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
AZALEA-TIMI 71 was a randomized, parallel-group, triple-masked phase 2 trial evaluating the safety and tolerability of abelacimab versus rivaroxaban in patients with atrial fibrillation and stroke as listed conditions. The registry reports three arms and an enrollment of 1287 participants.
| Feature | AZALEA-TIMI 71 |
|---|---|
| Phase | Phase 2 |
| Therapeutic area | Cardiovascular |
| Conditions | Atrial Fibrillation (AF); Stroke |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Triple |
| Primary purpose | Supportive care |
| Enrollment | 1287 |
| Arms | 3 |
| Lead sponsor | Anthos Therapeutics, Inc. |
| Start date | February 2, 2021 |
| Primary completion | February 15, 2024 |
| Current registry status | Active, not recruiting |
2. Clinical Question
The central statistical question was whether treatment with abelacimab was associated with a different time-to-first-bleeding-event profile than rivaroxaban for the registered composite of ISTH-defined major bleeding or clinically relevant non-major (CRNM) bleeding events.
Population
Patients enrolled in AZALEA-TIMI 71 with the registry-listed condition of atrial fibrillation; stroke is also listed as a condition.
Intervention
Abelacimab (MAA868), evaluated at 90 mg and 150 mg in the posted primary analyses.
Comparator
Rivaroxaban.
Primary question
How does the hazard of first composite major or CRNM bleeding compare between each abelacimab dose group and rivaroxaban?
3. Trial Design
Abelacimab (MAA868)
- Biological intervention
- 90 mg dose
- Primary analyses compared this group with rivaroxaban
- Primary composite bleeding endpoint analyzed using a stratified log-rank test and hazard ratio
Abelacimab (MAA868)
- Biological intervention
- 150 mg dose
- Primary analyses compared this group with rivaroxaban
- Primary composite bleeding endpoint analyzed using a stratified log-rank test and hazard ratio
4. Randomization, Masking, and Analysis Population
Randomization is important because it establishes the treatment comparison before outcomes occur. The trial was also triple-masked, which is a design feature intended to reduce the influence of treatment knowledge on trial conduct and outcome assessment.
The posted statistical analyses explicitly define their analysis population as the number of randomized participants who received at least one dose of study treatment. This distinction is important: the randomized population describes assignment, whereas the posted efficacy analyses use participants who actually received at least one dose.
5. Primary Endpoint
| Endpoint | Registry definition and time frame | Analysis |
|---|---|---|
| Incidence Rate of the First Occurrence of Composite of International Society on Thrombosis and Haemostasis (ISTH)-Defined Major Bleeding or Clinically Relevant Non-Major (CRNM) Bleeding Events | Assessed at every visit from randomization through the end of the randomized phase of the study, up to 33 months. | Stratified log-rank test; Cox proportional hazards model adjusted for stratification factors; hazard ratio. |
The registry defines the endpoint as the number of participants experiencing the first occurrence of the composite divided by the number of 100 person-years through the first event or end of treatment across all participants. For participants without an event, person-years are calculated through the end of treatment during the randomized phase.
This wording contains two related statistical ideas. The endpoint is reported as an incidence rate using person-time, while the posted inferential analysis is a time-to-event comparison summarized with a hazard ratio. The distinction matters because an incidence rate describes event occurrence per unit of observed person-time, whereas a hazard ratio compares modeled instantaneous event rates between treatment groups over time.
6. Statistical Methodology
Stratified log-rank test
The registry reports a stratified log-rank test for the primary endpoint. A log-rank test compares the event experience of treatment groups over the observed follow-up rather than comparing only a single event proportion at a fixed time point.
Stratification allows the comparison to account for the trial's stratification factors. The posted analysis notes describe the corresponding Cox proportional hazards model as adjusted for stratification factors.
Cox proportional hazards model
The posted analyses use a Cox proportional hazards model adjusted for stratification factors to obtain the hazard ratio and its confidence interval. Conceptually, the model compares the instantaneous event rate between groups while accounting for the prespecified stratification structure.
