← Clinical Trials
Stroke Phase 3 Completed NCT00887328

EXTEND: Complete Statistical Analysis of Alteplase in Stroke

An independent statistical review of the randomized phase 3 EXTEND trial evaluating tissue plasminogen activator (alteplase) versus placebo in patients with stroke, with the primary endpoint defined as modified Rankin Scale (mRS) 0–1 at 3 months.

EXTEND  ·  Phase 3  ·  180 participants  ·  Randomized, parallel-group, triple-masked
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

EXTEND was a randomized phase 3, parallel-group, triple-masked treatment trial evaluating tissue plasminogen activator (alteplase) versus placebo in patients with stroke. The registry lists 180 participants and a primary endpoint of modified Rankin Scale (mRS) 0–1 at 3 months.

180
Enrollment
Randomized trial
2
Arms
Alteplase vs placebo
3
Phase
Phase 3
3 mo
Primary endpoint
mRS 0–1
FeatureEXTEND
Trial acronymEXTEND
Full titleExtending the Time for Thrombolysis in Emergency Neurological Deficits
PhasePhase 3
ConditionStroke
StatusCompleted
AllocationRandomized
Design modelParallel
MaskingTriple
Primary purposeTreatment
Enrollment180
Number of arms2
InterventionsTissue Plasminogen Activator (Alteplase) and Placebo
Lead sponsorNeuroscience Trials Australia
Sponsor typeOther
Start date2010-06
Primary completion date2018-08-27
ClinicalTrials.gov identifierNCT00887328

2. Clinical Question

The central statistical question is whether treatment assignment to tissue plasminogen activator (alteplase), rather than placebo, is associated with a difference in the probability of achieving the registry-defined favorable functional outcome of modified Rankin Scale (mRS) 0–1 at 3 months.

Population

Participants enrolled in the phase 3 EXTEND trial for the condition of stroke.

Intervention

Tissue Plasminogen Activator (Alteplase).

Comparator

Placebo.

Primary question

Does randomized assignment to alteplase versus placebo affect the probability of mRS 0–1 at 3 months?

The randomized design is important because the comparison is defined by treatment assignment rather than by an observational comparison of people who happened to receive different treatments. With 2 arms and 180 enrolled participants, the trial was structured to compare the two treatment groups within the same protocol.

3. Trial Design

01
Enroll 180 participants
02
Randomize 2 treatment arms
03
Mask Triple-masked design
04
Treat Alteplase or placebo
05
Assess mRS 0–1 at 3 months
ARM 1

Alteplase

  • Tissue Plasminogen Activator (Alteplase)
  • Randomized treatment assignment
  • Triple-masked trial structure
ARM 2

Placebo

  • Placebo
  • Randomized comparator assignment
  • Triple-masked trial structure

The parallel-group structure means participants were assigned to one of the two treatment arms rather than being sequentially exposed to both interventions. That is statistically important for a treatment-effect question because the randomized groups can be compared at the prespecified assessment point without requiring within-person crossover comparisons.

4. Trial Architecture and Why It Matters Statistically

Allocation
Randomized. Randomization is the core mechanism for balancing measured and unmeasured prognostic factors in expectation across treatment groups.
Design model
Parallel. Each participant contributes to one randomized treatment group for the comparison.
Masking
Triple-masked. The registry identifies the trial as triple-masked, reducing opportunities for knowledge of treatment assignment to influence trial conduct or assessment.
Purpose
Treatment. The trial's primary purpose is recorded as evaluating treatment rather than diagnosis, prevention, screening, or another study purpose.

The combination of randomization, a parallel-group design, and triple masking establishes a framework in which the primary comparison can be interpreted as a randomized treatment comparison. The statistical analysis must nevertheless preserve the randomized groups and account appropriately for the binary nature of the primary endpoint.

