← Clinical Trials
mCRPC Phase 3 Time-to-Event Analysis NCT03395197

TALAPRO-2: Complete Statistical Analysis of Talazoparib + Enzalutamide in mCRPC

An independent statistical review of the randomized phase 3 TALAPRO-2 trial evaluating talazoparib with enzalutamide versus placebo with enzalutamide, with emphasis on blinded independent central review–assessed radiographic progression-free survival in Part 2.

Trial start: 2017-12-18  ·  Primary completion: 2022-10-03  ·  Status: Active, not recruiting
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the TALAPRO-2 ClinicalTrials.gov record.

1. Trial at a Glance

TALAPRO-2 is a randomized, parallel-group, quadruple-masked phase 3 trial in mCRPC. The registry reports an enrollment of 1054 participants and two interventions: talazoparib with enzalutamide versus placebo with enzalutamide.

1054
Enrollment
Registry enrollment
2
Arms
Parallel-group design
3
Phase
Phase 3
0.627
All-Comers rPFS HR
95% CI 0.506–0.777
FeatureTALAPRO-2
Trial nameTALAPRO-2
Brief titleTalazoparib + Enzalutamide vs. Enzalutamide Monotherapy in mCRPC
PhasePhase 3
ConditionmCRPC
AllocationRandomized
Design modelParallel
MaskingQuadruple
Primary purposeTreatment
Enrollment1054
Arms2
Lead sponsorPfizer
Sponsor typeIndustry
Trial statusActive, not recruiting
Start2017-12-18
Primary completion2022-10-03
ClinicalTrials.govNCT03395197

2. Clinical Question

The central statistical question in the reported Part 2 analyses is whether talazoparib added to enzalutamide produces a lower hazard of radiographic progression or death than placebo added to enzalutamide.

Population

Participants with mCRPC enrolled in the randomized Part 2 population. One reported analysis concerns all participants randomized to double-blind study treatment in Part 2 Cohort 1; the other concerns DDR-deficient participants randomized to double-blind study treatment in Part 2.

Intervention

Talazoparib with enzalutamide.

Comparator

Placebo with enzalutamide.

Primary question

Does talazoparib plus enzalutamide improve BICR-assessed radiographic progression-free survival relative to placebo plus enzalutamide?

Two distinct populations are important in the registry's formal statistical analyses: the all-comers population in Part 2 Cohort 1 and the population of participants with DDR deficiencies in Part 2. They should not be merged into a single estimate because they represent different analysis populations.

3. Trial Design

01
Enroll1054 participants
02
RandomizeTwo parallel groups
03
MaskQuadruple masking
04
TreatTalazoparib or placebo + enzalutamide
05
AssessRadiographic progression or death
INTERVENTION

Talazoparib + enzalutamide

  • Talazoparib with enzalutamide
  • Randomized study treatment
  • Double-blind treatment in the reported Part 2 efficacy analyses
COMPARATOR

Placebo + enzalutamide

  • Placebo with enzalutamide
  • Randomized study treatment
  • Double-blind treatment in the reported Part 2 efficacy analyses
Allocation
Randomized
Design model
Parallel
Masking
Quadruple
Primary purpose
Treatment

Randomization is statistically important because it establishes the treatment groups before outcome information is observed. In a properly conducted randomized comparison, baseline differences that arise by chance are not interpreted as treatment effects; instead, the treatment assignment defines the comparison from which the efficacy estimate is derived.

4. Endpoints

The registry lists eight primary endpoints. Six concern safety outcomes in Part 1, while two are time-to-event radiographic progression-free survival outcomes in Part 2. The two formal statistical analyses reported in the ClinicalTrials.gov record correspond to the two Part 2 rPFS endpoints.

