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-531 is a randomized, open-label, parallel-group phase 3 trial in medullary thyroid cancer. The registry describes 291 participants randomized to selpercatinib or to treatment with cabozantinib or vandetanib, with progression-free survival by blinded independent central review as the registered primary endpoint.
| Feature | LIBRETTO-531 |
|---|---|
| Phase | Phase 3 |
| Condition | Medullary Thyroid Cancer |
| Brief title | A Study of Selpercatinib (LY3527723) in Participants With RET-Mutant Medullary Thyroid Cancer |
| Design | Randomized, parallel-group, open-label |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 291 |
| Primary endpoint | Progression Free Survival (PFS) by Blinded Independent Central Review (BICR) |
| Lead sponsor | Loxo Oncology, Inc. |
| ClinicalTrials.gov | NCT04211337 |
| Registry status | Active, not recruiting |
2. Clinical Question
The central statistical question is whether assignment to selpercatinib produces a superior progression-free survival outcome compared with assignment to cabozantinib or vandetanib in the randomized population described by the registry.
Population
Participants with RET-mutant medullary thyroid cancer enrolled in the phase 3 randomized trial.
Intervention
Selpercatinib, designated Treatment A (TRT A) in the statistical analysis.
Comparator
Cabozantinib or vandetanib, designated Treatment B (TRT B). The registry analysis distinguishes intended treatment if randomized to Arm B.
Primary question
Does selpercatinib improve progression-free survival by blinded independent central review relative to cabozantinib or vandetanib?
3. Trial Design
Selpercatinib
- Drug intervention
- Analyzed as the selpercatinib treatment group
- Primary comparison against cabozantinib or vandetanib
Cabozantinib or Vandetanib
- Drug interventions
- Analyzed as the comparator group
- Stratification included intended treatment if randomized to Arm B
The registry reports an enrollment of 291 participants and two treatment groups. The statistical analyses of efficacy were based on all randomized participants, including participants who did not take the assigned treatment, did not receive the correct treatment, or otherwise did not follow the protocol.
4. Trial Timeline
Trial start
The registry lists February 11, 2020 as the study start date.
Primary completion
The registry lists May 22, 2023 as the primary completion date.
Active, not recruiting
The trial is listed as active, not recruiting in the ClinicalTrials.gov record.
5. Endpoints
The registry lists one primary endpoint, a time-to-event endpoint assessed from baseline to progressive disease or death from any cause, whichever occurs first, up to 39 months. The registry also reports secondary efficacy and tolerability outcomes.
| Endpoint | Registry definition / time frame | Type |
|---|---|---|
| Primary: Progression Free Survival (PFS) by Blinded Independent Central Review (BICR) | Baseline to Progressive Disease or Death from Any Cause, Whichever Occurs First, Up to 39 Months. PFS is defined as time from randomization until documented disease progression by BICR per RECIST 1.1 criteria, or death from any cause in the absence of BICR-documented progressive disease. | Time-to-event |
| Treatment Failure-Free Survival (TFFS) by BICR | Baseline to Progressive Disease, Unacceptable Toxicity or Death from Any Cause Up to 39 Months | Time-to-event |
| Overall Response Rate (ORR) by BICR | Baseline through Disease Progression or Death Up to 39 Months; percentage of participants with Complete Response (CR) or Partial Response (PR) | Binary |
| Duration of Response (DoR) by BICR | Date of CR or PR to Date of Disease Progression or Death Due to Any Cause Up to 39 Months | Time-to-event |
| Comparative Tolerability | Number of Weeks With High Side Effect Bother Based Score of 3 or 4 on the Functional Assessment of Cancer Therapy Item GP5 (FACT-GP5), from baseline to Progressive Disease, Unacceptable Toxicity or Death from Any Cause Up to 39 Months | Continuous |
6. Statistical Methodology
Primary PFS analysis
The primary PFS analysis compared selpercatinib with cabozantinib or vandetanib using a log-rank test. The analysis population was all randomized participants, regardless of whether they received the assigned treatment or followed the protocol.
The registry reports a hazard ratio of 0.280 with a two-sided 95% confidence interval of 0.165 to 0.475.
