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
CHAARTED was a randomized, open-label, parallel phase 3 treatment trial in metastatic hormone-sensitive prostate cancer. The registry reports 790 enrolled participants, two treatment arms, a superiority hypothesis, and overall survival as the single registered primary endpoint.
| Feature | CHAARTED |
|---|---|
| Trial name | CHAARTED |
| ClinicalTrials.gov identifier | NCT00309985 |
| Phase | Phase 3 |
| Condition | Metastatic Hormone-sensitive Prostate Cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 790 |
| Primary endpoint | Overall Survival |
| Primary endpoint type | Time-to-event |
| Hypothesis | Superiority |
| Trial status | Completed |
| Start | September 26, 2006 |
| Primary completion | December 23, 2013 |
| Lead sponsor | ECOG-ACRIN Cancer Research Group |
2. Clinical Question
The central statistical question was whether adding docetaxel to androgen-deprivation therapy improved overall survival compared with androgen-deprivation therapy alone in patients with metastatic hormone-sensitive prostate cancer.
Population
Patients with metastatic hormone-sensitive prostate cancer enrolled in the phase 3 CHAARTED trial.
Intervention
Androgen-deprivation therapy and docetaxel.
Comparator
Androgen-deprivation therapy alone.
Primary question
Does the addition of docetaxel produce a superior overall-survival result relative to androgen-deprivation therapy alone?
This question is naturally expressed as a time-to-event comparison rather than simply as a comparison of proportions. Overall survival incorporates both the occurrence of death and the amount of follow-up contributed by participants who have not died at the time their survival information is last available.
3. Trial Design
Androgen-Deprivation Therapy + Docetaxel
- Androgen-deprivation therapy
- Docetaxel
Androgen-Deprivation Therapy Alone
- Androgen-deprivation therapy
The registry describes the study as randomized, parallel, unmasked, and treatment-focused. Those design features matter statistically. Randomization provides the basis for comparing treatment assignments without relying on baseline prognostic variables to create the primary causal comparison. The absence of masking is a separate design characteristic and should not be confused with the validity of randomization itself.
4. Endpoints
| Endpoint | Registry definition / time frame | Type |
|---|---|---|
| Overall Survival | Overall survival is defined as the time from randomization to death or date last known alive. Survival data reflects the database as of December 23, 2013. Assessed every 3 months if patient is < 2 years from study entry; every 6 months if patient is 2 - 5 years from study entry; then annually if patient is 5 - 10 years from study entry. | Time-to-event |
| Time to Clinical Progression | Assessed every 3 months if patient is < 2 years from study entry; every 6 months if patient is 2 - 5 years from study entry; then annually if patient is 5 - 10 years from study entry. | Time-to-event |
| Time to Castration Resistant Prostate Cancer (Hormone Refractory Disease) | Assessed every 3 months if patient is < 2 years from study entry; every 6 months if patient is 2 - 5 years from study entry; then annually if patient is 5 - 10 years from study entry. | Time-to-event |
| Proportion of Patients With PSA Complete Response (CR) at 6 Months | Assessed at 6 months. | Binary |
| Proportion of Patients With PSA Complete Response (CR) at 12 Months | Assessed at 12 months. | Binary |
| QOL Change From Baseline to 3 Months | Assessed at baseline and 3 months. | Continuous |
5. Statistical Methodology
Primary endpoint: log-rank analysis
The posted primary statistical analysis compares overall survival between androgen-deprivation therapy plus docetaxel and androgen-deprivation therapy alone using the log-rank test. The analysis population is all randomized patients.
The registry identifies the hypothesis as superiority and reports a one-sided testing component in the analysis text. The posted P-value is two-sided in the statistical-analysis record's CI-sides field.
The log-rank test is appropriate for comparing two survival distributions when the outcome is time until an event and some participants may be censored. Rather than reducing each participant to a simple event/no-event indicator, the method uses the ordering of event times and the risk sets contributing information at those times.
Prespecified statistical design
The registry states that the study was designed to detect a 33.3% improvement in median survival time across treatments, using a one-sided type I error of 0.025 and 80% power.
Type I error
A one-sided type I error of 0.025 defines the prespecified false-positive error rate for the superiority testing framework described in the registry.
Power
The stated 80% power describes the probability of detecting the prespecified design effect under the assumptions used when the study was planned.
