← Clinical Trials
Non-small Cell Lung Cancer Phase 3 Disease-Free Survival NCT02504372

PEARLS: Complete Statistical Analysis of Pembrolizumab in Non-small Cell Lung Cancer

An independent statistical review of the randomized phase 3 PEARLS trial evaluating pembrolizumab versus placebo for participants with non-small cell lung cancer after resection with or without standard adjuvant therapy, with disease-free survival analyzed using multivariate Cox regression.

Trial status: COMPLETED  ·  Enrollment: 1177  ·  Primary completion: 2023-01-24
Scope of this record

This page separates reported trial results from statistical interpretation. Trial-specific numerical values and endpoint definitions are taken from the ClinicalTrials.gov record. The registry provides the official trial record.

Registry record: This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. View NCT02504372 on ClinicalTrials.gov.

1. Trial at a Glance

PEARLS was a randomized, parallel, triple-masked phase 3 trial evaluating pembrolizumab versus placebo in participants with non-small cell lung cancer after resection, with or without standard adjuvant therapy. The registry reports disease-free survival as the principal efficacy outcome and provides Cox-model hazard ratios for the primary analyses.

1177
Enrollment
Randomized trial
2
Arms
Parallel design
0.81
Primary DFS HR
95% CI 0.68–0.96
0.83
TPS ≥50% DFS HR
95% CI 0.59–1.16
FeaturePEARLS
Trial namePEARLS
Brief titleStudy of Pembrolizumab (MK-3475) vs Placebo for Participants With Non-small Cell Lung Cancer After Resection With or Without Standard Adjuvant Therapy (MK-3475-091/KEYNOTE-091)
PhasePhase 3
ConditionNon-small Cell Lung Cancer
AllocationRandomized
Design modelParallel
MaskingTriple
Primary purposeTreatment
Enrollment1177.0
InterventionsPembrolizumab; Placebo
Lead sponsorMerck Sharp & Dohme LLC
Sponsor typeIndustry
StatusCompleted
ClinicalTrials.govNCT02504372

2. Clinical Question

The statistical question is whether disease-free survival differs between participants randomized to pembrolizumab and those randomized to placebo after resection, with or without standard adjuvant therapy.

Population

Participants with non-small cell lung cancer after resection, with or without standard adjuvant therapy.

Intervention

Pembrolizumab.

Comparator

Placebo.

Primary question

Does pembrolizumab produce a different disease-free survival experience than placebo?

3. Trial Design

01
Randomize1177 participants
02
Two armsPembrolizumab vs placebo
03
ParallelConcurrent randomized groups
04
Triple maskedRegistry-design classification
05
Follow-upDFS through approximately 84 months
ARM 1

Pembrolizumab

  • Biological intervention.
  • Randomized treatment assignment.
  • Included in the primary pembrolizumab-versus-placebo comparison.
ARM 2

Placebo

  • Placebo comparator.
  • Randomized treatment assignment.
  • Included in the primary pembrolizumab-versus-placebo comparison.
Allocation
Randomized allocation was used to create the treatment comparison.
Masking
The registry classifies the trial as triple masked.
Design model
Parallel-group design with two intervention groups.
Primary purpose
Treatment.

4. Endpoints

EndpointTime frameRegistry definitionAnalysis
Disease-Free Survival (DFS) Up to approximately 84 months DFS was defined as the time from randomization to either the date of disease recurrence or death (whatever the cause) as assessed by the investigator. Recurrence of disease was defined as local regional recurrence, a distant (metastatic) recurrence, or a second primary cancer. Occurrence of a second extra-pulmonary malignancy was considered to be an event. Multivariate Cox proportional-hazards model; hazard ratio
DFS in Programmed Death Ligand-1 (PDL-1) Strong Positive Participants With Tumor Proportion Score (TPS) ≥50% Up to approximately 84 months DFS in PDL-1 strong positive participants with TPS ≥50% was defined as the time from randomization to either the date of disease recurrence or death (whatever the cause) as assessed by the investigator. Recurrence of disease was defined as local regional recurrence, a distant (metastatic) recurrence, or a second primary cancer. Occurrence of a second extra-pulmonary malignancy was considered to be an event. Multivariate Cox proportional-hazards model; hazard ratio

the ClinicalTrials.gov record identifies the primary endpoint types as binary; time-to-event, while the individual registered primary endpoints in the ClinicalTrials.gov record are defined as time from randomization to recurrence or death. The endpoint-specific statistical analyses use Cox proportional-hazards models, consistent with a time-to-event framework.

