This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
RE-COVER I was a randomized, double-blind, parallel phase 3 trial evaluating dabigatran etexilate 150 mg versus warfarin with an INR target of 2–3 for treatment of acute symptomatic venous thromboembolism. The primary endpoint was recurrent symptomatic VTE and deaths related to VTE, analyzed as a time-to-event outcome in the full analysis set.
| Feature | RE-COVER I |
|---|---|
| Phase | Phase 3 |
| Condition | Thromboembolism |
| Population | Patients with acute symptomatic venous thromboembolism |
| Design | Randomized, double-blind, parallel |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Primary endpoint | Number of participants with recurrent symptomatic VTE and deaths related to VTE |
| Primary endpoint type | Binary registry endpoint analyzed as a time-to-event outcome |
| Enrollment | 2564 |
| Start | 2006-02 |
| Primary completion | 2009-05 |
| Lead sponsor | Boehringer Ingelheim |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT00291330 |
2. Clinical Question
The central statistical question was whether dabigatran 150 mg was sufficiently similar to warfarin with respect to the time to recurrent symptomatic VTE or death related to VTE to satisfy a prespecified non-inferiority criterion. A separate analysis assessed superiority over a 6-month period.
Population
Participants with acute symptomatic venous thromboembolism enrolled in a phase 3 randomized treatment trial.
Intervention
Dabigatran etexilate 150 mg.
Comparator
Warfarin with an INR target of 2–3.
Primary question
Is dabigatran non-inferior to warfarin for recurrent symptomatic VTE and VTE-related death, using prespecified hazard-ratio and risk-difference margins?
3. Trial Design
Dabigatran 150 mg
- Dabigatran etexilate 150 mg.
- Participants were assigned to this group according to randomization for efficacy analyses.
Warfarin
- Warfarin.
- INR target of 2–3.
- Participants were assigned to this group according to randomization for efficacy analyses.
4. Endpoint Definition and Time Frames
The registry defines the primary endpoint as the number of participants with recurrent symptomatic VTE and deaths related to VTE. The registry states that all suspected recurrent VTEs and all deaths and bleeding events were evaluated by an independent central adjudication committee, and analyses were based on events centrally confirmed by that committee.
| Endpoint | Registry time frame | Statistical role |
|---|---|---|
| Recurrent symptomatic VTE and deaths related to VTE | Statistical analysis 1: from randomisation to end of post treatment period, planned to be up to day 224. Statistical analysis 2: from randomisation to 6 months, up to day 180. | Primary |
| Recurrent symptomatic VTE and all deaths | From randomisation to 6 months, up to day 180, and from randomisation to end of post-treatment period, planned to be up to day 224. | Secondary |
| Recurrent symptomatic DVT | From randomisation to 6 months, up to day 180, and from randomisation to end of post-treatment period, planned to be up to day 224. | Secondary |
| Recurrent symptomatic non-fatal PE | From randomisation to 6 months, up to day 180, and from randomisation to end of post-treatment period, planned to be up to day 224. | Secondary |
| Death due to VTE | From randomisation to 6 months, up to day 180, and from randomisation to end of post-treatment period, planned to be up to day 224. | Secondary |
| Death, any cause | From randomisation to 6 months, up to day 180, and from randomisation to end of post-treatment period, planned to be up to day 224. | Secondary |
| Bleeding events | From first intake of study drug to last intake of study drug + 6 days washout. | Secondary |
5. Analysis Populations
The registry distinguishes between the full analysis set used for efficacy analyses and the treated set used for bleeding analyses. This distinction is important because the treatment assignment rule is different from the exposure rule.
| Population | Definition | Role |
|---|---|---|
| Full analysis set (FAS) | All randomised patients documented to have taken at least one dose of study drug. Patients were assigned to treatment groups as randomised, regardless of the actual medication taken. | Primary and secondary efficacy analyses |
| Treated set (TS) | All randomised patients documented to have taken at least one dose of study drug. Patients were assigned to treatment groups as treated. | Bleeding-event analysis |
The efficacy analysis therefore retains the randomized treatment assignment even when actual treatment exposure may differ. This is an intention-to-treat principle applied within the registry's defined full analysis set.
