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
EARLY TAVR is a randomized, parallel-group clinical trial evaluating TAVR compared with surveillance in patients with asymptomatic severe aortic stenosis. The registry lists 901 participants and two arms, with the primary endpoint assessed when all patients have reached 2-year follow-up.
| Feature | EARLY TAVR |
|---|---|
| Trial name | EARLY TAVR: Evaluation of TAVR Compared to Surveillance for Patients With Asymptomatic Severe Aortic Stenosis |
| NCT identifier | NCT03042104 |
| Condition | Heart Diseases; Aortic Stenosis, Severe |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 901 |
| Number of arms | 2 |
| Intervention | Edwards SAPIEN 3 / SAPIEN 3 Ultra THV |
| Lead sponsor | Edwards Lifesciences |
| Sponsor type | Industry |
| Status | Active, not recruiting |
2. Clinical Question
The central statistical question is whether a strategy involving transcatheter aortic valve replacement differs from surveillance with respect to the registry's prespecified composite of all-cause death, all stroke, and unplanned cardiovascular hospitalization when all patients have reached 2-year follow-up.
Population
Patients with severe aortic stenosis who are asymptomatic, within the population enrolled in the EARLY TAVR randomized trial.
Intervention
Edwards SAPIEN 3 / SAPIEN 3 Ultra THV, a transcatheter heart valve device.
Comparator
Surveillance, as specified in the trial title and comparison.
Primary question
How does the randomized TAVR strategy compare with surveillance on the occurrence of the prespecified composite endpoint at the 2-year assessment?
3. Trial Design
Randomized allocation is the central design feature for comparing the two trial strategies.
The registry identifies the trial as a parallel-group design with two arms.
The registry lists the study as having no masking.
The registry identifies treatment as the primary purpose of the study.
Transcatheter aortic valve replacement
- Edwards SAPIEN 3 / SAPIEN 3 Ultra THV
- Device intervention
- Evaluated in patients with asymptomatic severe aortic stenosis
Surveillance
- Surveillance strategy
- Randomized comparison with TAVR
- Evaluated in the same trial population
4. Enrollment and Trial Timeline
Trial start
The registry lists 2017-07-12 as the study start date.
Registered enrollment
The study enrollment is listed as 901 participants across two randomized arms.
Primary completion date
The registry lists 2024-03-31 as the primary completion date.
Active, not recruiting
The registry status is ACTIVE_NOT_RECRUITING.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| All-cause death, all stroke, and unplanned cardiovascular hospitalization | The number of patients that have any of these conditions | When all patients have reached 2-year follow-up |
The endpoint is a composite endpoint: a patient is counted if the patient has any one of the listed component conditions. The statistical unit in the registry description is therefore the number of patients experiencing at least one component of the composite, rather than the total number of component events.
A composite endpoint answers a different question from separate analyses of death, stroke, or hospitalization. If a patient experiences more than one component, the registry's wording defines the primary measure around whether the patient has any of the specified conditions.
6. Planned Analysis
The ClinicalTrials.gov record identifies the primary endpoint as the number of patients who have any of the three component conditions when all patients have reached 2-year follow-up. No posted statistical analyses are listed in the registry record.
For an endpoint defined this way, the basic analysis would compare the proportion of patients in each randomized arm who experience at least one component of the composite by the specified 2-year assessment. A treatment effect can be expressed using an absolute risk difference, a relative risk, or another prespecified binary-outcome measure, with a confidence interval describing statistical precision.
The key distinction is between patients with the composite and the number of individual events. The registry's primary endpoint description uses the former.
If event dates are available in the underlying trial analysis, a time-to-first-event approach can additionally account for when the first component occurs rather than only whether it occurred by the 2-year landmark. That is a methodological option for this type of clinical endpoint; the registry information available here does not specify a particular time-to-event model, hazard ratio, or hypothesis-testing procedure.
7. Statistical Methodology
Composite endpoint analysis
The primary endpoint combines three clinically distinct outcomes: all-cause death, all stroke, and unplanned cardiovascular hospitalization. The primary binary question is whether a patient experiences any of these outcomes by the 2-year assessment.
A risk difference describes the absolute difference in the probability of experiencing at least one component of the composite by the specified follow-up time.
Risk ratio
A relative measure such as the risk ratio can describe how the probability of the composite in the TAVR arm compares with the probability in the surveillance arm. Relative measures are useful when the scientific question concerns proportional differences, while absolute measures show the magnitude of the difference in percentage-point terms.
An RR of 1 would indicate equal observed risks at the specified assessment time. Values below or above 1 describe relative differences in opposite directions.
