This page separates reported trial results from statistical interpretation. Trial facts and numerical results on this page are limited to the information reported in the ClinicalTrials.gov record.
1. Trial at a Glance
ACTT-2 was a randomized, double-masked, parallel phase 3 trial in COVID-19. The trial compared remdesivir plus baricitinib with remdesivir plus placebo and registered time to recovery as its primary endpoint, with additional primary analyses by race, ethnicity, and sex.
| Feature | ACTT-2 |
|---|---|
| Trial name | Adaptive COVID-19 Treatment Trial 2 (ACTT-2) |
| Phase | Phase 3 |
| Condition | COVID-19 |
| Design | Randomized, parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 1033 |
| Primary endpoints registered | 4 |
| Primary endpoint type | Time-to-event |
| Results posted | Yes |
| Outcome measures posted | 45 |
| Statistical analyses posted | 15 |
| Primary-endpoint analyses | 9 |
| Lead sponsor | National Institute of Allergy and Infectious Diseases (NIAID) |
| Sponsor type | NIH |
2. Clinical Question
The registered treatment comparison asks whether adding baricitinib to remdesivir changes the time to recovery and other clinical outcomes compared with remdesivir plus placebo in participants with COVID-19.
Population
Participants in a phase 3 randomized clinical trial for COVID-19. The ClinicalTrials.gov record reports enrollment of 1033 participants.
Intervention
Remdesivir plus baricitinib.
Comparator
Remdesivir plus placebo.
Primary question
Does remdesivir plus baricitinib produce a different time-to-recovery profile from remdesivir plus placebo?
3. Trial Design
Remdesivir Plus Baricitinib
- Remdesivir
- Baricitinib
Remdesivir Plus Placebo
- Remdesivir
- Placebo
Trial timing
Trial start
The registered study start date was 2020-05-08.
Primary completion
The registered primary completion date was 2020-07-31.
4. Endpoints
ClinicalTrials.gov registered four primary endpoints. All four were time-to-recovery endpoints evaluated from Day 1 through Day 29. The recovery definition is identical across the four registered primary endpoints.
| Primary endpoint | Time frame | Registry definition |
|---|---|---|
| Time to Recovery | Day 1 through Day 29 | Day of recovery is the first day on which the participant satisfies one of three ordinal-scale categories: 1) not hospitalized, no limitations on activities; 2) not hospitalized, limitation on activities and/or requiring home oxygen; or 3) hospitalized, not requiring supplemental oxygen and no longer requiring ongoing medical care. |
| Time to Recovery by Race | Day 1 through Day 29 | Same registered recovery definition as the overall endpoint, with analyses reported for race groups. |
| Time to Recovery by Ethnicity | Day 1 through Day 29 | Same registered recovery definition as the overall endpoint, with analyses reported for ethnicity groups. |
| Time to Recovery by Sex | Day 1 through Day 29 | Same registered recovery definition as the overall endpoint, with analyses reported for sex groups. |
Secondary endpoints with posted statistical analyses
| Secondary endpoint | Time frame | Analysis | Effect measure |
|---|---|---|---|
| Percentage of Participants Reporting Grade 3 and 4 Clinical and/or Laboratory Adverse Events (AEs) | Day 1 through Day 29 | Not reported | Risk difference |
| Percentage of Participants Reporting Serious Adverse Events (SAEs) | Day 1 through Day 29 | Not reported | Risk difference |
| Percentage of Participants at Each Clinical Status Using Ordinal Scale at Day 15 | Day 15 | Logistic regression | Odds ratio |
| Time to an Improvement of One Category Using an Ordinal Scale | Day 1 through Day 29 | Log-rank test | Hazard ratio |
| Time to an Improvement of Two Categories Using an Ordinal Scale | Day 1 through Day 29 | Log-rank test | Hazard ratio |
| Time to Discharge or to a NEWS of 2 or Less and Maintained for 24 Hours, Whichever Occurs First | Day 1 through Day 29 | Log-rank test | Hazard ratio |
5. Statistical Methodology
Intention-to-treat analysis
The primary time-to-recovery analysis used the intent-to-treat (ITT) population, defined in the registry as including all participants who were randomized. This is important because the treatment comparison remains tied to randomized assignment rather than being restricted to participants who completed treatment exactly as planned.