6. Statistical Methodology
Cox proportional-hazards model
The registry used Cox proportional-hazards regression to compare time to the first occurrence of the endpoint between dabigatran and warfarin. The primary model included treatment, active cancer at baseline, symptomatic PE at baseline, and the interaction between active cancer and symptomatic PE.
A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the dabigatran group; a value above 1 indicates a higher estimated instantaneous event rate. The HR is not a probability, risk difference, or proportion of participants experiencing the endpoint.
Kaplan-Meier estimation
Kaplan-Meier estimates were used to estimate the probability of remaining free from the relevant event over time. For the primary risk-difference analysis, the registry obtained the overall risk difference at 6 months from Kaplan-Meier estimates after adjustment for the stratification factors active cancer at baseline and symptomatic PE at baseline.
Here, di is the number of events at event time ti and ni is the number at risk immediately before that time.
Covariate adjustment and stratification
The primary analyses incorporated active cancer at baseline and symptomatic PE at baseline. The Kaplan-Meier risk-difference analysis explicitly adjusted for these stratification factors. The Cox model additionally included their interaction.
Censoring
For the post-treatment primary Cox analysis, participants without events were censored at the end of the post-treatment period, day 224. For the 6-month superiority analysis, participants without events were censored at day 180. Censoring allows participants who remain event-free during observed follow-up to contribute information without treating the absence of an observed event as indefinite follow-up.
7. Non-Inferiority Framework
The primary post-treatment Cox analysis used a prespecified non-inferiority margin of 2.75 for the hazard ratio. The Kaplan-Meier risk-difference analysis used a non-inferiority margin of 3.6%.
Hazard-ratio margin
Non-inferiority was evaluated against an upper hazard-ratio margin of 2.75. The reported 95% confidence interval was 0.65–1.70.
Risk-difference margin
Non-inferiority was evaluated against an upper risk-difference margin of 3.6%. The reported 95% confidence interval was −0.80–1.50.
For a non-inferiority analysis, the key question is whether the confidence interval remains within the prespecified region defining clinically acceptable loss of efficacy. This is different from asking only whether a conventional two-sided superiority test produces a small P-value.
A hazard ratio of 1.05 is not interpreted in isolation. The relevant non-inferiority comparison is between the confidence interval and the prespecified upper margin of 2.75. Similarly, the risk difference of 0.40 must be interpreted against the 3.6% non-inferiority margin rather than simply against zero.
8. Primary Results: Post-Treatment Hazard Ratio
The first primary statistical analysis compared the time to first occurrence of recurrent symptomatic VTE or VTE-related death from randomization through the end of the post-treatment period, planned to be up to day 224.
Primary Cox analysis
95% CI: 0.65–1.70 · P < 0.0001
Non-inferiority margin: HR 2.75
| Characteristic | Reported result |
|---|---|
| Comparison | Dabigatran 150 mg vs Warfarin |
| Analysis population | Full analysis set |
| Method | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Estimate | 1.05 |
| 95% CI | 0.65–1.70 |
| P-value | <0.0001 |
| Hypothesis | Non-inferiority or equivalence |
| Non-inferiority margin | 2.75 |
The estimated hazard ratio of 1.05 indicates that the fitted Cox model estimated a slightly higher instantaneous rate of the primary endpoint under dabigatran than under warfarin, although the estimate itself is close to 1. The estimate does not mean that participants receiving dabigatran had a 5% higher probability of experiencing the endpoint.
The 95% CI of 0.65–1.70 describes uncertainty around the hazard-ratio estimate. Importantly for the non-inferiority question, the upper confidence limit of 1.70 is below the prespecified non-inferiority margin of 2.75. That is the central statistical relationship supporting the registry's non-inferiority analysis.
The P-value <0.0001 should not be interpreted as the magnitude of treatment benefit or harm. A P-value addresses evidence against a specified statistical null framework; it does not say that the treatment effect is clinically large or small. The non-inferiority conclusion instead depends on the prespecified margin and confidence interval.
