← Clinical Trials
Melanoma Phase 3 Time-to-Event NCT02388906

CheckMate-238: Complete Statistical Analysis of Nivolumab in Melanoma

An independent statistical analysis of the randomized phase 3 CheckMate-238 trial evaluating nivolumab compared with ipilimumab for prevention of recurrence of melanoma after complete resection of stage IIIb/c or stage IV melanoma.

Trial status: Completed  ·  Enrollment: 906  ·  Primary endpoint: Recurrence-free survival
Scope of this record

This page separates reported trial results from statistical interpretation. The numerical results presented here are limited to the trial information reported in the ClinicalTrials.gov record. ClinicalTrials.gov provides the official trial registry record.

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

1. Trial at a Glance

CheckMate-238 was a randomized, parallel, quadruple-masked phase 3 trial comparing nivolumab 3 mg/kg with ipilimumab 10 mg/kg in participants with melanoma after complete resection of stage IIIb/c or stage IV melanoma. The registered primary endpoint was recurrence-free survival (RFS), a time-to-event endpoint followed for up to 36 months.

906
Enrollment
2 treatment arms
0.71
Primary RFS HR
95% CI 0.60–0.86
0.0003
Primary P-value
Two-sided
36 mo
Primary Time Frame
RFS
FeatureCheckMate-238
PhasePhase 3
ConditionMelanoma
Brief titleEfficacy Study of Nivolumab Compared to Ipilimumab in Prevention of Recurrence of Melanoma After Complete Resection of Stage IIIb/c or Stage IV Melanoma
AllocationRandomized
Design modelParallel
MaskingQuadruple
Primary purposeTreatment
Enrollment906.0
Arms2
Primary endpointRecurrence-free Survival (RFS)
Primary endpoint typeTime-to-event
Primary hypothesisSuperiority
Results postedYes
Statistical analyses posted3
Lead sponsorBristol-Myers Squibb
Sponsor typeIndustry
Trial statusCompleted
ClinicalTrials.govNCT02388906

2. Clinical Question

The central statistical question was whether nivolumab 3 mg/kg was superior to ipilimumab 10 mg/kg with respect to recurrence-free survival in participants with melanoma following complete resection of stage IIIb/c or stage IV melanoma.

Population

Participants with melanoma after complete resection of stage IIIb/c or stage IV melanoma.

Intervention

Nivolumab 3 mg/kg.

Comparator

Ipilimumab 10 mg/kg.

Primary question

Does nivolumab 3 mg/kg produce superior recurrence-free survival compared with ipilimumab 10 mg/kg?

3. Trial Design

01
Randomize 906 participants
02
Assign 2 parallel treatment groups
03
Mask Quadruple masking
04
Follow Recurrence-free survival
05
Analyze Stratified log-rank framework
Allocation
The registry identifies the trial as randomized.
Design model
The study used a parallel design with 2 arms.
Masking
The registry reports quadruple masking.
Primary purpose
The registered primary purpose was treatment.
ARM A · Nivolumab

Nivolumab 3 mg/kg

  • Nivolumab 3 mg/kg
  • Compared directly with ipilimumab 10 mg/kg
  • Primary analysis population: all randomized participants
  • Primary endpoint: recurrence-free survival
ARM B · Ipilimumab

Ipilimumab 10 mg/kg

  • Ipilimumab 10 mg/kg
  • Compared directly with nivolumab 3 mg/kg
  • Primary analysis population: all randomized participants
  • Primary endpoint: recurrence-free survival
Matching-placebo interventions: The registry lists placebo matching ipilimumab and placebo matching nivolumab among the interventions. This is consistent with the trial's reported quadruple-masked design, although the ClinicalTrials.gov record does not provide additional treatment-administration details.

4. Trial Timeline

March 16, 2015

Trial start

The registered trial start date was 2015-03-16.

November 26, 2018

Primary completion

The registered primary completion date was 2018-11-26.

Completed

Current registry status

The ClinicalTrials.gov record identifies the trial status as completed.

