This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical results on this page are restricted to the registry-reported TRIUMPH registry data.
1. Trial at a Glance
TRIUMPH was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating inhaled treprostinil against placebo inhalation solution in patients with severe pulmonary arterial hypertension. The registry reports 235 enrolled participants and statistical analyses for 9 posted outcome measures.
| Feature | TRIUMPH |
|---|---|
| Trial name | TRIUMPH |
| ClinicalTrials.gov | NCT00147199 |
| Phase | Phase 3 |
| Condition | Pulmonary Hypertension |
| Population | Patients with severe pulmonary arterial hypertension (PAH) |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 235 |
| Lead sponsor | United Therapeutics |
| Status | Completed |
| Trial dates | June 2005–October 2007 |
2. Clinical Question
The central statistical question was whether inhaled treprostinil produced a different change in peak 6-minute walk distance from baseline to Week 12 compared with placebo inhalation solution in patients with severe pulmonary arterial hypertension.
Population
Patients with severe pulmonary arterial hypertension, under a phase 3 treatment investigation.
Intervention
Inhaled treprostinil.
Comparator
Placebo inhalation solution.
Primary question
Does the treatment comparison produce a difference in change in peak 6-minute walk distance at Week 12?
3. Trial Design
The registry describes TRIUMPH as a randomized, parallel-group phase 3 trial with quadruple masking and a treatment purpose. There were 2 intervention arms and 235 enrolled participants.
Inhaled Treprostinil
- Inhaled treprostinil
- Study treatment arm in the randomized comparison
- Serious adverse events: 9/115
Placebo
- Placebo inhalation solution
- Comparator arm in the randomized comparison
- Serious adverse events: 13/120
4. Endpoints
The registry contains 1 registered primary endpoint and 8 secondary statistical analyses, for 9 posted statistical analyses in total. The primary endpoint was a continuous change-from-baseline measure assessed at Week 12.
| Endpoint | Role | Time frame | Registry definition / analysis |
|---|---|---|---|
| Peak 6-minute Walk Distance | Primary | 12 weeks | Change in peak 6-minute walk distance from baseline to Week 12. Peak 6MWD was defined as a 6-minute walk test within 10 to 60 minutes after study drug inhalation. |
| Borg Dyspnea Score | Secondary | 12 weeks | Continuous score; Wilcoxon rank sum test. |
| New York Heart Association (NYHA) Functional Classification | Secondary | 12 weeks | Participants; Wilcoxon rank sum test. |
| Trough 6MWD at Week 12 | Secondary | 12 Weeks | Continuous measure in meters; ANCOVA. |
| Peak 6MWD at Week 6 | Secondary | 6 weeks | Continuous measure in meters; Wilcoxon rank sum test. |
| Quality of Life (Minnesota Living With Heart Failure) — Global Score | Secondary | 12 weeks | Units on a scale; Wilcoxon rank sum test. |
| Quality of Life (Minnesota Living With Heart Failure) — Physical Dimension | Secondary | 12 weeks | Units on a scale; Wilcoxon rank sum test. |
| Quality of Life (Minnesota Living With Heart Failure) — Emotional Dimension Score | Secondary | 12 weeks | Units on a scale; Wilcoxon rank sum test. |
| N-terminal Pro-B-Type Natriuretic Peptide (NT Pro-BNP) | Secondary | 12 weeks | Continuous measure in pg/mL; Wilcoxon rank sum test using the per-protocol population. |
5. Statistical Methodology
Primary endpoint: ANCOVA
The primary endpoint was analyzed using analysis of covariance (ANCOVA) in the intention-to-treat population. ANCOVA is a linear-model approach that can compare treatment groups while accounting for a continuous baseline measurement when the outcome is a change from baseline or a follow-up measurement.
Here, the treatment coefficient represents the treatment-group difference after accounting for the baseline measurement. The exact model specification beyond the reported ANCOVA method is not reported in the ClinicalTrials.gov record.
