← Clinical Trials
Advanced Renal Cell Carcinoma Phase 3 Randomized NCT04338269

CONTACT-03: Complete Statistical Analysis of Atezolizumab + Cabozantinib in Advanced Renal Cell Carcinoma

An independent statistical review of the randomized phase 3 CONTACT-03 trial comparing atezolizumab plus cabozantinib with cabozantinib alone in participants with advanced renal cell carcinoma after immune checkpoint inhibitor treatment.

Trial start: July 28, 2020  ·  Primary completion: January 3, 2023  ·  Status: Terminated
Independent analysis: This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results are restricted to the statistical analyses reported in the ClinicalTrials.gov record. Where the registry does not provide a quantity, it is not introduced here.

1. Trial at a Glance

CONTACT-03 was a randomized, parallel, open-label phase 3 trial evaluating atezolizumab in combination with cabozantinib versus cabozantinib alone in participants with advanced renal cell carcinoma after immune checkpoint inhibitor treatment.

522
Enrollment
Randomized trial
2
Treatment arms
Parallel design
1.03
IRF-PFS HR
95% CI 0.83–1.28
0.94
OS HR
95% CI 0.70–1.27
FeatureCONTACT-03
PhasePhase 3
ConditionCarcinoma, Renal Cell
PopulationParticipants with advanced renal cell carcinoma after immune checkpoint inhibitor treatment
DesignRandomized, parallel, open-label
AllocationRandomized
Primary purposeTreatment
Primary endpointsIndependent-review-facility progression-free survival and overall survival
Enrollment522
Lead sponsorHoffmann-La Roche
Sponsor typeIndustry
ClinicalTrials.govNCT04338269

2. Clinical Question

The trial evaluated whether adding atezolizumab to cabozantinib produced different time-to-event outcomes than cabozantinib alone in participants with advanced renal cell carcinoma after immune checkpoint inhibitor treatment.

Population

Participants with advanced renal cell carcinoma after immune checkpoint inhibitor treatment.

Intervention

Atezolizumab in combination with cabozantinib.

Comparator

Cabozantinib alone.

Primary question

How do the randomized groups compare for independent-review-facility progression-free survival and overall survival?

3. Trial Design

01
Randomize522 participants
02
Assign2 treatment arms
03
TreatCombination vs control
04
AssessPFS / OS / response
05
AnalyzeITT and endpoint-specific populations
INTERVENTION

Atezolizumab + Cabozantinib

  • Atezolizumab
  • Cabozantinib
CONTROL

Cabozantinib

  • Cabozantinib alone
Allocation
Randomized
Design model
Parallel
Masking
None
Primary purpose
Treatment

The trial began on 2020-07-28 and had a primary completion date of 2023-01-03. The registry record is listed as TERMINATED. Enrollment was 522 participants.

4. Endpoints

EndpointRegistered definition / time frameEndpoint type
Progression Free Survival (PFS) as Assessed by an Independent Review Facility (IRF) (IRF-PFS) According to RECIST v1.1 From randomization to the first occurrence of PD according to RECIST v1.1, or death from any cause, whichever occurred first (up to 2 years 5 months) Time-to-event
Overall Survival (OS) From randomization to death due to any cause (up to 2 years 5 months). Time-to-event

IRF-PFS

Overall Survival

Overall survival was defined as the time from randomization to death due to any cause. Participants who were not reported as having died at the date of analysis were censored at the date when they were last known to be alive. Participants without post-baseline information were censored at the date of randomization. The registry states that OS was estimated using the Kaplan-Meier method.

Endpoint perspective: Both registered primary endpoints are time-to-event outcomes. That means the analysis must account not only for whether an event occurred, but also for the amount of follow-up available for each participant and for right-censoring.

5. Analysis Populations and Stratification

The primary efficacy analyses used the intention-to-treat population. The registry defines this population as all randomized participants, whether or not the participant received the assigned treatment.

