This page separates reported trial results from statistical interpretation. All numerical trial results presented here are taken from the registry-reported TRITON3 trial data. The registry provides the official trial record.
1. Trial at a Glance
TRITON3 was a randomized, open-label, parallel-group phase 3 trial evaluating rucaparib versus physician's choice of abiraterone acetate, enzalutamide, or docetaxel in participants with metastatic castration-resistant prostate cancer and homologous recombination gene deficiency.
| Feature | TRITON3 |
|---|---|
| Phase | Phase 3 |
| Condition | Metastatic Castration Resistant Prostate Cancer |
| Trial name | TRITON3 |
| Design | Randomized, parallel-group, open-label |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 405 |
| Primary endpoints | Radiographic progression-free survival by independent radiology review in participants with a BRCA alteration; and in participants with a BRCA or ATM alteration combined |
| Primary endpoint type | Time-to-event |
| Primary hypothesis | Superiority |
| Statistical method | Log-rank test |
| Effect measure | Hazard ratio |
| Study status | Completed |
| Study start | 2017-06-13 |
| Primary completion | 2022-08-25 |
| Lead sponsor | pharmaand GmbH |
| Sponsor type | Industry |
2. Clinical Question
The central statistical question was whether rucaparib would improve radiographic progression-free survival compared with physician's choice of abiraterone acetate, enzalutamide, or docetaxel in participants with metastatic castration-resistant prostate cancer and homologous recombination gene deficiency.
Population
Participants with metastatic castration-resistant prostate cancer and homologous recombination gene deficiency. The primary analyses separately identify participants with BRCA alterations and participants with BRCA or ATM alterations combined.
Intervention
Rucaparib.
Comparator
Physician's choice of abiraterone acetate, enzalutamide, or docetaxel.
Primary question
Does rucaparib produce a superior radiographic progression-free survival outcome relative to physician's choice of therapy?
3. Trial Design
Rucaparib
- Rucaparib was the experimental intervention.
- Primary efficacy analyses compared this arm with physician's choice of therapy.
- The ClinicalTrials.gov record identifies 81 serious-adverse-event-affected participants among 270 at risk during the treatment phase.
Physician's choice
- Abiraterone acetate, enzalutamide, or docetaxel.
- The choice of comparator therapy was made by the physician.
- The ClinicalTrials.gov record identifies 36 serious-adverse-event-affected participants among 130 at risk in this comparator group.
4. Analysis Populations
The registry-posted primary analyses were conducted in intention-to-treat populations. For the BRCA-specific analysis, the population was the ITT Population with BRCA mutated mCRPC. For the combined analysis, the population included all randomized participants with BRCA mutated mCRPC and participants with ATM mutated mCRPC.
| Analysis | Population | Role |
|---|---|---|
| BRCA rPFS | ITT Population with BRCA mutated mCRPC | Primary endpoint analysis |
| BRCA + ATM rPFS | ITT Population including all randomized participants with BRCA mutated mCRPC and participants with ATM mutated mCRPC | Primary endpoint analysis |
| BRCA OS | ITT Population with BRCA mutated mCRPC | Secondary endpoint analysis |
| BRCA + ATM OS | ITT Population including all randomized participants with BRCA mutated mCRPC and participants with ATM mutated mCRPC | Secondary endpoint analysis |
The ITT principle is particularly important here because the comparison is anchored to randomization rather than subsequent treatment exposure. In a randomized trial, analyzing participants according to their randomized assignment helps preserve the comparability created by randomization.
5. Primary Endpoints
| Endpoint | Registry time frame | Type | Primary analysis |
|---|---|---|---|
| Radiographic Progression-free Survival (rPFS) by IRR in Participants With a BRCA Alteration | From enrollment to primary completion of study (Total follow-up was up to approximately 4 years) | Time-to-event | Log-rank test; hazard ratio |
| Radiographic Progression-free Survival (rPFS) by IRR in Participants With a BRCA or ATM Alteration Combined | From enrollment to primary completion of study (Total follow-up was up to approximately 4 years) | Time-to-event | Log-rank test; hazard ratio |
Radiographic progression-free survival definition
The registry defines the primary efficacy endpoint as rPFS by independent radiology review, with time measured from randomization to the first objective evidence of radiographic progression, or death due to any cause, whichever occurs first.
