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
NSABP B-39/RTOG 0413 was a randomized, parallel-design trial in women who had undergone surgery for ductal carcinoma in situ or stage I or stage II breast cancer. The study evaluated radiation approaches including accelerated partial breast irradiation and whole breast irradiation, with in-breast tumor recurrence (IBTR) registered as the primary endpoint.
| Feature | NSABP B-39/RTOG 0413 |
|---|---|
| ClinicalTrials.gov identifier | NCT00103181 |
| Official trial title | Radiation Therapy (WBI Versus PBI) in Treating Women Who Have Undergone Surgery For Ductal Carcinoma In Situ or Stage I or Stage II Breast Cancer |
| Status | COMPLETED |
| Phase | NA |
| Condition | Breast Cancer |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 4216 |
| Number of arms | 2 |
| Lead sponsor | NSABP Foundation Inc |
| Sponsor type | NETWORK |
2. Clinical Question
The central clinical question is how radiation strategies used after surgery compare with respect to in-breast tumor recurrence in women who have undergone surgery for ductal carcinoma in situ or stage I or stage II breast cancer.
Population
Women who have undergone surgery for ductal carcinoma in situ or stage I or stage II breast cancer.
Radiation approaches
The registry lists 3-dimensional conformal accelerated partial breast irradiation, brachytherapy, whole breast irradiation, and MammoSite or another single-entry intracavitary device.
Allocation
Participants were randomized in a parallel two-arm design.
Primary question
How does the randomized radiation comparison relate to the occurrence of in-breast tumor recurrence during the registered follow-up period?
3. Trial Design
Partial breast irradiation approaches
- 3-dimensional conformal accelerated partial breast irradiation
- Brachytherapy
- MammoSite or another single-entry intracavitary device
- Radiation treatment
Whole breast irradiation
- Whole breast irradiation
- Radiation treatment
4. Trial Population and Treatment Structure
The registry identifies the condition as Breast Cancer and specifies a population of women who have undergone surgery for ductal carcinoma in situ or stage I or stage II breast cancer. The intervention description includes several radiation techniques rather than a single named treatment modality.
| Registry element | Recorded description |
|---|---|
| Population condition | Breast Cancer |
| Post-surgical setting | Women who have undergone surgery |
| Disease categories | Ductal carcinoma in situ; stage I breast cancer; stage II breast cancer |
| Intervention category | Radiation |
| Partial-breast techniques listed | 3-dimensional conformal accelerated partial breast irradiation; brachytherapy; MammoSite or other single-entry intracavitary device |
| Whole-breast technique listed | Whole breast irradiation |
This treatment structure is important statistically because the intervention field contains multiple radiation techniques. The appropriate interpretation of any future arm-level estimate depends on exactly how those techniques were grouped within the randomized comparison and how treatment receipt was represented in the final analysis.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| In-breast tumor recurrence (IBTR) | In-breast tumor recurrence (IBTR) | Every 6 months for 5 years; annually thereafter for up to 10 years |
IBTR is the registered primary endpoint. The follow-up schedule is explicitly longitudinal: assessments occur every 6 months for the first 5 years and annually thereafter for up to 10 years.
That structure matters for statistical analysis because recurrence is an event that can occur at different points during follow-up. A time-to-event framework is therefore a natural approach when the underlying dates of recurrence and censoring are available. Such an analysis uses not only whether recurrence occurred, but also when the event occurred and how long participants remained under observation without the event.
For a participant who develops IBTR, follow-up ends at the recurrence time. For a participant who remains free of the event through the available observation period, the follow-up time is censored at the appropriate observation endpoint.
6. Statistical Methodology
Why IBTR naturally leads to survival-analysis methods
The registered primary endpoint is a recurrence event followed over time rather than a single measurement taken once for every participant. In this setting, a time-to-event analysis can preserve information about the timing of recurrence and accommodate participants whose observation ends before recurrence.
Kaplan-Meier estimation
When the time to IBTR is available, the Kaplan-Meier method is commonly used to estimate the probability of remaining free of the event over time. It handles right-censored observations without treating them as if they experienced recurrence.
Here, di is the number of events at time ti and ni is the number at risk immediately before that time. The expression illustrates how the estimated event-free probability changes as observed events accumulate.