The principal statistical advantage is preservation of the treatment comparison created by randomization. ITT does not mean that every participant necessarily has an observed recovery event; time-to-event methods are designed to incorporate incomplete follow-up through censoring where appropriate.
Log-rank test
The registry reports a log-rank test for the overall time-to-recovery analysis and for three secondary time-to-event endpoints. The log-rank test compares the event-time distributions between treatment groups rather than comparing only whether an event occurred by a single fixed time.
For a recovery endpoint, the event is recovery rather than death or disease progression. A higher recovery hazard generally corresponds to recovery occurring sooner, provided the event definition and censoring framework are as specified.
Cox proportional-hazards effect measure
The primary analysis reports a Cox proportional-hazards effect measure. The resulting hazard ratio compares the estimated instantaneous rate of recovery between remdesivir plus baricitinib and remdesivir plus placebo over the analyzed period.
For this trial's recovery endpoint, an HR above 1 therefore corresponds to a faster estimated recovery process under the model. The HR is not a ratio of median recovery times and does not mean that every participant recovers a fixed percentage faster.
Logistic regression
The Day 15 ordinal-scale endpoint was analyzed using logistic regression. The registry reports an odds ratio as the effect measure. Logistic regression is appropriate when the analysis reduces an outcome to a binary status for the comparison being modeled; the odds ratio then describes the relative odds of the modeled outcome between treatment groups.
Risk difference
The two posted safety analyses use risk difference. Unlike a hazard ratio or odds ratio, a risk difference is an absolute measure: it describes the difference in the proportion of participants experiencing the specified event between the two groups.
Two-sided confidence intervals
The posted primary and secondary effect estimates with confidence intervals use 95% two-sided confidence intervals. A confidence interval communicates statistical precision around an estimate; it does not describe the range of individual participant outcomes.
6. Results: Time to Recovery
The primary time-to-recovery analysis compared remdesivir plus baricitinib with remdesivir plus placebo in the ITT population from Day 1 through Day 29. The registry reports a log-rank test and a Cox proportional-hazards effect estimate.
Primary time-to-recovery hazard ratio
95% CI: 1.00–1.31 · P = 0.047
Comparison: Remdesivir Plus Baricitinib vs Remdesivir Plus Placebo
An HR of 1.15 means that, under the reported Cox model, the estimated instantaneous rate of meeting the trial's recovery definition was 15% higher with remdesivir plus baricitinib than with remdesivir plus placebo over the Day 1 through Day 29 analysis period.
It does not mean that participants recovered 15% faster in an individual-level sense, that 15% more participants recovered, or that recovery time for every participant was reduced by 15%.
The 95% CI of 1.00–1.31 describes uncertainty around the estimated hazard ratio. Its lower endpoint is 1.00, so the interval reaches the conventional no-difference value for a hazard ratio.
The reported P = 0.047 is evidence from the specified statistical test against the null hypothesis under the trial's analysis framework. A p-value is not a measure of effect size and does not indicate the probability that the treatment works.