The registry definition further states that radiographic disease progression includes confirmed soft tissue disease progression and confirmed bone disease progression according to modified RECIST Version 1.1.
6. Statistical Methodology
Kaplan-Meier estimation
Radiographic progression-free survival and overall survival are time-to-event outcomes. Kaplan-Meier estimation is the standard descriptive framework for estimating the event-free probability over time while retaining information from participants who are censored before experiencing the event.
where di is the number of events at time ti and ni is the number at risk immediately before that time.
The registry-reported TRITON3 data identify the primary endpoints as time-to-event outcomes and report hazard ratios as the effect measure. The ClinicalTrials.gov record does not report Kaplan-Meier median estimates or survival probabilities at particular time points, so those quantities are not added to this analysis.
Log-rank test
The reported formal comparison for each of the four posted statistical analyses was the log-rank test. The log-rank test compares the observed and expected numbers of events between treatment groups across the observed follow-up.
The log-rank framework asks whether the pattern of event occurrence over follow-up is consistent with the null hypothesis of no difference between the randomized groups.
Hazard ratio
The effect measure reported for every posted analysis was the hazard ratio. A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the rucaparib group relative to physician's choice under the time-to-event model.
The hazard ratio is a relative time-to-event measure. It is not a probability, an absolute risk difference, or the proportion of participants who benefited.
Intention-to-treat analysis
The primary and secondary efficacy analyses reported in the ClinicalTrials.gov record uses ITT populations. This means the efficacy comparison is tied to randomized assignment rather than selectively redefining groups according to treatment actually received after randomization.
Superiority hypothesis
The registered hypothesis type for the posted primary analyses was superiority. Thus, the statistical question was whether the rucaparib treatment group differed favorably from physician's choice rather than whether the two strategies were sufficiently similar to satisfy a non-inferiority margin.
7. Primary Results: BRCA Alteration
The first primary endpoint was radiographic progression-free survival by independent radiology review in participants with a BRCA alteration. The analysis population was the ITT population with BRCA mutated metastatic castration-resistant prostate cancer.
rPFS hazard ratio
95% CI: 0.36–0.69 · P < 0.001
Log-rank test · Superiority hypothesis
An HR of 0.50 means that, under the reported time-to-event comparison, the estimated instantaneous rate of radiographic progression or death was one-half as high in the rucaparib group as in the physician's-choice group. Described another way, an HR of 0.50 corresponds to a 50% lower estimated hazard relative to the comparator.
The HR does not mean that 50% of participants avoided progression, that every participant had a 50% reduction in risk, or that median progression-free survival differed by a fixed percentage. Those interpretations would require absolute survival estimates or other measures that are not reported in the ClinicalTrials.gov record.
The 95% CI of 0.36–0.69 describes statistical uncertainty around the estimated hazard ratio. It does not describe the range of effects experienced by individual participants. Because the interval remains below 1, its values are consistent with a lower estimated hazard for rucaparib across the interval represented by the reported analysis.
The P < 0.001 result addresses the evidence against the null hypothesis under the reported log-rank testing framework. A p-value does not measure the size or clinical importance of an effect. Effect magnitude is conveyed by the HR, while the confidence interval conveys precision.
As with any hazard-ratio analysis, interpretation of a single HR should not be expanded into an assumption that the relative hazard is identical at every point in time unless the underlying time-to-event model supports that interpretation. The ClinicalTrials.gov record does not provide sufficient information to independently evaluate proportional hazards.
What the BRCA result establishes statistically
The posted analysis reports a hazard ratio below 1, a two-sided 95% confidence interval entirely below 1, and a p-value below 0.001 for the superiority comparison. These three quantities answer related but different questions: the HR describes the estimated relative effect, the confidence interval describes uncertainty around that estimate, and the p-value describes evidence against the null hypothesis.
