This page separates reported trial results from statistical interpretation. Numerical results are restricted to the ClinicalTrials.gov record and its posted statistical analyses. The ClinicalTrials.gov record is the official trial registry record.
1. Trial at a Glance
TRILUMINATE Pivotal was a randomized, parallel-design trial in patients with tricuspid regurgitation evaluating the TriClip device. The registry reports a total enrollment of 572, a single-masked design, and a primary analysis framework that included both a randomized cohort and a single-arm cohort.
| Feature | TRILUMINATE Pivotal |
|---|---|
| Trial name | TRILUMINATE Pivotal |
| Brief title | TRILUMINATE Pivotal Trial |
| ClinicalTrials.gov identifier | NCT03904147 |
| Condition | Tricuspid Regurgitation |
| Therapeutic area | Cardiovascular |
| Intervention | TriClipTM Device |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Primary purpose | Treatment |
| Phase | NA |
| Status | Active, not recruiting |
| Enrollment | 572 |
| Lead sponsor | Abbott Medical Devices |
| Sponsor type | Industry |
| Primary endpoints | 2 |
| Outcome measures posted | 11 |
| Statistical analyses posted | 9 |
2. Clinical Question
The statistical question is whether the TriClip device produces a superior outcome to the randomized control condition on the prespecified hierarchical composite endpoint at 12 months, and whether the single-arm cohort achieves the registry-defined combination of survival and quality-of-life improvement through 12 months.
Population
Participants enrolled in the TRILUMINATE Pivotal trial with the condition of tricuspid regurgitation. the ClinicalTrials.gov record reports total enrollment of 572.
Intervention
TriClipTM Device.
Comparator
The randomized cohort contains a Device Group and a Control Group. The ClinicalTrials.gov record does not provide additional treatment details for the control condition.
Primary questions
For the randomized cohort, does the Device Group have more favorable outcomes under the hierarchical composite at 12 months? For the single-arm cohort, what percentage meets the survival-plus-KCCQ definition through 12 months?
3. Trial Design
The registry ClinicalTrials.gov record lists 4 arms. The ClinicalTrials.gov record identifies the randomized Device Group, randomized Control Group, and Single Arm Group in the posted analyses and safety data. It does not provide enough information to reconstruct all 4 arm labels or their individual allocation structure without adding information beyond the ClinicalTrials.gov record.
4. Analysis Groups and Safety Denominators
Device Group
- Used in the primary randomized-cohort comparison.
- Compared with the Randomized - Control Group using the Finkelstein-Schoenfeld method for the primary hierarchical endpoint.
- Serious adverse events: 85 affected out of 175 at risk.
Control Group
- Used as the comparator in the randomized-cohort primary analysis.
- Serious adverse events: 73 affected out of 175 at risk.
- The ClinicalTrials.gov record does not specify additional treatment details for this group.
Single Arm Group
The single-arm cohort is analyzed separately for its 12-month survival-plus-KCCQ primary endpoint and several secondary outcomes. Serious adverse events were reported for 62 of 100 participants at risk.
Why denominators matter
The safety denominators differ from the 572 overall enrollment figure and from some endpoint-specific analysis populations. An endpoint estimate should therefore not be interpreted using the total enrollment denominator unless the analysis explicitly uses it.
5. Primary Endpoints
| Endpoint | Time frame | Registry definition / analysis |
|---|---|---|
| Randomized cohort: Hierarchical Composite of All-cause Death or Tricuspid Valve Surgery, Heart Failure Hospitalizations, and KCCQ Improvement | 12 Months | Analysis in the ITT population using the Finkelstein-Schoenfeld method. A win ratio was also calculated to descriptively quantify the magnitude of treatment benefits. |
| Single-arm cohort: Rate of Survival Through 12 Months With a Quality of Life Improvement Assessed Using KCCQ Overall Score of at Least 10 Points Compared to Baseline | 12 months | Assessed as the percentage of subjects meeting the definition of the endpoint. |
These two endpoints have different statistical structures. The randomized primary endpoint is a hierarchical composite requiring an ordering of patient-pair outcomes. The single-arm endpoint is a proportion defined by survival and a prespecified quality-of-life improvement criterion.
