← Clinical Trials
Prostate Cancer Phase 3 Time-to-Event Analysis NCT01308567

FIRSTANA: Complete Statistical Analysis of Cabazitaxel in Prostate Cancer

An independent statistical review of the randomized phase 3 FIRSTANA trial comparing cabazitaxel 20 mg/m2 and cabazitaxel 25 mg/m2 with docetaxel 75 mg/m2, with prednisone, in patients with metastatic castration-resistant prostate cancer.

Trial period: 2011-05-05 to 2015-09  ·  Enrollment: 1168  ·  Sponsor: Sanofi
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results and trial-specific design details are restricted to the ClinicalTrials.gov data posted on ClinicalTrials.gov for FIRSTANA. 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

FIRSTANA was a randomized, open-label, parallel-group phase 3 treatment trial enrolling 1168 participants. The three-arm study compared two cabazitaxel dose levels with docetaxel, with prednisone included as an intervention. The primary endpoint was overall survival, analyzed as a time-to-event outcome in the intention-to-treat population.

1168
Enrolled
Randomized participants
3
Arms
Parallel-group design
0.975
OS HR
Cabazitaxel 25 vs docetaxel
1.009
OS HR
Cabazitaxel 20 vs docetaxel
FeatureFIRSTANA
Trial nameFIRSTANA
PhasePhase 3
ConditionProstate Cancer
PopulationPatients with metastatic castration-resistant prostate cancer
DesignRandomized, parallel-group, open-label
AllocationRandomized
Primary purposeTreatment
Enrollment1168
Primary endpointOverall Survival (OS)
Primary endpoint typeTime-to-event
Primary hypothesis typeSuperiority
Primary analysis methodsLog-rank test and Cox proportional-hazards model
Lead sponsorSanofi
Sponsor typeIndustry
StatusCompleted
ClinicalTrials.govNCT01308567

2. Clinical Question

The primary statistical question was whether overall survival differed between patients randomized to docetaxel 75 mg/m2 and patients randomized to either of the two cabazitaxel dose levels. The registry classified the primary hypothesis as superiority.

Population

Patients with metastatic castration-resistant prostate cancer.

Intervention

Cabazitaxel at 20 mg/m2 or 25 mg/m2, with prednisone listed among the study interventions.

Comparator

Docetaxel 75 mg/m2, with prednisone listed among the study interventions.

Primary question

Does either cabazitaxel dose produce a different overall-survival hazard relative to docetaxel 75 mg/m2?

3. Trial Design

01
Randomize1168 participants
02
3 armsTwo cabazitaxel doses + docetaxel
03
TreatmentStudy treatment with prednisone
04
Follow-upTime-to-event outcomes
05
AnalysisLog-rank + Cox model
ARM 1

Docetaxel

  • Docetaxel 75 mg/m2
  • Prednisone was listed as an intervention.
ARM 2

Cabazitaxel 20 mg/m2

  • Cabazitaxel 20 mg/m2
  • Prednisone was listed as an intervention.
ARM 3

Cabazitaxel 25 mg/m2

  • Cabazitaxel 25 mg/m2
  • Prednisone was listed as an intervention.
Open-label design: the registry describes masking as none. Thus, treatment assignment was not masked in the trial design recorded by ClinicalTrials.gov. For an overall-survival endpoint, the event itself—death from any cause—is less dependent on subjective outcome assessment than endpoints requiring interpretation of symptoms or imaging, although the open-label design remains an important feature of the evidence base.

4. Trial Timeline

2011-05-05

Trial start

The registered study start date was May 5, 2011.

2015-09

Primary completion

The registered primary completion period was September 2015.

Completed

Final registry status

The trial is recorded as completed, with results posted to ClinicalTrials.gov.