The Cox interpretation also relies on the model's time-to-event framework and proportional-hazards assumption. Participants without events were censored at day 224, so the analysis concerns the event process over the specified follow-up rather than a simple comparison of cumulative proportions without regard to time.
9. Primary Results: Six-Month Risk Difference
The second primary analysis estimated the difference between groups in the proportion of participants with VTE or death related to VTE at 6 months, using Kaplan-Meier estimates at 180 days and adjustment for active cancer at baseline and symptomatic PE at baseline.
Primary Kaplan-Meier analysis
95% CI: −0.80% to 1.50% · P < 0.0001
Non-inferiority margin: 3.6%
| Characteristic | Reported result |
|---|---|
| Comparison | Dabigatran 150 mg vs Warfarin |
| Analysis population | Full analysis set |
| Method | Kaplan-Meier weighted estimates |
| Effect measure | Risk difference |
| Estimate | 0.40% |
| 95% CI | −0.80% to 1.50% |
| P-value | <0.0001 |
| Hypothesis | Non-inferiority or equivalence |
| Non-inferiority margin | 3.6% |
The estimated risk difference of 0.40% indicates that the Kaplan-Meier-estimated 6-month event probability was estimated to be 0.40 percentage points higher with dabigatran than with warfarin. This is an absolute difference, not a relative increase of 0.40%.
The 95% CI of −0.80% to 1.50% includes zero, so the interval is compatible with a small lower event probability, essentially no difference, or a small higher event probability. For the non-inferiority question, however, the upper limit of 1.50% remains below the prespecified 3.6% margin.
The P-value <0.0001 is not a measure of the size of the risk difference. The confidence interval and its relationship to the non-inferiority margin provide the more direct information for assessing whether the observed uncertainty is compatible with the prespecified non-inferiority criterion.
Because the estimate is based on Kaplan-Meier probabilities, it accounts for the time at which events and censoring occur rather than simply dividing the number of events by the number randomized.
10. Primary Results: Six-Month Superiority Analysis
A third primary analysis evaluated the same primary endpoint from randomization through day 180 using a Cox proportional-hazards model. Unlike the preceding analyses, this analysis was specified for superiority rather than non-inferiority.
Primary Cox superiority analysis
95% CI: 0.65–1.84 · P-value not below the conventional 0.05 threshold
Analysis period: randomisation through day 180
| Characteristic | Reported result |
|---|---|
| Comparison | Dabigatran 150 mg vs Warfarin |
| Analysis population | Full analysis set |
| Method | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Estimate | 1.10 |
| 95% CI | 0.65–1.84 |
| P-value | Not statistically significant at the conventional 0.05 level |
| Hypothesis | Superiority |
The estimated hazard ratio of 1.10 corresponds to an estimated instantaneous event rate that was higher under dabigatran than under warfarin in this particular six-month model, but the estimate is imprecise enough that the confidence interval also includes values below 1.
The 95% CI of 0.65–1.84 crosses 1.00, so the interval is compatible with either a lower or higher hazard under dabigatran. The result therefore does not establish superiority of dabigatran for this six-month analysis.
This analysis should also not be conflated with the non-inferiority analyses. Demonstrating non-inferiority and demonstrating superiority are different statistical questions with different reference points. A treatment can satisfy a non-inferiority criterion without demonstrating superiority.
The P-value is evidence about the specified superiority comparison, not a measure of clinical importance. Interpretation should focus jointly on the estimate, confidence interval, analysis population, follow-up period, and prespecified hypothesis.
