This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the ClinicalTrials.gov record; statistical explanations distinguish reported analyses from general interpretation.
1. Trial at a Glance
ACTT-4 was a completed phase 3, randomized, double-blind, parallel-group treatment trial in COVID-19. The trial enrolled 1010 participants and compared remdesivir plus baricitinib with remdesivir plus dexamethasone.
| Feature | ACTT-4 |
|---|---|
| Trial name | Adaptive COVID-19 Treatment Trial 4 (ACTT-4) |
| Phase | Phase 3 |
| Condition | COVID-19 |
| Design | Randomized, double-blind, parallel |
| Primary purpose | Treatment |
| Enrollment | 1010 |
| Lead sponsor | National Institute of Allergy and Infectious Diseases (NIAID) |
| Sponsor type | NIH |
| Start | 2020-12-02 |
| Primary completion | 2021-05-18 |
| ClinicalTrials.gov | NCT04640168 |
2. Clinical Question
The primary statistical question was whether remdesivir plus baricitinib differed from remdesivir plus dexamethasone with respect to the registered mechanical ventilation-free survival endpoint during Day 1 through Day 29. The registered hypothesis type was superiority.
Population
Participants enrolled in ACTT-4 with COVID-19. The ClinicalTrials.gov record identifies 1010 participants enrolled in the phase 3 treatment trial.
Intervention
Remdesivir plus baricitinib.
Comparator
Remdesivir plus dexamethasone.
Primary question
Does remdesivir plus baricitinib demonstrate superiority to remdesivir plus dexamethasone on the registered primary mechanical ventilation-free survival endpoint?
3. Trial Design
Remdesivir Plus Baricitinib
- Remdesivir
- Baricitinib
- Serious adverse events: 95/503
Remdesivir Plus Dexamethasone
- Remdesivir
- Dexamethasone
- Serious adverse events: 94/482
4. Primary Endpoints
The registry lists four primary endpoints. They share the same Day 1 through Day 29 mechanical ventilation-free survival definition but are additionally reported by race, ethnicity, and sex for three of the four entries.
| Registered primary endpoint | Time frame | Endpoint type | Reported analysis |
|---|---|---|---|
| The proportion of participants not meeting criteria for either death or hospitalization on invasive mechanical ventilation or ECMO | Day 1 through Day 29 | Time-to-event | Chi-squared analysis posted |
| The same mechanical ventilation-free survival endpoint by race | Day 1 through Day 29 | Time-to-event | Results posted; no separate formal statistical analysis reported |
| The same mechanical ventilation-free survival endpoint by ethnicity | Day 1 through Day 29 | Time-to-event | Results posted; no separate formal statistical analysis reported |
| The same mechanical ventilation-free survival endpoint by sex | Day 1 through Day 29 | Time-to-event | Results posted; no separate formal statistical analysis reported |
The registry definition states that mechanical ventilation-free survival was assessed through Day 29 and defined as the proportion of participants who had not died nor were hospitalized on invasive mechanical ventilation or ECMO as of Day 29. Results are reported as Kaplan-Meier estimates.
This is important statistically because the registry describes the endpoint using a time-to-event framework while the posted primary analysis is recorded as a chi-squared comparison of a binary outcome.
5. Analysis Populations
| Population | Definition / role |
|---|---|
| Modified intention-to-treat (mITT) | The registry-reported analysis text states that the mITT population includes all participants who were randomized. Participants were classified by randomized treatment assignment and baseline ordinal information. |
| Safety population | Participants with available data post baseline, analyzed as treated. |
The distinction between the mITT and safety populations is statistically important. Randomization provides the foundation for the primary efficacy comparison, whereas an as-treated safety population describes outcomes in relation to the treatment actually received. The two populations therefore answer related but different questions.
6. Primary Result
The formal statistical analysis posted for the principal mechanical ventilation-free survival endpoint compared remdesivir plus baricitinib with remdesivir plus dexamethasone in the mITT population.
Primary endpoint comparison
Chi-squared test · Superiority hypothesis
Analysis population: modified intention-to-treat (mITT)
| Endpoint | Groups compared | Method | P-value |
|---|---|---|---|
| Proportion not meeting criteria for death or hospitalization on invasive mechanical ventilation or ECMO, Day 1 through Day 29 | Remdesivir plus baricitinib vs remdesivir plus dexamethasone | Chi-squared | 0.909 |
What the result means: The registry-posted formal comparison used a chi-squared test and reports a P-value of 0.909 for the superiority hypothesis. A P-value of this magnitude does not provide statistical evidence against the null hypothesis under the specified test.
