← Clinical Trials
Small Cell Lung Cancer Phase 3 Time-to-Event Analysis NCT03703297

ADRIATIC: Complete Statistical Analysis of Durvalumab in Limited-Stage Small-Cell Lung Cancer

An independent statistical analysis of the randomized phase 3 ADRIATIC trial evaluating durvalumab, tremelimumab plus durvalumab, and placebo in patients with limited-stage small-cell lung cancer who had not progressed following concurrent chemoradiation therapy.

Trial start: 27 September 2018  ·  Primary completion: 15 January 2024  ·  Enrollment: 730
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the results posted in the ClinicalTrials.gov record.

1. Trial at a Glance

ADRIATIC is a randomized, triple-masked, phase 3 parallel-group trial in small-cell lung cancer. The study enrolled 730 participants and evaluated three treatment groups involving durvalumab, tremelimumab, and placebo. The posted primary analyses compare durvalumab with placebo for progression-free survival and overall survival.

730
Enrollment
Randomized phase 3 trial
3
Treatment Arms
Parallel-group design
0.76
PFS HR
95% CI 0.606–0.950
0.73
OS HR
95% CI 0.569–0.928
FeatureADRIATIC
TrialADRIATIC
ClinicalTrials.govNCT03703297
PhasePhase 3
ConditionSmall Cell Lung Cancer
PopulationPatients with limited-stage small-cell lung cancer who have not progressed following concurrent chemoradiation therapy
AllocationRandomized
Design modelParallel
MaskingTriple
Primary purposeTreatment
Enrollment730
Primary endpointsProgression-free survival and overall survival
Primary endpoint typeTime-to-event
Primary hypothesisSuperiority
Posted primary analyses2
Lead sponsorAstraZeneca
Status in the ClinicalTrials.gov recordActive, not recruiting

2. Clinical Question

The central statistical question is whether durvalumab improves time-to-event outcomes compared with placebo in patients with limited-stage small-cell lung cancer who have not progressed following concurrent chemoradiation therapy.

Population

Patients with limited-stage small-cell lung cancer who have not progressed following concurrent chemoradiation therapy.

Intervention of the posted comparison

Durvalumab.

Comparator

Placebo.

Primary question

Does durvalumab produce a superior time-to-event outcome relative to placebo for the registered primary endpoints?

Three-arm design, two-arm posted primary analysis. ADRIATIC has three treatment arms involving durvalumab, tremelimumab plus durvalumab, and placebo. The statistical analyses in the ClinicalTrials.gov record are specifically the two primary comparisons of durvalumab versus placebo. The registry caveat states that the durvalumab plus tremelimumab arm was blinded at the time of the interim results posting and that results for that arm will be posted at final analysis.

3. Trial Design

01
Randomize730 enrolled
02
Three armsDurvalumab / D+T / placebo
03
Follow-upTime-to-event outcomes
04
BICRPFS assessment
05
Primary analysisPFS and OS
Allocation
Randomized.
Structure
Parallel-group, three-arm phase 3 trial.
Masking
Triple masking.
Primary purpose
Treatment.
Study start
27 September 2018.
Primary completion
15 January 2024.
ARM 1 · POSTED PRIMARY COMPARISON

Durvalumab

  • Active treatment arm.
  • Compared directly with placebo in both posted primary endpoint analyses.
  • PFS assessed by blinded independent central review.
  • OS assessed from randomization until death due to any cause.
ARM 2 · BLINDED AT INTERIM POSTING

Durvalumab + Tremelimumab

  • Combination arm involving durvalumab and tremelimumab.
  • Results were not posted in the registry-reported interim-analysis data.
  • The registry caveat states that this arm remained blinded to the Sponsor according to the prespecified unblinding plan.
  • Results for this arm are to be posted at final analysis.
ARM 3 · POSTED PRIMARY COMPARATOR

Placebo

  • Comparator arm for the posted durvalumab analyses.
  • Used as the reference group for the reported hazard ratios.

4. Endpoints

The ClinicalTrials.gov record identifies two primary endpoints, both of which are time-to-event outcomes. Both have formal statistical analyses posted.

