This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
COV-BARRIER was a randomized, double-blind, phase 3, parallel-group trial evaluating baricitinib plus standard of care against placebo plus standard of care in participants with COVID-19. The registry reports 1,525 enrolled participants, two study arms, two registered primary endpoints, and formal statistical analyses for both primary endpoints.
| Feature | COV-BARRIER |
|---|---|
| Trial name | COV-BARRIER |
| Clinical condition | COVID-19 |
| Phase | Phase 3 |
| Design | Randomized, double-blind, parallel-group |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 1,525 |
| Interventions | Baricitinib and placebo |
| Primary endpoints | 2 binary endpoints, both assessed from Day 1 to Day 28 |
| Results | Posted, including 23 statistical analyses |
| Lead sponsor | Eli Lilly and Company |
| ClinicalTrials.gov | NCT04421027 |
2. Clinical Question
The principal question was whether adding baricitinib to standard of care changed the probability of death or the need for escalating respiratory support during the first 28 days compared with placebo plus standard of care in participants with COVID-19.
Population
Participants with COVID-19 enrolled in a randomized phase 3 treatment trial. The registry reports 1,525 participants enrolled.
Intervention
4 Milligrams (mg) Baricitinib plus standard of care.
Comparator
Placebo plus standard of care.
Primary question
Does baricitinib plus standard of care alter the Day 1 to Day 28 occurrence of death or clinically important ventilatory support requirements?
3. Trial Design
4 Milligrams (mg) Baricitinib + SOC
- Baricitinib: 4 Milligrams (mg)
- Standard of care
- Randomized, double-blind treatment assignment
- Registry serious adverse-event count: 110/750
- Follow-up serious adverse-event count: 18/645
Placebo + SOC
- Placebo
- Standard of care
- Randomized, double-blind treatment assignment
- Registry serious adverse-event count: 135/752
- Follow-up serious adverse-event count: 12/613
4. Endpoints
The registry contains two primary binary endpoints and a broader set of secondary endpoints spanning ordinal clinical status, ventilator-free days, time to recovery, hospitalization, mortality, oxygen-related outcomes, respiratory measures, and clinical deterioration.
| Endpoint role | Endpoint | Time frame | Type |
|---|---|---|---|
| Primary | Percentage of Participants Who Die or Require Non-Invasive Ventilation/High-Flow Oxygen or Invasive Mechanical Ventilation (Including ECMO) | Day 1 to Day 28 | Binary |
| Primary | Percentage of Participants Who Die or Require Non-Invasive Ventilation/High-Flow Oxygen or Invasive Mechanical Ventilation (Including ECMO) Population 2 | Day 1 to Day 28 | Binary |
| Secondary | Percentage of Participants With at Least 1-Point Improvement on NIAID-OS or Live Discharge From Hospital | Day 10 | Binary |
| Secondary | Number of Ventilator-Free Days | Day 1 to Day 28 | Continuous |
| Secondary | Time to Recovery | Day 1 to Day 28 | Time-to-event |
| Secondary | Percentage of Participants at Each Clinical Status Using NIAID-OS | Day 4, Day 7, Day 10, Day 14 | Ordinal |
| Secondary | Duration of Hospitalization | Days 1 to Day 28 | Continuous |
| Secondary | Overall Mortality | Day 1 to Day 28 | Time-to-event |
| Secondary | Time to Clinical Deterioration (One-category Increase on the NIAID-OS) | Day 1 to Day 28 | Time-to-event |
| Secondary | Time to Resolution of Fever in Participants With Fever at Baseline | Day 1 to Day 28 | Time-to-event |
| Secondary | Mean Change From Baseline on the NEWS | Baseline through Day 14 | Continuous / longitudinal |
Primary endpoint populations
The first primary analysis included all participants randomly assigned to study intervention, excluding participants with missing baseline ordinal scale values. The second primary endpoint used a narrower population: all randomly assigned participants who at baseline required oxygen supplementation and did not receive dexamethasone or other systemic corticosteroids for the primary study condition.
5. Statistical Methodology
Logistic regression for binary outcomes
The two primary endpoints were analyzed using logistic regression. The resulting effect measure was an odds ratio comparing baricitinib plus standard of care with placebo plus standard of care.
An odds ratio below 1 indicates lower estimated odds of the binary endpoint in the baricitinib group; an odds ratio above 1 indicates higher estimated odds. The odds ratio is not itself a risk ratio or an absolute percentage-point difference.
