← Clinical Trials
Nonmetastatic CRPC Phase 3 Time-to-Event Analysis NCT02003924

PROSPER: Complete Statistical Analysis of Enzalutamide in Nonmetastatic Castration-Resistant Prostate Cancer

An independent statistical review of the randomized phase 3 PROSPER trial evaluating enzalutamide versus placebo in patients with nonmetastatic castration-resistant prostate cancer, with emphasis on metastasis-free survival, secondary time-to-event endpoints, PSA response, and the statistical methods used to compare randomized groups.

Trial status: COMPLETED  ·  Enrollment: 1401  ·  Study period: 2013-10-31 to 2017-06-28
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results are restricted to the ClinicalTrials.gov record and the permitted linked sources. ClinicalTrials.gov provides the official trial registry 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

PROSPER was a randomized, parallel-group, phase 3 study of enzalutamide versus placebo in patients with nonmetastatic castration-resistant prostate cancer. The registry reports 1401 participants, two treatment arms, quadruple masking, and a primary time-to-event endpoint of metastasis-free survival.

1401
Enrollment
Randomized trial
2
Treatment arms
Parallel design
0.292
MFS HR
95% CI 0.241–0.352
<0.0001
MFS P-value
Two-sided
FeaturePROSPER
Trial namePROSPER
NCT IDNCT02003924
PhasePhase 3
StatusCOMPLETED
Therapeutic areaOncology
ConditionNonmetastatic Castration-Resistant Prostate Cancer; Prostate Cancer; Cancer of the Prostate
AllocationRANDOMIZED
Design modelPARALLEL
MaskingQUADRUPLE
Primary purposeTREATMENT
Enrollment1401.0
InterventionsEnzalutamide; Placebo
Lead sponsorPfizer
Results postedYes
Outcome measures posted47
Statistical analyses posted9

2. Clinical Question

The central statistical question was whether treatment assignment to enzalutamide rather than placebo was associated with a difference in the time from randomization to metastasis-free survival events in patients with nonmetastatic castration-resistant prostate cancer.

Population

Patients with nonmetastatic castration-resistant prostate cancer, as represented by the registered PROSPER study population.

Intervention

Enzalutamide 160 mg.

Comparator

Placebo.

Primary question

Does enzalutamide change metastasis-free survival relative to placebo under the prespecified superiority analysis?

3. Trial Design

01
Randomize1401 participants
02
Parallel armsEnzalutamide vs placebo
03
FollowTime-to-event outcomes
04
AssessMFS and secondary endpoints
05
CompareITT treatment assignment
Allocation
RANDOMIZED allocation with a parallel design.
Masking
QUADRUPLE masking.
Primary purpose
TREATMENT.
Hypothesis type
Superiority.
TREATMENT ARM

Enzalutamide

  • Enzalutamide 160 mg.
  • Serious adverse events were reported as 372 affected participants among 930 at risk.
CONTROL ARM

Placebo

  • Placebo.
  • Serious adverse events were reported as 100 affected participants among 465 at risk.
  • The registry also reports 12 affected participants among 87 at risk in the category "Placebo Patients Crossover to Enzalutami".

The ClinicalTrials.gov record identifies the study as randomized, parallel, and quadruple-masked. Those design features are important statistically because the treatment comparison is anchored to randomized assignment while masking is intended to reduce the influence of treatment knowledge on trial conduct and assessment.

4. Trial Timing and Registry Record

2013-10-31

Study start

The registered study start date was 2013-10-31.

2017-06-28

Primary completion

The registered primary completion date was 2017-06-28.

Results posted

Statistical results available

The ClinicalTrials.gov record contains 47 posted outcome measures and 9 posted statistical analyses, including one formal primary-endpoint analysis.

