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
EMERALD was a randomized, open-label, parallel-group phase 3 trial comparing elacestrant with standard of care in patients with ER+/HER2- advanced breast cancer. The registry reports two primary time-to-event endpoints and formal statistical analyses using stratified log-rank tests and stratified Cox proportional-hazards models.
| Feature | EMERALD |
|---|---|
| Trial name | EMERALD |
| Phase | Phase 3 |
| Condition | Breast Cancer |
| Brief title | Phase 3 Trial of Elacestrant Versus Standard of Care for the Treatment of ER+/HER2- Advanced Breast Cancer |
| Design | Randomized, parallel-group |
| Masking | None |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 478 |
| Arms | 2 |
| Interventions | Elacestrant; Standard of Care (SoC) |
| Status | COMPLETED |
| Study start | 2019-05-10 |
| Primary completion | 2021-08-24 |
| Lead sponsor | Stemline Therapeutics, Inc. |
| Sponsor type | INDUSTRY |
| ClinicalTrials.gov | NCT03778931 |
2. Clinical Question
The central statistical question was whether elacestrant produced a different time-to-event profile than standard of care for progression-free survival, with particular attention to participants identified as ESR1-mut and to the full randomized population.
Population
Participants in the phase 3 EMERALD trial with ER+/HER2- advanced breast cancer, including an ESR1-mut subgroup and the full participant population.
Intervention
Elacestrant.
Comparator
Standard of care (SoC).
Primary question
How does elacestrant compare with standard of care for progression-free survival in ESR1-mut participants and in all participants?
3. Trial Design
Elacestrant
- Elacestrant
- Randomized treatment arm
- Primary comparison against standard of care
Standard of Care
- Standard of care (SoC)
- Randomized comparator arm
- Primary comparison against elacestrant
The trial was open-label rather than masked. That feature is relevant to interpretation because treatment assignment was not concealed from participants or investigators after randomization. However, the primary progression-free survival endpoint was based on blinded imaging review committee assessment, which provides an important distinction between the open-label treatment administration and the blinded assessment of the imaging-based endpoint.
4. Primary Endpoints
| Endpoint | Registry definition / time frame | Statistical analysis |
|---|---|---|
| Progression-free Survival in ESR1-mut Participants | From Date of Randomization until Disease Progression or Death Due to Any Cause (up to 12 Months). Progression-free survival was based on blinded IRC assessment in ESR1-mut participants and defined as the length of time from randomization until the date of objective disease progression per RECIST v1.1 as assessed by the blinded IRC or death from any cause. | Stratified Cox proportional-hazards model; two-sided stratified log-rank test |
| Progression-free Survival in All Participants | From Date of Randomization until Disease Progression or Death Due to Any Cause (up to 12 Months). The registry defines this as progression-free survival based on blinded imaging review committee assessment in all ESR1-mut and ESR1-wt participants. | Stratified Cox proportional-hazards model; two-sided stratified log-rank test |
Both registered primary endpoints are time-to-event endpoints. That matters statistically because participants can have different follow-up times, and the analysis must account for both the timing of observed events and participants who have not experienced progression or death by the end of their observed follow-up.
5. Results: Progression-Free Survival in ESR1-mut Participants
The registry reports a formal primary analysis comparing elacestrant with standard of care in ESR1-mut participants. The reported effect measure is the hazard ratio from a stratified Cox proportional-hazards model, while the P-value was generated using a two-sided stratified log-rank test.
Hazard ratio for progression or death
95% CI: 0.387–0.768 · P = 0.0005
Elacestrant vs Standard of Care
| Feature | Reported analysis |
|---|---|
| Endpoint | Progression-free Survival in ESR1-mut Participants |
| Time frame | From Date of Randomization until Disease Progression or Death Due to Any Cause (up to 12 Months) |
| Groups compared | Elacestrant vs Standard of Care (SoC) |
| Method | Log Rank |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.546 |
| 95% CI | 0.387–0.768 |
| P-value | 0.0005 |
| CI approach | Profile likelihood |
| Cox model | Stratified Cox Proportional Hazards model with ties=Efron |
| Stratification factors | Prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
| P-value method | Two-sided stratified log-rank test |
The hazard ratio of 0.546 means that, under the fitted stratified Cox model, the estimated instantaneous hazard of progression or death was approximately 54.6% as large with elacestrant as with standard of care in the analyzed ESR1-mut population. Expressed as a simple relative-hazard interpretation, this corresponds to an estimated 45.4% lower hazard under the model.
