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
CompassHER2 is a phase 2 clinical trial sponsored by ECOG-ACRIN Cancer Research Group that evaluates recurrence-free survival in patients with invasive breast carcinoma and specified stage II or IIIA breast cancer. The registry describes a non-randomized, parallel, unmasked treatment study with 2175 participants and two arms.
| Feature | CompassHER2 |
|---|---|
| Official title | CompassHER2-pCR: Decreasing Chemotherapy for Breast Cancer Patients After Pre-surgery Chemo and Targeted Therapy |
| Phase | Phase 2 |
| Status | Active, not recruiting |
| Allocation | Non-randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 2175 |
| Number of arms | 2 |
| Lead sponsor | ECOG-ACRIN Cancer Research Group |
| Sponsor type | Network |
| Start date | March 13, 2020 |
| Primary completion date | October 31, 2027 |
| ClinicalTrials.gov | NCT04266249 |
2. Clinical Question
The registry describes a treatment study of breast cancer patients after pre-surgery chemotherapy and targeted therapy, with the central efficacy endpoint being recurrence-free survival. The statistical question is therefore centered on how recurrence-free survival is defined and estimated over the specified follow-up period.
Population
Patients with invasive breast carcinoma and registry-defined anatomic or prognostic stage II, stage IIA, stage IIB, or stage IIIA breast cancer.
Interventions
The registry lists docetaxel, nab-paclitaxel, paclitaxel, pertuzumab, trastuzumab, and trastuzumab emtansine, together with lumpectomy, mastectomy, and radiation therapy.
Design comparison
The trial has two arms and a parallel design, but allocation is explicitly recorded as non-randomized.
Primary question
How is recurrence-free survival estimated through up to 3 years after end of treatment, both in the overall analysis population and according to pretreatment clinical stage and estrogen receptor status?
3. Trial Design
The non-randomized allocation is one of the most important statistical features of the design. Unlike a randomized comparison, treatment assignment is not balanced by randomization. Consequently, observed differences between arms can be associated with differences in patient characteristics or other factors that are related to treatment assignment.
Interventions recorded in the registry
The registry lists the following interventions and procedures: docetaxel, lumpectomy, mastectomy, nab-paclitaxel, paclitaxel, pertuzumab, radiation therapy, trastuzumab, and trastuzumab emtansine. The trial title identifies the clinical context as decreasing chemotherapy after pre-surgery chemotherapy and targeted therapy.
| Intervention category | Registry-listed intervention |
|---|---|
| Drug | Docetaxel |
| Procedure | Lumpectomy |
| Procedure | Mastectomy |
| Drug | Nab-paclitaxel |
| Drug | Paclitaxel |
| Biological | Pertuzumab |
| Radiation | Radiation Therapy |
| Biological | Trastuzumab |
| Biological | Trastuzumab Emtansine |
4. Endpoints
The registered primary endpoint is recurrence-free survival (RFS). The registry defines the time frame as up to 3 years after end of treatment.
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Recurrence-free survival (RFS) | Events include recurrence of ipsilateral invasive breast tumor, recurrence of locoregional invasive breast tumor, and distant recurrence. | Up to 3 years after end of treatment |
What counts as an RFS event?
The endpoint is event-based. The registry identifies three event categories: recurrence of an ipsilateral invasive breast tumor, recurrence of a locoregional invasive breast tumor, and distant recurrence. The endpoint is therefore broader than a single anatomical recurrence category.
For a time-to-event endpoint, patients who have not experienced a registered event by the end of their available follow-up may contribute information through the time at which their observation ends. This is the role of censoring in survival analysis.
The registry specifies the RFS time frame as up to 3 years after end of treatment. The statistical analysis therefore concerns the estimated recurrence-free survival experience over that follow-up period.
5. Planned Analysis
No formal statistical results are posted in the ClinicalTrials.gov record. The registry states that the RFS survival curve will be estimated using the Kaplan-Meier method in the overall analysis population and by pretreatment clinical stage and estrogen receptor (ER) status. The registry also specifies use of the Greenwood method to estimate the 95% confidence interval for the 3-year rate.
