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
LIBRETTO-431 was a randomized, open-label, parallel-group phase 3 trial in advanced or metastatic RET fusion-positive non-small cell lung cancer. The registry reports two primary progression-free survival endpoints assessed by blinded independent central review, one specifically evaluating the comparison with pembrolizumab-containing control therapy and one evaluating the control regimen with or without pembrolizumab.
| Feature | LIBRETTO-431 |
|---|---|
| Phase | Phase 3 |
| Condition | Non-Small Cell Lung Cancer |
| Study population | Participants with advanced or metastatic RET fusion-positive non-small cell lung cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 261 |
| Primary endpoints | 2 |
| Results posted | Yes |
| Outcome measures posted | 25 |
| Statistical analyses posted | 10 |
| Lead sponsor | Eli Lilly and Company |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT04194944 |
2. Clinical Question
The central statistical question is whether selpercatinib changes progression-free survival relative to pemetrexed and platinum chemotherapy, with the control regimen evaluated both with pembrolizumab and with or without pembrolizumab according to the two registered primary endpoint definitions.
Population
Participants with advanced or metastatic RET fusion-positive non-small cell lung cancer.
Intervention
Selpercatinib.
Comparator
Pemetrexed and platinum therapy, with one primary comparison specifying pembrolizumab and the other specifying pembrolizumab with or without pembrolizumab.
Primary question
Does selpercatinib improve progression-free survival compared with the registered control regimen under the two primary analysis populations?
3. Trial Design
Selpercatinib
- Selpercatinib was the study intervention.
- The registry's primary efficacy comparisons identify this as TRT A.
- Serious adverse events were reported for 55 of 158 participants at risk.
Pemetrexed and platinum
- The control regimen included pemetrexed and platinum therapy.
- The primary comparisons specify pembrolizumab either as included or as optional according to the endpoint population.
- Serious adverse events were reported for 23 of 98 participants at risk.
The trial is registered as randomized and parallel-group, with no masking. The registry identifies the interventions as selpercatinib, carboplatin, cisplatin, pemetrexed, and pembrolizumab.
4. Enrollment and Trial Timeline
Trial start
The registry lists 2020-02-17 as the study start date.
Primary completion
The registry lists 2023-05-01 as the primary completion date.
Active, not recruiting
The trial is currently recorded as ACTIVE_NOT_RECRUITING in the ClinicalTrials.gov record.
5. Primary Endpoints
| Endpoint | Registry definition | Time frame | Endpoint type |
|---|---|---|---|
| Progression Free Survival (PFS) by Blinded Independent Central Review (BICR) (With Pembrolizumab) | PFS is defined as the time from randomization until the occurrence of documented disease progression by the BICR, per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 criteria, or death from any cause in the absence of BICR-documented progressive disease. | Baseline to Progressive Disease or Death from Any Cause Up to 31 Months | Time-to-event |
| PFS by BICR (With or Without Pembrolizumab) | PFS is defined as the time from randomization until the occurrence of documented disease progression by the BICR, per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 criteria, or death from any cause in the absence of BICR-documented progressive disease. | Baseline to Progressive Disease or Death from Any Cause Up to 31 Months | Other / unclear in registry classification |
Both primary endpoints are fundamentally time-to-event outcomes. The event is either BICR-documented disease progression according to RECIST version 1.1 or death from any cause when BICR-documented progression has not occurred.
6. Primary Results
PFS by BICR — With Pembrolizumab
Hazard ratio for progression or death
95% CI: 0.309–0.699 · P = 0.0002
Selpercatinib (TRT A) vs pemetrexed and platinum with pembrolizumab (TRT B).
| Feature | Reported analysis |
|---|---|
| Endpoint | Progression Free Survival (PFS) by Blinded Independent Central Review (BICR) (With Pembrolizumab) |
| Analysis population | ITT Pembrolizumab |
| Comparison | Selpercatinib (TRT A) vs Pemetrexed and Platinum With Pembrolizumab (TRT B) |
| Method | Log Rank |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.465 |
| 95% CI | 0.309–0.699 |
| P-value | 0.0002 |
| Hypothesis | Superiority |
The reported hazard ratio of 0.465 is below 1. In the context of this randomized time-to-event comparison, it represents an estimated relative hazard of progression or death that is approximately 46.5% of the corresponding hazard in the control group. Equivalently, 1 − 0.465 = 0.535, so the point estimate corresponds to an approximately 53.5% lower estimated hazard for the selpercatinib group.
