← Clinical Trials
Advanced Breast Cancer Phase 2 Non-Inferiority NCT01783444

BOLERO-6: Complete Statistical Analysis of Everolimus Plus Exemestane in Advanced Breast Cancer

An independent statistical analysis of the randomized phase 2 BOLERO-6 trial comparing everolimus 10 mg plus exemestane 25 mg with everolimus 10 mg alone and capecitabine 1250 mg/m2 in advanced breast cancer, with emphasis on progression-free survival, non-inferiority interpretation, quality of life, and safety.

Completed  ·  Enrollment 309  ·  Randomized, parallel, open-label
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical results on this page are restricted to the trial data posted on ClinicalTrials.gov for BOLERO-6.

1. Trial at a Glance

BOLERO-6 was a randomized phase 2 trial in advanced breast cancer with three parallel treatment groups. Its registered primary endpoint was progression-free survival comparing everolimus plus exemestane with everolimus alone, with the primary analysis framed as a non-inferiority comparison.

309
Enrollment
309.0 patients
3
Treatment Arms
Randomized parallel design
0.74
Primary PFS HR
90% CI 0.57–0.97
16
Analyses Posted
1 primary; 15 secondary
FeatureBOLERO-6
PhasePhase 2
ConditionBreast Cancer
Brief titleA Phase II Study of Everolimus in Combination With Exemestane Versus Everolimus Alone Versus Capecitabine in Advance Breast Cancer.
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Enrollment309.0
Primary endpointProgression Free Survival (PFS) - Everolimus Plus Exemestane Versus Everolimus Alone
Results postedYes
Outcome measures posted9
Statistical analyses posted16
Lead sponsorNovartis Pharmaceuticals
Sponsor typeIndustry

2. Clinical Question

The primary statistical question was whether progression-free survival with everolimus plus exemestane could be evaluated as non-inferior to everolimus alone in advanced breast cancer. A secondary PFS comparison evaluated the combination against capecitabine alone.

Population

Patients enrolled in a phase 2 study of advanced breast cancer.

Intervention

Everolimus 10 mg plus exemestane 25 mg.

Comparators

Everolimus 10 mg alone and capecitabine 1250 mg/m2.

Primary question

Is progression-free survival with everolimus plus exemestane non-inferior to everolimus alone?

3. Trial Design

01
Randomize309 enrolled
02
3 armsParallel randomized treatment groups
03
FollowPFS and other registered outcomes
04
AssessTime-to-event and continuous outcomes
05
ComparePrimary and secondary analyses
ARM 1

Everolimus + Exemestane

  • Everolimus 10 mg
  • Exemestane 25 mg
  • Primary comparator for the PFS non-inferiority analysis
ARM 2

Everolimus Alone

  • Everolimus 10 mg
  • Primary comparison group for PFS
  • Also used in secondary OS, ECOG deterioration, and TSQM comparisons
ARM 3

Capecitabine

  • Capecitabine 1250 mg/m2
  • Secondary comparison group for PFS, OS, ECOG deterioration, and TSQM analyses
Open-label design: the trial was not masked. This is particularly relevant when interpreting subjective outcomes such as treatment satisfaction and quality of life, because awareness of treatment assignment can potentially affect patient-reported assessments. The time-to-event primary endpoint is less directly dependent on subjective scoring, although endpoint ascertainment and assessment procedures remain part of the broader trial context.

4. Trial Timeline

MilestoneDate
Trial start2013-02-26
Primary completion2018-07-02
StatusCompleted

The ClinicalTrials.gov record identifies the study as completed, with a start date of February 26, 2013 and primary completion on July 2, 2018. The statistical interpretation below focuses on the analyses and endpoint definitions reported in the ClinicalTrials.gov record rather than reconstructing additional analyses from external publications.

