← Clinical Trials
Non-Small Cell Lung Cancer Phase 3 Disease-Free Survival NCT02486718

IMpower010: Complete Statistical Analysis of Atezolizumab in Non-Small Cell Lung Cancer

An independent statistical analysis of the randomized phase 3 IMpower010 trial evaluating atezolizumab versus best supportive care following chemotherapy in patients with non-small cell lung cancer, with emphasis on disease-free survival and the statistical interpretation of hazard ratios, confidence intervals, and time-to-event endpoints.

Phase 3  ·  Randomized  ·  Parallel design  ·  Results posted
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results and trial characteristics are restricted to the ClinicalTrials.gov data posted on ClinicalTrials.gov for IMpower010. Where the registry does not report a formal statistical method for an individual analysis, that limitation is stated explicitly.

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

IMpower010 is a randomized, parallel, open-label phase 3 trial evaluating atezolizumab versus best supportive care following chemotherapy in participants with non-small cell lung cancer. The registry reports 1,280 participants, two arms, three primary endpoints, and posted statistical analyses for all three primary endpoints.

1,280
Enrollment
Phase 3 trial
2
Arms
Atezolizumab vs BSC
3
Primary endpoints
All with estimates + CI
95
Months
Maximum registered time frame
FeatureIMpower010
Trial nameIMpower010
PhasePhase 3
ConditionNon-Small Cell Lung Cancer
DesignRandomized, parallel, open-label
AllocationRandomized
Primary purposeTreatment
Enrollment1,280
Primary endpointsThree registered endpoints focused on disease-free survival
Primary endpoint type(s)Binary; Time-to-event
Results postedYes
Statistical analyses posted10
Lead sponsorHoffmann-La Roche
Trial statusActive, not recruiting
ClinicalTrials.govNCT02486718

2. Clinical Question

The central statistical question is whether the disease-free survival experience differs between participants randomized to atezolizumab and those randomized to best supportive care (BSC) following chemotherapy.

Population

The randomized ITT population consisted of all randomized participants with resected Stage IB [tumors ≥ 4 centimetres (cm)]-IIIA non-small cell lung cancer, whether or not the participant received the assigned treatment.

Intervention

Atezolizumab. The registry ClinicalTrials.gov record also identifies cisplatin, vinorelbine, docetaxel, gemcitabine, and pemetrexed among the trial interventions.

Comparator

Best supportive care (BSC).

Primary question

Does atezolizumab produce a different disease-free survival experience from BSC in the prespecified randomized populations?

3. Trial Design

01
Enroll1,280 participants
02
RandomizeTwo parallel arms
03
TreatmentAtezolizumab or BSC
04
FollowDFS events and censoring
05
AnalyzeTime-to-event comparisons
Allocation
Randomized allocation with two parallel treatment groups.
Masking
None. The registry classifies the trial as open-label.
Primary purpose
Treatment.
Hypothesis type
Superiority for the posted primary analyses.
ARM 1

Atezolizumab

  • Atezolizumab is the randomized treatment group.
  • The posted efficacy analyses compare this group with BSC.
ARM 2

Best Supportive Care

  • BSC is the comparator group.
  • The posted efficacy analyses use the randomized comparison with atezolizumab.

The registry does not provide enough information in the ClinicalTrials.gov record to reconstruct a detailed treatment schedule, chemotherapy sequence, or operational follow-up schedule. Those details are therefore not inferred here.

4. Randomization and Analysis Populations

Randomization is central to the interpretation of the efficacy comparison because the primary analysis population is defined according to randomized treatment assignment rather than actual treatment received.

Analysis populationDefinition / role
Randomized ITT population All randomized participants with resected Stage IB [tumors ≥ 4 centimetres (cm)]-IIIA NSCLC, whether or not the participant received the assigned treatment.
Stage II-IIIA population All randomized participants with extent of disease as either Stage II or Stage IIIA; this population is a subset of the ITT population.
PD-L1 ≥1% subpopulation Stage II-IIIA participants with valid baseline PD-L1 SP263 measurement and ≥1% tumor-cell expression by the SP263 immunohistochemistry assay.
PD-L1 ≥50% subpopulation Stage II-IIIA participants with valid baseline PD-L1 SP263 measurement and ≥50% tumor-cell expression.