The HR does not mean that 53.5% of participants avoided progression, that 53.5% of participants benefited, or that each participant experienced the same reduction in risk. It is a relative time-to-event measure rather than an absolute probability.
The 95% confidence interval of 0.309–0.699 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of outcomes for individual patients.
The p-value of 0.0002 addresses evidence against the null hypothesis within the reported superiority analysis. A p-value is not a measure of the size of the treatment effect and should not be read as the probability that the treatment works.
Because the endpoint is time-to-event and the effect is summarized with a hazard ratio, interpretation also depends on the assumptions and censoring structure underlying the survival analysis. The registry reports a log-rank analysis and does not provide enough information in the ClinicalTrials.gov record to independently assess proportional hazards.
PFS by BICR — With or Without Pembrolizumab
Hazard ratio for progression or death
95% CI: 0.331–0.700 · P = 0.0001
Selpercatinib (TRT A) vs pemetrexed and platinum with or without pembrolizumab (TRT B).
| Feature | Reported analysis |
|---|---|
| Endpoint | PFS by BICR (With or Without Pembrolizumab) |
| Analysis population | ITT Population |
| Comparison | Selpercatinib (TRT A) vs Pemetrexed and Platinum With or Without Pembrolizumab (TRT B) |
| Method | Log Rank |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.482 |
| 95% CI | 0.331–0.700 |
| P-value | 0.0001 |
| Hypothesis | Superiority |
The hazard ratio of 0.482 means that the estimated hazard of progression or death was approximately 48.2% of the corresponding control hazard under the reported analysis. The point estimate therefore corresponds to an approximately 51.8% lower estimated hazard for the selpercatinib group.
This is not a statement that 51.8% of participants were protected from progression or death. A hazard ratio summarizes a relative difference in event rates over time; it does not directly provide an absolute survival probability or median time.
The 95% confidence interval of 0.331–0.700 quantifies uncertainty around the estimated hazard ratio. Its width matters: the data support a range of plausible relative effects rather than identifying the point estimate as exact.
The p-value of 0.0001 measures the strength of evidence against the relevant null hypothesis under the reported statistical test. It does not indicate that the effect is exactly 99.99% certain, nor does it measure clinical importance.
The analysis is based on an ITT population, which maintains the randomized treatment assignment as the basis of the efficacy comparison. Time-to-event interpretation still requires attention to censoring and the assumptions behind the hazard-ratio framework.
7. Secondary Endpoint Results
The registry reports eight secondary statistical analyses in addition to the two primary endpoint analyses. They cover disease control, objective response, duration of response, and intracranial response. The categorical endpoints were analyzed using the Cochran-Mantel-Haenszel method, while time-to-event duration of response endpoints were analyzed using the log-rank test.
Disease Control Rate
| Endpoint | Comparison | Method | OR | 95% CI | P-value |
|---|---|---|---|---|---|
| Percentage of Participant With Disease Control Rate (DCR) by BICR (With Pembrolizumab) | Selpercatinib vs Pemetrexed and Platinum With Pembrolizumab | Cochran-Mantel-Haenszel | 1.5 | 0.7–3.4 | 0.3996 |
| Percentage of Participant With DCR by BICR (With or Without Pembrolizumab) | Selpercatinib vs Pemetrexed and Platinum With or Without Pembrolizumab | Cochran-Mantel-Haenszel | 1.7 | 0.9–3.6 | 0.1390 |
Both DCR analyses use an odds ratio as the effect measure. An odds ratio above 1 indicates higher estimated odds of disease control in the selpercatinib group relative to the corresponding control group, but the confidence intervals and p-values quantify the uncertainty and evidence associated with those estimates.
Overall Response Rate
| Endpoint | Comparison | Method | OR | 95% CI | P-value |
|---|---|---|---|---|---|
| Overall Response Rate (ORR): Percentage of Participants With Complete Response (CR) or Partial Response (PR) by BICR (With Pembrolizumab) | Selpercatinib vs Pemetrexed and Platinum With Pembrolizumab | Cochran-Mantel-Haenszel | 2.7 | 1.4–5.1 | 0.0028 |
| ORR: Percentage of Participants With CR or PR by BICR (With or Without Pembrolizumab) | Selpercatinib vs Pemetrexed and Platinum With or Without Pembrolizumab | Cochran-Mantel-Haenszel | 2.9 | 1.6–5.2 | 0.0003 |
The ORR endpoint is binary: participants are classified according to whether they achieved a complete response or partial response. The odds ratio therefore compares the odds of meeting that response definition between treatment groups; it is not the same as a risk ratio or an absolute percentage-point difference.
