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
GEMINI-ACS-1 was a randomized, double-blind, parallel phase 2 trial evaluating the safety of rivaroxaban versus acetylsalicylic acid when used in addition to either clopidogrel or ticagrelor therapy in participants with acute coronary syndrome.
| Feature | GEMINI-ACS-1 |
|---|---|
| Trial name | GEMINI-ACS-1 |
| NCT ID | NCT02293395 |
| Therapeutic area | Cardiovascular |
| Condition | Acute Coronary Syndrome |
| Phase | Phase 2 |
| Status | Completed |
| Enrollment | 3037 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Lead sponsor | Janssen Research & Development, LLC |
| Sponsor type | Industry |
2. Clinical Question
The registered trial question was whether rivaroxaban, used in addition to P2Y12 inhibitor therapy, could be compared with acetylsalicylic acid used in addition to P2Y12 inhibitor therapy with respect to the occurrence of non coronary artery bypass graft-related Thrombolysis in Myocardial Infarction (TIMI) clinically significant bleeding events in participants with acute coronary syndrome.
Population
Participants with acute coronary syndrome.
Intervention
Rivaroxaban in addition to P2Y12 inhibitor therapy.
Comparator
Acetylsalicylic acid in addition to P2Y12 inhibitor therapy.
Primary question
How do rivaroxaban and acetylsalicylic acid compare for the registered bleeding endpoint over follow-up of up to 390 days?
3. Trial Design
Rivaroxaban
- Rivaroxaban 2.5 mg twice daily (BID)
- Used in addition to P2Y12 inhibitor therapy
- 1510 participants were at risk for the reported serious-adverse-event measure
Acetylsalicylic acid
- Acetylsalicylic acid 100 mg once daily (OD)
- Used in addition to P2Y12 inhibitor therapy
- 1506 participants were at risk for the reported serious-adverse-event measure
The trial also included clopidogrel and ticagrelor as interventions. Importantly, the registry states that randomization was not performed for use of the P2Y12 inhibitor. Consequently, comparisons across clopidogrel and ticagrelor strata are not randomized treatment comparisons.
4. Trial Timeline
Study start
The registered study start date was 2015-04-20.
Primary completion
The registered primary completion date was 2016-10-14.
Registry status
The study is listed as completed, with results posted on ClinicalTrials.gov.
5. Primary Endpoint
| Endpoint | Registry definition / time frame | Statistical analysis |
|---|---|---|
| Number of Participants With Non Coronary Artery Bypass Graft-Related (Non CABG-related) Thrombolysis in Myocardial Infarction (TIMI) Clinically Significant Bleeding Events | From start of study treatment until follow-up (up to 390 days) | Log-rank test; hazard ratio |
The registry classifies the primary endpoint as binary in its registered endpoint information, while the posted statistical analysis treats the outcome as a time-to-event endpoint and uses a log-rank test with a hazard ratio. The latter is the analysis framework directly represented by the posted statistical analysis.
What counted as clinically significant bleeding?
The registry defines non CABG-related TIMI clinically significant bleeding events as the sum of non CABG-related TIMI major bleeding events, TIMI minor bleeding events, and TIMI bleeding events requiring medical attention.
The registered definition identifies major bleeding as including symptomatic intracranial bleeding and clinically overt signs of hemorrhage with a hemoglobin drop of greater than or equal to 5 gram per deciliter (g/dl), among other criteria contained in the registry definition.
6. Statistical Methodology
Log-rank test
The posted analysis used the log-rank test to compare the time-to-event experience between the rivaroxaban and acetylsalicylic acid groups. The log-rank test is designed for comparing survival or event-time distributions when some participants are censored before experiencing the event.
The log-rank test evaluates whether the observed pattern of events over follow-up differs between groups, accounting for the timing of events rather than reducing the endpoint to a simple yes/no proportion at a single time point.