5. Primary Endpoint

EndpointRegistry definition / time frameEndpoint typePrimary analysis
Metastasis Free Survival (MFS) Time from randomization to first date of radiographic progression (RP) (by Blinded independent central radiology review [BICR]) at any time or death within 112 days of treatment discontinuation without evidence of RP. RP for bone disease: appearance of 1 or more metastatic lesions on bone scan. RP for soft tissue disease: per Response Evaluation Criteria in Solid Tumors, [RECIST 1.1])-at least a 20 percent (%) increase in the sum of diameters of target lesions,taking as reference the smallest sum on study (includes the baseline sum if smallest on study).Participants who did not have MFS event at the time of analysis data cut-off (28 June 2017) were censored at date of last assessment showing no objective evidence of RP prior to skeletal-related event or two or more consecutive missed tumor assessments. Participants who were randomized but later confirmed to have metastatic disease before randomization were censored on date of randomization. Analysis was based on Kaplan-Meier estimates. Time-to-event Log-rank test; hazard ratio from a Cox regression model

6. Analysis Population and Treatment Comparison

The primary MFS analysis used the intent-to-treat (ITT) population. The registry defines this population as all participants randomly assigned to study treatment and states that the analysis was based on randomized treatment assignment regardless of whether or not treatment was administered.

Analysis featureRegistry-reported approach
PopulationIntent-to-treat (ITT)
Groups comparedEnzalutamide 160 mg vs Placebo
Primary endpointMetastasis Free Survival (MFS)
Statistical testLog Rank
Effect measureHazard Ratio (HR)
HypothesisSuperiority
Model noteHR was based on a Cox regression model with treatment as the only covariate, stratified by factors defined in the registry analysis record, and was relative to placebo with < 1 favoring Enzalutamide.

The ITT principle is particularly important in a randomized trial. Once participants are randomized, analyzing them according to their assigned treatment preserves the treatment comparison created by randomization. It avoids changing the comparison based on what happened after randomization, such as treatment discontinuation or failure to receive treatment.

7. Primary Results: Metastasis-Free Survival

The registry reports a formal statistical analysis for MFS using a log-rank test and a hazard ratio derived from a Cox regression model. The analysis was performed in the ITT population comparing enzalutamide 160 mg with placebo.

Hazard ratio for metastasis-free survival

0.292

95% CI: 0.241–0.352   ·   P < 0.0001

Two-sided confidence interval; superiority hypothesis.

EndpointAnalysis populationMethodEffect estimate95% CIP-value
Metastasis Free Survival (MFS) ITT Log-rank test; stratified Cox regression HR 0.292 0.241–0.352 <0.0001
Clinical Biostats interpretation

The reported HR of 0.292 means that, under the Cox model used for this analysis, the estimated instantaneous rate of an MFS event in the enzalutamide group was about 29.2% of the corresponding rate in the placebo group. Expressed as a simple relative-hazard interpretation, this corresponds to an estimated 70.8% lower hazard because 1 − 0.292 = 0.708.

This does not mean that 70.8% of participants avoided metastasis, that 70.8% were cured, or that each individual participant experienced exactly a 70.8% reduction in risk. A hazard ratio is a relative time-to-event measure derived from a statistical model.

The 95% CI of 0.241–0.352 describes the statistical uncertainty around the estimated hazard ratio under the model and sampling framework. It is not a range containing the individual treatment effects experienced by patients.

The P-value of <0.0001 addresses the strength of evidence against the null hypothesis under the specified testing framework. It does not measure the size of the treatment effect. Effect size is communicated by the hazard ratio, while precision is communicated by the confidence interval.

Because this is a time-to-event analysis, interpretation also depends on censoring and the assumptions underlying the Cox model. In particular, a single hazard ratio is most straightforward to interpret when the relative hazards are reasonably described by a proportional-hazards model over the analyzed follow-up.

8. Secondary Time-to-Event Results

The registry contains formal statistical analyses for several secondary time-to-event endpoints. These analyses use the ITT population and compare enzalutamide 160 mg with placebo using log-rank testing and hazard ratios.

Secondary endpointTime frameEffect measureEstimate95% CIP-value
Time to Prostate-Specific Antigen (PSA) Progression From randomization until first PSA progression (until the data cut-off date of 28 June 2017, maximum duration of treatment: 42.8 months) Hazard ratio 0.066 0.054–0.081 <0.0001
Time to First Use of New Antineoplastic Therapy From randomization until first use of new antineoplastic therapy(until the data cut-off date of 28 June 2017, maximum duration of treatment: 42.8 months) Hazard ratio 0.208 0.168–0.258 <0.0001
Overall Survival From randomization until death (up to a maximum of 68.8 months) Hazard ratio 0.734 0.608–0.885 0.0011
Time to Pain Progression From randomization until onset of pain progression (until the data cut-off date of 28 June 2017, maximum duration of treatment: 42.8 months) Hazard ratio 0.959 0.801–1.149 0.6534
Time to First Use of Cytotoxic Chemotherapy From randomization up to the first use of cytotoxic chemotherapy (until the data cut-off date of 28 June 2017, maximum duration of treatment: 42.8 months) Hazard ratio 0.378 0.282–0.507 <0.0001