5. Endpoints

EndpointRegistry definitionTime frameType
Recurrence-free Survival (RFS) RFS is defined as the time between the date of randomization and the date of first recurrence (local, regional or distant metastasis), new primary melanoma, or death (whatever the cause), whichever occurs first. up to 36 months Time-to-event

The primary endpoint is therefore not simply a binary measure of whether recurrence occurred. It is a time-to-event endpoint: both the occurrence of an event and the time from randomization to that event contribute to the analysis.

RFS event definition
RFS event = first recurrence, new primary melanoma, or death, whichever occurs first

The registry definition combines several clinically relevant first-event outcomes into a single time-to-event endpoint.

6. Statistical Methodology

Primary analysis: stratified log-rank test

The registered primary analysis compared recurrence-free survival between nivolumab 3 mg/kg and ipilimumab 10 mg/kg using a log-rank test. The analysis notes identify the implementation as a stratified log-rank analysis.

The log-rank test is designed for comparing survival or event-time distributions between groups while accounting for the fact that participants can be followed for different lengths of time and may be censored before experiencing the event.

Effect measure: hazard ratio

The primary treatment effect was reported as a hazard ratio (HR). The primary estimate was 0.71, with a two-sided 95% confidence interval of 0.60 to 0.86.

Conceptual hazard-ratio interpretation
HR = hazard in nivolumab group ÷ hazard in ipilimumab group

An HR below 1 indicates a lower estimated instantaneous event rate in the numerator group, here nivolumab, relative to the comparator group under the time-to-event model.

Stratified analysis

The statistical analysis text specifically identifies the primary analysis as a stratified log-rank analysis. The ClinicalTrials.gov record does not specify the individual stratification factors, so none are added here.

Analysis population

The primary RFS analysis was conducted in all randomized participants. This is an important feature of the treatment comparison because the randomized treatment assignment defines the groups being compared in the efficacy analysis.

Superiority framework

The registered hypothesis type for the primary analysis was superiority. Thus, the statistical question was whether the observed time-to-event distributions provided evidence that nivolumab and ipilimumab differed in the direction specified by the superiority comparison.

What the registry does not specify: The ClinicalTrials.gov record does not provide a non-inferiority margin, crossover rule, factorial structure, Bayesian analysis, missing-data imputation method, or interim-analysis specification. Those topics are therefore not treated as features of the statistical design on this page.

7. Primary Results: Recurrence-Free Survival

The primary endpoint analysis compared recurrence-free survival for nivolumab 3 mg/kg versus ipilimumab 10 mg/kg over a time frame of up to 36 months. The analysis population was all randomized participants, and the method was a stratified log-rank analysis.

Primary RFS hazard ratio

0.71

95% CI: 0.60–0.86   ·   P = 0.0003

Two-sided 95% confidence interval · Superiority analysis

Primary endpointNivolumab 3 mg/kg vs Ipilimumab 10 mg/kg
OutcomeRecurrence-free Survival (RFS)
Time frameup to 36 months
Analysis populationAll randomized participants
MethodStratified log-rank test
Effect measureHazard ratio
Estimate0.71
95% CI0.60–0.86
P-value0.0003
Hypothesis typeSuperiority
Clinical Biostats interpretation

The estimated hazard ratio of 0.71 means that the estimated instantaneous rate of an RFS event in the nivolumab group was approximately 71% of that in the ipilimumab group, within the statistical framework used for the analysis. Equivalently, 1 − 0.71 = 0.29, so the estimate corresponds to an approximately 29% lower estimated hazard for the nivolumab group relative to the comparator.

The HR does not mean that 29% of participants avoided recurrence, that each individual participant experienced exactly a 29% reduction in risk, or that the absolute probability of recurrence was reduced by 29 percentage points. A hazard ratio is a relative time-to-event measure rather than an absolute risk difference.

The two-sided 95% confidence interval of 0.60 to 0.86 describes the uncertainty around the estimated hazard ratio under the statistical model and sampling framework. It is an interval for the treatment-effect parameter, not a range containing the outcomes of 95% of individual participants.

