← Clinical Trials
Breast Cancer Phase 3 Time-to-Event NCT00863655

BOLERO-2: Complete Statistical Analysis of Everolimus + Exemestane in Breast Cancer

An independent statistical analysis of the randomized phase 3 BOLERO-2 trial evaluating everolimus in combination with exemestane versus placebo plus exemestane in postmenopausal women with estrogen receptor positive locally advanced or metastatic breast cancer refractory to letrozole or anastrozole.

Phase 3  ·  Randomized  ·  Parallel design  ·  Quadruple masked
Scope of this record

This page separates reported trial results from statistical interpretation. Trial-specific numerical results and endpoint definitions are restricted to the ClinicalTrials.gov record for BOLERO-2. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

BOLERO-2 was a randomized, phase 3, parallel-design trial of everolimus plus exemestane versus placebo plus exemestane in postmenopausal women with estrogen receptor positive locally advanced or metastatic breast cancer who were refractory to letrozole or anastrozole. The registry reports 724 enrolled participants and a primary endpoint based on progression-free survival.

724
Enrolled
ClinicalTrials.gov record
2
Arms
Parallel design
0.43
PFS HR
95% CI 0.35–0.54
<0.0001
P-value
Log-rank test
FeatureBOLERO-2
Trial nameBOLERO-2
NCT identifierNCT00863655
PhasePhase 3
ConditionBreast Cancer
PopulationPostmenopausal women with estrogen receptor positive locally advanced or metastatic breast cancer who are refractory to letrozole or anastrozole
AllocationRandomized
Design modelParallel
MaskingQuadruple
Primary purposeTreatment
Enrollment724
Primary endpoint typeTime-to-event
Primary endpoint analyses posted1
Results postedYes
Lead sponsorNovartis Pharmaceuticals
Sponsor typeIndustry
Registry status: The ClinicalTrials.gov record identifies the study as completed, with a start date of June 3, 2009 and a primary completion date of December 4, 2014.

2. Clinical Question

The central question was whether adding everolimus to exemestane produced a different progression-free survival experience than exemestane with placebo in postmenopausal women with estrogen receptor positive locally advanced or metastatic breast cancer who were refractory to letrozole or anastrozole.

Population

Postmenopausal women with estrogen receptor positive locally advanced or metastatic breast cancer who were refractory to letrozole or anastrozole.

Intervention

Everolimus in combination with exemestane.

Comparator

Placebo in combination with exemestane.

Primary question

Does everolimus plus exemestane improve the time from randomization to first documented tumor progression or death compared with placebo plus exemestane?

3. Trial Design

The registry characterizes BOLERO-2 as a randomized, parallel-design, quadruple-masked phase 3 trial with treatment as its primary purpose. The intervention structure contains two treatment groups: everolimus plus exemestane and placebo plus exemestane.

01
Enroll724 participants
02
RandomizeTwo parallel arms
03
TreatEverolimus + exemestane or placebo + exemestane
04
AssessTumor progression and death
05
AnalyzeLog-rank test and hazard ratio
ARM A

Everolimus + Exemestane

  • Everolimus
  • Exemestane
  • Randomized treatment assignment
  • Primary comparison based on progression-free survival
ARM B

Placebo + Exemestane

  • Everolimus placebo
  • Exemestane
  • Randomized treatment assignment
  • Primary comparison based on progression-free survival
Why randomization matters: Randomization establishes the treatment assignment mechanism before outcomes are observed. For the primary efficacy comparison, the registry analysis states that all randomized patients were included in the Full Analysis Set. This preserves the randomized comparison rather than defining the efficacy population according to treatment actually received.

4. Endpoints

EndpointRegistry definition / time frameType
Progression-free Survival (PFS) Based on Local Radiology Review of Tumor Assessments date of randomization to the date of first documented tumor progression or death from any cause, whichever occurs first Time-to-event

The registry definition further describes progression-free survival as the time from the date of randomization to the date of first documented radiological progression or death due to any cause. Disease progression was based on tumor assessment by the local radiologist or investigator using RECIST 1.0 criteria.