The HR does not mean that 45.4% of participants avoided progression, that individual participants had exactly a 45.4% reduction in their personal risk, or that median progression-free survival was reduced or increased by 45.4%. A hazard ratio is a relative time-to-event measure, not an absolute probability.
The 95% confidence interval of 0.387–0.768 describes uncertainty around the estimated hazard ratio under the model and sampling framework. It does not describe the range of effects that individual patients would experience. Because the entire interval is below 1, the registry's estimated treatment effect is consistently below the null value across the reported interval.
The P-value of 0.0005 addresses the compatibility of the observed time-to-event comparison with the null hypothesis under the specified testing framework. It does not measure the magnitude of the treatment effect and should not be interpreted as the probability that the null hypothesis is true.
The interpretation also depends on the Cox model and its proportional-hazards framework. A single HR is most straightforward when the relative hazards are reasonably stable over time. The ClinicalTrials.gov record does not provide a separate assessment of that assumption, so the HR should be understood as the model-based summary reported for the analysis.
6. Results: Progression-Free Survival in All Participants
The second primary endpoint extended the progression-free survival analysis to all participants, defined by the registry as the ESR1-mut and ESR1-wt population combined.
Hazard ratio for progression or death
95% CI: 0.552–0.880 · P = 0.0018
Elacestrant vs Standard of Care
| Feature | Reported analysis |
|---|---|
| Endpoint | Progression-free Survival in All Participants |
| Time frame | From Date of Randomization until Disease Progression or Death Due to Any Cause (up to 12 Months) |
| Groups compared | Elacestrant vs Standard of Care (SoC) |
| Method | Log Rank |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.697 |
| 95% CI | 0.552–0.880 |
| P-value | 0.0018 |
| CI approach | Profile likelihood |
| Cox model | Stratified Cox Proportional Hazards model with ties=Efron |
| Stratification factors | Prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
| P-value method | Two-sided stratified log-rank test |
The hazard ratio of 0.697 means that, under the fitted stratified Cox model, the estimated instantaneous hazard of progression or death was approximately 69.7% as large with elacestrant as with standard of care in the all-participant analysis. A direct derived interpretation is an estimated 30.3% lower hazard under the model.
Again, this is not an absolute risk reduction and does not mean that 30.3% of participants were protected from progression or death. The hazard ratio summarizes the relative event rate over time under the specified model.
The 95% confidence interval, 0.552–0.880, quantifies uncertainty around the estimated HR. The interval remains below 1, so the reported uncertainty interval does not include the conventional no-difference value of 1.
The P-value of 0.0018 is evidence against the null hypothesis under the reported two-sided stratified log-rank testing procedure. It is not an estimate of the probability that elacestrant is effective, nor does it quantify clinical magnitude.
The analysis used the same two stratification factors reported for the ESR1-mut primary endpoint: prior treatment with fulvestrant and presence of visceral metastases. Stratification helps the comparison account for these prespecified factors rather than treating all participants as if those characteristics had been ignored.
7. Secondary Results: Overall Survival in ESR1-mut Participants
The registry also reports an overall survival analysis in ESR1-mut participants. This is a secondary endpoint rather than one of the two registered primary endpoints.
Hazard ratio for death
95% CI: 0.361–0.958 · P = 0.0325
Elacestrant vs Standard of Care
| Feature | Reported analysis |
|---|---|
| Endpoint | Overall Survival in ESR1-mut Participants |
| Time frame | From Date of Randomization until Death Due to Any Cause (Estimated up to 24 Months) |
| Groups compared | Elacestrant vs Standard of Care (SoC) |
| Method | Log Rank |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.592 |
| 95% CI | 0.361–0.958 |
| P-value | 0.0325 |
| Cox model | Stratified Cox Proportional Hazards model with ties=Efron |
| Stratification factors | Prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
| CI approach | Profile likelihood |
The HR of 0.592 corresponds to an estimated instantaneous hazard of death approximately 59.2% as large with elacestrant as with standard of care under the fitted stratified Cox model. A simple derived interpretation is an estimated 40.8% lower hazard of death.