The P-value of 0.0003 addresses evidence against the null hypothesis under the specified testing framework. It does not measure the size or clinical importance of the treatment effect. Effect size and statistical evidence are different quantities, which is why the HR and its confidence interval should be read alongside the P-value.

Because this is a time-to-event analysis, interpretation also depends on censoring and the assumptions underlying the hazard-ratio framework. The ClinicalTrials.gov record identifies a stratified log-rank analysis but does not provide sufficient detail here to independently assess proportional-hazards behavior over time.

What the primary result establishes statistically

The registry reports a formal superiority analysis of RFS with an HR of 0.71, a two-sided 95% CI of 0.60 to 0.86, and a P-value of 0.0003. The confidence interval lies below 1, while the P-value provides the reported statistical evidence for the superiority comparison.

These statistics establish the reported relative treatment comparison for the registered primary endpoint. They do not by themselves provide an absolute RFS difference at a particular month, a median RFS, or the number of participants experiencing each component of the composite event definition. Those quantities are not included in the ClinicalTrials.gov record.

Educational note: a Kaplan-Meier curve cannot be reconstructed accurately from the reported HR and confidence interval alone. Valid reconstruction requires participant-level event/censoring information or sufficiently detailed source data.

8. Secondary Endpoint Result: Overall Survival

The registry also reports a secondary analysis of overall survival (OS) for up to 106.6 months. This analysis compared nivolumab 3 mg/kg with ipilimumab 10 mg/kg among all randomized participants using a log-rank framework with stratification. The analysis notes specify a stratified Cox proportional-hazard model for the reported hazard ratio.

Overall survival hazard ratio

0.88

95.03% CI: 0.69–1.11

Stratified Cox proportional-hazard model · Two-sided CI · Superiority hypothesis

Secondary endpointReported analysis
OutcomeOverall Survival (OS)
Time frameup to 106.6 months
Analysis populationAll randomized Participants
Groups comparedNivolumab 3 mg/kg vs Ipilimumab 10 mg/kg
MethodLog-rank test
Analysis detailStratified Cox proportional hazard model
Effect measureHazard ratio
Estimate0.88
95.03% CI0.69–1.11
Hypothesis typeSuperiority
Clinical Biostats interpretation

The reported OS hazard ratio of 0.88 corresponds to an estimated instantaneous death rate in the nivolumab group that was approximately 88% of that in the ipilimumab group under the reported model. Expressed as a relative hazard reduction, 1 − 0.88 = 0.12, corresponding to an approximately 12% lower estimated hazard of death.

This does not mean that overall mortality was reduced by 12 percentage points, that 12% of participants benefited, or that an individual participant's probability of death was reduced by exactly 12%. The HR is a model-based relative time-to-event measure.

The 95.03% confidence interval of 0.69 to 1.11 is wider than the primary RFS interval and crosses 1.00. It therefore includes values representing a lower estimated hazard, as well as values above 1 representing a higher estimated hazard for nivolumab relative to ipilimumab. The interval communicates substantial uncertainty about the precise magnitude of the OS hazard ratio.

No P-value is reported for this secondary analysis in the ClinicalTrials.gov record. It would therefore be inappropriate to manufacture one from the confidence interval or to infer one from the point estimate.

The analysis is identified as a stratified Cox proportional-hazard model. As with other Cox-model hazard ratios, interpretation depends on the model's assumptions and on the treatment-group censoring and follow-up patterns.

9. Post-Hoc RFS Analysis

A separate post-hoc analysis of recurrence-free survival is reported for a time frame of up to 114 months. It again compares nivolumab 3 mg/kg with ipilimumab 10 mg/kg among all randomized participants using a stratified log-rank analysis.

