← Clinical Trials
Triple-Negative Breast Cancer Phase 3 Completed NCT03125902

IMpassion131: Complete Statistical Analysis of Atezolizumab in Triple-Negative Breast Cancer

An independent statistical analysis of the randomized phase 3 IMpassion131 trial evaluating atezolizumab plus paclitaxel versus placebo plus paclitaxel in participants with previously untreated locally advanced or metastatic triple-negative breast cancer.

Trial period: 2017-08-25 to 2019-11-15  ·  Enrollment: 653  ·  Results posted: Yes
Scope of this record

This page separates reported trial results from statistical interpretation. All trial-specific numerical results and design details on this page are taken from the ClinicalTrials.gov record. The registry provides the official trial record.

Registry note: 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

IMpassion131 was a randomized, double-blind, phase 3, parallel-group trial evaluating atezolizumab and paclitaxel versus placebo and paclitaxel in participants with previously untreated locally advanced or metastatic triple-negative breast cancer.

653
Enrolled
2-arm phase 3 trial
3
Phase
Treatment purpose
0.82
Primary PFS HR
PD-L1-positive, stratified
0.86
Primary ITT PFS HR
Stratified analysis
FeatureIMpassion131
PhasePhase 3
ConditionTriple-Negative Breast Cancer
DesignRandomized, double-blind, parallel
AllocationRandomized
Primary purposeTreatment
Enrollment653
Arms2
Primary endpoints2 registered endpoints, both progression-free survival
Primary endpoint typeTime-to-event
Results postedYes
Statistical analyses posted15
Primary-endpoint analyses4
Lead sponsorHoffmann-La Roche
Sponsor typeIndustry
Trial statusCompleted

2. Clinical Question

The central question was whether adding atezolizumab to paclitaxel changed progression-free survival compared with placebo plus paclitaxel in participants with previously untreated locally advanced or metastatic triple-negative breast cancer.

Population

Participants with previously untreated locally advanced or metastatic triple-negative breast cancer.

Intervention

Atezolizumab, an engineered anti-PDL1 antibody, together with paclitaxel.

Comparator

Atezolizumab placebo together with paclitaxel.

Primary question

Does atezolizumab plus paclitaxel improve progression-free survival relative to placebo plus paclitaxel?

3. Trial Design

01
Randomize653 enrolled
02
Double-blindMasked treatment assignment
03
Parallel groups2 treatment arms
04
Assess PFSRECIST v1.1
05
Follow outcomesEfficacy and safety
ARM 1

Atezolizumab + paclitaxel

  • Atezolizumab
  • Paclitaxel
  • Randomized treatment assignment
ARM 2

Placebo + paclitaxel

  • Atezolizumab placebo
  • Paclitaxel
  • Randomized treatment assignment

The registry describes the study as randomized, double-blind, parallel, with treatment as the primary purpose. The ClinicalTrials.gov record does not provide an allocation ratio or randomized sample size for each arm, so no arm-specific enrollment numbers are inferred here.

4. Trial Timeline

2017-08-25 · Start

Trial initiation

The registry lists August 25, 2017 as the trial start date.

2019-11-15 · Primary completion

Primary completion

The registry lists November 15, 2019 as the primary completion date.

Completed · Results posted

Statistical results available

Results are posted in the registry, including 15 statistical analyses and four primary-endpoint analyses.

5. Primary Endpoints

The registry contains two primary endpoints. Both are progression-free survival endpoints assessed using RECIST v1.1, one in a PD-L1-positive subpopulation and one in the full ITT population.

EndpointPopulationTime frameType
Progression-Free Survival (PFS) Assessed Using RECIST v1.1 in the Subpopulation With PD-L1-Positive Tumour Status Participants in the ITT population whose PD-L1 status was IC1/2/3 at randomization From Day 1 to disease progression or death from any cause, assessed up to primary completion date (approximately 26 months) Time-to-event
Progression-Free Survival (PFS) Assessed Using RECIST v1.1 in the ITT Population All randomized participants, whether or not assigned study treatment was received From Day 1 to disease progression or death from any cause, assessed up to primary completion date (approximately 26 months) Time-to-event

The registry definition states that PFS is the time from randomization to the first occurrence of investigator-determined progression using RECIST v1.1 or death from any cause during the study, whichever occurs first. The registry definition describes progression using changes in the sum of diameters of target lesions, including a relative increase of at least 20% and an absolute increase of at least 5 mm.

