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 ClinicalTrials.gov record.
1. Trial at a Glance
PEARL was a completed randomized phase 3 trial in metastatic breast cancer evaluating palbociclib with endocrine therapy against capecitabine in patients with hormone receptor-positive, HER2-negative disease and resistance to aromatase inhibitors. The registry reports a time-to-event primary endpoint, Progression-Free Survival (PFS), and posts secondary quality-of-life time-to-deterioration analyses using Cox proportional-hazards models.
| Feature | PEARL |
|---|---|
| Trial name | PEARL |
| NCT identifier | NCT02028507 |
| Phase | Phase 3 |
| Condition | Metastatic Breast Cancer |
| Population | HR+/HER2- metastatic breast cancer with resistance to aromatase inhibitors |
| Enrollment | 693 |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Primary endpoint | Progression-Free Survival (PFS) |
| Primary endpoint type | Time-to-event |
| Statistical method represented in posted analyses | Cox proportional-hazards model |
| Effect measure | Hazard ratio |
| Hypothesis type | Superiority |
| Trial status | Completed |
| Lead sponsor | Spanish Breast Cancer Research Group |
| Start | 2014-03-13 |
| Primary completion | 2019-01-14 |
2. Clinical Question
The central registered question concerns whether palbociclib combined with endocrine therapy should be compared with capecitabine for time to disease progression or death in patients with HR+/HER2- metastatic breast cancer whose disease has demonstrated resistance to aromatase inhibitors.
Population
Patients with metastatic breast cancer that is hormone receptor-positive and HER2-negative, with resistance to aromatase inhibitors.
Intervention
Palbociclib combined with endocrine therapy. The registry's posted analyses pool the exemestane and fulvestrant cohorts for the reported comparison.
Comparator
Capecitabine.
Primary question
Does the palbociclib-plus-endocrine-therapy strategy affect Progression-Free Survival relative to capecitabine?
3. Trial Design
Palbociclib + endocrine therapy
- Cohort 1: Palbociclib Plus Exemestane
- Cohort 2: Palbociclib Plus Fulvestrant
- The posted QoL analyses combine Cohort 1 and Cohort 2.
Capecitabine
- Capecitabine is the comparator in the posted secondary analyses.
- The posted analyses combine the capecitabine groups from Cohort 1 and Cohort 2.
The four-arm structure is important statistically because the intervention side is not a single endocrine therapy: the posted comparison combines two palbociclib-plus-endocrine-therapy cohorts, while the comparator is represented as capecitabine across the corresponding cohorts. The ClinicalTrials.gov record does not provide the randomization allocation ratio or the enrolled sample size for each randomized arm, so those quantities are not reconstructed on this page.
4. Trial Timing and Registry Status
Study start
The registry lists March 13, 2014 as the study start date.
Primary completion
The registry lists January 14, 2019 as the primary completion date.
Current registry status in the ClinicalTrials.gov record
The ClinicalTrials.gov record identifies PEARL as completed and reports results.
5. Primary Endpoint
| Endpoint | Registry definition and time frame | Analysis information reported |
|---|---|---|
| Progression-Free Survival (PFS) | The primary efficacy variable is PFS based on the investigator's assessment. PFS is defined as the time from randomization to the first documented progressive disease based on the investigator's assessment, using RECIST version 1.1, or death from any cause, whichever occurs first. The registered time frame is through study treatment, and average of 8 months. | Primary endpoint type: time-to-event. The ClinicalTrials.gov record identifies Cox proportional-hazards modeling as the statistical method used in the trial, but do not provide a statistical-analysis record with a PFS estimate, confidence interval, or P-value. |
6. Planned Analysis
Because the registered primary endpoint is a time-to-event measure, the natural statistical framework is survival analysis. The endpoint begins at randomization and ends at the first qualifying event: investigator-assessed RECIST version 1.1 progression or death from any cause, whichever occurs first.
A survival function describes the probability of remaining free of the event beyond time t. In a randomized comparison, Kaplan-Meier estimation can describe each treatment-group distribution, while a Cox proportional-hazards model can summarize the relative hazard as a hazard ratio.
The ClinicalTrials.gov record identifies the trial's the statistical method as a Cox proportional-hazards model, with hazard ratio as the effect measure and superiority as the hypothesis type. However, the ClinicalTrials.gov record does not provide the primary PFS estimate or its corresponding confidence interval and P-value.