Endpoint discipline: The ClinicalTrials.gov record identifies one registered primary endpoint and one posted formal statistical analysis. This page therefore does not add overall survival, response rate, median PFS, subgroup estimates, or other efficacy outcomes that are not contained in the ClinicalTrials.gov record.

5. Statistical Methodology

Primary analysis: log-rank test

The registry-reported method for the primary progression-free survival analysis was the log-rank test. The log-rank test is designed for comparing survival or time-to-event distributions between randomized groups while accounting for the fact that not every participant necessarily experiences the event during observation.

In this trial, the event is defined by the registered PFS endpoint: first documented tumor progression or death from any cause, whichever occurs first. The analysis therefore compares the timing of these events between the everolimus-plus-exemestane and placebo-plus-exemestane groups rather than simply comparing the proportion of participants who progressed.

Hazard ratio as the effect measure

The reported effect measure was the hazard ratio. The estimate was 0.43, with a two-sided 95% confidence interval of 0.35 to 0.54.

Primary effect estimate
HR = 0.43   ·   95% CI: 0.35–0.54   ·   P < 0.0001

The hazard ratio summarizes the relative event rate between the two randomized groups within the time-to-event analysis. An HR below 1 indicates a lower estimated hazard in the everolimus-plus-exemestane group relative to the placebo-plus-exemestane group.

Full Analysis Set

The registry-reported statistical analysis states that all randomized patients were included in the Full Analysis Set. This is important because the treatment comparison remains anchored to randomized assignment rather than being restricted to patients who completed a particular amount of treatment.

Superiority framework

The hypothesis type recorded for the primary analysis was superiority. The reported result therefore addresses whether the two randomized treatment groups differ in the direction specified by the superiority comparison. It is not a non-inferiority analysis, so there is no non-inferiority margin to interpret.

What was reported

A log-rank analysis of PFS with a hazard ratio of 0.43, a two-sided 95% confidence interval of 0.35 to 0.54, and P < 0.0001.

What was not reported

The ClinicalTrials.gov record does not provide a median PFS, event counts, Kaplan-Meier percentages, baseline table, subgroup analyses, interim-analysis details, or a formal multiplicity plan.

6. Primary Result: Progression-Free Survival

The primary endpoint was progression-free survival based on local radiology review of tumor assessments. The registry reports a formal statistical analysis comparing everolimus plus exemestane with placebo plus exemestane.

Hazard ratio for progression or death

0.43

95% CI: 0.35–0.54   ·   P < 0.0001   ·   Two-sided

Analysis: Log-rank test   ·   Hypothesis: Superiority

Primary endpointEverolimus + Exemestane vs Placebo + Exemestane
EndpointProgression-free Survival (PFS) Based on Local Radiology Review of Tumor Assessments
Time frameDate of randomization to first documented tumor progression or death from any cause, whichever occurs first
Analysis populationAll randomized patients in the Full Analysis Set
Statistical methodLog-rank test
Effect measureHazard ratio
Estimate0.43
95% confidence interval0.35–0.54
P-value<0.0001
Hypothesis typeSuperiority
Clinical Biostats interpretation

The reported hazard ratio of 0.43 means that the estimated instantaneous rate of the PFS event was approximately 43% of the corresponding rate in the comparator group under the time-to-event comparison. Expressed as a simple relative interpretation, this corresponds to an estimated 57% lower hazard of progression or death for everolimus plus exemestane relative to placebo plus exemestane.

The hazard ratio does not mean that 57% of patients avoided progression, that 57% of patients were cured, or that every patient experienced a 57% reduction in individual risk. It is a relative time-to-event effect measure, not an absolute probability.

The 95% confidence interval of 0.35 to 0.54 describes the statistical uncertainty around the estimated hazard ratio under the analysis framework. It is not a range containing 95% of individual patient outcomes. The interval is relatively narrow around the point estimate and remains below 1 throughout its reported range.

The P-value of <0.0001 addresses the evidence against the null hypothesis under the specified statistical test. It does not measure the size of the treatment effect, the probability that the treatment works, or the probability that the reported hazard ratio is exactly correct.

Because this is a time-to-event analysis, interpretation also depends on the censoring structure and the assumptions underlying the hazard-ratio framework. The ClinicalTrials.gov record does not report a formal assessment of proportional hazards, so the reported HR should not be treated as proof that the relative hazard was constant throughout follow-up.