Why the endpoint definition matters: PFS is not simply a measurement of tumor shrinkage. It is a time-to-event endpoint in which either documented progression or death can constitute the event. Participants who have not experienced an event at their relevant follow-up time contribute censored observations to the survival analysis.

6. Analysis Populations

PopulationRegistry definition / role
Intent-to-treat (ITT) All randomized participants, whether or not the assigned study treatment was received.
PD-L1-positive subpopulation Participants in the ITT population whose PD-L1 status was IC1/2/3 at the time of randomization.
Response-evaluable population Participants in the ITT population with measurable disease at baseline.
PRO-evaluable population Participants in the ITT population with baseline PRO assessment and at least one post-baseline PRO assessment in the questionnaire of interest.
Duration-of-response evaluable population Participants in the ITT population with measurable disease at baseline.
Confirmed-duration-of-response evaluable population Participants in the ITT population with measurable disease at baseline and with a confirmed objective response.

The distinction between these populations is statistically important. The primary ITT analysis preserves the randomized comparison, while secondary endpoints may be restricted to participants meeting additional assessment criteria. Such restrictions can change the population to which an estimate applies.

7. Primary Results: PD-L1-Positive PFS

The first primary endpoint evaluated progression-free survival in the PD-L1-positive subpopulation. The registry reports both a stratified and an unstratified log-rank analysis.

Stratified analysis

Hazard ratio for progression or death

0.82

95% CI: 0.60–1.12   ·   P = 0.2032

Comparison: Placebo + paclitaxel vs atezolizumab + paclitaxel.

FeatureReported analysis
EndpointPD-L1-positive PFS assessed using RECIST v1.1
Analysis populationPD-L1-positive subpopulation: ITT participants with PD-L1 status IC1/2/3 at randomization
MethodLog-rank test
AnalysisStratified
Effect measureHazard ratio
Estimate0.82
95% CI0.60–1.12
P-value0.2032
Hypothesis typeSuperiority
Clinical Biostats interpretation

An HR of 0.82 means that the estimated instantaneous rate of progression or death was approximately 18% lower with atezolizumab plus paclitaxel than with placebo plus paclitaxel, because 1 − 0.82 = 0.18.

The HR does not mean that 18% of participants avoided progression, that individual patients experienced exactly an 18% reduction in risk, or that median PFS was reduced or increased by 18%. It is a relative time-to-event measure.

The two-sided 95% CI of 0.60–1.12 describes the statistical uncertainty around the estimated HR. Because the interval includes 1, the data are compatible with both a lower and a higher instantaneous event rate under the model.

The P-value of 0.2032 addresses evidence against the null hypothesis used for the superiority analysis; it does not measure the size or clinical importance of the treatment effect. The registry reports the analysis as stratified but does not provide the specific stratification factors in the ClinicalTrials.gov record.

Unstratified analysis

Unstratified hazard ratio

0.84

95% CI: 0.62–1.14   ·   P = 0.2601

The unstratified estimate is close to the stratified estimate, but the two analyses are not interchangeable. Stratification can account for prespecified design factors when comparing survival experience, whereas an unstratified analysis treats the randomized groups without that adjustment.

Clinical Biostats interpretation

An HR of 0.84 corresponds to an estimated 16% lower instantaneous rate of progression or death in the atezolizumab-plus-paclitaxel group relative to the placebo-plus-paclitaxel group.

The 95% CI of 0.62–1.14 is wider than the point estimate itself can convey and includes 1. The P-value of 0.2601 should be interpreted as a measure of compatibility with the specified null hypothesis, not as a probability that the treatment has no effect.

8. Primary Results: ITT PFS

The second primary endpoint evaluated PFS in the full ITT population. The registry again reports both stratified and unstratified analyses.

