Apixaban Versus Warfarin for the Management of Post-operative Atrial Fibrillation: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Apixaban Versus Warfarin for the Management of Post-operative Atrial Fibrillation (NCT02889562), a randomized phase 2, phase 3 study. The primary outcome is Number of Participants With Strokes.

2. Clinical Question

The question is about the observed endpoint in this particular trial, not a general conclusion about anticoagulation or a guarantee about future patients. With no observed strokes in either group, the trial's event data do not distinguish the assigned treatments on this endpoint. The small number of enrolled participants and the short, specified follow-up period are important context when interpreting that observation.

3. Trial Design

This was a prospective, randomized pilot study with a parallel-group design. The registry lists the study as Phase 2 and Phase 3, records no masking, and describes the trial as open-label. Enrollment was 56 participants, with 28 assigned to apixaban and 28 to warfarin. The study status is completed. The registry lists Sanford Health as sponsor.

The study addressed atrial fibrillation occurring after isolated coronary artery bypass grafting. The registry summary describes eligibility candidates as patients with persistent atrial fibrillation lasting >12 hours or recurrent sustained atrial fibrillation, defined as >2 episodes lasting longer than 30 minutes. Candidates also had to meet inclusion and exclusion criteria and provide informed consent. The registry record does not report the full criteria or the analysis population details.

Apixaban and warfarin were the assigned treatments. The registry describes apixaban as 5 mg by mouth twice daily, subject to dose modifications, for at least 30 days. Warfarin dosing during hospitalization was to be guided by daily INR monitoring and hospital protocol. Routine postoperative care was planned in both groups, with anticoagulation managed by an anticoagulation clinic after discharge. The registry summary says participants would be followed for 30 days after surgery.

The recorded study start is 2016-09, and primary completion is 2019-05-11. Random assignment supports a comparison of the assigned groups, but the open-label design means treatment assignment was not masked to participants or care teams. The registry says stroke events would be adjudicated by independent committee members blinded to treatment assignment. Further information about concealment of allocation, adherence, departures from assigned treatment, and the detailed statistical analysis plan is not reported in the registry record.

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.

Overall Study

Reasons for not completing this period:

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.

6. Primary Endpoint

The registry lists the primary outcome as Number of Participants With Strokes, measured over 30 days.

The primary endpoint was the number of participants with strokes during 30 days. The registry describes efficacy as freedom from stroke during the study period and says stroke-related events would be adjudicated using predetermined definitions by independent committee members who remained blinded to treatment assignment. This endpoint therefore concerns whether a participant experienced a stroke in the specified follow-up period, rather than the number or severity of strokes per participant.

A binary event endpoint can make a direct comparison of the proportion of participants with at least one event in each randomized group. In general, stroke is a clinically consequential event, so counting its occurrence can address an important safety and efficacy question in a study of anticoagulation. The endpoint definition also specifies a time window, which is necessary because event counts depend on how long participants are observed. These are general reasons such an endpoint may be used; they do not establish the basis for the registry's particular endpoint choice.

The posted results record zero participants with strokes in the apixaban group and zero in the warfarin group. Thus, the observed data contain no between-group difference in event counts. With no observed events, the data cannot show how the treatments compare among participants who experienced a stroke, and they provide limited information about possible differences in the underlying event risk. The registry record does not report further endpoint definitions, analysis-population details, or the handling of participants with incomplete follow-up.

7. Results Posted in the Registry

• Warfarin: 0.0% (0/28) versus 0.0% (0/28) on control; risk difference 0.0 percentage points (95% CI 0.0 to 0.0); unadjusted two-proportion z-test p = n/a.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Total Post-operative Length of Stay

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Warfarin: mean 143.10 versus 167.50 on control; difference -24.40 (95% CI -58.94 to 10.14), standardized effect d = -0.37; z-test p = 0.166.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Warfarin and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists three secondary outcomes: units of blood given after initiation of anticoagulation medication, total postoperative length of stay, and time in therapeutic range of INR for participants on warfarin. The posted numerical secondary result discussed in the available summaries is total postoperative length of stay. Results for the other listed secondary outcomes are not reported in the registry record.

