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-IA was a randomized, parallel, unmasked phase 2 trial evaluating tenecteplase versus tissue plasminogen activator before endovascular therapy for ischemic stroke. The registry records 202 enrolled participants and one primary efficacy endpoint based on findings at the initial angiogram on day 0.
| Feature | EXTEND-IA |
|---|---|
| Trial acronym | EXTEND-IA |
| ClinicalTrials.gov identifier | NCT02388061 |
| Phase | Phase 2 |
| Status | COMPLETED |
| Condition | Ischemic Stroke |
| Enrollment | 202 |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Number of arms | 2 |
| Lead sponsor | Neuroscience Trials Australia |
| Sponsor type | OTHER |
2. Clinical Question
The clinical question recorded by the trial is a comparison of tenecteplase with tissue plasminogen activator before endovascular therapy in patients with ischemic stroke. The primary statistical question is whether the randomized treatment groups differ in the proportion of patients meeting the registry-defined angiographic endpoint at the initial angiogram on day 0.
Population
Patients enrolled in a phase 2 treatment trial for ischemic stroke.
Intervention
Tenecteplase before endovascular therapy.
Comparator
Tissue plasminogen activator before endovascular therapy.
Primary question
Do the randomized groups differ in the proportion achieving substantial angiographic reperfusion or absence of retrievable thrombus at the initial angiogram?
3. Trial Design
The registry describes EXTEND-IA as a randomized, parallel-group, unmasked phase 2 treatment trial. There were 2 intervention arms and 202 enrolled participants. The trial began on 2015-03-23 and reached its primary completion date on 2017-10-16.
Tenecteplase
- Tenecteplase
- Administered before endovascular therapy
Tissue Plasminogen Activator
- Tissue plasminogen activator
- Administered before endovascular therapy
The parallel design is statistically important because each participant is assigned to one randomized treatment group rather than receiving both interventions as part of a crossover. The treatment comparison can therefore be framed as a difference between two independently randomized proportions for the primary endpoint.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary endpoint | Proportion of patients with substantial angiographic reperfusion (mTICI) score of 2b/3 (restoration of blood flow to >50% of the affected arterial territory) or absence of retrievable thrombus at initial angiogram. | Initial angiogram (day 0) |
The primary endpoint is a binary composite-style angiographic outcome: a participant meets the endpoint if the initial angiogram demonstrates either substantial angiographic reperfusion, defined using the registry's mTICI 2b/3 criterion, or absence of a retrievable thrombus.
Why the endpoint is statistically different from a time-to-event endpoint
This endpoint is assessed at a specified procedural time rather than being measured as time until an event occurs. That distinction determines the appropriate statistical framework. There is no need for Kaplan-Meier estimation merely because the outcome is clinically important; the primary data structure is a binary response observed at the initial angiogram.
For each treatment group, the basic quantity is the proportion of participants who satisfy the registry-defined endpoint. The treatment comparison then concerns the difference between those proportions or another prespecified measure of relative effect.
5. Statistical Methodology
The ClinicalTrials.gov record identifies the primary endpoint but does not post a statistical analysis for it. For a binary endpoint comparing two randomized parallel groups, the principal statistical task would normally be to compare the proportions meeting the endpoint between the tenecteplase and tissue-plasminogen-activator groups.
| Statistical component | Application to this endpoint |
|---|---|
| Outcome type | Binary: endpoint met versus endpoint not met |
| Unit of analysis | Individual trial participant |
| Primary comparison | Proportion meeting the endpoint in the randomized treatment groups |
| Absolute effect | Difference in proportions |
| Relative effect | Risk ratio or odds ratio, if prespecified |
| Uncertainty | Confidence interval for the selected treatment-effect measure |
| Time frame | Initial angiogram (day 0) |
Difference in proportions
One natural measure for a binary endpoint is the absolute difference between treatment-group proportions:
Here, pTNK represents the proportion meeting the primary endpoint in the tenecteplase group and ptPA represents the corresponding proportion in the tissue-plasminogen-activator group.
An absolute difference is particularly interpretable because it expresses the comparison directly in percentage-point terms. For example, a difference of 0.10 would correspond to a 10-percentage-point difference. The direction of the difference depends on which treatment group has the larger observed proportion.
Risk ratio
A relative comparison could instead be expressed as a risk ratio:
A risk ratio above 1 indicates a larger observed proportion meeting the endpoint in the numerator group; a risk ratio below 1 indicates a smaller observed proportion.
The risk ratio and absolute difference answer different questions. The absolute difference describes the change in percentage points, whereas the risk ratio describes the multiplicative relationship between the two proportions. Both can be useful when the underlying event rates are available.
Odds ratio
An odds ratio is another possible effect measure for a binary endpoint:
The odds ratio compares the odds of meeting the endpoint rather than the probabilities themselves. It should not automatically be described as a risk ratio, particularly when the endpoint is common.
