← Clinical Trials
Stage IB-IIIA NSCLC Phase 3 Disease-Free Survival NCT02511106

ADAURA: Complete Statistical Analysis of AZD9291 in Stage IB-IIIA Non-Small Cell Lung Carcinoma

An independent statistical review of the randomized phase 3 ADAURA trial comparing AZD9291 with placebo in patients with stage IB-IIIA non-small cell lung carcinoma following complete tumour resection, with or without adjuvant chemotherapy.

Trial start: October 21, 2015  ·  Primary completion: April 11, 2022  ·  Sponsor: AstraZeneca
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

ADAURA was a randomized, parallel, triple-masked phase 3 trial evaluating AZD9291 versus placebo in patients with stage IB-IIIA non-small cell lung carcinoma following complete tumour resection, with or without adjuvant chemotherapy. The registered primary endpoint was disease-free survival (DFS), a time-to-event endpoint.

682
Enrolled
2 treatment arms
3
Phase
Randomized phase 3
0.23
DFS HR
Stage IIA-IIIA; 95% CI 0.18–0.30
0.27
DFS HR
Stage IB-IIIA; 95% CI 0.21–0.34
FeatureADAURA
PhasePhase 3
ConditionStage IB-IIIA non-small cell lung carcinoma
DesignRandomized, parallel, triple-masked
AllocationRandomized
Primary purposeTreatment
Enrollment682
Primary endpointDisease-free survival (DFS)
Primary endpoint typeTime-to-event
Primary analysis methodLog-rank test
Effect measureHazard ratio
Hypothesis typeSuperiority
StatusActive, not recruiting
ClinicalTrials.govNCT02511106
Lead sponsorAstraZeneca

2. Clinical Question

The central question was whether AZD9291 improves disease-free survival compared with placebo in patients with stage IB-IIIA non-small cell lung carcinoma after complete tumour resection, with or without adjuvant chemotherapy.

Population

Patients with stage IB-IIIA non-small cell lung carcinoma following complete tumour resection, with or without adjuvant chemotherapy.

Intervention

AZD9291 80 mg/40 mg, as listed in the registered intervention information.

Comparator

Placebo AZD9291 80 mg/40 mg.

Primary question

Does AZD9291 improve disease-free survival relative to placebo under a superiority framework?

3. Trial Design

01
Randomize682 enrolled
02
Parallel armsAZD9291 vs placebo
03
Triple maskingRegistered masking designation
04
DFS assessmentTime-to-event endpoint
05
OS follow-upSecondary endpoint
ARM 1

AZD9291

  • AZD9291 80 mg/40 mg
  • Listed as a drug intervention
  • Compared with placebo in the primary efficacy analyses
ARM 2

Placebo

  • Placebo AZD9291 80 mg/40 mg
  • Listed as a drug intervention
  • Comparator for the primary DFS analyses
Registered intervention structure: the trial data also list open-label AZD9291 80 mg/40 mg as an intervention. The formal primary statistical comparisons reported in the registry are specifically between the AZD9291 80 mg Tablet and Placebo Tablet groups.
Allocation
Randomized allocation was used to create the treatment comparison.
Design model
Parallel assignment with two treatment arms.
Masking
The registry designates the trial as triple-masked.
Primary purpose
Treatment.

4. Trial Timeline and Data Maturity

October 21, 2015

Trial start

The registered study start date was October 21, 2015.

Primary analysis

Early primary analysis

The registry states that the primary analysis was performed two years early following an IDMC recommendation.

2022

Exploratory final DFS analysis

The registry states that an exploratory analysis of final DFS data at the prespecified maturity of approximately 50% was performed in 2022.

April 11, 2022

Primary completion

The registered primary completion date was April 11, 2022.

Follow-up limitation: the registry states that after the final DFS data analysis and until the final OS analysis, only limited data were collected, including subsequent anti-cancer treatment, survival status, and limited safety data. This is important when interpreting the relationship between different analysis periods.

