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
OASIS-5 was a randomized, double-blind, parallel phase 3 treatment trial in thromboembolism comparing fondaparinux with enoxaparin. The registry records 20,078 enrolled participants and two primary endpoints assessed through Day 9.
| Feature | OASIS-5 |
|---|---|
| Trial acronym | OASIS-5 |
| Registry title | Michelangelo - Oasis 5 |
| Phase | Phase 3 |
| Condition | Thromboembolism |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 20,078 |
| Number of arms | 2 |
| Interventions | Fondaparinux and enoxaparin |
| Lead sponsor | GlaxoSmithKline |
| Sponsor type | Industry |
| Status | Completed |
| Start date | 2003-04 |
| Primary completion date | 2005-12 |
| ClinicalTrials.gov | NCT00139815 |
2. Clinical Question
The central statistical question is whether two randomized treatment groups differ with respect to the first occurrence of the registry-defined primary composite endpoint through Day 9, while also evaluating the incidence of adjudicated major bleeding through Day 9.
Population
The registry identifies the condition under study as thromboembolism and records an enrollment of 20,078 participants.
Intervention
Fondaparinux.
Comparator
Enoxaparin.
Primary question
How do the randomized treatment groups compare for the first occurrence of the primary composite endpoint and for adjudicated major bleeding through Day 9?
3. Trial Design
Fondaparinux
- Drug intervention.
- Randomized treatment assignment.
- Double-blind trial design.
- Parallel-group comparison.
Enoxaparin
- Drug comparator.
- Randomized treatment assignment.
- Double-blind trial design.
- Parallel-group comparison.
Randomization is the central design feature for the comparative statistical question. If treatment assignment is properly randomized, the treatment groups can be compared on outcomes without treating observed baseline differences as the primary source of the comparison. The double-blind structure is also important because knowledge of treatment assignment can influence outcome assessment, clinical management, reporting of events, and other aspects of trial conduct.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Primary endpoint 1 | death, myocardial infarction or refractory | up to and including Day 9 |
| Primary endpoint 2 | major bleeding | Up to Day 9 |
Composite endpoint
The first primary endpoint is recorded as “death, myocardial infarction or refractory”, with the time frame specified as up to and including Day 9. The registry description states that the endpoint concerns the first occurrence of any component of death, myocardial infarction or the recorded “refractory” component.
Because the registry wording available here ends with “refractory” rather than providing the full expanded term, the endpoint is presented using the registry wording rather than assigning an unstated clinical definition to that component.
Major bleeding
The second primary endpoint is the incidence of adjudicated major bleeding, with a time frame of Up to Day 9. The word “adjudicated” is statistically important because it indicates that the endpoint is based on a defined event-assessment process rather than simply an unreviewed count of reported bleeding events.
5. Statistical Structure of the Primary Questions
The two primary endpoints represent different statistical questions. The first is a time-to-first-event composite endpoint: participants can experience one or more component events, but the primary endpoint is defined around the first occurrence. Major bleeding is recorded separately as an incidence endpoint.
| Endpoint type | Statistical structure | Typical comparison |
|---|---|---|
| First occurrence of a composite event | Time from an appropriate trial origin until the first qualifying component event, with censoring for participants who remain event-free through the observation period. | Time-to-event methods such as Kaplan-Meier estimation, log-rank testing, and a Cox model may be appropriate when their assumptions and analysis plan are specified. |
| Incidence of adjudicated major bleeding | Occurrence of a defined safety event during the specified Day 9 window. | A risk or proportion comparison may be appropriate; a time-to-first-bleeding analysis may also be appropriate if event times and censoring are incorporated into the prespecified analysis. |
The distinction matters because an incidence proportion and a hazard ratio are not interchangeable quantities. A risk comparison asks about the proportion experiencing an event during a defined period. A hazard-based analysis instead compares instantaneous event rates over follow-up, conditional on remaining event-free immediately beforehand.
6. Planned Analysis
No formal statistical analyses were posted to ClinicalTrials.gov for the primary endpoints in the registry information available here. The registry therefore identifies the endpoints and their time frames, but does not provide posted treatment-effect estimates, confidence intervals, or p-values for those endpoints.
Primary composite endpoint
For a first-occurrence composite endpoint measured through Day 9, a standard time-to-event analysis would begin by defining the analysis time origin and counting each participant's first qualifying event. Participants without a qualifying event during the observation period would contribute follow-up until the appropriate censoring time.