The ITT definition is especially important for causal interpretation. Once randomization occurs, retaining participants in their randomized groups helps preserve the balance created by randomization, even if not every participant receives or continues the assigned treatment.

5. Primary Endpoints

Registered primary endpointTime frameDefinitionStatistical role
Disease-Free Survival (DFS) in Intent-to-treat (ITT) Population Up to 95 months Time from randomization to the first recurrence of NSCLC or occurrence of new primary NSCLC as determined by the investigator or death, whichever occurs first. Time-to-event; Kaplan-Meier methodology was used to estimate median DFS for each treatment arm.
DFS in All Randomized Stage II-IIIA Population Up to 95 months Time from randomization to the first recurrence of NSCLC or occurrence of new primary NSCLC as determined by the investigator or death, whichever occurs first. Time-to-event definition; Kaplan-Meier methodology was used to estimate median DFS for each treatment arm.
DFS in the PD-L1 SP263 ≥ 1% Tumor Cell (TC) Subpopulation Within the Stage II-IIIA Population Up to 95 months Time from randomization to the first recurrence of NSCLC or occurrence of new primary NSCLC as determined by the investigator or death, whichever occurs first. Time-to-event definition; Kaplan-Meier methodology was used to estimate median DFS for each treatment arm.
Endpoint terminology: Although the registry's normalized endpoint-type field classifies some entries as binary, the registered DFS definition is explicitly a time-to-event endpoint: the analysis follows participants from randomization until recurrence, new primary NSCLC, or death, whichever occurs first.

6. Statistical Methodology

Kaplan-Meier estimation

The registry explicitly states that Kaplan-Meier methodology was used to estimate median DFS for each treatment arm. This is appropriate for a time-to-event endpoint because not every participant necessarily experiences the event during the observation period.

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

Here, di represents the number of events at an observed event time and ni represents the number of participants at risk immediately before that time. The resulting survival function estimates the probability of remaining event-free beyond time t.

Log-rank testing

The registry identifies the log-rank test as the normalized statistical method for the posted analyses overall. A log-rank test compares the observed and expected numbers of events between randomized groups across follow-up time.

Conceptual comparison
H0: the treatment groups have the same event-time distribution

The test is based on the pattern of events over time rather than simply comparing the proportion of participants who have experienced an event at one fixed time point.

Hazard ratio

The principal effect measure reported for the primary time-to-event analyses is the hazard ratio (HR). The posted primary analyses all identify the comparison as atezolizumab versus BSC and report two-sided 95% confidence intervals.

Interpretation of the hazard ratio
HR < 1  →  lower estimated instantaneous event rate in the atezolizumab group

The hazard ratio is a relative time-to-event measure. It is not a percentage of patients who benefit, not a probability of being disease-free, and not an absolute difference in event-free rates.

Stratified analysis

The registry identifies each posted primary hazard-ratio analysis as a stratified analysis. The ClinicalTrials.gov record does not identify the exact stratification variables or provide the full model specification. Accordingly, this page does not infer which covariates were used for stratification.

Confidence intervals

Each of the three primary analyses reports a two-sided 95% confidence interval. The interval provides a range of values describing the uncertainty associated with the estimated treatment effect under the statistical framework used for the analysis.

What the registry does not specify

The ClinicalTrials.gov record does not report a formal Cox-model specification, treatment of tied event times, missing-data imputation method, Bayesian analysis, non-inferiority margin, crossover analysis, factorial design, or an interim-analysis alpha-spending scheme. These methods are therefore not attributed to IMpower010 on this page.

7. Primary Results: DFS in the ITT Population

The first primary endpoint was Disease-Free Survival (DFS) in Intent-to-treat (ITT) Population, with a registered time frame of up to 95 months.