Duration of Response
| Endpoint | Comparison | Method | HR | 95% CI | P-value |
|---|---|---|---|---|---|
| Duration of Response (DoR) by BICR (With Pembrolizumab) | Selpercatinib vs Pemetrexed and Platinum With Pembrolizumab | Log Rank | 0.377 | 0.224–0.633 | 0.0001 |
| DOR by BICR (With or Without Pembrolizumab) | Selpercatinib vs Pemetrexed and Platinum With or Without Pembrolizumab | Log Rank | 0.418 | 0.256–0.684 | 0.0004 |
Duration of response is a time-to-event endpoint defined in the registry from the date of complete response or partial response to the date of disease progression or death due to any cause. This makes it conceptually different from ORR: ORR asks whether a response occurred, while DOR concerns how long a response persists among the response-defined analysis population.
Intracranial Objective Response Rate
| Endpoint | Analysis population | Method | OR | 95% CI | P-value |
|---|---|---|---|---|---|
| Intracranial ORR by BICR (With Pembrolizumab) | CNS Pembrolizumab | Cochran-Mantel-Haenszel | 3.2 | 0.8–12.8 | 0.1809 |
| Intracranial ORR by BICR (With or Without Pembrolizumab) | CNS Overall | Cochran-Mantel-Haenszel | 5.2 | 1.4–19.6 | 0.0167 |
The intracranial endpoints focus on participants with the relevant baseline central nervous system assessment or CNS metastasis population specified by the registry. The confidence intervals are relatively wide, particularly for the first analysis, illustrating how estimates from a more restricted analysis population can be substantially less precise than the primary population-level comparison.
8. Results Summary
| Endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| Primary PFS by BICR, with pembrolizumab | Hazard ratio | 0.465 | 0.309–0.699 | 0.0002 |
| Primary PFS by BICR, with or without pembrolizumab | Hazard ratio | 0.482 | 0.331–0.700 | 0.0001 |
| DCR by BICR, with pembrolizumab | Odds ratio | 1.5 | 0.7–3.4 | 0.3996 |
| DCR by BICR, with or without pembrolizumab | Odds ratio | 1.7 | 0.9–3.6 | 0.1390 |
| ORR by BICR, with pembrolizumab | Odds ratio | 2.7 | 1.4–5.1 | 0.0028 |
| ORR by BICR, with or without pembrolizumab | Odds ratio | 2.9 | 1.6–5.2 | 0.0003 |
| DOR by BICR, with pembrolizumab | Hazard ratio | 0.377 | 0.224–0.633 | 0.0001 |
| DOR by BICR, with or without pembrolizumab | Hazard ratio | 0.418 | 0.256–0.684 | 0.0004 |
| Intracranial ORR by BICR, with pembrolizumab | Odds ratio | 3.2 | 0.8–12.8 | 0.1809 |
| Intracranial ORR by BICR, with or without pembrolizumab | Odds ratio | 5.2 | 1.4–19.6 | 0.0167 |
This pattern illustrates why a clinical-trial statistical assessment should not collapse all endpoints into one number. The primary endpoints are time-to-event comparisons, whereas several secondary endpoints are binary response outcomes. Each effect measure answers a different question and has a different interpretation.
9. Statistical Methodology
Log-rank test
The registry reports the log-rank test for both primary PFS analyses and both duration-of-response analyses. The log-rank framework compares the observed and expected pattern of events over follow-up between randomized groups.
The log-rank test evaluates whether the survival experience differs between groups across event times. The test itself is not the hazard ratio; the registry separately reports the hazard ratio as the effect measure.
Hazard ratio
The hazard ratio is the effect measure reported for PFS and DOR. Values below 1 correspond to a lower estimated instantaneous event rate in TRT A relative to TRT B under the reported analysis.
An HR of 0.465, for example, represents an estimated hazard in TRT A that is 46.5% of the corresponding hazard in TRT B. It does not directly provide an absolute probability of remaining progression-free.