Equivalently, 1 − 0.314 = 0.686, so the estimated hazard is approximately 68.6% lower for the comparison represented by the hazard ratio. This is a relative time-to-event interpretation, not a statement that 68.6% of patients avoided bleeding.
Time-to-event analysis
Because the primary endpoint concerns the time until the first bleeding event, the analysis can incorporate differing follow-up durations and censoring. A participant who has not experienced the endpoint contributes information until the end of the relevant observation period rather than being treated as though the participant had been observed indefinitely.
Covariate adjustment and stratification
The analysis notes state that the Cox proportional hazards model was adjusted for stratification factors. The posted analysis text also identifies covariate adjustment and stratified analysis as concepts used in the analysis.
Superiority framework
The posted hypothesis type is superiority for both primary comparisons. The statistical question is therefore whether the observed time-to-event experience provides evidence that the treatment groups differ, rather than whether one treatment remains within a prespecified non-inferiority margin.
7. Primary Results
The registry contains formal statistical analyses for both primary comparisons. Both compare an abelacimab dose group with rivaroxaban for the first occurrence of the composite of ISTH-defined major bleeding or CRNM bleeding.
Abelacimab 90 mg vs Rivaroxaban
Hazard ratio for first composite bleeding event
95% CI: 0.192–0.513 · P < 0.001
Stratified log-rank test with a Cox proportional hazards model adjusted for stratification factors.
The estimated hazard ratio of 0.314 means that, within the posted Cox model and analyzed follow-up, the estimated instantaneous hazard of the first composite bleeding event for the abelacimab 90 mg group was 31.4% of the rivaroxaban hazard. Expressed as a relative reduction, this corresponds to an estimated hazard approximately 68.6% lower than the comparator.
The hazard ratio does not mean that 68.6% of participants avoided bleeding, that the probability of bleeding was reduced by exactly 68.6%, or that every participant experienced the same treatment effect. It is a relative time-to-event measure derived from the fitted survival model.
The 95% confidence interval of 0.192–0.513 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of effects among individual participants.
The P-value of < 0.001 addresses evidence against the tested null hypothesis under the specified statistical framework. It does not measure the magnitude of the treatment effect. Effect magnitude is better conveyed by the hazard ratio and its confidence interval.
Because the analysis uses a Cox proportional hazards model, interpretation of a single hazard ratio also depends on the proportional-hazards framework. The registry does not report a separate assessment of that assumption in the information summarized here.
Abelacimab 150 mg vs Rivaroxaban
Hazard ratio for first composite bleeding event
95% CI: 0.243–0.602 · P < 0.001
Stratified log-rank test with a Cox proportional hazards model adjusted for stratification factors.
The estimated hazard ratio of 0.382 means that, within the posted Cox model and analyzed follow-up, the estimated instantaneous hazard of the first composite bleeding event for the abelacimab 150 mg group was 38.2% of the rivaroxaban hazard. Expressed as a relative reduction, this corresponds to an estimated hazard approximately 61.8% lower than the comparator.
Again, the hazard ratio is not an absolute risk reduction and does not imply that 61.8% of participants were protected from the endpoint. It summarizes a relative difference in the modeled event hazard over time.
The 95% confidence interval of 0.243–0.602 quantifies uncertainty around the estimated relative hazard. The interval remains below 1, which is consistent with the direction of the point estimate.
The P-value of < 0.001 describes the strength of statistical evidence against the relevant null hypothesis under the posted analysis. It is not an effect-size measure and should not be interpreted as saying that the treatment effect is "larger" than an effect associated with a different P-value.
As with the 90 mg comparison, interpretation of the Cox hazard ratio depends on the proportional-hazards framework and on the censoring and analysis-population definitions used by the registry.
