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
NEAT-HFpEF was a randomized, triple-masked phase 2 crossover trial evaluating whether isosorbide mononitrate increased daily activity compared with placebo in patients with heart failure.
| Feature | NEAT-HFpEF |
|---|---|
| Trial acronym | NEAT-HFpEF |
| NCT identifier | NCT02053493 |
| Phase | Phase 2 |
| Status | Completed |
| Condition | Heart Failure |
| Allocation | Randomized |
| Design model | Crossover |
| Masking | Triple |
| Primary purpose | Treatment |
| Enrollment | 110 |
| Interventions | Isosorbide Mononitrate; Placebo |
| Start date | 2014-04 |
| Primary completion date | 2015-02 |
| Lead sponsor | Adrian Hernandez |
| Sponsor type | Other |
2. Clinical Question
The central question was whether isosorbide mononitrate increases daily activity, assessed using 14-day averaged arbitrary accelerometry units, compared with placebo in patients with heart failure.
Population
Patients enrolled in the NEAT-HFpEF study for the treatment of heart failure.
Intervention
Isosorbide mononitrate.
Comparator
Placebo.
Primary question
Does isosorbide mononitrate increase daily activity compared with placebo?
3. Trial Design
The registry identifies NEAT-HFpEF as a randomized, triple-masked crossover trial with 2 treatment arms and 110 enrolled participants. A crossover design means that treatment comparison is made within a study in which participants receive the study interventions in different treatment periods rather than remaining in only one treatment group throughout the study.
Isosorbide Mononitrate
- Drug intervention
- Evaluated in a randomized crossover design
- Daily activity assessed with accelerometry
- Primary assessments occurred at 5-6 weeks and 11-12 weeks
Placebo
- Placebo intervention
- Evaluated within the same crossover framework
- Daily activity assessed with accelerometry
- Primary assessments occurred at 5-6 weeks and 11-12 weeks
4. Endpoints
The registry lists two primary endpoints, both based on arbitrary accelerometry units (AAU). The same activity construct was assessed at two time frames during the crossover study.
| Primary endpoint | Time frame | Registry definition |
|---|---|---|
| Arbitrary Accelerometry Units (AAU) (Phase I) | 5-6 weeks | To evaluate whether isosorbide mononitrate increases daily activity as assessed by 14-day averaged arbitrary accelerometry units in comparison to placebo. An arbitrary accelerometer unit is calculated within the accelerometer device that is worn by the patient and represents level of activity based on patient movement. Higher values indicate more movement. 0 indicates no movement. |
| Arbitrary Accelerometry Units (AAU) (Phase II) | 11-12 weeks | To evaluate whether isosorbide mononitrate increases daily activity as assessed by 14-day averaged arbitrary accelerometry units in comparison to placebo. An arbitrary accelerometer unit is calculated within the accelerometer device that is worn by the patient and represents level of activity based on patient movement. Higher values indicate more movement. 0 indicates no movement. |
What AAU measures
The registry defines an arbitrary accelerometer unit as a device-derived measure representing activity based on patient movement. Higher values indicate more movement, while 0 indicates no movement. The endpoint is therefore continuous rather than a binary event such as hospitalization or death.
5. Statistical Structure of the Primary Endpoints
The two primary endpoints measure the same underlying activity construct at different periods: 5-6 weeks and 11-12 weeks. Because the study is a crossover trial, the statistical structure is more complicated than a simple comparison of two unrelated groups.
6. Planned Analysis
The ClinicalTrials.gov record reports that results have been posted, but it does not post formal statistical analyses for either primary endpoint. The registry therefore provides the endpoint definitions and measurement windows without reporting an effect estimate, confidence interval, or p-value for the primary activity comparisons.
For a continuous outcome in a randomized crossover trial, the primary comparison would typically use a method that respects the within-participant pairing. A simple conceptual comparison is the within-participant treatment difference:
A positive difference would represent higher measured activity during the isosorbide mononitrate period, while a negative difference would represent lower measured activity during that period. A formal crossover analysis would generally also consider treatment period and treatment sequence where appropriate.
Depending on the prespecified statistical analysis plan, a paired analysis or a linear mixed-effects model could be used. A mixed model is particularly useful when the analysis needs to represent repeated measurements and other sources of within-participant correlation. The choice between these approaches should follow the prespecified analysis plan rather than being selected after seeing the outcome.
Phase I and Phase II should remain distinct
The registry defines separate primary endpoints at 5-6 weeks and 11-12 weeks. These are not simply duplicate labels for one measurement: they represent two specified assessment periods. Treating them as separate primary endpoints raises an important statistical question about how the multiplicity of primary comparisons was handled.