Comparing randomized groups
A standard comparison of two randomized time-to-event groups is the log-rank test. It compares the observed pattern of events over follow-up while accounting for the changing number of participants at risk.
Cox proportional-hazards model
A Cox proportional-hazards model is another common approach for a recurrence endpoint. It produces a hazard ratio describing the relative event rate between treatment groups under the model.
A hazard ratio is a relative, model-based measure. It is not the same as a relative risk, an absolute risk difference, or the probability that an individual participant will experience recurrence.
Absolute event probabilities
Time-specific recurrence estimates can complement a hazard ratio. For example, a cumulative incidence or an event-free survival estimate at a prespecified time point provides an absolute description that a relative hazard ratio alone cannot convey.
7. Analysis of a Longitudinal Recurrence Endpoint
The registry's assessment schedule provides an important statistical feature: participants are evaluated repeatedly rather than being observed only once. This creates a longitudinal follow-up structure in which recurrence status can be updated as participants move through the study period.
Early follow-up
IBTR is assessed every 6 months for the first 5 years.
Later follow-up
After the first 5 years, IBTR is assessed annually for up to 10 years.
Event timing
The statistical analysis can use the timing of recurrence rather than reducing every participant to a simple yes/no outcome.
Censoring
Participants without observed recurrence at the end of available follow-up contribute information through their censoring time in a standard time-to-event framework.
The distinction between recurrence status and recurrence timing is particularly important. If two participants both experience IBTR but at different follow-up times, a binary endpoint that simply records whether recurrence occurred would discard information about that difference. Time-to-event methods preserve it.
8. Planned Analysis
The planned statistical focus is the registered endpoint in-breast tumor recurrence (IBTR), assessed every 6 months for 5 years and annually thereafter for up to 10 years.
For an endpoint of this type, a conventional analysis would generally compare the time to IBTR between the randomized treatment groups. Kaplan-Meier methods can describe recurrence-free experience over time, while a log-rank test can provide a formal comparison of the event-time distributions. A Cox proportional-hazards model can estimate a hazard ratio and its confidence interval.
| Statistical question | Typical method for a time-to-IBTR endpoint |
|---|---|
| How does recurrence-free experience change over time? | Kaplan-Meier estimation |
| Do randomized groups differ in their event-time distributions? | Log-rank test |
| What is the relative event rate between groups? | Cox proportional-hazards model |
| How precise is the estimated treatment effect? | Confidence interval for the effect estimate |
| What is the absolute recurrence experience at a specified time? | Time-specific survival or cumulative-incidence estimate, as appropriate to the event definition and competing-event structure |
The appropriate final model depends on the prespecified statistical analysis plan and the structure of the observed event and censoring data. In particular, the proportional-hazards assumption should be assessed before interpreting a single Cox hazard ratio as an adequate summary of the treatment difference over the entire follow-up period.
9. Statistical Methods Explained
What makes IBTR a time-to-event endpoint?
IBTR is a recurrence that can occur at different points after treatment. The registry also specifies repeated follow-up over time. A time-to-event analysis therefore uses the duration until recurrence, rather than only whether recurrence occurred at some point.
Why use Kaplan-Meier estimation?
Not every participant necessarily experiences recurrence during observed follow-up. Kaplan-Meier estimation allows participants who have not experienced IBTR to contribute information until their censoring time without counting them as recurrence events.
What does the log-rank test compare?
The log-rank test compares the observed recurrence experience between randomized groups across follow-up. It evaluates the overall pattern of event times rather than comparing only one arbitrary time point.
What does a hazard ratio mean?
A hazard ratio compares the estimated instantaneous event rates between two groups within a time-to-event model. An HR below 1 indicates a lower estimated event rate for the numerator group, whereas an HR above 1 indicates a higher estimated event rate. The HR is not an absolute recurrence probability.
Why does censoring matter?
A participant may complete available follow-up without experiencing IBTR, or observation may otherwise end before recurrence. Treating every such participant as if recurrence never could occur would distort the analysis. Survival methods instead use the participant's observed follow-up and censor the observation when appropriate.
Why should the proportional-hazards assumption be checked?
A Cox hazard ratio is most naturally interpreted as a common relative hazard over time when the proportional-hazards assumption is reasonable. If the treatment effect changes substantially over follow-up, a single HR can conceal that pattern. Time-specific estimates and survival curves can then provide important complementary information.