11. Secondary Efficacy Results
The registry contains additional Cox and Kaplan-Meier analyses for recurrent symptomatic VTE and all deaths, recurrent symptomatic DVT, recurrent symptomatic non-fatal PE, VTE-related death, and death from any cause.
| Secondary endpoint | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Recurrent symptomatic VTE and all deaths | Kaplan-Meier | Risk difference 0.30% | −1.00% to 1.70% | 0.6220 |
| Recurrent symptomatic VTE and all deaths | Cox model | HR 1.00 | 0.69–1.46 | 0.9844 |
| Recurrent symptomatic DVT | Kaplan-Meier | Risk difference −0.20% | −1.10% to 0.70% | 0.6466 |
| Recurrent symptomatic DVT | Cox model | HR 0.76 | 0.40–1.42 | 0.3850 |
| Recurrent symptomatic non-fatal PE | Kaplan-Meier | Risk difference 0.30% | −0.30% to 1.00% | 0.2981 |
| Recurrent symptomatic non-fatal PE | Cox model | HR 2.00 | 0.86–4.68 | 0.1092 |
| Death due to VTE | Kaplan-Meier | Risk difference −0.30% | −1.20% to 0.60% | 0.5327 |
| Death due to VTE | Cox model | HR 0.33 | 0.03–3.15 | 0.3332 |
| Death, any cause | Kaplan-Meier | Risk difference −0.10% | −1.00% to 0.80% | 0.8018 |
| Death, any cause | Cox model | HR 0.94 | 0.54–1.63 | 0.8203 |
These secondary estimates illustrate why both point estimates and confidence intervals are needed. For example, the HR of 2.00 for recurrent symptomatic non-fatal PE is accompanied by a 95% CI of 0.86–4.68, which spans a wide range of possible relative hazards. Similarly, the HR of 0.33 for VTE-related death has a 95% CI of 0.03–3.15. The estimates alone therefore provide much less information than the estimates considered together with their uncertainty.
12. Bleeding Results
Bleeding events were analyzed in the treated set, with participants assigned to treatment groups as treated. The registry defined the bleeding-event observation period as beginning with first intake of study drug and continuing through last intake plus 6 days of washout, with the washout potentially reduced to 0 days if another anticoagulant therapy was taken on or after the last intake of active study drug.
| Bleeding analysis | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Major bleeding events | Cox proportional-hazards model | HR 0.82 | 0.45–1.48 | 0.5063 |
| Any bleeding events | Cox proportional-hazards model | HR 0.71 | 0.59–0.85 | 0.0002 |
The major-bleeding HR of 0.82 has a 95% CI of 0.45–1.48, which includes 1.00. The registry therefore does not establish a statistically significant difference in the hazard of major bleeding under the reported superiority analysis.
For any bleeding events, the reported HR is 0.71 with a 95% CI of 0.59–0.85 and P = 0.0002. Under the Cox model, this corresponds to an estimated instantaneous event rate under dabigatran that was approximately 71% of the rate under warfarin during the specified exposure and washout period.
The two bleeding analyses address different event definitions. The result for any bleeding should not be substituted for the result for major bleeding, and the hazard ratio should not be interpreted as an absolute percentage-point difference in the probability of bleeding.
13. Serious Adverse Events
The registry reports serious adverse events by treatment arm using affected participants over the stated at-risk population.
| Treatment group | Participants with serious adverse events | At risk |
|---|---|---|
| Dabigatran 150 mg | 165 | 1273 |
| Warfarin | 150 | 1266 |
Dabigatran
165 participants with serious adverse events among 1273 at risk.
Warfarin
150 participants with serious adverse events among 1266 at risk.
These counts are descriptive safety information. They should not be treated as a formal comparison of serious-adverse-event risk because the ClinicalTrials.gov record does not give a corresponding statistical analysis for this specific serious-adverse-event measure.
14. Statistical Methods Explained
Why was a Cox proportional-hazards model used?
The principal efficacy outcomes were time-to-event outcomes. Cox regression uses both whether an event occurred and when it occurred, while allowing participants who remain event-free to contribute information until censoring. The registry also incorporated treatment, active cancer at baseline, symptomatic PE at baseline, and their interaction in the primary Cox model.
What does an HR of 1.05 mean?
An HR of 1.05 means that the fitted model estimated the instantaneous event rate under dabigatran to be 1.05 times that under warfarin over the analyzed follow-up. It does not mean that 5% more participants experienced the event. For the non-inferiority question, the important comparison is between the confidence interval and the prespecified margin of 2.75.
