A multicentre, randomized, double-blinded, placebo-controlled Phase III study to investigate EXtending the time for Thrombolysis in Emergency Neurological Deficits (EXTEND)


  • Conflict of interest: None declared.

Correspondence: Geoffrey A. Donnan, National Stroke Research Institute, Austin Health, University of Melbourne, 245 Burgundy Street, Heidelberg, Vic 3084, Australia.

E-mail: gdonnan@unimelb.edu.au


Background and hypothesis

Thrombolytic therapy with tissue plasminogen activator is effective for acute ischaemic stroke within 4·5 h of onset. Patients who wake up with stroke are generally ineligible for stroke thrombolysis. We hypothesized that ischaemic stroke patients with significant penumbral mismatch on either magnetic resonance imaging or computer tomography at three- (or 4·5 depending on local guidelines) to nine-hours from stroke onset, or patients with wake-up stroke within nine-hours from midpoint of sleep duration, would have improved clinical outcomes when given tissue plasminogen activator compared to placebo.

Study design

EXtending the time for Thrombolysis in Emergency Neurological Deficits is an investigator-driven, Phase III, randomized, multicentre, double-blind, placebo-controlled study. Ischaemic stroke patients presenting after the three- or 4·5-h treatment window for tissue plasminogen activator and within nine-hours of stroke onset or with wake-up stroke within nine-hours from the midpoint of sleep duration, who fulfil clinical (National Institutes of Health Stroke Score ≥4–26 and prestroke modified Rankin Scale <2) will undergo magnetic resonance imaging or computer tomography. Patients who also meet imaging criteria (infarct core volume <70 ml, perfusion lesion : infarct core mismatch ratio >1·2, and absolute mismatch >10 ml) will be randomized to either tissue plasminogen activator or placebo.

Study outcome

The primary outcome measure will be modified Rankin Scale 0–1 at day 90. Clinical secondary outcomes include categorical shift in modified Rankin Scale at 90 days, reduction in the National Institutes of Health Stroke Score by 8 or more points or reaching 0–1 at day 90, recurrent stroke, or death. Imaging secondary outcomes will include symptomatic intracranial haemorrhage, reperfusion and or recanalization at 24 h and infarct growth at day 90.


In acute ischaemic stroke, intravenous thrombolysis (tissue plasminogen activator (tPA)) has been shown to result in improved outcome when given up to 4·5 h from symptom onset [1]. In spite of the clear benefit of tPA, approximately 5% of patients receive therapy, and this small proportion is mainly related to the short therapeutic time window [2]. Potentially increasing the number of patients eligible to receive tPA is to extend the time window for therapy. This approach is biologically plausible as there is evidence that potentially viable brain tissue in the ischaemic penumbra may exist for up to 48 h after symptom onset of ischaemic stroke, and in some cases [3, 4], magnetic resonance imaging (MRI) may give an indication of penumbral tissue as the mismatch between the hypoperfused perfusion-weighted image (PWI) and the infarct core seen on the diffusion-weighted image (DWI). Such PWI/DWI mismatch tissue may be present in up to half of all ischaemic stroke patients at 24 h [3].

Recent evidence confirms that the time window for thrombolysis may be extended beyond the 4·5-h limit [5]. When designing studies to confirm such an extension, there are two potential enrolment strategies that may be used. First, using an indiscriminant approach where all potential patients are included; however, this entails disadvantage of a large sample size requirement to compensate for the inclusion of patients who do not respond to therapy. The second approach is to enrol an enriched population of patients with greater likelihood of therapeutic responsiveness [6], using MRI, and in particular PWI and DWI to identify potentially penumbral tissue.

The relationship between PWI/DWI mismatch, thrombolytic therapy, and surrogate MR-based outcomes has been explored in a series of Phase II clinical trials. In the open–label, nonrandomized Diffusion and Perfusion Imaging Evaluation for Understanding Stroke Evolution (DEFUSE) study, patients presenting within a three–six-hour treatment window received tPA and had PWI, DWI, and MR angiography (MRA) performed at study entry and again within 24 h [7]. A strong relationship between recanalization, reperfusion, and attenuation of infarct growth was found. A sub-set of patients with ‘malignant mismatch’ where the DWI volume was greater than 100 ml and who were at high risk of symptomatic intracerebral haemorrhage was identified [8].