Reading the secondary hazard ratios

The secondary estimates span a wide range. An HR below 1 indicates a lower estimated event hazard for enzalutamide relative to placebo under the corresponding analysis, while an HR near 1 indicates a smaller estimated relative difference. The time-to-pain-progression analysis, for example, reported an HR of 0.959 with a 95% CI of 0.801–1.149 and P=0.6534. Unlike the other reported time-to-event estimates in this table, that confidence interval includes 1.

These endpoints should not be collapsed into a single treatment-effect statistic. They represent different clinical events occurring at different stages of disease and treatment. A treatment can affect PSA progression, use of subsequent therapy, survival, pain progression, and chemotherapy initiation differently because these events have different definitions, censoring mechanisms, and clinical determinants.

Multiplicity matters: the registry analysis notes that the family-wise two-sided type I error rate was maintained at 0.05 using a parallel testing strategy between overall survival, with an allocated type I error rate of 0.03, and the remaining key secondary endpoints, including time to PSA progression and time to first use of new antineoplastic therapy, with an allocated type I error rate of 0.02. Testing was performed only if the primary endpoint was statistically significant, and the analysis notes that subsequent testing depended on the preceding endpoint being statistically significant.

9. Secondary PSA Response Results

The registry also reports three analyses of the binary endpoint Percentage of Participants With Prostate Specific Antigen (PSA) Response. The analysis method was the Cochran-Mantel-Haenszel test, with the effect measure reported as a difference in response rate.

PSA response definitionEstimate: difference in response rate95% CIP-value
Decrease from Baseline ≥ 50% 73.96 70.91–77.02 <0.0001
Decrease from Baseline ≥ 90% 55.52 52.28–58.76 <0.0001
Decrease to Undetectable Level 9.65 7.75–11.54 <0.0001

Unlike MFS and the other time-to-event outcomes, these PSA response measures are binary outcomes. The reported effect measure is therefore a difference in response rate, rather than a hazard ratio. This distinction matters: the number 73.96 represents a difference in the percentage of participants meeting the registered PSA-response criterion, not a relative hazard and not a time-to-event measure.

The registry identifies the ITT population for these analyses and specifies the Cochran-Mantel-Haenszel method. The three response definitions should also be kept conceptually separate because a decrease of at least 50%, a decrease of at least 90%, and a decrease to an undetectable level represent progressively different PSA-response criteria.

10. Statistical Methodology

Kaplan-Meier estimation

Metastasis-free survival, PSA progression, new antineoplastic therapy, overall survival, pain progression, and cytotoxic chemotherapy use are all time-to-event outcomes. A standard way to describe such outcomes is with the Kaplan-Meier estimator, which estimates the probability of remaining event-free over time while accounting for right-censored observations.

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

where di is the number of events at time ti and ni is the number at risk immediately before that time.

The ClinicalTrials.gov record does not provide Kaplan-Meier estimates or curves themselves. The important statistical point is that the underlying time-to-event framework can accommodate participants whose event has not occurred by the time they leave the observable risk set. Such participants are censored rather than automatically treated as if they experienced the event.

Log-rank test

The registry reports the log-rank test for MFS and the listed secondary time-to-event endpoints. The log-rank procedure compares the observed and expected numbers of events between treatment groups over follow-up.

It is therefore fundamentally different from simply comparing the percentage of participants who eventually experienced an event. The timing of events contributes to the comparison, which is why the log-rank test is a natural method for randomized trials with time-to-event endpoints.

Cox proportional-hazards model

The MFS analysis notes that the hazard ratio was based on a Cox regression model with treatment as the only covariate, stratified by factors defined in the registry analysis record. The ClinicalTrials.gov record does not enumerate those stratification factors, so they are not independently reconstructed here.

Interpretation of the hazard ratio
HR < 1  →  lower estimated instantaneous event rate in the treatment group

A hazard ratio is a relative model-based measure. It is not the same as a relative risk, an absolute risk difference, or the percentage of participants who benefit.

Intention-to-treat analysis

The registry explicitly defines the ITT population as all participants randomly assigned to study treatment, with analyses based on randomized treatment assignment regardless of whether treatment was administered. This approach protects the original randomized comparison.