Nonparametric secondary analyses
Several secondary endpoints were analyzed using the Wilcoxon rank sum test. This is a two-group nonparametric comparison based on the ranks of observations rather than requiring the outcome itself to follow a normal distribution.
The registry reports a Hodges-Lehmann estimate for these analyses. The Hodges-Lehmann procedure provides a location-shift estimate associated with the rank-based comparison and can be accompanied by a confidence interval. Its interpretation depends on the scale and direction of the endpoint.
Intention-to-treat and per-protocol populations
The primary peak 6-minute walk distance analysis used the intention-to-treat population. Most secondary analyses also used an intention-to-treat population, whereas the NT Pro-BNP analysis used a per-protocol population. These are not interchangeable analysis populations.
Intention-to-treat
The treatment comparison is analyzed according to randomized treatment assignment. This maintains the connection between the efficacy comparison and the original randomization.
Per-protocol
The analysis is restricted according to protocol-related criteria. This can describe outcomes among participants meeting those criteria, but it does not preserve the same interpretation as an ITT comparison.
Sample-size assumptions
The registry's analysis notes state that sample size was calculated based on change in 6MWD at Week 12. The calculation assumed a between-treatment difference of 35m, a standard deviation of 75m, and a type I error of 0.05 using a two-sided p-value of less than 0.05. The stated target was 90% power to detect this difference, requiring 100 subjects per treatment group for a total of 200, with an allowance for dropout.
| Design assumption | Registry value |
|---|---|
| Target between-treatment difference | 35m |
| Assumed standard deviation | 75m |
| Type I error | 0.05 |
| Testing convention | Two-sided p-value of less than 0.05 |
| Target power | 90% |
| Required subjects per treatment group | 100 |
| Target total sample size | 200 |
| Actual enrollment | 235 |
6. Primary Result: Peak 6-Minute Walk Distance
The registered primary endpoint was change in peak 6-minute walk distance from baseline to Week 12. The registry reports an intention-to-treat ANCOVA analysis comparing inhaled treprostinil with placebo.
Hodges-Lehmann estimate
95% CI: 8.0–32.8 · P = 0.00044
Two-sided confidence interval · Superiority hypothesis
| Primary endpoint | Value |
|---|---|
| Outcome | Peak 6-minute Walk Distance |
| Time frame | 12 weeks |
| Outcome unit | meters |
| Analysis population | Intention to treat analysis |
| Groups compared | Inhaled Treprostinil vs Placebo |
| Method | ANCOVA |
| Effect measure | Hodges-Lehmann Estimate |
| Estimate | 20 |
| 95% CI | 8.0–32.8 |
| P-value | 0.00044 |
| Hypothesis type | Superiority |
The reported Hodges-Lehmann estimate of 20.0 meters represents the estimated location difference between the inhaled treprostinil and placebo groups for the analyzed peak 6-minute walk distance outcome. Because the endpoint is measured in meters, the estimate is expressed on that same measurement scale.
The estimate does not mean that every participant increased walking distance by exactly 20.0 meters, nor does it establish that every individual patient experienced that amount of improvement. It summarizes the between-group treatment comparison at the population level under the reported analysis.
The two-sided 95% confidence interval of 8.0–32.8 describes uncertainty around the estimated treatment difference. It is an interval for the statistical estimate, not a range containing 95% of individual patient responses.
The P-value of 0.00044 quantifies how incompatible the observed result would be with the null hypothesis under the specified testing framework. It does not measure the magnitude of the treatment effect, the probability that the treatment works, or the clinical importance of 20.0 meters. Effect size and uncertainty are better represented by the estimate and confidence interval.
The registry reports ANCOVA as the method but identifies the effect measure as a Hodges-Lehmann estimate.