EndpointRegistered definitionTime framePrimary analysis
Durvalumab Versus Placebo: Progression-Free Survival (PFS) Assessed by Blinded Independent Central Review (BICR) Per Response Evaluation Criteria in Solid Tumors Version 1.1 PFS per RECIST 1.1 assessed by BICR was defined as the time from the date of randomization until the date of objective disease progression (PD) or death (by any cause in the absence of progression) regardless of whether the participant withdrew from therapy or received another anticancer therapy prior to progression. Response evaluations performed every 8 weeks (q8w) ± 1 week up to 72 weeks, then every 12 weeks (q12w) ± 1 week up to 96 weeks. Stratified log-rank test; stratified Cox proportional-hazards model for HR and CI.
Durvalumab Versus Placebo: Overall Survival (OS) OS was defined as the time from the date of randomization until death due to any cause. Median OS was calculated using Kaplan-Meier method and its CI using Brookmeyer-Crowley method. From date of randomization until death due to any cause, up to DCO date 15 January 2024. Stratified log-rank test; stratified Cox proportional-hazards model for HR and CI.
Endpoint structure matters. Both primary endpoints begin at randomization and measure the time until an event. That means the analysis must account for patients who have not yet experienced the event at the time of analysis. These patients contribute follow-up information until they are censored rather than being treated as if they experienced the event.

5. Analysis Populations and Statistical Structure

The ClinicalTrials.gov record states that the full analysis set (FAS) included all randomized participants. Both posted primary endpoint analyses therefore use the randomized population rather than restricting the efficacy comparison to patients who completed treatment.

ElementRegistry-supported description
Analysis populationFAS included all randomized participants.
PFS comparisonDurvalumab versus placebo.
OS comparisonDurvalumab versus placebo.
Primary hypothesisSuperiority.
Primary methodLog-rank test.
Effect measureHazard ratio.
Confidence interval95%, two-sided.
ModelStratified Cox proportional-hazards model.

This distinction is important because a randomized trial's causal comparison is anchored to the treatment assignment generated by randomization. Restricting the primary efficacy analysis to patients who remain on treatment can introduce selection because continued treatment is itself influenced by post-randomization events.

6. Statistical Methodology

Stratified log-rank testing

The posted primary analyses used a stratified log-rank test. The log-rank test compares the observed and expected numbers of events between treatment groups over the follow-up period. Stratification allows the comparison to account for prespecified grouping variables without simply pooling all event times into one unstratified comparison.

Stratified Cox proportional-hazards model

The hazard ratio and its confidence interval were calculated using a stratified Cox proportional-hazards model. The registry-reported analysis notes identify covariate adjustment and stratified analysis as concepts in both primary analyses.

Conceptual hazard-ratio model
HR = estimated hazard in the durvalumab group ÷ estimated hazard in the placebo group

A hazard ratio below 1 indicates a lower estimated instantaneous event rate for durvalumab relative to placebo under the fitted time-to-event model. It is a relative measure of event hazard, not an absolute probability of experiencing the event.

Efron handling of ties

For both primary analyses, ties in event times were handled using the Efron approach. This is a detail of the Cox-model estimation procedure used when multiple participants have events at the same recorded time.

Covariate adjustment in the PFS model

For PFS, the stratified Cox model adjusted for tumor, node and metastasis (TNM) stage and receipt of prophylactic cranial irradiation (PCI), with treatment as the only covariate. This means the treatment hazard ratio is estimated within a model that accounts for those specified factors.

Covariate adjustment in the OS model

For OS, the stratified Cox model adjusted for receipt of PCI, with treatment as the only covariate. The difference between the PFS and OS model specifications is therefore relevant when interpreting the two reported hazard ratios.

Kaplan-Meier estimation

The registry explicitly states that median OS was calculated using the Kaplan-Meier method and that its confidence interval was calculated using the Brookmeyer-Crowley method. Kaplan-Meier estimation is also the natural descriptive framework for time-to-event data because it retains information from participants who are censored before an event occurs.

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

Here, di represents events at an event time and ni represents the number at risk immediately before that time. The estimator updates the probability of remaining event-free as observed events accumulate.

