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
EVA-3S is a completed phase 3 randomized trial registered under NCT00190398. Its registry title describes an endarterectomy-versus-angioplasty comparison in patients with severe symptomatic carotid stenosis, while the registry record lists carotid angioplasty and stenting with cerebral protection as the intervention and identifies the primary purpose as prevention.
| Feature | EVA-3S |
|---|---|
| Trial | EVA-3S |
| NCT identifier | NCT00190398 |
| Phase | Phase 3 |
| Status | Completed |
| Conditions | Transient Ischemic Attack; Carotid Stenosis; Atherosclerosis |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Prevention |
| Enrollment | 900 |
| Number of arms listed | 1 |
| Lead sponsor | Assistance Publique - Hôpitaux de Paris |
| Sponsor type | Other |
2. Clinical Question
The clinical question described by the registry title concerns the comparison of endarterectomy with angioplasty in patients with severe symptomatic carotid stenosis. The registry specifically identifies carotid angioplasty and stenting with cerebral protection as the intervention.
Population
Patients represented by the registry conditions of transient ischemic attack, carotid stenosis, and atherosclerosis, with the trial title specifying severe symptomatic carotid stenosis.
Intervention
Carotid angioplasty and stenting with cerebral protection.
Comparator
The trial title identifies endarterectomy as the comparator in the stated endarterectomy-versus-angioplasty comparison. The registry data listed here do not provide a separate comparator-arm description.
Primary question
For patients with severe symptomatic carotid stenosis, how does the intervention identified by the registry relate to the occurrence of stroke or death within 30 days of the procedure?
3. Trial Design
Intervention listed in the registry
Carotid angioplasty and stenting
- Carotid angioplasty and stenting
- Cerebral protection
- Device intervention
Endarterectomy
- Endarterectomy is identified in the registry trial title as the other side of the comparison.
- A separate comparator-arm description is not included in the registry data presented here.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary endpoint | Any stroke or death within 30 days of the procedure | during de study |
The primary endpoint is a composite binary outcome: a participant is counted as having experienced the endpoint if either stroke or death occurs within 30 days of the procedure. Because the registry describes the endpoint in terms of whether an event occurs within a fixed post-procedure window, the natural statistical representation is a proportion or event risk for each randomized group.
The composite combines two clinical events into one primary outcome. The registry does not provide separate event counts for stroke and death.
5. Planned Analysis
No formal statistical analyses are posted to ClinicalTrials.gov for the registered primary endpoint. The registry identifies the endpoint but does not report an effect estimate, confidence interval, p-value, event count, or other statistical result.
How an endpoint of this type is typically analysed
Because the primary endpoint is defined as whether any stroke or death occurs within a fixed 30-day period, a conventional analysis would begin by calculating the proportion of patients experiencing the composite endpoint in each randomized treatment group. The treatment comparison could then be expressed using an absolute risk difference, a relative risk, or an odds ratio, with a confidence interval describing statistical precision.
| Statistical quantity | Interpretation for this endpoint |
|---|---|
| Event proportion | The proportion of patients experiencing any stroke or death within 30 days of the procedure. |
| Risk difference | The absolute difference in the 30-day event proportions between treatment groups. |
| Relative risk | The ratio of the 30-day event probability in one treatment group to that in the other. |
| Odds ratio | The ratio of the odds of the composite endpoint between treatment groups. |
| Confidence interval | A measure of uncertainty around the estimated treatment contrast. |
| Hypothesis test | A statistical test could assess whether the observed difference is compatible with a specified null hypothesis. |
For a randomized comparison with a binary endpoint, a two-group proportion test or a contingency-table method could be used depending on the prespecified statistical analysis plan and the number of observed events. If event counts were small, an exact method could be preferable to a large-sample approximation. The appropriate method should ultimately follow the prespecified analysis plan rather than being selected after seeing the observed results.
6. Statistical Methodology
Binary endpoint analysis
The primary endpoint has a fixed assessment window: any stroke or death within 30 days of the procedure. This differs fundamentally from a time-to-event endpoint such as overall survival, where the exact timing of the event and censoring are central to the analysis. Here, the principal outcome is whether the composite event occurred during the defined window.
The risk difference expresses the absolute separation between the two 30-day event probabilities. The registry does not report those probabilities.
Relative treatment effects
A relative risk compares the probability of the composite event between randomized groups. An odds ratio instead compares the corresponding odds. These measures can differ materially when an outcome is not rare, so the choice should be specified before analysis rather than selected solely because it produces a preferred numerical presentation.
A relative risk below 1 would indicate a lower observed probability of the composite endpoint in the numerator group; it would not by itself describe the absolute clinical difference.
