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
RELAX-AHF-EU was a randomized, parallel-group, open-label phase 3 trial evaluating serelaxin plus standard of care versus standard of care in patients with acute heart failure. The registry reports a binary primary endpoint concerning worsening heart failure or all-cause death through Day 5 and seven posted statistical analyses.
| Feature | RELAX-AHF-EU |
|---|---|
| Trial | RELAX-AHF-EU |
| Phase | Phase 3 |
| Condition | Acute Heart Failure (AHF) |
| Design | Randomized, parallel-group, open-label |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 2666 |
| Interventions | Serelaxin; Standard of Care |
| Primary endpoint | Percentage of Participants With Worsening Heart Failure (WHF) / All Cause of Deaths Through Day 5 |
| Primary endpoint type | Binary; primary analysis posted as a time-to-event analysis |
| Trial status | Terminated |
| Study start | 2014-01-31 |
| Primary completion | 2017-03-26 |
| Lead sponsor | Novartis Pharmaceuticals |
2. Clinical Question
The central question was whether serelaxin added to standard of care affected the occurrence of worsening heart failure or all-cause death through Day 5 in patients with acute heart failure.
Population
Patients with acute heart failure (AHF) enrolled in a phase 3 randomized trial.
Intervention
Serelaxin plus standard of care.
Comparator
Standard of care alone.
Primary question
Does serelaxin plus standard of care change the time-to-event experience for worsening heart failure or all-cause death through Day 5?
3. Trial Design
Serelaxin combination
- Serelaxin
- Standard of care
- Serious adverse events reported as 214 affected participants among 1729 at risk
Standard-of-care comparator
- Standard of care
- Serious adverse events reported as 107 affected participants among 894 at risk
The registry reports an enrollment of 2666, while the serious-adverse-event summaries use arm-specific affected/at-risk counts of 214/1729 and 107/894. Those safety denominators are the populations reported for that safety measure and should not be substituted for the overall enrollment figure.
4. Endpoints
Primary Endpoint
| Endpoint | Time frame | Definition |
|---|---|---|
| Percentage of Participants With Worsening Heart Failure (WHF) / All Cause of Deaths Through Day 5 | 5 days | In-hospital WHF through Day 5 post-randomization included worsening signs and/or symptoms of heart failure that required an intensification of intravenous therapy for heart failure or mechanical ventilation, renal or circulatory support. A central event adjudication committee was appointed to oversee the WHF primary endpoint adjudication. |
Secondary Endpoints
| Endpoint | Time frame | Type | Analysis |
|---|---|---|---|
| Percentage of Participants With In-hospital Worsening Heart Failure/All-Cause Death/Readmission for Heart Failure Through Day 14 | 14 days | Time-to-event | Gehan's generalized Wilcoxon test; HR |
| Percentage of Participants With Persistent Sign or Symptoms of Heart Failure / Non-Improvement at Any Post Baseline Visit Through Day 5 | 5 days | Binary | Chi-squared |
| Percentage of Participants With Renal Deterioration at Any Post Baseline Visit Through Day 14 | 14 days | Binary | Chi-squared |
| Length of Index Hospital Stay | 30 Days | Continuous | Wilcoxon (Mann-Whitney) |
| Change From Baseline in Health-related Quality of Life Index Value, Assessed by EuroQoL EQ-5D-5L Questionnaire. | Baseline, Day 5, Day 14 | Continuous / longitudinal | Mixed Models Analysis |
The registry contains two statistical analyses for the EQ-5D-5L outcome, corresponding to Day 5 and Day 14 analysis notes.
5. Statistical Methodology
Primary time-to-event analysis
The primary analysis used Gehan's generalized Wilcoxon test in the Full Analysis Set, comparing serelaxin plus standard of care with standard of care. The registry classifies the method as nonparametric and reports a hazard ratio as the effect measure.
The hypothesis type was superiority. The analysis population was the Full Analysis Set, and the analysis text also references the intention-to-treat principle.
