This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Trial-specific numerical results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
DAPA-MI was a randomized, parallel-group, quadruple-masked, phase 3 trial evaluating dapagliflozin 10 mg versus placebo in participants with acute myocardial infarction and heart failure. The primary endpoint was analyzed using a hierarchical composite framework and the win-ratio method.
| Feature | DAPA-MI |
|---|---|
| Trial name | DAPA-MI |
| Phase | Phase 3 |
| Conditions | Acute Myocardial Infarction; Heart Failure |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 4017 |
| Interventions | Dapagliflozin (drug); Placebo (drug) |
| Primary endpoint type | Count / rate |
| Primary hypothesis | Superiority |
| Primary statistical method | Win ratio |
| Lead sponsor | AstraZeneca |
| Status | Completed |
| Start | 2020-12-22 |
| Primary completion | 2023-07-05 |
| ClinicalTrials.gov | NCT04564742 |
2. Clinical Question
The statistical question was whether treatment with dapagliflozin 10 mg, given once daily in addition to standard of care, produced a superior result compared with matching placebo when participants were evaluated using the registered hierarchical primary composite endpoint.
Population
Participants in the DAPA-MI trial with the registered conditions of acute myocardial infarction and heart failure.
Intervention
Dapagliflozin 10 mg, given once daily in addition to standard of care.
Comparator
Matching placebo, given once daily in addition to standard of care.
Primary question
Does dapagliflozin provide a superior result to placebo on the hierarchical primary composite endpoint?
3. Trial Design
Dapagliflozin
- Dapagliflozin 10 mg
- Given once daily
- In addition to standard of care
Placebo
- Matching placebo
- Given once daily
- In addition to standard of care
The registry reports a mean study duration of 12.0 months, measured as time in study until the last visit, with an accumulated 4023.9 participant-years. The maximum study duration for any participant was 29 months.
4. Endpoints
| Endpoint | Registered time frame | Type / analysis |
|---|---|---|
| Analysis of the Hierarchical Primary Composite Endpoint (Full Analysis Set) | 29 months | Binary endpoint in the posted statistical analysis; analyzed using a win-ratio method |
The registry describes the primary endpoint as a hierarchical primary composite endpoint. The registry definition states that participants received dapagliflozin 10 mg or matching placebo once daily in addition to standard of care. It also states that the number of events for NYHA corresponds to the number of participants with a non-missing value for NYHA.
5. Statistical Methodology
Full Analysis Set
The posted primary analysis is explicitly labeled Analysis of the Hierarchical Primary Composite Endpoint (Full Analysis Set). The statistical-analysis record states that all participants are included in the analysis and that each row item displays the number of participants for the given event without preceding events defined in the hierarchical order.
Intention-to-treat principle
The statistical-analysis record identifies intention-to-treat analysis as a concept used in the primary analysis. In an intention-to-treat framework, randomized participants remain associated with their assigned treatment group for the primary comparison. This preserves the treatment contrast created by randomization rather than redefining groups according to treatment actually received.
Win-ratio analysis
The registry reports Win Ratio Analysis as the method used for the primary endpoint. The win ratio is particularly suited to hierarchical composite endpoints because it can preserve the ordering of clinical outcomes rather than treating all component events as interchangeable.
This is a conceptual representation rather than a reconstruction of the DAPA-MI calculation. A win is determined using the trial's prespecified hierarchy and comparison rules. The ClinicalTrials.gov record does not provide the complete pairwise event table needed to reproduce the reported value.
Superiority testing
The registered hypothesis type was superiority. The objective was therefore to determine whether the clinical benefit of dapagliflozin was superior to placebo using the hierarchical composite endpoint and win-ratio method.
Multiplicity adjustment
The statistical-analysis record identifies multiplicity adjustment as an additional concept in the analysis text. This is important because hierarchical composite endpoints and clinical-trial analyses can involve multiple ordered comparisons. The ClinicalTrials.gov record does not specify the complete multiplicity procedure, so no additional adjustment scheme is inferred here.