7. Secondary Endpoint Results
The registry reports eight secondary statistical analyses. Most were conducted in an intention-to-treat population using the Wilcoxon rank sum test, while trough 6MWD used ANCOVA and NT Pro-BNP used a per-protocol population.
| Secondary endpoint | Population | Method | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Borg Dyspnea Score | Intention to treat | Wilcoxon rank sum | 0 | -0.5 to 0.0 | 0.623 |
| NYHA Functional Classification | Intention to treat | Wilcoxon rank sum | 0 | 0.0 to 0.0 | 0.807 |
| Trough 6MWD at Week 12 | Intention to treat | ANCOVA | 13.7 | 4.0 to 24.8 | 0.007 |
| Peak 6MWD at Week 6 | Intention to treat | Wilcoxon rank sum | 18.5 | 8.5 to 28.3 | 0.0003 |
| Quality of Life — Global Score | Intention to treat | Wilcoxon rank sum | -4.0 | -8.0 to 0 | 0.027 |
| Quality of Life — Physical Dimension | Intention to treat | Wilcoxon rank sum | -2.0 | -3.0 to 0.0 | 0.037 |
| Quality of Life — Emotional Dimension Score | Intention to treat | Wilcoxon rank sum | -1.0 | -2.0 to 0.0 | 0.173 |
| NT Pro-BNP | Per protocol | Wilcoxon rank sum | -167 | -333 to -64 | 0.001 |
Borg Dyspnea Score
Hodges-Lehmann estimate
95% CI: -0.5–0.0 · P = 0.623
Wilcoxon rank sum test · Intention-to-treat population
The registry notes that a Week 12 observation was not present for one subject and that this data point was not included in the analysis. That detail is important because the effective analysis set for this endpoint is therefore not literally identical to a dataset containing every expected Week 12 observation.
NYHA Functional Classification
Hodges-Lehmann estimate
95% CI: 0.0–0.0 · P = 0.807
Wilcoxon rank sum test · Intention-to-treat population
The registry classifies NYHA Functional Classification as a binary endpoint in its statistical analysis record while reporting a Wilcoxon rank sum test and a Hodges-Lehmann estimate. This page preserves that reported method rather than substituting a different statistical test based on the inferred measurement type.
Trough 6MWD at Week 12
Hodges-Lehmann estimate
95% CI: 4.0–24.8 · P = 0.007
ANCOVA · Intention-to-treat population
The estimate of 13.7 meters describes the reported between-group treatment comparison for trough 6MWD at Week 12. The confidence interval remains above 0, while the p-value is 0.007 under the reported superiority framework.
Peak 6MWD at Week 6
Hodges-Lehmann estimate
95% CI: 8.5–28.3 · P = 0.0003
Wilcoxon rank sum test · Intention-to-treat population
This secondary Week 6 analysis estimates a 18.5-meter difference between the randomized groups on the reported scale. Its confidence interval is 8.5–28.3 meters. Because this is a secondary endpoint assessed at a different time point from the primary Week 12 endpoint, it should be interpreted as a separate component of the trial evidence rather than as a replacement for the primary analysis.
Quality of Life: Global Score
Hodges-Lehmann estimate
95% CI: -8.0–0 · P = 0.027
Wilcoxon rank sum test · Intention-to-treat population
The global score analysis reports a Hodges-Lehmann estimate of -4.0 units, with a two-sided 95% confidence interval from -8.0 to 0. The ClinicalTrials.gov record identifies this as the Global Score but does not provide additional scale-direction information, so the statistical result is reported without assigning an independent favorable or unfavorable direction to the sign.
Quality of Life: Physical Dimension
Hodges-Lehmann estimate
95% CI: -3.0–0.0 · P = 0.037
Wilcoxon rank sum test · Intention-to-treat population
The physical-dimension analysis reports an estimated difference of -2.0 units. The confidence interval extends from -3.0 to 0.0, while the reported p-value is 0.037. As with the global score, the direction of a quality-of-life score should be interpreted according to the instrument's scoring convention rather than from the sign alone.
Quality of Life: Emotional Dimension
Hodges-Lehmann estimate
95% CI: -2.0–0.0 · P = 0.173
Wilcoxon rank sum test · Intention-to-treat population
The emotional-dimension estimate is -1.0 units, with a 95% confidence interval of -2.0 to 0.0 and a p-value of 0.173. The result illustrates why an estimated treatment difference should be considered together with its uncertainty rather than classified from the point estimate alone.
