This page separates reported trial information from statistical interpretation. The ClinicalTrials.gov record contains the registry endpoint definition, two formal statistical analyses of the Minnesota Living With Heart Failure score, and serious adverse-event counts by randomized arm. No formal statistical analysis for the registered primary endpoint was reported in the ClinicalTrials.gov recordset used here.
1. Trial at a Glance
I-PRESERVE was a randomized, double-blind, parallel phase 3 trial evaluating irbesartan versus placebo in participants with congestive heart failure. The ClinicalTrials.gov record reports enrollment of 4128 participants and identify a time-to-event primary endpoint involving death from any cause or protocol-specified cardiovascular hospitalization.
| Feature | I-PRESERVE |
|---|---|
| Trial name | I-PRESERVE |
| Brief title | Irbesartan in Heart Failure With Preserved Systolic Function (I-Preserve) |
| Phase | Phase 3 |
| Condition | Congestive Heart Failure |
| Design | Randomized, double-blind, parallel |
| Primary purpose | Treatment |
| Enrollment | 4128 |
| Arms | 2 |
| Interventions | Irbesartan; Placebo |
| Sponsor | Bristol-Myers Squibb |
| Sponsor type | Industry |
| Status | Completed |
| Primary endpoint type | Time-to-event |
| Results posted | Yes |
| Outcome measures posted | 19 |
| Statistical analyses posted | 2 |
2. Clinical Question
The trial's clinical question concerns whether randomized treatment with irbesartan, compared with placebo, differs with respect to outcomes in participants with congestive heart failure. The registered primary endpoint is a time-to-event composite involving all-cause death or protocol-specified cardiovascular hospitalization.
Population
Participants enrolled in the phase 3 I-PRESERVE trial with the condition listed as congestive heart failure.
Intervention
Irbesartan.
Comparator
Placebo.
Primary question
How does randomized irbesartan treatment compare with placebo for the registered time-to-first occurrence of the composite outcome of all-cause death or protocol-specified cardiovascular hospitalization?
3. Trial Design
Irbesartan
- Intervention: irbesartan
- Randomized treatment arm
- Double-blind trial design
Placebo
- Comparator: placebo
- Randomized control arm
- Double-blind trial design
The allocation was randomized and the design model was parallel. Together with double masking, this structure is intended to create comparable treatment groups while reducing the influence of treatment knowledge on participant behavior, assessment, and other aspects of trial conduct.
4. Randomization and Blinding
Randomization is the central design feature supporting a causal comparison between irbesartan and placebo. Because treatment assignment is randomized, baseline prognostic factors are expected to be distributed between groups through the randomization mechanism, although randomization does not guarantee numerical balance in every individual characteristic.
The registry describes the trial as double masked. In a double-blind trial, knowledge of assigned treatment is restricted for the relevant parties according to the trial's masking procedures. This can reduce the potential for differential behavior, reporting, treatment management, or assessment associated with knowing the assigned intervention.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame | Type |
|---|---|---|---|
| Percentage of Participants With First Occurrence of the Composite Outcome of Death (All Cause) or Protocol-Specified Cardiovascular (CV) Hospitalization at Given Timepoints | Treatment comparisons for time to first occurrence of composite outcome of all-cause death (composite outcome of death) or protocol-specified CV hospitalization. Protocol-specified CV hospitalizations include those ≥24 hrs or involving a calendar date change for a primary cause of worsening heart failure, unstable angina, myocardial infarction, ventricular or atrial dysrhythmia, or stroke, that also require intravenous or intramuscular therapy or a related procedure or significant augmentation of oral therapy. In addition, MI or stroke during any hospitalization are included. | Year 1, Year 2, Year 3, Year 4, Year 5 | Time-to-event |
The registry definition identifies a time-to-first-event composite. The two broad event components are all-cause death and protocol-specified cardiovascular hospitalization. The registry-reported definition further specifies qualifying cardiovascular hospitalizations by duration or calendar-date change and by their primary causes.
