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
CLASP IID is a randomized, parallel-design clinical trial evaluating the Edwards PASCAL System and Abbott MitraClip System in patients with mitral regurgitation. The registry identifies three primary endpoints spanning 30-day safety, 6-month mitral-regurgitation severity reduction, and time to first heart failure hospitalization or death through 5-year follow-up for the CLASP IIF Cohort.
| Feature | CLASP IID |
|---|---|
| Trial | CLASP IID |
| ClinicalTrials.gov identifier | NCT03706833 |
| Registry title | Edwards PASCAL CLASP IID/IIF Pivotal Clinical Trial |
| Status | Active, not recruiting |
| Start date | 2018-11-30 |
| Primary completion date | 2026-06-30 |
| Enrollment | 1247 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Lead sponsor | Edwards Lifesciences |
| Sponsor type | Industry |
2. Clinical Question
The central statistical question is whether the Edwards PASCAL System is not inferior to the Abbott MitraClip System on the registry-defined primary safety endpoint, on mitral-regurgitation severity reduction at 6 months in the CLASP IID Cohort, and on time to first heart failure hospitalization or death in the CLASP IIF Cohort.
Population
The registry lists degenerative mitral valve disease, mitral regurgitation, mitral insufficiency, and functional mitral regurgitation as the conditions under study.
Intervention
Edwards PASCAL System, a device intervention.
Comparator
Abbott MitraClip System, a device intervention.
Primary question
Does PASCAL meet the prespecified non-inferiority objective relative to MitraClip across the applicable primary endpoints?
3. Trial Design
Edwards PASCAL System
- Device intervention.
- Evaluated against the Abbott MitraClip System.
- Included in the randomized parallel trial design.
Abbott MitraClip System
- Device comparator.
- Serves as the comparison treatment for the non-inferiority questions.
- Included in the randomized parallel trial design.
The registry identifies 5 arms, but the information available here names two device interventions rather than providing arm-level enrollment or treatment assignments for all five arms. The statistical interpretation therefore focuses on the two named devices and the endpoint-specific cohort structure without assigning unreported sample sizes to individual arms.
4. Endpoints
The registry identifies three primary endpoints. They are not interchangeable: the first is a short-term safety composite, the second concerns echocardiographic mitral-regurgitation severity reduction, and the third is a time-to-event endpoint restricted to the CLASP IIF Cohort.
| Primary endpoint | Registry definition | Time frame |
|---|---|---|
| Primary safety | PASCAL is not inferior to MitraClip with respect to the proportion of patients with major adverse events (MAE). The primary safety endpoint is a composite of Major Adverse Events (MAEs). | 30 days. |
| Mitral-regurgitation reduction | PASCAL is not inferior to Mitraclip with respect to the proportion of patients with MR severity reduction as measured by echocardiography using a scale of 0-4+ for the CLASP IID Cohort. | 6 months |
| Heart failure hospitalization or death | PASCAL is not inferior to Mitraclip with respect to the time to first heart failure hospitalization or death for the CLASP IIF Cohort only. | From date of randomization until date of first heart failure hospitalization or death, through 5 year follow-up |
5. Non-Inferiority Framework
The registry explicitly frames all three primary questions as non-inferiority comparisons. This changes the statistical question from simply asking whether the two devices differ to asking whether the observed evidence is compatible with PASCAL retaining an acceptable level of performance relative to MitraClip.
For a non-inferiority trial, the critical reference is the prespecified margin. The direction and scale of the margin depend on how the endpoint is defined. A confidence interval is then compared with that margin to determine whether the non-inferiority criterion has been met.
The registry record identifies the endpoints as non-inferiority objectives but does not report the numerical non-inferiority margin in the information summarized here. That margin is therefore essential to any final numerical non-inferiority conclusion and cannot be substituted with an arbitrary threshold.
Why the p-value is not the central non-inferiority criterion
A conventional superiority test asks whether the evidence is inconsistent with an equality or null-effect hypothesis. Non-inferiority instead requires assessment against a clinically specified boundary. A small p-value by itself does not establish that a treatment is acceptably close to the comparator, just as a nonsignificant superiority test does not automatically establish non-inferiority.
