← Clinical Trials
Breast Cancer Randomized Time-to-Event NCT00072293

IBCSG 23-01: Complete Statistical Analysis of Axillary Surgery in Node-Negative Breast Cancer

An independent statistical analysis of IBCSG 23-01, a randomized parallel-group trial comparing axillary lymph node dissection with no axillary lymph node dissection in women with node-negative breast cancer and sentinel lymph node micrometastases.

Trial start: 2001-12  ·  Primary completion: 2012-09  ·  Enrollment: 931
Scope of this record

This page separates reported trial results from statistical interpretation. The numerical results presented here are restricted to the information contained in the ClinicalTrials.gov trial data posted on ClinicalTrials.gov for IBCSG 23-01 and its posted statistical analyses.

Registry note: 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

IBCSG 23-01 was a randomized, parallel-group clinical trial evaluating axillary lymph node dissection versus no axillary lymph node dissection in women with node-negative breast cancer and sentinel lymph node micrometastases. The registered primary endpoint was 5-year disease-free survival, analyzed as a time-to-event outcome.

931
Enrollment
Randomized trial
2
Arms
Parallel design
0.78
Primary HR
No dissection / dissection
0.004
Primary P-value
1-sided non-inferiority test
FeatureIBCSG 23-01
Trial nameIBCSG 23-01
NCT IDNCT00072293
ConditionBreast Cancer
PopulationWomen with node-negative breast cancer and sentinel lymph node micrometastases
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Enrollment931
Lead sponsorETOP IBCSG Partners Foundation
StatusCompleted

2. Clinical Question

The clinical question was whether women with node-negative breast cancer and sentinel lymph node micrometastases could undergo no axillary lymph node dissection rather than axillary lymph node dissection while maintaining disease-free outcomes according to the trial's prespecified time-to-event framework.

Population

Women with node-negative breast cancer and sentinel lymph node micrometastases.

Intervention

No axillary lymph node dissection.

Comparator

Axillary lymph node dissection.

Primary question

Does no axillary lymph node dissection preserve 5-year disease-free survival relative to axillary lymph node dissection under the registered non-inferiority framework?

3. Trial Design

01
Randomize931 participants
02
Two armsParallel-group allocation
03
CompareAxillary dissection vs no dissection
04
FollowTime-to-event outcomes
05
AnalyzeITT disease-free survival
ARM A

Axillary lymph node dissection

  • Axillary lymph node dissection procedure
  • Comparator treatment in the randomized parallel-group design
ARM B

No axillary lymph node dissection

  • No axillary lymph node dissection procedure
  • Intervention treatment in the randomized parallel-group design

The registry identifies the study as randomized, parallel, and unmasked. The ClinicalTrials.gov record does not provide a randomization ratio, detailed stratification factors, treatment schedules, or a treatment crossover description, so those design features are not added here.

4. Trial Timeline

2001-12

Trial start

The registry lists December 2001 as the trial start.

2012-09

Primary completion

The registry lists September 2012 as the primary completion date.

Completed

Results posted

ClinicalTrials.gov reports that results were posted, including three outcome measures and two statistical analyses.

5. Endpoints

EndpointRegistered definitionTime frame
5-year Disease-Free Survival Estimated percentage of patients alive and disease-free at 5 years from randomization, where disease-free survival is defined as the time from randomization to first evidence of invasive relapse at any site, second primary tumor (contralateral or non-breast) or death. 5-year estimate reported after a median follow-up of 60 months
5-year Overall Survival Secondary time-to-event outcome reported in the posted statistical analysis. 5-year estimate reported after a median follow-up of 60 months

The registry identifies the primary endpoint as 5-year Disease-Free Survival and classifies it as a time-to-event endpoint. The ClinicalTrials.gov record also report 5-year Overall Survival as a secondary time-to-event endpoint.

6. Analysis Populations and Treatment Comparison

The posted primary and secondary statistical analyses both use the intention-to-treat population. This means the treatment comparison is anchored to the randomized assignment rather than redefining groups according to treatment received after randomization.

AnalysisPopulationGroups comparedEffect measure
5-year Disease-Free Survival Intention-to-treat Axillary Dissection vs No Axillary Dissection Hazard ratio
5-year Overall Survival Intention-to-treat Axillary Dissection vs No Axillary Dissection Hazard ratio
Direction of the hazard ratio: for both posted analyses, the registry states that the hazard ratio is no axillary dissection / axillary dissection. Therefore, an HR below 1 represents a lower estimated event hazard in the no-dissection group relative to the dissection group under the reported model.