ANOVA for continuous outcomes
ANOVA was used for outcomes such as ventilator-free days, duration of hospitalization, duration of ICU stay, days with supplemental oxygen use, and days with resting respiratory rate below 24 breaths per minute. The reported effect measure was a least-squares mean difference or net mean difference.
Log-rank testing for time-to-event endpoints
Time to recovery, overall mortality, time to clinical deterioration, time to resolution of fever, time to independence from non-invasive mechanical ventilation, and time to independence from oxygen therapy were analyzed using log-rank methods. Hazard ratios were reported for these endpoints when estimates were available.
Mixed-effects models for NEWS
Mean change from baseline on the National Early Warning Score was evaluated at multiple postbaseline time points using a mixed-effects model. The registry reports least-squares mean differences at Day 4, Day 7, Day 10, and Day 14.
Proportional odds regression for NIAID-OS
Clinical status on the NIAID-OS was analyzed using a proportional odds model at Day 4, Day 7, Day 10, and Day 14. The registry reports odds ratios with two-sided 95% confidence intervals.
6. Primary Results
The registry reports formal analyses for both registered primary endpoints. Both were superiority analyses comparing placebo plus standard of care with baricitinib plus standard of care.
Primary Endpoint 1: Death or Escalation to Ventilatory Support
Odds ratio for the Day 1 to Day 28 binary endpoint
95% CI: 0.67–1.08 · P = 0.1800
Logistic regression; all randomly assigned participants, excluding participants with missing baseline ordinal scale values.
The estimated odds ratio of 0.85 means that the estimated odds of death or requiring non-invasive ventilation/high-flow oxygen or invasive mechanical ventilation, including ECMO, were 15% lower with baricitinib plus standard of care than with placebo plus standard of care under the fitted logistic model.
This does not mean that the probability of the endpoint was exactly 15% lower, because an odds ratio is not a risk ratio. It also does not describe an individual patient's probability of deterioration.
The two-sided 95% confidence interval of 0.67–1.08 spans 1.00. Thus, the estimate is compatible with a range extending from lower odds to slightly higher odds under the statistical model. The interval describes uncertainty around the estimated odds ratio; it does not describe the range of individual treatment effects.
The P = 0.1800 value is a measure of compatibility with the null hypothesis under the specified model and does not measure the magnitude or clinical importance of the estimated effect. The analysis was specified as a superiority test.
Primary Endpoint 2: Oxygen-Supplemented Population Without Baseline Corticosteroids
Odds ratio for the Day 1 to Day 28 binary endpoint
95% CI: 0.58–2.16 · P = 0.7280
Logistic regression; participants requiring oxygen supplementation at baseline who did not receive dexamethasone or other systemic corticosteroids for the primary study condition.
The estimated odds ratio of 1.12 corresponds to estimated odds that were 12% higher in the baricitinib-plus-standard-of-care group under the fitted model. Because this is an odds ratio, it should not be translated directly into a 12% higher probability.
The 95% confidence interval of 0.58–2.16 is substantially wider than the interval for the first primary endpoint. It spans values below and above 1.00, indicating considerable statistical uncertainty about the direction and magnitude of the estimated association in this population.
The P = 0.7280 value does not quantify the size of the treatment effect. It describes the statistical evidence against the null hypothesis under the specified logistic-regression analysis. It should be interpreted alongside the point estimate and confidence interval rather than as a standalone measure of clinical importance.
7. Secondary Efficacy Results
The registry contains a broad set of secondary analyses. These results illustrate why the statistical interpretation of a trial should consider the endpoint definition, effect measure, analysis population, and confidence interval rather than treating every P-value as an equivalent piece of evidence.
Ordinal Clinical Status
| Endpoint | Method | OR | 95% CI | P-value |
|---|---|---|---|---|
| NIAID-OS at Day 4 | Proportional odds regression | 1.21 | 1.00–1.47 | 0.0464 |
| NIAID-OS at Day 7 | Proportional odds regression | 1.25 | 1.04–1.49 | 0.0172 |
| NIAID-OS at Day 10 | Proportional odds regression | 1.17 | 0.97–1.41 | 0.0921 |
| NIAID-OS at Day 14 | Proportional odds regression | 1.28 | 1.05–1.56 | 0.0168 |
For a proportional odds model, an odds ratio above 1 indicates higher modeled odds of being at a more favorable position across the ordinal outcome categories, under the proportional-odds assumption. The registry labels the endpoint as percentage of participants at each clinical status, while the posted analysis uses an ordinal regression framework.