A PFS hazard ratio would quantify the relative instantaneous rate of progression or death between the randomized groups under the fitted Cox model. A value below 1 would correspond to a lower estimated event hazard for the numerator group, while a value above 1 would correspond to a higher estimated event hazard.
The hazard ratio would not be the same as the proportion of patients who progressed, the difference in median PFS, or an absolute percentage-point treatment effect. Those quantities answer different questions.
Because the ClinicalTrials.gov record does not contain the PFS result itself, any numerical PFS effect would require information outside the registry record and therefore is not reproduced here.
7. Secondary Quality-of-Life Analyses
The posted statistical analyses evaluate Time to Deterioration (TTD) in EORTC quality-of-life scales. The analysis population is defined as the QoL population: a subset of enrolled patients with available QoL questionnaires, specifically the baseline questionnaire and at least one more questionnaire.
The groups compared in every registry-reported analysis are Cohort 1 and 2: (Exemestane or Fulvestrant) Plus Palbociclib versus Cohort 1 and 2: Capecitabine. The registry reports Cox regression, a hazard ratio as the effect measure, two-sided 95% confidence intervals, and superiority as the hypothesis type.
EORTC QLQ-C30 Functional Scales
| Quality-of-life scale | HR | 95% CI | P-value |
|---|---|---|---|
| Global health status / quality of life | 0.70 | 0.55–0.89 | 0.003 |
| Physical functioning | 0.62 | 0.50–0.76 | <0.001 |
| Role functioning | 0.63 | 0.51–0.79 | <0.001 |
| Emotional functioning | 0.97 | 0.76–1.25 | 0.818 |
| Cognitive functioning | 0.70 | 0.54–0.89 | 0.004 |
| Social functioning | 0.62 | 0.49–0.78 | <0.001 |
These estimates describe time to deterioration rather than average questionnaire scores. An HR below 1 therefore corresponds to a lower estimated instantaneous rate of reaching the defined deterioration event in the palbociclib-plus-endocrine-therapy group relative to capecitabine, under the Cox model.
The global health status / quality-of-life analysis produced an HR of 0.70 with a 95% CI of 0.55–0.89 and P = 0.003. In model terms, this is an estimated 30% lower instantaneous hazard of the defined deterioration event because 1 − 0.70 = 0.30.
The confidence interval indicates uncertainty around the estimated relative hazard. It does not mean that individual patients experienced hazards confined to that interval, and it does not directly give the probability that the treatment is beneficial.
The P-value addresses evidence against the tested null hypothesis under the specified analysis; it is not a measure of effect size. The effect size is described by the HR, while the confidence interval describes its precision.
The physical functioning analysis produced an HR of 0.62 with a 95% CI of 0.50–0.76 and P < 0.001. The role functioning analysis produced an HR of 0.63 with a 95% CI of 0.51–0.79 and P < 0.001.
Both estimates are below 1, so the fitted models estimate lower hazards of deterioration for these scales in the palbociclib-plus-endocrine-therapy comparison. These results concern the specific QoL deterioration endpoints and should not be substituted for the registered primary PFS endpoint.
The emotional functioning result was different in statistical form: HR 0.97, 95% CI 0.76–1.25, P = 0.818. The point estimate is close to 1 and the confidence interval spans 1. This means the registry-reported analysis does not provide evidence of a clear difference in the instantaneous rate of emotional-functioning deterioration under this model.
Importantly, this does not establish that the two groups are identical. A nonsignificant P-value is not proof of no effect; the confidence interval provides the more informative description of the range of effects compatible with the analysis.
EORTC QLQ-C30 Symptom Scales
| Symptom scale | HR | 95% CI | P-value |
|---|---|---|---|
| Fatigue | 0.70 | 0.57–0.86 | 0.001 |
| Nausea and vomiting | 0.58 | 0.45–0.76 | <0.001 |
| Pain | 0.78 | 0.62–0.99 | 0.042 |
| Dyspnea | 1.08 | 0.81–1.44 | 0.599 |
| Insomnia | 0.84 | 0.64–1.10 | 0.196 |
| Appetite loss | 0.71 | 0.54–0.92 | 0.011 |
| Constipation | 1.08 | 0.83–1.41 | 0.564 |
| Diarrhea | 0.42 | 0.32–0.55 | <0.001 |
| Financial difficulties | 0.78 | 0.56–1.07 | 0.118 |
The symptom-scale results illustrate why a collection of hazard ratios should not be reduced to a single conclusion. For example, the diarrhea analysis has an HR of 0.42 with a 95% CI of 0.32–0.55 and P < 0.001, whereas dyspnea has an HR of 1.08 with a 95% CI of 0.81–1.44 and P = 0.599.
