← Clinical Trials
Prostatic Neoplasms Phase 3 Time-to-Event Analysis NCT00638690

COU-AA-301: Complete Statistical Analysis of Abiraterone Acetate in Castration-Resistant Prostate Cancer

An independent statistical analysis of the randomized phase 3 COU-AA-301 trial evaluating abiraterone acetate in castration-resistant prostate cancer previously treated with docetaxel-based chemotherapy, with overall survival as the registered primary endpoint.

Completed  ·  Enrollment 1195  ·  Start May 2008  ·  Primary completion August 2010
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical efficacy results and safety counts are taken from the ClinicalTrials.gov record. The registry record is the official source for the trial record.

Registry note: 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

COU-AA-301 was a randomized, parallel-group, quadruple-masked phase 3 treatment trial in prostatic neoplasms. The trial enrolled 1195 participants and compared abiraterone acetate with placebo, with overall survival registered as the primary time-to-event endpoint.

1195
Enrolled
Phase 3 trial
2
Arms
Parallel design
0.646
Overall Survival HR
95% CI 0.543–0.768
<0.0001
OS P-value
Two-sided
FeatureCOU-AA-301
Trial nameCOU-AA-301
NCT IDNCT00638690
PhasePhase 3
StatusCompleted
ConditionProstatic Neoplasms
Brief titleAbiraterone Acetate in Castration-Resistant Prostate Cancer Previously Treated With Docetaxel-Based Chemotherapy
AllocationRandomized
Design modelParallel
MaskingQuadruple
Primary purposeTreatment
Enrollment1195
Primary endpointOverall Survival
Primary endpoint typeTime-to-event
Hypothesis typeSuperiority
Lead sponsorCougar Biotechnology, Inc.

2. Clinical Question

The clinical question represented by the registered trial is whether treatment with abiraterone acetate produces a different overall-survival experience from placebo in patients described by the trial as having castration-resistant prostate cancer previously treated with docetaxel-based chemotherapy.

Population

Patients with castration-resistant prostate cancer, as specified by the trial's brief title, previously treated with docetaxel-based chemotherapy.

Intervention

Abiraterone acetate, with prednisone/prednisolone listed among the trial interventions.

Comparator

Placebo, with prednisone/prednisolone listed among the trial interventions.

Primary question

Does abiraterone acetate improve overall survival relative to placebo under the prespecified superiority framework?

3. Trial Design

01
Randomize1195 enrolled
02
Parallel armsAbiraterone vs placebo
03
Quadruple maskMasked trial design
04
Follow-upTime-to-event outcomes
05
AnalysisITT and stratified methods
INTERVENTION

Abiraterone Acetate

  • Abiraterone acetate
  • Prednisone/prednisolone listed as a trial intervention
COMPARATOR

Placebo

  • Placebo
  • Prednisone/prednisolone listed as a trial intervention
Allocation
Randomized
Design model
Parallel
Masking
Quadruple
Primary purpose
Treatment

The registry identifies the trial as randomized and parallel, with quadruple masking. These design features matter statistically because randomization establishes the treatment comparison, parallel allocation keeps participants in their assigned trial arm rather than assigning different treatments sequentially, and masking is intended to reduce the influence of treatment knowledge on trial conduct and assessment.

4. Endpoints

EndpointRoleDefinition / descriptionTime frame
Overall Survival Primary Overall survival is defined as the time interval from the date of randomization to the date of death from any cause. Up to 60 months
Time to Prostate-Specific Antigen Progression According to Prostate Specific Antigen Working Group Criteria Secondary Time-to-event endpoint as registered. Up to 12 months
Number of Patients Achieving a Prostate-Specific Antigen Decline ≥50% Secondary Number of patients achieving the registered PSA decline threshold. Up to 12 months
Radiographic Progression-free Survival Secondary Time-to-event endpoint as registered. Up to 11 months

The endpoint structure is statistically important because three of the four posted outcome measures are time-to-event endpoints. These endpoints are not adequately summarized by simply counting events: the timing of events and the information contributed by participants who have not yet experienced an event are both relevant.

5. Statistical Methodology

Intention-to-treat analysis

The posted analyses state that analysis was performed on the Intent-to-Treat (ITT) population. The registry describes this population as all patients randomized into the study, classified according to their assigned treatment.

For a randomized efficacy comparison, ITT analysis maintains the original treatment assignment as the basis for comparison. This preserves the treatment groups created by randomization rather than redefining the groups according to treatment actually received or later events.

Stratified log-rank testing

The primary overall-survival analysis used a log-rank test and is explicitly described as a stratified analysis. The same approach was used for the posted secondary time-to-event analyses.

