← Clinical Trials
Cardiovascular Phase 3 Time-to-Event Analysis NCT00526474

TRA 2P-TIMI 50: Complete Statistical Analysis of Vorapaxar in Atherosclerosis

An independent statistical analysis of the randomized phase 3 TRA 2P-TIMI 50 trial evaluating vorapaxar versus placebo for prevention of cardiovascular events in participants with atherosclerosis.

Trial status: COMPLETED  ·  Enrollment: 26449.0  ·  Primary completion: 2011-12-01
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.

1. Trial at a Glance

TRA 2P-TIMI 50 was a randomized, parallel-group, triple-masked, phase 3 prevention trial comparing vorapaxar with placebo in participants with atherosclerosis and related cardiovascular conditions. The registry reports a primary time-to-event endpoint analyzed with a Cox proportional-hazards model.

26449.0
Enrolled
Phase 3
2
Arms
Vorapaxar vs placebo
0.88
Primary HR
95% CI 0.82–0.95
0.001
Primary P-value
Two-sided
FeatureTRA 2P-TIMI 50
Trial nameTRA 2P-TIMI 50
NCT identifierNCT00526474
Brief titleTrial to Assess the Effects of Vorapaxar (SCH 530348; MK-5348) in Preventing Heart Attack and Stroke in Patients With Atherosclerosis (TRA 2°P - TIMI 50) (P04737)
Phase3
StatusCOMPLETED
Therapeutic areaCardiovascular
ConditionsAtherosclerosis; Ischemia; Myocardial Infarction; Cerebrovascular Accident; Peripheral Arterial Disease
AllocationRANDOMIZED
Design modelPARALLEL
MaskingTRIPLE
Primary purposePREVENTION
Enrollment26449.0
InterventionsVorapaxar; Placebo
Registered primary endpoints1
Outcome measures posted32
Statistical analyses posted32

2. Clinical Question

The primary statistical question was whether vorapaxar was superior to placebo for the time to first occurrence of cardiovascular death, myocardial infarction, stroke, or urgent coronary revascularization within 3 years from randomization.

Population

Participants enrolled in a phase 3 prevention trial involving atherosclerosis, ischemia, myocardial infarction, cerebrovascular accident, and peripheral arterial disease.

Intervention

Vorapaxar.

Comparator

Placebo.

Primary question

Does vorapaxar reduce the hazard of the registered composite cardiovascular endpoint relative to placebo?

3. Trial Design

01
Randomize26449.0 enrolled
02
Two armsVorapaxar / placebo
03
Triple maskingRandomized parallel design
04
FollowTime-to-event outcomes
05
AnalyzeCox proportional hazards
Allocation
RANDOMIZED
Participants were randomly assigned to one of two treatment arms.
Model
PARALLEL
The two intervention groups were followed as parallel randomized groups.
Masking
TRIPLE
The registry identifies the study as triple-masked.
Purpose
PREVENTION
The registered primary purpose was prevention.

Study timeline

2007-09-01

Study start

The registry lists 2007-09-01 as the study start date.

2011-12-01

Primary completion

The registry lists 2011-12-01 as the primary completion date.

Registry status

COMPLETED

The study is listed as completed, with results posted.

4. Endpoints

Primary endpoint

EndpointDefinition / analysis
Kaplan-Meier Estimate of the Percentage of Participants Who Experienced Cardiovascular (CV) Death, Myocardial Infarction (MI), Stroke, or Urgent Coronary Revascularization (UCR) Within 3 Years From Randomization Time frame: up to 3 years.

The time in days from study start to the first occurrence of CV death, MI, stroke, or UCR was recorded. A Clinical Endpoints Committee reviewed and adjudicated each suspected efficacy endpoint event while blinded to treatment. Participants without an endpoint event until last visit, or participants lost to follow-up without an event, were handled as time-to-event observations.

Secondary endpoints

The registry contains 15 secondary statistical analyses. All use time-to-event methodology and compare placebo with vorapaxar unless otherwise noted.

