← Clinical Trials
Urothelial Cancer Phase 3 Completed NCT04527991

TROPiCS-04: Complete Statistical Analysis of Sacituzumab Govitecan in Urothelial Cancer

An independent statistical analysis of the randomized phase 3 TROPiCS-04 trial comparing sacituzumab govitecan-hziy with physician's choice of treatment in participants with locally advanced or metastatic unresectable urothelial cancer.

Trial start: January 13, 2021  ·  Primary completion: July 4, 2025  ·  Enrollment: 711
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 results on this page are restricted to the trial data reported in the ClinicalTrials.gov record.

1. Trial at a Glance

TROPiCS-04 was a randomized, parallel-group, open-label phase 3 trial evaluating sacituzumab govitecan-hziy (SG) versus treatment of physician's choice in participants with locally advanced or metastatic unresectable urothelial cancer. The registered primary endpoint was overall survival (OS), a time-to-event endpoint analyzed using the Kaplan-Meier method and a stratified time-to-event comparison.

711
Enrollment
Randomized trial
3
Phase
Phase 3
0.86
Primary OS HR
95% CI 0.73–1.02
0.0870
Primary OS P-value
Two-sided
FeatureTROPiCS-04
PhasePhase 3
ConditionLocally Advanced or Metastatic Unresectable Urothelial Cancer
DesignRandomized, parallel-group, open-label
AllocationRandomized
Primary purposeTreatment
Primary endpointOverall Survival (OS)
Primary endpoint typeTime-to-event
Enrollment711
Arms2
StatusCompleted
Lead sponsorGilead Sciences
Sponsor typeIndustry
ClinicalTrials.govNCT04527991

2. Clinical Question

The central statistical question was whether sacituzumab govitecan-hziy should be compared with treatment of physician's choice with respect to overall survival in participants with locally advanced or metastatic unresectable urothelial cancer.

Population

Participants with locally advanced or metastatic unresectable urothelial cancer.

Intervention

Sacituzumab Govitecan-hziy (SG).

Comparator

Treatment of Physician's Choice. The registered interventions include paclitaxel, docetaxel, and vinflunine.

Primary question

Does SG produce a different overall-survival profile from treatment of physician's choice under the prespecified superiority framework?

3. Trial Design

01
Randomize711 participants
02
Two armsSG vs physician's choice
03
Follow-upTime-to-event outcomes
04
AssessmentOS, PFS, response, QoL
05
AnalysisStratified statistical methods
ARM 1

Sacituzumab Govitecan-hziy

  • Sacituzumab Govitecan-hziy (SG)
  • Primary comparison against treatment of physician's choice
  • Serious adverse events reported for 185 of 349 participants at risk
ARM 2

Treatment of Physician's Choice

  • Physician's choice of treatment
  • Registered drug interventions include paclitaxel, docetaxel, and vinflunine
  • Serious adverse events reported for 110 of 337 participants at risk
Open-label design: the registry describes the masking as none. This matters because treatment assignment was not masked to participants or investigators. For endpoints that involve investigator assessment, lack of masking can be a potential source of assessment-related bias; the registry separately reports progression-free survival by blinded independent central review (BICR), providing a distinct assessment framework for that endpoint.

4. Trial Timing and Registry Status

January 13, 2021

Trial start

The registered trial start date was January 13, 2021.

July 4, 2025

Primary completion

The registered primary completion date was July 4, 2025.

Registry status

Completed

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

5. Endpoints

EndpointRole / time frameRegistry definition or analysis
Overall Survival (OS) Primary
Up to 42 months
OS was defined as time from the date of randomization to the date of death, regardless of cause. Kaplan-Meier estimates were used for analysis.
Progression-Free Survival (PFS) by Investigator Assessment Secondary
Up to 42 months
Time-to-event endpoint analyzed with a stratified log-rank test and stratified hazard ratio.
Progression-Free Survival (PFS) by BICR Secondary
Up to 42 months
Time-to-event endpoint analyzed with blinded independent central review and a stratified log-rank framework.
Objective Response Rate (ORR) by Investigator Assessment Secondary
Up to 42 months
Binary endpoint analyzed using a Cochran-Mantel-Haenszel approach.
Objective Response Rate (ORR) by BICR Secondary
Up to 42 months
Binary endpoint analyzed using a Cochran-Mantel-Haenszel approach.
Clinical Benefit Rate (CBR) by Investigator Assessment Secondary
Up to 42 months
Binary endpoint analyzed using a Cochran-Mantel-Haenszel approach.
Clinical Benefit Rate (CBR) by BICR Secondary
Up to 42 months
Binary endpoint analyzed using a Cochran-Mantel-Haenszel approach.
Change From Baseline in EORTC-QLQ-C30 Domain Score Secondary
Cycle 5 Day 1; Cycle length = 21 days
Continuous outcome analyzed using a mixed-effects model for repeated measures (MMRM).