Stratified analysis

Hazard ratio for progression or death

0.86

95% CI: 0.70–1.05   ·   P = 0.1343

Comparison: Placebo + paclitaxel vs atezolizumab + paclitaxel.

FeatureReported analysis
EndpointITT PFS assessed using RECIST v1.1
Analysis populationAll randomized participants, whether or not assigned study treatment was received
MethodLog-rank test
AnalysisStratified
Effect measureHazard ratio
Estimate0.86
95% CI0.70–1.05
P-value0.1343
Hypothesis typeSuperiority
Clinical Biostats interpretation

An HR of 0.86 corresponds to an estimated 14% lower instantaneous rate of progression or death with atezolizumab plus paclitaxel relative to placebo plus paclitaxel.

The 95% CI of 0.70–1.05 includes 1. The interval therefore does not establish a single direction of effect with the conventional confidence-interval interpretation. It also shows that the point estimate alone should not be treated as a precise measure of treatment effect.

The P-value of 0.1343 is evidence from the specified statistical test, not a measure of effect magnitude. It should be considered alongside the HR, confidence interval, analysis population, and trial design.

Unstratified analysis

Unstratified hazard ratio

0.86

95% CI: 0.70–1.05   ·   P = 0.1285

Clinical Biostats interpretation

The unstratified estimate is also 0.86, with a 95% CI of 0.70–1.05. Its P-value is 0.1285. The identical point estimate but slightly different P-value illustrates that the reported inferential result depends on the analysis framework even when the estimated HR itself is unchanged.

Neither the HR nor its P-value should be interpreted without considering censoring and the time-to-event structure. A Cox-type interpretation of an HR also requires attention to the proportional-hazards assumption when such a model is used; the ClinicalTrials.gov record identifies the effect measure as an HR but do not provide a proportional-hazards diagnostic.

9. Primary Endpoint Results in One View

Primary endpointAnalysisHR95% CIP-value
PD-L1-positive PFS Stratified 0.82 0.60–1.12 0.2032
PD-L1-positive PFS Unstratified 0.84 0.62–1.14 0.2601
ITT PFS Stratified 0.86 0.70–1.05 0.1343
ITT PFS Unstratified 0.86 0.70–1.05 0.1285

Across the four primary analyses reported in the registry, the HR estimates are below 1, while all four two-sided 95% confidence intervals include 1. The ClinicalTrials.gov record does not provide median PFS, event counts, Kaplan-Meier estimates, or additional time-specific PFS probabilities, so those quantities are not presented here.

Do not confuse an HR below 1 with a statistically conclusive treatment effect. A point estimate below 1 describes the direction and magnitude of the estimated relative hazard. The corresponding confidence interval and prespecified inferential framework determine how much uncertainty surrounds that estimate.

10. Secondary Endpoint Results: Overall Survival

Overall survival was reported as a secondary endpoint in both the PD-L1-positive subpopulation and the ITT population.

PopulationAnalysisHR95% CIP-value
PD-L1-positive Stratified 1.11 0.76–1.64 0.5798
PD-L1-positive Unstratified 1.18 0.80–1.72 0.4035
ITT Stratified 1.12 0.88–1.43 0.3425
ITT Unstratified 1.16 0.92–1.48 0.2166

The registry defines OS as time from Day 1 to death from any cause. The PD-L1-positive analysis was assessed up to 36 months, while the ITT analysis was assessed up to end of study, up to approximately 36 months.

Clinical Biostats interpretation

The OS HR estimates are above 1 in all four reported analyses. For example, the stratified ITT estimate of 1.12 corresponds to an estimated instantaneous death rate approximately 12% higher in the atezolizumab-plus-paclitaxel group relative to the placebo-plus-paclitaxel group under the hazard-ratio interpretation.

This does not mean that treatment caused 12% more deaths in every population or that 12% of participants experienced an adverse survival outcome. It is a relative model-based time-to-event estimate.

The 95% CI of 0.88–1.43 for the stratified ITT analysis includes 1, indicating substantial uncertainty about the direction and magnitude of the relative hazard. The P-value of 0.3425 does not measure the size of the observed estimate.