6. Planned Analysis of the Primary Endpoint
A time-to-event endpoint of this form is ordinarily analyzed using methods that account for both event timing and right censoring. Kaplan-Meier methods can describe the estimated event-free distribution over time, while a log-rank framework can compare randomized groups. A Cox proportional-hazards model can provide a relative treatment-effect estimate such as a hazard ratio.
Here, T represents time from the relevant trial origin to the first qualifying component of the composite event. Participants without an observed event contribute information until their censoring time.
Because the ClinicalTrials.gov record does not specify the formal primary-endpoint statistical model, this page treats those methods as the typical analytical framework for an endpoint of this type, not as a claim about the exact analysis used in I-PRESERVE.
Why a composite time-to-event endpoint can be useful
Combining all-cause death with qualifying cardiovascular hospitalization can capture more clinically consequential events than death alone. Statistically, however, a composite endpoint is meaningful only if the component definitions are prespecified and clinically interpretable. The registry definition is therefore important: it specifies not simply "hospitalization," but a defined class of protocol-specified cardiovascular hospitalizations.
7. Secondary Endpoint Results: Minnesota Living With Heart Failure Score
The statistical analyses posted on ClinicalTrials.gov concern the Minnesota Living With Heart Failure (MLwHF) Total Score, defined as the sum of Questions 1–21. The endpoint was evaluated at baseline, Month 6, and Month 14. The analysis population consisted of participants with both a baseline score and a score at the relevant timepoint.
| Comparison | Method | Effect measure | Estimate | 95% CI |
|---|---|---|---|---|
| Placebo - Month 6 vs Irbesartan - Month 6 | Mixed Models Analysis | Mean Difference (Final Values) | 0.2 | -0.8 to 1.3 |
| Placebo - Month 14 vs Irbesartan - Month 14 | Mixed Models Analysis | Mean Difference (Final Values) | 0.6 | -0.5 to 1.7 |
Month 6 analysis
Adjusted mean difference in MLwHF total score
95% CI: -0.8 to 1.3 · two-sided 95% confidence interval
Comparison: Placebo - Month 6 vs Irbesartan - Month 6
The estimated mean difference in final MLwHF total score was 0.2, defined in the registry analysis as placebo at Month 6 minus irbesartan at Month 6. Thus, the point estimate places the adjusted mean score in the placebo group 0.2 units above the corresponding irbesartan mean under this analysis.
The estimate does not mean that an individual participant's score would differ by exactly 0.2 units, nor does it establish that the two treatments produce a clinically meaningful difference. The analysis is a comparison of group-level mean scores.
The two-sided 95% confidence interval of -0.8 to 1.3 describes uncertainty around the estimated mean difference. Because the interval includes 0, the ClinicalTrials.gov record does not establish a statistically distinguishable mean difference from zero at the conventional two-sided 95% confidence-interval level.
A confidence interval is not itself a measure of clinical importance, and it should not be interpreted as the range containing 95% of individual treatment effects.
Month 14 analysis
Adjusted mean difference in MLwHF total score
95% CI: -0.5 to 1.7 · two-sided 95% confidence interval
Comparison: Placebo - Month 14 vs Irbesartan - Month 14
The estimated mean difference at Month 14 was 0.6, again expressed as placebo minus irbesartan for the final values. The point estimate therefore represents a 0.6-unit higher adjusted mean score in the placebo group under the reported model.
The 95% confidence interval of -0.5 to 1.7 includes 0. This indicates that the ClinicalTrials.gov record is compatible with a range of mean differences extending from a modest negative value to a positive value. The interval should therefore be read as a measure of statistical precision rather than as a probability statement about the true effect.
The registry analysis does not supply a p-value in the ClinicalTrials.gov record. A p-value would address compatibility with a specified null hypothesis; it would not measure the magnitude or clinical importance of the observed mean difference.
