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
BIG 1-98 was a completed randomized phase 3 trial in postmenopausal women with breast cancer. The registry describes a parallel, quadruple-masked design with 8028 enrolled participants and four arms involving letrozole and tamoxifen citrate.
| Feature | BIG 1-98 |
|---|---|
| Trial acronym | BIG 1-98 |
| ClinicalTrials.gov identifier | NCT00004205 |
| Phase | Phase 3 |
| Status | Completed |
| Condition | Breast Cancer |
| Population | Postmenopausal women with breast cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 8028 |
| Number of arms | 4 |
| Interventions | Letrozole; tamoxifen citrate |
| Lead sponsor | ETOP IBCSG Partners Foundation |
| Sponsor type | Network |
2. Clinical Question
The registry identifies BIG 1-98 as a randomized treatment trial in postmenopausal women with breast cancer, with letrozole and tamoxifen citrate listed as the study interventions. The primary statistical question is centered on disease-free survival: how long participants remain free of the prespecified disease-related or mortality events after randomization.
Population
Postmenopausal women with breast cancer.
Intervention
Letrozole is identified in the registry as a study drug intervention.
Comparator
Tamoxifen citrate is identified in the registry as the other study drug intervention.
Primary question
How does treatment assignment relate to disease-free survival, as defined by the registry, through the specified follow-up period?
3. Trial Design
Trial timeline
Study start
The registry records March 1998 as the study start date.
Primary completion
The registry records December 2008 as the primary completion date.
Primary endpoint assessment frame
The registered disease-free survival endpoint is assessed up to the end of 2015, for a maximum of 17 years.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Disease-free survival | Time from randomization to recurrence, including recurrence restricted to the breast after breast conserving treatment, metastasis, appearance of a second primary tumor, or death from any cause, whichever occurs first. | Up to end of year 2015, for a maximum of 17 years |
Why the endpoint definition matters
Disease-free survival is a time-to-event endpoint. The analysis is not simply a comparison of the proportion of participants who experienced an event. It incorporates both the occurrence of the event and the time at which the event occurs.
The registry definition combines several clinically distinct events into a single first-event endpoint. Recurrence, metastasis, appearance of a second primary tumor, and death from any cause can therefore all terminate disease-free survival. The phrase "whichever occurs first" is statistically important: once one qualifying event occurs, that participant has reached the endpoint even if other potential events could occur later.
A participant who reaches the end of follow-up without a qualifying event contributes follow-up information without being treated as if the event occurred. This distinction is fundamental to survival analysis.
5. Planned Analysis
The ClinicalTrials.gov record defines disease-free survival as the primary endpoint but does not report posted statistical analyses or numerical efficacy results. For an endpoint of this type, the standard statistical framework is a time-to-event analysis that accounts for both event timing and censoring.
A conventional analysis would begin by estimating the disease-free survival function over time, commonly using the Kaplan-Meier method. Treatment groups can then be compared with a time-to-event hypothesis test such as the log-rank test, while a Cox proportional-hazards model can provide a hazard ratio and confidence interval describing the relative event rate over follow-up.
These methods are complementary. Kaplan-Meier estimation describes the disease-free survival experience over time; the log-rank test evaluates evidence of a difference between survival curves; and the Cox model summarizes the relative event rate through a hazard ratio.
Kaplan-Meier
Estimates the probability of remaining free of the composite endpoint over time while accounting for participants whose follow-up ends before an event occurs.
Log-rank comparison
Provides a standard way to compare time-to-event distributions between randomized treatment groups.
Cox model
Can express the treatment comparison as a hazard ratio with a confidence interval, provided the proportional-hazards framework is appropriate.
Censoring
Participants without a qualifying event by their last relevant follow-up time are typically censored rather than counted as events.
6. Statistical Methods Explained
What does disease-free survival measure?
Disease-free survival measures the time from randomization until the first qualifying event in the endpoint definition. In BIG 1-98, those events include recurrence, metastasis, appearance of a second primary tumor, or death from any cause.
This means DFS is not equivalent to simply asking whether a participant eventually developed recurrent disease. The timing of the event matters. Two participants who both experience a qualifying event contribute differently if one experiences it substantially earlier than the other.
Why use a time-to-event method rather than a simple proportion?