5. Primary Endpoint

EndpointRegistered definition / time frameAnalysis populationPrimary comparison
Progression Free Survival (PFS) - Everolimus Plus Exemestane Versus Everolimus Alone Date of randomization to the date of first documented tumor progression or death from any cause, whichever occurs first Full Analysis Set (FAS), which consisted of all randomized patients Everolimus 10 mg + Exemestane 25 mg vs Everolimus 10 mg

The registry definition further specifies that PFS is the time from randomization to the date of first radiologically documented progression or death due to any cause. If a patient did not have an event, PFS was censored at the date of the last adequate tumor assessment.

Analysis population: the registry-reported statistical analysis identifies the Full Analysis Set as all randomized patients and explicitly associates the analysis with the intention-to-treat principle. This is important because the treatment comparison remains anchored to randomization rather than being restricted to patients who completed treatment.

6. Primary Result: Progression-Free Survival

The registry contains a formal statistical analysis for the primary endpoint. The effect measure was a hazard ratio, with a two-sided 90% confidence interval. The analysis was identified as a non-inferiority analysis.

Everolimus + exemestane vs everolimus

HR 0.74

Two-sided 90% CI: 0.57–0.97

Hypothesis type: Non-inferiority

Primary endpointEstimateConfidence intervalAnalysis populationMethod information
Progression-free survival HR 0.74 90% two-sided CI: 0.57–0.97 Full Analysis Set; all randomized patients Hazard ratio obtained from a stratified Cox model, stratified by presence/absence of visceral metastasis as per IWRS
Clinical Biostats interpretation

What the estimate means: an HR of 0.74 means that the estimated instantaneous rate of progression or death was 0.74 times that of the everolimus-alone group under the fitted time-to-event model. Expressed as a relative model-based quantity, this corresponds to a 26% lower estimated hazard, calculated as 1 − 0.74.

What it does not mean: it does not mean that 26% of patients avoided progression, that every patient had a 26% reduction in risk, or that median PFS differed by 26%. A hazard ratio is a relative time-to-event measure, not an absolute risk difference.

Precision: the two-sided 90% confidence interval of 0.57–0.97 describes uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of effects that individual patients experienced.

The p-value: the registry analysis does not provide a p-value for the primary endpoint. A p-value should not be reconstructed from the reported confidence interval, particularly because this was explicitly framed as a non-inferiority analysis.

Non-inferiority is the key issue: the registry identifies the hypothesis type as non-inferiority, but the ClinicalTrials.gov record does not provide the prespecified non-inferiority margin. Therefore, the HR and confidence interval can be described precisely, but the non-inferiority criterion itself cannot be independently evaluated from the ClinicalTrials.gov record.

Model caution: the reported HR came from a stratified Cox model. Interpretation of a single Cox HR is most straightforward when the proportional-hazards representation is reasonably appropriate over the analyzed period. The ClinicalTrials.gov record does not provide diagnostics for that assumption.

7. What the Primary Hazard Ratio Tells Us

The primary HR of 0.74 is a relative comparison of the event rate over time. Because the endpoint combines two possible events—first documented tumor progression or death from any cause—the hazard ratio summarizes the randomized treatment comparison for this composite time-to-event endpoint rather than separately quantifying progression and death.

Conceptual hazard-ratio interpretation
HR = hazard in everolimus + exemestane ÷ hazard in everolimus alone

For the reported estimate, HR 0.74 indicates a lower estimated instantaneous event rate in the combination group relative to the everolimus-alone group within the fitted model.

The confidence interval adds information that the point estimate alone cannot provide. The reported interval extends from 0.57 to 0.97, so the registry-supported description should remain centered on the estimated HR and its uncertainty rather than treating 0.74 as a fixed underlying biological quantity.

8. Statistical Methodology

Time-to-event analysis

PFS, overall survival, ECOG performance deterioration, and definitive deterioration in global health status / quality of life were all represented as time-to-event outcomes in the analyses posted on ClinicalTrials.gov. Time-to-event methods are designed for situations in which participants can be followed for different lengths of time and some participants may not experience the event during observation.

Right censoring

The registered PFS definition explicitly states that participants without an event were censored at the date of the last adequate tumor assessment. Censoring allows information available before the last adequate assessment to contribute to the analysis without treating the participant as having experienced progression.

