Funding: This study is not supported by any funding source.
Recent evidence has suggested that intra-arterial thrombolysis may provide benefit beyond intravenous thrombolysis in ischemic stroke patients. Previous meta-analyses have only compared intra-arterial thrombolysis with standard treatment without thrombolysis. The objective was to review the benefits and harms of intra-arterial thrombolysis in ischemic stroke patients.
We undertook a meta-analysis of randomized controlled trials comparing the efficacy and safety of intra-arterial thrombolysis with either standard treatment or intravenous thrombolysis following acute ischemic stroke. Primary outcomes included poor functional outcomes (modified Rankin Scale 3–6), mortality, and symptomatic intracranial hemorrhage. Study quality was assessed, and outcomes were stratified by comparison treatment received.
Four trials (n = 351) comparing intra-arterial thrombolysis with standard treatment were identified. Intra-arterial thrombolysis reduced the risk of poor functional outcomes (modified Rankin Scale 3–6) [relative risk (RR) = 0·80; 95% confidence interval = 0·67–0·95; P = 0·01]. Mortality was not increased (RR = 0·82; 95% confidence interval = 0·56–1·21; P = 0·32); however, risk of symptomatic intracranial hemorrhage was nearly four times more likely (RR = 3·90; 95% confidence interval = 1·41–10·76; P = 0·006). Two trials (n = 81) comparing intra-arterial thrombolysis with intravenous thrombolysis were identified. Intra-arterial thrombolysis was not found to reduce poor functional outcomes (modified Rankin Scale 3–6) (RR = 0·68; 95% confidence interval = 0·46–1·00; P = 0·05). Mortality was not increased (RR = 1·12; 95% confidence interval = 0·47–2·68; P = 0·79); neither was symptomatic intracranial hemorrhage (RR = 1·13; 95% confidence interval = 0·32–3·99; P = 0·85). Differences in time from symptom onset-to-treatment and type of thrombolytic administered were found across the trials.
This analysis finds a modest benefit of intra-arterial thrombolysis over standard treatment, although it does not find a clear benefit of intra-arterial thrombolysis over intravenous thrombolysis in acute ischemic stroke patients. However, few trials, small sample sizes, and indirectness limit the strength of evidence.
After heart disease and cancer, stroke is the third most common cause of death in the United States and Canada and the leading cause of disability [1, 2]. More than 80% of all incident strokes are of ischemic type, caused by an occluding thrombus of a cerebral artery .
The intravenous administration of recombinant tissue plasminogen activator, or intravenous thrombolysis (IVT), is the most proven beneficial treatment for the emergency management of ischemic stroke . IVT is recommended in ischemic stroke patients with stroke onset up to 4·5 h . However, Health Canada and the Food and Drug Administration of the United States have approved only the original recommendation that IVT be given to ischemic stroke patients with symptom onset up to three-hours .
Local intra-arterial infusion of thrombolytic, or intra-arterial thrombolysis (IAT), may benefit late-presentation stroke patients with its longer recommended time window. IAT is currently recommended up to six-hours from stroke onset because of an occlusion of the middle cerebral artery in IVT ineligible patients [4, 6]. However, its comparative efficacy to IVT has not been well studied.
Thus far, six meta-analyses of IAT in ischemic stroke patients have been published [7-12]. A review of mostly nonrandomized studies  found IAT increased good functional outcomes over standard treatment without thrombolysis; and although mortality was not increased, there was a marked increased risk of symptomatic intracranial hemorrhage. Wardlaw et al.  conducted a review of thrombolytic therapy in acute ischemic stroke but made no conclusions on the use of IAT. Their results did not confirm a clinical benefit with IAT over standard treatment without thrombolysis except for increased excellent functional outcomes with intra-arterial urokinase. However, results were stratified by thrombolytic drug type, lending to imprecise and potentially underpowered estimates. Later, Lee et al.  also published a review of IAT in acute ischemic stroke that did not stratify by thrombolytic drug type and found IAT significantly increased good and excellent functional outcomes. IAT was also found to increase the risk of symptomatic intracranial hemorrhage, though with no corresponding increase in mortality. Included in their analysis was one trial of combined IAT + IVT vs. IVT . However, the marginal effect of IAT may differ depending on whether the base treatment is IVT or standard treatment. The potential benefit of IAT was reviewed again by Fields et al.  using randomized trials of IAT to standard treatment without thrombolysis for stroke due to a middle cerebral artery occlusion. They confirmed Lee et al.'s findings in this subset of middle cerebral artery occlusion strokes. Two reviews have compared IAT with IVT, and both reviews included nonrandomized studies [8, 11]. IAT was found to increase recanalization , but it did not increase good functional outcomes .