It is important not to reverse-engineer the design assumptions from the observed P-value. Power is a prospective design property tied to an assumed treatment effect, error rate, sample information, and analysis plan. A small observed P-value does not itself tell us what power the trial had for every possible effect size.
6. Primary Result: Overall Survival
Overall Survival
Log-rank test · all randomized patients · superiority hypothesis
The statistical-analysis record reports a two-sided confidence-interval convention, but the ClinicalTrials.gov record does not provide an overall-survival effect estimate or confidence interval.
The registry-posted analysis reports a P-value of 0.0003 for the comparison of androgen-deprivation therapy and docetaxel versus androgen-deprivation therapy alone using a log-rank test.
The P-value indicates that the observed separation in the time-to-event data was inconsistent with the null hypothesis under the statistical testing framework used for this comparison. It supports the trial's prespecified superiority hypothesis at the stated testing level.
The P-value does not quantify the size of the treatment effect. A P-value of 0.0003 is not a statement that the treatment reduced mortality by 99.97%, nor does it tell us the probability that the treatment is beneficial.
The trial's statistical-analysis record does not provide a hazard ratio, an absolute survival difference, a median-survival estimate, or a confidence interval for overall survival in the ClinicalTrials.gov record.
Because overall survival is a time-to-event endpoint, interpretation also depends on censoring and the pattern of event times. A log-rank P-value establishes evidence of a difference between the survival distributions under its assumptions; it is not itself a measure of clinical magnitude.
7. Secondary Time-to-Event Results
Time to Clinical Progression
Time to Clinical Progression
Log-rank test · all randomized patients · superiority hypothesis
Androgen-deprivation therapy and docetaxel vs androgen-deprivation therapy alone
The registry reports a log-rank analysis of time to clinical progression among all randomized patients. The posted P-value is <0.0001.
This P-value indicates strong statistical evidence against the null comparison of the time-to-clinical-progression distributions under the reported log-rank framework.
It does not provide the magnitude of the difference. The ClinicalTrials.gov record contains no hazard ratio, confidence interval, median time, or absolute event-rate estimate for this endpoint, so none is reported here.
Because this is a time-to-event endpoint, the result incorporates follow-up time and censoring rather than simply comparing the percentage of participants who eventually progressed.
Time to Castration Resistant Prostate Cancer
Time to Castration Resistant Prostate Cancer
Log-rank test · all randomized patients · superiority hypothesis
Registry wording: Time to Castration Resistant Prostate Cancer (Hormone Refractory Disease)
This secondary endpoint was analyzed with a log-rank test in all randomized patients. The registry-posted P-value is <0.0001.
The reported result provides statistical evidence of a difference between treatment groups in the time-to-event distribution for this endpoint under the stated superiority analysis.
Again, statistical significance should not be confused with effect size. No hazard ratio or confidence interval for this endpoint is present in the ClinicalTrials.gov record, so the magnitude and precision of the treatment difference cannot be quantified from the ClinicalTrials.gov record.
8. Secondary Binary Endpoints
PSA Complete Response at 6 Months
Proportion of Patients With PSA Complete Response (CR) at 6 Months
Fisher exact test · all randomized patients
The 6-month PSA complete-response endpoint is binary: each participant is classified according to whether the prespecified complete-response outcome occurred at the specified assessment time. The registry reports a Fisher exact test comparing the randomized treatment groups, with P < 0.0001.
Fisher's exact test is a categorical-data method that evaluates the treatment-group association with a binary outcome without relying on a large-sample approximation to the cell-count distribution.
The P-value indicates evidence of a difference in the proportions under the reported superiority framework. It does not tell us how large that difference was. The ClinicalTrials.gov record contains neither the response proportions nor an effect estimate or confidence interval for the 6-month endpoint.
PSA Complete Response at 12 Months
Proportion of Patients With PSA Complete Response (CR) at 12 Months
Fisher exact test · all randomized patients
The 12-month PSA complete-response endpoint was also analyzed with Fisher's exact test. The registry reports P < 0.0001 for the comparison between androgen-deprivation therapy plus docetaxel and androgen-deprivation therapy alone.
The statistical test addresses whether the observed binary response pattern differs between the two randomized groups. It does not quantify the clinical importance of the difference without the underlying proportions.