What it does not mean: The P-value is not an estimate of the size of the treatment effect. It does not say that the two groups were numerically identical, nor does it quantify the probability that the null hypothesis is true.
Confidence intervals: The registry-reported primary statistical-analysis record does not provide a confidence interval or an effect estimate for this primary comparison. Consequently, the precision and direction of the treatment difference cannot be summarized from the registry-reported formal analysis as an odds ratio, risk difference, or hazard ratio.
Why the P-value is not an effect size: A P-value measures compatibility of the observed data with the tested null hypothesis under the statistical procedure. It depends on the data and analysis structure; it is not a measure of clinical magnitude.
Time-to-event caution: The registry definition describes mechanical ventilation-free survival as a Day 1 through Day 29 time-to-event endpoint and states that results are reported as Kaplan-Meier estimates, while the posted formal analysis is classified as binary and uses a chi-squared test. Those descriptions should not be silently treated as interchangeable. The ClinicalTrials.gov record does not provide a Cox model, log-rank test, hazard ratio, or corresponding confidence interval for the primary endpoint.
7. Secondary Endpoint Results
Grade 3 and 4 Clinical and/or Laboratory Adverse Events
Reported risk difference
95% CI: 1.8–13.4 · Two-sided · Day 1 through Day 29
The registry reports a risk difference of 7.7 for the percentage of participants reporting Grade 3 and 4 clinical and/or laboratory adverse events. The ClinicalTrials.gov record does not state the statistical method used to construct this interval.
What the estimate means: A risk difference expresses an absolute difference between event proportions. The reported estimate is 7.7, with the comparison defined as remdesivir plus baricitinib versus remdesivir plus dexamethasone.
What it does not mean: The estimate is not a relative risk, odds ratio, or hazard ratio. It should not be interpreted as a percentage reduction or increase in instantaneous event rate.
Confidence interval: The two-sided 95% confidence interval is 1.8 to 13.4. The interval describes statistical uncertainty around the reported risk-difference estimate; it is not a range containing the outcomes that individual participants will experience.
Direction matters: Because the ClinicalTrials.gov record does not explicitly define the subtraction convention used for the reported risk difference, the numerical estimate is reported without assigning an additional directional clinical interpretation beyond the stated group comparison.
Safety-population caution: This analysis used the safety population and was analyzed as treated, rather than according to randomized assignment. That distinction matters when interpreting treatment-associated adverse events.
Serious Adverse Events
Reported risk difference
95% CI: −4.3–5.5 · Two-sided · Day 1 through Day 29
| Measure | Remdesivir + Baricitinib | Remdesivir + Dexamethasone | Reported effect |
|---|---|---|---|
| Serious adverse events | 95/503 affected/at risk | 94/482 affected/at risk | Risk difference 0.6 (95% CI −4.3 to 5.5) |
What the estimate means: The reported absolute risk-difference estimate for serious adverse events is 0.6 for remdesivir plus baricitinib compared with remdesivir plus dexamethasone.
Precision: The two-sided 95% confidence interval ranges from −4.3 to 5.5. Because the interval includes zero, the registry-reported interval is compatible with both a modest negative and a modest positive absolute difference.
P-values versus confidence intervals: No P-value is posted on ClinicalTrials.gov for this secondary analysis. The confidence interval itself provides information about the range of effect estimates compatible with the statistical analysis.
Safety interpretation: Serious adverse events are clinically important events, but their interpretation still depends on the analysis population, follow-up window, event definitions, and uncertainty interval.
Day 15 Ordinal Scale Categories
Reported odds ratio
95% CI: 0.80–1.27 · Two-sided · Day 15
The secondary endpoint was the percentage of participants meeting criteria for each of the 8 ordinal scale categories at Day 15. The ClinicalTrials.gov record reports an odds ratio of 1.01 with a two-sided 95% confidence interval of 0.80 to 1.27. The formal statistical method is not reported in the ClinicalTrials.gov record.
What the estimate means: An odds ratio of 1.01 is very close to 1.0, the value corresponding to equal odds under the usual odds-ratio parameterization.