The 95% CI of 0.361–0.958 indicates appreciable uncertainty around the point estimate. Although the interval is below 1, its upper boundary is relatively close to 1 compared with the point estimate, illustrating why the confidence interval is essential when interpreting an HR.
The P-value of 0.0325 is a two-sided inferential quantity associated with the reported analysis. It is not a measure of effect size, and it should not be used to convert the HR into an absolute clinical benefit.
This is a secondary endpoint. Its statistical interpretation therefore should remain tied to its prespecified role and the multiplicity structure of the overall trial. The ClinicalTrials.gov record does not provide an alpha-allocation scheme or multiplicity-adjustment procedure for the complete endpoint family, so no stronger claim about familywise error control should be inferred.
8. Secondary Results: Overall Survival in All Participants
The registry reports a second overall survival analysis covering all participants.
Hazard ratio for death
95% CI: 0.536–1.025 · P = 0.0697
Elacestrant vs Standard of Care
| Feature | Reported analysis |
|---|---|
| Endpoint | Overall Survival in All Participants |
| Time frame | From Date of Randomization until Death Due to Any Cause (Estimated up to 24 Months) |
| Groups compared | Elacestrant vs Standard of Care (SoC) |
| Method | Log Rank |
| Effect measure | Hazard Ratio (HR) |
| Estimate | 0.742 |
| 95% CI | 0.536–1.025 |
| P-value | 0.0697 |
| Cox model | Stratified Cox Proportional Hazards model with ties=Efron |
| Stratification factors | ESR1-mutational status (ESR1-mut vs ESR1-wt); prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
| P-value method | Two-sided stratified log-rank test |
The HR of 0.742 corresponds to an estimated instantaneous hazard of death approximately 74.2% as large with elacestrant as with standard of care under the reported stratified Cox model. A simple derived interpretation is an estimated 25.8% lower hazard.
The confidence interval of 0.536–1.025 is particularly important here because it includes 1. The point estimate is below 1, but the reported 95% confidence interval extends slightly above the conventional null value.
The P-value of 0.0697 likewise should not be interpreted as a measure of effect magnitude. A P-value does not tell us that the treatment effect is exactly the HR estimate, nor does it establish the probability that one treatment is superior.
The appropriate conclusion from the ClinicalTrials.gov record is descriptive: the estimated HR was below 1, but the reported two-sided 95% CI included 1. Because this was a secondary endpoint, interpretation also needs to account for the endpoint's place within the overall statistical testing strategy.
9. Summary of Reported Efficacy Analyses
| Endpoint | Role | HR | 95% CI | P-value |
|---|---|---|---|---|
| Progression-free Survival in ESR1-mut Participants | Primary | 0.546 | 0.387–0.768 | 0.0005 |
| Progression-free Survival in All Participants | Primary | 0.697 | 0.552–0.880 | 0.0018 |
| Overall Survival in ESR1-mut Participants | Secondary | 0.592 | 0.361–0.958 | 0.0325 |
| Overall Survival in All Participants | Secondary | 0.742 | 0.536–1.025 | 0.0697 |
The four posted analyses share a common statistical structure: treatment groups are compared using a log-rank framework, while the hazard ratio is estimated with a stratified Cox proportional-hazards model. The stratification factors differ between the two ESR1-focused analyses and the all-participant overall survival analysis, as specified in the registry analysis notes.
10. Safety
The ClinicalTrials.gov record reports serious adverse events by randomized arm using affected participants over participants at risk. This is a separate evidentiary domain from the time-to-event efficacy analyses.
| Safety measure | Elacestrant | Standard of Care |
|---|---|---|
| Serious adverse events | 29/237 | 25/230 |
The reported serious adverse-event figures are counts over the corresponding numbers at risk. They should not be converted into a comparative risk ratio or risk difference without additional assumptions or analysis. In particular, the ClinicalTrials.gov record does not provide a formal statistical comparison for these serious adverse-event counts.
11. Statistical Methodology
Kaplan-Meier estimation
Progression-free survival and overall survival are time-to-event outcomes. Kaplan-Meier estimation is the standard nonparametric framework for describing the probability of remaining event-free over time while accommodating right censoring.
Here, di is the number of events at event time ti and ni is the number at risk immediately before that time.