The ClinicalTrials.gov record does not provide the numbers of responders, response percentages, risk difference, risk ratio, odds ratio, or confidence interval. Reporting only the posted P-value therefore avoids implying an effect magnitude that is not contained in the registry ClinicalTrials.gov record.
9. Quality-of-Life Analysis
The registry includes QOL Change From Baseline to 3 Months as a secondary continuous endpoint. The analysis population consists of patients with both baseline and 3-month QOL assessments.
| Analysis group | Method | P-value | Analysis population |
|---|---|---|---|
| Androgen-Deprivation Therapy and Docetaxel | Wilcoxon signed rank test | 0.0009 | Patients with both baseline and 3-month QOL assessments |
| Androgen-Deprivation Therapy Alone | Wilcoxon signed rank test | 0.40 | Patients with both baseline and 3-month QOL assessments |
These are within-group analyses of change from baseline. The ClinicalTrials.gov record does not report a formal between-group comparison of QOL change. That distinction is essential: a statistically significant change within one arm and a nonsignificant change within another arm does not, by itself, demonstrate that the changes differ between arms.
The Wilcoxon signed rank test evaluates the distribution of paired changes within a treatment group. It is useful when a continuous change score may not satisfy the assumptions that would make a paired t-test attractive.
The reported P = 0.0009 for the androgen-deprivation therapy plus docetaxel group provides evidence against a null distribution of paired changes under that within-group test. The P = 0.40 for androgen-deprivation therapy alone does not provide evidence against its corresponding within-group null under the same framework.
Neither P-value answers the most direct comparative question: whether the average or distribution of QOL change differs between randomized treatment groups. A between-group comparison would require an analysis specifically designed for that question.
10. Statistical Methods Explained
Why was a log-rank test used for overall survival?
Overall survival is measured as time from randomization until death or the last date the participant is known to be alive. Some participants may therefore have incomplete event information at the analysis cutoff. The log-rank test is designed for comparing survival distributions while using the available follow-up from participants who remain at risk.
What does a log-rank P-value actually test?
The log-rank test evaluates whether the observed pattern of event times differs between treatment groups under the null hypothesis of no difference in the survival distributions. It is not an effect-size measure. A small P-value can establish statistical evidence of a difference without telling the reader whether that difference is small, moderate, or large.
Why is the analysis population described as all randomized patients?
Analyzing all randomized patients according to their randomized groups preserves the treatment comparison created by randomization. This principle is particularly important in superiority trials because excluding participants after randomization can introduce systematic differences between groups.
Why is censoring important?
A participant who is still alive at the last known follow-up has not experienced the event by that time, but that does not mean the participant has an observed survival time extending indefinitely. Survival methods use the information accumulated until the participant is censored. Consequently, the analysis uses both event times and the amount of follow-up contributed by participants without an observed event.
Why was Fisher exact testing used for PSA complete response?
The 6-month and 12-month PSA complete-response outcomes are binary. Fisher's exact test evaluates the relationship between treatment group and a binary outcome using the exact distribution associated with the observed contingency table. It is especially useful when a categorical comparison should not depend on a large-sample approximation.
Why does the QOL analysis use the Wilcoxon signed rank test?
The QOL endpoint is defined as change from baseline to 3 months, so the observations are paired within participants. The Wilcoxon signed rank test evaluates paired changes without requiring the change scores to follow a normal distribution. The important interpretive limitation is that a within-arm test does not substitute for a formal between-arm comparison.
What does the one-sided type I error mean here?
The registry states that the study was designed with a one-sided type I error of 0.025 for the superiority framework. A one-sided design allocates the prespecified false-positive probability to departures in the direction specified by the alternative hypothesis. This is distinct from the confidence-interval convention recorded for the posted statistical analysis.
11. Interpreting the Primary P-value Without an Effect Estimate
The primary result illustrates an important statistical reporting principle. The registry provides a formal comparison and a P-value, but the ClinicalTrials.gov record does not provide the corresponding effect estimate or confidence interval for overall survival.
What the P-value tells us
It quantifies the incompatibility of the observed test statistic with the null hypothesis under the specified statistical framework.
What it does not tell us
It does not quantify treatment effect magnitude, absolute benefit, median survival, or the probability that the treatment hypothesis is true.
What a confidence interval adds
A confidence interval would provide an interval description of uncertainty around an effect estimate, which is different information from a hypothesis-test P-value.
