This page separates reported registry information from statistical interpretation. The ClinicalTrials.gov record reports results for the registered primary endpoint and six outcome measures, but the ClinicalTrials.gov record contains no formal statistical-analysis estimates, confidence intervals, or p-values.
1. Trial at a Glance
PARTNER 3 is a randomized, parallel-group cardiovascular trial evaluating the safety and effectiveness of the SAPIEN 3 transcatheter heart valve in low-risk patients with aortic stenosis. The registry lists 1000 enrolled participants, two arms, no masking, and treatment as the primary purpose.
| Feature | PARTNER 3 |
|---|---|
| Trial name | PARTNER 3 |
| ClinicalTrials.gov identifier | NCT02675114 |
| Therapeutic area | Cardiovascular |
| Condition | Aortic Stenosis |
| Phase | NA |
| Status | ACTIVE_NOT_RECRUITING |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 1000 |
| Arms | 2 |
| Interventions | SAVR (procedure); SAPIEN 3 THV (device) |
| Lead sponsor | Edwards Lifesciences |
| Sponsor type | INDUSTRY |
| Start | 2016-03 |
| Primary completion | 2020-11-16 |
| Results posted | Yes |
2. Clinical Question
The registry describes PARTNER 3 as a trial of the safety and effectiveness of the SAPIEN 3 transcatheter heart valve in low-risk patients with aortic stenosis. The randomized comparison is between surgical aortic valve replacement and the SAPIEN 3 transcatheter heart valve.
Population
Low-risk patients with aortic stenosis, according to the trial's brief title and registry description.
Intervention
SAPIEN 3 THV, a transcatheter heart valve device.
Comparator
Surgical aortic valve replacement (SAVR).
Primary question
How does the randomized comparison of SAPIEN 3 THV and SAVR perform with respect to the registered 1-year composite endpoint of all-cause mortality, all stroke, and rehospitalization?
Because the ClinicalTrials.gov record does not provide the individual treatment-group event counts or a formal treatment-effect estimate for this composite, the statistical question can be defined precisely but cannot be assigned a numerical comparative result on this page.
3. Trial Design
Surgical Aortic Valve Replacement
- Intervention type: procedure
- Registry abbreviation: SAVR
- Serious adverse events: 221/454 affected/at risk
Transcatheter Aortic Valve Replacement
- Intervention: SAPIEN 3 THV
- Intervention type: device
- Serious adverse events: 153/496 affected/at risk
4. Endpoints
The registry lists 1 registered primary endpoint and reports results as available for that endpoint. The endpoint is a composite measured at 1 year.
| Endpoint | Registry definition | Time frame | Registry status |
|---|---|---|---|
| All-cause Mortality, All Stroke, and Rehospitalization | Number of patients that had any of these events | 1 year | Results posted |
Understanding the composite endpoint
The primary endpoint combines three clinically distinct event types: all-cause mortality, all stroke, and rehospitalization, with the registry specifying that the rehospitalization component is valve-related or procedure-related and includes heart failure.
The registry definition uses the wording "Number of patients that had any of these events." That phrase is important statistically. A patient experiencing more than one component is still counted as one patient for a patient-level composite endpoint. The ClinicalTrials.gov record does not provide the component-specific event counts, so it is not possible to determine from the registry record which component contributed most to the composite.
5. Statistical Methodology
The ClinicalTrials.gov record contains an important distinction: results are posted, but no formal statistical analyses were posted. Therefore, this page does not report a hazard ratio, risk ratio, odds ratio, risk difference, confidence interval, or p-value for the primary endpoint.
For a randomized two-arm trial with a binary patient-level composite endpoint assessed at a fixed 1-year time point, a natural primary analysis would compare the proportion of patients experiencing at least one component of the composite in each randomized group. Depending on the prespecified statistical analysis plan, this could be expressed as a risk difference, relative risk, odds ratio, or another prespecified treatment-effect measure. If time to the first composite event were the actual estimand, a time-to-event analysis could instead be appropriate.
Those are statistical methods that would ordinarily be considered, not analyses documented in the registry-reported PARTNER 3 data. The distinction matters because choosing an analysis after seeing the results can change the inferential properties of a clinical trial.
For a fixed 1-year binary endpoint, the treatment effect can be summarized by comparing the event risk in the SAPIEN 3 group with the event risk in the SAVR group. The ClinicalTrials.gov record does not provide the treatment-group event counts needed to calculate such an estimate.
Why randomization is statistically important
Randomization creates the framework for a causal comparison because treatment assignment, rather than a patient's clinical characteristics or treatment preference, determines the group to which the patient is assigned. In a sufficiently large randomized trial, baseline prognostic factors tend to be balanced in expectation between treatment groups.
