Safety and Effectiveness of Apixaban Compared to Warfarin in Patients With Non-valvular Atrial Fibrillation (a Type of Irregular Heart Rhythm) at Higher Chance of Bleeding: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Safety and Effectiveness of Apixaban Compared to Warfarin in Patients With Non-valvular Atrial Fibrillation (a Type of Irregular Heart Rhythm) at Higher Chance of Bleeding (NCT05471505), a registered phase not stated study. The primary outcome is Incidence Rate Per 1000 Participant-Years For First Occurrence of Composite Stroke and Systemic Embolism (SE) Events After Index Date: Balanced Cohorts.

2. Clinical Question

The clinical context is non-valvular atrial fibrillation, an irregular heart rhythm that the registry summary describes as potentially allowing blood to pool in the atria and increasing the chance of blood-clot formation. The summary frames anticoagulation as important for preventing thromboembolism in this population. It also notes that evidence for Japanese patients with higher bleeding risk was limited, motivating an evaluation using real-world data.

The question is comparative, but the design does not randomly assign participants to treatment. Instead, the analysis uses administrative claims and inverse probability treatment weighting (IPTW) to balance participant characteristics among reporting groups. Weighting can address measured characteristics included in the analysis, but the registry record does not report the full set of characteristics used or establish that all relevant differences were controlled.

The comparison therefore concerns observed event rates in weighted groups, not a randomized estimate. No direction or magnitude of the outcome difference can be stated from the registry record because the primary results are not reported there.

3. Trial Design

This completed study is identified as NCT05471505 and is sponsored by Pfizer. The registry lists an enrollment of 120722. The study start date is 2022-08-25, and the primary completion date is 2022-09-27. These study dates are distinct from the longer historical period in which claims data were observed.

The purpose was to compare effectiveness and safety of apixaban and warfarin among Japanese people with non-valvular atrial fibrillation at higher chance of bleeding, using a Japanese nationwide administrative claims database. The primary analysis describes eligible participants registered in the MDV database. The registry says participant characteristics were balanced using inverse probability treatment weighting. It does not report random assignment; allocation is not stated, and the record cautions that the treatment arms are not a like-for-like control comparison.

The primary outcome was the incidence rate per 1000 participant-years for the first occurrence of composite stroke and systemic embolism after the index date. Secondary outcomes covered first ischemic stroke, hemorrhagic stroke, systemic embolism, major intracranial bleeding, and major gastrointestinal bleeding events, also expressed as incidence rates per 1000 participant-years for balanced cohorts.

The described observation window for the primary outcome ran from Mar 2011 to Jun 2021 (approximately 10 years 4 months); extracted data were evaluated in approximately 1 month of this study. The index date was treatment initiation. Follow-up began the next day and ended at the first of the target outcome event, discontinuation of apixaban or warfarin, switching from either drug, or withdrawal from the database.

The registry lists 7 outcome measures and 7 analyses. It does not report detailed eligibility criteria, the exact number analyzed in each outcome analysis, the IPTW implementation, or a full statistical analysis plan.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: Data of eligible participants were extracted from Medical Data Vision Company Limited (MDV Co. Ltd.) database for duration of 01-Mar-2011 to 30-Jun-2021. Extracted data was evaluated for objectives of this study in approximately 1 month of this study.

Overall Study

Analysis population for the primary outcome, as defined in the registry: Eligible participants registered on MDV database, whose data was observed in the study. Analysis performed using Inverse probability treatment weighting (IPTW) method to balance participant characteristics among reporting groups. Here, "Overall Number of Participants Analyzed" is the number of participants after application of IPTW method to raw numbers and is different from the actual participants included in the reporting arm.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: All eligible participants whose data were extracted from database and observed in the study.

6. Primary Endpoint

The registry lists the primary outcome as Incidence Rate Per 1000 Participant-Years For First Occurrence of Composite Stroke and Systemic Embolism (SE) Events After Index Date: Balanced Cohorts, measured over Follow-up period during data observation period from Mar 2011 to Jun 2021 (approximately 10 years 4 months); extracted data evaluated in approximately 1 month of this study.

The primary endpoint is the incidence rate per 1000 participant-years for the first occurrence of a composite of stroke and systemic embolism after the index date. The stroke component includes any ischemic and any hemorrhagic stroke, excluding non-traumatic extradural hemorrhage. The registry describes identifying stroke and systemic embolism through hospital claims with the relevant diagnosis code as the first listed ICD-10 diagnosis code. Events on the index date are excluded.

