← Clinical Trials
Atrial Fibrillation Phase 3 Randomized NCT01830543

PIONEER AF-PCI: Complete Statistical Analysis of Rivaroxaban in Atrial Fibrillation Undergoing PCI

An independent statistical review of the randomized phase 3 PIONEER AF-PCI trial evaluating two rivaroxaban-based strategies against vitamin K antagonist therapy in patients with atrial fibrillation who undergo percutaneous coronary intervention.

Trial status: Completed  ·  Enrollment: 2124  ·  Primary completion: July 28, 2016
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical results on this page are restricted to the ClinicalTrials.gov record.

1. Trial at a Glance

PIONEER AF-PCI was a randomized, parallel, open-label phase 3 trial enrolling 2124 participants with atrial fibrillation who underwent percutaneous coronary intervention. The trial compared two rivaroxaban-based strategies with a vitamin K antagonist strategy, with clinically significant bleeding as the registered primary endpoint through Month 12.

2124
Enrolled
Randomized trial
3
Arms
Parallel design
0.59
Primary HR
Rivaroxaban 15 mg vs VKA
0.63
Primary HR
Rivaroxaban strategy vs VKA
FeaturePIONEER AF-PCI
PhasePhase 3
ConditionsAtrial Fibrillation; Percutaneous Coronary Intervention
DesignRandomized, parallel, unmasked
AllocationRandomized
Primary purposeTreatment
Enrollment2124
Registered primary endpointPercentage of Participants With Clinically Significant Bleeding
Primary endpoint time frameUp to Month 12
Registered endpoint typeBinary
Posted statistical analyses18
Posted outcome measures9
Trial statusCompleted

2. Clinical Question

The statistical question was whether either rivaroxaban-based treatment strategy differed from vitamin K antagonist therapy in the time to clinically significant bleeding during the registered follow-up period of up to Month 12.

Population

Participants with atrial fibrillation who undergo percutaneous coronary intervention.

Intervention strategies

Rivaroxaban 15 mg, or rivaroxaban 2.5 mg twice daily/15 mg once daily, within the trial's three-arm design.

Comparator

Vitamin K antagonist (VKA) therapy.

Primary question

How did the rivaroxaban-based strategies compare with VKA therapy for clinically significant bleeding up to Month 12?

3. Trial Design

01
Randomize2124 participants
02
3 strategiesTwo rivaroxaban-based strategies and VKA
03
DAPTDuration selected as 1, 6 or 12 months
04
Follow-upBleeding and cardiovascular events
05
AnalysisTime-to-event comparisons
RIVAROXABAN STRATEGY

Rivaroxaban 15 mg

  • Rivaroxaban 15 mg
  • One of the two rivaroxaban-based randomized strategies
  • Compared directly with the VKA strategy in the posted analyses
RIVAROXABAN STRATEGY

Rivaroxaban 2.5 mg BID / 15 mg QD

  • Rivaroxaban 2.5 mg twice daily/15 mg once daily
  • Second rivaroxaban-based randomized strategy
  • Compared directly with the VKA strategy in the posted analyses
COMBINATION THERAPY COMPONENTS

Antiplatelet therapy

  • Aspirin (ASA)
  • Clopidogrel
  • Prasugrel
  • Ticagrelor
COMPARATOR

Vitamin K antagonist

  • Vitamin K antagonist (VKA)
  • Randomized comparator strategy
Important design qualification: the registry states that stratification to 1, 6 or 12 months of dual antiplatelet therapy (DAPT) was based upon clinician choice and was not randomized. This distinction matters because clinician-selected treatment duration is not protected by randomization in the same way as the main randomized treatment assignment.

4. Randomization, Treatment Structure, and Analysis Population

The trial was randomized and parallel, with no masking. The ClinicalTrials.gov record identifies three arms and 2124 enrolled participants. The posted primary and secondary statistical analyses used a safety analysis set defined as all randomized participants who received at least 1 dose of study drug.