5. Statistical Methodology

Time-to-event analysis

DFS is a time-to-event endpoint because the analysis concerns both whether a participant experiences recurrence or death and when that event occurs. Participants who have not experienced the defined event by the end of their available follow-up can contribute information up to their censoring time.

Conceptual survival framework
S(t) = P(T > t)

Here, T represents the time from randomization to the DFS event defined by the registry. Survival methods describe the distribution of event times while accounting for right censoring.

Cox proportional-hazards model

The primary analyses use a multivariate Cox proportional-hazards model. The model estimates the relative hazard associated with treatment while adjusting for stage, PD-L1 status, adjuvant chemotherapy, region, histology, and smoking status.

Cox model concept
h(t | X) = h0(t) exp(βTX)

The hazard ratio associated with treatment summarizes the estimated relative instantaneous event rate between the randomized treatment groups, conditional on the covariates included in the model.

Adjusted analysis

The reported hazard ratios and 95% confidence intervals were based on multivariate Cox regression with treatment adjusted by stage, PD-L1 status, adjuvant chemotherapy, region, histology, and smoking status. This means the reported estimate is not simply an unadjusted comparison of event-time distributions.

Analysis populations

For the overall primary DFS analysis, the registry identifies the analysis population as all randomized participants. One participant was randomized to the study in error, did not provide informed consent, and was not included; no data was collected on that participant.

For the TPS ≥50% primary endpoint, the analysis population was all randomized PD-L1 strong positive participants with TPS ≥50%. The registry text likewise notes that one participant randomized in error did not provide informed consent and was not included.

One-sided testing

For the TPS ≥50% primary analysis, the registry reports that the one-sided p-value was based on a permutation test with the multivariate Cox regression model. Importantly, the confidence interval reported for the hazard ratio is explicitly 95% and two-sided. Thus, the direction of the hypothesis test and the confidence-interval convention should not be conflated.

6. Primary Results: Disease-Free Survival

The registry reports a formal statistical analysis for the overall DFS endpoint through approximately 84 months. The comparison was pembrolizumab versus placebo in all randomized participants, with the hazard ratio and 95% confidence interval derived from the multivariate Cox regression model.

Primary DFS hazard ratio

HR 0.81

95% CI: 0.68–0.96   ·   Two-sided 95% confidence interval

Multivariate Cox regression adjusted for stage, PD-L1 status, adjuvant chemotherapy, region, histology, and smoking status.

CharacteristicReported value
EndpointDisease-Free Survival (DFS)
Time frameUp to approximately 84 months
Analysis populationAll randomized participants, excluding one participant randomized in error who did not provide informed consent and for whom no data were collected
Groups comparedPembrolizumab vs Placebo
MethodMultivariate Cox proportional-hazards model
Effect measureHazard ratio
Estimate0.81
95% CI0.68–0.96
CI conventionTwo-sided
P-valueNot reported in the registry-reported statistical analysis
Clinical Biostats interpretation

A hazard ratio of 0.81 means that, under the fitted Cox model, the estimated instantaneous rate of a DFS event in the pembrolizumab group was 0.81 times that in the placebo group, conditional on the model's covariate adjustment. Expressed as a relative model-based quantity, this corresponds to an estimated 19% lower hazard because 1 − 0.81 = 0.19.

The HR does not mean that 19% of participants avoided recurrence, that 19% more participants were disease-free, or that every participant experienced the same reduction in event risk. It is a relative hazard measure from a time-to-event model.