An incidence rate relates the number of observed events to the amount of time participants contribute under observation. Expressing the result per 1000 participant-years accounts for differences in follow-up duration: participant-years represent accumulated time at risk, rather than simply counting people. This is useful in claims-based follow-up when observation can end at an event, treatment discontinuation, switching, or database withdrawal.

The endpoint focuses on the first recorded event after treatment initiation, rather than counting every recurrence. Combining stroke and systemic embolism yields a composite outcome, allowing the analysis to describe the rate of either type of event under one endpoint. The composite does not, by itself, show which component contributed most to the overall rate; separate secondary outcomes address ischemic stroke, hemorrhagic stroke, and systemic embolism.

The registry identifies a 95% confidence interval as the dispersion measure. A confidence interval communicates statistical uncertainty around an estimated rate, but no primary rate or interval is reported in the registry record. Consequently, the endpoint definition can be described, but its observed magnitude and precision cannot be assessed from the posted record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does an incidence rate represent?

In general, an incidence rate divides the number of events by accumulated time at risk. The primary outcome is expressed per 1000 participant-years, so the denominator represents time contributed by participants rather than the number of participants alone. This form can be useful when observation lengths differ. It is not the same quantity as the percentage of participants who experience an event, and it should not be interpreted as a probability without additional assumptions and a defined time horizon.

The endpoint counts the first qualifying composite event after the index date. The registry describes the composite as stroke or systemic embolism, with stroke comprising ischemic and hemorrhagic stroke and excluding non-traumatic extradural hemorrhage. Claims with the relevant diagnosis code first-listed in the hospital record were used to identify outcomes. These definitions make the endpoint operational in claims data, but a code-based definition is not the same as independent clinical adjudication. Whether adjudication or validation was performed is not reported in the registry record.

How does follow-up affect the rate?

Follow-up starts on the day after treatment initiation and ends at the first occurrence of the target event, discontinuation of apixaban or warfarin, switching from either treatment, or database withdrawal. In general, these stopping rules determine how much time enters the denominator. Participants with shorter observation contribute less time than those followed longer. The study describes its data observation period as Mar 2011 to Jun 2021 (approximately 10 years 4 months), while extracted data were evaluated in approximately 1 month of this study.

The registry does not provide details about how incomplete or unavailable claims were handled, how database withdrawal was distinguished from other loss of observation, or whether additional censoring methods were used. Such details matter because incomplete follow-up can influence both the number of observed events and the amount of person-time.

What is IPTW intended to do?

Inverse probability treatment weighting is a general method for adjusting an observational comparison using measured participant characteristics. Participants are assigned analysis weights related to the estimated probability of receiving their observed treatment. The purpose is to construct weighted groups with a more comparable distribution of recorded characteristics. The registry states that IPTW was used to balance participant characteristics among reporting groups.

Weighting does not turn a non-randomized study into a randomized trial. Its performance depends on which characteristics are measured and included, how treatment probabilities are estimated, and whether the groups have adequate overlap. Extreme weights can make estimates unstable, while omitted or unmeasured factors can leave residual confounding. The variables, model specification, weight handling, overlap assessment, and balance diagnostics are not reported in the registry record. The record also cautions that the arms are not a like-for-like control comparison.

What does a confidence interval convey?

In general, a confidence interval expresses uncertainty around an estimated quantity under the statistical method used. A narrower interval indicates greater statistical precision than a wider interval, while neither width nor location alone resolves potential bias from the observational design. The registry identifies a 95% confidence interval for the primary rate, but gives no estimate or interval limits. Statistical precision therefore cannot be evaluated from the posted results.

The record does not state how the confidence interval was constructed, whether the variance calculation accounted for IPTW, or what model was used to estimate rates. Those methodological details are not reported in the registry record. Without them and without numerical estimates, the interval method and its implications cannot be independently assessed.

Why can over-dispersion matter for rate analyses?

In general, event-count data may vary more than a simple model assumes. This is called over-dispersion. If present and not accommodated, standard errors and confidence intervals can be poorly calibrated. This is a general consideration for rate analyses; whether over-dispersion was assessed or how it was handled in this study is not reported in the registry record. No specific distributional assumption should be inferred from the endpoint description alone.