6. Statistical Methodology
Finkelstein-Schoenfeld hierarchical analysis
The randomized primary endpoint was analyzed using the Finkelstein-Schoenfeld method. This approach is designed for hierarchical composite outcomes in which more clinically important outcomes are considered before less important outcomes. Rather than reducing every component immediately to a simple yes/no composite, patient pairs are evaluated according to the prespecified hierarchy.
The registry definition orders all-cause death or tricuspid valve surgery, heart failure hospitalizations, and KCCQ improvement of at least 15 points. The analysis therefore incorporates the clinical hierarchy rather than treating every component as statistically interchangeable.
Win ratio
The registry reports a win ratio of 1.44 for the randomized primary endpoint, with a two-sided P-value of 0.0311. the ClinicalTrials.gov record identifies the hypothesis type as superiority and describes the win ratio as an estimate of the treatment effect.
A value above 1 indicates more favorable pairwise outcomes for the treatment group under the specified hierarchy. The registry's reported value of 1.44 therefore indicates a greater number of wins for the Device Group relative to losses under the Finkelstein-Schoenfeld framework.
Exact binomial analysis
The single-arm primary endpoint was analyzed using an exact test. The statistical method is Exact / Clopper-Pearson binomial. This is appropriate for a proportion when the analysis is framed around the number of participants meeting a binary endpoint definition and an exact rather than normal-approximation approach is used.
ANCOVA
Two secondary continuous outcomes in the randomized cohort used ANCOVA: change in KCCQ score from baseline to 12 months and change in 6 Minute Walk Test from baseline to 12 months. ANCOVA evaluates a follow-up outcome while accounting for baseline information, which can improve precision when baseline measurements are informative about the follow-up measurement.
Chi-squared testing
The randomized-cohort analysis of tricuspid regurgitation reduction to moderate or less at the 30-day visit used a chi-squared test. This is a categorical comparison between the randomized Device and Control Groups.
Wald / z-test
The randomized-device analysis of freedom from major adverse events within 30 days used a Z test. The single-arm 6 Minute Walk Test analysis used a normal approximation, normalized in the registry data to a Wald / z-test family.
7. Primary Results: Randomized Cohort
Hierarchical composite of death or tricuspid valve surgery, heart failure hospitalizations, and KCCQ improvement
Finkelstein-Schoenfeld win ratio
Two-sided P = 0.0311 · Superiority hypothesis
Randomized - Device Group vs Randomized - Control Group at 12 months
| Primary result component | Reported value |
|---|---|
| Analysis method | Finkelstein-Schoenfeld method |
| Method category | Win ratio / composite endpoint analysis |
| Effect measure | Win Ratio |
| Estimate | 1.44 |
| Confidence interval | A two-sided confidence interval is not included in the ClinicalTrials.gov record. |
| P-value | 0.0311 |
| Hypothesis | Superiority |
| Analysis population | ITT population, according to the registered endpoint definition |
The reported win ratio of 1.44 means that, under the prespecified hierarchical pairwise framework, the number of favorable patient-pair comparisons for the Device Group was greater than the corresponding number of unfavorable comparisons. A win ratio of 1.44 is therefore a relative measure of pairwise treatment benefit under this particular hierarchy.
It does not mean that 44% of patients improved, that mortality was reduced by 44%, or that every component of the composite individually improved by the same amount. The statistic combines information across the hierarchy and should be interpreted as a treatment-effect measure for the composite analysis rather than as a standalone effect estimate for death, surgery, hospitalization, or KCCQ.
The ClinicalTrials.gov record does not provide a confidence interval for the 1.44 estimate. Consequently, the precision of the win-ratio estimate cannot be characterized from the ClinicalTrials.gov record using a confidence interval. The two-sided P-value of 0.0311 addresses evidence against the null hypothesis within the reported testing framework; it is not a measure of the size of the treatment effect.
