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-38 was a completed randomized phase 3 trial in women who had undergone surgery for node-positive breast cancer. The trial used a parallel, unmasked design and evaluated three chemotherapy strategies.
| Feature | NSABP B-38 |
|---|---|
| Trial name | NSABP B-38 |
| ClinicalTrials.gov identifier | NCT00093795 |
| Phase | Phase 3 |
| Status | Completed |
| Condition | Breast Cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 4894 |
| Primary endpoint type | Time-to-event |
| Primary endpoint time frame | 5 years |
| Lead sponsor | NSABP Foundation Inc |
| Sponsor type | Network |
| Trial period | 2004-10 to 2012-03 primary completion |
2. Clinical Question
The central statistical question was whether different chemotherapy strategies produce different disease-free survival outcomes in women who had undergone surgery for node-positive breast cancer.
Population
Women who have undergone surgery for node-positive breast cancer.
Interventions
The registered interventions were cyclophosphamide, docetaxel, gemcitabine, paclitaxel, and doxorubicin.
Comparator structure
The trial randomized participants among three chemotherapy strategies rather than using a simple two-group comparison.
Primary question
How does disease-free survival compare across the randomized chemotherapy strategies over the registered 5-year time frame?
3. Trial Design
The ClinicalTrials.gov record identifies the allocation as randomized, the design model as parallel, masking as none, and the primary purpose as treatment. The three-arm structure is important statistically because a three-group randomized comparison can require a different multiplicity strategy from a single two-arm comparison.
TAC × 6
- Docetaxel
- Doxorubicin
- Cyclophosphamide
- 6 treatment cycles
- Serious AEs: 277/1601 affected/at risk
AC × 4 then P × 4
- Doxorubicin
- Cyclophosphamide
- Then paclitaxel
- 4 AC cycles followed by 4 P cycles
- Serious AEs: 158/1612 affected/at risk
AC × 4 then PG × 4
- Doxorubicin
- Cyclophosphamide
- Then paclitaxel plus gemcitabine
- 4 AC cycles followed by 4 PG cycles
- Serious AEs: 187/1604 affected/at risk
4. Endpoints
| Endpoint | Registry definition | Time frame | Endpoint type |
|---|---|---|---|
| Disease-free Survival | Any Recurrence, Contralateral Breast Cancer, Second Primary Cancer, Death From Any Cause Prior to Recurrence or Second Primary Cancer | 5 years | Time-to-event |
The registry definition is unusually important here because disease-free survival is not restricted to a single recurrence event. The registered composite includes any recurrence, contralateral breast cancer, second primary cancer, or death from any cause prior to recurrence or second primary cancer.
The registry further describes the endpoint as the percentage of patients alive and cancer-free.
5. Statistical Methodology
The ClinicalTrials.gov record identifies the primary endpoint as a time-to-event measure and define its 5-year assessment. However, the ClinicalTrials.gov record does not provide a formal statistical analysis entry specifying the test, model, effect measure, confidence interval, or multiplicity procedure used for the posted primary endpoint.
What would normally be used for this endpoint?
For a randomized trial with a disease-free survival endpoint, a conventional analysis would use a time-to-event framework. Kaplan-Meier methods can estimate the disease-free survival function over time, while a log-rank test can compare event-time distributions between randomized groups. A Cox proportional-hazards model can estimate a hazard ratio and its confidence interval when its modeling assumptions are appropriate.
Here, S(t) represents the probability of remaining free of a qualifying event beyond time t. The trial's registered endpoint specifies which events count; it does not, in the ClinicalTrials.gov record, specify the statistical model used to estimate S(t).
Why censoring matters
Time-to-event analysis can use information from patients who have not yet experienced a qualifying event at their last available assessment. Such observations are generally censored rather than treated as if the event never occurred. The validity of the resulting estimate depends on the censoring and follow-up assumptions used by the analysis.
Three-arm comparison
Because NSABP B-38 has three randomized treatment strategies, the analysis can involve more than one pairwise contrast. When several treatment comparisons are made, the prespecified multiplicity strategy becomes important because repeated hypothesis tests can increase the overall chance of a false-positive finding.
Randomization as the statistical foundation
Randomization creates the basis for comparing outcomes between treatment groups without relying on observational adjustment to establish the primary treatment contrast. In a randomized trial, the analysis should preserve the randomized comparison as closely as possible and should clearly identify the analysis population and prespecified contrasts.
6. Results
The registry data state that results were posted and that the primary endpoint has a reported result. However, the ClinicalTrials.gov record contains no formal statistical analyses. Consequently, the available information supports description of the endpoint and trial structure, but not reconstruction of a formal efficacy comparison.
Primary endpoint status
Primary endpoint: disease-free survival
Time frame: 5 years
Formal statistical analysis entries reported: none
No hazard ratio, confidence interval, p-value, event count, 5-year disease-free survival percentage, median disease-free survival, or pairwise comparison is included in the posted results. Those quantities therefore are not reported on this page.