The 95% confidence interval of 0.68–0.96 describes uncertainty around the estimated hazard ratio under the statistical model and sampling framework. It does not describe the range of individual patient outcomes.

No p-value is reported for this particular analysis in the ClinicalTrials.gov record. The confidence interval therefore provides the registry-reported inferential information about precision and the location of the estimated treatment effect.

Because this is a Cox-model estimate, interpretation also depends on the proportional-hazards framework. A single hazard ratio is most straightforward when the relative hazards are reasonably stable over time; the ClinicalTrials.gov record does not provide a separate assessment of that assumption.

7. Primary Results: DFS in PD-L1 Strong Positive Participants With TPS ≥50%

The second primary endpoint evaluates DFS among randomized participants with PD-L1 strong positive tumors and TPS ≥50%. The registry reports both a two-sided 95% confidence interval for the hazard ratio and a one-sided p-value based on a permutation test with multivariate Cox regression.

TPS ≥50% DFS hazard ratio

HR 0.83

95% CI: 0.59–1.16   ·   P = 0.13499

The reported p-value is one-sided; the reported 95% confidence interval is two-sided.

CharacteristicReported value
EndpointDFS in Programmed Death Ligand-1 (PDL-1) Strong Positive Participants With Tumor Proportion Score (TPS) ≥50%
Time frameUp to approximately 84 months
Analysis populationAll randomized PD-L1 strong positive participants with TPS ≥50%, excluding one participant randomized in error who did not provide informed consent and for whom no data were collected
Groups comparedPembrolizumab vs Placebo
MethodCox proportional-hazards model
Effect measureHazard ratio
Estimate0.83
95% CI0.59–1.16
CI conventionTwo-sided
P-value0.13499
P-value methodOne-sided permutation test with multivariate Cox regression model
Clinical Biostats interpretation

A hazard ratio of 0.83 corresponds to an estimated instantaneous DFS-event rate that is 0.83 times the placebo-group rate under the fitted Cox model. In purely descriptive relative terms, 1 − 0.83 corresponds to an estimated 17% lower hazard.

The HR does not mean that 17% of participants benefited, that 17% of participants avoided recurrence, or that the probability of being disease-free was reduced or increased by exactly 17 percentage points.

The 95% confidence interval is 0.59–1.16. Its width indicates substantially more uncertainty around this subgroup-specific estimate than would be conveyed by the point estimate alone. Because the interval extends across 1, the interval includes hazard-ratio values representing no relative difference as well as values in either direction.

The reported one-sided P = 0.13499 is a tail-area probability calculated under the specified testing framework; it is not a measure of the size of the treatment effect. A p-value should not be interpreted as the probability that the treatment works or as the probability that the null hypothesis is true.

A particularly important reporting detail is that the p-value and confidence interval use different conventions: the p-value is one-sided, whereas the confidence interval is 95% two-sided. They should therefore be described separately rather than treated as if they were generated from an identical inferential convention.

The analysis is also restricted to the PD-L1 TPS ≥50% subgroup, so its precision and interpretation are distinct from the overall randomized analysis.

8. Secondary Endpoint Result: DFS in PD-L1 Strong Positive Participants With TPS ≥1%

The ClinicalTrials.gov record includes one secondary endpoint analysis for DFS among PD-L1 strong positive participants with TPS ≥1%. The analysis again uses a multivariate Cox proportional-hazards model with treatment adjusted by stage, PD-L1 status, adjuvant chemotherapy, region, histology, and smoking status.