Time-to-Event Outcomes
| Endpoint | HR | 95% CI | P-value |
|---|---|---|---|
| Time to Recovery | 1.11 | 0.99–1.24 | 0.1453 |
| Overall Mortality | 0.57 | 0.41–0.78 | 0.0018 |
| Time to Clinical Deterioration | 0.887 | 0.741–1.061 | 0.1831 |
| Time to Resolution of Fever | 1.202 | 1.017–1.421 | 0.0243 |
| Time to Independence From Non-Invasive Mechanical Ventilation | 1.090 | Not reported | 0.6436 |
| Time to Independence From Oxygen Therapy | 1.124 | Not reported | 0.1270 |
Continuous and Longitudinal Outcomes
| Endpoint | Difference | 95% CI | P-value |
|---|---|---|---|
| Ventilator-free days | 0.75 days | -0.0 to 1.5 | 0.0586 |
| Duration of hospitalization | -0.76 days | -1.6 to 0.0 | 0.0626 |
| Duration of ICU stay | 0.02 days | -0.62 to 0.65 | 0.9526 |
| Days with supplemental oxygen use | -0.23 days | -0.68 to 0.21 | 0.3058 |
| Days with resting respiratory rate <24 breaths/minute | 0.11 days | -0.49 to 0.72 | 0.7073 |
The registry also reports the percentage of participants whose oxygen saturation changed from <94% to ≥94% from baseline at Day 10. The logistic-regression estimate was an odds ratio of 1.15 (95% CI 0.81–1.63; P = 0.4290).
8. Overall Mortality: A Key Time-to-Event Analysis
Hazard ratio for overall mortality
95% CI: 0.41–0.78 · P = 0.0018
Day 1 to Day 28; log-rank analysis; all participants randomly assigned to study intervention.
The hazard ratio of 0.57 means that, under the time-to-event comparison, the estimated instantaneous hazard of death was approximately 43% lower with baricitinib plus standard of care than with placebo plus standard of care over the analyzed period.
The HR is a relative time-to-event measure. It does not mean that 43% fewer participants died, and it does not specify an absolute mortality reduction. Absolute event probabilities would be needed to answer those questions.
The 95% CI of 0.41–0.78 indicates relatively greater precision than several of the other time-to-event estimates reported in the registry and does not cross 1.00. The P = 0.0018 value describes the statistical evidence under the log-rank analysis; it is not a measure of effect magnitude.
As with other hazard-ratio analyses, interpretation depends on the time-to-event framework and the relationship between hazards over time. A single HR summarizes a potentially dynamic process and should not automatically be interpreted as a constant reduction in individual risk at every time point.
9. Recovery and Clinical Deterioration
Time to Recovery
Hazard ratio
95% CI: 0.99–1.24 · P = 0.1453
The time-to-recovery analysis produced an HR of 1.11. In a time-to-event framework where the event is recovery, an HR above 1 corresponds to a higher estimated instantaneous rate of recovery in the baricitinib group. The confidence interval of 0.99–1.24 includes 1.00, and the P-value was 0.1453.
Time to Clinical Deterioration
Hazard ratio for one-category increase on the NIAID-OS
95% CI: 0.741–1.061 · P = 0.1831
Here the event is clinical deterioration rather than recovery. The HR of 0.887 corresponds to an estimated lower instantaneous hazard of the deterioration event in the baricitinib group, but the 95% CI includes 1.00. The direction of an HR must always be interpreted in relation to the event being analyzed.
Time to Resolution of Fever
Hazard ratio among participants with fever at baseline
95% CI: 1.017–1.421 · P = 0.0243
For this recovery-type endpoint, the HR of 1.202 corresponds to an estimated 20.2% higher instantaneous rate of fever resolution in the baricitinib group under the model. The confidence interval lies above 1.00, although the endpoint is one of many secondary analyses and therefore should be interpreted in the broader context of the trial's multiple outcomes.
10. Ventilator-Free Days and Hospitalization
| Outcome | Analysis | LS Mean Difference | 95% CI | P-value |
|---|---|---|---|---|
| Number of Ventilator-Free Days | ANOVA | 0.75 | -0.0 to 1.5 | 0.0586 |
| Duration of Hospitalization | ANOVA | -0.76 | -1.6 to 0.0 | 0.0626 |
| Duration of ICU Stay | ANOVA | 0.02 | -0.62 to 0.65 | 0.9526 |
The ventilator-free-days estimate was 0.75 days, with a 95% confidence interval from -0.0 to 1.5 and P = 0.0586. Because the confidence interval includes zero, the estimated mean difference is compatible with no difference under the reported model.