The hierarchical construction is also important. A win ratio depends on the ordering of the component outcomes. Changing the hierarchy can change the statistic, so the reported 1.44 should not be detached from the Finkelstein-Schoenfeld definition used by the registry.
8. Primary Results: Single-Arm Cohort
Survival through 12 months with KCCQ improvement of at least 10 points
Exact-test result
Single Arm Group · 12 months · Superiority hypothesis
The ClinicalTrials.gov record does not include the corresponding endpoint percentage or confidence interval.
| Primary result component | Reported value |
|---|---|
| Endpoint | Rate of survival through 12 months with KCCQ Overall Score improvement of at least 10 points compared with baseline |
| Analysis method | Exact test |
| Effect measure | Percentage of participants |
| Estimate | Not included in the ClinicalTrials.gov record |
| Confidence interval | Not included in the ClinicalTrials.gov record |
| P-value | 0.0008 |
| Hypothesis | Superiority |
The exact-test P-value of 0.0008 indicates the result was evaluated using an exact binomial framework rather than a normal approximation. The test addresses the statistical hypothesis specified for this single-arm endpoint; it does not itself provide the percentage of participants meeting the endpoint.
The registry analysis does not report the endpoint estimate or its confidence interval. Therefore, the P-value should not be converted into an implied response or survival rate. A P-value is evidence against a specified null hypothesis, not a direct measure of the magnitude or clinical importance of an outcome.
The endpoint itself combines survival through 12 months with a KCCQ improvement of at least 10 points. A participant must satisfy the complete endpoint definition to count as a success. The result therefore cannot be interpreted as a standalone KCCQ response rate or as a standalone 12-month survival estimate.
9. Secondary Endpoint Results
The registry provides formal statistical analyses for seven secondary endpoints. Several analyses report only a P-value in the registry-reported statistical-analysis fields; where an effect estimate or confidence interval is absent, this page does not infer one.
| Secondary endpoint | Time frame | Method | P-value |
|---|---|---|---|
| Randomized Cohort - Device Group Only: Rate of Kaplan-Meier Estimate (SE) of Freedom From Major Adverse Events (MAE) Occurring Within 30 Days Post-procedure | 30 Days | Z test | <0.0001 |
| Randomized Cohort: Change in Quality of Life as Assessed by KCCQ Score | 12 months minus baseline | ANCOVA model | <0.0001 |
| Randomized Cohort: Tricuspid Regurgitation Reduction to Moderate or Less at 30-Day Visit | 30 Days | Chi-square test | <0.0001 |
| Randomized Cohort: Change in 6 Minute Walk Test (6MWT) From Baseline to 12 Months | 12 Months | ANCOVA model | 0.2482 |
| Single-Arm Cohort: Percentage of Subjects With TR Reduction by at Least 1 Grade at 30-Day Follow-up | 30 days | Exact test | <0.0001 |
| Single-Arm Cohort: Rate of Freedom From MAE Through 30 Days | 30 days | Binomial exact method | <0.0001 |
| Single-Arm Cohort: Change in 6 Minute Walk Test (6MWT) From Baseline to 12 Months | 12 months | Normal approximation | 0.1090 |
KCCQ change in the randomized cohort
ANCOVA analysis
Device Group vs Control Group · 12 months minus baseline
The registry analysis does not provide an effect estimate or confidence interval in the ClinicalTrials.gov record.
The ANCOVA result indicates a statistically significant randomized-group comparison under the reported model, with P < 0.0001. Because the ClinicalTrials.gov record does not contain the adjusted treatment difference or confidence interval, the magnitude and precision of the KCCQ effect cannot be quantified from this record alone.
ANCOVA is particularly relevant here because the endpoint is a change from baseline to 12 months. The model can account for baseline KCCQ information while comparing follow-up outcomes. The P-value does not tell us how many KCCQ points separated the groups, nor whether a particular patient experienced a clinically meaningful improvement.
