This page separates reported trial results from statistical interpretation. Trial facts and numerical results on this page are restricted to the ClinicalTrials.gov record for BOLERO-3. The registry record is the official source for the trial information.
1. Trial at a Glance
BOLERO-3 was a randomized, double-blind, parallel phase 3 trial evaluating everolimus added to trastuzumab and vinorelbine versus placebo added to trastuzumab and vinorelbine in women with HER2/Neu-positive locally advanced or metastatic breast cancer. The registry reports a time-to-event primary endpoint and a formal hazard-ratio analysis using a log-rank test.
| Feature | BOLERO-3 |
|---|---|
| Trial name | BOLERO-3 |
| NCT identifier | NCT01007942 |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Oncology |
| Population | HER2/Neu-positive women with locally advanced or metastatic breast cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 569 |
| Lead sponsor | Novartis Pharmaceuticals |
| Sponsor type | Industry |
| Start | October 2009 |
| Primary completion | June 2015 |
2. Clinical Question
The central statistical question was whether adding everolimus to trastuzumab and vinorelbine changed the time to progression or death compared with placebo plus trastuzumab and vinorelbine in women with HER2/Neu-positive locally advanced or metastatic breast cancer.
Population
Women with HER2/Neu Over-expressing Locally Advanced Breast Cancer or Metastatic Breast Cancer.
Intervention
Everolimus in combination with trastuzumab and vinorelbine.
Comparator
Placebo in combination with trastuzumab and vinorelbine.
Primary question
Does everolimus plus trastuzumab and vinorelbine improve investigator-assessed progression-free survival relative to placebo plus trastuzumab and vinorelbine?
3. Trial Design
Everolimus combination
- Everolimus
- Trastuzumab
- Vinorelbine
Placebo combination
- Placebo
- Trastuzumab
- Vinorelbine
The registry identifies the allocation as randomized, the design model as parallel, and the masking as double. These features are important statistically because randomization establishes the framework for a treatment-group comparison, while masking can reduce the influence of knowledge of treatment assignment on trial conduct and assessment.
Patients were assigned to one of the two trial arms through randomization.
The two randomized groups were followed as separate treatment arms.
The registry identifies the study as double-masked.
The trial was designed to evaluate treatment effects.
4. Primary Endpoint
| Endpoint | Registry definition and time frame | Type |
|---|---|---|
| Progressive-free Survival (PFS) Per Investigator Assessment | Every 6 weeks until disease progression or death which ever occurred first up to about 41 months. | Time-to-event |
Registry definition of PFS
The registry defines PFS as the time from the date of randomization to the date of first radiologically documented tumor progression or death from any cause, whichever occurs first. The PFS primary analysis was performed when 415 events were reached. The registry definition describes progression using Response Evaluation Criteria In Solid Tumors Criteria (RECIST v1.0), including a 20% increase in the sum of the longest diameter of target lesions.
5. Statistical Methodology
The registry reports a log-rank test as the primary statistical method for the PFS comparison and a hazard ratio as the effect measure. The formal analysis used the Full Analysis Set (FAS), which consisted of all randomized patients. The registry also identifies the analysis as an intention-to-treat analysis.
| Statistical component | Reported approach |
|---|---|
| Endpoint | Progressive-free Survival (PFS) Per Investigator Assessment |
| Endpoint type | Time-to-event |
| Analysis population | Full Analysis Set; all randomized patients |
| Primary comparison | Everolimus + Vinorelbine + Trastuzumab vs Placebo + Vinorelbine + Trastuzumab |
| Test | Log-rank |
| Effect measure | Hazard ratio |
| Hypothesis type | Superiority |
| Primary analysis event target | 415 events |
| Assessment schedule | Every 6 weeks until disease progression or death which ever occurred first up to about 41 months |
Intention-to-treat analysis
The registry states that the Full Analysis Set consisted of all randomized patients and identifies intention-to-treat analysis as a concept in the primary analysis. This is statistically important because the treatment comparison remains anchored to the randomized assignment rather than being restricted to patients who completed treatment or adhered perfectly to the assigned regimen.
Log-rank testing
The log-rank test is designed for comparing time-to-event distributions between groups. Rather than comparing only one time point, it uses the observed ordering of events and the numbers at risk over follow-up.