6. Primary Result
Hierarchical Primary Composite Endpoint
Win ratio
95% CI: 1.20–1.50 · P < 0.001
Two-sided confidence interval; superiority hypothesis.
The primary analysis compared dapagliflozin 10 mg versus placebo in the full analysis set. The reported effect measure was the Win Ratio (WR), with an estimate of 1.34 and a two-sided 95% confidence interval of 1.20 to 1.50. The reported P-value was <0.001.
| Primary endpoint | Dapagliflozin 10 mg vs placebo |
|---|---|
| Endpoint | Analysis of the Hierarchical Primary Composite Endpoint (Full Analysis Set) |
| Time frame | 29 months |
| Analysis | Win Ratio Analysis |
| Effect measure | Win Ratio (WR) |
| Estimate | 1.34 |
| 95% CI | 1.20–1.50 |
| P-value | <0.001 |
| Hypothesis | Superiority |
| Analysis population | Full Analysis Set |
A win ratio of 1.34 means that, under the trial's hierarchical comparison framework, the treatment group had an estimated 34% as many wins above the comparator's wins when expressed as the ratio of treatment-group wins to control-group wins. Put differently, the estimated number of favorable pairwise outcomes for dapagliflozin was 1.34 times the corresponding number for placebo under the prespecified hierarchy.
The win ratio does not mean that 34% of participants improved, that 34% more participants benefited, or that an individual participant had a 34% greater probability of a favorable outcome. It is a population-level comparative statistic whose meaning depends on the hierarchy and pairwise rules used by the trial.
The 95% confidence interval of 1.20–1.50 describes statistical uncertainty around the estimated win ratio under the analysis framework. It does not describe the range of treatment effects experienced by individual participants.
The P-value <0.001 addresses the statistical evidence against the relevant null hypothesis under the prespecified analysis. It does not measure the size or clinical importance of the treatment effect. Effect size and uncertainty are better conveyed by the win ratio together with its confidence interval.
The interpretation also depends on the hierarchical structure of the composite endpoint. Because the ClinicalTrials.gov record does not provide the complete hierarchy or pairwise event counts, the numerical win ratio cannot be decomposed here into the contribution made by each individual endpoint component.
7. Understanding the Win Ratio
The win ratio differs fundamentally from a conventional risk ratio or hazard ratio. It is designed for situations in which several outcomes can be ordered by clinical importance. Rather than reducing all outcomes to a single undifferentiated event indicator, the analysis compares outcomes according to their hierarchy.
What a win means
A favorable pairwise comparison is a "win" for one treatment group according to the endpoint hierarchy and prespecified comparison rules.
What a loss means
The opposite treatment group has the more favorable outcome according to those same rules.
Why hierarchy matters
The clinical importance assigned to different outcomes determines which outcome takes precedence when comparing participants.
Why the ratio is useful
A ratio above 1 indicates more favorable pairwise outcomes for the treatment group under the defined comparison framework.
A value above 1 favors the treatment group in the win-ratio framework, while a value below 1 favors the comparator. The DAPA-MI estimate was 1.34.
Importantly, the win ratio should not be translated into a conventional percentage reduction in risk. A win ratio and a hazard ratio are different effect measures, and the numerical relationship between them is not a simple conversion.
8. Why the Confidence Interval Matters
The primary estimate of 1.34 is only one part of the statistical result. The associated 95% confidence interval, 1.20–1.50, provides information about the precision of the estimate.
The interval lies above 1, consistent with the reported superiority result. Its width reflects uncertainty in estimating the underlying treatment comparison from the available trial data. The interval does not imply that the true effect is equally likely to be any particular value within its limits, nor does it describe individual patient outcomes.
The confidence interval and P-value answer related but different questions. The confidence interval emphasizes the estimated effect and its precision; the P-value quantifies evidence against the null hypothesis under the specified statistical model and testing framework. Neither statistic alone describes clinical importance.
9. Trial Timing and Follow-Up
Trial start
The DAPA-MI trial began on December 22, 2020.