Likewise, participant follow-up and outcome occurrence may not be statistically independent in ways that a basic rate model assumes. The registry does not describe any clustering, repeated-event modeling, or sensitivity analyses. Because the primary endpoint uses the first occurrence, the stated analysis is about first events rather than a recurrent-event rate.

What can be inferred from the analysis description?

The record identifies the population as eligible participants registered on the MDV database and says the analysis used IPTW. It also gives a total enrollment of 120722, but does not state the number included in the primary analysis after eligibility criteria and analysis procedures. Enrollment should not be treated as the analyzed sample size. Outcome estimates, group-specific person-time, event counts, and confidence interval limits are not reported in the registry record.

The record lists 7 outcome measures and 7 analyses. The primary endpoint and secondary endpoints are rate outcomes. It does not provide a complete account of modeling, covariate selection, missing-data handling, sensitivity analysis, or multiplicity adjustment. These omissions prevent a full reconstruction of the statistical analysis. The posted record can therefore support a methods overview, but not an independent reanalysis or a quantitative appraisal of results.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists 7 outcome measures and 7 analyses: the primary composite of stroke and systemic embolism, alongside secondary outcomes for ischemic stroke, hemorrhagic stroke, systemic embolism, major intracranial bleeding, and major gastrointestinal bleeding. Considering multiple outcomes raises a general multiplicity issue: the more hypotheses examined, the greater the chance that at least one result may appear notable by chance if no adjustment or prespecified hierarchy is used. The record does not report whether outcomes were hierarchically ordered, whether multiplicity adjustment was planned or applied, or which analyses were designated confirmatory.

The primary outcome is identified in the registry, but the analysis plan, hypothesis-testing procedure, and any significance threshold are not reported in the registry record. No numerical results or p-values are reported, so the evidence across primary and secondary outcomes cannot be evaluated. The number of listed outcomes alone does not establish that every outcome was tested as an independent hypothesis or that a particular correction was required.

Missing-data methods are also not described. The record does not say how missing covariates, incomplete claims, or unavailable outcome information were handled, or whether participants with incomplete records were excluded or otherwise included. Follow-up may end at database withdrawal, but the consequences of withdrawal and the handling of incomplete observation are not specified. The number analyzed for each endpoint is not reported in the registry record.

The registry defines the eligible database population and reports total enrollment of 120722, but does not provide the analysis-specific population count. Enrollment and analyzed sample size are not interchangeable. Without details on exclusions, missingness, and endpoint-specific denominators, it is not possible to assess how these factors could affect the weighted comparisons.

11. What the Result Does, and Does Not, Mean

The registry record describes a real-world comparison of apixaban and warfarin among people with non-valvular atrial fibrillation at higher chance of bleeding. It specifies a primary rate outcome, the claims-based event definitions, the follow-up rules, and the use of IPTW to balance participant characteristics. These details explain what the analysis was intended to measure and how its groups were compared at a broad level.

The record does not report numerical primary or secondary outcome estimates, confidence interval limits, event counts, or person-time by group. It therefore does not show which group had a higher or lower event rate, how large any difference was, or how precise an estimate might be. No conclusion about statistical significance or comparative safety and effectiveness can be drawn from the posted outcome information.

Even where a weighted rate comparison is reported, general interpretation of an observational study differs from interpretation of randomized evidence. IPTW can adjust for measured characteristics, but cannot guarantee balance in every relevant factor or address unmeasured confounding. The registry explicitly warns that the arms are not a like-for-like control comparison. The omitted details about covariates, weighting diagnostics, missing data, and analysis methods further limit assessment.

The outcome definitions are based on hospital claims diagnosis coding. Such data can support large-scale observation, but the record does not report validation or adjudication procedures. The primary composite combines stroke and systemic embolism; its aggregate rate would not, on its own, reveal the contribution of each component.

The adverse-event fields list zero counts for the displayed categories in both groups and no serious-event terms. Those registry entries do not replace detailed comparative outcome results and cannot establish that no adverse events occurred in the underlying population. The most defensible interpretation is therefore about the study's question and design, not treatment effects: the record describes an observational, weighted comparison, while its quantitative findings and several methodological details are not reported in the registry record.