Secondary endpointPopulationHR (95% CI)P-value
CV death, MI, or stroke within 3 yearsITT0.87 (0.80–0.94)<0.001
GUSTO moderate or severe bleeding within 3 yearsAs Treated1.51 (1.31–1.74)<0.001
Clinically significant bleeding within 3 yearsAs Treated1.41 (1.31–1.51)<0.001
Death from any cause, MI, stroke, or UCR within 3 yearsITT0.91 (0.85–0.98)0.009
CV death or MI within 3 yearsITT0.86 (0.78–0.94)0.002
CV death, MI, stroke, UCR, or UH-VCIN within 3 yearsITT0.87 (0.81–0.93)<0.001
Death from any cause, MI, stroke, or any revascularization within 3 yearsITT0.91 (0.86–0.96)0.001
CV death within 3 yearsITT0.89 (0.76–1.04)0.151
MI within 3 yearsITT0.83 (0.74–0.93)0.001
UCR within 3 yearsITT0.88 (0.75–1.03)0.108
Stroke within 3 yearsITT0.97 (0.83–1.14)0.733
Death from any cause within 3 yearsITT0.95 (0.85–1.07)0.411
UH-VCIN within 3 yearsITT0.83 (0.74–0.93)0.001
Any revascularization within 3 yearsITT0.89 (0.83–0.95)<0.001
All secondary endpoint estimatesCox proportional-hazards regression with covariates for treatment and stratification factors, unless otherwise specified by the registry population definition.
Population distinction: the efficacy secondary endpoints above generally use the ITT population, defined by the registry as all participants randomly assigned to a treatment arm. The two bleeding endpoints use the As Treated population, defined as all participants who received at least 1 dose of study medication.

5. Statistical Methodology

Kaplan-Meier estimation

The primary endpoint is explicitly described as a Kaplan-Meier estimate of the percentage of participants experiencing the event within 3 years. This is appropriate for a time-to-event endpoint because it uses information from participants who have experienced the event as well as participants whose follow-up ends before an event is observed.

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

Here, di represents the number of events at event time ti, and ni represents the number at risk immediately before that time.

Cox proportional-hazards regression

The registry reports Cox proportional-hazards regression for the primary endpoint and the secondary and post-hoc time-to-event analyses. The corresponding effect measure is a hazard ratio.

Conceptual Cox model
h(t|X) = h0(t) exp(βX)

For a treatment indicator, the hazard ratio is represented by exp(β). A hazard ratio below 1 corresponds to a lower estimated hazard in the vorapaxar group relative to placebo under the fitted model.

Covariate adjustment and stratification

The registry analysis notes state that the hazard ratio was calculated with covariates for treatment and stratification factors. The registry also identifies covariate adjustment and stratified analysis as concepts in the primary analysis.

Intention-to-treat analysis

The primary endpoint was analyzed in the ITT population, defined by the registry as all participants randomly assigned to a treatment arm. This preserves treatment assignment as the basis of the primary comparison rather than redefining groups according to treatment exposure after randomization.

Superiority hypothesis

The primary analysis is explicitly classified as a superiority hypothesis. The reported confidence interval is two-sided at the 95% level, and the P-value is reported as 0.001.

6. Primary Result

Primary cardiovascular composite endpoint

HR 0.88

95% CI 0.82–0.95  ·  P = 0.001

ITT population  ·  Cox proportional-hazards regression  ·  Superiority

ElementReported result
EndpointCV death, MI, stroke, or UCR within 3 years from randomization
ComparisonPlacebo vs vorapaxar
Analysis populationIntent to Treat (ITT)
MethodCox Proportional Hazards Regression
Effect measureHazard ratio
Estimate0.88
95% CI0.82–0.95
P-value0.001
HypothesisSuperiority
Clinical Biostats interpretation

The reported hazard ratio of 0.88 means that, under the fitted Cox model, the estimated instantaneous hazard of the primary composite endpoint was 0.88 times that in the placebo group. Equivalently, 1 − 0.88 = 0.12, so the model-based relative hazard is approximately 12% lower with vorapaxar than with placebo.

The hazard ratio does not mean that exactly 12% of participants avoided an event, nor does it represent a 12-percentage-point difference in event probability. It is a relative time-to-event measure that incorporates the timing of events and censoring.

The 95% confidence interval of 0.82 to 0.95 describes statistical uncertainty around the estimated hazard ratio. Because the interval is entirely below 1, the reported interval is consistent with a lower hazard under the fitted model.

The P-value of 0.001 addresses evidence against the null hypothesis specified for the statistical test. It does not measure the magnitude of the treatment effect, the probability that vorapaxar is effective, or the probability that the null hypothesis is true.