8. Primary Results Side by Side
| Primary comparison | Hazard ratio | 95% CI | P-value | Method |
|---|---|---|---|---|
| Abelacimab 90 mg vs Rivaroxaban | 0.314 | 0.192–0.513 | < 0.001 | Stratified log-rank; Cox model adjusted for stratification factors |
| Abelacimab 150 mg vs Rivaroxaban | 0.382 | 0.243–0.602 | < 0.001 | Stratified log-rank; Cox model adjusted for stratification factors |
The graphic is a conceptual display of the reported hazard ratios relative to the rivaroxaban reference value of 1.000. It is not a Kaplan-Meier curve and does not represent cumulative event probabilities.
9. Secondary Endpoint Results
The registry also contains four posted secondary analyses. These use the same general survival-analysis framework: a stratified log-rank test and a Cox proportional hazards model adjusted for stratification factors.
| Secondary endpoint | Comparison | HR | 95% CI | P-value |
|---|---|---|---|---|
| Incidence Rate of the First Occurrence of ISTH-defined Major Bleeding Events | Abelacimab 90 mg vs Rivaroxaban | 0.263 | 0.121–0.572 | < 0.001 |
| Incidence Rate of the First Occurrence of ISTH-defined Major Bleeding Events | Abelacimab 150 mg vs Rivaroxaban | 0.325 | 0.159–0.663 | 0.001 |
| Incidence Rate of the First Occurrence of ISTH-defined Major or CRNM or Minor Bleeding Events | Abelacimab 90 mg vs Rivaroxaban | 0.460 | 0.332–0.638 | < 0.001 |
| Incidence Rate of the First Occurrence of ISTH-defined Major or CRNM or Minor Bleeding Events | Abelacimab 150 mg vs Rivaroxaban | 0.683 | 0.512–0.912 | 0.010 |
Major bleeding alone
90 mg comparison
The hazard ratio for first ISTH-defined major bleeding was 0.263, with a 95% CI of 0.121–0.572 and P < 0.001.
150 mg comparison
The hazard ratio for first ISTH-defined major bleeding was 0.325, with a 95% CI of 0.159–0.663 and P = 0.001.
These analyses narrow the composite endpoint to ISTH-defined major bleeding. The estimates remain below 1 in both posted comparisons, while the confidence intervals quantify the uncertainty associated with those estimates.
Major, CRNM, or minor bleeding
90 mg comparison
The hazard ratio for first ISTH-defined major, CRNM, or minor bleeding was 0.460, with a 95% CI of 0.332–0.638 and P < 0.001.
150 mg comparison
The hazard ratio for first ISTH-defined major, CRNM, or minor bleeding was 0.683, with a 95% CI of 0.512–0.912 and P = 0.010.
The broader bleeding endpoint includes minor bleeding in addition to major and CRNM bleeding. Its hazard ratios are therefore not expected to be numerically identical to those for the narrower major-bleeding endpoint.
10. Safety: Serious Adverse Events
The registry reports serious adverse events by treatment arm as affected participants divided by participants at risk.
| Treatment arm | Participants with serious adverse event | Participants at risk |
|---|---|---|
| Abelacimab 90 mg (MAA868) | 158 | 425 |
| Abelacimab 150 mg (MAA868) | 157 | 427 |
| Rivaroxaban | 167 | 428 |
The serious-adverse-event counts are a safety measure distinct from the primary bleeding endpoint. A serious adverse event may or may not be one of the bleeding events defining the primary efficacy endpoint, so the two sets of results should not be treated as interchangeable.
11. Statistical Methods Explained
Why was a stratified log-rank test used?
The primary endpoint is a time-to-event outcome, so the comparison needs to account for both whether an event occurred and when it occurred. The log-rank test compares the survival experience between groups across follow-up. Stratification incorporates the trial's stratification structure into that comparison.
What does a hazard ratio of 0.314 mean?
A hazard ratio of 0.314 indicates that the estimated instantaneous event hazard for the abelacimab 90 mg group was 31.4% of the rivaroxaban hazard under the fitted model. The complementary interpretation is an estimated 68.6% lower hazard. It does not mean a 68.6% reduction in cumulative probability for every patient.