Gehan's generalized Wilcoxon test
The Gehan generalized Wilcoxon approach is a rank-based comparison of time-to-event experience. Unlike a simple comparison of proportions at one fixed time point, it uses information from event timing and gives relatively greater emphasis to earlier events than the ordinary log-rank test.
That feature is particularly relevant when the clinical question concerns early worsening events during a short follow-up period. The registry's primary endpoint was defined through Day 5, so the statistical analysis is focused on the ordering of event times within that early window rather than on a long-term survival distribution.
Hazard ratio
The registry reports the effect measure for the primary analysis as a hazard ratio. An HR below 1 indicates a lower estimated instantaneous event rate in the serelaxin group relative to standard of care under the fitted analysis framework.
The statement follows directly from 1 − 0.71 = 0.29. It describes a relative hazard measure; it does not mean that 29% of participants avoided the endpoint or that each participant had exactly a 29% lower probability of an event.
Categorical analyses
The registry used chi-squared tests for the binary secondary endpoints concerning persistent signs or symptoms/non-improvement through Day 5 and renal deterioration through Day 14. These analyses compare categorical outcome distributions between the randomized treatment groups.
Wilcoxon / Mann-Whitney analysis
Length of index hospital stay was analyzed using the Wilcoxon (Mann-Whitney) test. This is a nonparametric two-group comparison that works from the ranks of observed values rather than requiring a normal-distribution assumption for the endpoint.
Mixed-effects analysis
The change from baseline in the EQ-5D-5L health-related quality-of-life index was analyzed using a mixed models approach at Day 5 and Day 14. Longitudinal models are useful when measurements are collected repeatedly from participants because observations from the same participant are correlated rather than independent.
The registry reports the EQ-5D-5L endpoint at baseline, Day 5, and Day 14, with separate posted analyses identified for Day 5 and Day 14.
6. Primary Result
The registry reports a formal statistical analysis for the primary endpoint in the Full Analysis Set using Gehan's generalized Wilcoxon test. The comparison was between serelaxin plus standard of care and standard of care, under a superiority hypothesis.
Worsening Heart Failure / All-Cause Death Through Day 5
Primary hazard ratio
95% CI: 0.51–0.98 · P = 0.0172
Two-sided confidence interval; Full Analysis Set; Gehan's generalized Wilcoxon test.
The estimated hazard ratio of 0.71 indicates an estimated instantaneous event rate approximately 29% lower with serelaxin plus standard of care than with standard of care in the analyzed comparison. The endpoint combined worsening heart failure and all-cause death through Day 5.
The HR does not mean that 29% of participants experienced a benefit, that 29% fewer participants necessarily had an event by Day 5, or that the absolute risk difference was 29 percentage points. A hazard ratio is a relative time-to-event measure.
The 95% confidence interval of 0.51–0.98 describes uncertainty around the estimated relative hazard under the analysis framework. Because the interval is below 1.00 at its upper boundary, the reported interval is consistent with a lower estimated hazard for the serelaxin comparison, although the interval remains sufficiently broad to include materially different effect sizes.
The P-value of 0.0172 quantifies the strength of evidence against the null hypothesis within the specified statistical test. It does not measure the size or clinical importance of the treatment effect. Effect size and precision are better described by the HR and confidence interval.
The analysis used Gehan's generalized Wilcoxon test rather than a simple fixed-time chi-squared comparison. Interpretation therefore depends on event timing as well as whether an event occurred during the analyzed period. As with any hazard-ratio interpretation, a single HR should not be treated as a complete description of the event process if the relative hazards vary substantially over time.
7. Secondary Time-to-Event Result
The registry also reports a time-to-event analysis through Day 14 incorporating in-hospital worsening heart failure, all-cause death, or readmission for heart failure.
Day 14 composite endpoint
95% CI: 0.61–1.02 · P = 0.0634
Full Analysis Set; Gehan's generalized Wilcoxon test; superiority hypothesis.
An HR of 0.79 corresponds to an estimated instantaneous event rate approximately 21% lower in the serelaxin plus standard-of-care group under the analyzed model. The endpoint differs from the primary endpoint because it extends through Day 14 and includes readmission for heart failure.