Superiority question
Is there evidence that the treatment effect differs from the null value in a specified direction?
Non-inferiority question
Is the treatment sufficiently close to the comparator that the prespecified loss of efficacy or safety performance is not exceeded?
6. Statistical Methodology
No formal statistical analyses are posted to ClinicalTrials.gov for the endpoints identified in the registry record. The following framework explains how endpoints of these types are typically evaluated without attributing unreported numerical analysis procedures to the registry.
Binary or proportion-based non-inferiority endpoints
The first primary endpoint is defined as the proportion of patients with major adverse events at 30 days. The second is defined as the proportion of patients with MR severity reduction at 6 months. For endpoints expressed as proportions, a non-inferiority analysis would ordinarily compare the event or success proportions between randomized groups using a prespecified effect scale and non-inferiority margin.
A proportion difference is only an illustrative effect scale. The actual non-inferiority analysis depends on the prespecified statistical analysis plan, including the chosen effect measure, margin, confidence level, and analysis population.
Time-to-event endpoint
The CLASP IIF primary endpoint is time from randomization to the first occurrence of heart failure hospitalization or death. A time-to-event analysis accounts for both whether an event occurs and when it occurs, while allowing patients without an observed event to contribute follow-up until censoring.
Typical methods for this type of endpoint include Kaplan-Meier estimation for event-free survival and a Cox proportional-hazards model for a relative hazard estimate. A non-inferiority time-to-event analysis would then compare the resulting confidence interval with the prespecified non-inferiority margin on the chosen hazard-ratio scale.
Composite endpoint structure
The CLASP IIF endpoint combines heart failure hospitalization or death into a time-to-first-event composite. The statistical endpoint is therefore triggered by whichever component occurs first. This can improve event accumulation, but interpretation requires attention to the relative contribution of each component because a composite can be driven disproportionately by one component.
Echocardiographic MR assessment
The CLASP IID primary MR endpoint uses echocardiography and a 0-4+ scale. The registry defines the endpoint in terms of the proportion of patients with MR severity reduction. Statistical analysis therefore needs to preserve the distinction between the underlying ordinal MR scale and the registry's proportion-based endpoint definition.
Randomization and analysis population
Randomization is important because it establishes the treatment comparison before outcomes occur. For a non-inferiority trial, preserving the benefits of randomization is particularly important because exclusions or post-randomization decisions can affect the treatment groups differently. Non-inferiority assessments commonly consider both randomized and treatment-exposure perspectives, but the specific analysis populations for this trial are not reported in the registry information summarized here.
7. Planned Analysis: 30-Day Major Adverse Events
The first primary endpoint asks whether PASCAL is not inferior to MitraClip with respect to the proportion of patients with major adverse events, where the primary safety endpoint is a composite of Major Adverse Events (MAEs).
Primary safety endpoint
Endpoint type: proportion of patients with MAEs
Question: non-inferiority of PASCAL versus MitraClip
For this endpoint, a conventional analysis would compare the two randomized treatment proportions and construct a confidence interval for the prespecified treatment-effect measure. Non-inferiority would then be judged against the trial's predefined margin.
The endpoint is designed to address short-term safety. A non-inferiority result would mean that the observed evidence is compatible with PASCAL meeting the predefined acceptable safety boundary relative to MitraClip. It would not by itself demonstrate superiority, equivalence, or the absence of any safety difference.
The confidence interval is central because it describes the uncertainty around the treatment comparison. The p-value, if reported in a future analysis, would not measure the magnitude or clinical importance of the safety difference and would not replace comparison with the non-inferiority margin.
8. Planned Analysis: Six-Month MR Severity Reduction
The second primary endpoint asks whether PASCAL is not inferior to MitraClip with respect to the proportion of patients with MR severity reduction, measured by echocardiography using a 0-4+ scale for the CLASP IID Cohort.