Reading the confidence interval

The point estimate and confidence interval should be considered together. An HR of 0.43 provides the central estimate, while the 95% CI of 0.35 to 0.54 shows the range of values compatible with the statistical uncertainty represented by the reported analysis. The interval remains below 1, which is consistent with the direction of the reported superiority result.

Educational note: a Kaplan-Meier curve cannot be reconstructed responsibly from the registry-reported summary statistics alone. The trial data provide the endpoint definition and hazard-ratio analysis, but they do not provide the event and censoring times needed to construct a valid patient-level Kaplan-Meier estimate.

7. Statistical Methods Explained

Why was a log-rank test used?

The primary endpoint is a time-to-event outcome. A simple comparison of event proportions would discard information about when progression or death occurred and would not naturally accommodate participants whose follow-up ends before an event. The log-rank test instead compares the observed timing of events between treatment groups while incorporating the risk sets over follow-up.

What does a hazard ratio of 0.43 mean?

A hazard ratio of 0.43 indicates that the estimated instantaneous event rate in the everolimus-plus-exemestane group was 0.43 times that of the placebo-plus-exemestane group in the reported analysis. The corresponding simple relative statement is that the estimated hazard was 57% lower. This is not equivalent to saying that 57% more patients remained progression-free at every time point.

Why is the confidence interval important?

A point estimate alone hides statistical uncertainty. The reported 95% CI of 0.35 to 0.54 places the estimated hazard ratio in the context of the uncertainty generated by the study data and analysis. A confidence interval also helps distinguish a precise estimate from one that is highly variable, although the ClinicalTrials.gov record does not provide enough information to characterize precision using any measure beyond the reported interval itself.

Why doesn't the P-value measure treatment effect size?

The P-value is a measure of compatibility between the observed data and a null hypothesis under the statistical model and test being used. It is affected by both effect magnitude and information available in the study. The hazard ratio and its confidence interval are the appropriate reported quantities for describing the estimated relative effect and its uncertainty; the P-value answers a different statistical question.

Why does the Full Analysis Set matter?

The registry states that all randomized patients were included in the Full Analysis Set. An analysis based on randomized assignment preserves the comparison created by randomization. Excluding participants after randomization based on later treatment behavior can introduce selection mechanisms that were not part of the original allocation process.

How should progression and death be handled in this endpoint?

The registered definition makes progression and death competing ways of reaching the PFS event: the endpoint occurs at the first documented tumor progression or death from any cause, whichever occurs first. Consequently, PFS is not simply a tumor-progression endpoint. A participant who dies before documented progression reaches the PFS event through death.

What does the superiority hypothesis imply?

The analysis was classified as a superiority analysis. That means the statistical question concerns evidence of a treatment difference rather than whether the experimental treatment is merely not unacceptably worse than control. A non-inferiority margin is therefore not part of the registry-reported analysis and should not be introduced into interpretation.

8. Intention-to-Treat and Time-to-Event Analysis

The statistical analysis identifies intention-to-treat as a concept in the analysis text and states that all randomized patients were included in the Full Analysis Set. This is particularly important in a randomized clinical trial because treatment assignment is the mechanism that creates the primary causal comparison.

Randomization
Participants were assigned through a randomized allocation process, creating the basis for comparison of the two treatment groups.
Analysis population
All randomized patients were included in the Full Analysis Set for the reported primary analysis.
Event definition
First documented tumor progression or death from any cause, whichever occurred first.
Effect measure
Hazard ratio, with a two-sided 95% confidence interval.

For a time-to-event endpoint, participants contribute information for as long as their event status is observable. Participants who have not experienced the event by the end of their available observation can be censored rather than treated as though they experienced the event at the end of follow-up. The ClinicalTrials.gov record does not specify the detailed censoring rules, so no additional censoring assumptions are attributed to BOLERO-2 here.

9. Understanding the Hazard Ratio More Precisely

Conceptual relationship
HR = hEverolimus + Exemestane(t) / hPlacebo + Exemestane(t)

Here, h(t) represents the instantaneous event rate at time t. The reported HR of 0.43 summarizes the relative event-rate comparison produced by the time-to-event analysis.