7. Statistical Methods Explained
Why is a crossover analysis different from an ordinary two-group comparison?
In an ordinary parallel-group trial, each participant contributes information to one randomized treatment group. In a crossover trial, participants can contribute information under more than one treatment condition. The observations from the same participant are correlated, so an analysis that ignores this pairing can misrepresent the amount of independent information in the study.
Why can within-participant comparisons be statistically useful?
Suppose one participant has generally high activity and another has generally low activity. A direct comparison of their raw activity levels contains substantial between-person variability. A crossover comparison instead focuses on how activity changes for the same participant between treatment periods. This can remove some stable person-level variation from the treatment comparison.
Why does the analysis need to consider treatment period?
Activity measured later in a study may differ from activity measured earlier for reasons unrelated to treatment. A period effect could therefore occur if outcomes systematically change between treatment periods. In a crossover model, treatment and period need to be distinguishable so that a treatment difference is not incorrectly attributed to the timing of measurement.
Why does treatment sequence matter?
Participants in a crossover trial receive treatments in different sequences. If the treatment received first affects the outcome measured during the subsequent period, sequence and carryover can complicate interpretation. The statistical model and study design therefore need to account for the order in which interventions are received when that is relevant to the estimand.
What does a 14-day average add to the analysis?
The registry defines the endpoint using 14-day averaged arbitrary accelerometry units. Averaging activity over multiple days can make the outcome less dependent on a single unusually active or inactive day. Statistically, however, the resulting endpoint remains a continuous measurement and should be analyzed using methods appropriate for its distribution and crossover structure.
Why are the 5-6 week and 11-12 week endpoints separate?
The registry specifies both time frames as primary endpoints. They represent activity assessments at different points in the treatment sequence. A complete confirmatory interpretation would therefore need to know how the statistical analysis plan handled the two primary comparisons, including any prespecified multiplicity strategy.
8. Interpreting a Crossover Treatment Difference
For a continuous crossover endpoint, the most direct effect measure is a difference in activity between the two treatment conditions within the same participants. The direction of that difference matters: because the registry states that higher AAU values indicate more movement, a positive isosorbide-mononitrate-minus-placebo difference would correspond to greater measured activity during isosorbide mononitrate exposure.
The difference would not by itself establish why activity changed. It would describe the measured treatment contrast under the statistical model and study conditions.
If a treatment difference were reported with a confidence interval, the interval would describe statistical uncertainty around the estimated population-level difference under the chosen model and assumptions. It would not describe the range of changes experienced by individual patients.
A p-value addresses the compatibility of the observed data with a specified null hypothesis under the statistical model. It does not tell the reader how large or clinically meaningful the treatment difference is. For a continuous activity endpoint, the estimated difference and its confidence interval are therefore essential parts of interpretation.
9. Crossover Design and Within-Participant Variability
The defining statistical feature of NEAT-HFpEF is its crossover structure. Randomization remains important because it determines the treatment sequence, while the crossover permits treatment conditions to be compared within participants.
| Feature | Statistical implication |
|---|---|
| Randomized | Treatment assignment or sequence is determined by randomization rather than participant choice. |
| Crossover | Participants are evaluated under more than one treatment condition, creating within-participant comparisons. |
| Triple masked | Knowledge of treatment assignment is limited across the masked study framework, reducing opportunities for differential behavior or assessment. |
| Continuous endpoint | AAU is a quantitative measure, so treatment effects can be expressed as differences in activity rather than only event proportions. |
| 14-day average | Each primary endpoint summarizes activity over 14 days rather than relying on a single activity measurement. |
10. Missing Data and Crossover Analysis Considerations
The registry information does not report a missing-data or imputation method. This is an important issue for a crossover trial because a participant may contribute complete information for one treatment period but not another.
A complete statistical analysis would need to define how incomplete treatment periods contribute to the analysis. Possible approaches include a model using all available repeated measurements under an appropriate missing-at-random assumption, a paired analysis restricted to participants with both treatment-period observations, or another prespecified method. These approaches can answer somewhat different practical questions and can make different assumptions about incomplete observations.
11. Safety Results
The registry reports serious adverse events by treatment sequence/arm. These figures describe affected participants relative to those at risk in each reported arm.
| Reported arm | Serious adverse events | Affected / at risk |
|---|---|---|
| Isosorbide Mononitrate Crossover to Plac | Serious adverse events | 2/51 |
| Placebo Crossover to Isosorbide Mononitr | Serious adverse events | 1/59 |
These are the serious-adverse-event figures reported by the registry for the two arms. Because the trial is a crossover study, the labels identify treatment sequences rather than functioning like simple permanent parallel-group treatment assignments.
