← Clinical Trials
Venous Thromboembolism Phase 3 Non-Inferiority NCT00986154

Hokusai-VTE: Complete Statistical Analysis of Edoxaban in Venous Thromboembolism

An independent statistical analysis of the randomized phase 3 Hokusai-VTE trial comparing low molecular weight heparin/edoxaban with low molecular weight heparin/warfarin for symptomatic deep-vein blood clots and/or lung blood clots.

Hokusai-VTE  ·  Phase 3  ·  Completed  ·  Enrollment 8292.0
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 trial data posted on ClinicalTrials.gov for Hokusai-VTE.

1. Trial at a Glance

Hokusai-VTE was a randomized, parallel-group, triple-masked phase 3 treatment trial enrolling 8292.0 participants with venous thromboembolism, including deep vein thrombosis and pulmonary embolism. The principal efficacy question was whether heparin/edoxaban was non-inferior to heparin/warfarin in preventing symptomatic recurrent venous thromboembolism.

8292.0
Enrollment
Total participants
2
Treatment arms
Parallel design
0.89
Primary HR
95% CI .703–1.128
<0.0001
Primary p-value
Non-inferiority analysis
FeatureHokusai-VTE
Trial nameHokusai-VTE
NCT IDNCT00986154
Therapeutic areaHematology
PhasePhase 3
StatusCompleted
Start2009-10
Primary completion2013-04
AllocationRandomized
Design modelParallel
MaskingTriple
Primary purposeTreatment
Enrollment8292.0
Lead sponsorDaiichi Sankyo
Sponsor typeIndustry

2. Clinical Question

The trial evaluated whether heparin/edoxaban was non-inferior to heparin/warfarin for preventing symptomatic recurrent venous thromboembolism following the initial event.

Population

Participants with venous thromboembolism, including deep vein thrombosis, pulmonary embolism, thromboembolism, and venous thrombosis.

Intervention

Heparin/edoxaban tosylate, with edoxaban identified in the registry as DU-176b.

Comparator

Heparin/warfarin.

Primary question

Is heparin/edoxaban non-inferior to heparin/warfarin in preventing symptomatic recurrent VTE?

3. Trial Design

01
Randomize8292.0 participants
02
Two armsHeparin/edoxaban vs heparin/warfarin
03
Follow-up12 months from randomization
04
Time-to-eventRecurrent VTE and bleeding
05
Cox analysisHazard ratios and confidence intervals
TREATMENT ARM

Heparin / Edoxaban Tosylate

  • Edoxaban tosylate (DU-176b)
  • Low molecular weight heparin / unfractionated heparin
COMPARATOR ARM

Heparin / Warfarin

  • Warfarin
  • Low molecular weight heparin / unfractionated heparin

The registry describes the allocation as randomized and the design model as parallel. Masking was triple. The primary purpose was treatment.

What the design establishes: randomization creates the basis for comparing outcomes between treatment assignments. Triple masking is a design feature intended to reduce knowledge of treatment assignment during the conduct and assessment of the trial. Neither feature by itself determines the magnitude of the treatment effect; those are questions addressed by the endpoint analyses.

4. Endpoints

EndpointRegistry definitionTime frameAnalysis
Primary Symptomatic Recurrent VTE, i.e., the Composite of DVT, Non-fatal PE, and Fatal PE 12 months from time of randomization Modified Intent To Treat Analysis Set; Cox proportional-hazards model
Secondary The Composite Clinical Outcome of Symptomatic Recurrent VTE and All-cause Mortality 12 months from time of randomization mITT Analysis set; Cox proportional-hazards model
Secondary safety Clinically Relevant Bleeding (i.e., Major or Clinically Relevant Non-major Bleeding) Occurring During Treatment 12 months from time of randomization Safety Analysis Set; Cox proportional-hazards model

The registered primary endpoint is a composite time-to-event endpoint. Its definition combines deep vein thrombosis, non-fatal pulmonary embolism, and fatal pulmonary embolism. The registry defines the overall study period as the time from the reference date to the last study follow-up visit.

