← Clinical Trials
Breast Cancer Phase 3 Randomized NCT05104866

TROPION-Breast01: Complete Statistical Analysis of Dato-DXd in HR-Positive, HER2-Negative Breast Cancer

An independent statistical review of the randomized phase 3 TROPION-Breast01 trial comparing Dato-DXd with investigator's choice of chemotherapy in participants with inoperable or metastatic HR-positive, HER2-negative breast cancer who had received one or two prior lines of systemic chemotherapy.

Phase 3  ·  Randomized, parallel-group  ·  Open-label  ·  Enrollment 732  ·  Results posted
Scope of this record

This page separates reported trial results from statistical interpretation. 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

TROPION-Breast01 is a phase 3, open-label, randomized, parallel-group study evaluating Dato-DXd versus investigator's choice of chemotherapy in participants with inoperable or metastatic HR-positive, HER2-negative breast cancer who have been treated with one or two prior lines of systemic chemotherapy.

732
Enrollment
2 treatment arms
0.63
PFS HR
99% CI 0.49–0.80
1.01
OS HR
95.97% CI 0.83–1.23
<0.0001
PFS P-value
Superiority analysis
FeatureTROPION-Breast01
PhasePhase 3
PopulationParticipants with inoperable or metastatic HR-positive, HER2-negative breast cancer who have been treated with one or two prior lines of systemic chemotherapy
DesignRandomized, open-label, parallel-group
AllocationRandomized
Primary purposeTreatment
Primary endpointsProgression-Free Survival and Overall Survival
Primary endpoint typeTime-to-event
Enrollment732
Arms2
Start2021-10-18
Primary completion2024-07-24
StatusActive, not recruiting
Lead sponsorAstraZeneca

2. Clinical Question

The primary statistical question was whether Dato-DXd would improve the two prespecified time-to-event endpoints, Progression-Free Survival and Overall Survival, compared with investigator's choice of chemotherapy.

Population

Participants with inoperable or metastatic HR-positive, HER2-negative breast cancer who have been treated with one or two prior lines of systemic chemotherapy.

Intervention

Dato-DXd.

Comparator

Investigator's Choice of Chemotherapy (ICC), consisting of capecitabine, gemcitabine, eribulin, or vinorelbine.

Primary question

Does Dato-DXd produce a superior time-to-event outcome compared with investigator's choice of chemotherapy?

3. Trial Design

01
Randomize732 participants
02
2 armsDato-DXd vs ICC
03
Open-labelNo masking
04
AssessTime-to-event and binary outcomes
05
CompareHazard and odds ratios
Allocation
Randomized allocation in a parallel-group design.
Masking
None. The trial is open-label.
Primary purpose
Treatment.
Hypothesis type
Superiority.
ARM 1

Dato-DXd

  • Dato-DXd
ARM 2

Investigator's Choice of Chemotherapy

  • Capecitabine
  • Gemcitabine
  • Eribulin
  • Vinorelbine

The statistical analysis record shows that the primary time-to-event comparisons were performed as Dato-DXd versus investigator's choice of chemotherapy, with analyses adjusted for stratification factors through stratified Cox proportional-hazards models.

4. Endpoints

Primary Endpoint: Progression-Free Survival

ItemRegistered information
EndpointProgression-Free Survival
TypeTime-to-event
Time frameOn-study tumor assessments occur every 6 weeks then every 9 weeks until disease progression, death or withdrawal of consent assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months
DefinitionPFS is defined as time from randomization until progression per RECIST 1.1, as assessed by BICR, or death due to any cause. The analysis will include all randomized participants as randomized regardless of whether the participant withdraws from therapy, receives another anti-cancer therapy, or clinically progresses prior to RECIST 1.1.

Primary Endpoint: Overall Survival

ItemRegistered information
EndpointOverall Survival
TypeTime-to-event
Time frameFrom date of randomization until death due to any cause. Assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months.
DefinitionOS is defined as time from randomization until the date of death due to any cause. The comparison will include all randomized participants as randomized, regardless of whether the participant withdraws from therapy or receives another anti-cancer therapy. The measure of interest is the hazard ratio of OS.