The Echoplanar Imaging Thrombolytic Evaluation Trial (EPITHET) was a double-blinded, randomized, placebo-controlled Phase II trial in ischaemic stroke patients presenting three–six-hours after symptom onset and were randomized to tPA or placebo after MRI scans had been obtained and which were then repeated at days 3–5 [9]. EPITHET found a relationship between tPA and reperfusion and attenuation of infarct growth on the serial PWI and DWI studies [10]. Furthermore, there was a 15% absolute improvement in clinical outcome measures (modified Rankin Scale (mRS) 0–1) in patients with penumbral tissue as demonstrated by the presence of PWI/DWI mismatch. When a further analysis was undertaken using co-registration techniques to more accurately determine mismatch volumes at study entry, the primary outcome measure of infarct growth was significantly attenuated [10]. The DEFUSE and EPITHET studies therefore provide strong biological support for the extension of the therapeutic window for tPA beyond the current 3–4·5 h in patients with PWI/DWI mismatch and also a means to limit study populations to those with the potential to benefit from putative reperfusion therapies.

Since the completion of DEFUSE and EPITHET, rapid and precise automated systems for the assessment of PWI/DWI mismatch has become available for research purpose (RAPID). Optimal perfusion measures and a clear definition of what constitutes PWI/DWI mismatch have been defined. A Tmax delay of plus six-seconds has been shown to be the PWI parameter that most accurately predicts final infarct volume, and a perfusion volume to infarct core ratio of 1·2 has been shown to predict response to tPA [11-13]. DWI volumes of more than 70 ml confer poor clinical outcome regardless of whether a patient receives tPA or not [14, 15].

Computer Tomography (CT) perfusion has recently been validated as an alternative modality to determine penumbral mismatch. The infarct core indicated by diffusion MRI is best represented by thresholded CT cerebral blood flow (CBF) [16-18]. The extent of salvageable tissue defined by MR-Tmax >6 s directly translates to CT-Tmax >6 s and has good agreement with perfusion–diffusion mismatch using the mismatch ratio >1·2, absolute mismatch volume >10 ml, and infarct core volume <70 ml criteria (Campbell et al., unpublished data). CT has practical advantages in 24-h accessibility and fewer practical restrictions compared with MRI.

Current guidelines limit tPA therapy to patients who can receive therapy 3–4·5 h after a known symptom onset. This time-linked treatment window excludes the majority of patients who wake up with stroke symptoms, the so-called ‘wake-up strokes’ (WUS). Recent clinical and imaging studies have suggested that a substantial number of patients with WUS have a stroke onset that is close to the time of waking so that many of these patients may still have potentially salvageable penumbral tissue [19, 20]. Penumbral mismatch is a logical method to identify WUS patients who may benefit from thrombolytic therapy with an acceptable level of risk.

On the basis of this information, we have designed the EXTEND, an international, Phase III, randomized, multicentre, double-blinded, placebo-controlled study. EXTEND will be an international collaborative study involving centres in Australia and countries around the world (EXTEND International).

Study objectives

The aim of this study was to test the hypothesis that ischaemic stroke patients of similar age and stroke severity with penumbral mismatch between 3 and 4·5 h (depending on local guidelines) and nine-hours after stroke onset will have better outcomes with intravenous tPA than patients receiving placebo.


Study design (Fig. 1)

Figure 1.

EXtending the time for Thrombolysis in Emergency Neurological Deficits study assessment flow chart. BI, Barthel Index; ICH, Intracranial haemorrhage; MRI, magnetic resonance imaging.

The EXTEND study will be a randomized, multicentre, double-blinded, placebo-controlled phase III trial (two arms with 1 : 1 randomization) in ischaemic stroke patients. Patients randomized to treatment will be stratified for geographical region and for time of randomization after stroke into one of the following strata:

  • Greater than three-hours (or 4·5 h depending on local practice) and up to six-hours
  • Greater than six-hours and up to nine-hours, and
  • WUS, where the time of stroke onset will be defined as the midpoint between the onset of sleep or when the patient was last known to be normal and the time of waking.

