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
MELODY-1 was a randomized, parallel-group, quadruple-masked phase 2 trial comparing macitentan with placebo in subjects with combined pre- and post-capillary pulmonary hypertension due to left ventricular dysfunction. The registry reports 63 enrolled participants and one primary endpoint focused on significant fluid retention or worsening in NYHA functional class.
| Feature | MELODY-1 |
|---|---|
| Phase | Phase 2 |
| Condition | Pulmonary Hypertension |
| Study population | Subjects with combined pre- and post-capillary pulmonary hypertension due to left ventricular dysfunction |
| Design | Randomized, parallel-group, quadruple-masked |
| Primary purpose | Treatment |
| Enrollment | 63 |
| Interventions | Macitentan and placebo |
| Study status | Completed |
| Start date | 2014-07-01 |
| Primary completion date | 2015-11-01 |
| Lead sponsor | Actelion |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT02070991 |
2. Clinical Question
The clinical question was whether treatment with macitentan, compared with placebo, affected the occurrence of significant fluid retention or worsening in NYHA functional class through the registry-defined end-of-study period in subjects with combined pre- and post-capillary pulmonary hypertension due to left ventricular dysfunction.
Population
Subjects with combined pre- and post-capillary pulmonary hypertension due to left ventricular dysfunction.
Intervention
Macitentan.
Comparator
Placebo.
Primary question
Does macitentan change the occurrence of significant fluid retention or worsening in NYHA functional class compared with placebo?
3. Trial Design
Macitentan
- Macitentan was the active drug intervention.
- The primary endpoint was evaluated through the registry-defined end-of-study period.
- Secondary analyses included NT-proBNP, PVR, mPAP, mRAP, PAWP, cardiac index, and DPG at or through Week 12.
Placebo
- Placebo was the randomized comparator.
- The same primary endpoint framework was used for the treatment comparison.
- Secondary analyses compared Week 12 biomarker and hemodynamic measures with the macitentan group.
4. Endpoints
The registry identifies one primary endpoint and several secondary endpoints. The endpoints span a clinically defined participant-level outcome, a biomarker measure, and invasive hemodynamic measures.
| Endpoint role | Endpoint | Time frame |
|---|---|---|
| Primary | Number of Participants Experiencing Significant Fluid Retention or Worsening in NYHA Functional Class (FC) up to End-of-treatment | From randomization up to End-of-Study (Week 12 + 30 days follow-up) plus 1 calendar day |
| Secondary | NT-proBNP at Week 12 Expressed as Percent of Baseline NT-proBNP at Rest | From randomization up to end of treatment period (Week 12) |
| Secondary | PVR at Rest at Week 12 Expressed as Percent of Baseline PVR at Rest | From randomization up to end of treatment period (Week 12) |
| Secondary | Change From Baseline to Week 12 in Mean Pulmonary Arterial Pressure (mPAP) | From randomization up to end of treatment period (Week 12) |
| Secondary | Change From Baseline to Week 12 in Mean Right Atrial Pressure (mRAP) | From randomization up to end of treatment period (Week 12) |
| Secondary | Change From Baseline to Week 12 in Pulmonary Artery Wedge Pressure (PAWP) | From randomization up to end of treatment period (Week 12) |
| Secondary | Change From Baseline to Week 12 in Cardiac Index (CI) | From randomization up to end of treatment period (Week 12) |
| Secondary | Change From Baseline to Week 12 in Diastolic Pulmonary Vascular Pressure Gradient (DPG) | From randomization up to end of treatment period (Week 12) |
Primary endpoint definition
The primary endpoint counted participants who had at least one of the following: significant fluid retention, defined as an increase in body weight at any time by ≥ 5% or ≥ 5 kg from baseline due to fluid overload and/or parenteral administration of diuretics; or worsening of NYHA functional class from baseline.
The endpoint therefore combines two clinically distinct components into a single participant-level outcome. A participant meeting either component contributes to the primary endpoint count. Statistically, that makes the endpoint a binary outcome for the purpose of the reported Fisher exact comparison.