Hazard ratio for disease-free survival

0.848

95% CI: 0.710–1.013   ·   P = 0.0683

Atezolizumab vs Best Supportive Care (BSC)

EndpointEstimate95% CIP-valueAnalysis
DFS in ITT population HR 0.848 0.710–1.013 0.0683 Stratified analysis; log-rank framework reported for the trial
Clinical Biostats interpretation

The estimated hazard ratio of 0.848 means that the estimated instantaneous rate of a DFS event was approximately 84.8% as high in the atezolizumab group as in the BSC group under the reported analysis. Expressed as a simple relative interpretation, this corresponds to an estimated 15.2% lower hazard.

The HR does not mean that 15.2% of participants avoided recurrence, that 15.2% of participants benefited, or that every participant had the same reduction in event risk. It is a relative time-to-event measure comparing the event rates represented by the fitted analysis.

The two-sided 95% CI of 0.710–1.013 indicates uncertainty around the estimate. Because the interval extends above 1, the data represented by this interval are compatible with an HR above 1 as well as values below 1.

The P = 0.0683 is a measure of statistical evidence against the null hypothesis under the specified testing framework. It does not measure the size of the treatment effect, the probability that the treatment works, or the probability that the null hypothesis is true.

The registry identifies the hypothesis type as superiority. Therefore, the relevant inferential question is whether the randomized groups differ in the prespecified superiority comparison; this is distinct from a non-inferiority question, for which a prespecified margin would be central.

8. Primary Results: DFS in All Randomized Stage II-IIIA Population

The second primary endpoint was DFS in All Randomized Stage II-IIIA Population, also evaluated over a registered time frame of up to 95 months.

Hazard ratio for disease-free survival

0.830

95% CI: 0.691–0.998

Atezolizumab vs Best Supportive Care (BSC)   ·   Superiority hypothesis

EndpointEstimate95% CIFormal method detail
DFS in all randomized Stage II-IIIA population HR 0.830 0.691–0.998 Method not reported for this individual analysis; stratified analysis identified
Clinical Biostats interpretation

The estimated HR of 0.830 corresponds to an estimated instantaneous DFS-event rate approximately 83.0% as high in the atezolizumab group as in the BSC group, or approximately a 17.0% lower estimated hazard.

This estimate applies to the randomized Stage II-IIIA population, which is a defined subset of the broader ITT population. It should therefore not be treated as though it were the estimate for every participant in the trial.

The two-sided 95% CI of 0.691–0.998 is relatively close to 1 at its upper boundary. That matters because statistical conclusions can be sensitive to the precise uncertainty represented by the confidence interval.

The registry does not provide a separate P-value for this individual primary analysis in the ClinicalTrials.gov record. The absence of a posted P-value here should not be converted into a calculated or reconstructed value.

The analysis is identified as a superiority analysis and as stratified. The ClinicalTrials.gov record does not specify the exact stratification factors or full regression-model details, so those features are not inferred.

9. Primary Results: DFS in the PD-L1 SP263 ≥1% TC Subpopulation

The third primary endpoint evaluated DFS in the Programmed Death-ligand 1 (PD-L1) SP263 ≥ 1% Tumor Cell (TC) Subpopulation Within the Stage II-IIIA Population, with a registered time frame of up to 95 months.

Hazard ratio for disease-free survival

0.704

95% CI: 0.545–0.910

Atezolizumab vs Best Supportive Care (BSC)   ·   Superiority hypothesis

EndpointEstimate95% CIFormal method detail
DFS in PD-L1 SP263 ≥1% TC Stage II-IIIA subpopulation HR 0.704 0.545–0.910 Method not reported for this individual analysis; stratified analysis identified
Clinical Biostats interpretation

The estimated HR of 0.704 corresponds to an estimated instantaneous DFS-event rate approximately 70.4% as high in the atezolizumab group as in the BSC group, or approximately a 29.6% lower estimated hazard.

The estimate applies specifically to the PD-L1 SP263 ≥1% TC subpopulation within Stage II-IIIA. It therefore should not be generalized numerically to participants outside that defined analysis population.