7. Results: Progression-Free Survival

The first posted primary analysis evaluated PFS for durvalumab versus placebo. The analysis used the full analysis set, which included all randomized participants, and the registry reports a stratified log-rank test with a stratified Cox proportional-hazards model for the hazard ratio and confidence interval.

Hazard ratio for progression or death

0.76

95% CI: 0.606–0.950   ·   P = 0.01608

Two-sided 95% confidence interval · Superiority hypothesis

ElementPFS result
ComparisonDurvalumab versus placebo
EndpointPFS assessed by BICR per RECIST 1.1
Analysis populationFull analysis set; all randomized participants
MethodStratified log-rank test
Effect measureHazard ratio
Estimate0.76
95% CI0.606–0.950
P-value0.01608
ModelStratified Cox proportional-hazards model, adjusting for TNM stage and receipt of PCI; treatment as only covariate; Efron method for ties
Clinical Biostats interpretation

The estimated hazard ratio of 0.76 means that, under the fitted time-to-event model, the estimated instantaneous rate of progression or death in the durvalumab group was approximately 24% lower than in the placebo group. This is a relative hazard interpretation, not a statement that 24% of patients avoided progression or death.

The 95% confidence interval of 0.606–0.950 describes statistical uncertainty around the estimated hazard ratio under the analysis model and sampling framework. Because the interval lies below 1, the interval is consistent with a lower estimated hazard for durvalumab relative to placebo across the interval represented by the model.

The P-value of 0.01608 addresses the statistical evidence against the null hypothesis under the specified testing framework. It does not measure the size of the treatment effect. A smaller P-value is not equivalent to a larger clinical effect, and the hazard ratio and its confidence interval are needed to describe the magnitude and precision of the estimated effect.

The analysis is also subject to the usual interpretation issues for Cox models. The hazard ratio summarizes relative event hazards through the model rather than directly reporting absolute differences in survival probability. Its interpretation is strongest when the proportional-hazards framework is a reasonable description of the data.

The ClinicalTrials.gov record does not provide median PFS, PFS event counts by treatment group, or a complete Kaplan-Meier curve. Those quantities are therefore not added to this page.

8. Results: Overall Survival

The second posted primary analysis evaluated overall survival, defined as time from randomization until death due to any cause. As with PFS, the analysis used the full analysis set and compared durvalumab with placebo.

Hazard ratio for death

0.73

95% CI: 0.569–0.928   ·   P = 0.01042

Two-sided 95% confidence interval · Superiority hypothesis

ElementOS result
ComparisonDurvalumab versus placebo
EndpointOverall survival
Analysis populationFull analysis set; all randomized participants
MethodStratified log-rank test
Effect measureHazard ratio
Estimate0.73
95% CI0.569–0.928
P-value0.01042
ModelStratified Cox proportional-hazards model, adjusting for receipt of PCI; treatment as only covariate; Efron method for ties
Clinical Biostats interpretation

The estimated OS hazard ratio of 0.73 means that, under the fitted Cox model, the estimated instantaneous rate of death in the durvalumab group was approximately 27% lower than in the placebo group. It does not mean that durvalumab reduces every patient's probability of death by 27%, nor does it mean that 27% of patients experience a particular survival benefit.

The 95% CI of 0.569–0.928 quantifies uncertainty around the estimated hazard ratio. The interval remains below 1, so the reported confidence interval is compatible with a lower estimated hazard of death for durvalumab relative to placebo under the specified analysis.

The P-value of 0.01042 quantifies evidence against the null hypothesis in the stated statistical framework. It does not quantify the magnitude, clinical importance, or probability that the treatment effect is exactly the observed hazard ratio.

OS is also an endpoint that can be influenced by events occurring after the initial randomized treatment period. The registry definition counts death from any cause, making OS a measure of the overall survival experience after randomization rather than a direct measure of treatment exposure alone.

The ClinicalTrials.gov record does not provide a median OS estimate, OS event counts by treatment group, or time-specific survival probabilities. Those quantities are not inferred from the hazard ratio.