The interpretation also depends on the proportional-hazards framework underlying the Cox model. A single hazard ratio is most straightforward to interpret when the relative hazards are reasonably represented by a common ratio over follow-up. The registry result alone does not provide a separate diagnostic assessment of that assumption.

7. Secondary Results

The registry reports 15 secondary time-to-event analyses. They provide a more detailed decomposition of the cardiovascular outcomes and also include two bleeding endpoints.

EndpointHR95% CIP-valuePopulation
CV death, MI, or stroke0.870.80–0.94<0.001ITT
GUSTO moderate or severe bleeding1.511.31–1.74<0.001As Treated
Clinically significant bleeding1.411.31–1.51<0.001As Treated
Death, MI, stroke, or UCR0.910.85–0.980.009ITT
CV death or MI0.860.78–0.940.002ITT
CV death, MI, stroke, UCR, or UH-VCIN0.870.81–0.93<0.001ITT
Death, MI, stroke, or any revascularization0.910.86–0.960.001ITT
CV death0.890.76–1.040.151ITT
MI0.830.74–0.930.001ITT
UCR0.880.75–1.030.108ITT
Stroke0.970.83–1.140.733ITT
Death from any cause0.950.85–1.070.411ITT
UH-VCIN0.830.74–0.930.001ITT
Any revascularization0.890.83–0.95<0.001ITT
Clinical Biostats interpretation

The secondary results show why composite endpoints should be examined component by component. Several cardiovascular composites have hazard ratios below 1, while individual endpoints such as CV death, UCR, stroke, and death from any cause have confidence intervals that include 1. The individual components therefore do not all reproduce the same statistical pattern as the primary composite.

The bleeding endpoints point in the opposite direction: the reported hazard ratios are 1.51 for GUSTO moderate or severe bleeding and 1.41 for clinically significant bleeding. A hazard ratio above 1 indicates a higher estimated hazard in the vorapaxar group relative to placebo for these endpoints.

These secondary P-values should not automatically be interpreted as though every endpoint represented an independent confirmatory hypothesis with a separate type I error budget. The ClinicalTrials.gov record identifies the analyses as superiority analyses but do not provide a multiplicity-adjustment framework for this set of secondary comparisons.

8. Safety

The registry supplies serious adverse-event counts by treatment arm:

ArmSerious adverse events affected / at risk
Placebo3419 / 13166
Vorapaxar3514 / 13186

The registry also includes bleeding as a secondary time-to-event outcome. GUSTO moderate or severe bleeding had a hazard ratio of 1.51 (95% CI 1.31–1.74; P <0.001), while clinically significant bleeding had a hazard ratio of 1.41 (95% CI 1.31–1.51; P <0.001), both in the As Treated population.

Safety interpretation: serious adverse-event counts and time-to-event bleeding analyses answer different statistical questions. The affected/at-risk counts describe participants with serious adverse events, whereas the Cox analyses compare the timing of first qualifying bleeding events over follow-up. They should not be treated as interchangeable measures.

9. Post-Hoc Analyses

The registry reports 16 post-hoc statistical analyses. These analyses include an intended label population consisting of all enrolled participants with coronary arterial disease (CAD) or peripheral arterial disease (PAD) and no history of stroke or transient ischemic attack (TIA), as well as an intended label safety population with the same CAD/PAD and no-stroke/TIA criteria who received at least 1 dose of study drug.

Post-hoc endpointPopulationHR95% CIP-value
CV death, MI, stroke, or UCRIntended Label Population0.830.76–0.90<0.001
CV death, MI, or strokeIntended Label Population0.800.73–0.89<0.001
GUSTO moderate or severe bleedingIntended Label Safety Population1.451.23–1.71<0.001
Clinically significant bleedingIntended Label Safety Population1.421.31–1.54<0.001
Death, MI, stroke, or UCRIntended Label Population0.860.79–0.93<0.001
CV death or MIIntended Label Population0.830.75–0.93<0.001
CV death, MI, stroke, UCR, or UH-VCINIntended Label Population0.830.77–0.90<0.001
Death, MI, stroke, or any revascularizationIntended Label Population0.890.83–0.95<0.001
CV death, intended label populationIntended Label Population0.860.71–1.030.108
MIIntended Label Population0.820.73–0.930.002
UCRIntended Label Population0.880.74–1.040.127
StrokeIntended Label Population0.670.52–0.870.002
Death from any causeIntended Label Population0.920.80–1.060.249
UH-VCINIntended Label Population0.860.76–0.980.019
Any revascularizationIntended Label Population0.890.83–0.960.003

All post-hoc analyses use Cox proportional-hazards regression with treatment and stratification-factor covariates or the corresponding registry-described covariate adjustment. The endpoint time frame is up to 3 years for each analysis.

