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 ClinicalTrials.gov record.
1. Trial at a Glance
INDIE-HFpEF was a randomized, triple-masked, phase 2 crossover trial evaluating nebulized sodium nitrite against placebo in patients with heart failure. The registered primary endpoint was peak VO2 after 4 weeks of treatment with inorganic nitrite compared with peak VO2 after 4 weeks of treatment with placebo, assessed by cardiopulmonary exercise testing at peak drug levels.
| Feature | INDIE-HFpEF |
|---|---|
| Trial name | INDIE-HFpEF |
| Brief title | Inorganic Nitrite Delivery to Improve Exercise Capacity in HFpEF |
| Phase | Phase 2 |
| Status | Completed |
| Condition | Heart Failure |
| Allocation | Randomized |
| Design model | Crossover |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 105 |
| Interventions | Nebulized Sodium Nitrite (drug); Placebo (drug) |
| Primary endpoint | Peak VO2 |
| Results posted | Yes |
| Statistical analyses posted | 10 |
| Registered start | 2016-08-10 |
| Primary completion | 2017-12-13 |
2. Clinical Question
The central statistical question was whether peak VO2 after treatment with nebulized inorganic nitrite differed from peak VO2 after placebo treatment in the crossover comparison.
Population
The trial enrolled 105 participants with the condition recorded as heart failure.
Intervention
Nebulized sodium nitrite, identified in the registry analyses as AIR001.
Comparator
Placebo.
Primary question
Does peak VO2 after 4 weeks of inorganic nitrite treatment differ from peak VO2 after 4 weeks of placebo treatment?
3. Trial Design
AIR001 Crossover to Placebo
- AIR001 treatment during one study phase
- Placebo treatment during the crossover phase
- Peak VO2 compared across the treatment periods
- Other outcomes analyzed using the same crossover comparison framework
Placebo Crossover to AIR001
- Placebo treatment during one study phase
- AIR001 treatment during the crossover phase
- Peak VO2 compared across the treatment periods
- Other outcomes analyzed using the same crossover comparison framework
The registry describes the comparison as AIR001 crossover to placebo versus placebo crossover to AIR001. This is important because each participant contributes information under both treatment conditions rather than remaining in only one treatment condition throughout the study.
4. Randomization, Masking, and Analysis Population
The combination of randomization, triple masking, and a crossover design creates a framework in which treatment-period observations can be compared while accounting for the repeated nature of the observations. The posted analysis used mixed models rather than treating all observations as independent.
5. Endpoints
| Endpoint | Registry definition / description | Time frame | Type |
|---|---|---|---|
| Peak VO2 | The primary endpoint will be the peak VO2 after 4 weeks treatment with inorganic nitrite as compared to the peak VO2 after 4 weeks treatment with placebo as assessed by cardiopulmonary exercise testing (CPET) performed at peak drug levels. | End of Phase 1 & End of Phase 2 | Continuous |
| Average Arbitrary Accelerometer Units (AAU) | Average accelerometry measurement. | End of Phase 1 & End of Phase 2 | Continuous |
| Medial E/e' Ratio as Measured by Echocardiography Core Lab | Medial E/e' ratio measured by echocardiography. | End of Phase 1 & End of Phase 2 | Continuous |
| Left Atrial Volume Index as Measured by Echocardiography | Left atrial volume index. | End of Phase 1 & End of Phase 2 | Continuous |
| Pulmonary Artery Systolic Pressure as Measured by Echocardiography | Pulmonary artery systolic pressure. | End of Phase 1 & End of Phase 2 | Continuous |
| Kansas City Cardiomyopathy Questionnaire (KCCQ) Clinical Score | KCCQ Clinical Score. | End of Phase 1 & End of Phase 2 | Continuous |
| N-terminal Pro-B-type Natriuretic Peptide Level (NT-proBNP) | NT-proBNP level. | End of Phase 1 & End of Phase 2 | Continuous |
| NYHA (New York Heart Association) Class | NYHA Class. | End of Phase 1 & End of Phase 2 | Binary |
| VE/VCO2 Slope (Ventilatory Efficiency) | VE/VCO2 slope as provided by the cardiopulmonary exercise testing core lab. | End of Phase 1 & End of Phase 2 | Continuous |
| VO2 at Ventilatory Threshold (Submaximal Exercise Capacity) | VO2 at ventilatory threshold as provided by the cardiopulmonary exercise testing core lab. | End of Phase 1 & End of Phase 2 | Continuous |
The posted statistical analyses compare patient-level data collected during the AIR001 period with data collected during the placebo period. The registry states that the outcome data are displayed as collected during each study phase and combined according to expected treatment assignment for analysis.