For duration of hospitalization, the least-squares mean difference was -0.76 days, with a 95% confidence interval from -1.6 to 0.0 and P = 0.0626. The negative direction corresponds to fewer days in the baricitinib group under the reported comparison, but the interval reaches zero.
The ICU-stay estimate was 0.02 days with a 95% confidence interval from -0.62 to 0.65 and P = 0.9526. The estimate is close to zero relative to the uncertainty reported by the registry.
11. National Early Warning Score: Longitudinal Analysis
The National Early Warning Score was assessed repeatedly from baseline through Day 14. The registry used a mixed-effects model and reported least-squares mean differences at each assessment time.
| Assessment | LS Mean Difference | 95% CI | P-value |
|---|---|---|---|
| Day 4 | -0.17 | -0.42 to 0.08 | 0.191 |
| Day 7 | -0.17 | -0.49 to 0.14 | 0.278 |
| Day 10 | -0.13 | -0.49 to 0.24 | 0.496 |
| Day 14 | -0.25 | -0.70 to 0.19 | 0.263 |
The negative point estimates consistently favor lower NEWS values in the baricitinib group, but each reported confidence interval includes zero. The estimates also vary in precision across the four time points.
Repeated measurements from the same participant are correlated: a participant's Day 4 score is not statistically independent of that participant's Day 7 score. A mixed-effects model provides a framework for analyzing longitudinal observations while accounting for within-participant dependence and estimating treatment differences over repeated assessments.
12. Statistical Methods Explained
Why was logistic regression used for the primary endpoints?
The two primary outcomes were binary: participants either experienced the defined composite event during Day 1 to Day 28 or did not. Logistic regression is designed for binary outcomes and models the odds of the event as a function of treatment and any included covariates. The registry reports the treatment effect as an odds ratio.
What does an odds ratio of 0.85 mean?
An odds ratio of 0.85 means that the estimated odds of the endpoint were 15% lower in the baricitinib group than in the placebo group under the fitted model. It does not mean that the event probability was 15% lower. The distinction becomes particularly important when the outcome is not rare.
Why was a proportional odds model used for NIAID-OS?
The NIAID-OS represents ordered clinical states. Proportional odds regression uses that ordering rather than treating every category as an unrelated binary outcome. The model estimates a common odds ratio across cumulative thresholds, subject to the proportional-odds assumption.
Why were log-rank methods used for mortality and recovery?
Mortality, recovery, deterioration, fever resolution, and independence from respiratory support are time-to-event outcomes. Some participants may not experience the event during the observation period, so their observations are censored rather than treated as if the event never occurred. The log-rank test compares the event-time distributions between randomized groups, while the hazard ratio summarizes their relative event rates.
What does an HR of 0.57 mean for mortality?
An HR of 0.57 indicates an estimated 43% lower instantaneous hazard of death in the baricitinib group relative to the placebo group over the analyzed period. It is not equivalent to a 43% reduction in the percentage of participants who died and does not directly provide an absolute mortality difference.
Why was a mixed-effects model used for NEWS?
NEWS was measured repeatedly for participants at multiple time points. A mixed-effects model accounts for the correlation among repeated observations within the same participant and provides treatment-group comparisons across the longitudinal assessment schedule.
13. Understanding Confidence Intervals in This Trial
Confidence intervals are particularly informative because the registry contains effects that point in different directions and vary substantially in precision.
| Endpoint | Estimate | 95% CI | What the interval indicates |
|---|---|---|---|
| Primary endpoint 1 | OR 0.85 | 0.67–1.08 | Includes both lower and slightly higher odds |
| Primary endpoint 2 | OR 1.12 | 0.58–2.16 | Wide uncertainty in both directions |
| Overall mortality | HR 0.57 | 0.41–0.78 | Does not include 1.00 |
| Time to recovery | HR 1.11 | 0.99–1.24 | Close to the null value of 1.00 |
| Ventilator-free days | 0.75 | -0.0 to 1.5 | Includes zero mean difference |
| NEWS Day 14 | -0.25 | -0.70 to 0.19 | Includes zero mean difference |
For odds ratios and hazard ratios, the null value is 1.00. For mean differences, the null value is 0. These different reference values are essential when reading the reported confidence intervals.