Why is the non-inferiority margin more important than a conventional P-value?
Non-inferiority asks whether the treatment is not unacceptably worse than the comparator by more than a prespecified amount. For the primary Cox analysis, the upper confidence limit of 1.70 was below the 2.75 margin. For the Kaplan-Meier risk-difference analysis, the upper confidence limit of 1.50% was below the 3.6% margin. Those comparisons directly address the non-inferiority criterion.
Why was a risk difference also reported?
The hazard ratio summarizes a relative time-to-event effect, whereas the risk difference expresses an absolute difference at a specified time point. Here, the registry estimated the 6-month risk difference using Kaplan-Meier estimates. Reporting both measures helps distinguish relative treatment effects from absolute differences in event probability.
Why were active cancer and symptomatic PE included in the analysis?
The primary Cox model included active cancer at baseline, symptomatic PE at baseline, and their interaction. The Kaplan-Meier risk-difference analysis adjusted for the two stratification factors. Incorporating these prespecified baseline factors can improve precision and aligns the analysis with the trial's stratified design.
Why are the efficacy and bleeding populations different?
The efficacy analyses used the full analysis set and assigned participants according to randomized treatment. Bleeding analyses used the treated set and assigned participants as treated. This distinction reflects the different statistical purposes: efficacy preserves the treatment assignment established by randomization, while exposure-related safety analyses are linked to actual treatment received.
What does the 95% confidence interval tell us?
The confidence interval describes statistical uncertainty around the estimated treatment effect under the specified model and analysis framework. It does not describe the range of individual treatment responses. A narrow interval indicates greater precision than a wide interval, but clinical interpretation still requires consideration of the relevant endpoint and prespecified margin.
15. Reading the Non-Inferiority Results Correctly
| Analysis | Estimate | 95% CI | Non-inferiority margin | Key comparison |
|---|---|---|---|---|
| Cox, post-treatment period | HR 1.05 | 0.65–1.70 | 2.75 | Upper CI 1.70 is below 2.75 |
| Kaplan-Meier, 6 months | Risk difference 0.40% | −0.80% to 1.50% | 3.6% | Upper CI 1.50% is below 3.6% |
This table captures the central statistical structure of the primary non-inferiority analyses. The point estimates are close to the null values, but the key non-inferiority evidence comes from the upper bounds of the confidence intervals remaining inside the prespecified acceptable-worse-than-comparator boundaries.
16. Why the P-values Need Context
The registry reports P < 0.0001 for both the post-treatment Cox analysis and the 6-month Kaplan-Meier risk-difference analysis. Those P-values are extremely small, but they do not quantify how large the treatment difference is.
Effect size
The effect size is represented by the hazard ratio or risk difference. These estimates describe the direction and magnitude of the observed treatment contrast.
Precision
The confidence interval describes uncertainty around the estimate and is particularly important for evaluating the non-inferiority margins.
P-value
The P-value addresses a statistical hypothesis. It is not the probability that the treatment is effective or the size of the treatment effect.
Clinical meaning
The clinical interpretation depends on the endpoint, absolute event probability, follow-up period, and prespecified non-inferiority criterion.
17. Proportional-Hazards Interpretation
The Cox model assumes a proportional-hazards structure for the treatment effect. In practical terms, the model represents the relative hazard between groups with a common hazard ratio over the modeled follow-up, conditional on the model specification.
A hazard ratio summarizes the relative instantaneous event rate under the model. It does not directly state the absolute probability that an individual participant will experience recurrent VTE or death.
The distinction matters particularly when communicating non-inferiority results. The statement that an HR is below a margin is a model-based statistical statement; it is not equivalent to saying that the treatments have the same absolute event probability at every point in time.
18. Censoring and Follow-Up Windows
The primary analyses used two related but distinct follow-up windows. The post-treatment analysis continued from randomization to the end of the post-treatment period, planned to be up to day 224. The second primary analysis evaluated the period through day 180.
