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
ELAN was a randomized, parallel, two-arm, assessor-blinded trial comparing early versus late initiation of direct oral anticoagulants in post-ischaemic stroke patients with atrial fibrillation. The registry lists a single primary endpoint: a composite of major bleeding, recurrent ischaemic stroke, systemic embolism and/or vascular death measured 30 ± 3 days after randomisation.
| Feature | ELAN |
|---|---|
| Study acronym | ELAN |
| ClinicalTrials.gov identifier | NCT03148457 |
| Condition | Ischaemic Stroke |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Primary purpose | Treatment |
| Number of arms | 2 |
| Enrollment | 2013 |
| Start date | 2017-11-06 |
| Primary completion date | 2022-12-21 |
| Lead sponsor | Insel Gruppe AG, University Hospital Bern |
| Sponsor type | OTHER |
| Status | COMPLETED |
2. Clinical Question
The clinical question is whether the timing of initiation of direct oral anticoagulant treatment after ischaemic stroke should be earlier or later. The registry compares two timing strategies using the same listed classes of direct oral anticoagulants in both arms.
Population
Post-ischaemic stroke patients with atrial fibrillation.
Intervention
Early treatment with rivaroxaban, dabigatran, apixaban or edoxaban.
Comparator
Late treatment with rivaroxaban, dabigatran, apixaban or edoxaban.
Primary question
How does early versus late initiation of direct oral anticoagulants compare for the registry-defined composite outcome at 30 ± 3 days after randomisation?
3. Trial Design
The trial is described by the registry as an international, multicentre, randomized-controlled, two-arm, assessor-blinded trial. The parallel design means participants are allocated to one of the two treatment-timing strategies rather than being assigned sequentially to both strategies.
4. Treatment Strategies
Early initiation of direct oral anticoagulants
- Rivaroxaban (Xarelto®)
- Dabigatran (Pradaxa®)
- Apixaban (Eliquis®)
- Edoxaban (Lixiana®)
Late initiation of direct oral anticoagulants
- Rivaroxaban (Xarelto®)
- Dabigatran (Pradaxa®)
- Apixaban (Eliquis®)
- Edoxaban (Lixiana®)
The key randomized contrast is therefore the timing of direct oral anticoagulant initiation. The listed anticoagulant options occur in both arms, so the central design comparison is early versus late treatment rather than one anticoagulant versus another.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary endpoint | Composite of major bleeding, recurrent ischaemic stroke, systemic embolism and/or vascular death | 30 ± 3 days after randomisation |
This is a composite clinical endpoint. A participant can contribute to the primary endpoint if at least one of the listed component events occurs during the specified assessment period. The components represent clinically distinct outcomes, which is important when interpreting a composite result.
Understanding the components
Major bleeding
A bleeding event included as one of the components of the registry-defined composite endpoint.
Recurrent ischaemic stroke
A recurrent ischaemic stroke occurring during the specified 30 ± 3 day assessment period.
Systemic embolism
A systemic embolic event included in the composite outcome.
Vascular death
Death from a vascular cause, included as the fourth component of the composite endpoint.
6. Planned Analysis
The primary endpoint is measured within a prespecified short follow-up window of 30 ± 3 days after randomisation. For an endpoint defined as whether at least one component event occurs during a fixed assessment window, the primary comparison would typically be expressed as a comparison of the proportion of participants experiencing the composite outcome in the two randomized groups.
Common statistical summaries for this type of endpoint include the event proportion in each arm, an absolute risk difference, and a relative measure such as a risk ratio or odds ratio. A two-group test for proportions may be used to assess whether the observed difference is compatible with the null hypothesis of equal event probabilities. The exact method depends on the prespecified statistical analysis plan, particularly when event counts are small.
A negative risk difference would correspond to a lower observed composite-event probability in the early-treatment group; a positive value would correspond to a higher probability. The numerical value cannot be determined from the registry information shown here.
The fixed 30 ± 3 day window also makes this endpoint different from a conventional long-term time-to-event endpoint. If the analysis is based on whether an event occurred within the specified window, Kaplan-Meier estimation and a Cox proportional-hazards model are not automatically required simply because the endpoint concerns clinical events. Their use would depend on the prespecified analysis and the actual timing and censoring structure of the data.
7. Statistical Methodology
Randomization
Randomization is the central statistical feature of the ELAN design. By assigning participants to early or late treatment strategies before the outcome is observed, randomization provides the basis for comparing the groups as treatment strategies rather than relying on adjustment for observed baseline characteristics alone.
For a randomized comparison, the primary analysis generally retains participants according to their randomized assignment. This preserves the treatment contrast created by randomization and avoids allowing post-randomization treatment decisions to redefine the primary comparison.
