This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
COAPT was a randomized, parallel-group, unmasked clinical trial in cardiology evaluating the MitraClip System in the setting of symptomatic heart failure and functional mitral regurgitation. The ClinicalTrials.gov record reports 776 participants, 3 arms, 2 registered primary endpoints, 338 posted outcome measures, and 3 posted statistical analyses.
| Feature | COAPT |
|---|---|
| Trial | Cardiovascular Outcomes Assessment of the MitraClip Percutaneous Therapy for Heart Failure Patients With Functional Mitral Regurgitation (The COAPT Trial) and COAPT CAS |
| NCT ID | NCT01626079 |
| Therapeutic area | Cardiology |
| Conditions | Mitral Regurgitation; Mitral Valve Regurgitation; Treatment of Functional Mitral Regurgitation in Symptomatic Heart Failure Subjects; Heart Failure |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Intervention | MitraClip System (device) |
| Lead sponsor | Abbott Medical Devices |
| Sponsor type | Industry |
| Results posted | Yes |
2. Clinical Question
The registry describes a randomized comparison centered on the use of the MitraClip System in symptomatic heart failure with functional mitral regurgitation. The registered primary endpoints address both device-related safety and recurrent heart-failure hospitalization.
Population
Participants in the COAPT trial with conditions including mitral regurgitation, functional mitral regurgitation in symptomatic heart failure, and heart failure.
Intervention
MitraClip System (device).
Comparator
The ClinicalTrials.gov record identifies a Control Group for the comparative analyses.
Primary questions
Does the device meet the prespecified primary safety performance goal, and how does the randomized comparison affect recurrent heart-failure hospitalization through 24 months?
3. Trial Design
4. Registered Primary Endpoints
| Endpoint | Time frame | Registry definition | Endpoint type |
|---|---|---|---|
| Primary Safety Endpoint - Percentage of Participants With Freedom From Device Related Complications at 12 Months | 12 months | Percentage of Participants with Freedom from Device related Complications at 12 Months. Composite of Single Leaflet Device Attachment (SLDA), device embolizations, endocarditis requiring surgery, Echocardiography Core Laboratory confirmed mitral stenosis requiring surgery, LVAD implant, heart transplant, and any device related complications requiring non-elective cardiovascular surgery. | Binary |
| Primary Effectiveness Endpoint | 24 months | Recurrent HF hospitalizations (HFH) through 24 months, analyzed when the last subject completes 12-month follow-up. | Other / unclear |
The two primary endpoints have different statistical structures. The safety endpoint is defined as freedom from a composite of device-related complications and was analyzed using a Kaplan-Meier survival estimate. The effectiveness endpoint is a recurrent-event outcome: the same participant can contribute more than one hospitalization, so a conventional analysis that counts only the first event would not fully represent the registered endpoint.
5. Primary Safety Analysis
The registry reports a formal statistical analysis for the primary safety endpoint among subjects randomized with an attempted procedure to the device group and with available data. The performance goal was 88%.
Freedom from device-related complications at 12 months
95% one-sided CI: 0.948 · P < 0.0001
Hypothesis type: non-inferiority
| Feature | Reported analysis |
|---|---|
| Analysis population | Subjects randomized with attempted procedure to the device group with available data |
| Performance goal | 88% |
| Method | Z test using Kaplan-Meier survival estimate |
| Variance estimation | Greenwood method estimated variance |
| Estimate | 0.966 |
| Confidence interval | 95% one-sided CI: 0.948 |
| P-value | <0.0001 |
| Hypothesis | Non-inferiority |
The Kaplan-Meier estimate of 0.966 represents the estimated freedom from the specified device-related complication composite at 12 months in the analyzed device-group population. Because the endpoint is defined as freedom from complications, an estimate closer to 1 indicates a larger estimated proportion remaining free of the specified event over the analysis period.
The result should not be interpreted as saying that every participant had the same probability of remaining complication-free, nor does it identify which component of the composite drove the estimate. The composite includes several clinically different events, ranging from device-related complications requiring non-elective cardiovascular surgery to events such as LVAD implantation or heart transplantation.
The one-sided 95% confidence bound of 0.948 describes the lower confidence limit for the estimated survival probability under the specified statistical framework. It is a statement about statistical uncertainty around the population-level estimate, not a prediction interval for individual patients.