Registered primary endpointTime frameTypeFormal analysis in the ClinicalTrials.gov record
Number of Participants With Treatment-Emergent Adverse Events (TEAEs) Occuring Within the First 66 Days of Dosing - Part 1 Post dose on Day 1 up to Day 66 in Part 1 Binary No formal statistical analysis reported
Number of Participants With All-Causality Clustered Treatment-Emergent Cytopenias by Preferred Term (PT) and Max CTCAE Grade Occuring Within the First 66 Days of Dosing - Part 1 Post dose on Day 1 up to Day 66 in Part 1 Binary No formal statistical analysis reported
Number of Participants With All-Causality TEAEs During the Overall Period of Part 1 Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy Binary No formal statistical analysis reported
Number of Participants With Treatment-Related TEAEs During the Overall Period of Part 1 Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy Binary No formal statistical analysis reported
Number of Participants With All-Causality Clustered Treatment-Emergent Cytopenias by PT and Max CTCAE Grade Occuring Anytime After Dosing - Part 1 Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy Binary No formal statistical analysis reported
Number of Participants With Treatment-Related Clustered Treatment-Emergent Cytopenias by PT and Max CTCAE Grade in >=10% of Participants Occuring Anytime After Dosing - Part 1 Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy Binary No formal statistical analysis reported
Blinded Independent Central Review (BICR) Assessed Radiographic Progression-Free Survival (rPFS) Per Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 for All-Comers - Part 2 Cohort 1 From the start of treatment to the time of first documented progression, or death (maximum up to 42 months) Time-to-event Yes
BICR Assessed rPFS Per RECIST 1.1 in Patients With DDR Deficiencies - Part 2 From the start of treatment to the time of first documented progression, or death (maximum up to 38 months) Other / unclear Yes

How rPFS was defined

For both reported rPFS analyses, the registry defines rPFS as the time from the date of randomization to first objective evidence of radiographic progression as assessed in soft tissue per RECIST 1.1, or death, whichever occurs first. Soft tissue disease status was assessed at regular intervals using computed tomography of the chest and computed tomography or magnetic resonance imaging of the abdomen and pelvis.

Important endpoint distinction: rPFS is a time-to-event endpoint. A participant contributes follow-up until radiographic progression, death, or censoring under the applicable analysis rules. The analysis therefore incorporates both the timing of events and information from participants who have not yet experienced an event at the end of their observed follow-up.

5. Statistical Methodology

Log-rank testing

The reported method for both primary rPFS analyses is the log-rank test. The log-rank framework compares the observed and expected numbers of events between randomized treatment groups over the observed follow-up period.

Conceptual comparison
Observed events  vs.  expected events under the null hypothesis

For a time-to-event comparison, the relevant information is not merely whether progression occurred. The timing of progression or death and the pattern of censoring across follow-up contribute to the statistical comparison.

Cox proportional-hazards model

The registry's analysis notes state that the hazard ratio was based on a Cox proportional-hazards model. The hazard ratio is therefore a model-based relative measure of the event hazard for talazoparib plus enzalutamide compared with placebo plus enzalutamide.

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

For a binary treatment indicator, the hazard ratio is represented by exp(β). An HR below 1 indicates a lower estimated instantaneous event hazard in the talazoparib-containing group relative to the comparator under the fitted model.

Superiority hypotheses

The reported hypothesis type is superiority. For the all-comers rPFS analysis, the registry text states the hypothesis as H01: HRrPFS ≥ 1 versus H11: HRrPFS < 1. The corresponding DDR-deficiency analysis uses the same directional superiority logic for its hazard ratio.

This matters because an HR below 1 is not merely being described; the prespecified statistical question was whether the treatment hazard ratio was below the null value of 1.

Analysis populations

The all-comers analysis included all participants randomized to double-blind study treatment in Part 2 Cohort 1 regardless of whether or not treatment was administered. The DDR-deficiency analysis included DDR-deficient participants randomized to double-blind study treatment in Part 2 regardless of whether or not treatment was administered.

AnalysisPopulationComparisonMethod
All-comers rPFS All participants randomized to double-blind study treatment in Part 2 Cohort 1 Talazoparib + enzalutamide vs placebo + enzalutamide Log-rank test; HR from Cox proportional-hazards model
DDR-deficient rPFS DDR-deficient participants randomized to double-blind study treatment in Part 2 Talazoparib + enzalutamide vs placebo + enzalutamide Log-rank test; HR from Cox proportional-hazards model

6. Results: All-Comers rPFS

The first formal primary analysis concerns BICR-assessed radiographic progression-free survival per RECIST 1.1 for all-comers in Part 2 Cohort 1. The registry reports a two-sided 95% confidence interval and a superiority hypothesis.