Why are confidence intervals important?
A point estimate summarizes the estimated treatment effect, but it does not show how precisely that effect has been estimated. A confidence interval gives an interval representation of statistical uncertainty under the model and sampling framework. It does not describe the range of effects experienced by individual participants.
10. Randomization and Causal Interpretation
Randomization is the central design feature that supports comparison of the two treatment groups. Because treatment assignment is randomized, differences in the primary endpoint can be interpreted within the causal framework created by the randomized comparison, subject to adherence, follow-up, censoring, missing information, and other aspects of the study's execution.
The unmasked design is also statistically relevant. Knowledge of treatment assignment can affect aspects of treatment delivery, follow-up, assessment, or other study processes. The implications depend on how the recurrence endpoint was ascertained and adjudicated, information that is not described in the registry data summarized here.
11. What an Effect Estimate Would Need to Show
For the primary IBTR endpoint, a statistically informative result would ideally include more than a single point estimate. A complete interpretation would consider the treatment-group event experience, the estimated relative effect, an appropriate confidence interval, and the statistical test used for the randomized comparison.
| Result component | What it contributes |
|---|---|
| Number of IBTR events | Shows the amount of observed event information contributing to the analysis. |
| Time-to-event estimate | Describes recurrence-free experience over follow-up. |
| Hazard ratio | Provides a relative measure of the event rate under a Cox model. |
| Confidence interval | Describes uncertainty around the estimated effect. |
| P-value | Addresses the compatibility of the observed data with a specified null hypothesis under the statistical testing framework; it does not measure the size or clinical importance of an effect. |
| Absolute recurrence estimates | Shows the magnitude of recurrence experience on an absolute scale at relevant time points. |
This distinction is important because statistical significance and effect magnitude are different concepts. A small p-value does not by itself establish that an effect is large, while a confidence interval gives information about the precision and plausible range of the estimated effect under the chosen model.
12. Follow-Up Structure and Long-Term Interpretation
Study start
The registry records March 2005 as the study start date.
Six-month assessments
IBTR is scheduled for assessment every 6 months during the first 5 years.
Annual assessments
IBTR is assessed annually thereafter for up to 10 years.
Primary completion
The registry records August 2018 as the primary completion date.
The long follow-up schedule has a direct statistical consequence: the study can capture recurrences that occur well after initial treatment. It also means that the amount of censoring and the number of participants remaining under observation can change substantially over time.
13. Why the Follow-Up Schedule Matters Statistically
Early events
Six-month assessment intervals provide repeated opportunities to identify IBTR during the first 5 years.
Late events
Annual assessments after 5 years extend observation for recurrence through the registered 10-year period.
Risk sets change
In time-to-event analysis, the participants contributing to the risk set can change as events and censoring occur.
Duration matters
A recurrence early in follow-up and a recurrence late in follow-up are different event-time observations even though both are classified as IBTR.
This is why a simple proportion of participants with recurrence can be an incomplete summary when follow-up differs between participants. Survival-analysis methods are designed to account for that unequal observation time.
14. Non-Inferiority and Margin Interpretation
The registry information identifies the study as a randomized comparison of radiation approaches but does not report a non-inferiority margin or a posted statistical analysis. Consequently, the registry record does not establish a numerical margin against which an effect estimate can be compared.
More generally, a non-inferiority analysis is not interpreted simply by asking whether a conventional p-value is greater than 0.05. The treatment-effect confidence interval must be evaluated against a prespecified non-inferiority margin, with the direction of the effect defined in advance.
The margin must be specified before the analysis and interpreted in the clinically appropriate direction. It is not created from the observed data.
For this trial, the registry does not provide a margin, so no numerical non-inferiority conclusion can be drawn from the registry record.
15. Multiplicity and Multiple Radiation Techniques
The intervention description contains more than one partial-breast radiation technique, while the registry identifies two randomized arms. This creates an important statistical distinction between the randomized comparison and the specific treatment techniques used within an intervention category.
| Statistical issue | Why it matters |
|---|---|
| Two randomized arms | The principal treatment comparison is defined at the randomized-arm level. |
| Multiple intervention techniques | Different radiation techniques may have different implementation characteristics even when represented within a broader intervention category. |
| Multiple potential comparisons | If several technique-specific hypotheses are tested, the family of tests must be defined before interpreting p-values. |
| Multiplicity adjustment | When several confirmatory hypotheses are tested, appropriate error control may be needed to maintain the intended overall type I error rate. |
The registry information does not report a multiplicity strategy. A final statistical interpretation should therefore distinguish the primary randomized comparison from any additional technique-specific or exploratory comparisons that might appear in a more detailed analysis plan.