For BOLERO-3, the registered hypothesis type was superiority. The reported PFS comparison used a log-rank test, with the hazard ratio providing a separate measure of the relative event hazard.
Hazard ratio
The hazard ratio summarizes the relative instantaneous event rate between treatment groups under a time-to-event model. For this analysis, an HR below 1 corresponds to a lower estimated hazard of progression or death in the everolimus-containing group relative to the placebo-containing group.
The hazard ratio is a relative time-to-event measure. It is not a median PFS difference, an absolute risk reduction, or the percentage of patients who benefit.
6. Results: Primary Progression-Free Survival Analysis
The registry reports a formal statistical analysis for the primary endpoint, Progressive-free Survival (PFS) Per Investigator Assessment. The comparison was based on the Full Analysis Set, consisting of all randomized patients.
Hazard ratio for progression or death
95% CI: 0.65–0.95 · P = 0.0067
Two-sided 95% confidence interval · Superiority hypothesis
| Primary endpoint | Everolimus + vinorelbine + trastuzumab | Placebo + vinorelbine + trastuzumab | Effect estimate |
|---|---|---|---|
| Progressive-free Survival (PFS) Per Investigator Assessment | Randomized patients in the Full Analysis Set | Randomized patients in the Full Analysis Set | HR 0.78 95% CI 0.65–0.95 P = 0.0067 |
What the estimate means: An HR of 0.78 means that, within the time-to-event comparison represented by the reported hazard ratio, the estimated instantaneous rate of progression or death in the everolimus-containing group was 0.78 times that of the placebo-containing group. Expressed as a simple relative interpretation, this corresponds to an estimated 22% lower hazard of progression or death.
What it does not mean: The HR of 0.78 does not mean that 22% of patients avoided progression, that every patient experienced a 22% reduction in risk, or that median PFS was 22% longer. It is a relative time-to-event measure rather than an absolute patient-level benefit measure.
What the confidence interval says: The two-sided 95% CI of 0.65–0.95 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. It does not describe the range of effects that individual patients experienced.
Why the P-value is not an effect size: The P-value of 0.0067 addresses the statistical evidence against the null hypothesis under the specified testing framework. It does not quantify how large or clinically important the treatment effect is. The HR and its confidence interval provide the effect-size information.
Important time-to-event cautions: A hazard ratio is a model-based relative measure and should not automatically be translated into an absolute difference in PFS. Interpretation also depends on censoring and on the behavior of the event hazards over time. The ClinicalTrials.gov record does not provide enough information to assess the proportional-hazards assumption directly.
Analysis population: The reported comparison used the Full Analysis Set consisting of all randomized patients. That preserves the randomized treatment assignment as the basis for the primary efficacy comparison.
7. Interpreting the Confidence Interval
The reported 95% confidence interval for the PFS hazard ratio extends from 0.65 to 0.95. Both endpoints are below 1, the value representing equal instantaneous event hazards between the randomized groups.
Point estimate
The HR of 0.78 is the single reported estimate of the relative event hazard for PFS.
Lower limit
The lower confidence-limit value is 0.65, representing a stronger relative reduction in the estimated event hazard than the point estimate.
Upper limit
The upper confidence-limit value is 0.95, relatively close to 1. This illustrates why the interval is important in addition to the point estimate.
Null value
For a hazard ratio, 1 is the conventional null value. The reported interval does not include 1.
A confidence interval should not be interpreted as a probability statement that the true hazard ratio has a 95% probability of falling between 0.65 and 0.95. Rather, it is an interval generated by a statistical procedure with stated long-run coverage properties under the applicable assumptions.
8. What the P-Value Adds
The reported two-sided P-value is 0.0067. In the context of the registered superiority hypothesis and the reported log-rank analysis, this is evidence against the null hypothesis of no difference in the time-to-event comparison under the specified statistical framework.
HR: describes the estimated relative treatment effect.
Confidence interval: describes uncertainty around that estimate.
P-value: quantifies the evidence against the null hypothesis within the specified testing framework.
These quantities should not be collapsed into a single concept. A small P-value does not imply a large treatment effect, and a hazard ratio does not by itself establish statistical evidence without considering its uncertainty and the testing procedure.
9. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm. These figures use the affected-patients-over-patients-at-risk format provided in the registry data.
| Safety measure | Everolimus + trastuzumab + vinorelbine | Placebo + trastuzumab + vinorelbine |
|---|---|---|
| Serious adverse events | 122 / 280 | 58 / 282 |
The registry also provides an important population distinction for interpreting safety. Seven randomized patients—4 in the everolimus arm and 3 in the placebo arm—never received treatment and were therefore excluded from the Safety Set. The registry states that all randomized patients were included in the Full Analysis Set.
10. Analysis Populations
| Population | Registry information | Statistical role |
|---|---|---|
| Full Analysis Set | All randomized patients | Primary PFS analysis |
| Safety Set | Randomized patients who received treatment; 7 randomized patients who never received treatment were excluded | Safety summaries |
The distinction is not merely administrative. Randomization protects the comparability created by treatment assignment, so an efficacy analysis based on all randomized patients is closely aligned with the intention-to-treat principle. A safety analysis, by contrast, is naturally connected to actual exposure to study treatment.
11. Statistical Methods Explained
Why was a log-rank test used?
PFS is a time-to-event endpoint. Patients can experience progression or death at different times, and some patients may be censored before an event occurs. A log-rank test is designed specifically to compare the event-time experience of two groups while incorporating the timing of events and the number of patients at risk over follow-up.
What does a hazard ratio of 0.78 mean?
A hazard ratio of 0.78 indicates an estimated instantaneous event hazard that is 78% as large in the everolimus-containing group as in the placebo-containing group, under the reported analysis. A simple relative interpretation is a 22% lower estimated hazard. This does not mean that 22% of patients avoided progression or that each patient had exactly a 22% reduction in probability of progression.
Why does the confidence interval matter?
The estimate 0.78 is not known without uncertainty. The reported two-sided 95% CI of 0.65–0.95 communicates the precision of the estimated hazard ratio. A point estimate without its interval would provide an incomplete description of the statistical result.
Why doesn't the P-value measure effect size?
The P-value of 0.0067 concerns the evidence against the null hypothesis in the specified test. It depends on both the observed data and the amount of information available. The hazard ratio describes the estimated relative effect, while the confidence interval describes its uncertainty.
Why was the Full Analysis Set important?
The registry states that the Full Analysis Set consisted of all randomized patients and identifies intention-to-treat analysis as a concept in the primary analysis. Analyzing randomized patients according to their assigned group preserves the treatment comparison created by randomization and avoids redefining the primary efficacy population according to treatment received.
What does censoring mean in a PFS analysis?
Censoring occurs when the precise event time is not observed within the available follow-up for a patient. Such a patient can still contribute information up to the point at which follow-up ends under the applicable censoring rules. Time-to-event methods such as Kaplan-Meier estimation and the log-rank test are designed to incorporate this incomplete event-time information.
Why should the hazard ratio not be treated as a median difference?
A hazard ratio compares relative event hazards over time; a median compares the time at which an estimated survival or event-free curve reaches a specified probability level. They summarize different features of the time-to-event distribution. The registry-reported BOLERO-3 data report the hazard ratio but do not provide median PFS.
12. Kaplan-Meier Estimation and Time-to-Event Thinking
The registry identifies PFS as a time-to-event endpoint, making survival-analysis concepts central to interpreting the result. A Kaplan-Meier analysis would estimate the probability of remaining progression-free over time while accounting for censoring.
Here, di represents events at an event time and ni represents patients at risk immediately before that time.
The ClinicalTrials.gov record does not report the complete sequence of event times, censoring times, or Kaplan-Meier estimates needed to reproduce a curve. The correct educational conclusion is therefore that BOLERO-3 illustrates Kaplan-Meier-type time-to-event reasoning, while the available numerical record supports the reported hazard ratio, confidence interval, and log-rank P-value rather than a reconstructed curve.
13. The Primary Analysis as a Statistical Workflow
This sequence illustrates why clinical-trial statistics should be read as a connected system rather than as isolated numbers. Randomization defines the comparison, the endpoint defines what event is being measured, follow-up determines when information is observed, the survival method handles event timing and censoring, and the effect estimate and confidence interval summarize the treatment comparison.