Stratified analysis
The primary analysis was stratified by RET mutation type (M918T vs. Others) and by intended treatment if randomized to Arm B (Vandetanib vs. Cabozantinib). This means the reported comparison was not simply an unstratified comparison of two pooled survival curves.
Log-rank testing
The log-rank test is designed to compare time-to-event distributions while accounting for the timing of events. It is therefore appropriate for endpoints such as PFS, TFFS, and DoR, where both events and censoring contribute to the analysis.
Exact binomial methodology for ORR
The registry reports the Clopper-Pearson method for ORR. This is an exact binomial confidence-interval approach. It is useful when the endpoint is a proportion because it does not rely on the same large-sample normal approximation used by simpler proportion intervals.
Wilcoxon / Mann-Whitney methodology
The comparative tolerability endpoint was analyzed using the Wilcoxon (Mann-Whitney) method. The reported effect measure was the mean difference in final values, with a two-sided 95% confidence interval.
Different estimands for different endpoint types
A central statistical feature of LIBRETTO-531 is that the trial does not use one universal statistical method for every outcome. Time-to-event endpoints use survival-analysis methods, ORR uses an exact binomial approach, and the continuous tolerability outcome uses a nonparametric comparison. The choice of method follows the structure of the endpoint rather than treating every outcome as though it were a simple continuous measurement.
7. Primary Result: Progression-Free Survival
The primary statistical analysis evaluates progression-free survival by blinded independent central review from baseline to progressive disease or death from any cause, whichever occurs first, up to 39 months.
Hazard ratio for progression or death
95% CI: 0.165–0.475 · P < 0.0001
Two-sided confidence interval; superiority hypothesis.
| Primary endpoint | Selpercatinib vs comparator | Method | Analysis population |
|---|---|---|---|
| Progression Free Survival by BICR | HR 0.280 (95% CI 0.165–0.475); P < 0.0001 | Stratified log-rank test | All randomized participants |
A hazard ratio of 0.280 means that the estimated instantaneous rate of progression or death under the selpercatinib assignment was approximately 28% of the corresponding estimated rate under Treatment B over the analyzed follow-up. Expressed as a relative complement, this corresponds to an estimated hazard that is approximately 72% lower.
The hazard ratio does not mean that 72% of participants avoided progression or death, that 72% of participants benefited, or that an individual patient's probability of progression was reduced by exactly 72%. It is a relative time-to-event measure comparing the event rates represented by the fitted analysis.
The 95% confidence interval of 0.165 to 0.475 describes statistical uncertainty around the estimated hazard ratio. It does not describe the range of treatment effects among individual participants.
The very small P-value, <0.0001, addresses the statistical evidence against the null hypothesis specified for the comparison. A P-value is not a measure of effect size and does not say that the probability the treatment works is greater than a particular percentage.
Because this is a time-to-event analysis, interpretation also depends on censoring and on the behavior of the hazards over time. A single hazard ratio compresses a potentially complex pair of event-time distributions into one relative measure.
Why the primary analysis is stratified
The registry identifies two stratification dimensions: RET mutation type and intended treatment within the comparator arm. Stratification allows the comparison to account for these prespecified factors rather than treating every randomized participant as if these characteristics were distributed without structure.
This sequence is important: the hazard ratio should be interpreted as the reported treatment effect from the prespecified stratified time-to-event analysis, not as a ratio obtained by dividing two crude percentages.
8. Secondary Result: Treatment Failure-Free Survival
Treatment Failure-Free Survival was defined from baseline to progressive disease, unacceptable toxicity, or death from any cause up to 39 months. The registry reports a stratified log-rank analysis in all randomized participants.
Hazard ratio for treatment failure
95% CI: 0.153–0.423 · P < 0.0001
Two-sided confidence interval; superiority hypothesis.
The reported HR of 0.254 indicates an estimated instantaneous rate of treatment failure approximately 25.4% as large in the selpercatinib group as in Treatment B, corresponding to an approximately 74.6% lower estimated hazard.
The important distinction from PFS is the event definition. TFFS includes unacceptable toxicity as an event in addition to progressive disease and death. Therefore, its treatment comparison incorporates both disease control and treatment-limiting toxicity into the time-to-event definition.