The registry reports 13 outcome measures and 11 statistical analyses. The ClinicalTrials.gov record includes the primary OS analysis, six additional efficacy analyses, and four health-related quality-of-life analyses.

6. Statistical Methodology

Overall survival and time-to-event analysis

Overall survival is a time-to-event endpoint because both the timing of death and whether death has occurred matter. The registry defines OS from randomization to death from any cause and specifies Kaplan-Meier estimation. The primary formal comparison used a log-rank test, with a stratified hazard ratio estimated from a Cox model adjusted for the randomization stratification factors reported in the electronic case report form (eCRF).

Conceptual survival quantity
S(t) = P(T > t)

Here, S(t) represents the probability of remaining event-free beyond time t. Kaplan-Meier estimation permits patients who have not experienced the event by their last available observation to contribute information through their censoring time.

Stratified log-rank test

The registry identifies the log-rank test as the primary OS comparison and the stratified log-rank test for the reported PFS analyses. Stratification allows the comparison to account for prespecified or recorded randomization factors rather than treating all participants as belonging to a single homogeneous risk set.

Cox proportional-hazards model

The primary OS hazard ratio and its confidence interval were calculated using a Cox model adjusted for randomization stratification factors as reported in the eCRF. The hazard ratio is therefore a model-based relative measure of the event hazard rather than a direct comparison of probabilities at one fixed time point.

Hazard-ratio interpretation
HR = 0.86  →  approximately 14% lower estimated hazard

This derived interpretation follows directly from 1 − 0.86 = 0.14. It does not mean that 14% of participants avoided death, nor does it imply that every participant experienced the same relative reduction.

Intention-to-treat analysis

The primary OS analysis used the Intent-to-treat (ITT) Analysis Set. The registry defines this population as all participants who had been randomized to the trial. An ITT analysis preserves treatment assignment as the basis for the comparison and therefore maintains the connection between the outcome analysis and the randomized design.

Cochran-Mantel-Haenszel analysis

The investigator-assessed and BICR response and clinical-benefit endpoints were analyzed using Cochran-Mantel-Haenszel statistics adjusted for randomization stratification factors. The effect measure reported was a stratified odds ratio.

MMRM for health-related quality of life

The quality-of-life analyses used an MMRM model. The registry text states that the model included all visits up to and including Cycle 5 Day 1. The analysis population was the HRQoL Evaluable Analysis Set with available data, defined in the registry as participants in the ITT Analysis Set who had an evaluable assessment.

7. Primary Result: Overall Survival

The registered primary endpoint was overall survival through up to 42 months. Participants in the ITT Analysis Set were analyzed. The comparison was sacituzumab govitecan-hziy versus treatment of physician's choice.

Stratified hazard ratio for death

0.86

95% CI: 0.73–1.02   ·   P = 0.0870

Two-sided confidence interval   ·   Superiority hypothesis

Primary endpointSGPhysician's choiceEffect estimate
Overall Survival ITT Analysis Set Stratified HR 0.86
95% CI 0.73–1.02
P = 0.0870
Clinical Biostats interpretation

What the estimate means: The stratified hazard ratio of 0.86 corresponds to an estimated hazard of death that is 14% lower in the SG group relative to the physician's-choice group under the fitted Cox model, after adjustment for the randomization stratification factors reported in the eCRF.

What it does not mean: A hazard ratio of 0.86 does not mean that 14% of patients benefited, that mortality was reduced by exactly 14% for every patient, or that the absolute probability of death was 14 percentage points lower.