The first result represents a lower estimated hazard of the specified deterioration event under the Cox model; the second has a point estimate above 1 but a confidence interval that includes 1. The two analyses therefore provide different patterns of statistical evidence.
Neither result, however, should be interpreted as establishing an overall treatment benefit or harm across all symptoms. These are distinct secondary endpoints, analyzed in a QoL population rather than the full randomized population.
EORTC QLQ-BR23 Scales
| Breast-cancer quality-of-life scale | HR | 95% CI | P-value |
|---|---|---|---|
| Body image | 0.95 | 0.74–1.21 | 0.659 |
| Future perspective | 1.01 | 0.74–1.39 | 0.928 |
| Systemic side-effects | 0.79 | 0.64–0.98 | 0.029 |
| Breast symptoms | 0.93 | 0.71–1.21 | 0.571 |
| Arm symptoms | 0.84 | 0.66–1.09 | 0.187 |
Across the registry-reported EORTC QLQ-BR23 analyses, the confidence intervals for body image, future perspective, breast symptoms, and arm symptoms include 1. The systemic side-effects scale has an HR of 0.79 with a 95% CI of 0.64–0.98 and P = 0.029.
The systemic side-effects result illustrates the distinction between an effect estimate and its uncertainty. An HR of 0.79 corresponds to an estimated lower instantaneous deterioration hazard in the palbociclib-plus-endocrine-therapy group under the fitted model, while the 95% CI of 0.64–0.98 indicates uncertainty around that estimate.
The P-value of 0.029 is evidence against the null hypothesis used for this individual analysis. It does not quantify how large the treatment effect is, nor does it establish that this secondary result should be interpreted with the same confirmatory status as the registered primary PFS endpoint.
8. Complete Posted Secondary Analysis Table
The following table preserves the full set of statistical analyses posted on ClinicalTrials.gov for PEARL. All 20 posted statistical analyses are represented by the ClinicalTrials.gov record; the listed analyses are secondary quality-of-life time-to-deterioration analyses.
| Scale | Instrument | HR | 95% CI | P-value |
|---|---|---|---|---|
| Global health status / quality of life | EORTC QLQ-C30 Functional Scale | 0.70 | 0.55–0.89 | 0.003 |
| Physical functioning | EORTC QLQ-C30 Functional Scale | 0.62 | 0.50–0.76 | <0.001 |
| Role functioning | EORTC QLQ-C30 Functional Scale | 0.63 | 0.51–0.79 | <0.001 |
| Emotional functioning | EORTC QLQ-C30 Functional Scale | 0.97 | 0.76–1.25 | 0.818 |
| Cognitive functioning | EORTC QLQ-C30 Functional Scale | 0.70 | 0.54–0.89 | 0.004 |
| Social functioning | EORTC QLQ-C30 Functional Scale | 0.62 | 0.49–0.78 | <0.001 |
| Fatigue | EORTC QLQ-C30 Symptom Scale | 0.70 | 0.57–0.86 | 0.001 |
| Nausea and vomiting | EORTC QLQ-C30 Symptom Scale | 0.58 | 0.45–0.76 | <0.001 |
| Pain | EORTC QLQ-C30 Symptom Scale | 0.78 | 0.62–0.99 | 0.042 |
| Dyspnea | EORTC QLQ-C30 Symptom Scale | 1.08 | 0.81–1.44 | 0.599 |
| Insomnia | EORTC QLQ-C30 Symptom Scale | 0.84 | 0.64–1.10 | 0.196 |
| Appetite loss | EORTC QLQ-C30 Symptom Scale | 0.71 | 0.54–0.92 | 0.011 |
| Constipation | EORTC QLQ-C30 Symptom Scale | 1.08 | 0.83–1.41 | 0.564 |
| Diarrhea | EORTC QLQ-C30 Symptom Scale | 0.42 | 0.32–0.55 | <0.001 |
| Financial difficulties | EORTC QLQ-C30 Symptom Scale | 0.78 | 0.56–1.07 | 0.118 |
| Body image | EORTC QLQ-BR23 Functional Scale | 0.95 | 0.74–1.21 | 0.659 |
| Future perspective | EORTC QLQ-BR23 Functional Scale | 1.01 | 0.74–1.39 | 0.928 |
| Systemic side-effects | EORTC QLQ-BR23 Symptom Scale | 0.79 | 0.64–0.98 | 0.029 |
| Breast symptoms | EORTC QLQ-BR23 Symptom Scale | 0.93 | 0.71–1.21 | 0.571 |
| Arm symptoms | EORTC QLQ-BR23 Symptom Scale | 0.84 | 0.66–1.09 | 0.187 |
9. Quality-of-Life Assessment Schedule
10. Statistical Methodology
Kaplan-Meier estimation
Time-to-deterioration endpoints are survival-analysis endpoints because each patient is followed until a defined deterioration event or until follow-up ends without an observed event. Kaplan-Meier estimation is a standard way to describe the event-time distribution while accounting for right censoring.