What it does not mean: An odds ratio of 1.01 is not a 1.01% improvement, nor is it a risk ratio of 1.01. Odds and probabilities are different quantities, especially when event probabilities are not small.
Confidence interval: The 95% CI of 0.80 to 1.27 spans 1.0. The registry-reported interval therefore includes values corresponding to lower and higher odds under the conventional treatment-versus-comparator interpretation.
Ordinal-scale caution: The endpoint contains eight categories. The ClinicalTrials.gov record reports the odds ratio but does not report the statistical model or its proportional-odds assumptions. It therefore would be inappropriate to infer a specific ordinal regression model from the effect measure alone.
Proportion Not Meeting Criteria for the Three Most Severe Ordinal Categories
Reported risk difference
95% CI: −0.09–0.02 · Two-sided · Day 1 through Day 29
The secondary endpoint assessed the proportion of participants not meeting criteria for one of the three most severe ordinal scale categories at any time. The reported risk difference was −0.03, with a two-sided 95% confidence interval of −0.09 to 0.02. The formal statistical method was not reported.
What the estimate means: The reported absolute effect measure is a risk difference of −0.03 for the stated comparison. A risk difference is an absolute difference in event proportions, not a relative measure.
Confidence interval: The 95% CI ranges from −0.09 to 0.02 and includes zero. The ClinicalTrials.gov record therefore do not establish a nonzero absolute difference at the conventional two-sided confidence-interval level.
Missing method detail: Because the registry does not report the statistical method for this secondary endpoint in the ClinicalTrials.gov record, the page does not infer a particular test or model.
Time to Improvement of One Category
Reported hazard ratio
95% CI: 0.94–1.23 · Two-sided · Day 1 through Day 29
The time-to-event endpoint measured time to an improvement of one category from baseline using an ordinal scale. The reported hazard ratio was 1.08, with a two-sided 95% confidence interval of 0.94 to 1.23.
What the estimate means: A hazard ratio of 1.08 means that, under the fitted time-to-event framework if the reported hazard ratio is interpreted in its conventional direction, the estimated instantaneous rate of reaching the specified improvement was 1.08 times the comparator rate.
What it does not mean: It does not mean that participants improved by 8%, nor that the median time to improvement was 8% shorter. A hazard ratio describes relative instantaneous event rates, not a direct ratio of elapsed times.
Confidence interval: The 95% CI of 0.94 to 1.23 includes 1.0. The interval therefore includes both modestly lower and modestly higher instantaneous improvement rates.
Model caution: The ClinicalTrials.gov record does not report the method used to obtain this hazard ratio. A conventional Cox interpretation would require attention to proportional-hazards assumptions, but the ClinicalTrials.gov record does not establish that a Cox model was used.
Time to Improvement of Two Categories
Reported hazard ratio
95% CI: 0.94–1.23 · Two-sided · Day 1 through Day 29
The time-to-event endpoint measured time to an improvement of two categories from baseline using an ordinal scale. The reported hazard ratio was 1.08, with a two-sided 95% confidence interval of 0.94 to 1.23.
What the estimate means: The reported hazard ratio is 1.08, indicating an estimated instantaneous event-rate ratio of 1.08 for the stated comparison under the conventional hazard-ratio interpretation.
Precision: The 95% CI of 0.94 to 1.23 includes 1.0, so the interval is compatible with effects in either direction around the null value.
Do not equate HR with time ratio: A hazard ratio is not a direct ratio of mean or median times. Interpreting it as "8% faster" would require assumptions that are not supported by the registry entry.
Time to Recovery
Reported hazard ratio
95% CI: 0.91–1.19 · Two-sided · Day 1 through Day 29
The registry reports a hazard ratio of 1.04 for time to recovery, with a two-sided 95% confidence interval of 0.91 to 1.19. The analysis population was the mITT population, and the formal statistical method was not reported in the ClinicalTrials.gov record.
What the estimate means: A hazard ratio of 1.04 is close to the conventional null value of 1.0. Under a standard hazard-ratio interpretation, it represents an estimated instantaneous recovery-event rate ratio of 1.04.
Confidence interval: The 95% CI of 0.91 to 1.19 includes 1.0, so the interval includes both modestly lower and modestly higher recovery-event rates.
What it does not mean: It does not mean that recovery occurred in 104% of participants or that recovery time differed by exactly 4%. Hazard ratios are relative time-to-event measures.