5. Statistical Methodology

Primary efficacy analysis: Cox proportional-hazards model

The primary analysis used a Cox proportional-hazards model in the modified intention-to-treat analysis set. The reported effect measure was a hazard ratio comparing heparin/edoxaban with heparin/warfarin.

Primary model
HR = hazard under heparin/edoxaban ÷ hazard under heparin/warfarin

The hazard ratio is a relative time-to-event measure. An HR below 1 indicates a lower estimated instantaneous event rate in the heparin/edoxaban group under the fitted model; an HR of 1 corresponds to equal estimated hazards.

Stratified analysis

The registry analysis text identifies intention-to-treat analysis and stratified analysis as concepts used in the primary analysis. For the secondary composite outcome, the Cox model explicitly included treatment group and randomization stratification factors: presenting diagnosis, baseline risk factors, and need for reduced dose.

Modified intention-to-treat analysis

The primary efficacy analysis population was the modified Intent To Treat Analysis Set. This is distinct from the safety analysis population. The analysis therefore needs to be interpreted in relation to the population actually specified for the endpoint rather than automatically assuming that every enrolled participant contributed to every analysis.

Safety analysis set

The clinically relevant bleeding analysis used the Safety Analysis Set. The registry specifies that this set included all randomized subjects who received at least one dose of study drug.

Confidence intervals

The primary hazard ratio was accompanied by a two-sided 95% confidence interval. In a non-inferiority framework, the confidence interval is especially important because the question is not simply whether the treatment groups differ. The relevant question is whether the observed uncertainty is sufficiently compatible with the prespecified non-inferiority criterion.

Important data constraint: the registry analysis text states that non-inferiority was assessed using the upper limit of the two-sided 95% confidence interval, but the numerical non-inferiority margin is not provided in the ClinicalTrials.gov record. This page therefore does not assign or infer a numerical margin.

6. Non-Inferiority Logic

Hokusai-VTE provides a useful example of why the interpretation of a clinical trial depends on its hypothesis structure, not only on the point estimate.

What non-inferiority asks

The question is whether the experimental treatment can be shown not to be unacceptably worse than the comparator according to a prespecified margin.

What the HR contributes

The hazard ratio describes the estimated relative event hazard. The confidence interval describes uncertainty around that estimate.

Why the upper CI matters

The registry states that non-inferiority was assessed by whether the upper limit of the two-sided 95% CI was below the prespecified margin.

Why the margin cannot be invented

The ClinicalTrials.gov record does not give the numerical margin. A valid statistical interpretation should not reconstruct it from the observed HR or confidence interval.

Generic non-inferiority rule
Upper limit of 95% CI < prespecified non-inferiority margin

This is the decision structure described in the registry analysis text. The exact numerical margin must come from the prespecified trial design or statistical analysis documentation.

7. Results: Primary Endpoint

The primary endpoint was symptomatic recurrent VTE, defined as the composite of DVT, non-fatal PE, and fatal PE, assessed over 12 months from randomization.

Hazard ratio for symptomatic recurrent VTE

0.89

95% CI: .703–1.128   ·   P < 0.0001

Analysis population: modified Intent To Treat Analysis Set

Primary endpointHeparin/Edoxaban vs Heparin/Warfarin
EndpointSymptomatic recurrent VTE: composite of DVT, non-fatal PE, and fatal PE
Time frame12 months from time of randomization
Analysis populationModified Intent To Treat Analysis Set
ModelCox proportional-hazards model
Effect measureHazard ratio
Estimate0.89
95% CI.703–1.128
P-value<0.0001
HypothesisNon-inferiority or equivalence
Clinical Biostats interpretation

What the estimate means: the HR of 0.89 means that the fitted Cox model estimated the instantaneous hazard of symptomatic recurrent VTE in the heparin/edoxaban group at approximately 89% of the hazard in the heparin/warfarin group. Expressed as a relative model-based quantity, this corresponds to an estimated 11% lower hazard.

What it does not mean: an HR of 0.89 does not mean that 11% fewer participants had VTE, nor does it mean that every participant had an 11% reduction in risk. The hazard ratio is a time-to-event measure, not an absolute risk difference.