What the confidence interval says: The two-sided 95% confidence interval extends from 0.73 to 1.02. It quantifies uncertainty around the estimated hazard ratio under the statistical model and sampling framework. Because the interval includes 1, the data are compatible with a range of relative effects that includes no hazard-ratio difference.

Why the P-value is not an effect size: The P-value of 0.0870 measures the strength of evidence against the null hypothesis under the specified testing framework; it does not describe the magnitude or clinical importance of the observed hazard ratio. The magnitude is described by the hazard ratio itself, while the confidence interval describes its precision.

Important caution: The Cox-model interpretation depends on the model's assumptions, including the proportional-hazards framework. Censoring, treatment discontinuation, subsequent treatment, and the exact composition of the analysis population can also affect the interpretation of a time-to-event estimate. The ClinicalTrials.gov record does not provide a separate proportional-hazards diagnostic or a time-varying treatment-effect analysis.

8. Secondary Time-to-Event Results

Progression-Free Survival by Investigator Assessment

Stratified hazard ratio

0.81

95% CI: 0.68–0.95   ·   P = 0.0117

Up to 42 months   ·   ITT Analysis Set

Clinical Biostats interpretation

The hazard ratio of 0.81 corresponds to an estimated 19% lower hazard of progression or death in the SG group under the reported stratified Cox-model framework. The 95% CI of 0.68–0.95 is entirely below 1, while the two-sided P-value is 0.0117.

The estimate describes a relative event rate over follow-up, not a difference in median PFS or a fixed percentage of participants who experienced benefit. The ClinicalTrials.gov record does not report median PFS or time-specific PFS probabilities, so those quantities are not substituted for the reported hazard ratio.

Progression-Free Survival by Blinded Independent Central Review

Stratified hazard ratio

0.86

95% CI: 0.72–1.03   ·   P = 0.1095

Up to 42 months   ·   ITT Analysis Set

Clinical Biostats interpretation

The BICR-assessed PFS hazard ratio of 0.86 corresponds to an estimated 14% lower hazard of progression or death in the SG group under the reported model. The 95% CI of 0.72–1.03 crosses 1, indicating appreciable uncertainty around the relative treatment effect. The P-value of 0.1095 is a measure of evidence under the specified testing framework, not a measure of effect size.

The difference between the investigator-assessed and BICR-assessed estimates illustrates why assessment method matters. These are two analyses of a related time-to-event outcome using different assessment frameworks; they should not be silently combined into a single estimate.

9. Secondary Response and Clinical-Benefit Results

EndpointMethodEffect measureEstimate95% CIP-value
ORR by Investigator Assessment Cochran-Mantel-Haenszel Stratified odds ratio 1.32 0.90–1.95 0.1594
ORR by BICR Cochran-Mantel-Haenszel Stratified odds ratio 1.84 1.24–2.73 0.0022
CBR by Investigator Assessment Cochran-Mantel-Haenszel Stratified odds ratio 1.67 1.19–2.34 0.0032
CBR by BICR Cochran-Mantel-Haenszel Stratified odds ratio 1.68 1.19–2.37 0.0034

How to interpret the odds ratios

The ORR and CBR analyses use odds ratios rather than hazard ratios because these endpoints are binary outcomes. An odds ratio of 1.84, for example, means that the estimated odds of the response outcome were 1.84 times as high in the SG group as in the physician's-choice group after adjustment for the reported randomization stratification factors.

An odds ratio is not the same as a risk ratio or a percentage-point difference. Without the underlying response counts or arm-specific response probabilities in the ClinicalTrials.gov record, it would not be appropriate to convert these odds ratios into absolute response rates.

Clinical Biostats interpretation

The investigator-assessed ORR analysis produced a stratified odds ratio of 1.32 with a 95% CI of 0.90–1.95 and P = 0.1594. The BICR analysis produced a stratified odds ratio of 1.84 with a 95% CI of 1.24–2.73 and P = 0.0022.

The two response analyses therefore give different estimates and levels of statistical evidence. That difference should be reported rather than averaged or treated as an inconsistency that can be resolved without knowing more about the assessment data. The BICR result has a confidence interval entirely above 1, while the investigator-assessed result has a confidence interval that includes 1.