Analysis populationRegistry definition / role
ITT populationIncluded all randomized participants, whether or not the participant received the assigned treatment. Used for the primary PFS and OS analyses.
ORR-evaluable populationIncluded all randomized participants with measurable disease at baseline.
DOR-evaluable populationIncluded all randomized participants with measurable disease at baseline who experienced a confirmed objective response (CR or PR).

The statistical analyses were described as stratified where applicable. The registry notes that stratified hazard ratios were estimated by Cox regression. If at least one stratum had fewer than 10 events at the time of analysis, the stratification factor containing the level with the smallest number of participants was removed from the stratified analysis.

Why the ITT population matters: An ITT analysis preserves the randomized comparison. Participants remain associated with the treatment assignment to which they were randomized even if they do not receive the assigned treatment. This is particularly important when interpreting a randomized efficacy comparison because departures from assigned treatment can otherwise alter the meaning of the comparison.

6. Statistical Methodology

Log-rank test for time-to-event endpoints

The registry reports the log-rank test for the primary PFS and OS analyses and for the secondary time-to-event analyses. The log-rank framework compares the observed pattern of event occurrence between treatment groups across follow-up while accounting for censoring.

For a time-to-event endpoint, the comparison is therefore not equivalent to simply comparing the proportion of participants who experienced an event at a single fixed time point. Participants contribute information over the period during which they remain under observation.

Cox proportional-hazards model

Hazard ratios were estimated using Cox regression. For the primary stratified analyses, the registry identifies a stratified Cox proportional-hazards model. For the corresponding unstratified analyses, the registry identifies a Cox proportional-hazards model without stratification.

Conceptual hazard-ratio interpretation
HR = estimated instantaneous event rate in one group ÷ estimated instantaneous event rate in the reference group

A hazard ratio below 1 indicates a lower estimated instantaneous event rate in the numerator treatment group relative to the reference group under the fitted model. A hazard ratio of 1 corresponds to equal estimated hazards.

Stratified analysis

Stratification allows the time-to-event comparison to account for the stratification structure used in the analysis. The registry specifically reports stratified and unstratified Cox estimates for the primary endpoints, allowing the reader to see how the estimated treatment comparison changes when the analysis is performed with or without the specified stratification.

Cochran-Mantel-Haenszel test

The investigator-assessed and independent-review-facility objective response rate analyses used the Cochran-Mantel-Haenszel test. This is a stratified method for comparing categorical outcomes while accounting for the analysis strata.

Odds ratio

For the binary objective response endpoints, the reported effect measure was an odds ratio. The registry states that the 95% confidence interval for the odds ratio was constructed using the Wald method.

Kaplan-Meier estimation

The registry explicitly states that overall survival was estimated using the Kaplan-Meier method. Kaplan-Meier estimation provides an estimate of the probability of remaining event-free through time while accounting for participants whose event status is censored during follow-up.

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

where di is the number of events at event time ti and ni is the number at risk immediately before that time.

7. Primary Results: Independent-Review-Facility PFS

The primary IRF-assessed PFS analysis used the ITT population and compared cabozantinib alone with atezolizumab plus cabozantinib. The reported method was the log-rank test, with hazard ratios estimated by Cox regression.

AnalysisGroupsHR95% CIP-value
Stratified primary analysis Cabozantinib vs Atezolizumab + Cabozantinib 1.03 0.83–1.28 0.7844
Unstratified primary analysis Cabozantinib vs Atezolizumab + Cabozantinib 1.04 0.84–1.29 0.7195

Primary IRF-PFS estimate

HR 1.03

95% CI: 0.83–1.28   ·   P = 0.7844

Stratified Cox proportional-hazards model; ITT population.

Clinical Biostats interpretation

The stratified hazard ratio of 1.03 means that, under the fitted Cox model, the estimated instantaneous rate of progression or death for the cabozantinib group was approximately 1.03 times that of the atezolizumab-plus-cabozantinib group in this analysis. Equivalently, the estimate is very close to 1.