The P < 0.0001 value addresses the hypothesis test; it does not quantify the magnitude or clinical importance of the observed estimate. Here, the relevant design logic is non-inferiority: the comparison is against the prespecified performance goal rather than against a conventional two-sided null hypothesis of no treatment difference.
The analysis is also dependent on the handling of censoring because Kaplan-Meier estimation is a time-to-event method. The ClinicalTrials.gov record does not provide enough detail to independently assess every censoring rule or the contribution of each component of the composite endpoint.
Why Kaplan-Meier was used for a percentage endpoint
Although the registry describes the primary safety endpoint as a percentage of participants, the formal analysis treats it as a time-to-event endpoint. That distinction matters. A participant who has not experienced a device-related complication by the time of last assessment can contribute follow-up information without necessarily having completed the full observation period. Kaplan-Meier estimation is designed for this setting because it accommodates right-censored observations.
At each event time, the estimated survival function updates according to the number of events and the number of participants still at risk immediately before that time.
6. Non-Inferiority Logic for the Safety Endpoint
The primary safety analysis is explicitly classified as non-inferiority. The ClinicalTrials.gov record states that the performance goal was 88% and that two thousand simulations were performed for sample-size and power calculations. Under the stated assumptions of 22% mortality and 7.5% attrition at 12 months, the registry states that a total of 305 subjects in the Device group would provide > 95% power to reject the null hypothesis at the one-sided significance level.
What is being tested?
The device-group Kaplan-Meier estimate of freedom from device-related complications is evaluated against a prespecified performance goal rather than a randomized control-group estimate.
Why one-sided?
Non-inferiority is directional: the design asks whether performance is sufficiently high relative to the prespecified standard. The registry-reported analysis reports a one-sided confidence interval.
Why the margin matters
The performance goal defines the benchmark against which the safety endpoint is evaluated. The inferential conclusion depends on that prespecified benchmark, not simply on whether a P-value is small.
Simulation-based planning
The registry states that two thousand simulations were performed to calculate sample size and power for the primary safety endpoint.
7. Primary Effectiveness Endpoint: Recurrent Heart-Failure Hospitalizations
The registered primary effectiveness endpoint is recurrent HF hospitalizations through 24 months, analyzed when the last subject completes 12-month follow-up.
How this endpoint would normally be analyzed
Because the endpoint is explicitly recurrent hospitalization, a participant may experience multiple HF hospitalizations during follow-up. Appropriate methods therefore need to account for repeated events within the same participant rather than treating every hospitalization as statistically independent.
For recurrent-event outcomes, possible approaches include recurrent-event models, joint frailty models, or other methods that explicitly represent within-participant dependence. The ClinicalTrials.gov record identifies a Joint Frailty Model for the secondary all-cause recurrent-hospitalization analysis, illustrating the type of framework used elsewhere in the record.
The registry wording also makes the time horizon important: the endpoint concerns recurrent HF hospitalizations through 24 months, with analysis tied to the specified follow-up milestone. The time window should therefore not be replaced by an unspecified "hospitalization rate" or by a first-hospitalization endpoint.
8. Secondary Results: Recurrent All-Cause Hospitalizations
The ClinicalTrials.gov record reports a formal secondary analysis of Recurrent Hospitalizations - All Cause through 24 months. This is distinct from the registered primary effectiveness endpoint because the secondary outcome includes all-cause hospitalizations.
Recurrent all-cause hospitalizations
95% CI: 0.60–0.96 · P < 0.02
Superiority hypothesis · MitraClip System vs Control Group
| Feature | Reported analysis |
|---|---|
| Outcome | Recurrent Hospitalizations - All Cause |
| Time frame | 24 Months |
| Population | All randomized subjects in Device and Control Group |
| Groups compared | MitraClip System vs Control Group |
| Method | Joint Fraility Model |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.76 |
| 95% CI | 0.60–0.96 |
| P-value | <0.02 |
| Hypothesis | Superiority |
The reported hazard ratio of 0.76 indicates a lower estimated recurrent-event hazard for the MitraClip System group relative to the Control Group under the reported joint-frailty modeling framework. Expressed as a simple relative interpretation, 0.76 corresponds to an estimated 24% lower hazard.
This does not mean that each participant had exactly 24% fewer hospitalizations, nor does it mean that 24% of participants avoided hospitalization. A recurrent-event hazard ratio summarizes a model-based comparison of event processes over follow-up.
The 95% confidence interval of 0.60–0.96 describes uncertainty around the estimated hazard ratio. Its width reflects the precision of the estimate; it does not describe the range of hospitalization counts that individual participants might experience.
