This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial facts on this page are restricted to the ClinicalTrials.gov record. The registry identifies both primary endpoints as having posted results, but the ClinicalTrials.gov record contains no formal statistical-analysis estimates, confidence intervals, or p-values for those endpoints.
1. Trial at a Glance
DARE-19 was a randomized, parallel-group, quadruple-masked phase 3 trial evaluating dapagliflozin 10 mg versus placebo in patients with COVID-19 and respiratory failure. The trial enrolled 1250 participants and specified two primary endpoints assessed from randomization through Day 30.
| Feature | DARE-19 |
|---|---|
| Trial name | DARE-19 |
| ClinicalTrials.gov identifier | NCT04350593 |
| Phase | Phase 3 |
| Condition | COVID-19 |
| Brief title | Dapagliflozin in Respiratory Failure in Patients With COVID-19 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 1250 |
| Interventions | Dapagliflozin 10 milligram (mg); placebo |
| Trial status | Completed |
| Start | 2020-04-22 |
| Primary completion | 2021-03-31 |
| Lead sponsor | Saint Luke's Health System |
2. Clinical Question
The central question was whether treatment with dapagliflozin 10 mg, compared with placebo, affected clinical outcomes during the 30-day treatment period in patients with COVID-19 and respiratory failure.
Population
Patients with COVID-19 and respiratory failure, as represented by the trial's registered condition and brief title.
Intervention
Dapagliflozin 10 milligram (mg).
Comparator
Placebo.
Primary question
How do the randomized groups compare with respect to the two registered primary endpoints during the 30-day period following randomization?
3. Trial Design
Random allocation is intended to balance prognostic factors across treatment groups in expectation and supports a causal comparison of assigned treatment.
Participants are assigned to one of two treatment strategies rather than being intentionally exposed to both treatments as part of a crossover design.
The registry identifies the study as quadruple-masked. The ClinicalTrials.gov record does not specify which four groups of individuals were masked.
The registered primary purpose is treatment rather than prevention, diagnostic evaluation, or another study purpose.
Dapagliflozin
- Dapagliflozin 10 milligram (mg)
- Randomized treatment arm
- Primary endpoints assessed through Day 30
Placebo
- Placebo
- Randomized comparator arm
- Primary endpoints assessed through Day 30
4. Trial Timeline
Trial start
The registered study start date was April 22, 2020.
Primary completion
The registered primary completion date was March 31, 2021.
Registry status
The ClinicalTrials.gov record identifies the trial as completed.
5. Endpoints
DARE-19 has two registered primary endpoints. Both use a time frame beginning at randomization and extending through Day 30, but they address different statistical questions.
| Primary endpoint | Registered time frame | Endpoint type |
|---|---|---|
| Prevention of COVID-19 Complications or Death: During the 30-day Treatment Period, Time to First Occurrence of New/Worsened Organ Dysfunction During Index Hospitalization or Death From Any Cause. | Randomization through Day 30 | Time-to-event |
| Improving Clinical Recovery: Hierarchical Composite Outcome Measure Including Death From Any Cause Through Day 30, New/Worsened Organ Dysfunction, Clinical Status at Day 30 and Hospital Discharge Before Day 30 and Alive at Day 30. | Randomization through Day 30 | Other / unclear |
Primary Endpoint 1: Time to first organ dysfunction or death
The first registered endpoint is explicitly a time-to-event outcome. It measures time from randomization to the first occurrence of new or worsened organ dysfunction during index hospitalization or death from any cause.
The registry definition specifies that event rates are presented as the number of subjects with an event per 100 patient-months (30 days) of follow-up. The registry-reported definition begins to enumerate the components of new or worsened organ dysfunction, and the ClinicalTrials.gov record lists “Respiratory decompensation requiring initiation of mechanical ventilation (includes invasive or non-invasive ventilation, CPAP, or BiPAP), and/or initiation of extracorporeal membrane oxygenation (ECMO) * New or worsening congestive heart failure * Requirement for vasopressor therapy and/or inotropic or mechanical circulatory support * Ventricular tachycardia or fibrillation lasting at least 30 seconds and/or associated with hemodynamic instability or pulseless electrical activity, or resuscitated cardiac arrest * Doubling of s-Creatinine or initiation of renal replacement therapy”. This page therefore does not extend the definition beyond the registry-reported text.
Primary Endpoint 2: Hierarchical clinical recovery composite
The second endpoint is a hierarchical composite outcome measure. The registry definition identifies death from any cause through Day 30, new/worsened organ dysfunction, clinical status at Day 30, and hospital discharge before Day 30 with survival at Day 30 as components of the hierarchy.
The registry describes improvement by Day 30 in relation to baseline as being discharged from hospital without a worsening event and alive, or still in hospital without a worsening event and without oxygen support.