Primary MR endpoint
Assessment: echocardiographic MR severity using a 0-4+ scale
Question: non-inferiority of PASCAL versus MitraClip
The registry expresses this endpoint as a proportion of patients achieving MR severity reduction. A typical non-inferiority analysis would therefore summarize the proportion in each randomized treatment group and construct an appropriate confidence interval for their difference or another prespecified effect measure.
This endpoint concerns an intermediate physiologic or imaging outcome rather than mortality or hospitalization. A finding of non-inferiority would address whether the PASCAL device meets the prespecified MR-reduction criterion relative to MitraClip; it would not automatically establish equivalence across every level of the 0-4+ MR scale.
The confidence interval indicates the precision of the estimated treatment comparison. Interpretation depends critically on the numerical non-inferiority margin and the exact endpoint-analysis rules, neither of which is specified in the registry information summarized here.
9. Planned Analysis: CLASP IIF Heart Failure Hospitalization or Death
The third primary endpoint applies to the CLASP IIF Cohort only. It evaluates time from randomization until the date of first heart failure hospitalization or death, through 5-year follow-up.
Primary time-to-event endpoint
Event: first heart failure hospitalization or death
Population: CLASP IIF Cohort only
A typical analysis would use time-to-event methods because patients can experience the endpoint at different times and some patients may remain free of the endpoint at the end of observed follow-up. Kaplan-Meier methods can describe the event-free probability over time, while a Cox model can provide a relative hazard estimate.
In a Cox model, the exponentiated treatment coefficient can be interpreted as a hazard ratio under the model. The model does not directly estimate a five-year absolute risk difference.
The endpoint combines hospitalization and death and records the first of those events. The resulting treatment comparison therefore addresses time to the first component event, not separate effects on hospitalization and mortality.
If a hazard ratio were reported, it would represent a relative rate of experiencing the composite endpoint under the fitted model. It would not mean that the same percentage reduction applies separately to heart failure hospitalization and death.
For a non-inferiority analysis, the confidence interval would need to be evaluated against the prespecified hazard-ratio margin. The proportional-hazards assumption would also be relevant if a Cox model were used.
10. Statistical Methods Explained
Why is this a non-inferiority trial rather than a conventional superiority trial?
The registry explicitly defines the three primary questions in terms of non-inferiority. The purpose is therefore to determine whether PASCAL remains within an acceptable prespecified performance boundary relative to MitraClip rather than simply asking whether the two devices differ.
Why does the non-inferiority margin matter so much?
The margin defines how much difference from the comparator can be accepted while still calling the new treatment non-inferior. A confidence interval that crosses the non-inferiority boundary does not establish non-inferiority, even if another statistical test produces a seemingly favorable p-value.
Why are the first two endpoints expressed as proportions?
The registry defines the 30-day safety endpoint and the 6-month MR endpoint in terms of the proportion of patients meeting their respective endpoint definitions. That makes these questions fundamentally different from the CLASP IIF endpoint, which is defined by the time until an event occurs.
Why is the CLASP IIF endpoint a time-to-event analysis?
Heart failure hospitalization or death can occur at different times after randomization. A time-to-event framework retains information about when events occur and can appropriately account for patients who remain event-free through their observed follow-up.
What does a hazard ratio mean for this trial?
If the CLASP IIF endpoint is analyzed with a Cox model, a hazard ratio below 1 would represent a lower estimated instantaneous rate of the composite endpoint in the PASCAL group relative to MitraClip, under the model. It would not be equivalent to a percentage difference in five-year event-free survival.
Why does the 0-4+ MR scale matter?
The MR measurement is inherently ordered: the 0-4+ scale contains severity levels rather than a continuous measurement on an unrestricted numerical scale. The registry, however, defines the primary endpoint as the proportion of patients with MR severity reduction. The analysis therefore needs to follow the endpoint definition rather than treating the raw 0-4+ score as though it were a conventional continuous variable.
Why should the two cohorts not simply be pooled?
The registry assigns the six-month MR endpoint to the CLASP IID Cohort and the five-year heart failure hospitalization-or-death endpoint to the CLASP IIF Cohort only. Those endpoint definitions establish different analysis populations and follow-up structures. Combining them would obscure the prespecified questions.