Cochran-Mantel-Haenszel test
The registry reports the Cochran-Mantel-Haenszel method for the binary DCR, ORR, and intracranial ORR analyses. This method can combine evidence across strata while accounting for stratification in a categorical treatment comparison. The ClinicalTrials.gov record identifies the analysis as stratified but does not specify the individual stratification factors.
Odds ratio
The odds ratio is used for the categorical response endpoints. An OR above 1 indicates higher odds of the defined binary outcome in TRT A relative to TRT B, while an OR below 1 indicates lower odds.
Because odds and probability are different quantities, an odds ratio should not automatically be described as a percentage increase in the response probability.
Intention-to-treat analysis
The primary analyses use ITT populations. For the overall ITT analysis, the registry defines the population as all randomized participants, even if a participant did not take the assigned treatment, did not receive the correct treatment, or otherwise did not follow the protocol. The pembrolizumab-specific ITT population is defined as participants included in the ITT population who were stratified with the intent to receive pembrolizumab in the event of control-arm assignment.
Blinded independent central review
Although the overall study is unmasked, the primary PFS endpoint is assessed by BICR. Central review provides an independent assessment of radiographic progression using RECIST version 1.1, reducing the possibility that knowledge of randomized treatment assignment directly determines the recorded central-review progression event.
10. Analysis Populations and Their Statistical Importance
| Population | Registry role | Why it matters statistically |
|---|---|---|
| ITT Population | All randomized participants for the overall primary analysis | Preserves randomized treatment assignment as the basis of the efficacy comparison. |
| ITT Pembrolizumab | Participants in the ITT population stratified with intent to receive pembrolizumab if assigned to the control arm | Defines the population for the pembrolizumab-specific primary comparison and related secondary analyses. |
| CNS Pembrolizumab | Participants in the ITT population with baseline CNS assessment in the pembrolizumab-specific analysis population | Defines a more restricted population for intracranial response assessment. |
| CNS Overall | Participants in the ITT population with baseline CNS assessment and CNS metastasis at baseline | Defines the population for the overall intracranial ORR analysis. |
The distinction between these populations is statistically important. A treatment effect estimated in all randomized participants is not automatically interchangeable with an estimate from a CNS-specific or pembrolizumab-intent population. Changing the population changes the estimand being described.
11. Statistical Methods Explained
Why was a log-rank test used for PFS?
PFS records the time until progression or death, so participants may be censored if the event has not occurred by the relevant observation time. The log-rank test is designed for comparing time-to-event experience between groups while using information across follow-up rather than reducing each participant to a simple yes/no event indicator.
What does an HR of 0.465 mean?
It means the estimated hazard in the selpercatinib group was 0.465 times the corresponding hazard in the comparator group under the reported analysis. The reciprocal interpretation should not be confused with an absolute risk reduction: the HR is a relative time-to-event measure.
Why is the confidence interval important?
The estimate 0.465 is a point estimate. The corresponding 95% confidence interval, 0.309–0.699, communicates uncertainty around that estimate. A confidence interval that is wider indicates less precision than a narrower interval, even when the point estimate is similar.
Why doesn't the p-value measure treatment effect size?
The p-value measures the compatibility of the observed data with a specified null hypothesis under the statistical model and test. It depends on both the observed effect and the amount of information available. The effect size itself is communicated by the hazard ratio or odds ratio, while the confidence interval provides an accompanying measure of precision.
Why are ORR and DCR analyzed differently from PFS?
ORR and DCR are binary outcomes: participants either meet the registered response definition or they do not. PFS and DOR contain timing information and can involve censoring. Consequently, the registry uses the Cochran-Mantel-Haenszel method for the binary endpoints and the log-rank method for the time-to-event endpoints.
Why does the BICR designation matter in an open-label trial?
The study itself is not masked, but the primary progression assessment is performed by blinded independent central review. This separates knowledge of treatment assignment at the trial level from the central radiologic assessment used to determine the primary endpoint event.
Why should the CNS results be interpreted separately?
The CNS analyses use specifically defined populations rather than the entire ITT population. The smaller analysis population can produce less precise estimates, as illustrated by the 95% CI of 0.8–12.8 for the pembrolizumab-specific intracranial ORR analysis.