Patient population

Inclusion criteria

  • Patients presenting with acute hemispheric ischaemic stroke
  • Patient, family member, or legally responsible person, depending on local ethics requirements, has given informed consent
  • Age ≥18 years
  • Treatment onset can commence after three-hours and up to and including nine-hours after stroke onset according to registered product information, or greater than 4·5 h up to and including nine-hours according to locally accepted guidelines
  • WUS patients are defined as having no symptoms at sleep onset, but stroke symptoms on waking; time of stroke onset is to be taken as the midpoint between sleep onset (or last known to be normal) and time of waking. The maximum time window for randomization is then nine-hours from the midpoint as described, and
  • National Institutes of Health Stroke Score (NIHSS) of ≥4–26 with clinical signs of hemispheric infarction.
Imaging inclusion criteria
  • Penumbral mismatch – a ‘penumbra to core’ lesion volume ratio of greater than 1·2 and an absolute difference greater than 10 ml (Using a MR or CT Tmax >6-s delay perfusion lesion and MR-DWI or CT-CBF core lesion), and
  • An infarct core lesion volume of less than or equal to 70 ml using MR-DWI or CT-CBF

Exclusion criteria

  • Intracranial haemorrhage (ICH) identified by CT or MRI
  • Rapidly improving symptoms, particularly if in the judgment of the managing clinician the improvement is likely to result in the patient having an NIHSS of <4 at randomization
  • Prestroke mRS score of ≥2 (indicating previous disability)
  • Contraindication to the use of imaging contrast agents
  • Infarct core >1/3 middle cerebral artery (MCA) territory qualitatively on CT
  • Participation in any investigational study in the previous 30 days
  • Any terminal illness such that the patient would not be expected to survive more than one-year
  • Any condition that could impose hazards to the patient or effect the participation of the patient in the study (this applies to patients with severe microangiopathy such as haemolytic uremic syndrome or thrombotic thrombocytopenic purpura) (at the discretion of the investigator)
  • Pregnancy
  • Stroke within the last three-months; recent past history or clinical presentation of ICH, sub-arachnoid haemorrhage, arteriovenous malformation, aneurysm, or cerebral neoplasm (at the discretion of each investigator)
  • Current use of oral anticoagulants and a prolonged prothrombin time (International Normalised Ratio (INR) >1·6)
  • Use of heparin, except for low-dose sub-cutaneous heparin, in the previous 48 h and a prolonged activated partial thromboplastin time exceeding the upper limit of the local laboratory normal range
  • Use of glycoprotein IIb–IIIa inhibitors within the past 72 h; use of single or dual agent oral platelet inhibitors (clopidogrel and/or or low-dose aspirin) prior to study entry is permitted
  • Clinically significant hypoglycaemia
  • Uncontrolled hypertension defined by a blood pressure >185 mmHg systolic or >110 mmHg diastolic on at least two separate occasions at least 10 mins apart, or requiring aggressive treatment to reduce the blood pressure to within these limits, the definition of ‘aggressive treatment’ (at the discretion of the investigator)
  • Hereditary or acquired haemorrhagic diathesis
  • Gastrointestinal or urinary bleeding within the preceding 21 days
  • Major surgery within the preceding 14 days which poses risk in the opinion of the investigator, and
  • Exposure to a thrombolytic agent within the previous 72 h.

Clinical assessment

A health care professional trained in administration and blinded to the treatment designation will measure neurological impairment and functional scores. The NIHSS will be performed before randomization, and repeated at 12–24 h, three–days, and 90 days after treatment. At day 90, clinical assessments will also include the mRS and the Barthel Index.

Imaging assessment and parameters

Patients will have standardized MRI or CT sequences performed prior to randomization, and, where possible, a second MRI scan will be performed using the same parameters 12–24 h later. MRI sequences include, DWI, PWI, MRA, T2* Gradient Echo, and Fluid Attenuated Inversion Recovery (FLAIR) at both time points. Baseline CT includes noncontrast CT, CT perfusion, and CT angiography. A mismatch between the volume of the perfusion lesion and the irreversibly damaged infarct core is used for penumbral estimation. The perfusion lesion is defined as Tmax >6 s for MRI and CT. Infarct core is defined using MRI diffusion imaging or CT-CBF imaging. Mismatch is defined as a perfusion : core lesion volume ratio of greater than 1·2. Absolute mismatch volume will be greater than 10 ml. Maximum infarct core volume will be 70 ml.