What is missing from the ClinicalTrials.gov record
The ClinicalTrials.gov record does not contain an overall-survival hazard ratio, confidence interval, median survival estimate, or survival percentage.
This distinction prevents a common error in clinical-trial interpretation: treating statistical evidence against a null hypothesis as though it were a direct measure of treatment benefit. The two concepts are related, but they answer different questions.
12. Secondary Endpoint Structure
| Endpoint | Type | Method | P-value | Interpretive role |
|---|---|---|---|---|
| Overall Survival | Time-to-event | Log-rank test | 0.0003 | Registered primary endpoint |
| Time to Clinical Progression | Time-to-event | Log-rank test | <0.0001 | Secondary efficacy endpoint |
| Time to Castration Resistant Prostate Cancer | Time-to-event | Log-rank test | <0.0001 | Secondary efficacy endpoint |
| PSA CR at 6 Months | Binary | Fisher exact test | <0.0001 | Secondary response endpoint |
| PSA CR at 12 Months | Binary | Fisher exact test | <0.0001 | Secondary response endpoint |
| QOL Change From Baseline to 3 Months — ADT + Docetaxel | Continuous | Wilcoxon signed rank test | 0.0009 | Within-group change |
| QOL Change From Baseline to 3 Months — ADT Alone | Continuous | Wilcoxon signed rank test | 0.40 | Within-group change |
The methods therefore span three distinct statistical structures: time-to-event outcomes, binary outcomes, and paired continuous change. Matching the analysis method to the endpoint structure is one of the central statistical features of the trial.
13. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk:
| Treatment arm | Serious adverse events | At risk |
|---|---|---|
| Arm A: Androgen-Deprivation Therapy and Docetaxel | 116 | 390 |
| Arm B: Androgen-Deprivation Therapy Alone | 12 | 392 |
These are the serious-adverse-event counts reported in the ClinicalTrials.gov record. The page does not convert them into percentages or relative measures because the instruction for this record is to report numbers exactly as reported in the registry and not introduce derived results unless they follow directly from the ClinicalTrials.gov record.
The serious-adverse-event information should be interpreted separately from efficacy testing. Safety outcomes describe harms or clinically important events occurring during treatment exposure and follow-up; they are not simply another expression of the primary survival endpoint.
The denominators also matter. The registry data explicitly provides affected participants and participants at risk for each arm. Without defining a different analysis population or follow-up rule, the registry-reported counts should not be silently replaced with the overall enrollment of 790.
14. Missing Data, Censoring, and Analysis Populations
Censoring in overall survival
The registered definition of overall survival explicitly includes the date last known alive. That wording establishes the role of censoring in the endpoint: when a participant has not been observed to die by the available follow-up date, the survival information contributes up to the last known-alive date.
QOL assessment availability
The QOL analysis uses patients with both baseline and 3-month QOL assessments. This is different from the all-randomized analysis population used for the primary overall-survival analysis. The distinction is important because a paired analysis cannot be performed for a participant who lacks one of the two required measurements.
15. Multiplicity and Multiple Secondary Endpoints
The trial has one registered primary endpoint and multiple secondary endpoint analyses. This structure creates an important distinction between the primary confirmatory question and the collection of additional statistical comparisons.
| Level | Endpoints in the ClinicalTrials.gov record | Statistical consideration |
|---|---|---|
| Primary | Overall Survival | Primary superiority comparison using the log-rank test. |
| Secondary time-to-event | Time to Clinical Progression; Time to Castration Resistant Prostate Cancer | Each has its own posted log-rank P-value. |
| Secondary binary | PSA CR at 6 months; PSA CR at 12 months | Each has its own posted Fisher exact P-value. |
| Secondary continuous | QOL Change From Baseline to 3 Months | Separate within-arm Wilcoxon signed rank tests are reported. |
Because the ClinicalTrials.gov record does not specify a multiplicity-adjustment procedure for the secondary endpoints, their P-values should not automatically be interpreted as though they represent a single multiplicity-controlled familywise test. Each P-value is reported exactly as posted, while its inferential role should remain distinct from the primary endpoint.
16. Non-Inferiority, Crossover, Factorial Design, and Bayesian Methods
Several common clinical-trial statistical topics are not supported by the registry-reported CHAARTED data and therefore are not treated as trial features here.
Non-inferiority
The ClinicalTrials.gov record identifies the hypothesis as superiority. No non-inferiority margin is reported.
