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
ACOSOG Z0011 was a randomized, parallel-group phase 3 treatment trial enrolling 605 women with stage I or stage IIA breast cancer. The registry identifies two arms and the interventions axillary lymph node dissection and whole breast irradiation. The registered primary endpoints were overall survival, morbidity, and disease-free survival, each with a time frame of up to 10 years.
| Feature | ACOSOG Z0011 |
|---|---|
| Trial acronym | ACOSOG Z0011 |
| NCT identifier | NCT00003855 |
| Phase | Phase 3 |
| Status | COMPLETED |
| Condition | Breast Cancer |
| Enrollment | 605 |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Number of arms | 2 |
| Lead sponsor | Alliance for Clinical Trials in Oncology |
| Sponsor type | OTHER |
2. Clinical Question
The registry describes a phase 3 randomized treatment trial in women who have stage I or stage IIA breast cancer. The central statistical question is how outcomes differ between the two randomized treatment arms involving the registered interventions of axillary lymph node dissection and whole breast irradiation.
Population
Women who have stage I or stage IIA breast cancer.
Interventions
The registry identifies axillary lymph node dissection as a procedure and whole breast irradiation as a radiation intervention.
Comparator
The record identifies two randomized arms but does not provide enough arm-level treatment detail here to specify the complete regimen assigned to each arm.
Primary question
How do overall survival, morbidity, and disease-free survival compare across the randomized treatment groups over a period of up to 10 years?
3. Trial Design
A parallel randomized design creates two treatment groups that can be compared prospectively. Randomization is the key design feature for causal comparison because treatment assignment is determined by the trial rather than by subsequent patient or investigator choice.
The absence of masking means that the trial is registered as having no masking. That feature is particularly relevant when outcomes may be affected by treatment knowledge. For time-to-event endpoints such as overall survival, death is objectively defined, whereas other outcomes can potentially be more sensitive to how treatment is known or assessed.
4. Trial Timeline
Trial start
The registry lists April 1999 as the study start date.
Primary completion
The registry lists August 2007 as the primary completion date.
Registered endpoint follow-up
Overall survival, morbidity, and disease-free survival are each registered with a time frame of up to 10 years.
5. Endpoints
| Registered primary endpoint | Time frame | Statistical role |
|---|---|---|
| Overall survival | Up to 10 years | Primary time-to-event endpoint |
| Morbidity | Up to 10 years | Primary clinical outcome requiring an appropriate comparison of morbidity between randomized groups |
| Disease-free survival | Up to 10 years | Primary time-to-event endpoint |
Overall survival
Overall survival is registered with a time frame of up to 10 years. For a conventional survival endpoint, the event is death, and patients who have not experienced the event at their last observation contribute follow-up until censoring.
Morbidity
Morbidity is registered as a primary endpoint with a time frame of up to 10 years. The registry information available here does not provide a more detailed operational definition of morbidity. Consequently, the appropriate statistical analysis would depend on whether morbidity is represented as a binary outcome, a count, a time-to-event outcome, or another prespecified measure.
Disease-free survival
Disease-free survival is registered as a primary endpoint with a time frame of up to 10 years. Disease-free survival is generally analyzed as a time-to-event endpoint, with the event definition determined by the trial's prespecified endpoint criteria.
6. Planned Analysis
No formal statistical analyses were posted to ClinicalTrials.gov. The registry does, however, identify the three primary endpoints and their common time frame of up to 10 years. The appropriate statistical framework can therefore be described without assigning an outcome to either randomized group.
Overall survival
Overall survival would ordinarily be analyzed as a time-to-event endpoint. The Kaplan-Meier method provides an estimate of the survival function over time, allowing patients with incomplete follow-up to contribute information until censoring.
Here, di represents the number of events at time ti, while ni represents the number of patients at risk immediately before that time.
A comparison of randomized groups for overall survival would commonly use a log-rank test, with a Cox proportional-hazards model providing a hazard ratio and confidence interval when the proportional-hazards framework is appropriate.
Disease-free survival
Disease-free survival is also a time-to-event endpoint. A Kaplan-Meier analysis can estimate the disease-free survival curve, while a log-rank comparison and Cox model are commonly used to compare randomized groups and quantify the relative event rate.
Morbidity
The statistical method for morbidity cannot be specified from the registry description alone because the operational definition and measurement structure are not included. If morbidity were defined as the occurrence of at least one event by a fixed time, a risk difference, risk ratio, or odds ratio could be appropriate. If the timing of morbidity events were central to the endpoint, a time-to-event method could instead be required.