6. Primary Result: Peak VO2
The primary endpoint was peak VO2, measured in ml/kg/min at the end of Phase 1 and Phase 2. The registry reports a mixed-model analysis among all patients randomized, comparing the AIR001 and placebo treatment periods.
Mean difference in final peak VO2
95% CI: −0.56 to 0.16 · P = 0.2650
Analysis: mixed-effects model · Superiority hypothesis · Two-sided 95% CI
| Primary endpoint | AIR001 vs placebo | 95% CI | P-value | Analysis |
|---|---|---|---|---|
| Peak VO2 | Mean difference −0.20 ml/kg/min | −0.56 to 0.16 | 0.2650 | Mixed-effects model |
The estimated mean difference was −0.20 ml/kg/min for the AIR001 treatment period relative to the placebo treatment period under the reported analysis. Because the estimate is negative, the point estimate is in the direction of a lower final peak VO2 with AIR001 under this comparison.
The estimate does not mean that every participant experienced a decrease of 0.20 ml/kg/min. It is a population-level model-based difference in final values. It also does not by itself describe individual treatment response.
The 95% confidence interval of −0.56 to 0.16 describes the uncertainty around the estimated mean difference under the reported statistical framework. Because the interval spans zero, the data are compatible with both a negative and a positive mean difference of the sizes represented within the interval.
The P = 0.2650 is a measure of the compatibility of the observed data with the tested null hypothesis under the model and testing framework. It is not a measure of the size or clinical importance of the effect. A p-value should therefore not be interpreted as a probability that the treatment works or does not work.
The analysis used a mixed-effects model and included patients with incomplete data. The ClinicalTrials.gov record does not report the model's complete covariance structure, period-effect specification, sequence-effect specification, or a formal carryover analysis. Those details matter when interpreting a crossover model and cannot be reconstructed from the posted summary alone.
7. Secondary Endpoint Results
The registry contains nine posted secondary statistical analyses in addition to the primary peak VO2 analysis. All use mixed-model analysis, compare the AIR001 and placebo treatment periods, and analyze all patients randomized. The confidence intervals are two-sided 95% intervals.
| Secondary endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Average Arbitrary Accelerometer Units (AAU) | Mean difference | −14.68 | −263.65 to 234.29 | 0.9069 |
| Medial E/e' Ratio | Mean difference | 0.05 | −1.16 to 1.27 | 0.9338 |
| Left Atrial Volume Index | Mean difference | −0.29 | −2.79 to 2.20 | 0.8151 |
| Pulmonary Artery Systolic Pressure | Mean difference | 0.82 | −1.41 to 3.05 | 0.4658 |
| KCCQ Clinical Score | Mean difference | 1.06 | −1.37 to 3.49 | 0.3902 |
| NT-proBNP | Mean difference | 10.69 | −53.21 to 74.6 | 0.7404 |
| NYHA Class | Median difference | 0.05 | −0.08 to 0.18 | 0.4282 |
| VE/VCO2 Slope | Mean difference | −0.49 | −1.08 to 0.11 | 0.1067 |
| VO2 at Ventilatory Threshold | Mean difference | −0.10 | −0.37 to 0.16 | 0.4411 |
What the secondary results show statistically
The posted secondary estimates vary in direction and magnitude, but every registry-reported 95% confidence interval spans zero. The p-values range from 0.1067 to 0.9338. These observations describe the posted statistical results; they do not by themselves establish equivalence of AIR001 and placebo for any endpoint.
For several endpoints, the confidence intervals are substantially wider than the point estimates. That distinction is important: a non-small p-value and an interval spanning zero can reflect uncertainty about the magnitude and direction of an effect rather than proof that the treatment effect is exactly zero.
Secondary endpoint detail
Average AAU
The estimated mean difference was −14.68 accelerometry units, with a 95% CI of −263.65 to 234.29 and P = 0.9069. The interval is much wider than the point estimate.
Medial E/e' Ratio
The estimated mean difference was 0.05, with a 95% CI of −1.16 to 1.27 and P = 0.9338.