11. Secondary Endpoint Results: Time to Deterioration

The registry also reports time to deterioration in global health status/health-related quality of life in the PRO-evaluable population.

Time to deterioration in global health status / HRQoL

HR 0.94

95% CI: 0.71–1.24   ·   P = 0.6465

From Day 1 to deterioration, assessed up to 64 months.

The analysis used a log-rank test and was reported as stratified. The PRO-evaluable population consisted of participants in the ITT population with a baseline PRO assessment and at least one post-baseline PRO assessment in the questionnaire of interest.

Clinical Biostats interpretation

An HR of 0.94 is close to 1 and corresponds to an estimated 6% lower instantaneous rate of deterioration in the atezolizumab-plus-paclitaxel group under the hazard-ratio interpretation.

The 95% CI of 0.71–1.24 spans both sides of 1, so the point estimate should not be treated as a precise estimate of a treatment difference. The P-value of 0.6465 provides the inferential result reported for this secondary endpoint but does not quantify the clinical importance of the observed HR.

12. Secondary Endpoint Results: Objective Response

Objective response was evaluated as a binary endpoint using RECIST v1.1. The registry reports confirmed and unconfirmed response analyses in both the PD-L1-positive population and the response-evaluable population.

Response analysisPopulationMethodOR95% CIP-value
Confirmed, investigator-assessed PD-L1-positive Cochran-Mantel-Haenszel 1.44 0.87–2.37 0.1526
Unconfirmed, investigator-assessed PD-L1-positive Cochran-Mantel-Haenszel 1.40 0.85–2.31 0.1834
Confirmed, investigator-assessed Response-evaluable Cochran-Mantel-Haenszel 1.42 1.00–2.02 0.0513
Unconfirmed, investigator-assessed Response-evaluable Cochran-Mantel-Haenszel 1.30 0.93–1.81 0.1226

The response endpoints were assessed from Day 1 to progression, with assessment up to the primary completion date, approximately 26 months. The response-evaluable population consisted of ITT participants with measurable disease at baseline.

Clinical Biostats interpretation

An odds ratio of 1.42 means that the estimated odds of the specified confirmed objective-response outcome were 1.42 times as high in the atezolizumab-plus-paclitaxel group relative to the comparator group under the reported comparison.

An OR is not the same as a risk ratio or a difference in response percentages. The distinction becomes especially important when the outcome is not rare.

The 95% CI of 1.00–2.02 reaches 1.00 at its lower boundary. The associated P-value of 0.0513 should not be converted into an effect-size judgment. The estimate, confidence interval, endpoint definition, and analysis population provide complementary information.

13. Secondary Endpoint Results: Duration of Response

The registry contains two duration-of-response analyses, both time-to-event endpoints.

EndpointPopulationAnalysisHR95% CIP-value
Duration of Objective Response, unconfirmed DoR-evaluable population Unstratified log-rank 0.74 0.54–1.02 0.0641
Duration of Confirmed Response C-DoR-evaluable population Unstratified log-rank 0.62 0.42–0.90 0.01227

Both endpoints measure time from objective response to progression. The unconfirmed duration-of-response analysis was assessed up to primary completion date, approximately 26 months. The confirmed-duration-of-response analysis used participants with measurable disease at baseline and a confirmed objective response.

Clinical Biostats interpretation

For confirmed duration of response, the HR of 0.62 corresponds to an estimated 38% lower instantaneous rate of progression among the analyzed confirmed-response population, because 1 − 0.62 = 0.38.

This estimate applies to the confirmed-response-evaluable population; it is not a treatment effect estimate for all randomized participants. Conditioning on having a confirmed response changes the question being answered.

The 95% CI of 0.42–0.90 quantifies uncertainty around the HR and remains below 1. The P-value of 0.01227 is the registry-reported result for this secondary analysis. It does not establish that the treatment effect is of a particular absolute magnitude, nor does it replace consideration of the endpoint's restricted analysis population.