5. Primary Result
Significant Fluid Retention or Worsening in NYHA Functional Class
Difference between macitentan and placebo
95% CI: -15.07 to 33.26 · P = 0.3372
Fisher exact test · Two-sided 95% CI · Superiority hypothesis
| Primary endpoint | Macitentan vs placebo |
|---|---|
| Effect measure | Difference between maci. and placebo |
| Estimate | 10.08 |
| 95% CI | -15.07 to 33.26 |
| P-value | 0.3372 |
| Statistical method | Fisher Exact |
| Hypothesis | Superiority |
| Time frame | From randomization up to End-of-Study (Week 12 + 30 days follow-up) plus 1 calendar day |
The reported effect estimate is 10.08 for the difference between macitentan and placebo. In a two-group binary endpoint analysis, a difference measure describes separation between the groups rather than the absolute frequency of the outcome within either group. The registry does not provide the arm-specific primary endpoint counts in the statistical-analysis record, so the estimate should be interpreted in the form in which it was reported.
The 95% confidence interval of -15.07 to 33.26 is relatively wide and includes zero. Thus, the reported data are compatible with a difference in either direction as well as relatively little difference between the treatment groups. The interval describes uncertainty around the estimated group difference; it is not a range containing the individual outcomes of participants.
The P-value of 0.3372 is a measure of the evidence against the specified superiority null hypothesis under the Fisher exact framework. It is not a measure of the size, clinical importance, or probability of the treatment effect. A P-value should therefore be read alongside the effect estimate and its confidence interval rather than as a standalone measure of treatment benefit or harm.
Because the endpoint is defined as having at least one of two component events, interpretation also depends on the composition of that combined endpoint. Fluid retention and worsening NYHA functional class are clinically related but not identical events, and a composite can conceal differences in which component drives the overall result.
6. Secondary Results
The registry also reports seven secondary statistical comparisons between macitentan and placebo. Two endpoints are expressed as ratios of geometric means, while five use treatment effects based on mean change from baseline. All reported confidence intervals are two-sided 95% intervals.
NT-proBNP at Week 12
Ratio of geometric means
95% CI: 0.55 to 1.08
Superiority hypothesis · Percentage of baseline NT-proBNP
The estimated treatment effect was a ratio of geometric means of 0.77. A ratio below 1 indicates that the geometric mean outcome in the macitentan group was estimated to be lower relative to the placebo group on the reported scale. The 95% CI extends from 0.55 to 1.08, so the interval includes 1, the value corresponding to no ratio difference.
The registry notes that the values of 6 patients were missing in both the macitentan and placebo group. Missing observations can reduce the amount of information available for an endpoint and can affect the interpretation of an analysis if the patients with missing measurements differ systematically from those with observed measurements.
PVR at Rest at Week 12
Ratio of geometric means
95% CI: 0.64 to 1.36
Superiority hypothesis · Percentage of baseline PVR
The reported ratio of geometric means was 0.93. This is close to 1 on the ratio scale, while the confidence interval of 0.64 to 1.36 spans both below and above 1. The estimate therefore represents relatively modest separation on the reported ratio scale, with substantial uncertainty around it.
The registry reports missing PVR values for 11 patients in the macitentan group and 8 patients in the placebo group. The amount and distribution of missing data are important considerations when interpreting any Week 12 comparison based on observed measurements.
Mean Pulmonary Arterial Pressure
Treatment effect in mean change from baseline
95% CI: -4.3 to 4.9 mmHg
Superiority hypothesis · Week 12
The treatment effect for change from baseline in mPAP was 0.3 mmHg, with a 95% CI from -4.3 to 4.9 mmHg. Because the interval includes zero, the data are compatible with both a negative and positive treatment difference on the reported mean-change scale.
The registry reports that mPAP values were missing for 10 patients in the macitentan group and 7 patients in the placebo group.
Mean Right Atrial Pressure
Treatment effect in mean change from baseline
95% CI: -2.2 to 3.6 mmHg
Superiority hypothesis · Week 12
The estimated treatment effect for mRAP was 0.7 mmHg, with a 95% CI from -2.2 to 3.6 mmHg. The interval includes zero and therefore reflects uncertainty spanning both directions of treatment difference.