The P < 0.02 value addresses the statistical test under the reported superiority framework. It is not a measure of effect size, clinical importance, or probability that the treatment "works."
Because recurrent hospitalizations can occur repeatedly within the same participant, the joint-frailty framework is important. A simple first-event Kaplan-Meier analysis would discard information from later hospitalizations and would not represent the registered recurrent-event structure in the same way.
9. Secondary Results: Death or Heart-Failure Hospitalization
The registry also reports a Finkelstein-Schoenfeld analysis of all-cause death or recurrent HF hospitalization through 24 months. The reported effect measure is a win ratio.
Finkelstein-Schoenfeld win ratio
95% CI: 1.29–2.04 · P < 0.0001
Superiority hypothesis · Randomized Group
| Feature | Reported analysis |
|---|---|
| Outcome | Death or HF Hospitalization Within 24 Months (Finkelstein-Schoenfeld Analysis of All-Cause Death or Recurrent HF Hospitalization Through 24 Months) |
| Time frame | 24 months |
| Outcome unit | Win Ratio |
| Groups compared | Randomized Group |
| Method | Finkelstein-Schoenfeld Analysis |
| Effect measure | Win Ratio |
| Estimate | 1.61 |
| 95% CI | 1.29–2.04 |
| P-value | <0.0001 |
| Hypothesis | Superiority |
A win ratio of 1.61 means that, under the Finkelstein-Schoenfeld comparison rule used for this analysis, the number of favorable pairwise comparisons was estimated to be 1.61 times the number of unfavorable comparisons. The interpretation depends on the hierarchical comparison procedure and therefore differs fundamentally from interpreting a hazard ratio.
The win ratio does not mean that 61% of participants benefited, that mortality was reduced by 61%, or that hospitalization risk was reduced by 61%. It is a summary of pairwise treatment comparisons under the specified endpoint hierarchy.
The 95% confidence interval of 1.29–2.04 describes uncertainty around the estimated win ratio. It does not describe the distribution of individual participant outcomes.
The P < 0.0001 value addresses the statistical comparison; it does not measure the magnitude of clinical benefit. The effect size is represented by the win ratio itself and should be interpreted alongside the endpoint definition and comparison hierarchy.
10. Statistical Methodology
Kaplan-Meier estimation
The primary safety analysis used a Kaplan-Meier survival estimate. Kaplan-Meier methods are useful when the timing of an event matters and some participants are censored before the end of follow-up.
For COAPT's safety endpoint, the event is a component of the prespecified device-related complication composite. Participants who remain free of the event contribute follow-up until their event, censoring time, or the relevant analysis time.
Greenwood variance estimation
The registry analysis states that the P-value was calculated from a Z test using the Kaplan-Meier survival estimate together with the Greenwood method estimated variance. Greenwood's formula provides an estimated variance for the Kaplan-Meier survival probability and is commonly used to quantify uncertainty around a survival estimate.
Joint frailty model
The secondary recurrent all-cause hospitalization analysis used a Joint Fraility Model, as spelled in the registry method field. A joint-frailty approach is designed for settings where recurrent events within the same participant are correlated and where a terminal event may also be relevant to the event process.
The key statistical issue is dependence: if one participant experiences several hospitalizations, those events are not independent observations. A model that recognizes participant-level dependence can use the repeated-event structure more appropriately than a method that treats every hospitalization as an independent observation.
Finkelstein-Schoenfeld analysis
The death-or-recurrent-HF-hospitalization endpoint was analyzed using a Finkelstein-Schoenfeld analysis with a win ratio as the effect measure. This class of method compares participants pairwise using a prespecified hierarchy of clinically ordered outcomes.
The numerical value depends on the comparison hierarchy and the rules used to determine whether one participant "wins," "loses," or is tied in a pairwise comparison.
Non-inferiority testing
The primary safety analysis was explicitly identified as a non-inferiority analysis. Its statistical benchmark was a performance goal of 88%, and the reported confidence interval was one-sided.
Superiority testing
The reported recurrent-hospitalization and win-ratio secondary analyses were classified as superiority analyses. These analyses therefore ask whether the randomized groups differ in the favorable direction rather than whether one group remains above a predefined performance standard.
11. Statistical Methods Explained
Why was Kaplan-Meier estimation used for the primary safety endpoint?