Here, di represents events at an observed event time and ni represents patients at risk immediately before that time.
The ClinicalTrials.gov record does not provide Kaplan-Meier estimates or reconstructed survival curves for the primary PFS endpoint or the QoL analyses. Therefore, no numerical Kaplan-Meier quantities are added here.
Cox proportional-hazards model
The posted analyses use Cox regression, normalized in the ClinicalTrials.gov record as a Cox proportional-hazards model. This model estimates the relationship between treatment assignment and the instantaneous event rate while allowing the baseline hazard to remain unspecified.
For a binary treatment comparison, the hazard ratio is represented by exp(β). The HR is a relative measure of the modeled event hazard, not a probability that an individual patient will experience an event.
Hazard ratio
A hazard ratio of 0.70, for example, means that the fitted model estimates the instantaneous event hazard in the numerator group at approximately 70% of that in the comparator group, subject to the model and the definition of the event. It does not mean that 30% of patients avoid deterioration or that the median time to deterioration is 30% longer.
Confidence intervals
The posted analyses use two-sided 95% confidence intervals. The interval quantifies statistical uncertainty around the estimated hazard ratio under the model and sampling framework. A confidence interval that includes 1 is compatible with no relative hazard difference under the corresponding interval procedure; it does not prove equality between groups.
Analysis population
The QoL analyses were not described as analyses of all 693 enrolled patients. Instead, the registry specifies a QoL population consisting of enrolled patients with a baseline questionnaire and at least one additional questionnaire. This distinction matters because eligibility for the analysis depends on availability of post-baseline QoL information.
Open-label design
The study was not masked. For patient-reported outcomes, an open-label design can be particularly relevant because knowledge of treatment assignment may influence how participants report symptoms, functioning, or overall health status. This is a design consideration rather than evidence that a particular reported estimate is biased in one direction.
11. Statistical Methods Explained
Why is a Cox model appropriate for these endpoints?
Time to deterioration includes both the occurrence of the event and the time at which it occurs. Some participants may finish observation without experiencing the defined deterioration event, creating right-censored observations. Cox regression is designed for this type of time-to-event information and provides a relative hazard measure that can compare treatment groups without requiring a fully specified baseline hazard distribution.
What does an HR of 0.62 mean?
An HR of 0.62 means that, under the fitted Cox model, the estimated instantaneous event rate for the numerator group is 62% of the comparator's estimated event rate. Equivalently, the estimated relative reduction in the instantaneous hazard is 38%. That statement is about the modeled hazard, not about the percentage of patients who ultimately deteriorate.
Why is the confidence interval important?
The point estimate alone does not describe precision. For the physical-functioning endpoint, the HR is 0.62 and the 95% CI is 0.50–0.76. The interval communicates uncertainty around 0.62. A narrower interval generally provides more precise statistical information than a wider interval, although precision and clinical importance are separate concepts.
Why does the P-value not measure effect size?
The P-value describes how compatible the observed result is with the tested null hypothesis under the specified statistical framework. It does not say how large the treatment effect is. The HR describes the relative effect, while the confidence interval provides information about uncertainty around that effect.
Why does the QoL population matter?