The registry reports missing mRAP values for 10 patients in the macitentan group and 7 patients in the placebo group.
Pulmonary Artery Wedge Pressure
Treatment effect in mean change from baseline
95% CI: -4.2 to 3.7 mmHg
Superiority hypothesis · Week 12
The reported treatment effect for PAWP was -0.3 mmHg, with a 95% CI from -4.2 to 3.7 mmHg. The point estimate is close to zero relative to the width of the confidence interval, and the interval includes zero.
The registry reports missing PAWP values for 11 patients in the macitentan group and 8 patients in the placebo group.
Cardiac Index
Treatment effect in mean change from baseline
95% CI: 0.11 to 0.69 L/min/m2
Superiority hypothesis · Week 12
The treatment effect for cardiac index was 0.40 L/min/m2, with a 95% CI of 0.11 to 0.69 L/min/m2. Unlike the other reported mean-change hemodynamic endpoints, this confidence interval does not include zero. The point estimate and interval therefore indicate separation from zero on the reported treatment-effect scale.
The registry reports missing cardiac-index values for 11 patients in the macitentan group and 8 patients in the placebo group.
Diastolic Pulmonary Vascular Pressure Gradient
Treatment effect in mean change from baseline
95% CI: -4.5 to 3.6 mmHg
Superiority hypothesis · Week 12
The estimated treatment effect for DPG was -0.4 mmHg, with a 95% CI from -4.5 to 3.6 mmHg. The confidence interval includes zero and is substantially wider than the magnitude of the point estimate.
The registry reports missing DPG values for 11 patients in the macitentan group and 8 patients in the placebo group.
| Secondary endpoint | Effect | 95% CI | Missing values noted by registry |
|---|---|---|---|
| NT-proBNP at Week 12, percent of baseline | Ratio of geometric means: 0.77 | 0.55 to 1.08 | 6 patients in both groups |
| PVR at Rest at Week 12, percent of baseline | Ratio of geometric means: 0.93 | 0.64 to 1.36 | 11 macitentan; 8 placebo |
| Change in mPAP | 0.3 mmHg | -4.3 to 4.9 | 10 macitentan; 7 placebo |
| Change in mRAP | 0.7 mmHg | -2.2 to 3.6 | 10 macitentan; 7 placebo |
| Change in PAWP | -0.3 mmHg | -4.2 to 3.7 | 11 macitentan; 8 placebo |
| Change in cardiac index | 0.40 L/min/m2 | 0.11 to 0.69 | 11 macitentan; 8 placebo |
| Change in DPG | -0.4 mmHg | -4.5 to 3.6 | 11 macitentan; 8 placebo |
7. Statistical Methodology
Fisher exact test for the primary endpoint
The primary endpoint is a participant-level binary outcome: each participant either experienced at least one qualifying event or did not. The registry reports Fisher Exact as the statistical method for the macitentan-versus-placebo comparison.
Fisher's exact test evaluates the treatment-group association for a binary outcome using the exact distribution of a 2 × 2 contingency table. It is particularly useful when sample sizes or event counts are small enough that large-sample approximations may be unreliable.
With an enrollment of 63 participants across two arms, an exact test provides a natural framework for a binary endpoint when the number of observed endpoint events may be limited. The reported effect estimate and confidence interval add information about the magnitude and precision of the group difference beyond the hypothesis test itself.
Ratio of geometric means
The NT-proBNP and PVR secondary analyses report treatment effects as ratios of geometric means. A ratio of 1 represents equality on the ratio scale. A ratio below 1 indicates a lower estimated geometric mean in the macitentan group relative to placebo, whereas a ratio above 1 indicates a higher estimated geometric mean.
The ratio is multiplicative rather than additive. Consequently, uncertainty is interpreted relative to 1 rather than relative to 0. The reported 95% confidence intervals for NT-proBNP and PVR can therefore be assessed by asking whether they contain 1.