The safety endpoint is defined at 12 months, but the underlying event is time-dependent. Participants can experience a device-related complication before 12 months or remain event-free until they are censored. Kaplan-Meier estimation uses the timing of events and censoring rather than treating every participant as if they had identical follow-up.
Why is the primary safety analysis non-inferiority rather than superiority?
The registry analysis evaluates the device-group safety estimate against a prespecified performance goal of 88%. That is a non-inferiority framework: the question is whether the observed safety performance is sufficiently high relative to the benchmark. It is conceptually different from asking whether one randomized treatment group has a better outcome than another.
What does the estimate 0.966 mean?
It is the reported Kaplan-Meier estimate for freedom from the specified device-related complication composite at 12 months. It should be interpreted as an estimated event-free proportion under the survival-analysis framework, not as an individual-level guarantee and not as a measure of the frequency of each separate component of the composite.
Why was a joint frailty model used for recurrent hospitalizations?
Recurrent hospitalization data contain multiple observations from the same participant. Those observations are correlated. A frailty component can represent unobserved participant-level susceptibility and can therefore account for within-participant dependence that would be missed by treating each hospitalization as an independent event.
What does a hazard ratio of 0.76 mean for recurrent hospitalization?
Under the reported joint-frailty model, the MitraClip System group had an estimated recurrent-event hazard of 0.76 relative to the Control Group. The simple relative interpretation is a 24% lower estimated hazard. It does not mean 24% fewer participants were hospitalized or that every participant experienced a 24% reduction.
What does a win ratio of 1.61 mean?
A win ratio of 1.61 means the favorable pairwise comparisons were estimated to outnumber unfavorable comparisons by a ratio of 1.61 under the Finkelstein-Schoenfeld rules. It is not interchangeable with a hazard ratio, relative risk, odds ratio, or percentage of participants who benefited.
Why should the P-value not be treated as the effect size?
A P-value describes evidence against a specified statistical null under the analysis framework. It depends on both the observed data and the amount of information available. The effect estimate, confidence interval, endpoint definition, and analysis population provide the information needed to understand the size and precision of an observed treatment comparison.
12. Primary Safety Endpoint: Composite-Endpoint Considerations
The primary safety endpoint combines several distinct device-related complications: Single Leaflet Device Attachment (SLDA), device embolizations, endocarditis requiring surgery, Echocardiography Core Laboratory confirmed mitral stenosis requiring surgery, LVAD implant, heart transplant, and device-related complications requiring non-elective cardiovascular surgery.
| Composite component | Included in registered safety endpoint |
|---|---|
| Single Leaflet Device Attachment (SLDA) | Yes |
| Device embolizations | Yes |
| Endocarditis requiring surgery | Yes |
| Echocardiography Core Laboratory confirmed mitral stenosis requiring surgery | Yes |
| LVAD implant | Yes |
| Heart transplant | Yes |
| Device related complications requiring non-elective cardiovascular surgery | Yes |
A composite endpoint can increase the number of observed events and thereby improve statistical information, but the components may differ substantially in clinical meaning and frequency. The reported estimate therefore applies to the composite endpoint as defined; it should not be interpreted as a separate estimate for every component.
13. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized arm using affected participants over participants at risk.
| Arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| MitraClip System | 281 | 302 |
| Control Group | 291 | 312 |
The registry data provide affected and at-risk counts for serious adverse events but do not provide a formal between-group statistical analysis for this safety measure in the ClinicalTrials.gov record. Accordingly, this page reports the counts without assigning a comparative P-value, confidence interval, or effect estimate.
14. Randomization and Analysis Populations
Randomization is the principal design feature that supports causal comparison between the randomized groups. In the statistical analyses posted on ClinicalTrials.gov, the recurrent all-cause hospitalization analysis explicitly uses all randomized subjects in the Device and Control Group.
| Analysis | Analysis population reported in the ClinicalTrials.gov record |
|---|---|
| Primary safety endpoint | Subjects randomized with attempted procedure to the device group with available data |
| Recurrent all-cause hospitalization | All randomized subjects in Device and Control Group |
| Death or recurrent HF hospitalization | Randomized Group |
| Serious adverse events | Affected / at-risk counts registry-reported separately by arm |
These population definitions are not interchangeable. In particular, the primary safety analysis is based on the device-group population meeting the registry-reported analysis criteria, whereas the recurrent hospitalization analysis explicitly uses randomized participants in the comparative groups.