8. Primary Results: BRCA or ATM Alterations Combined
The second primary endpoint broadened the analysis population to participants with a BRCA or ATM alteration combined. The registry-reported analysis population includes all randomized participants with BRCA mutated metastatic castration-resistant prostate cancer and participants with ATM mutated metastatic castration-resistant prostate cancer.
rPFS hazard ratio
95% CI: 0.47–0.80 · P < 0.001
Log-rank test · Superiority hypothesis
An HR of 0.61 means that the estimated instantaneous rate of radiographic progression or death was approximately 61% of the corresponding rate in the physician's-choice group under the reported time-to-event analysis. Equivalently, an HR of 0.61 corresponds to a 39% lower estimated hazard relative to the comparator.
The HR is not a 39% absolute reduction in the probability of progression or death, and it does not mean that 39% of participants benefited. It is a relative measure of event rate over the analyzed follow-up.
The 95% CI of 0.47–0.80 provides the reported uncertainty interval around the estimated HR. Because the entire interval is below 1, the interval is consistent with a lower estimated hazard for rucaparib throughout the range represented by the confidence interval.
The P < 0.001 result is evidence against the null hypothesis used for the superiority comparison. It should not be interpreted as a 0.1% probability that the observed treatment effect occurred by chance, nor as a measure of the magnitude of treatment benefit.
The combined BRCA-or-ATM analysis also illustrates why the analysis population must always be identified. An effect estimate in a combined molecular population is not automatically interchangeable with the estimate from the BRCA-only population.
| Primary endpoint | Analysis population | HR | 95% CI | P-value |
|---|---|---|---|---|
| rPFS by IRR in participants with a BRCA alteration | ITT population with BRCA mutated mCRPC | 0.50 | 0.36–0.69 | <0.001 |
| rPFS by IRR in participants with a BRCA or ATM alteration combined | ITT population including BRCA mutated and ATM mutated mCRPC participants | 0.61 | 0.47–0.80 | <0.001 |
9. Secondary Endpoint Results: Overall Survival
Overall survival was reported as a secondary time-to-event endpoint. The same general statistical framework was used for both the BRCA-specific and BRCA-or-ATM combined populations: ITT analysis, log-rank testing, and hazard-ratio estimation.
Overall Survival in Participants With a BRCA Alteration
OS hazard ratio
95% CI: 0.68–1.20 · P = 0.5044
Log-rank test · Superiority hypothesis
The BRCA-specific OS estimate of 0.91 is much closer to 1 than the corresponding primary rPFS estimate of 0.50. The registry-reported confidence interval extends on both sides of 1, from 0.68 to 1.20, and the reported p-value is 0.5044.
Overall Survival in Participants With a BRCA or ATM Alteration Combined
OS hazard ratio
95% CI: 0.78–1.26 · P = 0.9368
Log-rank test · Superiority hypothesis
The combined-population OS estimate was 0.99, which is very close to the null value of 1. The 95% confidence interval extends from 0.78 to 1.26, and the reported p-value is 0.9368.
| Secondary endpoint | Analysis population | HR | 95% CI | P-value |
|---|---|---|---|---|
| Overall Survival in Participants With a BRCA Alteration | ITT Population with BRCA mutated mCRPC | 0.91 | 0.68–1.20 | 0.5044 |
| Overall Survival in Participants With a BRCA or ATM Alteration Combined | ITT Population including BRCA mutated and ATM mutated mCRPC participants | 0.99 | 0.78–1.26 | 0.9368 |
10. Comparing the rPFS and OS Results
The four posted analyses provide an instructive contrast between two time-to-event endpoints. The primary rPFS analyses reported HRs of 0.50 and 0.61, whereas the secondary OS analyses reported HRs of 0.91 and 0.99.
| Endpoint family | BRCA | BRCA + ATM |
|---|---|---|
| rPFS HR | 0.50 | 0.61 |
| rPFS 95% CI | 0.36–0.69 | 0.47–0.80 |
| rPFS P-value | <0.001 | <0.001 |
| OS HR | 0.91 | 0.99 |
| OS 95% CI | 0.68–1.20 | 0.78–1.26 |
| OS P-value | 0.5044 | 0.9368 |
Statistically, these results should be treated as separate endpoint analyses rather than as contradictory measurements of one quantity. rPFS measures time until radiographic progression or death, while OS measures time until death from any cause. The endpoints can therefore produce different treatment-effect estimates because they capture different stages of the clinical course.