Why is the confidence interval important?
The point estimate is only one estimate of the underlying treatment effect. The 95% confidence interval provides a range reflecting statistical uncertainty under the model and study design. For the 90 mg comparison, the interval is 0.192–0.513; for the 150 mg comparison, it is 0.243–0.602.
Why does the P-value not measure effect size?
A P-value describes the compatibility of the observed data with a null hypothesis under the specified statistical test. It depends on the amount of information as well as the observed difference. A treatment effect should therefore be described using its effect estimate and confidence interval rather than by the P-value alone.
Why is the analysis population important?
The registry specifies that the posted analyses use randomized participants who received at least one dose of study treatment. This differs conceptually from simply referring to everyone assigned to an arm. The distinction is important when interpreting efficacy and safety because treatment exposure enters the definition of the posted analysis population.
Why use a time-to-event endpoint instead of only counting events?
Participants can have different amounts of observable follow-up. A time-to-event analysis uses the timing of events and handles censoring, allowing participants to contribute information until the end of their observed treatment period or until the first event. This is particularly relevant when the registry expresses incidence using person-years.
What does "superiority" mean here?
The registry identifies the posted primary hypotheses as superiority hypotheses. Thus, the inferential framework is aimed at demonstrating a difference between the treatment groups rather than demonstrating that an experimental treatment is not unacceptably worse than a comparator according to a predefined non-inferiority margin.
12. Kaplan-Meier and Censoring Concepts
The primary endpoint is a time-to-first-event endpoint assessed repeatedly from randomization through the end of the randomized phase, up to 33 months. Kaplan-Meier methods are commonly used to estimate the event-free survival function for this type of endpoint.
Here, di represents the number of events at event time ti, while ni is the number at risk immediately before that time.
The registry's posted results do not include the underlying event-time and censoring records needed to reconstruct a Kaplan-Meier curve. The hazard ratios therefore provide the principal inferential summary available from the posted statistical analyses.
13. Stratification and Covariate Adjustment
The registry identifies the primary method as a stratified log-rank test and states that the corresponding Cox proportional hazards model was adjusted for stratification factors. This is important because the treatment comparison was not simply an unadjusted comparison of two pooled event histories.
Stratified comparison
The log-rank comparison accounts for the trial's stratification structure rather than treating all participants as belonging to one undifferentiated risk set.
Adjusted Cox model
The hazard-ratio estimation incorporates the stratification factors in the Cox proportional hazards framework.
Adjustment can improve the alignment between the inferential analysis and the randomization structure. It does not transform a hazard ratio into an absolute risk difference, nor does it eliminate uncertainty around the estimate.
14. Interpreting the Secondary Analyses
The four secondary analyses extend the primary composite endpoint in two directions: one examines ISTH-defined major bleeding alone, while the other broadens the event definition to major, CRNM, or minor bleeding.
| Endpoint scope | 90 mg HR | 150 mg HR |
|---|---|---|
| Major or CRNM bleeding | 0.314 | 0.382 |
| Major bleeding | 0.263 | 0.325 |
| Major, CRNM, or minor bleeding | 0.460 | 0.683 |
The numerical differences across these endpoints illustrate why endpoint definition matters. A narrower endpoint captures a more specific category of clinical events, whereas adding minor bleeding broadens the event definition. The hazard ratio is always interpreted relative to the particular endpoint being analyzed.
The presence of several posted hypothesis tests means that the collection of P-values should not automatically be interpreted as though every result had an independently controlled type I error rate. The registry identifies the hypothesis type and methods for each analysis but does not provide, in the information summarized here, a multiplicity-adjustment scheme that would support treating the six posted analyses as a single formally adjusted family.
15. Early Trial Termination
Early termination is an important design feature when interpreting a randomized trial. It can affect the amount of accumulated follow-up and the amount of information available for estimating treatment effects. The registry records the recommendation date, but the information summarized here does not provide the committee's detailed statistical rationale for that recommendation.