TPS ≥1% DFS hazard ratio

HR 0.78

95% CI: 0.62–0.97   ·   One-sided P = 0.01327

CharacteristicReported value
EndpointDFS in PDL-1 Strong Positive Participants With TPS ≥1%
Time frameUp to approximately 84 months
Analysis populationAll randomized PD-L1 strong positive participants with TPS ≥1%
Groups comparedPembrolizumab vs Placebo
MethodCox proportional-hazards model
Effect measureHazard ratio
Estimate0.78
95% CI0.62–0.97
CI conventionTwo-sided
P-value0.01327
P-value methodOne-sided permutation test with multivariate Cox regression model
Clinical Biostats interpretation

The reported HR of 0.78 means that the estimated instantaneous DFS-event rate under the fitted model was 0.78 times that of the placebo group. As a simple derived relative statement, 1 − 0.78 corresponds to an estimated 22% lower hazard.

The HR is not an absolute risk reduction and does not tell us what fraction of participants experienced a recurrence or death. Those questions require absolute event probabilities or survival estimates at specified times, which are not included in the registry-reported statistical analysis.

The two-sided 95% confidence interval of 0.62–0.97 provides the reported uncertainty around the HR estimate. Its upper limit remains below 1, while the interval nevertheless spans a range of potentially meaningful treatment effects rather than identifying a single true effect size.

The one-sided P = 0.01327 is a measure of compatibility with the null hypothesis under the specified one-sided permutation-testing framework. It does not measure the magnitude of the HR and should not be read as a percentage probability that the treatment effect is real.

As with the other Cox analyses, the interpretation depends on the time-to-event model and its assumptions, including the proportional-hazards framework and appropriate handling of censoring.

9. Other Pre-Specified Analysis: DFS at 68 Months

The registry also reports a pre-specified analysis of DFS at 68 months. Although this is not identified as one of the two primary endpoint analyses in the ClinicalTrials.gov record, it provides an additional Cox-model estimate using the all-randomized analysis population.

DFS at 68 months

HR 0.76

95% CI: 0.63–0.91   ·   One-sided P = 0.00143

CharacteristicReported value
EndpointDFS at 68 Months
Time frameUp to approximately 68 months
Analysis populationAll randomized participants, excluding one participant randomized in error who did not provide informed consent and for whom no data were collected
Groups comparedPembrolizumab vs Placebo
MethodCox proportional-hazards model
Effect measureHazard ratio
Estimate0.76
95% CI0.63–0.91
CI conventionTwo-sided
P-value0.00143
P-value methodOne-sided permutation test with multivariate Cox regression model
Clinical Biostats interpretation

An HR of 0.76 corresponds to an estimated instantaneous DFS-event rate that is 0.76 times the placebo-group rate under the fitted model. In simple relative terms, 1 − 0.76 corresponds to an estimated 24% lower hazard.

This does not mean that the absolute probability of recurrence or death was 24% lower, nor does it mean that 24% of participants benefited. The hazard ratio is a relative model-based time-to-event measure.

The 95% CI of 0.63–0.91 quantifies uncertainty around the estimate. It does not describe individual-level variability and should not be interpreted as saying that individual patients have hazards confined to this interval.

The reported one-sided P = 0.00143 is a tail-area measure under the stated testing framework. It does not measure effect size. In particular, a small p-value does not make an HR numerically larger or smaller; effect magnitude is conveyed by the HR itself and its confidence interval.

This analysis is separately identified in the registry as an Other_Pre_Specified analysis. Its inferential interpretation should therefore be kept distinct from the two analyses identified as primary in the ClinicalTrials.gov record.

10. Results Summary

EndpointRoleHR95% CIP-valueTesting detail
Disease-Free Survival (DFS)Primary0.810.68–0.96Not reportedTwo-sided 95% CI
DFS, TPS ≥50%Primary0.830.59–1.160.13499One-sided permutation test; two-sided 95% CI
DFS, TPS ≥1%Secondary0.780.62–0.970.01327One-sided permutation test; two-sided 95% CI
DFS at 68 MonthsOther pre-specified0.760.63–0.910.00143One-sided permutation test; two-sided 95% CI

Across the four registry-reported analyses, the reported point estimates are all below 1. The amount of statistical uncertainty differs by endpoint, and the TPS ≥50% analysis has a confidence interval that extends above 1. These estimates should be interpreted according to their endpoint role, analysis population, and testing convention rather than treating all four analyses as interchangeable tests.