14. Secondary Results Summary

EndpointHR / OR95% CIP-valueMethod
OS, PD-L1-positive, stratified1.110.76–1.640.5798Log-rank
OS, PD-L1-positive, unstratified1.180.80–1.720.4035Log-rank
OS, ITT, stratified1.120.88–1.430.3425Log-rank
OS, ITT, unstratified1.160.92–1.480.2166Log-rank
TTD in global health status / HRQoL0.940.71–1.240.6465Log-rank
Confirmed ORR, PD-L1-positive1.440.87–2.370.1526Cochran-Mantel-Haenszel
Unconfirmed ORR, PD-L1-positive1.400.85–2.310.1834Cochran-Mantel-Haenszel
Confirmed ORR, response-evaluable1.421.00–2.020.0513Cochran-Mantel-Haenszel
Unconfirmed ORR, response-evaluable1.300.93–1.810.1226Cochran-Mantel-Haenszel
Duration of objective response, unconfirmed0.740.54–1.020.0641Log-rank
Duration of confirmed response0.620.42–0.900.01227Log-rank

15. Statistical Methodology

Log-rank testing for time-to-event endpoints

The registry identifies the log-rank test as the statistical method for PFS, OS, time to deterioration, and duration-of-response analyses. The log-rank test compares the event-time experience of two groups across follow-up while accounting for the timing of events and censoring.

Conceptual question
H0: the survival distributions do not differ between treatment groups

For a time-to-event endpoint, the analysis incorporates information about when events occur rather than reducing each participant to a simple event/no-event indicator.

Hazard ratios

The principal effect measure for the time-to-event analyses was the hazard ratio. An HR below 1 indicates a lower estimated instantaneous event rate in the first-named treatment comparison direction; an HR above 1 indicates a higher estimated instantaneous event rate.

Interpretation
HR = 0.82  →  approximately 18% lower estimated instantaneous event rate

This is a relative time-to-event interpretation. It is not a statement about the percentage of participants who experience an event and is not equivalent to a relative risk or absolute risk difference.

Cochran-Mantel-Haenszel testing

The registry identifies the Cochran-Mantel-Haenszel test for objective-response analyses. This framework is useful for comparing a binary outcome across treatment groups while accounting for stratification when strata are part of the analysis structure.

Odds ratios

Objective response was summarized with an odds ratio. If the probability of response is p, its odds are p/(1-p). The odds ratio compares those odds between groups.

Why an odds ratio is not a response-rate ratio
OR = (p1 / (1 − p1)) ÷ (p0 / (1 − p0))

An OR of 1.42 therefore cannot be read as "42% more patients responded." It compares odds, not probabilities directly.

Intention-to-treat analysis

The ITT population was explicitly defined as all randomized participants, whether or not the assigned treatment was received. This approach maintains the connection between the efficacy analysis and the randomized treatment assignment.

Stratified versus unstratified analysis

For several endpoints, the registry reports both stratified and unstratified analyses. A stratified analysis incorporates the relevant stratification structure rather than treating the entire study population as one homogeneous comparison. The ClinicalTrials.gov record identifies several analyses as stratified but do not list the underlying stratification factors.

16. Statistical Methods Explained

Why use a log-rank test for PFS and OS?

PFS and OS are time-to-event outcomes. The log-rank test is designed to compare event-time distributions while accounting for the timing of events and censored observations. This is more informative than simply comparing the proportion of participants who experienced progression by an arbitrary cutoff.

What does an HR of 0.82 mean?

An HR of 0.82 corresponds to an estimated 18% lower instantaneous event rate in the treatment group under the hazard-ratio interpretation. It does not mean that 18% of participants avoided progression, nor does it mean that each participant experienced an 18% reduction in individual risk.

Why is the confidence interval important?

A point estimate summarizes one estimated effect, but the confidence interval describes the uncertainty around that estimate under the statistical model and sampling framework. For the stratified PD-L1-positive PFS analysis, the HR was 0.82 with a 95% CI of 0.60–1.12. The interval communicates considerably more information than the point estimate alone.

Why does the P-value not measure effect size?