Secondary Endpoint Definitions and Time Frames

EndpointTypeTime frame
Objective Response Rate (ORR)BinaryFrom date of randomization until event. Assessed up to data cut-off (17Jul2023) to a maximum of approximately 21 months for BICR assessment and assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months for investigator assessment.
Progression-Free Survival by Investigator AssessmentTime-to-eventOn-study tumor assessments occur every 6 weeks then every 9 weeks until disease progression, death or withdrawal of consent assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months
Disease Control Rate (DCR)BinaryAssessed up to data cut-off (17Jul2023) to a maximum of 21 months for BICR assessment and assessed up to data cut-off (24Jul2024) to a maximum of 33 months for investigator assessment.
Time to First Subsequent Therapy (TFST)Time-to-eventFrom randomization to start of first subsequent anti-cancer therapy. Assessments occur at every visit after study treatment has been discontinued. Assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months.
Time to Second Subsequent Therapy (TSST)Time-to-eventFrom randomization to start of second subsequent anti-cancer therapy. Assessments occur at every visit after study treatment has been discontinued. Assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months.
Time From Randomization to Second Progression or Death (PFS2)Time-to-eventFrom date of randomization to second progression or death. PFS2 assessments occur every 3 months after disease progression. Assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months
Clinical Outcome Assessment- TTD in PainTime-to-eventFrom date of randomization to 18 weeks post-progression. Assessments occur every 3 weeks then every 6 weeks until 18 weeks post-progression and at end of treatment visit. Assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months.
Clinical Outcome Assessment- TTD in Physical FunctioningTime-to-eventFrom date of randomization to 18 weeks post-progression. Assessments occur every 3 weeks then every 6 weeks until 18 weeks post-progression and at end of treatment visit. Assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months.
Clinical Outcome Assessment- TTD in GHSTime-to-eventFrom date of randomization to 18 weeks post-progression. Assessments occur every 3 weeks then every 6 weeks until 18 weeks post-progression and at end of treatment visit. Assessed up to data cut-off (24Jul2024) to a maximum of approximately 33 months.

5. Statistical Methodology

Primary time-to-event analysis

The primary PFS and OS analyses used the log-rank test for treatment comparison and a stratified Cox proportional-hazards model for estimation of the hazard ratio. The analysis text states that the model was adjusted for stratification factors.

Core survival-analysis quantities
HR = estimated hazard in Dato-DXd ÷ estimated hazard in ICC

An HR below 1 favors Dato-DXd according to the registry analysis text. An HR above 1 favors ICC for these analyses.

Binary endpoint analysis

For ORR and DCR, the registry reports logistic regression with treatment and the stratification factors. The reported effect measure is an odds ratio. The analysis text states that an odds ratio greater than 1 favors Dato-DXd.

Confidence intervals for response-rate differences

The registry also reports differences in raw response or disease-control rates using the Miettinen-Nurminen method. These are two-sided 95% confidence intervals. A positive difference favors Dato-DXd according to the analysis text.

Stratification

Several analyses are explicitly described as adjusted for stratification factors or as stratified analyses. The ClinicalTrials.gov record does not specify the individual stratification factors, so this page does not infer or add them.

Superiority framework

The statistical analyses are identified as superiority hypotheses. This matters because the interpretation of an estimate is relative to a null hypothesis of no treatment difference rather than a non-inferiority margin. No non-inferiority margin is reported in the ClinicalTrials.gov record.

6. Primary Results

Progression-Free Survival

Hazard ratio for progression or death

0.63

99% two-sided CI: 0.49–0.80   ·   P < 0.0001

Log-rank comparison with a stratified Cox proportional-hazards model adjusted for stratification factors.

Clinical Biostats interpretation

The estimated hazard ratio of 0.63 means that, under the fitted Cox model, the estimated instantaneous rate of progression or death was approximately 37% lower with Dato-DXd than with investigator's choice of chemotherapy. The 37% figure is a direct interpretation of 1 − 0.63; it is not a statement that 37% of participants avoided progression or death.

The two-sided 99% confidence interval of 0.49–0.80 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of individual patient outcomes.