The length-of-stay comparison is a continuous outcome and addresses a different question from the primary binary stroke endpoint. Its recalculated mean difference is -24.4 hours, with a 95% confidence interval from -58.94 to 10.14; the unadjusted recalculated p-value is 0.166. The sponsor reports a one-sided t-test with p = 0.17. The interval includes zero, and the summary is compatible with uncertainty in either direction. These findings should not be used to infer a difference in stroke occurrence, because the endpoints measure different outcomes.

Secondary outcomes are often useful for understanding other aspects of a trial, but their results generally carry less weight than the primary outcome when they are exploratory or one of several analyses. Testing multiple outcomes increases the chance of observing at least one apparently unusual result by chance if no adjustment or prespecified hierarchy is used. Whether such an adjustment or hierarchy was planned in this trial is not reported in the registry record. The registry also does not report the analysis populations or missing-data handling for the secondary outcomes.

The posted secondary findings therefore offer limited additional evidence. The length-of-stay estimate is imprecise, while numeric results for the other listed secondary outcomes are not reported in the registry record. The primary result remains a zero-event comparison over 30 days; the secondary information neither resolves the uncertainty around that endpoint nor provides a basis for a broad conclusion across outcomes.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Units of Blood Given After Initiation of Anticoagulation Medication
• Total Post-operative Length of Stay
• Time in Therapeutic Range of INR, if on Warfarin

9. Statistical Interpretation

10. Statistical Methods Explained

Several statistical ideas help put the results in context. The methods below distinguish the trial's posted analysis from general explanations of how common statistical summaries are interpreted. Details of the full protocol and statistical analysis plan are not reported in the registry record.

What does a risk difference describe?

For a binary outcome, a risk difference compares the event proportion in one group with the event proportion in another. A value of zero means the observed event proportions are the same. In this trial, no participants in either group were recorded as having a stroke, so the posted risk difference is 0.0. This is a statement about observed events in the randomized groups, not proof that the underlying risks are identical.

When event counts are very small or absent, a risk difference can provide little information about how the underlying event probabilities compare. The trial's observed result is not evidence that a difference could never occur; it indicates only that no stroke was recorded in either group during the stated timeframe. The registry reports a chi-squared test with p = 1.0. A p-value measures how compatible the data are with a specified statistical null model, under the assumptions of the test. It is not the probability that the null hypothesis is true, and it is not a measure of the probability that one treatment is better.

What does the length-of-stay comparison describe?

Total postoperative length of stay was recorded as a continuous outcome in hours. The group summaries are means with standard deviations: 167.5 hours with a standard deviation of 76.6 for apixaban, and 143.1 hours with a standard deviation of 53.2 for warfarin. A mean describes the arithmetic average, while a standard deviation describes the spread of observations around that average. The comparison is a mean difference, rather than a difference in the number of participants with an event.

The recalculated mean difference is -24.4 hours for warfarin relative to apixaban, with a 95% confidence interval from -58.94 to 10.14. The negative direction follows the posted comparison: the warfarin mean is lower than the apixaban mean. The interval includes zero, so the results are compatible with no mean difference as well as with differences in either direction within the interval. The recalculated unadjusted p-value is 0.166. These calculations are unadjusted comparisons from the posted group summaries; they are not adjusted for other variables.

The registry's own result for length of stay is a one-sided t-test with p = 0.17. A t-test evaluates a mean comparison under its model assumptions. A one-sided test addresses a direction specified for the comparison, whereas a two-sided test considers departures in either direction. The registry record does not report the rationale for using a one-sided test or the direction specified in advance. The recalculated two-sided result and the sponsor's reported one-sided result should not be treated as interchangeable tests.

Why can over-dispersion matter?

General explanation: a statistical model may assume that observed variation is described by a particular probability distribution. Over-dispersion means the data vary more than that model expects. When present, it can make standard errors too small and confidence intervals too narrow if the extra variation is ignored. This matters especially for count outcomes, for which a basic model may assume a particular relationship between the mean and variance. The registry record does not report an over-dispersion assessment, and the posted primary outcome is a binary participant count rather than a repeated event count. No claim about over-dispersion in this trial can be made from the posted results.