There are no reviews of randomized trials comparing IAT with IVT. Previous reviews of IAT have lacked a consistent comparator or restricted their analyses to subsets of the stroke population or by the type of thrombolytic drug received. In order to address these issues, we conducted a systematic review of all randomized trials to determine the potential benefit of IAT and analyzed multiple outcomes stratified by type of comparison treatment received.
Using controlled vocabulary and keyword searches surrounding stroke, thrombolysis, intra-arterial administration, and randomized controlled trial (RCT), studies were identified by searching electronic databases and trial registries. Only English-language articles were included for final review. The MEDLINE (1948 to week 5 2011), EMBASE (1980 to week 5 2011), Cochrane Central Register of Controlled Trials (to fourth quarter 2010), and BIOSIS Previews (1995 to February 2011) databases as well as the Stroke Trials Registry, ClinicalTrials.gov, and the Current Controlled Trials Registry were reviewed. EMBASE, MEDLINE, and the Cochrane Central Register of Controlled trials were searched via Ovid; BIOSIS Previews was searched using ISI Web of Knowledge. Our search strategy can be found in Appendix S1.
Eligibility assessment was performed independently in a standardized manner by two reviewers (JN and JH) using a predetermined screening form. We included RCTs where the patient population was diagnosed with ischemic stroke that compared IAT with standard treatment or IVT. No restrictions on drug type, dose, mechanical device, or mechanical disruption were imposed. All disagreements between reviewers were resolved by consensus.
One review author (JN) abstracted data. Information was extracted from each included trial on: (1) characteristics of the trial (number of patients, age, stroke location, time from onset to treatment, method of diagnosis, and length of follow-up); (2) type of intervention (type, dose, and duration), type of medical treatment (type, dose, and duration), and adjunctive therapies; and (3) outcome measures (good and excellent functional outcome, mortality, intracranial hemorrhage, and death due to intracranial hemorrhage). Functional outcomes included good functional outcomes and excellent outcomes, defined as an mRS score  of ≤2 and ≤1, respectively, or by trial definition if a different scale was used. Both outcomes indicate independence in daily activities; as such, good functional outcomes and excellent functional outcomes are not independent of each other. Other outcomes of interest included: mortality at 90 days or end of follow-up; symptomatic intracranial hemorrhage as defined by the National Institutes of Neurological Disorders and Stroke (NINDS) trial  or by the trial's definition; and recanalization at the end of infusion.
We analyzed the data using Review Manager 5·1 (Copenhagen, Denmark). We tested for heterogeneity between trials using the I2 statistic. Fixed-effects Mantel-Haenszel risk ratios and their corresponding 95% confidence intervals were calculated for each outcome measure where possible. Results were stratified by type of comparison treatment received. Statistical significance was judged at the 5% level. Good functional outcomes and excellent functional outcomes were swapped for poor functional outcomes (the nonevents) – poor functional outcomes [modified Rankin Scale (mRS) 3–6] and death or disability (mRS 2-6), respectively – so that a risk reduction favored IAT. Similarly, recanalization was swapped for failure to recanalize.
Risk of bias
The validity of eligible RCTs was assessed using the Cochrane risk of bias tool  in Review Manager 5·1. Egger's tests and funnel plots were planned to evaluate the risk of bias across studies.
The original search produced 543 citations. Once duplicates were removed, 413 citations remained for title and abstract screening. After first-level screening, 33 articles were included for full-text screening of which 27 were excluded because they were not an RCT [17-36], comparator included IAT , no comparator [38-40], or active treatment was confounded by IVT . The full text of two articles could not be retrieved as attempts to contact the authors were unsuccessful [41, 42]. A PRISMA diagram of study selection is presented in Fig. 1. A total of six studies were identified for inclusion in this review [43-48]. The kappa estimate for agreement was 0·59 for first-level screening; for second-level screening, we had perfect agreement.