ElementRegistry information
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Arms3
Safety analysis setAll randomized participants who received at least 1 dose of study drug
Endpoint analysis frameworkTime-to-event analysis using the log-rank test and hazard ratio

The analysis population is important because the primary posted estimates are not described as an all-randomized intention-to-treat analysis. Instead, the registry explicitly defines the safety analysis set as randomized participants who received at least one dose of study drug. That population definition should remain attached to the reported hazard ratios when interpreting them.

5. Primary Endpoint

The registered primary endpoint was Percentage of Participants With Clinically Significant Bleeding, with a time frame of Up to Month 12. Although the registry classifies the registered endpoint type as binary, the posted statistical analyses treat the endpoint as a time-to-event outcome and use log-rank testing with hazard ratios.

EndpointRegistry definition / analysis
Clinically significant bleedingComposite of TIMI major bleeding, minor bleeding, and bleeding requiring medical attention (BRMA)
TIMI major bleedingAny symptomatic intracranial hemorrhage, or clinically overt signs of hemorrhage, including imaging, associated with a drop in hemoglobin of ≥5 grams per deciliter (g/dL) (or when the hemoglobin concentration is not available, an absolute drop in hematocrit of >=15 percent (%)). TIMI minor bleeding event is defined as any clinically overt sign of hemorrhage (including imaging) that is associated with a fall in hemoglobin concentration of 3 to less than (\<) 5 g/dL (or, when hemoglobin concentration is not available, a fall in hematocrit of 9 percent to \<15 percent). A BRMA event is defined as any bleeding event that requires medical treatment, surgical treatment, or laboratory evaluation, and does not meet criteria for a major or minor bleeding event.
Time frameUp to Month 12
Registered endpoint typeBinary
Posted analysis typeTime-to-event
Statistical methodLog-rank test
Effect measureHazard ratio

The composite construction is statistically important. A participant can experience clinically significant bleeding by meeting one of several component definitions. The composite therefore summarizes time to the first qualifying event under the endpoint definition rather than representing a single biological event type.

6. Primary Results

Rivaroxaban 15 mg vs Vitamin K Antagonist

Hazard ratio for clinically significant bleeding

0.59

95% CI: 0.47–0.76   ·   P < 0.001

Time frame: Up to Month 12  ·  Two-sided 95% confidence interval

The posted analysis compared rivaroxaban 15 mg with VKA using the log-rank test in the safety analysis set. The estimated hazard ratio was 0.59, with a two-sided 95% confidence interval from 0.47 to 0.76 and P < 0.001.

Clinical Biostats interpretation

An HR of 0.59 means that, within the time-to-event model used for this comparison, the estimated instantaneous hazard of clinically significant bleeding was approximately 59% as high in the rivaroxaban 15 mg group as in the VKA group. Expressed as a relative model-based quantity, this corresponds to an estimated 41% lower hazard.

It does not mean that 41% of participants avoided bleeding, that the absolute probability of bleeding was reduced by 41 percentage points, or that every participant experienced the same relative reduction.

The 95% CI of 0.47–0.76 describes statistical uncertainty around the estimated hazard ratio. Because the entire interval is below 1, the interval is consistent with a lower estimated hazard in the rivaroxaban 15 mg group under this analysis.

The P < 0.001 value addresses the statistical evidence against the tested null comparison under the analysis framework. A p-value does not measure the magnitude of the treatment effect and should not be read as the probability that the treatment effect is real.

The interpretation also depends on censoring, the analysis population, and the assumptions underlying a hazard-ratio representation of the event process. The registry reports a hazard ratio but does not provide enough information in the ClinicalTrials.gov record to independently assess the proportional-hazards assumption.