Hazard ratio
The effect measure reported for the primary analysis was the hazard ratio (HR). The estimated hazard ratio for rivaroxaban versus acetylsalicylic acid was 1.09.
An HR of 1 would correspond to equal estimated hazards under the model. An HR of 1.09 corresponds to an estimated hazard that is 9% higher in the rivaroxaban group relative to the acetylsalicylic acid group, within the model and analysis framework used for the comparison.
Confidence interval
The reported 95% confidence interval was 0.8 to 1.5. A confidence interval gives a measure of statistical precision around the estimated hazard ratio. It is not a range containing 95% of individual patient outcomes.
P-value
The reported two-sided p-value was 0.584. In the context of the registered superiority hypothesis, this is the probability measure produced by the specified statistical test under the null hypothesis; it is not a measure of the size, clinical importance, or probability of the observed effect.
Analysis population
The registry describes the analysis population as including all randomized participants who had received at least one dose of study agent and had events occurring between randomization and the last dose of the study agent plus the applicable follow-up described in the registry analysis record.
7. Primary Result
Non CABG-related TIMI Clinically Significant Bleeding Events
Hazard ratio for clinically significant bleeding
95% CI: 0.8–1.5 · P = 0.584
Rivaroxaban 2.5 mg BID vs acetylsalicylic acid 100 mg OD
| Primary endpoint | Comparison | Method | Effect estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Number of Participants With Non CABG-related TIMI Clinically Significant Bleeding Events | Rivaroxaban 2.5 mg BID vs Acetylsalicylic Acid 100 mg OD | Log-rank | HR 1.09 | 0.8–1.5 | 0.584 |
The estimated hazard ratio of 1.09 means that the fitted relative event rate for non CABG-related TIMI clinically significant bleeding was estimated to be 9% higher with rivaroxaban than with acetylsalicylic acid in this randomized comparison.
That estimate does not mean that 9% more participants experienced bleeding, nor does it mean that an individual participant had a 9% higher probability of bleeding. A hazard ratio is a relative time-to-event measure, not an absolute risk difference or risk ratio.
The 95% CI of 0.8–1.5 indicates substantial uncertainty around the point estimate. The interval includes 1, so the data represented by this analysis are compatible with a range of relative hazard values on both sides of equal hazard.
The p-value of 0.584 does not quantify the magnitude of the treatment difference. It describes the statistical evidence against the null hypothesis under the specified two-sided log-rank testing framework. It should therefore be read alongside the HR and confidence interval rather than as a standalone measure of effect.
Because the endpoint was analyzed as a time-to-event outcome, interpretation also depends on the event timing, censoring, and the assumptions underlying the hazard-based representation. A single HR summarizes a complex event-time process and should not automatically be interpreted as a constant difference in risk at every point in follow-up.
8. Safety Results
The ClinicalTrials.gov record reports serious adverse events by arm for the primary randomized treatment comparison.
| Serious adverse-event measure | Affected / at risk |
|---|---|
| Rivaroxaban 2.5 mg Twice Daily (BID) | 125 / 1510 |
| Acetylsalicylic Acid 100 mg Once Daily | 138 / 1506 |
These counts describe affected participants relative to the participants at risk in the corresponding groups. They should not be substituted for the primary bleeding endpoint: serious adverse events and non CABG-related TIMI clinically significant bleeding are distinct outcome concepts in the ClinicalTrials.gov record.
Rivaroxaban
125 affected participants among 1510 participants at risk for the reported serious-adverse-event measure.
Acetylsalicylic acid
138 affected participants among 1506 participants at risk for the reported serious-adverse-event measure.
9. Statistical Methods Explained
Why was a log-rank test used?
The primary outcome was analyzed as a time-to-event endpoint, with follow-up extending from the start of study treatment until follow-up of up to 390 days. A log-rank test is suited to comparing event-time distributions between randomized groups when participants may have different observed follow-up times because of censoring.
What does a hazard ratio of 1.09 mean?