A log-rank test compares the observed pattern of event occurrence between treatment groups over follow-up. A stratified version performs the comparison while accounting for prespecified or relevant strata rather than treating all participants as belonging to a single homogeneous risk set.

Hazard ratio

The primary endpoint used a hazard ratio as the effect measure. A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the abiraterone acetate group relative to placebo under the fitted time-to-event comparison.

Conceptual interpretation
HR < 1  →  lower estimated instantaneous event rate in the abiraterone acetate group

A hazard ratio is a relative time-to-event measure. It is not itself an absolute survival probability, an absolute risk difference, or the percentage of patients who benefit.

Chi-squared testing and risk ratio

The PSA-response endpoint was analyzed with a chi-squared test and summarized using a risk ratio. Unlike overall survival, this endpoint reduces the outcome to whether a participant achieved the registered PSA decline threshold within the specified time frame.

The risk ratio compares the probability of the specified outcome between the two groups. A value above 1 therefore indicates a higher observed probability of achieving the PSA response in the abiraterone acetate group relative to placebo.

6. Results

The registry data contain four posted statistical analyses: one primary analysis of overall survival and three secondary analyses. The primary and secondary analyses use the ITT population, and the posted time-to-event analyses are identified as stratified analyses.

Primary Endpoint: Overall Survival

Hazard ratio for overall survival

0.646

95% CI: 0.543–0.768   ·   P < 0.0001

Two-sided confidence interval · Superiority hypothesis · Stratified log-rank analysis

Primary endpointAbiraterone Acetate vs Placebo
OutcomeOverall Survival
Time frameUp to 60 months
Analysis populationIntent-to-Treat
Statistical methodLog-rank test
Analysis characteristicStratified analysis
Effect measureHazard ratio
Estimate0.646
95% CI0.543–0.768
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

The reported overall-survival hazard ratio of 0.646 means that, under the time-to-event analysis, the estimated instantaneous rate of death in the abiraterone acetate group was 64.6% of that in the placebo group. Expressed as a simple relative interpretation, this corresponds to an estimated 35.4% lower hazard because 1 − 0.646 = 0.354.

This does not mean that 35.4% of patients avoided death, that individual patients experienced exactly a 35.4% reduction in their probability of dying, or that the absolute survival probability differed by 35.4 percentage points.

The two-sided 95% confidence interval of 0.543–0.768 describes uncertainty around the estimated relative hazard under the statistical model and analysis framework. It does not describe the range of effects that individual patients can experience.

The P < 0.0001 result addresses evidence against the null hypothesis in the specified testing framework. A p-value is not a measure of the size of the treatment effect and should not be substituted for the hazard ratio or its confidence interval.

Because this is a time-to-event analysis, interpretation also depends on censoring and on the assumptions underlying the hazard-ratio framework. The registry data identify this analysis as stratified, but the ClinicalTrials.gov record does not provide further detail about the specific strata used.

Secondary Endpoint: Time to PSA Progression

Hazard ratio for time to PSA progression

0.580

95% CI: 0.462–0.728   ·   P < 0.0001

Two-sided confidence interval · Stratified log-rank analysis · ITT population

FeatureReported result
EndpointTime to Prostate-Specific Antigen Progression According to Prostate Specific Antigen Working Group Criteria
Time frameUp to 12 months
MethodLog-rank test
Analysis populationIntent-to-Treat
Analysis characteristicStratified analysis
Effect measureHazard ratio
Estimate0.580
95% CI0.462–0.728
P-value<0.0001
Clinical Biostats interpretation

The hazard ratio of 0.580 indicates a lower estimated instantaneous rate of PSA progression in the abiraterone acetate group relative to placebo under the reported survival-analysis framework. A direct relative interpretation is approximately 42.0% lower estimated hazard because 1 − 0.580 = 0.420.

The estimate is specific to the registered endpoint: time to PSA progression according to Prostate Specific Antigen Working Group criteria. It should not be interpreted as an overall-survival effect or as a statement that 42.0% of participants experienced prevention of PSA progression.

The 95% confidence interval of 0.462–0.728 quantifies uncertainty around the estimated hazard ratio. The P < 0.0001 value provides evidence against the null hypothesis in the reported testing framework, but it does not measure the magnitude or clinical importance of the effect.