Kaplan-Meier estimation can describe the event-free probability over time. A log-rank test can provide a treatment-group comparison across the follow-up period, while a Cox proportional-hazards model can express the relative treatment effect as a hazard ratio. These methods are typical approaches for this endpoint structure; the ClinicalTrials.gov record does not specify that any particular one was prespecified.
Here, S(t) represents the probability of remaining free of the qualifying first event beyond time t. A treatment comparison can therefore be expressed through differences in survival curves or through a relative hazard measure, depending on the prespecified statistical analysis.
Major bleeding
For adjudicated major bleeding through Day 9, a simple analysis could compare the proportion of participants with at least one qualifying bleeding event in each randomized group. If the exact event time and censoring information are incorporated into the analysis, a time-to-first-bleeding framework could instead be used.
The distinction between these approaches should be established prospectively. A risk ratio, risk difference, odds ratio, and hazard ratio answer related but different questions and should not be substituted for one another without an explicit statistical rationale.
7. Statistical Methods Explained
Why does the composite endpoint focus on the first occurrence?
A first-occurrence composite endpoint assigns the primary event time to the earliest qualifying component event. This prevents multiple subsequent events in the same participant from being counted as independent primary events. The resulting analysis is therefore a comparison of time to the first qualifying event rather than a comparison of the total number of events experienced by participants.
What is gained by combining several clinical events into one endpoint?
A composite endpoint can increase the number of observed primary events and can capture several clinically relevant ways in which participants may experience an unfavorable outcome. Statistically, however, the components must be interpreted carefully. If components differ substantially in clinical importance or frequency, the composite treatment effect can be driven disproportionately by one component.
What does a hazard ratio measure?
A hazard ratio compares the estimated instantaneous event rates between treatment groups over the analyzed follow-up. An HR below 1 would indicate a lower estimated hazard in the numerator treatment group relative to the reference group. It does not directly equal a risk ratio or risk difference.
The expression illustrates the usual interpretation of a Cox-model hazard ratio. The exact numerical estimate for OASIS-5 is not reported in the registry information available here.
Why is censoring important?
Participants may complete the Day 9 observation period without experiencing the endpoint, or they may have follow-up that ends before the full observation window. Time-to-event methods account for the information contributed before censoring rather than automatically treating every participant without an observed event as equivalent.
Why is adjudication important for major bleeding?
Adjudication applies a defined assessment process to determine whether a reported event satisfies the endpoint definition. This can improve consistency in outcome classification and reduce dependence on unreviewed investigator descriptions when an event definition requires clinical judgment.
Why should the two primary endpoints be considered separately?
The composite efficacy endpoint and adjudicated major bleeding represent different outcome domains. The first concerns death, myocardial infarction, or the registry's recorded “refractory” component; the second concerns major bleeding. A complete statistical interpretation therefore requires separate effect estimates for each endpoint rather than collapsing them into a single measure.
8. Composite Endpoints and Component Interpretation
Composite endpoints are useful statistical designs but require more than simply reporting one overall p-value or hazard ratio. The interpretation should begin with the composite itself and then examine the individual components when those component results are available.
First-event definition
The primary composite is based on the first occurrence of any recorded component through Day 9.
Component frequency
A component that occurs much more frequently than the others can contribute disproportionately to the composite result.
Clinical importance
Components of different severity or clinical meaning should not automatically be interpreted as equivalent simply because they share one statistical endpoint.
Statistical power
Combining events can increase the total number of observed events, which can improve precision when the composite is clinically and statistically coherent.
For OASIS-5, the registry identifies the first occurrence of death, myocardial infarction or the recorded “refractory” component. Because no component-level event counts or effect estimates are posted in the ClinicalTrials.gov record, the relative contribution of each component cannot be quantified from the registry information alone.
9. Time-to-Event Analysis Through Day 9
The Day 9 time frame is short relative to many conventional survival analyses, but the underlying statistical principles are the same. Each participant contributes information over the prespecified observation interval, and the primary composite endpoint is based on the timing of the first qualifying event.
Randomized treatment comparison
Participants are assigned to fondaparinux or enoxaparin under the randomized, double-blind, parallel design.
Monitor for the primary composite
The first occurrence of death, myocardial infarction or the registry-recorded “refractory” component determines the primary composite event time.
Evaluate the primary endpoints
The primary composite is assessed up to and including Day 9, while adjudicated major bleeding is assessed up to Day 9.
When follow-up is short, the absolute event probability over the specified period can be particularly intuitive. A time-to-event model adds another dimension by using the timing of events rather than only whether an event occurred before the end of follow-up.