Rivaroxaban 2.5 mg BID / 15 mg QD vs Vitamin K Antagonist

Hazard ratio for clinically significant bleeding

0.63

95% CI: 0.5–0.8   ·   P < 0.001

Time frame: Up to Month 12  ·  Two-sided 95% confidence interval

The second primary comparison evaluated the rivaroxaban 2.5 mg twice daily/15 mg once daily strategy against VKA. The log-rank analysis produced an estimated hazard ratio of 0.63, with a two-sided 95% confidence interval from 0.5 to 0.8 and P < 0.001.

Clinical Biostats interpretation

An HR of 0.63 indicates an estimated instantaneous hazard of clinically significant bleeding approximately 63% as high in the rivaroxaban strategy as in the VKA group. As a relative model-based interpretation, that corresponds to an estimated 37% lower hazard.

This is a relative time-to-event measure, not an absolute risk reduction. It does not tell us directly how many fewer participants experienced clinically significant bleeding, nor does it imply that the same percentage reduction applied to every individual.

The 95% CI of 0.5–0.8 gives a range of statistical uncertainty around the estimated hazard ratio. Its endpoints remain below 1, so the reported interval is compatible with a lower estimated hazard under the specified comparison.

The P < 0.001 result provides evidence against the null hypothesis used in the superiority analysis, but it does not quantify clinical importance or effect size. The hazard ratio and its confidence interval provide that effect-size information.

As with the first primary comparison, interpretation requires attention to the safety analysis population and the time-to-event framework. The ClinicalTrials.gov record does not provide a reconstructed Kaplan-Meier curve or enough individual-level event and censoring information to independently reproduce the estimate.

7. Secondary Bleeding Results

The registry contains additional time-to-event analyses for TIMI major bleeding, TIMI minor bleeding, and bleeding requiring medical attention. These analyses use the same safety analysis set and the log-rank test, with hazard ratios and two-sided 95% confidence intervals.

TIMI Major Bleeding

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA0.660.33–1.310.234
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA0.570.28–1.160.114

Both estimates are below 1, but the corresponding confidence intervals include 1. The registry reports P = 0.234 for the first comparison and P = 0.114 for the second. These are superiority analyses, and the ClinicalTrials.gov record does not provide evidence that should be interpreted independently of the broader multiplicity and precision considerations of a trial containing multiple endpoint comparisons.

TIMI Minor Bleeding

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA0.510.2–1.280.144
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA0.50.2–1.260.134

The estimated hazards were approximately one-half of the VKA comparison hazard for both rivaroxaban strategies, but the confidence intervals are wide and include 1. The width of these intervals is an important part of the interpretation: a point estimate alone can give a false impression of precision when relatively uncommon events are being analyzed.

Bleeding Requiring Medical Attention

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA0.610.47–0.8<0.001
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA0.670.52–0.860.002

For bleeding requiring medical attention, both posted estimates are below 1 and both confidence intervals exclude 1. The reported P-values are <0.001 and 0.002, respectively. These analyses are secondary endpoints rather than the registered primary endpoint, so their statistical interpretation should remain distinct from the primary hypothesis tests.

8. Cardiovascular Event Results

The registry also reports time-to-event analyses of a composite of adverse cardiovascular events, cardiovascular death, myocardial infarction, stroke, and stent thrombosis. These estimates are particularly useful for illustrating why a trial can show different statistical behavior across bleeding and cardiovascular endpoints.

Composite of Cardiovascular Death, Myocardial Infarction, and Stroke

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA1.080.69–1.680.75
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA0.930.59–1.480.765

The first comparison has an estimated HR above 1, whereas the second has an estimate below 1. Both confidence intervals include 1. The estimates therefore do not establish a statistically distinguishable difference in the composite endpoint under these posted superiority analyses.

Reading an HR near 1

An HR of 1.08 should not be interpreted as proof that the treatment has an 8% higher absolute cardiovascular risk. It is a relative hazard estimate, and its 95% CI of 0.69–1.68 spans values below and above 1. The uncertainty interval is therefore essential to understanding the estimate.

Similarly, an HR of 0.93 is close to 1, but the appropriate interpretation is not simply "no effect." The confidence interval of 0.59–1.48 indicates that the data are compatible with a range of relative hazard values under the model.