5. Primary Endpoint

EndpointRegistry definitionTime frame
Modified Rankin Scale (mRS) 0-1 Modified Rankin Scale (mRS) 0-1 3 months

The primary endpoint is a binary functional-outcome endpoint: participants are classified according to whether their modified Rankin Scale outcome is 0–1 at the specified 3-month assessment. This creates a natural two-group comparison of the proportion achieving the favorable endpoint.

Binary endpoint framework
pA = P(mRS 0–1 at 3 months | alteplase)
pP = P(mRS 0–1 at 3 months | placebo)

The treatment comparison can then be expressed using an absolute risk difference, a risk ratio, an odds ratio, or another prespecified binary-outcome measure. The registry record identifies the endpoint but does not report a posted statistical analysis specifying which effect measure was used.

6. Statistical Methodology

The ClinicalTrials.gov record identifies the primary endpoint and its 3-month time frame, but no formal statistical analyses are posted for this endpoint. For a randomized two-arm trial with a binary primary endpoint such as mRS 0–1, the principal analysis would ordinarily compare the proportion of participants achieving mRS 0–1 between the randomized treatment groups.

Binary treatment-effect estimation

A binary endpoint can be summarized in several complementary ways. The risk difference measures the difference in the observed probability of the favorable outcome between groups. The risk ratio expresses the probability in the alteplase group relative to the probability in the placebo group. The odds ratio compares the corresponding odds.

Common effect measures
Risk Difference = pA − pP
Risk Ratio = pA / pP
Odds Ratio = [pA/(1−pA)] / [pP/(1−pP)]

These measures answer related but different questions. A risk difference is expressed in percentage points, whereas a risk ratio and odds ratio are relative measures.

Hypothesis testing

For a conventional two-arm binary endpoint, a null hypothesis can be formulated around equality of the treatment-group probabilities. A suitable test may be a chi-square test, Fisher's exact test when appropriate, or a model-based test depending on the prespecified statistical analysis plan and observed data structure.

Confidence intervals

A confidence interval should accompany the primary treatment-effect estimate. It provides information about the precision of the estimated treatment effect and is more informative than a p-value alone because it displays a range of effect sizes compatible with the specified statistical framework.

Regression adjustment

If baseline covariates or prespecified stratification factors were incorporated into the analysis, a regression model such as logistic regression could be used to estimate an adjusted treatment effect. The ClinicalTrials.gov record does not report such covariates or a regression specification, so no particular adjusted model is attributed to EXTEND.

Intention-to-treat principle

For a randomized efficacy analysis, the usual primary framework is intention-to-treat: participants are analyzed according to the treatment group to which they were randomized. This preserves the comparison generated by randomization and avoids redefining treatment groups according to subsequent treatment exposure.

7. Planned Analysis

The primary endpoint is Modified Rankin Scale (mRS) 0-1 at 3 months. Because this endpoint records whether a participant reaches a specified favorable functional outcome, the primary statistical comparison would ordinarily be a two-group analysis of binary proportions.

Primary estimand

The treatment effect concerns the difference in the probability of achieving mRS 0–1 at 3 months under randomized assignment to alteplase versus placebo.

Primary outcome type

Binary. Participants are classified according to whether the mRS outcome falls within the prespecified 0–1 category at 3 months.

Typical effect measures

Risk difference, risk ratio, and odds ratio can each describe the treatment contrast. The appropriate prespecified measure depends on the trial's statistical analysis plan.

Uncertainty

A confidence interval should accompany the effect estimate to show the statistical precision of the treatment comparison.

Registry reporting status: no formal statistical analyses are posted to ClinicalTrials.gov for the registered primary endpoint. The registry therefore identifies what was measured and when it was measured, but does not provide a posted numerical treatment-effect estimate, confidence interval, or p-value for the primary endpoint.

The absence of a posted analysis is distinct from the absence of a prespecified endpoint. EXTEND has a clearly identified primary endpoint and time frame. What is not reported in the registry record is the numerical result and formal inferential analysis for that endpoint.

8. Statistical Methods Explained

What makes mRS 0–1 a binary endpoint?