The most important distinction is between a hazard and a probability. Hazard describes the instantaneous event rate among participants who remain event-free immediately before a given time. A hazard ratio therefore cannot be read as though it were a ratio of cumulative probabilities.

For example, an HR of 0.43 does not imply that the probability of progression or death was exactly 43% of the control probability at every time point. The relationship between hazards and survival probabilities is cumulative and depends on how the hazard evolves over time.

What HR = 0.43 supports

A lower estimated event hazard for the everolimus-plus-exemestane group in the reported time-to-event comparison.

What HR = 0.43 does not establish

A 57% absolute reduction in progression, a 57% increase in cure, or identical treatment effects at every point in time.

10. What the P-Value Adds

The reported P-value was <0.0001 for the log-rank comparison. The P-value should be interpreted in conjunction with the effect estimate and confidence interval rather than as a standalone measure of clinical importance.

QuantityWhat it tells the readerWhat it does not tell the reader
Hazard ratio 0.43Direction and relative magnitude of the reported time-to-event effectAbsolute probability of progression or death
95% CI 0.35–0.54Statistical uncertainty around the estimated hazard ratioRange of individual patient treatment effects
P < 0.0001Evidence against the null hypothesis under the reported log-rank analysisProbability that the treatment works or magnitude of benefit

This distinction is especially important when communicating clinical trial results. A very small P-value can coexist with an effect that is clinically modest in another setting, while a clinically important effect estimate can have a wide confidence interval if the available information is limited. The three quantities answer different questions.

11. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm. These are presented as affected participants divided by participants at risk, exactly as reported in the ClinicalTrials.gov record.

Safety measureAffected / at risk
Everolimus + Exemestane158/482
Placebo + Exemestane37/238

These safety data should be interpreted separately from the primary PFS analysis. The efficacy analysis concerns time to progression or death and uses randomized treatment assignment in the Full Analysis Set. Serious adverse events describe a safety outcome and are reported here as affected participants relative to the corresponding number at risk.

Do not combine the endpoints: a treatment comparison for progression-free survival cannot be converted into a global benefit-risk conclusion without a prespecified framework for considering efficacy, adverse events, treatment exposure, follow-up, and other clinically relevant outcomes.

12. Statistical Interpretation of the Primary Result

Effect size

The reported HR of 0.43 indicates a substantially lower estimated hazard of the PFS event in the everolimus-plus-exemestane group relative to placebo plus exemestane. The simple derived interpretation is a 57% lower estimated hazard, not a 57% absolute reduction in the probability of progression or death.

Precision

The two-sided 95% CI of 0.35 to 0.54 provides the reported uncertainty interval around the HR. Because the entire interval is below 1, the range of values represented by the interval remains on the same side of the no-hazard-difference value.

Statistical evidence

The reported P < 0.0001 indicates strong statistical evidence against the null hypothesis under the specified log-rank test. It should not be interpreted as a percentage probability that the observed effect is real or as a direct measure of clinical magnitude.

Endpoint meaning

PFS combines first documented tumor progression and death from any cause into a single time-to-event endpoint. Therefore, the HR describes the relative hazard of reaching either of those endpoint events first.

Analysis population

The registry analysis explicitly states that all randomized patients were included in the Full Analysis Set. This makes the reported efficacy estimate a randomized-group comparison rather than a comparison based only on patients who remained on therapy.

13. What Cannot Be Concluded From the Supplied Result

A statistically strong primary result still has boundaries. The ClinicalTrials.gov record does not provide enough information to answer several questions that readers may naturally ask about a survival analysis.

Why this matters: statistical interpretation should stop where the evidence reported by the registry stops. Adding familiar features of other oncology trials would make the page appear more complete but would not make the analysis more accurate.