Cardiovascular Death

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA1.290.59–2.80.523
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA1.190.54–2.620.664

These cardiovascular-death estimates have substantially wide confidence intervals relative to the point estimates. That is a reminder that precision depends on the amount of information available for a particular endpoint, not merely on the total randomized sample size.

Myocardial Infarction

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA0.860.46–1.590.625
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA0.750.4–1.420.374

Stroke

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA1.070.39–2.960.891
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA1.360.52–3.580.53

Stent Thrombosis

ComparisonHR95% CIP-value
Rivaroxaban 15 mg vs VKA1.20.32–4.450.79
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA1.440.4–5.090.574

The stent-thrombosis estimates illustrate an especially important statistical principle: a wide confidence interval can contain both a potentially lower and a potentially higher hazard. A nonsignificant P-value does not demonstrate equivalence, and a point estimate above 1 does not by itself establish harm.

9. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized treatment group using affected participants over participants at risk. These counts should be interpreted as safety summaries rather than as the same time-to-event estimands used for the primary bleeding analyses.

Randomized strategySerious adverse eventsAt risk
Rivaroxaban 15 mg237696
Rivaroxaban 2.5 mg BID / 15 mg225706
Vitamin K Antagonist (VKA)271697

The registry reports these figures as 237/696, 225/706, and 271/697, respectively. They are descriptive arm-level safety counts. They should not be converted here into new percentages or compared through a newly calculated statistical test because the ClinicalTrials.gov record does not report such an analysis.

Why this distinction matters: the primary endpoint is analyzed as a time-to-event outcome, whereas the serious-adverse-event figures reported here are reported as affected/at-risk counts. Treating every safety number as though it were generated by the same statistical model would obscure rather than clarify the trial's analysis structure.

10. Statistical Methodology

Time-to-event analysis

The primary and posted secondary analyses treat clinically significant bleeding and the other listed outcomes as time-to-event endpoints. Instead of considering only whether an event occurred by Month 12, the analysis uses information about when the event occurred and accommodates participants whose event status is censored during follow-up.

Conceptual survival function
S(t) = P(T > t)

Here T represents the time to the event of interest. The survival function is the probability of remaining event-free beyond time t.

This framework is especially useful when participants have different follow-up times. A binary endpoint discards the timing information; a time-to-event analysis retains it and can incorporate right-censored observations.

Log-rank test

The registry identifies the log-rank test as the reported method for all 18 posted statistical analyses. Conceptually, the log-rank test compares the observed and expected numbers of events between treatment groups across event times.

Core idea
Observed events vs expected events over follow-up

The comparison is based on the ordering of event times and the numbers at risk, rather than simply comparing final percentages at a single time point.

Because the analysis uses the full follow-up structure, the log-rank test is not equivalent to a simple chi-square comparison of the percentage of participants with an event by Month 12.

Hazard ratio

The effect measure reported for every registry-reported statistical analysis is the hazard ratio. An HR below 1 indicates a lower estimated instantaneous event hazard in the numerator treatment group relative to the comparator; an HR above 1 indicates a higher estimated instantaneous event hazard.

Interpretive rule
HR = hazard in treatment group / hazard in comparator group

The hazard ratio is a relative time-to-event measure. It is not a relative risk, an absolute risk difference, or a probability that an individual patient will experience an event.

Confidence intervals

The reported primary analyses use two-sided 95% confidence intervals. The interval describes uncertainty around the estimated hazard ratio under the statistical model and sampling framework. It does not describe the range of effects that individual participants experienced.

Superiority testing

The analyses posted on ClinicalTrials.gov are identified as superiority hypotheses. That means the statistical question is whether the treatment groups differ under the prespecified superiority framework, rather than whether one strategy meets a predefined non-inferiority margin.

Safety analysis population

The posted analyses use the safety analysis set: all randomized participants who received at least 1 dose of study drug. This population definition is part of the estimand context and should be reported alongside the hazard ratios rather than treated as a minor technical detail.