Registered endpoint time frame
The hierarchical primary composite endpoint was registered with a 29-month time frame, which was also the maximum study duration for any participant.
Primary completion
The registry lists July 5, 2023 as the primary completion date.
The registry reports a mean study duration of 12.0 months, measured as time in study until the last visit, and 4023.9 accumulated participant-years. The distinction between mean study duration and the maximum follow-up is important: not every participant necessarily contributed the maximum duration of observation.
10. Safety
The ClinicalTrials.gov record reports serious adverse events by treatment arm. These data are presented separately from the efficacy analysis because safety and efficacy address different statistical questions.
| Safety measure | Dapagliflozin 10 mg | Placebo |
|---|---|---|
| Serious adverse events, affected / at risk | 449 / 2019 | 404 / 1998 |
The registry provides the affected and at-risk counts but does not provide a formal statistical comparison for serious adverse events in the ClinicalTrials.gov record. The counts should therefore not be converted into an unreported treatment-effect estimate or P-value.
11. Statistical Methods Explained
Why was a win ratio used?
The registered primary endpoint is hierarchical and composite. A win-ratio analysis can preserve the ordering of outcomes by comparing participants pairwise according to that hierarchy. This can be useful when some clinical outcomes are considered more important than others and should therefore take precedence in the analysis.
What does a win ratio of 1.34 mean?
A win ratio of 1.34 indicates that, under the prespecified hierarchical comparison framework, treatment-group wins were estimated to occur 1.34 times as often as control-group wins. It is not a 34% response rate, a 34% reduction in risk, or a 34% increase in survival.
Why does the confidence interval matter?
The 95% confidence interval of 1.20–1.50 quantifies uncertainty around the estimated win ratio. The point estimate alone can conceal how precisely the treatment effect has been estimated. The interval also remains tied to the statistical framework used to generate it.
Why does the P-value not measure effect size?
The P-value of <0.001 indicates strong statistical evidence against the relevant null hypothesis under the prespecified analysis. It does not say whether the effect is small, moderate, or large in clinical terms. Effect size is communicated by the win ratio itself, while precision is communicated by the confidence interval.
Why is the analysis based on the Full Analysis Set?
The registry explicitly labels the primary endpoint analysis as being performed in the Full Analysis Set and states that all participants are included in the analysis. This anchors the primary comparison to the trial population defined for that analysis rather than restricting the analysis to participants who completed treatment.
Why does the hierarchy matter when interpreting the result?
A hierarchical composite does not treat every component as equivalent. The ordering determines which outcome takes precedence in pairwise comparisons. Consequently, the same numerical win ratio cannot be interpreted without reference to the endpoint hierarchy. Because the complete hierarchy is not included in the ClinicalTrials.gov record, this page does not infer component-specific contributions.
Does a win ratio of 1.34 mean that 34% more patients benefited?
No. A win ratio is constructed from favorable and unfavorable pairwise comparisons. It is therefore not equivalent to a proportion of patients benefiting, an absolute risk difference, or a relative risk. Translating it into one of those measures without the underlying data would be statistically incorrect.
12. Intention-to-Treat Analysis
The registry's statistical-analysis record specifically identifies intention-to-treat analysis among the concepts used in the primary analysis. The principle is important because randomization establishes the basis for a causal comparison before treatment adherence, discontinuation, or other post-randomization events occur.
Preserves randomization
Participants remain associated with their randomized treatment assignment for the primary efficacy comparison.
Reduces post-randomization selection
Restricting analysis to participants who behave exactly as planned can introduce selection into a randomized comparison.
In this trial, the statistical-analysis record states that all participants were included in the analysis. That is consistent with the purpose of an intention-to-treat approach: maintaining the treatment contrast generated by randomization.