The P < 0.0001 value addresses evidence against the relevant null hypothesis under the statistical test. It does not measure the magnitude or clinical importance of the treatment effect. The hazard ratio and its confidence interval provide the effect-size information.

Because the analysis uses a Cox proportional-hazards model, interpretation of a single hazard ratio is tied to the proportional-hazards modeling framework. The ClinicalTrials.gov record does not report a diagnostic assessment of that assumption. The analysis is also based on randomized participants as randomized, and censoring and subsequent treatment can affect the observed time-to-event data.

Overall Survival

Hazard ratio for death

1.01

95.97% two-sided CI: 0.83–1.23   ·   P = 0.9445

Log-rank comparison with a stratified Cox proportional-hazards model adjusted for stratification factors.

Clinical Biostats interpretation

The estimated OS hazard ratio of 1.01 is very close to 1.00. Under the fitted model, this corresponds to an estimated instantaneous hazard of death that is approximately 1% higher with Dato-DXd than with investigator's choice of chemotherapy. This is a model-based relative estimate, not a statement that 1% more participants died.

The 95.97% confidence interval of 0.83–1.23 spans 1.00 and therefore includes hazard ratios corresponding to lower or higher estimated hazards with Dato-DXd. The interval is the more informative description of statistical precision than the point estimate alone.

The P = 0.9445 value indicates that the observed result provides little evidence against the relevant no-difference null hypothesis under this test. A p-value does not establish that two treatments are clinically equivalent, nor does it measure the probability that either treatment is beneficial.

The analysis was specified as a superiority comparison. Therefore, it should not be recast as a formal non-inferiority or equivalence analysis. No non-inferiority margin is reported in the ClinicalTrials.gov record.

Important CI detail: The ClinicalTrials.gov record lists the OS confidence interval as 95.97%: 0.83–1.23, while its analysis notes separately state “95% Confidence Interval 2-Sided 0.83 to 1.22.” This page reports the formal CI fields exactly as reported in the registry for the statistical analysis: 95.97%, 0.83–1.23. The separate analysis-note values are not substituted for those fields.

7. Secondary Endpoint Results

Objective Response Rate — BICR Assessment

Odds ratio for response

1.95

95% two-sided CI: 1.41–2.71   ·   P < 0.0001

Confirmed response by BICR assessment, based on the IA1 data cut-off.

MeasureDato-DXd vs ICC
Odds ratio1.95
95% CI1.41–2.71
P-value<0.0001
Raw response-rate difference13.55
95% CI for difference6.96–20.05

The raw response-rate difference was calculated using the Miettinen-Nurminen method. A positive difference favors Dato-DXd according to the registry analysis text.

Objective Response Rate — Investigator Assessment

Odds ratio for response

2.06

95% two-sided CI: 1.49–2.87   ·   P < 0.0001

Confirmed response by investigator assessment, based on the FA data cut-off.

MeasureDato-DXd vs ICC
Odds ratio2.06
95% CI1.49–2.87
P-value<0.0001
Raw response-rate difference14.64
95% CI for difference8.08–21.11

Progression-Free Survival by Investigator Assessment

Hazard ratio

0.64

95% two-sided CI: 0.55–0.76   ·   P < 0.0001

Stratified Cox proportional-hazards model adjusted for the stratification factors.

Disease Control Rate — BICR Assessment

MeasureEstimate
Odds ratio1.75
95% CI1.27–2.42
P-value0.0006
Raw disease-control-rate difference11.58
95% CI for difference4.93–18.16

The BICR disease-control analysis used logistic regression with treatment and the stratification factors. The raw difference used the Miettinen-Nurminen method.

Disease Control Rate — Investigator Assessment

MeasureEstimate
Odds ratio2.12
95% CI1.51–3.00
P-value<0.0001
Raw disease-control-rate difference14.06
95% CI for difference7.74–20.33

Time to First Subsequent Therapy

Hazard ratio

0.58

95% two-sided CI: 0.50–0.68

Stratified Cox proportional-hazards model. A hazard ratio below 1 favors Dato-DXd.

