This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. The registry identifies the primary endpoint and reports that results are posted, but the ClinicalTrials.gov record contains no formal statistical-analysis estimates for the primary endpoint.
1. Trial at a Glance
NSABP B-40 was a randomized, parallel, open-label phase 3 treatment trial in breast cancer. The trial enrolled 1206.0 participants and used six treatment arms involving chemotherapy regimens with or without bevacizumab.
| Feature | NSABP B-40 |
|---|---|
| Trial name | NSABP B-40 |
| NCT identifier | NCT00408408 |
| Phase | Phase 3 |
| Condition | Breast Cancer |
| Population | Women with stage I, stage II, or stage IIIA breast cancer that can be removed by surgery |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1206 |
| Arms | 6 |
| Primary endpoint | Pathologic Complete Response (pCR) of the Primary Tumor in the Breast |
| Results posted | Yes |
| Outcome measures posted | 10 |
| Lead sponsor | NSABP Foundation Inc |
| Sponsor type | NETWORK |
| Trial start | 2006-11 |
| Primary completion | 2011-03 |
2. Clinical Question
The central clinical question was whether adding bevacizumab to chemotherapy affected the probability of achieving a pathologic complete response of the primary breast tumor in women with surgically removable stage I, stage II, or stage IIIA breast cancer.
Population
Women with stage I, stage II, or stage IIIA breast cancer that can be removed by surgery.
Intervention
Chemotherapy regimens containing bevacizumab, with the six-arm design also varying the chemotherapy backbone.
Comparator
Chemotherapy regimens without bevacizumab.
Primary question
Does treatment assignment affect the binary outcome of pathologic complete response of the primary tumor in the breast?
3. Trial Design
The registry describes the allocation as randomized, the design model as parallel, masking as none, and the primary purpose as treatment. The six-arm structure is important statistically because the trial is not simply a single two-group comparison.
Three chemotherapy-only groups
- Docetaxel Then AC
- Docetaxel + Capecitabine Then AC
- Docetaxel + Gemcitabine Then AC
Three bevacizumab-containing groups
- Docetaxel + Bevacizumab Then AC + Bevacizumab
- Docetaxel + Capecitabine + Bevacizumab Then AC + Bevacizumab
- Docetaxel + Gemcitabine + Bevacizumab Then AC + Bevacizumab
The ClinicalTrials.gov record identifies bevacizumab as a biological intervention and capecitabine, cyclophosphamide, docetaxel, doxorubicin hydrochloride (Adriamycin), and gemcitabine hydrochloride as drug interventions.
4. Primary Endpoint
| Endpoint | Registry definition | Time frame | Type |
|---|---|---|---|
| Pathologic Complete Response (pCR) of the Primary Tumor in the Breast | Percentage of patients absent of histologic evidence of invasive tumor cells in the surgical breast specimen. | Time of surgery, on average 6 or 13 months | Binary |
The registry lists one primary endpoint. Its endpoint type is identified as Binary, and results are posted for this endpoint. The ClinicalTrials.gov record does not provide the arm-specific pCR percentages, treatment-effect estimates, confidence intervals, or p-values.
5. Understanding the pCR Endpoint
A binary endpoint records whether a prespecified event or state occurred. Here, the relevant state is pathologic complete response of the primary tumor in the breast at surgery.
The primary endpoint is therefore fundamentally a comparison of proportions across treatment groups rather than a time-to-event endpoint.
For a randomized comparison, the proportion with pCR can be summarized separately in each treatment group. A treatment effect can then be expressed using measures such as an absolute risk difference, risk ratio, or odds ratio, each with an appropriate confidence interval.
The important point is that the registry's designation of the endpoint as binary does not by itself specify which effect measure or inferential procedure was used for the posted results. The registry-reported statistical-analysis section is empty, so a specific formal analysis cannot be attributed to NSABP B-40 from the ClinicalTrials.gov record.
6. Statistical Methodology
Comparing binary proportions
For a binary endpoint such as pCR, a standard treatment comparison begins with the observed proportion of patients achieving pCR in each randomized group. The difference between those proportions can be estimated directly, while relative measures can provide complementary perspectives.
These measures answer related but different questions. None can be calculated from the ClinicalTrials.gov record because the arm-specific pCR results are not included.
Contingency-table methods
A two-group binary endpoint can commonly be evaluated using a chi-square test or, when expected cell counts are small, Fisher's exact test. For a multi-arm trial, the analysis must also account for the fact that more than two randomized groups exist.
Regression-based analysis
Logistic regression is another standard approach for a binary endpoint. It models the probability of pCR and can produce an odds ratio while allowing prespecified covariates or treatment-group indicators to be incorporated. The registry-reported NSABP B-40 data do not state that logistic regression was used.
Six-arm structure
The six-arm design creates an important distinction between the overall question of whether treatment assignment is associated with pCR and specific pairwise contrasts between individual treatment groups. If multiple pairwise comparisons are performed, the resulting multiplicity must be addressed according to the prespecified statistical plan.