The two-sided 95% CI of 0.545–0.910 remains below 1. This indicates that the uncertainty interval represented by the registry's analysis is entirely below the null hazard ratio of 1.

The confidence interval describes uncertainty around the population-level treatment-effect estimate. It does not indicate that individual participants' treatment effects must fall between 0.545 and 0.910.

The registry does not provide a separate P-value for this individual primary analysis in the ClinicalTrials.gov record. A P-value should not be reconstructed from the HR and confidence interval unless the statistical analysis specification needed for that calculation is known.

10. Secondary Disease-Free Rate Results

The registry also posts disease-free rates at Year 3 and Year 5 for the ITT population, the randomized Stage II-IIIA population, and the PD-L1 SP263 ≥1% TC subpopulation within Stage II-IIIA. These are presented as differences in event-free rates between atezolizumab and BSC.

Secondary endpointTime frameDifference in event-free rates95% CI
Disease-free Rate at Year 3 in ITT Population Year 3 5.98 percentage points -0.28 to 12.23
Disease-free Rate at Year 3 in All Randomized Stage II-IIIA Population Year 3 6.65 percentage points -0.06 to 13.36
Disease-free Rate at Year 3 in PD-L1 SP263 ≥1% TC Subpopulation Year 3 10.67 percentage points 1.56 to 19.79
Disease-free Rate at Year 5 in ITT Population Year 5 5.48 percentage points -0.94 to 11.90
Disease-free Rate at Year 5 in All Randomized Stage II-IIIA Population Year 5 4.88 percentage points -1.94 to 11.70
Disease-free Rate at Year 5 in PD-L1 SP263 ≥1% TC Subpopulation Year 5 10.46 percentage points 1.16 to 19.76
How to read these differences: A positive difference indicates a higher estimated disease-free rate in the atezolizumab group under the registry's reported comparison. The confidence intervals quantify uncertainty around those differences. Several intervals cross zero, whereas the intervals reported for the PD-L1 SP263 ≥1% TC population at Year 3 and Year 5 remain above zero.

Time-to-event estimates versus fixed-time rates

The two types of results answer related but different questions. A hazard ratio summarizes a relative treatment effect over the time-to-event analysis, whereas a Year 3 or Year 5 disease-free rate describes the estimated event-free proportion at one specified time. Neither should be substituted for the other.

11. Additional Posted DFS Analysis: PD-L1 SP263 ≥50% TC

The registry also reports a secondary DFS analysis in the PD-L1 (SP263 ≥ 50% TC) Subpopulation Within the Stage II-IIIA Population.

Hazard ratio for disease-free survival

0.503

95% CI: 0.332–0.761

Atezolizumab vs Best Supportive Care (BSC)

Clinical Biostats interpretation

The estimated HR of 0.503 corresponds to an estimated instantaneous DFS-event rate approximately 50.3% as high in the atezolizumab group as in the BSC group, or approximately a 49.7% lower estimated hazard.

The 95% CI of 0.332–0.761 remains below 1, indicating that the full reported confidence interval is on the lower-hazard side of the null value.

This analysis is a defined PD-L1 subgroup analysis rather than a comparison of independently randomized trials. A difference between subgroup estimates cannot by itself establish that PD-L1 level modifies the treatment effect; a formal interaction analysis would be needed for that question.

The registry does not provide a separate P-value for this individual analysis in the ClinicalTrials.gov record. The analysis is identified as a stratified analysis, but the ClinicalTrials.gov record does not specify the complete statistical model.

12. Safety Results

The ClinicalTrials.gov record reports serious adverse events by arm using affected participants over participants at risk. The safety information available here is limited to the reported serious-adverse-event counts and denominators.

Safety groupSerious adverse eventsParticipants at riskReported affected / at risk
Enrollment Phase 259 1,269 259/1269
Best Supportive Care (BSC) 42 495 42/495
Atezolizumab 88 495 88/495

The registry's registry-reported safety field describes these as affected participants over participants at risk. The "Enrollment Phase" entry should not be silently treated as a third randomized treatment arm: the ClinicalTrials.gov record specifies two arms, Atezolizumab and BSC.