11. Safety Results

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

Safety measurePembrolizumabPlacebo
Serious adverse events142/58090/581
Serious adverse events by reported affected / at-risk counts
Pembrolizumab
142/580
Placebo
90/581

The ClinicalTrials.gov record does not provide a formal statistical analysis, confidence interval, or p-value for serious adverse events. The counts therefore should be presented descriptively rather than treated as a formal hypothesis test.

Safety denominator matters: the reported serious-adverse-event values are expressed as affected participants over participants at risk: 142/580 for pembrolizumab and 90/581 for placebo. The denominators are not the overall enrollment of 1177, so the safety figures should not be silently converted into percentages or compared as if they represented identical analysis populations without additional registry information.

12. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

DFS is defined as the time from randomization until recurrence or death. A Cox model is designed for this type of censored time-to-event outcome and estimates a relative hazard while allowing adjustment for covariates. The registry analysis adjusted treatment by stage, PD-L1 status, adjuvant chemotherapy, region, histology, and smoking status.

What does an HR of 0.81 mean?

An HR of 0.81 means that the fitted model estimates the instantaneous DFS-event rate in the pembrolizumab group to be 0.81 times that in the placebo group, conditional on the model. It can be expressed as a 19% lower estimated hazard, but it is not a 19-percentage-point improvement in the probability of being disease-free.

Why does the confidence interval matter?

A point estimate such as 0.81 is only one estimate from the data. The 95% CI of 0.68–0.96 shows the uncertainty associated with that estimate under the stated model and sampling framework. A confidence interval also makes the precision of the estimate visible rather than reducing the result to a single number.

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

A p-value describes how compatible the observed result is with a specified null hypothesis under the stated testing procedure. It does not quantify how large the treatment effect is. The HR describes relative effect magnitude, while the confidence interval describes uncertainty around that magnitude.

Why is the TPS ≥50% analysis especially important to interpret carefully?

The TPS ≥50% analysis is restricted to a subgroup rather than the complete randomized population. Its HR is 0.83, but the 95% CI is 0.59–1.16. The registry also reports a one-sided P = 0.13499. These features illustrate why a subgroup point estimate should not be interpreted without its confidence interval and the precise analysis population.

Why distinguish one-sided p-values from two-sided confidence intervals?

The ClinicalTrials.gov record explicitly state that the p-values for the TPS ≥50%, TPS ≥1%, and 68-month analyses are based on one-sided permutation tests, while their reported 95% confidence intervals are two-sided. These are different inferential conventions. A reader should not assume that the reported p-value and confidence interval represent identical tail allocations.

What does randomization contribute to the analysis?

Randomization assigns participants to the intervention groups by design rather than allowing treatment choice to determine assignment. This creates the foundation for a comparative treatment analysis. The primary overall DFS analysis is explicitly defined using all randomized participants, apart from the single participant randomized in error who did not provide informed consent and for whom no data were collected.

13. Confidence Intervals, Censoring, and the Hazard Ratio

Relative effect

The hazard ratio compares modeled instantaneous event rates between pembrolizumab and placebo. Values below 1 favor a lower estimated event hazard for pembrolizumab within the model.

Precision

The confidence interval communicates uncertainty around the HR estimate. Narrower intervals generally provide more precise estimates than wider intervals.

Absolute effect

A hazard ratio does not provide an absolute difference in the probability of being disease-free. Absolute survival estimates at specified times would answer a different question.

Censoring

Time-to-event methods can incorporate participants whose event status is not observed through the end of their available follow-up by treating their last known follow-up as a censoring point under the analysis framework.

A useful distinction
HR ≠ risk ratio ≠ risk difference

A hazard ratio is a time-to-event model parameter. It should not be substituted for an absolute risk difference or a ratio of cumulative event probabilities.

14. Covariate Adjustment in the Cox Model

The registry-reported analysis notes identify six variables used to adjust the treatment effect estimate: stage, PD-L1 status, adjuvant chemotherapy, region, histology, and smoking status.