12. Trial Timeline
Study start
The NEAT-HFpEF trial began in 2014-04.
Phase I primary assessment
Daily activity was assessed using 14-day averaged arbitrary accelerometry units.
Phase II primary assessment
Daily activity was again assessed using 14-day averaged arbitrary accelerometry units.
Primary completion
The registry lists 2015-02 as the primary completion date.
13. What the Registry Reports — and What It Does Not
The ClinicalTrials.gov record identifies a completed randomized crossover trial with 110 enrolled participants and two primary activity endpoints. It also reports serious adverse events by arm. Results are marked as posted, but formal statistical analyses are not posted for the two primary endpoints.
| Information category | Registry status |
|---|---|
| Trial design | Reported |
| Enrollment | 110 |
| Primary endpoints | Reported for 5-6 weeks and 11-12 weeks |
| Endpoint definitions | Reported |
| Results posted | Yes |
| Formal statistical analyses for primary endpoints | Not posted |
| Primary endpoint effect estimates | Not reported in the registry information |
| Primary endpoint confidence intervals | Not reported in the registry information |
| Primary endpoint p-values | Not reported in the registry information |
| Serious adverse events by arm | Reported |
This distinction is important statistically. A registry can identify the outcome that a study was designed to measure without providing the numerical analysis needed to quantify the observed treatment effect. The endpoint definition therefore establishes the estimand of interest, while the statistical estimate and uncertainty determine what can be concluded from the observed data.
14. Why This Trial Matters Statistically
NEAT-HFpEF is a useful teaching example because it combines randomization, triple masking, crossover treatment exposure, repeated activity measurement, continuous outcomes, and two prespecified primary assessment windows. Those features illustrate why trial design and statistical analysis cannot be separated.
| Concept | How it appears in NEAT-HFpEF |
|---|---|
| Randomization | The registry identifies the allocation as randomized. |
| Crossover design | The study model is explicitly identified as crossover. |
| Masking | The registry identifies triple masking. |
| Continuous outcome | Daily activity is measured using arbitrary accelerometry units. |
| Repeated measurement | Activity is summarized over 14 days within each primary assessment period. |
| Within-participant comparison | The crossover structure permits treatment-condition comparisons within participants. |
| Period effects | Different treatment periods can introduce timing-related differences that should be represented in the analysis. |
| Sequence effects | The order of treatment exposure can matter in crossover trials. |
| Missing data | Incomplete treatment periods can disrupt the intended paired comparison. |
| Multiplicity | Two primary endpoints are specified at different time frames. |
| Safety analysis | Serious adverse events are reported separately by treatment sequence/arm. |
15. Limitations and Interpretation Issues
- Incomplete statistical reporting: the registry reports that results have been posted, but formal statistical analyses for the primary endpoints are not posted.
- No primary effect estimates: the registry information does not provide numerical treatment-effect estimates, confidence intervals, or p-values for the two primary AAU endpoints.
- Crossover complexity: analysis must account for the correlation created when participants are evaluated under different treatment conditions.
- Period effects: activity can vary with time, making treatment period an important component of crossover interpretation.
- Sequence effects: the order in which participants receive treatments can influence interpretation if treatment effects persist or interact with study period.
- Missing observations: incomplete crossover data can prevent simple within-participant comparisons and may require model-based handling.
- Two primary time frames: the registry defines primary AAU endpoints at both 5-6 weeks and 11-12 weeks. The confirmatory interpretation of two primary comparisons depends on the prespecified multiplicity strategy.
- Measurement interpretation: AAU is a device-derived measure of movement. A numerical change in AAU describes measured activity and should not automatically be equated with a change in clinical outcomes such as hospitalization or survival.
- Safety denominator: the reported serious-adverse-event counts are presented within treatment sequence/arm denominators and should be interpreted in the context of the crossover design.
- Generalizability: the registry identifies the condition as heart failure, but the information presented here does not provide a detailed baseline-characteristics table from which broader population comparisons could be made.
16. A Statistical Walkthrough of the Primary Question
The cleanest way to understand the primary analysis is to start with the measurement rather than the hypothesis test. Each primary endpoint is a 14-day average of arbitrary accelerometry units. Higher values mean more movement, and 0 means no movement.
Step 1 · Measure
Activity is captured by an accelerometer and summarized as 14-day averaged AAU.