NT Pro-BNP
Hodges-Lehmann estimate
95% CI: -333 to -64 · P = 0.001
Wilcoxon rank sum test · Per-protocol population
The NT Pro-BNP analysis used a per-protocol population rather than the ITT population used for the primary endpoint. The reported estimate is -167 pg/mL, with a 95% confidence interval from -333 to -64 pg/mL. Because the analysis population differs from the primary efficacy analysis, the result should not be treated as though it arose from the same estimand and population.
8. Confidence Intervals and Effect Estimates
TRIUMPH provides a useful example of why a statistical result is more informative when the point estimate, confidence interval, analysis method, and analysis population are considered together.
A 95% confidence interval is a statement about the statistical procedure used to estimate the treatment difference. It should not be interpreted as a probability distribution for the true effect or as a range containing 95% of individual patient outcomes.
For the primary endpoint, the estimated difference is 20.0 meters and the 95% confidence interval is 8.0–32.8 meters. The interval therefore provides substantially more information than the p-value alone: it describes both the estimated direction and the precision of the treatment comparison on the outcome scale.
The secondary analyses show why confidence intervals are particularly useful when multiple endpoints are considered. For example, the Week 6 peak 6MWD estimate is 18.5 meters with a 95% confidence interval of 8.5–28.3 meters, while the Borg Dyspnea Score estimate is 0.0 with a 95% confidence interval of -0.5–0.0. These estimates are on different outcome scales and therefore should not be compared by magnitude alone.
9. Statistical Methods Explained
Why was ANCOVA used for the primary endpoint?
The primary endpoint is a continuous change in peak 6-minute walk distance from baseline to Week 12. ANCOVA is well suited to continuous outcomes because it can model the follow-up outcome or change while incorporating baseline information. Baseline adjustment can improve precision when baseline measurements explain some of the variation in the outcome. The registry identifies ANCOVA as the reported method; the complete model specification is not reported.
What does a 20.0-meter Hodges-Lehmann estimate mean?
It is the reported estimated location difference between the treatment groups for the primary endpoint. Because the outcome is measured in meters, the estimate is expressed in meters. It is not a statement that every treated participant gained 20.0 meters relative to a corresponding placebo participant.
Why use the Wilcoxon rank sum test?
The Wilcoxon rank sum test is a nonparametric method for comparing two groups. It uses the ordering or ranks of observations rather than relying on a normal-distribution assumption for the raw outcome. TRIUMPH used it for several secondary endpoints, including Borg Dyspnea Score, peak 6MWD at Week 6, quality-of-life dimensions, and NT Pro-BNP.
Why does the analysis population matter?
The primary endpoint used the intention-to-treat population, while NT Pro-BNP used a per-protocol population. An ITT analysis retains the randomized treatment assignment as the basis for comparison. A per-protocol analysis instead focuses on participants meeting protocol-defined criteria. Consequently, the two analyses answer somewhat different statistical questions and should not be combined into one undifferentiated efficacy estimate.
What does the primary 95% confidence interval tell us?
The 95% confidence interval of 8.0–32.8 meters describes the uncertainty associated with the estimated 20.0-meter treatment difference under the reported statistical framework. It does not imply that individual treatment effects fall between 8.0 and 32.8 meters, and it does not mean that there is a 95% probability that the true treatment effect lies inside this particular realized interval.
Why should the secondary p-values not be interpreted in isolation?
TRIUMPH has one registered primary endpoint and multiple secondary analyses. The ClinicalTrials.gov record identifies the methods and p-values but does not provide a multiplicity-adjustment procedure or endpoint hierarchy beyond the primary/secondary designations. Therefore, the secondary p-values should be interpreted as the reported results of their respective analyses rather than automatically treated as a separately error-controlled family of confirmatory tests.
10. Intention-to-Treat Analysis
The primary peak 6MWD analysis was performed in the intention-to-treat population. This is a central principle of randomized clinical-trial analysis because it preserves the treatment assignment created by randomization.