11. Statistical Methods Explained

Why was a log-rank test used?

Because the registry analyses treat the outcomes as time-to-event endpoints. A log-rank test uses information about event timing and numbers at risk throughout follow-up, making it appropriate for comparing event-time distributions between randomized groups.

What does an HR of 0.59 mean?

An HR of 0.59 means the estimated instantaneous hazard in the rivaroxaban 15 mg group was 0.59 times the corresponding hazard in the VKA group under the fitted time-to-event framework. It can be expressed as an estimated 41% lower hazard, but it should not be translated into a 41% absolute reduction in the number of participants experiencing bleeding.

Why is the confidence interval as important as the point estimate?

The HR is only one estimate from the observed data. The 95% CI communicates how precisely that effect was estimated. For example, the primary HR of 0.59 has a 95% CI of 0.47–0.76, whereas the TIMI major bleeding HR of 0.66 has a much wider 95% CI of 0.33–1.31. The latter point estimate alone would conceal substantial uncertainty.

Why doesn't P < 0.001 measure effect size?

A p-value addresses evidence against a null hypothesis within a specified statistical framework. It does not say whether an effect is large or small. Effect size is described by the hazard ratio, while its precision is described by the confidence interval. Sample size and event information can also influence the p-value.

Why can an HR above 1 still be statistically inconclusive?

Consider the cardiovascular-death HR of 1.29 for rivaroxaban 15 mg versus VKA. Its 95% CI is 0.59–2.8 and P = 0.523. Because the interval includes 1, the data do not establish a statistically distinguishable difference under this superiority analysis. The point estimate alone is insufficient.

Why does the registered endpoint type say binary when the posted analysis is time-to-event?

The ClinicalTrials.gov record classifies the registered primary endpoint as Binary while separately reporting that the statistical analyses use a time-to-event framework. These are different descriptions of the endpoint: one concerns the registry's endpoint classification, while the posted analysis incorporates event timing and uses a log-rank test with a hazard ratio.

12. Interpreting the Primary Results Together

Primary comparisonHR95% CIP-valueStatistical reading
Rivaroxaban 15 mg vs VKA0.590.47–0.76<0.001Estimated hazard below VKA; CI excludes 1
Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA0.630.5–0.8<0.001Estimated hazard below VKA; CI excludes 1

The two primary estimates are directionally similar: both are below 1, both have two-sided 95% confidence intervals entirely below 1, and both have P < 0.001. The statistical interpretation is therefore internally consistent across the two posted primary comparisons.

That consistency should not be expanded into claims about endpoints that were not statistically significant. For example, the cardiovascular composite analyses produced HRs of 1.08 and 0.93 with confidence intervals that include 1. The primary bleeding result and the cardiovascular secondary analyses answer different questions and should be kept separate.

Relative effect is not absolute effect

The primary HRs summarize relative event hazards. Without event-free survival estimates, cumulative incidence values, or the underlying participant-level event and censoring information, the ClinicalTrials.gov record does not support calculation of absolute risk differences or numbers needed to treat.

The comparison is not randomized DAPT duration

The registry specifically states that assignment to 1, 6, or 12 months of DAPT was based on clinician choice and was not randomized. Consequently, that treatment-duration decision should not be interpreted as though it were another randomized comparison embedded within the trial.

13. Multiplicity and Multiple Endpoint Comparisons

The ClinicalTrials.gov record contains 18 statistical analyses across the primary and secondary endpoints. The primary endpoint has two posted superiority comparisons against VKA, while multiple secondary bleeding and cardiovascular outcomes were also analyzed.