Because the endpoint is time-to-recovery, interpretation also depends on the handling of censoring and on the assumptions underlying the Cox proportional-hazards effect measure. The ClinicalTrials.gov record does not provide enough information to independently evaluate those assumptions.
| Primary analysis feature | Reported value |
|---|---|
| Analysis population | ITT population; all participants randomized |
| Comparison | Remdesivir Plus Baricitinib vs Remdesivir Plus Placebo |
| Time frame | Day 1 through Day 29 |
| Endpoint | Time to Recovery |
| Test | Log-rank |
| Effect measure | Cox proportional-hazards hazard ratio |
| Estimate | 1.15 |
| 95% CI | 1.00–1.31 |
| P-value | 0.047 |
| Hypothesis type | Superiority |
7. Primary Results by Race
ClinicalTrials.gov also posts Cox proportional-hazards estimates for racial groups under the registered Time to Recovery by Race endpoint. The ClinicalTrials.gov record identifies these as primary analyses, but do not report the analysis method field for these subgroup analyses beyond the reported Cox proportional-hazards effect measure.
| Race subgroup | Hazard ratio | 95% CI |
|---|---|---|
| Asian participants | 1.11 | 0.73–1.68 |
| Black or African American participants | 1.06 | 0.75–1.50 |
| White participants | 1.13 | 0.93–1.37 |
| Race of Other participants | 1.34 | 1.03–1.74 |
The subgroup estimates range from 1.06 to 1.34. An HR above 1 is directionally consistent with a higher estimated instantaneous recovery rate for remdesivir plus baricitinib, but the confidence intervals show different levels of statistical precision across the reported groups.
The Asian estimate of 1.11 has a 95% CI of 0.73–1.68, while the Black or African American estimate of 1.06 has a 95% CI of 0.75–1.50. Both intervals include 1.00.
The White estimate is 1.13 with a 95% CI of 0.93–1.37. The Race of Other estimate is 1.34 with a 95% CI of 1.03–1.74.
These subgroup estimates should not be interpreted as proof that treatment effects differ between racial groups merely because their individual confidence intervals differ. A formal claim of treatment-effect heterogeneity ordinarily requires an interaction test or another prespecified comparison of treatment effects. No such interaction result is reported in the ClinicalTrials.gov record.
8. Primary Results by Ethnicity
The registered Time to Recovery by Ethnicity endpoint was analyzed in an ITT population consisting of all randomized participants for whom ethnicity was reported. Two ethnicity-specific Cox proportional-hazards estimates are posted.
| Ethnicity subgroup | Hazard ratio | 95% CI |
|---|---|---|
| Not Hispanic or Latino participants | 1.31 | 1.08–1.60 |
| Hispanic or Latino participants | 1.08 | 0.89–1.31 |
The estimated recovery hazard ratio is 1.31 among participants classified as Not Hispanic or Latino and 1.08 among Hispanic or Latino participants.
The 95% CI for the Not Hispanic or Latino analysis is 1.08–1.60, whereas the interval for the Hispanic or Latino analysis is 0.89–1.31. These intervals quantify uncertainty within each subgroup; they do not directly test whether the two subgroup effects are statistically different from each other.
The ethnicity analysis also uses a somewhat different analysis-population definition from the overall primary analysis because the registry specifies that ethnicity must be reported. That distinction matters when interpreting subgroup estimates.
9. Primary Results by Sex
The registered Time to Recovery by Sex endpoint was analyzed in the ITT population. The ClinicalTrials.gov record includes separate Cox proportional-hazards estimates for male and female participants.
| Sex subgroup | Hazard ratio | 95% CI |
|---|---|---|
| Male participants | 1.23 | 1.04–1.46 |
| Female participants | 1.06 | 0.85–1.32 |
The estimated recovery hazard ratio is 1.23 for male participants and 1.06 for female participants. Both point estimates are above 1, but their confidence intervals differ in width and position.
The male-participant estimate has a 95% CI of 1.04–1.46, while the female-participant estimate has a 95% CI of 0.85–1.32. The female interval includes 1.00.
A subgroup confidence interval crossing 1.00 does not establish that there is no treatment effect in that subgroup, just as an interval not crossing 1.00 does not by itself establish that the treatment effects differ between subgroups. The appropriate question for effect modification is whether there is evidence of an interaction between treatment and subgroup.
10. Secondary Results: Ordinal-Scale Outcomes
Day 15 Clinical Status
The percentage of participants at each clinical status using the ordinal scale at Day 15 was analyzed using logistic regression in the ITT population. The registry reports an odds ratio as the effect measure.