The 95% CI of 0.61–1.02 is wider than the point estimate alone suggests and crosses 1.00. The interval therefore includes both a reduction and a value close to no relative difference under the model.
The reported P = 0.0634 is a hypothesis-test result, not a measure of effect magnitude. It should not be interpreted as a 6.34% probability that the treatment works or as evidence that the observed HR is clinically unimportant. The appropriate description is the reported estimate, its uncertainty interval, and the stated hypothesis-testing result.
8. Secondary Categorical Results
Persistent Signs or Symptoms / Non-Improvement Through Day 5
Chi-squared analysis
Full Analysis Set; only evaluable patients with non-missing information included.
Serelaxin + standard of care vs standard of care; superiority hypothesis.
The registry reports a chi-squared test with P < 0.0001 for persistent signs or symptoms of heart failure/non-improvement at any post-baseline visit through Day 5. This is evidence from the specified categorical comparison against its null hypothesis, but the registry result does not provide an effect estimate or confidence interval for this analysis.
A very small P-value does not establish the size of the between-group difference. Without a reported risk difference, odds ratio, or group-specific percentages in the registry analysis, the magnitude and precision of the categorical effect cannot be summarized from this result alone.
Renal Deterioration Through Day 14
Chi-squared analysis
Full Analysis Set; only evaluable patients with non-missing information included.
Serelaxin + standard of care vs standard of care; superiority hypothesis.
The registry reports P = 0.0002 for renal deterioration at any post-baseline visit through Day 14 using a chi-squared test. The result indicates statistical evidence of a difference under the specified categorical test.
As with the Day 5 categorical endpoint, the P-value does not itself indicate how large the difference was. The registry analysis does not provide an effect estimate or confidence interval for this posted result, so the P-value should not be used as a substitute for an effect-size measure.
9. Hospital Stay Result
Length of Index Hospital Stay
Time frame: 30 Days · Wilcoxon (Mann-Whitney)
Full Analysis Set; serelaxin + standard of care vs standard of care.
The Wilcoxon (Mann-Whitney) analysis produced P = 0.1392. The test compares the distributions of length of index hospital stay through the specified 30-day time frame using ranks.
The registry does not report a treatment-group difference, confidence interval, or other effect estimate for this posted analysis. The P-value therefore provides the hypothesis-testing result but does not quantify the magnitude of any distributional difference.
It is also important not to interpret a nonsignificant P-value as proof that the two groups have identical hospital-stay distributions. It indicates that this particular test did not provide the specified level of statistical evidence against its null hypothesis.
10. Health-Related Quality of Life Results
The registry reports change from baseline in the EQ-5D-5L health-related quality-of-life index value at baseline, Day 5, and Day 14. A mixed models analysis was used, with separate posted analyses identified for Day 5 and Day 14.
| Analysis time point | Method | Analysis population | P-value |
|---|---|---|---|
| Day 5 | Mixed Models Analysis | Full Analysis Set; only evaluable patients with non-missing information included | 0.3115 |
| Day 14 | Mixed Models Analysis | Full Analysis Set; only evaluable patients with non-missing information included | 0.1236 |
The Day 5 mixed-model analysis reported P = 0.3115, while the Day 14 analysis reported P = 0.1236. These are longitudinal analyses of change from baseline in the EQ-5D-5L index value rather than time-to-event analyses.
The mixed-model framework is appropriate conceptually for repeated measurements because multiple observations from an individual are correlated. The model can account for the longitudinal structure more directly than treating every observation as independent.
The ClinicalTrials.gov record contains P-values but do not provide the corresponding treatment-effect estimates or confidence intervals for the EQ-5D-5L analyses. Consequently, the P-values should be interpreted as hypothesis-test results rather than as measures of the magnitude of change in health-related quality of life.