14. What the Primary Result Does — and Does Not — Establish
It establishes a reported statistical comparison
The registry reports a formal log-rank comparison of PFS between the randomized treatment groups, with HR 0.78, 95% CI 0.65–0.95, and P = 0.0067.
It does not provide a median PFS
The ClinicalTrials.gov record does not contain a median PFS estimate or its confidence interval, so none is reported here.
It does not quantify every patient's benefit
The hazard ratio is a population-level relative time-to-event measure. It does not imply that every patient experienced the same treatment effect.
It does not summarize safety by itself
Efficacy and safety are distinct dimensions. Serious adverse events are reported separately by treatment arm.
15. Limitations and Interpretation Issues
- Limited reported statistical analyses: the ClinicalTrials.gov record contains one formal statistical analysis, corresponding to the primary PFS endpoint. It does not provide a comparable set of formal statistical analyses for the other posted outcome measures.
- No median PFS in the ClinicalTrials.gov record: the ClinicalTrials.gov record reports the hazard ratio, confidence interval, and P-value but does not provide median PFS.
- No Kaplan-Meier estimates reported: the available data do not contain the underlying event and censoring information needed to reconstruct the survival curve.
- No subgroup results reported: subgroup-specific hazard ratios and confidence intervals are not part of the ClinicalTrials.gov record and therefore are not reported.
- No baseline table reported: detailed baseline demographic and disease characteristics are not included in the ClinicalTrials.gov record.
- No multiplicity information reported: the ClinicalTrials.gov record do not describe a multiplicity adjustment strategy or endpoint hierarchy beyond identifying the primary endpoint and a superiority hypothesis.
- No interim-analysis information reported: the ClinicalTrials.gov record do not specify whether interim efficacy analyses were performed or how any interim testing was controlled.
- No missing-data or imputation strategy reported: the registry data do not specify an imputation method for missing assessments.
- No Bayesian analysis reported: the reported methodology is frequentist, centered on the log-rank test and hazard ratio; no Bayesian method is identified in the ClinicalTrials.gov record.
- Safety population differs from the randomized population: seven randomized patients who never received treatment were excluded from the Safety Set, whereas all randomized patients were included in the Full Analysis Set.
- Proportional-hazards interpretation: the ClinicalTrials.gov record does not provide diagnostics for the proportional-hazards assumption. The hazard ratio should therefore be understood as the reported summary measure rather than treated as a complete description of how treatment effects behave at every time point.
16. Design Topics Not Specified in the Supplied Registry Data
| Topic | What the ClinicalTrials.gov record establishes |
|---|---|
| Non-inferiority margin | Not specified; the registered hypothesis type is superiority. |
| Crossover | Not specified in the ClinicalTrials.gov record. |
| Factorial design | Not specified; the design model is parallel with two arms. |
| Multiplicity adjustment | Not specified in the ClinicalTrials.gov record. |
| Interim analysis | Not specified in the ClinicalTrials.gov record. |
| Missing-data / imputation method | Not specified in the ClinicalTrials.gov record. |
| Stratification factors | Not specified in the ClinicalTrials.gov record. |
| Bayesian methods | Not specified; the posted analysis uses a log-rank test and hazard ratio. |
These omissions are important because a statistical analysis should distinguish between what is documented and what might commonly be used in a similar trial. The absence of a method from the ClinicalTrials.gov record is not a basis for assuming that method was used.
17. Safety Population and Denominators
The serious-adverse-event counts illustrate another recurring issue in clinical-trial statistics: the denominator matters. The registry reports 122/280 for the everolimus-containing arm and 58/282 for the placebo-containing arm.
| Arm | Affected | At risk | Reported format |
|---|---|---|---|
| Everolimus + trastuzumab + vinorelbine | 122 | 280 | 122/280 |
| Placebo + trastuzumab + vinorelbine | 58 | 282 | 58/282 |
The at-risk counts are not the same as the total randomized enrollment of 569. That is consistent with the registry's distinction between the Full Analysis Set and Safety Set. The seven randomized patients who never received treatment were excluded from the Safety Set.
For safety interpretation, it is therefore essential to use the denominator specified for the safety endpoint rather than substituting the overall randomized enrollment.