That does not mean randomization guarantees identical groups. Nor does it remove the need to define the analysis population, endpoint, follow-up period, missing-data strategy, and estimand in advance. Those elements determine how the randomized comparison is translated into a statistical estimate.
Why the absence of a formal analysis matters
The empty statistical-analysis record means the ClinicalTrials.gov record does not support a numerical claim about the magnitude or precision of the treatment difference.
This is particularly important for a composite endpoint. Without the actual event counts and the prespecified inferential method, even a seemingly simple calculation could answer a different question from the trial's formal analysis.
6. Planned Analysis
What the registry establishes
All-cause mortality, all stroke, and rehospitalization, defined as the number of patients having any of these events.
The registry reports that results have been posted, but the ClinicalTrials.gov record contains no statistical analyses for the primary endpoint.
| Analysis question | Typical statistical approach | What the ClinicalTrials.gov record permits |
|---|---|---|
| How many patients experienced at least one component by 1 year? | Compare event proportions between randomized groups. | The endpoint definition is available, but treatment-group event counts are not. |
| What is the magnitude of the treatment difference? | Prespecified risk difference, relative risk, odds ratio, or another defined estimand. | No formal effect estimate is reported. |
| How precise is the treatment estimate? | Construct a confidence interval around the prespecified effect measure. | No confidence interval is reported. |
| Is there statistical evidence against the null hypothesis? | Apply the prespecified hypothesis test and type-I-error framework. | No p-value or formal test result is reported. |
7. Statistical Methods Explained
Why is randomization important in PARTNER 3?
Randomization is the central design feature that permits the two interventions to be compared as randomized groups. If randomization is implemented correctly, systematic differences in prognosis are not intentionally assigned to one treatment group. This supports causal interpretation of a prespecified treatment contrast.
Randomization does not guarantee balance in every observed or unobserved variable. Chance imbalance remains possible, which is why baseline characteristics and the prespecified analysis population still matter.
What does a composite endpoint measure?
A composite endpoint combines several clinically relevant outcomes into a single endpoint. In PARTNER 3, the registry defines the primary endpoint as patients who have any of all-cause mortality, all stroke, or the specified rehospitalization event within 1 year.
The main statistical advantage is that more patients may experience at least one component than would experience a particular component alone, potentially increasing the amount of information available. The main interpretive challenge is that the components can have different frequencies and clinical importance.
Why should component results be separated from the composite?
Suppose a composite endpoint changes between randomized groups. That change could be driven mostly by one component while the others show little difference. The composite therefore answers the prespecified combined question; it does not automatically establish the same effect for mortality, stroke, and rehospitalization separately.
For PARTNER 3, the ClinicalTrials.gov record does not contain component-specific statistical estimates. Consequently, no component should be described as driving the primary result on the basis of the registry record.
What would a risk difference mean?
For a fixed 1-year endpoint, a risk difference would subtract the event probability in one randomized group from the event probability in the other. It is an absolute measure, so it answers a different question from a relative measure such as a risk ratio.
The sign and magnitude would depend on the prespecified direction of the comparison and the observed event proportions. No numerical risk difference is reported here because the necessary event counts are absent from the ClinicalTrials.gov record.
Why would a confidence interval be reported?
A point estimate alone does not communicate how precisely the treatment effect has been estimated. A confidence interval provides an interval representation of sampling uncertainty under the statistical model and analysis procedure used.
For example, a narrow interval would generally indicate greater statistical precision than a wide interval, all else equal. But the confidence interval does not describe the range of effects that must occur in individual patients.
Why does the p-value not measure treatment benefit?
A p-value addresses the compatibility of the observed data with a specified null hypothesis under the chosen statistical model. It does not tell us the size of the treatment effect, its clinical importance, or the probability that the treatment is effective.
For a clinical-trial endpoint such as this one, the effect estimate and its confidence interval are therefore essential complements to any formal hypothesis test.
Why does the 1-year time frame matter?
The registry specifies the primary endpoint at 1 year. A result defined at one year is a fixed-horizon statement and should not silently be converted into a longer-term outcome. Longer follow-up can answer additional questions, but it represents a different time frame and potentially a different estimand.
8. Serious Adverse Events
The ClinicalTrials.gov record provides serious adverse-event counts by arm as affected patients divided by patients at risk.
| Safety measure | SAVR | SAPIEN 3 THV |
|---|---|---|
| Serious adverse events, affected / at risk | 221 / 454 | 153 / 496 |
The denominators also differ from the overall enrollment of 1000. That is not inherently contradictory: a safety analysis may use a defined population of patients at risk for a particular outcome. However, the ClinicalTrials.gov record does not provide the population-selection rule, so the 454 and 496 denominators should not be treated as if they were automatically the full randomized groups.