Precision: the two-sided 95% CI of .703–1.128 describes uncertainty around the estimated hazard ratio under the analysis model and sampling framework. It is not a range containing the treatment effect for individual patients.

Why the p-value needs context: the reported P < 0.0001 is associated with the registry's non-inferiority hypothesis, not automatically with a conventional superiority claim. A very small p-value in a non-inferiority analysis should not be read as proof that edoxaban is superior to warfarin.

Non-inferiority caution: the ClinicalTrials.gov record states that non-inferiority depended on the upper confidence limit being below a prespecified margin, but does not provide that numerical margin. The result should therefore be understood through the stated non-inferiority framework rather than by comparing the CI only with 1.00.

Model caution: the Cox interpretation also relies on the model's time-to-event framework and its proportional-hazards assumption. Censoring and the specified analysis population are additional components of the interpretation.

8. Results: Symptomatic Recurrent VTE Plus All-Cause Mortality

A secondary endpoint combined symptomatic recurrent VTE with all-cause mortality. It was evaluated over 12 months from randomization in the mITT Analysis set using a Cox proportional-hazards model.

Hazard ratio for recurrent VTE or all-cause mortality

1.00

95% CI: .832–1.200   ·   P = .9933

Analysis population: mITT Analysis set

Secondary endpointReported analysis
OutcomeComposite Clinical Outcome of Symptomatic Recurrent VTE and All-cause Mortality
Time frame12 months from time of randomization
Analysis populationmITT Analysis set
ModelCox proportional-hazards model
Effect measureHazard ratio
Estimate1.00
95% CI.832–1.200
P-value.9933
HypothesisSuperiority
Clinical Biostats interpretation

What the estimate means: an HR of 1.00 is the model's point estimate of equal instantaneous hazard between the two treatment groups for the composite endpoint.

What the confidence interval says: the 95% CI of .832–1.200 spans values below and above 1.00. It therefore includes both a possible lower estimated hazard and a possible higher estimated hazard for heparin/edoxaban relative to heparin/warfarin under this model.

What the p-value means: the reported P = .9933 is the formal result posted on ClinicalTrials.gov for the superiority hypothesis. It provides no evidence of a statistically detectable superiority difference under that analysis. It is not a measure of the size of the effect and should not be interpreted as the probability that the null hypothesis is true.

Endpoint interpretation: because this is a composite, its result reflects the time to the first event represented by the combined endpoint. It does not separately quantify the treatment effect on each component.

9. Results: Clinically Relevant Bleeding

Clinically relevant bleeding was a secondary safety endpoint defined as major or clinically relevant non-major bleeding occurring during treatment. It was analyzed over 12 months from randomization in the Safety Analysis Set.

Hazard ratio for clinically relevant bleeding

0.81

95% CI: .705–.936   ·   P = .0040

Analysis population: Safety Analysis Set

Safety endpointReported analysis
OutcomeClinically Relevant Bleeding: Major or Clinically Relevant Non-major Bleeding Occurring During Treatment
Time frame12 months from time of randomization
Analysis populationSafety Analysis Set
ModelCox proportional-hazards model
Effect measureHazard ratio
Estimate0.81
95% CI.705–.936
P-value.0040
HypothesisSuperiority
Clinical Biostats interpretation

What the estimate means: the HR of 0.81 indicates that the estimated instantaneous hazard of clinically relevant bleeding was approximately 81% of the corresponding hazard in the heparin/warfarin group under the fitted Cox model. This is equivalent to an estimated 19% lower hazard.

What it does not mean: it does not mean that 19% of participants avoided bleeding or that each participant's bleeding probability was reduced by 19%. It is a relative time-to-event estimate.

Precision: the 95% CI of .705–.936 remains below 1.00, indicating that the reported interval is compatible with a lower hazard in the heparin/edoxaban group under this analysis.

P-value: P = .0040 is the reported result for the superiority hypothesis. The p-value addresses statistical evidence against the relevant null hypothesis; it does not measure clinical importance, effect magnitude, or the probability that the observed effect will be reproduced.