Study characteristics are presented in Table 1. The included studies randomized 431 patients. Four trials compared IAT with standard treatment (without IVT) [44, 46-48], while two trials compared IAT with IVT [43, 45]. Location of occlusion was restricted to middle cerebral artery in four trials [44-46, 48] and posterior vessels in one trial ; one trial did not restrict to any location . Median National Institutes of Health Stroke Scale  ranged from 14 to 26. Only two trials included participants above 80 years of age [46, 47]. Mechanical clot retrieval was allowed in one trial .
Table 1. Characteristics of included studies evaluating IAT vs. standard treatment or IVT
Time from symptom onset to treatment (median)
Intervention (type, dose)
aNo time reported.
IAT, intra-arterial thrombolysis; IVT, intravenous thrombolysis; NIHSS, National Institutes of Health Stroke Scale; IA, Intra-arterial; SSS, Scandinavian Stroke Scale; MCA, middle cerebral artery.
IA prourokinase (120,000 IU/5 min + max 600,000 IU)
IV heparin (5000 IU)
No specific treatment. Antithrombotic activities not permitted.
mRS 0–2 (90 days)
mRS 0–1 (90 days)
mortality (90 days)
symptomatic intracranial hemorrhage (24 h)
Five trials measured good functional outcomes (mRS 0–2) [43, 45-48], and five trials measured excellent functional outcomes (mRS 0–1) [43, 44, 46-48]; of these trials, four trials reported both good functional outcomes and excellent functional outcomes [43, 46-48]. Four trials reported mortality at 90 days [44-46, 48], one trial at seven-days , and one trial at six-months . Reporting of hemorrhagic complication was varied in name, definition, and time-point reported. Three trials followed the National Institute of Neurological Disorders and Stroke definition of symptomatic intracranial hemorrhage – a computed tomography-documented hemorrhage that caused neurological deterioration [44, 46, 48]. Other trials were less clear on their definition. Recanalization was reported in just two studies, both of which compared IAT with standard treatment [44, 46].
The time from symptom onset was not the same across all trials. Three trials reported a median treatment delay from symptom onset around 5–5·5 h [44-46]. The other three trials had considerably different delays from symptom onset to treatment. Both Ogawa et al.  and Ciccone et al.  reported shorter treatment delays (3·8 h and 2·58–3·28 h, respectively), whereas Macleod et al.  reported a very late treatment time (+11 h from symptom onset). Ogawa et al. used the mean time delay from symptom onset .
The type of thrombolytic drug was also not the same across all trials. Three trials of IAT vs. standard treatment used prourokinase [44, 46, 48]. The remaining trial of IAT vs. standard treatment use urokinase . Of the two trials that compared IAT with IVT, one used alteplase  and the other used prourokinase . Therefore, of six identified trials, only one trial used the recommended thrombolytic drug type of alteplase.
Risk of bias
Results of the Cochrane risk of bias tool are presented in Supporting Figs S1 and S2. In general, threats to the internal validity of included studies were low. Low risk of selection bias was detected; three studies did not report how random sequence generation was performed [44, 45, 47]; however, all included studies conducted appropriate allocation concealment so the overall risk of selection bias was determined to be of low risk. One study had high risk of detection bias as outcome assessors were not blinded .
One study was found to have unclear risk of attrition bias . In this study, 11% in the treatment group did not receive treatment, and 8% in the control group crossed over. In addition, 9% in the treatment group and 10% in the control group had observations that were carried forward, whereas a further 1% and 3% in the treatment and control groups, respectively, were imputed as failure because of missing observations. However, because both good and excellent functional outcome were reported as intention to treat, these potential biases may have worked in the direction opposite to the point estimate, which showed a benefit for the treatment group.
A high risk of reporting bias was detected in one study . Symptomatic intracranial hemorrhage was not reported. As well, recanalization was only reported for the treatment arm.
Funnel plots and Egger's tests were not completed due to the small number of trials.
Synthesis of results
IAT vs. standard treatment
Data for poor functional outcomes (mRS 3–6) were available from three trials, representing 310 patients. IAT reduced poor functional outcomes (mRS 3–6) by 20% over standard treatment [relative risk (RR) = 0·80; 95% confidence interval (CI) = 0·67–0·95; P = 0·01; I2 = 0%] (Fig. 2), a statistically significant finding. A statistically significant reduction in the risk of poor functional outcomes (mRS 2–6) was also found to be associated with IAT. The RR of poor functional outcomes (mRS 2–6) was 0·83 with IAT compared with standard treatment (four trials; n = 350; 95% CI = 0·73–0·94; P = 0·004; I2 = 0%) (Fig. 3).