9. Statistical Interpretation of the Available Safety Data
The registry data report 221/454 affected patients for SAVR and 153/496 for SAPIEN 3 THV. These are descriptive affected/at-risk counts.
They do not, by themselves, establish a statistically significant difference, a causal effect estimate, or the relative contribution of individual serious adverse-event categories.
Because the at-risk denominators are different, simply comparing the numerators would be misleading. Statistical comparison of proportions or rates must account for the number of patients contributing risk in each group and must follow the prespecified safety-analysis framework.
A safety comparison needs a clearly defined time window, population, event definition, and statistical method. Without those details, the available counts should remain descriptive rather than being converted into an unreported treatment-effect claim.
10. Missing Data, Censoring, and Analysis Population
The ClinicalTrials.gov record does not specify the missing-data or imputation strategy, censoring rules, analysis population definitions, or estimand used for the primary endpoint.
These details are especially important in clinical trials because a one-year composite endpoint can be affected by incomplete follow-up. For a fixed-horizon binary analysis, investigators need to define how patients with incomplete information at the one-year landmark are handled. For a time-to-event analysis, the treatment of incomplete follow-up is instead expressed through censoring rules.
Neither framework should be assumed from the registry's endpoint wording alone.
Missing data
The ClinicalTrials.gov record does not report an imputation method or a prespecified rule for missing primary-endpoint information.
Censoring
No censoring rules are posted on ClinicalTrials.gov for the primary endpoint.
Analysis population
The ClinicalTrials.gov record does not define an intention-to-treat, per-protocol, modified intention-to-treat, or safety population for the primary endpoint.
Estimand
The registry endpoint identifies what is measured, but the ClinicalTrials.gov record does not specify the formal estimand framework.
11. Multiplicity and Interim Analysis
The ClinicalTrials.gov record does not report an interim-analysis plan, alpha-spending method, multiplicity adjustment, endpoint hierarchy beyond identifying one registered primary endpoint, or a formal type-I-error allocation.
This absence matters because a clinical trial can measure many outcomes while designating only some as confirmatory. A formal p-value cannot be interpreted correctly without knowing which hypothesis was tested, whether the test was prespecified, and how the overall type-I-error rate was controlled.
| Design topic | Information available in the ClinicalTrials.gov record |
|---|---|
| Registered primary endpoints | 1 |
| Primary endpoint | All-cause mortality, all stroke, and rehospitalization at 1 year |
| Outcome measures posted | 6 |
| Interim-analysis method | Not reported |
| Alpha-spending method | Not reported |
| Multiplicity adjustment | Not reported |
| One-sided vs two-sided testing | Not reported |
12. Non-Inferiority and Superiority: What Can Be Said From the Registry Data?
The ClinicalTrials.gov record does not identify the PARTNER 3 hypothesis as a non-inferiority, superiority, equivalence, or other formal testing framework. They also do not provide a non-inferiority margin.
That distinction is important because a randomized device trial may be designed around a specific hypothesis, but the statistical interpretation depends on the prespecified estimand and decision rule. A non-inferiority analysis, for example, requires a clinically justified margin and a confidence-interval rule. A superiority analysis uses a different null hypothesis.
13. Crossover, Factorial Design, and Bayesian Methods
| Topic | What the ClinicalTrials.gov record shows |
|---|---|
| Crossover | Not reported in the ClinicalTrials.gov record. |
| Factorial design | No; the registry identifies a randomized parallel design with 2 arms. |
| Bayesian methods | Not reported in the ClinicalTrials.gov record. |
| Stratification | Not reported in the ClinicalTrials.gov record. |
| Interim analysis | Not reported in the ClinicalTrials.gov record. |
A parallel two-arm randomized design is structurally different from a factorial design because the latter simultaneously evaluates combinations of multiple experimental factors. The registry-reported PARTNER 3 record identifies only two arms and a parallel design; it does not describe multiple crossed treatment factors.
14. Results Availability and Interpretation Boundary
ClinicalTrials.gov results are identified as posted, and the primary endpoint is identified as having results posted. However, the ClinicalTrials.gov recordset contains no statistical analyses.
That creates a useful distinction between an outcome being reported and a formal comparative analysis being available for reconstruction. The first is supported here. The second is not.
| Information requested for a statistical result | registry? |
|---|---|
| Primary endpoint definition | Yes |
| Primary endpoint time frame | Yes — 1 year |
| Results posted indicator | Yes |
| Formal statistical-analysis record | No |
| Treatment-effect estimate | No |
| Confidence interval | No |
| P-value | No |
| Component-specific primary-endpoint estimates | No |
15. What a Complete Primary Analysis Would Normally Contain
For an educational statistical workflow, a complete analysis of the PARTNER 3 primary endpoint would ordinarily make several layers explicit.