Tricuspid regurgitation reduction to moderate or less
Chi-squared test
Device Group vs Control Group · 30-Day Visit
The registry-reported analysis does not report the corresponding group percentages or confidence interval.
The chi-squared test evaluates the categorical comparison between the randomized groups for whether tricuspid regurgitation was reduced to moderate or less at the 30-day visit. The reported P < 0.0001 indicates strong statistical evidence against the null comparison under the reported test.
The result does not reveal the absolute difference between groups because the registry-reported analysis does not include the underlying percentages. A P-value alone cannot establish whether the difference was small or large in absolute terms.
6 Minute Walk Test in the randomized cohort
ANCOVA analysis
Device Group vs Control Group · Baseline to 12 months
The reported P-value of 0.2482 does not provide evidence of a statistically significant randomized-group difference under the reported ANCOVA analysis. It does not demonstrate that the treatment groups were identical, nor does it quantify the size of any difference.
The registry states that 83 (24%) of the 350 subjects, comprising 44 Device and 39 Control participants, were not included because paired 6MWT values at baseline and 12 months were missing. This degree of missing paired information is relevant when interpreting the precision and representativeness of a change-from-baseline analysis.
Secondary outcomes in the single-arm cohort
| Endpoint | Analysis note | P-value |
|---|---|---|
| TR reduction by at least 1 grade at 30-day follow-up | Twelve subjects were not included because paired TR severity at baseline and 30 days was missing. | <0.0001 |
| Freedom from MAE through 30 days | One subject withdrew before 30 days without experiencing an MAE and was excluded from the analysis. | <0.0001 |
| Change in 6MWT from baseline to 12 months | Twenty-nine percent of subjects were not included because 6MWT values at baseline and/or 12 months were missing. | 0.1090 |
The single-arm secondary analyses use exact binomial methods for the two binary outcomes and a normal approximation for the 6MWT outcome. The two exact analyses have reported P-values below 0.0001, while the 6MWT analysis reports P = 0.1090.
Because these are single-arm analyses, the P-values do not represent a randomized treatment-versus-control comparison. They address the null hypotheses specified for their respective single-arm statistical tests. Their interpretation therefore differs fundamentally from the randomized-cohort win ratio.
10. Missing Data and Analysis Populations
Endpoint-specific exclusions are important in this trial because the denominator can change depending on the measurement being analyzed. The statistical analyses posted on ClinicalTrials.gov explicitly identify several sources of missing information.
| Endpoint | Reported missing / excluded information |
|---|---|
| Primary single-arm survival + KCCQ endpoint | Nine subjects were stated to be excluded; the registry note separately lists KCCQ unavailable at 12 months (n=6), missed 12-month visit (n=3), and declined KCCQ (n=2). |
| Randomized KCCQ change | Forty (11%) of the 350 subjects, comprising 20 Device and 20 Control, were not included because paired KCCQ values at baseline and 12 months were missing. |
| Randomized TR reduction | Forty-one (12%) of the 350 subjects, comprising 13 Device and 28 Control, were not included because TR was not available at 30 days. |
| Randomized 6MWT change | Eighty-three (24%) of the 350 subjects, comprising 44 Device and 39 Control, were not included because paired 6MWT values at baseline and 12 months were missing. |
| Single-arm TR reduction | Twelve subjects were not included because paired TR severity at baseline and 30 days was missing. |
| Single-arm freedom from MAE | One subject, identified in the registry as Subject 1300, withdrew before 30 days without experiencing an MAE and was excluded. |
| Single-arm 6MWT change | Twenty-nine percent of subjects were not included because 6MWT values at baseline and/or 12 months were missing. |
These exclusions illustrate why an endpoint should always be interpreted together with its analysis population. A P-value or effect estimate computed among participants with complete paired measurements does not automatically describe every enrolled participant.