6. Planned Analysis
The registry identifies a single primary endpoint: the proportion of patients with substantial angiographic reperfusion (mTICI) score of 2b/3, defined as restoration of blood flow to >50% of the affected arterial territory, or absence of retrievable thrombus at the initial angiogram. The time frame is the initial angiogram on day 0.
For this type of endpoint, the usual primary analysis would compare the proportion meeting the endpoint between the two randomized treatment groups. The analysis could be expressed using an absolute risk difference, risk ratio, odds ratio, or another prespecified binary-outcome measure, with a corresponding confidence interval and hypothesis test.
Observed proportion
The first descriptive quantity is the percentage of participants in each randomized group who meet the angiographic endpoint.
Absolute comparison
The risk difference expresses how many additional or fewer participants per 100 would meet the endpoint under one treatment compared with the other.
Relative comparison
A risk ratio or odds ratio expresses the treatment comparison on a relative scale rather than a percentage-point scale.
Statistical uncertainty
A confidence interval would quantify uncertainty around the selected treatment-effect estimate rather than simply reporting whether a hypothesis test crossed a threshold.
7. Statistical Methods Explained
What makes the primary endpoint a binary outcome?
The endpoint is based on whether each participant satisfies a defined angiographic condition at the initial angiogram. Each participant therefore contributes a response that can be classified as meeting or not meeting the endpoint. The resulting analysis is fundamentally a comparison of two proportions.
Why is randomization important for the comparison?
Randomization creates the intended treatment contrast before the outcome is observed. In a parallel randomized trial, the primary comparison is between participants assigned to tenecteplase and participants assigned to tissue plasminogen activator. Randomization is therefore the foundation for interpreting the between-group difference as a treatment comparison rather than simply an association between treatment received and outcome.
Why report an absolute difference?
An absolute difference preserves the clinical scale of the endpoint. If one group has a response proportion of p1 and another has p2, their difference directly describes the change in the percentage of participants meeting the endpoint. Relative measures can appear large even when the absolute change is small, so absolute and relative measures should be distinguished.
What would a confidence interval add?
A point estimate alone does not communicate how precisely the treatment effect has been estimated. A confidence interval gives a range of effect estimates compatible with the statistical model and observed data under the chosen confidence level. A narrow interval indicates greater statistical precision than a wide interval, all else equal.
Why does a p-value not measure treatment effect size?
A p-value addresses evidence against a specified null hypothesis under the statistical model. It does not state how large the treatment difference is. Effect size should instead be described using a quantity such as an absolute difference, risk ratio, or odds ratio, together with its confidence interval.
Why does the composite endpoint require careful interpretation?
The primary endpoint can be satisfied through either substantial angiographic reperfusion or absence of retrievable thrombus. Those two pathways are clinically related but not numerically interchangeable. A participant meeting either criterion contributes to the overall binary endpoint, so the overall proportion represents the combined endpoint rather than a separate estimate for each component.
8. Interpreting the Primary Endpoint
The registry's endpoint is deliberately tied to a specific procedural assessment: the initial angiogram (day 0). This makes the timing of measurement an important part of the estimand. The analysis is not describing a general probability of recovery over an extended follow-up period; it concerns whether the predefined angiographic condition is present at that particular assessment.
If an analysis reports a positive difference in proportions for tenecteplase versus tissue plasminogen activator, that means a larger proportion of participants in the tenecteplase group met the registry-defined angiographic endpoint at the initial angiogram. The difference would be expressed in percentage points when written as a risk difference.
If a risk ratio were used, a value above 1 would indicate that the observed probability of meeting the endpoint was higher in the tenecteplase group than in the tissue-plasminogen-activator group. An odds ratio would require the more specific interpretation of odds rather than probability.
An angiographic endpoint at day 0 is not itself a measure of long-term survival, neurological recovery, functional status, or quality of life. Statistical interpretation should remain aligned with the endpoint that was actually defined and the time at which it was assessed.
9. Composite Endpoint Considerations
The primary endpoint combines two qualifying conditions: an mTICI score of 2b/3, defined by the registry as restoration of blood flow to >50% of the affected arterial territory, or absence of retrievable thrombus at the initial angiogram.
| Component | Registry-defined role |
|---|---|
| mTICI 2b/3 | Substantial angiographic reperfusion, defined as restoration of blood flow to >50% of the affected arterial territory |
| Absence of retrievable thrombus | Alternative qualifying condition for the primary endpoint |
| Combined endpoint | Participant meets the endpoint if either qualifying condition is present |
| Assessment | Initial angiogram (day 0) |
Composite endpoints can increase the number of participants contributing to an analysis because more than one clinical or procedural pathway can satisfy the endpoint. But interpretation depends on the components. If the components have very different clinical importance, the overall composite proportion may conceal differences in which component is driving the result.