Left Atrial Volume Index
The estimated mean difference was −0.29 ml/m2, with a 95% CI of −2.79 to 2.20 and P = 0.8151.
Pulmonary Artery Systolic Pressure
The estimated mean difference was 0.82 mmHg, with a 95% CI of −1.41 to 3.05 and P = 0.4658.
KCCQ Clinical Score
The estimated mean difference was 1.06 units, with a 95% CI of −1.37 to 3.49 and P = 0.3902.
NT-proBNP
The estimated mean difference was 10.69 pg/mL, with a 95% CI of −53.21 to 74.6 and P = 0.7404.
NYHA Class
The posted effect measure was a median difference of 0.05, with a 95% CI of −0.08 to 0.18 and P = 0.4282.
VE/VCO2 Slope
The estimated mean difference was −0.49, with a 95% CI of −1.08 to 0.11 and P = 0.1067.
VO2 at Ventilatory Threshold
The estimated mean difference was −0.10 ml/min, with a 95% CI of −0.37 to 0.16 and P = 0.4411.
8. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm using the number affected and the number at risk.
| Treatment arm | Serious adverse events | At risk |
|---|---|---|
| AIR001 | 5 | 103 |
| Placebo | 4 | 102 |
These counts are descriptive safety results. The ClinicalTrials.gov record does not provide a statistical comparison, confidence interval, exposure-adjusted rate, event timing, or adverse-event severity classification beyond identifying these as serious adverse events. Accordingly, the counts should not be converted into a comparative safety conclusion from the registry summary alone.
9. Statistical Methodology
Mixed-effects model
The central statistical method reported for the primary and secondary analyses was a mixed-effects model. This is particularly relevant to a crossover trial because observations collected from the same participant under different treatment periods are not generally independent.
The participant-level random component provides a way to account for correlation among repeated observations contributed by the same individual. The exact fixed-effect and covariance specifications used in INDIE-HFpEF are not fully provided in the ClinicalTrials.gov record.
Why a mixed model fits a crossover design
In a conventional parallel-group trial, one participant usually contributes outcome information under one randomized treatment assignment. In a crossover design, the same participant contributes outcome information during more than one treatment period. A model that accounts for within-participant correlation can therefore use the repeated observations without incorrectly treating them as independent observations from different people.
Final-value comparison
The registry reports the effect measure for the continuous outcomes as Mean Difference (Final Values). The statistical comparison therefore focuses on the final measured value during the AIR001 period versus the final measured value during the placebo period, with the mixed model used to generate the treatment comparison and p-value.
Incomplete observations
The registry explicitly states that the mixed models used to generate the p-values included all patients, including those with incomplete data. This is important because a complete-case analysis would exclude participants with missing observations, whereas the reported analysis did not simply discard all such participants from the model.
The ClinicalTrials.gov record does not specify the imputation mechanism or a separate missing-data sensitivity analysis. Therefore, it would be inappropriate to characterize the approach as a particular multiple-imputation method or to claim a specific missing-at-random assumption from the registry summary.
Superiority framework
The posted analyses identify the hypothesis type as superiority. This means the statistical question was whether the treatment periods differed, rather than whether AIR001 was no worse than placebo by a prespecified non-inferiority margin.
For a continuous endpoint such as peak VO2, the reported mean difference is evaluated against zero. The ClinicalTrials.gov record does not report a non-inferiority margin or equivalence margin.
Confidence intervals
The posted confidence intervals are 95% and two-sided. A confidence interval gives an uncertainty range around the estimated treatment difference under the statistical model. It is more informative than the point estimate alone because it shows how precisely the treatment effect has been estimated.
P-values
The p-value measures how compatible the observed data are with the null hypothesis under the specified statistical model and testing framework. It is not a measure of the size of an effect, its clinical importance, or the probability that the null hypothesis is true.
10. Crossover Design and Within-Participant Comparisons
The crossover structure is one of the most statistically important features of INDIE-HFpEF. The two comparison groups are defined by treatment sequence: AIR001 crossover to placebo and placebo crossover to AIR001.
A crossover design can be statistically efficient because each participant can serve as their own comparison across treatment conditions. That can reduce some sources of between-person variability. At the same time, crossover studies introduce design questions that do not arise in the same way in a simple parallel-group trial, including possible period effects, sequence effects, and treatment carryover.
