Funding: The study is managed by the steering committee and executive sub-committees and financially supported by the National Health and Medical Research Council (NHMRC) of Australia, National Heart Foundation of Australia (NHF), National Stroke Foundation of Australia (NSF), a Royal Melbourne Hospital Home Lottery Project Grant, and the Royal Australasian College of Physicians (RACP). In kind logistic support is provided by the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Neuroscience Trials Australia, the Melbourne Brain Centre at the Royal Melbourne Hospital, and the Florey Institute of Neuroscience and Mental Health. Covidien (manufacturer of the Solitaire FR device) will provide the study device free of charge and additional infrastructure funding but have no role in the design or analysis of the study. The sponsor of the study is the National Stroke Research Institute (NSRI), a division of the Florey Institute of Neuroscience and Mental Health.
Thrombolysis with tissue plasminogen activator is proven to reduce disability when given within 4·5 h of ischemic stroke onset. However, tissue plasminogen activator only succeeds in recanalizing large vessel arterial occlusion in a minority of patients. We hypothesized that anterior circulation ischemic stroke patients, selected with ‘dual target’ vessel occlusion and evidence of salvageable brain using computed tomography or magnetic resonance imaging ‘mismatch’ within 4·5 h of onset, would have improved reperfusion and early neurological improvement when treated with intra-arterial clot retrieval after intravenous tissue plasminogen activator compared with intravenous tissue plasminogen activator alone.
EXTEND-IA is an investigator-initiated, phase II, multicenter prospective, randomized, open-label, blinded-endpoint study. Ischemic stroke patients receiving standard 0·9 mg/kg intravenous tissue plasminogen activator within 4·5 h of stroke onset who have good prestroke functional status (modified Rankin Scale <2, no upper age limit) will undergo multimodal computed tomography or magnetic resonance imaging. Patients who also meet dual target imaging criteria: vessel occlusion (internal carotid or middle cerebral artery) and mismatch (perfusion lesion : ischemic core mismatch ratio >1·2, absolute mismatch >10 ml, ischemic core volume <70 ml) will be randomized to either clot retrieval with the Solitaire FR device after full dose intravenous tissue plasminogen activator, or tissue plasminogen activator alone.
The coprimary outcome measure will be reperfusion at 24 h and favorable clinical response (reduction in National Institutes of Health Stroke Scale by ≥8 points or reaching 0–1) at day 3. Secondary outcomes include modified Rankin Scale at day 90, death, and symptomatic intracranial hemorrhage.
Treatment with intravenous tissue plasminogen activator (IV tPA) within 4·5 h of stroke onset is proven to reduce disability  but only rapidly recanalizes approximately 30% of occluded arteries [2, 3]. Recanalization rates are dependent on the location of arterial occlusion: <10% in internal carotid artery (ICA) occlusion, around 30% for proximal middle cerebral artery (MCA), and 42% for distal MCA occlusions (M2) at 2 h. This leaves significant room for improvement in reperfusion rates for patients with major vessel occlusion.
Intra-arterial (IA) clot retrieval promises increased rates of recanalization but is resource intensive and recent randomized trials have shown no benefit compared with tPA [4, 5]. The Interventional Management of Stroke 3 (IMS-3) randomized trial comparing 0·9 mg/kg IV tPA with a bridging strategy of tPA (0·6 mg/kg for most of the trial) followed by IA therapy was recently halted due to futility, after 656 patients had been enrolled . The Local versus Systemic Thrombolysis for Acute Ischemic Stroke (SYNTHESIS Expansion) trial was unable to demonstrate any difference in outcome between IV tPA and direct IA therapy . Neither study raised any safety concerns with symptomatic hemorrhage rates of ∼6% in both IV and IA arms.