8. Results Summary
| Endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Primary mechanical ventilation-free survival endpoint | Chi-squared comparison | Not reported | Not reported | 0.909 |
| Grade 3 and 4 clinical and/or laboratory AEs | Risk difference | 7.7 | 1.8–13.4 | Not reported |
| Serious adverse events | Risk difference | 0.6 | −4.3–5.5 | Not reported |
| Day 15 ordinal scale categories | Odds ratio | 1.01 | 0.80–1.27 | Not reported |
| Not meeting criteria for three most severe ordinal categories | Risk difference | −0.03 | −0.09–0.02 | Not reported |
| Time to improvement of one category | Hazard ratio | 1.08 | 0.94–1.23 | Not reported |
| Time to improvement of two categories | Hazard ratio | 1.08 | 0.94–1.23 | Not reported |
| Time to recovery | Hazard ratio | 1.04 | 0.91–1.19 | Not reported |
This table illustrates why an overall interpretation should not collapse all ACTT-4 outcomes into a single number. The primary endpoint has a formal chi-squared P-value but no registry-reported effect estimate, while the secondary endpoints use different effect measures and have different analysis populations.
9. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm and provide a separate secondary analysis for Grade 3 and 4 clinical and/or laboratory adverse events.
| Safety measure | Remdesivir + Baricitinib | Remdesivir + Dexamethasone | Reported analysis |
|---|---|---|---|
| Serious adverse events | 95/503 | 94/482 | Risk difference 0.6 (95% CI −4.3 to 5.5) |
| Grade 3 and 4 clinical and/or laboratory adverse events | Secondary endpoint | Risk difference 7.7 (95% CI 1.8–13.4) | |
The serious-adverse-event counts are reported as affected participants over participants at risk: 95/503 for remdesivir plus baricitinib and 94/482 for remdesivir plus dexamethasone. The ClinicalTrials.gov record does not provide a P-value for this safety comparison.
10. Statistical Methodology
Chi-squared test for the primary comparison
The formal primary analysis posted in the ClinicalTrials.gov record used a chi-squared test. This is a categorical-data method for comparing observed outcome distributions or proportions between groups.
The chi-squared framework compares observed cell counts with the counts expected under the null hypothesis. For a two-group binary endpoint, the test can be used to assess whether the observed proportions differ between treatment groups.
For ACTT-4, the primary analysis record describes the endpoint as binary for the formal comparison and compares remdesivir plus baricitinib with remdesivir plus dexamethasone in the mITT population. The reported P-value is 0.909.
Kaplan-Meier estimation
The registered primary endpoint definition states that mechanical ventilation-free survival was assessed through Day 29 and that results are reported as Kaplan-Meier estimates. Kaplan-Meier estimation is designed for time-to-event data and accounts for participants whose event status is not observed over the complete follow-up period.
Here, di represents events at an observed event time and ni represents participants at risk immediately before that time.
The ClinicalTrials.gov record does not provide the underlying event and censoring data needed to reconstruct a Kaplan-Meier curve. They also do not provide a formal log-rank or Cox-model analysis for the primary endpoint.
Risk difference
Several secondary endpoints are reported using the risk difference. Conceptually, the risk difference compares two event probabilities on an absolute scale.
A risk difference of zero corresponds to equal event probabilities under the stated comparison. Positive and negative values indicate opposite directions depending on the prespecified subtraction convention.
For ACTT-4, the ClinicalTrials.gov record reports risk-difference estimates of 7.7, 0.6, and −0.03 for different secondary endpoints. The registry data do not explicitly provide the subtraction convention for each estimate, so this page does not add an inferred direction to those numbers.
Odds ratio
The Day 15 ordinal-scale endpoint reports an odds ratio of 1.01 with a 95% confidence interval of 0.80 to 1.27. An odds ratio compares odds rather than probabilities.
An OR of 1 corresponds to equal odds. The odds ratio is not generally interchangeable with a risk ratio.
Hazard ratio
Three secondary time-to-event outcomes are reported using hazard ratios: time to improvement of one category, time to improvement of two categories, and time to recovery.
Under the conventional interpretation, HR = 1 represents equal instantaneous event rates. HR values above or below 1 describe relative differences in instantaneous event rates, not direct differences in elapsed time.
The ClinicalTrials.gov record does not identify the formal model used for these hazard-ratio analyses. Therefore, although Cox regression is a common method for such endpoints, this page does not state that a Cox model was used for these particular ACTT-4 analyses.