5. Endpoints

EndpointRegistry definition / time frameRole
Overall Survival (OS) OS was defined as the time interval from the date of randomization to the date of death due to any cause. In the absence of confirmation of death, survival time was censored at the last date participant was known to be alive, or at the cut-off date if the participant's last contact was after the cut-off date. The study cut-off date for the final analysis of OS was the date when the 774th death had occurred. Time frame: Baseline up to death or study cut-off date, whichever was earlier (maximum duration: 51 months). Primary
Progression Free Survival (PFS) Baseline up to tumor progression, PSA progression, pain progression or death (maximum duration: 51 months). Secondary
Time to Tumor Progression Free Survival Baseline up to tumor progression or death due to any cause or study cut-off date, whichever was earlier (maximum duration as posted in the registry record). Secondary
Time to Prostate Serum Antigen Progression Free Survival (PSA-PFS) Baseline up to PSA progression or death due to any cause or study cut-off date, whichever was earlier (maximum duration as posted in the registry record). Secondary
Time to Pain Progression Free Survival (Pain PFS) Baseline until disease progression, death or study cut-off date (maximum duration: 51 months). Secondary
Skeletal Related Events (SRE) Free Survival Baseline until occurrence of first SRE or death (maximum duration: 51 months). Secondary

The registry contains one primary endpoint—overall survival—and multiple secondary time-to-event outcomes. This creates a coherent statistical structure: randomization establishes the comparison, follow-up generates event and censoring times, and survival methods summarize the distribution of those times between treatment groups.

6. Analysis Populations

PopulationRegistry descriptionRole in reported analyses
Intention-to-treat All randomized participants. Primary OS analyses and the posted secondary efficacy analyses.
Why ITT matters: analyzing randomized participants according to their randomized treatment assignment preserves the comparison created by randomization. It also means that the OS estimate is a comparison of the treatment strategies represented by the randomized arms rather than a comparison restricted to patients who remained on treatment.

7. Primary Results: Overall Survival

ClinicalTrials.gov reports two formal primary analyses of overall survival. Both use the intention-to-treat population and compare each cabazitaxel dose separately with docetaxel 75 mg/m2. The reported hypothesis type is superiority, and both analyses use a log-rank test with the hazard ratio estimated using a Cox proportional-hazards regression model.

Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2

Hazard ratio for overall survival

0.975

95% CI: 0.819–1.16   ·   P = 0.7574

Two-sided 95% confidence interval · ITT population

Primary OS comparisonEstimate95% CIP-valueAnalysis
Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2 HR 0.975 0.819–1.16 0.7574 Log-rank test; Cox proportional-hazards model
Clinical Biostats interpretation

An HR of 0.975 means that the fitted model estimated the instantaneous rate of death in the cabazitaxel 25 mg/m2 group at approximately 97.5% of the corresponding rate in the docetaxel 75 mg/m2 group. Expressed as a simple relative-hazard interpretation, this corresponds to an estimated hazard approximately 2.5% lower under the fitted model.

That does not mean that 2.5% fewer patients died, that individual patients experienced a 2.5% reduction in their personal probability of death, or that survival times differed by 2.5%. A hazard ratio is a relative time-to-event measure, not an absolute risk difference.

The two-sided 95% CI of 0.819–1.16 describes uncertainty around the estimated hazard ratio. Its width indicates that the observed HR is not a highly precise estimate of a single underlying relative hazard. The interval also includes 1, the value corresponding to equal hazards under the model.

The P-value of 0.7574 addresses the statistical evidence against the null hypothesis specified for the comparison; it does not measure the size or clinical importance of the treatment effect. A P-value is therefore not interchangeable with the HR or its confidence interval.

The Cox model was adjusted for ECOG PS score at baseline, measurable disease at baseline, and region with commercial availability of cabazitaxel at the time of randomization. Interpretation of the HR also depends on the proportional-hazards framework used by the Cox model. Censoring is part of the analysis because participants without confirmed death were censored according to the registry definition.

Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2

Hazard ratio for overall survival

1.009

95% CI: 0.85–1.197   ·   P = 0.9967

Two-sided 95% confidence interval · ITT population

Primary OS comparisonEstimate95% CIP-valueAnalysis
Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2 HR 1.009 0.85–1.197 0.9967 Log-rank test; Cox proportional-hazards model
Clinical Biostats interpretation

An HR of 1.009 means that the estimated instantaneous rate of death under the fitted Cox model was approximately 1.009 times that of the docetaxel group for the cabazitaxel 20 mg/m2 comparison. The point estimate is therefore very close to 1.

This does not establish that the two treatments produce exactly identical survival, nor does it mean that individual patients have identical outcomes. The estimate is subject to sampling uncertainty, which is summarized by the confidence interval.

The two-sided 95% CI of 0.85–1.197 spans 1. It therefore includes values corresponding to a lower estimated hazard as well as values corresponding to a higher estimated hazard for cabazitaxel relative to docetaxel. The confidence interval is more informative about that uncertainty than the point estimate alone.

The P-value of 0.9967 is a measure of statistical evidence under the specified hypothesis-testing framework; it is not a measure of effect size. The HR and its confidence interval should remain the principal quantitative description of the estimated treatment effect.

As with the 25 mg/m2 comparison, the Cox model was adjusted for ECOG PS score at baseline, measurable disease at baseline, and region with commercial availability of cabazitaxel at the time of randomization. The interpretation remains conditional on the model's proportional-hazards framework and the registry's censoring definition.

8. Comparing the Two Primary Hazard Ratios

Comparison vs Docetaxel 75 mg/m2OS HR95% CIP-value
Cabazitaxel 25 mg/m20.9750.819–1.160.7574
Cabazitaxel 20 mg/m21.0090.85–1.1970.9967

The two comparisons answer related but distinct questions because each evaluates a different cabazitaxel dose against the same docetaxel reference group. The point estimates are close to 1 in both comparisons, while both confidence intervals include 1. The registry does not provide a direct statistical comparison of the two cabazitaxel doses with one another in the registry-reported primary analyses, so the two HRs should not be treated as a formal head-to-head test between cabazitaxel 20 mg/m2 and 25 mg/m2.

Important multiplicity point: two primary comparisons against the same control create a multiple-comparison structure. The ClinicalTrials.gov record identifies both comparisons as primary superiority analyses, but do not provide an alpha-allocation or multiplicity-adjustment scheme in the ClinicalTrials.gov record. The two reported P-values should therefore be interpreted exactly as posted rather than assuming an unstated multiplicity procedure.

9. Secondary Results: Progression-Free Survival

Progression-free survival was a secondary time-to-event endpoint. The registry defines its time frame as baseline up to tumor progression, PSA progression, pain progression or death, with a maximum duration of 51 months. Both analyses were performed in the ITT population using Cox proportional-hazards models adjusted for ECOG PS score at baseline, measurable disease at baseline, and region with commercial availability of cabazitaxel at the time of randomization.

ComparisonHR95% CIP-value
Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2 0.989 0.849–1.152 Not reported in registry-reported statistical analysis
Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2 1.063 0.913–1.236 Not reported in registry-reported statistical analysis
How to read these PFS estimates

The HR of 0.989 for cabazitaxel 25 mg/m2 versus docetaxel is very close to 1, while the HR of 1.063 for cabazitaxel 20 mg/m2 versus docetaxel is modestly above 1. These are estimates of relative event hazards, not direct differences in median PFS or percentages of patients progressing.

The corresponding confidence intervals—0.849–1.152 and 0.913–1.236—both include 1. Thus, the intervals encompass both lower and higher estimated hazards relative to docetaxel.

The statistical analyses posted on ClinicalTrials.gov do not contain P-values for these PFS comparisons. Accordingly, no formal P-value-based conclusion is added here. The hazard ratios and confidence intervals are the quantities actually reported in the ClinicalTrials.gov record.