11. Safety Results
the ClinicalTrials.gov record reports serious adverse events by arm. These data are presented as affected participants out of participants at risk; no additional safety event definitions or comparative statistical test are reported in the ClinicalTrials.gov record.
| Group | Serious adverse events | At risk |
|---|---|---|
| Randomized - Device Group | 85 | 175 |
| Randomized - Control Group | 73 | 175 |
| Randomized - Single Arm Group | 62 | 100 |
The serious-adverse-event data describe the number of affected participants and the corresponding at-risk denominator. They are descriptive safety results, not an adjusted treatment-effect estimate.
The randomized Device and Control Groups have the same registry-reported denominator of 175, which makes their affected-participant counts directly comparable as raw event counts. However, the ClinicalTrials.gov record does not provide a confidence interval, P-value, exposure time, event timing, or definition beyond the label "serious adverse events." Those missing elements limit what can be concluded statistically from these counts alone.
The Single Arm Group should not be treated as another randomized comparator. Its serious-adverse-event count comes from a separate cohort with 100 participants at risk.
12. Statistical Methods Explained
Why use a win ratio rather than a simple composite event rate?
The primary randomized endpoint is hierarchical. The Finkelstein-Schoenfeld framework permits patient pairs to be compared first on the higher-priority component and, if the pair cannot be distinguished there, on subsequent components. This preserves the intended ordering of clinical outcomes. A conventional composite that simply counts whether any component occurred would treat all components as equivalent unless additional weighting were introduced.
What does a win ratio of 1.44 mean?
A win ratio of 1.44 means the Device Group had more favorable pairwise outcomes than unfavorable outcomes under the specified hierarchy, with the ratio of wins to losses equal to 1.44. It does not mean that the device reduced a particular event by 44%, because the statistic summarizes the complete hierarchical endpoint rather than one isolated component.
Why does the hierarchy matter?
Death or tricuspid valve surgery, heart failure hospitalization, and KCCQ improvement are not statistically interchangeable events. The Finkelstein-Schoenfeld method evaluates them in a defined order. Consequently, the resulting win ratio depends on the endpoint hierarchy specified in the analysis.
Why was ANCOVA used for KCCQ and 6MWT change?
Both KCCQ and 6MWT are continuous measurements with baseline and follow-up observations. ANCOVA provides a framework for comparing follow-up outcomes between randomized groups while accounting for baseline measurements. This can be more informative than comparing raw changes without considering baseline values.
Why use an exact binomial method?
The single-arm primary endpoint is expressed as a percentage of participants meeting a binary definition. Exact binomial methods, including the Clopper-Pearson approach, avoid relying on the normal approximation when calculating inference for a proportion. The reported analysis therefore uses an exact framework for its binary endpoint.
What does a P-value such as 0.0311 tell us?
A P-value quantifies how unusual the observed test statistic would be under the null hypothesis and the assumptions of the statistical test. It does not measure clinical importance, treatment-effect magnitude, or the probability that the null hypothesis is true. The win ratio of 1.44 supplies the effect estimate; the P-value supplies a different piece of information about statistical evidence.
Why does missing paired information matter?
Change-from-baseline endpoints require both baseline and follow-up observations. Participants without one of those measurements cannot contribute directly to the paired change analysis. When a substantial proportion is excluded, the resulting estimate can depend on which participants remain available for analysis and on the mechanism producing the missing data.
13. Understanding the Primary Win Ratio
The registry-reported estimate of 1.44 is therefore directional: it favors the Device Group within the statistical structure of the randomized primary endpoint. The statistic is not a percentage improvement and should not be transformed into one.
What the 1.44 does mean
It is the reported treatment-effect estimate from the Finkelstein-Schoenfeld hierarchical comparison of the randomized Device and Control Groups.
What the 1.44 does not mean
It is not a 44% reduction in mortality, a 44% increase in survival, or a statement that 44% of participants benefited.
Why the P-value is separate
The P-value of 0.0311 addresses the statistical test. It does not quantify the size of the treatment effect.
Why precision matters
The ClinicalTrials.gov record does not contain a confidence interval for the win ratio, so precision cannot be summarized with a reported interval from these data.