9. Primary Results Side by Side

Primary endpointEstimate95% CIP-valuePrimary method
PFS HR 0.76 0.606–0.950 0.01608 Stratified log-rank test
OS HR 0.73 0.569–0.928 0.01042 Stratified log-rank test

Both posted primary analyses have hazard ratios below 1, and both reported two-sided P-values are below 0.05. Statistically, the two analyses therefore point in the same direction: the durvalumab group had a lower estimated event hazard than the placebo group for the two posted primary time-to-event endpoints.

That concordance is informative, but the two estimates should not be collapsed into a single overall treatment statistic. PFS and OS are different endpoints, have different event definitions, and use somewhat different covariate-adjustment specifications in the reported Cox models.

PFS: what it captures

PFS captures the time from randomization to objective disease progression or death in the absence of progression, with progression assessed by BICR according to RECIST 1.1.

OS: what it captures

OS captures the time from randomization to death due to any cause. It is not dependent on radiologic progression assessment.

Relative effect

The hazard ratio summarizes the relative event hazard between randomized groups under the fitted model.

Precision

The confidence interval communicates uncertainty around each estimated hazard ratio and should be read alongside the point estimate.

10. Statistical Methods Explained

Why was a log-rank test used?

The primary endpoints are time-to-event outcomes. A log-rank test is designed to compare event-time distributions between randomized groups while incorporating the timing of events and allowing observations to be censored. It is therefore more appropriate for these endpoints than a simple comparison of proportions at one arbitrary time point.

What does an HR of 0.76 mean for PFS?

An HR of 0.76 indicates an estimated instantaneous event hazard for progression or death that is approximately 24% lower for durvalumab than placebo under the fitted model. It is not the same as saying that PFS time is 24% longer, nor does it directly provide the percentage of patients who remain progression-free at a particular time.

What does an HR of 0.73 mean for OS?

An HR of 0.73 indicates an estimated instantaneous hazard of death approximately 27% lower for durvalumab relative to placebo under the reported Cox model. The interpretation concerns relative hazard over follow-up, not an absolute reduction in mortality probability.

Why are confidence intervals important?

A point estimate such as 0.73 is only one estimate of the underlying treatment effect. The 95% CI of 0.569–0.928 shows the statistical uncertainty around that estimate under the specified model and study framework. It provides information about precision that a P-value alone cannot provide.

Why doesn't the P-value measure treatment effect size?

The P-value evaluates evidence against a null hypothesis within a specified testing framework. It depends on both the estimated effect and the amount of information in the data. The hazard ratio tells us the estimated relative effect; the confidence interval tells us about its precision; the P-value addresses statistical evidence against the null.

Why use a stratified Cox model?

The analyses posted on ClinicalTrials.gov identify stratification and covariate adjustment as components of the analysis. Stratification allows the survival comparison to account for specified factors while preserving the randomized treatment comparison. For PFS, the model adjusts for TNM stage and receipt of PCI; for OS, it adjusts for receipt of PCI.

Why does the analysis population matter?

The full analysis set includes all randomized participants. Analyzing randomized participants according to the treatment comparison helps preserve the balance created by randomization. In time-to-event trials, excluding patients after randomization can change the composition of the groups and potentially compromise the original comparison.

11. Confidence Intervals and the Meaning of the Reported Precision

The two primary estimates have confidence intervals that extend below 1 but are not especially narrow. This provides a useful statistical lesson: an analysis can provide evidence of a treatment difference while still leaving meaningful uncertainty about the exact magnitude of that difference.

EndpointHRLower 95% CIUpper 95% CIWidth of reported interval
PFS0.760.6060.9500.344
OS0.730.5690.9280.359

The confidence intervals should not be read as a range in which an individual patient's treatment effect is expected to fall. They instead describe uncertainty around the population-level hazard-ratio estimate under the statistical model.

It is also important not to treat the lower endpoint of a confidence interval as a minimum guaranteed benefit. For example, an upper confidence limit of 0.950 does not mean that every future study would necessarily observe a hazard ratio below 0.950. It describes the uncertainty associated with this particular analysis.

12. Stratification and Covariate Adjustment

The registry analysis notes explicitly identify covariate adjustment and stratified analysis. The PFS and OS models do not use exactly the same adjustment specification, which is an important methodological detail.