The estimate does not mean that 3% more participants progressed or died, nor does it describe an individual's probability of progression or death. A hazard ratio is a relative time-to-event measure rather than an absolute event probability.

The 95% confidence interval of 0.83–1.28 describes statistical uncertainty around the estimated hazard ratio under the model and sampling framework. It spans 1, so the interval includes both a lower estimated hazard and a higher estimated hazard for cabozantinib relative to the combination.

The P-value of 0.7844 is a measure used in the hypothesis-testing framework represented by this analysis; it is not a measure of effect size. A P-value should not be interpreted as the probability that one treatment is effective or ineffective, and it does not quantify the magnitude or clinical importance of the treatment difference.

The analysis was based on the ITT population and used stratified Cox regression. Interpretation of a single hazard ratio also depends on the proportional-hazards model being an appropriate description of the treatment effect over time. The registry does not state a non-inferiority margin or identify this analysis as a non-inferiority test.

Stratified versus unstratified PFS analysis

The unstratified analysis produced an HR of 1.04 with a 95% CI of 0.84–1.29 and a P-value of 0.7195. The close numerical similarity between the stratified and unstratified estimates is descriptive; it does not establish that stratification is unnecessary in other settings.

8. Primary Results: Overall Survival

Overall survival was the second registered primary endpoint. The analysis population was the ITT population, and the registry reports both stratified and unstratified Cox estimates.

AnalysisGroupsHR95% CIP-value
Stratified primary analysis Cabozantinib vs Atezolizumab + Cabozantinib 0.94 0.70–1.27 0.6902
Unstratified primary analysis Cabozantinib vs Atezolizumab + Cabozantinib 0.96 0.71–1.29 0.7853

Primary OS estimate

HR 0.94

95% CI: 0.70–1.27   ·   P = 0.6902

Stratified Cox proportional-hazards model; ITT population.

Clinical Biostats interpretation

The stratified OS hazard ratio of 0.94 means that the fitted model estimated the instantaneous rate of death in the cabozantinib group at approximately 0.94 times that of the atezolizumab-plus-cabozantinib group during the analyzed follow-up.

This does not mean that the cabozantinib group had exactly 6% lower mortality, that 6% of participants benefited, or that an individual participant's probability of death was reduced by 6%. The hazard ratio is a model-based relative time-to-event measure.

The 95% CI of 0.70–1.27 indicates uncertainty around the estimated hazard ratio. It includes 1 and therefore encompasses values corresponding to a lower estimated hazard and values corresponding to a higher estimated hazard for cabozantinib relative to the combination.

The P-value of 0.6902 does not measure the size of the treatment effect. It summarizes the compatibility of the observed data with the null hypothesis represented by the statistical test; it should not be interpreted as the probability that the treatment effect is zero.

The analysis was conducted in the ITT population and used a stratified Cox proportional-hazards model. As with other Cox analyses, interpretation of one summary hazard ratio depends on the model's assumptions, including the relevance of a proportional-hazards representation over the analysis period.

Unstratified OS analysis

The unstratified Cox analysis produced an HR of 0.96, with a 95% CI of 0.71–1.29 and a P-value of 0.7853. The direction and magnitude of the point estimate were similar to the stratified analysis.

9. Secondary Results: Investigator-Assessed PFS

Investigator-assessed PFS was a secondary time-to-event endpoint. The analysis used the ITT population and RECIST v1.1. The registry reports both stratified and unstratified Cox analyses.

AnalysisHR95% CIP-valueMethod
Stratified 1.03 0.83–1.27 0.8037 Log-rank; stratified Cox
Unstratified 1.03 0.83–1.27 0.7894 Log-rank; unstratified Cox

The stratified and unstratified point estimates were both 1.03. Their confidence intervals were also 0.83–1.27, while the reported P-values were 0.8037 and 0.7894, respectively.