How should the planning illustrations be read?

The sample-size illustrations are based on the observed length-of-stay effect, with two-sided significance level 0.05, equal allocation, and a normal approximation. They are planning illustrations rather than observed power calculations or evidence that the trial was designed around those assumptions. Planning for 100% of the observed effect gives 115 participants per arm for 80% power and 154 participants per arm for 90% power. Planning for 75% of the observed effect gives 204 participants per arm for 80% power and 273 participants per arm for 90% power. Planning for 50% of the observed effect gives 459 participants per arm for 80% power and 615 participants per arm for 90% power.

These figures illustrate how the number of participants needed depends on the effect size targeted under the stated assumptions. They do not show the trial's observed power, and they are not the trial's original sample-size justification. The registry record does not report the trial's own sample-size justification. Because the planning illustration uses an observed effect from this small study, it should not be treated as a reliable forecast of a future study's true effect or required enrollment.

What is not specified?

The registry record does not report whether the primary analysis was intention-to-treat, per-protocol, or another analysis population; how missing observations were handled; whether any interim analyses were planned or conducted; or whether covariate adjustment was prespecified. It also does not report a detailed multiplicity strategy. Those details affect how analysis results should be interpreted, so the reported numbers should not be assumed to reflect procedures that are not documented in the registry.

11. Confidence Intervals and Precision

Precision concerns how narrowly an estimate identifies a range of values compatible with the data and the statistical method. A confidence interval is a general statistical summary of uncertainty around an estimate. In repeated use under the method's assumptions, intervals at a stated confidence level have a long-run coverage interpretation; an individual interval is not a probability statement that the true value lies within its endpoints.

For the primary outcome, the posted risk difference is 0.0, with an interval from 0.0 to 0.0. Both groups had zero recorded strokes among 28 participants each. The posted interval gives no visible width around the point estimate, but that should not be interpreted as perfect knowledge of the underlying risk difference. With no events in either arm, the observed data do not provide an event-based contrast, and the registry record does not report another primary-outcome interval or uncertainty method. The zero-width interval is therefore not sufficient evidence that clinically meaningful differences have been ruled out.

For postoperative length of stay, the recalculated mean difference is -24.4 hours, and the 95% confidence interval extends from -58.94 to 10.14. The interval includes zero and spans negative and positive values. It is compatible with a range of possible mean differences, including no difference; it does not establish that all values in that range are equally plausible. The interval also reflects the unadjusted comparison based on the posted summary statistics, not a covariate-adjusted analysis.

Precision is shaped by the amount of information in the data, including sample size, outcome variability, and event frequency. In general, low event counts limit what can be learned about binary outcomes, while variability in a continuous outcome affects the precision of a mean comparison. The study's enrollment was 56, and the primary endpoint had no observed events. These features call for caution when interpreting the precision of the treatment comparisons. No interval for the sponsor's length-of-stay analysis is reported in the registry record.

12. Analyses Reported by the Sponsor

The registry's reported primary analysis is a chi-squared test with p = 1.0. The posted group counts are zero participants with strokes in each arm, and the recalculated risk difference is 0.0. These summaries agree on the absence of an observed between-group difference in stroke counts. The registry record does not report a confidence interval for the sponsor's primary analysis beyond the interval attached to the posted group summaries.

For total postoperative length of stay, the sponsor's reported result is a one-sided t-test with p = 0.17. The recalculated comparison from the posted group summaries is an unadjusted mean difference of -24.4 hours, with a 95% confidence interval from -58.94 to 10.14 and an unadjusted p-value of 0.166. The recalculation uses the posted means and standard deviations and is two-sided; it does not replace the sponsor's analysis. The registry record does not provide more detail about the sponsor's test assumptions or the reason for the one-sided test.

The sponsor's analysis is the primary reference for what the registry reports as its own analysis. The recalculated comparison is useful as a transparent, simple check based on the posted summaries, but it may not reproduce unreported features of the sponsor's analysis. The posted numbers do not show a discrepancy flag, and no claim of a conflicting result is warranted. Neither p-value establishes the presence or absence of a meaningful effect.

13. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when investigators make several statistical comparisons, such as testing multiple endpoints, time points, treatment contrasts, or subgroups. In general, as the number of tests increases, the chance of at least one small p-value can increase even if no real effects exist. Prespecifying a primary endpoint and a testing hierarchy, or adjusting for multiple comparisons, can help control how the collection of results is interpreted. Whether the trial used such a strategy is not reported in the registry record.

The registry identifies number of participants with strokes as the primary outcome and lists three secondary outcomes. The posted results include a primary chi-squared test and a secondary one-sided t-test for length of stay. The registry record does not report whether additional analyses were conducted, whether the secondary outcomes were treated as confirmatory or exploratory, or whether any multiplicity adjustment was applied. Accordingly, an isolated secondary p-value should be interpreted cautiously rather than as a stand-alone confirmation.

Missing-data handling is another important analysis detail. Participants may have incomplete outcome information or may not contribute all planned observations; the way such cases are handled can affect estimates and uncertainty. The registry record does not report the number of participants with missing primary or secondary data, the handling of missing observations, or whether follow-up was complete for all participants. The reported group sizes should not be taken as evidence that all participants contributed data to every analysis unless that is specified.

The analysis population also affects interpretation. An intention-to-treat analysis generally compares participants according to randomized assignment, while a per-protocol analysis focuses on participants who followed specified treatment procedures. These are general distinctions; the registry record does not say which population was used for the primary or secondary analyses, nor does it describe treatment switching, adherence, or exclusions after randomization.

The record also does not report interim analyses, stopping rules, or a full statistical analysis plan. It is therefore not possible to determine from the registry record whether these issues were addressed in prespecified procedures. These unreported details do not invalidate the posted results, but they limit assessment of how fully the analyses align with the original plan.

14. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: required transfusions after discharge (3.6%), subconjunctival hemorrhage (3.6%).

The adverse-event figures record serious adverse events in 1 of 28 participants in the apixaban group and 1 of 28 participants in the warfarin group. Deaths are recorded as 0 in each group. Other adverse events are recorded in 14 of 28 participants in the apixaban group and 7 of 28 participants in the warfarin group. The listed serious-event terms are “required transfusions after discharge” and “Subconjunctival hemorrhage,” each reported as 3.6%.

These are descriptive counts from the study, not estimates that establish a causal safety difference. With few participants and few serious events, the counts are limited in their ability to characterize uncommon harms or to distinguish between treatment groups. The registry record does not report further information about event timing, severity grading, relatedness assessments, or whether an individual participant experienced more than one event. The listed terms should not be treated as an exhaustive account of all safety outcomes.

Open-label treatment assignment is recorded, while the registry describes independent blinded adjudication for stroke events. That statement concerns the stroke endpoint; it does not establish that adverse events were assessed by blinded reviewers. These safety figures are descriptive trial observations and do not by themselves support a broader treatment recommendation.

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

The clearest statement supported by the primary results is that no participants in either randomized group were recorded as having a stroke during the 30-day timeframe. The posted risk difference is 0.0, and the registry reports p = 1.0 for the chi-squared analysis. This is an observation about the enrolled participants and recorded follow-up, not proof that apixaban and warfarin have identical effects on stroke risk.

The trial enrolled 56 participants and is described as a pilot study. With no observed strokes in either group, the study has little direct information about how stroke risks might differ. The posted interval from 0.0 to 0.0 should not be read as excluding a possible difference in the underlying risks. The registry record does not report a primary-outcome uncertainty analysis beyond that interval.

The secondary length-of-stay result is a separate outcome. Its recalculated estimate and interval are compatible with values in both directions, and the sponsor's reported analysis is a one-sided test. This result does not answer the primary stroke question. Other listed secondary outcomes do not have numeric results reported in the registry record, so the overall pattern across all planned outcomes cannot be fully assessed.

The available results do not establish equivalence, prove absence of risk, or show that one assigned treatment is superior. Nor do they support medical advice for an individual. Interpretation is constrained by the pilot enrollment, zero primary events, open-label design, 30-day follow-up, and the absence of registry entries about missing data, analysis populations, and the full statistical plan. These limits are reasons to describe the findings narrowly, not to replace the observed results with assumptions.