Safety-population caution: this analysis used the Safety Analysis Set, which the registry defines as randomized subjects who received at least one dose of study drug. That differs from the mITT population used for the primary efficacy endpoint.

10. Serious Adverse Events

The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk.

Treatment armSerious adverse eventsAt risk
Heparin/Edoxaban Tosylate8014118
Heparin/Warfarin8524122

These counts provide an arm-specific safety description. They should not be substituted for the clinically relevant bleeding time-to-event analysis, because the two summaries represent different safety concepts and the bleeding endpoint was analyzed using a Cox model.

Safety interpretation: serious adverse-event counts are descriptive here. The ClinicalTrials.gov record does not provide a formal between-arm statistical analysis for these serious adverse-event counts, so no hazard ratio or p-value is assigned to them on this page.

11. Analysis Populations

PopulationRole in Hokusai-VTE analysis
Modified Intent To Treat Analysis SetPrimary analysis population for symptomatic recurrent VTE.
mITT Analysis setAnalysis population for the composite of symptomatic recurrent VTE and all-cause mortality.
Safety Analysis SetAnalysis population for clinically relevant bleeding; includes randomized subjects who received at least one dose of study drug.

The distinction matters because the same treatment comparison can produce different estimands depending on who contributes to an analysis. Efficacy analyses anchored to randomized assignment preserve the treatment-comparison framework of randomization, while exposure-based safety analyses can use a population defined by treatment receipt.

12. Stratification in the Cox Models

The registry-reported analysis text identifies stratified analysis as an important component of the time-to-event methodology. For the secondary composite of symptomatic recurrent VTE and all-cause mortality, the registry specifies the following randomization stratification factors:

Stratification factorRegistry description
Presenting diagnosisPE with/without DVT; DVT only
Baseline risk factorsTemporary factors; all others
Need for reduced doseIncluded as a randomization stratification factor

Stratification allows the time-to-event comparison to account for prespecified factors used during randomization. Conceptually, the analysis compares treatment groups within the structure created by those strata rather than treating the entire randomized population as an undifferentiated sample.

13. Kaplan-Meier Estimation and Time-to-Event Endpoints

The ClinicalTrials.gov record classifies all three posted statistical analyses as time-to-event analyses. This is important because a participant's contribution depends not only on whether an event occurred but also on when the event or censoring occurred.

Kaplan-Meier concept
S(t) = ∏ti ≤ t (1 − di/ni)

A Kaplan-Meier estimator uses the observed event times and numbers at risk to estimate the probability of remaining event-free through time. The registry analysis specifies Cox regression rather than providing a separate Kaplan-Meier estimate in the posted statistical-analysis fields.

For an endpoint such as recurrent VTE over 12 months, time-to-event analysis is useful because participants can have different lengths of observed follow-up. Participants who have not experienced the endpoint by their last available observation can contribute information up to that point rather than simply being treated as if they had the same follow-up duration as every other participant.

14. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

The registered outcomes are time-to-event endpoints. A Cox model estimates the relative hazard between treatment groups while using the timing of events and allowing censored observations to contribute information up to their censoring time. The registry identifies Cox regression as the analysis method for all three posted analyses.

What does an HR of 0.89 mean?

An HR of 0.89 means that the fitted model estimated the instantaneous event hazard under heparin/edoxaban at 89% of that under heparin/warfarin. The corresponding 11% figure is a relative hazard interpretation, not an 11-percentage-point reduction in event probability.

Why is the non-inferiority margin more important than comparing the HR with 1?

In a superiority analysis, the value 1 is the conventional reference for a hazard ratio. Non-inferiority asks a different question: whether the experimental treatment is not worse than the comparator by more than a prespecified clinically acceptable amount. The registry text explicitly describes the decision rule using the upper confidence limit and the non-inferiority margin. Because the numerical margin is not reported here, it should not be inferred.

Why does the primary p-value not establish superiority?

The primary analysis is explicitly classified in the ClinicalTrials.gov record as a non-inferiority or equivalence hypothesis. Therefore, the reported P < 0.0001 must be interpreted in the context of that hypothesis. It cannot simply be relabeled as a conventional superiority p-value.