12. Why This Trial Matters Statistically

This study addresses a recognizable statistical challenge: comparing treatments in routine-care data when treatment is not randomly assigned. The registry describes a Japanese administrative claims database and a population with non-valvular atrial fibrillation considered at higher chance of bleeding. It uses IPTW to balance participant characteristics, a method intended to improve comparability on measured factors while retaining an observational design.

The choice of an incidence rate per 1000 participant-years is also relevant to real-world follow-up. Participants may contribute different amounts of observation, and the registry defines stopping points tied to an outcome, treatment change, discontinuation, or database withdrawal. A rate uses accumulated observation time, although its interpretation depends on event ascertainment, follow-up rules, and the analysis method.

Statistically, this design highlights the distinction between balance on recorded characteristics and causal comparability. Weighting can address measured differences only under assumptions about adequate measurement, model specification, and overlap between treatment groups. The record cautions that the arms are not a like-for-like control comparison, and does not report the covariate list or balance diagnostics needed to assess the adjustment.

The registry reports enrollment of 120722 and lists 7 outcome measures and 7 analyses, but does not report outcome estimates or the number analyzed for the primary endpoint. A large enrollment figure alone does not establish precision, eliminate bias, or show the magnitude of an effect. The absent numerical results mean the substantive comparison cannot be evaluated from the record.

The example is therefore statistically useful as a reminder that design description, outcome definition, follow-up, adjustment, and reporting all matter alongside sample size. It also shows why an observational estimate, if reported, needs to be interpreted with attention to residual confounding and uncertainty rather than treated as equivalent to a randomized comparison.

13. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The most direct limitation for interpreting this record is that numerical results for the primary and secondary outcomes are not reported in the registry record. Rates, confidence interval limits, event counts, and group-specific person-time are unavailable, preventing assessment of magnitude, direction, or precision.

The study is not described as randomized. Allocation is not stated, and the registry cautions that the arms are not a like-for-like control comparison. Although IPTW was used, the covariates, treatment-probability model, handling of weights, balance diagnostics, and overlap assessment are not reported. Residual confounding therefore cannot be evaluated from the available description.

The primary outcome relies on hospital claims diagnosis codes, including a first-listed ICD-10 diagnosis code. The registry does not report validation or adjudication of these claims-based definitions. The number of participants analyzed for each endpoint, group-specific follow-up time, and endpoint event counts are also not reported. The enrollment figure of 120722 does not establish the size of the analysis population.

The handling of missing data and incomplete observation is not reported in the registry record. Follow-up ends at database withdrawal among other events, but details of withdrawal and its analytical treatment are absent. The record also does not report sensitivity analyses, the full statistical analysis plan, or multiplicity procedures across the listed outcomes.

Finally, the adverse-event fields show zero counts in the listed categories for both groups and no serious-event terms, but provide no accompanying explanation. These fields are insufficient to characterize comparative adverse-event experience. Together, these reporting gaps limit what can be concluded from the registry entry, even though the study question and broad endpoint framework are described.

14. Related Tutorials

15. Related Statistical Calculators

16. Sources

• ClinicalTrials.gov: Safety and Effectiveness of Apixaban Compared to Warfarin in Patients With Non-valvular Atrial Fibrillation (a Type of Irregular Heart Rhythm) at Higher Chance of Bleeding — NCT05471505.
• Registry record last updated 2024-07-11; status last verified 2024-06.
• Results first posted 2024-07-11.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

17. Record Summary

NCT05471505 is a completed, Pfizer-sponsored observational study comparing apixaban with warfarin among people with non-valvular atrial fibrillation at higher chance of bleeding. The registry reports enrollment of 120722, a study start date of 2022-08-25, and a primary completion date of 2022-09-27. The comparison used Japanese administrative claims data, with IPTW described as the approach for balancing participant characteristics.

The primary outcome was the rate per 1000 participant-years of first composite stroke and systemic embolism events after treatment initiation. Secondary outcomes covered ischemic stroke, hemorrhagic stroke, systemic embolism, major intracranial bleeding, and major gastrointestinal bleeding. The described claims observation period was Mar 2011 to Jun 2021 (approximately 10 years 4 months).

Numerical outcome estimates and analysis-specific sample sizes are not reported in the registry record. Allocation is not stated, and the registry cautions that the arms are not a like-for-like control comparison. The record therefore supports an account of the question and broad methods, but not a quantitative conclusion about comparative outcomes.