Hazard ratio for radiographic progression or death

0.627

95% CI: 0.506–0.777   ·   P < 0.0001

BICR-assessed rPFS, Part 2 Cohort 1 all-comers population

MeasureReported result
EndpointBICR Assessed Radiographic Progression-Free Survival Per RECIST 1.1 for All-Comers - Part 2 Cohort 1
Analysis populationAll participants randomized to double-blind study treatment in Part 2 Cohort 1 regardless of whether or not treatment was administered
ComparisonTalazoparib + enzalutamide vs placebo + enzalutamide
MethodLog-rank test
Effect measureHazard ratio
Estimate0.627
95% CI0.506–0.777
P-value<0.0001
HypothesisSuperiority
Maximum reported follow-up windowUp to 42 months
Clinical Biostats interpretation

An HR of 0.627 means that, under the Cox proportional-hazards model used for the analysis, the estimated instantaneous hazard of radiographic progression or death was 0.627 times the corresponding hazard in the placebo-plus-enzalutamide group. Expressed as a relative complement, 1 − 0.627 = 0.373, so the estimated hazard was approximately 37.3% lower in relative terms.

The HR does not mean that 62.7% of patients remained progression-free, that 37.3% of patients were cured, or that every individual patient experienced a 37.3% reduction in risk. It is a relative time-to-event measure derived from the model.

The 95% CI of 0.506–0.777 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of outcomes that individual patients can experience. Because the entire interval is below 1, the interval is consistent with a lower estimated hazard in the talazoparib-containing group relative to the comparator.

The P < 0.0001 result addresses the statistical evidence against the reported null hypothesis. A p-value does not measure the magnitude of the treatment effect and should not be read as a probability that the treatment effect is real or as a measure of clinical importance.

Interpretation also depends on the proportional-hazards model used for the HR. A single HR is most naturally interpreted when the relative hazards are reasonably summarized by a common hazard ratio over follow-up. The ClinicalTrials.gov record does not provide enough information to independently assess the proportional-hazards assumption, nor do they provide a reconstructed Kaplan-Meier curve or median rPFS for this analysis.

7. Results: rPFS in Patients With DDR Deficiencies

The second formal primary analysis concerns BICR-assessed rPFS per RECIST 1.1 among participants with DDR deficiencies in Part 2. This is a separate analysis population and should be interpreted independently from the all-comers result.

Hazard ratio for radiographic progression or death

0.447

95% CI: 0.328–0.610   ·   P < 0.0001

BICR-assessed rPFS in participants with DDR deficiencies

MeasureReported result
EndpointBICR Assessed rPFS Per RECIST 1.1 in Patients With DDR Deficiencies - Part 2
Analysis populationDDR-deficient participants randomized to double-blind study treatment in Part 2 regardless of whether or not treatment was administered
ComparisonTalazoparib + enzalutamide vs placebo + enzalutamide
MethodLog-rank test
Effect measureHazard ratio
Estimate0.447
95% CI0.328–0.610
P-value<0.0001
HypothesisSuperiority
Maximum reported follow-up windowUp to 38 months
Clinical Biostats interpretation

An HR of 0.447 means that, under the fitted Cox model, the estimated instantaneous hazard of radiographic progression or death in the DDR-deficient analysis population was 0.447 times that of the comparator group. In relative-complement terms, 1 − 0.447 = 0.553, corresponding to an estimated hazard approximately 55.3% lower relative to the comparator.

Again, the HR is not a percentage of patients benefiting. It does not mean that 55.3% of patients avoided progression, nor does it translate directly into a median difference or an absolute increase in the probability of remaining progression-free.

The two-sided 95% CI of 0.328–0.610 quantifies uncertainty around the estimated hazard ratio. The interval remains below 1, so the reported uncertainty interval is consistent with a lower estimated event hazard for talazoparib plus enzalutamide under the model.