Safety interpretation: Serious adverse events and efficacy endpoints answer different statistical questions. The ClinicalTrials.gov record does not provide enough detail to determine the exact adverse-event definitions, exposure windows, severity grading, or formal between-arm hypothesis test. Therefore, no additional safety comparison is calculated here.

13. Statistical Methods Explained

Why was Kaplan-Meier estimation used?

DFS is defined from randomization until recurrence, new primary NSCLC, or death, whichever occurs first. Because participants can have different follow-up times and some can remain event-free at the end of observation, Kaplan-Meier estimation is suited to constructing the event-free survival curve while accommodating right censoring.

What does an HR of 0.848 mean?

An HR of 0.848 means that the estimated instantaneous rate of a DFS event in the atezolizumab group was 0.848 times that in the BSC group under the reported analysis. It can also be described as approximately a 15.2% lower estimated hazard. It does not mean a 15.2% absolute increase in disease-free survival.

Why is the confidence interval important?

The point estimate alone hides uncertainty. For the ITT analysis, the HR is 0.848, but the two-sided 95% CI is 0.710–1.013. The interval therefore provides important context about the precision of the estimate and about which treatment-effect values remain compatible with the statistical uncertainty represented by the analysis.

What does a P-value of 0.0683 tell us?

The P-value describes the degree of statistical evidence against a null hypothesis under the specified testing framework. It does not tell us the magnitude of the HR, the probability that the treatment is effective, or the clinical importance of the observed estimate.

Why does randomization matter for the ITT analysis?

The ITT population keeps randomized participants associated with their assigned groups regardless of whether they received the assigned treatment. This preserves the comparison generated by randomization more directly than reassigning participants according to treatment actually received.

Why are the Stage II-IIIA and PD-L1 populations analyzed separately?

The registry defines these populations more narrowly than the overall ITT population. The Stage II-IIIA population is a subset based on disease stage, while the PD-L1 analyses further restrict the population according to baseline SP263 tumor-cell expression. An effect estimate in one population should therefore be interpreted within that population.

What does a hazard ratio not tell us?

A hazard ratio does not directly provide the absolute probability of remaining disease-free at Year 3 or Year 5. It also does not imply that the treatment effect is identical for every participant. Time-to-event interpretation depends on the underlying event-time distributions and the assumptions of the statistical model used to obtain the estimate.

14. Understanding the Three Primary Populations

PopulationWhat is selected?Statistical implication
ITT All randomized participants meeting the registered ITT definition Broadest primary randomized efficacy population in the ClinicalTrials.gov record
Stage II-IIIA Randomized participants with Stage II or IIIA disease Subset of ITT population; treatment effect is estimated within a narrower disease-stage population
PD-L1 SP263 ≥1% TC Stage II-IIIA participants with valid baseline measurement and ≥1% TC expression Biomarker-defined subset; estimate should not automatically be generalized to the full ITT population

The hierarchy matters statistically. The same randomized treatment comparison is being examined in increasingly restricted populations. Smaller analysis populations generally contain less information than the full randomized population, although the ClinicalTrials.gov record does not provide the participant counts for each of these subsets.

15. Comparing Relative and Absolute Effects

Relative effect

The hazard ratio expresses the relative difference in the event rate represented by the time-to-event analysis. For example, an HR of 0.704 is a relative measure, not an absolute percentage-point difference.

Absolute effect

The Year 3 and Year 5 disease-free-rate analyses report differences in event-free rates directly in percentage points. These estimates provide a fixed-time perspective.

Precision

Confidence intervals should accompany both types of estimates because a point estimate without uncertainty can give a misleading impression of precision.

Different questions

A hazard ratio and a fixed-time rate difference should not be treated as interchangeable effect measures. Each describes a different aspect of the observed time-to-event experience.