The ClinicalTrials.gov record does not report a formal carryover test, washout duration, or sequence-by-treatment interaction estimate. These omissions do not invalidate the posted results; they simply limit how specifically the crossover structure can be reconstructed from the registry summary.
11. Blinding and Randomization
INDIE-HFpEF is recorded as triple-masked and randomized. These design features operate at different levels.
Randomization
Random allocation helps make treatment assignment less systematically related to participant characteristics, supporting the validity of the treatment comparison.
Triple masking
Triple masking reduces the opportunity for knowledge of treatment assignment to influence behavior, assessment, or other aspects of the study process.
Crossover
Each participant contributes data under both treatment conditions, making within-participant dependence an important part of the statistical analysis.
Mixed-effects analysis
The reported model accounts for the repeated structure rather than treating all period-level measurements as independent.
12. Statistical Methods Explained
Why was a mixed-effects model used?
The trial used a crossover design, so participants contributed outcome measurements during more than one treatment period. Those observations from the same person are correlated. A mixed-effects model can account for that within-participant dependence while estimating the treatment-period difference.
What does a mean difference of −0.20 mean for peak VO2?
It means the estimated average final peak VO2 was 0.20 ml/kg/min lower during the AIR001 treatment period than during the placebo treatment period under the reported model. It is a population-level estimate, not a statement that every participant experienced a decrease of exactly 0.20 ml/kg/min.
What does the 95% confidence interval tell us?
The 95% confidence interval for peak VO2 was −0.56 to 0.16. It describes uncertainty around the estimated mean difference. Because zero lies inside the interval, the posted result is compatible with a range of treatment differences that includes both directions.
Why does P = 0.2650 not mean that the treatments are equivalent?
A non-small p-value does not establish equivalence. The trial was analyzed under a superiority framework, and no equivalence or non-inferiority margin is reported in the ClinicalTrials.gov record. Failing to reject a zero-difference null hypothesis is not the same as proving that two treatments have sufficiently similar effects.
Why is incomplete data relevant?
The registry states that the mixed models included all patients, including those with incomplete data. This means the posted analysis was not simply a complete-case comparison. However, the ClinicalTrials.gov record does not specify the model's missing-data assumptions or any sensitivity analysis, so the missing-data mechanism cannot be reconstructed in detail.
What is distinctive about the crossover comparison?
Unlike a parallel-group analysis, treatment exposure changes within participant. That makes the participant a natural source of within-person comparison, but it also means period and sequence considerations matter. The registry summary does not provide enough information to determine exactly how every possible crossover effect was modeled.
How should the p-values for the many secondary endpoints be interpreted?
Each posted p-value addresses its corresponding endpoint under the reported model. The ClinicalTrials.gov record does not report a multiplicity adjustment or an endpoint hierarchy for these secondary analyses. Therefore, the collection of secondary p-values should not automatically be interpreted as a family of independently confirmatory tests.
13. Multiplicity and the Secondary Endpoint Set
The registry reports one primary endpoint and nine secondary statistical analyses in the ClinicalTrials.gov recordset. All of the posted analyses use a superiority hypothesis and two-sided 95% confidence intervals.
| Analysis set | Number in the ClinicalTrials.gov record | Statistical interpretation |
|---|---|---|
| Primary endpoint analyses | 1 | Peak VO2 |
| Secondary endpoint analyses | 9 | Additional physiologic, functional, biomarker, and clinical outcomes |
| Total statistical analyses posted | 10 | One primary plus nine secondary analyses |
The ClinicalTrials.gov record does not identify a multiplicity adjustment across the secondary endpoints. This matters because examining many endpoints creates more opportunities for apparently unusual findings to occur by chance. The appropriate interpretation is therefore endpoint-specific unless the prespecified multiplicity strategy is available from additional statistical documentation.
14. Planned Analysis Features Not Reported in the ClinicalTrials.gov record
The registry data supports several important conclusions about the statistical approach, but it does not provide every component of a complete statistical analysis plan.
| Feature | What is supported by the ClinicalTrials.gov record |
|---|---|
| Non-inferiority margin | Not reported; the hypothesis type is superiority. |
| Equivalence margin | Not reported. |
| Carryover analysis | Not reported. |
| Period effect specification | Not reported in the registry-reported analysis summary. |
| Sequence effect specification | Not reported in the registry-reported analysis summary. |
| Multiplicity adjustment | Not reported in the ClinicalTrials.gov record. |
| Bayesian methods | Not reported. |
| Formal interim analysis | Not reported in the ClinicalTrials.gov record. |
| Specific imputation method | Not reported; the registry states that mixed models included patients with incomplete data. |
| Exact mixed-model covariance structure | Not reported. |
This distinction is important for educational interpretation. A statistical analysis page should not fill gaps in a registry record by assuming that a conventional crossover method, missing-data method, or multiplicity strategy must have been used.