Secondary Endpoint: PSA Decline ≥50%

Risk ratio for achieving a PSA decline ≥50%

5.266

95% CI: 3.459–8.018   ·   P < 0.001

Two-sided confidence interval · Chi-squared analysis · ITT population

FeatureReported result
EndpointNumber of Patients Achieving a Prostate-Specific Antigen Decline ≥50%
Time frameUp to 12 months
MethodChi-squared test
Analysis populationIntent-to-Treat
Effect measureRisk ratio / relative risk
Estimate5.266
95% CI3.459–8.018
P-value<0.001
Clinical Biostats interpretation

The risk ratio of 5.266 indicates that the probability of achieving the specified PSA decline threshold was estimated to be 5.266 times as high in the abiraterone acetate group as in the placebo group over the registered time frame.

This is a binary response measure, so it differs fundamentally from the hazard ratios for overall survival and PSA progression. It does not describe how quickly the response occurred, how long a response lasted, or how survival differed between groups.

The 95% confidence interval of 3.459–8.018 describes uncertainty around the risk-ratio estimate. The P < 0.001 value addresses statistical evidence against the null hypothesis; it does not say that the effect is "0.001" in magnitude or quantify the absolute percentage-point difference between groups.

Secondary Endpoint: Radiographic Progression-free Survival

Hazard ratio for radiographic progression-free survival

0.673

95% CI: 0.585–0.776   ·   P < 0.0001

Two-sided confidence interval · Stratified log-rank analysis · ITT population

FeatureReported result
EndpointRadiographic Progression-free Survival
Time frameUp to 11 months
MethodLog-rank test
Analysis populationIntent-to-Treat
Analysis characteristicStratified analysis
Effect measureHazard ratio
Estimate0.673
95% CI0.585–0.776
P-value<0.0001
Clinical Biostats interpretation

The hazard ratio of 0.673 indicates a lower estimated instantaneous rate of radiographic progression or the relevant event in the abiraterone acetate group relative to placebo under the reported time-to-event analysis. A simple relative interpretation is approximately 32.7% lower estimated hazard because 1 − 0.673 = 0.327.

The result does not mean that 32.7% of patients were progression-free, nor does it provide an absolute difference in progression-free survival probabilities. It is a relative time-to-event measure.

The 95% confidence interval of 0.585–0.776 indicates the statistical uncertainty around the estimated hazard ratio. The P < 0.0001 value indicates strong statistical evidence against the null hypothesis under the reported test, but it does not quantify the size of the effect.

7. Results Summary

EndpointRoleMethodEffect95% CIP-value
Overall Survival Primary Stratified log-rank HR 0.646 0.543–0.768 <0.0001
Time to PSA Progression Secondary Stratified log-rank HR 0.580 0.462–0.728 <0.0001
PSA Decline ≥50% Secondary Chi-squared RR 5.266 3.459–8.018 <0.001
Radiographic Progression-free Survival Secondary Stratified log-rank HR 0.673 0.585–0.776 <0.0001

Across the four posted analyses, the effect estimates all point in the same direction for the abiraterone acetate comparison: the three time-to-event hazard ratios are below 1, while the PSA-response risk ratio is above 1. The estimates answer different statistical questions and therefore should not be treated as interchangeable measures of treatment effect.

Important interpretation point: a hazard ratio and a risk ratio have different denominators, meanings, and time structures. The HRs summarize relative event hazards over time, whereas the RR for PSA decline compares the probability of achieving a binary response within the registered time frame.

8. Safety Results

The ClinicalTrials.gov record reports serious adverse events by arm using affected participants divided by the corresponding at-risk population. These are reported as counts and denominators rather than as a separately modeled comparative efficacy endpoint.

GroupSerious adverse eventsAffected / at risk
Abiraterone AcetateSerious adverse events365/791
PlaceboSerious adverse events175/394
Placebo to Abiraterone AcetateSerious adverse events29/67

The third category, Placebo to Abiraterone Acetate, is reported separately in the ClinicalTrials.gov record and should not be silently combined with either randomized treatment group. Similarly, the denominators shown here should not be treated as the overall randomized enrollment unless the underlying registry data explicitly define them that way.

9. Statistical Methods Explained

Why was a log-rank test used for overall survival?

Overall survival is a time-to-event endpoint: the outcome is not simply whether death occurred, but when it occurred relative to randomization. Some participants can also remain alive at the end of their observed follow-up, creating censored observations. The log-rank test is designed to compare time-to-event distributions while using information across the follow-up period.

What does an HR of 0.646 mean?

An HR of 0.646 means that the estimated instantaneous rate of death in the abiraterone acetate group was 64.6% of that in the placebo group under the reported analysis. It can be expressed as approximately a 35.4% lower estimated hazard, but it should not be translated into a 35.4-percentage-point improvement in survival or interpreted as an individual patient's probability.