The registry defines the QoL population as patients with the baseline questionnaire and at least one additional questionnaire. That means the secondary analyses are based on a subset selected by availability of QoL data. The results therefore should not automatically be treated as though they were estimated from every randomized participant.
Why does the open-label design matter for quality of life?
Patient-reported outcomes are susceptible to reporting behavior. Because PEARL was unmasked, participants knew their treatment assignment. This does not establish that the QoL findings are biased, but it is a relevant interpretive limitation when the endpoint depends on participant-reported functioning and symptoms.
Why should the secondary analyses not be substituted for PFS?
PFS is the registered primary efficacy endpoint and has a specific definition based on investigator-assessed RECIST version 1.1 progression or death. The posted analyses in the ClinicalTrials.gov record evaluate different endpoints: time to deterioration in QoL scales. A statistically informative secondary endpoint does not become the primary efficacy endpoint simply because its numerical results are available.
12. Multiplicity Across Quality-of-Life Analyses
The ClinicalTrials.gov record contains 20 statistical analyses across multiple EORTC QLQ-C30 and QLQ-BR23 scales. This creates an important interpretive issue: each additional hypothesis test provides another opportunity for a nominally small P-value to occur by chance when many tests are considered together.
| Feature | PEARL the ClinicalTrials.gov record | Statistical implication |
|---|---|---|
| Primary endpoint | PFS | Primary efficacy question |
| Posted statistical analyses | 20 | Multiple secondary comparisons |
| Secondary endpoint family represented | QoL time to deterioration | Multiple scales and domains |
| Effect measure | Hazard ratio | Relative time-to-event effect |
| Confidence interval | Two-sided 95% | Uncertainty for each reported HR |
| Hypothesis type | Superiority | Tests are framed around superiority |
The ClinicalTrials.gov record does not provide a multiplicity-adjustment procedure for the 20 posted QoL analyses. Consequently, the individual P-values should be read as results of their respective statistical analyses rather than automatically interpreted as a familywise-error-controlled collection of confirmatory tests.
13. Missing Data and Censoring
Time-to-deterioration analysis naturally incorporates censoring because not every participant necessarily experiences a deterioration event during observation. The ClinicalTrials.gov record explicitly define the QoL analysis population according to questionnaire availability, requiring the baseline questionnaire and at least one more questionnaire.
The ClinicalTrials.gov record does not specify an imputation method for missing QoL observations. Accordingly, no particular imputation procedure is attributed to PEARL.
Missing data can affect patient-reported outcome analyses because the availability of follow-up questionnaires is part of the analysis population definition. The Cox model can accommodate right censoring, but censoring assumptions still matter. The ClinicalTrials.gov record does not provide enough information to determine whether missing questionnaire data were missing completely at random, missing at random, or missing for reasons related to treatment or patient status.
14. Safety
The ClinicalTrials.gov record reports serious adverse events by arm as affected patients divided by patients at risk. These figures are presented exactly as reported in the registry and should not be converted into different denominators.
| Cohort / treatment | Serious adverse events | Affected / at risk |
|---|---|---|
| Cohort 1: Palbociclib Plus Exemestane | Serious adverse events | 34 / 150 |
| Cohort 2: Palbociclib Plus Fulvestrant | Serious adverse events | 21 / 149 |
| Cohort 1 and 2: Capecitabine | Serious adverse events | 72 / 289 |
The serious-adverse-event counts provide a descriptive safety summary by treatment cohort. They should not be directly compared as though the denominators were identical or as though these figures represented a formal time-to-event safety analysis.
The ClinicalTrials.gov record does not provide confidence intervals, P-values, exposure-adjusted rates, or a formal statistical comparison for serious adverse events. Therefore, no inferential safety comparison is added.
15. Limitations and Interpretation Issues
- Primary PFS result not reported: the registered primary endpoint is PFS, but the ClinicalTrials.gov record does not contain a PFS hazard ratio, confidence interval, or P-value.
- Open-label design: the registry classifies the study as unmasked. This is particularly relevant to patient-reported quality-of-life outcomes because participants know their treatment assignment.
- Quality-of-life analysis population: the QoL analyses use a subset of enrolled patients with a baseline questionnaire and at least one additional questionnaire, rather than simply the full enrollment population.