Stratified Cox model

The registry states that the reported hazard ratio was obtained from a stratified Cox model, with stratification by presence or absence of visceral metastasis as per IWRS. Stratification allows the baseline hazard to differ across the specified strata while estimating the treatment comparison within the Cox framework.

Cox model concept
h(t | X) = h0(t) × exp(βX)

The coefficient associated with treatment is translated into a hazard ratio through exp(β). In the BOLERO-6 analysis, the reported treatment effect was presented as a hazard ratio rather than as a difference in survival probabilities.

Intention-to-treat analysis

The registry-reported analysis explicitly identifies the Full Analysis Set as consisting of all randomized patients and notes the intention-to-treat analysis concept. This preserves the treatment assignment created by randomization and avoids redefining the comparison according to treatment exposure or completion.

Continuous outcome analysis

Treatment satisfaction was analyzed using a mean difference between groups for several TSQM domains. Unlike a hazard ratio, a mean difference operates on the scale of the questionnaire outcome and is interpreted as an average difference between the compared groups among participants meeting the analysis requirement.

Incomplete patient-reported measurements

For the TSQM analyses, the registry definition states that only participants who had both post-baseline assessments were included. This is an important distinction from the PFS analysis, which used all randomized patients in the Full Analysis Set. The ClinicalTrials.gov record does not specify an imputation method for missing TSQM assessments.

9. Non-Inferiority Design

Non-inferiority changes the statistical question. Instead of asking simply whether the treatment groups are different, a non-inferiority analysis asks whether the new treatment's effect is not unacceptably worse than the comparator according to a prespecified margin.

What the registry provides

The primary PFS analysis is explicitly classified as non-inferiority, and the estimated HR is 0.74 with a two-sided 90% CI of 0.57–0.97.

What is needed for the formal decision

The ClinicalTrials.gov record does not state the prespecified non-inferiority margin. Without that margin, the formal non-inferiority boundary cannot be independently reconstructed.

Why the margin matters: for a non-inferiority hazard-ratio analysis, the relevant question is whether the confidence interval remains within the prespecified non-inferiority boundary. The fact that a confidence interval does or does not include a particular value such as 1 is not, by itself, the complete non-inferiority decision rule.

10. Secondary Progression-Free Survival

The trial also evaluated PFS for everolimus plus exemestane versus capecitabine alone. This comparison used the same broad time-to-event definition and the Full Analysis Set.

Everolimus + exemestane vs capecitabine

HR 1.26

Two-sided 90% CI: 0.96–1.66

Hypothesis type: Non-inferiority

EndpointEstimate90% two-sided CIPopulationStratification
PFS: everolimus + exemestane vs capecitabine HR 1.26 0.96–1.66 Full Analysis Set; all randomized patients Presence/absence of visceral metastasis as per IWRS
Clinical Biostats interpretation

An HR of 1.26 means that the estimated instantaneous rate of progression or death was 1.26 times the corresponding rate in the capecitabine group under the reported Cox-model framework. Unlike the primary comparison, the point estimate is above 1.

The two-sided 90% CI of 0.96–1.66 communicates substantial uncertainty around the point estimate. The interval extends on both sides of 1, so the reported interval itself should not be converted into a simple statement that one treatment has a definitively higher or lower event rate.

The registry labels this analysis as non-inferiority, but the ClinicalTrials.gov record again do not state the non-inferiority margin. No p-value is reported. Therefore, the formal non-inferiority conclusion cannot be reconstructed solely from the ClinicalTrials.gov record.

11. Overall Survival

Overall survival was evaluated as a secondary time-to-event endpoint. The registered time frame was every 3 months following the end of treatment visit, assessed for approximately 54 months.

ComparisonHazard ratioTwo-sided 90% CIAnalysis population
Everolimus + exemestane vs everolimus 1.27 0.95–1.70 Full Analysis Set
Everolimus + exemestane vs capecitabine 1.33 0.99–1.79 Full Analysis Set

Both OS analyses used hazard ratios from a stratified Cox model, with stratification by presence or absence of visceral metastasis as per IWRS. The ClinicalTrials.gov record classifies both analyses as non-inferiority hypotheses.