14. P-values and Multiple Secondary Endpoints
The registry reports 23 statistical analyses across the trial, including two primary endpoint analyses and numerous secondary analyses. Several secondary P-values are below 0.05, including the analyses of NIAID-OS at Day 4, Day 7, and Day 14, overall mortality, and time to resolution of fever.
A P-value is not a probability that the null hypothesis is true, nor is it a measure of the size or clinical importance of an effect. For example, an odds ratio of 1.25 and an odds ratio of 1.25 can have very different evidentiary interpretations depending on their confidence intervals, sample size, endpoint definition, and analysis plan.
15. Analysis Populations and Missing Data
The registry uses several analysis populations depending on the endpoint. This matters because the denominator and eligibility criteria can change the interpretation of an effect estimate.
| Analysis population feature | Examples in the registry |
|---|---|
| All randomized participants | Primary endpoint 1, time to recovery, overall mortality |
| Randomized participants excluding missing baseline ordinal scale | Primary endpoint 1 and several ordinal and continuous analyses |
| Baseline oxygen supplementation and no baseline corticosteroids for primary condition | Primary endpoint 2 |
| Baseline fever | Time to resolution of fever |
| Baseline NIAID-OS of 6 and non-invasive mechanical ventilation | Time to independence from non-invasive mechanical ventilation |
| Baseline ordinal scale score of 5 or 6 | Time to independence from oxygen therapy |
| Baseline and at least one postbaseline observation | NEWS and several continuous outcomes |
Several analyses explicitly exclude participants with missing baseline measures or require postbaseline observations. Such restrictions can change the population being analyzed and should therefore be read alongside each effect estimate rather than assuming that every result uses exactly the same denominator.
16. Secondary Endpoint: Oxygen Independence and Respiratory Support
| Endpoint | HR / Difference | 95% CI | P-value |
|---|---|---|---|
| Time to independence from non-invasive mechanical ventilation | HR 1.090 | Not reported | 0.6436 |
| Time to independence from oxygen therapy | HR 1.124 | Not reported | 0.1270 |
| Number of days with supplemental oxygen use | -0.23 days | -0.68 to 0.21 | 0.3058 |
| Number of days with resting respiratory rate <24 breaths/minute | 0.11 days | -0.49 to 0.72 | 0.7073 |
These endpoints illustrate the importance of matching the direction of the effect measure to the event. For an independence or resolution event, an HR above 1 represents a higher instantaneous rate of reaching that favorable event. For a deterioration event, an HR below 1 represents a lower instantaneous rate of deterioration.
17. Safety Results
The registry reports serious adverse events by randomized treatment group and also provides separate follow-up counts.
| Safety population / period | Placebo | 4 Milligrams (mg) Baricitinib |
|---|---|---|
| Serious adverse events | 135/752 | 110/750 |
| Follow-up serious adverse events | 12/613 | 18/645 |
The reported figures are counts affected over the corresponding populations at risk. They should not be conflated with incidence rates, exposure-adjusted rates, or the probability of experiencing any adverse event unless the corresponding denominator and follow-up definition are specified.
The randomized treatment groups provide a direct comparison of serious adverse-event counts within the populations reported by the registry. The separate follow-up figures use different at-risk denominators and therefore represent a different analysis context. Comparing the four fractions as though they were interchangeable measures would obscure that distinction.
18. What the Major Effect Measures Mean
Odds ratio
The odds ratio compares the odds of a binary outcome between treatment groups. An OR of 1 indicates equal odds. An OR below 1 indicates lower odds in the numerator treatment group; an OR above 1 indicates higher odds.
where pT and pC represent event probabilities in the treatment and comparator groups, respectively.
Hazard ratio
A hazard ratio compares instantaneous event rates over time. For an endpoint such as mortality, an HR below 1 indicates a lower estimated hazard in the treatment group. For a recovery endpoint, an HR above 1 indicates a higher estimated recovery hazard.
Mean difference
A mean difference compares average outcome values between groups. In the registry's ANOVA analyses, the reported quantity is a least-squares mean difference or net mean difference. The null value is zero.
19. Statistical Interpretation of the Primary and Key Secondary Findings
Primary endpoint 1
The estimated OR was 0.85 with a 95% CI of 0.67–1.08 and P = 0.1800. The confidence interval includes the null value of 1.00.
Primary endpoint 2
The estimated OR was 1.12 with a 95% CI of 0.58–2.16 and P = 0.7280. The interval is wide and includes 1.00.