Clinical Benefit Rate

Clinical Biostats interpretation

The investigator-assessed CBR analysis reported a stratified odds ratio of 1.67 (95% CI 1.19–2.34; P = 0.0032). The BICR analysis reported a stratified odds ratio of 1.68 (95% CI 1.19–2.37; P = 0.0034).

The close numerical similarity of the two CBR estimates is descriptive of these two reported analyses. It does not by itself establish that the two assessment approaches are interchangeable, because the underlying classifications and assessment processes are different.

10. Health-Related Quality of Life Results

The registry reports four secondary analyses of change from baseline in EORTC-QLQ-C30 domain scores at Cycle 5 Day 1, with a cycle length of 21 days. These analyses used MMRM and compared SG with treatment of physician's choice using least-square mean differences.

Domain / analysisMethodLS mean difference95% CIP-value
Physical Functioning MMRM 1.1 -2.5–4.7 0.5638
Global Health Status / QoL MMRM 0.0 -3.9–3.9 0.9944
Pain MMRM -2.7 -7.8–2.4 0.3007
Fatigue MMRM -2.0 -6.6–2.6 0.3871
Clinical Biostats interpretation

The MMRM estimates are between-group differences in least-square mean change from baseline at Cycle 5 Day 1, not hazard ratios or odds ratios. For example, the physical-functioning estimate of 1.1 indicates a modeled SG-versus-physician's-choice difference of 1.1 score units in change from baseline at the specified time point.

The 95% confidence intervals for all four reported analyses include 0. The corresponding P-values range from 0.3007 to 0.9944. These P-values should not be interpreted as measures of the magnitude of the quality-of-life differences; the estimates and confidence intervals provide the direct information about the size and precision of the modeled differences.

11. Safety

The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm using affected participants over participants at risk. These are the safety figures available for this page.

Safety measureSacituzumab Govitecan-hziyTreatment of Physician's Choice
Serious adverse events 185/349 110/337

The denominators in these safety results are the reported numbers at risk for the respective arms. They should not be replaced by the overall enrollment of 711, and the ClinicalTrials.gov record does not provide sufficient detail to construct a broader adverse-event table.

Safety interpretation: serious adverse events are a safety outcome and should be interpreted separately from the efficacy endpoints. The ClinicalTrials.gov record does not specify event severity distributions beyond the serious-adverse-event classification, individual event types, attribution, exposure duration, or discontinuation rates. Those quantities are therefore not inferred here.

12. Analysis Populations and Why They Matter

Analysis populationDefinition / role in the ClinicalTrials.gov record
Intent-to-treat (ITT) Analysis Set Included all participants who had been randomized to the trial; used for the primary OS analysis and the reported efficacy analyses.
HRQoL Evaluable Analysis Set Participants in the ITT Analysis Set who had an evaluable assessment; used for the registry-reported EORTC-QLQ-C30 MMRM analyses.
Safety populations Serious adverse events are reported using arm-specific numbers at risk: 349 for SG and 337 for physician's choice.

The distinction between populations is statistically important. The ITT population preserves the randomized comparison, whereas an evaluable quality-of-life population requires observed assessment data. These populations answer related but not identical questions.

13. Statistical Methods Explained

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

Overall survival records both whether death occurs and when it occurs. A log-rank test is designed for comparing time-to-event distributions while accounting for differing follow-up and right censoring. In this trial, the primary OS method was reported as a log-rank test, with a stratified hazard ratio obtained from a Cox model.

What does the OS hazard ratio of 0.86 mean?

A hazard ratio of 0.86 means that the estimated instantaneous hazard of death was 0.86 times that in the comparator group under the fitted model. Equivalently, 1 − 0.86 = 0.14, or a 14% lower estimated hazard. It is not a 14-percentage-point reduction in mortality and does not mean that 14% of participants benefited.

Why is the confidence interval important?

The 95% CI of 0.73–1.02 shows the uncertainty around the estimated OS hazard ratio. A point estimate alone can give a false impression of precision. The confidence interval shows that the observed estimate is compatible with a range of underlying relative effects, including values slightly above 1.

Why use the Cochran-Mantel-Haenszel test for ORR and CBR?

ORR and CBR are binary outcomes, so a time-to-event method is not required for the basic comparison. The Cochran-Mantel-Haenszel approach allows the analysis to incorporate randomization stratification factors while estimating a stratified odds ratio and its associated confidence interval and P-value.

