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
RELATIVITY-047 was a randomized, parallel, quadruple-masked phase 2/3 trial evaluating relatlimab plus nivolumab versus nivolumab alone in participants with advanced melanoma. The registered primary endpoint was progression-free survival, a time-to-event endpoint analyzed using a stratified log-rank test with a hazard ratio as the effect measure.
| Feature | RELATIVITY-047 |
|---|---|
| Trial name | RELATIVITY-047 |
| NCT ID | NCT03470922 |
| Phase | Phase 2/3 |
| Condition | Melanoma |
| Enrollment | 714 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Primary endpoint type | Time-to-event |
| Primary hypothesis | Superiority |
| Primary statistical method | Log-rank test |
| Effect measure | Hazard ratio |
| Lead sponsor | Bristol-Myers Squibb |
| Sponsor type | Industry |
2. Clinical Question
The central question was whether relatlimab plus nivolumab improved progression-free survival compared with nivolumab alone in participants with advanced melanoma.
Population
Participants with advanced melanoma.
Intervention
Relatlimab plus nivolumab.
Comparator
Nivolumab alone.
Primary question
Does relatlimab plus nivolumab produce superior progression-free survival relative to nivolumab alone?
3. Trial Design
Relatlimab plus nivolumab
- Relatlimab
- Nivolumab
- Biological interventions
Nivolumab alone
- Nivolumab
- Biological intervention
The registered design is randomized, parallel, and quadruple masked. Randomization provides the principal framework for comparing the two treatment strategies, while masking is intended to reduce the influence of knowledge of treatment assignment on trial conduct and assessment.
4. Endpoints
| Endpoint | Registry definition and time frame | Analysis |
|---|---|---|
| Progression Free Survival (PFS) | From randomization to date of first documented tumor progression or death (up to approximately 33 months). Progression-free survival is defined as the time between the date of randomization and the date of first documented tumor progression, assessed by a blinded independent central review (BICR) (per RECIST v1.1 criteria), or death due to any cause, whichever occurs first. Subjects who die without a reported progression will be considered to have progressed on the date of their death. | Stratified log-rank test; hazard ratio |
The registry identifies one registered primary endpoint and classifies it as a time-to-event endpoint. Because either documented tumor progression or death can define the event, the analysis incorporates the timing of events rather than reducing each participant to a simple yes/no outcome.
5. Results
Primary Endpoint: Progression-Free Survival
The posted statistical analysis evaluated PFS from randomization to first documented tumor progression or death, with the analysis population defined as all randomized participants. The treatment groups were relatlimab plus nivolumab versus nivolumab.
Hazard ratio for progression or death
95% CI: 0.62–0.92 · P = 0.0055
Two-sided confidence interval · Superiority hypothesis
| Primary endpoint | Relatlimab + Nivolumab | Nivolumab | Effect estimate | P-value |
|---|---|---|---|---|
| Progression Free Survival (PFS) | Arm A | Arm B | HR 0.75 (95% CI 0.62–0.92) | 0.0055 |
The hazard ratio of 0.75 means that, within the Cox proportional-hazards framework used for the analysis, the estimated instantaneous rate of PFS events was 0.75 times that of the nivolumab group. Expressed as a simple relative interpretation, this corresponds to an estimated 25% lower hazard of progression or death for the relatlimab-plus-nivolumab group.
The hazard ratio does not mean that 25% of participants avoided progression, that every participant experienced a 25% reduction in risk, or that median PFS was reduced or increased by 25%. It is a relative time-to-event measure derived from the statistical model.
The 95% confidence interval of 0.62–0.92 describes uncertainty around the estimated hazard ratio. It is an interval for the treatment-effect estimate under the analysis framework, not a range describing individual patients' outcomes.
The P-value of 0.0055 addresses the statistical evidence against the null hypothesis used for the comparison. It does not measure the size, clinical importance, or probability of the treatment effect. The effect size is conveyed by the hazard ratio and its confidence interval.
Because the effect measure is a Cox proportional-hazards estimate, interpretation of a single hazard ratio also depends on the proportional-hazards framework. The ClinicalTrials.gov record does not report a separate assessment of that assumption.