12. Primary Endpoint Interpretation
Relative effect
The two primary HR estimates are 0.465 and 0.482. Both are below 1, indicating lower estimated hazards for progression or death in the selpercatinib group under their respective analyses.
Precision
The 95% CIs are 0.309–0.699 and 0.331–0.700. These intervals describe uncertainty around the corresponding point estimates rather than individual patient outcomes.
Statistical evidence
The reported p-values are 0.0002 and 0.0001. These quantify evidence against the relevant null hypotheses and are not measures of effect magnitude.
Endpoint definition
PFS combines BICR-documented RECIST version 1.1 progression with death from any cause when BICR-documented progression has not occurred.
The similarity of the two primary point estimates is informative descriptively, but the two analyses should remain tied to their registered populations and comparator definitions. The first analysis is specifically for participants stratified with the intent to receive pembrolizumab in the event of control assignment; the second uses the overall ITT population.
13. Secondary Endpoint Interpretation
Response endpoints
The ORR analyses report odds ratios of 2.7 and 2.9, with corresponding 95% confidence intervals of 1.4–5.1 and 1.6–5.2. The associated p-values are 0.0028 and 0.0003. These analyses provide a categorical response perspective that complements the primary time-to-event analysis.
The DCR analyses report odds ratios of 1.5 and 1.7, with 95% confidence intervals of 0.7–3.4 and 0.9–3.6. Their p-values are 0.3996 and 0.1390. The distinction between DCR and ORR is important: disease control can include a broader set of response classifications than the CR-or-PR definition used for ORR.
Duration of response
The DOR hazard ratios are 0.377 and 0.418. Both are below 1, with 95% confidence intervals of 0.224–0.633 and 0.256–0.684. These are time-to-event analyses beginning at the date of CR or PR and ending at disease progression or death due to any cause.
Intracranial response
The intracranial ORR analyses illustrate the importance of uncertainty. The pembrolizumab-specific analysis reports an OR of 3.2 with a 95% CI of 0.8–12.8 and a p-value of 0.1809. The overall CNS analysis reports an OR of 5.2 with a 95% CI of 1.4–19.6 and a p-value of 0.0167. The wide intervals show that these estimates are considerably less precise than the primary PFS estimates.
14. Multiplicity and Multiple Endpoints
The registry identifies 2 primary endpoints and reports 10 statistical analyses in total: two primary analyses and eight secondary analyses. The analyses include multiple related measures of disease control, response, duration of response, and intracranial response.
| Endpoint family | Number of reported analyses | Statistical framework |
|---|---|---|
| Primary PFS | 2 | Log-rank test; hazard ratio |
| DCR | 2 | Cochran-Mantel-Haenszel; odds ratio |
| ORR | 2 | Cochran-Mantel-Haenszel; odds ratio |
| DOR | 2 | Log-rank test; hazard ratio |
| Intracranial ORR | 2 | Cochran-Mantel-Haenszel; odds ratio |
Multiple endpoints create a statistical interpretation issue because each additional hypothesis test creates another opportunity for a nominally unusual result. The ClinicalTrials.gov record identifies the hypothesis type as superiority, but they do not provide an alpha-spending plan, multiplicity-adjustment hierarchy, or detailed familywise error-control procedure.
15. Stratified Analysis
Several registry analyses explicitly identify stratified analysis as an analysis concept. The primary PFS analysis includes stratified analysis in its analysis text, and the pembrolizumab-specific DCR, ORR, DOR, and intracranial ORR analyses also identify stratified analysis.
Stratification can improve alignment between the statistical analysis and the randomized trial design when important baseline factors were used to structure the randomization. The ClinicalTrials.gov record does not specify the individual stratification variables, so they are not listed here.
The Cochran-Mantel-Haenszel framework is particularly suited to categorical comparisons when stratification is part of the analysis. For time-to-event endpoints, the registry reports the log-rank method and identifies stratified analysis as an associated concept.
16. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm. These figures should be kept separate from the efficacy results because safety is an exposure-related outcome rather than a time-to-progression efficacy measure.
| Treatment arm | Serious adverse events affected | Participants at risk | Reported format |
|---|---|---|---|
| Selpercatinib (TRT A) | 55 | 158 | 55/158 |
| Carboplatin or Cisplatin + Pemetrexed +/- Pembrolizumab (TRT B) | 23 | 98 | 23/98 |
The registry supplies the serious-adverse-event counts in affected/at-risk form. The bars above are therefore a visual representation of those reported counts rather than an independently recomputed rate.