Composite endpoint analysis
The primary endpoint combines four possible clinical outcomes: major bleeding, recurrent ischaemic stroke, systemic embolism and/or vascular death. Statistically, the composite can be represented as an indicator variable for whether at least one component occurred during the primary assessment period.
The composite endpoint therefore creates one primary binary outcome from several component events within the specified follow-up window.
Absolute effect
An absolute risk difference is particularly intuitive for a short fixed follow-up period. It answers the question: how many percentage points does the probability of experiencing the composite endpoint differ between the two randomized strategies during the specified window?
Relative effect
A risk ratio compares the probability of the composite endpoint between treatment groups. An odds ratio instead compares the odds of the endpoint. These measures are related but are not interchangeable, particularly when event probabilities are not small.
Confidence intervals
A confidence interval around an estimated treatment effect communicates statistical precision. A narrow interval indicates that the estimate is relatively precise under the assumptions of the analysis, whereas a wider interval indicates greater uncertainty. The interval does not describe the range of effects experienced by individual patients.
P-values
A p-value addresses compatibility of the observed data with a specified null hypothesis under the statistical model. It is not a measure of clinical importance and does not quantify the magnitude of the treatment effect. An effect estimate and its confidence interval should therefore accompany any formal hypothesis test.
8. Statistical Methods Explained
Why is the endpoint a composite?
A composite endpoint can capture several clinically relevant outcomes within a single prespecified outcome measure. In ELAN, the registry combines major bleeding, recurrent ischaemic stroke, systemic embolism and vascular death. This means the primary analysis asks about the occurrence of any qualifying component rather than treating each component as the sole primary endpoint.
Why does the 30 ± 3 day window matter?
The follow-up window defines the period over which the primary endpoint is assessed. A treatment effect measured at 30 ± 3 days answers a different question from an effect measured over a substantially longer period. The time frame should therefore remain attached to the endpoint whenever the result is reported.
What does a risk difference tell us?
The risk difference expresses the treatment contrast in absolute probability terms. For example, a risk difference of −0.03 would represent a 3 percentage-point lower event probability in the first group.
What is the difference between risk ratio and odds ratio?
A risk ratio compares probabilities directly, whereas an odds ratio compares odds. When events are uncommon, the two measures can be numerically similar, but they can diverge as event probabilities increase. Reporting the measure that corresponds to the prespecified analysis is therefore important.
Why is the composite endpoint not the same as each component?
A statistically defined composite can be driven by one or more of its components. Two treatments could have similar effects on one component but different effects on another. Consequently, a composite result should be interpreted together with clinically important component outcomes when those analyses are available.
Why does randomization matter?
Randomization creates the primary basis for causal comparison between early and late treatment strategies. Because treatment assignment is randomized, systematic differences in measured and unmeasured baseline characteristics are expected to be reduced between groups at the point of assignment.
9. The Statistical Meaning of the Trial Design
| Design feature | Statistical implication |
|---|---|
| Randomized allocation | Provides the basis for a causal comparison of the assigned treatment strategies. |
| Two arms | Creates a direct comparison between early and late treatment initiation. |
| Parallel design | Participants remain assigned to one treatment-timing strategy rather than serving as their own randomized crossover comparison. |
| Single masking | The registry identifies the trial as single-masked, with assessor blinding described in the full study title. |
| Composite primary endpoint | Combines four clinically distinct event types into one primary outcome. |
| Fixed primary time frame | Limits the primary endpoint to 30 ± 3 days after randomisation. |
| 2013 enrolled participants | Provides the randomized trial population from which the primary comparison is made. |
10. Interpreting a Future Primary Result
When results for ELAN are reported, the most informative presentation would show the composite-event proportion in each randomized arm together with an effect measure and its confidence interval. The interpretation should distinguish the absolute event probability from the relative treatment effect.
The risk difference would describe the absolute difference in the probability of experiencing the composite endpoint during the 30 ± 3 day assessment period. It would not indicate that every component of the composite changed by the same amount.
The risk ratio would compare the probability of the composite endpoint between the early and late treatment groups. A value below 1 would indicate a lower estimated probability in the numerator group; a value above 1 would indicate a higher estimated probability.
The odds ratio would compare the odds rather than the probabilities of experiencing the composite endpoint. It should not be described as a risk ratio unless the numerical relationship between the measures is justified.
For a risk difference, the null value is 0. For a risk ratio or odds ratio, the null value is 1. A confidence interval containing the relevant null value indicates that the data are compatible with no difference under that analysis at the corresponding confidence level. It does not establish that the treatment strategies are clinically identical.
11. Composite Endpoints: A Deeper Statistical View
Composite endpoints are attractive because they can increase the number of observed events and combine outcomes that matter to patients. But the statistical meaning of the composite depends on the relationship among its components.