Time zero
Participants enter the primary time-to-event risk set and are analyzed according to randomized treatment assignment in the full analysis set.
Six-month analysis point
The registry specifies a primary Kaplan-Meier risk-difference analysis and a separate Cox superiority analysis through 6 months.
End of post-treatment period
The post-treatment primary Cox analysis censored participants without events at the end of the post-treatment period.
Using different censoring times can produce different numerical estimates even when the underlying endpoint is the same. Therefore, the 1.05 hazard ratio and the 1.10 hazard ratio should not be treated as contradictory estimates; they correspond to different analysis periods and hypotheses.
19. Primary Endpoint Interpretation
The post-treatment Cox analysis estimated an HR of 1.05 with a 95% CI of 0.65–1.70, while the 6-month Kaplan-Meier analysis estimated a risk difference of 0.40% with a 95% CI of −0.80% to 1.50%. The upper confidence limits were below the prespecified non-inferiority margins of 2.75 and 3.6%, respectively.
These are non-inferiority statements, not statements that dabigatran and warfarin have identical effects. The confidence intervals permit a range of treatment effects, and the margins define the amount of potential inferiority considered acceptable under the prespecified statistical framework.
The six-month Cox analysis produced an HR of 1.10 with a 95% CI of 0.65–1.84. Because the confidence interval includes 1.00, this analysis does not establish superiority of dabigatran over warfarin for the primary endpoint during the six-month period.
20. Secondary Endpoint Interpretation
The secondary analyses provide a broader picture of individual components and related outcomes, but their estimates are not interchangeable.
| Finding | Statistical reading |
|---|---|
| Recurrent symptomatic VTE and all deaths, HR 1.00 | The point estimate is at the null value, with a 95% CI of 0.69–1.46. |
| Recurrent symptomatic DVT, HR 0.76 | The point estimate is below 1, but the 95% CI of 0.40–1.42 includes 1. |
| Recurrent symptomatic non-fatal PE, HR 2.00 | The point estimate is above 1, but the 95% CI of 0.86–4.68 is wide and includes 1. |
| VTE-related death, HR 0.33 | The point estimate is below 1, while the 95% CI of 0.03–3.15 is very wide and includes 1. |
| All-cause death, HR 0.94 | The point estimate is close to 1, with a 95% CI of 0.54–1.63. |
The pattern illustrates an important principle of clinical-trial interpretation: an individual point estimate should not be interpreted without its confidence interval. Several secondary estimates span a broad range of plausible treatment effects.
21. Multiplicity and Multiple Analyses
The registry contains 15 posted statistical analyses, including three analyses of the primary endpoint and multiple secondary endpoint analyses. Because many statistical comparisons are reported, the nominal P-value for any single secondary analysis should not automatically be treated as if it represented an independently powered confirmatory hypothesis.
The primary endpoint itself was evaluated under distinct non-inferiority and superiority frameworks. These should remain conceptually separate: a non-inferiority analysis asks whether the treatment's effect remains within a prespecified acceptable boundary, whereas a superiority analysis asks whether the treatment produces a statistically demonstrable improvement relative to the comparator.
22. Safety Interpretation
The bleeding analyses show different patterns for major bleeding and any bleeding. The reported HR for major bleeding was 0.82 with a 95% CI of 0.45–1.48, whereas the HR for any bleeding was 0.71 with a two-sided 95% CI of 0.59–0.85.
Major bleeding
The confidence interval crosses 1.00, so the registry's reported analysis does not establish a statistically significant difference in the hazard of major bleeding.
Any bleeding
The HR of 0.71 and 95% CI of 0.59–0.85 indicate a lower estimated hazard of any bleeding under dabigatran during the specified treatment and washout period.
Serious adverse events
165/1273 participants were affected in the dabigatran group and 150/1266 in the warfarin group.
Different endpoints
Major bleeding, any bleeding, and serious adverse events are distinct safety measures and should not be collapsed into a single outcome.
23. Important Limitations and Interpretation Issues
- Non-inferiority margin: interpretation depends on the prespecified margins of 2.75 for the hazard ratio and 3.6% for the risk difference.