The P < 0.0001 value provides evidence against the reported superiority null hypothesis. It should not be interpreted as an effect-size metric. A very small p-value can coexist with an effect that has uncertain clinical magnitude, while a p-value itself does not tell us the absolute number of progression-free months gained.

The DDR-deficiency result also should not be treated as if it were simply another measurement of the all-comers HR. It comes from a distinct analysis population. Any comparison between the two HRs would require an appropriate formal assessment of treatment-effect heterogeneity; the ClinicalTrials.gov record does not report such an interaction analysis.

8. Comparing the Two Reported rPFS Estimates

The two formal estimates point in the same directional relationship, but they answer questions in different analysis populations. The all-comers estimate is 0.627, while the DDR-deficient estimate is 0.447.

PopulationHR95% CIP-valueMaximum time frame
All-comers, Part 2 Cohort 1 0.627 0.506–0.777 <0.0001 Up to 42 months
DDR deficiencies, Part 2 0.447 0.328–0.610 <0.0001 Up to 38 months

The numerical difference between the point estimates should not automatically be interpreted as proof that DDR deficiency modifies the treatment effect. The correct statistical question for effect modification is whether the treatment effect differs between prespecified populations beyond what would be expected from sampling variation. That requires an appropriate interaction or heterogeneity analysis.

Do not compare HRs by eye alone. The DDR-deficient HR of 0.447 is numerically smaller than the all-comers HR of 0.627, but separate confidence intervals do not constitute a formal test that the treatment effects differ. The registry-reported TALAPRO-2 data do not report an interaction test, so the two estimates should be presented as population-specific results rather than as evidence of a statistically demonstrated difference between populations.

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

Relative effect

A hazard ratio compares estimated instantaneous event hazards between two groups over time. For TALAPRO-2, both reported HRs are below 1, meaning the fitted model estimates a lower hazard of radiographic progression or death for the talazoparib-plus-enzalutamide group relative to placebo plus enzalutamide in the corresponding analysis population.

Not an absolute risk measure

An HR does not tell us the absolute probability of progression by a particular date. Absolute survival or progression-free estimates require the underlying time-to-event distribution, typically summarized using a Kaplan-Meier curve or specific time-point estimates. Those values are not included in the ClinicalTrials.gov record.

Not a median difference

An HR of 0.627 does not imply that one treatment adds a fixed number of months to rPFS. Likewise, an HR of 0.447 cannot be converted directly into a median rPFS difference. Median survival or progression-free time must be obtained from the estimated time-to-event distribution itself.

Not a probability of causation

The p-value of <0.0001 is evidence against the reported null hypothesis within the statistical framework. It is not the probability that the null hypothesis is true, the probability that the treatment works, or a measure of the clinical importance of the effect.

10. Kaplan-Meier Estimation and Censoring

Radiographic progression-free survival is a time-to-event endpoint, so a natural descriptive tool is the Kaplan-Meier estimator. The basic idea is to estimate the probability of remaining event-free as follow-up progresses, updating the estimate at observed event times while retaining participants who remain under observation without an event.

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

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

For TALAPRO-2 rPFS, the event is the first documented radiographic progression or death, whichever occurs first. This definition makes death part of the endpoint rather than treating it as an ordinary loss to follow-up.

Censoring is equally important. A participant who has not experienced progression or death by the end of observable follow-up can contribute information up to the censoring point. The validity of a Kaplan-Meier analysis depends on the censoring process and the assumptions underlying the analysis rules. The ClinicalTrials.gov record does not provide the detailed censoring rules or participant-level event data needed to reconstruct the curves independently.

Educational note: the page intentionally does not draw a fabricated Kaplan-Meier curve from the reported HRs and confidence intervals. A valid Kaplan-Meier reconstruction requires event and censoring information or sufficiently detailed digitized source data.

11. Why a Log-Rank Test Was Used

The log-rank test is designed for comparing time-to-event distributions between groups. Unlike a simple binary comparison, it incorporates the timing of events throughout follow-up. That is appropriate for rPFS because progression or death can occur at different times for different participants.