Confidence intervals
A confidence interval around a risk difference or risk ratio quantifies statistical uncertainty associated with the estimated treatment comparison. A narrow interval indicates greater statistical precision than a wide interval, all else being equal.
Time-to-event analysis
Clinical trials with death, stroke, and hospitalization may also analyze time to the first component event. Kaplan-Meier estimation can describe the distribution of time to first event, while a Cox proportional-hazards model can provide a hazard ratio. Those methods are appropriate when event timing and censoring are part of the prespecified analysis, but the registry record summarized here does not identify such methods as the posted primary analysis.
Competing events within a composite
Death, stroke, and hospitalization do not have identical clinical meanings. Death also prevents a later nonfatal event from occurring. Consequently, an analysis of the composite should not be interpreted as though its three components were interchangeable. The composite provides one patient-level outcome, while the component outcomes answer different clinical questions.
8. Statistical Methods Explained
Why use a composite endpoint?
A composite can combine several clinically relevant outcomes into a single prespecified measure. This can provide a broader assessment of whether patients experience any important event during follow-up. The tradeoff is that the components may differ substantially in severity, frequency, and treatment responsiveness.
Why count patients rather than total events?
The registry defines the primary endpoint as the number of patients who have any of the listed conditions. This means the primary count is patient-based. Recurrent hospitalizations or multiple events in the same patient are not equivalent to additional patients meeting the registry's stated composite definition.
What does a risk difference tell us?
A risk difference compares the probability of experiencing the composite between the two randomized strategies. Unlike a relative measure, it directly describes the difference in absolute probability at the specified follow-up time.
What does a risk ratio tell us?
A risk ratio compares the probability of the composite in one randomized group with the probability in the other. A value of 1 represents equal risks; a value below 1 represents a lower observed probability in the numerator group, and a value above 1 represents a higher probability.
Why might time-to-event methods also be useful?
Two patients can both experience the composite by 2 years but experience it at very different times. Time-to-event analysis uses that timing information and can also account for patients who have not experienced the endpoint by their last available follow-up through censoring.
Why should the composite components also be examined separately?
A composite can obscure important differences among its components. A treatment effect on hospitalization does not necessarily imply the same effect on stroke or death. Separate component analyses help explain what is driving the overall composite result, although the statistical interpretation of those analyses depends on the prespecified testing framework.
Why does the absence of masking matter statistically?
The registry lists masking as none. For an objective component such as all-cause death, lack of masking is less directly relevant to event ascertainment than it may be for outcomes requiring clinical judgment. For outcomes such as hospitalization or stroke classification, knowledge of treatment assignment can potentially influence aspects of assessment or reporting, making prespecified definitions and consistent adjudication important design considerations.
9. Understanding the Composite Endpoint
All-cause death
This component captures death from any cause and is therefore not restricted to cardiovascular mortality.
All stroke
This component is specified broadly as all stroke rather than a narrower stroke subtype in the registry endpoint name.
Unplanned cardiovascular hospitalization
This component captures an unplanned hospitalization categorized as cardiovascular within the endpoint definition.
Composite event
A patient meeting any one of the three conditions contributes to the primary composite endpoint.
The statistical interpretation of the composite should therefore remain tied to its exact definition. A statement about the primary endpoint is a statement about experiencing at least one of these specified outcomes, not a statement that every component changes to the same degree.
10. What a Future Result Would Need to Show
For a transparent statistical presentation of the primary endpoint, the most informative result table would report the number of patients with at least one composite event in each randomized arm, the corresponding denominators, the resulting risks, and a prespecified measure of treatment effect with its confidence interval.
| Result element | Why it matters |
|---|---|
| Number with the composite | Shows how many patients experienced at least one component. |
| Number analyzed | Provides the denominator needed to interpret the event count as a risk. |
| Risk in each arm | Provides the absolute probability of the composite by the specified assessment. |
| Absolute treatment difference | Shows the percentage-point difference between randomized strategies. |
| Relative treatment effect | Shows the proportional relationship between the two risks. |
| Confidence interval | Shows statistical precision around the treatment-effect estimate. |
| Component outcomes | Shows whether the composite is driven primarily by death, stroke, hospitalization, or a combination. |
The registry record represented here does not provide those statistical estimates, confidence intervals, or p-values.
11. Statistical Interpretation of the Primary Endpoint
The primary endpoint asks whether a patient experienced at least one of all-cause death, all stroke, or unplanned cardiovascular hospitalization by the time all patients had reached 2-year follow-up.
The composite is not a measure of the total number of events. It also does not imply that death, stroke, and hospitalization have equal clinical importance or that a treatment effect on one component necessarily applies equally to the others.
An absolute risk difference describes how many additional or fewer patients per unit population experienced the composite, whereas a relative measure describes the proportional relationship between the two risks. Presenting both avoids relying on a single summary of treatment effect.