Day 15 ordinal-scale analysis
95% CI: 1.01–1.57 · P = 0.44
Analysis: logistic regression; ITT population
An odds ratio of 1.26 indicates that the modeled odds for the outcome represented by the logistic regression were estimated to be 1.26 times those in the remdesivir-plus-placebo group.
An odds ratio is not a risk ratio. The odds ratio also cannot be interpreted as saying that the probability of the clinical status was 26% higher without knowing the underlying event probabilities and the exact binary contrast used by the regression model.
The ClinicalTrials.gov record reports a 95% CI of 1.01–1.57 and a P-value of 0.44. Those fields should be reproduced as posted rather than recalculated from the summary estimate. The apparent tension between the reported confidence interval and p-value is a reason to avoid independently deriving significance from one field while disregarding another.
Time to One-Category Improvement
The registry reports a time-to-event analysis for time to an improvement of one category using the ordinal scale, from Day 1 through Day 29. The ITT population was analyzed using a log-rank test with a Cox proportional-hazards effect measure.
One-category improvement
95% CI: 1.06–1.39 · P = 0.002
The HR of 1.21 corresponds to a 21% higher estimated instantaneous rate of achieving the defined one-category improvement under the Cox model for remdesivir plus baricitinib versus remdesivir plus placebo.
The 95% CI of 1.06–1.39 describes uncertainty around that relative hazard estimate. It does not mean that individual participants had outcomes confined to that range.
The P-value of 0.002 provides evidence against the corresponding null hypothesis under the reported testing framework. It does not quantify the magnitude or clinical importance of the improvement.
Time to Two-Category Improvement
The time to an improvement of two categories using the ordinal scale was also analyzed from Day 1 through Day 29 using the ITT population, a log-rank test, and a Cox proportional-hazards effect measure.
Two-category improvement
95% CI: 1.05–1.38 · P = 0.005
The HR of 1.20 means that the estimated instantaneous rate of reaching the specified two-category improvement was 20% higher in the remdesivir-plus-baricitinib group under the reported model.
The 95% CI of 1.05–1.38 indicates uncertainty around the estimated relative hazard. The P-value of 0.005 is evidence against the null hypothesis used for this comparison; it is not itself an effect-size measure.
Because this is another time-to-event endpoint, the estimate should be interpreted in the context of the Day 1 through Day 29 window and the underlying censoring and proportional-hazards assumptions.
Time to Discharge or NEWS of 2 or Less
The registry reports a further time-to-event endpoint: time to discharge or to a NEWS of 2 or less and maintained for 24 hours, whichever occurs first. The analysis was restricted to participants with a baseline NEWS score of 2 or greater and otherwise used the ITT population.
Discharge / NEWS endpoint
95% CI: 1.07–1.44 · P = 0.003
The HR of 1.24 corresponds to a 24% higher estimated instantaneous rate of reaching the specified discharge-or-NEWS endpoint with remdesivir plus baricitinib under the reported Cox model.
The 95% CI of 1.07–1.44 describes the statistical uncertainty around the relative hazard estimate. It does not describe the range of discharge times for individual participants.
The analysis population is important: the registry specifies that participants were included according to randomization but restricted to those with a baseline NEWS score of 2 or greater. Thus this estimate answers a narrower question than the overall time-to-recovery analysis.
11. Safety Results
The ClinicalTrials.gov record contains two secondary safety analyses. Both use the safety population, defined as participants with available data post baseline and analyzed as treated. The effect measure is risk difference.
| Safety endpoint | Risk difference | 95% CI | Analysis population |
|---|---|---|---|
| Grade 3 and 4 clinical and/or laboratory adverse events | -6 | -12 to 0 | Safety population; analyzed as treated |
| Serious adverse events | -5 | -10 to 0 | Safety population; analyzed as treated |
Serious adverse events by arm
| Treatment group | Participants with serious AEs | At risk |
|---|---|---|
| Remdesivir Plus Baricitinib | 88 | 507 |
| Remdesivir Plus Placebo | 109 | 509 |
The posted serious-adverse-event risk difference is -5 with a 95% CI of -10 to 0. A negative risk difference indicates fewer events in the first group when the risk difference is defined as intervention minus comparator. The registry data do not provide a separate p-value for this safety analysis.