The 95% CI of 0.153 to 0.423 expresses uncertainty around the relative hazard estimate. The interval should not be interpreted as a range of individual patient outcomes.
The P-value of <0.0001 indicates strong statistical evidence under the reported hypothesis-testing framework, but it does not quantify the magnitude or clinical importance of the effect. The hazard ratio and its confidence interval provide that effect-size information.
9. Secondary Result: Overall Response Rate
Overall Response Rate was defined as the percentage of participants with a complete response or partial response by BICR, measured from baseline through disease progression or death up to 39 months.
Odds ratio for response
95% CI: 2.2–6.3 · P < 0.0001
Two-sided confidence interval; superiority hypothesis.
| Endpoint | Effect measure | Estimate | 95% CI | P-value | Method |
|---|---|---|---|---|---|
| Overall Response Rate by BICR | Odds ratio | 3.7 | 2.2–6.3 | <0.0001 | Clopper-Pearson method |
An odds ratio of 3.7 means that the estimated odds of achieving a complete or partial response were 3.7 times the corresponding odds under Treatment B, based on the reported comparison.
An odds ratio is not a risk ratio. It should not be restated as saying that participants were 3.7 times as likely to respond unless the distinction between odds and probability is preserved.
The 95% CI of 2.2 to 6.3 describes uncertainty around the reported odds-ratio estimate. Because the confidence interval is entirely above 1, the interval is consistent with an odds ratio favoring the selpercatinib comparison under the reported statistical framework.
The P-value of <0.0001 addresses evidence against the relevant null hypothesis. It does not measure how large the response difference is. Effect size is communicated by the odds ratio and its confidence interval.
The registry reports the Clopper-Pearson method for the binary response endpoint. This method is an exact binomial interval approach; it is distinct from the odds ratio itself, which is the reported comparative effect measure.
10. Secondary Result: Duration of Response
Duration of Response was measured from the date of complete or partial response to the date of disease progression or death due to any cause, up to 39 months. The analysis used a stratified log-rank test.
Hazard ratio for loss of response
95% CI: 0.129–0.587 · P = 0.0004
Two-sided confidence interval; superiority hypothesis.
The HR of 0.275 indicates an estimated instantaneous rate of progression or death among responders approximately 27.5% as large in the selpercatinib group as in Treatment B, corresponding to an approximately 72.5% lower estimated hazard within the response-duration analysis.
DoR is conditional on achieving a response. It therefore answers a different question from PFS: PFS begins at randomization, whereas DoR begins when a complete or partial response occurs.
The 95% CI of 0.129 to 0.587 reflects uncertainty around the estimated relative hazard. Its width also illustrates that an endpoint defined among responders can have a different precision profile from an endpoint evaluated in the full randomized population.
The P-value of 0.0004 is evidence against the reported null hypothesis under the stated analysis, but it is not itself a measure of durability or magnitude. The hazard ratio and confidence interval provide the effect-size interpretation.
11. Secondary Result: Comparative Tolerability
The registry reports a comparative tolerability endpoint measuring the number of weeks with a high side-effect-bother score of 3 or 4 on the Functional Assessment of Cancer Therapy Item GP5 (FACT-GP5). The time frame extends from baseline to progressive disease, unacceptable toxicity, or death from any cause up to 39 months.
Mean difference in final values
95% CI: -0.23 to -0.10 · P < 0.0001
Two-sided confidence interval; superiority hypothesis.
The reported mean difference in final values is -0.16, with a 95% confidence interval from -0.23 to -0.10. The negative direction indicates that the reported final-value mean difference favored the direction represented by the negative estimate, but the ClinicalTrials.gov record does not provide a more detailed scale interpretation that would justify translating this estimate into a particular clinical threshold.
The confidence interval provides information about precision around the reported difference. Because the interval remains below zero, it is consistent with a difference in the negative direction under the reported two-sided analysis.
The P-value of <0.0001 should not be treated as a measure of the size of the tolerability difference. The estimate and confidence interval describe magnitude and uncertainty; the P-value describes evidence against the null hypothesis.