5. Endpoints

EndpointRegistry definition / time frameType
Disease-free survival (DFS) Defined as the time from the date of randomization until the date of disease recurrence or death (by any cause in the absence of recurrence). Time frame: up to approximately 5 years after the first patient is randomized, with a maximum follow-up of 70 months. Primary; time-to-event
Overall survival (OS) Time frame: up to approximately 7 years after the first patient is randomized, with a maximum follow-up of 86 months. Secondary; time-to-event

The registry lists one registered primary endpoint: assessing the efficacy of AZD9291 compared with placebo as measured by disease-free survival. Four formal statistical analyses are posted: two for the primary DFS endpoint and two for the secondary OS endpoint.

6. Statistical Methodology

Primary endpoint: disease-free survival

Disease-free survival is a time-to-event endpoint. In ADAURA, the registry defines DFS as the time from randomization until disease recurrence or death from any cause in the absence of recurrence. This definition combines two possible event types into a single endpoint.

The posted primary analyses use a log-rank test and report the treatment effect as a hazard ratio. Both primary analyses are described as superiority analyses.

Hazard-ratio framework
HR = estimated hazard in AZD9291 group ÷ estimated hazard in placebo group

For this trial, the registry notes that a hazard ratio below 1 favours AZD9291. The hazard ratio is a relative time-to-event measure; it is not a percentage of patients who are disease-free and it is not an absolute risk difference.

Stratified analysis

The posted DFS and OS analyses state that the log-rank test was stratified by stage, race and mutation type. Stratification allows the treatment comparison to account for these prespecified factors rather than treating every patient as though those characteristics were distributed identically across the comparison.

Intention-to-treat principle

The registry analysis text identifies intention-to-treat analysis as an analysis concept for the primary and secondary time-to-event comparisons. The underlying principle is to preserve the randomized comparison by evaluating patients according to their randomized treatment assignment rather than allowing post-randomization treatment experience to redefine the original groups.

Kaplan-Meier estimation

Kaplan-Meier estimation is a standard descriptive method for time-to-event data because it can accommodate right-censored observations. The ADAURA registry specifically reports the log-rank test as the statistical method for its posted comparisons; it does not list Kaplan-Meier estimation as a method. Kaplan-Meier curves are therefore a useful way to understand the type of endpoint, but they should not be presented as a registry-reported analysis method unless the source explicitly identifies them.

Confidence intervals

The primary analyses report two-sided 95% confidence intervals for the hazard ratio. The interval describes statistical uncertainty around the estimated relative effect under the analysis framework. It does not describe the range of individual patient outcomes.

P-values

The p-value assesses evidence against the null hypothesis within the specified testing framework. It is not a measure of the magnitude of the treatment effect. In ADAURA, the magnitude is more directly described by the hazard ratio and its confidence interval.

7. Primary Results: Disease-Free Survival in Stage IIA-IIIA Patients

The first posted primary analysis evaluated DFS in the Full Analysis Set (FAS) of stage IIA-IIIA patients. The comparison was AZD9291 80 mg Tablet versus Placebo Tablet.

DFS hazard ratio

0.23

95% CI: 0.18–0.30   ·   P < 0.0001

Two-sided 95% confidence interval; log-rank test stratified by stage, race and mutation type.

ElementPosted analysis
EndpointDisease-free survival (DFS)
PopulationFull Analysis Set (FAS), stage IIA-IIIA patients
ComparisonAZD9291 80 mg Tablet vs Placebo Tablet
MethodLog-rank test
Effect measureHazard ratio
Estimate0.23
95% CI0.18–0.30
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

A hazard ratio of 0.23 means that the estimated hazard of a DFS event in the AZD9291 group was 23% of the estimated hazard in the placebo group under the reported time-to-event comparison. Equivalently, 1 − 0.23 = 0.77, so the estimate corresponds to an approximately 77% lower estimated hazard for the AZD9291 group relative to placebo.

The hazard ratio does not mean that 77% of patients avoided recurrence, that 77% of patients were cured, or that each individual patient had exactly a 77% reduction in risk. It is a relative model-based time-to-event measure.

The two-sided 95% confidence interval of 0.18–0.30 gives a range of values describing uncertainty around the estimated hazard ratio. The interval is relatively narrow compared with the magnitude of the point estimate, although it still represents sampling and model uncertainty rather than the range of individual patient responses.

The p-value of <0.0001 indicates very strong evidence against the null hypothesis under the reported testing framework. It does not tell us that the effect is clinically large, nor does it provide the probability that the null hypothesis is true.

