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
FLIPPER was a randomized phase 2 parallel-group trial evaluating fulvestrant plus palbociclib versus fulvestrant plus placebo as first-line treatment for postmenopausal patients with hormone receptor-positive advanced breast cancer. The registered primary endpoint was the rate of progression-free survival at 1 year, assessed according to RECIST version 1.1 by the investigator.
| Feature | FLIPPER |
|---|---|
| Trial acronym | FLIPPER |
| ClinicalTrials.gov identifier | NCT02690480 |
| Phase | Phase 2 |
| Status | UNKNOWN |
| Condition | Breast Neoplasms |
| Enrollment | 189 |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | TRIPLE |
| Primary purpose | TREATMENT |
| Number of arms | 2 |
| Lead sponsor | Spanish Breast Cancer Research Group |
| Sponsor type | OTHER |
2. Clinical Question
The registered clinical question concerns whether adding palbociclib to fulvestrant changes progression-free survival in postmenopausal patients with hormone receptor-positive advanced breast cancer receiving first-line treatment.
Population
Postmenopausal patients with hormone receptor-positive advanced breast cancer.
Intervention
Fulvestrant plus PD-0332991 (palbociclib).
Comparator
Fulvestrant plus placebo.
Primary question
What is the rate of progression-free survival at 1 year in the randomized comparison?
3. Trial Design
Participants were assigned to one of two parallel treatment groups.
The registry classifies the study as triple-masked.
The registry classifies the primary purpose as treatment.
The registry records two treatment arms within the randomized parallel design.
Fulvestrant + Palbociclib
- PD-0332991 (palbociclib)
- Fulvestrant
Fulvestrant + Placebo
- Placebo
- Fulvestrant
4. Trial Timeline
Study start
The registry records 2016-02-17 as the study start date.
Primary completion
The registry records 2020-01-11 as the primary completion date.
5. Endpoints
The registry lists one registered primary endpoint. Its wording is important because the endpoint is a fixed-time PFS measure rather than a generic request to compare all progression-free survival over the entire follow-up period.
| Endpoint | Time frame | Registry definition / assessment |
|---|---|---|
| Efficacy in terms of the rate of Progression-Free Survival (PFS) | at 1 year | Assessed according to Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 by the investigator |
What the primary endpoint means
A progression-free survival rate at 1 year is a time-specific measure. Conceptually, it asks what proportion of the randomized population remains alive without progression at the 1-year landmark, with the exact statistical handling of patients whose status is not observed through that time depending on the prespecified analysis method.
This distinction matters. A 1-year PFS rate is not the same quantity as median PFS, an instantaneous hazard ratio, or the percentage of patients who ever experience tumor response. Each summarizes a different aspect of time-to-event behavior.
6. Planned Analysis
Primary endpoint
Time frame: at 1 year
Assessment: RECIST version 1.1 by the investigator
No formal statistical analyses were posted to ClinicalTrials.gov for the registered primary endpoint. The registry therefore identifies the endpoint and its assessment framework but does not report a numerical treatment comparison, confidence interval, or p-value.
How a 1-year PFS endpoint is typically analysed
Because progression-free survival is a time-to-event endpoint, a common approach is to estimate the progression-free survival function with the Kaplan-Meier method and obtain the estimated survival probability at the prespecified 1-year time point. Patients who have not experienced progression or death by their last adequate assessment can contribute follow-up information up to censoring.
For PFS, T represents the time to the defined progression-free survival event. At the 1-year landmark, S(1 year) represents the estimated probability of remaining progression-free at that time under the analysis framework.
If two randomized groups are compared, the analysis can also report an absolute difference between their 1-year PFS estimates, a ratio of the estimates, or a time-to-event effect measure such as a hazard ratio. The appropriate estimand should be specified before analysis because these quantities answer different questions.
7. Statistical Methodology
Randomized comparison
The randomized design is the central statistical feature of FLIPPER. Randomization creates the framework for comparing outcomes between the two treatment strategies without relying on observational adjustment to establish the primary treatment contrast.
Time-to-event analysis
Progression-free survival is fundamentally a time-to-event outcome. Unlike a simple binary response measured at a single visit, PFS incorporates both whether an event occurs and when it occurs. This makes methods that account for censoring particularly relevant.
Kaplan-Meier estimation
For a time-to-event endpoint such as PFS, the Kaplan-Meier estimator is commonly used to estimate the probability of remaining event-free over time. Its major advantage is that it can incorporate patients whose event status is not observed for the entire follow-up period.