The registry-reported EMERALD data do not report Kaplan-Meier medians or time-specific survival estimates, so the statistical results on this page focus on the hazard ratios, confidence intervals, and P-values that are explicitly posted in the registry.
Stratified log-rank test
The log-rank test compares time-to-event experience between randomized groups across follow-up rather than reducing every participant to a single binary outcome. A stratified version performs that comparison while accounting for specified stratification factors.
For the two primary progression-free survival analyses, the registry states that the P-value was generated using a two-sided stratified log-rank test. The stratification factors were prior treatment with fulvestrant and presence of visceral metastases.
Stratified Cox proportional-hazards model
The registry reports that the hazard ratios were estimated using a stratified Cox Proportional Hazards model with ties=Efron. Stratification permits the baseline hazard to differ across the specified strata while estimating a common treatment-effect parameter across them.
The hazard ratio is a model-based relative measure. It is not the same as a relative risk, an absolute risk difference, a probability of benefit, or a percentage of patients who respond.
Efron handling of ties
Event times in clinical-trial datasets can contain ties, meaning more than one event can be recorded at the same observed time. The registry specifically reports ties=Efron for the Cox models. The Efron approach provides a way to handle tied event times when constructing the partial likelihood used by the Cox model.
Profile-likelihood confidence intervals
For the primary analyses, the registry states that the confidence interval was calculated using a profile likelihood approach. This is distinct from simply reporting the point estimate alone: the confidence interval provides an interval measure of statistical uncertainty around the fitted hazard ratio.
Stratification
Stratification is important because the registry explicitly identifies factors that were incorporated into the time-to-event analyses. For the ESR1-mut primary analysis and the all-participant PFS analysis, these were prior treatment with fulvestrant and presence of visceral metastases. For all-participant overall survival, ESR1-mutational status was additionally included as a stratification factor.
12. Statistical Methods Explained
Why was a log-rank test used?
Progression-free survival and overall survival are time-to-event endpoints, so the timing of events and censoring matter. The log-rank test is designed to compare event-time distributions between treatment groups while using the follow-up information available at each event time. In EMERALD, the registry reports a stratified version so the comparison accounts for specified stratification factors.
What does an HR of 0.546 mean?
An HR of 0.546 means that the fitted model estimates the instantaneous progression-or-death hazard in the elacestrant group to be 54.6% of the corresponding hazard in the standard-of-care group for the ESR1-mut primary analysis. The direct derived difference is 1 − 0.546 = 0.454, or an estimated 45.4% lower hazard. It does not mean that 45.4% of participants avoided progression or that each participant experienced a 45.4% reduction in personal risk.
Why is the confidence interval important?
The point estimate is only one estimate from the observed trial data. The 95% confidence interval describes statistical uncertainty around that estimate under the specified analysis framework. For the ESR1-mut PFS analysis, the interval is 0.387–0.768. For all-participant PFS, it is 0.552–0.880. Looking only at the HR without its interval hides important information about precision.
Why doesn't the P-value measure effect size?
A P-value measures how compatible the observed data are with a specified null hypothesis under the statistical testing procedure. It depends on both the observed effect and the amount of information in the analysis. The HR is the quantity that summarizes the estimated relative treatment effect; the confidence interval describes uncertainty around it. Therefore a smaller P-value does not automatically mean a larger or more clinically important effect.
What does stratification do in a Cox model?
A stratified Cox model allows the baseline hazard to vary across the specified strata rather than forcing one common baseline hazard for all participants. The treatment-effect parameter is then estimated while respecting the stratification structure. In EMERALD, the registry explicitly reports prior fulvestrant treatment and visceral metastases as stratification factors for the primary PFS analyses.
Why does censoring matter?
In a time-to-event analysis, not every participant necessarily experiences the event during observed follow-up. A censored participant contributes information up to the time at which their event status is no longer observed. Kaplan-Meier and Cox methods are designed to use this partial follow-up rather than simply excluding such participants. The validity of that approach depends on assumptions concerning the censoring mechanism and the broader study design.
Why is the proportional-hazards assumption relevant?
The Cox model summarizes the relative event hazard with a hazard ratio. If the relative hazard changes substantially over time, a single HR can become less descriptive of the full time course. The ClinicalTrials.gov record does not report a formal proportional-hazards diagnostic, so the reported HR should be interpreted as the model-based summary rather than as proof that the hazards were proportional throughout follow-up.