IAT was not found to increase mortality over standard treatment (four trials; n = 350; RR = 0·82; 95% CI = 0·56–1·21; P = 0·32; I2 = 0%) (Fig. 4). Although mortality was not increased, IAT did raise the risk of symptomatic intracranial hemorrhage compared with standard treatment. Three of the four trials of IAT vs. standard treatment reported symptomatic intracranial hemorrhage. Hemorrhagic complication was nearly four times more likely following IAT compared with standard treatment (three trials; n = 316; RR = 3·90; 95% CI = 1·41–10·76; P = 0·006; I2 = 0%) (Fig. 5).
Only two trials reported recanalization. Both trials compared IAT with standard treatment (n = 198). IAT reduced the failure to recanalize by 21% (RR = 0·79, 95% CI = 0·70–0·89; P = 0·0001) (Fig. 6). Large heterogeneity was detected I2 = 83%.
IAT vs. IVT
Poor functional outcomes (mRS 3–6) were reduced by 32% with IAT; however, the reduction was not found to be statistically significant (two trials; n = 81; RR = 0·68; 95% CI = 0·46–1·00; P = 0·05; I2 = 0%) (Fig. 7). Only one trial, randomizing 54 patients, reported poor functional outcomes (mRS 2–6). IAT reduced poor functional outcomes (mRS 2–6) by 28% over IVT; however, this was not a statistically significant reduction (RR = 0·72; 95% CI = 0·46–1·11; P = 0·14) (Fig. 8).
No statistically significant increase of mortality (two trials; n = 83; RR = 1·12; 95% CI = 0·47–2·68; P = 0·79; I2 = 0%) (Fig. 9) or symptomatic intracranial hemorrhage was found following IAT (two trials; n = 81; RR = 1·13; 95% CI = 0·32–3·99; P = 0·85) (Fig. 10). However, low to moderate heterogeneity was detected in the latter (I2 = 42%). None of the included trials reported failure to recanalize.
We conducted a comprehensive meta-analysis that evaluates the performance of IAT for patients with acute ischemic stroke compared with other strategies. We identified four trials of IAT vs. standard treatment that randomized 351 patients. IAT was found to reduce poor functional outcomes (mRS 3–6) while it increased the risk of symptomatic intracranial hemorrhage. However, mortality was not increased. Finally, IAT was associated with a reduction in failure to recanalize, though results from only two trials were available and heterogeneity was large. We also identified two trials of IAT vs. IVT that randomized 81 patients. IAT was not found to reduce poor functional outcomes (mRS 3–6) or poor functional outcomes (mRS 2–6). However, only one trial was available for the latter outcome. Regarding safety, IAT did not increase mortality or symptomatic intracranial hemorrhage.
The only trial that used alteplase compared IAT with IVT . Other trials used prourokinase [44-46, 48] or urokinase . However, it is alteplase that is the recommended thrombolytic drug type for acute ischemic stroke treatment according to the American Heart Association/American College of Cardiology and American College of Chest Physicians guidelines [4, 6]. Evidence therefore suffers from a degree of indirectness.
Experience with IVT has shown that the treatment effect improves with shorter delay from symptom onset [50, 51]. It is important to note that the time from symptom onset was not the same across all trials. Despite the differences in treatment delay in the present review, no heterogeneity was detected for poor functional outcomes or mortality when IAT was compared with either standard treatment or IVT. However, without sub-group results with specific time strata, definitive conclusions on the effect of timing on the treatment benefit following IAT are premature.
Risks of internal biases in the present review were generally found to be low. One of the largest studies (n = 180) had an unclear risk of attrition bias though the direction of the potential bias was not clear; there were a number of crossovers, patients who failed to receive treatment, and imputed observations . Another study had a high risk of selective reporting bias, although its overall impact on the pooled estimate was low because of its small sample size .