Mean change from baseline
The mPAP, mRAP, PAWP, cardiac index, and DPG analyses use a treatment effect expressed as mean change from baseline. On this additive scale, zero represents no difference between treatment groups in the reported treatment effect.
A confidence interval containing zero therefore indicates that the data are compatible with a range of treatment differences that includes no difference. A confidence interval entirely above or below zero indicates separation from zero on that statistical scale.
Confidence intervals
Every posted statistical analysis in the registry uses a two-sided 95% confidence interval. The interval is a measure of statistical precision around the reported estimate under the corresponding analysis framework. It should not be interpreted as a probability distribution for the true treatment effect or as a range covering individual participant responses.
Missing observations
Several Week 12 secondary endpoints have substantial numbers of missing values relative to the total enrollment. The registry explicitly identifies the number of missing values for each of these analyses. Missingness matters because the precision and interpretation of a treatment comparison depend not only on the observed values but also on which participants contribute measurements.
The registry does not report an imputation method for these secondary endpoints in the statistical-analysis information summarized here. Accordingly, the reported treatment estimates should be interpreted using the analysis information that is actually specified in the registry.
8. Statistical Methods Explained
Why was Fisher's exact test used for the primary endpoint?
The primary outcome is naturally represented as a binary participant-level event: a participant either experienced significant fluid retention or worsening in NYHA functional class, or did not. Fisher's exact test directly evaluates a two-group binary comparison using an exact probability calculation rather than relying on a large-sample approximation. That is particularly relevant for a phase 2 trial with 63 enrolled participants.
What does a difference of 10.08 mean?
The registry identifies the primary effect measure as the difference between macitentan and placebo and reports an estimate of 10.08. The difference is a between-group effect measure, not a statement that 10.08% of participants experienced the endpoint. Without the arm-specific endpoint counts in the statistical-analysis record, the safest interpretation is to retain the registry's reported effect measure rather than reconstructing separate event rates.
Why does the primary confidence interval include zero?
The reported 95% CI is -15.07 to 33.26. Zero represents no difference on an additive difference scale. Because zero lies within the interval, the data are compatible with both a negative and a positive difference under the statistical framework used for the reported estimate.
What does a ratio of geometric means of 0.77 mean?
For NT-proBNP, the reported ratio is 0.77. A ratio below 1 means that the estimated geometric mean in the macitentan group was lower relative to placebo on the reported percentage-of-baseline scale. The associated 95% CI is 0.55 to 1.08, which includes 1, so the uncertainty interval spans the value representing no ratio difference.
Why is zero different from one in these analyses?
The answer depends on the effect measure. A difference is additive, so no difference is represented by zero. A ratio is multiplicative, so no ratio difference is represented by one. This distinction is essential when reading confidence intervals: the mPAP interval of -4.3 to 4.9 is evaluated relative to zero, whereas the NT-proBNP interval of 0.55 to 1.08 is evaluated relative to one.
How should the missing Week 12 measurements be interpreted?
The registry identifies missing measurements for several secondary endpoints. Missing data reduce the amount of observed information and can affect precision. More importantly, the statistical implications depend on why measurements are missing and how the missingness relates to treatment and outcome. A count of missing observations alone does not establish whether a particular missing-data mechanism is present.
9. Safety Results
The registry reports serious adverse events by randomized arm as affected participants divided by participants at risk. These figures provide an arm-level safety summary distinct from the efficacy endpoints.
| Safety measure | Macitentan | Placebo |
|---|---|---|
| Serious adverse events, affected / at risk | 11 / 31 | 6 / 32 |
The reported counts show that serious adverse events affected 11 of 31 participants at risk in the macitentan arm and 6 of 32 participants at risk in the placebo arm. These are descriptive safety data; they should not be substituted for the primary efficacy analysis or interpreted as an adjusted treatment effect.
The denominators also illustrate why safety summaries need their own analysis framework. The safety population represented by the reported at-risk denominators is not identical to simply quoting the total enrolled population of 63. Safety interpretation therefore depends on the population represented by the registry's safety measure.