How to read the OS estimates

For the everolimus-plus-exemestane versus everolimus comparison, the HR of 1.27 corresponds to a higher estimated instantaneous rate of death in the combination group under the fitted model. The 90% CI of 0.95–1.70 reflects uncertainty around that estimate.

For the comparison with capecitabine, the HR was 1.33 with a two-sided 90% CI of 0.99–1.79. This is again a relative model-based estimate, not a percentage of patients who died or a direct measure of absolute survival.

No p-values are posted on ClinicalTrials.gov for either OS analysis, and the non-inferiority margins are not included in the ClinicalTrials.gov record. Accordingly, the estimates should be reported without constructing an unsupported formal hypothesis-test conclusion.

12. ECOG Performance Deterioration

The registry included time to Eastern Cooperative Oncology Group (ECOG) Performance Deterioration as a secondary endpoint. The registered assessment schedule was baseline, every 6 weeks up to about 43 months.

ComparisonHazard ratioTwo-sided 90% CIHypothesis type
Everolimus + exemestane vs everolimus 1.09 0.72–1.66 Non-inferiority
Everolimus + exemestane vs capecitabine 1.18 0.78–1.77 Non-inferiority

Both comparisons were analyzed in the Full Analysis Set. The registry-reported analysis text again identifies a stratified Cox model with stratification by presence/absence of visceral metastasis as per IWRS.

Interpretation

An HR of 1.09 for the comparison with everolimus alone represents a relatively small point estimate above 1, while its 90% CI of 0.72–1.66 is substantially wider than the point estimate. The comparison with capecitabine has an HR of 1.18 and a 90% CI of 0.78–1.77.

These intervals illustrate why a point estimate should not be interpreted in isolation. The confidence intervals encompass a range of plausible relative effects under the model, and no p-values or non-inferiority margins are posted on ClinicalTrials.gov for these analyses.

13. Time to Definitive Deterioration in Global Health Status / Quality of Life

The registry evaluated time to 10% definitive deterioration in the Global Health Status / Quality of Life, measured from baseline every 6 weeks up to about 43 months.

ComparisonHazard ratioTwo-sided 90% CIHypothesis type
Everolimus + exemestane vs everolimus 0.64 0.46–0.88 Non-inferiority
Everolimus + exemestane vs capecitabine 1.33 0.93–1.91 Non-inferiority

The HR of 0.64 for the combination versus everolimus alone corresponds to a 36% lower estimated hazard of the specified definitive deterioration, using the reported hazard-ratio framework. The corresponding 90% CI was 0.46–0.88.

For the capecitabine comparison, the HR was 1.33 with a two-sided 90% CI of 0.93–1.91. Because these are time-to-event patient-reported outcomes, the interpretation concerns the timing of the prespecified deterioration event, not an average change in questionnaire score.

Clinical Biostats interpretation

The two quality-of-life time-to-event comparisons illustrate an important statistical distinction. The combination's HR of 0.64 versus everolimus alone describes the relative timing of a prespecified deterioration event. It does not mean that quality of life was 36% higher, nor does it quantify the average magnitude of a patient's change in quality of life.

For both comparisons, the confidence interval is more informative than the point estimate alone because it expresses uncertainty around the estimated hazard ratio. The ClinicalTrials.gov record does not provide p-values or the non-inferiority margin for these analyses.

14. Treatment Satisfaction: TSQM

The Treatment Satisfaction Questionnaire for Medication (TSQM) was assessed at Week 3 and Week 12. The registry analysis used mean change between those assessments. Only participants with both post-baseline assessments were included.