Step 2 · Compare
The treatment condition of interest is isosorbide mononitrate, with placebo as the comparator.
Step 3 · Respect pairing
The crossover design means observations from the same participant are related and should not be treated as independent observations.
Step 4 · Model time
The 5-6 week and 11-12 week assessment windows make treatment period a relevant part of the statistical structure.
Once these design features are represented correctly, the treatment contrast can be estimated with a method appropriate for a continuous crossover outcome. The inferential result would then consist of an effect estimate, a measure of uncertainty, and a hypothesis test if specified by the statistical analysis plan.
17. Crossover Design: A Deeper Statistical View
A simplified crossover model can be expressed conceptually as:
The participant component represents stable differences between people, while the treatment and period components represent systematic features of the crossover comparison. A formal model may include additional terms when required by the prespecified analysis plan.
The key point is that the participant effect is not merely statistical decoration. Participants differ substantially in their usual activity levels. A crossover design can use each participant's measurements under different treatment conditions to reduce the influence of those stable between-person differences.
That advantage comes with a tradeoff: the design relies on the validity of comparisons across treatment periods. If treatment effects persist into the next period, if disease status changes materially with time, or if participants do not complete both periods, the simple within-person contrast can become more difficult to interpret.
18. Multiplicity of the Two Primary Endpoints
The registry specifies two primary endpoints: AAU (Phase I) at 5-6 weeks and AAU (Phase II) at 11-12 weeks. Because both are identified as primary endpoints, the statistical interpretation should distinguish the number of prespecified primary comparisons from the number of measurements made on each participant.
| Endpoint | Role | Time frame | Statistical issue |
|---|---|---|---|
| AAU (Phase I) | Primary | 5-6 weeks | Continuous crossover outcome |
| AAU (Phase II) | Primary | 11-12 weeks | Continuous crossover outcome at a later period |
If both endpoints were intended to support a single confirmatory conclusion, the prespecified multiplicity strategy would be important for interpreting statistical significance. Possible strategies in clinical trials include requiring both endpoints to meet a criterion, allocating alpha between hypotheses, or specifying a hierarchical sequence. The registry information presented here does not report which strategy was used.
19. Safety and Efficacy Are Different Statistical Questions
The primary endpoints concern daily activity, whereas the registry separately reports serious adverse events. These outcomes should not be combined into a single numerical assessment without a prespecified benefit-risk framework.
Efficacy endpoint
AAU measures daily activity through accelerometry, with higher values indicating more movement.
Safety endpoint
Serious adverse events are reported as affected participants relative to the at-risk denominator for each reported treatment sequence/arm.
The statistical properties are different as well. AAU is continuous, while serious adverse events are event counts or proportions. A continuous treatment difference and an adverse-event risk comparison require different effect measures and potentially different models.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registry defines two randomized crossover comparisons of 14-day averaged activity, one at 5-6 weeks and one at 11-12 weeks. A valid analysis would need to preserve the within-participant structure and account for the timing of the treatment periods.
Clinical interpretation
AAU is a measure of physical movement captured by an accelerometer. Its clinical interpretation depends on the magnitude, precision, and context of any observed treatment difference rather than on the endpoint label alone.
21. Sources
- ClinicalTrials.gov: NEAT-HFpEF, NCT02053493.
- PubMed: PubMed record 32150314.
- PubMed: PubMed record 28588021.
- PubMed: PubMed record 26549714.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
Continue through the Clinical Biostats trial-analysis library
Explore additional clinical-trial statistical methods, study designs, endpoints, and analytical frameworks.
24. Record Summary
NEAT-HFpEF provides a useful example of the statistical issues created when a randomized trial uses a crossover design and a continuous activity endpoint. The registry identifies 110 enrolled participants, two treatment arms, triple masking, and two primary AAU endpoints assessed at 5-6 weeks and 11-12 weeks. Because the primary outcome is a 14-day average of accelerometry-derived movement, the treatment comparison is naturally expressed as a continuous difference rather than an event ratio. Because the study is a crossover trial, that difference should preserve the within-participant structure and account for treatment period and sequence as appropriate.
The registry reports that results have been posted and provides serious adverse-event counts by arm, but formal statistical analyses for the primary endpoints are not posted. The resulting statistical lesson is particularly important: identifying a prespecified endpoint and collecting the data are distinct from estimating the treatment effect and quantifying its uncertainty. For a crossover trial, the analysis must also reflect the paired nature of the observations rather than reducing the design to a conventional independent two-group comparison.