16. Interpreting the Confidence Intervals

AnalysisEstimate95% CINull value
DFS, ITT0.8480.710–1.013HR = 1
DFS, Stage II-IIIA0.8300.691–0.998HR = 1
DFS, PD-L1 ≥1% TC0.7040.545–0.910HR = 1
DFS, PD-L1 ≥50% TC0.5030.332–0.761HR = 1

For hazard ratios, the conventional null value is 1. A confidence interval entirely below 1 is consistent with a lower estimated hazard for atezolizumab throughout the interval represented by the reported uncertainty. The ITT interval extends slightly above 1, whereas the other listed HR intervals remain below 1.

For the disease-free-rate differences, the corresponding null value is 0. A confidence interval crossing zero includes both a positive and negative difference under the uncertainty represented by the interval.

17. Multiplicity and Multiple Primary Endpoints

IMpower010 has three registered primary endpoints, all involving DFS but evaluated in distinct analysis populations. The posted analyses identify a superiority hypothesis for each primary comparison.

Primary analysisPopulationEffect measureHypothesis
Primary 1 ITT Hazard ratio Superiority
Primary 2 All randomized Stage II-IIIA Hazard ratio Superiority
Primary 3 PD-L1 SP263 ≥1% TC within Stage II-IIIA Hazard ratio Superiority

Multiple primary endpoints raise an important statistical-design question: how is the overall type I error controlled across the testing sequence? The ClinicalTrials.gov record identifies the endpoints and superiority framework but do not specify a multiplicity-adjustment strategy, testing hierarchy, alpha allocation, or gatekeeping procedure. Those details are therefore not reconstructed here.

Important: A P-value from one primary endpoint cannot automatically be interpreted independently of the other primary endpoints when several confirmatory hypotheses are tested. The exact error-control strategy must come from the prespecified statistical design. The ClinicalTrials.gov record does not provide that strategy.

18. Interim Analysis, Crossover, and Other Design Features

Several design features commonly require special statistical treatment in clinical trials. The registry-reported IMpower010 registry data do not report enough information to attribute these methods to this trial.

Design topicWhat the ClinicalTrials.gov record establishesWhat is not reported
Interim analysis Not specified in the ClinicalTrials.gov record No interim-analysis schedule or alpha-spending method is reported.
Crossover Not specified in the ClinicalTrials.gov record No crossover policy or crossover-adjusted analysis is reported.
Factorial design Parallel two-arm design is reported No factorial structure is reported.
Non-inferiority margin Superiority hypothesis is reported No non-inferiority margin is applicable to the registry-reported superiority analyses.
Bayesian methods Not reported No Bayesian analysis is identified.
Missing-data imputation Not reported No imputation strategy is specified in the ClinicalTrials.gov record.

This distinction is important. A statistical analysis page should not fill gaps in the registry record with methods that would merely be conventional for similar trials. The method must be attributed to IMpower010 only when the ClinicalTrials.gov record supports that attribution.

19. Censoring and the Meaning of DFS

DFS is a time-to-event endpoint because participants are followed from randomization until the first qualifying event. Participants who have not experienced recurrence, new primary NSCLC, or death by the relevant end of observation can contribute censored follow-up rather than a DFS event.

Conceptual time-to-event structure
DFS time = time from randomization → first qualifying DFS event

The registry defines the qualifying event as first recurrence of NSCLC, occurrence of new primary NSCLC as determined by the investigator, or death, whichever occurs first.

Censoring is not equivalent to saying that a participant remained disease-free forever. It means that the analysis has usable event-free follow-up up to the censoring point. This distinction is one reason Kaplan-Meier methodology is preferable to simply dividing the number of events by the number randomized.

20. Hazard-Ratio Interpretation Across the Primary Analyses

Reported hazard ratios
ITT
0.848
Stage II-IIIA
0.830
PD-L1 ≥1% TC
0.704
PD-L1 ≥50% TC
0.503

The visual above is an educational representation of the reported hazard-ratio values relative to a reference scale of 1. It should not be read as a Kaplan-Meier curve or as a display of individual participant outcomes.

The sequence of estimates shows that the reported HR becomes smaller in the narrower PD-L1-defined populations. That observation is descriptive. It does not by itself establish that PD-L1 expression modifies the treatment effect. A formal interaction test is needed to address effect modification.