Analysis familyNumber represented in the ClinicalTrials.gov recordInterpretive role
Primary endpoint comparisons2Primary superiority comparisons
TIMI major bleeding2Secondary endpoint comparisons
TIMI minor bleeding2Secondary endpoint comparisons
BRMA2Secondary endpoint comparisons
Cardiovascular death / MI / stroke composite2Secondary endpoint comparisons
Cardiovascular death2Secondary endpoint comparisons
Myocardial infarction2Secondary endpoint comparisons
Stroke2Secondary endpoint comparisons
Stent thrombosis2Secondary endpoint comparisons

When many statistical comparisons are performed, the interpretation of individual p-values depends on the prespecified multiplicity strategy. The ClinicalTrials.gov record identifies the analyses as superiority tests but do not provide an alpha-allocation or multiplicity-adjustment procedure. Therefore, the reported p-values should be presented as reported rather than retrospectively treated as though a specific multiplicity correction had been documented in the ClinicalTrials.gov record.

Do not count significant secondary p-values as independent primary findings. The fact that BRMA has P < 0.001 and P = 0.002 in the two comparisons does not by itself establish a separate confirmatory claim unless the statistical hierarchy and multiplicity procedure support that interpretation.

14. Crossover, Treatment Duration, and Causal Interpretation

The registry caveat identifies an important nonrandomized component of the trial: the duration of DAPT at 1, 6, or 12 months was determined by clinician choice. This creates an important distinction between the randomized treatment strategy and treatment-duration decisions made within the clinical setting.

Randomized component

The main trial allocation was randomized across three treatment strategies.

Nonrandomized component

DAPT duration of 1, 6 or 12 months was based upon clinician choice and was not randomized.

Statistical implication

Associations involving clinician-selected DAPT duration cannot automatically be given the causal interpretation of the randomized treatment comparison.

Interpretive boundary

The ClinicalTrials.gov record does not provide a separate randomized analysis of DAPT duration.

The registry also notes that there was not sufficient power to definitively assess efficacy because of the modest number of adverse cardiovascular events. This is a precision issue: a trial can have a substantial randomized enrollment while still having limited information for relatively uncommon cardiovascular outcomes.

15. Why Event Counts Matter More Than Enrollment Alone

The PIONEER AF-PCI data illustrate an important distinction in clinical-trial statistics: the number randomized is not the same as the amount of information available for every endpoint.

There were 2124 enrolled participants, but cardiovascular death, myocardial infarction, stroke, and stent thrombosis are individual event categories with comparatively uncertain estimates in the analyses posted on ClinicalTrials.gov. The resulting confidence intervals are notably broad for several of these outcomes.

EndpointRivaroxaban 15 mg vs VKARivaroxaban 2.5 mg BID / 15 mg QD vs VKA
Primary clinically significant bleeding0.59 (0.47–0.76)0.63 (0.5–0.8)
TIMI major bleeding0.66 (0.33–1.31)0.57 (0.28–1.16)
Cardiovascular death1.29 (0.59–2.8)1.19 (0.54–2.62)
Stroke1.07 (0.39–2.96)1.36 (0.52–3.58)
Stent thrombosis1.2 (0.32–4.45)1.44 (0.4–5.09)

The contrast is educational. The primary bleeding estimates are comparatively precise, whereas several individual cardiovascular outcomes have wide intervals. The registry's own caveat that there was not sufficient power to definitively assess efficacy due to the modest number of adverse cardiovascular events is therefore directly relevant to interpretation.

16. What the Hazard Ratio Does — and Does Not — Mean in PIONEER AF-PCI

Example: HR 0.59

The primary HR of 0.59 is a relative estimate of the event hazard for clinically significant bleeding. It means the estimated instantaneous hazard was about 59% of that in the VKA comparator under the reported time-to-event analysis.

It does not mean that 59% of participants bled, that 41% of participants were protected from bleeding, or that the absolute probability of bleeding fell by 41 percentage points.

Confidence interval

The 95% CI of 0.47–0.76 expresses uncertainty around the HR estimate. The interval does not describe individual patient responses. It also does not mean that there is a 95% probability that the true HR lies inside this particular interval under every interpretation of probability.