14. Time-to-Event and Composite Endpoint Considerations
The trial's primary randomized endpoint contains all-cause death and tricuspid valve surgery as its highest-priority component, followed by heart failure hospitalizations and then KCCQ improvement. This creates a fundamentally different statistical object from a conventional time-to-first-event endpoint.
The ClinicalTrials.gov record does not provide a Kaplan-Meier estimate, hazard ratio, median event time, or formal proportional-hazards analysis for the primary randomized endpoint. It would therefore be inappropriate to substitute a conventional Cox-model interpretation for the reported Finkelstein-Schoenfeld analysis.
Kaplan-Meier curves and Cox hazard ratios are designed for particular time-to-event structures. The reported primary endpoint instead uses a hierarchical pairwise comparison. The appropriate educational interpretation is therefore centered on pairwise wins and losses under the hierarchy, not on an inferred hazard ratio.
15. Secondary Endpoint Statistical Interpretation
| Method | Endpoints in this trial | What it answers |
|---|---|---|
| ANCOVA | KCCQ change; randomized 6MWT change | Whether adjusted continuous outcomes differ between randomized groups. |
| Chi-squared test | Randomized TR reduction to moderate or less | Whether the categorical outcome distribution differs between randomized groups. |
| Exact / Clopper-Pearson binomial | Single-arm TR reduction; single-arm freedom from MAE; single-arm primary endpoint | Inference for a binary proportion using an exact framework. |
| Wald / z-test | Device-group freedom from MAE; single-arm 6MWT analysis | Inference using a normal-approximation framework. |
| Win ratio | Randomized primary hierarchical composite | Pairwise treatment benefit across the prespecified outcome hierarchy. |
The diversity of methods is a useful reminder that a clinical trial rarely has one universal statistical test. The analysis should follow the endpoint's measurement scale, study design, and estimand.
16. Multiplicity and Hypothesis Testing
the ClinicalTrials.gov record identifies the hypothesis type for the posted statistical analyses as superiority. It also reports two primary endpoints: one for the randomized cohort and one for the single-arm cohort.
The ClinicalTrials.gov record does not provide an alpha-allocation scheme, multiplicity-adjustment procedure, hierarchical testing sequence across the primary endpoints, or interim-analysis spending rule.
17. Randomization and Causal Interpretation
The randomized component provides the principal basis for comparing the Device and Control Groups. Randomization is important because, in expectation, it balances measured and unmeasured baseline characteristics between treatment groups, allowing differences in subsequent outcomes to be attributed to treatment assignment under the assumptions of the randomized trial.
The single-arm cohort does not provide the same randomized comparator. Its primary and secondary results can describe outcomes in that cohort, but they cannot by themselves establish a randomized treatment effect against an untreated or alternative-treatment population.
Randomized analysis
The primary randomized endpoint compares the Device Group directly with the Control Group using the Finkelstein-Schoenfeld method.
Single-arm analysis
The single-arm endpoints evaluate the percentage or change observed within the Single Arm Group using exact or normal-approximation methods as specified.
18. Endpoint-Specific Missingness: Why It Changes the Interpretation
The analyses posted on ClinicalTrials.gov show that missingness is not uniform across outcomes. The randomized KCCQ analysis excludes 40 of 350 participants, while the randomized 6MWT analysis excludes 83 of 350. The TR analysis excludes 41 of 350. The single-arm 6MWT analysis states that 29% of subjects were not included.
These differences mean that one should not assume that every secondary endpoint represents the same analysis population. An observed difference in KCCQ, TR, and 6MWT can involve different sets of participants because the required measurements differ.
Missing-data handling is part of the estimand and analysis strategy. The ClinicalTrials.gov record identifies who was excluded for several endpoints but does not provide an imputation strategy. It is therefore not appropriate to claim that a particular imputation method, missing-at-random assumption, or sensitivity analysis was used.