Post-hoc RFS hazard ratio

0.76

95% CI: 0.63–0.90

Stratified log-rank analysis · Two-sided CI · Superiority hypothesis

Post-hoc endpointReported analysis
OutcomeRecurrence-free Survival (RFS)
Time frameup to 114 months
Analysis populationAll randomized participants
Groups comparedNivolumab 3 mg/kg vs Ipilimumab 10 mg/kg
MethodStratified log-rank test
Effect measureHazard ratio
Estimate0.76
95% CI0.63–0.90
Hypothesis typeSuperiority
Analysis rolePost-hoc
Clinical Biostats interpretation

The post-hoc RFS HR of 0.76 corresponds to an estimated instantaneous RFS event rate approximately 76% as high in the nivolumab group as in the ipilimumab group, under the reported analysis framework. In relative terms, this corresponds to an approximately 24% lower estimated hazard.

The HR should not be interpreted as a 24-percentage-point reduction in recurrence probability or as an individual-level prediction. It summarizes a relative time-to-event comparison.

The two-sided 95% confidence interval of 0.63 to 0.90 provides the reported precision around the post-hoc estimate and remains below 1.00. Because the analysis is explicitly identified as post-hoc, its inferential role should be distinguished from the registered primary endpoint analysis.

No P-value is reported for this post-hoc analysis in the ClinicalTrials.gov record. The absence of a reported P-value means that no formal P-value should be added or reconstructed.

Primary versus post-hoc analysis: The primary RFS analysis and the later post-hoc RFS analysis should not be treated as interchangeable. They have different registered roles and different time frames. The primary analysis addresses the registered endpoint over up to 36 months; the post-hoc analysis extends the reported RFS analysis to up to 114 months.

10. Putting the Time-to-Event Results Together

AnalysisRoleTime frameHRConfidence intervalP-value
RFS Primary up to 36 months 0.71 95% CI 0.60–0.86 0.0003
OS Secondary up to 106.6 months 0.88 95.03% CI 0.69–1.11 Not reported
RFS Post-hoc up to 114 months 0.76 95% CI 0.63–0.90 Not reported

These three analyses illustrate why the role of an endpoint and the time frame of an analysis matter. The primary RFS analysis has its own registered endpoint definition and 36-month time frame. The secondary OS analysis addresses a different event over a longer reported time frame. The post-hoc RFS analysis returns to RFS but uses a later time frame.

The hazard ratios also should not be compared as though they were measurements of the same quantity. RFS and OS have different event definitions, and the analyses use different time frames and, for OS, a different model description. A smaller or larger HR across endpoints is not automatically evidence that one endpoint is more clinically important or that the underlying treatment effects are directly interchangeable.

11. Statistical Methods Explained

Why was a log-rank test used?

The primary endpoint, RFS, is a time-to-event outcome. Participants may experience recurrence, a new primary melanoma, or death at different times, while others may be censored before an event. The log-rank test is designed to compare event-time distributions between randomized groups while accounting for the timing of events and censoring.

What does an RFS hazard ratio of 0.71 mean?

An HR of 0.71 means that the estimated instantaneous rate of the defined RFS event was approximately 71% as high in the nivolumab group as in the ipilimumab group under the reported analysis. The corresponding relative hazard reduction is approximately 29%. It does not mean that 29% of participants avoided recurrence or that each participant had exactly a 29% lower probability of recurrence.

Why does the confidence interval matter?

The estimate 0.71 is only one point on a range of plausible values under the statistical model. The 95% CI of 0.60 to 0.86 describes uncertainty around that estimate. A narrower interval generally conveys more precision than a wider interval, while the location of the interval relative to 1.00 helps describe the direction and statistical uncertainty of a hazard-ratio comparison.

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

The primary P-value of 0.0003 describes the strength of evidence against the null hypothesis under the specified testing framework. It does not tell us whether the HR is 0.71, 0.50, or 0.90. The effect size is represented by the HR and its confidence interval. A P-value should therefore be read alongside, rather than instead of, the effect estimate.

Why is the OS result different from the primary RFS result?

RFS and OS are different endpoints. RFS counts the first recurrence, new primary melanoma, or death, whereas OS is based on death. They therefore measure different clinical events and can produce different hazard-ratio estimates. The ClinicalTrials.gov record reports an RFS HR of 0.71 and an OS HR of 0.88; these should not be treated as contradictory measurements of the same endpoint.

What does stratification add to the analysis?