16. Missing Data and Censoring
Longitudinal recurrence studies must distinguish between a participant who has an observed IBTR event and a participant whose observation ends without an observed event. The latter contributes follow-up information but is censored rather than classified as an event.
Censoring
Censoring records the period during which a participant was observed without an IBTR event before follow-up ended.
Incomplete follow-up
Shorter observation does not automatically mean that a participant experienced recurrence.
Informative censoring
Survival methods generally rely on assumptions about the relationship between censoring and future event risk. Violations can bias estimates.
Imputation
The registry information does not report an imputation strategy for missing recurrence information.
For a time-to-event endpoint, treating every missing or incomplete observation as an event would generally produce a different estimand from treating it as censored. The handling of incomplete observations should therefore be prespecified and aligned with the clinical question.
17. Interim Analysis and Sequential Testing
The registry information identifies the primary endpoint and follow-up schedule but does not report an interim-analysis plan, interim efficacy boundary, alpha-spending function, or stopping rule.
In a randomized clinical trial, an interim analysis allows accumulating data to be examined before the final analysis. If formal hypothesis testing is performed repeatedly, the statistical design normally needs to account for those repeated looks so that the overall type I error remains controlled.
Interim look
An interim analysis examines data before the planned final analysis.
Alpha spending
An alpha-spending approach allocates the type I error across planned information times.
Stopping boundary
A prespecified boundary can define when accumulating evidence is sufficiently strong for an early decision.
Interpretive caution
18. Safety and Adverse Events
The registry information identifies the trial's treatment purpose and radiation interventions but does not provide posted statistical analyses or serious-adverse-event counts by randomized arm.
| Safety information | ClinicalTrials.gov record |
|---|---|
| Serious adverse events by arm | Not reported in the ClinicalTrials.gov record |
| Arm-specific adverse-event estimates | Not reported in the ClinicalTrials.gov record |
| Posted statistical safety analyses | None |
For a radiation-treatment trial, safety interpretation is conceptually distinct from IBTR efficacy. A treatment comparison can be described using recurrence outcomes while safety requires its own event definitions, observation windows, denominators, and severity classifications.
19. Bayesian Methods
The registry information does not report a Bayesian analysis for NSABP B-39/RTOG 0413.
In a Bayesian analysis, evidence is combined with a prior distribution to produce a posterior distribution for the treatment effect. This differs from a conventional frequentist confidence interval, which is interpreted through repeated-sampling properties rather than as a probability statement about a fixed parameter.
The distinction concerns the statistical framework used to quantify uncertainty. The registry does not report a Bayesian framework for this trial.
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial's registered primary endpoint is IBTR, a longitudinal recurrence endpoint. Its structure naturally supports time-to-event methods when event and censoring times are available.
Clinical interpretation
The clinical question concerns recurrence within the breast following surgery for ductal carcinoma in situ or stage I or stage II breast cancer, under randomized radiation treatment assignment.
These two perspectives should remain connected but distinct. Statistical methods describe how evidence about recurrence is quantified, while clinical interpretation asks what that recurrence pattern means within the population and treatment setting studied.
21. Important Limitations and Interpretation Issues
- No posted statistical analyses: the registry lists the primary endpoint but does not provide a posted statistical analysis or numerical outcome estimate for IBTR.
- Multiple radiation techniques: the intervention description includes several partial-breast approaches, so interpretation depends on how these techniques were grouped in the randomized comparison and analysis.
- Longitudinal follow-up: recurrence is assessed repeatedly, making the timing and censoring of events important components of statistical interpretation.
- Unmasked design: the registry describes the study as NONE for masking. The implications for endpoint ascertainment depend on the procedures used to identify and classify IBTR.
- Unreported analysis details: the registry record does not report a statistical analysis plan, multiplicity strategy, interim-analysis boundary, missing-data method, or non-inferiority margin.
- Time-to-event assumptions: if a Cox proportional-hazards model is used, proportional hazards should be evaluated before relying on a single hazard ratio as a complete description of the treatment difference.