Time to Second Subsequent Therapy

Hazard ratio

0.83

95% two-sided CI: 0.70–0.98

Stratified Cox proportional-hazards model. A hazard ratio below 1 favors Dato-DXd.

Time From Randomization to Second Progression or Death (PFS2)

Hazard ratio

0.76

95% two-sided CI: 0.63–0.93

Stratified Cox proportional-hazards model. A hazard ratio below 1 favors Dato-DXd.

Clinical Outcome Assessment — TTD in Pain

Hazard ratio

0.84

95% two-sided CI: 0.67–1.06   ·   P = 0.1519

Stratified Cox proportional-hazards model adjusted for stratification factors.

Clinical Outcome Assessment — TTD in Physical Functioning

Hazard ratio

0.79

95% two-sided CI: 0.62–1.01   ·   P = 0.0596

Stratified Cox proportional-hazards model adjusted for stratification factors.

Clinical Outcome Assessment — TTD in GHS

Hazard ratio

0.83

95% two-sided CI: 0.67–1.04   ·   P = 0.1155

Stratified Cox proportional-hazards model adjusted for stratification factors.

Reading the secondary analyses

The secondary results use different estimands. Odds ratios describe relative odds of a binary outcome, while hazard ratios describe relative instantaneous event rates in time-to-event analyses. These estimates should therefore not be compared as though they were interchangeable measures of treatment effect.

The confidence intervals also vary in width because the endpoints contain different amounts of information. For example, the PFS investigator-assessment estimate is 0.64 with a 95% CI of 0.55–0.76, whereas TTD in pain is 0.84 with a 95% CI of 0.67–1.06. The latter interval crosses 1.00, so the result is compatible with a range of effects on either side of the no-difference value under the model.

8. Safety

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

Serious adverse eventsAffected / at risk
Dato-DXd62/360
Investigator's Choice of Chemotherapy67/351

These are serious adverse-event counts by arm as reported in the ClinicalTrials.gov record. The ClinicalTrials.gov record does not supply a formal between-arm statistical test for serious adverse events, so no comparative p-value or effect estimate is added.

Statistical distinction: efficacy analyses are anchored to randomized treatment assignment, whereas safety summaries are commonly organized around treatment exposure. The ClinicalTrials.gov record specifically reports the serious-adverse-event affected/at-risk quantities above; it does not provide additional safety definitions or comparative analyses for this page.

9. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

PFS, OS, TFST, TSST, PFS2, and the three clinical outcome assessment endpoints are time-to-event outcomes. A Cox model provides a way to estimate a relative hazard while accounting for differing follow-up times and censoring. In TROPION-Breast01, the registry analysis text specifies a stratified Cox proportional-hazards model for the relevant analyses.

What does an HR of 0.63 mean for PFS?

An HR of 0.63 means the estimated instantaneous hazard of progression or death under the fitted model is 63% of that in the comparator group. Equivalently, 1 − 0.63 gives an estimated 37% lower hazard. It does not mean that 37% of participants avoided progression or that each participant experienced exactly a 37% reduction in risk.

Why is the PFS confidence interval 99% while many secondary intervals are 95%?

The registry statistical-analysis record reports a 99% two-sided confidence interval for the primary PFS hazard ratio and a 95.97% two-sided interval for OS. Most of the registry-reported secondary analyses use 95% two-sided confidence intervals. Confidence level is part of the statistical specification and should be reported rather than silently converted to a common level.

What does an odds ratio of 1.95 mean for ORR?

An odds ratio of 1.95 means the estimated odds of confirmed response are 1.95 times as high in the Dato-DXd group as in the investigator's-choice group under the logistic regression model. Odds are not the same as probabilities, so an odds ratio cannot be read directly as a 95% higher response rate.

Why also report the raw response-rate difference?

The odds ratio is a relative measure on the odds scale. The raw difference is on the percentage-point scale and directly describes the separation between response rates. TROPION-Breast01 reports both, with the difference confidence interval calculated using the Miettinen-Nurminen method.

Why are stratification factors included in the models?

The registry-reported analysis text states that the PFS and OS analyses were adjusted for stratification factors and that the ORR and DCR logistic-regression models included treatment and the stratification factors. Incorporating those factors aligns the analysis with the randomized design and can improve the precision of the treatment comparison.