8. How the Mixed-Effects Model Was Used
Both reported statistical analyses used a mixed-effects model, described in the registry as "Mixed Models Analysis." This is appropriate for longitudinal measurements because the same participant can contribute repeated scores over time, creating within-participant correlation.
The registry-reported analysis notes state that baseline score was used as a covariate and that treatment, visit, treatment-by-visit, and baseline angiotensin-converting enzyme (ACE) inhibitor use were used as predictors.
This structure is more informative than simply comparing raw Month 6 or Month 14 means. Baseline score is incorporated into the model, which adjusts the comparison for the participant's starting MLwHF score. The visit and treatment-by-visit terms allow the model to represent change over the longitudinal assessment schedule and whether treatment differences vary by visit.
Why baseline adjustment matters
For a continuous outcome measured before and after treatment, baseline adjustment can improve precision by accounting for variation in participants' starting values. It also provides an estimate of the treatment-group difference in follow-up values conditional on the baseline covariate included in the model.
Why the treatment-by-visit term matters
A treatment-by-visit interaction allows the treatment contrast to differ across visits. That is important when a study measures the same outcome repeatedly: the Month 6 treatment difference need not be assumed to be identical to the Month 14 treatment difference.
9. Statistical Methodology
Mixed-effects models
A mixed-effects model combines fixed effects, which represent systematic predictors such as treatment and visit, with random effects that account for participant-level or other sources of correlation when specified by the model.
For longitudinal clinical-trial data, the key statistical issue is that observations from the same participant are generally not independent. A mixed model provides a framework for representing this dependence while estimating treatment-related differences.
The exact covariance and random-effects specification is not provided in the ClinicalTrials.gov record. The expression above is therefore a conceptual representation rather than a reconstruction of the exact fitted model.
Covariate adjustment
The registry analysis notes explicitly identify covariate adjustment. Baseline MLwHF score was used as a covariate, and baseline ACE inhibitor use was included among the predictors. This means the reported mean differences are model-adjusted estimates rather than simple unadjusted differences between the two randomized groups.
Mean difference
The effect measure for both statistical analyses is the mean difference in final values. The direction is important: the registry-reported group comparison is written as placebo minus irbesartan.
A positive estimate therefore corresponds to a higher adjusted final score in the placebo group under the registry's reported comparison order.
Confidence intervals
Both reported estimates have two-sided 95% confidence intervals. A confidence interval communicates statistical precision around the estimated treatment contrast. Narrower intervals generally indicate greater precision, while wider intervals indicate greater uncertainty.
Superiority hypothesis
The statistical analyses posted on ClinicalTrials.gov identify the hypothesis type as superiority. This is distinct from a non-inferiority framework: the objective is not to demonstrate that irbesartan remains within a prespecified acceptable loss of effect, but to evaluate a treatment difference under a superiority hypothesis.
10. Secondary Endpoint Analysis Framework
| Component | Reported specification |
|---|---|
| Outcome | MLwHF Total Score, sum of Questions 1-21 |
| Time frame | Baseline, Month 6, Month 14 |
| Outcome type | Continuous |
| Analysis population | Participants with baseline score and score at timepoint |
| Method | Mixed Models Analysis |
| Effect measure | Mean difference |
| Hypothesis | Superiority |
| Covariate adjustment | Baseline score, treatment, visit, treatment-by-visit, baseline ACE inhibitor use |
The analysis population is noteworthy. The registry text defines it as participants who have both a baseline score and a score at the relevant timepoint. This is narrower than a statement that every randomized participant necessarily contributed to each longitudinal comparison.
11. Statistical Methods Explained
Why was a mixed-effects model used?
The MLwHF score was measured longitudinally at baseline, Month 6, and Month 14. Repeated observations from the same participant are correlated, so treating every observation as independent would generally be inappropriate. A mixed-effects framework allows the analysis to model longitudinal structure while estimating treatment differences.
Why was the baseline score included as a covariate?