The modified Rankin Scale is converted here into a prespecified favorable-outcome category: mRS 0–1. For the primary analysis, the relevant classification is therefore whether each participant does or does not meet that threshold at 3 months. This differs from analyzing the full ordinal scale, because an ordinal analysis retains more of the information contained in the individual mRS categories.

Why compare proportions between the two randomized groups?

The primary endpoint defines a favorable outcome at a fixed follow-up time. The most direct question is therefore whether the proportion with mRS 0–1 differs between participants randomized to alteplase and those randomized to placebo.

What does a risk difference tell us?

A risk difference is expressed in percentage points. For example, if the favorable-outcome probability were 40% in one group and 30% in another, the risk difference would be 10 percentage points. This is an absolute measure and can be easier to interpret clinically than a relative measure alone.

Why might an odds ratio differ from a risk ratio?

Odds and probabilities are related but are not interchangeable. The odds of an event are the probability divided by one minus the probability. As the event becomes more common, the odds ratio can depart substantially from the risk ratio. For that reason, the effect measure should always be identified explicitly.

Why is randomization important for the analysis?

Randomization establishes the treatment groups before outcomes are observed and, in expectation, balances prognostic factors across groups. This allows the primary comparison to be interpreted as a randomized treatment contrast rather than simply an association between treatment received and outcome.

Why is the 3-month time frame part of the endpoint?

The endpoint is not simply "mRS 0–1." It is mRS 0–1 at 3 months. The timing defines when the functional outcome is assessed and prevents different follow-up times from being treated as though they represented the same endpoint.

9. Interpreting the Primary Endpoint Without a Posted Result

Statistical interpretation

A formal result for the primary endpoint would normally consist of an estimated treatment effect comparing the probability of mRS 0–1 at 3 months between the randomized groups, together with a confidence interval and a hypothesis-test result. None of those numerical results are posted in the ClinicalTrials.gov record.

What an effect estimate would mean

If expressed as a risk difference, the estimate would describe the absolute difference in the probability of achieving mRS 0–1 between the alteplase and placebo groups. If expressed as a risk ratio, it would describe the relative probability of the favorable outcome. An odds ratio would describe the relative odds rather than the relative probability.

Why the confidence interval matters

A confidence interval would describe statistical uncertainty around the estimated treatment effect under the chosen analysis framework. A narrow interval would indicate greater precision than a wide interval, although precision is not the same as clinical importance.

Why a p-value is not an effect size

A p-value addresses compatibility with a specified null hypothesis under the statistical model. It does not state how large the treatment effect is. Effect magnitude should therefore be communicated with an appropriate effect estimate and its confidence interval.

10. Trial Timeline

2010-06

Trial start

The EXTEND trial is recorded as beginning in June 2010.

Phase 3

Randomized treatment evaluation

The trial is registered as a phase 3 randomized, parallel-group, triple-masked treatment study with 2 arms.

3 months

Primary endpoint assessment

The registered primary endpoint is Modified Rankin Scale (mRS) 0-1 at 3 months.

2018-08-27

Primary completion

The registry records a primary completion date of August 27, 2018.

11. Analysis Population and Randomization

The registry records 180 participants and 2 treatment arms but does not provide arm-specific enrollment counts in the information reported here. Therefore, no numerical allocation between the alteplase and placebo groups is presented.

Design elementRegistry informationStatistical significance
Randomization Randomized Supports comparison by treatment assignment.
Parallel design Parallel Participants contribute to one randomized treatment group.
Masking Triple Reduces opportunities for knowledge of assignment to influence trial conduct or outcome assessment.
Total enrollment 180 Defines the overall enrolled population available for the randomized comparison.
Number of arms 2 Creates a two-group treatment comparison.

For the primary analysis, maintaining the randomized comparison is especially important. Reclassifying participants according to treatment actually received can introduce post-randomization selection and weaken the protection provided by randomization.

12. Missing Data and Outcome Assessment

The primary endpoint is assessed at a fixed 3-month time point. In a complete analysis, participants without an evaluable mRS assessment at that time would require a prespecified approach to missing outcome data.