A simple proportion at the end of follow-up would discard much of the information about when events occurred and would be difficult to interpret when participants have different amounts of follow-up. Survival methods preserve the ordering and timing of observed events while handling incomplete observation appropriately.
What is censoring?
Censoring occurs when a participant's disease-free survival time is not observed all the way to a qualifying event during the period of observation. For example, if a participant reaches the end of relevant follow-up without an event, the analysis can use the information that the participant remained event-free through the last observed time.
Censoring does not mean that the participant is assumed to have remained disease-free forever. It means that the observed follow-up establishes a lower bound on the participant's disease-free survival time.
What does a Kaplan-Meier curve show?
A Kaplan-Meier curve estimates the probability of remaining event-free as time progresses. Each observed event can produce a downward step, while censoring changes the number of participants considered at risk for subsequent times without itself being treated as an endpoint event.
Here, di represents the number of events at event time ti, while ni represents the number at risk immediately before that time.
What would a hazard ratio mean for disease-free survival?
A hazard ratio compares the estimated instantaneous event rates between treatment groups under a fitted survival model. A hazard ratio below 1 would indicate a lower estimated instantaneous rate of reaching the composite disease-free survival endpoint in the numerator group relative to the reference group.
It would not mean that the same percentage of participants avoided recurrence, that every participant experienced the same proportional reduction in risk, or that the hazard ratio equals a difference in disease-free survival percentages.
Why does the proportional-hazards assumption matter?
A Cox proportional-hazards model summarizes the treatment comparison with a hazard ratio under an assumption that the relative hazards are reasonably stable over time. If the relative treatment effect changes substantially during follow-up, a single hazard ratio may not fully describe the pattern seen in the survival curves.
7. Interpreting Disease-Free Survival Correctly
The registry-defined DFS endpoint combines recurrence, metastasis, second primary tumor, and death from any cause. Consequently, a DFS event does not necessarily mean that breast cancer recurrence occurred. The event could be any one of the components specified in the endpoint definition.
DFS is a time-to-event outcome. An analysis should therefore preserve information about when events occur rather than reducing the entire study to a single end-of-follow-up proportion.
Because the registry describes randomized allocation, the treatment comparison is anchored to treatment assignment rather than to the treatment participants may subsequently receive or the characteristics they have after randomization.
If a future analysis reported a hazard ratio, that estimate would provide a relative comparison of event rates. It would not by itself tell a reader the absolute probability of remaining disease-free at a particular time. Kaplan-Meier estimates at clinically meaningful time points would provide that complementary perspective.
8. What a Complete DFS Analysis Would Report
A statistically informative presentation of this trial's primary endpoint would ideally bring several complementary quantities together rather than relying on a single test statistic.
| Component | Statistical role | What it communicates |
|---|---|---|
| Number randomized | Defines the analysis population | Scale of the randomized comparison |
| Number of DFS events | Defines observed information | How much event information contributes to the analysis |
| Kaplan-Meier curve | Descriptive time-to-event estimate | How disease-free survival changes over time |
| Time-specific DFS estimate | Absolute effect description | Estimated proportion remaining event-free at a specified time |
| Median DFS, if estimable | Summary of event timing | Time by which approximately half the analysis population has experienced the endpoint |
| Hazard ratio | Relative treatment effect | Relative instantaneous event-rate comparison under the model |
| 95% confidence interval | Precision | Uncertainty around the estimated treatment effect |
| Statistical test | Evidence against the null hypothesis | How compatible the observed comparison is with the specified null model |
These quantities answer different questions. A confidence interval addresses precision, while a p-value addresses evidence against a specified null hypothesis. Neither one is a direct measure of clinical importance. The hazard ratio describes a relative effect, whereas absolute disease-free survival estimates describe the outcome on the original probability scale.
9. Analysis of the Four-Arm Structure
The registry classifies BIG 1-98 as a four-arm randomized parallel trial. That design feature is statistically important because the primary comparison cannot automatically be treated as though the study had only two randomized groups.
Multiple randomized groups
Four arms create more than one possible pairwise treatment comparison. The appropriate comparison structure depends on the prespecified statistical objectives.
Multiple testing
If several confirmatory comparisons are performed, the interpretation of statistical significance depends on the prespecified approach to multiplicity and type I error control.
Overall trial question
The registry identifies letrozole and tamoxifen citrate as interventions but does not specify the complete arm-level comparison structure in the record summarized here.