Because DFS is a time-to-event endpoint, interpretation also depends on censoring and the underlying event-time structure. A single hazard ratio summarizes the relative event hazard and should not automatically be interpreted as a constant relative risk at every time point without considering the proportional-hazards assumption.

8. Primary Results: Disease-Free Survival in Stage IB-IIIA Patients

The second posted primary analysis evaluated DFS in the Full Analysis Set (FAS) of stage IB-IIIA patients. This analysis uses the broader stage population specified in the registry result.

DFS hazard ratio

0.27

95% CI: 0.21–0.34   ·   P < 0.0001

Two-sided 95% confidence interval; log-rank test stratified by stage, race and mutation type.

ElementPosted analysis
EndpointDisease-free survival (DFS)
PopulationFull Analysis Set (FAS), stage IB-IIIA patients
ComparisonAZD9291 80 mg Tablet vs Placebo Tablet
MethodLog-rank test
Effect measureHazard ratio
Estimate0.27
95% CI0.21–0.34
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

A hazard ratio of 0.27 means that the estimated hazard of a DFS event in the AZD9291 group was 27% of the estimated hazard in the placebo group under the reported analysis. As a simple interpretation of the point estimate, this corresponds to an approximately 73% lower estimated hazard.

Again, this is not a statement that 73% of patients were protected from recurrence or death. Hazard ratios compare event rates over time; they do not directly report absolute survival probabilities or the number of patients who benefit.

The two-sided 95% confidence interval of 0.21–0.34 describes uncertainty around the estimated relative treatment effect. It is important to retain the distinction between statistical precision and clinical meaning: a precise estimate can still require clinical context about the endpoint, follow-up, adverse effects, and applicability to other populations.

The p-value of <0.0001 provides evidence against the null hypothesis under the reported superiority testing framework. It should not be read as a measure of how large or clinically important the treatment effect is.

The analysis was stratified by stage, race and mutation type. That detail matters because the reported hazard ratio and log-rank result are not simply an unadjusted comparison that ignores those factors.

9. Comparing the Two Primary DFS Analyses

The two primary analyses address overlapping but different analysis populations. The stage IIA-IIIA analysis reports an HR of 0.23, while the stage IB-IIIA analysis reports an HR of 0.27.

Primary DFS analysisAnalysis populationHR95% CIP-value
Analysis 1FAS stage IIA-IIIA0.230.18–0.30<0.0001
Analysis 2FAS stage IB-IIIA0.270.21–0.34<0.0001

The estimates are close in direction and magnitude, but they should not be treated as two independent patient populations without recognizing their overlap. The broader stage IB-IIIA analysis necessarily contains the stage IIA-IIIA population plus the stage IB and stage I patients represented in that analysis set.

Important statistical distinction: comparing 0.23 and 0.27 informally is not the same as testing whether the treatment effect differs between the two stage populations. A formal comparison of treatment effects would require an appropriate interaction or heterogeneity analysis, which is not reported in the ClinicalTrials.gov record.

10. Secondary Results: Overall Survival

Overall survival was reported as a secondary time-to-event endpoint. The registry provides two formal analyses, one in stage IIA-IIIA patients and one in the broader stage IB-IIIA population.

Overall Survival: Stage IIA-IIIA

OS hazard ratio

0.4913

95.03% CI: 0.3307–0.7299   ·   P = 0.0004

Two-sided confidence interval; log-rank test stratified by stage, race and mutation type.

ElementPosted analysis
EndpointOverall survival (OS)
PopulationFull Analysis Set (FAS), stage IIA-IIIA patients
ComparisonAZD9291 80 mg Tablet vs Placebo Tablet
MethodLog-rank test
Effect measureHazard ratio
Estimate0.4913
Confidence interval95.03% CI: 0.3307–0.7299
P-value0.0004
HypothesisSuperiority

Overall Survival: Stage IB-IIIA

OS hazard ratio

0.4912

95.03% CI: 0.3439–0.7017   ·   P < 0.0001

Two-sided confidence interval; log-rank test stratified by stage, race and mutation type.