Confidence intervals
The secondary endpoint analyses use two-sided 95% confidence intervals. Confidence intervals are useful because they display both the point estimate and the uncertainty around it. For example, the serious-adverse-event risk difference of 0.6 has a 95% CI of −4.3 to 5.5, whereas the Grade 3 and 4 adverse-event risk difference of 7.7 has a 95% CI of 1.8 to 13.4.
11. Statistical Methods Explained
Why use a chi-squared test for a binary endpoint?
A chi-squared test is designed for categorical data and can compare the observed frequencies of binary outcomes across treatment groups. In ACTT-4, the registry-reported formal primary analysis classified the endpoint as binary and used a chi-squared comparison in the mITT population.
Why is the primary endpoint described as time-to-event while the posted analysis is binary?
The registry definition describes mechanical ventilation-free survival through Day 29 and states that results are reported as Kaplan-Meier estimates. The statistical-analysis entry separately classifies the posted analysis as binary and reports a chi-squared test. These are different analytical descriptions. A careful statistical review should preserve that distinction rather than assume that a Cox model or log-rank test was performed.
What does a hazard ratio of 1.08 mean?
A hazard ratio of 1.08 represents an estimated instantaneous event-rate ratio of 1.08 under the conventional interpretation. It does not mean that participants recovered 8% faster, because a hazard ratio is not a direct ratio of elapsed times.
Why does the confidence interval matter?
The confidence interval shows the uncertainty surrounding the point estimate. For example, the hazard ratio of 1.04 for time to recovery has a 95% CI of 0.91 to 1.19. The interval includes 1.0, so the ClinicalTrials.gov record is compatible with both modestly lower and modestly higher instantaneous recovery-event rates.
Why is a P-value not an effect size?
The primary P-value of 0.909 is evidence from a statistical test against its null hypothesis; it is not a measure of how large or clinically important a treatment difference is. Effect size requires an estimate such as a risk difference, odds ratio, or hazard ratio together with its uncertainty.
Why distinguish mITT from the safety population?
The mITT analysis preserves the randomized treatment assignment for efficacy comparisons, while the safety population is analyzed as treated among participants with available post-baseline data. These choices answer different questions and can produce different denominators.
12. Interpreting the Endpoint Family
ACTT-4 illustrates an important feature of clinical-trial analysis: a single randomized trial can contain endpoints that require different statistical representations. The primary endpoint is framed around mechanical ventilation-free survival, while secondary outcomes include binary adverse-event measures, an ordinal-scale outcome summarized with an odds ratio, absolute risk differences, and time-to-event outcomes summarized with hazard ratios.
| Endpoint type | ACTT-4 example | Reported measure | Key interpretation issue |
|---|---|---|---|
| Binary | Primary mechanical ventilation-free survival analysis | Chi-squared P-value | No primary effect estimate reported in the ClinicalTrials.gov record |
| Binary safety | Grade 3 and 4 AEs | Risk difference | Absolute rather than relative measure |
| Binary safety | Serious AEs | Risk difference | Safety population analyzed as treated |
| Ordinal scale | Day 15 categories | Odds ratio | Formal statistical model not reported |
| Time-to-event | Time to improvement of one category | Hazard ratio | Hazard ratio is not a direct time ratio |
| Time-to-event | Time to improvement of two categories | Hazard ratio | Confidence interval includes 1.0 |
| Time-to-event | Time to recovery | Hazard ratio | Formal statistical model not reported |
13. Multiplicity and Multiple Endpoints
The ClinicalTrials.gov record identifies 4 registered primary endpoints and 44 posted outcome measures, with 8 statistical analyses posted. The primary hypothesis type is recorded as superiority.
Why multiplicity matters
When several hypotheses are evaluated, the probability of obtaining at least one apparently unusual result can increase if every test is interpreted independently at the same nominal threshold.
What the ClinicalTrials.gov record does not show
The ClinicalTrials.gov record does not specify a multiplicity-adjustment procedure, alpha allocation, hierarchical testing strategy, or a prespecified familywise error-control plan.
Accordingly, the reported P-value of 0.909 should be described exactly as the posted primary analysis P-value rather than being combined with an inferred multiplicity framework that is not contained in the ClinicalTrials.gov record.