1. Analysis population
Identify exactly which randomized patients contribute to the primary analysis and how deviations from the randomized population are handled.
2. Endpoint construction
Define whether a patient is counted once after the first qualifying component and specify the assessment window.
3. Treatment effect
Report the prespecified absolute or relative measure of treatment effect, with its direction defined in advance.
4. Uncertainty
Provide a confidence interval or other prespecified uncertainty measure around the treatment-effect estimate.
5. Hypothesis test
Report the prespecified test and p-value only in the context of its type-I-error framework.
6. Component analysis
Describe mortality, stroke, and rehospitalization separately when those component results are available, without treating them as interchangeable outcomes.
The registry-reported PARTNER 3 dataset supports the endpoint definition and overall trial design but does not provide the remaining numerical components needed to reproduce such an analysis.
16. Why the Safety Denominators Need Careful Reading
The serious-adverse-event information is reported as affected / at risk, with 221/454 for SAVR and 153/496 for SAPIEN 3 THV. This format communicates both the number of affected patients and the population against which those patients are counted.
A denominator is not a minor reporting detail. It determines the scale against which an event count should be interpreted. Two numerators cannot be compared directly when their populations at risk differ.
At the same time, a simple proportion is not necessarily the final analysis. Safety outcomes may be defined over specific observation periods, may involve recurrent events, or may be subject to competing risks or exposure differences. The ClinicalTrials.gov record does not specify which framework was used for these serious-adverse-event results.
The registry supplies the numerator and denominator for each arm. The appropriate inferential comparison, however, requires the underlying safety-analysis definition and prespecified statistical method.
17. Limitations
- No formal statistical analyses reported: no formal statistical analyses were posted, so no treatment-effect estimate, confidence interval, or p-value can be reported from the ClinicalTrials.gov recordset.
- Composite endpoint: the primary endpoint combines all-cause mortality, all stroke, and specified rehospitalization. The clinical meaning of the composite cannot be reduced to a single component without component-specific evidence.
- Component results unavailable: the ClinicalTrials.gov record does not provide separate numerical results for mortality, stroke, or rehospitalization.
- Analysis population unavailable: the ClinicalTrials.gov record does not define the primary efficacy population or explain exclusions from the overall enrollment.
- Missing-data strategy unavailable: no imputation, missing-endpoint, or incomplete-follow-up rule is reported.
- Testing framework unavailable: the ClinicalTrials.gov record does not identify the formal null hypothesis, alpha level, sidedness, non-inferiority margin, or multiplicity strategy.
- Interim-analysis information unavailable: no interim-analysis or alpha-spending approach is provided.
- Safety definitions incomplete: serious adverse events are reported as affected/at-risk counts, but the event definition and statistical-analysis framework are not included.
- Denominator differences: the serious-adverse-event denominators are 454 and 496, rather than the overall enrollment of 1000. The ClinicalTrials.gov record does not explain the population-selection rule behind these denominators.
- External reconstruction is intentionally excluded: although linked publications are identified in the registry data, the numerical trial results in those publications are not incorporated here because the registry-reported task specifies that trial numbers must come only from the ClinicalTrials.gov record.
18. Why This Trial Matters Statistically
PARTNER 3 is a useful teaching case because it illustrates an important feature of clinical-trial statistics: the scientific question, endpoint definition, randomization, analysis population, estimand, and inferential procedure form a chain. If one link is missing, a numerical result cannot necessarily be reconstructed reliably from the remaining information.
| Concept | How it appears in PARTNER 3 |
|---|---|
| Randomization | The registry identifies a randomized allocation. |
| Parallel-group design | The registry identifies a parallel design with 2 arms. |
| Composite endpoint | The primary endpoint combines all-cause mortality, all stroke, and specified rehospitalization. |
| Fixed time frame | The registered primary endpoint is assessed at 1 year. |
| Binary patient-level outcome | The endpoint definition counts patients having any of the specified events. |
| Safety analysis | Serious adverse events are reported as affected/at-risk counts by arm. |
| Missing statistical analysis | No formal statistical-analysis entries are reported despite the results-posted indicator. |
| Analysis population | The primary analysis population is not specified in the ClinicalTrials.gov record. |
| Multiplicity | One registered primary endpoint is identified, but the multiplicity framework is not reported. |
| Non-inferiority framework | No non-inferiority margin or testing framework is reported. |
| Bayesian methods | No Bayesian analysis is reported in the ClinicalTrials.gov record. |
19. Statistical Methods Explained: A Deeper Walkthrough
Why is a composite endpoint attractive in trial design?