Confidence intervals
A confidence interval complements the point estimate by describing its statistical precision. For a risk difference, the interval is expressed in absolute probability units; for a relative risk or odds ratio, it is expressed on a ratio scale. A narrow interval indicates greater statistical precision than a wide interval, all else being equal.
Randomization and inference
Randomization is important because it establishes the basis for comparing outcome frequencies between treatment assignments. In a randomized trial, the principal inferential question concerns the difference attributable to treatment assignment under the trial's prespecified analysis framework, rather than simply the association between treatment received and outcome.
Composite endpoints
The endpoint combines stroke and death. A composite endpoint can increase the number of patients experiencing the primary event and can therefore provide greater statistical information than either component alone. At the same time, interpretation depends on the clinical importance and frequency of each component. A composite result does not automatically imply that its individual components have identical treatment effects.
7. Statistical Methods Explained
Why is a proportion appropriate for this endpoint?
The endpoint asks whether an event occurs during a prespecified 30-day period. Each patient can therefore be classified as having experienced or not experienced the composite endpoint during that window. This naturally produces a binary outcome that can be summarized as a proportion.
Why might a risk difference be useful?
A risk difference directly describes the absolute separation between treatment groups. For a procedural outcome such as stroke or death within a defined period, an absolute difference can be easier to interpret clinically than a ratio alone because it preserves the original probability scale.
What does a relative risk add?
A relative risk expresses the event probability in one treatment group relative to another. It describes proportional separation, whereas a risk difference describes absolute separation. Both can be informative because a large relative change can correspond to a small absolute change when baseline risk is low.
Why is a confidence interval necessary?
A point estimate alone does not communicate how precisely the treatment effect has been estimated. The confidence interval provides a range of values compatible with the statistical model and observed data under the chosen confidence level. Wider intervals indicate greater uncertainty in the estimate.
Why does randomization matter for the statistical comparison?
Randomization provides the foundation for comparing outcomes according to treatment assignment. It helps balance measured and unmeasured prognostic factors in expectation, allowing differences in outcomes between randomized groups to be interpreted within the causal framework of the trial.
Why should the composite endpoint be interpreted carefully?
Stroke and death are distinct clinical events. A combined endpoint answers the question of whether either event occurred, but it does not by itself describe the treatment effect on each component. Separate component analyses are therefore useful when they are prespecified and available.
8. What the Primary Endpoint Means Statistically
The primary endpoint is a binary composite outcome observed within a fixed 30-day period after the procedure. The fundamental comparison is therefore between the proportions of randomized patients experiencing any stroke or death in the respective treatment groups.
The endpoint does not by itself measure the severity of a stroke, the timing of an event within the 30-day window, long-term survival, or outcomes beyond the specified period. Those would require separate definitions and analyses.
Restricting the endpoint to 30 days creates a standardized period for comparing early procedural outcomes. Events outside that window are not part of this registered primary endpoint unless they are captured by another prespecified outcome.
9. Results Status
The ClinicalTrials.gov record identifies the primary endpoint but contains no posted statistical analyses. Consequently, the registry does not report a numerical treatment effect, confidence interval, p-value, or event count for the primary endpoint.
Reported
The primary endpoint is any stroke or death within 30 days of the procedure.
Not reported
No formal statistical analysis or numerical primary-endpoint result is posted in the registry record.
Interpretive consequence
The registry record supports discussion of the endpoint and its appropriate statistical framework, but not a numerical estimate of the treatment effect.
Analysis population
The ClinicalTrials.gov record does not identify a separate efficacy analysis population or provide participant-level outcome information.
10. Statistical Issues That Would Matter in a Full Analysis
Several design and analysis questions would normally be specified before evaluating a randomized trial with this endpoint. The registry information presented here does not provide those details, so they should be treated as methodological considerations rather than reported features of EVA-3S.
| Issue | Why it matters |
|---|---|
| Analysis population | The analysis should specify whether the primary comparison follows the intention-to-treat principle, a per-protocol population, or another prespecified population. |
| Missing outcome data | The analysis should specify how patients without an ascertainable 30-day outcome are handled. |
| Component events | Stroke and death should be considered separately as well as within the composite when component-level data are available. |
| Effect measure | The statistical analysis plan should specify whether the primary treatment contrast is a risk difference, relative risk, odds ratio, or another measure. |
| Multiplicity | If multiple primary or secondary hypotheses are formally tested, the analysis plan should specify how type I error is controlled. |
| Covariate adjustment | Any adjusted analysis should be prespecified, with the covariates and model defined independently of the observed outcome. |
These considerations illustrate why a complete statistical analysis cannot be reconstructed from an endpoint definition alone. The endpoint determines the broad class of appropriate methods, while the protocol and statistical analysis plan determine the exact inferential framework.
11. Trial Timeline
Trial start
The registry lists 2000-11 as the study start date.