The ClinicalTrials.gov record does not provide median rPFS, median OS, Kaplan-Meier survival probabilities, event counts for the efficacy analyses, or subgroup estimates. Those quantities are therefore not introduced here.
11. Statistical Methods Explained
Why was a log-rank test used?
Both primary endpoints and both secondary endpoints are time-to-event outcomes. A log-rank test is designed to compare event-time distributions between randomized groups while incorporating the timing of events and allowing for right-censoring. It is therefore aligned with the structure of rPFS and OS data.
What does an HR of 0.50 mean?
An HR of 0.50 means the estimated instantaneous event rate in the rucaparib group was one-half that in the physician's-choice group under the reported analysis. It does not mean that half as many participants necessarily experienced the endpoint, because a hazard ratio is not a simple ratio of cumulative event probabilities.
Why is the confidence interval important?
The confidence interval communicates precision. For the BRCA rPFS analysis, the HR was 0.50 with a 95% CI of 0.36–0.69. The interval indicates that the point estimate should not be treated as exact. The wider the interval, the greater the statistical uncertainty around the point estimate.
Why doesn't the p-value measure treatment effect size?
The p-value addresses compatibility with the null hypothesis under the specified statistical test. It does not quantify how large the treatment effect is. Two analyses can have similar p-values but different effect sizes, or similar effect estimates but different p-values because of differences in statistical precision.
Why does the ITT population matter?
Randomization creates the basis for comparing treatment groups. An ITT analysis retains participants according to their randomized assignment, helping preserve that basis for inference. In TRITON3, the posted efficacy analyses explicitly identify ITT populations for both the BRCA-specific and BRCA-or-ATM analyses.
Why can rPFS and OS have different hazard ratios?
They are different endpoints. rPFS records the first objective radiographic progression or death, whereas OS records death. A treatment can therefore affect the timing of progression differently from the timing of death. Subsequent events and treatments can also contribute to differences between these endpoint-specific estimates.
Does a nonsignificant OS p-value prove no treatment effect?
No. A nonsignificant superiority test does not establish equality. For example, the BRCA OS HR was 0.91 with a 95% CI of 0.68–1.20 and P = 0.5044. The appropriate statistical statement is that the reported analysis did not provide evidence of superiority at the tested level; it is not a proof that the true hazard ratio is exactly 1.
12. Confidence Intervals and the Null Value
For hazard ratios, the null value is 1. An HR of 1 represents equal estimated instantaneous event rates between the two groups. The position of the confidence interval relative to 1 therefore provides an important visual and numerical summary.
| Analysis | HR | 95% CI | Relationship to 1 |
|---|---|---|---|
| BRCA rPFS | 0.50 | 0.36–0.69 | Entire interval below 1 |
| BRCA + ATM rPFS | 0.61 | 0.47–0.80 | Entire interval below 1 |
| BRCA OS | 0.91 | 0.68–1.20 | Interval includes 1 |
| BRCA + ATM OS | 0.99 | 0.78–1.26 | Interval includes 1 |
The two rPFS confidence intervals lie entirely below 1, whereas both OS confidence intervals include 1. This distinction is consistent with the corresponding p-values reported in the registry analyses. It does not, by itself, establish that the rPFS effect and OS effect are statistically different from one another; such a claim would require an appropriate formal comparison of the endpoint-specific effects.
13. Safety Results
The ClinicalTrials.gov record reports serious adverse events by arm using affected participants and the corresponding participants at risk. These figures should be interpreted as descriptive safety counts rather than as efficacy measures.
| Study phase / arm | Serious adverse events affected | At risk |
|---|---|---|
| Rucaparib — Treatment Phase | 81 | 270 |
| Abiraterone Acetate or Enzalutamide or Docetaxel | 36 | 130 |
| Rucaparib — Cross-over Phase | 18 | 70 |
The ClinicalTrials.gov record does not provide a formal comparative p-value, confidence interval, exposure-adjusted incidence rate, or statistical test for serious adverse events. Accordingly, the safety data are presented descriptively rather than transformed into an inferential treatment comparison.
14. Crossover and Its Statistical Implications
The ClinicalTrials.gov record identifies a rucaparib cross-over phase with 18 serious-adverse-event-affected participants among 70 at risk. The presence of a cross-over phase is statistically important because treatment received after the original randomized comparison can affect subsequent outcomes, particularly overall survival.