A P-value describes the degree of evidence against a specified null hypothesis under the statistical testing framework. It depends on both the estimated effect and the information available for estimating it. It is therefore not a percentage benefit, a probability that the treatment works, or a measure of clinical importance.

Why does ITT matter?

Randomization creates the foundation for a causal comparison between assigned treatment strategies. An ITT analysis preserves that assignment by analyzing participants according to the group to which they were randomized, regardless of whether they received the assigned treatment.

Why report both stratified and unstratified analyses?

The registry contains both versions for several endpoints. Reporting both can show how sensitive the numerical estimate is to the specified analysis framework. In IMpassion131, for example, the PD-L1-positive PFS HR was 0.82 in the stratified analysis and 0.84 in the unstratified analysis.

Why is an odds ratio different from a hazard ratio?

An odds ratio describes a binary outcome such as objective response. A hazard ratio describes a relative instantaneous event rate over time. Although both are relative effect measures, they answer different statistical questions and should not be interpreted interchangeably.

17. Multiplicity and Multiple Analyses

The ClinicalTrials.gov record identifies two registered primary endpoints and four primary-endpoint statistical analyses: stratified and unstratified analyses for each PFS endpoint. The registry also contains secondary endpoint analyses across OS, time to deterioration, objective response, and duration of response.

Analysis familyRegistry informationStatistical implication
Primary endpoints2 PFS endpointsTwo distinct primary efficacy questions were registered.
Primary analyses4Each primary endpoint has stratified and unstratified reported analyses.
Secondary analyses11 of the 15 posted statistical analysesMultiple additional efficacy analyses were reported.
Formal multiplicity adjustmentNot provided in the ClinicalTrials.gov recordNo alpha-adjustment procedure should be inferred.
Multiplicity caution: The presence of multiple reported P-values does not by itself establish that each P-value represents an independently confirmatory hypothesis test. The ClinicalTrials.gov record does not specify an alpha-allocation or multiplicity-adjustment strategy, so none is inferred here.

18. Interim Analysis and Other Design Features

The ClinicalTrials.gov record does not report an interim-analysis procedure, alpha-spending rule, non-inferiority margin, crossover plan, Bayesian analysis, or missing-data/imputation method. Those topics therefore cannot be characterized as features of the IMpassion131 statistical design from the ClinicalTrials.gov record.

No non-inferiority margin reported

The registered analyses identify superiority or other hypotheses, not a non-inferiority margin.

No Bayesian method reported

The statistical analyses posted on ClinicalTrials.gov use log-rank and Cochran-Mantel-Haenszel methods; no Bayesian method is listed.

No crossover information

The ClinicalTrials.gov record does not identify a treatment crossover strategy.

No imputation method

The ClinicalTrials.gov record does not report a formal missing-data or imputation procedure.

19. Safety Results

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

Treatment groupSerious adverse eventsAffected / at risk
Atezolizumab + paclitaxel Serious adverse events 112 / 431
Placebo + paclitaxel Serious adverse events 40 / 220

These figures should be read as reported counts over the corresponding numbers at risk. They are not converted here into an unreported statistical comparison, and no relative risk, odds ratio, or confidence interval is calculated from them.

Clinical Biostats interpretation

Safety and efficacy answer different questions. A serious-adverse-event count describes an important component of treatment safety, whereas PFS and OS describe time-to-event efficacy outcomes. The two domains should therefore be interpreted separately rather than collapsed into a single statistical measure.

The ClinicalTrials.gov record does not provide the definitions, exposure time, grading distribution, or formal statistical comparison needed for a more detailed serious-adverse-event analysis.

20. Reading the Primary PFS Results Correctly

What the HR tells us

The HR summarizes the relative instantaneous event rate under the time-to-event analysis. Values below 1 favor the atezolizumab-containing group in the reported comparison.

What the HR does not tell us

It does not provide the median PFS, absolute risk difference, probability of cure, or percentage of participants who benefit.

What the CI tells us

The 95% CI describes uncertainty around the estimated relative effect. The primary PFS intervals reported here all include 1.