Why is an odds ratio not the same as a response-rate difference?

Odds are defined as probability divided by one minus probability. An odds ratio therefore compares odds, not probabilities directly. An odds ratio of 1.84 means the odds are 1.84 times as large; it does not by itself tell us the absolute difference in response percentages.

Why was MMRM used for the quality-of-life analyses?

Quality-of-life scores are repeatedly measured over visits. MMRM is designed for longitudinal continuous outcomes and can model the relationship between repeated observations rather than reducing each participant to a single measurement. In this registry analysis, the model included visits through Cycle 5 Day 1.

Why does investigator assessment versus BICR matter?

Investigator assessment and blinded independent central review are different measurement processes. In an open-label trial, investigator assessment can potentially be influenced by knowledge of treatment assignment. BICR introduces an independent review process that can provide a distinct assessment of progression or response. The two sets of estimates should therefore be reported separately.

14. Understanding the Primary P-value

Primary OS test
HR = 0.86   |   95% CI 0.73–1.02   |   P = 0.0870

The three quantities answer different questions. The hazard ratio describes the estimated relative treatment effect, the confidence interval describes statistical precision, and the P-value describes evidence against the null hypothesis under the specified testing framework.

A P-value should not be converted into a probability that the treatment has no effect. Likewise, a P-value of 0.0870 does not mean that the treatment effect has an 8.70% probability of being due to chance. Those interpretations confuse the frequentist definition of a P-value with a posterior probability.

A useful reading sequence

For a time-to-event result, read the hazard ratio first to understand the estimated direction and relative magnitude; inspect the 95% confidence interval to understand precision and whether it includes the null value of 1; then consider the P-value within the prespecified hypothesis-testing framework. Finally, examine the analysis population, censoring, stratification, assessment method, and proportional-hazards assumptions before drawing broader conclusions.

15. Stratification and Covariate Adjustment

The analyses posted on ClinicalTrials.gov repeatedly state that effect estimates were adjusted for randomization stratification factors. For the OS analysis, the Cox model used the randomization stratification factors as reported in the eCRF. The PFS analyses used the same type of adjustment, while the ORR and CBR analyses used Cochran-Mantel-Haenszel statistics adjusted for randomization stratification factors.

Why stratify?

Stratification can account for important factors used in the randomized design when comparing outcomes between treatment groups.

Why adjust the analysis?

The adjusted estimate is aligned with the trial's stratified design rather than treating the randomized population as completely unstructured.

The ClinicalTrials.gov record does not identify the individual randomization stratification factors. They should therefore not be reconstructed from external publications or assumed from common urothelial-cancer trial designs.

16. Multiplicity and Multiple Secondary Endpoints

The registry contains one registered primary endpoint and multiple secondary endpoints, including two PFS analyses, two ORR analyses, two CBR analyses, and four quality-of-life analyses in the ClinicalTrials.gov record.

Endpoint familyReported analysesStatistical method
Overall survival1 primary analysisLog-rank test; stratified Cox hazard ratio
Progression-free survival2 secondary analysesStratified log-rank test; stratified Cox hazard ratio
Objective response rate2 secondary analysesCochran-Mantel-Haenszel; stratified odds ratio
Clinical benefit rate2 secondary analysesCochran-Mantel-Haenszel; stratified odds ratio
Quality of life4 secondary analysesMMRM; LS mean difference
Multiplicity caution: the ClinicalTrials.gov record identifies the endpoint roles and individual P-values but do not specify an alpha-allocation or multiplicity-adjustment procedure for the collection of secondary analyses. Therefore, the individual P-values should be reported as reported in the registry rather than assigning them a confirmatory interpretation that is not documented in the ClinicalTrials.gov record.

17. Censoring and Proportional-Hazards Considerations

Because OS and PFS are time-to-event outcomes, some participants may contribute follow-up without experiencing the event during the observation period. Kaplan-Meier methods account for right censoring under their usual assumptions, while Cox-model hazard ratios summarize relative event hazards under a proportional-hazards framework.

The primary OS result should therefore not be read as though it were a simple comparison of two proportions. A hazard ratio is a model-based summary of event rates over time. If the relative hazard changes materially during follow-up, a single hazard ratio may not fully describe the treatment-effect pattern.