For educational purposes, this is an important distinction from an analysis that excludes participants after randomization because of treatment discontinuation or other post-randomization events. Excluding such participants can change the population being compared and potentially weaken the protection provided by randomization.

Cochran-Mantel-Haenszel testing

The PSA-response analyses use the Cochran-Mantel-Haenszel test. This family of methods is designed for categorical data and can account for stratification when the analysis includes relevant strata. In PROSPER, the registry identifies the method and the ITT population, while the reported effect measure is a difference in response rate.

11. Statistical Methods Explained

Why was a log-rank test used for MFS?

MFS is a time-to-event endpoint. Participants can experience the event at different times, while others may be censored before an event is observed. The log-rank test uses the timing of events across follow-up rather than reducing the endpoint to a simple yes/no status at a single time point.

What does an MFS hazard ratio of 0.292 mean?

It means the fitted model estimated the instantaneous event hazard in the enzalutamide group to be 0.292 times that in the placebo group. In relative terms, that is an estimated 70.8% lower hazard. It does not mean that 70.8% of participants avoided metastasis or that every patient experienced that same reduction.

Why is the confidence interval important?

The 95% CI of 0.241–0.352 communicates precision around the MFS hazard-ratio estimate. A narrow interval indicates that the estimate is relatively concentrated under the statistical framework used, while the interval itself still should not be interpreted as the range of effects for individual patients.

Why doesn't the P-value measure effect size?

A P-value measures how compatible the observed data are with a specified null hypothesis under the statistical model and testing procedure. It is influenced by both the magnitude of an effect and the amount of information available. The hazard ratio is the effect measure; the confidence interval describes its statistical precision.

Why is ITT important in this trial?

ITT preserves the randomized treatment assignment as the basis of the efficacy comparison. This means that post-randomization events do not determine which randomized group a participant belongs to for the primary efficacy analysis.

Why are PSA response and MFS analyzed differently?

MFS records a time until an event and therefore uses survival-analysis methods. PSA response is a binary outcome based on whether a participant meets a specified response criterion. The registry consequently uses a Cochran-Mantel-Haenszel test and reports a difference in response rate rather than a hazard ratio.

Why does multiplicity matter for the secondary endpoints?

When several hypotheses are tested, the chance of obtaining at least one apparently positive result can increase if every test is treated as an independent 0.05 test. The registry describes an allocated type I error framework and a sequential testing strategy, so the P-values should be interpreted in the context of that prespecified testing structure rather than in isolation.

12. Confidence Intervals and Statistical Precision

The reported confidence intervals show how the same statistical framework communicates uncertainty for different endpoint types.

EndpointEstimate95% confidence intervalMeasure type
MFSHR 0.2920.241–0.352Hazard ratio
Time to PSA progressionHR 0.0660.054–0.081Hazard ratio
Time to new antineoplastic therapyHR 0.2080.168–0.258Hazard ratio
Overall survivalHR 0.7340.608–0.885Hazard ratio
Time to pain progressionHR 0.9590.801–1.149Hazard ratio
Time to cytotoxic chemotherapyHR 0.3780.282–0.507Hazard ratio
PSA response: decrease from baseline ≥ 50%73.9670.91–77.02Difference in response rate
PSA response: decrease from baseline ≥ 90%55.5252.28–58.76Difference in response rate
PSA response: decrease to undetectable level9.657.75–11.54Difference in response rate

A useful statistical habit is to read an estimate and its confidence interval together. For example, the MFS HR of 0.292 should not be considered independently of its 95% CI of 0.241–0.352. Similarly, the time-to-pain-progression estimate of 0.959 should be read together with its 95% CI of 0.801–1.149 rather than judged solely by whether the point estimate is below 1.

13. Multiplicity and Hierarchical Testing

The registry analysis notes a specific approach to maintaining the family-wise two-sided type I error rate at 0.05. A parallel testing strategy was used between overall survival, with an allocated type I error rate of 0.03, and the remaining key secondary endpoints, including time to PSA progression and time to first use of new antineoplastic therapy, with an allocated type I error rate of 0.02.

Primary endpoint first

The registry states that testing of the relevant secondary endpoints was performed only if the primary endpoint was statistically significant.