- Different follow-up windows: the primary analyses use day 180 and the end of the post-treatment period, planned to be up to day 224, so their estimates are not expected to be numerically identical.
- Analysis population: the full analysis set included randomized patients documented to have taken at least one dose, while bleeding analyses used the treated set.
- Censoring: participants without events were censored at the specified end of follow-up. Time-to-event estimates therefore depend on both observed events and censoring information.
- Cox model assumptions: hazard-ratio interpretation depends on the proportional-hazards framework and the model specification.
- Secondary endpoints: multiple secondary analyses create a broader multiplicity context than a single confirmatory hypothesis.
- Wide secondary confidence intervals: several secondary outcomes have confidence intervals spanning substantial ranges of possible effects.
- Serious adverse events: the reported affected/at-risk counts are descriptive and do not by themselves provide a formal treatment-effect estimate.
24. Why This Trial Matters Statistically
RE-COVER I is a useful teaching case because it demonstrates how a randomized clinical trial can evaluate non-inferiority using more than one time-to-event representation. The trial combines Cox regression, Kaplan-Meier estimation, absolute risk differences, hazard ratios, covariate adjustment, stratification, censoring, and a prespecified non-inferiority margin.
| Concept | How it appears in RE-COVER I |
|---|---|
| Randomization | Participants were randomly assigned in a parallel-group phase 3 design. |
| Double blinding | The registry identifies the trial as double masked. |
| Intention-to-treat principle | Efficacy participants were assigned to groups as randomized in the full analysis set. |
| Kaplan-Meier estimation | Used to estimate 6-month event probabilities and risk differences. |
| Cox model | Used for time-to-event comparisons and hazard-ratio estimation. |
| Hazard ratio | Used for primary and secondary time-to-event comparisons. |
| Risk difference | Used to express the adjusted 6-month absolute difference. |
| Non-inferiority design | Primary analyses used prespecified upper margins for HR and risk difference. |
| Stratified analysis | Active cancer at baseline and symptomatic PE at baseline were incorporated into the primary analyses. |
| Censoring | Event-free participants were censored at the specified follow-up boundaries. |
| Safety analysis | Bleeding outcomes were analyzed in the treated set rather than the randomized full analysis set. |
25. Statistical Takeaways
Non-inferiority
The post-treatment HR estimate was 1.05 with a 95% CI of 0.65–1.70, below the prespecified upper margin of 2.75.
Absolute difference
The 6-month Kaplan-Meier risk difference was 0.40%, with a 95% CI of −0.80% to 1.50%, below the 3.6% non-inferiority margin at its upper bound.
Superiority
The six-month superiority Cox analysis produced HR 1.10 with a 95% CI of 0.65–1.84, which includes 1.00.
Bleeding
Any bleeding had HR 0.71 (95% CI 0.59–0.85), while major bleeding had HR 0.82 (95% CI 0.45–1.48).
26. Related Tutorials
Learn more about the methods used in this trial:
27. Related Statistical Calculators
28. Sources
- ClinicalTrials.gov: NCT00291330.
- PubMed record: PMID 27807306.
- PubMed record: PMID 24344086.
- PubMed record: PMID 24081972.
- PubMed record: PMID 19966341.
Continue through the Clinical Biostats statistical library
Explore the statistical methods and calculators that support clinical-trial analysis, survival analysis, treatment-effect estimation, and trial design.
29. Record Summary
RE-COVER I provides a clear example of how non-inferiority evidence is constructed for a time-to-event endpoint. The primary post-treatment Cox analysis reported HR 1.05 with a 95% CI of 0.65–1.70 against a non-inferiority margin of 2.75, while the six-month Kaplan-Meier analysis reported a risk difference of 0.40% with a 95% CI of −0.80% to 1.50% against a 3.6% margin. The separate six-month superiority analysis produced HR 1.10 with a 95% CI of 0.65–1.84. Secondary analyses extended the statistical picture to recurrent DVT, non-fatal PE, VTE-related death, all-cause death, and bleeding.