An HR of 1.09 means that the estimated hazard in the rivaroxaban group was 1.09 times the estimated hazard in the acetylsalicylic acid group under the analysis model. Equivalently, it represents an estimated 9% higher hazard. It does not directly give the percentage of participants who experienced bleeding.
Why does the confidence interval matter?
The point estimate is only one estimate of the treatment contrast. The 95% CI of 0.8–1.5 shows the uncertainty around that estimate. Because the interval spans 1, the registry-reported analysis does not establish a statistically distinguishable difference from equal hazard under the two-sided superiority framework.
Why does the p-value not measure effect size?
The p-value of 0.584 describes the statistical evidence against the null hypothesis under the specified test. It does not tell us that the treatment effect is "58.4%" or that there is a 58.4% probability that the treatments are equivalent. Effect magnitude is described by the HR, while its uncertainty is described by the confidence interval.
Why is randomization important here?
Randomization is the design feature that supports a causal comparison between the randomized treatment groups, subject to the trial's eligibility criteria, conduct, analysis population, and other limitations. It is different from the nonrandomized choice of P2Y12 inhibitor described in the registry.
Why should clopidogrel and ticagrelor not be compared as if they were randomized arms?
The registry explicitly states that randomization was not performed for use of the P2Y12 inhibitor. Therefore, differences between participants receiving clopidogrel and ticagrelor can be confounded by factors associated with how that therapy was selected. The registry also states that such strata comparisons were confounded and were not planned.
What does "double-blind" contribute statistically?
Double masking reduces the opportunity for knowledge of treatment assignment to influence trial conduct and assessment. In a trial involving a clinically defined bleeding endpoint, maintaining masking can help reduce differential assessment or management related to knowledge of assigned treatment.
10. Confidence Intervals and the Superiority Question
The registered hypothesis type for the posted primary analysis is superiority. The reported hazard ratio of 1.09 is therefore interpreted as an estimate of the relative event hazard under the superiority comparison.
The point estimate lies above 1, but the confidence interval extends below and above 1. The registry-reported p-value is 0.584 from a two-sided log-rank analysis.
This distinction is important. A point estimate above 1 is not, by itself, evidence of a statistically established difference. Conversely, a confidence interval that includes 1 does not prove that the two treatments have exactly identical effects. It indicates that the data are compatible with equal hazard as well as with values on either side of 1 within the interval.
11. Time-to-Event Analysis in GEMINI-ACS-1
The primary endpoint combines two pieces of information: whether a clinically significant bleeding event occurred and when it occurred during the observed follow-up. That makes the posted analysis different from simply calculating the percentage of participants who experienced an event by a fixed calendar date.
Event timing
The analysis retains information about when clinically significant bleeding events occurred during follow-up.
Censoring
Participants who do not experience the event during their observed period can contribute information before their censoring time.
Log-rank comparison
The test compares the event-time experience between the two randomized treatment groups.
Hazard ratio
The HR summarizes the relative hazard between the rivaroxaban and acetylsalicylic acid groups.
Why not reduce the result to a single percentage?
A fixed-time percentage can be useful, but it can discard information about event timing and follow-up. The posted GEMINI-ACS-1 analysis instead uses a survival-analysis framework, which is consistent with a trial in which participants may have different amounts of observed follow-up.
12. Randomization and the P2Y12 Inhibitor Issue
the ClinicalTrials.gov record identifies the overall allocation as randomized and the design as parallel. At the same time, the registry provides an important qualification: randomization was not performed for use of the P2Y12 inhibitor.
| Feature | What the ClinicalTrials.gov record supports |
|---|---|
| Overall allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Rivaroxaban vs acetylsalicylic acid | Randomized treatment comparison represented in the posted primary analysis |
| P2Y12 inhibitor selection | Not randomized |
| Clopidogrel vs ticagrelor strata | Confounded comparison; registry states strata comparisons were not planned |
This distinction is central to interpreting the trial's four listed interventions. The presence of four intervention names does not mean that every pairwise contrast among them represents a randomized causal comparison.