ElementPosted analysis
EndpointOverall survival (OS)
PopulationFull Analysis Set (FAS), stage IB-IIIA patients
ComparisonAZD9291 80 mg Tablet vs Placebo Tablet
MethodLog-rank test
Effect measureHazard ratio
Estimate0.4912
Confidence interval95.03% CI: 0.3439–0.7017
P-value<0.0001
HypothesisSuperiority
Clinical Biostats interpretation

The OS hazard ratios of 0.4913 and 0.4912 indicate that the estimated hazard of death was approximately one-half in the AZD9291 group relative to the placebo group under the respective analyses. As a simple transformation of the point estimates, these correspond to approximately 50.87% and 50.88% lower estimated hazards, respectively.

The confidence intervals quantify uncertainty around those estimates. For the stage IIA-IIIA analysis, the 95.03% CI is 0.3307–0.7299. For the stage IB-IIIA analysis, it is 0.3439–0.7017. Neither interval represents the range of survival times for individual patients.

The p-values provide evidence against the null hypothesis within the reported superiority testing framework. They should not be interpreted as the probability that AZD9291 works, nor as a measure of the clinical size of the treatment effect.

As with DFS, OS is subject to censoring and time-to-event modeling considerations. A hazard ratio summarizes a relative event hazard and should not automatically be interpreted as a constant relative reduction at every point in follow-up without considering the proportional-hazards assumption.

11. Primary and Secondary Results at a Glance

EndpointAnalysis populationHRCIP-value
DFS FAS stage IIA-IIIA 0.23 95% CI 0.18–0.30 <0.0001
DFS FAS stage IB-IIIA 0.27 95% CI 0.21–0.34 <0.0001
OS FAS stage IIA-IIIA 0.4913 95.03% CI 0.3307–0.7299 0.0004
OS FAS stage IB-IIIA 0.4912 95.03% CI 0.3439–0.7017 <0.0001

All four posted analyses favor AZD9291 according to the registry's stated interpretation that a hazard ratio below 1 favors AZD9291. The two DFS analyses are primary-endpoint analyses; OS is a secondary endpoint.

12. Safety Results

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

Safety measureAZD9291Placebo
Serious adverse events68 / 33747 / 343

The denominators differ from the overall enrollment of 682 because the serious-adverse-event measure is reported using the affected and at-risk counts reported by the registry. The ClinicalTrials.gov record does not provide enough additional detail to reconstruct a broader adverse-event table or to determine the specific types of serious adverse events.

Safety interpretation: serious adverse events are a separate evidence domain from the time-to-event efficacy analyses. The DFS and OS hazard ratios cannot be combined mathematically with serious-adverse-event counts to produce a single benefit-risk statistic.

13. Statistical Methods Explained

Why was a log-rank test used?

The registered primary endpoint is time-to-event DFS, so the comparison must account not only for whether an event occurred but also for when it occurred and for censored observations. The log-rank test is a standard method for comparing survival distributions between randomized groups. ADAURA's posted analyses specifically identify the log-rank test and state that it was stratified by stage, race and mutation type.

What does a hazard ratio of 0.23 mean?

A hazard ratio of 0.23 means that the estimated hazard of the DFS event in the AZD9291 group was 23% of the estimated hazard in the placebo group under the reported analysis. It does not mean that 23% of patients had an event, and it does not mean that every patient experienced the same relative reduction.

Why are confidence intervals important?

A point estimate is only one estimate of the treatment effect. The 95% confidence interval provides information about the uncertainty surrounding that estimate. In ADAURA, the stage IIA-IIIA DFS estimate of 0.23 has a 95% CI of 0.18–0.30, while the stage IB-IIIA estimate of 0.27 has a 95% CI of 0.21–0.34.

Why doesn't the p-value measure effect size?

The p-value measures the strength of evidence against a null hypothesis under the specified statistical model and testing procedure. It depends on both the observed data and the amount of information available. Effect size is instead conveyed by the hazard ratio, while its uncertainty is conveyed by the confidence interval.

Why was the analysis stratified?

The registry states that the log-rank analyses were stratified by stage, race and mutation type. Stratification permits the comparison to account for those factors when evaluating the time-to-event difference. It is different from reporting separate treatment effects for each stratum.

What does intention-to-treat mean here?