17. Why the Analysis Population Matters
LIBRETTO-431 illustrates a central principle of clinical-trial statistics: an effect estimate is inseparable from the population in which it was estimated.
Overall ITT
The second primary PFS analysis uses all randomized participants according to the registry definition, even when treatment was not taken exactly as assigned.
ITT Pembrolizumab
The first primary PFS analysis uses an ITT population defined according to intent to receive pembrolizumab in the event of control-arm assignment.
CNS populations
Intracranial ORR analyses use specifically defined baseline CNS assessment populations rather than the full ITT population.
Response analyses
Response endpoints are binary and use odds ratios, whereas PFS and DOR are time-to-event outcomes summarized with hazard ratios.
This structure prevents an important analytical error: treating all reported estimates as though they were generated from exactly the same participants under exactly the same estimand. The population definition is part of the meaning of the result.
18. Understanding Hazard Ratios and Odds Ratios Together
| Measure | Used in LIBRETTO-431 for | Core question | What it does not directly provide |
|---|---|---|---|
| Hazard ratio | PFS and DOR | How do event rates over time compare between randomized groups? | An absolute risk difference or median survival time. |
| Odds ratio | DCR, ORR, intracranial ORR | How do the odds of meeting a binary response definition compare? | A direct percentage-point difference in response probability. |
The two effect measures should not be placed on the same numerical scale. For example, an OR of 2.7 and an HR of 0.465 describe different statistical quantities. Neither should be interpreted as a generic percentage improvement without returning to the endpoint definition and the underlying model.
19. Clinical Biostats Interpretation of the Overall Statistical Pattern
Both registered primary PFS analyses report hazard ratios below 1: 0.465 for the pembrolizumab-specific comparison and 0.482 for the comparison with or without pembrolizumab. Their 95% confidence intervals are 0.309–0.699 and 0.331–0.700, respectively.
The ORR analyses report odds ratios of 2.7 and 2.9, while the DCR analyses report odds ratios of 1.5 and 1.7. These measures provide a categorical response perspective rather than a time-to-event perspective.
The DOR hazard ratios of 0.377 and 0.418 are below 1, providing a separate time-to-event assessment of duration among the registered response populations.
The intracranial ORR estimates are 3.2 and 5.2, but their confidence intervals are wide: 0.8–12.8 and 1.4–19.6. The width of these intervals is important when interpreting precision.
The statistical story is therefore multidimensional: the primary endpoint is time-to-event, several secondary endpoints are binary, and CNS analyses use restricted populations. The appropriate interpretation is to preserve those distinctions rather than reduce the trial to a single p-value.
20. Important Limitations and Interpretation Issues
- Open-label design: the registry identifies the study as unmasked. This is distinct from the blinded independent central review used for the primary PFS assessment.
- Hazard-ratio interpretation: a hazard ratio is a relative time-to-event measure and should not be treated as an absolute risk reduction or the percentage of participants who benefit.
- Proportional-hazards considerations: the ClinicalTrials.gov record reports hazard ratios but do not provide sufficient information to independently verify the proportional-hazards assumption.
- Analysis-population differences: the two primary analyses use different population definitions, so their estimates should not be treated as if they came from an identical analysis population.
- CNS precision: the intracranial ORR confidence intervals are wide, particularly the 0.8–12.8 interval in the pembrolizumab-specific analysis.
- Multiplicity: the registry reports two primary endpoints and multiple secondary analyses, but the ClinicalTrials.gov record does not specify the complete multiplicity-adjustment strategy.
- Missing numerical details: the ClinicalTrials.gov record does not provide median PFS, Kaplan-Meier time-point estimates, event counts for the primary analyses, baseline characteristics, subgroup forest plots, or participant-level censoring information.
- Safety scope: the ClinicalTrials.gov record is limited to serious adverse events by arm and does not support a complete adverse-event profile.
- Secondary endpoint interpretation: DCR, ORR, DOR, and intracranial ORR answer different questions and use different analysis populations and effect measures.
21. What the Primary Hazard Ratios Do — and Do Not — Mean
The reported HR of 0.465 means the estimated hazard of progression or death in the selpercatinib group was 46.5% of the corresponding hazard in the pemetrexed-and-platinum-with-pembrolizumab group under the reported analysis.