10. Secondary Results: Time to Tumor Progression

ComparisonHR95% CIAnalysis population
Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2 0.958 0.785–1.17 ITT
Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2 0.916 0.75–1.118 ITT
Statistical interpretation

The point estimates are below 1 for both cabazitaxel dose comparisons. In a Cox model, an HR below 1 corresponds to a lower estimated instantaneous event rate for cabazitaxel relative to docetaxel. However, the 95% confidence intervals are 0.785–1.17 and 0.75–1.118, respectively, and both include 1.

The ClinicalTrials.gov record does not report P-values for these analyses, so the interpretation should remain focused on the estimated hazard ratios and their uncertainty rather than adding an unreported hypothesis-test result.

11. Secondary Results: PSA-PFS

ComparisonHR95% CIAnalysis population
Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2 0.948 0.8–1.123 ITT
Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2 1.047 0.886–1.238 ITT

PSA-PFS is defined in the ClinicalTrials.gov record as baseline up to PSA progression or death due to any cause or study cut-off date, whichever was earlier. The analyses used Cox proportional-hazards models in the ITT population.

Statistical interpretation

The HR of 0.948 for cabazitaxel 25 mg/m2 represents a fitted hazard estimate slightly below 1, whereas the HR of 1.047 for cabazitaxel 20 mg/m2 is slightly above 1. Neither point estimate should be interpreted without its confidence interval.

The 95% CIs are 0.8–1.123 and 0.886–1.238. Both intervals include 1, demonstrating the uncertainty in the estimated relative hazards. The analyses posted on ClinicalTrials.gov do not report P-values for these comparisons.

12. Secondary Results: Pain PFS

ComparisonHR95% CIAnalysis population
Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2 1.189 0.986–1.434 ITT
Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2 1.189 0.985–1.435 ITT

The registry defines Pain PFS as baseline until disease progression, death or study cut-off date, with a maximum duration of 51 months. The Cox model was adjusted for ECOG PS score at baseline, measurable disease at baseline, and region with commercial availability of cabazitaxel at the time of randomization.

Statistical interpretation

Both cabazitaxel dose comparisons have the same reported point estimate of 1.189. Under the Cox model, that corresponds to an estimated instantaneous event rate approximately 18.9% higher than the docetaxel reference group for the Pain PFS endpoint.

This does not mean that 18.9% more patients experienced pain progression, nor does it quantify an absolute difference in time without pain progression. The hazard ratio summarizes relative event rates over the analyzed follow-up.

The confidence intervals are 0.986–1.434 and 0.985–1.435. Each interval is close to 1 at its lower boundary and extends above 1, so the point estimate should not be treated as a precise estimate of a higher hazard.

13. Secondary Results: Skeletal Related Events Free Survival

ComparisonHR95% CIAnalysis population
Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2 1.121 0.886–1.417 ITT
Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2 1.014 0.798–1.288 ITT

SRE-free survival was defined as baseline until occurrence of the first skeletal related event or death, with a maximum duration of 51 months. Both dose comparisons were evaluated with Cox proportional-hazards models.

Statistical interpretation

The estimated HRs of 1.121 and 1.014 are both close enough to 1 that their interpretation depends strongly on the corresponding confidence intervals. The 95% CIs—0.886–1.417 and 0.798–1.288—both include 1.

As with the other secondary time-to-event analyses, the HR should not be converted into an absolute event probability or a difference in median survival without the underlying survival estimates. The ClinicalTrials.gov record does not provide those quantities.

14. Consolidated Time-to-Event Results

EndpointCabazitaxel 25 vs Docetaxel95% CICabazitaxel 20 vs Docetaxel95% CI
Overall Survival0.9750.819–1.161.0090.85–1.197
Progression Free Survival0.9890.849–1.1521.0630.913–1.236
Time to Tumor Progression Free Survival0.9580.785–1.170.9160.75–1.118
PSA-PFS0.9480.8–1.1231.0470.886–1.238
Pain PFS1.1890.986–1.4341.1890.985–1.435
SRE Free Survival1.1210.886–1.4171.0140.798–1.288

This consolidated view illustrates why a trial with several time-to-event endpoints should not be summarized by one hazard ratio alone. Each endpoint represents a different event definition. A hazard ratio for OS answers a different question from a hazard ratio for PSA-PFS, Pain PFS, or SRE-free survival.

15. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk.

Treatment armSerious adverse events affectedAt riskReported affected / at risk
Docetaxel 75 mg/m2 126 387 126/387
Cabazitaxel 20 mg/m2 127 369 127/369
Cabazitaxel 25 mg/m2 188 391 188/391
Serious adverse events: affected / at risk
Docetaxel 75 mg/m2
126/387
Cabazitaxel 20 mg/m2
127/369
Cabazitaxel 25 mg/m2
188/391

The bars above are a visual representation of the registry-reported affected/at-risk counts. The page does not introduce additional safety endpoints or adverse-event categories beyond those contained in the trial data. Safety should also be interpreted separately from the ITT efficacy analyses because the registry's serious-adverse-event figures are presented by treatment arm and at-risk population.

Safety denominator matters: an affected/at-risk count is not automatically interchangeable with an ITT efficacy denominator. When comparing safety information with efficacy results, the analysis population and denominator should always be identified explicitly.

16. Statistical Methodology

Log-rank test

The primary OS comparisons used the log-rank test. The log-rank test is designed for comparing survival distributions between groups when observations can be right-censored. Rather than comparing only a single time point, it uses the ordering of observed event times over follow-up.

Conceptual question
H0: survival experience is equivalent between the comparison groups

For FIRSTANA, the registry-reported primary analyses use the log-rank method to compare each cabazitaxel dose with docetaxel 75 mg/m2. The registry reports two-sided 95% confidence intervals for the associated hazard ratios.

Cox proportional-hazards model

The hazard ratio was estimated using a Cox proportional-hazards regression model. For the primary OS analyses, the model was adjusted by ECOG PS score at baseline, measurable disease at baseline, and region with commercial availability of cabazitaxel at the time of randomization. The same adjustment variables are reported in the registry-reported secondary analyses.

Conceptual form
h(t|X) = h0(t) exp(βTX)

The exponentiated treatment coefficient, exp(β), is interpreted as a hazard ratio under the proportional-hazards model. An HR below 1 indicates a lower estimated instantaneous event rate for the treatment group relative to the reference group; an HR above 1 indicates a higher estimated event rate.

Adjustment variables

Adjustment can improve statistical precision and account for prespecified covariates in the fitted model. In FIRSTANA's registry-reported analyses, the Cox model includes baseline ECOG performance status, measurable disease at baseline, and region with commercial availability of cabazitaxel at randomization.

Intention-to-treat analysis

The primary and registry-reported secondary efficacy analyses use the ITT population, defined as all randomized participants. This is particularly important in a randomized trial because the treatment comparison remains anchored to the randomization process rather than to subsequent treatment exposure.

Censoring

For OS, the registry specifies that participants without confirmed death were censored at the last date they were known to be alive, or at the cut-off date when the participant's last contact was after the cut-off date. Censoring allows participants who do not experience the event during observed follow-up to contribute information up to the point at which their event status is no longer observed.

17. Statistical Methods Explained

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

Overall survival is a time-to-event endpoint. Some participants may still be alive at the analysis cut-off, so their exact event time is unknown and their observations are censored. The log-rank test is specifically designed for comparing time-to-event distributions while accounting for such censoring. In FIRSTANA, it was the reported method for both primary OS comparisons.

What does an HR of 0.975 mean?

An HR of 0.975 means the fitted Cox model estimated the instantaneous death rate for cabazitaxel 25 mg/m2 at approximately 97.5% of the rate for docetaxel 75 mg/m2. It does not mean that survival probability was 97.5%, that 2.5% fewer patients died, or that each patient experienced a 2.5% reduction in mortality risk.

Why does an HR need a confidence interval?