One possible reason for the lack of a positive result in these studies was the lack of patient selection using advanced imaging to ensure vessel occlusion and salvageable tissue. This may have led to inclusion of patients with large, irreversibly injured ischemic core volumes who had little chance of good functional recovery [6, 7]. Unselected patients also demonstrate a strong relationship between time and clinical outcome, which may have confounded any benefit from increased recanalization . In contrast, patients selected with good collateral flow show less time-dependence in their clinical response to reperfusion .These neutral trials also employed intra-arterial tPA and early generation MERCI and Penumbra devices. In IMS-3, the rate of effective reperfusion (TICI 2B/3) for ICA or MCA occlusion was ∼40% . The development of retrievable stents (‘stentrievers’) such as the Solitaire FR device (Covidien, Mansfield, Massachusetts, USA) has increased expected rates of effective reperfusion to >80% . The recent SWIFT randomized trial demonstrated significantly greater recanalization without symptomatic hemorrhage and improved clinical outcomes in patients treated with Solitaire vs. the MERCI device .
In EXTEND-IA, we hypothesize that the optimal selection strategy for a clot retrieval trial is the dual target of an occluded major vessel amenable to device retrieval, together with a small ischemic core and significant ‘mismatch’ as an estimate of ischemic penumbra identified using multimodal CT or MR imaging (Fig. 1). A large mismatch and small ischemic core correlates with good collateral blood flow [12, 13]. Ischemic core volume is directly related to functional outcome, and a large core at baseline limits the potential response to reperfusion [6, 7]. Reperfusion (visualized using CT or MRI perfusion imaging) is associated with attenuation of ischemic core growth [14-16], which otherwise tends to expand to the limits of the penumbra .
The DEFUSE , EPITHET  and DEFUSE-2  studies have demonstrated a clear association between major reperfusion and good clinical outcome in patients with mismatch in time windows extending beyond the current 4·5 h shown to be clinically effective for IV tPA in unselected patients. Despite this association, mismatch-based treatments have not yet been proven to improve clinical outcome compared with placebo . However, the existing randomized trials have used intravenous thrombolytic agents with modest reperfusion rates and also employed variable mismatch criteria. Confounding factors in the use of mismatch have included qualitative visual assessment that has been shown to be inaccurate  and inclusion of patients with substantial volumes of benign oligemia rather than true penumbral tissue. Recently published reanalyses of the The Desmoteplase in Acute Ischemic Stroke Trial 2 (DIAS-2) trial have demonstrated benefits of desmoteplase in the sub-groups with either major vessel occlusion  or a large mismatch . The most recent mismatch-based trial was Mechanical Retrieval and Recanalization of Stroke Clots Using Embolectomy (MR-RESCUE) , which failed to show any benefit of clot retrieval over standard therapy. Limitations of the study design included use of less effective embolectomy devices, a complex automatic penumbral selection methodology that failed in 42% of cases, inclusion of patients with large ischemic core lesions and late assessment of recanalization at seven days.
The optimal definition of mismatch has been refined in recent years. Increasing the time to maximum (Tmax) perfusion threshold from ≥2 to >5–6 s improves identification of tissue destined for infarction without reperfusion [25-27] and enriches the rate of favorable clinical response to reperfusion . MRI diffusion-weighted-imaging (DWI) has been confirmed as a clinically reliable indicator of ischemic core [29, 30]. An optimal CT perfusion mismatch paradigm using relative cerebral blood flow (rCBF) to estimate ischemic core has also been developed and validated against multimodal MRI [31, 32]. An automated mismatch selection package RAPID (RApid processing of PerfusIon and Diffusion, Stanford University, California, USA)  will be used in the EXTEND-IA trial to generate standardized, quantitative mismatch estimation using CT or MRI within minutes of image acquisition. CT has practical advantages over MRI in accessibility, speed, ease of monitoring of unstable patients, and reduced contraindications (e.g. pacemakers and stents). A recent trial using ‘dual target’ CT vessel occlusion and mismatch was able to demonstrate superiority of tenecteplase over tPA for both clinical and imaging end-points with only 25 patients per group . This suggests that well-designed multimodal CT mismatch criteria can indeed identify an enriched population of patients with the potential for a major clinical response to reperfusion.
The combination of new devices such as the Solitaire FR with high rates of procedural success and refined CT and MR imaging selection criteria based on fully automated software processing create an ideal environment for EXTEND-IA.
