality alone. Laypersons as well as stroke experts perceived mRS 5 disability as bad as or even worse than death [21,22], and therefore extreme disability and death are frequently collapsed into a single worst-outcome category in acute stroke trials [10,23].

It would be instructive to compare the magnitude of the ERT’s benefits with those of coronary interventions. In a metaanalysis of RCTs comparing primary percutaneous coronary intervention with intravenous fibrinolytic therapy in ST-segment-elevation myocardial infarction, the NNT was 45 to prevent one short-term mortality and 29 to prevent long-term mortality [24]. When pooling five recent stent-retriever trials, the NNT to avoid one mortality at 3 months with ERT compared to standard therapy (IV-TPA in most cases) was 29, although the reduction was not statistically significant. Moreover, the NNT to prevent one patient from having an mRS 5-6 outcome was only 9, and this reduction was highly statistically significant. Therefore, the benefits of ERT with stent-retriever in severe AIS appear to be greater than the benefits of primary percutaneous coronary intervention in ST-segment-elevation myocardial infarction.

Our findings accord with and extend the results of an earlier meta-analysis [5]. The earlier meta-analysis compared ERT with no thrombolytic therapy using data from five RCTs involving 395 patients. Our study included 15 RCTs involving 2,899 patients, and also compared ERT with contemporary standard therapy largely composed of IV-TPA. Therefore, our results more convincingly indicate the benefits of ERT in the current clinical setting.

For most efficacy outcomes, heterogeneity was present among trials. However, this heterogeneity was mainly driven by variations in the magnitude rather than in the direction of the treatment effect. For SICH, there was no heterogeneity for extreme disability or death and mortality outcomes. In addition, when restricting analysis to stent-retriever trials, we found no significant heterogeneity across the trials for most outcomes except for any ICH and partial or complete recanalization rates. Therefore, the benefit of ERT appears generally consistent across trials, particularly among the stent-retriever trials.

The validity of meta-analysis results can only be guaranteed when the quality of included trials is high. Most RCTs included in this meta-analysis were at low risk of bias for each item except for treatment blinding. However, blinding of treatment assignments in trials comparing endovascular thrombectomy versus medical therapy would not be feasible for practical reason. Most trials used appropriate allocation concealment and blinded outcome assessment, and no trial contained attrition bias, reporting bias, or other forms of bias.

One critique states that only a minority of patients with AIS would be eligible for ERT but the implementation of ERT programs requires tremendous resource allocation. In the USA, only 5%-13% of patients with AIS arrived within the commonly recommended ERT time window of 3-6 hours, and the proportion of patients eligible for ERT was estimated to be even lower [25]. In Korea, about 51% of patients with AIS arrived within 6 hours according to the Report of Assessment for Quality of Acute Stroke Care in Korea 2010 [26]. In a multicenter study, among 27,851 patients with AIS, 4.6% were treated with ERT (ERT alone or combined with IV-TPA) [27]. However, patients with severe stroke are an important group which requires more effective interventions. In recent stent-retriever trials, only 26.4% of patients in control arms, despite the high IV-TPA rate of 77%-100% across the trials, achieved mRS 0-2 outcomes at 3 months (Table 3). In Korea, cardioembolic stroke, which usually presents as severe stroke and might be indicated for ERT in a large proportion of patients arriving within time window, has increased and currently accounts for more than 20% of all AIS cases [28,29], and will further increase in the near future due to a rapid increase of elderly population. With current standard therapies, approximately one-third of patients with cardioembolic stroke are bed-ridden or die at 3 months [30]. According to a multicenter registry study in Korea, about 20% of patients had a severe stroke with an initial NIHSS score of 10 or more likely caused by an acute large artery occlusion [27]. Therefore, patients eligible for ERT do not represent the minority of all ischemic stroke patients.

Our study has several limitations. As with all meta-analyses, there is a risk of publication bias and of failing to identify all relevant studies. However, we included all large RCTs, and missing some small trials would therefore not alter our results. In addition, we could not identify publication bias for any outcome except for SICH in recent stent-retriever trials. Inspecting a funnel plot showed that bias was more likely to be driven by extremely low SICH rates in the ERT arms of two trials [9,10] rather than a true significant bias. Given the variation in the definitions of SICH used across trials, the current pooled estimate of the SICH risk might be imprecise. However, the variation in SICH definition was not substantial. This study-level meta-analysis was not able to indicate the benefits of ERT in important subgroups. A pooled analysis of individual patient data from recent stent-retriever thrombectomy trials is warranted to provide more insights into the efficacy of ERT according to age, initial stroke severity, occlusion site, additional benefit or safety of prior IV-TPA use, onset-to-treatment and onset-to-recanalization times, and final reperfusion grade. Our findings were generated from data from clinical trials that generally enrolled patients at centers equipped with efficient critical pathways and experienced neurointerventionists. Therefore, these results may not be applicable to inexperienced centers.

In conclusion, the current updated meta-analysis shows that ERT compared to the current standard therapy substantially improves clinical outcomes and reduces extreme disability or death. Despite an increased risk of any ICH, ERT does not increase mortality or SICH. The time has come to reorganize our acute stroke care system to provide ERT for eligible patients more frequently and faster.