13. Multiplicity and Hierarchical Testing
The ClinicalTrials.gov record identifies multiplicity adjustment as a concept in the primary analysis text. Multiplicity is a central issue whenever a study involves multiple statistical opportunities to obtain evidence against a null hypothesis.
| Concept | Role in DAPA-MI |
|---|---|
| Primary endpoint | Hierarchical primary composite endpoint |
| Primary hypothesis | Superiority |
| Primary effect measure | Win Ratio (WR) |
| Multiplicity | Identified in the posted analysis text |
| Analysis population | Full Analysis Set |
A hierarchical endpoint can itself provide a structured way to prioritize outcomes, but that does not eliminate the need to understand the trial's formal error-control strategy. The ClinicalTrials.gov record identifies multiplicity adjustment as part of the analysis but do not provide enough detail to reconstruct the full procedure. Accordingly, no additional alpha-spending or multiplicity algorithm is attributed to DAPA-MI here.
14. What the Primary Result Does — and Does Not — Establish
The posted primary analysis reports a win ratio of 1.34 with a two-sided 95% confidence interval of 1.20–1.50 and P < 0.001 for dapagliflozin 10 mg versus placebo on the hierarchical primary composite endpoint. The registered hypothesis type was superiority.
The result does not provide a percentage of participants who benefited, an individual patient's probability of benefit, an absolute risk reduction, a number needed to treat, or a component-specific effect estimate. Those quantities require different data and/or effect measures.
The interpretation of a win ratio is inseparable from the hierarchy used to define wins and losses. Without the complete hierarchy and pairwise event data, the reported 1.34 should be interpreted as the overall registered win-ratio result rather than decomposed into individual components.
15. Design Features That Matter Statistically
| Design feature | DAPA-MI | Statistical relevance |
|---|---|---|
| Randomization | Randomized | Creates the treatment comparison used for the primary efficacy analysis. |
| Parallel design | Parallel | Participants are compared between two concurrently randomized treatment groups. |
| Masking | Quadruple | Masking can reduce the potential influence of treatment knowledge on trial conduct and assessment. |
| Primary endpoint | Hierarchical composite | Requires an analysis that respects the ordering of component outcomes. |
| Primary method | Win ratio | Provides a ratio of treatment-group wins to comparator-group wins under the hierarchy. |
| Hypothesis | Superiority | The analysis tests whether the treatment produces a superior result rather than establishing non-inferiority. |
| Multiplicity | Adjustment identified in analysis text | Indicates that multiplicity was considered in the statistical framework. |
| ITT concept | Identified in analysis text | Supports analysis according to randomized treatment assignment. |
16. Planned vs Reported Statistical Information
The ClinicalTrials.gov record for DAPA-MI contains a formal primary statistical analysis with an effect estimate, confidence interval, and P-value, so the results can be interpreted directly rather than only described as planned.
Reported
The registry reports the primary endpoint, its 29-month time frame, the Full Analysis Set, win-ratio method, estimate of 1.34, 95% CI of 1.20–1.50, and P < 0.001.
Not reconstructed
The ClinicalTrials.gov record does not provide enough information to reconstruct component-level wins, losses, the complete hierarchy, or an independent calculation of the reported win ratio.
This distinction is important in statistical reporting. A published or registry-reported summary statistic should not be reverse-engineered into additional quantities unless the underlying information supports the calculation.
17. Limitations
- Incomplete component detail: The ClinicalTrials.gov record identifies the endpoint as hierarchical but do not provide the complete hierarchy needed to explain the contribution of every component to the win ratio.
- No component-specific effect estimates: The posted statistical analysis provides the overall win ratio, but the ClinicalTrials.gov record does not include separate estimates for individual components of the hierarchy.
- Win-ratio interpretation: The win ratio cannot be interpreted as a conventional risk ratio, hazard ratio, relative risk reduction, or proportion of participants benefiting.
- Multiplicity detail: Multiplicity adjustment is identified in the analysis text, but the ClinicalTrials.gov record does not specify the complete adjustment procedure.
- Safety comparison: Serious adverse-event counts are reported by arm, but the registry-reported statistical analysis does not provide a formal comparison or confidence interval for those safety counts.
- Analysis reconstruction: The reported win ratio cannot be independently recalculated from the registry-reported summary because the underlying pairwise comparisons and complete event hierarchy are not provided.