15. Understanding the Primary Confidence Interval
Peak VO2 treatment difference
95% CI: −0.56 to 0.16
P = 0.2650
The point estimate is relatively close to zero compared with the width of the confidence interval. That makes the interval especially important for interpretation. The analysis does not identify a single precise value as the only plausible treatment difference; rather, the reported uncertainty extends from a negative difference through zero to a positive difference.
The confidence interval and p-value answer related but different questions. The interval shows the range of treatment differences compatible with the statistical uncertainty of the estimate, while the p-value summarizes evidence against the specified null hypothesis. Neither quantity measures clinical importance by itself.
16. Interpreting the Secondary Results Without Overreading Them
The secondary results illustrate why a table of p-values is not enough to understand a clinical trial.
| Endpoint | Estimate | 95% CI | What the interval contributes |
|---|---|---|---|
| Average AAU | −14.68 | −263.65 to 234.29 | Shows substantial uncertainty around the point estimate. |
| Medial E/e' Ratio | 0.05 | −1.16 to 1.27 | Includes zero and both directions of effect. |
| Left Atrial Volume Index | −0.29 | −2.79 to 2.20 | Includes zero and a broad range around the estimate. |
| Pulmonary Artery Systolic Pressure | 0.82 | −1.41 to 3.05 | Includes zero and both positive and negative differences. |
| KCCQ Clinical Score | 1.06 | −1.37 to 3.49 | Does not isolate a single direction with high precision. |
| NT-proBNP | 10.69 | −53.21 to 74.6 | Wide relative to the point estimate. |
| NYHA Class | 0.05 | −0.08 to 0.18 | Includes zero under the reported median-difference measure. |
| VE/VCO2 Slope | −0.49 | −1.08 to 0.11 | Includes zero and remains uncertain in direction. |
| VO2 at Ventilatory Threshold | −0.10 | −0.37 to 0.16 | Includes zero and both directions of effect. |
The statistical lesson is broader than this particular trial: an estimated effect should always be read together with its uncertainty interval and the analysis population and model that produced it.
17. Safety Interpretation
The serious-adverse-event counts were 5 of 103 for AIR001 and 4 of 102 for placebo.
The chart is intentionally descriptive rather than a formal comparative safety analysis. The ClinicalTrials.gov record gives affected participants and participants at risk but does not provide a confidence interval or hypothesis test for the serious-adverse-event comparison.
18. Trial Timeline
Trial start
The registry lists August 10, 2016 as the study start date.
Primary completion
The registry lists December 13, 2017 as the primary completion date.
Results posted
The trial is recorded as completed, with results posted and 10 statistical analyses available in the ClinicalTrials.gov record.
19. What This Analysis Can and Cannot Establish
What is directly reported
The registry reports a randomized, triple-masked crossover design, a mixed-effects analytical method, a primary peak VO2 comparison, nine secondary analyses, confidence intervals, p-values, and serious-adverse-event counts.
What the primary estimate says
The estimated mean difference in final peak VO2 was −0.20 ml/kg/min, with a 95% CI of −0.56 to 0.16 and P = 0.2650.
What it does not say
The primary estimate does not describe individual responses, prove equivalence, or establish a probability that the treatment is effective or ineffective.
What the registry record does not include
The ClinicalTrials.gov record does not provide all details of the crossover model, missing-data assumptions, multiplicity strategy, or period and carryover analyses.
20. Important Limitations and Interpretation Issues
- Crossover-specific effects: the trial compares treatment periods within a crossover design, but the ClinicalTrials.gov record does not report a formal carryover analysis, washout duration, or complete period and sequence modeling details.
- Mixed-model specification: the registry identifies the analysis as a mixed-effects model but does not provide the complete model specification or covariance structure in the ClinicalTrials.gov record.
- Incomplete observations: the registry states that all randomized patients, including those with incomplete data, were included in the mixed models. The specific missing-data assumptions and sensitivity analyses are not reported in the ClinicalTrials.gov recordset.