P-value

The primary P < 0.001 is evidence against the null hypothesis in the reported superiority analysis. It does not measure clinical importance, does not measure the probability that the treatment is effective, and should not replace the HR and confidence interval when communicating the magnitude and precision of the result.

Proportional-hazards caution

A single HR is easiest to interpret as a common relative hazard over time when the proportional-hazards assumption is reasonable. The ClinicalTrials.gov record reports the HR but do not provide enough information to independently test that assumption. A hazard ratio should therefore be understood as a model-based summary rather than a universal percentage reduction applying identically at every time point.

17. Limitations

18. Why This Trial Matters Statistically

PIONEER AF-PCI is a useful statistical teaching case because it combines randomized treatment allocation, multiple treatment strategies, a composite bleeding endpoint, time-to-event analysis, hazard ratios, secondary cardiovascular outcomes, safety-population definitions, and a clinically selected component of treatment duration.

ConceptHow it appears in PIONEER AF-PCI
RandomizationThree-arm randomized parallel design
Time-to-event endpointsPrimary and secondary posted analyses use event timing through follow-up
Log-rank testReported method for all registry-reported statistical analyses
Hazard ratioEffect measure for every registry-reported statistical analysis
Confidence intervalTwo-sided 95% intervals reported for the primary and secondary comparisons
Composite endpointClinically significant bleeding combines three bleeding categories
Superiority testingPrimary and secondary analyses are identified as superiority hypotheses
Safety populationRandomized participants receiving at least 1 dose of study drug
Nonrandomized treatment componentDAPT duration was selected by clinician choice
Limited event informationRegistry caveat notes insufficient power for definitive efficacy assessment because of modest cardiovascular event counts
Multiplicity18 statistical analyses span the primary and secondary outcomes

19. Statistical Interpretation of Bleeding vs Cardiovascular Outcomes

One of the most instructive features of the ClinicalTrials.gov record is that the statistical estimates are not uniform across endpoint families. The two primary clinically significant bleeding comparisons have HRs below 1 with confidence intervals excluding 1. Several individual cardiovascular outcomes have confidence intervals that are much wider and include 1.

Bleeding endpoints

The primary HRs are 0.59 and 0.63, with two-sided 95% confidence intervals of 0.47–0.76 and 0.5–0.8, respectively.

Cardiovascular endpoints

Individual cardiovascular outcomes have estimates ranging across values below and above 1, with several broad confidence intervals.

What this demonstrates

A trial's statistical conclusions can differ by endpoint because the underlying event processes and amount of event information differ.

What it does not demonstrate

A nonsignificant secondary endpoint does not by itself establish equivalence, and a significant endpoint does not automatically become a separate primary claim.

The registry caveat provides an explicit reason for caution on efficacy: there was not sufficient power to definitively assess efficacy because of the modest number of adverse cardiovascular events. Statistically, this is consistent with the broad confidence intervals observed for several cardiovascular outcomes.

20. Understanding Censoring in This Trial

Because the posted analyses are time-to-event analyses, participants do not necessarily need to experience an event for their data to contribute information. A participant who remains event-free through the available observation period can contribute follow-up time until censoring.

Conceptual example
Event-free follow-up → contributes time at risk → censored if event status becomes unknown

This is why a time-to-event analysis can use different follow-up durations without simply treating participants with shorter observation as though they had completed the entire study period.

The ClinicalTrials.gov record does not provide individual censoring dates or event histories, so a Kaplan-Meier curve cannot be reconstructed from the registry summaries alone. The appropriate statistical interpretation is therefore based on the reported hazard ratios, confidence intervals, p-values, and documented analysis method.