It does not mean that 46.5% of patients progressed, that 53.5% of patients were cured, or that every individual experienced exactly a 53.5% reduction in risk.
The reported HR of 0.482 means the estimated hazard of progression or death in the selpercatinib group was 48.2% of the corresponding hazard in the pemetrexed-and-platinum-with-or-without-pembrolizumab group.
The corresponding point estimate represents an approximately 51.8% lower estimated hazard, but that statement remains a relative model-based interpretation rather than an absolute probability.
The 95% CIs of 0.309–0.699 and 0.331–0.700 describe uncertainty around the two estimates. They do not describe the range of individual patient responses or guarantee that the true effect lies inside the interval.
The p-values of 0.0002 and 0.0001 quantify evidence against the relevant null hypotheses under the reported tests. They are not effect-size measures and do not indicate the probability that a clinical hypothesis is true.
22. Why This Trial Matters Statistically
LIBRETTO-431 is a useful teaching case because it combines randomized treatment comparison, time-to-event endpoints, blinded central review within an open-label study, stratified analyses, categorical response measures, restricted CNS populations, and multiple endpoint families.
| Concept | How it appears in LIBRETTO-431 |
|---|---|
| Randomization | The trial is registered as randomized with 261 participants and 2 parallel arms. |
| Intention-to-treat analysis | Both primary analyses use ITT-defined populations. |
| Time-to-event endpoints | Both primary endpoints are PFS defined from randomization to progression or death. |
| Blinded independent review | Primary PFS progression is assessed by BICR using RECIST version 1.1. |
| Log-rank test | Used for both primary PFS analyses and both DOR analyses. |
| Hazard ratio | Used to quantify the relative PFS and DOR treatment effects. |
| Cochran-Mantel-Haenszel test | Used for DCR, ORR, and intracranial ORR analyses. |
| Odds ratio | Used for categorical disease-control and response endpoints. |
| Stratified analysis | Identified in the analysis text for several registry analyses. |
| CNS-specific analysis | Intracranial ORR is evaluated in specifically defined CNS populations. |
| Multiplicity | Two primary endpoints and eight secondary statistical analyses are reported. |
23. Statistical Methods in Context
Question 1: Does treatment change time to progression or death?
The primary answer uses PFS, a time-to-event endpoint, analyzed with the log-rank test and expressed with a hazard ratio.
Question 2: Does treatment change response probability?
ORR and DCR are binary endpoints analyzed with the Cochran-Mantel-Haenszel method and expressed using odds ratios.
Question 3: Are responses durable?
DOR is a time-to-event endpoint beginning at CR or PR and ending at progression or death due to any cause.
Question 4: Is CNS response different?
Intracranial ORR is evaluated separately in registry-defined CNS populations, with odds ratios and confidence intervals.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Statistical Calculators
26. Sources
- ClinicalTrials.gov: NCT04194944 — LIBRETTO-431.
- PubMed: PMID 41996652.
- PubMed: PMID 38828957.
- PubMed: PMID 37870973.
- PubMed: PMID 35378489.
- PubMed: PMID 33150799.
Continue through the Clinical Biostats statistical pathway
Use the related tutorials and calculators to explore the survival-analysis, categorical-data, and clinical-trial methods illustrated by LIBRETTO-431.
27. Record Summary
LIBRETTO-431 provides a compact example of several important clinical-trial statistical methods. The randomized phase 3 study enrolled 261 participants in two parallel arms and evaluated two registered primary PFS endpoints using BICR assessment, the log-rank test, and hazard ratios. The reported primary HRs were 0.465 and 0.482, with corresponding 95% confidence intervals of 0.309–0.699 and 0.331–0.700. Secondary analyses used both time-to-event and categorical methods, including hazard ratios for DOR and odds ratios for DCR, ORR, and intracranial ORR.
The most useful statistical interpretation keeps the estimands distinct. PFS asks about time to progression or death after randomization; DOR asks about the time from response to progression or death; ORR and DCR ask whether a defined categorical response or disease-control state was achieved; and intracranial ORR focuses on a specifically defined CNS population. Confidence intervals describe precision, p-values describe evidence against specified null hypotheses, and the analysis population determines exactly which participants the resulting estimate represents.