EXTEND Trial: Towards a More Inclusive But Complex Thrombolysis

 

Written By: Zach Ravnitzky, DO; Edited By: Brian Smith, DO, MA, MMSc-Med

 

What is the study question?

Does IV alteplase improve functional outcomes in acute ischemic stroke patients presenting 4.5–9.0 hours after onset when selected by perfusion imaging showing salvageable brain tissue?

What study design is used to answer the question?

Multicenter, randomized, double-blinded, placebo-controlled

What is the exposure or intervention?

IV alteplase 0.9 mg/kg (max 90 mg) (10% bolus followed by 90% infusion over 1 hour)  vs placebo.

What is the outcome (or disease) and how is it measured (or defined)?

Primary outcome: modified rankin score 0–1 at 90 days, adjusted for age and baseline NIHSS. (secondary outcomes: modified rankin score 0–2, and early neurological improvement (NIHSS change at 24 hours), symptomatic ICH (parenchymal hematoma with ≥4-point NIHSS increase within 36 hours) and 90-day mortality.)

How were participants selected or recruited for the study?

Patients presenting to participating centers in Australia, New Zealand, Taiwan, and Finland underwent automated CT or MR perfusion imaging (RAPID software). Only those meeting predefined perfusion mismatch criteria were eligible for randomization.

Who was included in the study?

Patients with acute ischemic stroke 4.5–9.0 hours from onset (or wake-up stroke), NIHSS ≥ 4, pre-stroke mRS 0–1, and perfusion mismatch on RAPID imaging: ischemic core < 70 mL, mismatch volume > 10 mL, and mismatch ratio > 1.2 (Tmax > 6 seconds defining critically hypoperfused tissue).

Who was excluded from the study?

Standard thrombolysis exclusions: rapidly improving deficits, high bleeding risk, uncontrolled hypertension (>185/110), large infarct core (≥70 mL), no perfusion mismatch, pre-stroke mRS > 1, and standard alteplase contraindications.

If randomization occurred, at what point did it occur?

After perfusion imaging confirmed the patient met mismatch criteria. Patients were randomized 1:1 to alteplase or placebo.

What do the authors say about power/sample size? Is the power adequate?

Planned enrollment was 310 patients, but only 225 were enrolled (73%) before early termination due to the WAKE-UP trial showing benefit to thrombolytics, making it unethical to continue the study by giving patients placebo (standard medical management.) Power was inadequate — the primary outcome was only marginally significant (p = 0.04) and the ordinal mRS analysis was not significant.

Are the study groups' characteristics comparable at baseline?

Yes, baseline demographics, stroke severity, and imaging characteristics were generally well balanced. 65% had wake-up strokes and 70% had LVO’s, but these proportions were similar between groups.

What statistical tests/methods are used? Is ITT maintained?

To analyze the main result, the researchers used a type of regression model (modified Poisson regression) that accounts for differences in age and stroke severity between the two groups. For the secondary outcomes, they used other standard statistical methods (ordinal logistic regression) to look at shifts across the entire disability scale. ITT was maintained: all 225 randomized patients were analyzed per assigned group. The statistical analysis plan was published before database lock.

What are the results of the study?

Patients who got the drug were roughly twice as likely to achieve functional independence and three times more likely to show early neurological improvement within the first day. However, the overall shift across the full disability scale was not statistically significant.

On the safety side, symptomatic brain bleeding was higher with alteplase (about 6%) compared to placebo (about 1%), but death rates were similar between the two groups.

(data below copied from study)

  • mRS 0–1: 35.4% vs. 29.5% (adj. RR 1.44, 95% CI 1.01–2.06; p = 0.04)

  • Ordinal mRS shift: OR 1.55, 95% CI 0.96–2.49; p = 0.382

  • mRS 0–2: OR 2.02; p = 0.031

  • Early neurological improvement: OR 2.76; p = 0.006

  • sICH: 6.2% vs. 0.9% (adj. RR 7.22; p = 0.05)

  • Death: 11.5% vs. 8.9% (not significant)

Are the results statistically significant?

The primary outcome was marginally significant (p = 0.04, CI lower bound 1.01). Functional independence and early neurological improvement were significant. The ordinal mRS shift was not significant. sICH was borderline (p = 0.05). These marginal results raise concerns about robustness given early termination.

Are the results clinically significant?

The absolute benefit was 5.9 percentage points (NNT ~17), comparable to conventional-window thrombolysis. However, the sICH rate yields an NNH of ~19, meaning the benefit-to-risk ratio is narrow. Mortality did not differ, suggesting sICH did not translate to excess deaths, though the trial was underpowered for this.

What are the biases/limitations?

  • Early termination (73% enrollment) which may overestimate treatment effect

  • Small sample size which underpowered the study for secondary and safety outcomes

  • ~70% had LVO (most would now receive EVT, limiting current applicability)

  • Imaging selection bias — requires RAPID software, limiting generalizability

  • Software evolution over 8-year recruitment — potential variability in patient selection

  • Limited geographic diversity (4 countries)

Which limitations are unique to the study design, and how are they addressed?

Early termination bias: Stopped for loss of equipoise (ethical), not efficacy; statistical analysis plan published before unblinding. 

What is the overall internal validity.

Moderate. Bias and confounding were well addressed through randomization, double-blinding, ITT analysis, and adjustment for key prognostic variables. However, chance remains a plausible partial explanation — the primary outcome was marginally significant (CI lower bound 1.01), the ordinal analysis was negative, and the trial was underpowered due to early termination. The pooled meta-analysis with ECASS4-EXTEND and EPITHET (n = 414, OR 1.86, p = 0.011) provides stronger evidence against chance.

What is the external validity. Would these results change your practice?

External validity is limited: the trial required specialized perfusion imaging not universally available, ~70% of patients had LVO (now treated with EVT), and recruitment was from only four countries. EXTEND alone would warrant cautious adoption, but combined with subsequent confirmatory trials (HOPE, TRACE-III) and incorporation into the 2021 ESO and 2026 AHA/ASA guidelines, the evidence now supports extended-window thrombolysis for imaging-selected patients.

How does this fit our current knowledge?

EXTEND was pivotal in shifting stroke treatment from a time-based to a tissue-based paradigm. It demonstrated that perfusion imaging can identify patients who benefit from thrombolysis beyond 4.5 hours. Its findings have been confirmed by TRACE-III (2023), HOPE (2025), and OPTION (2026), and are now incorporated into major international guidelines. The TIMELESS trial (2024) added nuance by showing limited added benefit of thrombolysis when rapid EVT is available. A pooled meta-analysis of 8 extended-window RCTs (1,742 patients) confirmed improved outcomes with extended-window IVT despite increased sICH, with no mortality increase. EXTEND was underpowered; its findings have been validated by subsequent evidence.

 
Booth EM