What ITT preserves
Randomized treatment assignment remains the basis of the comparison, helping preserve the causal structure created by randomization.
What ITT does not guarantee
ITT does not eliminate missing observations, nonadherence, treatment discontinuation, or other practical complications. Those issues still require appropriate handling and interpretation.
The ClinicalTrials.gov record explicitly identify one missing Week 12 Borg Dyspnea Score observation and state that it was not included in that analysis. This is an example of why "ITT population" and "complete data for every endpoint" are not synonymous concepts.
11. Per-Protocol Analysis
NT Pro-BNP was analyzed using a per-protocol population. This distinction is important because restricting the analysis population can change the composition of the compared groups and can therefore change the interpretation of the treatment effect.
A per-protocol result can provide information about outcomes among participants meeting protocol criteria, but it should not be described as equivalent to an ITT treatment effect. In TRIUMPH, the NT Pro-BNP result is therefore best read alongside its stated population, method, estimate, confidence interval, and p-value.
12. Safety Results
The ClinicalTrials.gov record provides serious adverse event counts by randomized treatment arm. They do not provide additional adverse-event categories or comparative statistical tests for safety.
| Safety measure | Inhaled Treprostinil | Placebo |
|---|---|---|
| Serious adverse events | 9/115 | 13/120 |
The denominator is important when reading these figures: the registry reports 9 affected out of 115 for inhaled treprostinil and 13 out of 120 for placebo. These are descriptive serious-adverse-event counts by arm. The ClinicalTrials.gov record does not provide a formal hypothesis test, confidence interval, exposure-adjusted rate, or detailed adverse-event classification, so none is inferred here.
13. Limitations and Interpretation Issues
- Limited model detail: the registry identifies ANCOVA and Wilcoxon rank sum methods but does not supply the full statistical model specifications.
- Effect-measure terminology: the primary analysis is reported as ANCOVA while the effect measure is listed as a Hodges-Lehmann estimate. The ClinicalTrials.gov record does not provide enough information to reconstruct the exact estimation procedure beyond the registry fields.
- Multiple secondary analyses: eight secondary statistical analyses are reported. The ClinicalTrials.gov record does not identify a multiplicity-adjustment procedure, so their p-values should not automatically be interpreted as individually error-controlled confirmatory findings.
- Different analysis populations: the primary analysis uses ITT, whereas NT Pro-BNP uses per protocol. Differences between these populations affect how their estimates should be interpreted.
- Missing observation: one Week 12 Borg Dyspnea Score observation was not present and was not included in that analysis.
- Endpoint scale: quality-of-life estimates are reported in units on a scale, but the ClinicalTrials.gov record does not provide additional scoring-direction information. The sign of those estimates should therefore not be given an independent clinical interpretation here.
- NYHA analysis: the registry records NYHA Functional Classification as a binary endpoint while reporting a Wilcoxon rank sum test and Hodges-Lehmann estimate. This page preserves the reported method rather than substituting an inferred alternative.
- Safety detail: only serious adverse-event counts by arm are reported. More detailed safety conclusions would require additional data.
- Generalizability: the trial record describes a specific phase 3 population of patients with severe pulmonary arterial hypertension; the ClinicalTrials.gov record does not provide detailed baseline characteristics or broader population descriptors.
14. Why This Trial Matters Statistically
TRIUMPH is a useful teaching case because the registry results bring together several common clinical-trial concepts within a relatively compact design: randomization, quadruple masking, an ANCOVA primary analysis, nonparametric secondary analyses, ITT and per-protocol populations, Hodges-Lehmann effect estimates, confidence intervals, and multiple endpoint analyses.