7. Statistical Methodology

Primary analysis: one-sided non-inferiority test

The primary endpoint, 5-year disease-free survival, was analyzed using a 1-sided non-inferiority test in the intention-to-treat population. The posted effect measure was a hazard ratio, with an estimate of 0.78 and a two-sided 95% confidence interval of 0.55 to 1.11.

5-year Disease-Free Survival

HR 0.78

95% CI: 0.55–1.11   ·   P = 0.004

Intention-to-treat analysis · 1-sided non-inferiority test

The registry analysis notes that the hazard ratio is calculated as no axillary dissection/axillary dissection. Thus, the point estimate of 0.78 corresponds to an estimated hazard approximately 22% lower in the no-dissection group under the fitted comparison.

Clinical Biostats interpretation

The HR of 0.78 is a relative time-to-event measure. Because the ratio is below 1, the estimated instantaneous event hazard was lower for no axillary lymph node dissection than for axillary lymph node dissection under the reported analysis.

It does not mean that 22% of patients avoided an event, that 22% more patients were disease-free at 5 years, or that every patient experienced a 22% reduction in risk. A hazard ratio is a model-based relative measure of event rates over follow-up.

The 95% confidence interval of 0.55–1.11 describes statistical uncertainty around the estimated hazard ratio. It is not an interval containing the outcomes that individual patients can experience.

The P = 0.004 value is evidence against the null hypothesis specified for the reported one-sided non-inferiority test. A p-value does not measure the magnitude of the treatment effect and should not be read as the probability that the treatment effect is real.

Non-inferiority requires comparison with a prespecified non-inferiority margin. The registry text identifies the analysis as a non-inferiority test and reports that the original design targeted a one-sided statistical significance level of 10%, but the exact margin is not contained in the ClinicalTrials.gov record. Therefore, the margin itself should not be reconstructed from the observed HR or confidence interval.

Secondary analysis: 5-year Overall Survival

The secondary 5-year Overall Survival analysis used the log-rank test in the intention-to-treat population. The effect measure was again a hazard ratio, with the same stated direction of no axillary dissection relative to axillary dissection.

5-year Overall Survival

HR 0.89

90% CI: 0.52–1.54   ·   P = 0.73

Intention-to-treat analysis · Log-rank test · Superiority hypothesis

Clinical Biostats interpretation

The HR of 0.89 means that the estimated hazard of death was lower in the no-dissection group than in the dissection group under the reported hazard-ratio comparison, with the estimated relative hazard approximately 89% of the comparator hazard.

This does not mean that mortality was 11% lower in an absolute sense, nor does it mean that 11% of patients survived because of the intervention. The hazard ratio is not an absolute survival difference.

The 90% confidence interval of 0.52–1.54 is relatively broad and includes 1.00, indicating substantial statistical uncertainty around the estimated relative hazard. The interval describes uncertainty around the estimated treatment comparison, not individual patient outcomes.

The reported P = 0.73 does not provide evidence of a statistically significant superiority difference under the posted analysis. Importantly, a non-significant p-value is not proof that the two treatments have exactly identical effects; it indicates that the observed data did not provide evidence for the superiority hypothesis under the stated analysis.

8. Results Summary

EndpointAnalysisEffect estimateConfidence intervalP-value
5-year Disease-Free Survival 1-sided non-inferiority test HR 0.78 95% CI 0.55–1.11 0.004
5-year Overall Survival Log-rank; superiority hypothesis HR 0.89 90% CI 0.52–1.54 0.73

The two analyses answer different statistical questions. The primary disease-free survival analysis was framed as non-inferiority, whereas the secondary overall survival analysis was framed as superiority. Their p-values therefore cannot be interpreted as though they arose from the same hypothesis.

9. Statistical Methods Explained

Why was an intention-to-treat analysis used?

In a randomized trial, the intention-to-treat principle preserves the comparison created by randomization. Participants remain associated with the group to which they were randomized when efficacy outcomes are analyzed. This helps maintain the causal interpretation supported by random allocation and avoids redefining the comparison according to later treatment behavior.

What does an HR of 0.78 mean?