10. Understanding the Primary Confidence Interval
The primary estimate is 10.08, while its 95% CI extends from -15.07 to 33.26. The interval is considerably wider than the point estimate itself. That means the observed estimate should not be treated as a precise measurement of the underlying treatment difference.
Because the interval extends on both sides of zero, the statistical uncertainty includes treatment differences in either direction. The point estimate alone therefore gives an incomplete picture of the evidence.
The 0.3372 P-value addresses the prespecified superiority hypothesis using Fisher's exact test. It does not tell us that the true treatment difference is 0.3372, nor does it quantify the magnitude of the observed difference. Effect estimates and confidence intervals remain essential for understanding the result.
11. Interpreting the Secondary Hemodynamic Findings
The secondary endpoints provide several different views of cardiovascular physiology at Week 12. Their statistical scales are important because they are not directly interchangeable.
| Measure | Reported effect scale | No-effect reference value | 95% CI includes no-effect value? |
|---|---|---|---|
| NT-proBNP | Ratio of geometric means | 1 | Yes: 0.55 to 1.08 |
| PVR | Ratio of geometric means | 1 | Yes: 0.64 to 1.36 |
| mPAP | Mean change from baseline | 0 | Yes: -4.3 to 4.9 |
| mRAP | Mean change from baseline | 0 | Yes: -2.2 to 3.6 |
| PAWP | Mean change from baseline | 0 | Yes: -4.2 to 3.7 |
| Cardiac index | Mean change from baseline | 0 | No: 0.11 to 0.69 |
| DPG | Mean change from baseline | 0 | Yes: -4.5 to 3.6 |
This table illustrates a central statistical principle: an estimate cannot be interpreted correctly without knowing its scale. For ratio measures, the reference value is 1; for additive treatment effects, the reference value is 0.
The cardiac-index analysis is the one reported secondary mean-change comparison whose 95% CI does not include zero. Its estimate is 0.40 L/min/m2, with a 95% CI of 0.11 to 0.69 L/min/m2. The other reported mean-change intervals include zero. The registry does not provide P-values for these secondary statistical analyses in the posted analysis information, so the confidence intervals are the principal inferential summaries available here.
12. Multiplicity and the Secondary Endpoints
MELODY-1 has one registered primary endpoint and multiple secondary endpoints. This distinction matters because the statistical evidence attached to a secondary endpoint should not automatically be treated as having the same confirmatory status as the primary endpoint.
Seven secondary statistical analyses are posted: NT-proBNP, PVR, mPAP, mRAP, PAWP, cardiac index, and DPG. When multiple secondary endpoints are examined, the chance of observing an apparently unusual result somewhere among the collection of analyses can increase if each is interpreted independently at the same nominal threshold.
Primary endpoint
The primary comparison has a clearly identified superiority hypothesis and a reported Fisher exact test with P = 0.3372.
Secondary endpoints
The secondary analyses provide supportive information across biomarkers and hemodynamic variables, but their interpretation should remain distinct from the primary endpoint.
The registry information does not specify an adjustment procedure for the family of secondary endpoints. The presence of several analyses is therefore an important consideration when interpreting any individual secondary confidence interval as evidence of a treatment effect.
13. Missing Data and Analysis Populations
The Week 12 secondary endpoints have varying amounts of missing data. The registry explicitly describes the missing observations for each analysis.
| Endpoint | Missing observations |
|---|---|
| NT-proBNP | The values of 6 patients are missing in both the macitentan and the placebo group. |
| PVR | The values of 11 patients in the macitentan group and of 8 patients in the placebo group are missing. |
| mPAP | The values of 10 patients in the macitentan group and of 7 patients in the placebo group are missing. |
| mRAP | The values of 10 patients in the macitentan group and of 7 patients in the placebo group are missing. |
| PAWP | The values of 11 patients in the macitentan group and of 8 patients in the placebo group are missing. |
| Cardiac index | The values of 11 patients in the macitentan group and of 8 patients in the placebo group are missing. |
| DPG | The values of 11 patients in the macitentan group and of 8 patients in the placebo group are missing. |
The unequal number of missing observations across some treatment groups is also relevant. Missingness itself is not proof of bias, but differential missingness can become an important concern when the probability of having an observed Week 12 value differs between randomized groups.