6. Results Availability and Planned Analysis
Primary Endpoint 1: Time to first new/worsened organ dysfunction or death
For a time-to-event endpoint of this form, a conventional analysis would typically estimate the time-to-event distribution using the Kaplan-Meier method and compare randomized groups using a suitable time-to-event test. A Cox proportional-hazards model may be used to estimate a hazard ratio and its confidence interval when that model is prespecified and appropriate.
A hazard ratio below 1 would conventionally indicate a lower estimated instantaneous event rate in the dapagliflozin group relative to placebo over the analysed period. That interpretation cannot be assigned to DARE-19 without a reported treatment-effect estimate.
The ClinicalTrials.gov record does not report the formal model, test statistic, confidence interval, p-value, censoring rules, proportional-hazards assessment, or analysis population for this endpoint.
Primary Endpoint 2: Hierarchical clinical recovery
A hierarchical composite endpoint can be analysed using a method that respects the ordering of outcomes rather than reducing the hierarchy to an ordinary binary endpoint. Depending on the prespecified statistical analysis plan, possible approaches include rank-based or win-based comparisons in which more clinically favourable outcomes are compared against less favourable outcomes according to the hierarchy.
Because the ClinicalTrials.gov record does not identify the formal statistical procedure, no particular method is attributed to DARE-19 here. The registry does not provide a formal comparative estimate in the ClinicalTrials.gov record.
A complete statistical result would ordinarily identify the treatment effect being estimated, the analysis population, the statistical model or test, the uncertainty measure, and the corresponding hypothesis-test result where applicable.
7. Clinical Biostats Interpretation of the Primary Endpoints
The first endpoint asks when the first qualifying event occurs, not merely whether a participant ever experiences one. This distinction matters because two treatment groups can have similar proportions of patients with events while differing in the timing of those events.
A hazard ratio, if reported, would summarize a relative difference in the estimated instantaneous event rate under the chosen survival model. It would not be the same as a risk ratio, an absolute risk difference, or the proportion of patients who avoid an event.
If a hazard ratio or another comparative estimate were available, its confidence interval would describe statistical uncertainty around that estimate under the corresponding model and sampling framework. A narrow interval would generally indicate greater precision than a wide interval, all else equal.
The confidence interval would not describe the range of effects experienced by individual patients.
A p-value, if reported, would quantify how incompatible the observed data and specified test statistic are with the null hypothesis under the statistical testing framework. It would not measure the size of the treatment effect, the probability that the treatment works, or the clinical importance of an observed difference.
The second primary endpoint requires particular care because the hierarchy combines several types of clinical outcomes. Interpretation depends on the prespecified ordering and statistical procedure. A single summary result should not be interpreted as though it were simply the percentage of patients who survived or were discharged.
8. Safety Results
The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm using affected participants divided by participants at risk.
| Safety measure | Dapagliflozin 10mg | Placebo |
|---|---|---|
| Serious adverse events, affected / at risk | 70 / 613 | 87 / 616 |
The ClinicalTrials.gov record does not provide a formal statistical comparison of serious adverse-event rates. Accordingly, the values above are reported descriptively rather than converted into an unreported risk ratio, risk difference, confidence interval, or p-value.
9. Statistical Methodology
Randomization
DARE-19 used randomized allocation. Randomization is the central design mechanism for separating treatment assignment from baseline prognostic factors in expectation. Under appropriate conduct and analysis, this supports a causal comparison between the randomized groups.
Randomization does not guarantee identical baseline characteristics in a particular sample. Its principal statistical value is that, before treatment is assigned, treatment allocation is not deliberately determined by participants' observed or unobserved prognostic characteristics.
Time-to-event analysis
The first primary endpoint is explicitly classified as time-to-event. Time-to-event methods are designed for outcomes in which both whether an event occurs and when it occurs provide information.
Kaplan-Meier estimation is commonly used to describe the event-free probability over time. A log-rank-type comparison can be used to compare survival distributions, while a Cox proportional-hazards model can provide a hazard ratio when its assumptions and prespecification are appropriate.
Here di represents events at an event time and ni represents participants at risk immediately before that time. This formula is presented as a general statistical explanation, not as a reconstruction of the DARE-19 results.
Cox proportional-hazards model
If a Cox model were prespecified for the first endpoint, it could estimate the relative hazard between the two randomized groups while accommodating right-censored observations. The proportional-hazards assumption would need to be considered when interpreting a single hazard ratio across the follow-up period.
This is a general interpretation. The registry-reported DARE-19 data do not provide a DARE-19 hazard ratio.