The analysis was stratified by LAG-3 status, BRAF status, and AJCC M-stage. The resulting estimate therefore reflects the prespecified stratified time-to-event comparison rather than an unstratified comparison that ignores those factors.
Stratification used in the primary analysis
The registry states that the log-rank test was stratified by three factors:
| Stratification factor | Categories |
|---|---|
| LAG-3 | ≥ 1% vs < 1% |
| BRAF | Mutation positive vs mutation wild-type |
| AJCC M-stage | M0/M1any[0] vs M1any[1] |
Stratification is important because the log-rank comparison accounts for these prespecified factors rather than treating all participants as belonging to one undifferentiated risk set. The same factors therefore form part of the statistical structure underlying the reported primary comparison.
6. Safety Results
The registry provides serious adverse-event counts by treatment arm. These are reported as affected participants over the participants at risk for the corresponding arm.
| Safety measure | Relatlimab + Nivolumab | Nivolumab |
|---|---|---|
| Serious adverse events | 155/355 | 140/359 |
The serious-adverse-event information is descriptive. The ClinicalTrials.gov record does not provide a formal statistical comparison of serious adverse events, a confidence interval, or a P-value, so no inferential comparison is added here.
7. Statistical Methodology
Time-to-event analysis
PFS is a time-to-event endpoint because the outcome is defined not only by whether progression or death occurs, but also by when the first qualifying event occurs. Participants who have not yet experienced a qualifying event at their last evaluable follow-up contribute information up to that point through censoring.
The registry definition makes progression and death competing ways for a participant to reach the PFS event endpoint. A participant who dies without a reported progression is considered to have progressed on the date of death.
Kaplan-Meier estimation
Kaplan-Meier estimation is the standard descriptive framework for displaying the distribution of a time-to-event endpoint such as PFS. It accounts for right-censoring and produces an estimated event-free probability over time.
Here, di represents events at event time ti, while ni is the number at risk immediately before that time.
The ClinicalTrials.gov record does not provide a Kaplan-Meier table or numerical survival probabilities at specific time points. Accordingly, no reconstructed curve or unreported PFS percentages are presented.
Stratified log-rank test
The reported primary comparison used a log-rank test, specifically with stratification by LAG-3 status, BRAF status, and AJCC M-stage. The log-rank framework compares the observed and expected numbers of events across randomized treatment groups over follow-up.
Stratification allows the comparison to account for the prespecified factors while maintaining the randomized treatment comparison. It is particularly useful when the distribution of these factors may be associated with the underlying event process.
Cox proportional-hazards effect measure
The reported effect measure was a Cox proportional-hazard estimate, normalized in the registry data to a hazard ratio. A hazard ratio compares the modeled instantaneous event rates between treatment groups.
For this analysis, HR = 0.75 indicates a lower estimated hazard of the PFS event in the relatlimab-plus-nivolumab group relative to nivolumab alone.
Confidence interval
The reported confidence interval is 95% and two-sided. It provides an uncertainty range around the estimated hazard ratio. The interval of 0.62–0.92 lies below 1, which is consistent with the direction of the reported superiority comparison.
Analysis population
The primary analysis population is explicitly reported as all randomized participants. This is important because treatment assignment at randomization defines the comparison for the primary efficacy analysis rather than restricting the analysis to a subset selected after randomization.
8. Statistical Methods Explained
Why is PFS a time-to-event endpoint?
PFS incorporates both the occurrence of disease progression and death, together with the time from randomization to the first qualifying event. This preserves information about follow-up duration that would be lost in a simple binary endpoint.
Why was a log-rank test used?
The log-rank test is designed for comparing time-to-event distributions between treatment groups. It uses information accumulated throughout follow-up rather than comparing only one fixed time point. In RELATIVITY-047, the reported comparison was additionally stratified by LAG-3 status, BRAF status, and AJCC M-stage.
What does a hazard ratio of 0.75 mean?
A hazard ratio of 0.75 indicates that the estimated instantaneous rate of the PFS event was 0.75 times the corresponding rate in the comparator group under the fitted hazard-ratio model. A convenient relative interpretation is a 25% lower estimated hazard. It is not a 25% increase in median PFS and does not imply that exactly 25% of participants benefited.
Why is the confidence interval important?