16. Why This Trial Matters Statistically

This trial illustrates a central statistical challenge in clinical research: an important outcome can be too uncommon within a small study to support a precise comparison. Here, the primary endpoint is stroke, but the posted data record no strokes in either group. Equal zero counts are a real description of what was observed, yet they provide little ability to distinguish the groups' underlying event risks.

The contrast between the primary and secondary outcomes also matters. A binary endpoint with no events and a continuous endpoint such as length of stay have different data structures and different uncertainties. A zero-event primary comparison does not make a secondary mean comparison a substitute for the primary question, and a secondary result cannot supply information about strokes that were not observed.

The registry describes the study as randomized and as a pilot, with an open-label design and blinded adjudication of stroke events. These features help readers identify the intended comparison and some safeguards, while also highlighting what remains undocumented. In particular, analysis populations, missing-data procedures, multiplicity handling, and the original sample-size justification are not reported in the registry record.

For biostatisticians and clinical researchers, the case underscores why a point estimate and p-value should be read alongside event counts, confidence intervals, design details, and the limits of the available record. It also shows why sample-size illustrations based on an observed pilot effect are planning aids rather than evidence of observed power or proof of the design's adequacy. The defensible conclusion is bounded by the data actually recorded and the uncertainty that remains.

17. 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.

Several limitations specific to this study affect interpretation. The registry records enrollment of 56 participants, with 28 in each arm, and zero strokes in both groups. Because there were no observed primary events, the trial offers no event contrast for estimating a treatment difference in stroke occurrence. The posted interval of 0.0 to 0.0 does not eliminate uncertainty about the underlying risks.

Follow-up for the primary endpoint was 30 days. This defines the period represented by the primary result; it cannot describe events outside that period. The registry record does not report additional follow-up results. The study is described as a pilot, and its sample-size justification is not reported in the registry record. The planning illustrations based on the observed length-of-stay effect should not be mistaken for the trial's own design calculations.

The trial was open-label, with no masking recorded. The registry states that stroke events would be adjudicated by independent committee members blinded to treatment assignment, but does not report whether other outcome assessments were blinded. The record also does not report allocation concealment, adherence, treatment changes, missing outcome data, or analysis populations. These are relevant details for judging potential bias and the scope of the group comparisons.

Only one secondary outcome has posted numerical results in the record: total postoperative length of stay. Results for units of blood given after anticoagulation initiation and time in therapeutic range of INR are not reported in the registry record. Multiplicity procedures and the full statistical analysis plan are also not reported. The one-sided sponsor test for length of stay and the unadjusted recalculation answer different testing formulations; the registry does not report the rationale for the one-sided test.

Finally, the safety counts are limited by the small study size and the sparse serious-event data. The registry record does not provide further details about adverse-event ascertainment or analysis. These limitations support a narrow description of the posted results rather than a broad conclusion about comparative outcomes.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Apixaban Versus Warfarin for the Management of Post-operative Atrial Fibrillation — NCT02889562.
• Registry record last updated 2021-01-22; status last verified 2020-12.
• Results first posted 2021-01-22.
• 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.

21. Record Summary

NCT02889562 is a completed, prospective, randomized, open-label, parallel pilot study of apixaban versus warfarin for postoperative atrial fibrillation after isolated coronary artery bypass grafting. The registry lists Phase 2 and Phase 3, enrollment of 56, and 28 participants in each group. The recorded study start is 2016-09, and primary completion is 2019-05-11.

The primary outcome was number of participants with strokes during 30 days. Zero participants had a recorded stroke in each arm; the reported chi-squared p-value is 1.0. The posted numerical secondary result is postoperative length of stay. Other registry-listed secondary outcomes are units of blood given after anticoagulation initiation and time in therapeutic range of INR for warfarin, with numerical results not reported in the registry record.

The results describe a small pilot with no primary events, so they do not establish equivalence or resolve possible differences in underlying stroke risk. Important analysis details, including missing-data handling and analysis populations, are not reported in the registry record.