Randomized clinical trial
The registry classifies EVA-3S as a phase 3, randomized, parallel-design study with no masking.
Primary completion
The registry lists 2009-12 as the primary completion date.
Registry status
The study is listed as completed in the ClinicalTrials.gov record.
12. Limitations
- No posted statistical analyses: the registry does not provide a numerical analysis of the primary endpoint.
- No event counts: the record does not provide the number of patients experiencing stroke, death, or the combined endpoint.
- No effect estimate: a risk difference, relative risk, odds ratio, or other treatment-effect estimate is not reported.
- No confidence interval: the registry does not report statistical uncertainty around a treatment-effect estimate.
- No p-value: the registry does not report a formal hypothesis-test result for the primary endpoint.
- No component analysis: the registry defines the composite of stroke or death but does not provide separate component results.
- Limited arm-level detail: the registry information presented here lists one arm and the carotid angioplasty and stenting intervention, while the trial title identifies an endarterectomy-versus-angioplasty comparison.
- No detailed analysis-plan information: the registry data presented here do not specify the primary statistical test, analysis population, missing-data method, multiplicity strategy, or covariate-adjustment strategy.
- Fixed-window interpretation: the primary endpoint concerns events within 30 days of the procedure and should not be interpreted as a measure of long-term outcome.
13. Why This Trial Matters Statistically
EVA-3S is a useful statistical teaching case because its registered primary endpoint illustrates a different analytical problem from the time-to-event endpoints commonly encountered in oncology and other longitudinal trials. The endpoint is defined over a fixed post-procedure interval and combines two clinically important events into a single binary outcome.
| Concept | How it appears in EVA-3S |
|---|---|
| Randomization | The registry classifies allocation as randomized. |
| Parallel design | The registry identifies the design model as parallel. |
| Binary endpoint | The primary endpoint asks whether any stroke or death occurs within a defined period. |
| Composite endpoint | Stroke and death are combined into the primary outcome. |
| Risk difference | Provides an absolute treatment contrast for the 30-day event probability. |
| Relative risk | Provides a proportional comparison of the 30-day event probability. |
| Confidence interval | Describes uncertainty around the treatment-effect estimate. |
| Intention-to-treat principle | A potential framework for preserving the randomized comparison in the primary analysis. |
| Component analysis | Stroke and death may require separate interpretation in addition to the composite. |
| Missing data | A complete analysis would need prespecified rules for patients whose 30-day status is unavailable. |
14. Composite Endpoints: A Deeper Statistical View
The phrase “any stroke or death” creates a composite endpoint by combining two possible events. Statistically, this can be represented as an indicator variable equal to one when at least one component occurs during the specified window and zero otherwise.
A patient who experiences either component is counted as having experienced the primary endpoint. The registry does not provide the component frequencies needed to examine their individual contributions.
Composite endpoints can be statistically efficient because several types of clinically relevant events contribute to the same primary outcome. However, efficiency does not remove the need to examine the components. If one component is much more frequent than the other, the composite treatment effect may largely reflect that more frequent component.
There is also a distinction between statistical and clinical interpretation. A statistically detectable difference in the composite does not automatically establish that every component is affected to the same extent. Component-specific estimates and their uncertainty are therefore important when they are available.
15. Fixed-Time Outcomes Versus Time-to-Event Outcomes
The 30-day definition also determines the statistical structure of the primary analysis. A fixed-time binary endpoint asks whether an event occurred by a specified deadline. A time-to-event endpoint instead retains information about when the event occurred and can account explicitly for censoring.
Fixed-time analysis
The primary EVA-3S endpoint can be summarized as the proportion of patients with any stroke or death within the specified 30-day window.
Time-to-event analysis
A survival analysis would instead model the time from a defined origin to the event and would require a different endpoint definition and censoring framework.
This distinction is important because a treatment can affect both the probability of an event and its timing. A fixed-window endpoint intentionally focuses the analysis on whether the event occurs during the clinically defined period rather than on the exact time at which it occurs.
16. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The registered primary endpoint is a binary composite outcome assessed within 30 days of the procedure. A randomized comparison would ordinarily estimate and compare the corresponding event probabilities using a prespecified inferential method.
Clinical interpretation
The endpoint focuses on early stroke or death following the procedure. Its clinical meaning therefore concerns short-term procedural outcomes rather than long-term vascular or survival outcomes.
Keeping these two levels of interpretation separate is important. Statistical analysis quantifies the treatment comparison and its uncertainty; clinical interpretation considers what the endpoint represents for patients and how it relates to the broader course of disease.
17. Sources
- ClinicalTrials.gov: EVA-3S, NCT00190398.
Explore the Clinical Biostats approach to trial statistics
Clinical trial interpretation is strongest when endpoint definitions, study design, effect measures, uncertainty, and analysis populations are considered together.