For the primary endpoint, the registry defines the HR as no axillary dissection divided by axillary dissection. An HR of 0.78 therefore indicates a lower estimated instantaneous disease-free-survival event hazard in the no-dissection group under the fitted time-to-event comparison. A useful derived interpretation is that 0.78 is 22% below 1.00, but this should not be converted into a statement that 22% of patients benefited.

Why is non-inferiority different from superiority?

A superiority trial asks whether one treatment provides evidence of a better outcome than another. A non-inferiority trial asks whether the new or alternative strategy is not unacceptably worse than the comparator according to a prespecified margin. The distinction is important because a hazard ratio below 1 alone does not establish non-inferiority, and a conventional two-sided confidence interval should not be interpreted without reference to the prespecified non-inferiority framework.

Why does the primary confidence interval extend above 1?

The reported two-sided 95% CI for the primary HR is 0.55–1.11. Because this interval includes 1, it should not be interpreted using the simple rule that a two-sided 95% CI excluding 1 is required for every non-inferiority conclusion. Non-inferiority is evaluated against the prespecified margin using the appropriate one-sided hypothesis framework. The exact margin is not contained in the ClinicalTrials.gov record, so it cannot be inferred here.

What does the 5-year time frame add?

The registered endpoint is a time-to-event outcome summarized at 5 years. This is different from a simple proportion observed at a single visit because disease-free survival accounts for the timing of events and censoring during follow-up. The registry defines the endpoint from randomization to the first qualifying invasive relapse, second primary tumor, or death.

Why use a log-rank test for overall survival?

The log-rank test is designed to compare time-to-event distributions between groups while accounting for the timing of events and censoring. It is therefore a natural method for an overall-survival endpoint. In the posted secondary analysis, the log-rank method was used together with the hazard-ratio effect measure.

Why should the p-value not be treated as an effect size?

A p-value summarizes how compatible the observed data are with a specified null hypothesis under the statistical model. It depends on the effect, variability, information, and sample size. The hazard ratio and its confidence interval provide information about the magnitude and precision of the estimated treatment comparison; the p-value does not replace them.

10. Understanding the Primary Non-Inferiority Analysis

The primary statistical feature of IBCSG 23-01 is the non-inferiority framework. This is important because the clinical interpretation of an alternative treatment strategy can depend on whether it preserves an acceptable amount of the comparator's efficacy rather than whether it demonstrates a statistically significant advantage.

Non-inferiority logic
H0: alternative treatment is worse than the comparator by more than the prespecified margin
H1: alternative treatment is not worse than the comparator by more than the prespecified margin

The exact numerical margin is not reported in the ClinicalTrials.gov record. It should therefore not be reverse-engineered from the observed hazard ratio, confidence interval, or p-value.

The registry-reported analysis also states that, as originally designed, the target accrual was 1960 patients with analysis planned after 558 events, based on 90% power and a one-sided statistical significance level of 10%. These were design targets rather than the observed enrollment and event results reported in the primary analysis reported here.

Design versus observed sample size: the ClinicalTrials.gov record reports enrollment of 931, while the original design description in the posted analysis refers to a target accrual of 1960. These quantities answer different questions: one describes the trial's reported enrollment and the other describes an original design target. They should not be treated as interchangeable.

11. Hazard Ratios and Confidence Intervals

Both posted analyses use a hazard ratio, but the uncertainty intervals have different confidence levels. The primary disease-free survival analysis reports a 95% two-sided CI, while the secondary overall survival analysis reports a 90% two-sided CI.

EndpointHRCI levelCIDirection
5-year Disease-Free Survival 0.78 95% 0.55–1.11 No dissection / dissection
5-year Overall Survival 0.89 90% 0.52–1.54 No dissection / dissection
Reading the intervals correctly

The primary estimate is 0.78, but the plausible values represented by its reported 95% interval extend from 0.55 to 1.11. The secondary estimate is 0.89, with a reported 90% interval from 0.52 to 1.54. These intervals show that point estimates should not be interpreted without considering their precision.

A confidence interval crossing 1 does not by itself answer a non-inferiority question. Conversely, a confidence interval excluding 1 in a superiority analysis would not automatically answer a non-inferiority question. The hypothesis framework and prespecified margin determine the relevant decision rule.

12. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants divided by those at risk. The reported figures are:

ArmSerious adverse eventsAffected / at risk
Axillary DissectionSerious adverse events1 / 447
No Axillary DissectionSerious adverse events0 / 453
Clinical Biostats interpretation

The ClinicalTrials.gov record reports one affected participant among 447 at risk in the axillary-dissection group and zero among 453 at risk in the no-dissection group. These are the reported serious-adverse-event counts and denominators; they should not be expanded into a broader safety profile that is not contained in the ClinicalTrials.gov record.

The safety denominators also should not be confused with the overall trial enrollment of 931. The registry provides 447 and 453 as the affected/at-risk figures for this specific serious-adverse-event measure.

13. Multiplicity and Hypothesis Testing

The ClinicalTrials.gov record identifies two posted statistical analyses: one primary and one secondary. The primary hypothesis type is non-inferiority or equivalence, while the secondary hypothesis type is superiority.

AnalysisRoleHypothesis typeMethod
5-year Disease-Free Survival Primary Non-inferiority or equivalence 1-sided non-inferiority test
5-year Overall Survival Secondary Superiority Log-rank test

The ClinicalTrials.gov record does not describe an alpha-allocation procedure between the primary and secondary endpoints, a multiplicity-adjustment strategy, or an interim-analysis plan. Those features therefore cannot be reconstructed from the registry information provided.

14. Crossover, Stratification, and Missing Data

Crossover

The ClinicalTrials.gov record does not report treatment crossover. No crossover effect on the overall-survival comparison is inferred.

Stratification

The ClinicalTrials.gov record does not identify randomization or analysis stratification factors. No stratified Cox or stratified log-rank procedure is assumed.

Missing data

The ClinicalTrials.gov record does not specify a missing-data or imputation method for the time-to-event analyses.

Bayesian methods

No Bayesian method is reported in the ClinicalTrials.gov record.

This distinction is important for an independent statistical review. Time-to-event analyses naturally incorporate censoring, but the ClinicalTrials.gov record does not provide enough information to describe the exact censoring rules, missing-data handling, or imputation procedures used in the underlying statistical analysis.

15. Statistical Assumptions Behind the Hazard Ratio

A hazard ratio is most naturally interpreted through a proportional-hazards framework when the relative hazard is reasonably stable over time. The ClinicalTrials.gov record reports hazard ratios but do not provide diagnostic information about the proportional-hazards assumption.

Conceptual hazard-ratio interpretation
HR = hNo ALND(t) / hALND(t)

For IBCSG 23-01, the posted analysis defines the numerator as no axillary lymph node dissection and the denominator as axillary lymph node dissection.

This formula helps explain the direction of the reported estimates. An HR below 1 indicates a lower estimated event hazard for the numerator group; an HR above 1 would indicate a higher estimated event hazard. It does not directly provide a difference in 5-year percentages or a difference in median survival.

Important limitation: the ClinicalTrials.gov record does not include Kaplan-Meier estimates, median disease-free survival, median overall survival, event counts for the efficacy endpoints, or proportional-hazards diagnostics. Those quantities are therefore not reported or reconstructed on this page.

16. Limitations

17. Why This Trial Matters Statistically

IBCSG 23-01 is a useful statistical teaching example because the primary question is framed around non-inferiority rather than superiority. That changes how the hazard ratio, confidence interval, and p-value must be interpreted.

ConceptHow it appears in IBCSG 23-01
RandomizationRandomized parallel-group comparison of two surgical strategies
Intention-to-treat analysisBoth posted efficacy analyses use the ITT population
Time-to-event endpoint5-year disease-free survival is the registered primary endpoint
Hazard ratioReported for both disease-free survival and overall survival
Confidence interval95% CI for the primary HR and 90% CI for the secondary HR
Non-inferiorityPrimary disease-free survival analysis uses a 1-sided non-inferiority test
Log-rank testReported method for the secondary overall-survival analysis
Hypothesis testingPrimary and secondary analyses use different hypothesis types
Design targetsOriginal design description includes target accrual, event target, power, and one-sided significance level
Safety denominatorsSerious adverse events are reported as affected participants among those at risk by arm

The trial therefore illustrates why a statistical result should always be read in the context of the estimand, endpoint, analysis population, hypothesis, confidence interval, and decision rule. A hazard ratio of 0.78 means something different when evaluated in a non-inferiority framework than it would in a conventional superiority analysis.

18. What the Primary Result Does — and Does Not — Establish

What the estimate says

The reported primary HR of 0.78 indicates a lower estimated disease-free-survival event hazard for no axillary lymph node dissection relative to axillary lymph node dissection under the posted analysis.