TSQM domainComparisonMean Difference (Net)Two-sided 90% CI
Side-effects Everolimus + exemestane vs everolimus 4.3 -3.3 to 11.9
Side-effects Everolimus + exemestane vs capecitabine -2.2 -9.9 to 5.5
Effectiveness Everolimus + exemestane vs everolimus -3.3 -10.4 to 3.8
Effectiveness Everolimus + exemestane vs capecitabine -3.3 -9.7 to 3.0
Convenience Everolimus + exemestane vs everolimus -1.6 -5.8 to 2.5
Convenience Everolimus + exemestane vs capecitabine -1.1 -5.5 to 3.3
Global Satisfaction Everolimus + exemestane vs everolimus -2.7 -8.3 to 2.9
Global Satisfaction Everolimus + exemestane vs capecitabine -3.3 -8.4 to 1.9

These are mean differences rather than hazard ratios. The sign therefore depends on the direction of the questionnaire scale and the definition of the reported net difference; it should not be interpreted using the same rules as an HR below or above 1.

Analysis population matters: although the registry identifies the Full Analysis Set as the overarching population, the TSQM analyses specifically included only participants who had both post-baseline assessments. That requirement means the estimand is not simply the average result among every randomized participant.

15. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized treatment arm using affected participants divided by the corresponding number at risk.

Treatment armSerious adverse eventsAffected / at risk
Everolimus 10 mg + Exemestane 25 mg Serious adverse events 37/104
Everolimus 10 mg Serious adverse events 30/103
Capecitabine 1250 mg/m2 Serious adverse events 30/102

Combination arm

37 of 104 participants were affected by serious adverse events.

Everolimus arm

30 of 103 participants were affected by serious adverse events.

Capecitabine arm

30 of 102 participants were affected by serious adverse events.

Interpretive caution

These are serious-adverse-event counts by arm; the ClinicalTrials.gov record does not provide a formal between-group statistical test for these safety data.

Safety should be considered separately from efficacy. The serious-adverse-event figures describe observed safety events among the reported at-risk populations; they do not provide a single numerical measure that can be combined with the PFS hazard ratio.

16. Statistical Methods Explained

Why was a hazard ratio used for PFS?

PFS is a time-to-event endpoint in which participants can experience progression or death at different times, while some remain event-free at their last adequate tumor assessment. A hazard ratio provides a relative comparison of the instantaneous event rate over time and can accommodate right-censored observations within the Cox framework.

What does HR 0.74 mean?

For the primary comparison, HR 0.74 means that the fitted model estimated the instantaneous rate of progression or death in the everolimus-plus-exemestane group at 0.74 times the rate in the everolimus-alone group. It does not mean that 74% of patients avoided progression or that individual patients experienced a 26% reduction.

Why was the Cox model stratified?

The registry states that the hazard ratio was obtained from a stratified Cox model, stratified by presence or absence of visceral metastasis as per IWRS. Stratification allows the baseline hazard to differ across those strata while estimating the treatment comparison within the model.

Why is non-inferiority judged against a margin?

Non-inferiority asks whether a treatment is not worse than its comparator by more than a prespecified clinically acceptable amount. Therefore, the relevant statistical boundary is the non-inferiority margin, not simply whether a confidence interval crosses 1. The registry-reported BOLERO-6 data identify the hypothesis as non-inferiority but do not provide the margin.

Why use the Full Analysis Set for efficacy?

The registry-reported analysis defines the Full Analysis Set as all randomized patients. An intention-to-treat approach preserves the original randomized comparison and therefore avoids changing the treatment groups according to subsequent treatment exposure or completion.

Why are the TSQM analyses different from PFS?

TSQM is a continuous patient-reported outcome, so the registry reports mean differences rather than hazard ratios. In addition, only participants with both post-baseline assessments were included. That differs from the PFS analysis, which used the Full Analysis Set of all randomized patients.

Why should a p-value not be inferred from the confidence interval?

The ClinicalTrials.gov record provides two-sided 90% confidence intervals but does not provide p-values. The primary endpoint is also explicitly framed as a non-inferiority hypothesis, so a conventional significance interpretation based on a two-sided null value of 1 would not necessarily reproduce the prespecified non-inferiority decision rule.

17. Understanding the Confidence Intervals

BOLERO-6 provides a useful example of why an effect estimate should always be accompanied by its confidence interval.