Randomization and causal interpretation
Randomization is the central design feature supporting comparison of treatment groups. Because the registry identifies the trial as randomized and parallel, treatment assignment—not a patient's choice of treatment—is the basis for the intended comparison. That does not eliminate statistical uncertainty or every possible source of bias, but it provides the foundation for causal inference when the randomized comparison is analyzed according to the prespecified design.
7. What the Registry Reports — and What It Does Not
| Item | registry data | Interpretation |
|---|---|---|
| Results posted | Yes | ClinicalTrials.gov reports that results are posted. |
| Outcome measures posted | 10 | Ten outcome measures are identified in the ClinicalTrials.gov record. |
| Primary endpoint | pCR of the primary tumor in the breast | Results are posted for the primary endpoint. |
| Formal statistical analyses | None reported | No treatment-effect estimate, confidence interval, or p-value can be reproduced from the ClinicalTrials.gov record. |
| Primary endpoint type | Binary | The endpoint is analyzed conceptually as a binary outcome. |
8. Planned Analysis
Because the ClinicalTrials.gov record reports that results are posted but does not supply formal statistical-analysis estimates, the most defensible educational approach is to distinguish the reported endpoint from the analysis that would normally be used for this type of endpoint.
Observed outcome
For each treatment group, calculate the percentage of patients with absent histologic evidence of invasive tumor cells in the surgical breast specimen.
Primary comparison
Compare the binary pCR outcome between prespecified treatment groups using an appropriate two-group or multi-group categorical-data procedure.
Effect estimate
Report an absolute and/or relative effect measure, such as a risk difference, risk ratio, or odds ratio, with a confidence interval.
Multiplicity
If several pairwise treatment comparisons are made, interpretation should follow the prespecified multiplicity strategy. The ClinicalTrials.gov record does not specify that strategy.
ClinicalTrials.gov does not report a formal comparison in the ClinicalTrials.gov record. Consequently, this page does not state whether any particular comparison met a statistical significance threshold.
9. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm as the number affected divided by the number at risk. These figures provide an arm-level safety description, but they do not provide a formal statistical comparison.
| Treatment arm | Serious adverse events | Interpretation of registry-reported measure |
|---|---|---|
| Docetaxel Then AC | 4/199 | 4 affected among 199 at risk |
| Docetaxel + Bev Then AC + Bev | 15/196 | 15 affected among 196 at risk |
| Docetaxel + Capecitabine Then AC | 8/203 | 8 affected among 203 at risk |
| Docetaxel + Cape + Bev Then AC + Bev | 30/196 | 30 affected among 196 at risk |
| Docetaxel + Gem Then AC | 2/194 | 2 affected among 194 at risk |
| Docetaxel + Gem + Bev Then AC + Bev | 15/202 | 15 affected among 202 at risk |
These are descriptive safety counts. They should not be converted into a formal between-arm inference without knowing the exact safety analysis population, endpoint definition, exposure handling, and prespecified statistical method.
10. Statistical Methods Explained
Why is pCR a binary endpoint?
Because the registry defines the endpoint as whether the surgical breast specimen contains histologic evidence of invasive tumor cells. Each patient therefore contributes a yes/no outcome to the primary endpoint.
Why compare proportions rather than means?
A mean is appropriate for a quantitative continuous measurement. pCR is instead a categorical outcome, so the primary descriptive quantity is the proportion of patients meeting the response definition.
What does a risk difference tell us?
A risk difference expresses the absolute difference in the probability of pCR between two treatment groups. For example, a positive risk difference would mean that the first group had a higher observed pCR proportion than the comparison group. The ClinicalTrials.gov record does not contain the pCR proportions needed to calculate one for NSABP B-40.
What does an odds ratio tell us?
An odds ratio compares the odds of pCR between two groups. An odds ratio above 1 would indicate higher odds of pCR in the numerator group, whereas an odds ratio below 1 would indicate lower odds. An odds ratio is not the same as a risk ratio, particularly when the outcome is common.
Why does the six-arm design matter?
With six randomized treatment groups, there can be multiple clinically interesting comparisons. Testing several comparisons increases the importance of prespecified contrast definitions and multiplicity control. The ClinicalTrials.gov record identifies the six arms but does not specify a formal multiplicity strategy.
Why is randomization important?
Randomization assigns treatment according to the trial design rather than according to observed prognosis or treatment preference. This is what allows differences in outcomes between randomized groups to be interpreted as evidence about treatment assignment, subject to random error, adherence, missing information, and the other limitations of the trial.
Why can the posted results not simply be reconstructed here?
The ClinicalTrials.gov record states that results are posted and identifies the primary endpoint, but the formal statistical-analysis section contains no entries. Without arm-specific pCR results and the corresponding inferential quantities, calculating a treatment effect would require information not contained in the ClinicalTrials.gov record.
11. Clinical Biostats Interpretation
The pCR endpoint measures a pathologic response at surgery. It is a binary measure describing whether the specified histologic finding was present or absent in the surgical breast specimen. It is not itself a measure of overall survival, recurrence time, or long-term disease control.
If a formal risk difference, risk ratio, or odds ratio were available, it would describe the difference or relative association in pCR between specified randomized treatment groups. It would not mean that every patient experiences the estimated effect, nor would it automatically describe long-term clinical outcomes.