Possible approaches in randomized trials include complete-case analysis, conservative assumptions, multiple imputation, or model-based methods. Each makes different assumptions about the mechanism and implications of missingness. The registry record does not report a missing-data or imputation strategy for EXTEND.

Why this matters: a binary functional endpoint can be sensitive to how participants with missing 3-month assessments are handled. A complete statistical analysis therefore needs both the observed outcome data and the prespecified rules governing missing observations.

13. Multiplicity and Interim Analysis

The ClinicalTrials.gov information reported here identifies one registered primary endpoint: Modified Rankin Scale (mRS) 0-1 at 3 months. It does not provide a posted multiplicity strategy, alpha-spending method, interim-analysis schedule, or other formal error-control procedure.

Single registered primary endpoint

The registry identifies mRS 0–1 at 3 months as the primary endpoint, which gives the trial a clearly defined primary outcome in the registry record.

Interim monitoring

No interim-analysis method is reported in the registry information presented here.

Multiplicity

No formal multiplicity-adjustment procedure is reported in the registry information presented here.

Type I error

A confirmatory interpretation would ordinarily require the prespecified significance level and any relevant multiplicity or interim-monitoring rules.

These distinctions matter because the nominal p-value from a statistical test cannot be interpreted independently of the analysis plan when multiple endpoints, multiple looks at the data, or multiple treatment comparisons are involved.

14. Crossover, Factorial Design, and Bayesian Methods

The registered design model is parallel, with 2 arms. The registry information does not identify a crossover design, factorial design, non-inferiority margin, or Bayesian analysis.

Design topicRegistry statusStatistical implication
Crossover Not reported No within-participant treatment-period comparison is identified by the registered parallel design.
Factorial design Not reported The trial is described as a 2-arm parallel study rather than a factorial experiment.
Non-inferiority margin Not reported The registry information does not identify a non-inferiority framework or margin.
Bayesian methods Not reported No Bayesian prior, posterior analysis, or Bayesian decision rule is identified.
Interim analysis Not reported No interim-monitoring procedure is identified in the registry information.

15. Safety and Adverse Events

The registry information provided for EXTEND identifies the treatment interventions and primary efficacy endpoint but does not report serious adverse events by treatment arm.

Safety interpretation: no arm-specific serious-adverse-event counts or percentages are presented because the ClinicalTrials.gov record information available here does not report them. Safety conclusions require the corresponding event data, denominators, definitions, and analysis population.

This distinction is important in randomized trials. Safety is not summarized by the primary efficacy endpoint. An efficacy comparison of mRS 0–1 addresses functional outcome, whereas adverse-event analyses address treatment-emergent harms and require their own prespecified definitions and populations.

16. Why This Trial Matters Statistically

EXTEND is a useful teaching example because its primary endpoint illustrates a common clinical-trial problem: converting a multidimensional clinical outcome into a prespecified binary endpoint and then comparing that endpoint between randomized groups.

ConceptHow it appears in EXTEND
Randomization Participants were randomized to 2 treatment arms.
Parallel-group design The trial uses a parallel design rather than a crossover structure.
Triple masking The registry identifies the study as triple-masked.
Binary endpoint The primary endpoint is mRS 0–1 at a fixed 3-month time point.
Absolute treatment effect A risk difference would describe the absolute difference in favorable functional outcome between groups.
Relative treatment effect A risk ratio or odds ratio could describe the relative treatment contrast.
Confidence intervals Uncertainty around the treatment effect should be quantified rather than relying on a p-value alone.
Intention-to-treat Randomized treatment assignment provides the natural basis for the primary efficacy comparison.
Missing data Fixed-time functional assessment makes prespecified handling of missing 3-month outcomes important.
Clinical endpoint thresholding The full mRS is represented by the favorable 0–1 category for the registered primary endpoint.