| Concept | How it appears in TRIUMPH |
|---|---|
| Randomization | The trial is recorded as randomized with 2 parallel treatment arms. |
| Blinding | The registry records quadruple masking. |
| Continuous outcome analysis | The primary peak 6MWD endpoint is continuous and analyzed using ANCOVA. |
| Nonparametric analysis | Several secondary outcomes use the Wilcoxon rank sum test. |
| Hodges-Lehmann estimate | The primary and secondary statistical analyses report Hodges-Lehmann effect estimates. |
| Confidence intervals | The reported analyses include two-sided 95% confidence intervals. |
| Intention-to-treat | The primary endpoint and most secondary endpoints use ITT populations. |
| Per-protocol analysis | NT Pro-BNP is analyzed using the per-protocol population. |
| Missing data | One missing Week 12 Borg Dyspnea Score observation is explicitly identified. |
| Multiplicity | One primary endpoint and multiple secondary analyses are reported; no adjustment procedure is reported in the ClinicalTrials.gov record. |
| Sample-size planning | The primary endpoint calculation specified an assumed difference, standard deviation, alpha, power, and required sample size. |
15. Reading the Primary Result Correctly
The primary reported estimate is 20.0 meters. This is the estimated treatment-group difference for the peak 6-minute walk distance endpoint under the reported analysis.
The 95% CI of 8.0–32.8 meters communicates uncertainty around that estimate. The interval is more informative than a p-value alone because it retains the outcome's measurement scale.
The reported P = 0.00044 indicates strong statistical evidence against the null hypothesis under the stated two-sided testing framework. It does not quantify clinical importance or the probability that the treatment is effective.
The result comes from an intention-to-treat ANCOVA analysis. The analysis population and method are essential parts of the result; the number 20.0 should not be detached from those conditions when interpreting it.
The primary result does not establish an individual-level effect of exactly 20.0 meters, does not provide a distribution of individual treatment responses, and does not by itself summarize the full safety or secondary-endpoint evidence.
16. Primary vs Secondary Evidence
The statistical structure of TRIUMPH makes the distinction between primary and secondary endpoints especially important. The primary endpoint was defined prospectively as change in peak 6-minute walk distance from baseline to Week 12. Secondary analyses examined additional measures at Week 6 and Week 12.
| Evidence layer | Endpoint examples | Interpretive role |
|---|---|---|
| Primary | Peak 6-minute Walk Distance at 12 weeks | Central prespecified efficacy comparison reported by the registry. |
| Secondary functional measures | Borg Dyspnea Score; NYHA Functional Classification; Trough 6MWD; Peak 6MWD at Week 6 | Additional measures of functional status and symptoms. |
| Quality of life | Minnesota Living With Heart Failure Global, Physical, and Emotional dimensions | Patient-reported outcome dimensions analyzed separately. |
| Biomarker | NT Pro-BNP | Secondary biochemical measure analyzed in the per-protocol population. |
This hierarchy does not make secondary endpoints uninformative. It means that their statistical role differs from the registered primary endpoint. In particular, the ClinicalTrials.gov record does not establish that all secondary p-values were subject to a common multiplicity-control strategy.
17. Understanding the Wilcoxon / Mann-Whitney Framework
The Wilcoxon rank sum test is often described as a comparison of two independent groups using ranks. Rather than directly modeling the arithmetic difference between group means, the method considers how observations from the two groups are ordered relative to one another.
The method is useful when a rank-based comparison is appropriate. The Hodges-Lehmann estimate adds an interpretable location-shift summary to the rank-based analysis.
In TRIUMPH, this method was used for Borg Dyspnea Score, NYHA Functional Classification, peak 6MWD at Week 6, all three reported Minnesota Living With Heart Failure dimensions, and NT Pro-BNP. The registry therefore provides a clear example of using different statistical methods for different outcome analyses within the same randomized trial.
18. Sample Size and Statistical Power
The registry states that the sample-size calculation was based on the primary endpoint, change in 6MWD at Week 12. The design assumptions were a 35m between-treatment difference, a standard deviation of 75m, a two-sided type I error of 0.05, and 90% power. The calculation required 100 subjects per treatment group, or 200 total, with an allowance for dropout.
Planning quantity
The assumed treatment difference was 35m. This was a planning assumption used to determine the required sample size, not the observed treatment estimate.
Observed quantity
The reported primary treatment estimate was 20.0 meters. A planned effect size and an observed effect estimate answer different questions.