Primary statistical estimand
Estimated from the Kaplan-Meier survival curve.
95% confidence interval estimated using the Greenwood method.
For a conventional Kaplan-Meier analysis, the survival curve estimates the probability of remaining event-free through each time point while accounting for right-censored observations. The 3-year RFS rate is obtained from the estimated survival function at the 3-year time point.
Overall analysis population
The registry specifies Kaplan-Meier estimation in the overall analysis population.
Clinical stage
The registry specifies estimation by pretreatment clinical stage.
Estrogen receptor status
The registry specifies estimation by pretreatment ER status.
Uncertainty
The 95% confidence interval for the 3-year rate is to be estimated using Greenwood's method.
6. Statistical Methodology
Kaplan-Meier estimation
The Kaplan-Meier method is designed for time-to-event outcomes such as recurrence-free survival. It estimates the probability of remaining event-free as follow-up progresses, while allowing patients to have different lengths of observable follow-up.
Here, di represents the number of events at a particular event time and ni represents the number of participants at risk immediately before that time.
In this trial, the Kaplan-Meier curve is the natural representation of the registered RFS endpoint because recurrence can occur at different times and some participants may have follow-up that ends before a recurrence is observed.
Greenwood confidence interval
The registry specifically identifies the Greenwood method for estimating the 95% confidence interval for the 3-year RFS rate. Greenwood's variance estimate uses the event counts and numbers at risk across the Kaplan-Meier curve to quantify uncertainty in the estimated survival probability.
The resulting variance estimate is used to construct a confidence interval around the Kaplan-Meier survival estimate. The registry specifically applies this approach to the 3-year rate.
Analysis by clinical stage and ER status
The registry specifies that the survival curve will be estimated not only in the overall analysis population but also by pretreatment clinical stage and estrogen receptor status. These analyses describe recurrence-free survival across clinically defined groups.
A subgroup-specific Kaplan-Meier curve is descriptive of the event-free experience within that subgroup. A difference between subgroup-specific curves should not automatically be interpreted as evidence that clinical stage or ER status modifies a treatment effect, particularly in a non-randomized study. A formal effect-modification question requires a prespecified comparative model and an appropriate interaction assessment.
No randomized treatment contrast is specified by the allocation field
The registry records allocation as non-randomized. This matters because the principal statistical protection reported by randomization is the balancing of measured and unmeasured prognostic factors in expectation. Without randomization, an observed difference between arms may reflect differences in the patients entering those arms as well as differences associated with the treatments themselves.
7. Statistical Methods Explained
What is recurrence-free survival?
Recurrence-free survival is a time-to-event endpoint. In this trial, the registered event definition includes ipsilateral invasive breast tumor recurrence, locoregional invasive breast tumor recurrence, and distant recurrence. The endpoint therefore combines several clinically relevant recurrence pathways into one event-time analysis.
Why use Kaplan-Meier analysis?
Patients in a clinical trial are not necessarily observed for exactly the same amount of time. Some may complete the planned observation window without an event, while others may have their observation end earlier. Kaplan-Meier estimation incorporates these differing follow-up times rather than treating all participants as if they had identical observation periods.
What does the 3-year RFS rate represent?
The 3-year RFS rate is the estimated probability of remaining free of a qualifying recurrence through the 3-year time point, according to the Kaplan-Meier survival curve. It is an absolute time-specific measure rather than a hazard ratio or a relative risk.
Why is Greenwood's method used?
A Kaplan-Meier estimate is subject to sampling uncertainty. Greenwood's method provides an estimate of the variance of the Kaplan-Meier survival probability, which can then be used to construct a confidence interval around the estimated 3-year RFS rate.
Why does the non-randomized design matter?
In a randomized trial, treatment assignment is determined by the randomization process. Here, the registry records the allocation as non-randomized. Consequently, differences between treatment groups may be confounded by factors associated with treatment selection, even if those factors are not measured or reported.
What does a confidence interval around the RFS rate tell us?
The 95% confidence interval describes statistical uncertainty around the estimated population-level RFS rate under the specified estimation framework. A narrower interval generally indicates greater precision, while a wider interval indicates more uncertainty. It does not describe the range of outcomes that individual patients will experience.