The absence of a registry-reported statistical-analysis entry does not mean that the trial lacked a statistical analysis. It means that the ClinicalTrials.gov record does not contain the formal analysis results needed to report an effect estimate responsibly.
For a time-to-event endpoint, an effect estimate such as a hazard ratio would describe the relative event rate under a specified model. A confidence interval would describe uncertainty around that estimate, while a p-value would address a prespecified hypothesis under the relevant testing framework. None of those quantities can be reported here without additional trial data.
Similarly, a 5-year percentage would be an absolute time-specific estimate and would answer a different question from a hazard ratio.
7. Safety Results
The ClinicalTrials.gov record provides serious adverse events by randomized treatment strategy as affected patients divided by the number at risk. These are the safety results that can be reported from the provided record.
| Treatment strategy | Serious adverse events | Affected / at risk |
|---|---|---|
| TAC × 6 | Serious adverse events | 277/1601 |
| AC × 4 Then P × 4 | Serious adverse events | 158/1612 |
| AC × 4 Then PG × 4 | Serious adverse events | 187/1604 |
What the numerator represents
The affected count is the number of patients identified as having serious adverse events in the registry-reported arm-level safety field.
What the denominator represents
The denominator is the number described in the ClinicalTrials.gov record as being at risk for the reported serious-adverse-event measure.
What is not reported here
The ClinicalTrials.gov record does not provide severity grades, individual adverse-event terms, discontinuations, deaths due to adverse events, or a formal statistical comparison of safety between arms.
Interpretation
Safety and efficacy answer different questions. The serious-adverse-event counts should therefore be considered separately from the disease-free survival endpoint.
8. Statistical Methods Explained
What does randomization accomplish?
Randomization assigns participants to treatment groups using a chance-based allocation process. Its central statistical purpose is to make treatment assignment independent of measured and unmeasured baseline characteristics in expectation. This allows differences in subsequent outcomes to be interpreted as treatment-group comparisons rather than simply differences between pre-existing patient groups.
Why is disease-free survival a time-to-event endpoint?
The registered endpoint is not merely whether a patient is disease-free at one fixed point. It incorporates the time until a qualifying event. Two patients could both be disease-free at a particular assessment while having had very different follow-up histories. Time-to-event methods preserve information about when events occur and can accommodate censoring.
Why might Kaplan-Meier methods be used?
Kaplan-Meier estimation is designed for time-to-event data in which some observations may be censored. It estimates the probability of remaining event-free as follow-up progresses. For NSABP B-38, such a method would naturally correspond to the registered disease-free survival endpoint.
What would a hazard ratio mean?
A hazard ratio compares the estimated instantaneous event rates between treatment groups under a specified time-to-event model. A hazard ratio below 1 would indicate a lower estimated event rate for the numerator group relative to the reference group. It would not mean that the same percentage of patients avoid an event, nor would it directly equal the difference in 5-year disease-free survival.
Why does the confidence interval matter?
A confidence interval communicates statistical uncertainty around an effect estimate. A narrow interval generally indicates greater precision than a wide interval, although precision depends on the underlying study information and analysis. The interval is not a prediction interval for individual patient outcomes.
Why does a three-arm trial raise multiplicity issues?
With three randomized groups, investigators may want to make multiple pairwise comparisons or compare one strategy against a common reference. If several hypotheses are tested, the probability of at least one false-positive result can increase unless the analysis plan accounts for multiplicity. The ClinicalTrials.gov record does not specify the multiplicity procedure used in NSABP B-38, so no particular adjustment is attributed to the trial here.
9. Missing Analysis Information and What It Means
The ClinicalTrials.gov record contains an important distinction between results being posted and formal statistical-analysis entries being available. For NSABP B-38, the former is marked “Yes,” while no formal statistical analyses were posted.
| Information | Status in the ClinicalTrials.gov record |
|---|---|
| Results posted | Yes |
| Primary endpoint | Disease-free survival |
| Primary endpoint type | Time-to-event |
| Primary endpoint time frame | 5 years |
| Formal statistical analyses reported | None |
| Hazard ratio reported | No |
| Confidence interval reported | No |
| P-value reported | No |
| Primary endpoint event count reported | No |
| 5-year disease-free survival estimate reported | No |
This distinction prevents a common error in trial-result summaries: treating the existence of a posted registry outcome as if it automatically registry-reported the complete inferential statistics needed to reproduce the trial's primary comparison.
10. Trial Timeline
| Milestone | Date |
|---|---|
| Trial start | 2004-10 |
| Primary completion | 2012-03 |
| Status | Completed |
The ClinicalTrials.gov record identifies the study period through its start and primary-completion dates. No additional interim or long-term follow-up dates are provided in the trial data used for this page.