What it tests

The log-rank test evaluates whether the observed pattern of event occurrence over time differs between the randomized treatment groups under the null hypothesis.

What it does not estimate

The log-rank test itself does not produce the hazard ratio. The reported HR comes from the Cox proportional-hazards model.

Why time matters

A participant progressing early contributes a different amount of information from a participant remaining progression-free for a long period.

Why randomization matters

The randomized treatment assignment provides the basis for comparing the time-to-event experience of the two groups.

The combination of a log-rank comparison and a Cox-derived HR is therefore complementary: the log-rank test addresses the evidence for a difference in the time-to-event distributions, while the HR provides a relative effect estimate with a confidence interval.

12. Statistical Methods Explained

Why was a log-rank test used?

Because the primary efficacy outcomes reported here are time-to-event endpoints. The log-rank test uses information about when events occur rather than reducing each participant to a simple yes/no status at a fixed date. This makes it suitable for comparing rPFS between randomized groups when follow-up times differ.

What does an HR of 0.627 mean?

It means the estimated instantaneous hazard of radiographic progression or death was 0.627 times the comparator hazard under the fitted Cox model. Equivalently, the model-based relative complement is 37.3%. It does not mean that exactly 37.3% of patients benefited or that every patient had a 37.3% reduction in individual risk.

What does an HR of 0.447 mean?

For the DDR-deficient analysis population, it means the estimated instantaneous hazard was 0.447 times that of the comparator under the Cox model. Its relative complement is 55.3%. The number is population-specific and should not be treated as a direct estimate of the treatment effect in every participant enrolled in the trial.

Why does the confidence interval matter?

A point estimate alone does not communicate sampling uncertainty. The 95% confidence intervals of 0.506–0.777 and 0.328–0.610 show the statistical precision associated with the two HR estimates under their respective analysis frameworks. They do not describe the range of individual patient outcomes.

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

The p-value measures the compatibility of the observed data with the specified null hypothesis within the test's assumptions. It is influenced by both effect magnitude and the amount of information in the analysis. The HR and its confidence interval are therefore needed to describe the estimated treatment effect and its uncertainty.

Why must the two HRs not be treated as a subgroup interaction test?

The all-comers and DDR-deficient analyses have different analysis populations. A smaller HR in the DDR-deficient population does not by itself establish treatment-effect modification. A formal interaction or heterogeneity analysis would be needed to support that conclusion, and no such analysis is reported in the ClinicalTrials.gov record.

Why is censoring important?

Participants may reach the end of observable follow-up without experiencing progression or death. Their partial follow-up still contributes information. Time-to-event methods such as Kaplan-Meier estimation, log-rank testing, and Cox modeling are designed to use this information while accounting for differing observation times.

13. Safety Endpoints and Reported Serious Adverse Events

The registry lists six Part 1 primary endpoints concerning treatment-emergent adverse events and clustered cytopenias. These are binary participant-level outcomes defined over specified periods. The ClinicalTrials.gov record does not contain formal statistical analyses for those six endpoints, so no treatment-effect estimate or p-value is reported for them here.

The ClinicalTrials.gov record does contain affected/at-risk counts for serious adverse events by several Part 1 and Part 2 groups. These counts should be kept distinct because they refer to different parts, cohorts, populations, and denominators.

GroupSerious adverse events affected / at risk
Part 1: Talazoparib 1 mg QD + Enzalutamide6/13
Part 1: Talazoparib 0.5 mg QD + Enzalutamide3/6
Part 2 Cohort 1: Talazoparib + Enzalutamide157/398
Part 2 Cohort 1: Placebo + Enzalutamide107/401
Part 2 Participants With DDR Deficiencies: Talazoparib + Enzalutamide60/198
Part 2 Participants With DDR Deficiencies: Placebo + Enzalutamide40/199

These counts are descriptive safety data, not the same statistical object as the primary rPFS hazard ratios. In particular, the denominators differ between Part 1 and Part 2 groups, so the figures should not be pooled or compared as though they came from one common analysis population.