- Multiple secondary analyses: 20 statistical analyses are reported across multiple QoL domains. Without a registry-reported multiplicity procedure, individual nominal P-values should not automatically be interpreted as independently confirmatory.
- Missing-data information: the ClinicalTrials.gov record does not specify an imputation method for missing QoL observations.
- Censoring: time-to-deterioration analyses depend on assumptions about censoring and the information contributed by patients who do not experience the event during observation.
- Proportional-hazards assumption: Cox hazard ratios are model-based. A single HR is most naturally interpreted under the proportional-hazards framework; the ClinicalTrials.gov record does not report a formal proportional-hazards diagnostic.
- Exploratory interpretation: a result for an individual QoL domain should not be generalized automatically to the primary PFS endpoint or to all dimensions of quality of life.
- Registry limitations: the registry notes that capecitabine outcome was better than initially anticipated, with 9 months compared with 6 months in the protocol assumptions. It also notes that the open-label design may lead to biased interpretations.
- Subtype classification: the registry notes that subtype classification for an exploratory objective was carried out in 70% of patients in the primary tumour and might have changed in metastatic disease.
16. Why This Trial Matters Statistically
PEARL is a useful statistical teaching case because the ClinicalTrials.gov record contains a clear primary time-to-event endpoint but the posted formal statistical analyses available in the ClinicalTrials.gov record is concentrated on secondary patient-reported quality-of-life outcomes. This creates an important distinction between what the trial was designed to answer and which numerical analyses are available in the registry ClinicalTrials.gov record.
| Concept | How it appears in PEARL |
|---|---|
| Randomization | The study uses randomized allocation. |
| Parallel-group design | The registry identifies a parallel design model. |
| Time-to-event endpoint | PFS is the registered primary endpoint. |
| RECIST-based progression | PFS uses investigator assessment with RECIST version 1.1. |
| Cox proportional-hazards model | Used for the posted secondary QoL time-to-deterioration analyses. |
| Hazard ratio | The posted analyses use HR as the effect measure. |
| Confidence interval | Each registry-reported secondary HR has a two-sided 95% CI. |
| Quality-of-life endpoints | Time to deterioration is evaluated across EORTC QLQ-C30 and QLQ-BR23 domains. |
| Analysis population | QoL analyses use patients with baseline and at least one additional QoL questionnaire. |
| Open-label design | No masking is reported. |
| Multiplicity | 20 posted statistical analyses span multiple QoL domains. |
| Missing-data considerations | The ClinicalTrials.gov record identifies a questionnaire-based QoL population but do not specify an imputation procedure. |
17. Reading the Secondary Hazard Ratios Correctly
For the nausea-and-vomiting scale, the HR is 0.58 with a 95% CI of 0.45–0.76. Under the fitted Cox model, the estimated instantaneous hazard of the defined deterioration event is lower in the palbociclib-plus-endocrine-therapy group relative to capecitabine.
The HR does not say that 58% of patients avoided deterioration, nor does it mean that the time to deterioration increased by exactly 42%. It is a relative hazard measure.
For future perspective, the HR is 1.01 with a 95% CI of 0.74–1.39 and P = 0.928. The point estimate is close to 1, while the interval extends both below and above 1. The registry-reported analysis therefore does not establish a clear difference in the modeled deterioration hazard for this scale.
For dyspnea, the HR is 1.08 with a 95% CI of 0.81–1.44 and P = 0.599. The point estimate is above 1, but the confidence interval includes 1. The appropriate interpretation is not that palbociclib plus endocrine therapy definitively increases deterioration; rather, the registry-reported analysis does not provide clear statistical evidence of a difference for this endpoint.
18. Primary Endpoint vs Quality-of-Life Endpoints
| Feature | Primary PFS | Posted QoL analyses |
|---|---|---|
| Role | Registered primary endpoint | Secondary analyses |
| Event | First documented RECIST version 1.1 progression or death | Time to deterioration in specified EORTC QoL scales |
| Starting point | Randomization | QoL assessment schedule specified by the registry |
| Assessment | Investigator assessment | Questionnaire-based QoL assessment |
| Analysis population reported | Not specified in the ClinicalTrials.gov record | QoL population with baseline and at least one more questionnaire |
| Posted effect estimate | Not reported | Hazard ratio |
| Posted CI | Not reported | Two-sided 95% |
| Posted P-value | Not reported | Reported for each listed analysis |
This distinction is central to interpreting the registry results. A trial can have a primary efficacy question and a large collection of secondary patient-reported outcomes. The statistical evidence from one family should not be substituted for the other.