Participants can begin the trial with different MLwHF scores. Including baseline score as a covariate accounts for this starting-point variation when estimating follow-up treatment differences. This can improve statistical precision compared with an unadjusted comparison of final values.
What does a mean difference of 0.2 mean?
For the Month 6 analysis, the reported contrast is placebo minus irbesartan. A value of 0.2 therefore means that the model-estimated final MLwHF score was 0.2 units higher in the placebo group than the irbesartan group under the reported model. It is a group-level adjusted difference, not an individual-level treatment effect.
Why does the confidence interval cross zero?
For Month 6, the 95% confidence interval is -0.8 to 1.3. For Month 14, it is -0.5 to 1.7. Because zero lies inside both intervals, the registry-reported confidence intervals do not exclude no mean difference at the two-sided 95% level. The interval also shows that the direction and magnitude of the underlying mean difference are uncertain.
Why is treatment-by-visit included?
A treatment-by-visit term allows the treatment contrast to change over time. Without such a term, the model would impose a more restrictive relationship between treatment and visit. The registry-reported analysis notes explicitly identify treatment-by-visit as one of the predictors.
Why should a p-value not be reconstructed?
The ClinicalTrials.gov record reports the estimates and 95% confidence intervals but do not provide p-values. A p-value is a separate inferential quantity tied to a null hypothesis and testing procedure. Reconstructing one from rounded summary information would introduce assumptions not contained in the ClinicalTrials.gov record.
12. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants divided by participants at risk.
| Safety measure | Irbesartan | Placebo |
|---|---|---|
| Serious adverse events | 1146 / 2064 | 1155 / 2062 |
The ClinicalTrials.gov record does not provide a formal statistical comparison, confidence interval, or p-value for serious adverse events. The appropriate interpretation is therefore descriptive: serious adverse events were reported in both randomized arms, with the affected/at-risk counts shown above.
13. Missing Data and Analysis Population
The registry-reported MLwHF analyses define the analysis population as participants with a baseline score and a score at the relevant timepoint. This is an important qualification because longitudinal clinical-trial outcomes can be missing for reasons related to discontinuation, illness, follow-up, or other trial processes.
The ClinicalTrials.gov record does not specify a formal imputation method for missing MLwHF values. They also do not provide sufficient information to determine whether a particular missing-data mechanism was assumed or whether sensitivity analyses were conducted.
14. Multiplicity and Multiple Outcome Measures
The trial record reports 19 outcome measures and 2 statistical analyses in the ClinicalTrials.gov recordset. The two statistical-analysis records both concern the same MLwHF Total Score at different follow-up timepoints.
| Feature | What the ClinicalTrials.gov record establishes | Statistical implication |
|---|---|---|
| Primary endpoint | One registered time-to-event composite endpoint | Formal primary-endpoint analysis is not included in the ClinicalTrials.gov record |
| Outcome measures | 19 posted | Multiple outcomes create a broader analytical context |
| Formal statistical analyses | 2 posted | Both concern the MLwHF score |
| MLwHF visits | Month 6 and Month 14 comparisons | Multiple longitudinal contrasts should be interpreted with awareness of the analysis plan and any multiplicity controls |
The ClinicalTrials.gov record does not identify a multiplicity-adjustment strategy for the two MLwHF comparisons. Therefore, the two confidence intervals are presented as reported, without assigning them a familywise interpretation that is not documented in the ClinicalTrials.gov record.
15. Censoring and Time-to-Event Interpretation
The primary endpoint is a time-to-event outcome, which introduces a statistical issue that does not arise in the same way for a single continuous measurement: censoring. If a participant has not experienced the composite event by the end of their available observation period, the exact future event time is unknown.
Survival-analysis methods use the information available up to the censoring time rather than treating an unobserved future event as if it had never occurred. This is why time-to-event analyses are preferable to simply comparing the proportion of participants with events without regard to follow-up duration.