11. Safety: Serious Adverse Events
The registry reports serious adverse events by treatment arm using affected participants and participants at risk.
| Safety measure | Serelaxin + Standard of Care | Standard of Care |
|---|---|---|
| Serious adverse events | 214 / 1729 | 107 / 894 |
| Format | Affected / at risk | Affected / at risk |
The registry reports 214/1729 serious adverse events in the serelaxin plus standard-of-care group and 107/894 in the standard-of-care group. These are affected/at-risk counts as reported by the registry.
These figures are descriptive safety information. They should not be converted into a comparative risk estimate without knowing that the denominators correspond to the same prespecified safety population and endpoint definition and without the appropriate statistical analysis. The reported serious-adverse-event counts also use arm-specific at-risk denominators that differ from the overall trial enrollment of 2666.
12. Statistical Methods Explained
Why was Gehan's generalized Wilcoxon test used for the primary endpoint?
The primary endpoint was analyzed as a time-to-event outcome over a short five-day period. Gehan's generalized Wilcoxon test is a rank-based method that gives greater emphasis to earlier event times than a conventional log-rank comparison. That characteristic can matter when early worsening events are central to the clinical question.
What does an HR of 0.71 mean?
An HR of 0.71 means that the estimated instantaneous event rate for the serelaxin comparison was 71% of that in the standard-of-care group under the analysis framework. Equivalently, 1 − 0.71 = 0.29, so the estimated hazard was approximately 29% lower. It does not mean a 29% absolute reduction in the proportion of participants experiencing the endpoint.
Why is the confidence interval important?
The primary 95% confidence interval was 0.51–0.98. The point estimate of 0.71 is only one estimate from the data. The interval communicates the uncertainty surrounding that estimate under the statistical model and sampling framework. A confidence interval is therefore more informative about precision than the point estimate alone.
Why does the P-value not measure effect size?
The primary P-value of 0.0172 describes evidence against the null hypothesis under the specified test. P-values depend on both the observed data and the amount of information available. A small P-value can occur with a modest effect when a study is sufficiently informative, while a clinically meaningful effect can produce a less impressive P-value in a smaller study. The HR and confidence interval are needed to describe the estimated effect and its precision.
Why were chi-squared tests used for some secondary endpoints?
Persistent signs or symptoms/non-improvement and renal deterioration were registered as binary endpoints. A chi-squared test provides a direct comparison of categorical outcome distributions between the two treatment groups. The registry reports P-values for these analyses but not corresponding effect estimates or confidence intervals.
Why was a mixed model used for EQ-5D-5L?
The EQ-5D-5L outcome was measured longitudinally at baseline, Day 5, and Day 14. A mixed-effects framework can account for the correlation among repeated observations from the same participant and model the longitudinal outcome rather than treating each measurement as an independent observation.
What does the primary endpoint actually combine?
The primary endpoint combines worsening heart failure and all-cause death through Day 5. Registry-defined worsening heart failure included worsening signs and/or symptoms that required an intensification of intravenous therapy for heart failure or mechanical ventilation, renal or circulatory support. A central event adjudication committee oversaw WHF adjudication.
13. Intention-to-Treat and Analysis Populations
The registry's primary analysis identifies the Full Analysis Set as the analysis population and references the intention-to-treat principle in the analysis text. Several secondary binary and EQ-5D-5L analyses additionally specify that only evaluable patients with non-missing information were included.
| Analysis | Population reported |
|---|---|
| Primary WHF / all-cause death through Day 5 | Full Analysis Set |
| Day 14 WHF / death / readmission | Full Analysis Set |
| Persistent signs or symptoms / non-improvement | Full Analysis Set; only evaluable patients with non-missing information included |
| Renal deterioration | Full Analysis Set; only evaluable patients with non-missing information included |
| Length of index hospital stay | Full Analysis Set |
| EQ-5D-5L Day 5 and Day 14 | Full Analysis Set; only evaluable patients with non-missing information included |
This distinction matters statistically. Randomized comparisons derive their causal interpretation from maintaining treatment assignment as the basis for comparison. At the same time, an analysis that excludes participants with missing outcome information can represent a more restricted evaluable population for that particular endpoint.