Suppose one component is substantially more common than the others. The overall composite event rate may then be driven largely by that component. A statistically clear difference in the composite does not automatically imply that each individual component shows the same magnitude or direction of effect.
Conversely, if component events occur with similar frequency and the treatment affects them in a reasonably consistent way, the composite can provide a useful summary of the overall clinical outcome pattern. The registry definition in ELAN explicitly identifies the four components, making that distinction important when the results are interpreted.
Component frequency
The most frequent component can contribute substantially to the overall composite event count.
Component importance
Components can differ substantially in clinical meaning even when they are combined into one statistical endpoint.
Consistency
Similar treatment effects across components strengthen the interpretability of a composite, whereas divergent effects require more careful examination.
Timing
The 30 ± 3 day assessment period means that component events outside that primary window answer a different question.
12. Assessor Blinding and Outcome Assessment
The registry describes ELAN as assessor-blinded. Blinding of outcome assessment is particularly relevant when clinical endpoints require judgment or classification. It can reduce the opportunity for knowledge of treatment assignment to influence how an outcome is assessed.
Assessor blinding does not make a randomized trial completely insensitive to other sources of bias. It addresses the outcome-assessment process; it does not eliminate differences that could arise after randomization through treatment discontinuation, adherence, additional care, or other post-randomization events.
13. Missing Data and Follow-Up
The primary endpoint has a defined assessment period of 30 ± 3 days after randomisation. For a fixed-window binary endpoint, missing outcome information can affect the denominator and potentially introduce bias if missingness is related to treatment assignment or outcome risk.
A complete statistical analysis would therefore need to specify how participants with incomplete endpoint ascertainment are handled. Possible approaches include prespecified rules for endpoint ascertainment, sensitivity analyses under alternative assumptions, or other methods appropriate to the pattern and mechanism of missingness.
The registry information presented here identifies the endpoint and its time frame but does not report a posted statistical analysis specifying a missing-data or imputation strategy.
14. Multiplicity
The registry lists one primary endpoint: the composite of major bleeding, recurrent ischaemic stroke, systemic embolism and/or vascular death. The available record does not provide a posted statistical analysis describing a multiplicity adjustment for additional endpoints or multiple testing procedures.
That distinction matters because a single prespecified primary endpoint is statistically different from conducting many unadjusted primary tests. If several secondary or component analyses are subsequently examined, their inferential status depends on the prespecified hierarchy and error-control strategy.
The purpose of multiplicity adjustment is to control the probability of making false-positive conclusions across a defined family of confirmatory hypotheses.
15. Interim Analysis
The available registry information does not report a posted statistical analysis describing an interim efficacy analysis, alpha-spending strategy, group-sequential boundary, or other interim-testing procedure for the primary endpoint.
For a randomized trial, interim analyses can be useful for monitoring accumulating evidence, but repeated unadjusted looks at the primary outcome can increase the overall probability of a false-positive finding. A formal interim design therefore requires prespecified information times and an appropriate error-control strategy.
Without a posted statistical analysis specifying such a procedure, the existence or characteristics of an interim hypothesis test cannot be inferred from the trial's enrollment or completion dates.
16. Non-Inferiority, Bayesian Methods, and Other Design Features
| Method or design topic | Registry information available here |
|---|---|
| Non-inferiority margin | Not reported in the available registry information. |
| Bayesian methods | Not reported in the available registry information. |
| Crossover | The design is listed as parallel; no crossover strategy is reported in the available registry information. |
| Interim analysis | No posted statistical analysis describing an interim procedure is available. |
| Missing-data/imputation method | No posted statistical analysis describing a method is available. |
| Stratification factors | Not reported in the available registry information. |
The absence of a listed method should not be interpreted as evidence that the investigators did not use a particular statistical procedure in the full protocol or statistical analysis plan. It means that the available registry information does not specify that method.
17. Why the Primary Endpoint Is Statistically Interesting
ELAN provides a useful example of how trial design and endpoint construction interact. The intervention is defined by timing, while the primary endpoint combines both thromboembolic and bleeding outcomes. That creates a clinically meaningful statistical question: the overall effect of an early versus late treatment strategy is captured through a composite that includes outcomes on different sides of the treatment-risk balance.
This structure makes the trial particularly useful for understanding why the choice of endpoint is part of statistical design rather than merely a reporting decision. A composite endpoint has to be defined before the analysis, its time frame has to be specified, and the interpretation of its treatment effect has to remain tied to its component events.
18. What the Registry Does and Does Not Establish
What is established by the record
The registry identifies ELAN as a completed, randomized, parallel, two-arm, single-masked trial enrolling 2013 participants and compares early versus late initiation of listed direct oral anticoagulants.
