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
PREVENT-HD was a randomized, double-blind, parallel-group phase 3 trial evaluating rivaroxaban against placebo in medically ill outpatients with acute, symptomatic COVID-19 infection. The registered primary efficacy and safety outcomes were both time-to-event endpoints assessed through Day 35.
| Feature | PREVENT-HD |
|---|---|
| Phase | Phase 3 |
| Condition | Coronavirus Disease 2019 (COVID-19) |
| Design | Randomized, double-blind, parallel-group |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 1,284 |
| Primary endpoints | 2 time-to-event endpoints |
| Trial status | Terminated |
| Study start | August 13, 2020 |
| Primary completion | June 1, 2022 |
| ClinicalTrials.gov | NCT04508023 |
| Lead sponsor | Janssen Research & Development, LLC |
2. Clinical Question
The central question was whether rivaroxaban could reduce the time to first occurrence of a composite of major thrombotic, hospitalization, and mortality events compared with placebo in medically ill outpatients with acute, symptomatic COVID-19 infection.
Population
Medically ill outpatients with acute, symptomatic Coronavirus Disease 2019 (COVID-19) infection.
Intervention
Rivaroxaban 10 milligrams (mg).
Comparator
Placebo, with standard of care represented among the registered interventions.
Primary question
Does rivaroxaban change the time to first occurrence of the primary efficacy composite endpoint through Day 35?
3. Trial Design
Rivaroxaban
- Rivaroxaban 10 milligrams (mg)
- Randomized treatment assignment
- Double-blind trial design
Placebo
- Placebo comparator
- Randomized treatment assignment
- Double-blind trial design
The registry describes the study as randomized, double-blind, and parallel-group. The primary efficacy analysis used the intention-to-treat population, preserving treatment assignment as the basis for the comparison.
4. Endpoints
| Endpoint | Registry definition | Time frame | Endpoint type |
|---|---|---|---|
| Primary efficacy composite | Number of participants with time to first occurrence of symptomatic venous thromboembolism (VTE), myocardial infarction (MI), ischemic stroke, acute limb ischemia, non-CNS systemic embolization, all-cause hospitalization, or all-cause mortality. | Up to Day 35 | Time-to-event |
| Principle safety outcome | Number of participants with time to first occurrence of fatal bleeding or critical site bleeding based on a modification of the International Society on Thrombosis and Haemostasis (ISTH) criteria. | Up to Day 35 | Time-to-event |
Primary efficacy endpoint
The primary efficacy endpoint was defined as the time from randomization to the first occurrence of any component of the composite. The components included symptomatic VTE, MI, ischemic stroke, acute limb ischemia, non-CNS systemic embolization, all-cause hospitalization, and all-cause mortality.
Principle safety outcome
The principle safety outcome was the time to first occurrence of fatal bleeding or critical site bleeding. Critical sites listed in the registry definition include intracranial, intra-spinal, intraocular, pericardial, intra-articular, intra-muscular with compartment syndrome, and retroperitoneal bleeding.
Both primary endpoints are therefore time-to-event outcomes. This is statistically important because the analysis must account not only for whether an event occurred, but also for when it occurred and for participants whose observation ended before an event was observed.
5. Statistical Methodology
Intention-to-treat analysis
The posted primary efficacy analysis used the Intent-To-Treat (ITT) analysis set, defined in the registry as including all randomized participants who provided informed consent. Participants were compared according to randomized treatment assignment rather than according to treatment actually received.
Stratified log-rank test
The primary efficacy comparison used a log-rank test stratified by the time from COVID-19 positive test to randomization. The log-rank test compares the observed pattern of event occurrence between treatment groups over follow-up rather than comparing only a single event proportion at Day 35.
The hazard ratio was reported from a Cox proportional-hazards analysis, while the treatment-group comparison was specified as a stratified log-rank test.
Cox proportional-hazards effect measure
The effect measure reported for the primary efficacy endpoint was a Cox proportional-hazards estimate expressed as a hazard ratio. The hazard ratio compares the estimated instantaneous event rates between the randomized groups over the analyzed follow-up.
HR > 1 → higher estimated instantaneous event rate in the rivaroxaban group
HR < 1 → lower estimated instantaneous event rate in the rivaroxaban group
The hazard ratio is a relative time-to-event measure. It is not a risk ratio, an absolute risk difference, or a statement about the proportion of participants who benefit.
Stratification
The posted analysis specifies stratification by the time from COVID-19 positive test to randomization. Stratification allows the comparison to account for this timing variable without requiring a single unstratified treatment comparison to ignore it.
Superiority hypothesis
The registered hypothesis type for the posted primary analysis was superiority. The inferential question was therefore whether the randomized treatment groups differed in the time-to-event outcome, rather than whether rivaroxaban met a non-inferiority margin.
6. Primary Efficacy Result
The registry reports one formal statistical analysis for the primary efficacy composite endpoint. The analysis population was the ITT set, and rivaroxaban 10 mg was compared with placebo.
Time to first occurrence of the primary efficacy composite endpoint
95% CI: 0.63–2.15 · P = 0.626
Stratified log-rank test by time from COVID-19 positive test to randomization
| Analysis feature | Posted result |
|---|---|
| Endpoint | Time to first occurrence of the primary efficacy composite endpoint |
| Time frame | Up to Day 35 |
| Analysis population | Intent-To-Treat (ITT) |
| Comparison | Rivaroxaban 10 mg vs placebo |
| Statistical test | Stratified log-rank test |
| Effect measure | Cox proportional-hazards hazard ratio |
| Hazard ratio | 1.16 |
| 95% confidence interval | 0.63–2.15 |
| P-value | 0.626 |
| Hypothesis | Superiority |
The estimated hazard ratio of 1.16 means that the fitted analysis estimated a higher instantaneous rate of first occurrence of the primary efficacy composite in the rivaroxaban group than in the placebo group, by a relative factor of 1.16. Equivalently, the point estimate is 16% above 1. A hazard ratio above 1 should not be interpreted as meaning that 16% more participants necessarily experienced the composite.
The 95% confidence interval of 0.63–2.15 is wide and crosses 1. It indicates substantial uncertainty around the point estimate: the data are compatible with an estimated hazard below that of placebo as well as with an estimated hazard above it. The interval therefore provides important information about precision that cannot be conveyed by the point estimate alone.
The P-value of 0.626 is a measure of the statistical evidence against the null hypothesis under the specified testing framework. It is not a measure of effect size, clinical importance, or the probability that rivaroxaban is effective or ineffective.
Because the endpoint is time-to-event, interpretation also depends on censoring and on the assumptions underlying the proportional-hazards model used for the hazard ratio. A single hazard ratio summarizes the relative event-rate comparison through follow-up; it does not describe every individual's experience or establish a constant treatment effect at every time point.
7. How to Read the Primary Result
The most important statistical feature of the posted efficacy result is the combination of its point estimate, confidence interval, and P-value. These three quantities answer different questions.
The point estimate
HR 1.16 is the estimated relative event rate from the fitted time-to-event analysis. It is the single most compact summary of the estimated treatment effect, but it does not describe the uncertainty by itself.
The confidence interval
The 95% CI of 0.63–2.15 shows that the estimate is imprecise. It spans values below and above 1, so the data are compatible with both directions of relative treatment effect.
The P-value
P = 0.626 quantifies statistical evidence under the specified hypothesis-testing framework. It does not tell us the size of the treatment effect or the probability that the null hypothesis is true.
The endpoint
The composite includes thrombotic events, hospitalization, and death. Consequently, the hazard ratio applies to the composite endpoint as defined rather than to any single component in isolation.
The distinction between a composite endpoint and its individual components is particularly important. A composite combines several clinically different events into one time-to-first-event outcome. The posted analysis gives one hazard ratio for the composite; it does not provide separate estimates for each listed component.
8. Statistical Methods Explained
Why was a log-rank test used?
The primary efficacy endpoint records the time until the first occurrence of a composite event, making survival-analysis methods appropriate. The log-rank test compares the observed event-time experience between randomized groups while accommodating right-censored observations. In this trial, the log-rank comparison was stratified by the time from a positive COVID-19 test to randomization.
What does a hazard ratio of 1.16 mean?
A hazard ratio of 1.16 means that the estimated instantaneous rate of first occurrence of the primary efficacy composite was 1.16 times that in the placebo group under the fitted model. The value is a relative measure. It does not mean that 16% of participants experienced an event, nor does it mean that the absolute event probability was 16 percentage points higher.
Why does the confidence interval matter?
The 95% confidence interval of 0.63–2.15 shows how uncertain the estimated hazard ratio is. Because the interval is broad and includes 1, the point estimate should not be treated as a precise description of the underlying treatment effect. The interval is especially important in a trial that was terminated prematurely because of enrollment challenges.
Why doesn't the P-value measure treatment effect size?
A P-value describes the statistical evidence against a null hypothesis under a specified testing procedure. It depends on both the observed effect and the amount of information available. A P-value therefore cannot substitute for the hazard ratio and its confidence interval when describing the magnitude and precision of an effect.
Why was the analysis stratified?
The posted log-rank test was stratified by the time from COVID-19 positive test to randomization. Stratification incorporates this prespecified factor into the time-to-event comparison and avoids treating participants with different intervals between testing and randomization as though that feature were irrelevant to the comparison.
What does ITT contribute to the analysis?
The ITT analysis set included all randomized participants who provided informed consent. An ITT analysis retains the treatment assignment established by randomization, which preserves the central design feature that makes randomized treatment groups comparable in expectation.
9. Composite Endpoint Considerations
The primary efficacy endpoint combined seven possible first events: symptomatic VTE, MI, ischemic stroke, acute limb ischemia, non-CNS systemic embolization, all-cause hospitalization, and all-cause mortality.
| Component | Role in primary composite |
|---|---|
| Symptomatic venous thromboembolism (VTE) | Component of the first-event composite |
| Myocardial infarction (MI) | Component of the first-event composite |
| Ischemic stroke | Component of the first-event composite |
| Acute limb ischemia | Component of the first-event composite |
| Non-CNS systemic embolization | Component of the first-event composite |
| All-cause hospitalization | Component of the first-event composite |
| All-cause mortality | Component of the first-event composite |
Statistically, the composite endpoint is analyzed according to the first qualifying event. This creates a single time-to-event outcome from several clinically distinct possibilities. The resulting hazard ratio therefore summarizes the combined endpoint rather than establishing an effect on each individual component.
The inclusion of all-cause hospitalization is also important for interpretation. Hospitalization is conceptually different from death or a thrombotic event, so a composite result should be understood in the exact terms of the registered endpoint rather than being described simply as a measure of thrombosis.
10. Safety Endpoint
The second registered primary endpoint was the time to first occurrence of the principle safety outcome: fatal bleeding or critical site bleeding based on a modification of the ISTH criteria.
| Safety endpoint | Definition | Time frame |
|---|---|---|
| Principle safety outcome | Fatal bleeding or critical site bleeding based on a modification of ISTH criteria | Up to Day 35 |
The registry defines fatal bleeding as any bleeding event that leads to a fatal outcome. Critical site bleeding includes bleeding occurring at critical sites such as intracranial, intra-spinal, intraocular, pericardial, intra-articular, intra-muscular with compartment syndrome, and retroperitoneal sites.
Although this was a registered primary endpoint and results were posted, the registry data contain no formal statistical analysis for this endpoint comparable to the posted efficacy analysis. The reported safety information available in the record includes serious adverse events by randomized arm.
Serious adverse events by arm
| Arm | Affected | At risk |
|---|---|---|
| Rivaroxaban 10 mg | 1 | 599 |
| Placebo | 0 | 598 |
These counts describe serious adverse events by arm as reported in the registry. They should not be substituted for the formally defined principle safety outcome because the two concepts are not identical: serious adverse events are a broader safety category, whereas the principle safety outcome is specifically defined around fatal and critical-site bleeding.
11. Analysis Population and Censoring
The primary efficacy analysis used the ITT population. In a time-to-event analysis, participants contribute information while they remain under observation and event-free. If follow-up ends before the primary event occurs, the observation may be censored rather than treated as if the participant had experienced the event.
The analysis uses the available event-free follow-up rather than requiring every participant to experience the event before contributing information.
The registry does not provide additional censoring rules, missing-data procedures, or imputation specifications for the posted analysis. Those details therefore cannot be characterized beyond the information reported in the record.
12. Proportional-Hazards Interpretation
The reported effect measure was a Cox proportional-hazards hazard ratio. The proportional-hazards framework summarizes the relative instantaneous event rates between treatment groups under the model's proportionality structure.
If the relative hazard changes substantially over time, a single hazard ratio can compress a more complicated pattern of treatment effects into one number. The HR of 1.16 should therefore be interpreted as the model-based summary reported for the primary time-to-event comparison, rather than as a guarantee that the instantaneous event-rate ratio was exactly 1.16 at every point through Day 35.