The registry does not identify a formal imputation strategy for these posted secondary analyses. Consequently, the interpretation should focus on the reported analysis population and missingness information rather than assuming that an unreported imputation method was used.
14. Why the Primary Endpoint Is Statistically Interesting
The MELODY-1 primary endpoint is an example of a composite binary endpoint. Instead of analyzing fluid retention and NYHA functional-class worsening separately as the primary outcome, the registry combines them into one endpoint defined by occurrence of at least one qualifying component.
The logical "OR" means that a participant meeting either component qualifies as having experienced the primary endpoint. This produces a single outcome suitable for a two-group categorical analysis.
Composite endpoints can improve efficiency by counting more than one clinically relevant type of event, but they require careful interpretation. The components can differ in clinical importance, frequency, and susceptibility to measurement or treatment effects. A statistically neutral composite can arise even if its individual components move in different directions.
For this trial, the registry's primary statistical result is therefore best understood as a comparison of the combined endpoint, not as a separate estimate for fluid retention and a separate estimate for NYHA functional-class worsening.
15. Effect Measures: Difference, Ratio, and Mean Change
One of the most useful statistical lessons from this trial is that several effect measures coexist in the same study. They answer related but different questions.
| Effect measure | Used for | Reference value | Interpretive question |
|---|---|---|---|
| Difference | Primary endpoint | 0 | How far apart are the groups on the reported additive scale? |
| Ratio of geometric means | NT-proBNP, PVR | 1 | How does the estimated geometric mean in macitentan compare multiplicatively with placebo? |
| Mean change from baseline | mPAP, mRAP, PAWP, cardiac index, DPG | 0 | How different are the groups in average change from baseline? |
Confusing these scales can produce incorrect conclusions. For example, the NT-proBNP estimate of 0.77 is not equivalent to a difference of 0.77, and the cardiac-index estimate of 0.40 L/min/m2 is not a ratio. The statistical meaning comes from the effect measure specified for each endpoint.
16. Limitations
- Sample size: the study enrolled 63 participants, so estimates can be sensitive to a relatively small number of endpoint observations.
- Primary confidence interval: the primary estimate of 10.08 has a 95% CI from -15.07 to 33.26, indicating substantial uncertainty around the treatment difference.
- Composite endpoint: the primary outcome combines significant fluid retention and worsening in NYHA functional class. The overall result does not separately quantify the treatment difference for each component.
- Missing Week 12 data: multiple secondary endpoints have missing measurements, including 11 macitentan and 8 placebo observations for several hemodynamic variables.
- Secondary multiplicity: several secondary endpoints were analyzed. The registry information does not specify a multiplicity-adjustment procedure for these posted analyses.
- Secondary endpoint uncertainty: most reported secondary confidence intervals include their respective no-effect values. The cardiac-index interval is the exception among the posted mean-change analyses.
- Limited statistical detail: the registry identifies the statistical methods and estimates but does not provide a complete statistical analysis plan for every secondary analysis in the posted information summarized here.
- Safety denominator: serious adverse events are reported as affected participants divided by participants at risk, with denominators of 31 and 32, rather than simply using the overall enrollment of 63.
17. Why This Trial Matters Statistically
MELODY-1 is a useful teaching case because it places several fundamental clinical-trial concepts into a compact randomized phase 2 design: a binary composite primary endpoint, an exact categorical test, additive and multiplicative secondary effect measures, confidence intervals, missing Week 12 measurements, and a separate safety analysis.