Safety interpretation: the ClinicalTrials.gov record reports serious adverse-event counts as affected participants divided by participants at risk. They do not supply a formal statistical comparison, confidence interval, or p-value for these safety counts. Accordingly, the appropriate interpretation here is descriptive rather than a formal comparative inference.

14. The Six Part 1 Safety Primary Endpoints

Six registered primary endpoints are binary safety outcomes from Part 1. Their definitions and time frames are important because a binary adverse-event endpoint depends not only on the event definition but also on the observation window.

EndpointTime frame
Treatment-emergent adverse events occurring within the first 66 days of dosing Post dose on Day 1 up to Day 66 in Part 1
All-causality clustered treatment-emergent cytopenias by preferred term and maximum CTCAE grade within the first 66 days Post dose on Day 1 up to Day 66 in Part 1
All-causality treatment-emergent adverse events during the overall period Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy
Treatment-related treatment-emergent adverse events during the overall period Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy
All-causality clustered treatment-emergent cytopenias by preferred term and maximum CTCAE grade anytime after dosing Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy
Treatment-related clustered treatment-emergent cytopenias by preferred term and maximum CTCAE grade in >=10% of participants anytime after dosing Post dose on Day 1 up to 28 days after the last dose of study intervention, or before new systemic antineoplastic therapy

The registry definitions posted on ClinicalTrials.gov for these endpoints describe an adverse event as any untoward medical occurrence in a participant who received study intervention without regard to the possibility of a causal relationship. TEAEs are newly occurring adverse events or events that worsen after first dose. The registry also identifies CTCAE version 4 grading for the safety definitions.

Because these endpoints are binary, a conventional comparative analysis would generally involve a comparison of event proportions or an appropriate regression model for binary outcomes. However, the registry-reported TALAPRO-2 data do not provide formal statistical analyses for these six endpoints. No comparative estimate should therefore be inferred from the endpoint definitions alone.

15. Multiplicity and the Two Reported Primary rPFS Analyses

The ClinicalTrials.gov record identifies two formal primary endpoint analyses, both testing superiority with hazard ratios below the null value of 1 as the alternative direction. They concern related but distinct populations: all-comers in Part 2 Cohort 1 and participants with DDR deficiencies in Part 2.

AnalysisRoleNull hypothesis reportedAlternative direction
All-comers rPFS Primary H01: HRrPFS ≥ 1 H11: HRrPFS < 1
DDR-deficient rPFS Primary H02: HRrPFS+ ≥ 1 H12: HRrPFS+ < 1

Multiplicity is important whenever more than one formal hypothesis is tested. The existence of two primary analyses means that their relationship within the trial's overall type-I-error framework matters for confirmatory interpretation. The ClinicalTrials.gov record identifies the hypotheses and results but do not provide the complete alpha-allocation or multiplicity procedure.

Interpretation boundary: the reported P < 0.0001 values are reproduced exactly as reported in the registry. This page does not infer an additional multiplicity adjustment, alpha allocation, or sequential testing procedure that is not present in the registry-reported TALAPRO-2 data.

16. Blinding and Its Statistical Role

TALAPRO-2 is registered as quadruple-masked. The reported primary efficacy endpoint is also assessed by blinded independent central review. These design features are statistically relevant because knowledge of treatment assignment can influence outcome assessment, particularly when assessment requires interpretation of radiographic findings.

Randomization

Creates the treatment comparison before outcome information is observed and provides the foundation for the causal comparison.

Quadruple masking

Reduces the opportunity for treatment knowledge to influence conduct or assessment within the masked trial framework.

Independent central review

Provides a centralized assessment framework for the radiographic endpoint rather than relying solely on local investigator assessment.

RECIST 1.1

Provides the stated framework for defining objective radiographic progression in the reported rPFS endpoint.

These features do not eliminate every source of uncertainty. Statistical inference still depends on the analysis population, follow-up, censoring, endpoint definitions, model assumptions, and the prespecified testing framework.

17. Results Posted vs. Statistical Analyses Posted

The ClinicalTrials.gov record indicates that results are posted, with eight outcome measures posted and two statistical analyses posted. The two formal analyses correspond to the two Part 2 rPFS endpoints described above.