The HR is an estimate obtained from observed trial data. A confidence interval communicates the statistical uncertainty surrounding that estimate. For the cabazitaxel 25 mg/m2 OS comparison, the HR is 0.975 and the 95% CI is 0.819–1.16. For cabazitaxel 20 mg/m2, the HR is 1.009 and the 95% CI is 0.85–1.197. Looking only at the point estimates would conceal that uncertainty.

Why is the P-value not an effect-size measure?

A P-value measures the compatibility of the observed data with a specified null hypothesis under the statistical testing framework. It depends on the amount of information and the statistical model as well as on the magnitude of the observed difference. The HR describes the estimated relative effect; the confidence interval describes uncertainty around it. These quantities answer different questions.

Why does the Cox model adjust for baseline variables?

The registry-reported Cox model adjusts for ECOG PS score at baseline, measurable disease at baseline, and region with commercial availability of cabazitaxel at randomization. These variables enter the regression model alongside treatment assignment, allowing the treatment hazard ratio to be estimated conditional on the specified covariate structure.

What does censoring mean in OS?

For a participant who has not been observed to die, censoring means the analysis uses the participant's information through the last date they were known to be alive, or the applicable study cut-off. Censoring is not the same as treating the participant as having survived indefinitely. The survival model uses the observed follow-up time without assigning an unobserved death date.

Why are the secondary endpoints analyzed separately?

PFS, tumor-progression time, PSA-PFS, Pain PFS, and SRE-free survival use different event definitions. Even though all are time-to-event outcomes and are analyzed with Cox models in the ClinicalTrials.gov record, they measure different clinical events. A hazard ratio for one endpoint should therefore not be substituted for another.

18. Confidence Intervals and Statistical Precision

The confidence intervals across the reported analyses illustrate an important principle of clinical-trial interpretation: the point estimate is only one part of the result.

EndpointCabazitaxel 25 mg/m2 HR95% CICabazitaxel 20 mg/m2 HR95% CI
OS0.9750.819–1.161.0090.85–1.197
PFS0.9890.849–1.1521.0630.913–1.236
Time to tumor progression0.9580.785–1.170.9160.75–1.118
PSA-PFS0.9480.8–1.1231.0470.886–1.238
Pain PFS1.1890.986–1.4341.1890.985–1.435
SRE-free survival1.1210.886–1.4171.0140.798–1.288

Every registry-reported 95% confidence interval includes 1. This does not mean that the true effect must equal exactly 1. Rather, it means that the interval of plausible values under the stated confidence procedure includes the no-hazard-difference value. The width of each interval also provides information about precision.

Do not confuse statistical uncertainty with individual variability. A confidence interval around an HR describes uncertainty about the estimated population-level relative hazard under the model. It is not a prediction interval for the outcome of an individual patient.

19. The Proportional-Hazards Assumption

The Cox proportional-hazards model is the principal regression method reported for the FIRSTANA time-to-event analyses. A central modeling concept is proportionality of hazards: the relative hazard between treatment groups is assumed to be reasonably stable over the modeled time scale.

What an HR summarizes
HR(t) ≈ htreatment(t) / hreference(t)

Under proportional hazards, the treatment-to-reference hazard ratio is represented by a single parameter rather than a separate ratio at every follow-up time.

If hazards are not approximately proportional, a single HR can become a less complete description of the survival experience. In that situation, graphical survival estimates, time-specific measures, or alternative models may provide additional insight. The registry-reported FIRSTANA data do not report a formal proportional-hazards diagnostic, so no such diagnostic is asserted here.

20. Primary Analysis Logic

Step 1 · Randomization

Participants were randomized into three parallel treatment arms, establishing the basis for the between-group efficacy comparison.

Step 2 · Define the event

For OS, the event was death due to any cause. Participants without confirmed death were censored according to the registry definition.

Step 3 · Compare survival

The log-rank test provides the reported primary survival comparison, while the Cox model estimates the hazard ratio.