Registry-data limitation: the ClinicalTrials.gov record does not report a diagnostic assessment of the proportional-hazards assumption, time-specific OS probabilities, median OS, or the shape of the Kaplan-Meier curves. Those quantities are not reconstructed here.

18. Results Summary

EndpointRoleEstimate95% CIP-value
Overall SurvivalPrimaryHR 0.860.73–1.020.0870
PFS, Investigator AssessmentSecondaryHR 0.810.68–0.950.0117
PFS, BICRSecondaryHR 0.860.72–1.030.1095
ORR, Investigator AssessmentSecondaryOR 1.320.90–1.950.1594
ORR, BICRSecondaryOR 1.841.24–2.730.0022
CBR, Investigator AssessmentSecondaryOR 1.671.19–2.340.0032
CBR, BICRSecondaryOR 1.681.19–2.370.0034
Physical FunctioningSecondaryLS mean difference 1.1-2.5–4.70.5638
Global Health Status / QoLSecondaryLS mean difference 0.0-3.9–3.90.9944
PainSecondaryLS mean difference -2.7-7.8–2.40.3007
FatigueSecondaryLS mean difference -2.0-6.6–2.60.3871

This table intentionally keeps effect measures in their native statistical forms. Hazard ratios, odds ratios, and mean differences cannot be interpreted as interchangeable quantities.

19. Limitations

20. Why This Trial Matters Statistically

TROPiCS-04 provides a useful teaching example because the registry combines several common clinical-trial analysis frameworks within a single randomized phase 3 study. The primary endpoint is time-to-event, while secondary outcomes span survival, binary response, clinical benefit, and repeated quality-of-life measurements.

ConceptHow it appears in TROPiCS-04
RandomizationParticipants were randomized in a two-arm parallel design.
Intention-to-treat analysisThe primary OS analysis used all randomized participants in the ITT Analysis Set.
Kaplan-Meier estimationThe registered OS endpoint specifies Kaplan-Meier estimates.
Log-rank testThe primary OS comparison used a log-rank test.
Stratified log-rank testThe registry-reported PFS analyses used stratified log-rank tests.
Hazard ratioOS and PFS were summarized with stratified hazard ratios.
Cox modelThe OS hazard ratio and confidence interval were calculated using a Cox model adjusted for randomization stratification factors.
Cochran-Mantel-Haenszel testORR and CBR analyses used Cochran-Mantel-Haenszel statistics.
Odds ratioBinary efficacy outcomes were summarized using stratified odds ratios.
MMRMEORTC-QLQ-C30 change-from-baseline analyses used mixed-effects models for repeated measures.
Open-label assessmentThe trial had no masking, while some endpoints were separately evaluated by BICR.

21. Related Tutorials

Learn more about the methods used in this trial:

22. Related Statistical Calculators

23. Sources

Continue through Clinical Biostats

Explore statistical tutorials and calculators related to survival analysis, categorical outcomes, confidence intervals, and repeated-measures methods.

24. Record Summary

TROPiCS-04 is a randomized phase 3 trial with 711 participants and a single registered primary time-to-event endpoint: overall survival through up to 42 months. The primary analysis used the ITT population, a log-rank test, and a stratified Cox-model hazard ratio. The reported OS estimate was 0.86, with a two-sided 95% CI of 0.73–1.02 and P = 0.0870.

The secondary analyses illustrate several additional statistical frameworks. Investigator-assessed PFS had a stratified HR of 0.81, while BICR-assessed PFS had a stratified HR of 0.86. Binary ORR and CBR outcomes were analyzed with Cochran-Mantel-Haenszel methods and stratified odds ratios. Quality-of-life outcomes were analyzed using MMRM and reported as least-square mean differences at Cycle 5 Day 1.

The most useful statistical reading of this trial therefore combines the effect measure, confidence interval, P-value, analysis population, assessment method, and model assumptions. These elements provide more information than any single number in isolation.

Clinical Biostats methodology: This page separates registry-reported numerical results from statistical interpretation. Where the ClinicalTrials.gov record does not provide a quantity such as median survival, absolute response rates, subgroup estimates, or individual stratification factors, that quantity is not reconstructed from outside sources.