14. Multiplicity and Multiple Secondary Endpoints
The registry reports one primary endpoint and multiple secondary endpoint analyses. The posted statistical analyses include time-to-event, binary, continuous, and longitudinal outcomes.
| Endpoint family | Number represented in posted analyses | Method |
|---|---|---|
| Primary time-to-event endpoint | 1 | Gehan's generalized Wilcoxon test |
| Secondary time-to-event endpoint | 1 | Gehan's generalized Wilcoxon test |
| Binary secondary endpoints | 2 | Chi-squared |
| Continuous secondary endpoint | 1 | Wilcoxon (Mann-Whitney) |
| Longitudinal EQ-5D-5L analyses | 2 | Mixed Models Analysis |
15. Why the Primary and Secondary Time-to-Event Analyses Are Different
The primary endpoint and the Day 14 secondary endpoint both use Gehan's generalized Wilcoxon test and report hazard ratios, but they represent different clinical outcome definitions and observation windows.
| Feature | Primary | Secondary |
|---|---|---|
| Time frame | 5 days | 14 days |
| Outcome | WHF / all-cause deaths | In-hospital WHF / all-cause death / readmission for heart failure |
| Effect measure | HR 0.71 | HR 0.79 |
| 95% CI | 0.51–0.98 | 0.61–1.02 |
| P-value | 0.0172 | 0.0634 |
| Statistical test | Gehan's generalized Wilcoxon | Gehan's generalized Wilcoxon |
These estimates should not be treated as two measurements of exactly the same outcome. The second endpoint adds readmission for heart failure and extends observation through Day 14. Changes in endpoint definition and follow-up window can change both the event process and the estimated treatment effect.
16. Understanding the Mixed-Effects Model Analysis
The EQ-5D-5L endpoint is structurally different from the primary time-to-event endpoint. Rather than asking whether and when a participant experienced worsening heart failure or death, it measures change in a health-related quality-of-life index over repeated assessment times.
Repeated observations
Participants were assessed at baseline, Day 5, and Day 14. Measurements from the same participant are expected to be correlated.
Modeling objective
The mixed-model framework evaluates longitudinal change while accounting for the repeated-measure structure.
Day 5 result
The registry reports P = 0.3115 for the Day 5 analysis.
Day 14 result
The registry reports P = 0.1236 for the Day 14 analysis.
The absence of a reported effect estimate and confidence interval in the posted analysis means the P-values should not be interpreted as quantitative measures of the size of the health-related quality-of-life difference.
17. Statistical Interpretation of P-values Across the Trial
| Endpoint | P-value | What the number represents |
|---|---|---|
| Primary WHF / all-cause death through Day 5 | 0.0172 | Evidence against the null hypothesis from the Gehan generalized Wilcoxon analysis |
| Day 14 WHF / death / readmission | 0.0634 | Evidence against the null hypothesis from the corresponding time-to-event analysis |
| Persistent signs or symptoms / non-improvement | <0.0001 | Evidence against the null hypothesis from the chi-squared analysis |
| Renal deterioration | 0.0002 | Evidence against the null hypothesis from the chi-squared analysis |
| Length of index hospital stay | 0.1392 | Evidence against the null hypothesis from the Wilcoxon analysis |
| EQ-5D-5L Day 5 | 0.3115 | Evidence against the null hypothesis from the mixed-model analysis |
| EQ-5D-5L Day 14 | 0.1236 | Evidence against the null hypothesis from the mixed-model analysis |
The table illustrates why P-values should not be compared as if they were a common scale of treatment benefit. The underlying endpoints, statistical models, effect measures, and observation windows differ. A P-value is tied to its particular statistical question.
18. What the Hazard Ratio Does — and Does Not — Mean
The primary HR of 0.71 means that, within the analyzed time-to-event framework, the estimated instantaneous event rate for the serelaxin plus standard-of-care group was approximately 71% of that in the standard-of-care group.
It does not mean that 29% of participants avoided worsening heart failure or death, that the absolute risk was reduced by 29 percentage points, or that every individual participant experienced a 29% reduction in risk.