14. Design Topics Not Reported in the Supplied Data

Design topicWhat can be stated from the registry-reported BOLERO-2 data
Non-inferiority marginNot applicable to the registry-reported primary analysis; the hypothesis type is superiority.
CrossoverThe ClinicalTrials.gov record does not report crossover.
Factorial designThe design model is parallel, with two arms; no factorial design is reported.
MultiplicityNo multiplicity strategy is reported in the ClinicalTrials.gov record.
Interim analysisNo interim-analysis plan or boundary is reported in the ClinicalTrials.gov record.
Missing-data / imputation rulesNo detailed missing-data or imputation strategy is reported in the ClinicalTrials.gov record.
StratificationNo stratification factors are reported in the ClinicalTrials.gov record.
Bayesian methodsNo Bayesian method is reported; the primary analysis uses a log-rank test.

This distinction is important because the absence of a reported method in the registry extract is not evidence that a particular method was definitely not used elsewhere in the trial's full statistical documentation. It means only that the method is not available in the data used for this page.

15. Limitations

16. Why This Trial Matters Statistically

BOLERO-2 is a useful teaching example because its primary analysis illustrates the core workflow for a randomized time-to-event endpoint: define an event precisely, preserve the randomized comparison, account for incomplete event observation, compare survival experience between groups, and summarize the relative effect with a hazard ratio and confidence interval.

ConceptHow it appears in BOLERO-2
RandomizationThe trial is randomized and uses two parallel treatment groups.
MaskingThe registry classifies the study as quadruple masked.
Time-to-event endpointPFS is defined from randomization to first documented progression or death.
Full Analysis SetAll randomized patients were included in the reported Full Analysis Set.
Log-rank testThe reported primary statistical comparison uses a log-rank test.
Hazard ratioThe primary effect measure is HR 0.43.
Confidence intervalThe HR is accompanied by a two-sided 95% CI of 0.35–0.54.
P-valueThe reported log-rank P-value is <0.0001.
Superiority testingThe registry classifies the primary hypothesis as superiority.
Safety analysisSerious adverse events are reported by treatment arm as affected/at-risk counts.

The deeper statistical lesson

The central lesson is that a time-to-event result is more than a single number. The HR of 0.43 is meaningful because it is attached to a defined endpoint, a randomized comparison, an analysis population, a statistical test, and a confidence interval. Removing any of those pieces makes the result easier to misinterpret.

The endpoint definition is particularly important. PFS is not simply "time until progression." Death from any cause is also an event, and the endpoint occurs at whichever event happens first. This means the statistical estimand incorporates both radiologically documented progression and mortality into the same time-to-event outcome.

Likewise, the Full Analysis Set matters because the analysis is anchored to randomization. The hazard ratio is therefore not merely a comparison of patients who happened to remain on their assigned treatment. It is a treatment-group comparison established by the randomized design and analyzed according to the registry definition.

17. From Event Times to the Log-Rank Test

To understand the primary analysis conceptually, imagine follow-up progressing from the earliest observed event to the latest. At each event time, the analysis considers which randomized patients remain at risk and compares the observed number of events with what would be expected if the treatment groups had the same underlying survival experience.

Conceptual log-rank comparison
Observed events  −  Expected events  →  test statistic

The log-rank framework accumulates information across event times rather than treating the final event status of every participant as a simple binary outcome.

This is one reason survival analysis is useful for clinical trials. A participant who remains progression-free for a long period contributes information about the treatment comparison even if the participant has not experienced the endpoint by the end of observation. Censoring allows that partial follow-up information to be represented without pretending that an unobserved future event did or did not occur.

The ClinicalTrials.gov record identifies the log-rank test as the reported primary method but do not provide the underlying event-by-event risk sets. Consequently, this page explains the method conceptually rather than attempting to reconstruct the test statistic.

18. Kaplan-Meier Estimation and the Primary Endpoint

Kaplan-Meier estimation is the standard descriptive framework for displaying a time-to-event endpoint such as PFS. Although the registry-reported statistical analysis field specifically identifies the log-rank test and hazard ratio, Kaplan-Meier estimation is useful for understanding what the underlying survival distributions represent.

Kaplan-Meier concept
S(t) = ∏ti ≤ t (1 − di/ni)

Here, di is the number of events at time ti and ni is the number at risk immediately before that event time.

The Kaplan-Meier curve estimates the probability of remaining event-free beyond each point in time. For BOLERO-2, that event-free state corresponds to remaining free of the registered PFS event: first documented tumor progression or death from any cause.