19. What a Hazard Ratio Does — and Does Not — Mean
A hazard ratio compares estimated instantaneous event rates under a time-to-event model. An HR of 0.70 corresponds to an estimated 30% lower instantaneous hazard for the numerator group relative to the comparator, provided the proportional-hazards interpretation is appropriate.
A hazard ratio is not the probability that an individual patient will deteriorate, remain progression-free, or benefit. It is also not a direct substitute for an absolute risk difference or a difference in median event time.
The HR describes the estimated magnitude and direction of the relative treatment effect. The P-value addresses compatibility with a specified null hypothesis. The two quantities should always be interpreted separately.
The 95% confidence interval gives a range of parameter values compatible with the statistical model and sampling process at the stated confidence level. It does not describe the range of individual patient outcomes.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registry-reported secondary analyses estimate hazard ratios for time to deterioration across multiple EORTC quality-of-life domains. Some estimates are below 1 with confidence intervals that exclude 1, while others have confidence intervals that include 1.
Clinical interpretation
These analyses describe specific patient-reported functioning and symptom endpoints. They should be interpreted domain by domain and should not be converted into a global claim about the registered primary PFS endpoint.
21. Statistical Lessons From the Posted Results
The PEARL registry results demonstrate several recurring principles of clinical-trial statistics.
- Start with the endpoint definition: the meaning of an HR depends on exactly what event is being modeled.
- Separate primary and secondary endpoints: PFS is the registered primary endpoint, whereas the registry-reported numerical analyses concern QoL deterioration.
- Use the analysis population correctly: the QoL analyses are restricted to patients with sufficient questionnaire information.
- Read the entire confidence interval: the point estimate alone is not enough to characterize uncertainty.
- Do not equate P < 0.05 with a large effect: statistical significance and effect magnitude are distinct concepts.
- Do not equate P > 0.05 with equivalence: a nonsignificant test does not establish that treatment groups are identical.
- Account for multiplicity: 20 secondary analyses across multiple QoL domains create a broader inferential context than any one P-value conveys.
- Consider open-label reporting: patient-reported outcomes can be influenced by awareness of treatment assignment.
- Distinguish censoring from missing questionnaires: Cox methods address time-to-event censoring, but the ClinicalTrials.gov record does not establish a particular missing-data mechanism or imputation method.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: PEARL, NCT02028507. The registry is the source for the trial design, registered endpoint definitions, posted statistical analyses, safety counts, and registry limitations represented on this page.
- PubMed record: PMID 33385521.
- PubMed record: PMID 35429901.
- PubMed record: PMID 34425406.
- PubMed record: PMID 36749874.
Continue with the underlying statistical methods
Explore the survival-analysis concepts behind PEARL, including hazard ratios, Cox proportional-hazards models, confidence intervals, Kaplan-Meier estimation, randomization, P-values, and time-to-event endpoints.
25. Record Summary
PEARL is a randomized phase 3, parallel, unmasked trial with 693 enrolled participants evaluating palbociclib with endocrine therapy versus capecitabine in HR+/HER2- metastatic breast cancer with resistance to aromatase inhibitors. The registered primary endpoint is investigator-assessed Progression-Free Survival, defined from randomization to the first documented RECIST version 1.1 progression or death from any cause. The ClinicalTrials.gov record identifies Cox proportional-hazards modeling and hazard ratios as the trial's statistical framework, but they do not provide a formal numerical PFS analysis.
The numerical statistical ClinicalTrials.gov record are secondary time-to-deterioration analyses of EORTC QLQ-C30 and QLQ-BR23 quality-of-life domains. These analyses consistently use Cox regression, hazard ratios, two-sided 95% confidence intervals, and a QoL analysis population requiring baseline and at least one additional questionnaire. Some domains have HRs below 1 with confidence intervals excluding 1, while others have estimates near or above 1 with intervals that include 1.
The most important statistical lesson is therefore not to collapse these different endpoints into one result. PFS, time to deterioration in physical functioning, time to deterioration in pain, and time to deterioration in systemic side-effects are different estimands. Each requires its own endpoint definition, analysis population, effect measure, uncertainty assessment, and multiplicity context.