15. Multiplicity and Endpoint Hierarchy
The registry identifies 2 primary endpoints and reports statistical analyses classified as non-inferiority and superiority. However, the ClinicalTrials.gov record does not provide a complete multiplicity-adjustment strategy for all primary and secondary hypotheses.
| Endpoint / analysis | Role | Hypothesis type | Reported formal analysis |
|---|---|---|---|
| Freedom from device-related complications at 12 months | Primary safety | Non-inferiority | Yes |
| Recurrent HF hospitalizations through 24 months | Primary effectiveness | Not reported in statistical-analysis record | No formal analysis reported |
| Recurrent all-cause hospitalizations through 24 months | Secondary | Superiority | Yes |
| Death or recurrent HF hospitalization through 24 months | Secondary | Superiority | Yes |
Because the ClinicalTrials.gov record does not describe a complete multiplicity procedure, the individual reported P-values should be interpreted in the context of their endpoint roles rather than assumed to represent a single uniformly adjusted family of hypotheses.
16. Crossover, Interim Analysis, and Missing Data
Crossover
The ClinicalTrials.gov record does not report a crossover design or crossover results. No crossover effect is inferred.
Factorial design
The trial is described as a parallel design. The ClinicalTrials.gov record does not identify a factorial structure.
Interim analysis
The ClinicalTrials.gov record does not provide an interim-analysis plan or interim efficacy boundary.
Missing data / imputation
The ClinicalTrials.gov record does not specify an imputation method for missing observations. The primary safety analysis does specify participants with available data.
This distinction is important for statistical interpretation. The absence of a registry-reported method description does not justify assuming that a particular imputation, censoring, or interim-monitoring strategy was used.
17. Bayesian Methods and Stratification
The registry's posted analyses use a z-test based on Kaplan-Meier survival estimates, a joint frailty model, and the Finkelstein-Schoenfeld (win ratio) analysis. No Bayesian method is identified in the ClinicalTrials.gov record.
Likewise, the ClinicalTrials.gov record does not identify randomization stratification factors. No stratified analysis variables are therefore added to this page.
| Methodological topic | What the ClinicalTrials.gov record supports |
|---|---|
| Kaplan-Meier estimation | Used for the primary safety endpoint. |
| Greenwood variance | Used for the primary safety endpoint's Kaplan-Meier estimate. |
| Joint frailty model | Used for recurrent all-cause hospitalization. |
| Finkelstein-Schoenfeld analysis | Used for death or recurrent HF hospitalization. |
| Win ratio | Reported as the effect measure for the Finkelstein-Schoenfeld analysis. |
| Bayesian analysis | Not identified in the ClinicalTrials.gov record. |
| Stratification factors | Not identified in the ClinicalTrials.gov record. |
18. Understanding the Two Main Effect Measures
Hazard ratio
The recurrent all-cause hospitalization analysis reports a hazard ratio of 0.76. A hazard ratio compares event rates over time under the specified model. A value below 1 indicates a lower estimated hazard in the numerator group relative to the reference group.
Thus, 0.76 can be described as an estimated 24% lower hazard under the reported model. This arithmetic does not mean 24% fewer participants experienced hospitalization.
Win ratio
The death-or-recurrent-HF-hospitalization analysis reports a win ratio of 1.61. Unlike a hazard ratio, the win ratio is constructed from pairwise comparisons under the Finkelstein-Schoenfeld framework.
The two estimates should therefore not be compared as though they were measurements on the same scale. A hazard ratio describes a model-based relative event hazard, whereas a win ratio describes the relative number of favorable versus unfavorable pairwise comparisons under the prespecified comparison rules.
19. Confidence Intervals and P-values
The primary safety endpoint reports a 95% one-sided confidence interval with lower bound 0.948. The recurrent all-cause hospitalization analysis reports a 95% two-sided confidence interval of 0.60–0.96, while the win-ratio analysis reports a 95% two-sided confidence interval of 1.29–2.04.
These intervals describe statistical uncertainty around their respective estimates. They do not describe the range of outcomes that an individual patient may experience.
The primary safety analysis reports P < 0.0001, the recurrent all-cause hospitalization analysis reports P < 0.02, and the win-ratio analysis reports P < 0.0001. A P-value quantifies evidence against a specified null hypothesis under the relevant statistical model; it does not measure effect size or clinical importance.