11. Non-Inferiority and Confidence Intervals
Non-inferiority interpretation is fundamentally interval-based. The treatment effect estimate describes the observed difference or ratio, while the confidence interval describes the range of effect values compatible with the statistical uncertainty under the chosen analysis.
Estimate
Describes the observed treatment comparison on the chosen effect scale.
Confidence interval
Describes uncertainty around the estimated comparison and determines how close the data remain to the non-inferiority boundary.
Non-inferiority margin
Defines the largest clinically acceptable loss relative to the comparator for the prespecified endpoint.
P-value
Addresses a specified hypothesis under a particular testing framework; it does not independently define non-inferiority or clinical relevance.
The registry information available here does not provide the numerical margins. Consequently, a final non-inferiority determination requires the actual prespecified margin for each endpoint and the corresponding formal statistical analysis.
12. Time-to-Event Analysis: Key Statistical Issues
Kaplan-Meier estimation
For the CLASP IIF endpoint, Kaplan-Meier estimation would typically describe the probability of remaining free of first heart failure hospitalization or death over time.
Here, di represents events at an event time and ni represents patients at risk immediately before that time.
Censoring
Patients who have not experienced heart failure hospitalization or death by their last evaluable follow-up can contribute information up to that time. Proper censoring rules are important because biased censoring can distort survival estimates.
Cox proportional-hazards modeling
A Cox model is commonly used to estimate a hazard ratio for a randomized time-to-event comparison. Its interpretation relies on the proportional-hazards framework. If the relative hazard changes materially over time, a single hazard ratio may summarize the treatment difference less completely than time-specific or absolute survival measures.
Composite endpoints
The first-event definition means that hospitalization and death are treated as components of one endpoint. A statistically significant or non-inferior composite result does not necessarily imply that each component has the same magnitude or direction of effect. Component-specific event counts and analyses are important when interpreting the clinical meaning of a composite endpoint.
13. Safety Endpoint: Statistical Interpretation
The 30-day safety endpoint is a composite of Major Adverse Events and is expressed as the proportion of patients with MAEs. Because the endpoint is a composite, its statistical interpretation depends on the precise components and endpoint adjudication rules used in the trial.
For a non-inferiority safety question, this distinction is especially important. The overall composite can satisfy the prespecified criterion while individual components may have different observed patterns. Component-level interpretation therefore complements the primary composite analysis.
14. Echocardiographic Endpoint: Statistical Interpretation
The six-month CLASP IID endpoint uses echocardiographic assessment of MR severity on a 0-4+ scale and asks about the proportion of patients with MR severity reduction.
The statistical analysis should preserve the registry's endpoint definition. It should not automatically treat a change from one MR grade to another as a normally distributed continuous measurement. The analysis also needs to account for the fact that echocardiographic measurements can involve assessment variability.
Measurement scale
MR severity is recorded on an ordered 0-4+ scale.
Primary endpoint definition
The registry defines the endpoint using the proportion of patients with MR severity reduction.
Assessment timing
The primary endpoint time frame is 6 months.
Interpretation
Non-inferiority addresses the prespecified reduction criterion relative to MitraClip, not every possible difference across the entire MR scale.
15. Design Topics
Randomization
Randomization assigns treatment before the primary outcomes are observed and is the principal design mechanism supporting an unbiased comparison between the named device strategies, subject to adherence to the randomized assignment and the prespecified analysis.
Parallel design
The registry identifies the study as a parallel design. Participants are therefore compared across treatment strategies rather than being described as sequentially receiving the two named devices in a crossover framework.
Masking
The registry identifies the study as having no masking. This is relevant when interpreting outcomes that involve clinical assessment or potentially subjective decisions. The extent to which lack of masking affects a particular endpoint depends on how that endpoint is measured and adjudicated.
Five registered arms
The registry identifies 5 arms. The information available here specifies the Edwards PASCAL System and Abbott MitraClip System as the interventions but does not provide arm-level sample sizes or a complete mapping of all five arms. Statistical interpretation should therefore avoid assigning unreported enrollment numbers to specific arms.