21. What the P-Value Does — and Does Not — Mean

P-value interpretation

The ITT primary analysis reports P = 0.0683. This value quantifies the evidence against the relevant null hypothesis under the statistical testing framework. It is not the probability that the null hypothesis is true.

It also does not measure effect size. The effect size is represented by the HR of 0.848, while the confidence interval of 0.710–1.013 describes uncertainty around that estimate.

A P-value should therefore be read together with the effect estimate, confidence interval, endpoint definition, analysis population, and prespecified multiplicity framework rather than as a standalone measure of clinical importance.

22. Statistical Interpretation vs Clinical Interpretation

Statistical interpretation

The registry reports hazard-ratio estimates below 1 for the primary DFS analyses. The ITT estimate is 0.848 with a two-sided 95% CI of 0.710–1.013 and P = 0.0683. The Stage II-IIIA and PD-L1 ≥1% analyses report HRs of 0.830 and 0.704, respectively.

Clinical interpretation

Clinical meaning requires consideration of the absolute disease-free rates, duration of follow-up, treatment burden, safety, patient characteristics, and the specific population being analyzed. The ClinicalTrials.gov record does not provide enough information to make a broader benefit-risk assessment.

The distinction is deliberate. A statistically estimated treatment effect is not itself a treatment recommendation. Statistical evidence describes the observed randomized comparison; clinical decisions require a wider evidence base.

23. Important Limitations and Interpretation Issues

24. Why This Trial Matters Statistically

IMpower010 is a useful teaching example because the same randomized comparison is examined through several complementary statistical lenses: a broad ITT population, a Stage II-IIIA population, and biomarker-defined PD-L1 populations.

ConceptHow it appears in IMpower010
RandomizationRandomized, parallel phase 3 design with two arms.
Intention-to-treat analysisPrimary ITT population includes all randomized participants meeting the registered definition whether or not assigned treatment was received.
Time-to-event endpointDFS is measured from randomization to recurrence, new primary NSCLC, or death.
Kaplan-Meier estimationUsed to estimate median DFS for each treatment arm.
Log-rank testIdentified as the normalized statistical method for the trial analyses.
Hazard ratioUsed as the principal reported effect measure for DFS analyses.
Confidence intervalTwo-sided 95% CIs are reported for the primary HR estimates and secondary event-free-rate differences.
Stratified analysisEach posted primary HR analysis is identified as a stratified analysis.
Biomarker-defined analysisDFS is evaluated in PD-L1 SP263 ≥1% TC and ≥50% TC populations.
Multiple primary endpointsThree DFS primary endpoints are registered across distinct analysis populations.
Absolute versus relative effectsHazard ratios are complemented by Year 3 and Year 5 disease-free-rate differences.

25. Related Tutorials

Learn more about the methods used in this trial:

26. Related Calculators

27. Sources

Continue through the Clinical Biostats statistical pathway

Explore the underlying survival-analysis concepts, statistical calculators, and other clinical-trial analyses that build on the methods illustrated by IMpower010.

28. Record Summary

IMpower010 provides a useful statistical case study in randomized time-to-event analysis. The ClinicalTrials.gov record defines three primary DFS endpoints across the ITT, Stage II-IIIA, and PD-L1 SP263 ≥1% TC populations; report hazard ratios with two-sided 95% confidence intervals; identify stratified analyses and the log-rank test; and provide secondary Year 3 and Year 5 disease-free-rate differences. The additional PD-L1 SP263 ≥50% TC analysis provides another example of interpreting a biomarker-defined time-to-event estimate.

The most informative reading of the results combines the hazard ratio, confidence interval, P-value where reported, analysis population, endpoint definition, and fixed-time disease-free-rate differences. Just as importantly, interpretation should distinguish what the registry actually specifies from statistical methods that would merely be conventional for similar trials.

Clinical Biostats methodology: A trial-results page should not merely repeat the registry. The goal is to explain the statistical structure of the randomized comparison, clarify what each estimate means, identify uncertainty and design considerations, and avoid attributing unreported methods or results to the trial.