10. Interpreting a Potential Hazard Ratio
If a prespecified Cox analysis produced a hazard ratio below 1 for fondaparinux versus enoxaparin, that would indicate a lower estimated instantaneous rate of the primary event in the fondaparinux group over the analyzed follow-up. The registry does not post a numerical hazard ratio for this endpoint.
A hazard ratio would not mean that the same percentage of participants experienced a reduction in risk, nor would it directly provide the absolute number of participants who avoided the event. Absolute risks over the Day 9 period would provide that complementary information.
A 95% confidence interval would quantify statistical uncertainty around an estimated treatment effect under the specified model and sampling framework. A narrow interval indicates greater precision than a wide interval, but neither interval width nor the point estimate alone describes the clinical importance of every individual participant's outcome.
A p-value addresses the compatibility of the observed data with a specified null hypothesis under the statistical model. It does not describe the magnitude of a treatment effect. Effect estimates and their confidence intervals are needed to describe magnitude and precision.
11. Analysis of Major Bleeding
Major bleeding is a primary endpoint distinct from the thromboembolic composite. The registry defines it as the incidence of adjudicated major bleeding through Day 9.
| Feature | Major bleeding endpoint |
|---|---|
| Endpoint type | Safety endpoint |
| Definition | Incidence of adjudicated major bleeding |
| Time frame | Up to Day 9 |
| Randomized comparison | Fondaparinux versus enoxaparin |
| Posted treatment estimate | Not reported in the ClinicalTrials.gov record |
For a binary Day 9 safety outcome, the most direct descriptive quantities would be the number and proportion of participants with adjudicated major bleeding in each treatment group. An absolute risk difference would describe the difference in event probability, while a relative risk would describe the ratio of those probabilities.
If exact event times were available and a time-to-first-bleeding analysis were prespecified, the statistical framework could instead incorporate both event timing and censoring. That distinction is important because two studies can have the same Day 9 event proportions while having different distributions of when events occurred.
12. Randomization and Causal Interpretation
The randomized allocation is central to the causal interpretation of a treatment comparison. Randomization is intended to balance measured and unmeasured prognostic factors in expectation, so that systematic differences in outcomes between treatment groups can be attributed to the assigned intervention under the assumptions of the randomized trial.
What randomization supports
A comparison of outcomes according to assigned treatment, with treatment assignment separated from baseline prognostic factors by the randomization process.
What randomization does not guarantee
Exact numerical equality of every baseline characteristic between the two groups in the realized sample.
Why blinding matters
Double blinding can reduce the influence of treatment knowledge on outcome assessment and other aspects of trial conduct.
Why analysis populations matter
The interpretation of an effect depends on whether participants are analyzed according to randomized assignment, treatment received, or another prespecified population.
The registry information available here identifies the allocation as randomized and the masking as double, but it does not provide the detailed analysis-population definitions or statistical analysis plan needed to characterize those populations further.
13. What Is and Is Not Reported in the Registry
| Statistical element | Registry information available here |
|---|---|
| Enrollment | 20,078 |
| Primary endpoints | Two endpoints are registered |
| Primary endpoint time frames | Up to and including Day 9; Up to Day 9 |
| Treatment groups | Fondaparinux and enoxaparin |
| Randomized design | Yes |
| Parallel design | Yes |
| Double masking | Yes |
| Posted statistical analyses | None |
| Posted efficacy estimates | Not reported |
| Posted confidence intervals | Not reported |
| Posted p-values | Not reported |
The absence of posted treatment estimates means that the registry supports a detailed analysis of the trial's design and endpoint structure, but not a numerical estimate of the treatment effect for either primary endpoint.
14. Important Limitations and Interpretation Issues
- No posted statistical analyses: the ClinicalTrials.gov record does not provide treatment-effect estimates, confidence intervals, or p-values for the registered primary endpoints.
- Incomplete composite wording: the first primary endpoint is recorded as “death, myocardial infarction or refractory,” and the available registry description does not expand the recorded “refractory” component further.
- No component-level results: the registry information available here does not report event counts for the individual components of the primary composite.
- No treatment-group event counts: the available record does not provide the number of participants experiencing either primary endpoint in each arm.
- No detailed statistical analysis plan: the available registry information does not specify the exact inferential procedure, randomization ratio, stratification factors, multiplicity strategy, interim-analysis plan, or missing-data method.
- Short endpoint window: both primary endpoints are defined through Day 9, so their interpretation is specifically tied to that prespecified observation period rather than to longer-term outcomes.