Registry elementSupplied value
Primary endpoints registered8
Outcome measures posted8
Statistical analyses posted2
Primary-endpoint analyses2
Primary analyses with estimate + CI2
Effect measureHazard ratio
Hypothesis typeSuperiority

This distinction is important for a statistical-results page. A posted outcome measure is not automatically equivalent to a posted inferential comparison. For the six Part 1 safety endpoints, the ClinicalTrials.gov record identifies the outcomes and their time frames but do not supply formal comparative estimates, confidence intervals, or p-values.

18. Timeline

2017-12-18 · Trial start

TALAPRO-2 begins

The registry lists 2017-12-18 as the trial start date.

Part 1 · Safety assessment

Early and overall safety endpoints

Part 1 includes safety endpoints covering the first 66 days and the overall period through 28 days after the last dose or before new systemic antineoplastic therapy, depending on the endpoint.

Part 2 · Primary efficacy analysis

BICR-assessed rPFS

The registry reports formal log-rank analyses and Cox-model hazard ratios for all-comers and participants with DDR deficiencies.

2022-10-03 · Primary completion

Registry primary completion

the ClinicalTrials.gov record lists 2022-10-03 as the primary completion date.

19. Important Limitations and Interpretation Issues

20. Why This Trial Matters Statistically

TALAPRO-2 is a useful statistical teaching case because it connects randomized trial design, masked outcome assessment, time-to-event endpoints, log-rank testing, Cox regression, confidence intervals, and population-specific treatment-effect estimates within one study.

ConceptHow it appears in TALAPRO-2
RandomizationThe trial uses randomized allocation.
Parallel-group designThe registry identifies a parallel design with two arms.
Quadruple maskingThe trial is registered as quadruple-masked.
Time-to-event endpointBICR-assessed rPFS is defined by time to progression or death.
RECIST 1.1Radiographic progression is assessed per RECIST 1.1.
Independent reviewThe primary rPFS endpoint uses blinded independent central review.
Kaplan-Meier frameworkrPFS is a time-to-event endpoint for which Kaplan-Meier estimation is a natural descriptive approach.
Log-rank testThe registry-reported formal method for both rPFS analyses.
Hazard ratioThe reported effect measure for both formal rPFS analyses.
Cox modelThe registry-reported analysis notes state that the HR was based on a Cox proportional-hazards model.
Confidence intervalBoth primary rPFS analyses provide two-sided 95% confidence intervals.
Superiority testingBoth formal analyses test a directional superiority hypothesis with HR below 1.
Population-specific analysisSeparate results are reported for all-comers and DDR-deficient participants.
Safety endpoint analysisSix registered Part 1 primary endpoints are binary safety outcomes, although formal comparative analyses are not reported here.

21. A Statistical Reading of the TALAPRO-2 Results

The most important statistical feature of the reported efficacy results is the distinction between effect magnitude, statistical uncertainty, and absolute clinical outcome.

Effect magnitude

The all-comers HR of 0.627 and the DDR-deficient HR of 0.447 are relative measures. Both are below the null value of 1, and the reported estimates therefore describe lower modeled event hazards for talazoparib plus enzalutamide than for placebo plus enzalutamide in their respective analysis populations.

Statistical uncertainty

The corresponding 95% confidence intervals quantify uncertainty around each HR. The all-comers interval is 0.506–0.777, while the DDR-deficient interval is 0.328–0.610. Neither interval crosses 1.

Evidence against the null

Both reported p-values are <0.0001. Under the stated superiority framework, these results provide strong statistical evidence against the corresponding null hypotheses. The p-values do not, however, replace the HR and its confidence interval when communicating effect magnitude.

What is not available from the ClinicalTrials.gov record

The ClinicalTrials.gov record does not provide median rPFS, Kaplan-Meier time-point estimates, numbers of progression events, numbers censored, or participant-level event times. Consequently, the HR should not be transformed into an invented median difference or absolute progression-free survival percentage.