Adjustment variableRole in the reported analysis
StageIncluded as a treatment-adjustment variable in the multivariate Cox model
PD-L1 statusIncluded as a treatment-adjustment variable in the multivariate Cox model
Adjuvant chemotherapyIncluded as a treatment-adjustment variable in the multivariate Cox model
RegionIncluded as a treatment-adjustment variable in the multivariate Cox model
HistologyIncluded as a treatment-adjustment variable in the multivariate Cox model
Smoking statusIncluded as a treatment-adjustment variable in the multivariate Cox model

Adjustment does not change the fundamental treatment comparison into a nonrandomized comparison. Rather, the reported Cox model estimates the treatment hazard ratio conditional on the listed covariates. This can improve statistical control for prognostic variables and account for their contribution to the modeled hazard.

Interpretation boundary: the ClinicalTrials.gov record identifies these variables as model-adjustment factors. They do not provide enough information to reconstruct the full model specification, coefficient estimates for each covariate, proportional-hazards diagnostics, or the complete statistical analysis plan.

15. One-Sided Testing and Permutation Testing

The statistical analyses posted on ClinicalTrials.gov for the TPS ≥50%, TPS ≥1%, and 68-month DFS endpoints report one-sided p-values based on a permutation test with multivariate Cox regression. This is a specific feature of the reported inferential framework.

One-sided test

A one-sided test concentrates the rejection region in one prespecified direction. It should be interpreted according to the hypothesis and testing convention specified for the analysis.

Permutation test

A permutation procedure evaluates the test statistic using a reference distribution generated by rearranging treatment assignments or an equivalent resampling mechanism under the null framework.

The p-value itself does not reveal the magnitude of the treatment effect. For that, the HR and its confidence interval are required. In this dataset, the explicit distinction between one-sided p-values and two-sided 95% confidence intervals is an important part of the statistical reporting.

16. What the Registry Does Not Report in the Supplied Data

The ClinicalTrials.gov record is sufficient to reconstruct the principal statistical comparison, but they do not provide several quantities that would normally appear in a comprehensive clinical-trial results report.

These omissions are not evidence that the corresponding procedures were absent from the full protocol or statistical analysis plan. They simply cannot be characterized from the ClinicalTrials.gov record.

17. Comparing the Primary Analyses

FeatureOverall DFSTPS ≥50% DFS
RolePrimaryPrimary
Analysis populationAll randomized participantsAll randomized PD-L1 strong positive participants with TPS ≥50%
HR0.810.83
95% CI0.68–0.960.59–1.16
P-valueNot reported0.13499
P-value conventionNot reportedOne-sided
CI conventionTwo-sidedTwo-sided
Cox adjustment variablesStage, PD-L1 status, adjuvant chemotherapy, region, histology, smoking statusStage, PD-L1 status, adjuvant chemotherapy, region, histology, smoking status

The point estimates of 0.81 and 0.83 are numerically similar, but the corresponding confidence intervals differ substantially in width. The TPS ≥50% analysis is based on a more restricted population, and its interval extends from 0.59 to 1.16. Statistical interpretation should therefore focus on the complete estimate-and-uncertainty pair rather than comparing point estimates alone.

18. Understanding the Analysis Population

The overall primary DFS analysis uses all randomized participants, with one explicit exception: one participant was randomized in error, did not provide informed consent, and had no data collected. This detail matters because the denominator for an efficacy analysis is determined by the prespecified analysis population, not simply by the headline enrollment number.

Overall DFS
All randomized participants, excluding the one participant randomized in error without informed consent and without collected data.
TPS ≥50%
All randomized PD-L1 strong positive participants with TPS ≥50%, with the same registry-noted exclusion of the participant randomized in error.
TPS ≥1%
All randomized PD-L1 strong positive participants with TPS ≥1%, with the same registry-noted exclusion.
68-month DFS
All randomized participants, with the same registry-noted exclusion.