A hazard is not a probability of death or hospitalization over the entire follow-up period. It is an instantaneous rate conditional on remaining at risk.
16. Hazard Ratios and the Proportional-Hazards Assumption
If a Cox proportional-hazards model were used for the primary time-to-event comparison, its treatment effect would commonly be summarized with a hazard ratio. The ClinicalTrials.gov record does not provide a primary-endpoint hazard ratio, so none is reported here.
A hazard ratio is a relative instantaneous-rate measure. It should not be interpreted as a relative risk, a risk difference, a probability of benefiting, or a percentage reduction in the number of events.
The proportional-hazards assumption is also relevant. A single Cox hazard ratio is most straightforward to interpret when the relative hazard is reasonably stable over time. The ClinicalTrials.gov record does not report an assessment of proportional hazards, so this page does not claim that the assumption was satisfied.
The fact that the registered endpoint is time-to-event does not by itself tell us which survival model was used. The ClinicalTrials.gov record identifies the endpoint type but does not provide the primary statistical-analysis record. The methods discussed here are therefore educational explanations of the standard analysis framework rather than reconstructed trial results.
17. Confidence Intervals in the Reported MLwHF Analyses
| Visit | Mean difference | 95% CI | Interpretation of precision |
|---|---|---|---|
| Month 6 | 0.2 | -0.8 to 1.3 | The interval spans negative and positive mean differences. |
| Month 14 | 0.6 | -0.5 to 1.7 | The interval also spans negative and positive mean differences. |
Both confidence intervals include zero. This is the central inferential feature of the registry-reported MLwHF results. The point estimates are positive in both comparisons, but the intervals show uncertainty in the estimated direction and magnitude of the underlying mean difference.
The confidence intervals also illustrate why a point estimate should not be interpreted in isolation. An estimate of 0.2 or 0.6 may appear precise as a single number, but the corresponding intervals demonstrate that substantially different underlying values remain compatible with the analysis.
18. What the Reported Mean Differences Do — and Do Not — Mean
The reported mean difference of 0.2 is the adjusted final-value contrast for placebo minus irbesartan. It does not mean that every placebo-treated participant had a score exactly 0.2 units higher than an irbesartan-treated participant.
The reported mean difference of 0.6 is the adjusted final-value contrast for placebo minus irbesartan. It does not describe an individual participant's change, nor does it establish the clinical importance of the difference without additional information about the score's clinical interpretation.
The 95% confidence intervals of -0.8 to 1.3 at Month 6 and -0.5 to 1.7 at Month 14 describe uncertainty around the estimated mean differences. They are not ranges for individual patient outcomes.
19. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The two reported MLwHF analyses estimate small positive placebo-minus-irbesartan mean differences, with two-sided 95% confidence intervals that include zero. Both analyses use mixed models with baseline score and other specified predictors.
Clinical interpretation
The ClinicalTrials.gov record does not provide enough information to determine the clinical importance of either MLwHF difference. The registry record should therefore be read alongside the scale definition, the analysis population, and the uncertainty intervals rather than relying on the point estimates alone.
The primary endpoint is a separate question. It concerns time to first occurrence of a composite of all-cause death or protocol-specified cardiovascular hospitalization, whereas the MLwHF analyses concern a continuous patient-reported outcome measured at specified visits. These endpoints answer different clinical questions and should not be substituted for one another.
20. Important Limitations and Interpretation Issues
- Primary statistical result not reported: the trial data identify the registered primary time-to-event endpoint and state that results are posted, but they do not provide a formal statistical-analysis record for that endpoint. No primary hazard ratio, confidence interval, p-value, or event estimate is therefore reported here.
- Secondary analyses only: the two formal statistical analyses registry-reported concern the MLwHF Total Score rather than the registered primary endpoint.
- Analysis population: the MLwHF analyses include participants with a baseline score and a score at the relevant timepoint. This should not automatically be equated with an all-randomized analysis population.