18. Why This Trial Matters Statistically
BOLERO-3 provides a compact example of how a randomized oncology trial translates a clinical question into a time-to-event analysis. Its statistical profile connects randomization, double masking, intention-to-treat analysis, an event-driven PFS endpoint, log-rank testing, hazard-ratio estimation, confidence intervals, and safety-population definitions.
| Concept | How it appears in BOLERO-3 |
|---|---|
| Randomization | Randomized allocation to two parallel treatment groups. |
| Double masking | The registry identifies the trial as double-masked. |
| Time-to-event endpoint | PFS measured from randomization until progression or death. |
| Event-driven analysis | Primary PFS analysis performed when 415 events were reached. |
| Intention-to-treat analysis | Primary analysis population was the Full Analysis Set consisting of all randomized patients. |
| Log-rank test | Reported statistical method for the primary PFS comparison. |
| Hazard ratio | 0.78 for the primary PFS comparison. |
| Confidence interval | Two-sided 95% CI of 0.65–0.95. |
| P-value | Two-sided P = 0.0067. |
| Superiority testing | The registered hypothesis type was superiority. |
| Safety population | Seven randomized patients who never received treatment were excluded from the Safety Set. |
19. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The randomized comparison of investigator-assessed PFS produced a hazard ratio of 0.78 with a two-sided 95% CI of 0.65–0.95 and a P-value of 0.0067 using a log-rank test in the Full Analysis Set.
Clinical interpretation
The ClinicalTrials.gov record supports describing the relative PFS treatment effect reported by the registry. They do not provide median PFS, absolute PFS probabilities, or other time-specific efficacy measures from which additional clinical-effect summaries could be calculated without introducing information outside the ClinicalTrials.gov record.
This distinction prevents a common statistical mistake: treating a statistically significant time-to-event comparison as though it automatically supplies every clinically relevant measure of benefit. The reported HR, confidence interval, and P-value answer specific statistical questions; they do not replace absolute event-time summaries or other clinical outcomes that are not included in the ClinicalTrials.gov record.
20. A Closer Look at the Primary Endpoint Definition
The registry's PFS definition combines two possible events: radiologically documented tumor progression and death from any cause. The first event to occur determines the endpoint event time.
This composite endpoint structure is important when interpreting the hazard ratio. A lower PFS hazard can arise through differences in the timing of progression, differences in the timing of death before documented progression, or both. The registry-reported aggregate hazard ratio does not decompose the effect into those components.
21. Results Summary
Primary PFS result
95% CI 0.65–0.95 · Two-sided P = 0.0067
Log-rank analysis · Full Analysis Set · Superiority hypothesis
The primary result is therefore best read as a coordinated set of statistics rather than as a single number: the hazard ratio describes the estimated relative event hazard, the confidence interval describes uncertainty around that estimate, and the P-value describes evidence against the null hypothesis under the reported testing framework.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Statistical Calculators
24. Sources
- ClinicalTrials.gov: BOLERO-3 — NCT01007942. Official registry record and source for the trial data presented on this page.
- Linked publication: PubMed record for PMID 24742739.
Continue through the Clinical Biostats statistical pathway
Connect the trial's time-to-event endpoint to deeper tutorials, statistical calculators, and other clinical-trial analyses.
25. Record Summary
BOLERO-3 is a useful teaching example of randomized time-to-event analysis. The ClinicalTrials.gov record describes a completed, double-masked, randomized phase 3 parallel trial with 569 enrolled patients and two treatment arms. Its primary endpoint was investigator-assessed progression-free survival, defined from randomization to radiologically documented tumor progression or death, assessed every 6 weeks until disease progression or death up to about 41 months. The primary analysis used the Full Analysis Set of all randomized patients, a log-rank test, and a hazard ratio under a superiority hypothesis.
The reported primary result was an HR of 0.78 with a two-sided 95% CI of 0.65–0.95 and P = 0.0067. Statistically, the three quantities should be interpreted together: the HR provides the relative effect estimate, the confidence interval provides its uncertainty, and the P-value provides evidence against the null hypothesis within the reported testing framework. The registry also reports serious adverse events separately by treatment arm and identifies a distinction between the randomized Full Analysis Set and the Safety Set.