To test the hypothesis that anterior circulation ischemic stroke patients, selected with ‘dual target’ vessel occlusion and CT or MRI mismatch within 4·5 h of onset would have improved reperfusion and early neurological improvement when treated with intra-arterial clot retrieval after intravenous tPA compared with intravenous tPA alone.
The EXTEND-IA study is a phase II, multicenter, prospective, randomized, open-label, blinded-endpoint (PROBE) study (two arms with 1:1 randomization via a centralized website) in ischemic stroke patients. Patients will be stratified by site of vessel occlusion into one of the following strata: (a) ICA occlusion; (b) proximal middle cerebral artery (MCA-M1); (c) distal middle cerebral artery (MCA-M2). Approximately 14 centers in Australia and New Zealand will participate.
1.Patients presenting with anterior circulation acute ischemic stroke eligible using standard criteria to receive IV tPA within 4·5 h of stroke onset.
2.Patient, family member, or legally responsible person depending on local ethics requirements has given informed consent.
3.Patient's age is ≥18 years (with no upper age limit).
4.Intra-arterial clot retrieval treatment can commence (groin puncture) within six-hours of stroke onset.
Imaging inclusion criteria
5.Arterial occlusion on CT or MR angiography of the ICA, M1, or M2
6.Mismatch – using CT or MRI with a Tmax >6 s delay perfusion volume and either CT-rCBF or DWI ischemic core volume.
(a) Mismatch ratio >1·2;
(b) Absolute mismatch volume >10 ml; and
(c) Ischemic core lesion volume <70 ml.
Standard contraindications to intravenous tPA.
Rapidly improving symptoms at the discretion of the investigator.
Prestroke mRS score of ≥2 (indicating previous disability)
Inability to access the cerebral vasculature in the opinion of the neurointerventional team or contraindication to use of the Solitaire FR device.
Contraindication to imaging with contrast agents.
Participation in any investigational study in the previous 30 days.
Any terminal illness such that patient would not be expected to survive more than one-year.
Neurological impairment and functional scores will be measured by a healthcare professional trained in their administration and blinded to the treatment assignment. The National Institutes of Health Stroke Score (NIHSS) will be performed before randomization and repeated at 18–30 h, three-days, and 90 days after stroke onset. The modified Rankin Score (mRS) will be assessed at day 90.
Imaging assessment and parameters
Centers will perform standard multimodal CT or MR before or immediately after commencing intravenous tPA. Eligibility will then be determined using ‘dual target’ vessel occlusion on CT/MR angiography and mismatch using RAPID automated software. An MRI will be performed 18–30 h poststroke onset to assess reperfusion, recanalization, and ischemic core growth . The imaging protocols will follow current international consensus guidelines .
Radiological outcome measures will be centrally analyzed, blinded to treatment allocation. The presence and degree of reperfusion will be determined as the percentage difference between the coregistered 24 h and acute perfusion lesion volumes. Recanalization will be determined based on 24 h MRA and classified using the Thrombolysis in Myocardial Infarction (TIMI) scale . Ischemic core growth is calculated as the absolute and relative difference in volume between CT-rCBF estimated ischemic core at baseline and 24 h DWI lesion volume.
The investigational device (Solitaire FR, Covidien) is a retrievable stent delivered at the site of intracranial vessel occlusion and then removed under negative pressure aspiration. Patients randomized to intra-arterial clot retrieval will receive full dose IV tPA and be transferred for neurointervention. If dramatic clinical recovery occurs after randomization, the patient should still undergo diagnostic angiography. Recovery does not necessarily imply full recanalization, and angiography is the best method to establish whether there is an ongoing target for therapy. The use of conscious sedation or general anesthesia for the procedure is at the discretion of the individual site neurointerventionalist. Close attention will be paid to maintaining stable blood pressure and minimizing delays in starting the procedure, which must commence (groin puncture) within six-hours of stroke onset. Before deploying the Solitaire device, the site of vessel occlusion should be confirmed using digital subtraction angiography. If there is no lesion amenable to clot retrieval in the ICA, M1, or M2, the procedure is complete and intervention would not proceed. Up to two passes of the Solitaire device are permitted within the manufacturer's instructions. A second device could be used for further passes up to a maximum of three per arterial segment. An angiogram should be performed after each pass of the device. Proximal balloon occlusion is recommended as per manufacturer's instructions. The procedure must be completed within eight-hours of stroke onset. During the procedure, catheters may be flushed with heparinized saline at a concentration of 1000 units heparin per litre 0·9% sodium chloride. Use of other devices, lytic agents, angioplasty, or intracranial stenting is not permitted within the protocol. Stenting of the extracranial internal carotid artery is permitted when absolutely necessary to obtain access to distal occlusion or to prevent acute reocclusion. This may require the use of antiplatelet agents. Otherwise, no antiplatelets/anticoagulants should be given until at least 24 h after the procedure. A control angiogram should be obtained at the conclusion of the procedure and will be centrally graded for angiographic revascularization using the modified Treatment In Cerebral Ischemia (mTICI) classification  and any embolization into new territories.