A stratified log-rank analysis compares treatment groups while accounting for prespecified strata in the analysis framework. The ClinicalTrials.gov record confirms that the RFS analyses were stratified, but it does not identify the individual stratification factors. The statistical interpretation should therefore not assume specific factors that are not present in the registry data.

Why should the post-hoc analysis be separated from the primary result?

The post-hoc RFS analysis has a different analytical role from the registered primary endpoint analysis. Its HR of 0.76 is informative as a separately reported analysis over up to 114 months, but it should not silently replace the primary RFS analysis of 0.71 over up to 36 months. Different analysis roles can carry different inferential interpretations.

12. Understanding the Hazard Ratio More Deeply

Primary RFS result
HR = 0.71   →   estimated hazard ratio below 1

The estimated event hazard in the nivolumab group was lower than the estimated event hazard in the ipilimumab group within the reported primary analysis.

A hazard is an instantaneous event rate conditional on remaining event-free up to a given time. The hazard ratio compares those rates between treatment groups. It is therefore fundamentally different from comparing two fixed-time risks.

For example, an HR of 0.71 does not tell us the probability of being recurrence-free at 12 months, 24 months, or 36 months. Those quantities require survival estimates at specific time points. Likewise, the HR does not directly provide the median RFS. Because those statistics are not included in the ClinicalTrials.gov record, they are not presented here.

What HR = 0.71 tells us

It gives a relative comparison of the estimated event hazard between the randomized groups in the primary RFS analysis.

What HR = 0.71 does not tell us

It does not provide an absolute recurrence probability, median RFS, number needed to treat, or an individual participant's outcome.

13. Confidence Intervals and Statistical Precision

The primary RFS estimate is 0.71 with a two-sided 95% CI of 0.60 to 0.86. The interval is entirely below 1.00, which is consistent with the reported direction of the treatment effect and the reported superiority analysis.

The OS estimate is 0.88 with a 95.03% CI of 0.69 to 1.11. Unlike the primary RFS interval, this interval extends across 1.00. The ClinicalTrials.gov record does not report an OS P-value, so the interpretation should remain centered on the reported estimate and confidence interval rather than adding an unreported test statistic.

The post-hoc RFS estimate is 0.76 with a 95% CI of 0.63 to 0.90. It is reported separately because it is a post-hoc analysis with a time frame of up to 114 months.

EstimatePoint estimateConfidence intervalHow to read it
Primary RFS 0.71 0.60–0.86 The reported interval is below 1.00 and quantifies uncertainty around the primary HR.
Secondary OS 0.88 0.69–1.11 The reported interval includes 1.00, indicating uncertainty about the direction of the OS HR.
Post-hoc RFS 0.76 0.63–0.90 The reported interval is below 1.00 for this later RFS analysis.

14. P-values and Hypothesis Testing

The primary RFS analysis reports a P-value of 0.0003 under a superiority hypothesis. The value is associated specifically with the primary RFS analysis; it should not be transferred to the secondary OS analysis or the post-hoc RFS analysis.

P-value

The P-value quantifies statistical evidence against the null hypothesis under the specified analysis framework. It is not a measure of effect magnitude.

Effect estimate

The hazard ratio describes the estimated relative treatment effect. It should be considered together with its confidence interval.

Confidence interval

The interval communicates uncertainty around the estimated hazard ratio and provides information that a P-value alone does not.

Analysis role

Primary, secondary, and post-hoc analyses have different inferential roles and should be labeled accordingly.

Importantly, a P-value should not be interpreted as the probability that the treatment is effective or as the probability that the null hypothesis is true. Neither interpretation follows from the definition of a frequentist P-value.

15. Censoring and Kaplan-Meier Estimation

RFS is a time-to-event endpoint, so the statistical analysis must accommodate participants who have not yet experienced a qualifying event at the time their available follow-up ends. Such observations are typically handled as right-censored observations in survival analysis.

The Kaplan-Meier estimator is the standard nonparametric framework for estimating the event-free survival function from time-to-event data. Conceptually, at each observed event time, the estimator updates the probability of remaining event-free according to the number of events and the number of participants at risk.