For EXTEND-IA, the registry definition should therefore be retained whenever the primary endpoint is described. Treating the endpoint simply as "reperfusion" would omit the second qualifying condition, while treating it as two separate primary endpoints would change the statistical question.
10. Statistical Precision and Sample Size
The registry records an enrollment of 202 participants. Sample size is directly relevant to the precision of a binary treatment comparison: with more participants, estimated proportions generally become more precise, although the actual precision also depends on the observed proportions and the analysis method.
The expression illustrates why the amount of information depends on both the estimated proportion and the number of observations. The corresponding treatment comparison requires its own variance calculation.
The number 202 describes enrollment, not the width of a confidence interval and not the magnitude of a treatment effect. Those quantities require the observed endpoint data and the specified statistical analysis. The registry does not post those statistical results in the record described here.
11. Randomization and Causal Interpretation
Randomization is one of the central statistical features of EXTEND-IA. Because participants were randomized to the two treatment arms, the treatment groups are defined by assignment rather than by an observational decision to receive one treatment or another.
What randomization supports
It establishes the framework for a causal comparison between the assigned interventions, subject to the usual assumptions and conduct of a randomized trial.
What randomization does not guarantee
Randomization does not make the groups numerically identical on every characteristic, nor does it remove statistical uncertainty from the treatment comparison.
Why allocation matters
The treatment contrast is defined prospectively by assignment, which is preferable to comparing groups created by post-treatment characteristics.
Why the endpoint timing matters
The causal question must be linked to the prespecified measurement point: the initial angiogram on day 0.
12. Masking and Measurement
The registry classifies masking as NONE. This means the trial is recorded as an unmasked study rather than a blinded comparison. That feature is particularly relevant when interpreting outcomes that involve assessment procedures.
The primary endpoint is angiographic and is assessed at the initial angiogram. Because the registry records no masking, readers should distinguish the statistical advantage created by randomization from the separate issue of whether outcome assessment was blinded. Randomization and masking address different sources of potential bias.
| Design feature | Statistical implication |
|---|---|
| Randomized | Supports comparison according to treatment assignment |
| Parallel | Each participant belongs to one randomized treatment group |
| Unmasked | The registry reports no masking |
| Primary endpoint | Objective angiographic classification at the initial angiogram |
| Endpoint time frame | Day 0 procedural assessment |
13. Statistical Interpretation of a Two-Group Proportion
A useful way to understand the primary analysis is to imagine the trial's data arranged as a two-by-two table. Each randomized treatment group forms a column, while the endpoint status forms two rows.
| Endpoint met | Endpoint not met | |
|---|---|---|
| Tenecteplase | a | b |
| Tissue plasminogen activator | c | d |
The tenecteplase proportion would be a/(a+b), while the tissue-plasminogen-activator proportion would be c/(c+d). A treatment comparison could then be expressed as a difference, ratio, or odds ratio. The appropriate confidence interval and hypothesis test depend on the prespecified statistical method.
This framework also clarifies why the enrollment number alone cannot generate an efficacy result. The total number of enrolled participants provides the denominator framework, but the treatment effect requires the number of participants who actually meet the endpoint in each randomized group.
14. What a Formal Hypothesis Test Would Ask
For a two-group binary endpoint, a conventional null hypothesis might state that the underlying endpoint proportions are equal:
The corresponding alternative hypothesis would depend on whether the study was designed to test for any difference or a directional difference. The registry record does not post the statistical analysis specifying that testing framework.
A hypothesis test and an effect estimate should be reported together when formal results are available. The test addresses compatibility with the null hypothesis, while the effect estimate describes the magnitude and direction of the observed treatment difference.
15. Why Day 0 Timing Matters
The primary endpoint is measured at the initial angiogram (day 0). This creates a sharply defined analysis time point and avoids ambiguity about when the primary outcome is supposed to be evaluated.
For statistical interpretation, timing is part of the endpoint definition. Changing the assessment window could change which observations qualify and could therefore change the estimand. An endpoint assessed immediately during the procedural period is not interchangeable with an endpoint assessed after subsequent clinical follow-up.
Trial start
The registry records the study start date as 2015-03-23.
Primary endpoint assessment
The primary endpoint is assessed at the initial angiogram.
Primary completion
The registry records 2017-10-16 as the primary completion date.
16. Limitations
- No posted statistical analysis: the ClinicalTrials.gov record does not post formal statistical analyses for the primary endpoint.
- No posted treatment-effect estimates: the registry record described here does not provide a numerical estimate, confidence interval, or p-value for the primary endpoint.
- No analysis population details: the registry information does not specify a statistical analysis population for the primary endpoint beyond the enrolled trial population.