Crossover
No crossover design or crossover analysis is reported in the ClinicalTrials.gov record.
Factorial design
The registry identifies a two-arm parallel design rather than a factorial design.
Bayesian methods
No Bayesian analysis is identified in the registry-reported statistical-method fields.
These omissions are statistically meaningful. A clinical trial should not be described as using a design or inferential framework merely because that framework is common in other studies. The analysis presented here is restricted to the methods contained in the ClinicalTrials.gov record.
17. What the Different Tests Contribute
| Method | Endpoint structure | Question addressed |
|---|---|---|
| Log-rank test | Time-to-event | Do the treatment groups have different event-time distributions under the reported survival-testing framework? |
| Fisher exact test | Binary | Is the binary outcome associated with treatment group under the exact categorical framework? |
| Wilcoxon signed rank test | Paired continuous change | Does the distribution of within-participant changes differ from the specified null within an analysis group? |
This progression is a useful statistical lesson. The same randomized trial can require different analytical tools because different endpoints encode information differently. A survival time is not analyzed like a binary response, and a paired change score is not analyzed like an independent two-group survival comparison.
18. Trial Timeline
Trial start
The registry lists September 26, 2006 as the study start date.
Primary completion
The registry lists December 23, 2013 as the primary completion date.
Primary survival database date
The registered overall-survival definition states that survival data reflects the database as of December 23, 2013.
Registry status
the ClinicalTrials.gov record classifies CHAARTED as COMPLETED.
19. Why the Primary Result Is a Time-to-Event Result
Overall survival is defined from randomization to death or the date last known alive. This definition contains two essential pieces of information: when the event occurred and how long each participant was observed without the event.
The survival function represents the probability of remaining event-free beyond time t. In an observed trial, estimation must account for participants whose event time is not observed because their follow-up ends before an event.
This is why simply asking how many participants died by the database cutoff would discard information. A participant known to be alive for a long period contributes more follow-up information than someone who is observed only briefly before censoring.
The log-rank test extends this time-to-event framework to the treatment comparison by evaluating observed versus expected events over the sequence of event times. Its strength is that it uses the longitudinal structure of the survival data rather than collapsing the endpoint into a single binary status.
20. Why This Trial Matters Statistically
CHAARTED is a useful teaching example because the ClinicalTrials.gov record brings together several core ideas in clinical-trial statistics without requiring the reader to treat every endpoint as though it were analyzed in the same way.
| Concept | How it appears in CHAARTED |
|---|---|
| Randomization | Participants were randomized to two parallel treatment arms. |
| Superiority testing | The trial's hypothesis type is superiority. |
| Time-to-event analysis | Overall survival, time to clinical progression, and time to castration resistant prostate cancer. |
| Log-rank test | Used for the reported time-to-event comparisons. |
| One-sided design | The study was designed with a one-sided type I error of 0.025. |
| Power | The design specified 80% power for the stated treatment-effect assumption. |
| Binary outcomes | PSA complete response at 6 and 12 months. |
| Fisher exact test | Used for the reported PSA complete-response comparisons. |
| Paired analysis | QOL change from baseline to 3 months. |
| Wilcoxon signed rank test | Used for the reported within-arm QOL analyses. |
| Analysis populations | All randomized patients for the primary and several secondary analyses; patients with both QOL assessments for the QOL analysis. |
| Safety denominators | Serious adverse events are reported as affected participants divided by participants at risk for each arm. |
The strongest statistical lesson is not simply that several P-values are small. It is that each P-value belongs to a specific endpoint, analysis population, test, and hypothesis. Interpreting a clinical trial requires keeping those pieces connected.
21. Statistical Methods Explained: Reading the Results Correctly
The overall-survival P-value of 0.0003 indicates strong statistical evidence under the reported test, but the ClinicalTrials.gov record does not contain an effect estimate. A treatment effect cannot be reconstructed from the P-value alone.
The QOL results illustrate an especially important distinction. P = 0.0009 within one arm and P = 0.40 within the other do not constitute a formal test that the two arms changed differently. The comparative question requires a between-arm analysis of change.
Overall survival retains information about event timing and censoring, whereas PSA complete response at a fixed month reduces the endpoint to a binary classification at that assessment time. Different information structures justify different statistical methods.
A sufficiently small P-value can establish evidence against a null hypothesis without establishing whether the observed difference is large enough to matter clinically. Effect estimates and confidence intervals are essential for quantifying magnitude and precision when they are available.