The registry reports a Wilcoxon (Mann-Whitney) method for this endpoint while reporting mean difference as the effect measure. These are not interchangeable concepts: the test is nonparametric, whereas the reported effect measure is a mean difference in final values.
12. Safety: Serious Adverse Events
The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk.
| Treatment group | Serious adverse events affected / at risk |
|---|---|
| Selpercatinib — Treatment A (TRT A) | 43 / 193 |
| Cabozantinib or Vandetanib — Treatment B | 27 / 97 |
The denominator difference is important when reading safety counts. A raw count alone cannot be interpreted as a treatment-group rate because the numbers at risk differ between the reported groups.
13. Statistical Methods Explained
Why was a log-rank test used for PFS?
PFS is a time-to-event endpoint. Participants can experience progression or death at different times, while some participants may be censored before an event occurs. The log-rank test is designed to compare the survival experience of two groups across follow-up rather than comparing only a single percentage at one time point.
What does a hazard ratio of 0.280 mean?
It is a relative time-to-event effect measure. In the reported primary analysis, 0.280 means that the estimated instantaneous rate of progression or death associated with the selpercatinib assignment was about 28% of that associated with Treatment B. It does not mean that 28% of participants progressed or died.
Why was the PFS analysis stratified?
The registry specifies stratification by RET mutation type and intended treatment if randomized to Treatment B. Stratification allows the time-to-event comparison to incorporate these prespecified factors rather than ignoring them in the primary analysis.
Why is the confidence interval important?
The point estimate alone does not show how precisely the treatment effect has been estimated. The 95% CI of 0.165 to 0.475 around the primary PFS HR of 0.280 communicates the statistical uncertainty surrounding that estimate. A confidence interval is not a prediction interval for individual patients.
Why does the ORR analysis use Clopper-Pearson?
ORR is a binary endpoint: a participant either meets the complete-response or partial-response definition or does not. The Clopper-Pearson method provides an exact binomial confidence interval for a proportion, avoiding reliance on a normal approximation to the binomial distribution.
Why is the ORR effect measure an odds ratio rather than a risk ratio?
The registry reports an odds ratio of 3.7. Odds are defined as probability of response divided by probability of nonresponse. A ratio of odds therefore differs mathematically from a ratio of probabilities. Preserving that distinction prevents an odds ratio from being incorrectly described as a risk ratio.
Why should P-values not be treated as effect sizes?
A P-value addresses evidence against a null hypothesis under a specified statistical model and analysis framework. It does not quantify the size of a treatment effect. In LIBRETTO-531, effect size is communicated by measures such as the PFS HR of 0.280, the ORR odds ratio of 3.7, and the tolerability mean difference of -0.16, each accompanied by its confidence interval.
14. Analysis Populations and Randomization
The primary and secondary efficacy analyses reported in the ClinicalTrials.gov record uses all randomized participants. The registry explicitly states that participants remain in this population even if they did not take the assigned treatment, did not receive the correct treatment, or otherwise did not follow the protocol.
| Analysis feature | Registry-supported interpretation |
|---|---|
| Primary efficacy population | All randomized participants |
| Treatment assignment | Selpercatinib versus cabozantinib or vandetanib |
| Protocol deviations | Participants remain in the randomized analysis population even when treatment or protocol adherence differs from assignment |
| Primary hypothesis | Superiority |
| Primary endpoint type | Time-to-event |
This approach preserves the principal comparison created by randomization. The treatment assignment, rather than subsequent treatment behavior, defines the groups for the primary efficacy analysis. That is important because treatment switching, discontinuation, or protocol deviations occurring after randomization can otherwise introduce differences that are not part of the original randomized comparison.
15. Interim Analysis and Alpha Spending
The ClinicalTrials.gov record explicitly identify interim analysis / alpha spending as a concept in the statistical analysis of the comparative tolerability endpoint. The registry therefore indicates that interim-analysis considerations were part of the analysis framework.
Why interim analysis matters
An interim analysis evaluates accumulating information before the end of the planned observation period. Repeated efficacy looks can affect the probability of a false-positive result if they are not incorporated into the trial's error-control framework.