Hierarchical composite analysis
The second primary endpoint is explicitly described as hierarchical. Hierarchical composites are designed to retain ordering information across clinically different outcomes. The statistical method should therefore reflect the ordering rather than assuming that every component has equivalent clinical meaning.
Because the ClinicalTrials.gov record does not specify the formal statistical test or model, the precise estimand cannot be reconstructed from the endpoint description alone.
Analysis populations
The ClinicalTrials.gov record provides the overall enrollment of 1250 and safety denominators of 613 and 616 for serious adverse events, but do not define the formal efficacy analysis population. It would therefore be inappropriate to label the primary analysis as intention-to-treat, per-protocol, modified intention-to-treat, or another population without supporting registry data.
10. Statistical Methods Explained
Why does randomization matter?
Randomization creates treatment groups through an allocation mechanism rather than through investigator or participant choice. This reduces systematic allocation bias and provides the foundation for interpreting between-group differences as treatment effects, subject to the trial's conduct and analysis.
Why is the first endpoint a time-to-event outcome?
The endpoint is defined as the time to first occurrence of new or worsened organ dysfunction during index hospitalization or death from any cause. Time contains information that would be lost if the analysis simply classified every participant as “event” or “no event” without regard to when the event occurred.
Why might Kaplan-Meier methods be used?
In a time-to-event study, some participants may reach Day 30 without an event. Their exact event time is then unknown beyond their last observed time. Kaplan-Meier estimation incorporates those right-censored observations rather than treating them as if they had experienced the event at Day 30.
What would a hazard ratio mean here?
A hazard ratio would compare the estimated instantaneous event rates between dapagliflozin and placebo under the specified time-to-event model. It would not mean that the same percentage of participants experienced fewer events, and it would not directly provide an absolute treatment benefit.
Why is the second endpoint called hierarchical?
The endpoint combines several clinical states and outcomes and explicitly places them within a hierarchy. A hierarchical analysis can distinguish more favourable outcomes from less favourable outcomes rather than treating every component as interchangeable.
Why can the posted result not be reconstructed from the ClinicalTrials.gov record?
The registry data establish that results are posted, but the registry-reported statistical-analysis field contains no estimates, confidence intervals, or p-values. Without those values or the underlying participant-level or sufficiently detailed event data, a numerical treatment-effect analysis would require assumptions that are not supported by the ClinicalTrials.gov record.
11. Limitations
- No formal statistical estimates reported: the registry indicates that results are posted, but the ClinicalTrials.gov record contains no treatment-effect estimates, confidence intervals, or p-values.
- Analysis population not specified: the ClinicalTrials.gov record does not identify the formal primary efficacy analysis population.
- Statistical model not specified: the ClinicalTrials.gov record does not identify the exact model or test used for either primary endpoint.
- Composite endpoint complexity: the second primary endpoint combines several clinical outcomes, so its interpretation depends on the hierarchy and formal analysis procedure.
- Safety denominators differ from total enrollment: serious adverse events are reported using denominators of 613 and 616 rather than the overall enrollment of 1250.
- No subgroup data reported: the ClinicalTrials.gov record does not contain subgroup estimates, confidence intervals, or interaction tests.
- No multiplicity information reported: the registry data identify two primary endpoints but do not provide the formal multiplicity-control procedure.
- No interim-analysis information reported: the ClinicalTrials.gov record does not describe interim looks, stopping boundaries, or alpha-spending procedures.
- No missing-data or imputation strategy reported: the ClinicalTrials.gov record does not identify how missing endpoint information was handled.
12. Design Topics Not Specified in the ClinicalTrials.gov record
| Topic | What the registry-reported DARE-19 data establish | What cannot be concluded |
|---|---|---|
| Non-inferiority margin | No non-inferiority margin is reported. | A non-inferiority analysis should not be attributed to the trial from the available data. |
| Crossover | No crossover information is reported. | No crossover effect on interpretation can be assessed. |
| Factorial design | The design model is listed as parallel. | No factorial structure is reported. |
| Multiplicity | Two primary endpoints are registered. | The formal multiplicity-control strategy is not reported. |
| Interim analysis | No interim-analysis procedure is reported. | No stopping boundary or alpha-spending method can be attributed to DARE-19. |
| Missing data / imputation | No missing-data method is reported. | No imputation method can be inferred. |
| Stratification | No stratification factors are reported. | No stratified analysis should be attributed to the trial. |
| Bayesian methods | No Bayesian method is reported. | No Bayesian analysis should be inferred. |
13. Why This Trial Matters Statistically
DARE-19 is a useful teaching case because the registry record places two substantially different endpoint structures within the same randomized phase 3 trial: a conventional time-to-event endpoint and a hierarchical composite clinical-recovery endpoint.