An ITT analysis preserves the original randomized comparison, but OS is measured across the participant's subsequent clinical course. If participants receive a different treatment after progression or cross over, the eventual OS experience can reflect more than the initially randomized treatment.
What ITT preserves
The treatment comparison remains anchored to the randomized assignment, preserving the principal advantage created by randomization.
What crossover can change
Subsequent treatment can influence later outcomes, potentially making OS less directly attributable to the originally assigned treatment alone.
The ClinicalTrials.gov record does not provide enough information to quantify a crossover-adjusted OS effect, so no such adjustment is attempted here.
15. Multiplicity and Endpoint Interpretation
TRITON3 has two registered primary endpoints and additional secondary endpoints. The ClinicalTrials.gov record identifies both primary analyses as superiority tests and also provide two secondary OS analyses.
| Endpoint | Role | Population | Reported P-value |
|---|---|---|---|
| BRCA rPFS | Primary | ITT BRCA mutated mCRPC | <0.001 |
| BRCA + ATM rPFS | Primary | ITT BRCA or ATM mutated mCRPC | <0.001 |
| BRCA OS | Secondary | ITT BRCA mutated mCRPC | 0.5044 |
| BRCA + ATM OS | Secondary | ITT BRCA or ATM mutated mCRPC | 0.9368 |
Multiple endpoints create an important interpretive issue: the existence of several statistical tests can affect how nominal p-values should be interpreted if a formal multiplicity-control strategy has been specified. The registry-reported TRITON3 data identify the endpoints, methods, effect measures, and hypothesis type but do not provide a multiplicity-adjustment procedure or alpha-allocation scheme. Therefore, this page does not infer one.
16. Interim Analysis, Missing Data, Stratification, and Bayesian Methods
Interim analysis
The registry-reported TRITON3 data do not report an interim-analysis procedure, stopping boundary, or alpha-spending method. None is inferred here.
Missing data / imputation
The ClinicalTrials.gov record does not report a specific missing-data or imputation method for the efficacy analyses. No imputation procedure is attributed to the trial.
Stratification
The ClinicalTrials.gov record identifies the log-rank test but do not provide stratification factors. No specific stratification variables are added.
Bayesian methods
The statistical analyses posted on ClinicalTrials.gov do not identify a Bayesian method. The reported primary method is the log-rank test with hazard-ratio effect measures.
This distinction is important for a standardized statistical review. A method should be included because it is supported by the trial record, not because it is common in similar trials.
17. Time-to-Event Analysis: What Is Actually Being Compared?
Time-to-event analysis is not simply a comparison of whether an event happened. It uses both whether an event occurred and when it occurred, while accommodating participants whose event status is not observed through the full follow-up period.
A participant contributes information until the event occurs or until the participant is censored. The statistical analysis uses these event and censoring times rather than reducing every participant to a simple yes/no outcome.
For TRITON3, this framework is particularly relevant because both primary endpoints are rPFS outcomes and both secondary endpoints are OS outcomes. The hazard ratio summarizes the relative event rate over the analyzed follow-up, while the log-rank test evaluates the separation of event-time distributions.
18. Why the BRCA-Specific and Combined Analyses Should Be Kept Separate
The registry defines two distinct primary populations: participants with a BRCA alteration and participants with a BRCA or ATM alteration combined. These are related but not identical estimands.
| Population | Primary rPFS HR | 95% CI | P-value |
|---|---|---|---|
| BRCA alteration | 0.50 | 0.36–0.69 | <0.001 |
| BRCA or ATM alteration combined | 0.61 | 0.47–0.80 | <0.001 |
The combined analysis should not be interpreted as merely a larger version of the BRCA analysis. Adding participants with ATM alterations changes the analysis population and therefore changes the population to which the resulting hazard ratio applies.
Likewise, the numerical difference between 0.50 and 0.61 does not by itself establish that rucaparib has a statistically different effect in BRCA and ATM populations. A formal interaction or heterogeneity analysis would be required to make that claim, and no such analysis is provided in the ClinicalTrials.gov record.