11. Design Topics Supported by the Registry Data
Randomization
The allocation is explicitly identified as randomized. This is the central design feature supporting causal treatment-group comparison.
Parallel design
The registered design model is parallel, meaning the three treatment strategies form concurrent randomized groups rather than sequential treatment periods.
No masking
The registry identifies masking as none. The ClinicalTrials.gov record does not specify additional details about outcome-assessor blinding.
Three-arm structure
Three treatment strategies create a multi-group comparison and make the planned contrast structure important for statistical interpretation.
Topics not specified in the ClinicalTrials.gov record
The ClinicalTrials.gov record does not specify a non-inferiority margin, crossover procedure, factorial structure, Bayesian analysis, missing-data imputation method, stratification factors, or interim-analysis method. Those topics are therefore not attributed to NSABP B-38 on this page.
12. Limitations
- Formal efficacy statistics are unavailable in the registry-reported analysis field: no hazard ratio, confidence interval, p-value, event count, or 5-year disease-free survival estimate is provided.
- Three-arm interpretation: a multi-arm randomized design can require explicit definition of the primary contrasts and multiplicity control. The ClinicalTrials.gov record does not identify those procedures.
- Time-to-event censoring: disease-free survival requires appropriate handling of patients whose event status is not observed through the end of follow-up. The ClinicalTrials.gov record does not specify the censoring rules or statistical implementation.
- Model assumptions: if a Cox model were used, proportional hazards would be an important modeling consideration. The ClinicalTrials.gov record does not establish that a Cox model was used.
- Safety interpretation: the registry-reported serious-adverse-event counts do not include enough detail to reconstruct the full safety profile or a formal between-arm safety analysis.
- Endpoint definition: disease-free survival is a composite time-to-event endpoint that includes several types of recurrence, second primary cancer, contralateral breast cancer, and death under the registered definition. These components should not be treated as interchangeable clinical events.
- Registry-data scope: the analysis presented here is limited to the ClinicalTrials.gov record. It does not import numerical findings from publications or other sources.
13. Why This Trial Matters Statistically
NSABP B-38 is a useful statistical teaching case because its design combines randomization, three parallel treatment strategies, and a clinically composite time-to-event endpoint.
| Concept | How it appears in NSABP B-38 |
|---|---|
| Randomization | Participants were randomized among three treatment strategies. |
| Parallel design | The registry identifies a parallel trial design. |
| Three-arm comparison | Three chemotherapy strategies are identified. |
| Time-to-event analysis | The registered primary endpoint is disease-free survival. |
| Composite endpoint | Disease-free survival includes recurrence, contralateral breast cancer, second primary cancer, and qualifying death. |
| Kaplan-Meier estimation | A conventional approach for describing disease-free survival over time. |
| Log-rank testing | A conventional approach for comparing randomized time-to-event distributions. |
| Cox regression | A conventional model for estimating hazard ratios when appropriate. |
| Multiplicity | Three randomized groups can create multiple possible treatment contrasts. |
| Safety analysis | Serious adverse events are reported by treatment strategy. |
14. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The trial provides a randomized, three-arm framework for comparing a 5-year disease-free survival endpoint. The ClinicalTrials.gov record does not provide the formal inferential statistics needed to quantify the primary treatment comparisons.
Clinical interpretation
The trial addresses an important treatment question in women who underwent surgery for node-positive breast cancer, while the registry-reported safety field provides serious-adverse-event counts for each randomized strategy.
These are deliberately separate statements. Statistical significance, effect size, clinical importance, and safety are related but distinct dimensions of evidence and should not be collapsed into one conclusion.
15. Sources
- ClinicalTrials.gov: NCT00093795 — NSABP B-38.
- PubMed: PubMed record associated with NSABP B-38.
16. Related Tutorials
Learn more about the methods used in this trial:
17. Related Calculators
Continue through the Clinical Biostats clinical-trial pathway
Explore statistical tutorials and calculators that develop the concepts used to understand randomized clinical trials and their time-to-event endpoints.
18. Record Summary
NSABP B-38 was a completed randomized phase 3 trial enrolling 4894 participants with three chemotherapy strategies in a parallel, unmasked treatment design. Its registered primary endpoint was 5-year disease-free survival, defined as the percentage of patients alive and cancer-free and incorporating any recurrence, contralateral breast cancer, second primary cancer, or death from any cause prior to recurrence or second primary cancer.
The ClinicalTrials.gov record provides arm-level serious-adverse-event counts of 277/1601 for TAC × 6, 158/1612 for AC × 4 Then P × 4, and 187/1604 for AC × 4 Then PG × 4. Although results are marked as posted, no formal statistical-analysis entries are included in the ClinicalTrials.gov recordset. A responsible statistical review therefore explains the appropriate time-to-event framework without fabricating hazard ratios, confidence intervals, p-values, event counts, or 5-year efficacy estimates that are not present in the ClinicalTrials.gov record.