The trial's primary completion date was February 15, 2024. The distinction between an independent data monitoring committee recommendation for early termination and the registry's recorded primary completion date is important when describing the trial timeline.
16. Trial Timeline
Trial start
AZALEA-TIMI 71 began with a randomized, parallel-group design and triple masking.
Early termination recommendation
The independent data monitoring committee recommended early termination of the trial.
Primary completion
The registry records February 15, 2024 as the primary completion date.
Registered endpoint follow-up
The primary and posted secondary bleeding endpoints were assessed at every visit from randomization through the end of the randomized phase, up to 33 months.
17. What the Hazard Ratios Do — and Do Not — Mean
The primary hazard ratio of 0.314 for abelacimab 90 mg versus rivaroxaban means that the fitted model estimated the instantaneous hazard of the first composite bleeding event to be 31.4% of the comparator hazard. The corresponding relative hazard reduction is approximately 68.6%.
The 150 mg comparison produced an HR of 0.382, corresponding to an estimated instantaneous hazard of 38.2% of the comparator hazard and an approximately 61.8% lower estimated hazard.
Neither hazard ratio tells a patient the probability of experiencing a bleeding event. Absolute risk depends on the baseline event rate, follow-up, censoring, and other features of the event-time distribution. A hazard ratio should therefore not be translated directly into a percentage of patients who will or will not experience the event.
The 90 mg estimate has a 95% CI of 0.192–0.513, while the 150 mg estimate has a 95% CI of 0.243–0.602. These intervals communicate uncertainty around the corresponding point estimates and should be considered alongside the hazard ratios rather than replaced by the P-values.
The two primary comparisons both have P < 0.001. This provides the registry-reported measure of statistical evidence under the posted tests. It does not establish the magnitude of benefit and should not be used as a substitute for the confidence interval.
18. Limitations and Interpretation Issues
- Early termination: the independent data monitoring committee recommended early termination on September 14, 2023. Early termination can influence the amount of accumulated information and follow-up available for estimation.
- Analysis population: the posted analyses include randomized participants who received at least one dose of study treatment, rather than simply all participants assigned to an arm.
- Time-to-event assumptions: the hazard ratios come from Cox proportional hazards models. Interpretation of a single hazard ratio depends on the proportional-hazards framework.
- Endpoint definitions: major bleeding, major or CRNM bleeding, and major/CRNM/minor bleeding are distinct endpoints. Their hazard ratios should not be compared as though they represented identical outcomes.
- Multiple analyses: the registry contains two primary and four secondary posted statistical analyses. The information summarized here does not provide a multiplicity-adjustment scheme sufficient to interpret all six P-values as a single adjusted family.
- Limited absolute information: the posted analyses provide hazard ratios and confidence intervals, but the registry information summarized here does not report median event-free times or Kaplan-Meier survival probabilities.
- Limited subgroup information: the registry information summarized here does not report subgroup-specific hazard ratios or forest-plot results.
- Generalizability: the registry record describes a randomized phase 2 trial; applicability to populations or settings outside the enrolled trial population cannot be established from the posted statistical estimates alone.
19. Why This Trial Matters Statistically
AZALEA-TIMI 71 is a useful teaching case because it connects randomized-trial design with a clinically important time-to-event endpoint and demonstrates how the same statistical framework can be applied to nested endpoint definitions.