19. Safety and Efficacy Should Be Kept Separate
The trial reports both efficacy outcomes and serious adverse events. These represent different statistical questions. The primary win ratio addresses a hierarchical clinical outcome, whereas the ClinicalTrials.gov record is affected-participant counts among specified at-risk groups.
| Domain | Reported analysis structure |
|---|---|
| Primary randomized efficacy | Finkelstein-Schoenfeld hierarchical comparison; win ratio 1.44; P = 0.0311. |
| Primary single-arm efficacy | Exact binomial test; P = 0.0008. |
| Quality of life | ANCOVA for randomized KCCQ change; P < 0.0001. |
| Functional capacity | ANCOVA for randomized 6MWT change; P = 0.2482; normal approximation for single-arm change; P = 0.1090. |
| TR severity | Chi-squared test in randomized cohort; exact test in single-arm cohort. |
| Safety | Serious adverse events reported descriptively by group in the ClinicalTrials.gov record. |
A statistical analysis should not combine these domains into a single informal benefit-risk score. Each outcome has its own definition, analysis population, and inferential framework.
20. Limitations
- Incomplete effect estimates: Several posted statistical analyses contain P-values but do not provide the corresponding effect estimate or confidence interval in the ClinicalTrials.gov record.
- Win-ratio confidence interval unavailable: The primary randomized win ratio is reported as 1.44 with P = 0.0311, but the ClinicalTrials.gov record does not provide its confidence interval.
- Single-arm interpretation: The single-arm cohort does not provide a randomized comparator, so its P-values should not be interpreted as randomized treatment-effect tests.
- Endpoint-specific missingness: Several analyses exclude participants because baseline, follow-up, or paired measurements were unavailable.
- Registry discrepancy: The registry-reported primary single-arm analysis note reports nine exclusions while listing reasons whose counts sum to 11. The discrepancy cannot be resolved from the ClinicalTrials.gov record.
- Safety information is limited: Serious adverse events are reported as affected/at-risk counts, without additional timing, event definitions, exposure adjustment, confidence intervals, or comparative testing.
- Multiplicity information is incomplete: The ClinicalTrials.gov record identifies superiority hypotheses and posted analyses but does not specify an overall alpha-control strategy.
- Four-arm structure is not fully specified: the ClinicalTrials.gov record reports 4 arms, while the ClinicalTrials.gov record identifies three group labels. The fourth arm cannot be reconstructed without adding external information.
- No inferred survival analysis: Although one primary endpoint contains survival and the trial's statistical concepts includes Kaplan-Meier estimation, the registry-reported primary randomized analysis is a Finkelstein-Schoenfeld analysis rather than a reported Cox hazard-ratio analysis.
21. Why This Trial Matters Statistically
TRILUMINATE Pivotal is a useful teaching example because its statistical architecture differs from the conventional pattern of a two-arm randomized trial analyzed with a single time-to-event hazard ratio. The trial combines a hierarchical composite endpoint, a win-ratio effect measure, a separate single-arm primary endpoint, continuous patient-reported and functional outcomes, categorical structural outcomes, and endpoint-specific missingness.
| Concept | How it appears in TRILUMINATE Pivotal |
|---|---|
| Randomization | Randomized allocation in a parallel-design trial. |
| Hierarchical composite | Death or tricuspid valve surgery, heart failure hospitalizations, and KCCQ improvement are incorporated in a defined hierarchy. |
| Finkelstein-Schoenfeld method | Primary randomized endpoint analysis. |
| Win ratio | Primary treatment-effect measure, estimated at 1.44. |
| Exact binomial testing | Primary single-arm endpoint and selected binary secondary endpoints. |
| ANCOVA | Randomized KCCQ and 6MWT change analyses. |
| Chi-squared test | Randomized TR reduction analysis. |
| Wald / z-test | Device-group MAE analysis and single-arm 6MWT analysis. |
| Missing data | Different endpoint-specific exclusions affect several secondary analyses. |
| Single-arm analysis | Provides a separate primary endpoint and secondary analyses that should not be interpreted as randomized comparisons. |
22. Statistical Methods Explained in Practical Terms
The most important lesson from this trial is that the statistical method must match the endpoint. The same trial can legitimately require several different methods because its outcomes are measured on different scales and answer different questions.