A point estimate alone does not show how precisely the randomized comparison has been estimated. The confidence interval provides the corresponding uncertainty under the selected statistical model and analysis framework.
12. Limitations
- Composite interpretation: death, stroke, and unplanned cardiovascular hospitalization are clinically different outcomes, so a combined result can conceal differences among components.
- Component weighting: a simple patient-level composite gives each patient one endpoint status regardless of which component occurred; it does not assign greater numerical weight to death than to hospitalization.
- Endpoint timing: the registry specifies assessment when all patients have reached 2-year follow-up, but the available record does not specify a posted time-to-event analysis.
- Statistical method specification: the registry information available here does not identify a particular hypothesis test, confidence-interval method, hazard-ratio model, stratification scheme, or multiplicity procedure for the primary endpoint.
- No posted statistical analyses: the ClinicalTrials.gov record contains no posted statistical analyses for the primary endpoint.
- Unmasked design: the registry lists no masking, which is an important design consideration for outcomes that can involve clinical assessment or classification.
- Generalizability: interpretation applies to the population enrolled in this randomized study of asymptomatic severe aortic stenosis and should not automatically be extended to populations outside the trial's eligibility criteria.
13. Why This Trial Matters Statistically
EARLY TAVR provides a useful statistical teaching case because the primary endpoint combines several clinically important outcomes and evaluates them at a prespecified follow-up milestone within a randomized parallel-group design.
| Concept | How it appears in EARLY TAVR |
|---|---|
| Randomization | The registry identifies randomized allocation to two trial arms. |
| Parallel-group design | The study uses a parallel design rather than sequential treatment assignment. |
| Composite endpoint | The primary endpoint combines all-cause death, all stroke, and unplanned cardiovascular hospitalization. |
| Patient-level analysis | The registry defines the primary endpoint as the number of patients having any listed condition. |
| Fixed follow-up assessment | The primary endpoint is assessed when all patients have reached 2-year follow-up. |
| Absolute risk | The proportion experiencing the composite can describe the event burden in each randomized arm. |
| Relative risk | A risk ratio can describe the proportional difference between randomized groups when used in a prespecified analysis. |
| Confidence intervals | Interval estimates provide information about the precision of the treatment comparison. |
| Component analysis | Separate examination of death, stroke, and hospitalization helps explain the composite result. |
| Unmasked design | The registry identifies the study as having no masking. |
14. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoint is a patient-level composite assessed at 2-year follow-up. A statistically complete analysis would compare the risks between randomized arms and report an effect estimate together with its confidence interval and appropriate hypothesis-testing procedure.
Clinical interpretation
The meaning of a treatment comparison depends on which components contribute to the composite. Death, stroke, and unplanned cardiovascular hospitalization represent different clinical outcomes and should not be treated as interchangeable simply because they share one endpoint definition.
15. Questions a Statistical Reader Should Ask
When evaluating the eventual results of a composite endpoint such as the EARLY TAVR primary outcome, several questions are especially important:
- What proportion of patients in each randomized arm experienced at least one component by 2 years?
- What absolute risk difference and relative treatment effect were prespecified?
- What confidence interval accompanies the primary treatment-effect estimate?
- How many events came from death, stroke, and unplanned cardiovascular hospitalization?
- Were the component outcomes directionally consistent?
- Were event times analyzed, and if so, what censoring and time-to-event methods were prespecified?
- Was the statistical analysis based on all randomized participants or another defined analysis population?
- How were missing outcome assessments handled?
- Were component and secondary analyses subject to a prespecified multiplicity strategy?
These questions distinguish the statistical result of the composite from the broader clinical interpretation of the trial.
16. Sources
- ClinicalTrials.gov: EARLY TAVR, NCT03042104.
- PubMed record: PubMed.
- PubMed record: PubMed.
- PubMed record: PubMed.
- PubMed record: PubMed.
- PubMed record: PubMed.
Continue through the Clinical Biostats knowledge graph
Explore the broader statistical framework for randomized clinical trials, composite endpoints, risk measures, and time-to-event analysis.
17. Record Summary
EARLY TAVR is a randomized, parallel-group trial enrolling 901 participants with severe aortic stenosis, comparing TAVR using Edwards SAPIEN 3 / SAPIEN 3 Ultra THV with surveillance. Its registered primary endpoint is the number of patients experiencing any of all-cause death, all stroke, or unplanned cardiovascular hospitalization when all patients have reached 2-year follow-up. Statistically, the central challenge is interpreting a patient-level composite: the overall endpoint can be summarized with absolute and relative measures, but its clinical meaning depends on understanding the individual components as well as the combined result.