Clinical Biostats interpretation

The investigator-assessed PFS estimates are close to 1, and the corresponding confidence intervals include 1. These results describe the treatment comparison reported by the registry; they should not be converted into an absolute probability of progression or death without the underlying survival estimates.

The use of investigator assessment also illustrates why endpoint definition matters. IRF-PFS and investigator-assessed PFS are related endpoints but are not interchangeable simply because both use RECIST v1.1.

10. Secondary Results: Objective Response Rate

Objective response rate was evaluated in the ORR-evaluable population, defined in the registry as all randomized participants with measurable disease at baseline. The binary endpoint was analyzed using the Cochran-Mantel-Haenszel test, with odds ratios and Wald 95% confidence intervals.

EndpointOR95% CIP-value
Investigator-assessed ORR (INV-ORR) 0.87 0.61–1.24 0.4306
IRF-assessed ORR (IRF-ORR) 1.00 0.70–1.43 0.9893

Investigator-assessed ORR

Odds ratio

0.87

95% CI: 0.61–1.24   ·   P = 0.4306

Cochran-Mantel-Haenszel analysis; ORR-evaluable population.

Clinical Biostats interpretation

An odds ratio of 0.87 compares the odds of objective response between the groups as specified in the registry analysis. It is not a risk ratio and should not be interpreted as saying that the probability of response was 13% lower.

The 95% CI of 0.61–1.24 spans 1, indicating uncertainty that includes odds ratios below and above the null value. The P-value of 0.4306 is not an effect-size measure; it does not tell us how large or clinically important the response difference is.

IRF-assessed ORR

Odds ratio

1.00

95% CI: 0.70–1.43   ·   P = 0.9893

Cochran-Mantel-Haenszel analysis; ORR-evaluable population.

Clinical Biostats interpretation

An odds ratio of 1.00 is exactly the null value for an odds-ratio comparison. The reported 95% CI of 0.70–1.43 shows that the data are compatible with odds ratios below and above 1.

The P-value of 0.9893 should not be read as a probability that the two treatments are identical. It is a hypothesis-test quantity, not a posterior probability of equality and not a measure of clinical equivalence.

11. Secondary Results: Duration of Response

Duration of response was analyzed among participants in the DOR-evaluable population: randomized participants with measurable disease at baseline who experienced a confirmed objective response (CR or PR).

EndpointAnalysisHR95% CIP-value
Investigator-assessed DOR Stratified 0.70 0.47–1.05 0.0816
Investigator-assessed DOR Unstratified 0.72 0.49–1.08 0.1099
IRF-assessed DOR Stratified 1.04 0.70–1.54 0.8354
IRF-assessed DOR Unstratified 1.05 0.71–1.55 0.8159

The investigator-assessed DOR analyses produced hazard ratios below 1, while the IRF-assessed DOR analyses produced hazard ratios close to 1. The confidence intervals for all four analyses include 1.

Clinical Biostats interpretation

DOR is conditional on having achieved a confirmed objective response, so it addresses a different population and question from PFS and OS. The DOR population is therefore not equivalent to the ITT population used for the primary endpoints.

The stratified investigator-assessed DOR HR of 0.70 has a 95% CI of 0.47–1.05 and a P-value of 0.0816. The estimate alone should not be treated as proof of a treatment effect because the confidence interval includes 1 and the registry does not identify a non-inferiority or equivalence framework for this analysis.

The IRF-assessed DOR HR of 1.04 has a 95% CI of 0.70–1.54 and a P-value of 0.8354. Again, the confidence interval represents uncertainty around the estimated relative event rate and is not a range of individual patient outcomes.

12. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm as the number affected among the number at risk.

Treatment groupSerious adverse eventsAffected / at risk
Cabozantinib (Control) Serious adverse events 95 / 256
Atezolizumab + Cabozantinib Serious adverse events 141 / 262

These are the serious-adverse-event counts reported in the ClinicalTrials.gov record. The page does not convert them into percentages because the instruction for this record is to preserve reported numbers rather than derive additional numerical results.