Risk difference is an absolute rather than relative effect measure. A value of -5 indicates a five-percentage-point lower event proportion under the conventional intervention-minus-comparator definition.
The 95% CI of -10 to 0 expresses uncertainty around that absolute difference. Because the interval reaches 0, the posted interval includes the possibility of no absolute difference.
The safety population differs from the ITT population used for primary efficacy analysis. Safety was analyzed as treated among participants with available post-baseline data, so treatment exposure rather than randomized assignment is the organizing principle for this analysis.
12. Statistical Methods Explained
Why was a log-rank test used for time to recovery?
Time to recovery records not only whether recovery occurred, but also when it occurred. A log-rank test uses information across the observed event times and compares the time-to-event experience between groups. This is more informative for a recovery-time endpoint than a simple comparison of the percentage recovered at one arbitrary time point.
What does an HR of 1.15 mean for a recovery endpoint?
For this trial, the event is recovery. An HR of 1.15 therefore indicates a higher estimated instantaneous recovery rate in the remdesivir-plus-baricitinib group. This differs from the familiar interpretation of a hazard ratio for death, where an HR below 1 generally indicates a lower event rate. The direction of interpretation depends on what the event actually is.
Why is the hazard ratio not the same as a percentage of patients recovered?
A hazard ratio is a relative time-to-event measure. It summarizes the relative event rate under a statistical model over the analyzed period. It does not directly tell us the proportion of participants who recovered, the average number of days saved, or the outcome for any particular participant.
Why does the confidence interval matter?
The point estimate is only one representation of the treatment comparison. The 95% confidence interval communicates how precisely the effect was estimated within the specified statistical framework. For example, the primary HR of 1.15 has a 95% CI of 1.00–1.31, whereas the one-category improvement HR of 1.21 has a 95% CI of 1.06–1.39. The intervals provide information that the point estimates alone cannot.
Why is an odds ratio different from a risk difference?
An odds ratio compares odds, whereas a risk difference compares probabilities directly. An OR of 1.26 therefore cannot simply be described as a 26-percentage-point increase. To obtain an absolute risk difference from an odds ratio, the underlying event probability is needed.
Why does ITT matter in this trial?
The overall time-to-recovery analysis used the ITT population, which includes all randomized participants. Analyzing according to randomization helps preserve the comparability created by the randomization process and avoids redefining the treatment groups based on subsequent treatment behavior.
Why should subgroup hazard ratios not be compared casually?
Subgroup estimates can vary because of sampling variability, especially when groups are smaller than the overall trial population. The fact that one subgroup has an HR of 1.34 and another has an HR of 1.06 does not itself demonstrate that the treatment effect is different between those groups. A formal interaction analysis is ordinarily needed for that conclusion.
13. Confidence Intervals, P-values, and Effect Size
ACTT-2 provides a useful illustration of why three related but distinct statistical quantities should be read together: the effect estimate, the confidence interval, and the p-value.
Effect estimate
The hazard ratio, odds ratio, or risk difference describes the estimated direction and magnitude of the treatment comparison.
Confidence interval
The 95% CI describes uncertainty around the estimated effect under the specified statistical framework.
P-value
The p-value measures compatibility of the observed data with a specified null hypothesis; it is not a measure of effect magnitude.
Clinical meaning
Statistical evidence does not by itself determine clinical importance. Relative effects should be considered alongside the endpoint definition, time frame, absolute outcomes, and safety results.
14. Interpreting the Primary Time-to-Recovery Result
The primary result is a useful example of interpreting a time-to-event analysis without overextending what the statistic can say.