Why the distinction matters
A four-arm design can contain treatment comparisons that answer different clinical questions. Combining them without preserving their original definitions can obscure the intended estimand.
10. Masking and Statistical Bias
The registry describes BIG 1-98 as quadruple masked. Masking is primarily a design feature rather than a statistical test, but it can influence the risk of bias in trial conduct and outcome assessment.
For a time-to-event endpoint such as DFS, masking can be particularly relevant to how events are assessed and classified. The registry's endpoint includes several clinically defined events, including recurrence and metastasis. Keeping assessment procedures appropriately separated from knowledge of treatment assignment can reduce opportunities for treatment knowledge to influence event classification.
Randomization addresses a different problem. It establishes the framework for comparing treatment assignments, while masking helps reduce the potential influence of treatment knowledge on trial conduct and assessment. These are complementary protections rather than interchangeable ones.
11. Censoring and Follow-Up
The primary endpoint is assessed from randomization through the end of 2015, with a maximum stated duration of 17 years. Such a long observation window makes censoring an important part of the statistical structure.
| Observation status | Statistical meaning |
|---|---|
| Qualifying event occurs | The participant reaches the DFS endpoint at the event time. |
| No qualifying event by last follow-up | The participant contributes disease-free follow-up through the observed time and is censored thereafter under a conventional survival analysis. |
| Multiple possible events | The first qualifying event determines the DFS endpoint because the registry specifies "whichever occurs first." |
Censoring requires assumptions for standard survival analysis. In particular, the reason a participant leaves observation should not be systematically related to the participant's unobserved future event risk in a way that invalidates the analysis. The registry information summarized here does not report the detailed censoring rules or missing-data procedures used for the endpoint.
12. Composite Endpoint Considerations
Composite endpoints can improve statistical efficiency by allowing several clinically relevant events to contribute to a single endpoint. They also require careful interpretation because a treatment effect on the composite can arise from its effects on different components.
| DFS component | Role in the endpoint |
|---|---|
| Recurrence | A qualifying event that ends disease-free survival. |
| Recurrence restricted to the breast after breast conserving treatment | Explicitly included within the registry's recurrence definition. |
| Metastasis | A qualifying event that ends disease-free survival. |
| Second primary tumor | A qualifying event that ends disease-free survival. |
| Death from any cause | A qualifying event that ends disease-free survival regardless of cause. |
A composite endpoint should therefore be interpreted as the endpoint that was actually defined, not as a synonym for any one component. In particular, the inclusion of death from any cause means that DFS and a narrower recurrence-only endpoint would not necessarily answer the same statistical question.
13. Confidence Intervals and P-Values
If a formal treatment-effect estimate is reported for DFS, its confidence interval and p-value should be interpreted as different statistical quantities.
Confidence interval
A confidence interval communicates uncertainty around an estimated treatment effect. Narrower intervals generally indicate greater statistical precision than wider intervals, all else equal.
P-value
A p-value evaluates how compatible the observed data are with a specified null hypothesis under the statistical model. It is not a measure of treatment-effect magnitude.
Effect size
A hazard ratio, absolute survival difference, or other effect estimate describes the size and direction of an observed treatment comparison.
Clinical importance
Statistical evidence and clinical importance are related but distinct questions. A small p-value does not automatically establish that an effect is large or clinically important.
14. Multiplicity in a Four-Arm Trial
Multiplicity becomes relevant whenever a study evaluates multiple hypotheses, endpoints, treatment comparisons, or interim looks in a manner that could increase the probability of at least one false-positive conclusion.
BIG 1-98 has four randomized arms according to the registry. That design alone does not establish which comparisons were confirmatory or how type I error was allocated. Those details would need to come from the prespecified statistical analysis plan.
When multiple hypotheses are tested, the overall probability of making at least one false-positive rejection can exceed the nominal level used for a single test unless an appropriate multiplicity strategy is applied.
Common approaches include hierarchical testing, Bonferroni-type adjustments, alpha allocation, or other prespecified procedures. The appropriate approach depends on the trial's intended confirmatory questions rather than on the number of arms alone.