A composite endpoint can increase the number of patients reaching the primary endpoint when several clinically relevant outcomes are combined. More observed events can provide greater statistical information than a rarer individual component alone.
But the gain in event frequency comes with an interpretive cost. If one component is much more common than another, the overall composite may be driven disproportionately by that component. Therefore, the components should be examined individually whenever their results are available.
Why is the phrase "any of these events" statistically important?
It establishes the patient-level nature of the registered endpoint. The patient is counted as having the composite event if at least one qualifying component occurs. A patient with multiple components is not automatically counted multiple times in the primary patient-level event count.
This distinction separates a composite binary endpoint from an analysis of recurrent events or a count of all events.
Could the primary endpoint be analyzed with a Kaplan-Meier method?
Possibly, but the registry definition alone does not establish that this was the prespecified method. If investigators define the endpoint as time to first occurrence of any component, a time-to-event analysis can be appropriate. If the analysis is instead a fixed 1-year binary endpoint, an analysis of the proportion of patients with an event by 1 year is a natural framework.
The exact method must come from the trial's statistical analysis plan rather than being inferred solely from the endpoint name.
Why is intention-to-treat important?
In a randomized trial, intention-to-treat analysis generally preserves the comparison created by randomization by analyzing patients according to their assigned treatment. However, the registry-reported PARTNER 3 data do not state that an intention-to-treat population was used for the primary endpoint, so this page does not label it as the documented primary analysis.
Why can safety analysis populations differ from efficacy populations?
Safety is often tied to actual exposure to an intervention, while efficacy is commonly anchored to randomized assignment. Those are different scientific questions. A safety analysis may therefore use an at-risk or treated population rather than exactly the same denominator used for efficacy.
For PARTNER 3, the registry-reported serious-adverse-event denominators of 454 and 496 demonstrate that the reported safety information uses defined at-risk populations, but the ClinicalTrials.gov record does not provide their precise population definitions.
Why should statistical significance not be inferred from event counts?
Two groups can have different observed event proportions without the difference meeting a prespecified statistical threshold. Conversely, a statistically detectable difference may be clinically small. Statistical inference requires the sample size, endpoint definition, analysis method, null hypothesis, and uncertainty calculation.
Because the ClinicalTrials.gov record omit the formal analysis, the appropriate conclusion is descriptive rather than inferential.
20. Overall Statistical Interpretation
PARTNER 3 is registered as a randomized, parallel-group, unmasked cardiovascular trial with 1000 enrolled participants, two arms, and a primary 1-year composite endpoint involving all-cause mortality, all stroke, and specified rehospitalization.
The ClinicalTrials.gov record provides serious-adverse-event counts of 221/454 for SAVR and 153/496 for SAPIEN 3 THV. They also identify the trial as having posted results.
No formal primary-endpoint treatment effect, confidence interval, or p-value is reported. No component-specific efficacy results are reported either.
A rigorous statistical analysis should distinguish a registry's statement that results exist from the actual numerical evidence needed to reproduce an inferential comparison. Filling the missing fields from memory or from an external publication would violate the source boundary for this analysis.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: PARTNER 3 — NCT02675114.
- PubMed: Publication record — PMID 42524810.
- PubMed: Publication record — PMID 42508852.
- PubMed: Publication record — PMID 42340728.
- PubMed: Publication record — PMID 41144631.
- PubMed: Publication record — PMID 40243974.
Continue through the Clinical Biostats knowledge graph
Connect randomized-trial design and endpoint concepts with deeper statistical tutorials and practical analysis tools.
24. Record Summary
PARTNER 3 provides a useful statistical teaching example because its registry record clearly establishes the randomized parallel-group design, the two interventions, the 1000-participant enrollment, and the 1-year composite primary endpoint, while the ClinicalTrials.gov record does not provide the numerical treatment-effect analysis needed to reproduce a formal efficacy comparison.
The primary endpoint is defined as the number of patients who had any of all-cause mortality, all stroke, and specified rehospitalization at 1 year. That definition illustrates why endpoint construction matters: a composite produces a single patient-level outcome while combining events that can have different frequencies and clinical meanings.
The safety data provide affected/at-risk counts of 221/454 for SAVR and 153/496 for SAPIEN 3 THV. These figures are useful descriptively, but their proper inferential interpretation requires the safety population definition, event definition, follow-up framework, and statistical method.
The central statistical lesson is therefore methodological: