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.
1. Trial at a Glance
COMPASS was a completed, randomized, quadruple-masked, parallel-group phase 3 trial evaluating rivaroxaban and aspirin strategies in the prevention setting for coronary or peripheral artery disease. The registry reports 27,395 participants and three antithrombotic treatment groups.
| Feature | COMPASS |
|---|---|
| Trial name | COMPASS |
| NCT identifier | NCT01776424 |
| Brief title | Rivaroxaban for the Prevention of Major Cardiovascular Events in Coronary or Peripheral Artery Disease |
| Phase | Phase 3 |
| Status | COMPLETED |
| Therapeutic area | Cardiology |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | QUADRUPLE |
| Primary purpose | PREVENTION |
| Enrollment | 27,395 |
| Lead sponsor | Bayer |
| Sponsor type | INDUSTRY |
| Start | 2013-02-28 |
| Primary completion | 2017-07-21 |
2. Clinical Question
The central statistical question was whether rivaroxaban-based antithrombotic strategies changed the time to clinically important cardiovascular events compared with rivaroxaban placebo plus aspirin 100 mg in participants randomized to the initial antithrombotic study part.
Population
Participants in a prevention trial for coronary or peripheral artery disease.
Intervention
Rivaroxaban 2.5mg + Aspirin 100mg, or Rivaroxaban 5mg + Aspirin Placebo.
Comparator
Rivaroxaban Placebo + Aspirin 100mg.
Primary questions
How did each rivaroxaban strategy compare with the aspirin-only strategy for the first occurrence of the composite efficacy outcome and for the first occurrence of modified ISTH major bleeding?
3. Trial Design
Participants were allocated to a parallel-group treatment design.
The registry classifies the trial as QUADRUPLE masked.
The trial was designed for prevention and control.
The formal statistical analyses treat the first occurrence of each endpoint as a time-to-event outcome.
Rivaroxaban 2.5mg + Aspirin 100mg
- Rivaroxaban (Xarelto, BAY59-7939)
- Aspirin 100mg
Rivaroxaban 5mg + Aspirin Placebo
- Rivaroxaban (Xarelto, BAY59-7939)
- Aspirin placebo
Rivaroxaban Placebo + Aspirin 100mg
- Rivaroxaban placebo
- Aspirin 100mg
4. Randomization, Stratification, and Analysis Population
The formal primary and secondary analyses used an intention-to-treat (ITT) population. The registry defines the ITT population as all participants randomized to antithrombotic treatment for the initial study part, including participants randomized to pantoprazole/placebo and participants not randomized to pantoprazole/placebo.
| Analysis feature | Registry-supported description |
|---|---|
| Primary efficacy population | ITT |
| Primary safety population | ITT for the posted formal analyses |
| Stratification | Analyses were stratified by pantoprazole randomization. |
| Formal test | Log-rank test |
| Model-based estimate | Stratified Cox proportional hazards model |
| Effect measure | Hazard ratio |
| Hypothesis type | Superiority |
| Confidence interval | 95%, two-sided |
This analysis structure is well matched to endpoints defined by the first occurrence of an event. Participants who have not experienced the event by the relevant analysis cutoff contribute follow-up information without requiring an event to have occurred. The hazard ratio therefore summarizes the relative instantaneous event rate over follow-up rather than simply comparing proportions of participants who eventually experienced an event.
5. Primary Endpoints
| Endpoint | Registered definition / time frame | Analysis |
|---|---|---|
| Composite primary efficacy outcome | The First Occurrence of the Composite Primary Efficacy Outcome, Myocardial Infarction (MI), Stroke, or Cardiovascular (CV) Death. For each participant, the first occurrence of the composite primary efficacy outcome after randomization was evaluated. | Stratified log-rank test; stratified Cox proportional hazards model; hazard ratio |
| Primary safety outcome | The First Occurrence of the Primary Safety Outcome Major Bleeding Based on a Modification of the International Society on Thrombosis and Haemostasis (ISTH) Criteria. For each participant, the first occurrence of modified ISTH major bleeding after randomization was evaluated. | Stratified log-rank test; stratified Cox proportional hazards model; hazard ratio |
The registry defines the composite efficacy endpoint around the first occurrence of myocardial infarction, stroke, or cardiovascular death. The primary safety endpoint is the first occurrence of modified ISTH major bleeding.
6. Statistical Methodology
Primary analytical framework
The registry reports the same core method for all four primary comparisons:
- Analysis population: intention-to-treat.
- Test: log-rank test.
- Stratification: pantoprazole randomization.
- Model: stratified Cox proportional hazards model.
- Effect measure: hazard ratio.
- Hypothesis: superiority.
The statistical strategy is coherent with the endpoint definitions. A participant can have different amounts of observed follow-up, and the analysis uses the timing of the first event rather than requiring every participant to have identical follow-up. The log-rank test provides the hypothesis test for separation in event-time distributions, while the Cox model supplies a hazard-ratio estimate and confidence interval.
Two treatment comparisons for each primary endpoint
The registry reports two separate antithrombotic comparisons against the same reference group:
| Comparison | Interpretation of the comparison |
|---|---|
| Rivaroxaban 2.5mg + Aspirin 100mg vs Rivaroxaban Placebo + Aspirin 100mg | Tests the rivaroxaban 2.5mg plus aspirin strategy against aspirin 100mg without active rivaroxaban. |
| Rivaroxaban 5mg + Aspirin Placebo vs Rivaroxaban Placebo + Aspirin 100mg | Tests the rivaroxaban 5mg strategy without active aspirin against aspirin 100mg without active rivaroxaban. |
Both comparisons were analyzed as superiority hypotheses. The ClinicalTrials.gov record does not provide a multiplicity-adjustment procedure, interim-analysis procedure, missing-data/imputation strategy, crossover analysis, or non-inferiority margin, so those topics are not inferred for this page.
7. Primary Efficacy Result: Rivaroxaban 2.5mg + Aspirin 100mg
For the first occurrence of the composite primary efficacy outcome of myocardial infarction, stroke, or cardiovascular death, the posted analysis compared rivaroxaban 2.5mg plus aspirin 100mg with rivaroxaban placebo plus aspirin 100mg.
Composite primary efficacy outcome
95% CI: 0.66–0.86 · P = 0.00004
Stratified log-rank test with a stratified Cox proportional hazards model; ITT population.
| Feature | Reported result |
|---|---|
| Comparison | Rivaroxaban 2.5mg + Aspirin 100mg vs Rivaroxaban Placebo + Aspirin 100mg |
| Endpoint | First occurrence of MI, stroke, or CV death |
| Method | Log Rank; stratified by pantoprazole randomization |
| Model | Stratified Cox proportional hazards model |
| Hazard ratio | 0.76 |
| 95% CI | 0.66–0.86 |
| P-value | 0.00004 |
| Hypothesis | Superiority |
What the estimate means: An HR of 0.76 means that, under the proportional-hazards interpretation of the model, the estimated hazard of the first composite efficacy event was 24% lower in the rivaroxaban 2.5mg plus aspirin group relative to the rivaroxaban-placebo plus aspirin group.
What it does not mean: It does not mean that exactly 24% fewer participants experienced an event, nor does it mean that every individual participant had a 24% reduction in personal risk. A hazard ratio is a relative time-to-event measure.
Precision: The 95% confidence interval extends from 0.66 to 0.86. The interval describes the statistical uncertainty around the estimated hazard ratio; it is not a range containing 95% of individual treatment effects.
P-value: The P-value of 0.00004 addresses the statistical evidence against the tested null hypothesis under the specified analysis. It is not a measure of effect size, clinical importance, or probability that the treatment hypothesis is true.
Caution: The interpretation of a Cox hazard ratio as a common relative hazard over time depends on the proportional-hazards model. The analysis was stratified by pantoprazole randomization and used the ITT population. The ClinicalTrials.gov record does not provide additional information needed to assess the proportional-hazards assumption or the treatment effect at specific time points.
8. Primary Efficacy Result: Rivaroxaban 5mg + Aspirin Placebo
The second primary efficacy analysis compared rivaroxaban 5mg plus aspirin placebo with rivaroxaban placebo plus aspirin 100mg for the first occurrence of myocardial infarction, stroke, or cardiovascular death.
Composite primary efficacy outcome
95% CI: 0.79–1.03 · P = 0.11490
Stratified log-rank test with a stratified Cox proportional hazards model; ITT population.
| Feature | Reported result |
|---|---|
| Comparison | Rivaroxaban 5mg + Aspirin Placebo vs Rivaroxaban Placebo + Aspirin 100mg |
| Endpoint | First occurrence of MI, stroke, or CV death |
| Method | Log Rank; stratified by pantoprazole randomization |
| Model | Stratified Cox proportional hazards model |
| Hazard ratio | 0.90 |
| 95% CI | 0.79–1.03 |
| P-value | 0.11490 |
| Hypothesis | Superiority |
What the estimate means: An HR of 0.90 corresponds to an estimated 10% lower hazard of the first composite efficacy event for rivaroxaban 5mg plus aspirin placebo relative to rivaroxaban placebo plus aspirin 100mg, under the model's proportional-hazards interpretation.
What it does not mean: The HR is not a percentage reduction in the number of participants with events, and it does not describe the probability of an event for an individual participant.
Precision: The 95% CI of 0.79–1.03 spans 1.00. Thus the interval includes both values below 1 and a value slightly above 1, indicating uncertainty about the direction and magnitude of the relative hazard under this analysis.
P-value: The P-value of 0.11490 is evidence from the specified superiority test, not a measure of the size or clinical relevance of the estimated effect.
Caution: The result is based on an ITT time-to-event analysis stratified by pantoprazole randomization. As with the other Cox analyses, the hazard-ratio interpretation relies on the model assumptions. The ClinicalTrials.gov record does not report a multiplicity-adjustment procedure for the two primary efficacy comparisons.
9. Primary Safety Result: Rivaroxaban 2.5mg + Aspirin 100mg
The primary safety endpoint was the first occurrence of major bleeding based on a modification of the International Society on Thrombosis and Haemostasis (ISTH) criteria. The registry defines modified ISTH major bleeding to include fatal bleeding, symptomatic bleeding in a critical area or organ, specified surgical-site bleeding requiring reoperation, and other qualifying major bleeding circumstances described in the registry definition.
Modified ISTH major bleeding
95% CI: 1.40–2.05 · P < 0.00001
Rivaroxaban 2.5mg + Aspirin 100mg vs Rivaroxaban Placebo + Aspirin 100mg.
| Feature | Reported result |
|---|---|
| Comparison | Rivaroxaban 2.5mg + Aspirin 100mg vs Rivaroxaban Placebo + Aspirin 100mg |
| Endpoint | First occurrence of modified ISTH major bleeding |
| Method | Log Rank; stratified by pantoprazole randomization |
| Model | Stratified Cox proportional hazards model |
| Hazard ratio | 1.70 |
| 95% CI | 1.40–2.05 |
| P-value | <0.00001 |
| Hypothesis | Superiority |
What the estimate means: An HR of 1.70 means that the estimated hazard of first modified ISTH major bleeding was 70% higher in the rivaroxaban 2.5mg plus aspirin group than in the rivaroxaban-placebo plus aspirin group, under the proportional-hazards interpretation.
What it does not mean: It does not mean that 70% of participants experienced bleeding, nor that each participant's absolute probability of major bleeding increased by 70 percentage points.
Precision: The 95% CI of 1.40–2.05 lies above 1.00. The width of the interval describes uncertainty around the relative hazard estimate.
P-value: The P-value of <0.00001 indicates strong statistical evidence against the superiority-test null hypothesis for this endpoint. It does not quantify the clinical seriousness of bleeding or the magnitude of absolute risk.
Caution: This is a time-to-first-event analysis. The hazard ratio does not by itself communicate the absolute number of bleeding events, cumulative incidence at a particular time, or the clinical distribution of bleeding events.
10. Primary Safety Result: Rivaroxaban 5mg + Aspirin Placebo
The second primary safety comparison evaluated the first occurrence of modified ISTH major bleeding for rivaroxaban 5mg plus aspirin placebo versus rivaroxaban placebo plus aspirin 100mg.
Modified ISTH major bleeding
95% CI: 1.25–1.84 · P = 0.00003
Rivaroxaban 5mg + Aspirin Placebo vs Rivaroxaban Placebo + Aspirin 100mg.
| Feature | Reported result |
|---|---|
| Comparison | Rivaroxaban 5mg + Aspirin Placebo vs Rivaroxaban Placebo + Aspirin 100mg |
| Endpoint | First occurrence of modified ISTH major bleeding |
| Method | Log Rank; stratified by pantoprazole randomization |
| Model | Stratified Cox proportional hazards model |
| Hazard ratio | 1.51 |
| 95% CI | 1.25–1.84 |
| P-value | 0.00003 |
| Hypothesis | Superiority |
What the estimate means: An HR of 1.51 corresponds to an estimated 51% higher hazard of first modified ISTH major bleeding for rivaroxaban 5mg plus aspirin placebo relative to rivaroxaban placebo plus aspirin 100mg.
What it does not mean: It is not a 51-percentage-point increase in bleeding probability and does not indicate that 51% more participants necessarily experienced bleeding.
Precision: The 95% CI extends from 1.25 to 1.84. Because the entire interval is above 1.00, the estimated relative hazard is above the reference hazard throughout the reported confidence interval.
P-value: The P-value of 0.00003 describes the statistical evidence from the specified superiority test. It should not be interpreted as the probability that the treatment causes bleeding or as a measure of clinical importance.
Caution: Interpretation remains conditional on the ITT population, the stratification by pantoprazole randomization, and the proportional-hazards framework used by the registry analysis.
11. Secondary Endpoint Results
The registry also reports formal time-to-event analyses for several secondary outcomes. Each used the ITT population, a log-rank test stratified by pantoprazole randomization, and a stratified Cox proportional hazards model.
| Secondary endpoint | Comparison | HR | 95% CI | P-value |
|---|---|---|---|---|
| First occurrence of MI, ischemic stroke, acute limb ischemia, or CHD death | Rivaroxaban 2.5mg + Aspirin 100mg vs Rivaroxaban Placebo + Aspirin 100mg | 0.72 | 0.63–0.83 | 0.00001 |
| First occurrence of MI, ischemic stroke, acute limb ischemia, or CHD death | Rivaroxaban 5mg + Aspirin Placebo vs Rivaroxaban Placebo + Aspirin 100mg | 0.88 | 0.77–1.01 | 0.06437 |
| First occurrence of MI, ischemic stroke, acute limb ischemia, or CV death | Rivaroxaban 2.5mg + Aspirin 100mg vs Rivaroxaban Placebo + Aspirin 100mg | 0.74 | 0.65–0.85 | 0.00001 |
| First occurrence of MI, ischemic stroke, acute limb ischemia, or CV death | Rivaroxaban 5mg + Aspirin Placebo vs Rivaroxaban Placebo + Aspirin 100mg | 0.88 | 0.77–0.99 | 0.03995 |
| All-cause mortality | Rivaroxaban 2.5mg + Aspirin 100mg vs Rivaroxaban Placebo + Aspirin 100mg | 0.82 | 0.71–0.96 | 0.01062 |
| All-cause mortality | Rivaroxaban 5mg + Aspirin Placebo vs Rivaroxaban Placebo + Aspirin 100mg | 0.97 | 0.84–1.12 | 0.66418 |
The secondary results illustrate why a statistical analysis should distinguish the effect estimate from the evidence against the null hypothesis. For example, the HR of 0.72 for the MI/ischemic-stroke/acute-limb-ischemia/CHD-death endpoint in the rivaroxaban 2.5mg plus aspirin comparison is accompanied by a 95% CI of 0.63–0.83, whereas the corresponding 5mg comparison has an HR of 0.88 with a 95% CI of 0.77–1.01. These are different estimates with different precision, and the confidence intervals communicate more information about uncertainty than the point estimates alone.
The all-cause mortality analyses provide another useful contrast. The 2.5mg plus aspirin comparison has an HR of 0.82 (95% CI 0.71–0.96; P = 0.01062), while the 5mg plus aspirin placebo comparison has an HR of 0.97 (95% CI 0.84–1.12; P = 0.66418). The ClinicalTrials.gov record does not provide a multiplicity-adjustment procedure, so these secondary P-values should not be treated as though they were the only statistical tests performed in the trial.
12. Safety: Serious Adverse Events by Arm
The ClinicalTrials.gov record reports serious adverse events as affected participants over participants at risk for the three antithrombotic treatment groups.
| Arm | Serious adverse events | Format |
|---|---|---|
| Rivaroxaban 2.5mg + Aspirin 100mg | 692/9134 | Affected / at risk |
| Rivaroxaban 5mg + Aspirin Placebo | 664/9109 | Affected / at risk |
| Rivaroxaban Placebo + Aspirin 100mg | 630/9107 | Affected / at risk |
| LTOLE Rivaroxaban Rivaroxaban 2.5mg + As | 292/12903 | Affected / at risk |
The serious-adverse-event counts above should not be substituted for the primary safety endpoint. Serious adverse events and modified ISTH major bleeding are different outcome definitions. The primary safety analysis specifically concerns the first occurrence of modified ISTH major bleeding and estimates its relative hazard using time-to-event methods.
13. Statistical Methods Explained
Why was a log-rank test used?
The primary outcomes were defined by the first occurrence of clinical events after randomization. A log-rank test is designed to compare event-time distributions between treatment groups while incorporating the timing of events and censored follow-up. That makes it more appropriate for these endpoints than a simple comparison of event proportions at one arbitrary time point.
Why was a Cox proportional hazards model used?
The registry reports a stratified Cox proportional hazards model in addition to the log-rank test. The Cox model provides a hazard ratio that summarizes the relative event hazard between treatment groups while respecting the time-to-event structure. In this trial, the model was stratified by pantoprazole randomization.
What does a hazard ratio below 1 mean?
For an endpoint such as the composite efficacy outcome, an HR below 1 indicates a lower estimated event hazard in the numerator treatment group relative to the comparator. Thus HR 0.76 corresponds to a 24% lower estimated hazard. It does not mean a 24% absolute reduction in risk or that every participant experiences the same relative effect.
What does a hazard ratio above 1 mean?
For the modified ISTH major-bleeding endpoint, an HR above 1 indicates a higher estimated event hazard in the treatment group relative to the comparator. HR 1.70 corresponds to a 70% higher estimated hazard under the model interpretation. Again, this is not an absolute risk increase of 70 percentage points.
Why is the confidence interval important?
A point estimate such as HR 0.76 is only one estimate from the observed data. The 95% confidence interval of 0.66–0.86 communicates the statistical precision around that estimate. A confidence interval that crosses 1.00 indicates that a no-difference hazard ratio is compatible with the interval; an interval entirely below 1.00 or above 1.00 indicates that the reported interval does not contain 1.00.
Why doesn't the P-value measure the size of the treatment effect?
The P-value evaluates the evidence against the specified null hypothesis under the statistical model. It depends on both the observed effect and the amount of information available. Consequently, a very small P-value can accompany a modest effect, while a larger P-value can occur with a larger point estimate when the estimate is imprecise. Effect size and confidence interval should therefore be examined alongside the P-value.
Why was the analysis stratified by pantoprazole randomization?
The registry explicitly states that the log-rank tests and stratified Cox proportional hazards models were stratified by pantoprazole randomization. Stratification allows the baseline hazard structure to differ across the strata while estimating the treatment comparison within the stratified analysis framework. The ClinicalTrials.gov record does not provide additional details about the statistical rationale beyond this reported method specification.
14. Understanding the Four Primary Analyses Together
| Primary endpoint | 2.5mg Rivaroxaban + Aspirin | 5mg Rivaroxaban + Aspirin Placebo |
|---|---|---|
| Composite efficacy: MI, stroke, or CV death | HR 0.76 (95% CI 0.66–0.86; P = 0.00004) | HR 0.90 (95% CI 0.79–1.03; P = 0.11490) |
| Modified ISTH major bleeding | HR 1.70 (95% CI 1.40–2.05; P < 0.00001) | HR 1.51 (95% CI 1.25–1.84; P = 0.00003) |
Statistically, the four analyses should be read as distinct treatment-endpoint comparisons rather than collapsed into a single number. The efficacy endpoint and safety endpoint measure different clinical phenomena, and the two rivaroxaban strategies are different regimens. The table makes the central structure visible: each active strategy was compared with the same rivaroxaban-placebo plus aspirin reference group, while the outcomes were analyzed using the same general time-to-event framework.
An important feature is that the direction of the estimated hazard ratio differs by endpoint. For the 2.5mg plus aspirin strategy, the efficacy HR is below 1 while the major-bleeding HR is above 1. The same directional pattern appears for the 5mg strategy. That contrast is precisely why an efficacy-only statistical analysis would be incomplete: the primary safety endpoint was formally specified and analyzed alongside the primary efficacy endpoint.
15. What the Registry Data Do Not Establish
The ClinicalTrials.gov record supports a detailed analysis of the registered endpoints and posted hazard-ratio analyses, but they do not support several additional claims that might appear in a more extensive clinical-trial review.
- The ClinicalTrials.gov record does not report median survival or median time-to-event values.
- The ClinicalTrials.gov record does not provide Kaplan-Meier event probabilities at specific time points.
- No subgroup-specific hazard ratios or forest-plot results are reported.
- No formal multiplicity-adjustment procedure is reported.
- No interim-analysis procedure is reported.
- No missing-data or imputation strategy is reported.
- No crossover analysis is reported.
- No non-inferiority margin is reported; the registered hypothesis type is superiority.
- No Bayesian analysis is reported.
16. Limitations
Hazard ratios are relative measures
The primary effect measures are hazard ratios. They are useful for comparing event hazards over follow-up, but they do not directly provide absolute risk differences, numbers needed to treat, or event probabilities at a particular time.
Proportional-hazards interpretation
The Cox model is a proportional-hazards model. Interpreting one HR as a common relative hazard over follow-up therefore depends on the proportional-hazards framework. The ClinicalTrials.gov record does not report a formal assessment of that assumption.
Multiple statistical comparisons
There are two primary efficacy comparisons and two primary safety comparisons, followed by multiple secondary analyses. The ClinicalTrials.gov record identifies the hypothesis type as superiority but do not specify a multiplicity-adjustment procedure. Consequently, individual P-values should not automatically be interpreted as though they represent a single isolated hypothesis test.
Limited absolute-risk information
The ClinicalTrials.gov record provides hazard ratios, confidence intervals, and P-values but not the underlying event counts for the primary efficacy and primary safety analyses. That prevents calculation of absolute risk differences or numbers needed to treat from the ClinicalTrials.gov record alone.
Safety-set reconciliation
The registry includes a correction concerning one participant in the Safety Analysis Set who never took study drug. This demonstrates why version-specific registry data and analysis-population definitions matter when interpreting safety results.
17. Why This Trial Matters Statistically
COMPASS is a useful teaching example because its primary statistical structure makes several core survival-analysis concepts visible in one trial.
Time-to-event analysis
The primary outcomes are defined by the first occurrence of clinical events, making event timing central to the analysis rather than merely whether an event ever occurred.
Hazard ratios
The trial demonstrates how a relative hazard can summarize a treatment comparison and why the HR should be interpreted alongside its confidence interval.
Stratified testing
The registry explicitly specifies stratification by pantoprazole randomization for both the log-rank tests and Cox models.
Efficacy and safety together
The primary efficacy and safety endpoints move in different directions, illustrating why a treatment comparison cannot be understood from efficacy P-values alone.
The trial is also a useful example of the distinction between statistical significance and effect estimation. For example, the 5mg efficacy comparison has HR 0.90 with a 95% CI of 0.79–1.03 and P = 0.11490, while its major-bleeding comparison has HR 1.51 with a 95% CI of 1.25–1.84 and P = 0.00003. Looking only at whether a P-value is below a threshold would discard much of the information conveyed by the effect estimates and confidence intervals.
Finally, the repeated use of the same statistical framework across efficacy and safety outcomes makes COMPASS a particularly clear example of a prespecified time-to-event analysis pipeline: define a first event, follow participants over time, compare event-time distributions with a stratified log-rank test, and estimate the relative hazard with a stratified Cox model.
Learn the statistical methods behind COMPASS
Explore the core survival-analysis concepts used to interpret randomized clinical-trial time-to-event results.
20. Sources
- ClinicalTrials.gov — COMPASS, NCT01776424
- PubMed — PMID 29540326
- PubMed — PMID 29132880
- PubMed — PMID 29132879
- PubMed — PMID 28844192
- PubMed — PMID 28754388