Why the post-hoc label matters: post-hoc analyses can be useful for understanding how the registered results relate to specific populations or endpoint definitions, but their statistical role is different from a prespecified primary analysis. They should therefore be interpreted as additional analyses rather than silently promoted to primary confirmatory evidence.

10. Important Registry Caveat: Stroke

The registry contains an important study-level caveat: prior to completion of the study, based on a communication from the Data and Safety Monitoring Board, all participants who had experienced a stroke either before or during the study had study drug discontinued.

Why this matters statistically: the treatment-exposure history is therefore not necessarily equivalent to a simple uninterrupted randomized treatment assignment for participants who experienced stroke. This is particularly relevant when comparing efficacy analyses based on the ITT population with safety or exposure-based analyses.

11. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

The endpoints in this registry record are time-to-event outcomes. A Cox model is designed to compare event hazards between groups while incorporating the timing of events and censoring. The TRA 2P-TIMI 50 registry analyses explicitly report Cox proportional-hazards regression.

What does a hazard ratio of 0.88 mean?

A hazard ratio of 0.88 means the estimated instantaneous hazard in the vorapaxar group is 0.88 times the corresponding hazard in the placebo group under the fitted model. It does not mean that 12% fewer participants necessarily experienced the endpoint.

Why is the Kaplan-Meier method paired with the Cox model?

They answer related but different questions. Kaplan-Meier estimation describes the event-free survival pattern over time, whereas the Cox model provides a relative hazard measure comparing treatment groups. Using both provides complementary information about the time-to-event distribution.

Why does the analysis population matter?

The primary endpoint uses the ITT population, preserving randomized assignment. The bleeding endpoints use the As Treated population, consisting of participants who received at least 1 dose. These populations answer different questions and should not be conflated.

What does the 95% confidence interval tell us?

The 95% confidence interval quantifies uncertainty around the estimated hazard ratio under the statistical model and sampling framework. A narrow interval indicates greater precision than a wider interval, but the interval is not a range containing the individual treatment effects experienced by patients.

Why does a P-value not measure effect size?

A P-value measures the compatibility of the observed data with a specified null hypothesis under the statistical testing framework. It depends on both the estimated effect and the amount of information in the data. Effect size is described by the hazard ratio and its confidence interval, not by the P-value alone.

What should be considered when interpreting many secondary and post-hoc P-values?

Multiple statistical tests create opportunities for apparently small P-values to occur by chance even when individual null hypotheses are true. The ClinicalTrials.gov record identifies the hypothesis type as superiority but do not specify a complete multiplicity-adjustment strategy for all secondary and post-hoc analyses. These results should therefore be interpreted in the context of the primary endpoint and the overall analysis plan.

12. Interpreting the Primary Hazard Ratio

Relative effect

The primary HR of 0.88 corresponds to an estimated hazard approximately 12% lower in the vorapaxar group relative to placebo, because 1 − 0.88 = 0.12.

Precision

The 95% CI of 0.82–0.95 indicates that the reported estimate is accompanied by uncertainty, while the entire interval remains below 1.

What it does not mean

The HR is not an absolute risk reduction, a number needed to treat, a probability of individual benefit, or a statement that all patients experience the same relative reduction in event hazard.

Model dependence

The HR is generated by a Cox proportional-hazards model. Its interpretation therefore depends on the model framework, including the proportional-hazards assumption and the handling of censoring.

13. Comparing Efficacy and Bleeding Endpoints

Endpoint classHazard ratio95% CIInterpretive direction
Primary CV death / MI / stroke / UCR0.880.82–0.95HR below 1 favors lower event hazard with vorapaxar
CV death / MI / stroke0.870.80–0.94HR below 1
MI0.830.74–0.93HR below 1
Stroke0.970.83–1.14CI includes 1
Death from any cause0.950.85–1.07CI includes 1
GUSTO moderate or severe bleeding1.511.31–1.74HR above 1
Clinically significant bleeding1.411.31–1.51HR above 1

This table illustrates an important feature of clinical-trial interpretation: different endpoints can move in different directions. A statistical analysis should therefore preserve the endpoint definitions rather than collapsing efficacy and safety into a single undifferentiated conclusion.