Why is the overall-survival analysis described as stratified?

Stratification allows a time-to-event comparison to account for the structure of the trial's analysis strata. The ClinicalTrials.gov record explicitly identify the primary analysis as stratified, but they do not specify the identities of those strata in the ClinicalTrials.gov record.

What does the 95% confidence interval tell us?

The interval gives a range of values that reflects statistical uncertainty around the estimated effect under the specified model and sampling framework. For the primary HR, 0.543–0.768 indicates uncertainty around the point estimate of 0.646. It does not describe a range of outcomes for individual patients.

Why is the PSA decline endpoint analyzed differently?

Achieving a PSA decline of at least 50% is a binary endpoint. Each participant is classified according to whether the registered response criterion was achieved within the specified period. The registry therefore reports a chi-squared test and a risk ratio rather than a time-to-event hazard ratio.

What does a risk ratio of 5.266 mean?

A risk ratio of 5.266 means that the probability of the specified PSA response was estimated to be 5.266 times as high in the abiraterone acetate group as in the placebo group over the registered time frame. It does not mean that the absolute probability was 526.6%, nor does it describe the magnitude of an overall-survival effect.

Why does the 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. Effect size is communicated by measures such as the HR or RR, while the confidence interval communicates uncertainty around that estimate. A very small p-value can accompany either a modest or a large effect depending on sample size and variability.

10. Intention-to-Treat and the Role of Randomization

The posted statistical analyses use the Intent-to-Treat population. The registry definition identifies this as all patients randomized into the study, classified according to their assigned treatment.

Randomization

Randomization creates the treatment comparison before outcomes are observed. This is the foundation for interpreting differences between randomized groups as treatment effects under the trial's design.

ITT principle

Analyzing participants according to randomized assignment maintains that original comparison rather than changing group membership based on later treatment exposure or outcome status.

Time-to-event analysis

ITT analysis does not eliminate censoring. Participants can still contribute information until an event or censoring time, which is why survival-analysis methods remain necessary.

Binary response analysis

The PSA decline endpoint uses the same ITT population but a different statistical framework because the outcome is binary rather than time-to-event.

11. Interpreting the Hazard Ratios Together

EndpointHRSimple relative interpretationWhat it measures
Overall Survival 0.646 Approximately 35.4% lower estimated hazard Time from randomization to death from any cause
Time to PSA Progression 0.580 Approximately 42.0% lower estimated hazard Time to PSA progression according to registered criteria
Radiographic Progression-free Survival 0.673 Approximately 32.7% lower estimated hazard Radiographic progression-free survival

The three hazard ratios should be read endpoint by endpoint. Although all are below 1, they refer to different events and time-to-event definitions. The fact that one HR is numerically smaller than another does not by itself establish that the corresponding treatment effects are statistically different.

Relative effects are not absolute effects
Relative measure ≠ absolute risk difference ≠ survival probability

A hazard ratio summarizes a relative time-to-event effect. To understand absolute clinical outcomes, one would ordinarily also examine survival probabilities, event rates, or other absolute measures at prespecified time points when those data are available.

12. Multiplicity and Multiple Endpoints

The ClinicalTrials.gov record identifies one primary endpoint and three secondary endpoints. The primary endpoint is overall survival, while the secondary endpoints address PSA progression, PSA decline, and radiographic progression-free survival.

EndpointRoleStatistical framework
Overall SurvivalPrimaryStratified log-rank; hazard ratio
Time to PSA ProgressionSecondaryStratified log-rank; hazard ratio
PSA Decline ≥50%SecondaryChi-squared; risk ratio
Radiographic Progression-free SurvivalSecondaryStratified log-rank; hazard ratio

Multiple endpoints create an important statistical distinction between the primary confirmatory question and additional outcomes. A result on a secondary endpoint is not automatically equivalent to the primary endpoint merely because its p-value is small. Formal multiplicity control depends on the prespecified testing strategy.

Data limitation: the ClinicalTrials.gov record identifies the hypothesis type as superiority and provide p-values for all four analyses, but they do not provide an alpha-allocation or multiplicity-adjustment scheme. This page therefore does not infer one.

13. Stratified Analysis and Proportional-Hazards Interpretation

The registry describes all three posted time-to-event analyses as stratified analyses. Stratification can account for differences between analysis strata while preserving a common treatment comparison within the survival-analysis framework.

Hazard ratios are relative measures of event rates over time. In a Cox-type proportional-hazards framework, interpretation commonly relies on an assumption that the relative hazards are reasonably represented by a common hazard ratio over the relevant follow-up. The ClinicalTrials.gov record identifies hazard ratios and stratified log-rank tests, but they do not provide additional model diagnostics for the proportional-hazards assumption.