Why is the analysis population important?

The primary endpoint was analyzed in the modified Intent To Treat Analysis Set, while clinically relevant bleeding was analyzed in the Safety Analysis Set. The populations answer related but different statistical questions and should not be silently combined.

What does a confidence interval contribute?

A confidence interval communicates the statistical precision of an estimated effect. For example, the primary HR of 0.89 has a two-sided 95% CI of .703–1.128. The interval shows that the point estimate alone does not capture the uncertainty in the treatment comparison.

Why does stratification matter?

The registry-reported analysis text identifies stratification factors used in the Cox models. Stratified analysis can account for prespecified features incorporated into randomization, helping the treatment comparison respect the structure of the trial design.

15. Reading the Primary Confidence Interval

Point estimate

0.89 is the estimated relative hazard from the fitted Cox model.

Lower bound

.703 is the lower endpoint of the two-sided 95% confidence interval.

Upper bound

1.128 is the upper endpoint and is the critical quantity identified by the registry's non-inferiority description.

Reference value

1.00 represents equal estimated hazards. But non-inferiority is not determined solely by whether the interval crosses 1.00.

This distinction is central. The primary confidence interval extends above 1.00, but the registry text does not define the numerical non-inferiority margin. Therefore, a statistically careful reading cannot substitute a simple “CI excludes 1” rule for the prespecified non-inferiority criterion.

16. P-Values and Effect Sizes

The three posted analyses illustrate why p-values and effect estimates should always be read together.

EndpointHR95% CIP-valueHypothesis
Symptomatic recurrent VTE0.89.703–1.128<0.0001Non-inferiority or equivalence
Recurrent VTE + all-cause mortality1.00.832–1.200.9933Superiority
Clinically relevant bleeding0.81.705–.936.0040Superiority

A p-value is evidence against a specified null hypothesis under the statistical model; it is not a direct measure of effect size. The HR describes the estimated relative treatment effect, while the confidence interval communicates uncertainty around that estimate. The hypothesis type determines how the p-value should be interpreted.

17. Primary vs Secondary Analyses

AnalysisRoleInterpretive framework
Symptomatic recurrent VTEPrimaryNon-inferiority or equivalence
Symptomatic recurrent VTE + all-cause mortalitySecondarySuperiority
Clinically relevant bleedingSecondary safetySuperiority

The distinction between endpoint roles matters. A primary non-inferiority question and a secondary superiority question are not interchangeable statistical tests. Each has its own estimand, analysis population, and hypothesis structure.

Multiplicity: the ClinicalTrials.gov record does not specify a multiplicity-adjustment procedure or an endpoint hierarchy beyond identifying the primary and secondary analyses. This page therefore does not infer an alpha-allocation scheme or claim a particular familywise-error strategy.

18. What the Hazard Ratio Does — and Does Not — Mean

Primary efficacy

The HR of 0.89 represents the estimated relative hazard of symptomatic recurrent VTE for heparin/edoxaban versus heparin/warfarin under the Cox model. It is not an absolute probability, an absolute risk difference, or the proportion of participants who benefit.

Secondary composite

The HR of 1.00 represents equal estimated instantaneous hazard for the composite of symptomatic recurrent VTE and all-cause mortality. It does not establish that the underlying component events were individually identical.

Safety

The HR of 0.81 for clinically relevant bleeding represents a lower estimated instantaneous hazard in the heparin/edoxaban group under the fitted model. It should not be translated directly into an absolute percentage reduction without the underlying event-time data.

19. Limitations and Interpretation Issues

20. Why This Trial Matters Statistically

Hokusai-VTE is a useful teaching case because the same randomized comparison illustrates several distinct principles of clinical-trial statistics: non-inferiority testing, time-to-event analysis, Cox regression, stratification, confidence intervals, analysis populations, and the separation of efficacy from safety.