Median percentage reperfusion at 24 h post stroke, adjusted for site of arterial occlusion.
NIHSS reduction ≥8 points or reaching 0–1 at 3 days (favorable clinical response) adjusted for baseline NIHSS and age.
Secondary outcomes are mRS at 3 months – ordinal full scale analysis, mRS 0–1 and mRS 0–2; symptomatic intracranial hemorrhage [symptomatic intracranial hemorrhage includes any sub-arachnoid bleeding associated with clinical symptoms and symptomatic intracerebral hemorrhage (SICH). SICH is defined as parenchymal hematoma type 2 (PH2) within 36 h of treatment combined with ≥4 point increase in NIHSS from baseline, or the lowest NIHSS value between baseline and 24 h ]; and death due to any cause.
Tertiary outcomes are reperfusion on perfusion imaging at 24 h poststroke without symptomatic intracerebral hemorrhage; revascularization at 24 h poststroke; ischemic core growth at 24 h; median reduction in stroke severity (NIHSS) at 24 h; median reduction in stroke severity (NIHSS) at day 3; NIHSS reduction ≥8 points or reaching 0–1 at three-months; and home time (number of days spent at home in the first 90 days).
The intra-arterial treatment is open-label. All those involved in the subsequent clinical and imaging assessment of outcomes will be blinded to treatment allocation. The independent Data Safety Monitoring Board (DSMB) will have access to unblinded grouped data.
All randomized subjects will be included in analyses on an intention-to-treat basis. The original sample size estimation is based on the assumption that the patient mix is broadly reflective of population-wide prevalence (i.e. ICA: 30%, M1: 50%, M2: 20%). An estimated total sample size of 100 patients (with 50 patients in each of treatment and control arms) should yield 80% power to detect both:
(a) a significant difference of 24% in strata-weighted median reperfusion at 24 h (80% in treatment vs. 56% in control arm) at two-sided statistical significance threshold of P = 0·025; and
(b) a significant difference of 33% (63% in treatment vs. 30% in control arm) in the proportion of patients with ≥8 point reduction in NIHSS or reaching 0–1 at three-days (favorable clinical response) adjusted for baseline NIHSS and age at two-sided statistical significance threshold of P = 0·025.
Although the preplanned sample size is 100 patients, adaptive increase in sample size will be performed if the results of interim analysis using data from the first 60 patients are promising as per the methodology of Mehta and Pocock . The conditional power will be evaluated in the interim analysis by the blinded study statistician, and if it falls in the prespecified ‘promising’ range the sample size will be increased, subject to a predetermined upper limit (150 patients) to increase the conditional power to 0·80 level (independently for both coprimary end-points using alpha 0·025).
The primary efficacy analysis will be based on an intention-to-treat basis. Two coprimary outcomes will be compared using two-sided significance tests. Statistical significance level adjustment will be made using the Bonferroni–Holm step-down procedure .
For the coprimary outcome analysis:
(a) the reperfusion outcomes will be compared between treatment and control arms of the trial adjusted for site of baseline arterial occlusion (all three strata) using the van Elteren test (a stratified version of the Wilcoxon rank-sum test); and
(b) the proportion of patients with a favorable clinical response indicated by an NIHSS reduction ≥8 points or reaching 0–1 at three-days will be compared between treatment and control arms of the trial adjusted for age and baseline NIHSS score using binary logistic regression.