Safety versus efficacy: Serious adverse events and time-to-event efficacy endpoints answer different questions. A safety comparison should not be folded into an efficacy hazard ratio, and an efficacy result should not be interpreted as a summary measure of overall benefit-risk without a separate clinical assessment.

13. Statistical Methods Explained

Why was a log-rank test used?

The primary PFS and OS endpoints are time-to-event outcomes. The log-rank test is designed to compare survival-type event-time distributions between randomized groups while incorporating the timing of events and allowing for censoring. This is more informative for a time-to-event endpoint than a simple comparison of event proportions at one arbitrary time.

What does a hazard ratio of 1.03 mean?

A hazard ratio of 1.03 means that the fitted model estimated the instantaneous event rate in the cabozantinib group at about 1.03 times the event rate in the atezolizumab-plus-cabozantinib group for that analysis. It does not mean that 3% of participants had an event, that the absolute risk differed by 3%, or that each participant experienced a 3% difference in risk.

Why are confidence intervals reported with hazard ratios?

The point estimate is only one estimate of the treatment contrast. The confidence interval communicates the statistical uncertainty around that estimate. For example, the stratified IRF-PFS estimate was HR 1.03 with a 95% CI of 0.83–1.28. The interval therefore provides substantially more information than the point estimate alone.

Why is a P-value not an effect size?

A P-value is generated from a statistical testing framework. It depends on the observed data and the model used for the test, including the amount of information available. It does not directly quantify the magnitude of a treatment effect. Effect estimates such as hazard ratios and odds ratios, together with their confidence intervals, are the appropriate quantities for describing the estimated size and uncertainty of the comparison.

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

ORR is a binary endpoint, so a categorical-data method is appropriate. The Cochran-Mantel-Haenszel procedure provides a way to compare treatment groups while accounting for stratification rather than ignoring the stratified structure of the analysis.

Why does the analysis population change for DOR?

DOR is defined beginning at the date of first documented objective response. Participants who never experience a confirmed response do not enter that response-duration population. The registry therefore defines a DOR-evaluable population restricted to randomized participants with measurable disease at baseline who experienced a confirmed objective response. This makes DOR fundamentally different from an ITT endpoint such as OS.

Why report both stratified and unstratified analyses?

The registry reports both forms for the primary endpoints and several secondary time-to-event endpoints. The stratified analysis incorporates the specified stratification structure into the Cox model, whereas the unstratified analysis does not. Comparing them can be descriptively useful, but the prespecified analysis should remain the principal reference for interpreting the trial's formal statistical design.

14. Reading the Primary Results Together

EndpointPrimary stratified estimate95% CIP-valueAnalysis population
IRF-PFS HR 1.03 0.83–1.28 0.7844 ITT
OS HR 0.94 0.70–1.27 0.6902 ITT

The two primary endpoints point in different numerical directions: the IRF-PFS point estimate is 1.03, while the OS point estimate is 0.94. Both confidence intervals include 1. The appropriate interpretation is therefore to describe the estimates, their uncertainty, and their prespecified statistical context rather than treating a point estimate as a definitive description of every participant's experience.

Relative measure

The hazard ratio describes a relative comparison of event rates under the fitted Cox model.

Absolute measure

The ClinicalTrials.gov record does not report median PFS, median OS, or fixed-time survival probabilities, so those quantities are not presented here.

Precision

The confidence interval describes uncertainty around the estimated treatment contrast.

Statistical evidence

The P-value belongs to the specified hypothesis-testing framework and should not be interpreted as an effect-size measure.

15. Stratification and Small Event Counts

The registry provides an important implementation detail for the stratified Cox analyses: if at least one stratum had <10 events at the time of analysis, the stratification factor containing the level with the smallest number of participants was removed from the stratified analysis.