AnalysisEstimateTwo-sided 90% CI
Primary PFS: combination vs everolimusHR 0.740.57–0.97
PFS: combination vs capecitabineHR 1.260.96–1.66
OS: combination vs everolimusHR 1.270.95–1.70
OS: combination vs capecitabineHR 1.330.99–1.79
ECOG deterioration: combination vs everolimusHR 1.090.72–1.66
ECOG deterioration: combination vs capecitabineHR 1.180.78–1.77
Quality-of-life deterioration: combination vs everolimusHR 0.640.46–0.88
Quality-of-life deterioration: combination vs capecitabineHR 1.330.93–1.91

The intervals vary considerably in width. This reflects differences in the amount of information available for each endpoint and comparison. A narrow interval provides greater statistical precision around the estimated effect than a wider interval, although precision is not the same thing as clinical importance.

Confidence interval versus effect size: the confidence interval answers an uncertainty question, while the point estimate describes the estimated effect. Neither one by itself answers whether a result is clinically important. For non-inferiority, the prespecified margin is an additional essential component of interpretation.

18. Multiplicity and Multiple Comparisons

The ClinicalTrials.gov record contains one registered primary endpoint and multiple secondary analyses. The primary endpoint is PFS comparing everolimus plus exemestane with everolimus alone. Secondary analyses include PFS and OS comparisons with capecitabine, ECOG deterioration, quality-of-life deterioration, and several TSQM domains.

Analysis familyRole in the ClinicalTrials.gov recordEffect measure
PFS: combination vs everolimusPrimary endpointHazard ratio
PFS: combination vs capecitabineSecondaryHazard ratio
OSSecondaryHazard ratio
ECOG deteriorationSecondaryHazard ratio
Quality-of-life deteriorationSecondaryHazard ratio
TSQMSecondaryMean difference

The ClinicalTrials.gov record does not specify a multiplicity-adjustment procedure or an alpha-allocation hierarchy across the secondary endpoints. Consequently, the secondary estimates should be presented as reported analyses rather than treated as a sequence of independently confirmed hypotheses without additional information about the prespecified testing strategy.

19. Missing Data and Censoring

The primary PFS definition explicitly specifies censoring at the date of the last adequate tumor assessment when no event occurred. This is a central feature of time-to-event analysis: participants can contribute follow-up information even when the event has not been observed.

The TSQM analyses use a different rule: only participants who had both post-baseline assessments were included. This creates an analysis population defined partly by availability of questionnaire measurements.

PFS

Participants without an event were censored at the date of the last adequate tumor assessment.

TSQM

Only participants with both post-baseline assessments were included in the reported analysis.

The ClinicalTrials.gov record does not specify an imputation procedure for missing TSQM values. That distinction matters because complete-case inclusion and statistical imputation answer different missing-data problems and can rely on different assumptions.

20. Randomization and Stratification

Randomization is the structural foundation of the treatment comparison. Because participants were randomized into three parallel groups, the treatment assignments were determined before the outcome comparisons were made.

For the reported hazard-ratio analyses, the registry specifies stratification by presence or absence of visceral metastasis as per IWRS. The stratified Cox model therefore accounts for this stratification variable when estimating the treatment hazard ratio.

Why stratification matters
Treatment effect estimate + prespecified stratification structure → more faithful analysis of the randomized comparison

Stratification does not turn an observational comparison into a randomized one. Rather, it incorporates the specified stratification structure into the time-to-event model used for the randomized comparison.

21. What This Trial Does Not Establish From the Supplied Data

Several common trial-reporting details are not contained in the registry-reported BOLERO-6 statistical record. They should not be reconstructed from the point estimates alone.

Statistical discipline: these are not missing numbers to be estimated. They are design or analysis details that require source documentation. A complete statistical analysis should distinguish between what the registry actually reports and what might normally appear in a full protocol or statistical analysis plan.

22. Important Limitations and Interpretation Issues

23. Why This Trial Matters Statistically

BOLERO-6 is a useful teaching case because it combines randomized treatment allocation with a non-inferiority primary question, multiple time-to-event endpoints, stratified Cox modeling, censoring, and patient-reported continuous outcomes.