21. Results in Context of the Registered Endpoint

QuestionAnswer from the ClinicalTrials.gov record
What was the registered primary endpoint?Percentage of Participants With Clinically Significant Bleeding
What was the registered time frame?Up to Month 12
How was clinically significant bleeding defined?Composite of TIMI major bleeding, TIMI minor bleeding, and BRMA
How was the posted endpoint analyzed?Time-to-event analysis using the log-rank test
What effect measure was reported?Hazard ratio
What was the first primary comparison?Rivaroxaban 15 mg vs VKA
What was the second primary comparison?Rivaroxaban 2.5 mg BID / 15 mg QD vs VKA
Were results posted?Yes

This alignment between the registered endpoint and the posted statistical analysis is useful for teaching: the endpoint can be described as the occurrence of a clinically defined composite while the statistical analysis uses the additional dimension of time until the first qualifying event.

22. Longitudinal Trial History

2013-05-10

Trial start

PIONEER AF-PCI began enrollment under the registered randomized phase 3 design.

2016-07-28

Primary completion

The registry lists July 28, 2016 as the primary completion date.

Completed

Results posted

The registry contains 9 posted outcome measures and 18 posted statistical analyses, including 2 primary-endpoint analyses.

23. Important Statistical Distinctions

Hazard vs probability

A hazard ratio compares instantaneous event hazards. It is not the same as comparing cumulative probabilities at Month 12.

Point estimate vs precision

The HR is the point estimate; the 95% CI shows its statistical precision. Both should be read together.

Primary vs secondary

The two clinically significant bleeding analyses are primary; the TIMI and cardiovascular analyses are secondary.

Randomized vs clinician-selected

The three-arm treatment allocation was randomized, whereas DAPT duration was based on clinician choice.

24. Limitations of the Available Registry Results

The ClinicalTrials.gov record is sufficient to describe the reported statistical analyses, but it does not contain individual-level event histories, Kaplan-Meier coordinates, or enough information to reconstruct the full survival curves. It also does not provide a documented multiplicity-adjustment procedure, detailed model diagnostics, or a separate randomized comparison of DAPT duration.

Accordingly, this page deliberately does not manufacture absolute event rates, reconstructed survival curves, additional subgroup analyses, or new statistical tests. Doing so would create numbers that are not present in the ClinicalTrials.gov record and could give the appearance of greater precision than the registry record supports.

Registry-first principle: reported estimates are presented exactly as reported in the registry. Educational interpretation explains what those estimates mean statistically without adding unreported calculations or outcome claims.

25. Related Tutorials

Learn more about the methods used in this trial:

26. Related Calculators

27. Sources

Continue through Clinical Biostats

Build from the statistical concepts used in this trial into deeper tutorials, calculators, and clinical-trial analysis workflows.

28. Record Summary

PIONEER AF-PCI provides a compact teaching example of how randomized clinical-trial data can be analyzed across several related but distinct endpoint families. The registered primary endpoint was clinically significant bleeding through Month 12, defined as a composite of TIMI major bleeding, TIMI minor bleeding, and bleeding requiring medical attention. The posted analyses treated that endpoint as time-to-event and used log-rank testing with hazard ratios. The two primary comparisons produced HRs of 0.59 and 0.63, each with a two-sided 95% confidence interval entirely below 1 and P < 0.001.

The secondary results demonstrate why statistical interpretation cannot stop at the primary p-value. TIMI major and minor bleeding estimates were less precise, while bleeding requiring medical attention showed HRs of 0.61 and 0.67. Cardiovascular outcomes had different estimates and, for several individual endpoints, substantially wider confidence intervals. The registry explicitly cautions that the trial did not have sufficient power to definitively assess efficacy because of the modest number of adverse cardiovascular events.

The trial also illustrates the importance of separating randomized from nonrandomized components. The principal three-arm allocation was randomized, whereas the choice of 1, 6 or 12 months of DAPT was based upon clinician choice. For statistical interpretation, that distinction is essential because randomization is what supports the strongest causal comparison between the randomized treatment strategies.

Clinical Biostats methodology: A trial-results page should not merely repeat reported numbers. The goal is to reconstruct the statistical story of the trial: define the estimand, identify the analysis population and method, report the effect estimate and uncertainty, explain what the result does and does not mean, and preserve the limitations that determine how far the evidence can be interpreted.