Here, di is the number of events at time ti and ni is the number at risk immediately before that time.
RECIST version 1.1 assessment
The registry specifies RECIST version 1.1 and investigator assessment for the primary endpoint. RECIST provides a standardized framework for classifying tumor response and progression from imaging-based disease assessments.
Fixed-time estimation
The registered time frame of 1 year gives the primary endpoint a prespecified landmark. Statistically, this is useful because it defines the point at which the progression-free survival function is evaluated rather than leaving the primary time horizon unspecified.
Comparing the randomized groups
For a conventional time-to-event comparison, a stratified or unstratified log-rank test and a Cox proportional-hazards model are commonly considered when their assumptions and the prespecified statistical analysis plan support them. The registry information available here does not specify which inferential model was used for FLIPPER.
8. Statistical Methods Explained
Why is PFS a time-to-event endpoint?
PFS contains timing information. Two patients can both eventually experience progression, but the statistical meaning is different if one progresses substantially earlier than the other. A time-to-event analysis retains that timing information rather than reducing the entire follow-up to a simple yes/no indicator.
Why use a 1-year PFS rate?
A fixed-time PFS rate provides an easily interpretable landmark: the estimated probability of remaining progression-free at 1 year. It can complement other measures such as a median PFS or a hazard ratio, although those measures should not be treated as interchangeable.
Why is censoring important?
Not every participant will necessarily have a documented progression event before the end of observation. A patient may remain progression-free when follow-up ends. Survival-analysis methods can use the information available up to that point while accounting for the fact that the eventual event time is unknown.
What does RECIST version 1.1 contribute?
RECIST version 1.1 provides a standardized framework for determining whether measurable disease has progressed or responded according to imaging assessments. This is particularly important because a time-to-progression endpoint depends on consistent event definitions.
Why does randomization matter statistically?
Randomization determines treatment assignment rather than allowing treatment choice to be driven by patient or clinician characteristics. It therefore provides the foundation for a causal comparison of the assigned treatment strategies, subject to adherence, follow-up, endpoint measurement, and other features of the trial.
Why should a PFS rate not be interpreted as a hazard ratio?
A PFS rate describes an absolute estimated probability at a particular time point. A hazard ratio is a relative comparison of event rates over time under a proportional-hazards model. A numerical relationship between the two cannot be assumed without the underlying survival data and model.
9. Analysis Population and Interpretation
The registry information identifies enrollment of 189 participants and a randomized, two-arm parallel design. It does not provide detailed definitions of efficacy analysis populations, per-protocol populations, or safety populations in the information presented here.
| Statistical element | What is established by the registry | Why it matters |
|---|---|---|
| Enrollment | 189 | Defines the registered trial size. |
| Allocation | Randomized | Provides the basis for a randomized treatment comparison. |
| Design model | Parallel | Participants remain assigned within one of two concurrent treatment strategies rather than crossing between treatments as part of the registered design. |
| Masking | Triple | Indicates triple masking in the registry classification. |
| Primary endpoint | PFS rate at 1 year | Defines the principal efficacy quantity and its time horizon. |
| Assessment | RECIST version 1.1 by investigator | Defines the registered framework for assessing progression. |
10. Interpreting a Future PFS Estimate
Suppose a randomized comparison reports a 1-year PFS estimate for each arm. The first question should be whether the estimates describe the same estimand, analysis population, censoring rules, and assessment framework. Only then should their numerical difference or ratio be interpreted.
Absolute difference
An absolute difference between two 1-year PFS estimates describes how far apart the estimated probabilities are at that specific time point.
Relative comparison
A relative measure compares the estimates proportionally. It should not be confused with an absolute percentage-point difference.
Hazard ratio
If a Cox model is used, the hazard ratio summarizes a relative event-rate contrast over follow-up under the model. It is not the same as the 1-year PFS rate.
Confidence interval
A confidence interval communicates statistical uncertainty around an estimated effect. It does not describe the range of outcomes that an individual patient can experience.
11. Censoring and the 1-Year Landmark
The 1-year PFS rate depends not only on observed progression events but also on how participants are handled when complete follow-up to 1 year is unavailable. This is why survival-analysis methods are generally preferable to simply counting patients who are known to be progression-free at the landmark.