EndpointStratified Cox model adjustmentTreatment variableTie handling
PFS TNM stage and receipt of PCI Treatment as only covariate Efron approach
OS Receipt of PCI Treatment as only covariate Efron approach

Why adjustment can improve a time-to-event analysis

Covariate adjustment can account for prognostic factors included in the prespecified analysis model. In a randomized trial, randomization remains the foundation of the treatment comparison; adjustment does not transform the study into an observational analysis.

For PFS, adjustment for TNM stage and receipt of PCI means that the reported hazard ratio comes from a model incorporating those factors. For OS, the registry-reported analysis notes specify adjustment for receipt of PCI. Consequently, the PFS and OS hazard ratios are not generated from identical model specifications.

Interpretive caution: adjustment does not make the hazard ratio a patient-level prediction. It produces an adjusted estimate under a particular statistical model. The meaning of the estimate remains tied to the endpoint definition, analysis population, censoring rules, and model assumptions.

13. The Full Analysis Set and Randomization

The ClinicalTrials.gov record defines the FAS as including all randomized participants. This is important because the trial is designed to answer a comparative question created by random assignment.

Randomization creates comparability

Randomization is intended to balance treatment groups with respect to measured and unmeasured characteristics in expectation.

FAS preserves the randomized comparison

Including all randomized participants keeps the efficacy analysis tied to the treatment assignment rather than subsequent treatment exposure.

Time-to-event follow-up

Participants can contribute information even when an event has not occurred by the end of their available follow-up.

Censoring is not an event

A censored participant contributes follow-up through the censoring time but is not counted as having experienced the endpoint event.

14. Safety Results

The ClinicalTrials.gov record reports serious adverse events for the durvalumab and placebo groups. These figures are presented as affected participants divided by participants at risk.

ArmParticipants with serious adverse eventsParticipants at riskRegistry expression
Durvalumab7826278/262
Placebo6426564/265

The registry therefore reports 78/262 participants affected in the durvalumab arm and 64/265 in the placebo arm. The ClinicalTrials.gov record does not provide a corresponding serious-adverse-event figure for the durvalumab plus tremelimumab arm in the posted results used here.

Do not overextend the safety comparison. The serious-adverse-event counts are safety observations, not efficacy outcomes. They should not be combined with the PFS or OS hazard ratios into a single numerical "benefit-risk" statistic. The denominator also matters: 78/262 and 64/265 describe affected participants relative to their respective reported at-risk populations.

15. Interim Analysis and Blinding of the Combination Arm

The registry caveat is particularly important for interpreting the scope of the posted results. It states that, based on results of the secondary objectives at the interim analysis reviewed by an unblinded independent data monitoring committee (IDMC), data for the durvalumab plus tremelimumab arm were to remain blinded to the Sponsor according to the prespecified study unblinding plan.

2018 · Trial start

ADRIATIC begins

The ClinicalTrials.gov record records a study start date of 27 September 2018.

Interim analysis · IDMC review

Combination-arm data remained blinded

The registry states that an unblinded IDMC reviewed secondary-objective results and that the durvalumab plus tremelimumab data remained blinded to the Sponsor under the prespecified unblinding plan.

15 January 2024 · Primary completion

Registry primary-completion date

The ClinicalTrials.gov record records 15 January 2024 as the primary completion date and uses this date as the OS data-cutoff date in the posted endpoint description.

Results posting

Durvalumab versus placebo results available

The posted primary analyses provide formal PFS and OS comparisons of durvalumab versus placebo.

Why the blinding caveat matters: the posted results should not be interpreted as a complete three-arm efficacy comparison. The ClinicalTrials.gov record explicitly states that results for the durvalumab plus tremelimumab arm will be posted at final analysis because that arm was blinded at the time of the interim results posting.

16. What the Hazard Ratio Does — and Does Not — Mean

PFS hazard ratio

The PFS HR of 0.76 corresponds to an estimated 24% lower instantaneous hazard of progression or death for durvalumab versus placebo under the reported model.

It does not mean that 24% of patients were prevented from progressing, that PFS duration increased by exactly 24%, or that every patient received the same proportional benefit.