- Multiplicity: nine secondary endpoint analyses were posted in addition to the primary endpoint, but the ClinicalTrials.gov record does not identify a multiplicity adjustment.
- Secondary endpoints: the secondary analyses should not automatically be treated as confirmatory evidence simply because they have reported p-values and confidence intervals.
- Endpoint measurement: several outcomes depend on specialized measurement procedures, including cardiopulmonary exercise testing and echocardiography core-lab assessments.
- Safety: the ClinicalTrials.gov record is limited to serious-adverse-event counts and denominators; it does not provide a full adverse-event profile or comparative safety analysis.
- Generalizability: the ClinicalTrials.gov record identifies heart failure as the condition but does not provide a detailed baseline-characteristics table, so the representativeness of the enrolled population cannot be fully assessed from this record alone.
21. Why This Trial Matters Statistically
INDIE-HFpEF is a useful statistical teaching case because it combines randomization, triple masking, crossover treatment exposure, repeated measurements, mixed-effects modeling, incomplete observations, continuous and non-continuous endpoints, confidence intervals, p-values, and multiple secondary outcomes.
| Concept | How it appears in INDIE-HFpEF |
|---|---|
| Randomization | The registry identifies the allocation as randomized. |
| Blinding | The study is recorded as triple-masked. |
| Crossover design | Participants contribute treatment-period data under AIR001 and placebo. |
| Repeated observations | Outcome data are collected at the end of Phase 1 and Phase 2. |
| Mixed-effects model | Used for the primary and posted secondary analyses. |
| Mean difference | Used as the reported effect measure for the continuous outcomes. |
| Median difference | Used as the reported effect measure for NYHA Class. |
| Confidence intervals | All registry-reported statistical analyses report two-sided 95% confidence intervals. |
| P-values | Reported for the primary and secondary statistical analyses. |
| Incomplete data | The registry states that mixed models included patients with incomplete data. |
| Multiplicity | Nine secondary analyses accompany the primary analysis; no adjustment is reported in the ClinicalTrials.gov record. |
| Safety analysis | Serious adverse events are reported by treatment arm with affected and at-risk counts. |
22. A Practical Framework for Reading the Results
A useful way to read the INDIE-HFpEF results is to move through the evidence in a fixed order rather than starting with the p-value.
For peak VO2, this framework produces a more complete interpretation than simply saying that the result was or was not statistically significant. The endpoint was peak VO2; the estimated mean difference was −0.20 ml/kg/min; the 95% CI was −0.56 to 0.16; the p-value was 0.2650; and the estimate came from a mixed-effects model applied to a crossover treatment comparison.
That final design step is essential. The same numerical difference could have a different interpretation if it arose from a parallel-group model, a paired analysis, or a model with different missing-data assumptions.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: NCT02742129 — INDIE-HFpEF.
- PubMed: PMID 32150314.
- PubMed: PMID 30398602.
- PubMed: PMID 28476756.
Continue through Clinical Biostats
Use the related tutorials and statistical calculators to explore the design, inference, and modeling concepts illustrated by this trial.
26. Record Summary
INDIE-HFpEF provides a compact but statistically rich example of a randomized crossover clinical trial. The registry identifies a triple-masked phase 2 design with 105 enrolled participants, two treatment conditions, and a primary peak VO2 endpoint evaluated at the end of Phase 1 and Phase 2. The primary analysis used a mixed-effects model and estimated a mean difference of −0.20 ml/kg/min, with a two-sided 95% confidence interval of −0.56 to 0.16 and P = 0.2650.
The nine secondary analyses use the same broad mixed-model framework and provide additional information across accelerometry, echocardiographic measures, KCCQ, NT-proBNP, NYHA Class, ventilatory efficiency, and submaximal exercise capacity. Their confidence intervals all include zero in the ClinicalTrials.gov record. Serious adverse events were reported as 5/103 for AIR001 and 4/102 for placebo.
The most important statistical lesson is not simply the primary p-value. The trial illustrates how randomization, crossover design, within-participant correlation, mixed-effects modeling, confidence intervals, missing observations, and multiple secondary endpoints interact when a clinical trial is interpreted. The registry supplies the principal numerical results but does not provide enough detail to reconstruct every element of the crossover model, multiplicity strategy, or missing-data analysis.