What the P-value tells us

The P-value summarizes evidence against a specified null hypothesis under the reported testing procedure. It is not an effect-size measure.

Important time-to-event caution: A hazard ratio is a model-based summary of event rates over time. When a proportional-hazards model is used, the interpretation depends on the proportional-hazards framework. The ClinicalTrials.gov record identifies hazard ratios as the effect measure but do not provide a diagnostic assessment of that assumption.

21. What Is Not Reported in the Supplied Registry Data

Several details that are often useful on a complete clinical-trial analysis page are not present in the registry-reported IMpassion131 data. They are intentionally not reconstructed here.

This distinction matters because a statistically detailed page should not manufacture precision that is absent from the public registry data.

22. Why This Trial Matters Statistically

IMpassion131 provides a useful teaching case because the registry combines randomized treatment allocation, double blinding, two primary time-to-event endpoints, a biomarker-defined analysis population, stratified and unstratified survival analyses, binary response analyses, and duration-of-response analyses.

ConceptHow it appears in IMpassion131
RandomizationRandomized phase 3 parallel-group comparison.
BlindingDouble-blind design.
ITT analysisPrimary PFS analysis in all randomized participants, whether or not assigned treatment was received.
Time-to-event analysisPFS, OS, time to deterioration, and duration of response.
Log-rank testReported method for the time-to-event analyses.
Hazard ratioPrimary effect measure for time-to-event analyses.
Cochran-Mantel-Haenszel testReported method for objective-response analyses.
Odds ratioReported effect measure for objective response.
Stratified analysisReported for several primary and secondary time-to-event analyses.
Biomarker-defined populationPD-L1-positive PFS and OS analyses restricted to the IC1/2/3 population.
Analysis-population effectsResponse and duration-of-response analyses use populations narrower than the full ITT population.

23. Statistical Interpretation of the Overall Evidence

The ClinicalTrials.gov record shows a consistent statistical pattern within the reported primary PFS analyses: all four HR estimates are below 1, ranging from 0.82 to 0.86, while each two-sided 95% confidence interval includes 1. The corresponding P-values range from 0.1285 to 0.2601.

The secondary endpoints are more heterogeneous. The reported OS HRs range from 1.11 to 1.18, the time-to-deterioration HR is 0.94, the objective-response ORs range from 1.30 to 1.44, and the duration-of-response HRs are 0.74 and 0.62. These estimates answer different questions and arise from different analysis populations, so they should not be collapsed into a single overall treatment-effect statistic.

A useful statistical reading

The most important discipline when reading these results is to keep three dimensions separate: effect size, precision, and statistical evidence under the specified test. The HR or OR provides the estimated effect, the confidence interval describes its uncertainty, and the P-value addresses the corresponding hypothesis test. None of the three should be used as a substitute for the others.

24. Limitations

25. Related Tutorials

Learn more about the statistical methods used in this trial:

26. Related Statistical Calculators

27. Sources

Continue with Clinical Biostats statistical methods

Explore the survival-analysis, categorical-data, confidence-interval, and clinical-trial concepts that appear throughout IMpassion131.

28. Record Summary

IMpassion131 is a useful example of how a randomized phase 3 trial can generate several related but distinct statistical questions. The primary endpoints are both PFS measures, with analyses reported in a PD-L1-positive population and the full ITT population. The registry reports stratified and unstratified log-rank analyses, hazard ratios with two-sided 95% confidence intervals, and P-values. Secondary analyses extend the statistical framework to OS, time to deterioration, objective response, and duration of response using both time-to-event and binary-outcome methods.

The central statistical lesson is that an HR, OR, confidence interval, and P-value each convey different information. In particular, an HR below 1 does not by itself establish a conclusive treatment effect, an OR is not a response-rate ratio, and a P-value is not an effect-size measure. Interpretation should remain tied to the endpoint definition, analysis population, censoring structure, and prespecified statistical framework.

Clinical Biostats methodology: A trial-results page should not merely repeat the registry. The goal is to reconstruct the statistical story of the trial while clearly separating reported evidence from educational interpretation and avoiding unsupported reconstruction of missing results.