OS hazard ratio

The OS HR of 0.73 corresponds to an estimated 27% lower instantaneous hazard of death for durvalumab versus placebo under the reported model.

It does not mean that the probability of death was reduced by exactly 27% for every patient or that 27% of participants avoided death.

Confidence intervals

The PFS 95% CI of 0.606–0.950 and OS 95% CI of 0.569–0.928 describe uncertainty around the respective population-level hazard-ratio estimates. They do not describe the range of treatment effects among individual patients.

P-values

The PFS P-value of 0.01608 and OS P-value of 0.01042 quantify statistical evidence against their respective null hypotheses under the reported testing framework. They are not measures of effect size, clinical importance, or the probability that the null hypothesis is true.

17. Proportional-Hazards Considerations

The reported hazard ratios come from Cox proportional-hazards models. The proportional-hazards framework concerns how the relative hazard between treatment groups behaves over time.

A single hazard ratio is particularly easy to interpret when the relative hazards are reasonably represented by a common ratio over follow-up. If the treatment effect changes materially over time, one number can compress a more complicated pattern.

This is why a complete time-to-event interpretation normally considers the hazard ratio together with Kaplan-Meier curves, time-specific survival estimates, median event times where available, and the censoring pattern. The ClinicalTrials.gov record does not provide enough information to reproduce those additional displays, so this page does not infer them from the hazard ratios.

No Kaplan-Meier curve is reconstructed here. A valid curve requires the underlying event and censoring information or appropriately digitized source data; it should not be fabricated from the reported hazard ratio and P-value.

18. Multiplicity and the Three-Arm Structure

ADRIATIC has three treatment arms, but the statistical analyses posted on ClinicalTrials.gov contain two primary endpoint comparisons, both specifically comparing durvalumab versus placebo. The ClinicalTrials.gov record identifies the hypothesis type as superiority.

FeatureRegistry-supported interpretation
Number of arms3
Primary endpoints2
Posted primary comparisonsDurvalumab versus placebo for PFS and OS
Hypothesis typeSuperiority
Primary methodLog-rank test
Combination-arm status at interim postingBlinded to Sponsor; results to be posted at final analysis

Multiplicity is important whenever a trial contains multiple formal opportunities to declare an effect. However, the ClinicalTrials.gov record does not specify an alpha-allocation procedure, multiplicity-adjustment method, or endpoint-testing hierarchy. Those details should therefore not be inferred.

Statistical restraint is important here. The two reported P-values should be described exactly as posted. The available data does not establish how familywise type I error was allocated across the two primary endpoints or across the three-arm design. This page therefore does not assign an unstated multiplicity procedure to the trial.

19. Censoring and Time-to-Event Interpretation

Both primary endpoints depend on follow-up over time. Participants who have not experienced progression, death, or the relevant endpoint event by the time their follow-up ends may be censored rather than counted as events.

Why censoring is useful

It allows participants to contribute the information available before their follow-up ends rather than discarding their entire observation.

Why censoring needs assumptions

Time-to-event inference depends on assumptions concerning the relationship between censoring and the event process.

PFS event definition

The registered PFS definition counts objective progression or death in the absence of progression.

OS event definition

The registered OS definition counts death due to any cause.

The PFS definition also explicitly states that progression or death remains relevant regardless of whether a participant withdrew from therapy or received another anticancer therapy before progression. That feature makes PFS an endpoint based on the prespecified disease-event definition rather than simply time on assigned treatment.

20. Why This Trial Matters Statistically

ADRIATIC is a useful statistical teaching case because it combines a randomized three-arm phase 3 structure with two primary time-to-event endpoints and a posted analysis focused on a specific active-treatment-versus-placebo comparison.