The reported HR of 1.15 indicates a 15% higher estimated instantaneous recovery rate for remdesivir plus baricitinib relative to remdesivir plus placebo under the reported Cox model.
The 95% CI of 1.00–1.31 places the reported estimate within an interval that extends from the conventional no-difference value to a higher estimated recovery hazard.
The reported two-sided P-value is 0.047. This is a statement about the statistical test, not a probability that the treatment effect is real and not a direct measure of how large or clinically meaningful the effect is.
Because recovery is the event, an HR above 1 has a favorable directional interpretation for time to recovery within the statistical model. That interpretation should not be transferred automatically to other endpoints, where the event may have a different meaning.
15. Multiplicity and Multiple Primary Endpoints
The registry identifies 4 primary endpoints: overall time to recovery, time to recovery by race, time to recovery by ethnicity, and time to recovery by sex. The ClinicalTrials.gov record does not report a multiplicity-adjustment procedure or an alpha-allocation strategy for these four registered endpoints.
| Endpoint family | Registered role | Posted result | Multiplicity information reported |
|---|---|---|---|
| Time to Recovery | Primary | HR 1.15; 95% CI 1.00–1.31; P = 0.047 | Not reported |
| Time to Recovery by Race | Primary | Four race-specific HR estimates | Not reported |
| Time to Recovery by Ethnicity | Primary | Two ethnicity-specific HR estimates | Not reported |
| Time to Recovery by Sex | Primary | Two sex-specific HR estimates | Not reported |
16. Stratification, Crossover, and Interim Analysis
Stratification
The ClinicalTrials.gov record does not report randomization stratification factors or a stratified analysis procedure for ACTT-2. The primary analysis is identified as a log-rank test with a Cox proportional-hazards effect measure, but no stratification variables are provided.
Crossover
The ClinicalTrials.gov record does not report a crossover procedure. No crossover-adjusted treatment effect should therefore be inferred from the posted results.
Interim analysis
The ClinicalTrials.gov record does not report an interim-analysis method, alpha-spending procedure, stopping boundary, or information fraction. No such design feature is inferred here.
Bayesian methods
No Bayesian method is identified in the registry-reported normalized statistical-method fields. The reported methods are log-rank testing and logistic regression, with hazard ratio, odds ratio, and risk difference as effect measures.
17. Missing Data and Censoring
Time-to-event analyses require rules for observations that do not experience the event during the analysis period or whose follow-up ends before the event. The ClinicalTrials.gov record identifies the ITT populations and the Day 1 through Day 29 windows, but they do not provide the detailed censoring rules, missing-data strategy, or imputation procedure.
This distinction is important because a hazard ratio cannot be interpreted independently of how incomplete follow-up is handled. A statistical analysis page should not invent a missing-data method simply because one would normally be expected in a clinical trial.
18. Comparing the Main Statistical Effect Measures
| Endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Time to Recovery | Hazard ratio | 1.15 | 1.00–1.31 | 0.047 |
| Day 15 ordinal-scale clinical status | Odds ratio | 1.26 | 1.01–1.57 | 0.44 |
| Time to one-category improvement | Hazard ratio | 1.21 | 1.06–1.39 | 0.002 |
| Time to two-category improvement | Hazard ratio | 1.20 | 1.05–1.38 | 0.005 |
| Time to discharge or NEWS endpoint | Hazard ratio | 1.24 | 1.07–1.44 | 0.003 |
| Grade 3 and 4 AEs | Risk difference | -6 | -12 to 0 | Not reported |
| Serious AEs | Risk difference | -5 | -10 to 0 | Not reported |
These effect measures should not be placed on a single numerical scale. A hazard ratio describes a time-to-event comparison, an odds ratio describes relative odds, and a risk difference describes an absolute difference in event proportions. Their numerical values therefore answer different statistical questions.