Overall mortality
The mortality HR was 0.57 with a 95% CI of 0.41–0.78 and P = 0.0018. The interval does not include 1.00.
Time to recovery
The recovery HR was 1.11 with a 95% CI of 0.99–1.24 and P = 0.1453. The interval is close to but includes 1.00.
The combination of endpoints demonstrates why a trial cannot be summarized adequately by one number. The two registered primary endpoints are binary composite outcomes with logistic-regression odds ratios, whereas mortality is a time-to-event endpoint analyzed using a log-rank framework and expressed as a hazard ratio. The estimands, assumptions, and interpretation differ.
20. Limitations
- Composite primary endpoints: each primary endpoint combines death with escalating respiratory-support requirements. A composite can capture clinically important events efficiently, but the components may differ in clinical meaning and frequency.
- Different analysis populations: several secondary endpoints require specific baseline characteristics or non-missing follow-up measurements, so their populations are not necessarily identical.
- Multiple endpoints: the registry contains 23 posted statistical analyses. Secondary P-values therefore require attention to multiplicity and the broader analysis plan.
- Odds ratios are not risk ratios: the two primary estimates of 0.85 and 1.12 should not be translated directly into percentage changes in event probability.
- Hazard-ratio assumptions: a single HR summarizes a time-to-event comparison and is most naturally interpreted within the model used to estimate it. It should not automatically be interpreted as a constant relative risk at every time.
- Ordinal-model assumptions: proportional odds regression relies on a common odds-ratio structure across cumulative ordinal thresholds. The registry identifies the method but does not provide the full diagnostic assessment of that assumption.
- Longitudinal correlation: repeated NEWS measurements within a participant are correlated, which is why a mixed-effects framework is appropriate. The interpretation depends on the model specification.
- Safety denominators: serious adverse-event counts and follow-up counts use different at-risk populations and should not be treated as the same estimand.
21. Why This Trial Matters Statistically
COV-BARRIER is a useful statistical teaching case because its registry results span several major families of clinical-trial methodology within one randomized study.
| Concept | How it appears in COV-BARRIER |
|---|---|
| Randomization | Randomized allocation to baricitinib plus standard of care or placebo plus standard of care |
| Blinding | Double-blind design |
| Binary regression | Logistic regression for both primary endpoints and oxygen-saturation improvement |
| Odds ratio | Primary endpoint and ordinal analyses |
| Ordinal regression | Proportional odds analyses of NIAID-OS at Days 4, 7, 10, and 14 |
| Time-to-event analysis | Recovery, mortality, deterioration, fever resolution, and independence endpoints |
| Log-rank test | Formal comparison of time-to-event outcomes |
| Hazard ratio | Reported for mortality, recovery, deterioration, fever resolution, and respiratory independence |
| ANOVA | Ventilator-free days, hospitalization, ICU stay, oxygen use, and respiratory-rate outcomes |
| Mixed-effects model | Repeated NEWS assessments from baseline through Day 14 |
| Confidence intervals | Precision estimates for the reported treatment effects |
The statistical lesson is that endpoint type should drive the analysis method. Binary outcomes naturally lead to logistic regression; ordered categories can be analyzed with proportional odds regression; repeated continuous observations can require mixed-effects models; and time-to-event endpoints require methods that account for censoring and event timing.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: NCT04421027 — COV-BARRIER.
- Publication: PubMed record 35695334.
- Publication: PubMed record 35123660.
- Publication: PubMed record 34480861.
Continue through Clinical Biostats
Connect the statistical methods used in COV-BARRIER with deeper tutorials and practical statistical tools.
25. Record Summary
COV-BARRIER provides a broad example of clinical-trial statistical analysis across binary, ordinal, continuous, longitudinal, and time-to-event outcomes. The two registered primary endpoints used logistic regression and produced odds ratios of 0.85 and 1.12, respectively. Among the secondary analyses, overall mortality was evaluated with a log-rank method and produced an HR of 0.57 (95% CI 0.41–0.78; P = 0.0018), while time to recovery produced an HR of 1.11 (95% CI 0.99–1.24; P = 0.1453).
The trial is particularly useful for understanding that statistical interpretation depends on the estimand and analysis method. Odds ratios describe binary outcomes, hazard ratios describe relative event rates over time, mean differences describe continuous outcomes, proportional odds models use ordinal information, and mixed-effects models address repeated measurements. Confidence intervals then provide the essential measure of uncertainty around each reported estimate.