ConceptHow it appears in ADRIATIC
RandomizationThe study uses randomized allocation.
Three-arm designDurvalumab, durvalumab plus tremelimumab, and placebo are the three intervention components represented in the registry profile.
BlindingThe trial is triple-masked.
Time-to-event endpointsPFS and OS are both primary endpoints.
Kaplan-Meier estimationMedian OS is specified to be calculated using the Kaplan-Meier method.
Log-rank testingBoth posted primary analyses use a stratified log-rank test.
Hazard ratioHR is the reported effect measure for both primary endpoints.
Cox modelHRs and CIs are calculated using stratified Cox proportional-hazards models.
Covariate adjustmentPFS adjusts for TNM stage and PCI; OS adjusts for PCI.
Analysis populationThe FAS includes all randomized participants.
Interim analysisThe registry-reported caveat describes an interim analysis reviewed by an unblinded IDMC.
Blinded armThe durvalumab plus tremelimumab arm remained blinded to the Sponsor when the interim results were posted.
Safety analysisSerious adverse events are reported for the durvalumab and placebo arms.

21. Important Limitations and Interpretation Issues

22. A Statistical Reading of the Two Primary Results

The most direct reading of the posted efficacy results is that both primary time-to-event analyses favor the durvalumab group relative to placebo in the reported statistical framework.

Two primary endpoints

PFS HR 0.76 · OS HR 0.73

PFS: 95% CI 0.606–0.950 · P = 0.01608

OS: 95% CI 0.569–0.928 · P = 0.01042

Several distinctions should be preserved. First, the estimates are relative hazards, not absolute treatment effects. Second, the confidence intervals describe statistical uncertainty around the estimated hazards. Third, the P-values describe evidence against the relevant null hypotheses, not the size of the benefit. Fourth, the results apply to the specific randomized comparison that was analyzed: durvalumab versus placebo.

The statistical evidence is therefore best represented as a set of related quantities rather than one number. The hazard ratio communicates relative effect, the confidence interval communicates precision, and the P-value communicates evidence under the prespecified testing framework. The endpoint definition determines what event is being compared.

23. Secondary Objectives and the Scope of Posted Results

The registry profile states that 25 outcome measures and 2 statistical analyses were posted. For this page, the registry-reported formal statistical analyses correspond to the two primary endpoints: PFS and OS for durvalumab versus placebo.

The registry caveat also states that the interim analysis was reviewed by an unblinded IDMC and that the durvalumab plus tremelimumab arm remained blinded to the Sponsor. Consequently, the absence of a posted combination-arm result in the registry-reported statistical-analyses data should not be interpreted as an outcome for that arm.

Interpretation rule: absence of a posted numerical result is not evidence of no treatment effect. Here, the registry explicitly supplies a design-based explanation: the durvalumab plus tremelimumab arm was intentionally kept blinded at the interim posting and is expected to have results posted at final analysis.

24. Related Tutorials

Learn more about the methods used in this trial:

25. Related Calculators

26. Sources

Continue through the Clinical Biostats statistical pathway

Use the trial's endpoints as a starting point for deeper study of survival analysis, hazard ratios, confidence intervals, randomization, and related clinical-trial methods.

27. Record Summary

ADRIATIC provides a clear example of modern randomized time-to-event analysis. The ClinicalTrials.gov record describes a randomized, triple-masked, three-arm phase 3 trial enrolling 730 participants, with PFS and OS as primary time-to-event endpoints. The posted formal analyses compare durvalumab with placebo using stratified log-rank tests and stratified Cox proportional-hazards models.

The reported PFS hazard ratio was 0.76 with a 95% CI of 0.606–0.950 and P = 0.01608. The reported OS hazard ratio was 0.73 with a 95% CI of 0.569–0.928 and P = 0.01042. Both estimates are below 1, corresponding to lower estimated event hazards for durvalumab relative to placebo under the reported models.

The statistical interpretation should remain anchored to what was actually analyzed. The FAS included all randomized participants; PFS used BICR assessment under RECIST 1.1 and adjustment for TNM stage and PCI; OS used adjustment for PCI; and ties were handled with the Efron approach. The combination arm involving durvalumab and tremelimumab remained blinded to the Sponsor at the interim results posting, with its results to be posted at final analysis.

Clinical Biostats methodology: A rigorous trial-results page separates the reported statistical evidence from interpretation. For ADRIATIC, that means preserving the exact endpoint definitions, analysis population, hazard ratios, confidence intervals, P-values, model specifications, and registry caveats rather than filling missing results from outside sources.