13. Primary vs Secondary Endpoints
| Endpoint | Role | Time frame | Analysis reported |
|---|---|---|---|
| Progression-free Survival in ESR1-mut Participants | Primary | From Date of Randomization until Disease Progression or Death Due to Any Cause (up to 12 Months) | Stratified Cox HR; two-sided stratified log-rank P-value |
| Progression-free Survival in All Participants | Primary | From Date of Randomization until Disease Progression or Death Due to Any Cause (up to 12 Months) | Stratified Cox HR; two-sided stratified log-rank P-value |
| Overall Survival in ESR1-mut Participants | Secondary | From Date of Randomization until Death Due to Any Cause (Estimated up to 24 Months) | Stratified Cox HR; log-rank analysis |
| Overall Survival in All Participants | Secondary | From Date of Randomization until Death Due to Any Cause (Estimated up to 24 Months) | Stratified Cox HR; two-sided stratified log-rank P-value |
The distinction between primary and secondary endpoints is essential. A result on a secondary endpoint should not automatically be interpreted with the same confirmatory status as a primary endpoint. The ClinicalTrials.gov record identifies the endpoint roles but do not provide a complete multiplicity-adjustment or alpha-allocation scheme for the entire endpoint family.
14. Multiplicity and Inferential Interpretation
EMERALD has two registered primary endpoints and additional secondary analyses. When a trial evaluates multiple endpoints, repeated formal testing can increase the probability of observing at least one statistically unusual result under a global null hypothesis unless the design accounts for multiplicity.
Two primary endpoints
The registry identifies both ESR1-mut progression-free survival and all-participant progression-free survival as primary endpoints.
Secondary endpoints
Overall survival in ESR1-mut participants and overall survival in all participants are posted as secondary analyses.
Nominal P-values
A reported P-value describes the specified individual test. It does not, by itself, establish the familywise error rate for all analyses in the trial.
What the registry does not provide here
The ClinicalTrials.gov record does not specify an alpha-spending, hierarchical testing, or other complete multiplicity-control strategy for the full endpoint family.
Consequently, the most defensible presentation is to preserve the registry's endpoint roles and reported P-values without assigning an additional confirmatory interpretation that is not supported by the ClinicalTrials.gov record.
15. Stratification and Why It Matters
The EMERALD analyses are explicitly stratified. For the primary PFS analyses, the stratification factors were prior treatment with fulvestrant (yes vs no) and presence of visceral metastases (yes vs no). For overall survival in all participants, the registry additionally specifies ESR1-mutational status (ESR1-mut vs ESR1-wt).
| Analysis | Stratification factors |
|---|---|
| Progression-free Survival in ESR1-mut Participants | Prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
| Progression-free Survival in All Participants | Prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
| Overall Survival in ESR1-mut Participants | Prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
| Overall Survival in All Participants | ESR1-mutational status (ESR1-mut vs ESR1-wt); prior treatment with fulvestrant (yes vs no); presence of visceral metastases (yes vs no) |
Stratification does not mean that the treatment effect is separately estimated and tested within every stratum. Rather, the Cox and log-rank procedures incorporate the strata into the overall comparison. This can improve alignment between the analysis and the randomized trial design when the stratification variables are prognostically relevant or were prespecified for balancing purposes.
16. Interpreting the Four Hazard Ratios Together
| Endpoint | HR | Simple derived interpretation | 95% CI includes 1? |
|---|---|---|---|
| ESR1-mut PFS | 0.546 | Approximately 45.4% lower estimated hazard | No |
| All-participant PFS | 0.697 | Approximately 30.3% lower estimated hazard | No |
| ESR1-mut OS | 0.592 | Approximately 40.8% lower estimated hazard | No |
| All-participant OS | 0.742 | Approximately 25.8% lower estimated hazard | Yes |
The table provides a compact statistical description rather than a ranking of the endpoints. The four HRs answer different questions in different analysis populations and endpoint roles. In particular, progression-free survival and overall survival are not interchangeable outcomes, and the ESR1-mut analyses are not the same estimand as the all-participant analyses.
The fact that the ESR1-mut PFS HR is 0.546 and the all-participant PFS HR is 0.697 does not by itself establish that treatment effect differs by ESR1 status. A formal claim of heterogeneity would require an appropriate interaction or treatment-by-subgroup analysis. The ClinicalTrials.gov record does not report such an interaction test.