Allocated error

The analysis describes an overall two-sided type I error rate of 0.05 distributed through the prespecified testing strategy.

Why this matters

The reported secondary P-values should be interpreted within the trial's testing strategy rather than as a collection of unrelated hypothesis tests.

Different endpoints

MFS, overall survival, PSA progression, pain progression, and treatment-use endpoints represent different clinical questions and should not be treated as interchangeable.

Multiplicity is often misunderstood as a technical adjustment applied only after results are obtained. In a well-specified confirmatory trial, it is instead part of the design. The allocation of type I error determines how evidence is evaluated across the family of prespecified hypotheses.

14. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment category. These figures should be read as affected participants divided by the corresponding number at risk, rather than as a comparison of time-to-event efficacy outcomes.

Safety categoryAffected / at risk
Enzalutamide 160 mg372 / 930
Placebo100 / 465
Placebo Patients Crossover to Enzalutami12 / 87

The ClinicalTrials.gov record does not provide a statistical hypothesis test, confidence interval, or comparative effect measure for these serious-adverse-event figures. Accordingly, the safety information is reported descriptively rather than converted into an unreported statistical comparison.

The crossover category also needs to be kept distinct from the randomized placebo category. Participants who later crossed over represent a post-randomization treatment pathway, whereas the primary efficacy analysis is based on randomized treatment assignment.

15. Crossover and Treatment Assignment

The ClinicalTrials.gov record explicitly identify a category of Placebo Patients Crossover to Enzalutami, with 12 affected participants among 87 at risk for the reported safety measure.

From a statistical perspective, crossover creates an important distinction between assignment and treatment received. The ITT efficacy analysis asks what happened according to the randomized treatment assignment. A treatment-received analysis instead asks what happened according to subsequent exposure. These are different estimands and answer different causal questions.

Interpretation caution: the ClinicalTrials.gov record does not provide enough information to quantify how crossover affected the overall survival estimate. The presence of a crossover category should therefore not be used to calculate an adjustment or to infer an alternative survival effect.

16. What the MFS Hazard Ratio Does — and Does Not — Mean

Relative effect

The MFS HR of 0.292 is a model-based relative comparison of event hazards. Because the registry specifies that values below 1 favor enzalutamide, the estimate indicates a lower estimated MFS event hazard for enzalutamide relative to placebo.

Not an absolute risk reduction

The HR is not the difference between two percentages of participants who experienced metastasis or death. Without corresponding absolute survival probabilities at a specified time point, the HR alone cannot tell a reader the absolute number of events prevented.

Not a patient-level prediction

A hazard ratio describes the randomized groups at the population level. It should not be interpreted as saying that an individual participant's personal risk is reduced by exactly the same percentage.

Why follow-up matters

MFS is explicitly defined as a time from randomization to an event or censoring-related endpoint. The interpretation therefore depends on the observed follow-up period, event timing, and censoring rules rather than simply on whether an event occurred at some unspecified point.

17. Time-to-Event Endpoints: Why the Timing of Events Matters

PROSPER provides a useful illustration of why time-to-event analysis is different from ordinary binary analysis. Suppose two treatment groups eventually had the same number of events. That alone would not establish that their clinical courses were equivalent: one group could experience events substantially earlier than the other.

Survival methods preserve information about when events occur. The Kaplan-Meier framework describes the event-free probability over time, the log-rank test compares the treatment groups across follow-up, and the Cox model summarizes the relative event hazard using a hazard ratio.

Statistical componentQuestion it helps answer
Kaplan-Meier estimationHow does the probability of remaining event-free change over time?
Log-rank testIs there evidence that the time-to-event distributions differ between randomized groups?
Cox modelWhat is the estimated relative hazard between treatment groups?
95% confidence intervalHow precise is the estimated relative effect?
P-valueHow compatible are the data with the specified null hypothesis under the testing framework?

18. Statistical Interpretation of the Secondary Endpoints

The secondary results illustrate why statistical interpretation should consider both the endpoint and its estimand.

For time to PSA progression, the HR was 0.066 with a 95% CI of 0.054–0.081 and P<0.0001. The point estimate is far below 1, indicating a large relative difference in the estimated event hazard under the reported model.

For time to first use of new antineoplastic therapy, the HR was 0.208 with a 95% CI of 0.168–0.258 and P<0.0001. This measures the time to a subsequent treatment event, not metastasis itself.