19. What the Hazard Ratio Does — and Does Not — Mean
The HR of 0.50 corresponds to a 50% lower estimated hazard of radiographic progression or death in the rucaparib group relative to physician's choice under the reported analysis.
It does not mean that exactly 50% fewer participants progressed, that 50% of participants were cured, or that every participant experienced the same proportional reduction.
The HR of 0.61 corresponds to a 39% lower estimated hazard of radiographic progression or death in the rucaparib group relative to physician's choice.
Again, this is a relative hazard measure rather than an absolute reduction in the probability of an event.
The HR of 0.91 corresponds to an estimated instantaneous death rate approximately 9% lower in the rucaparib group relative to physician's choice under the reported model.
The 95% CI of 0.68–1.20 includes the null value of 1, and the reported P = 0.5044 does not provide statistical evidence for superiority under the reported test.
The HR of 0.99 is very close to the null value of 1. The 95% CI of 0.78–1.26 includes 1, and P = 0.9368 does not provide statistical evidence for superiority under the reported test.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The two registered primary rPFS analyses reported hazard ratios below 1, with 95% confidence intervals entirely below 1 and P < 0.001. The secondary OS analyses reported hazard ratios of 0.91 and 0.99, with confidence intervals that included 1.
Clinical interpretation
The ClinicalTrials.gov record supports describing a difference in the reported rPFS time-to-event analyses and a lack of statistical evidence for superiority in the posted OS analyses. The ClinicalTrials.gov record does not provide median survival or absolute survival probabilities, so those measures are not used to extend the interpretation.
This separation is deliberate. Statistical evidence and clinical interpretation are related but not interchangeable. A hazard ratio describes a relative time-to-event comparison; a p-value describes evidence under a hypothesis test; and a confidence interval describes uncertainty. None of these quantities alone summarizes the full clinical value of a treatment strategy.
21. Important Limitations and Interpretation Issues
- No median survival estimates reported: the ClinicalTrials.gov record does not report median rPFS or median OS, so no median event times are presented.
- No absolute survival probabilities reported: the ClinicalTrials.gov record does not provide Kaplan-Meier survival probabilities at specified time points.
- No efficacy event counts reported: the statistical-analysis records provide HRs, confidence intervals, and p-values but do not provide the underlying efficacy event counts.
- Open-label design: masking was reported as none. This is relevant to outcomes that can be influenced by treatment knowledge, although the primary rPFS endpoint was assessed by independent radiology review.
- Independent radiology review: the primary endpoint is defined by independent radiology review, which is distinct from investigator assessment.
- Population-specific estimates: the BRCA-specific and BRCA-or-ATM analyses apply to different analysis populations and should not be treated as interchangeable.
- Hazard-ratio interpretation: an HR is not an absolute risk difference and does not describe individual treatment benefit.
- Proportional-hazards caution: the ClinicalTrials.gov record does not provide enough information to assess whether the proportional-hazards assumption is appropriate throughout follow-up.
- Multiplicity: the ClinicalTrials.gov record identifies two primary endpoints and two secondary OS analyses but do not specify a multiplicity-adjustment procedure.
- Crossover: a rucaparib cross-over phase is identified in the safety data, which can complicate interpretation of outcomes measured after the initial randomized treatment.
- Safety denominators: serious-adverse-event counts are posted on ClinicalTrials.gov for treatment and cross-over phases with different denominators and should not be combined.
- Unreported methods: the ClinicalTrials.gov record does not specify interim-analysis procedures, missing-data imputation, stratification factors, or Bayesian methods.
22. Why This Trial Matters Statistically
TRITON3 is a useful teaching case because it places several core clinical-trial concepts into a single randomized phase 3 framework: intention-to-treat analysis, time-to-event endpoints, independent radiology review, log-rank testing, hazard ratios, confidence intervals, superiority testing, molecularly defined analysis populations, crossover, and the distinction between progression-free and overall survival.