22. Important Limitations and Interpretation Issues
- Effect estimates are incompletely represented in the registry analysis: the primary statistical analysis reports a P-value but does not provide an overall-survival hazard ratio or confidence interval in the ClinicalTrials.gov record.
- Absolute survival results are not reported: median overall survival and time-specific survival estimates are not included in the ClinicalTrials.gov record and therefore are not reported here.
- Secondary endpoint effect sizes are also absent: the ClinicalTrials.gov record provides P-values for the secondary analyses but not corresponding effect estimates or confidence intervals.
- QOL analyses are within-arm: the statistical analyses posted on ClinicalTrials.gov compare baseline with 3-month QOL within each treatment group rather than reporting a formal between-group treatment comparison.
- Missing QOL assessments: the QOL analysis is restricted to patients with both baseline and 3-month assessments. The ClinicalTrials.gov record does not specify an imputation procedure.
- Multiplicity: several secondary endpoints were tested, but the ClinicalTrials.gov record does not specify a multiplicity-adjustment procedure for those comparisons.
- One-sided design: the study was designed with a one-sided type I error of 0.025, while the statistical-analysis record identifies two-sided confidence intervals. These should not be conflated.
- Statistical significance is not effect size: small P-values do not establish the magnitude or practical importance of an observed treatment difference.
- Safety and efficacy use different information: serious adverse-event counts should be evaluated independently from efficacy endpoints rather than combined into a single statistical conclusion.
- Registry scope: this analysis is restricted to the ClinicalTrials.gov record. It does not add results from external publications beyond the source links.
23. Clinical Biostats Interpretation of the Evidence
What is directly supported
The registered primary overall-survival comparison was analyzed with a log-rank test in all randomized patients and produced P = 0.0003 under a superiority hypothesis.
What is also reported
Time to clinical progression and time to castration resistant prostate cancer each have log-rank P-values <0.0001; PSA complete response at 6 and 12 months each has Fisher exact P <0.0001.
What requires caution
The ClinicalTrials.gov record does not provide effect estimates or confidence intervals for these endpoints, so statistical significance cannot be translated into a numerical treatment-effect magnitude here.
What QOL adds
QOL change was tested within each arm using Wilcoxon signed rank tests, with P = 0.0009 and P = 0.40 respectively; these are not a formal between-arm comparison.
The most defensible interpretation is therefore deliberately narrower than a publication-style efficacy summary. The ClinicalTrials.gov record establishes statistical evidence for the primary overall-survival comparison and several secondary outcomes, but it does not provide enough numerical information to quantify the size or precision of those treatment effects. Maintaining that distinction is part of rigorous statistical reporting.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Calculators
26. Sources
- ClinicalTrials.gov: NCT00309985 — CHAARTED.
- PubMed: PMID 26244877.
- PubMed: PMID 39633601.
- PubMed: PMID 38429210.
- PubMed: PMID 34007017.
- PubMed: PMID 32313142.
Continue with the statistical methods behind CHAARTED
Explore the survival-analysis, categorical-data, nonparametric, hypothesis-testing, and clinical-trial methods represented in this trial.
27. Record Summary
CHAARTED provides a useful example of endpoint-specific clinical-trial statistics. The trial was a randomized phase 3 parallel study of androgen-deprivation therapy with docetaxel versus androgen-deprivation therapy alone in metastatic hormone-sensitive prostate cancer, with overall survival registered as the primary time-to-event endpoint. The primary analysis used a log-rank test in all randomized patients and reported P = 0.0003 under a superiority framework.
The secondary analyses demonstrate how statistical methods change with endpoint structure: log-rank tests were used for time to clinical progression and time to castration resistant prostate cancer; Fisher exact tests were used for PSA complete response at 6 and 12 months; and Wilcoxon signed rank tests were used for within-arm QOL change from baseline to 3 months. Serious adverse events were reported separately by treatment arm as affected participants and participants at risk.
The key statistical limitation of the ClinicalTrials.gov record is that the posted analyses provide P-values without corresponding effect estimates and confidence intervals for the principal efficacy endpoints. The appropriate interpretation is that the reported statistical tests provide evidence against their respective null hypotheses, while the magnitude and precision of the treatment effects require numerical estimates that are not contained in the ClinicalTrials.gov record.