14. Why This Trial Matters Statistically

TRA 2P-TIMI 50 is a useful teaching case because the registry combines randomized treatment assignment with a large set of time-to-event outcomes, covariate-adjusted Cox models, stratified analyses, ITT efficacy analyses, As Treated safety analyses, and post-hoc population definitions.

ConceptHow it appears in TRA 2P-TIMI 50
RandomizationParticipants were randomly assigned to vorapaxar or placebo.
Triple maskingThe registry classifies the study as triple-masked.
Parallel designThe registry classifies the design model as PARALLEL.
Kaplan-Meier estimationThe primary and secondary endpoint definitions are expressed as Kaplan-Meier estimates.
Cox proportional-hazards modelThe registry reports Cox proportional-hazards regression for the statistical analyses.
Hazard ratioThe primary and other time-to-event analyses use hazard ratios as the effect measure.
Confidence intervalsThe primary and posted analyses include two-sided 95% confidence intervals.
ITT analysisThe primary endpoint is analyzed in the ITT population.
As Treated analysisThe two bleeding secondary endpoints use the As Treated population.
Covariate adjustmentThe registry analysis notes describe adjustment for treatment and stratification factors.
Stratified analysisStratified analysis is identified as an analysis concept in the registry.
Superiority testingThe primary and posted analyses are classified under a superiority hypothesis.
Post-hoc analysisSixteen post-hoc time-to-event analyses are posted.
Safety analysisSerious adverse events are reported by treatment arm, alongside time-to-event bleeding analyses.

15. Limitations

16. A Statistical Reading of the Complete Results

The primary result is a Cox-model hazard ratio of 0.88 with a two-sided 95% CI of 0.82–0.95 and P = 0.001 for the composite of CV death, MI, stroke, or UCR within 3 years. The corresponding analysis is based on the ITT population and adjusts for treatment and stratification factors.

The secondary cardiovascular endpoints generally report hazard ratios below 1, including 0.87 for CV death, MI, or stroke; 0.86 for CV death or MI; 0.83 for MI; 0.83 for UH-VCIN; and 0.89 for any revascularization. At the same time, some individual outcomes have confidence intervals that include 1, including CV death, UCR, stroke, and death from any cause.

The bleeding analyses provide an important counterpoint. GUSTO moderate or severe bleeding has an HR of 1.51 and clinically significant bleeding has an HR of 1.41, both with P <0.001. Thus, the registry contains efficacy and safety endpoints with different estimated hazard directions.

The post-hoc analyses narrow the population for several cardiovascular endpoints to participants with CAD or PAD and no history of stroke or TIA. Within those analyses, the registry reports HRs of 0.83 for the primary composite, 0.80 for CV death, MI, or stroke, and 0.67 for stroke, among other results. Because these analyses are post-hoc, their statistical role should remain distinct from that of the prespecified primary analysis.

17. Related Tutorials

Learn more about the methods used in this trial:

18. Related Calculators

19. Sources

Continue through the Clinical Biostats statistical pathway

Use the trial's methods as a starting point for deeper study of survival analysis, hazard ratios, randomized clinical-trial design, and statistical inference.

20. Record Summary

TRA 2P-TIMI 50 provides a detailed example of time-to-event analysis in a large randomized cardiovascular prevention trial. The registry reports a primary composite endpoint analyzed with a covariate-adjusted Cox proportional-hazards model in the ITT population, with a hazard ratio of 0.88, a two-sided 95% confidence interval of 0.82–0.95, and P = 0.001.

The broader registry results illustrate several important statistical principles: composite endpoints must be interpreted alongside their components; efficacy and safety can show different hazard directions; ITT and As Treated populations answer different questions; confidence intervals provide information about precision that P-values do not; and post-hoc analyses should remain clearly separated from prespecified primary analyses.

Clinical Biostats methodology: This page separates the numerical results reported in the trial record from statistical explanation. The purpose is to show how the design, analysis population, endpoint definition, effect measure, confidence interval, and P-value fit together rather than reducing the trial to a single numerical result.