The point estimate alone does not describe statistical precision. The 95% CI of 0.62–0.92 shows the range of hazard-ratio values supported by the analysis under its statistical assumptions and sampling framework. A narrower interval would indicate greater precision; the interval reported here provides the uncertainty information available in the registry analysis.
What does the P-value of 0.0055 tell us?
The P-value quantifies the evidence against the null hypothesis used for the statistical comparison, under the specified analysis framework. It does not tell us the probability that the treatment works, the magnitude of the treatment effect, or whether the effect is clinically important. Those questions require the hazard ratio, confidence interval, absolute outcomes, and clinical context.
Why does stratification matter?
The primary log-rank test was stratified by LAG-3 status, BRAF status, and AJCC M-stage. Stratification incorporates these prespecified factors into the time-to-event comparison, so the reported result is not simply an unadjusted comparison of all participants pooled together.
9. Understanding the Primary PFS Result
The primary hazard ratio of 0.75 is a relative measure. Within the reported Cox framework, the estimated hazard of progression or death was lower with relatlimab plus nivolumab than with nivolumab alone.
The 95% CI of 0.62–0.92 provides the uncertainty interval around the estimated hazard ratio. It is important to report the interval alongside the point estimate because the point estimate alone does not show how precisely the treatment effect was estimated.
The reported P = 0.0055 is evidence against the null comparison under the stated statistical test. The P-value should not be substituted for the hazard ratio or confidence interval when describing the magnitude or precision of the treatment effect.
The registered hypothesis type is superiority. The analysis therefore asks whether the treatment strategies differ in the direction specified by the superiority hypothesis rather than whether one treatment is merely not worse than the other within a prespecified non-inferiority margin.
10. What the PFS Definition Means Statistically
The registry defines PFS as the time from randomization until the first documented tumor progression assessed by blinded independent central review according to RECIST v1.1, or death from any cause, whichever occurs first.
Randomization is time zero
The PFS clock begins at randomization, which provides a common origin for the randomized treatment comparison.
Progression is one event route
First documented tumor progression is a qualifying event when it occurs before death.
Death is also an event
Death from any cause is itself a PFS event, even when no prior progression has been reported.
Independent review matters
The registry specifies blinded independent central review for the tumor-progression assessment.
This definition prevents a participant who dies without documented progression from being treated as permanently progression-free. The death date becomes the PFS event date in that situation.
11. Blinding and Endpoint Assessment
The trial is registered as quadruple masked. The PFS definition also specifies assessment of tumor progression by a blinded independent central review using RECIST v1.1 criteria.
| Design feature | Statistical relevance |
|---|---|
| Quadruple masking | Reduces the opportunity for knowledge of treatment assignment to influence trial conduct and assessment. |
| Blinded independent central review | Provides a blinded framework for determining documented tumor progression. |
| RECIST v1.1 | Provides the registered criteria for assessing tumor progression. |
| Death from any cause | Counts as a PFS event when it occurs before a reported progression. |
These features are particularly relevant for PFS because progression assessment involves clinical and radiologic determinations. The statistical analysis cannot be separated entirely from the way the endpoint is defined and assessed.
12. Randomization and the Primary Comparison
The allocation method is registered as randomized, with a parallel design and two treatment arms. The primary analysis compares all randomized participants in the two treatment groups.
Randomization establishes the treatment comparison before subsequent PFS outcomes occur. The primary analysis then compares time-to-event experience according to randomized treatment assignment.
This structure is why the analysis population matters. Conditioning the primary efficacy analysis on events or post-randomization treatment decisions could change the comparison being estimated. The registry instead identifies all randomized participants as the analysis population.
13. What the Registry Results Do Not Establish
The posted primary analysis is informative about the reported PFS comparison, but several quantities are not included in the ClinicalTrials.gov record and therefore are not reported here.
- No median PFS is reported in the trial data used for this page.
- No time-specific PFS probabilities are reported, so no additional survival percentages are presented.
- No subgroup hazard ratios are reported, so no subgroup treatment-effect estimates are inferred.
- No formal safety comparison is reported for the serious-adverse-event counts, so the safety section remains descriptive.
- No missing-data or imputation method is reported for the primary PFS analysis.
- No Bayesian method is reported in the registry-reported statistical analysis.
- No crossover analysis is reported in the ClinicalTrials.gov record.