13. What the Primary Result Does — and Does Not — Mean
The HR of 1.09 is a relative time-to-event estimate comparing rivaroxaban 2.5 mg BID with acetylsalicylic acid 100 mg OD for the registered non CABG-related TIMI clinically significant bleeding endpoint.
It does not mean that 9% of participants experienced bleeding, that the absolute bleeding probability increased by 9 percentage points, or that every participant had exactly the same relative change in risk.
The 95% CI of 0.8–1.5 indicates uncertainty around the estimated HR. It includes values below 1, equal to 1, and above 1, so the point estimate should not be treated as a precise determination of the direction or magnitude of the underlying treatment effect.
The two-sided p-value of 0.584 is a test statistic summary under the specified null hypothesis. It does not measure clinical relevance, equivalence, or the probability that the null hypothesis is true.
14. Safety and Efficacy Are Different Statistical Questions
GEMINI-ACS-1 illustrates why a trial can have more than one important statistical question even when the primary endpoint is focused on safety.
| Question | Supplied data | Statistical interpretation |
|---|---|---|
| Primary bleeding endpoint | HR 1.09; 95% CI 0.8–1.5; P = 0.584 | Formal time-to-event comparison using a log-rank test |
| Serious adverse events | 125/1510 vs 138/1506 | Counts by arm are reported; no formal comparison is reported |
The distinction matters because the primary endpoint has an explicitly posted inferential analysis, whereas the registry-reported serious-adverse-event information consists of affected and at-risk counts. It would be inappropriate to attach the primary endpoint's HR or p-value to the serious-adverse-event counts.
15. Limitations
- Homogeneous study population: The registry states that the study population was homogeneous with limited participation of non-Caucasian participants. This limits how broadly the findings can be generalized across populations that were less represented.
- Nonrandomized P2Y12 inhibitor use: Randomization was not performed for use of the P2Y12 inhibitor. Comparisons between clopidogrel and ticagrelor strata are therefore confounded and were not planned.
- Analysis-population qualification: The posted analysis population included randomized participants who had received at least one dose of study agent and had events occurring within the registry-specified analysis window. This should be distinguished from an unrestricted all-randomized ITT analysis.
- Time-to-event assumptions: Hazard-ratio interpretation depends on the survival-analysis framework and the way hazards summarize event timing over follow-up. A single HR should not automatically be interpreted as a constant difference in absolute risk.
- Confidence-interval width: The 95% CI of 0.8–1.5 spans a relatively broad range of possible hazard ratios around the point estimate, so the estimate should be interpreted with attention to uncertainty.
- Endpoint distinction: Serious adverse events and the primary non CABG-related TIMI clinically significant bleeding endpoint are different measures and should not be conflated.
- Generalizability: Limited participation of non-Caucasian participants means that the trial population does not equally represent all demographic groups.
16. Why This Trial Matters Statistically
GEMINI-ACS-1 is a useful teaching case because it shows how a randomized trial can combine a safety-focused endpoint, survival-analysis methodology, masked treatment assignment, and an important distinction between randomized and nonrandomized treatment factors.
| Concept | How it appears in GEMINI-ACS-1 |
|---|---|
| Randomization | The trial allocation is randomized for the treatment comparison represented in the primary analysis. |
| Blinding | The trial is double-masked. |
| Parallel design | The registry identifies a parallel design model. |
| Time-to-event endpoint | The posted primary analysis treats the bleeding endpoint as a time-to-event outcome. |
| Kaplan-Meier framework | The primary endpoint is analyzed in a survival-analysis framework, for which Kaplan-Meier estimation is the standard descriptive companion. |
| Log-rank test | The registry-posted statistical method is a log-rank test. |
| Hazard ratio | The treatment effect is reported as an HR of 1.09. |
| Confidence interval | The 95% CI is 0.8–1.5. |
| P-value | The reported two-sided p-value is 0.584. |
| Superiority testing | The posted hypothesis type is superiority. |
| Nonrandomized factor | P2Y12 inhibitor use was not randomized, limiting interpretation of strata comparisons. |
| Generalizability | The registry identifies a homogeneous population with limited participation of non-Caucasian participants. |
17. A Practical Reading of the Statistical Result
A disciplined interpretation of the primary result can be organized into four layers.