- Follow-up duration: participants may contribute different amounts of observed time, which is why recurrence proportions alone may not fully represent the available event-time information.
22. Why This Trial Matters Statistically
NSABP B-39/RTOG 0413 is a useful statistical teaching case because it combines randomized treatment allocation with a recurrence endpoint followed longitudinally over an extended period. It illustrates how clinical trial design, event definitions, follow-up schedules, censoring, and time-to-event methodology fit together.
| Concept | How it appears in NSABP B-39/RTOG 0413 |
|---|---|
| Randomization | Participants were randomized. |
| Parallel design | The registry specifies a parallel design model. |
| Two-arm comparison | The study has two arms. |
| Time-to-event analysis | IBTR is a recurrence endpoint followed over time. |
| Repeated follow-up | IBTR is assessed every 6 months for 5 years and annually thereafter for up to 10 years. |
| Censoring | Participants without an observed event can contribute follow-up until their observation ends. |
| Kaplan-Meier estimation | A standard method for describing event-free experience over time. |
| Log-rank testing | A standard method for comparing two randomized time-to-event groups. |
| Cox modeling | A standard model for estimating relative event rates for time-to-event outcomes. |
| Absolute vs relative effects | Time-specific recurrence estimates and hazard ratios describe different aspects of treatment effect. |
23. Statistical Reasoning: From Endpoint to Analysis
The statistical reasoning starts with the clinical endpoint rather than the statistical test. Once IBTR is defined and its observation schedule established, the analyst determines how recurrence timing and censoring should be represented. Only then is the appropriate comparison and effect measure selected.
This sequence prevents a common analytical mistake: choosing a statistical test first and attempting to fit the clinical endpoint to it afterward. For recurrence outcomes, the temporal structure is part of the information being analyzed.
24. Interpreting Relative and Absolute Effects
Suppose a future analysis reports a hazard ratio for IBTR. That estimate would answer a relative question about event rates between randomized groups under the fitted model. It would not directly state the absolute probability of recurrence by a particular year.
A hazard ratio summarizes the relative difference in the instantaneous rate of recurrence between groups under a Cox model. It should be interpreted together with its confidence interval and the model assumptions.
An absolute recurrence estimate describes the magnitude of recurrence experience on a clinically interpretable scale at a specified time point. Absolute and relative measures answer different questions.
A p-value addresses evidence against a specified null hypothesis within the chosen testing framework. It does not measure the size of the treatment effect, its clinical importance, or the probability that one treatment is superior.
For a long-term recurrence endpoint, presenting both relative and absolute measures is particularly useful because the hazard ratio compresses a potentially complex time pattern into one number, whereas time-specific estimates show how recurrence-free experience evolves.
25. Statistical Analysis Questions for This Trial
What is the estimand?
The principal clinical quantity concerns the difference in IBTR experience between the randomized radiation strategies over the registered follow-up period.
What is the event?
The registered primary event is in-breast tumor recurrence.
What is the follow-up?
Assessment occurs every 6 months for 5 years and annually thereafter for up to 10 years.
What is the comparison?
The study is randomized, has two arms, and uses a parallel design.
26. Record Summary
NSABP B-39/RTOG 0413 is a randomized, parallel, two-arm treatment trial enrolling 4216 participants. The registry identifies women who have undergone surgery for ductal carcinoma in situ or stage I or stage II breast cancer and lists radiation interventions including 3-dimensional conformal accelerated partial breast irradiation, brachytherapy, whole breast irradiation, and MammoSite or another single-entry intracavitary device.
The registered primary endpoint is in-breast tumor recurrence (IBTR), assessed every 6 months for 5 years and annually thereafter for up to 10 years. Because IBTR is a recurrence observed longitudinally, time-to-event methodology provides a natural statistical framework: Kaplan-Meier estimation can describe recurrence-free experience, log-rank testing can compare randomized groups, and Cox regression can quantify a relative event rate through a hazard ratio when its assumptions are appropriate.
The registry record identifies the study as completed, with a start date of 2005-03 and a primary completion date of 2018-08. No posted statistical analyses are present in the ClinicalTrials.gov record, so the statistical interpretation centers on the registered endpoint, design, follow-up structure, and the methods that would ordinarily be appropriate for its analysis rather than on a numerical treatment-effect estimate.