- Composite endpoint interpretation: the overall composite can conceal differences among its individual components unless component-specific analyses are also examined.
- Safety interpretation: major bleeding should be evaluated separately from the efficacy composite because it represents a different clinical outcome and statistical question.
15. Why This Trial Matters Statistically
OASIS-5 is a useful statistical teaching case because its registry design combines randomized allocation, double masking, a parallel-group structure, a composite first-event endpoint, and a separately defined adjudicated bleeding endpoint.
| Concept | How it appears in OASIS-5 |
|---|---|
| Randomization | Participants were randomized to fondaparinux or enoxaparin. |
| Parallel-group design | The registry identifies the design model as parallel. |
| Double masking | The registry identifies the trial as double masked. |
| Composite endpoint | The first primary endpoint combines death, myocardial infarction or the recorded “refractory” component. |
| Time-to-event analysis | The primary composite is defined around first occurrence through Day 9. |
| Safety endpoint | Major bleeding is registered as a separate primary endpoint. |
| Adjudication | Major bleeding is defined as an incidence of adjudicated major bleeding. |
| Multiplicity | Two primary endpoints require an explicitly defined approach to confirmatory error control. |
| Censoring | Time-to-event methods would need to account for participants who remain event-free or otherwise leave observation before the endpoint occurs. |
| Effect measures | Hazard ratios, risks, risk differences, and other measures answer different statistical questions and should not be conflated. |
16. Statistical Interpretation Framework
Relative versus absolute effects
A relative treatment measure such as a hazard ratio summarizes the relationship between event rates, while an absolute measure such as a risk difference describes the change in event probability over a specified period. Both can be informative, but they answer different questions.
These are simple representations for a fixed-period binary endpoint. A hazard ratio instead compares modeled instantaneous event rates over time.
Why the Day 9 window matters
An endpoint measured through Day 9 describes what happened during that particular observation period. A treatment effect observed over a short window should not automatically be generalized to longer follow-up without corresponding data.
Why a composite result needs context
Suppose a treatment comparison showed a difference in the composite endpoint. That result would establish a difference in the combined endpoint under the prespecified analysis, but it would not automatically establish an identical effect for every component. Component-specific event frequencies and treatment effects would be needed to understand what drove the composite.
Why statistical significance and clinical magnitude are different
A small p-value can arise from a precise estimate in a large trial, while a clinically important estimate can be imprecisely estimated in a smaller trial. Enrollment of 20,078 participants makes precision an important consideration, but the registry's lack of posted estimates means the actual precision of the OASIS-5 treatment comparisons cannot be assessed from the available record alone.
17. Multiplicity: Two Primary Endpoints
OASIS-5 has two registered primary endpoints: the first occurrence of the composite endpoint through Day 9 and adjudicated major bleeding through Day 9. When more than one endpoint is designated primary, the statistical analysis plan normally needs to define how the overall type I error will be controlled across the primary hypotheses.
Why multiplicity matters
Testing multiple primary hypotheses can increase the chance of at least one false-positive conclusion if each hypothesis is tested independently at the same nominal significance level.
Possible prespecified strategies
Common approaches include hierarchical testing, alpha allocation, or other procedures designed to preserve the desired familywise error rate.
The available registry information does not specify which multiplicity procedure was used for the two primary endpoints. A numerical significance claim therefore cannot be reconstructed from the registry endpoint listing alone.
18. Missing Data and Censoring
For a fixed Day 9 incidence analysis, missing endpoint information can create uncertainty about whether a participant experienced the event during the observation window. For a time-to-event analysis, participants who do not experience the event can be censored at an appropriate time, allowing their available follow-up to contribute information.
The statistical treatment of incomplete observations should be defined before the results are examined. The registry information available here does not specify an imputation method or detailed missing-data strategy.
19. Statistical Methods Explained: A Deeper View
Why might Kaplan-Meier estimation be useful even with a Day 9 endpoint?
Kaplan-Meier estimation uses the timing of first events and accounts for censoring. Even over a relatively short observation period, it can provide a curve describing the probability of remaining event-free over time rather than reducing every participant to a simple yes/no indicator at the end of the window.
When would a risk ratio be preferable to a hazard ratio?
If the primary question is simply the proportion of participants with major bleeding by Day 9, a risk ratio directly compares those fixed-period probabilities. A hazard ratio incorporates event timing and therefore answers a different question.
What does a confidence interval add to a point estimate?