This rule illustrates a practical issue in stratified survival analysis. Stratification can be useful because it allows baseline hazard functions to differ across strata, but very sparse event information within a stratum can make the analysis unstable or less informative. The registry therefore specifies a rule for modifying the stratification structure when event counts are small.

Why this matters
Stratified Cox model → account for analysis strata → monitor sparse-event strata

The treatment effect remains estimated through the Cox model, but the precise stratification structure used at analysis can depend on the observed event information under the registry's stated rule.

16. Censoring and Time-to-Event Interpretation

The OS registry definition explicitly describes censoring. Participants who were alive at the analysis date were censored at the date they were last known to be alive. Participants without post-baseline information were censored at randomization.

Censoring is not the same as an event. A censored participant contributes information up to the censoring time, after which the participant's event status is not observed within the analysis window. Kaplan-Meier estimation and Cox regression are designed to incorporate this structure.

Important distinction: a censored observation does not mean that the participant experienced the endpoint at the censoring time. It means that follow-up for the endpoint ended there under the analysis rules.

17. Odds Ratios and Time-to-Event Hazard Ratios Are Not Interchangeable

CONTACT-03 uses two different effect measures because the trial contains different types of endpoints.

Endpoint typeEffect measureInterpretation
Time-to-eventHazard ratioRelative comparison of estimated instantaneous event rates under the Cox model.
Binary objective responseOdds ratioRelative comparison of the odds of response.

An odds ratio of 1.00 and a hazard ratio of 1.00 both correspond to their respective null values, but they describe different statistical quantities. A hazard ratio should not be interpreted as an odds ratio, and an odds ratio should not be interpreted as a hazard ratio.

18. Confidence Intervals and the Null Value

For the reported hazard ratios, the null value is 1. The same is true for the reported odds ratios. This makes the confidence intervals particularly useful for understanding the direction and precision of each estimate.

AnalysisEstimate95% CINull value contained?
IRF-PFS, stratifiedHR 1.030.83–1.28Yes
OS, stratifiedHR 0.940.70–1.27Yes
INV-ORROR 0.870.61–1.24Yes
IRF-ORROR 1.000.70–1.43Yes
INV-DOR, stratifiedHR 0.700.47–1.05Yes
IRF-DOR, stratifiedHR 1.040.70–1.54Yes

A confidence interval containing the null value does not prove that the treatments have identical effects. It indicates that the estimated treatment contrast is uncertain enough that the null value is included within the interval under the specified statistical framework.

19. Primary Endpoint Results: What They Do and Do Not Establish

IRF-PFS

The stratified HR of 1.03 with 95% CI 0.83–1.28 and P = 0.7844 is the reported primary IRF-PFS comparison. It does not provide median PFS, an absolute probability of progression, or an individual-level prediction.

Overall Survival

The stratified OS HR of 0.94 with 95% CI 0.70–1.27 and P = 0.6902 is the reported primary OS comparison. It does not provide a median survival time or a fixed-time survival probability because those quantities are not contained in the ClinicalTrials.gov record.

Precision versus significance

The confidence interval is useful for understanding the range of treatment-effect estimates compatible with the analysis. The P-value addresses a hypothesis-testing question. Neither quantity should be used alone as a substitute for the complete effect estimate, endpoint definition, analysis population, and study design.

20. Limitations

21. Why This Trial Matters Statistically

CONTACT-03 is a useful teaching example because the same randomized trial contains several common clinical-trial statistical structures: time-to-event primary endpoints, ITT analysis, stratified and unstratified Cox models, log-rank testing, binary response endpoints, Cochran-Mantel-Haenszel analysis, odds ratios, and endpoint-specific analysis populations.