Step 4 · Quantify uncertainty

The hazard ratio is accompanied by a two-sided 95% confidence interval, allowing the estimated effect to be interpreted in context.

This sequence is useful because it separates the roles of different statistical quantities. Randomization addresses treatment allocation. The endpoint definition determines what event is being measured. The log-rank test addresses a group comparison. The Cox model supplies a relative effect estimate. The confidence interval communicates uncertainty around that estimate.

21. Multiplicity and Multiple Primary Comparisons

FIRSTANA has one registered primary endpoint—overall survival—but the statistical analyses posted on ClinicalTrials.gov contain two primary OS comparisons: cabazitaxel 25 mg/m2 versus docetaxel 75 mg/m2, and cabazitaxel 20 mg/m2 versus docetaxel 75 mg/m2.

Primary comparisonHypothesis typeTwo-sided CIP-value
Cabazitaxel 25 mg/m2 vs Docetaxel 75 mg/m2Superiority95%0.7574
Cabazitaxel 20 mg/m2 vs Docetaxel 75 mg/m2Superiority95%0.9967

Multiple formal comparisons can affect the overall probability of making at least one false-positive claim if each is tested independently at the same nominal significance threshold. The ClinicalTrials.gov record does not specify an alpha-allocation or multiplicity adjustment. Consequently, this page does not impose a correction that is not documented in the ClinicalTrials.gov record.

Interpretive discipline: the existence of two primary comparisons does not justify selecting whichever HR is numerically smaller and treating it as the trial's single primary result. Each comparison has its own treatment contrast, estimate, confidence interval, and P-value.

22. Randomization and Causal Interpretation

Randomization is the central design feature that permits a treatment comparison to be interpreted causally under the trial framework. Because treatment assignment is randomized rather than selected by investigators or participants based on prognosis, baseline prognostic factors are balanced in expectation across randomized groups.

The FIRSTANA registry identifies the allocation as randomized and the design model as parallel. The efficacy analyses were conducted in the ITT population, preserving the randomized treatment assignment as the basis of the comparison.

However, randomization does not eliminate every statistical issue. Censoring, model assumptions, multiple comparisons, missing information, endpoint definitions, and the distinction between efficacy and safety populations can all affect interpretation of the resulting estimates.

23. Why This Trial Matters Statistically

FIRSTANA is a useful teaching example because its registry results show how a three-arm randomized trial can generate multiple treatment contrasts while maintaining a consistent time-to-event analytical framework.

ConceptHow it appears in FIRSTANA
RandomizationThe study used randomized allocation in a three-arm parallel design.
Three-arm comparisonTwo cabazitaxel dose levels were each compared with docetaxel 75 mg/m2 in the registry-reported primary analyses.
Intention-to-treatPrimary and registry-reported secondary efficacy analyses used the ITT population.
Time-to-event endpointOverall survival was the primary endpoint, with multiple secondary time-to-event outcomes.
Log-rank testThe reported method for both primary OS comparisons.
Cox modelUsed to estimate hazard ratios for primary and secondary time-to-event analyses.
Covariate adjustmentThe Cox model was adjusted for baseline ECOG PS, measurable disease, and region with commercial availability of cabazitaxel.
Confidence intervalsPrimary HRs were reported with two-sided 95% confidence intervals.
MultiplicityOne primary endpoint generated two formal primary treatment comparisons.
CensoringOS censoring was explicitly defined by the last date known alive or the study cut-off rule.
Safety denominatorsSerious adverse events were reported separately by arm using affected and at-risk counts.

24. What the Hazard Ratio Does — and Does Not — Mean

Example: OS HR 0.975

The OS HR of 0.975 for cabazitaxel 25 mg/m2 versus docetaxel 75 mg/m2 is a relative measure of the modeled instantaneous rate of death. It can be described as approximately 97.5% of the reference hazard under the fitted model.

It does not mean that 97.5% of patients survived, that 2.5% fewer patients died, or that every patient experienced a 2.5% reduction in risk.