16. Interim Analysis, Multiplicity, Stratification, Bayesian Methods, and Missing Data
| Design topic | What the registry information establishes | Statistical implication |
|---|---|---|
| Interim analysis | No interim-analysis procedure is reported in the information summarized here. | An interim-monitoring scheme should not be assumed without the prespecified statistical analysis documentation. |
| Multiplicity | Three primary endpoints are registered, with different endpoint definitions and cohort applicability. | The formal relationship among these hypotheses and any type I error allocation requires the prespecified statistical analysis plan. |
| Stratification | No stratification factors are reported in the information summarized here. | Stratified analyses should not be attributed to the registry without documentation. |
| Bayesian methods | No Bayesian analysis is reported. | A Bayesian prior or posterior probability should not be assumed. |
| Missing data / imputation | No missing-data or imputation method is reported. | The handling of missing assessments requires the prespecified analysis rules. |
The distinction between these design features is important. A randomized non-inferiority trial can have multiple primary endpoints without necessarily using a simple one-test-follows-another hierarchy. The statistical analysis plan determines how multiplicity, missing observations, interim looks, and analysis populations are handled.
17. Analysis Populations and Non-Inferiority Bias
Non-inferiority trials require particular care with analysis populations. In a superiority trial, deviations from assigned treatment can make it harder to detect a difference. In a non-inferiority trial, some forms of non-adherence or protocol deviation can instead make treatment groups appear more similar and therefore potentially make non-inferiority easier to declare.
For that reason, the randomized treatment comparison and the treatment actually received can answer different questions. A rigorous non-inferiority evaluation generally gives careful attention to the prespecified primary population, protocol deviations, treatment exposure, and sensitivity analyses.
The registry information summarized here does not specify the exact analysis populations or protocol-deviation rules for the three primary endpoints. Those details are therefore essential to a complete formal non-inferiority assessment.
18. Limitations and Interpretation Issues
- No posted statistical analyses: ClinicalTrials.gov does not report formal statistical analyses for the registered primary endpoints in the record summarized here.
- Non-inferiority margins: the numerical margins are not reported in the information summarized here, yet they are essential for determining whether a non-inferiority criterion has been satisfied.
- No numerical results: the registry information summarized here does not provide endpoint estimates, confidence intervals, p-values, event counts, or treatment-specific safety results.
- Different endpoint types: the 30-day MAE endpoint and six-month MR-reduction endpoint are proportion-based, whereas the CLASP IIF endpoint is a time-to-first-event outcome.
- Different cohorts: the six-month MR endpoint applies to the CLASP IID Cohort, while the heart failure hospitalization-or-death endpoint applies to the CLASP IIF Cohort only.
- Composite safety endpoint: the 30-day MAE endpoint combines major adverse events, so component-specific behavior cannot be determined from the composite definition alone.
- Composite time-to-event endpoint: heart failure hospitalization and death are combined as a first-event endpoint, so a single treatment effect would not necessarily describe both components separately.
- Unmasked design: the registry identifies no masking, which can matter for outcomes involving clinical assessment or treatment decisions.
- Five-arm design: the registry identifies five arms, while the summarized intervention information names two devices without providing a complete arm-level allocation.
- Missing-data rules: the statistical handling of missing echocardiographic assessments and other endpoint data is not reported in the information summarized here.
19. Why This Trial Matters Statistically
CLASP IID is a useful statistical teaching case because it combines randomization, non-inferiority testing, device comparison, a composite safety endpoint, an echocardiographic endpoint, and a long-term time-to-event endpoint within the same pivotal trial program.