The registry does not report a formal proportional-hazards diagnostic or a time-varying treatment-effect analysis. The appropriate interpretation is therefore the reported Cox hazard ratio within the stated analysis framework.
13. Why Premature Termination Matters Statistically
The registry states that the study was terminated prematurely because of enrollment challenges. This is a direct limitation of the information available for statistical estimation.
Precision
A smaller-than-planned information base can produce wider confidence intervals. Here, the 95% CI extends from 0.63 to 2.15, illustrating substantial uncertainty around the point estimate.
Information accumulation
Time-to-event analyses gain information from observed events and follow-up. Premature termination can restrict both the number of participants enrolled and the amount of event information accumulated.
Interpretation
The point estimate should be considered together with its confidence interval rather than interpreted in isolation. The interval allows a wide range of plausible relative effects.
Generalizability
Enrollment challenges can also affect the population ultimately represented in the completed study, although the registry does not provide enough information here to quantify that effect.
14. Statistical Interpretation of the Confidence Interval
The interval from 0.63 to 2.15 is particularly informative because it crosses the null hazard ratio of 1.00 in both directions.
A value below 1 would correspond to a lower estimated instantaneous event rate with rivaroxaban, while a value above 1 would correspond to a higher estimated instantaneous event rate. The confidence interval encompasses both possibilities. Thus, the precision of the estimate is a central part of the statistical interpretation.
The 95% confidence interval is not a range containing 95% of individual patient effects, and it is not a statement that there is a 95% probability that the true hazard ratio lies between 0.63 and 2.15. It is an interval constructed under the specified statistical model and repeated-sampling framework.
15. P-value and Effect Size Are Different Quantities
The primary analysis reported P = 0.626. It is tempting to treat the P-value as the main result, but that would discard information reported by the effect estimate and confidence interval.
| Quantity | Question it addresses | PREVENT-HD result |
|---|---|---|
| Hazard ratio | What relative treatment effect was estimated? | 1.16 |
| 95% CI | How precisely was that effect estimated? | 0.63–2.15 |
| P-value | How much statistical evidence is present against the null under the test? | 0.626 |
The three quantities should be read together. The hazard ratio supplies the direction and magnitude of the point estimate, the confidence interval describes its uncertainty, and the P-value summarizes evidence under the specified hypothesis-testing procedure.
16. Design Features Relevant to Interpretation
| Design feature | PREVENT-HD | Statistical relevance |
|---|---|---|
| Allocation | Randomized | Supports comparison according to assigned treatment |
| Masking | Double | Reduces the potential influence of treatment knowledge on trial conduct and assessment |
| Design model | Parallel | Participants remain in their randomized treatment group rather than crossing between treatment sequences as part of a crossover design |
| Primary endpoint type | Time-to-event | Requires methods that account for event timing and censoring |
| Primary efficacy population | ITT | Maintains randomized treatment assignment for the primary efficacy comparison |
| Statistical test | Stratified log-rank | Compares event-time distributions while incorporating the stated stratification factor |
| Effect measure | Hazard ratio | Provides a relative time-to-event treatment effect |
| Hypothesis type | Superiority | Tests for evidence of a treatment difference rather than against a non-inferiority margin |
The registry does not report a factorial design, crossover design, Bayesian analysis, non-inferiority margin, or a specific interim-analysis framework for the posted efficacy result.
17. Planned Analysis and Posted Results
The registry contains posted results and one formal statistical analysis. The posted formal analysis is for the primary efficacy composite endpoint and uses a stratified log-rank test with a Cox proportional-hazards hazard ratio.
Study start
The registered study start date was August 13, 2020.
Primary completion
The registered primary completion date was June 1, 2022.
Terminated
The study was terminated prematurely because of enrollment challenges, according to the registry.
18. Limitations
- Premature termination: the registry states that the study was terminated prematurely due to enrollment challenges. This limits the amount of information available for the treatment-effect estimate.
- Wide confidence interval: the primary efficacy HR of 1.16 has a 95% CI of 0.63–2.15, indicating substantial uncertainty around the point estimate.
- Composite endpoint: the primary efficacy endpoint combines symptomatic VTE, MI, ischemic stroke, acute limb ischemia, non-CNS systemic embolization, all-cause hospitalization, and all-cause mortality. A composite result should not automatically be interpreted as an effect on every component.