- Composite endpoint: the primary endpoint can be satisfied through either substantial angiographic reperfusion or absence of retrievable thrombus, so an overall endpoint proportion combines these two qualifying pathways.
- Unmasked design: the registry classifies masking as NONE, so randomization should not be conflated with blinded assessment.
- Single time point: the primary endpoint is defined at the initial angiogram on day 0 and should not be interpreted as a long-term clinical outcome.
- Statistical specification: without a posted analysis plan in the registry record, the precise choice among possible binary-outcome effect measures and hypothesis-testing procedures is not established by the record.
17. Why This Trial Matters Statistically
EXTEND-IA is a useful teaching example because it illustrates a form of clinical-trial analysis that is easy to overlook when attention is concentrated on survival analysis. The primary endpoint is a binary procedural outcome assessed at a defined time point, so the central statistical problem is the comparison of proportions between randomized parallel groups.
| Concept | How it appears in EXTEND-IA |
|---|---|
| Randomization | Participants were randomly allocated to the two treatment arms. |
| Parallel-group design | The registry identifies the design model as PARALLEL. |
| Binary endpoint | The primary outcome is the proportion meeting a defined angiographic condition. |
| Composite endpoint | mTICI 2b/3 or absence of retrievable thrombus qualifies as endpoint achievement. |
| Fixed assessment time | The primary endpoint is measured at the initial angiogram (day 0). |
| Absolute effect | A difference in endpoint proportions gives a percentage-point treatment comparison. |
| Relative effect | A risk ratio or odds ratio can express the treatment contrast on a relative scale. |
| Confidence interval | Quantifies uncertainty around the estimated treatment effect. |
| Hypothesis testing | Can assess a prespecified null hypothesis about the two treatment-group proportions. |
| Unmasked design | The registry reports masking as NONE. |
18. Statistical Concepts in This Trial
The statistical structure of EXTEND-IA is particularly useful for understanding how randomized trials with binary endpoints are analyzed. The primary endpoint is neither a continuous measurement nor a time-to-event outcome; it is a yes/no outcome assessed at a defined procedural time.
Binary outcomes
Each participant either meets or does not meet the registry-defined primary endpoint.
Risk difference
Compares treatment groups in absolute percentage-point terms.
Risk ratio
Compares the probabilities of endpoint achievement on a multiplicative scale.
Odds ratio
Compares endpoint odds rather than endpoint probabilities.
Confidence intervals
Describe statistical uncertainty around an estimated treatment effect.
Composite endpoints
Require attention to the individual components that can qualify a participant as an endpoint responder.
19. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoint is a two-group binary comparison. The appropriate analysis focuses on the proportions meeting the endpoint, the treatment-effect measure selected for comparison, its uncertainty, and the hypothesis-testing framework.
Clinical interpretation
The endpoint concerns substantial angiographic reperfusion or absence of retrievable thrombus at the initial angiogram. Its clinical meaning should therefore remain tied to this procedural outcome rather than being treated as a direct measure of every later patient outcome.
20. Reading a Future Treatment Effect Estimate
If a formal analysis reports a treatment effect for EXTEND-IA, several quantities should be read together rather than in isolation.
| Quantity | Question it answers |
|---|---|
| Tenecteplase proportion | What proportion met the primary endpoint in the tenecteplase group? |
| Tissue plasminogen activator proportion | What proportion met the primary endpoint in the comparator group? |
| Risk difference | How many percentage points separate the groups? |
| Risk ratio | How does the probability of endpoint achievement compare multiplicatively? |
| Odds ratio | How do the odds of endpoint achievement compare? |
| Confidence interval | How precisely has the treatment effect been estimated? |
| p-value | How compatible are the observed data with the specified null hypothesis? |
This separation is important because an effect estimate can be clinically interpretable even when statistical uncertainty is substantial, while a small p-value does not by itself establish that an effect is large or clinically important.
21. Sources
- ClinicalTrials.gov: NCT02388061 — EXTEND-IA.
Continue through the Clinical Biostats knowledge graph
Explore the broader statistical methods used to design, analyze, and interpret randomized clinical trials.
22. Record Summary
EXTEND-IA is a completed randomized phase 2 trial with 202 enrolled participants, 2 treatment arms, a parallel design, and no masking. The registry defines one primary endpoint: the proportion of patients with substantial angiographic reperfusion, specified as an mTICI score of 2b/3 with restoration of blood flow to >50% of the affected arterial territory, or absence of retrievable thrombus at the initial angiogram on day 0.
Statistically, this makes the primary question a comparison of two randomized proportions rather than a time-to-event analysis. The most informative formal presentation would pair the observed endpoint proportion in each treatment group with an absolute and/or relative treatment-effect estimate, its confidence interval, and the corresponding hypothesis-testing framework. The composite nature of the endpoint and its fixed day-0 assessment are central to interpreting what the resulting statistic would actually mean.