14. Subgroup-Related Primary Endpoints
Three of the four registered primary endpoints repeat the mechanical ventilation-free survival definition by race, ethnicity, and sex. The ClinicalTrials.gov record states that results are posted for each but do not provide separate formal statistical-analysis entries for those three subgroup-defined endpoints.
| Subgroup dimension | Time frame | Formal analysis reported? |
|---|---|---|
| Race | Day 1 through Day 29 | No separate formal analysis reported |
| Ethnicity | Day 1 through Day 29 | No separate formal analysis reported |
| Sex | Day 1 through Day 29 | No separate formal analysis reported |
Subgroup results require additional caution because subgroup sample sizes can be smaller than the overall randomized population. More importantly, observing different estimates across subgroups does not itself demonstrate that treatment effects differ. A formal interaction analysis is generally required to assess treatment-effect heterogeneity, but the registry-reported ACTT-4 data do not report such an analysis.
15. Censoring and Time-to-Event Interpretation
The registry describes mechanical ventilation-free survival and several secondary outcomes as time-to-event endpoints. Time-to-event methods are useful when participants can have different amounts of observed follow-up or when the exact timing of an event matters.
Kaplan-Meier estimation, as identified in the primary endpoint definition, uses the available event-time information while accounting for censoring. However, the ClinicalTrials.gov record for the primary endpoint is a chi-squared comparison of a binary outcome. These two analytical representations answer somewhat different questions: a Day 29 binary analysis focuses on final status by a specified time, whereas a survival analysis uses the timing of events throughout follow-up.
16. Statistical Interpretation of the Main Effect Measures
The secondary hazard ratios are 1.08, 1.08, and 1.04. Their respective 95% confidence intervals are 0.94–1.23, 0.94–1.23, and 0.91–1.19. Each interval includes 1.0. These estimates should be interpreted as relative instantaneous event-rate measures rather than direct ratios of recovery times.
The Day 15 ordinal-scale analysis reports an odds ratio of 1.01 with a 95% CI of 0.80–1.27. The interval includes the conventional null value of 1.0. The ClinicalTrials.gov record does not identify the model used to generate this odds ratio.
The reported risk differences are 7.7, 0.6, and −0.03, with corresponding 95% CIs of 1.8–13.4, −4.3–5.5, and −0.09–0.02. These measures are on an absolute scale and therefore complement relative measures such as odds ratios and hazard ratios.
The primary formal analysis reports P = 0.909 from a chi-squared test. The ClinicalTrials.gov record does not provide the corresponding primary effect estimate or confidence interval, so the magnitude and precision of the primary treatment difference cannot be reconstructed from the registry-reported analysis entry.
17. Limitations
- Primary effect estimate not reported: The formal primary analysis reports P = 0.909 but does not provide the corresponding primary endpoint proportions, risk difference, odds ratio, hazard ratio, or confidence interval in the ClinicalTrials.gov record.
- Different endpoint descriptions: The primary endpoint is described as time-to-event in the registered endpoint section and as binary in the formal statistical-analysis entry. The page preserves both descriptions rather than inferring an unreported survival model.
- Missing method detail for secondary outcomes: The ClinicalTrials.gov record provides effect measures for several secondary endpoints but label the statistical method as not reported.
- Multiplicity information is incomplete: The ClinicalTrials.gov record identifies four registered primary endpoints and numerous posted outcomes but do not specify an alpha-adjustment or multiplicity-control procedure.
- Subgroup analysis detail is limited: Race, ethnicity, and sex appear in the registered primary endpoints, but separate formal subgroup analyses are not reported.
- Safety and efficacy populations differ: Efficacy analyses use the mITT population, while safety analyses use participants with available post-baseline data analyzed as treated.
- No individual-level time-to-event data: Event times, censoring times, and risk sets are not reported, preventing independent reconstruction of Kaplan-Meier curves or alternative survival estimates.
- No baseline table reported: The ClinicalTrials.gov record does not contain baseline demographic or clinical characteristics by treatment group.
- No treatment recommendation follows from the statistical page: Statistical estimates describe the reported randomized comparison and its uncertainty; clinical treatment decisions require the full clinical evidence and context.