Type I error
The sample-size notes specify an alpha of 0.05 using a two-sided p-value of less than 0.05.
Power
The design targeted 90% power under the stated sample-size assumptions.
This distinction between planning and results is fundamental. The 35m value should not be substituted for the observed 20.0-meter estimate, and the observed 20.0-meter estimate should not be treated as though it were the effect assumed when the study was designed.
19. Timeline
Trial begins
The registry lists the TRIUMPH start date as 2005-06.
Primary completion
The registry lists primary completion as 2007-10.
Outcome analyses available
The registry reports results for 9 outcome measures and 9 statistical analyses, including 1 primary-endpoint analysis with an estimate and confidence interval.
20. Limitations of the Reported Statistical Record
The registry provides enough information to reconstruct the main statistical structure and reported numerical results, but not every detail needed to reproduce the complete analysis from raw data.
- The ClinicalTrials.gov record does not include the complete ANCOVA model specification.
- The ClinicalTrials.gov record does not identify the exact covariates included in the ANCOVA models.
- The ClinicalTrials.gov record does not provide the full data-processing rules for every endpoint.
- The ClinicalTrials.gov record does not provide a formal multiplicity-adjustment strategy for the secondary analyses.
- The ClinicalTrials.gov record does not provide detailed baseline characteristics.
- The ClinicalTrials.gov record does not provide detailed adverse-event categories beyond serious adverse events by arm.
- The ClinicalTrials.gov record does not provide enough information to reconstruct individual-level distributions or participant-level treatment responses.
These limitations are reasons to distinguish a registry-based statistical interpretation from a full statistical-programming replication. The reported estimates and p-values can be described exactly as reported in the registry without implying access to the underlying patient-level dataset or complete statistical analysis plan.
21. Why the Analysis Population Changes the Question
| Population | TRIUMPH endpoint | Statistical implication |
|---|---|---|
| Intention to treat | Primary peak 6MWD | Maintains randomized assignment as the basis of the treatment comparison. |
| Intention to treat | Borg Dyspnea Score | One missing Week 12 observation was excluded from that endpoint's analysis. |
| Intention to treat | NYHA Functional Classification | Uses the randomized comparison under the reported nonparametric method. |
| Intention to treat | Trough 6MWD, Week 6 peak 6MWD, quality-of-life dimensions | Secondary efficacy analyses retain the ITT framework. |
| Per protocol | NT Pro-BNP | Restricts the analysis population and therefore has a different interpretation from an ITT estimate. |
When reading the trial, the analysis population should be treated as part of the result itself. An estimate without its population can be misleading because the population defines which participants contribute to the comparison and therefore which treatment question is being estimated.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: NCT00147199 — TRIUMPH.
- Linked publication: PubMed PMID 20430262.
Continue through the Clinical Biostats statistical pathway
Explore the underlying trial-design and statistical methods used to understand randomized clinical-trial results, from ANCOVA and confidence intervals to nonparametric testing and analysis populations.
25. Record Summary
TRIUMPH provides a useful example of how a randomized phase 3 trial can combine different statistical approaches across endpoints. Its primary analysis evaluated change in peak 6-minute walk distance at Week 12 using ANCOVA in the intention-to-treat population and reported a Hodges-Lehmann estimate of 20.0 meters, with a two-sided 95% confidence interval of 8.0–32.8 meters and a P-value of 0.00044. Secondary endpoints used both ANCOVA and Wilcoxon rank sum methods, with most analyzed in ITT populations and NT Pro-BNP analyzed per protocol.
The statistical lesson is not simply that one p-value crossed a conventional threshold. The more complete interpretation combines the estimated treatment difference, its confidence interval, the analysis method, the analysis population, the endpoint definition and time frame, and the distinction between the registered primary endpoint and the multiple secondary analyses. The reported serious adverse-event counts of 9/115 for inhaled treprostinil and 13/120 for placebo provide an additional descriptive safety dimension, while the ClinicalTrials.gov record does not support more detailed safety inference.