Why alpha spending matters
Alpha spending is a framework for controlling the overall type I error across planned looks at accumulating data. The ClinicalTrials.gov record identifies the concept but does not provide enough detail to reconstruct a specific spending function or boundary.
16. Multiplicity and Hypothesis Testing
The ClinicalTrials.gov record identifies one registered primary endpoint and several secondary statistical analyses. The primary endpoint is tested under a superiority hypothesis. Several secondary endpoints also have reported P-values.
| Endpoint | Role | Effect measure | Reported P-value |
|---|---|---|---|
| Progression Free Survival by BICR | Primary | Hazard ratio | <0.0001 |
| Treatment Failure-Free Survival by BICR | Secondary | Hazard ratio | <0.0001 |
| Overall Response Rate by BICR | Secondary | Odds ratio | <0.0001 |
| Duration of Response by BICR | Secondary | Hazard ratio | 0.0004 |
| Comparative Tolerability | Secondary | Mean difference | <0.0001 |
These P-values should not automatically be interpreted as though every secondary endpoint were an independent confirmatory hypothesis with its own error budget. The ClinicalTrials.gov record identifies the endpoints and their analyses but does not provide a complete multiplicity-adjustment hierarchy. Accordingly, this page reports the registry values without constructing an unreported multiplicity procedure.
17. Understanding the Primary Hazard Ratio
The primary PFS HR of 0.280 is the most direct summary of the reported randomized time-to-event comparison. It is a relative measure of the event rate represented by progression or death.
The 95% CI of 0.165–0.475 indicates the statistical precision of the primary estimate under the reported analysis. The interval remains below 1 throughout, which is consistent with the direction of the reported superiority comparison.
The P-value of <0.0001 provides evidence against the null hypothesis in the reported two-sided log-rank analysis. It should not be translated into the probability that the null hypothesis is true or false.
The hazard ratio does not give a median PFS, a fixed-time PFS probability, an absolute risk difference, or the probability that an individual participant will benefit. Those quantities require different information and different estimands.
18. Why Time-to-Event Analysis Is Central to LIBRETTO-531
Three of the five posted statistical analyses concern time-to-event endpoints: PFS, TFFS, and DoR. This creates an important statistical distinction from ordinary binary comparisons.
Time matters
A participant progressing early and a participant progressing much later do not provide the same information as a simple yes/no endpoint at the end of follow-up.
Censoring matters
Participants who have not experienced the event by their last evaluable observation can contribute information up to their censoring time.
Event definition matters
PFS, TFFS, and DoR use different starting points and different event definitions, so their hazard ratios answer different questions.
Relative and absolute effects differ
A hazard ratio summarizes relative event rates. It cannot by itself provide the absolute probability of remaining progression-free at a particular time.
The primary LIBRETTO-531 analysis uses this time-to-event structure rather than reducing follow-up to a single binary outcome.
19. Comparing the Reported Effect Measures
| Endpoint | Effect measure | Estimate | 95% CI | What the measure describes |
|---|---|---|---|---|
| PFS | Hazard ratio | 0.280 | 0.165–0.475 | Relative event hazard for progression or death |
| TFFS | Hazard ratio | 0.254 | 0.153–0.423 | Relative event hazard for progression, unacceptable toxicity, or death |
| ORR | Odds ratio | 3.7 | 2.2–6.3 | Relative odds of complete or partial response |
| DoR | Hazard ratio | 0.275 | 0.129–0.587 | Relative event hazard from response to progression or death |
| Comparative tolerability | Mean difference | -0.16 | -0.23 to -0.10 | Difference in final values for the reported tolerability measure |
The estimates should not be placed on a common numerical scale. A hazard ratio, odds ratio, and mean difference have different mathematical meanings. The most important statistical discipline is therefore to retain the effect-measure label whenever quoting an estimate.
20. Limitations
- Open-label design: the registry describes no masking. Knowledge of treatment assignment can influence some aspects of patient management and subjective assessments, although the primary PFS endpoint was assessed by blinded independent central review.
- Primary endpoint scope: the registered primary endpoint is PFS by BICR, not overall survival or a composite measure of all clinical outcomes.
- Different endpoint definitions: PFS, TFFS, ORR, DoR, and tolerability measure different clinical constructs and should not be interpreted as interchangeable outcomes.