Randomized analysis principle
For the primary treatment comparison, preserving the randomized groups is generally important because randomization is what establishes the comparability of the groups at baseline in expectation. An intention-to-treat analysis would ordinarily analyze participants according to their randomized assignment rather than allowing later treatment changes to redefine the comparison.
7. Statistical Methods Explained
Why use Kaplan-Meier estimation for overall survival?
Not every participant will necessarily experience death during the observation period. Kaplan-Meier estimation handles this right censoring while retaining the information available before a patient's last known follow-up. It therefore provides a way to estimate survival over the registered follow-up period without treating every censored patient as if death had occurred.
What does a hazard ratio mean in this type of trial?
A hazard ratio compares the modeled instantaneous event rates between two treatment groups over time. In general, an HR below 1 indicates a lower estimated event rate in the numerator group, whereas an HR above 1 indicates a higher estimated event rate.
The hazard ratio is a relative time-to-event measure. It is not the same as a relative risk, an absolute risk difference, or a probability that an individual patient will experience the event.
Why is disease-free survival different from overall survival?
Overall survival concerns survival itself, whereas disease-free survival captures a disease-related event process defined by the protocol. The two endpoints can therefore provide different information about treatment effects and should not be treated as interchangeable outcomes.
Why does the morbidity endpoint require its exact definition?
Morbidity can be measured in several statistically different ways. A simple proportion of patients with morbidity is analyzed differently from a count of events per patient or a time-to-first-event endpoint. Without the operational definition, choosing one analysis would risk attaching a method to an endpoint that may have been defined differently.
Why does randomization matter for statistical interpretation?
Randomization separates treatment assignment from the prognostic factors that might otherwise influence treatment selection. This allows differences observed between randomized groups to be interpreted as treatment-group differences under the trial's design, subject to the usual issues of adherence, missing follow-up, subsequent treatment, and endpoint measurement.
Why should overall survival and disease-free survival be interpreted separately?
The endpoints measure different clinical processes. Disease-free survival can capture events that occur before death, while overall survival focuses on death from any cause. A treatment comparison can therefore have different statistical characteristics for the two endpoints even when both are analyzed using time-to-event methods.
8. Statistical Interpretation of Time-to-Event Endpoints
A Kaplan-Meier curve describes the estimated probability of remaining event-free over time. The vertical separation between curves can vary during follow-up, so a single time-point estimate does not completely describe the treatment comparison.
A Cox-model hazard ratio summarizes a relative difference in event rates under the model. It does not mean that the same percentage difference applies to every patient, and it does not directly provide the absolute number of patients who benefit.
A confidence interval quantifies statistical uncertainty around an estimated treatment effect. A narrower interval indicates greater precision than a wider interval, all else being equal. The interval should be interpreted together with the effect estimate rather than treated as a substitute for the estimate itself.
A p-value addresses the compatibility of the observed data with a specified null hypothesis under the statistical model. It does not measure the magnitude of a treatment effect, the clinical importance of an effect, or the probability that a treatment hypothesis is true.
9. Multiplicity and Three Primary Endpoints
ACOSOG Z0011 registers three primary endpoints: overall survival, morbidity, and disease-free survival. Multiple primary endpoints create a multiplicity issue because evaluating several hypotheses can increase the chance of obtaining at least one statistically significant result by chance if each hypothesis is tested independently at the same nominal significance level.
| Endpoint | Statistical type | Key interpretation issue |
|---|---|---|
| Overall survival | Time-to-event | Requires attention to censoring and the time-to-event model. |
| Morbidity | Depends on endpoint definition | The measurement scale determines the appropriate comparison. |
| Disease-free survival | Time-to-event | Requires a prespecified event definition and censoring rules. |
The registry information available here does not specify an alpha-allocation strategy, hierarchical testing procedure, or other formal multiplicity adjustment. Those details are important when interpreting a trial with multiple primary endpoints because the evidentiary threshold for a family of hypotheses depends on the prespecified statistical design.
10. Censoring and Follow-Up
Both overall survival and disease-free survival are naturally analyzed over time. Patients may reach the end of observation without experiencing the endpoint, producing right-censored observations. Proper handling of censoring allows those patients to contribute their observed follow-up without treating the censoring time as an event.
Administrative censoring
A participant can remain event-free when the planned observation period ends. The patient contributes information through the last available follow-up.
Loss to follow-up
If a participant's outcome becomes unknown before the end of follow-up, the statistical analysis must apply prespecified censoring rules.