What is defined
The primary endpoint is a composite of major bleeding, recurrent ischaemic stroke, systemic embolism and/or vascular death at 30 ± 3 days after randomisation.
What is not posted
No formal statistical analysis is posted on ClinicalTrials.gov for the primary endpoint.
What requires an analysis result
The treatment effect, confidence interval, event counts, p-value, and component-specific outcome estimates require reported outcome data and a corresponding statistical analysis.
19. Limitations
- No posted statistical analysis: the ClinicalTrials.gov record available here lists the primary endpoint but does not contain a posted statistical analysis with an effect estimate, confidence interval, or p-value.
- Composite endpoint: the primary outcome combines four clinically distinct events, so a composite estimate would not by itself describe the effect on each individual component.
- Fixed assessment window: the primary endpoint is defined at 30 ± 3 days after randomisation, so it represents a short-term outcome measure rather than an unrestricted lifetime or long-term event probability.
- Missing component detail: without posted component-level results, the relative contribution of major bleeding, recurrent ischaemic stroke, systemic embolism and vascular death to any eventual composite result cannot be assessed from the registry record alone.
- Statistical-plan detail: the available registry information does not specify the full inferential framework, including any stratification, multiplicity, interim-analysis, or missing-data procedures.
- Timing versus drug identity: the same four listed direct oral anticoagulant options appear in both treatment strategies, so interpretation should focus on the randomized timing strategy rather than treating the trial as a comparison among individual anticoagulants.
20. Why This Trial Matters Statistically
ELAN illustrates several important principles in clinical-trial statistics without requiring a complex survival endpoint. The primary outcome is a fixed-window composite, the treatment contrast is randomized, and the intervention is a treatment-timing strategy rather than a single drug-versus-placebo comparison.
| Concept | How it appears in ELAN |
|---|---|
| Randomization | Participants are randomized to early or late treatment strategies. |
| Parallel design | The two treatment strategies form separate randomized arms. |
| Composite endpoint | Four clinical outcomes are combined into one primary endpoint. |
| Fixed follow-up | The primary endpoint is assessed 30 ± 3 days after randomisation. |
| Absolute risk | A natural measure for a fixed-window event outcome. |
| Relative effect | Risk ratios or odds ratios can describe relative differences between randomized groups when specified by the analysis plan. |
| Confidence intervals | Quantify uncertainty around an estimated treatment effect. |
| Type I error | Multiple confirmatory analyses require appropriate control if more than one hypothesis is formally tested. |
| Assessor blinding | Helps reduce potential influence of treatment knowledge on outcome assessment. |
| Analysis populations | The randomized comparison should remain linked to the treatment assignment created by randomization. |
21. A Practical Statistical Reading of ELAN
A reader approaching the eventual results should start with the endpoint definition rather than with a p-value. The first question is: what proportion of participants experienced at least one of the four qualifying events within 30 ± 3 days?
The second question is the size of the difference between the randomized groups. An absolute risk difference provides a direct percentage-point comparison, while a relative measure places the difference on a proportional scale.
The third question is precision. The confidence interval shows how much uncertainty surrounds the estimated treatment effect. A statistically significant p-value, if reported, would not replace that information.
The fourth question is what is driving the composite. A difference in the combined endpoint is most interpretable when the component outcomes are also understood, because major bleeding and recurrent ischaemic stroke are not interchangeable clinical events.
Finally, the analysis should remain tied to the randomized treatment strategy and its prespecified time frame. The primary result would describe the comparative outcome associated with the early versus late initiation strategy during the 30 ± 3 day assessment period.
22. Trial Timeline
Study start
The ClinicalTrials.gov record lists 2017-11-06 as the study start date.
2013 participants
The registry lists an enrollment of 2013 participants across the two randomized arms.
Primary endpoint assessment
The primary composite endpoint is defined for 30 ± 3 days after randomisation.
Primary completion
The ClinicalTrials.gov record lists 2022-12-21 as the primary completion date.
23. Sources
- ClinicalTrials.gov: NCT03148457 — ELAN.
Continue through Clinical Biostats
Explore additional clinical trial results, statistical tutorials, and analysis tools.
24. Record Summary
ELAN is a completed, randomized, parallel, two-arm, assessor-blinded trial enrolling 2013 participants with ischaemic stroke and evaluating early versus late initiation of direct oral anticoagulants. The registry-defined primary endpoint is a composite of major bleeding, recurrent ischaemic stroke, systemic embolism and/or vascular death, assessed at 30 ± 3 days after randomisation.
Statistically, the trial is a useful example of a randomized comparison built around a treatment-timing strategy and a fixed-window composite endpoint. The eventual primary result should be interpreted through the event proportions, absolute and relative treatment effects, confidence interval, and component-level outcomes, while keeping the 30 ± 3 day time frame attached to the interpretation.