A simple observed proportion can exclude or classify patients differently when follow-up is incomplete. A time-to-event estimator can use partial follow-up through censoring, subject to the assumptions of the analysis.
The validity of censoring-based inference depends on assumptions about why observations are censored. Informative loss to follow-up can create bias if the probability of censoring is related to the unobserved progression experience in a way not adequately addressed by the analysis.
12. Investigator Assessment and Measurement
The primary endpoint is specified as being assessed by the investigator according to RECIST version 1.1. The distinction between investigator assessment and an independent central review can matter statistically because assessment procedures can influence the timing and classification of progression events.
For a PFS endpoint, even modest differences in assessment timing can affect event dates and therefore the estimated survival curve. A prespecified and consistently applied assessment framework is consequently an important part of the statistical definition of the endpoint.
13. Randomization and Masking
FLIPPER is registered as randomized with a parallel design and triple masking. These design features address different sources of potential bias.
| Design feature | Statistical role |
|---|---|
| Randomization | Creates comparable treatment-assignment groups in expectation and supports causal treatment comparisons. |
| Parallel design | Each randomized group follows its assigned treatment strategy rather than forming a crossover comparison. |
| Triple masking | Reduces the opportunity for knowledge of treatment assignment to influence relevant trial procedures or assessments. |
| RECIST version 1.1 | Standardizes the framework used for progression assessment. |
14. Multiplicity, Interim Analysis, and Other Design Features
The registry information identifies one registered primary endpoint: efficacy in terms of the rate of PFS at 1 year. It does not provide a posted statistical analysis specifying an endpoint hierarchy, multiplicity adjustment, interim efficacy boundary, alpha-spending strategy, or Bayesian analysis.
These distinctions are important because statistical interpretation depends on the prespecified testing framework. If multiple endpoints, repeated interim looks, or multiple treatment contrasts are tested formally, the type I error structure should be defined in advance.
Multiplicity
When several hypotheses are tested, the overall false-positive error rate can exceed the nominal level unless the testing strategy accounts for the multiplicity.
Interim analysis
Repeatedly inspecting accumulating efficacy data can affect type I error unless interim monitoring is incorporated into the statistical design.
Bayesian methods
The registry information does not identify a Bayesian primary analysis, so Bayesian posterior probabilities should not be assumed as the trial's inferential framework.
Missing data
The registry information does not specify an imputation strategy. For a time-to-event endpoint, censoring rules are generally more central than conventional single-value imputation of missing continuous outcomes.
15. Planned Analysis of the Primary Endpoint
The primary endpoint is the rate of PFS at 1 year. A statistically coherent analysis would begin by defining the event and censoring rules, estimating the PFS function over time, and evaluating the function at the 1-year landmark.
| Analysis step | Statistical purpose |
|---|---|
| Define progression-free survival | Establish exactly what constitutes an event and what constitutes censoring. |
| Construct the time-to-event dataset | Represent each participant's follow-up time and event/censoring status. |
| Estimate PFS over time | Use an appropriate survival estimator such as Kaplan-Meier when prespecified and appropriate. |
| Evaluate PFS at 1 year | Obtain the estimated progression-free probability at the registered landmark. |
| Quantify uncertainty | Report an appropriate confidence interval around the 1-year PFS estimate. |
| Compare treatment groups | Use the prespecified inferential method appropriate to the randomized comparison. |
The registry does not post the formal statistical analysis for these steps. In particular, it does not provide a numerical 1-year PFS estimate, treatment-effect estimate, confidence interval, or p-value.
16. Why a P-Value Would Not Be Enough
If a formal comparison were reported, the p-value would address evidence against a specified null hypothesis under the chosen statistical model. It would not describe the size of the treatment effect.
A small p-value can provide evidence against a null hypothesis, but it does not tell the reader how large the treatment difference is. Effect estimates and their confidence intervals are therefore essential for understanding the magnitude and precision of a randomized comparison.
A difference in 1-year PFS should be considered in the context of the absolute probabilities, duration of follow-up, endpoint definition, uncertainty, and the broader clinical evidence. Statistical significance alone does not establish the practical importance of a difference.