- Time-to-event assumptions: the reported Cox hazard ratio is model-based and should be interpreted within the proportional-hazards framework.
- Limited posted analyses: the registry reports one formal statistical analysis, for the primary efficacy endpoint. The posted record does not provide comparable formal statistical estimates for the primary safety endpoint.
- Missing analysis details: the registry does not report additional censoring rules, imputation procedures, or formal proportional-hazards diagnostics for the posted analysis.
- Population and enrollment: enrollment totaled 1,284 participants, but the study's premature termination due to enrollment challenges limits the amount of information accumulated relative to a fully completed study.
19. Why This Trial Matters Statistically
PREVENT-HD provides a useful teaching example of how randomized clinical trials with short-term time-to-event outcomes are analyzed. The statistical story is not contained in the hazard ratio alone: the endpoint definition, ITT population, stratification factor, confidence interval, P-value, censoring framework, and premature termination all affect interpretation.
| Concept | How it appears in PREVENT-HD |
|---|---|
| Randomization | Participants were randomized to rivaroxaban or placebo. |
| Blinding | The trial was double-blind. |
| Time-to-event endpoint | The primary efficacy and safety outcomes were defined as time to first occurrence through Day 35. |
| ITT analysis | The posted primary efficacy analysis used all randomized participants who provided informed consent. |
| Stratified analysis | The log-rank test was stratified by time from COVID-19 positive test to randomization. |
| Log-rank test | Used for the primary efficacy time-to-event comparison. |
| Hazard ratio | The primary efficacy effect measure was a Cox proportional-hazards hazard ratio of 1.16. |
| Confidence interval | The 95% CI was 0.63–2.15, illustrating uncertainty around the estimate. |
| P-value | The primary efficacy analysis reported P = 0.626. |
| Composite endpoint | The primary efficacy endpoint combined seven possible first-event components. |
| Premature termination | The study was terminated prematurely because of enrollment challenges. |
20. Clinical Biostats Interpretation
The posted primary efficacy analysis estimated a hazard ratio of 1.16 for rivaroxaban versus placebo, with a 95% confidence interval of 0.63–2.15 and a P-value of 0.626. The point estimate is above 1, but the confidence interval spans both below and above 1, indicating that the estimate is not precise enough to distinguish clearly between a lower and higher instantaneous event rate under the model.
The result should therefore be read as a time-to-event estimate with substantial uncertainty rather than as a simple comparison of percentages. The P-value does not measure the magnitude of the estimated effect, while the confidence interval does not represent the range of individual patient outcomes.
The HR of 1.16 does not mean that 16% more participants necessarily experienced the primary composite endpoint. It also does not mean that every participant had a 16% higher risk. Hazard ratios summarize relative instantaneous event rates under a time-to-event model.
The primary endpoint includes hospitalization as well as several thrombotic and mortality events. Consequently, the HR applies to the first occurrence of the entire registered composite. It cannot be interpreted as a pure measure of venous or arterial thrombosis without additional component-specific analyses.
The registry's statement that the study was terminated prematurely because of enrollment challenges is important context for the wide confidence interval. Statistical precision depends on the information accumulated in the study, and premature termination can restrict that information.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: PREVENT-HD, NCT04508023.
- PubMed: PMID 37154020.
- PubMed: PMID 37591523.
Continue through the Clinical Biostats knowledge graph
Explore the statistical methods that underlie randomized clinical-trial analysis, from time-to-event endpoints and hazard ratios to confidence intervals and hypothesis testing.
24. Record Summary
PREVENT-HD provides a focused example of randomized time-to-event analysis in an acute infectious-disease setting. The primary efficacy endpoint was the time to first occurrence of a broad composite through Day 35, and the posted formal analysis used an ITT population, a log-rank test stratified by time from COVID-19 positive test to randomization, and a Cox proportional-hazards hazard ratio.
The primary efficacy estimate was HR 1.16, with a 95% CI of 0.63–2.15 and P = 0.626. The confidence interval is central to interpretation because it spans both sides of the null value of 1 and is relatively wide. The registry also identifies premature termination because of enrollment challenges as an important limitation. Together, these features make PREVENT-HD a useful case study in interpreting a hazard ratio, confidence interval, log-rank test, composite endpoint, ITT analysis, stratification, and the effect of incomplete information on statistical precision.