1. Effect estimate
The HR was 1.09 for rivaroxaban versus acetylsalicylic acid.
2. Precision
The 95% CI was 0.8–1.5, indicating uncertainty around the point estimate.
3. Hypothesis test
The two-sided log-rank p-value was 0.584 under the registered superiority framework.
4. Design context
The comparison was randomized and double-blind, while P2Y12 inhibitor use was not randomized.
This four-part reading is more informative than simply reporting that the p-value was above or below a conventional threshold. It keeps the estimated effect, its uncertainty, the inferential test, and the design context connected.
18. How to Read the Serious-Adverse-Event Counts
The ClinicalTrials.gov record reports 125 affected participants among 1510 at risk in the rivaroxaban group and 138 among 1506 at risk in the acetylsalicylic acid group.
These are descriptive affected/at-risk counts. The ClinicalTrials.gov record does not provide a formal hypothesis test, confidence interval, or hazard ratio for this serious-adverse-event measure.
It is therefore appropriate to report the counts as reported in the registry, while avoiding the appearance that they represent the formal statistical analysis of the primary endpoint. The primary analysis is specifically the log-rank comparison of non CABG-related TIMI clinically significant bleeding events.
19. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The posted analysis estimated an HR of 1.09 with a 95% CI of 0.8–1.5 and a two-sided p-value of 0.584 using a log-rank test.
Clinical interpretation
The ClinicalTrials.gov record describes the relative time-to-event comparison and selected serious-adverse-event counts, but they do not provide enough additional outcome information to construct a broader clinical benefit-risk assessment.
The distinction is deliberate. Statistical evidence describes the observed randomized comparison under a specified analysis. A broader clinical interpretation would require considering the full set of efficacy, safety, follow-up, and patient-level outcomes, and those additional results are not included in the ClinicalTrials.gov record.
20. What Is Not Supported by the Supplied Data
The ClinicalTrials.gov record does not provide results for additional endpoints beyond the posted primary analysis and the reported serious-adverse-event counts. Accordingly, this page does not introduce median event times, subgroup effect estimates, Kaplan-Meier percentages, additional p-values, additional confidence intervals, or other outcome measures from outside sources.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: GEMINI-ACS-1, NCT02293395.
- PubMed: PMID 31535314.
- PubMed: PMID 31141104.
- PubMed: PMID 28325638.
- PubMed: PMID 26995378.
Continue through the Clinical Biostats statistical pathway
Connect this trial's randomized design, time-to-event endpoint, hazard ratio, confidence interval, and log-rank analysis to deeper statistical tutorials and calculation tools.
24. Record Summary
GEMINI-ACS-1 provides a useful example of how a randomized, double-blind phase 2 trial can be analyzed when its primary endpoint is represented as a time-to-event outcome. The registry analysis compares rivaroxaban 2.5 mg twice daily with acetylsalicylic acid 100 mg once daily using a log-rank test and reports a hazard ratio of 1.09, a two-sided 95% confidence interval of 0.8–1.5, and a p-value of 0.584. The most appropriate reading of that result keeps the effect estimate, its uncertainty, and the hypothesis test together rather than reducing the analysis to the p-value alone.
The trial also illustrates an important design principle: overall randomization does not make every treatment factor randomized. The registry states that P2Y12 inhibitor use was not randomized and that comparisons between clopidogrel and ticagrelor strata were confounded and not planned. Finally, the limited participation of non-Caucasian participants is an explicit limitation on generalizability.