- No non-inferiority margin is reported; the registered hypothesis type is superiority.
14. Limitations
- Single primary statistical analysis posted: the ClinicalTrials.gov record contains one formal statistical analysis for the primary endpoint.
- Limited reported efficacy detail: the posted analysis provides the hazard ratio, confidence interval, and P-value but does not provide median PFS or time-specific PFS estimates.
- Hazard-ratio interpretation: a Cox hazard ratio is a model-based relative measure and should not be interpreted as an absolute risk difference or an individual-level treatment effect.
- Proportional-hazards framework: interpretation of a single Cox hazard ratio depends on the model's assumptions. The ClinicalTrials.gov record does not report a separate proportional-hazards diagnostic.
- Stratified analysis: the reported PFS comparison is stratified by LAG-3 status, BRAF status, and AJCC M-stage. The resulting estimate should therefore be interpreted as the specified stratified comparison.
- Endpoint definition: PFS combines tumor progression and death, so the hazard ratio describes the composite event defined by the registry rather than tumor progression alone.
- Safety information is descriptive: serious adverse-event counts are reported by arm, but no inferential safety analysis is posted in the ClinicalTrials.gov record.
- External generalizability: the registry data identify the trial population as participants with advanced melanoma. The ClinicalTrials.gov record does not provide a broader characterization of eligibility or baseline clinical characteristics.
15. Why This Trial Matters Statistically
RELATIVITY-047 is a useful teaching case because its primary analysis brings together the core components of randomized time-to-event methodology: randomization, masking, a precisely defined composite endpoint, independent assessment of progression, stratified log-rank testing, Cox hazard-ratio estimation, confidence intervals, and a superiority hypothesis.
| Concept | How it appears in RELATIVITY-047 |
|---|---|
| Randomization | The trial uses randomized allocation between two parallel treatment arms. |
| Blinding | The registry classifies the study as quadruple masked. |
| Time-to-event endpoint | PFS measures time from randomization to progression or death. |
| Independent review | Tumor progression is assessed by blinded independent central review. |
| RECIST v1.1 | The registry specifies RECIST v1.1 criteria for documented tumor progression. |
| Kaplan-Meier framework | Time-to-event endpoints of this type are naturally summarized using Kaplan-Meier estimation; the ClinicalTrials.gov record does not provide the numerical KM estimates. |
| Log-rank test | The primary statistical method reported by the registry. |
| Stratified analysis | The log-rank test was stratified by LAG-3, BRAF, and AJCC M-stage. |
| Hazard ratio | The reported treatment-effect measure was a Cox proportional-hazard estimate. |
| Confidence interval | The primary estimate is accompanied by a two-sided 95% CI. |
| Superiority testing | The registered hypothesis type is superiority. |
| Safety analysis | Serious adverse events are reported as affected participants over those at risk by arm. |
16. Related Statistical Methods
Learn more about the methods used in this trial:
17. Related Statistical Calculators
18. Sources
- ClinicalTrials.gov: RELATIVITY-047, NCT03470922.
- PubMed: PMID 41811823.
- PubMed: PMID 40940136.
- PubMed: PMID 40513285.
- PubMed: PMID 39671533.
- PubMed: PMID 38320023.
Continue with the statistical methods behind clinical trials
Explore the survival-analysis, clinical-trial, and statistical-inference concepts that underpin randomized time-to-event analyses such as RELATIVITY-047.
19. Record Summary
RELATIVITY-047 provides a focused example of randomized time-to-event analysis. The trial enrolled 714 participants in a randomized, parallel, quadruple-masked phase 2/3 design comparing relatlimab plus nivolumab with nivolumab alone. Its registered primary endpoint was progression-free survival, defined from randomization to first documented tumor progression or death. The posted primary analysis used a stratified log-rank test and reported a Cox proportional-hazard estimate of 0.75 with a two-sided 95% CI of 0.62–0.92 and P = 0.0055.
The most useful statistical interpretation combines the hazard ratio with its confidence interval and P-value while keeping the endpoint definition and analysis population in view. The result describes a relative time-to-event comparison, not an individual-level probability of benefit. The stratification factors, blinded independent central review, and randomized analysis population are integral parts of how the reported PFS result should be understood.