ConceptHow it appears in Hokusai-VTE
RandomizationThe registry identifies the allocation as randomized.
Parallel designThe trial has a parallel design model with 2 arms.
Triple maskingThe registry identifies masking as triple.
Non-inferiorityThe primary symptomatic recurrent VTE analysis uses a non-inferiority or equivalence hypothesis.
Time-to-event endpointThe primary and secondary posted analyses are time-to-event analyses.
Cox modelAll three posted statistical analyses use Cox regression.
Hazard ratioEach analysis reports a hazard ratio as the effect measure.
Confidence intervalEach posted analysis provides a two-sided 95% confidence interval.
Stratified analysisThe analysis text identifies stratification, including presenting diagnosis, baseline risk factors, and need for reduced dose.
Intention-to-treat analysisThe primary efficacy analysis uses a modified Intent To Treat Analysis Set.
Safety populationClinically relevant bleeding is analyzed in the Safety Analysis Set.
Composite endpointThe primary endpoint combines DVT, non-fatal PE, and fatal PE.

21. A Statistical Walk-Through of the Primary Result

Step 1 · Define the endpoint

Symptomatic recurrent VTE

The primary endpoint is a composite of DVT, non-fatal PE, and fatal PE over 12 months from randomization.

Step 2 · Define the comparison

Heparin/edoxaban vs heparin/warfarin

The two randomized treatment strategies form the comparison used in the primary analysis.

Step 3 · Model time to event

Cox proportional-hazards regression

The registry identifies Cox regression as the statistical method and hazard ratio as the effect measure.

Step 4 · Estimate uncertainty

HR 0.89 with 95% CI .703–1.128

The point estimate and confidence interval summarize the relative treatment effect and its statistical uncertainty.

Step 5 · Apply the hypothesis

Non-inferiority framework

The registry states that non-inferiority was evaluated using the upper limit of the two-sided 95% confidence interval relative to a prespecified margin.

22. Primary Analysis: Statistical Interpretation vs Clinical Interpretation

Statistical interpretation

The primary Cox analysis estimated an HR of 0.89 with a two-sided 95% CI of .703–1.128 and a reported P < 0.0001 under a non-inferiority or equivalence hypothesis.

Clinical interpretation

The trial compares two treatment strategies for symptomatic venous thromboembolism. The statistical result should be considered together with the endpoint definition, analysis population, confidence interval, non-inferiority margin, and safety findings.

The distinction is important because statistical evidence and clinical interpretation answer related but different questions. The HR is a relative model-based quantity. The confidence interval describes uncertainty. The non-inferiority framework determines the relevant decision rule. Safety is evaluated separately rather than being folded into the primary efficacy statistic.

23. Related Tutorials

Learn more about the methods used in this trial:

24. Related Statistical Calculators

25. Sources

Continue through the Clinical Biostats statistical library

Explore the underlying survival-analysis, clinical-trial, and statistical-inference methods used to interpret randomized time-to-event studies.

26. Record Summary

Hokusai-VTE provides a compact example of how a randomized clinical trial can require several layers of statistical interpretation. The primary endpoint was symptomatic recurrent VTE, analyzed over 12 months using a Cox proportional-hazards model in a modified Intent To Treat Analysis Set. The reported HR was 0.89 with a two-sided 95% CI of .703–1.128 and a reported P < 0.0001 under a non-inferiority or equivalence hypothesis. Secondary analyses examined the composite of symptomatic recurrent VTE and all-cause mortality and clinically relevant bleeding, again using Cox regression but with different hypotheses and, for bleeding, a different analysis population.

The most important statistical lesson is that the primary result cannot be interpreted correctly by looking at the HR or p-value in isolation. The endpoint definition, modified intention-to-treat population, two-sided confidence interval, stratification, proportional-hazards framework, and especially the non-inferiority margin determine how the result should be understood. Because the ClinicalTrials.gov record does not provide the numerical margin, this page deliberately does not infer it.

Clinical Biostats methodology: A trial-results page should distinguish the reported statistical result from its interpretation. For Hokusai-VTE, that means preserving the registry's non-inferiority framework, reporting the exact hazard ratios and confidence intervals, separating efficacy from safety populations, and avoiding unsupported assumptions about margins, multiplicity, interim monitoring, missing-data methods, or additional outcomes.