Event definition
The meaning of an event differs between overall survival and disease-free survival, making endpoint definitions central to interpretation.
Time frame
The registry specifies a follow-up time frame of up to 10 years for all three primary endpoints.
11. Proportional-Hazards Assumption
If a Cox proportional-hazards model is used for overall survival or disease-free survival, the hazard ratio is based on a proportional-hazards framework. This assumes that the relative hazards are reasonably stable over time on the scale required by the model.
When proportional hazards do not hold, a single hazard ratio can compress a more complicated time-varying treatment effect into one summary number. In that situation, Kaplan-Meier estimates at clinically meaningful times, restricted mean survival time, or other prespecified measures can provide additional information.
12. Intention-to-Treat Analysis
For a randomized clinical trial, the primary efficacy comparison is ordinarily anchored to the randomized treatment assignment. This is the intention-to-treat principle. It preserves the comparison created by randomization and avoids allowing post-randomization treatment decisions to redefine the treatment groups.
The registry identifies the allocation as RANDOMIZED but does not provide the analysis populations or a detailed statistical analysis plan. Consequently, the exact population used for any eventual efficacy analysis cannot be characterized from the registry fields presented here.
13. Missing Data and Imputation
Missingness can affect both survival and morbidity analyses, but the statistical consequences depend on the endpoint and on why observations are missing.
For time-to-event outcomes, incomplete observation is often handled through censoring rather than conventional single-value imputation. For morbidity, the appropriate approach depends on whether the endpoint is binary, repeated, count-based, or time-to-event.
The ClinicalTrials.gov record does not specify a missing-data or imputation strategy for the registered primary endpoints. That distinction matters because different assumptions about missing observations can lead to different estimands and different uncertainty around treatment effects.
14. Stratification and Covariate Adjustment
The registry identifies the trial as randomized and parallel-group but does not specify randomization strata or covariates for an adjusted analysis. Therefore, a specific stratified Cox model or covariate-adjusted treatment estimate cannot be assigned to the registered design from the information available here.
In a randomized trial, an unadjusted comparison can provide a direct estimate of the difference between randomized groups. Covariate adjustment can improve precision when important baseline prognostic variables are prespecified and appropriately modeled, but the adjustment strategy should be defined before examining treatment outcomes.
15. Safety and Morbidity
Morbidity is itself one of the three registered primary endpoints, with a time frame of up to 10 years. That makes morbidity a central part of the statistical evaluation rather than merely a supplementary safety outcome.
| Outcome domain | Registry status | Statistical consideration |
|---|---|---|
| Overall survival | Primary endpoint; up to 10 years | Time-to-event comparison |
| Morbidity | Primary endpoint; up to 10 years | Method depends on the endpoint definition and measurement scale |
| Disease-free survival | Primary endpoint; up to 10 years | Time-to-event comparison |
The ClinicalTrials.gov record does not include serious adverse-event counts or arm-specific safety estimates. No arm-level morbidity estimates are reported here because the registry record does not provide numerical results for the endpoint.
16. Results Status
This distinction is important. A completed trial and a statistically analyzed trial are not synonymous within a registry interface. Completion describes study status, whereas a posted statistical analysis would provide numerical estimates, uncertainty measures, and formal hypothesis-test results.
17. How the Trial Design Determines the Analysis
| Design feature | Statistical consequence |
|---|---|
| Randomized allocation | Supports comparison of outcomes according to assigned treatment groups. |
| Parallel design | Participants remain associated with one randomized treatment pathway rather than serving sequentially as their own randomized comparator. |
| Two arms | Creates a direct between-group comparison for the registered endpoints. |
| No masking | Treatment assignment is not masked; endpoint-specific susceptibility to assessment bias depends on how each outcome is measured. |
| Three primary endpoints | Requires explicit consideration of multiplicity and endpoint interpretation. |
| Up to 10 years of follow-up | Allows time-to-event endpoints to be evaluated over an extended observation period. |
18. Limitations
- No posted statistical analyses: the registry contains no numerical treatment-effect estimates, confidence intervals, or p-values in its posted statistical-analysis information.
- Limited arm-level detail: the registry information identifies two arms and the two interventions but does not specify the complete treatment assignment for each arm in the information presented here.
- Morbidity definition: morbidity is registered as a primary endpoint, but the available record does not provide the operational definition needed to select one specific statistical model.
- No analysis-population details: the registry information presented here does not identify an intention-to-treat, per-protocol, safety, or other analysis population.