| Concept | How it appears in DARE-19 |
|---|---|
| Randomization | Randomized allocation to dapagliflozin 10 mg or placebo. |
| Parallel design | Two parallel treatment arms. |
| Blinding | Quadruple-masked design. |
| Time-to-event analysis | Primary endpoint measuring time to first new/worsened organ dysfunction or death through Day 30. |
| Hierarchical composite | Primary clinical-recovery endpoint combines death, organ dysfunction, clinical status, and discharge/survival status. |
| Fixed follow-up window | Both registered primary endpoints use randomization through Day 30. |
| Safety analysis | Serious adverse events are reported by treatment arm using affected and at-risk counts. |
| Analysis transparency | The ClinicalTrials.gov record distinguishes posted results from the absence of registry-reported formal statistical-analysis estimates. |
14. How to Read the DARE-19 Statistical Record
Start with the estimand
Before interpreting a p-value or effect estimate, identify exactly what outcome is being compared, over what time period, and in which analysis population.
Separate endpoint structures
A time-to-event endpoint and a hierarchical composite endpoint should not be interpreted as though they were the same type of outcome.
Check the denominator
Safety results use 613 and 616 participants at risk. The overall enrollment is 1250. These figures answer different questions.
Do not infer missing statistics
A posted result does not justify reconstructing a treatment effect when the ClinicalTrials.gov record does not provide the underlying estimate or sufficient event-level information.
15. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial uses randomized parallel-group allocation and specifies two primary endpoints through Day 30. The first is a time-to-event endpoint; the second is a hierarchical composite. The ClinicalTrials.gov record does not contain formal comparative estimates.
Clinical interpretation
The clinical question concerns whether dapagliflozin 10 mg differs from placebo in patients with COVID-19 and respiratory failure with respect to the registered clinical outcomes. The ClinicalTrials.gov record does not support a conclusion about the direction or magnitude of the treatment effect.
16. A Statistical Reading of the Safety Data
The serious-adverse-event counts provide a straightforward descriptive starting point: 70 affected participants among 613 at risk in the dapagliflozin 10mg arm and 87 among 616 at risk in the placebo arm.
Those counts should not be treated as a formal efficacy or safety conclusion by themselves. A comparative safety analysis would need to specify the estimand, denominator, event definition, follow-up, analysis population, and statistical method. The ClinicalTrials.gov record does not provide those details beyond the affected/at-risk counts.
Descriptive safety data show the number of participants affected and the number considered at risk in each arm. They preserve the raw information reported by the registry without introducing an unreported statistical model. Formal comparative inference should be presented only when the underlying analysis is available.
17. What Is Known and What Remains Unspecified
| Question | Answer from the ClinicalTrials.gov record |
|---|---|
| Was the trial randomized? | Yes. |
| Was it parallel? | Yes. |
| Was it masked? | Yes; quadruple-masked. |
| How many participants were enrolled? | 1250. |
| How many arms? | 2. |
| What was the intervention? | Dapagliflozin 10 milligram (mg). |
| What was the comparator? | Placebo. |
| What were the primary endpoints? | Time to first new/worsened organ dysfunction or death; hierarchical clinical recovery composite. |
| What was the primary time frame? | Randomization through Day 30. |
| Are results posted? | Yes. |
| Are formal statistical analyses reported here? | No; the registry-reported statistical-analysis array is empty. |
| Are subgroup estimates reported? | No. |
| Are serious adverse events reported? | Yes; 70/613 versus 87/616. |
18. Related Tutorials
Learn more about the methods used in this trial:
19. Related Calculators
20. Sources
- ClinicalTrials.gov: DARE-19, NCT04350593.
- Linked PubMed record: PubMed 40739638.
- Linked PubMed record: PubMed 38770818.
- Linked PubMed record: PubMed 35699910.
- Linked PubMed record: PubMed 34302745.
Continue through the Clinical Biostats statistical tutorials
Use the trial's design and endpoint structure as a starting point for deeper study of randomized trials, blinding, and statistical analysis.
21. Record Summary
DARE-19 is a randomized, parallel-group, quadruple-masked phase 3 trial of dapagliflozin 10 mg versus placebo in patients with COVID-19 and respiratory failure. The trial enrolled 1250 participants and registered two primary endpoints, both assessed from randomization through Day 30: a time-to-first-event endpoint covering new or worsened organ dysfunction or death, and a hierarchical composite endpoint addressing clinical recovery. The ClinicalTrials.gov record confirms that results are posted for both endpoints but do not provide the formal statistical-analysis estimates needed to state treatment effects, confidence intervals, or p-values. The most statistically appropriate interpretation is therefore to distinguish the registered endpoint definitions and design from the unreported comparative estimates, while preserving the reported serious-adverse-event counts of 70/613 for dapagliflozin 10mg and 87/616 for placebo.