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

Here, di represents events at event time ti, while ni represents the number at risk immediately before that time.

The statistical analyses posted on ClinicalTrials.gov identify log-rank methods and hazard ratios, but they do not provide the underlying event and censoring records or specific Kaplan-Meier survival probabilities. Consequently, no survival curve or fixed-time RFS estimate is reconstructed here.

16. Stratified Analysis

The primary RFS analysis is explicitly described as a stratified log-rank analysis. Stratification is useful when the trial's design includes factors that are important to account for in the treatment comparison.

In a stratified log-rank analysis, the comparison is constructed within strata and then combined across strata rather than treating the entire study population as a single unstratified risk set. This can preserve the structure specified by the trial's analysis plan.

Registry limitation: The ClinicalTrials.gov record says that the analysis was stratified but does not identify the individual stratification variables. It would therefore be inappropriate to name or interpret particular stratification factors without additional trial documentation.

17. Analysis Populations

The primary RFS analysis and the post-hoc RFS analysis both specify all randomized participants as the analysis population. The secondary OS analysis likewise specifies all randomized participants.

EndpointAnalysis populationWhy it matters
Primary RFS All randomized participants The primary efficacy comparison is anchored to randomized treatment assignment.
Secondary OS All randomized Participants The OS comparison is also based on the randomized groups.
Post-hoc RFS All randomized participants The later RFS analysis retains the randomized analysis population.

Using all randomized participants for efficacy analyses is important because randomization establishes the basis for the treatment comparison. It also means that subsequent treatment exposure, discontinuation, or other post-randomization events should not automatically lead to participants being reassigned to a different efficacy group.

18. Multiplicity and Multiple Analyses

The ClinicalTrials.gov record identifies one registered primary endpoint, one secondary statistical analysis, and one post-hoc RFS analysis. It does not provide a detailed multiplicity-adjustment strategy, alpha-allocation scheme, or hierarchical testing procedure.

AnalysisRole in the ClinicalTrials.gov recordReported P-valueInterpretive point
RFS, up to 36 months Primary 0.0003 Formal primary superiority analysis.
OS, up to 106.6 months Secondary Not reported Reported HR and CI should be presented without inventing an unreported P-value.
RFS, up to 114 months Post-hoc Not reported Later analysis should remain explicitly labeled as post-hoc.

Multiplicity is important whenever multiple endpoints, time points, subgroups, or analyses are considered. A collection of analyses cannot automatically be interpreted as though every result had the same confirmatory status as the registered primary endpoint. The ClinicalTrials.gov record does not provide enough information to describe a formal familywise-error strategy beyond identifying the endpoint roles.

19. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm. These figures describe the number affected and the number at risk for each arm.

Safety measureNivolumab 3 mg/kgIpilimumab 10 mg/kg
Serious adverse events 106/452 215/453
Serious adverse events: affected participants
Nivolumab 3 mg/kg
106
Ipilimumab 10 mg/kg
215

The reported denominators are 452 for nivolumab and 453 for ipilimumab. The ClinicalTrials.gov record does not identify the safety-analysis population definition beyond these affected/at-risk counts, nor does it provide additional adverse-event categories.

Do not substitute an efficacy statistic for a safety statistic. The RFS hazard ratio of 0.71 describes a time-to-event efficacy comparison. The serious-adverse-event figures describe affected participants relative to the reported at-risk denominators. They answer different statistical questions and should not be combined into a single numerical benefit-risk measure.

20. What the Primary RFS Result Does — and Does Not — Mean

What it means

The primary RFS HR of 0.71 indicates a lower estimated instantaneous rate of the registry-defined RFS event in the nivolumab group relative to the ipilimumab group under the reported stratified analysis.

What it does not mean

It does not mean that 29% of participants avoided recurrence, that every participant experienced a 29% reduction in probability of recurrence, or that the absolute RFS difference was 29 percentage points.

Why the confidence interval matters

The 95% CI of 0.60 to 0.86 quantifies uncertainty around the HR estimate. It does not describe the distribution of individual treatment effects or individual patient outcomes.