17. Statistical Assumptions
Different analyses of PFS rely on different assumptions. A fixed-time Kaplan-Meier estimate relies on appropriate handling of censoring and event definitions. A Cox proportional-hazards model additionally relies on the proportional-hazards assumption for the interpretation of a single hazard ratio as a common relative effect over time.
| Method | Key consideration |
|---|---|
| Kaplan-Meier | Requires appropriate event and censoring definitions and assumptions supporting noninformative censoring. |
| Log-rank test | Provides a time-to-event comparison whose behavior is most favorable when treatment effects are reasonably represented through the follow-up. |
| Cox proportional-hazards model | Uses a proportional-hazards framework; a single HR can be difficult to interpret when hazards are substantially nonproportional. |
| Fixed-time PFS rate | Provides an absolute estimate at the prespecified 1-year landmark rather than a summary of the entire survival curve. |
18. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record does not provide numerical primary-endpoint results, confidence intervals, or p-values for the registered PFS endpoint.
- Limited endpoint information: the registered information identifies the PFS rate at 1 year and its RECIST version 1.1 investigator assessment, but does not provide a detailed statistical analysis plan.
- Analysis population details: detailed definitions of efficacy, per-protocol, and safety populations are not identified in the available registry information.
- Event and censoring details: the registry information does not provide individual progression times, censoring times, or the detailed rules used to determine those observations.
- Unreported inferential framework: the registry information does not identify a specific log-rank, Cox, Bayesian, multiplicity, or interim-analysis framework.
- Investigator assessment: progression was specified for investigator assessment according to RECIST version 1.1; assessment methodology is therefore an important component of the endpoint definition.
- Fixed-time interpretation: a 1-year PFS rate describes one landmark and should not be treated as a complete summary of the PFS trajectory.
19. Why This Trial Matters Statistically
FLIPPER is a useful teaching example because its registered primary endpoint illustrates how a clinical question can be translated into a specific time-to-event estimand. The distinction between a 1-year PFS rate and broader survival measures is particularly important when interpreting randomized oncology trials.
| Concept | How it appears in FLIPPER |
|---|---|
| Randomization | The trial is registered as randomized. |
| Parallel-group design | The trial has two parallel treatment arms. |
| Triple masking | The registry classifies the study as triple-masked. |
| Time-to-event analysis | The primary endpoint is progression-free survival. |
| Fixed-time estimand | The primary PFS time frame is at 1 year. |
| RECIST assessment | Progression is assessed according to RECIST version 1.1 by the investigator. |
| Censoring | Potentially incomplete follow-up is a central consideration for time-to-event estimation. |
| Kaplan-Meier estimation | A standard approach for estimating a PFS probability at a fixed time point. |
| Effect estimation | Absolute 1-year differences and model-based time-to-event measures answer different statistical questions. |
| Uncertainty | Confidence intervals are needed to communicate precision around an estimated PFS rate or treatment effect. |
20. Statistical Concepts in This Trial
The primary statistical lesson from FLIPPER is the importance of matching the analysis to the endpoint. A 1-year PFS rate requires attention to time-to-event structure, censoring, event definitions, and the distinction between an absolute landmark probability and a relative treatment-effect measure.
Endpoint definition
The registry specifies PFS rate at 1 year rather than an unspecified overall PFS comparison.
Event definition
RECIST version 1.1 provides the registered framework for investigator-assessed progression.
Time horizon
The 1-year landmark gives the primary efficacy question a specific time point.
Statistical uncertainty
An estimated PFS rate should be accompanied by an appropriate measure of uncertainty when results are available.
21. Record Summary
FLIPPER is a randomized, triple-masked, phase 2 parallel-group trial with 189 enrolled participants and two treatment arms: fulvestrant plus palbociclib and fulvestrant plus placebo. The registered primary endpoint is efficacy in terms of the rate of progression-free survival at 1 year, assessed according to RECIST version 1.1 by the investigator.
From a statistical perspective, the trial illustrates the difference between a fixed-time survival probability and other time-to-event measures. A 1-year PFS rate is an absolute landmark quantity, whereas a hazard ratio would represent a relative event-rate comparison under a model such as Cox regression. These quantities should not be substituted for one another.
The registry records a study start date of 2016-02-17, a primary completion date of 2020-01-11, and a current status of UNKNOWN. The ClinicalTrials.gov record does not post formal statistical analyses or numerical results for the registered primary endpoint.
22. Sources
- ClinicalTrials.gov: FLIPPER (NCT02690480).
Continue through the Clinical Biostats clinical-trial library
Clinical trial analysis connects study design, endpoint definitions, survival methods, uncertainty, and interpretation into a single statistical framework.