Example: OS HR 1.009

The OS HR of 1.009 for cabazitaxel 20 mg/m2 versus docetaxel 75 mg/m2 is approximately 1.009 times the modeled instantaneous death rate in the reference group.

It does not imply a 0.9% absolute increase in deaths or a 0.9% difference in survival probability. The HR is a time-to-event model parameter, not an absolute risk measure.

Why the confidence interval matters

The 95% CIs of 0.819–1.16 and 0.85–1.197 show that the estimated OS effects are uncertain. Both intervals include 1, so the point estimates should not be interpreted in isolation as definitive evidence of a directional difference in OS.

25. Clinical Interpretation vs Statistical Interpretation

Statistical interpretation

The registry-reported primary analyses estimate OS hazard ratios of 0.975 and 1.009 for the two cabazitaxel dose comparisons, with two-sided 95% confidence intervals that include 1. The corresponding reported P-values are 0.7574 and 0.9967.

Endpoint interpretation

The secondary analyses extend the same time-to-event framework to PFS, tumor progression, PSA progression, pain progression, and skeletal related events. Each endpoint has its own event definition and hazard ratio.

The statistical results should not be collapsed into a single qualitative statement about the entire trial. Overall survival, progression-free survival, PSA-PFS, Pain PFS, and SRE-free survival represent different outcomes. Similarly, safety is summarized using a different analysis perspective from ITT efficacy.

26. Important Limitations and Interpretation Issues

27. Planned Statistical Interpretation Framework

For any additional FIRSTANA endpoint not accompanied by a formal statistical analysis in the ClinicalTrials.gov record, the appropriate framework depends on the endpoint type. For time-to-event endpoints, Kaplan-Meier estimation is commonly used to describe the survival distribution, the log-rank test can compare groups, and the Cox proportional-hazards model can estimate a hazard ratio with a confidence interval.

The registry-reported FIRSTANA data already provide formal statistical analyses for the primary endpoint and the listed secondary time-to-event endpoints. Accordingly, this page reports those estimates directly rather than replacing them with hypothetical analyses.

Methodological principle: a statistical analysis page should distinguish what the registry actually reports from what would normally be considered an appropriate analytical method. That distinction is especially important when a registry provides an effect estimate and confidence interval but omits a P-value or other secondary summary statistic.

28. Related Tutorials

Learn more about the methods used in this trial:

29. Related Calculators

30. Sources

Continue through Clinical Biostats

Connect this trial's time-to-event endpoints with deeper statistical tutorials and analysis tools.

31. Record Summary

FIRSTANA is a randomized, parallel-group phase 3 trial with 1168 enrolled participants and three treatment arms: docetaxel 75 mg/m2, cabazitaxel 20 mg/m2, and cabazitaxel 25 mg/m2, with prednisone listed among the study interventions. Overall survival was the single registered primary endpoint and was analyzed as a time-to-event outcome in the ITT population using log-rank testing and Cox proportional-hazards regression.

The two primary OS comparisons produced HRs of 0.975 and 1.009, with two-sided 95% confidence intervals of 0.819–1.16 and 0.85–1.197, respectively. The corresponding reported P-values were 0.7574 and 0.9967. The secondary analyses extended the same Cox-model framework to PFS, tumor progression, PSA-PFS, Pain PFS, and SRE-free survival.

The most important statistical lesson is that these results should be read as a collection of prespecified treatment contrasts and endpoint-specific estimates rather than as a single number. Hazard ratios quantify relative event rates, confidence intervals quantify uncertainty around those estimates, P-values address hypothesis-testing evidence rather than effect magnitude, and the endpoint definition determines what clinical event the model is actually analyzing.

Clinical Biostats methodology: This page distinguishes reported numerical evidence from statistical interpretation. Where the registry-reported FIRSTANA registry data provide an estimate and confidence interval, those values are reported exactly as reported in the registry. Where the ClinicalTrials.gov record does not provide a statistic, it is not inferred or reconstructed from external information.