Why should the p-value not be treated as an effect-size measure?

A p-value describes the compatibility of the observed data with a specified null hypothesis under the statistical model. It depends on both the estimated effect and the amount of information in the analysis. The hazard ratio or odds ratio provides the effect estimate, while the confidence interval describes its statistical precision.

10. Primary Result Interpretation in Context

PFS signal

The PFS hazard ratio was 0.63, with a 99% two-sided confidence interval of 0.49–0.80 and P < 0.0001. The entire reported 99% interval is below 1.00.

OS estimate

The OS hazard ratio was 1.01, with a 95.97% two-sided confidence interval of 0.83–1.23 and P = 0.9445. The interval includes 1.00.

Response

BICR-confirmed ORR produced an odds ratio of 1.95 and a raw response-rate difference of 13.55, both with confidence intervals reported in the registry analysis.

Subsequent disease control

TFST, TSST, and PFS2 were analyzed with Cox proportional-hazards methods and produced hazard ratios of 0.58, 0.83, and 0.76, respectively.

The central statistical feature is the difference between the two primary time-to-event estimates. PFS shows a hazard ratio below 1 with a confidence interval entirely below 1, whereas OS is estimated at 1.01 with a confidence interval spanning 1.00. Those are distinct statistical findings and should not be collapsed into a single overall treatment-effect statement.

11. Understanding the Confidence Intervals

PFS
HR 0.63   |   99% CI 0.49–0.80

The interval quantifies uncertainty around the PFS hazard-ratio estimate under the specified analysis. Every value in the registry-reported interval is below 1.00.

OS
HR 1.01   |   95.97% CI 0.83–1.23

The interval contains 1.00, so the ClinicalTrials.gov record is compatible with either a lower or higher estimated hazard under the Cox model at that confidence level.

BICR ORR
OR 1.95   |   95% CI 1.41–2.71

The odds-ratio interval lies above 1.00, while the separate Miettinen-Nurminen response-rate difference is 13.55 with a 95% CI of 6.96–20.05.

A confidence interval should not be interpreted as the range in which the true effect must fall with a particular probability after the study has been completed. It is better understood as an interval produced by a statistical procedure with a specified long-run coverage property under repeated sampling and the assumptions of the analysis.

12. Time-to-Event Analysis: What Is Being Compared?

Time-to-event endpoints contain more information than a simple yes/no outcome because they incorporate both whether an event occurred and when it occurred. They also allow participants who have not yet experienced the event to contribute follow-up information until censoring.

EndpointEvent frameworkReported effect
Progression-Free SurvivalProgression per RECIST 1.1 assessed by BICR, or death due to any causeHR 0.63
Overall SurvivalDeath due to any causeHR 1.01
PFS by Investigator AssessmentTime-to-event assessmentHR 0.64
TFSTStart of first subsequent anti-cancer therapyHR 0.58
TSSTStart of second subsequent anti-cancer therapyHR 0.83
PFS2Second progression or deathHR 0.76
TTD in PainTime to the specified clinical outcome eventHR 0.84
TTD in Physical FunctioningTime to the specified clinical outcome eventHR 0.79
TTD in GHSTime to the specified clinical outcome eventHR 0.83

The different endpoints address different stages of the clinical pathway. PFS focuses on progression or death, OS focuses only on death, TFST and TSST incorporate subsequent treatment, and PFS2 extends the disease-progression framework beyond the first progression event.

13. Binary Outcomes: ORR and DCR

ORR and DCR are binary outcomes: each participant is classified according to whether the prespecified outcome occurred. Logistic regression is therefore a natural model for the odds of the outcome, and that is the method reported in the trial data.

OutcomeAssessmentOdds ratio95% CIP-value
ORRBICR1.951.41–2.71<0.0001
ORRInvestigator2.061.49–2.87<0.0001
DCRBICR1.751.27–2.420.0006
DCRInvestigator2.121.51–3.00<0.0001

The agreement between BICR and investigator-assessed analyses is useful descriptively, but the estimates are still separate analyses of the same randomized comparison. The odds ratios should not be treated as independent pieces of evidence that can simply be multiplied or counted.