The MRI protocol will follow current international consensus guidelines [21], although the follow-up imaging will be performed at 12–24 h as recently validated to accurately reflect final infarct volume [22]. An initial scout view will be followed by isotropic DWI (created from DWI images obtained with diffusion-sensitizing gradients applied in three orthogonal planes) using b-values between 0 s/mm2, equivalent to a T2-weighted image, and 1000 s/mm2. Whole brain imaging will use 25 contiguous axial slices each 5 mm in thickness. Perfusion images will be derived from the concentration–time curve obtained after the administration of intravenous gadolinium (0·1 mmol/kg) given over three–five-seconds with gradient echo images acquired every 1·8 s for at least 100 s (15 axial slices at each time point). The imaging times for PWI and DWI are each approximately three-minutes. Time-of-flight MRA will be obtained to determine the presence or absence of MCA stem or major branch occlusion. A gradient echo (T2*) sequence will be performed to assess for the presence of ICH. A FLAIR sequence is also acquired to identify lesion expansion to the sub-acute imaging time point.

The imaging data will be processed by an automated program RAPID (supplied by Stanford University, United States), and the results will be displayed on the predetermined workstation or display unit of the recruiting centres. The processed images and results will be reviewed by the responsible clinician of the recruiting centre before randomization. All imaging data will be electronically transferred and read at the coordinating centre on a weekly basis, to quality assure the MRI-based decision making at the recruiting centres. The coordinating centre will be blinded to the clinical information and treatment designation until after the radiological outcome measures have been determined.

The presence and degree of reperfusion will be determined as the difference between the 24-h and acute PWI lesion volumes (percentage and cm3 change). Recanalization will be determined based on initial and 24 h MRA, and classified according to the Thrombolysis in Myocardial Infarction (TIMI) system. The presence of symptomatic hemorrhagic transformation will be assessed on follow-up MR or CT according to Safe Implementation of Thrombolysis in Stroke-Monitoring Study (SIT-MOST) criteria (i.e. parenchymal haematoma type 2 (PH2) occurring within three–six-hours of treatment, combined with neurological deterioration leading to an increase of four points on the NIHSS from baseline, or the lowest NIHSS value after baseline to 24 h) [23].

Investigational product

The investigational product (tPA or placebo, blinded) is supplied as 50 mg lyophilized powder in glass vials. The dose of tPA to be administered is 0·6 mg/kg or 0·9 mg/kg (maximum 60 mg or 90 mg according to local legally approved practice and guidelines) given as 10% of total bolus dose over one-minute, then the remaining 90% as an infusion over 60 mins. The lower dose of 0·6 mg/kg is the legally approved dose based on its efficacy and safety profile in the Japanese population.


All those involved in the conduct of the study will be blinded to treatment allocation. The Data Safety Monitoring Committee (DSMC) will have access to data that are grouped, but not unblinded in terms of treatment allocation.

Sample size

Approximately 400 participants randomized to receive either intravenous tPA or placebo in a 1 : 1 ratio will be enrolled in the EXTEND trial. This sample size is to be achieved through recruitment both in Australia and internationally (EXTEND (International)). Based on 0·8 power to detect a significant difference of 15% in the proportion of patients with mRS 0–1 outcome (3% in tPA arm, 21% in placebo arm [9]), P = 0·05, two sided, and to compensate for nonevaluable patients, a total sample size of up to 400 patients would be required with about 200 in each of treatment and placebo arms.

Randomization and statistical analysis

Patients will be randomized to receive the blinded investigational product according to a centralized procedure coordinated via the online EXTEND electronic case report form. The randomization system for investigational product will be based on computer-generated randomization code lists, with stratification for geographical region and, further, for time from onset of stroke to randomization.