This is an important distinction between enrollment, randomized population, and endpoint-specific analysis population. The overall trial enrollment is 1177.0, while the safety data reported by the registry use separate denominators of 580 and 581.

19. Important Limitations and Interpretation Issues

20. Why This Trial Matters Statistically

PEARLS is a useful statistical teaching case because it combines randomized allocation, masking, a time-to-event endpoint, multivariate Cox regression, covariate adjustment, subgroup-defined primary analysis, one-sided permutation testing, and confidence-interval interpretation.

ConceptHow it appears in PEARLS
RandomizationThe trial uses randomized allocation with two parallel arms.
BlindingThe registry classifies the trial as triple masked.
Time-to-event endpointDFS measures time from randomization to recurrence or death.
Cox modelPrimary and additional DFS analyses use Cox proportional-hazards models.
Hazard ratioTreatment effects are reported as HRs comparing pembrolizumab with placebo.
Confidence intervalAll registry-reported formal HR estimates include two-sided 95% confidence intervals.
Covariate adjustmentStage, PD-L1 status, adjuvant chemotherapy, region, histology, and smoking status are included in the multivariate model.
One-sided testingThe TPS ≥50%, TPS ≥1%, and 68-month analyses report one-sided permutation-test p-values.
Analysis populationsOverall DFS and biomarker-defined DFS analyses use different populations.
Safety denominatorsSerious adverse events are reported as affected participants over participants at risk for each arm.

21. Statistical Interpretation in One View

Overall DFS

The reported HR of 0.81 indicates a lower estimated DFS-event hazard for pembrolizumab than placebo under the multivariate Cox model. The 95% CI is 0.68–0.96. The registry-reported analysis does not report a p-value for this endpoint.

TPS ≥50% DFS

The reported HR is 0.83, with a two-sided 95% CI of 0.59–1.16 and a one-sided P = 0.13499. The wider interval demonstrates the importance of examining uncertainty rather than relying on the point estimate alone.

TPS ≥1% DFS

The reported HR is 0.78, with a two-sided 95% CI of 0.62–0.97 and a one-sided P = 0.01327. The HR corresponds descriptively to a 22% lower estimated hazard under the fitted model, but it is not an absolute risk reduction.

DFS at 68 months

The reported HR is 0.76, with a two-sided 95% CI of 0.63–0.91 and a one-sided P = 0.00143. This is an additional pre-specified analysis and should be distinguished from the two primary endpoint analyses.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue through Clinical Biostats

Use the related tutorials and statistical calculators to explore the survival-analysis concepts illustrated by this randomized trial.

25. Record Summary

PEARLS provides a compact example of how a randomized phase 3 trial can be analyzed when the principal outcome is disease-free survival. The ClinicalTrials.gov record uses multivariate Cox proportional-hazards models and report hazard ratios with two-sided 95% confidence intervals. The overall DFS analysis reports HR 0.81 (95% CI 0.68–0.96), while the PD-L1 TPS ≥50% primary analysis reports HR 0.83 (95% CI 0.59–1.16; one-sided P = 0.13499). A secondary TPS ≥1% analysis reports HR 0.78 (95% CI 0.62–0.97; one-sided P = 0.01327), and the other pre-specified 68-month DFS analysis reports HR 0.76 (95% CI 0.63–0.91; one-sided P = 0.00143).

The most important statistical lesson is that these numbers should not be read independently. A complete interpretation considers the analysis population, endpoint definition, Cox-model adjustment, hazard-ratio scale, confidence interval, one-sided versus two-sided inference, and endpoint role. The ClinicalTrials.gov record also show why safety results require attention to their denominators: serious adverse events are reported as 142/580 for pembrolizumab and 90/581 for placebo.

Clinical Biostats methodology: A trial-results page should distinguish the reported statistical evidence from the educational interpretation of that evidence. Where the ClinicalTrials.gov record does not provide a result, estimate, or methodological detail, this page does not substitute an inferred value.