| Concept | How it appears in CLASP IID |
|---|---|
| Randomization | The registry identifies randomized allocation. |
| Parallel design | The registry identifies a parallel design model. |
| Non-inferiority | All three registered primary endpoints are explicitly framed as non-inferiority questions. |
| Composite endpoint | The primary safety endpoint is a composite of Major Adverse Events. |
| Proportion endpoint | The safety and MR-reduction endpoints are defined in terms of proportions of patients. |
| Ordinal clinical measurement | MR severity is measured by echocardiography using a 0-4+ scale. |
| Time-to-event analysis | The CLASP IIF endpoint measures time to first heart failure hospitalization or death. |
| Long-term follow-up | The CLASP IIF primary endpoint is followed through 5 years. |
| Multiple primary endpoints | Three primary endpoints are registered, with different populations and time frames. |
| Unmasked design | The registry identifies no masking. |
| Missing-data considerations | Endpoint-specific assessment and follow-up create a need for prespecified rules for incomplete observations. |
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial is structured around non-inferiority questions. Each conclusion must be tied to its endpoint-specific effect measure, confidence interval, and prespecified non-inferiority margin.
Clinical interpretation
The trial addresses whether the PASCAL System performs within the predefined acceptable range relative to MitraClip for short-term safety, MR severity reduction, and the CLASP IIF time-to-event outcome.
These two levels should remain distinct. Statistical non-inferiority is a statement about the prespecified comparison and its uncertainty. Clinical interpretation additionally depends on what constitutes a meaningful difference for patients and clinicians, the components of composite outcomes, the durability of effects, and the practical context of the device comparison.
21. What a Complete Non-Inferiority Result Would Need to Show
A complete statistical report for CLASP IID would need to present, for each primary endpoint, the observed treatment-specific outcome, the chosen effect measure, the confidence interval, and the prespecified non-inferiority margin. The interpretation would then depend on whether the confidence interval remained within the allowable non-inferiority boundary.
| Primary endpoint | Essential statistical information | Core interpretation |
|---|---|---|
| 30-day MAE | MAE proportion by treatment, treatment-effect estimate, confidence interval, non-inferiority margin | Whether PASCAL meets the prespecified safety non-inferiority criterion |
| 6-month MR severity reduction | Proportion with MR severity reduction by treatment, treatment-effect estimate, confidence interval, non-inferiority margin | Whether PASCAL meets the prespecified MR-reduction non-inferiority criterion |
| CLASP IIF time to HF hospitalization or death | Time-to-event estimates, treatment-effect estimate, confidence interval, non-inferiority margin, analysis population | Whether PASCAL meets the prespecified non-inferiority criterion for the composite time-to-event endpoint |
This structure prevents a common statistical error: treating a favorable point estimate as sufficient evidence of non-inferiority. The uncertainty interval and the prespecified margin are integral to the conclusion.
22. Planned Analysis Summary
Primary safety endpoint
Compare the proportion of patients with Major Adverse Events and evaluate PASCAL versus MitraClip against the prespecified non-inferiority criterion.
CLASP IID MR endpoint
Evaluate the proportion of patients with MR severity reduction measured by echocardiography on the 0-4+ scale and assess non-inferiority.
CLASP IIF time-to-event endpoint
Analyze time from randomization to first heart failure hospitalization or death and evaluate the non-inferiority criterion.
23. Sources
- ClinicalTrials.gov: NCT03706833 — CLASP IID.
Continue through the Clinical Biostats trial library
Explore additional clinical trial statistical analyses and educational resources covering randomized designs, endpoints, survival analysis, non-inferiority testing, and clinical-trial methodology.
24. Record Summary
CLASP IID is a randomized, parallel clinical trial with 1247 enrolled participants and 5 registered arms, comparing the Edwards PASCAL System with the Abbott MitraClip System in the setting of mitral regurgitation and related listed conditions. Its statistical structure is centered on three non-inferiority primary endpoints: a 30-day composite Major Adverse Event safety endpoint, a six-month echocardiographic MR severity-reduction endpoint for the CLASP IID Cohort, and a time-to-first heart failure hospitalization or death endpoint through 5-year follow-up for the CLASP IIF Cohort.
The principal statistical lesson is that these endpoints require different analytical perspectives. Proportion-based endpoints require an appropriate effect measure and confidence interval evaluated against a non-inferiority margin, while the CLASP IIF endpoint requires time-to-event methodology. Across all three endpoints, the numerical non-inferiority margin and the prespecified analysis rules are central to interpretation.