ConceptHow it appears in CONTACT-03
RandomizationRandomized phase 3 parallel-group design with 522 participants enrolled.
Intention-to-treat analysisPrimary PFS and OS analyses included all randomized participants, whether or not they received assigned treatment.
Time-to-event endpointsIRF-PFS and OS were registered as primary time-to-event endpoints.
Kaplan-Meier estimationThe registry states that OS was estimated using the Kaplan-Meier method.
Log-rank testUsed for the reported primary PFS and OS comparisons and secondary time-to-event analyses.
Hazard ratioUsed for PFS, OS, and duration-of-response analyses.
Stratified analysisStratified Cox analyses were reported for multiple time-to-event endpoints.
Cochran-Mantel-Haenszel testUsed for investigator-assessed and IRF-assessed objective response rate.
Odds ratioReported for the binary ORR endpoints.
Confidence intervals95% two-sided confidence intervals accompanied all reported effect estimates.
Endpoint-specific populationsORR and DOR used evaluable populations defined separately from the ITT population.
Independent reviewPrimary PFS and one ORR/DOR analysis framework used independent review facility assessment.

22. Statistical Methods: A Practical Workflow

01
DefineEndpoint and event
02
RandomizePreserve comparison
03
FollowEvents and censoring
04
ModelLog-rank / Cox / CMH
05
InterpretEstimate + CI + P-value

The statistical workflow can be understood as a sequence rather than as a collection of isolated tests. First, the endpoint defines what constitutes an event. Second, randomization establishes the treatment comparison. Third, follow-up produces event and censoring information. Fourth, an analysis method appropriate to the endpoint is applied. Finally, the estimate, confidence interval, and P-value are interpreted together with the analysis population and design.

23. Results Summary

EndpointEffect measureEstimate95% CIP-value
Primary IRF-PFS, stratifiedHazard ratio1.030.83–1.280.7844
Primary IRF-PFS, unstratifiedHazard ratio1.040.84–1.290.7195
Primary OS, stratifiedHazard ratio0.940.70–1.270.6902
Primary OS, unstratifiedHazard ratio0.960.71–1.290.7853
Secondary INV-PFS, stratifiedHazard ratio1.030.83–1.270.8037
Secondary INV-PFS, unstratifiedHazard ratio1.030.83–1.270.7894
Secondary INV-ORROdds ratio0.870.61–1.240.4306
Secondary IRF-ORROdds ratio1.000.70–1.430.9893
Secondary INV-DOR, stratifiedHazard ratio0.700.47–1.050.0816
Secondary INV-DOR, unstratifiedHazard ratio0.720.49–1.080.1099
Secondary IRF-DOR, stratifiedHazard ratio1.040.70–1.540.8354
Secondary IRF-DOR, unstratifiedHazard ratio1.050.71–1.550.8159

24. Related Tutorials

Learn more about the methods used in this trial:

25. Related Calculators

26. Sources

Continue through the Clinical Biostats statistical library

Connect the endpoints and methods in this trial to deeper statistical tutorials and analysis tools.

27. Record Summary

CONTACT-03 provides a compact example of how a randomized phase 3 clinical trial can use different statistical methods for different endpoint types. The two primary endpoints were time-to-event outcomes analyzed in the ITT population using log-rank testing and Cox regression, with both stratified and unstratified hazard ratios reported. Secondary objective response endpoints were binary and were analyzed using the Cochran-Mantel-Haenszel test with odds ratios, while duration of response returned to a time-to-event framework in a response-evaluable population.

The most useful statistical reading therefore keeps several distinctions in view: time-to-event versus binary endpoints, hazard ratios versus odds ratios, ITT versus endpoint-specific evaluable populations, stratified versus unstratified models, and point estimates versus their confidence intervals and P-values. The ClinicalTrials.gov record reports results for each of these analyses but do not provide several quantities commonly used for a fuller clinical interpretation, including median survival estimates, fixed-time survival probabilities, subgroup results, crossover information, and a detailed multiplicity or interim-analysis plan.

Clinical Biostats methodology: The purpose of an independent trial-analysis page is to reconstruct the statistical structure of the reported evidence, explain what each method and estimate means, and keep reported results separate from assumptions or conclusions that are not supported by the ClinicalTrials.gov record.