- Hazard-ratio interpretation: a hazard ratio is a relative time-to-event measure and should not be translated into an absolute probability or percentage of patients benefiting.
- Proportional-hazards caution: interpretation of a single hazard ratio is most straightforward when the relative hazard is reasonably stable over time. The ClinicalTrials.gov record does not provide a diagnostic assessment of proportional hazards.
- Analysis-population distinction: the reported efficacy analyses use all randomized participants, preserving the randomized comparison but potentially including participants who did not receive or adhere to assigned treatment.
- Multiplicity: multiple secondary analyses are reported, but the ClinicalTrials.gov record does not specify a complete multiplicity hierarchy or adjustment procedure.
- Interim analysis: the ClinicalTrials.gov record identifies interim-analysis and alpha-spending considerations but does not provide sufficient information to reconstruct the exact statistical boundary or spending function.
- Safety denominator: serious adverse-event counts are reported with their at-risk denominators, but the ClinicalTrials.gov record does not provide a formal statistical comparison of serious adverse events.
- Incomplete numerical detail: the ClinicalTrials.gov record does not report medians, fixed-time survival probabilities, baseline characteristics, subgroup estimates, or a Kaplan-Meier event table, so those quantities are not inferred here.
21. Why This Trial Matters Statistically
LIBRETTO-531 is a useful teaching example because several core clinical-trial methods appear within a single randomized comparison. The trial combines a time-to-event primary endpoint, blinded independent central review, stratified survival analysis, an odds-ratio response analysis, a nonparametric tolerability analysis, and interim-analysis considerations.
| Statistical concept | How it appears in LIBRETTO-531 |
|---|---|
| Randomization | 291 participants were randomized in a parallel-group phase 3 design. |
| Primary time-to-event endpoint | PFS by BICR, from baseline to progressive disease or death, whichever occurs first, up to 39 months. |
| Log-rank test | Used for the primary PFS analysis and the reported TFFS and DoR analyses. |
| Stratified analysis | Primary PFS analysis stratified by RET mutation type and intended Treatment B. |
| Hazard ratio | Used for PFS, TFFS, and DoR treatment effects. |
| Confidence intervals | Reported with the treatment-effect estimates to quantify statistical uncertainty. |
| Exact binomial method | Clopper-Pearson method reported for ORR. |
| Odds ratio | Reported as the comparative effect measure for ORR. |
| Wilcoxon / Mann-Whitney | Reported for comparative tolerability. |
| Interim analysis / alpha spending | Identified as a statistical-analysis concept for the tolerability analysis. |
| ITT-style efficacy analysis | All randomized participants are included even when assigned treatment was not received or protocol adherence differed. |
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: LIBRETTO-531, NCT04211337.
- PubMed: PMID 40617171.
- PubMed: PMID 39694978.
- PubMed: PMID 37870969.
- PubMed: PMID 35969032.
- PubMed: PMID 34292174.
Continue with Clinical Biostats
Connect the statistical methods in this trial to deeper tutorials, calculators, and additional clinical-trial analyses.
25. Record Summary
LIBRETTO-531 provides a compact example of several important principles in clinical-trial statistics. The primary endpoint is a time-to-event outcome analyzed in all randomized participants using a stratified log-rank framework and summarized with a hazard ratio. The reported primary PFS estimate was 0.280, with a two-sided 95% CI of 0.165–0.475 and a P-value of <0.0001.
The secondary analyses illustrate why endpoint-specific methodology matters. TFFS and DoR use time-to-event methods but have different event definitions and starting points. ORR is binary and uses the Clopper-Pearson method with an odds ratio of 3.7. Comparative tolerability is analyzed using a Wilcoxon / Mann-Whitney method with a reported mean difference of -0.16.
The most useful statistical interpretation therefore does not reduce the trial to a single P-value. It distinguishes the randomized population, endpoint definitions, analysis methods, effect measures, confidence intervals, and assumptions underlying time-to-event interpretation. It also keeps efficacy and safety on separate analytical tracks and avoids treating reported secondary P-values as automatically equivalent to a prespecified primary confirmatory test.