The distinction between a curve and an HR is useful. The Kaplan-Meier curve is a descriptive representation of the estimated event-free experience over time. The hazard ratio is a relative effect measure summarizing the treatment comparison through a time-to-event model or analysis framework. They complement one another rather than being interchangeable.

19. Confidence Intervals in Context

The reported confidence interval is 0.35 to 0.54 around the HR estimate of 0.43. A confidence interval should be read as an uncertainty statement about the estimated parameter under the specified statistical framework, not as a prediction interval for future individual patients.

Point estimate

HR 0.43 is the central estimate from the reported primary analysis.

Lower limit

0.35 is the lower endpoint of the two-sided 95% confidence interval.

Upper limit

0.54 is the upper endpoint of the two-sided 95% confidence interval.

No-effect reference

For a hazard ratio, 1 represents equal estimated hazard between groups.

The interval is especially informative because it communicates both direction and precision. Every value in the reported interval is below 1, consistent with the direction of the superiority result. At the same time, the interval is not a statement that the true effect for every individual patient lies somewhere between 0.35 and 0.54.

20. Trial Timeline

2009-06-03

Trial start

The registry profile identifies June 3, 2009 as the study start date.

Phase 3

Randomized treatment comparison

The study was conducted as a randomized, parallel-design phase 3 trial with quadruple masking and treatment as its primary purpose.

2014-12-04

Primary completion

The registry profile identifies December 4, 2014 as the primary completion date.

Completed

Registry status

the ClinicalTrials.gov record identifies the study status as completed and reports that results were posted.

21. Primary Result in One Statistical Summary

BOLERO-2 primary PFS analysis

HR 0.43

95% CI 0.35–0.54   ·   P < 0.0001

Two-sided log-rank test   ·   Superiority hypothesis   ·   Full Analysis Set

In statistical terms, the reported primary analysis provides evidence of a difference in progression-free survival between the two randomized treatment groups, with the estimated hazard favoring everolimus plus exemestane. The result should be interpreted as a relative time-to-event effect and not as an absolute probability or median survival difference.

Bottom line for the statistical reader: the key result is not simply "P < 0.0001." The complete statistical statement is that a randomized comparison of PFS, defined as time from randomization to first documented tumor progression or death from any cause, produced an HR of 0.43 with a two-sided 95% CI of 0.35–0.54 using a log-rank test, with all randomized patients included in the Full Analysis Set.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue with the underlying statistical methods

Use the related tutorials and calculators to explore the survival-analysis concepts represented in the BOLERO-2 primary endpoint.

25. Record Summary

BOLERO-2 is a randomized phase 3, parallel-design trial with quadruple masking that evaluated everolimus plus exemestane versus placebo plus exemestane in postmenopausal women with estrogen receptor positive locally advanced or metastatic breast cancer who were refractory to letrozole or anastrozole. The ClinicalTrials.gov record identifies progression-free survival based on local radiology review of tumor assessments as the primary endpoint.

The formal primary analysis used a log-rank test, with all randomized patients included in the Full Analysis Set. The reported effect measure was a hazard ratio of 0.43, with a two-sided 95% confidence interval of 0.35 to 0.54 and a P-value of <0.0001 under a superiority hypothesis.

The statistical interpretation is strongest when these quantities are kept together: the endpoint defines what event is being measured; randomization defines the treatment comparison; the Full Analysis Set preserves the randomized population; the log-rank test evaluates the time-to-event distributions; the hazard ratio describes the relative event hazard; the confidence interval describes uncertainty around that estimate; and the P-value addresses evidence against the null hypothesis. The ClinicalTrials.gov record does not provide median PFS, absolute PFS probabilities, subgroup estimates, event counts, or detailed interim, multiplicity, censoring, or imputation procedures, so those features are not inferred here.

Clinical Biostats methodology: A rigorous trial-results page should distinguish what the registry actually reports from what statistical principles allow us to explain. For BOLERO-2, the core teaching point is the interpretation of a randomized time-to-event analysis: a defined PFS endpoint, a log-rank comparison, a hazard ratio of 0.43, a two-sided 95% CI of 0.35–0.54, and P < 0.0001.