| Concept | How it appears in AZALEA-TIMI 71 |
|---|---|
| Randomization | Participants were randomized across three parallel arms. |
| Blinding | The trial used triple masking. |
| Time-to-event endpoint | The primary endpoint is the first occurrence of a bleeding composite assessed over follow-up. |
| Person-time | The registered incidence-rate definition uses participants with events per 100 person-years. |
| Stratified log-rank test | Used for the posted primary and secondary time-to-event comparisons. |
| Cox model | Used to estimate hazard ratios with adjustment for stratification factors. |
| Hazard ratio | Provides the relative treatment effect for the time-to-event comparison. |
| Confidence interval | Quantifies uncertainty around each hazard-ratio estimate. |
| Superiority testing | The posted analyses are identified as superiority hypotheses. |
| Endpoint hierarchy | The registry distinguishes one primary composite endpoint from four posted secondary analyses. |
| Early termination | An independent data monitoring committee recommended early termination on September 14, 2023. |
20. A Deeper Statistical Reading of the Endpoint
The primary endpoint combines two categories: ISTH-defined major bleeding and clinically relevant non-major bleeding. The analysis focuses on the first occurrence of the composite. This means that the posted primary analysis does not simply count every bleeding episode accumulated by a participant; it identifies the first qualifying event for the time-to-event comparison.
This distinction affects interpretation. A participant who experiences multiple bleeding events is represented by the first qualifying event in the primary time-to-first-event analysis. The resulting hazard ratio therefore addresses the relative rate of experiencing a first event, not the total burden of all bleeding episodes.
The use of person-years in the registry's incidence-rate definition provides a complementary way to express event occurrence. Participants contribute exposure time until their first event or the end of treatment during the randomized phase, so participants with different lengths of observation can contribute different amounts of person-time.
This structure explains why the primary endpoint is naturally analyzed using survival methods rather than only a simple comparison of crude event proportions.
21. Primary vs Secondary Endpoint Interpretation
| Endpoint family | Clinical scope | 90 mg HR | 150 mg HR |
|---|---|---|---|
| Primary composite | Major or CRNM bleeding | 0.314 | 0.382 |
| Secondary | Major bleeding | 0.263 | 0.325 |
| Secondary | Major, CRNM, or minor bleeding | 0.460 | 0.683 |
These estimates should be interpreted within their respective endpoint definitions. A hazard ratio of 0.263 for major bleeding is not a contradiction of a hazard ratio of 0.460 for the broader major/CRNM/minor endpoint. They answer different questions because the event definitions differ.
The same principle applies when comparing the 90 mg and 150 mg results. The posted analyses are separate comparisons against rivaroxaban. Their numerical hazard ratios do not constitute a formal direct statistical comparison of the two abelacimab doses unless such a comparison is explicitly specified and analyzed.
22. Related Tutorials
Learn more about the statistical methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: NCT04755283 — AZALEA-TIMI 71.
- PubMed: PMID 39842011.
- PubMed: PMID 42213637.
- PubMed: PMID 41637102.
- PubMed: PMID 40546068.
Continue through Clinical Biostats
Connect the trial's endpoints and statistical methods to deeper tutorials, calculators, and additional clinical-trial analyses.
25. Record Summary
AZALEA-TIMI 71 provides a useful example of a randomized time-to-event analysis in cardiovascular research. The primary endpoint is the first occurrence of ISTH-defined major or CRNM bleeding, assessed repeatedly through the end of the randomized phase for up to 33 months. The registry reports two primary comparisons: abelacimab 90 mg versus rivaroxaban with HR 0.314 (95% CI 0.192–0.513; P < 0.001), and abelacimab 150 mg versus rivaroxaban with HR 0.382 (95% CI 0.243–0.602; P < 0.001).
The same statistical framework was used for secondary analyses of major bleeding and of major, CRNM, or minor bleeding. The interpretation of these results depends on understanding the distinction between a hazard ratio and an absolute event probability, the role of censoring and follow-up time, the analysis population, the stratified log-rank test, and the adjusted Cox proportional hazards model.
The trial also illustrates why design features belong in statistical interpretation. Triple masking and randomization support the internal structure of the comparison, while the independent data monitoring committee's recommendation for early termination on September 14, 2023 is an important feature of the trial history. Taken together, the registry results are best understood as a set of prespecified time-to-event comparisons with effect estimates, uncertainty intervals, and corresponding hypothesis tests rather than as isolated P-values.