- Time-frame interpretation: The registered endpoint time frame is 29 months, while the reported mean study duration was 12.0 months. These are different summaries and should not be treated as interchangeable.
- Generalizability: The ClinicalTrials.gov record describes the trial population only at the level of its listed conditions and design. More detailed eligibility and baseline characteristics are not included in the ClinicalTrials.gov record.
18. Why This Trial Matters Statistically
DAPA-MI is a useful teaching example because it illustrates a different approach to composite clinical-trial endpoints from the more familiar time-to-event framework. The primary analysis uses a hierarchical composite endpoint and a win ratio, while retaining the randomized comparison and intention-to-treat principles.
| Concept | How it appears in DAPA-MI |
|---|---|
| Randomization | Randomized parallel-group trial |
| Masking | Quadruple masking |
| Intention-to-treat | Identified as a concept in the primary analysis |
| Composite endpoint | Hierarchical primary composite endpoint |
| Win ratio | Primary statistical method |
| Superiority | Registered hypothesis type |
| Confidence interval | 95% two-sided CI of 1.20–1.50 around the primary estimate |
| P-value | <0.001 for the primary analysis |
| Multiplicity | Identified in the posted analysis text |
| Safety analysis | Serious adverse-event counts reported by randomized arm |
The most important statistical lesson is that the choice of endpoint determines the interpretation of the effect measure. A win ratio should be taught as a win-ratio result, not translated casually into a hazard ratio or risk reduction. Understanding the endpoint construction is therefore just as important as understanding the numerical estimate.
19. A Practical Reading Framework for the DAPA-MI Result
Step 1 · Identify the estimand
The primary question concerns dapagliflozin 10 mg versus placebo using the hierarchical primary composite endpoint.
Step 2 · Identify the analysis
The registry reports a Full Analysis Set and a Win Ratio Analysis.
Step 3 · Read the effect
The reported win ratio is 1.34, indicating more treatment-group wins than control-group wins under the hierarchy.
Step 4 · Read uncertainty
The two-sided 95% CI is 1.20–1.50, describing uncertainty around the estimated ratio.
This four-step approach prevents a common statistical mistake: interpreting the P-value first and the estimand second. The correct order is to understand what was compared, how it was compared, what the effect measure means, and then how precisely it was estimated.
20. Related Tutorials
Learn more about the methods used in this trial:
21. Related Statistical Calculators
22. Sources
- ClinicalTrials.gov: DAPA-MI, NCT04564742.
- PubMed: PMID 38320489.
- PubMed: PMID 37648579.
Continue through the Clinical Biostats statistical pathway
Use the trial's statistical concepts as a starting point for deeper study of randomization, confidence intervals, intention-to-treat analysis, multiplicity, P-values, and win-ratio methodology.
23. Record Summary
DAPA-MI was a randomized, parallel, quadruple-masked phase 3 trial with 4017 enrolled participants and two treatment arms: dapagliflozin 10 mg and placebo. Its primary endpoint was a hierarchical composite analyzed in the Full Analysis Set using a win-ratio method. The registry reports a primary win ratio of 1.34, with a two-sided 95% confidence interval of 1.20–1.50 and P < 0.001 under a superiority hypothesis.
The central statistical lesson is that the win ratio must be interpreted in the context of its hierarchical endpoint. A value of 1.34 represents the relative frequency of favorable pairwise outcomes under the prespecified comparison framework; it is not interchangeable with a hazard ratio, risk ratio, absolute risk reduction, or percentage of patients benefiting. The confidence interval provides the corresponding measure of statistical precision, while the P-value addresses evidence against the null hypothesis rather than effect magnitude.
The registry also identifies intention-to-treat analysis and multiplicity adjustment among the concepts incorporated into the primary statistical analysis. Serious adverse events were reported as 449/2019 for dapagliflozin 10 mg and 404/1998 for placebo. Because the ClinicalTrials.gov record does not provide the complete endpoint hierarchy, pairwise event data, or a formal safety comparison, this page deliberately does not derive additional component-specific estimates.