8. Analysis of Clinical Stage and ER Status
The registry specifically identifies pretreatment clinical stage and estrogen receptor status as dimensions for Kaplan-Meier estimation. These variables are therefore part of the planned descriptive statistical presentation of recurrence-free survival.
| Planned view | Statistical approach | Purpose |
|---|---|---|
| Overall analysis population | Kaplan-Meier survival curve | Estimate recurrence-free survival in the overall analysis population |
| Pretreatment clinical stage | Kaplan-Meier survival curves by stage | Describe recurrence-free survival according to pretreatment clinical stage |
| Pretreatment ER status | Kaplan-Meier survival curves by ER status | Describe recurrence-free survival according to pretreatment estrogen receptor status |
| 3-year RFS rate | Kaplan-Meier estimate with Greenwood 95% CI | Quantify the estimated recurrence-free survival probability at 3 years |
These planned analyses are primarily about describing the recurrence-free survival distribution and its uncertainty. Because the registry records the trial as non-randomized, comparisons between treatment arms require particular care: a difference in observed RFS does not, by itself, establish that treatment assignment caused the difference.
9. Censoring and Follow-up
Kaplan-Meier analysis depends on censoring assumptions. A participant who has not experienced a qualifying recurrence by the end of observed follow-up can contribute information up to that point without being counted as an event.
The important statistical distinction is between an event and the end of observation. Censoring is not equivalent to recurrence. It indicates that the participant's event status is not observed beyond the censoring time within the relevant follow-up framework.
Event
A recurrence meeting the registered RFS event definition: ipsilateral invasive, locoregional invasive, or distant recurrence.
Censoring
Observation ends without a qualifying recurrence being recorded during the observed period.
Time horizon
The registered RFS time frame is up to 3 years after end of treatment.
Statistical consequence
Kaplan-Meier estimation allows participants with different observed follow-up durations to contribute information appropriately.
10. Why the Non-Randomized Design Matters Statistically
The allocation field is explicitly recorded as NON_RANDOMIZED. This is not a minor descriptive detail; it changes the interpretation of any between-arm comparison.
Randomization creates comparability between treatment groups in expectation by assigning treatment independently of baseline prognostic characteristics. In a non-randomized design, treatment assignment can instead be related to clinical characteristics, treatment selection, institutional practice, or other factors.
Suppose two treatment groups later display different recurrence-free survival curves. The statistical observation would be real as a comparison of the observed groups, but its causal interpretation would be more limited. The difference could contain both treatment-associated effects and differences in the underlying patient populations.
11. What the 95% Confidence Interval Means
The registered primary endpoint calls for a 95% confidence interval around the 3-year RFS rate using the Greenwood method. This interval is intended to quantify uncertainty around the estimated survival probability.
A confidence interval describes the statistical precision of an estimated 3-year RFS rate. If the interval is relatively narrow, the estimate is more precisely determined within the statistical framework. If it is wide, there is greater uncertainty about the underlying population value.
The interval is not a prediction interval for individual patients and does not mean that 95% of patients will have outcomes falling inside its numerical limits.
Time-to-event precision depends on the information accumulated over follow-up. The number and timing of events, together with the number remaining at risk, influence the statistical uncertainty of the Kaplan-Meier estimate.
12. Primary Completion and Trial Timing
Trial start
The registry records March 13, 2020 as the study start date.
Active, not recruiting
The ClinicalTrials.gov record currently identifies the study as active, not recruiting.
Primary completion
The registry lists October 31, 2027 as the primary completion date.
The timing is important when interpreting the absence of posted efficacy results. A trial can have a registered endpoint and a planned analysis without having a posted statistical result available in the registry.
13. Limitations
- No posted statistical results: the ClinicalTrials.gov record contains the registered primary endpoint and its planned statistical approach but no posted statistical analyses.
- Non-randomized allocation: treatment groups are not created through random assignment, so between-arm differences can be affected by confounding and selection.
- No arm-specific enrollment counts: the registry provides total enrollment of 2175 and identifies two arms, but it does not provide arm-specific enrollment counts in the information used here.