15. Why This Trial Matters Statistically
BIG 1-98 is a useful teaching example because the registry combines a large randomized phase 3 population, a four-arm parallel design, quadruple masking, and a long-term composite time-to-event endpoint.
| Concept | How it appears in BIG 1-98 |
|---|---|
| Randomization | The registry classifies allocation as randomized. |
| Parallel design | The registry identifies a parallel design model with four arms. |
| Masking | The registry describes the study as quadruple masked. |
| Time-to-event analysis | The primary endpoint is disease-free survival measured from randomization to the first qualifying event. |
| Composite endpoint | DFS includes recurrence, metastasis, second primary tumor, and death from any cause. |
| Censoring | Longitudinal follow-up requires appropriate handling of participants who have not experienced the endpoint by their last observation. |
| Hazard ratio | A conventional Cox analysis can summarize relative event rates if its assumptions are appropriate. |
| Multiple arms | Four randomized arms can create multiple treatment comparisons and therefore potential multiplicity considerations. |
| Long follow-up | The registered DFS assessment extends through the end of 2015, for a maximum of 17 years. |
16. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record does not contain posted statistical analyses or numerical results for the registered primary endpoint.
- Limited arm-level detail: the registry information summarized here identifies four arms and the two study drugs but does not provide the arm-level treatment schedules or sample sizes.
- Composite endpoint: DFS combines several event types, so a DFS event does not uniquely identify recurrence, metastasis, second primary tumor, or death.
- Potential multiplicity: a four-arm randomized design can involve multiple possible treatment comparisons. The statistical interpretation of multiple comparisons depends on the prespecified testing strategy.
- Censoring assumptions: standard time-to-event methods rely on assumptions concerning censoring and the relationship between observed follow-up and future event risk.
- Proportional hazards: a single Cox hazard ratio may not adequately summarize treatment effects if relative hazards change substantially over time.
- Long observation window: the maximum 17-year endpoint frame creates substantial opportunity for follow-up patterns, censoring, and changes in the observed event experience to affect the statistical analysis.
- Endpoint definition: the inclusion of death from any cause means DFS should not be interpreted as a pure measure of breast cancer recurrence.
17. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The primary endpoint is a randomized time-to-event outcome. Appropriate analysis focuses on the distribution of time to the first qualifying DFS event, with treatment comparisons expressed through survival functions, event-time comparisons, and potentially hazard ratios with confidence intervals.
Clinical interpretation
The endpoint reflects a broad disease-free outcome that includes recurrence, metastasis, second primary tumor, and death from any cause. Clinical interpretation therefore requires attention to which type of event contributes to any observed difference.
18. A Practical Framework for Reading the Primary Analysis
When numerical DFS results become available, a reader can evaluate them in a structured sequence rather than focusing on the p-value alone.
First, verify the exact endpoint definition. Second, examine the Kaplan-Meier curves or equivalent time-specific estimates. Third, evaluate the effect estimate, such as a hazard ratio, together with its confidence interval. Fourth, consider the event count and amount of follow-up. Finally, determine whether the magnitude and pattern of the observed effect are clinically meaningful rather than treating statistical significance as the entire conclusion.
19. Statistical Concepts in This Trial
Readers can use this trial as a practical example of randomized survival analysis, composite endpoints, censoring, and multi-arm trial design.
20. Sources
- ClinicalTrials.gov: BIG 1-98, NCT00004205.
Continue through Clinical Biostats
Explore additional clinical trial analyses, statistical tutorials, and quantitative tools for clinical research.
21. Record Summary
BIG 1-98 is a randomized phase 3, parallel, quadruple-masked trial enrolling 8028 participants in four arms to study treatment with letrozole and tamoxifen citrate in postmenopausal women with breast cancer. Its registered primary endpoint is disease-free survival, defined as time from randomization to recurrence, including recurrence restricted to the breast after breast conserving treatment, metastasis, appearance of a second primary tumor, or death from any cause, whichever occurs first.
Statistically, the key feature is that DFS is a composite time-to-event endpoint. A rigorous analysis must therefore account for event timing, censoring, and the multiple possible components of the endpoint. Kaplan-Meier estimation provides a way to describe disease-free survival over time; a log-rank framework can compare treatment groups; and a Cox model can express a relative event-rate comparison through a hazard ratio when its assumptions are appropriate.
The four-arm randomized structure also makes the distinction between individual treatment comparisons and the overall trial design important. If several confirmatory hypotheses are evaluated, multiplicity and type I error control become part of the statistical interpretation. Finally, the long registered follow-up period means that the timing and completeness of follow-up are integral to understanding the eventual disease-free survival analysis.