| Concept | How it appears in TRITON3 |
|---|---|
| Randomization | Randomized phase 3 parallel-group design with 405 enrolled participants. |
| ITT analysis | Primary and secondary efficacy analyses use specified ITT populations. |
| Time-to-event endpoint | Both primary endpoints are radiographic progression-free survival; both secondary analyses are overall survival. |
| Independent review | Primary rPFS endpoints are reported by independent radiology review. |
| Log-rank test | Reported formal method for all four statistical analyses. |
| Hazard ratio | Reported effect measure for both rPFS and OS analyses. |
| Confidence interval | Two-sided 95% CIs accompany all four posted hazard-ratio estimates. |
| Superiority | Both primary analyses and both secondary analyses are identified with a superiority hypothesis. |
| Analysis populations | BRCA-specific and BRCA-or-ATM combined populations are analyzed separately. |
| Crossover | The ClinicalTrials.gov record includes a distinct rucaparib cross-over phase. |
| Safety analysis | Serious adverse events are reported descriptively by treatment and cross-over phase. |
| Multiplicity | Two primary endpoints and additional secondary endpoints require attention to endpoint hierarchy and error control. |
23. Longitudinal Trial History
Study start
The TRITON3 study began on June 13, 2017, according to the ClinicalTrials.gov record.
Randomized comparison
The completed phase 3 trial randomized participants in a parallel-group, open-label design comparing rucaparib with physician's choice of abiraterone acetate, enzalutamide, or docetaxel.
rPFS in BRCA population
The BRCA-specific primary analysis reported an HR of 0.50 with a 95% CI of 0.36–0.69 and P < 0.001.
rPFS in BRCA + ATM population
The combined primary analysis reported an HR of 0.61 with a 95% CI of 0.47–0.80 and P < 0.001.
Overall survival
The BRCA-specific OS analysis reported HR 0.91, 95% CI 0.68–1.20, P = 0.5044; the combined BRCA-or-ATM analysis reported HR 0.99, 95% CI 0.78–1.26, P = 0.9368.
Primary completion
The ClinicalTrials.gov record identifies August 25, 2022 as the primary completion date, with total rPFS follow-up up to approximately 4 years.
24. A Practical Reading of the Four Reported Analyses
A useful way to read the TRITON3 results is to move through the statistical hierarchy rather than focusing on a single number.
Step 1 — Identify the endpoint
First determine whether the analysis concerns rPFS or OS. These are different estimands and should not be interpreted as interchangeable.
Step 2 — Identify the population
Next determine whether the estimate applies to participants with a BRCA alteration or to the combined BRCA-or-ATM population.
Step 3 — Read the HR
The HR describes the relative time-to-event effect. Values below 1 favor a lower estimated event hazard in the rucaparib group.
Step 4 — Read the CI
The 95% CI communicates uncertainty around the HR and shows whether the interval includes the null value of 1.
Step 5 — Read the p-value
The p-value addresses evidence against the null hypothesis. It is not an effect-size measure.
Step 6 — Check the design
Finally, consider randomization, open-label status, independent review, ITT analysis, crossover, and the available information about multiplicity and other statistical procedures.
25. Related Tutorials
Learn more about the methods used in this trial:
26. Related Statistical Calculators
27. Sources
- ClinicalTrials.gov: TRITON3 — NCT02975934.
- Linked publication: PubMed record — PMID 36795891.
- Linked publication: PubMed record — PMID 36898948.
- Linked publication: PubMed record — PMID 32203306.
Continue through the Clinical Biostats statistical pathway
Explore the statistical concepts that connect randomized clinical-trial design with survival analysis, inference, and clinical-trial methodology.
28. Record Summary
TRITON3 provides a useful example of a randomized phase 3 time-to-event analysis in a molecularly defined oncology population. The two primary rPFS analyses used intention-to-treat populations, independent radiology review, log-rank testing, hazard-ratio estimation, two-sided 95% confidence intervals, and superiority hypotheses. The BRCA-specific rPFS analysis reported HR 0.50 (95% CI 0.36–0.69; P < 0.001), while the combined BRCA-or-ATM analysis reported HR 0.61 (95% CI 0.47–0.80; P < 0.001). The secondary OS analyses reported HR 0.91 (95% CI 0.68–1.20; P = 0.5044) in the BRCA population and HR 0.99 (95% CI 0.78–1.26; P = 0.9368) in the combined population.
The most statistically informative reading therefore combines the endpoint, analysis population, hazard ratio, confidence interval, p-value, and trial design. The differences between rPFS and OS results should be interpreted as endpoint-specific findings rather than reduced to a single summary statistic.