Although both baseline covariate adjusted and unadjusted results will be reported, baseline covariate adjusted analysis is prespecified as the primary outcome analysis for this trial. Despite best efforts, occasional patients are unable to undergo reperfusion imaging at 24 h. To maintain intention-to-treat principles, these patients will be imputed with 0% reperfusion. To avoid potential bias from investigator's discretion in not obtaining imaging, an independent, blinded clinician is required to determine cases where 24-h imaging of the patient is felt not to be feasible. The results will also be reported for the ‘target group’ who received IA clot retrieval as per protocol compared with the group randomized to IV tPA to adjust for effects such as recanalization before cerebral angiogram and any off-protocol interventions.
For the secondary outcome analysis, assumption-free, ordinal analysis of mRS will be undertaken on the full range (0–6) of the mRS. The proportions of mRS 0–1 and mRS 0–2 outcomes will also be compared between IV-IA and IV only arms of the trial adjusted for age and baseline NIHSS score using a binary logistic regression model.
Data safety monitoring
Interim safety analysis will be undertaken by the independent DSMB when 60 patients have completed the three-month assessment. The Haybittle–Peto procedure for generating early stopping boundaries will be used. Rates of mortality at three-months and the incidence of symptomatic intracranial hemorrhage within 36 h of intervention between groups will be tested independently. A recommendation of early termination due to safety reasons will be considered by the DSMB if the corresponding Haybittle–Peto boundary (P = 0·003, Z = 3) is crossed. No formal interim analyses for efficacy or futility are planned.
Intra-arterial therapy is intuitively attractive due to the demonstrated high rates of recanalization achievable with modern devices, paired with the relatively modest rates of symptomatic intracranial hemorrhage in more recent studies. However, equipoise persists given the current neutral trial data [4, 5] and the more invasive, resource-intensive nature of intra-arterial therapy. EXTEND-IA combines refined imaging selection with the most effective intra-arterial device currently available in a randomized trial using the current standard of care (IV tPA) as comparator. The ‘bridging’ IV then IA design ensures all patients receive the proven therapy without delay and capitalizes on the ability to rapidly institute IV therapy while arranging the more logistically complex IA technique. The use of up-front imaging selection rather than delaying until IV thrombolysis is complete, allows early activation of the neurointerventional team. This saves ∼1 h compared with delaying imaging until tPA is complete to assess ‘tPA failure’. Time has a critical influence on the rate of good outcomes. Previous experience suggests that <15% of our patients will recanalize within 2 h of IV tPA so waiting to image and thus delaying clot retrieval for all patients in order to prevent diagnostic angiography (which carries minimal risk) in <15% seems inadvisable.
The choice of reperfusion as a coprimary end-point is based on its key physiologic relevance and consistent strong association with clinical outcome across the literature. Currently, no randomized head-to-head comparison of IV and IA therapy has demonstrated increased reperfusion with IA therapy. Early neurological recovery assessed using NIHSS is also strongly associated with functional outcome [19, 34]. Day 3 assessment was chosen to avoid any residual influence of peri-procedural anesthesia.
The recent IMS-3  and SYNTHESIS  results emphasize the ongoing equipoise regarding the efficacy of intra-arterial therapy. EXTEND-IA is unique and differs from IMS-3 in the use of advanced imaging selection, a homogeneous intervention with a single device that has the highest reported recanalization success (rather than a mixture of IA lysis and several different devices), extension of eligibility from 3 to 4·5 h from stroke onset consistent with current clinical practice and no upper age limit (rather than an arbitrary 82 years). The combination of imaging selection and a highly effective device has the potential to maximize clinical benefit and reduce hemorrhage risk with the minimum sample size.
We gratefully acknowledge Prof. Gregory Albers, Prof. Roland Bammer, and the Stanford University Stroke Center for providing RAPID software free of charge for the EXTEND-IA trial; Simon McBride, Christopher Stanbridge, and Karen Harrap of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) for creating the electronic case record and web randomization and Elise Cowley of Neuroscience Trials Australia for study management.