The registry identifies intention-to-treat analysis as a concept in the posted analyses. In a randomized trial, the intention-to-treat principle preserves the treatment assignment created by randomization. This helps maintain the causal comparison established at randomization even when patients' subsequent treatment experience differs.

Why does the proportional-hazards assumption matter?

A hazard ratio is a compact summary of relative event hazards over time. If the relative hazards change substantially over follow-up, a single hazard ratio can hide important features of the survival distributions. The ClinicalTrials.gov record reports hazard ratios but do not provide a formal proportional-hazards diagnostic, so that assumption should be regarded as an interpretive consideration rather than an established finding from the ClinicalTrials.gov record.

14. Analysis Populations and Their Interpretation

The posted statistical analyses use the Full Analysis Set (FAS), with separate analyses for stage IIA-IIIA and stage IB-IIIA patients. This distinction is central to interpreting the numerical results.

EndpointAnalysis populationWhy it matters
DFSFAS stage IIA-IIIAPrimary analysis restricted to the specified stage population.
DFSFAS stage IB-IIIAPrimary analysis using the broader registered stage population.
OSFAS stage IIA-IIIASecondary analysis using the same narrower stage population.
OSFAS stage IB-IIIASecondary analysis using the broader stage population.

Because the two stage populations overlap, their hazard ratios should not be treated as if they arose from completely independent experiments. Differences between 0.23 and 0.27, or between 0.4913 and 0.4912, should be interpreted in light of the populations being analyzed rather than as direct evidence of a change in treatment effect.

15. Interim Analysis and Data Collection

The registry contains an important design caveat: the primary analysis was performed two years early following an IDMC recommendation. It also states that an exploratory analysis of final DFS data at the prespecified maturity of approximately 50% was performed in 2022.

Early primary analysis

The primary analysis occurred two years earlier than planned according to the registry's stated chronology, following an IDMC recommendation.

Exploratory final DFS analysis

The registry states that final DFS data were examined exploratorily at the prespecified maturity of approximately 50% in 2022.

Later data collection

After the final DFS analysis and until the final OS analysis, only limited data were collected.

Types of later data

The registry identifies subsequent anti-cancer treatment, survival status and limited safety data as the later collected information.

This chronology matters because clinical-trial results are tied to information accumulated at particular data cutoffs. Combining estimates from different analysis periods without identifying their respective maturities can create a misleading impression of a single continuously observed dataset.

16. Censoring and Time-to-Event Interpretation

DFS and OS are not ordinary binary endpoints. Each patient contributes information over time, and patients who have not experienced the event by the end of available follow-up may be censored.

Conceptual survival framework
Time-to-event outcome = event occurrence + event timing + censoring information

This is why survival methods can use information from patients who have not yet experienced an event without treating those patients as though they had experienced no event indefinitely.

For DFS, the event definition reported by the registry is disease recurrence or death by any cause in the absence of recurrence. For OS, the endpoint is overall survival, with the registry specifying a follow-up period of up to approximately 7 years after the first patient is randomized and a maximum follow-up of 86 months.

The ClinicalTrials.gov record does not provide median DFS, median OS, event counts, Kaplan-Meier estimates at specific time points, or detailed censoring patterns. Those quantities therefore are not presented here.

17. Multiplicity, Superiority, and What the Registry Does Not Report

The posted analyses identify superiority as the hypothesis type for both DFS and OS analyses. The ClinicalTrials.gov record does not specify a non-inferiority margin, factorial design, crossover analysis, Bayesian model, missing-data imputation procedure, or formal multiplicity-adjustment strategy.

Design topicWhat can be established from the ClinicalTrials.gov record
SuperiorityReported for the posted DFS and OS analyses.
Non-inferiority marginNot provided; this was not a reported non-inferiority analysis.
Factorial designNot reported; the design model is parallel.
CrossoverNot reported in the ClinicalTrials.gov record.
Bayesian methodsNot reported in the statistical analyses posted on ClinicalTrials.gov.
Missing-data / imputation methodNot reported in the statistical analyses posted on ClinicalTrials.gov.
StratificationReported: stage, race and mutation type.
Interim analysisRegistry states that the primary analysis was performed two years early following an IDMC recommendation.