19. Limitations
- Registry-only scope: this analysis is limited to the ClinicalTrials.gov record and does not add results from external publications.
- Incomplete design detail: the ClinicalTrials.gov record does not report randomization strata, detailed censoring rules, missing-data methods, crossover procedures, interim-analysis boundaries, or Bayesian methods.
- Multiple primary endpoints: four primary endpoints are registered, but the ClinicalTrials.gov record does not specify a multiplicity-adjustment strategy.
- Subgroup uncertainty: race, ethnicity, and sex estimates are subgroup analyses and generally carry more sampling uncertainty than the overall randomized comparison.
- Interaction testing: differences between subgroup point estimates do not establish treatment-effect heterogeneity without a formal interaction analysis.
- Cox-model assumptions: interpretation of the hazard ratio relies on the assumptions underlying the reported proportional-hazards effect measure. The ClinicalTrials.gov record does not provide an assessment of those assumptions.
- Safety population differs from ITT: the safety analyses use participants with available post-baseline data and analyze them as treated, whereas the primary efficacy analysis uses randomized participants.
- Day 15 result: the registry fields reported here report an odds ratio of 1.26 with a 95% CI of 1.01–1.57 and a P-value of 0.44. Those values are reproduced as posted rather than recalculated.
- No participant-level reconstruction: the posted summary statistics do not permit reconstruction of the full recovery-time distribution or an independent Kaplan-Meier curve.
20. Why This Trial Matters Statistically
ACTT-2 is a useful statistical teaching case because its registry results combine several major clinical-trial concepts within one randomized comparison. The primary endpoint is a time-to-event outcome, while the posted secondary analyses extend the analysis to ordinal clinical status and safety outcomes.
| Concept | How it appears in ACTT-2 |
|---|---|
| Randomization | Participants were randomized in a parallel phase 3 design. |
| Blinding | The registry reports double masking. |
| ITT analysis | The primary time-to-recovery analysis includes all randomized participants. |
| Time-to-event analysis | Time to recovery and several secondary clinical outcomes are analyzed as event times. |
| Log-rank test | Used for the overall time-to-recovery analysis and three secondary time-to-event endpoints. |
| Hazard ratio | Used as the effect measure for the recovery and other time-to-event analyses. |
| Logistic regression | Used for the posted Day 15 ordinal-scale analysis. |
| Odds ratio | Reported for the Day 15 logistic-regression analysis. |
| Risk difference | Used for the two posted safety analyses. |
| Subgroup analysis | Primary endpoint results are reported by race, ethnicity, and sex. |
| Safety population | Safety endpoints use participants with available post-baseline data, analyzed as treated. |
| Multiple endpoints | Four primary endpoints are registered, while the registry-reported multiplicity strategy is not reported. |
21. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
22. Related Statistical Calculators
23. Sources
- ClinicalTrials.gov: ACTT-2 (NCT04401579).
- PubMed: PMID 38618926.
- PubMed: PMID 38408357.
- PubMed: PMID 36442063.
- PubMed: PMID 35695334.
- PubMed: PMID 34473343.
Continue through the Clinical Biostats statistical pathway
Explore the statistical concepts behind randomized trials, time-to-event endpoints, regression models, confidence intervals, and clinical safety analyses.
24. Record Summary
ACTT-2 provides a compact example of how a randomized phase 3 clinical trial can generate several different statistical questions from the same treatment comparison. Its primary time-to-recovery analysis used an ITT population, a log-rank test, and a Cox proportional-hazards effect measure, producing a reported HR of 1.15 with a 95% CI of 1.00–1.31 and P = 0.047. Additional primary analyses reported recovery hazard ratios by race, ethnicity, and sex.
The secondary analyses illustrate why endpoint definition determines statistical interpretation. Time to ordinal-scale improvement and time to the discharge-or-NEWS endpoint were analyzed with survival methods, while the Day 15 clinical-status analysis used logistic regression and an odds ratio. Safety outcomes used risk differences in an as-treated safety population.