The primary safety endpoint uses a one-sided confidence interval because its formal hypothesis is non-inferiority. The two reported secondary comparative analyses use two-sided 95% confidence intervals and are classified as superiority analyses.
20. Limitations
- Incomplete primary-effectiveness statistics: the ClinicalTrials.gov record identifies recurrent HF hospitalizations through 24 months as a primary endpoint, but it does not supply a formal statistical analysis for that endpoint.
- Composite safety endpoint: the primary safety result combines several clinically different device-related complications. The ClinicalTrials.gov record does not provide component-specific estimates.
- Analysis-population differences: the primary safety analysis uses subjects with an attempted device procedure and available data, while the recurrent hospitalization analysis uses all randomized subjects in the Device and Control Group.
- Recurrent-event interpretation: recurrent hospitalization analyses require methods that account for repeated events within participants; the reported joint-frailty model is therefore not directly interchangeable with a first-event Kaplan-Meier analysis.
- Win-ratio interpretation: the win ratio depends on the pairwise comparison framework and endpoint hierarchy. It should not be translated directly into a hazard ratio or percentage treatment effect.
- Multiplicity: the ClinicalTrials.gov record does not provide a complete multiplicity-adjustment strategy across all primary and secondary analyses.
- Missing-data details: the ClinicalTrials.gov record does not specify a complete imputation strategy or all censoring rules.
- Stratification details: the registry-reported statistical records do not identify stratification factors, so none are inferred.
- Registry scope: the ClinicalTrials.gov recordset contains 338 posted outcome measures but only 3 statistical analyses in the statistical-analysis field. This page therefore focuses on the analyses explicitly provided rather than reconstructing unreported statistical results.
21. Why This Trial Matters Statistically
COAPT is a useful statistical teaching case because the ClinicalTrials.gov record places several distinct analysis frameworks in the same randomized trial. The primary safety endpoint uses Kaplan-Meier estimation and a non-inferiority framework; recurrent hospitalization uses a joint-frailty model; and the combined death/recurrent-HF-hospitalization endpoint uses a Finkelstein-Schoenfeld win ratio.
| Concept | How it appears in COAPT |
|---|---|
| Randomization | Randomized parallel-group design with 776 enrolled participants. |
| Kaplan-Meier estimation | Primary safety endpoint at 12 months. |
| Non-inferiority | Primary safety hypothesis against a performance goal of 88%. |
| One-sided confidence interval | Primary safety analysis reports a 95% one-sided lower confidence bound of 0.948. |
| Greenwood variance | Used with the Kaplan-Meier survival estimate for the primary safety analysis. |
| Recurrent-event analysis | Recurrent all-cause hospitalization analyzed with a joint-frailty model. |
| Hazard ratio | Reported as 0.76 for recurrent all-cause hospitalizations. |
| Win ratio | Reported as 1.61 for death or recurrent HF hospitalization. |
| Finkelstein-Schoenfeld analysis | Used for the death-or-recurrent-HF-hospitalization endpoint. |
| Superiority | Reported hypothesis type for the two secondary statistical analyses. |
| Composite endpoint | Primary safety endpoint combines multiple device-related complications. |
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Statistical Calculators
24. Sources
- ClinicalTrials.gov: NCT01626079 — COAPT.
- PubMed record: PMID 40357542.
- PubMed record: PMID 39290680.
- PubMed record: PMID 38795108.
- PubMed record: PMID 38788821.
- PubMed record: PMID 38060997.
Continue through Clinical Biostats
Connect this trial's endpoints and statistical methods with deeper tutorials and practical statistical calculators.
25. Record Summary
COAPT provides a compact example of how different clinical-trial questions can require different statistical frameworks. Its primary safety endpoint uses a Kaplan-Meier estimate with Greenwood variance and a one-sided non-inferiority framework against a performance goal of 88%. Its recurrent all-cause hospitalization analysis uses a joint-frailty model and reports a hazard ratio of 0.76, while the death-or-recurrent-HF-hospitalization analysis uses a Finkelstein-Schoenfeld framework and reports a win ratio of 1.61.
The most important statistical lesson is that these estimates are not interchangeable. The Kaplan-Meier estimate describes freedom from a time-to-event composite, the hazard ratio summarizes a model-based relative event hazard for recurrent hospitalization, and the win ratio summarizes favorable versus unfavorable pairwise comparisons under a specified hierarchy. Each requires its own interpretation of the endpoint definition, analysis population, uncertainty interval, and hypothesis-testing framework.