14. Multiplicity and Multiple Endpoints

TROPION-Breast01 has two registered primary endpoints, both time-to-event outcomes, and numerous secondary analyses. Multiple endpoints create an important statistical issue: if many hypotheses are tested, the probability of obtaining at least one apparently positive result by chance can increase unless the testing strategy controls the relevant error rate.

Endpoint familyRole in the ClinicalTrials.gov recordReported analysis
Progression-Free SurvivalPrimaryLog-rank test; stratified Cox model
Overall SurvivalPrimaryLog-rank test; stratified Cox model
Objective Response RateSecondaryLogistic regression; Miettinen-Nurminen difference
Disease Control RateSecondaryLogistic regression; Miettinen-Nurminen difference
Other time-to-event outcomesSecondaryCox proportional-hazards model or log-rank test
Interpretation caution: the ClinicalTrials.gov record identifies the analyses as superiority analyses but does not provide a complete multiplicity-control strategy or alpha-allocation scheme. Accordingly, this page does not claim that every secondary p-value is independently confirmatory or that the collection of secondary analyses has a specified familywise error rate.

15. Interim Analysis and Data Cutoffs

The ClinicalTrials.gov record contains different data cutoffs for different endpoints. For example, the primary OS analysis was assessed up to the data cut-off of 24Jul2024, while the ORR analyses use a 17Jul2023 data cut-off. This distinction matters because estimates from different data cutoffs do not necessarily describe the same information set.

2021-10-18

Trial start

The registered trial start date was 2021-10-18.

17Jul2023

IA1 / FA outcome analyses

The registry-reported ORR and DCR analyses use the 17Jul2023 data cut-off, with BICR and investigator assessment identified separately.

24Jul2024

OS data cut-off

The registered OS time frame specifies assessment up to data cut-off 24Jul2024.

2024-07-24

Primary completion

The registered primary completion date was 2024-07-24.

The presence of multiple data cutoffs is not itself a statistical problem. It becomes important when interpreting results because an estimate is always tied to the information available at the specified cutoff.

16. Analysis Populations and Censoring

The registered PFS definition explicitly states that the analysis includes all randomized participants as randomized, regardless of whether a participant withdraws from therapy, receives another anti-cancer therapy, or clinically progresses before RECIST 1.1 assessment. The OS definition likewise states that the comparison includes all randomized participants as randomized regardless of treatment withdrawal or subsequent anti-cancer therapy.

Randomization anchor

The primary PFS and OS definitions begin at randomization, preserving the randomized comparison as the foundation of the efficacy analysis.

Therapy discontinuation

Withdrawal from therapy does not remove a randomized participant from the primary PFS or OS comparison according to the registered definitions.

Subsequent therapy

Receipt of another anti-cancer therapy does not remove a randomized participant from the registered primary efficacy populations.

Censoring

The ClinicalTrials.gov record identifies time-to-event methodology but does not provide the complete censoring rules beyond the endpoint definitions reported here.

No specific missing-data or imputation method is reported in the ClinicalTrials.gov record. This page therefore does not add an imputation procedure that is not documented in the record provided.

17. Stratified Analysis

The analysis notes repeatedly state that comparisons were adjusted for stratification factors or performed with stratified models. This approach is especially important in a randomized trial because the randomization process may have deliberately balanced important prognostic factors between treatment groups.

Conceptual interpretation
Stratified Cox model → treatment effect estimated while accounting for prespecified strata

The treatment hazard ratio remains the principal effect measure, while the stratification structure accounts for the factors specified in the analysis.

The ClinicalTrials.gov record does not identify the individual stratification factors. It would therefore be inappropriate to name or reconstruct them from external knowledge of the trial.

18. No Non-Inferiority Margin, Factorial Design, or Crossover Analysis Reported

The registry-reported statistical profile identifies the primary hypotheses as superiority. It does not report a non-inferiority margin, factorial structure, or crossover analysis.