- No stratification details: specific randomization strata and covariates for adjusted analyses are not identified.
- No multiplicity procedure reported: the record identifies three primary endpoints but does not specify an alpha-allocation or multiplicity-adjustment strategy.
- No missing-data strategy reported: the registry does not specify imputation or censoring rules beyond the endpoint time frames provided.
- No numerical safety results: arm-specific serious adverse-event counts and other safety estimates are not provided in the record information presented here.
19. Why This Trial Matters Statistically
ACOSOG Z0011 is a useful statistical teaching case because its registered design combines randomized treatment assignment with three primary outcomes that have different statistical structures. Overall survival and disease-free survival are time-to-event endpoints, while morbidity requires an analysis tailored to its precise definition.
| Concept | How it appears in ACOSOG Z0011 |
|---|---|
| Randomization | Patients were assigned using a randomized allocation. |
| Parallel-group design | The study uses a parallel design with two arms. |
| Time-to-event analysis | Overall survival and disease-free survival are registered primary endpoints. |
| Censoring | Relevant to analyses of survival and disease-free survival over time. |
| Kaplan-Meier estimation | A standard method for estimating survival or disease-free survival in the presence of censoring. |
| Hazard ratio | A standard relative measure for comparing time-to-event hazards between randomized groups. |
| Multiplicity | Three registered primary endpoints require attention to the family of statistical hypotheses. |
| Endpoint definition | Morbidity illustrates why the measurement structure determines the statistical method. |
| Long follow-up | All three primary endpoints have a registered time frame of up to 10 years. |
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The study is a randomized, parallel-group phase 3 trial with two arms and three registered primary endpoints. Two endpoints are time-to-event outcomes, while the appropriate method for morbidity depends on its precise definition.
Clinical interpretation
The clinical domains registered for evaluation are overall survival, morbidity, and disease-free survival in women with stage I or stage IIA breast cancer, with each outcome followed for up to 10 years.
21. What Can and Cannot Be Concluded From the Registry Record
| Question | What the record establishes |
|---|---|
| Was the trial randomized? | Yes. The allocation is listed as RANDOMIZED. |
| Was it a parallel-group study? | Yes. The design model is PARALLEL. |
| How many participants were enrolled? | 605. |
| How many arms were registered? | 2. |
| What were the registered interventions? | Axillary lymph node dissection and whole breast irradiation. |
| What were the primary endpoints? | Overall survival, morbidity, and disease-free survival. |
| What was the endpoint time frame? | Up to 10 years for each registered primary endpoint. |
| Are numerical statistical results posted? | No posted statistical analyses are present in the ClinicalTrials.gov record. |
22. Statistical Reading Guide
A useful way to read a randomized clinical trial such as ACOSOG Z0011 is to move from design to estimand to analysis and then to interpretation.
1. Start with randomization
Identify the randomized groups and determine exactly which treatment contrast is being estimated.
2. Define the endpoint
Determine whether the outcome is a time-to-event measure, a binary outcome, a count, or another endpoint type.
3. Match the method
Use an analysis appropriate to the endpoint structure, including Kaplan-Meier and Cox methods for conventional time-to-event outcomes.
4. Examine uncertainty
Interpret effect estimates together with confidence intervals rather than relying on a p-value alone.
This framework prevents an important statistical mistake: selecting a statistical method based only on the name of an endpoint. The endpoint's operational definition, timing, censoring rules, analysis population, and prespecified hypothesis structure all influence the appropriate analysis.
23. Sources
- ClinicalTrials.gov: ACOSOG Z0011, NCT00003855.
Continue through the Clinical Biostats knowledge graph
Explore statistical methods, clinical-trial design concepts, and analysis workflows related to randomized time-to-event studies.
24. Record Summary
ACOSOG Z0011 is registered as a completed randomized phase 3 parallel-group treatment trial enrolling 605 women with stage I or stage IIA breast cancer. The registry identifies two arms and the interventions axillary lymph node dissection and whole breast irradiation. Its three primary endpoints are overall survival, morbidity, and disease-free survival, each assessed over a time frame of up to 10 years.
From a statistical perspective, the trial illustrates the importance of matching the analysis to the endpoint. Overall survival and disease-free survival are naturally handled as time-to-event outcomes, for which Kaplan-Meier estimation, log-rank comparisons, and Cox proportional-hazards models are standard approaches. Morbidity requires a method determined by its exact operational definition. The presence of three primary endpoints also makes multiplicity an important component of the statistical interpretation.