For overall survival, the HR was 0.734 with a 95% CI of 0.608–0.885 and P=0.0011. The registry defines this endpoint simply as time from randomization until death, with a maximum duration of 68.8 months in the reported time frame.

For time to pain progression, the HR was 0.959 with a 95% CI of 0.801–1.149 and P=0.6534. The confidence interval extends across 1, so the reported point estimate should not be interpreted independently of that uncertainty.

For time to first use of cytotoxic chemotherapy, the HR was 0.378 with a 95% CI of 0.282–0.507 and P<0.0001. Again, this is a time-to-treatment-use endpoint rather than a direct measure of metastasis-free survival.

19. Limitations and Interpretation Issues

20. Why This Trial Matters Statistically

PROSPER is a useful teaching example because it connects several fundamental clinical-trial methods in a single randomized study. Its primary endpoint is a time-to-event outcome, the primary analysis uses a log-rank framework and Cox-derived hazard ratio, the efficacy population follows the ITT principle, secondary binary outcomes use the Cochran-Mantel-Haenszel test, and the registry describes a multiplicity strategy for key secondary testing.

ConceptHow it appears in PROSPER
RandomizationParticipants were randomly assigned to enzalutamide or placebo.
BlindingThe registry describes the study as quadruple-masked.
ITT analysisThe MFS and secondary efficacy analyses use randomized treatment assignment in the ITT population.
Time-to-event analysisMFS, PSA progression, new antineoplastic therapy, overall survival, pain progression, and cytotoxic chemotherapy use are time-to-event endpoints.
Log-rank testUsed for the formal time-to-event comparisons reported in the registry.
Hazard ratioUsed to quantify relative treatment effects for the time-to-event endpoints.
Cox regressionThe MFS HR was based on a Cox regression model with treatment as the only covariate and stratification by registry-defined factors.
Confidence interval95% two-sided confidence intervals accompany the reported effect estimates.
Cochran-Mantel-Haenszel testUsed for the reported binary PSA-response analyses.
MultiplicityA parallel testing strategy allocated type I error across OS and key secondary endpoints.
CrossoverThe registry includes a category for placebo patients crossing over to enzalutamide.

21. A Practical Reading Framework for PROSPER

A statistically disciplined reading of the PROSPER results can follow a sequence rather than beginning with the P-value.

1. Identify the estimand

Ask what event is being measured and from what starting point. MFS begins at randomization and uses a specific radiographic-progression or death definition.

2. Identify the population

For the reported efficacy analyses, the population is ITT, preserving randomized treatment assignment.

3. Identify the effect measure

Time-to-event outcomes use hazard ratios, while the PSA-response analyses use differences in response rate.

4. Read the interval

The confidence interval provides information about statistical precision and should be read together with the point estimate.

5. Read the P-value in context

The P-value is part of a prespecified hypothesis-testing framework and does not itself quantify clinical effect size.

6. Check design complications

Multiplicity, censoring, crossover, and the assumptions of the Cox model all affect interpretation of time-to-event results.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue through the Clinical Biostats statistical pathway

Use the trial as a practical example of randomized treatment comparison, time-to-event analysis, hazard ratios, confidence intervals, categorical-data methods, and multiplicity.

25. Record Summary

PROSPER provides a compact but rich example of clinical-trial time-to-event statistics. The study was a randomized, parallel-group, quadruple-masked phase 3 trial with 1401 participants and a primary endpoint of metastasis-free survival. The primary analysis used the ITT population, a log-rank test, and a Cox-model hazard ratio. The reported MFS HR was 0.292, with a two-sided 95% CI of 0.241–0.352 and P<0.0001.

The secondary analyses demonstrate why endpoint-specific interpretation matters. Time to PSA progression, time to first use of new antineoplastic therapy, overall survival, time to pain progression, and time to first use of cytotoxic chemotherapy all use time-to-event methodology, while PSA-response outcomes use the Cochran-Mantel-Haenszel test and differences in response rate. The registry also describes a prespecified multiplicity strategy for key secondary testing and reports a crossover category among placebo patients.

Clinical Biostats methodology: The purpose of a trial-results page is not simply to reproduce reported numbers. It is to connect the randomized design, endpoint definition, analysis population, statistical method, effect measure, uncertainty, multiplicity structure, and interpretation into one coherent statistical story while keeping reported evidence separate from educational explanation.