The point estimate is the single value produced by the statistical procedure. The confidence interval communicates sampling uncertainty around that estimate under the specified inferential framework. Two studies can have similar point estimates but substantially different precision.
Why can a composite endpoint improve statistical efficiency?
When multiple clinically related events are combined, the total number of first events can be greater than the number of any single component. More events can provide greater statistical information. The benefit is strongest when the components are clinically coherent and treatment effects are reasonably aligned.
Why can the composite endpoint also complicate interpretation?
A composite treatment effect can be driven primarily by its most frequent component. If one component is substantially more common than another, the overall estimate may tell us relatively little about the less frequent component. Component-specific results are therefore important whenever they are available.
20. Limitations of the Available Statistical Record
The ClinicalTrials.gov record establishes the major design and endpoint framework for OASIS-5, but the statistical information available here is not sufficient to reproduce a complete numerical analysis.
| Question | What can be established |
|---|---|
| How large was the trial? | Enrollment was 20,078. |
| What were the randomized treatments? | Fondaparinux and enoxaparin. |
| What was the design? | Randomized, double-blind, parallel, phase 3. |
| What were the primary endpoints? | The registered composite first-occurrence endpoint and adjudicated major bleeding. |
| When were they assessed? | Through Day 9. |
| What were the observed treatment effects? | Not posted in the ClinicalTrials.gov record available here. |
| What were the confidence intervals? | Not posted. |
| What were the p-values? | Not posted. |
This distinction is important for statistical education. A trial can have a highly informative design while the registry record itself does not contain enough numerical information to reproduce the final inferential results.
21. Why the Trial Design Supports a Direct Treatment Comparison
The randomized parallel-group structure creates a direct comparison between fondaparinux and enoxaparin. Double masking further reduces the possibility that knowledge of treatment assignment will systematically influence outcome assessment or other aspects of participant or investigator behavior.
Because the primary efficacy endpoint is based on first occurrence, each participant contributes one primary event time if an eligible event occurs. Participants without a qualifying event during the relevant period contribute information as event-free observations under an appropriate time-to-event framework.
The separate major-bleeding endpoint provides a distinct safety dimension. Statistically, this is important because efficacy and bleeding outcomes can move in different directions. A complete interpretation therefore requires separate estimates for the efficacy and safety questions rather than relying on a single combined summary.
22. Trial Timeline
Trial start
The OASIS-5 trial began in April 2003 according to the registry.
Randomized comparative design
The registry identifies OASIS-5 as a phase 3 randomized, double-blind, parallel treatment trial comparing fondaparinux with enoxaparin.
Primary completion
The registry records December 2005 as the primary completion date.
Registry status
The current trial record identifies the study status as completed.
23. Why This Trial Is a Useful Statistical Example
OASIS-5 illustrates how several core clinical-trial concepts interact within one randomized study. The primary efficacy endpoint is a first-occurrence composite, while major bleeding is a separately defined adjudicated safety endpoint. Both are measured over a short, prespecified period.
Endpoint construction
The composite endpoint demonstrates why the definition of an event must be established before statistical testing begins.
Time-to-event methods
The first-event structure naturally raises questions about event timing, censoring, survival curves, and hazard ratios.
Safety analysis
Major bleeding illustrates why adverse outcomes should be analyzed as distinct endpoints rather than folded into a generic efficacy summary.
Multiplicity
Two primary endpoints require an explicit inferential framework if confirmatory type I error control is intended.
24. Sources
- ClinicalTrials.gov: NCT00139815, OASIS-5 (Michelangelo - Oasis 5).
- PubMed record: PubMed 16537663.
- PubMed record: PubMed 20124123.
- PubMed record: PubMed 19825809.
Continue through Clinical Biostats
Explore additional clinical trial analyses, statistical tutorials, and practical statistical tools.
25. Record Summary
OASIS-5 was a completed randomized, double-blind, parallel phase 3 trial with 20,078 participants comparing fondaparinux with enoxaparin for thromboembolism. The registry identifies two primary endpoints: the first occurrence of death, myocardial infarction or refractory through and including Day 9, and the incidence of adjudicated major bleeding through Day 9.
From a statistical perspective, the trial provides a useful framework for understanding first-event composite endpoints, time-to-event analysis, censoring, hazard ratios, fixed-period risk measures, adjudicated safety outcomes, and multiplicity across multiple primary endpoints. The available ClinicalTrials.gov record does not post formal statistical analyses or numerical treatment-effect results, so the appropriate interpretation centers on the registered design and endpoint definitions rather than an unreported efficacy estimate.