Practical reading rule

When reading an HR, ask three questions: What event? What population? What time frame? For COU-AA-301, the answers differ across overall survival, PSA progression, and radiographic progression-free survival. Keeping those definitions attached to the HR prevents an effect estimate from being generalized beyond the endpoint it actually represents.

14. Censoring and Time-to-Event Endpoints

Overall survival, PSA progression, and radiographic progression-free survival are all time-to-event endpoints. Their analysis must accommodate participants whose event has not occurred by the time their observed follow-up ends.

Event time

For a participant who experiences the event, the analysis uses the observed time from the relevant starting point to that event.

Censoring

A participant without an observed event can still contribute information up to the point at which their outcome becomes censored.

Why simple proportions are insufficient

Counting events alone ignores when events occurred and how long participants were observed without an event.

Why HR is useful

The hazard ratio provides a relative comparison that incorporates event timing within the survival-analysis framework.

15. Limitations

16. Why This Trial Matters Statistically

COU-AA-301 is a useful teaching example because its registry data connect randomized clinical-trial design with several core statistical methods: ITT analysis, stratified log-rank testing, hazard ratios, confidence intervals, p-values, and a chi-squared analysis of a binary response endpoint.

ConceptHow it appears in COU-AA-301
RandomizationThe trial is registered as randomized.
Parallel designThe design model is parallel with two arms.
Quadruple maskingThe registry identifies the trial as quadruple masked.
ITT analysisAll four posted statistical analyses use the ITT population.
Time-to-event endpointsOverall survival, PSA progression, and radiographic progression-free survival.
Log-rank testingUsed for the primary endpoint and both posted secondary time-to-event analyses.
Stratified analysisThe three posted time-to-event analyses are described as stratified.
Hazard ratioUsed for overall survival, PSA progression, and radiographic progression-free survival.
Confidence interval95% two-sided intervals are reported for all four posted effect estimates.
Chi-squared testUsed for the binary PSA decline ≥50% endpoint.
Risk ratioUsed to summarize the PSA decline ≥50% endpoint.
Superiority testingThe analyses posted on ClinicalTrials.gov identify superiority as the hypothesis type.

17. Statistical Interpretation vs Clinical Interpretation

Statistical interpretation

The posted primary analysis estimated an overall-survival HR of 0.646 with a 95% CI of 0.543–0.768 and P < 0.0001. The analysis used an ITT population and a stratified log-rank method.

Clinical interpretation

The statistical result describes the relative time-to-event comparison observed in the randomized trial. Its clinical meaning requires keeping the endpoint definition, follow-up, absolute outcomes, safety, and patient population in view.

The distinction matters because a statistically precise relative effect does not, by itself, provide every quantity needed for clinical decision-making. A complete interpretation normally considers both relative and absolute measures, the certainty of the estimate, the endpoint's clinical meaning, and the safety experience.

18. Sources

19. Related Tutorials

Learn more about the methods used in this trial:

20. Related Calculators

Continue through the Clinical Biostats statistical pathway

Use the trial's endpoints and methods as a practical route into survival analysis, categorical-data methods, confidence intervals, and randomized-trial interpretation.

21. Record Summary

COU-AA-301 provides a compact example of how a randomized phase 3 clinical trial can combine different statistical frameworks for different endpoint types. The primary overall-survival endpoint was analyzed in the ITT population using a stratified log-rank test and summarized with a hazard ratio of 0.646 (95% CI 0.543–0.768; P < 0.0001). Secondary analyses used the same general distinction between time-to-event and binary outcomes: stratified log-rank methods produced HRs of 0.580 for time to PSA progression and 0.673 for radiographic progression-free survival, while the PSA decline ≥50% endpoint used a chi-squared test and reported an RR of 5.266.

The most important statistical lesson is that these estimates should be interpreted in their proper context. Hazard ratios describe relative event hazards over time; risk ratios describe relative probabilities for a binary outcome; confidence intervals quantify uncertainty around the estimates; and p-values address statistical evidence against a null hypothesis rather than effect magnitude. Randomization, ITT analysis, stratification, censoring, endpoint definition, and the distinction between primary and secondary outcomes all shape how the reported results should be understood.

Clinical Biostats methodology: A trial-results page should not merely repeat a registry entry. The goal is to reconstruct the statistical story of the trial while clearly separating reported evidence from educational interpretation and avoiding unsupported assumptions about analyses that are not contained in the available trial data.