This distinction is important. Absence of a method from the ClinicalTrials.gov record should not be converted into a claim that the method was not used. It means only that the ClinicalTrials.gov record does not establish it.

18. Understanding the DFS Effect Estimates

Stage IIA-IIIA

The DFS HR of 0.23 corresponds to an estimated hazard approximately 77% lower in the AZD9291 group than in the placebo group, using the simple transformation 1 − HR. The 95% CI is 0.18–0.30 and the p-value is <0.0001.

Stage IB-IIIA

The DFS HR of 0.27 corresponds to an estimated hazard approximately 73% lower in the AZD9291 group than in the placebo group. The 95% CI is 0.21–0.34 and the p-value is <0.0001.

Relative versus absolute effects: the ClinicalTrials.gov record reports hazard ratios, confidence intervals and p-values, but do not provide median DFS, median OS, absolute DFS probabilities, or absolute OS probabilities. It is therefore not appropriate to manufacture an absolute treatment benefit from the hazard ratios alone.

19. What the OS Results Add

DFS is the registered primary endpoint, while OS is a secondary endpoint. The OS analyses provide a different time-to-event perspective because the event is death rather than disease recurrence or death.

PopulationDFS HROS HR
FAS stage IIA-IIIA0.230.4913
FAS stage IB-IIIA0.270.4912

The fact that the OS hazard ratios are approximately 0.49 while the DFS hazard ratios are approximately 0.23–0.27 does not indicate a statistical inconsistency. DFS and OS are different endpoints with different event definitions and different follow-up structures. A treatment can have different relative effects on different clinical endpoints.

Likewise, the OS results should not be used to retroactively redefine DFS as the trial's primary endpoint. The registry explicitly identifies DFS as the registered primary endpoint and OS as a secondary endpoint.

20. Limitations

21. Why This Trial Matters Statistically

ADAURA is a useful teaching case because it combines a randomized phase 3 design with a primary time-to-event endpoint, stratified survival testing, hazard-ratio estimation, multiple analysis populations, an early primary analysis, and later exploratory DFS and OS analyses.

ConceptHow it appears in ADAURA
RandomizationThe study uses randomized allocation in a parallel design.
BlindingThe registered masking designation is triple.
Time-to-event endpointDFS is the registered primary endpoint; OS is secondary.
Log-rank testReported as the method for the posted DFS and OS analyses.
Hazard ratioUsed as the reported effect measure for all four posted analyses.
Confidence intervalTwo-sided confidence intervals accompany the reported effect estimates.
Stratified analysisAnalyses were stratified by stage, race and mutation type.
Intention-to-treatIdentified as an analysis concept in the posted statistical analyses.
SuperiorityThe registry identifies superiority as the hypothesis type.
Interim analysisThe registry states that the primary analysis occurred two years early following an IDMC recommendation.
Analysis maturityAn exploratory final DFS analysis was performed at approximately 50% maturity in 2022.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue exploring clinical-trial statistics

Connect the endpoints and methods in ADAURA with deeper statistical tutorials, calculators, and other clinical trial analyses.

25. Record Summary

ADAURA provides a clear example of randomized time-to-event analysis. The registered primary endpoint was disease-free survival, and the posted primary analyses used stratified log-rank tests with hazard ratios as the effect measure. In the Full Analysis Set, the stage IIA-IIIA analysis reported a DFS HR of 0.23 (95% CI 0.18–0.30; P < 0.0001), while the stage IB-IIIA analysis reported an HR of 0.27 (95% CI 0.21–0.34; P < 0.0001). Secondary overall-survival analyses reported HRs of 0.4913 and 0.4912 in the corresponding populations. The statistical interpretation must also account for the different analysis populations, the early primary analysis described by the registry, the exploratory final DFS analysis at approximately 50% maturity, and the limited later data collection described in the registry.

Clinical Biostats methodology: A trial-results page should distinguish reported numerical evidence from statistical interpretation. For ADAURA, that means preserving the registered endpoint definition, analysis population, stratification factors, effect estimates, confidence intervals and p-values while avoiding unsupported reconstruction of median survival, subgroup results, baseline characteristics, or other statistical details not contained in the ClinicalTrials.gov record.