17. Clinical Endpoint vs Statistical Endpoint

Clinical endpoint

The registered endpoint is modified Rankin Scale (mRS) 0–1 at 3 months, representing a prespecified favorable functional outcome category at a defined follow-up time.

Statistical endpoint

The analysis compares the probability of reaching that category between the randomized alteplase and placebo groups.

Effect-size question

The key quantitative question is how much the probability of mRS 0–1 differs between the two treatment assignments.

Precision question

The confidence interval indicates how precisely the treatment effect has been estimated under the chosen statistical framework.

This distinction prevents an important statistical mistake: treating the clinical endpoint definition itself as though it were the treatment effect. "mRS 0–1 at 3 months" defines what is measured. The treatment effect requires a comparison of the outcome between the randomized groups.

18. Important Limitations and Interpretation Issues

19. A Deeper Statistical Reading of mRS 0–1

The mRS is commonly treated as an ordinal functional outcome, but the registered EXTEND primary endpoint uses a threshold: mRS 0–1. That decision has direct statistical consequences.

Threshold analysis

A threshold analysis reduces the outcome to two categories: favorable, defined as mRS 0–1, and not favorable, defined as values outside that category. This makes the endpoint straightforward to analyze with binary-outcome methods, but it also means that observations elsewhere on the scale are not distinguished within the primary endpoint.

Absolute versus relative effects

Suppose a hypothetical trial produced favorable-outcome probabilities of 45% and 35%. The absolute treatment difference would be 10 percentage points. The corresponding relative probability would be 45% / 35% = 1.29. The odds ratio would be different again. These measures would all describe the same underlying comparison from different perspectives.

Why the denominator matters

Every binary-outcome estimate depends on the number of participants contributing evaluable outcomes. If some participants do not have a 3-month mRS assessment, the treatment-effect estimate can depend materially on the prespecified handling of those observations. This is why the analysis population and missing-data rules belong alongside the headline treatment effect.

20. What a Complete Statistical Report Would Contain

For this type of randomized phase 3 trial, a complete statistical presentation of the primary endpoint would normally include several linked pieces of information.

ComponentPurpose
Number randomized per arm Defines the randomized denominators for the treatment comparison.
Number with mRS 0–1 Shows the observed favorable-outcome counts.
Favorable-outcome percentages Provides the most direct descriptive comparison.
Effect estimate Quantifies the treatment contrast using a specified measure such as risk difference, risk ratio, or odds ratio.
95% confidence interval Quantifies uncertainty around the treatment-effect estimate.
Hypothesis test Assesses the prespecified null hypothesis under the selected statistical framework.
Analysis population Clarifies which randomized participants contributed to the primary analysis.
Missing-data handling Explains how participants without an evaluable 3-month endpoint were treated analytically.

The ClinicalTrials.gov record identifies the first essential element of this framework—the primary endpoint and its timing—but does not post the numerical statistical analysis for that endpoint.

21. Sources

Continue through the Clinical Biostats clinical trial library

Explore additional randomized clinical-trial analyses and statistical methods across the Clinical Biostats library.

22. Record Summary

EXTEND is a completed phase 3 randomized, parallel-group, triple-masked treatment trial involving 180 participants with stroke and 2 treatment arms: tissue plasminogen activator (alteplase) and placebo. Its registered primary endpoint is Modified Rankin Scale (mRS) 0–1 at 3 months.

Statistically, the central issue is a binary randomized treatment comparison: the probability of achieving the prespecified favorable functional outcome at the specified 3-month assessment. A complete analysis would ordinarily report the outcome proportions, an appropriate treatment-effect estimate, its confidence interval, and a hypothesis test, together with the analysis population and missing-data strategy. The ClinicalTrials.gov record does not post those formal numerical analyses, so the statistical interpretation can be described at the design and methodology level without assigning an unreported treatment effect.

Clinical Biostats perspective: The statistical value of a trial record is not limited to its numerical results. The endpoint definition, randomization structure, masking, analysis population, missing-data rules, and choice of effect measure determine how the eventual treatment comparison should be interpreted.