17. What the Confidence Intervals Tell Us
Confidence intervals are particularly informative when the point estimate alone could give an exaggerated impression of precision.
ESR1-mut PFS
The HR is 0.546 with a 95% CI of 0.387–0.768. The interval is entirely below 1, while still showing meaningful uncertainty around the point estimate.
All-participant PFS
The HR is 0.697 with a 95% CI of 0.552–0.880. The interval remains below 1 but is centered closer to the null than the ESR1-mut PFS estimate.
ESR1-mut OS
The HR is 0.592 with a 95% CI of 0.361–0.958. The upper confidence limit is below 1, but the interval is comparatively broad.
All-participant OS
The HR is 0.742 with a 95% CI of 0.536–1.025. The interval extends across 1, so the reported uncertainty interval includes the conventional no-difference value.
A confidence interval should not be read as the probability that the true HR lies inside the displayed interval. Instead, it is a frequentist interval constructed by a specified statistical procedure. The profile-likelihood approach reported for the primary analyses is the method used to calculate those primary confidence intervals.
18. P-Values and What They Do Not Tell You
The registry reports P-values of 0.0005, 0.0018, 0.0325, and 0.0697 for the four formal analyses. These values are associated with the respective statistical testing procedures and should be read alongside the effect estimates and confidence intervals.
Keeping these roles separate prevents the common error of treating a P-value as if it were a measure of treatment magnitude or clinical importance.
For example, the ESR1-mut PFS analysis has an HR of 0.546 and a P-value of 0.0005. The P-value does not mean that the effect is "stronger" by a numerical factor than the all-participant PFS result with a P-value of 0.0018. The HRs and their confidence intervals are the quantities that describe the estimated relative treatment effects.
19. Analysis Population and Estimand Considerations
The ClinicalTrials.gov record distinguishes analyses by endpoint population rather than providing a single universal efficacy population. Two primary endpoints specifically concern ESR1-mut participants, while the other primary endpoint and one secondary overall survival endpoint concern all participants.
| Population scope | Endpoints reported |
|---|---|
| ESR1-mut participants | Progression-free survival; overall survival |
| All participants | Progression-free survival; overall survival |
This distinction matters because an HR estimated in an ESR1-mut subgroup is an effect estimate for that population, whereas an all-participant HR is an aggregate treatment effect over the broader population represented in that analysis. The two estimates should not be treated as if they were interchangeable.
The ClinicalTrials.gov record does not provide detailed analysis-population counts for each endpoint, treatment discontinuation patterns, censoring counts, missing-data procedures, or formal estimand definitions. Those details are therefore not added to this page.
20. Censoring and Time-to-Event Interpretation
Progression-free survival is defined from randomization until disease progression or death from any cause. Overall survival is defined from randomization until death from any cause. These definitions create a time-to-event structure in which the precise timing of an event matters.
A participant who has not experienced the specified event by the end of observable follow-up does not simply become a zero. Instead, the participant contributes observed follow-up information and may be censored. Survival methods use that information to estimate the event-time distribution and compare treatment groups.
21. Limitations
- Open-label design: the trial was not masked. Treatment assignment was therefore known during treatment, which can matter for subjective assessments and treatment behavior. The primary PFS assessment was nevertheless based on blinded IRC assessment as specified in the registry.
- Hazard-ratio assumptions: the Cox model provides a single relative hazard summary. If hazards are not approximately proportional over time, that single number can be less descriptive of the full survival experience.
- Secondary-endpoint multiplicity: the ClinicalTrials.gov record identifies secondary analyses but do not provide a complete multiplicity-control strategy for all posted analyses.
- Subgroup interpretation: the ESR1-mut analyses concern a defined subgroup and should not automatically be interpreted as formal evidence that the treatment effect differs from the ESR1-wt population.
- Confidence-interval precision: the width of an interval matters. Point estimates alone do not communicate the uncertainty of the statistical estimate.
- Limited registry detail: the ClinicalTrials.gov record does not provide median PFS, median OS, time-specific survival probabilities, event counts, baseline characteristics, subgroup forest plots, or detailed missing-data procedures. Those quantities are not inferred here.