- No allocation ratio: the registry identifies allocation as non-randomized and does not specify a numerical allocation ratio.
- No numerical RFS estimates: the registry does not provide a posted 3-year RFS estimate or its confidence interval.
- No event counts: the registry information does not provide the number of RFS events required to reproduce a Kaplan-Meier analysis.
- No detailed subgroup estimates: clinical stage and ER status are identified as analysis dimensions, but numerical subgroup results are not posted.
- No baseline comparison table: the registry information does not provide the baseline characteristics needed to assess numerical balance between the two arms.
- Limited causal interpretation: a non-randomized parallel design requires caution when interpreting observed treatment-group differences as causal effects.
14. Why This Trial Matters Statistically
CompassHER2 is a useful statistical teaching case because it combines a clinically important time-to-event endpoint with a non-randomized treatment design. That combination illustrates a central principle of clinical-trial statistics: the method used to estimate an outcome and the design used to assign treatment answer different questions.
| Concept | How it appears in CompassHER2 |
|---|---|
| Time-to-event analysis | Recurrence-free survival is the registered primary endpoint. |
| Kaplan-Meier estimation | The registry specifies Kaplan-Meier estimation of the RFS survival curve. |
| Greenwood variance | The registry specifies Greenwood estimation for the 95% confidence interval of the 3-year RFS rate. |
| Absolute survival probability | The primary endpoint focuses on the estimated RFS rate at the 3-year time point. |
| Censoring | Time-to-event methodology accommodates participants whose observation ends without a recorded recurrence. |
| Subgroup estimation | RFS curves are planned by pretreatment clinical stage and ER status. |
| Non-randomized design | Allocation is recorded as non-randomized, affecting the causal interpretation of between-arm comparisons. |
| Parallel design | The trial is recorded as a parallel design with two arms. |
15. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The registered primary analysis estimates recurrence-free survival using Kaplan-Meier methods and quantifies uncertainty in the 3-year rate with a Greenwood-based 95% confidence interval.
Clinical interpretation
The endpoint is intended to capture recurrence of ipsilateral invasive breast tumor, locoregional invasive breast tumor, or distant recurrence during the registered follow-up period.
Design interpretation
The trial is non-randomized, so an observed difference between treatment groups cannot be interpreted in the same way as a difference produced by randomized allocation.
Evidence interpretation
The registry currently provides the planned endpoint and statistical method rather than numerical efficacy results.
16. What a Future RFS Analysis Would Show
A completed primary RFS analysis would ordinarily present the estimated Kaplan-Meier curve over the registered follow-up period, the number of participants at risk at relevant time points, the qualifying recurrence events, and the estimated RFS rate at 3 years with its 95% confidence interval.
Because the registry specifically calls for analyses by pretreatment clinical stage and ER status, a complete statistical presentation would also show how the estimated recurrence-free survival curves differ across those prespecified clinical dimensions.
For the two treatment arms, interpretation would additionally need to account for the non-randomized allocation. A numerical difference between the curves would describe the observed difference between groups, while a causal treatment interpretation would require consideration of the factors associated with treatment selection.
This sequence mirrors the statistical structure specified by the registry for the primary recurrence-free survival endpoint.
17. Sources
- ClinicalTrials.gov: CompassHER2 (NCT04266249).
- PubMed: CompassHER2 publication record.
Continue through the Clinical Biostats trial library
Clinical trial analyses connect study design, endpoints, survival methods, uncertainty, and interpretation into a structured statistical framework.
18. Record Summary
CompassHER2 is a phase 2, non-randomized, parallel clinical trial with two arms and an enrollment of 2175 participants. The registered primary endpoint is recurrence-free survival, defined through qualifying ipsilateral invasive, locoregional invasive, or distant recurrence and assessed for up to 3 years after end of treatment.
The statistical framework specified by the registry is centered on Kaplan-Meier estimation of the RFS survival curve in the overall analysis population and according to pretreatment clinical stage and ER status. Greenwood's method is specified for the 95% confidence interval around the 3-year RFS rate.