The 95% CI of 0.51–0.98 shows the uncertainty surrounding the primary HR estimate. The lower and upper limits represent plausible values under the stated statistical framework rather than a range of individual treatment responses.
The primary analysis used a time-to-event method rather than simply comparing the percentage of participants with an event at Day 5. Participants who experience an event earlier contribute information about event timing, and the Gehan generalized Wilcoxon test gives relatively greater emphasis to earlier events.
19. Important Limitations and Interpretation Issues
- Short primary follow-up: the primary endpoint was defined through Day 5, so the primary treatment effect describes an early clinical period rather than a long-term outcome.
- Composite endpoint: the primary endpoint combines worsening heart failure and all-cause death. A composite can be influenced by the frequency and timing of its individual components.
- Different endpoint definitions: the Day 14 secondary endpoint adds readmission for heart failure and extends the time frame to 14 days, so its HR should not be treated as interchangeable with the primary HR.
- Effect estimates are incomplete for several secondary outcomes: the posted analyses for categorical, hospital-stay, and EQ-5D-5L endpoints report P-values without corresponding effect estimates and confidence intervals.
- Missing information in evaluable analyses: several secondary analyses specify that only evaluable participants with non-missing information were included.
- Hazard-ratio interpretation: a single HR summarizes relative event rates within the statistical framework but does not provide an absolute risk difference or describe individual treatment response.
- Nonparametric methods: the primary Gehan analysis and hospital-stay analysis are rank-based methods. Their interpretations differ from model-based mean comparisons or simple binary risk comparisons.
- Multiplicity: multiple secondary endpoints were analyzed under superiority hypotheses. Individual P-values should be interpreted within the context of the overall analysis plan rather than as isolated measures of evidence.
- Safety denominators: the serious-adverse-event summary uses arm-specific at-risk populations rather than the overall enrollment denominator.
20. Why This Trial Matters Statistically
RELAX-AHF-EU is useful as a teaching case because its posted analyses span several major classes of clinical-trial methods within the same randomized comparison.
| Concept | How it appears in RELAX-AHF-EU |
|---|---|
| Randomization | Randomized parallel-group comparison of serelaxin plus standard of care versus standard of care |
| Superiority testing | The primary and posted secondary analyses are identified with a superiority hypothesis |
| Time-to-event analysis | Primary Day 5 and secondary Day 14 composite outcomes |
| Gehan generalized Wilcoxon test | Primary and secondary time-to-event analyses |
| Hazard ratio | Effect measure for the posted time-to-event analyses |
| Confidence interval | 95% two-sided CI reported for the primary and Day 14 HR estimates |
| Chi-squared test | Binary analyses of persistent signs/symptoms or non-improvement and renal deterioration |
| Wilcoxon / Mann-Whitney test | Analysis of length of index hospital stay |
| Mixed-effects model | Longitudinal EQ-5D-5L analyses at Day 5 and Day 14 |
| Intention-to-treat principle | Referenced in the statistical analysis text |
| Missing data / evaluability | Several secondary analyses include only evaluable patients with non-missing information |
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: NCT02064868 — RELAX-AHF-EU.
- PubMed: PMID 35184572.
- PubMed: PMID 34514815.
Continue through the Clinical Biostats library
Connect the endpoints and statistical methods used in this trial with deeper tutorials and practical statistical tools.
24. Record Summary
RELAX-AHF-EU provides a compact example of how several statistical approaches can coexist within one randomized clinical trial. The primary endpoint was a short-term time-to-event composite analyzed using Gehan's generalized Wilcoxon test, with a reported HR of 0.71, 95% CI 0.51–0.98, and P = 0.0172. Secondary analyses extended the time-to-event framework to Day 14 and used chi-squared tests, a Wilcoxon (Mann-Whitney) test, and mixed-effects models for different endpoint structures.
The most useful statistical reading therefore distinguishes among effect estimates, confidence intervals, P-values, endpoint definitions, analysis populations, and statistical methods. The primary hazard ratio provides a relative time-to-event measure, while the categorical and longitudinal analyses address different clinical questions and should not be interpreted as interchangeable measures of treatment effect.