A confidence interval around a binary-outcome effect estimate would describe statistical uncertainty around that estimate under the selected analysis. A narrow interval would indicate greater statistical precision than a wide interval, but the interval would not describe the range of individual patient responses.
If a p-value were reported, it would address the evidence against a specified null hypothesis under the statistical model. It would not tell us how large the treatment effect is. Effect size and uncertainty should therefore be examined directly rather than inferred from the p-value alone.
12. Limitations and Interpretation Issues
- Incomplete formal analysis data: the registry-reported statistical-analysis section is empty, so numerical treatment-effect estimates for the primary endpoint cannot be independently reproduced from the ClinicalTrials.gov record.
- Six-arm design: multiple treatment groups create several possible comparisons, making the choice of prespecified contrasts important.
- Multiplicity: the ClinicalTrials.gov record does not specify how multiple comparisons were handled. Formal claims based on several tests therefore cannot be reconstructed from the ClinicalTrials.gov record.
- Binary endpoint: pCR is a surgical pathologic outcome and should not be treated as interchangeable with a time-to-event endpoint or a survival measure.
- Safety interpretation: the registry-reported serious-adverse-event counts are descriptive. They do not establish statistical differences between arms.
- Missing numerical results: although the registry indicates that results are posted, the ClinicalTrials.gov record does not include the arm-specific primary-endpoint results needed for a quantitative efficacy comparison.
- Trial status: the ClinicalTrials.gov record lists the trial status as UNKNOWN, so this page does not infer a current operational status beyond the information provided.
13. Why This Trial Matters Statistically
NSABP B-40 is a useful teaching example because it combines a randomized phase 3 design with a binary pathologic endpoint and a six-arm treatment structure. That combination highlights a different set of statistical issues from trials whose primary endpoints are overall survival or progression-free survival.
| Concept | How it appears in NSABP B-40 |
|---|---|
| Randomization | The registry identifies the allocation as randomized. |
| Parallel design | The trial uses a parallel design model with six treatment arms. |
| Binary endpoint | The primary endpoint is pCR of the primary tumor in the breast. |
| Proportion comparison | The natural descriptive analysis is the percentage achieving pCR in each treatment group. |
| Effect measures | Risk difference, risk ratio, and odds ratio are possible ways to quantify a binary treatment effect. |
| Multi-arm inference | Six treatment groups create multiple possible pairwise and composite comparisons. |
| Multiplicity | Multiple formal comparisons require a prespecified strategy for controlling the relevant error rate. |
| Safety analysis | Serious adverse events are reported descriptively as affected/at-risk counts for each arm. |
| Evidence boundaries | Results are posted, but the registry-reported formal statistical-analysis data contain no estimates or p-values. |
14. Trial Timeline
Trial start
The ClinicalTrials.gov record identifies 2006-11 as the trial start.
Primary completion
The ClinicalTrials.gov record identifies 2011-03 as the primary completion date.
15. What Can and Cannot Be Concluded From the registry Results
Supported
The trial was randomized, parallel, phase 3, unmasked, enrolled 1206.0 participants, used six arms, and designated pCR of the primary tumor in the breast as its single registered primary endpoint.
Not numerically supported
The ClinicalTrials.gov record does not provide pCR percentages by arm, a treatment-effect estimate, a confidence interval, or a p-value for the primary endpoint.
Safety information available
Serious adverse-event counts are posted on ClinicalTrials.gov for all six treatment arms.
Formal inference unavailable
The empty statistical-analysis array means that a formal efficacy conclusion cannot be reconstructed from the ClinicalTrials.gov record alone.
16. Related Tutorials
Learn more about the methods used in this trial:
17. Related Calculators
18. Sources
- ClinicalTrials.gov: NSABP B-40 (NCT00408408). Official trial registry record.
- PubMed: Publication associated with NSABP B-40, PMID 22276821.
- PubMed: Publication associated with NSABP B-40, PMID 26272770.
Continue through the Clinical Biostats trial-analysis pathway
Connect the trial's design and binary endpoint to deeper tutorials and statistical calculation tools.
19. Record Summary
NSABP B-40 is a randomized phase 3, parallel, unmasked treatment trial enrolling 1206.0 participants across six arms. Its single registered primary endpoint was pathologic complete response of the primary tumor in the breast, defined as the percentage of patients absent of histologic evidence of invasive tumor cells in the surgical breast specimen, assessed at the time of surgery, on average 6 or 13 months. The ClinicalTrials.gov record identifies the endpoint as binary and states that results are posted.
The principal statistical lesson is the distinction between having a registered binary endpoint and having the numerical evidence needed to quantify its treatment effect. The ClinicalTrials.gov record does not contain the formal statistical-analysis estimates for pCR. Consequently, this page explains the appropriate statistical framework—proportions, risk differences, risk ratios, odds ratios, categorical-data testing, multi-arm comparisons, and multiplicity—without attributing unreported numerical results to the trial.
The ClinicalTrials.gov record provides serious-adverse-event counts for all six treatment arms, but those counts are descriptive and are not accompanied by a formal statistical comparison in the ClinicalTrials.gov record.