- Safety comparison: the registry-reported serious adverse-event data provide affected/at-risk counts but not a formal between-group statistical comparison.
- Time-frame interpretation: the primary endpoints are reported over a time frame of up to 12 months, while the secondary overall survival endpoints use an estimated time frame of up to 24 months. Results from different endpoint time frames should not be conflated.
22. Why This Trial Matters Statistically
EMERALD is a useful teaching example because the registry results illustrate several central concepts in modern clinical-trial survival analysis without requiring the reader to infer statistics that were not reported.
| Concept | How it appears in EMERALD |
|---|---|
| Randomization | The trial is randomized with two parallel treatment arms. |
| Time-to-event endpoints | Both registered primary endpoints are progression-free survival outcomes. |
| Kaplan-Meier framework | Time-to-event endpoints are naturally represented using survival-function methods. |
| Log-rank testing | The registry reports Log Rank as the analysis method for the formal comparisons. |
| Stratified log-rank test | The primary P-values were generated using two-sided stratified log-rank tests. |
| Hazard ratio | Each posted statistical analysis reports an HR comparing elacestrant with standard of care. |
| Cox model | HR estimation used a stratified Cox proportional-hazards model with ties=Efron. |
| Confidence intervals | The primary HR confidence intervals were calculated using a profile likelihood approach. |
| Stratification | Prior fulvestrant treatment and visceral metastases were used as stratification factors for the primary PFS analyses. |
| Subgroup population | ESR1-mut participants form the population for one primary and one secondary endpoint analysis. |
| Multiplicity | Two primary endpoints coexist with secondary overall survival analyses, requiring careful interpretation of individual P-values. |
| Safety analysis | Serious adverse events are reported separately from efficacy time-to-event outcomes. |
23. Clinical Biostats Interpretation
The two primary progression-free survival analyses both produced HR estimates below 1: 0.546 in ESR1-mut participants and 0.697 in all participants. Their respective 95% confidence intervals were 0.387–0.768 and 0.552–0.880, and their reported P-values were 0.0005 and 0.0018.
These are model-based relative treatment-effect estimates. They do not provide an absolute probability of progression-free survival, a median survival time, or a percentage of patients who benefit.
The ESR1-mut overall survival analysis reported an HR of 0.592 with a 95% CI of 0.361–0.958 and P = 0.0325. The all-participant overall survival analysis reported an HR of 0.742 with a 95% CI of 0.536–1.025 and P = 0.0697.
The all-participant OS confidence interval includes 1, so its reported uncertainty interval spans both the conventional no-difference value and HR values below 1. That distinction is more informative than simply describing the point estimate as being below 1.
The most informative reading of EMERALD is to keep the treatment effect, precision, testing procedure, endpoint role, and analysis population separate. Hazard ratios describe relative event hazards; confidence intervals describe uncertainty; P-values describe evidence against a specified null; and endpoint definitions determine exactly what event is being analyzed.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Calculators
26. Sources
- ClinicalTrials.gov: EMERALD, NCT03778931.
- PubMed: PMID 38381994.
- PubMed: PMID 35584336.
- PubMed: PMID 31426673.
Continue through Clinical Biostats
Connect this trial's survival-analysis methods with deeper statistical tutorials and practical calculation tools.
27. Record Summary
EMERALD provides a focused example of randomized clinical-trial survival analysis. The trial enrolled 478 participants in two parallel randomized arms comparing elacestrant with standard of care. Its two registered primary endpoints were progression-free survival in ESR1-mut participants and progression-free survival in all participants, both measured from randomization until disease progression or death due to any cause over a time frame of up to 12 months.
The registry reports hazard ratios of 0.546 and 0.697 for those two primary analyses, with 95% confidence intervals of 0.387–0.768 and 0.552–0.880, respectively. Both analyses used stratified Cox proportional-hazards models with ties handled by Efron, while the reported P-values came from two-sided stratified log-rank tests. The primary confidence intervals were calculated using a profile likelihood approach.
The secondary overall survival analyses reported HRs of 0.592 in ESR1-mut participants and 0.742 in all participants. The corresponding 95% confidence intervals were 0.361–0.958 and 0.536–1.025. These results illustrate why a statistical interpretation should consider the endpoint role, analysis population, effect estimate, confidence interval, testing method, and stratification structure together rather than relying on a P-value alone.