Win ratio
Best understood as a pairwise comparison under a hierarchy. It is not interchangeable with a risk ratio, odds ratio, or hazard ratio.
ANCOVA
Useful for comparing continuous follow-up outcomes between randomized groups while incorporating baseline information.
Chi-squared test
Addresses whether categorical outcome distributions differ between groups.
Exact binomial test
Provides inference for a binary proportion without relying on a normal approximation.
23. Reported Results in One View
| Endpoint | Population | Method | Effect / result |
|---|---|---|---|
| Hierarchical composite | Randomized cohort | Finkelstein-Schoenfeld | Win ratio 1.44; P = 0.0311 |
| Survival + KCCQ improvement | Single-arm cohort | Exact test | P = 0.0008; percentage and CI not reported |
| Freedom from MAE within 30 days | Randomized Device Group | Z test | P < 0.0001 |
| KCCQ change | Randomized cohort | ANCOVA | P < 0.0001 |
| TR reduction to moderate or less | Randomized cohort | Chi-squared | P < 0.0001 |
| 6MWT change | Randomized cohort | ANCOVA | P = 0.2482 |
| TR reduction by at least 1 grade | Single-arm cohort | Exact test | P < 0.0001 |
| Freedom from MAE through 30 days | Single-arm cohort | Binomial exact | P < 0.0001 |
| 6MWT change | Single-arm cohort | Normal approximation | P = 0.1090 |
24. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The randomized primary analysis produced a win ratio of 1.44 with a two-sided P-value of 0.0311 under the Finkelstein-Schoenfeld hierarchy. Several secondary endpoints also produced reported P-values, using methods matched to their measurement scale.
Clinical interpretation
The primary composite incorporates mortality or tricuspid valve surgery, heart failure hospitalization, and KCCQ improvement. The registry also reports quality-of-life, tricuspid-regurgitation, functional-capacity, and safety outcomes. Their clinical meaning should be considered separately because they measure different aspects of patient experience and clinical status.
A statistically significant result does not automatically establish the magnitude of clinical benefit. Conversely, a non-significant P-value does not prove that two treatments have identical effects. The most informative interpretation uses the endpoint definition, effect estimate where available, confidence interval where available, P-value, analysis population, and missing-data context together.
25. Record Summary
TRILUMINATE Pivotal provides a detailed example of how modern clinical-trial statistics can move beyond a single conventional endpoint. Its randomized primary analysis uses a Finkelstein-Schoenfeld hierarchical composite and reports a win ratio of 1.44 with a two-sided P-value of 0.0311. A separate single-arm primary endpoint uses an exact binomial framework and reports P = 0.0008, while secondary outcomes use ANCOVA, chi-squared, exact binomial, and Wald / z-test approaches.
The key statistical discipline is to keep these analyses distinct. The win ratio is not a hazard ratio; a single-arm exact-test P-value is not a randomized treatment comparison; an ANCOVA P-value does not reveal the magnitude of the adjusted KCCQ difference; and raw serious-adverse-event counts do not by themselves establish a comparative safety effect. The appropriate interpretation follows the endpoint-specific estimand and analysis method.
26. Related Tutorials
Learn more about the methods used in this trial:
27. Related Calculators
28. Sources
- ClinicalTrials.gov: TRILUMINATE Pivotal, NCT03904147.
- Linked publication: PubMed record — PMID 40159089.
- Linked publication: PubMed record — PMID 39903153.
- Linked publication: PubMed record — PMID 39570246.
- Linked publication: PubMed record — PMID 39222896.
- Linked publication: PubMed record — PMID 36876753.
Continue with the statistical methods
Explore the underlying biostatistical methods and calculators used to understand randomized clinical-trial endpoints, categorical outcomes, continuous outcomes, and hierarchical treatment-effect measures.