What the estimate does not say

It does not directly report a 5-year absolute difference, a median disease-free-survival difference, the fraction of patients who benefit, or the probability that an individual patient will remain disease-free.

What the p-value says

The reported P = 0.004 belongs to the one-sided non-inferiority test. It is evidence against the null hypothesis defined by that testing framework; it is not a measure of the size of the treatment effect.

What remains necessary

The exact non-inferiority margin is essential for independently reconstructing the numerical decision rule. Because that margin is not contained in the ClinicalTrials.gov record, this page does not manufacture or infer it.

19. Overall Survival in Context

The secondary overall-survival analysis reports an HR of 0.89, with a two-sided 90% CI of 0.52–1.54 and P = 0.73. The wide interval shows that the point estimate alone is not a complete description of the uncertainty in the comparison.

The overall-survival analysis also uses a different hypothesis type from the primary analysis. It is labeled as a superiority analysis and uses the log-rank test. Therefore, its p-value should be interpreted as a test of superiority rather than as a second test of the primary non-inferiority question.

Statistical distinction: the primary endpoint and secondary endpoint should be kept conceptually separate. The primary analysis addresses non-inferiority for disease-free survival, while the secondary analysis evaluates superiority for overall survival. The ClinicalTrials.gov record does not provide a multiplicity-adjustment scheme connecting these hypotheses.

20. Serious Adverse Events: Statistical Perspective

The serious-adverse-event data are sparse: one affected participant is reported among 447 at risk in the axillary-dissection group, compared with zero among 453 at risk in the no-dissection group.

MeasureAxillary DissectionNo Axillary Dissection
Affected participants10
At risk447453
Reported format1/4470/453

With such small counts, the absence of a reported serious adverse event in one group should not be interpreted as proof that the true event rate is zero. The ClinicalTrials.gov record does not provide a confidence interval or a formal between-group statistical test for this safety measure, so no such inferential comparison is added.

21. A Practical Reading of the IBCSG 23-01 Results

A disciplined reading of the posted results can follow five steps.

  1. Start with the endpoint. The primary endpoint is 5-year disease-free survival, defined as time from randomization to invasive relapse, second primary tumor, or death.
  2. Identify the analysis population. The primary analysis uses intention-to-treat.
  3. Identify the hypothesis. The primary comparison is a one-sided non-inferiority test rather than a superiority test.
  4. Read the effect estimate with its interval. The HR is 0.78 with a 95% CI of 0.55–1.11.
  5. Interpret the p-value within its testing framework. The reported P = 0.004 belongs to the one-sided non-inferiority analysis and should not be interpreted as an effect-size measure.

The same process can then be applied to the secondary overall-survival result, while recognizing that its hypothesis type, confidence level, and statistical method differ from those of the primary endpoint.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue through Clinical Biostats

Explore the statistical concepts behind randomized trials, time-to-event endpoints, hazard ratios, confidence intervals, and non-inferiority designs.

25. Record Summary

IBCSG 23-01 is a randomized parallel-group trial in women with node-negative breast cancer and sentinel lymph node micrometastases, comparing axillary lymph node dissection with no axillary lymph node dissection. Its registered primary endpoint is 5-year disease-free survival, defined as time from randomization to invasive relapse at any site, second primary tumor, or death. The posted primary analysis uses an intention-to-treat population and a one-sided non-inferiority test, reporting an HR of 0.78, a two-sided 95% CI of 0.55–1.11, and P = 0.004. The secondary overall-survival analysis reports an HR of 0.89, a two-sided 90% CI of 0.52–1.54, and P = 0.73 using a log-rank method and a superiority hypothesis.

The central statistical lesson is that these numbers cannot be interpreted independently of their hypothesis frameworks. The primary result is a non-inferiority analysis, so its decision rule depends on a prespecified margin that is not contained in the ClinicalTrials.gov record. The secondary overall-survival result is a superiority analysis and therefore answers a different question. Together, the results illustrate how hazard ratios, confidence intervals, p-values, intention-to-treat analysis, and non-inferiority logic fit together in a randomized time-to-event trial.

Clinical Biostats methodology: A trial-results page should distinguish the numerical evidence reported by the registry from the statistical interpretation applied to that evidence. Where the ClinicalTrials.gov record does not contain a design feature, endpoint estimate, or methodological detail, this page does not infer it from external sources.