For IAT vs. standard treatment, although heterogeneity was low for all pooled estimates, the precision was also low. Pooled estimates for both mortality and symptomatic intracranial hemorrhage exhibited wide confidence intervals. As heterogeneity is low, imprecision may have been due to an overall lack of power. Pooled estimates for poor functional outcomes (mRS 3–6) and poor functional outcomes (mRS 2–6) also exhibited wide confidence intervals. Relatively small sample sizes were used to generate relative risk estimates and thus the pooled estimates should be interpreted with caution.
Evidence of a net benefit of IAT over IVT is lacking. The lower bound of the confidence interval for the pooled estimate of poor functional outcomes (mRS 3–6) reached unity while it crossed unity for the pooled estimate of poor functional outcomes (mRS 2–6). Two studies (n = 81) were included in the former pooled estimate, and only one study (n = 54) was included in the latter. In addition, one study did not blind outcome assessors. This had an obvious effect on the precision and fragility of the pooled estimates, especially as the study sizes were relatively small. Pooled estimates for mortality and symptomatic intracranial hemorrhage were balanced around unity, although they also showed considerable imprecision. Imprecise, fragile, and nonsignificant clinical benefits provide little evidence of a net benefit of IAT over IVT.
Three reviews of RCTs comparing IAT with standard treatment were identified [7, 9, 12]. Wardlaw et al. stratified results by thrombolytic drug type received . Their results showed a statistically significant reduction in death or dependency defined as mRS 2–6 following IAT with urokinase compared with standard treatment. Two reviews confirm statistically significant increases in good and excellent functional outcomes with no increase in mortality [7, 12]. No reviews of IAT vs. IVT were identified.
Randomizing IAT in early time windows may be controversial. One of the included trials in this review, the SYNTHESIS trial by Ciccone et al. , randomized ischemic stroke patients with stroke onset less than three-hours to IVT or IAT with or without mechanical intervention . The authors concluded that IAT may perform better in earlier time windows. Our results confirm that IAT may outperform IVT. However, estimates are highly imprecise and fragile for poor functional outcomes; in addition, only two trials (n = 81) were available for analysis. IAT has the potential to treat patients in both early and late time windows. With additional evidence, there may be strong evidence of benefit following IAT. The successor to SYNTHESIS, the SYNTHESIS EXP , is continuing the investigation in patients with stroke onset less than 4·5 h. It will be the largest randomized study to date to compare IAT with IVT, with 350 participants, and the first large study to compare IVT with IAT head to head, albeit with or without mechanical intervention.
SYNTHESIS EXP is powered to detect an absolute difference of 15% in excellent functional outcomes. In our review, IAT was associated with an approximate 20% absolute increase in either good or excellent functional outcomes, over IVT. Given our point estimates, SYNTHESIS EXP is well powered for its primary outcome; however, it should be noted that our point estimates are based on one or two trials with small sample sizes. Although SYNTHESIS EXP is well powered to detect expected differences in its primary efficacy outcome, it is underpowered to detect differences in safety outcomes. Our pooled estimates of mortality and symptomatic intracranial hemorrhage were nonsignificant, very fragile, and very imprecise. They cannot be used to make inferences on sample size calculations except to emphasize that any strong evidence of safety will require a very large sample size, likely in excess of 10 000. As this is probably not feasible, the acceptance of IAT is likely to depend on a demonstration of large benefit in good/excellent functional outcomes.
In addition to an adequate sample size to detect differences in safety outcomes with greater precision, future trials should consider a few additional items in trial design and reporting. Trials should use alteplase. Thus far, most trials comparing IAT to either standard treatment or IVT have not used alteplase, yet this is the recommended drug type [4, 6]. As well, when defining symptomatic intracranial hemorrhage, trials should be clearer regarding the time frame, extent of hemorrhage, and criteria for neurological/functional deterioration. Finally, trials should also publish the entire distribution of mRS scores. With such information, a pooled distribution may be possible.
In patients with acute ischemic stroke following treatment with IAT compared with standard treatment, there is a modest reduction in poor functional outcomes with no increase in mortality, albeit with a considerable risk increase in symptomatic intracranial hemorrhage. However, few trials, small sample sizes, and indirectness limit the strength of evidence. Evidence of a reduction in poor functional outcomes following IAT over IVT is lacking, which may be due to a lack of power. To improve the evidence base, future trials should employ alteplase, use a more precise definition of symptomatic intracranial hemorrhage, and publish the entire distribution of mRS scores.