All randomized subjects will be included in randomised controlled trial (RCT) analyses on an intention-to-treat basis. Missing outcome data will be handled through multiple imputation procedures subject to the validity of missing-at-randomness assumptions. For the primary outcome analysis, the proportions of mRS 0–1 outcomes will be compared between treatment and placebo arms of the RCT adjusted for geographical region and time between stroke onset and treatment (greater than three-hours (or 4·5 h depending on local practice) up to six-hours; greater than six-hours and up to nine-hours; WUS as defined earlier, age, and baseline NIHSS, using binary logistic regression model). Although both adjusted and unadjusted results will be reported, adjusted analysis is prespecified as the primary outcome analysis for this RCT.

A secondary analysis of the categorical shift in mRS will be undertaken on the full range (0–6) of the mRS using Cochran–Mantel–Haenszel shift test and proportional odds logistic regression subject to the validity of shift analysis model assumptions. Other secondary outcome analyses will be carried out according to standard statistical principles for comparison of parametric or nonparametric distributions as appropriate [24].

Primary outcome

  • modified Rankin Scale (mRS) 0–1 at day 90.

Secondary outcomes

  • categorical shift in mRS at day 90
  • reduction in the NIHSS by eight or more points or an NIHSS of 0 or 1.
  • death due to any cause
  • symptomatic ICH
  • reperfusion at 24 h after onset
  • recanalization at 24 h after onset
  • infarct growth on DWI within 24 h
  • recurrent stroke by day 90

Symptomatic intracerebral haemorrhage (SICH) is defined as ‘Intracerebral haemorrhage (PH2) occurring within six-hours of treatment, combined with neurological deterioration leading to an increase of 4 points on the NIHSS from baseline, or the lowest NIHSS value after baseline to 24 hours’ [23] and adjudicated by a centralized committee blinded to treatment allocation.

Data safety monitoring

Safety interim analyses will be undertaken when 100, 200, and 300 patients have completed the three-month assessments. It will be conducted by an independent DSMC. The Haybittle–Peto procedure for generating early stopping boundaries will be used. To compare the safety of the tPA therapy and placebo, noninferiority hypotheses tests for the two primary safety parameters – mortality at three-months and the incidence of symptomatic ICH within 72 h of intervention – are conducted independently. The margin of noninferiority (δ) chosen is tentatively set at absolute 5% for both safety parameters. A recommendation of early termination due to experimental treatment inferiority on deaths or symptomatic haemorrhages within 72 h from intervention onset will be considered by the DSMC if the corresponding Haybittle–Peto boundary (P = 0·003, Z = 3) at a given interim analysis is crossed. If there are concerns about the safety of participants, the DSMC will make a recommendation to the trial Steering Committee about continuing, stopping, or modifying the trial. No formal interim analyses for efficacy or futility are planned.

Study organization and funding

Both the steering committee and executive sub-committees manage this study. The study is financially supported by Commonwealth Scientific and Industrial Research Organisation, Florey Neuroscience Institutes (FNI), and National Health and Medical Research Council of Australia. The National Stroke Research Institute, a member of the FNI, is the study sponsor.


EXTEND is the first, randomized, multicentre, double-blinded, placebo-controlled Phase III trial assessing the efficacy of thrombolytic therapy in an enriched population of acute stroke patients based on the presence of clinically significant ischaemic penumbra (MRI or CT mismatch) between three- to nine-hours after stroke onset. A novel feature of EXTEND is the inclusion of patients who wake up with stroke. The allowance of different tPA dosages (0·6 or 0·9 mg/kg) reflects the practical nature and worldwide generalizability of this study. If the study outcome is positive, it will significantly widen the therapeutic time window of acute stroke treatment and increase the number of patients who can benefit. It is an important step towards reducing the burden of stroke worldwide.


EXTEND is supported in part by Commonwealth Scientific and Industrial Research Organisation (CSIRO) of Australia through the CSIRO Preventative Health Flagship Cluster as part of the START-EXTEND Collaborative Study with funding from the CSIRO National Research Flagship Collaboration Fund.

The FNI acknowledges the strong support from the Victorian Government and, in particular, the funding from the Operational Infrastructure Support Grant. In addition, we like to acknowledge the help in the preparation of this manuscript from the EXTEND study manager Ms Sue Bates and study coordinator Ms Elise Cowley.