| Concept | How it appears in MELODY-1 |
|---|---|
| Randomization | Participants were randomly allocated to macitentan or placebo in a parallel-group design. |
| Masking | The registry describes the trial as quadruple-masked. |
| Binary endpoint | The primary endpoint counts participants experiencing significant fluid retention or worsening in NYHA functional class. |
| Fisher exact test | The registry reports Fisher Exact for the primary comparison. |
| Difference effect measure | The primary treatment effect is reported as a difference between macitentan and placebo. |
| Ratio of geometric means | NT-proBNP and PVR are reported using ratios of geometric means. |
| Mean change | mPAP, mRAP, PAWP, cardiac index, and DPG use treatment effects based on mean change from baseline. |
| Confidence intervals | All posted statistical analyses report two-sided 95% confidence intervals. |
| Missing data | The registry explicitly reports missing Week 12 measurements for the secondary endpoints. |
| Safety analysis | Serious adverse events are reported separately by arm as affected participants divided by participants at risk. |
18. What the Primary P-value Does — and Does Not — Mean
The primary P-value is 0.3372 from a Fisher exact test under a superiority hypothesis. It measures how compatible the observed categorical treatment comparison is with the null hypothesis used by that test. It does not measure the probability that macitentan works or does not work.
The effect estimate and P-value answer different questions. The 10.08 estimate describes the reported treatment difference, whereas 0.3372 describes statistical evidence under the specified hypothesis-testing framework. Neither number by itself describes the clinical importance of the finding.
The 95% CI from -15.07 to 33.26 is essential because it shows the uncertainty surrounding the estimate. A point estimate without its interval would hide the degree to which the treatment difference is statistically uncertain.
19. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoint was analyzed using Fisher's exact test. The reported difference was 10.08, with a two-sided 95% CI of -15.07 to 33.26 and P = 0.3372. The interval includes zero.
Clinical interpretation
The primary endpoint combines significant fluid retention and worsening in NYHA functional class. Its interpretation should remain focused on that combined clinical outcome rather than treating the statistical estimate as a direct measure of every component or every patient-level experience.
The secondary findings add physiological and biomarker information but should be interpreted according to their individual effect scales. The cardiac-index estimate has a 95% CI entirely above zero, while the confidence intervals for NT-proBNP, PVR, mPAP, mRAP, PAWP, and DPG include their respective no-effect values.
20. Planned and Reported Endpoint Framework
The registry distinguishes the primary endpoint from the secondary Week 12 measures. The primary endpoint extends beyond Week 12 through the specified end-of-study period, while the secondary endpoints are assessed through the end of the treatment period at Week 12.
| Endpoint family | Timing | Statistical role |
|---|---|---|
| Significant fluid retention or worsening in NYHA FC | Randomization through End-of-Study (Week 12 + 30 days follow-up) plus 1 calendar day | Primary superiority comparison |
| NT-proBNP | Through Week 12 | Secondary superiority comparison |
| PVR | Through Week 12 | Secondary superiority comparison |
| mPAP | Through Week 12 | Secondary superiority comparison |
| mRAP | Through Week 12 | Secondary superiority comparison |
| PAWP | Through Week 12 | Secondary superiority comparison |
| Cardiac index | Through Week 12 | Secondary superiority comparison |
| DPG | Through Week 12 | Secondary superiority comparison |
This difference in follow-up windows is important. The primary endpoint incorporates events through the registry-defined end-of-study period, whereas the secondary analyses describe Week 12 measurements. The estimates therefore summarize different portions and types of trial information.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: NCT02070991 — MELODY-1.
Continue through the Clinical Biostats knowledge graph
Connect the statistical methods used in clinical trials with deeper tutorials and practical statistical calculators.
24. Record Summary
MELODY-1 is a randomized phase 2 comparison of macitentan versus placebo in subjects with combined pre- and post-capillary pulmonary hypertension due to left ventricular dysfunction. Its primary endpoint is a clinically defined composite of significant fluid retention or worsening in NYHA functional class, analyzed with Fisher's exact test. The reported primary treatment difference is 10.08, with a two-sided 95% CI of -15.07 to 33.26 and P = 0.3372.
The secondary analyses demonstrate why effect measures must be interpreted on their proper statistical scales. NT-proBNP and PVR use ratios of geometric means, whereas mPAP, mRAP, PAWP, cardiac index, and DPG use mean change from baseline. The cardiac-index estimate is 0.40 L/min/m2 with a 95% CI of 0.11 to 0.69 L/min/m2; the other reported secondary confidence intervals include their respective no-effect values. Several Week 12 endpoints also have explicitly reported missing observations.