A useful mental model

Think of the HR as answering "How does the event hazard compare over time?", the confidence interval as answering "How precisely has that relative effect been estimated?", and the p-value as answering "How incompatible are the observed data with the specified null hypothesis?". None of these alone answers "How many additional months did an individual patient gain?"

22. Interpreting the DDR-Deficient Analysis Carefully

The DDR-deficient analysis is particularly useful for illustrating why a subgroup estimate must be interpreted according to its population definition. The reported HR of 0.447 is estimated specifically among DDR-deficient participants randomized to double-blind study treatment in Part 2.

The result cannot be generalized automatically to every participant in the trial. Conversely, it would also be inappropriate to infer that the difference between 0.447 and 0.627 proves that DDR deficiency changes the treatment effect. The statistical comparison required for that claim is an interaction analysis or another formal test of heterogeneity.

What the result supports

A lower estimated hazard of rPFS or death for the talazoparib-containing group within the reported DDR-deficient analysis population.

What the result does not establish

That the treatment effect is statistically different in DDR-deficient versus non-DDR-deficient participants.

What the CI contributes

The 0.328–0.610 interval communicates uncertainty around the DDR-deficient HR.

What is missing

The ClinicalTrials.gov record does not report a formal interaction or treatment-effect heterogeneity analysis.

23. Why This Is a Time-to-Event Trial, Not Just a Binary Trial

Several TALAPRO-2 endpoints are binary safety outcomes, but the principal formal efficacy analyses are time-to-event outcomes. That distinction changes the statistical machinery.

FeatureBinary safety endpointrPFS endpoint
Basic outcomeWhether an event occurred within a defined periodTime until progression or death
Time informationReduced to the specified observation windowCentral to the endpoint
Typical descriptive measureEvent count / proportionKaplan-Meier survival function
Reported TALAPRO-2 formal methodNone posted on ClinicalTrials.gov for the six Part 1 primary safety endpointsLog-rank test
Reported effect measureNone posted on ClinicalTrials.gov for those six endpointsHazard ratio
Model reportedNone reportedCox proportional-hazards model

This distinction is more than terminology. If two patients both eventually progress, a binary endpoint records the same outcome for both, while a time-to-event analysis retains information about whether progression occurred early or much later.

24. Related Tutorials

Learn more about the methods used in this trial:

25. Related Calculators

26. Sources

Continue through the Clinical Biostats statistical methods

Connect clinical-trial endpoints to focused tutorials and statistical calculators for deeper study of survival analysis, inference, and trial design.

27. Record Summary

TALAPRO-2 provides a clear example of how a randomized phase 3 oncology trial can combine masked treatment assignment, independent radiographic assessment, a time-to-event primary efficacy endpoint, log-rank testing, and Cox-model hazard ratios. The ClinicalTrials.gov record reports an HR of 0.627 (95% CI 0.506–0.777; P < 0.0001) for BICR-assessed rPFS in the all-comers Part 2 Cohort 1 population and an HR of 0.447 (95% CI 0.328–0.610; P < 0.0001) in participants with DDR deficiencies.

The statistically appropriate interpretation is to keep those estimates tied to their respective analysis populations and to distinguish relative hazard from absolute clinical outcome. The HR describes a modeled relative event hazard; the confidence interval describes uncertainty around that estimate; and the p-value addresses evidence against the specified superiority null hypothesis. None of those quantities by itself supplies a median rPFS, a time-specific progression-free probability, or an individual patient's treatment benefit.

The safety portion of the registry contains six Part 1 primary binary endpoints and serious-adverse-event counts across several trial populations. Because the ClinicalTrials.gov record does not provide formal comparative statistical analyses for those six safety endpoints, the safety information is presented descriptively rather than supplemented with inferred p-values or effect estimates.

Clinical Biostats methodology: A trial-results page should distinguish the reported statistical evidence from educational interpretation. For TALAPRO-2, that means preserving the registry's endpoint definitions, analysis populations, HR estimates, confidence intervals, p-values, and stated methods while avoiding unsupported reconstruction of median outcomes, Kaplan-Meier curves, subgroup interactions, or multiplicity procedures.