Why the P-value matters differently

The P-value of 0.0003 describes evidence against the null hypothesis within the primary superiority testing framework. It does not measure the magnitude of the HR or the practical importance of the treatment difference.

21. Primary vs Secondary vs Post-Hoc Evidence

Primary RFS

RFS through up to 36 months, analyzed in all randomized participants with a stratified log-rank test. HR 0.71, 95% CI 0.60–0.86, P = 0.0003.

Secondary OS

OS through up to 106.6 months, analyzed in all randomized participants. HR 0.88 with 95.03% CI 0.69–1.11.

Post-hoc RFS

RFS through up to 114 months, analyzed in all randomized participants. HR 0.76 with 95% CI 0.63–0.90.

Interpretive discipline

The three estimates should be kept tied to their endpoint, time frame, and analytical role rather than treated as interchangeable evidence.

22. Limitations

23. Why This Trial Matters Statistically

CheckMate-238 is a useful statistical teaching case because the ClinicalTrials.gov record brings together randomized allocation, quadruple masking, a time-to-event primary endpoint, stratified log-rank testing, hazard-ratio estimation, confidence intervals, a formal superiority hypothesis, a secondary survival analysis, and a later post-hoc RFS analysis.

ConceptHow it appears in CheckMate-238
RandomizationThe trial is registered as randomized with 906 enrolled participants.
Parallel designThe design model is parallel with 2 arms.
Quadruple maskingThe registry identifies the masking as quadruple.
Time-to-event endpointRFS is the registered primary endpoint and is defined by time from randomization to the first qualifying event.
Log-rank testThe primary RFS comparison uses a stratified log-rank analysis.
Hazard ratioThe primary RFS effect is HR 0.71; the secondary OS effect is HR 0.88; the post-hoc RFS effect is HR 0.76.
Confidence intervalEach reported hazard-ratio analysis includes a two-sided confidence interval.
P-valueThe primary RFS analysis reports P = 0.0003.
SuperiorityThe primary RFS analysis is identified as a superiority hypothesis.
Analysis populationAll randomized participants are specified for the reported efficacy analyses.
Secondary endpointOverall survival is reported as a secondary time-to-event analysis.
Post-hoc analysisA later RFS analysis is explicitly identified as post-hoc and uses a time frame of up to 114 months.
Safety analysisSerious adverse events are reported as affected/at-risk counts for each treatment arm.

24. Related Tutorials

Learn more about the methods used in this trial:

25. Related Statistical Calculators

26. Sources

Continue through the Clinical Biostats statistical pathway

Use the related tutorials and calculators to explore the survival-analysis concepts illustrated by CheckMate-238, including hazard ratios, confidence intervals, log-rank testing, Kaplan-Meier estimation, and time-to-event endpoints.

27. Record Summary

CheckMate-238 provides a clear example of randomized time-to-event analysis. The registered primary endpoint was recurrence-free survival, defined as the time from randomization to first recurrence, new primary melanoma, or death, whichever occurred first, over up to 36 months. The primary analysis used a stratified log-rank test in all randomized participants and reported an HR of 0.71 with a two-sided 95% CI of 0.60–0.86 and a P-value of 0.0003.

The same trial record also reports a secondary overall-survival analysis through up to 106.6 months, with an HR of 0.88 and a 95.03% CI of 0.69–1.11, using a stratified Cox proportional-hazard model. A separate post-hoc RFS analysis through up to 114 months reports an HR of 0.76 with a 95% CI of 0.63–0.90.

The statistical lesson is that these estimates must remain connected to their specific endpoint, time frame, analysis population, method, and analysis role. The primary RFS result, secondary OS result, and post-hoc RFS result are not interchangeable statistics. Reading the hazard ratio together with its confidence interval, reported P-value where available, and the underlying trial design provides a more complete statistical interpretation than any single number alone.

Clinical Biostats methodology: A trial-results page should not merely repeat a headline result. The goal is to present the reported statistical evidence in a standardized format while clearly separating the registry's documented results from educational interpretation and avoiding unsupported reconstruction of missing trial data.