18. Why This Trial Matters Statistically
ACTT-4 is a useful teaching case because it combines randomized treatment allocation with several different forms of clinical-trial endpoints. The same trial record illustrates categorical testing, absolute risk differences, odds ratios, hazard ratios, Kaplan-Meier estimation, intention-to-treat principles, and safety analysis.
| Concept | How it appears in ACTT-4 |
|---|---|
| Randomization | Participants were randomized in a phase 3 parallel-group trial. |
| Double blinding | The registered design is double-masked. |
| Intention-to-treat analysis | The mITT population includes all randomized participants in the registry-reported analysis description. |
| Chi-squared test | Used for the posted formal primary endpoint comparison. |
| Kaplan-Meier estimation | Specified in the registry definition of mechanical ventilation-free survival. |
| Risk difference | Used for Grade 3 and 4 AEs, serious AEs, and the three-most-severe-category endpoint. |
| Odds ratio | Reported for the Day 15 ordinal-scale endpoint. |
| Hazard ratio | Reported for time to one-category improvement, two-category improvement, and recovery. |
| Confidence intervals | Two-sided 95% intervals accompany the reported secondary effect estimates. |
| Safety analysis | Uses a safety population with participants analyzed as treated. |
| Subgroup endpoints | Primary endpoint definitions are separately listed by race, ethnicity, and sex. |
| Multiplicity | Four registered primary endpoints and 44 posted outcome measures create a setting in which endpoint hierarchy and error control are important considerations. |
19. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The formal primary comparison uses a chi-squared test and reports P = 0.909. Secondary analyses provide risk differences, an odds ratio, and hazard ratios with two-sided 95% confidence intervals. Several secondary intervals include their conventional null values.
Clinical interpretation
The reported endpoints describe mechanical ventilation-free survival, clinical recovery, ordinal improvement, and adverse events over defined follow-up windows. Their clinical meaning depends on the endpoint definition, absolute event probabilities, analysis population, and uncertainty around each estimate.
The distinction is important. A statistical result does not by itself determine clinical importance. Conversely, a clinically interesting endpoint should not be interpreted without considering its analysis population, effect measure, confidence interval, and prespecified role in the trial.
20. Overall Statistical Reading of ACTT-4
The registry-reported ACTT-4 record contains a clear randomized comparison and a formal primary analysis, but it does not provide enough detail to reproduce every aspect of the primary statistical comparison independently. The most important reported primary result is the chi-squared P-value of 0.909.
The secondary results provide more information about effect magnitude. The Grade 3 and 4 adverse-event analysis reports a risk difference of 7.7 with a 95% CI of 1.8–13.4. Serious adverse events have a reported risk difference of 0.6 with a 95% CI of −4.3–5.5. The Day 15 ordinal-scale endpoint reports an odds ratio of 1.01 with a 95% CI of 0.80–1.27. Time to one-category improvement and two-category improvement each has an HR of 1.08 with a 95% CI of 0.94–1.23, while time to recovery has an HR of 1.04 with a 95% CI of 0.91–1.19.
These results demonstrate why clinical-trial interpretation should combine the endpoint definition, analysis population, effect measure, confidence interval, and statistical method rather than relying on a single P-value.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: ACTT-4, NCT04640168.
- Publication: PubMed record, PMID 36442063.
- Publication: PubMed record, PMID 35617986.
- Publication: PubMed record, PMID 34473343.
Continue through the Clinical Biostats library
Explore statistical tutorials and calculators covering the categorical, survival, and clinical-trial methods illustrated by ACTT-4.
24. Record Summary
ACTT-4 provides a useful example of how randomized clinical-trial evidence can involve several statistical frameworks within the same study. The trial was a phase 3, randomized, double-blind, parallel comparison of remdesivir plus baricitinib versus remdesivir plus dexamethasone in COVID-19, with 1010 participants enrolled.
The formal primary analysis used a chi-squared test in the modified intention-to-treat population and reported P = 0.909. The registry separately defines mechanical ventilation-free survival as a Day 1 through Day 29 time-to-event endpoint and states that results are reported as Kaplan-Meier estimates. Because the ClinicalTrials.gov record does not provide the primary effect estimate or confidence interval, the primary result should be reported as the posted statistical test rather than reconstructed into an unsupported effect measure.
The secondary analyses demonstrate several complementary approaches: risk differences for adverse-event and ordinal outcomes, an odds ratio for the Day 15 ordinal-scale endpoint, and hazard ratios for time-to-improvement and time-to-recovery endpoints. The reported confidence intervals provide information about precision and should be interpreted alongside the point estimates rather than treated as substitutes for effect size or clinical importance.