Design topicSupported by the ClinicalTrials.gov record?Interpretation
SuperiorityYesThe primary and secondary analyses are identified as superiority analyses.
Non-inferiority marginNoNo margin is reported, so non-inferiority logic is not applied.
Factorial designNoThe design model is parallel, not factorial.
CrossoverNoNo crossover analysis is reported in the ClinicalTrials.gov record.
Bayesian methodsNoNo Bayesian method is reported.
Interim analysisData-cutoff labels are reportedThe ClinicalTrials.gov record contains IA1/FA terminology but does not provide a complete interim-monitoring or alpha-spending specification.

19. What the Primary Results Do — and Do Not — Establish

PFS

The PFS result provides evidence of a treatment-group difference under the specified stratified survival analysis: HR 0.63, 99% CI 0.49–0.80, P < 0.0001. It does not provide a median PFS, an absolute probability of progression-free survival at a particular time, or the proportion of individual patients benefiting because those quantities are not reported in the ClinicalTrials.gov record.

OS

The OS estimate is HR 1.01 with a 95.97% CI of 0.83–1.23 and P = 0.9445. This is an estimate of the relative hazard of death under the Cox model. It does not establish equivalence, because equivalence or non-inferiority would require a prespecified framework and margin that are not reported here.

Response

The BICR ORR odds ratio of 1.95 and raw response-rate difference of 13.55 describe binary tumor-response outcomes at the specified analysis cutoff. Neither measure is a substitute for a time-to-event endpoint.

20. Limitations

21. Why This Trial Matters Statistically

TROPION-Breast01 is a useful teaching case because it places several common clinical-trial methods side by side within one randomized study. The primary endpoints are time-to-event outcomes analyzed with log-rank testing and stratified Cox models, while response and disease-control endpoints use logistic regression and score-based confidence intervals for raw differences.

ConceptHow it appears in TROPION-Breast01
RandomizationRandomized, parallel-group phase 3 design
Time-to-event analysisPFS and OS are the two primary endpoints
Log-rank testUsed for the primary PFS and OS comparisons and selected secondary time-to-event outcomes
Cox proportional-hazards modelUsed for hazard-ratio estimation, including stratified analyses
Hazard ratioPrimary effect measure for PFS and OS and several secondary time-to-event endpoints
Logistic regressionUsed for ORR and DCR
Odds ratioReported effect measure for ORR and DCR
Miettinen-Nurminen methodUsed for confidence intervals around raw response-rate and disease-control-rate differences
Confidence intervalsReported at different specified confidence levels across analyses
Stratified analysisExplicitly incorporated into multiple analyses
Multiple endpointsTwo primary endpoints plus numerous secondary analyses

The trial therefore illustrates why a clinical-trial results page should not reduce all evidence to a single p-value. The appropriate interpretation depends on the endpoint, estimand, effect measure, confidence interval, analysis population, data cutoff, and model assumptions.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue through Clinical Biostats

Connect this trial's endpoints and statistical methods to deeper biostatistics tutorials and practical statistical calculators.

25. Record Summary

TROPION-Breast01 provides a clear example of how modern randomized clinical-trial results require endpoint-specific statistical interpretation. Its two primary endpoints are time-to-event outcomes analyzed with log-rank testing and stratified Cox proportional-hazards models. The registry-reported PFS analysis reports HR 0.63 with a 99% two-sided CI of 0.49–0.80 and P < 0.0001, while the registry-reported OS analysis reports HR 1.01 with a 95.97% two-sided CI of 0.83–1.23 and P = 0.9445. Secondary analyses add logistic-regression odds ratios, raw response-rate differences using the Miettinen-Nurminen method, and additional Cox-model time-to-event estimates.

The statistical lesson is that these measures answer different questions. A hazard ratio describes a relative time-to-event effect under a model; an odds ratio describes relative odds for a binary endpoint; a raw rate difference describes absolute separation on the percentage-point scale; a confidence interval describes statistical precision; and a p-value addresses evidence against a specified null hypothesis. Reading the trial correctly requires keeping those quantities distinct rather than treating them as interchangeable.

Clinical Biostats methodology: The purpose of an independent statistical analysis page is to reconstruct the statistical structure of the trial from the reported record while clearly separating documented results from educational interpretation. Where the ClinicalTrials.gov record does not report a value or method, this page does not infer one.