- Missing-data method not reported: the ClinicalTrials.gov record do not specify an imputation strategy or a formal sensitivity analysis for missing MLwHF values.
- Multiplicity: 19 outcome measures are reported in the ClinicalTrials.gov record, but no multiplicity-adjustment strategy is posted on ClinicalTrials.gov for the two reported MLwHF analyses.
- Model specification: the registry identifies a mixed-effects model and lists its predictors, but the complete covariance and random-effects structure is not provided in the ClinicalTrials.gov record.
- Clinical meaning of the score difference: the ClinicalTrials.gov record provides the statistical mean differences and confidence intervals but do not provide a threshold for clinical importance of the MLwHF score.
- Time-to-event methodology: because the ClinicalTrials.gov record does not provide the primary statistical-analysis record, standard survival methods described on this page should not be mistaken for a verified reconstruction of the exact primary analysis.
- Safety analysis: serious adverse-event counts are reported descriptively, without a formal statistical comparison.
21. Why This Trial Matters Statistically
I-PRESERVE provides a useful teaching case because it places two different statistical structures within the same randomized clinical-trial framework: a time-to-event composite endpoint and a longitudinal continuous patient-reported outcome.
| Concept | How it appears in I-PRESERVE |
|---|---|
| Randomization | Participants were randomized to irbesartan or placebo. |
| Blinding | The trial was double masked. |
| Parallel design | The design model was parallel. |
| Time-to-event analysis | The registered primary endpoint is a time-to-event composite. |
| Composite endpoint | All-cause death or protocol-specified cardiovascular hospitalization. |
| Longitudinal analysis | MLwHF was assessed at baseline, Month 6, and Month 14. |
| Mixed-effects model | The two registry-reported formal analyses used Mixed Models Analysis. |
| Covariate adjustment | Baseline score and baseline ACE inhibitor use were among the specified predictors. |
| Treatment-by-visit | The reported model included treatment-by-visit. |
| Mean difference | MLwHF results were expressed as placebo minus irbesartan mean differences. |
| Confidence intervals | Both reported MLwHF estimates include two-sided 95% confidence intervals. |
| Safety denominators | Serious adverse events are reported as affected participants divided by participants at risk. |
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
Use the statistical calculator collection to explore related trial-analysis concepts:
24. Sources
- ClinicalTrials.gov: NCT00095238 — I-PRESERVE.
- Linked publication: PubMed record: PMID 19001508.
- Linked publication: PubMed record: PMID 20231531.
- Linked publication: PubMed record: PMID 18279770.
- Linked publication: PubMed record: PMID 16230259.
- Linked publication: PubMed record: PMID 20142024.
Continue exploring clinical-trial statistics
Connect the design, endpoints, and analytical methods in I-PRESERVE with deeper Clinical Biostats tutorials and statistical calculators.
25. Record Summary
I-PRESERVE is a randomized, double-blind, parallel phase 3 trial enrolling 4128 participants with congestive heart failure and comparing irbesartan with placebo. Its registered primary endpoint is a time-to-event composite of all-cause death or protocol-specified cardiovascular hospitalization, evaluated at Year 1 through Year 5.
The formal statistical analyses in the ClinicalTrials.gov record instead concern the Minnesota Living With Heart Failure Total Score at Month 6 and Month 14. Both use mixed-effects models with baseline score as a covariate and include treatment, visit, treatment-by-visit, and baseline ACE inhibitor use as predictors. The reported mean differences are 0.2 at Month 6 with a 95% CI of -0.8 to 1.3, and 0.6 at Month 14 with a 95% CI of -0.5 to 1.7.
The statistical lesson is the importance of matching the analysis to the endpoint. The registered primary outcome is a time-to-event composite and would ordinarily require survival-analysis methods that account for censoring and event timing. The reported MLwHF outcome is continuous and longitudinal, making a mixed-effects framework appropriate for repeated measurements and covariate adjustment. These are different estimands and should not be interpreted interchangeably.