ConceptHow it appears in BOLERO-6
RandomizationThree-arm randomized parallel design
Intention-to-treat analysisFull Analysis Set consisting of all randomized patients
Time-to-event endpointsPFS, OS, ECOG deterioration, and quality-of-life deterioration
Hazard ratioPrimary and secondary time-to-event effect measure
Stratified Cox modelHR obtained with stratification by visceral-metastasis status as per IWRS
Non-inferiorityPrimary PFS and several secondary analyses identified as non-inferiority hypotheses
Confidence intervalsTwo-sided 90% intervals accompany the reported estimates
CensoringPFS censored at the date of the last adequate tumor assessment when no event occurred
Continuous outcomesTSQM reported using mean differences
Analysis populationsAll randomized patients for FAS efficacy analyses; TSQM additionally required both post-baseline assessments

The trial is particularly instructive for non-inferiority education. A hazard ratio below 1 can look favorable in isolation, but the formal question is not simply whether the estimate is below 1. The prespecified non-inferiority margin determines how much worsening would have been considered acceptable, and that margin is not contained in the ClinicalTrials.gov record.

24. Clinical Biostats Interpretation of the Overall Statistical Picture

Primary PFS comparison

The reported HR was 0.74 with a two-sided 90% CI of 0.57–0.97 for everolimus plus exemestane versus everolimus alone.

Comparator PFS analysis

Against capecitabine, the reported HR was 1.26 with a two-sided 90% CI of 0.96–1.66.

Overall survival

The reported HRs were 1.27 versus everolimus and 1.33 versus capecitabine, with corresponding two-sided 90% confidence intervals of 0.95–1.70 and 0.99–1.79.

Quality of life

Time to 10% definitive deterioration had an HR of 0.64 versus everolimus and 1.33 versus capecitabine.

These results should not be compressed into a single overall statistic. The primary PFS comparison, secondary PFS comparison, OS analyses, quality-of-life deterioration analyses, and TSQM measurements represent different estimands and different statistical questions.

The strongest statistical reading is therefore endpoint-specific: identify the population, comparison, effect measure, confidence interval, hypothesis type, and analysis framework for each result. For the non-inferiority analyses in particular, the missing margin prevents the formal decision rule from being reconstructed from the ClinicalTrials.gov record alone.

25. Related Tutorials

Learn more about the methods used in this trial:

26. Related Calculators

27. Sources

Continue with the statistical methods

Explore the broader methods behind randomized trials, time-to-event analysis, hazard ratios, confidence intervals, and non-inferiority designs.

28. Record Summary

BOLERO-6 was a randomized, parallel phase 2 study of everolimus plus exemestane, everolimus alone, and capecitabine in advanced breast cancer. Its primary endpoint was progression-free survival comparing the combination with everolimus alone, analyzed in the Full Analysis Set using a hazard ratio from a stratified Cox model. The reported primary estimate was HR 0.74 with a two-sided 90% CI of 0.57–0.97.

The trial also reported secondary PFS, OS, ECOG deterioration, quality-of-life deterioration, and TSQM analyses. The time-to-event endpoints were summarized with hazard ratios and two-sided 90% confidence intervals, while TSQM used mean differences. Serious adverse events were reported as 37/104, 30/103, and 30/102 across the three treatment arms.

The central statistical lesson is that the numerical estimate is only one component of the analysis. For BOLERO-6, interpretation also requires attention to the Full Analysis Set, intention-to-treat principle, censoring, stratified Cox modeling, the distinction between hazard ratios and mean differences, and—most importantly for the primary question—the non-inferiority margin. Because the ClinicalTrials.gov record does not provide that margin or p-values, those elements should not be inferred from the reported estimates.

Clinical Biostats methodology: A trial-results page should distinguish reported evidence from statistical interpretation. The goal is to explain what each estimate means, what it does not mean, how uncertainty affects interpretation, and which design details are necessary before making a formal statistical conclusion.