This page presents the registered design and primary endpoint of NSABP C-07 together with statistical interpretation of how the endpoint would ordinarily be analyzed. The ClinicalTrials.gov record does not currently contain posted statistical analyses or numerical outcome results.
1. Trial at a Glance
NSABP C-07 was a randomized, parallel phase 3 trial evaluating fluorouracil plus leucovorin with or without oxaliplatin in patients with stage II or stage III colon cancer. The registered primary endpoint was disease-free survival (DFS) rate, defined using time to first documented evidence of colon cancer recurrence, second primary cancer, or death from any cause without prior evidence of colon cancer recurrence or second primary cancer, assessed up to 3 years.
| Feature | NSABP C-07 |
|---|---|
| Phase | Phase 3 |
| Condition | Stage II Colon Cancer; Stage III Colon Cancer |
| Design | Randomized, parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 2472 |
| Number of arms | 2 |
| Start date | 2000-02 |
| Primary completion date | 2005-09 |
| Lead sponsor | National Cancer Institute (NCI) |
| Sponsor type | NIH |
| ClinicalTrials.gov | NCT00004931 |
2. Clinical Question
The central clinical question was whether adding oxaliplatin to fluorouracil plus leucovorin changed the disease-free survival experience of patients with stage II or stage III colon cancer compared with fluorouracil plus leucovorin without oxaliplatin.
Population
Patients with stage II or stage III colon cancer.
Intervention
Fluorouracil plus leucovorin with oxaliplatin.
Comparator
Fluorouracil plus leucovorin without oxaliplatin.
Primary question
Does adding oxaliplatin alter disease-free survival through the registered 3-year assessment period?
3. Trial Design
Fluorouracil + leucovorin + oxaliplatin
- Leucovorin calcium
- Fluorouracil
- Oxaliplatin
Fluorouracil + leucovorin
- Leucovorin calcium
- Fluorouracil
- No oxaliplatin
The registry also lists laboratory biomarker analysis as an intervention categorized as “other.” The registry information available here does not provide additional details that would define this component as a separate randomized treatment arm.
4. Trial Timeline
Study initiation
The registered start date for NSABP C-07 was 2000-02.
Primary completion
The registered primary completion date was 2005-09.
Completed trial
The ClinicalTrials.gov record identifies the study status as COMPLETED.
5. Primary Endpoint
The registry identifies a single registered primary endpoint: DFS rate. The definition is time to first documented evidence of colon cancer recurrence, second primary cancer, or death from any cause without prior evidence of colon cancer recurrence or second primary cancer, assessed up to 3 years.
| Endpoint | Registry definition | Time frame |
|---|---|---|
| DFS rate | Time to first documented evidence of colon cancer recurrence, second primary cancer, or death from any cause without prior evidence of colon cancer recurrence or second primary cancer | Assessed up to 3 years |
6. Planned Analysis
No formal statistical analyses are posted to ClinicalTrials.gov for this record. The registry does, however, define the primary outcome sufficiently to identify the general statistical framework appropriate for the endpoint.
What would ordinarily be measured
For each randomized participant, follow-up would begin at the appropriate trial origin and continue until the first qualifying DFS event or censoring. A qualifying event is the first documented evidence of colon cancer recurrence, a second primary cancer, or death from any cause without prior evidence of colon cancer recurrence or second primary cancer.
Kaplan-Meier estimation
Because DFS is defined as a time to first event, the Kaplan-Meier estimator is a standard way to describe the disease-free survival distribution. It uses the observed event times while retaining information from participants who remain event-free until their last evaluable follow-up.
Here, di is the number of events at time ti, and ni is the number of participants at risk immediately before that time.
Comparing the randomized groups
A time-to-event comparison between the two randomized treatment groups is commonly performed with a log-rank test. A Cox proportional-hazards model can additionally summarize the relative event rate through a hazard ratio and provide a confidence interval around that estimate.
Three-year assessment
The registry specifies assessment up to 3 years. A time-specific DFS estimate at 3 years would describe the estimated proportion remaining free of a qualifying event at that time, while the underlying time-to-event analysis uses the complete event and censoring information available through follow-up.
7. Statistical Methodology
Randomization
Randomization is the central design feature for comparing the oxaliplatin-containing regimen with fluorouracil plus leucovorin without oxaliplatin. In a randomized parallel trial, treatment assignment is determined before outcome follow-up, allowing the treatment groups to be compared according to their assigned interventions.
Time-to-event analysis
The primary endpoint is naturally represented as a time-to-event variable. This matters because participants may have different lengths of observable follow-up and because some participants may remain free of a qualifying event when their observation ends.
Right censoring
When a participant has not experienced the DFS event by the last evaluable observation, the participant's event-free follow-up can contribute information without treating the participant as if the event had occurred. This is the role of right censoring in survival analysis.
Kaplan-Meier estimation
The Kaplan-Meier estimator provides a nonparametric estimate of the probability of remaining event-free over time. It is particularly useful for displaying how the disease-free survival experience develops throughout follow-up rather than reducing the endpoint to a single percentage.
Hazard ratio
If a Cox model is used, the hazard ratio would summarize the relative instantaneous event rate between the randomized groups. A value below 1 would indicate a lower estimated hazard in the numerator treatment group relative to the reference group; a value above 1 would indicate a higher estimated hazard.
The hazard ratio is a relative time-to-event measure. It is not itself a three-year absolute risk, a risk difference, or the percentage of participants who benefit from treatment.
Confidence intervals
A confidence interval around a treatment-effect estimate quantifies statistical uncertainty under the model and sampling framework. It should be considered alongside the point estimate rather than treated as an alternative to the estimate itself.
8. Statistical Methods Explained
Why is DFS a time-to-event endpoint?
The registry definition is explicitly based on the time until the first documented recurrence, second primary cancer, or qualifying death. Two participants can therefore both remain event-free at a particular calendar time while contributing different amounts of follow-up. A time-to-event framework preserves that timing information.
Why use Kaplan-Meier estimation?
Kaplan-Meier estimation is designed for situations in which not every participant experiences the event during observed follow-up. Participants who remain event-free when observation ends can still contribute information through their censoring time.
What does a hazard ratio represent?
A hazard ratio compares estimated instantaneous event rates between treatment groups within a time-to-event model. It summarizes a relative treatment effect over the analyzed follow-up; it does not directly state the absolute probability of remaining disease-free at 3 years.
Why does censoring matter?
Censoring determines how much event-free follow-up each participant contributes. The validity of standard survival methods depends on assumptions about the relationship between censoring and the future event process. Informative or systematically different censoring can complicate interpretation.
Why is a 3-year DFS rate different from a median DFS?
A 3-year DFS rate is a time-specific survival estimate. A median DFS is the time at which the estimated disease-free survival probability reaches 50%, if that point is observed. They describe different aspects of the survival distribution.
Why might a Cox model be used in addition to Kaplan-Meier curves?
Kaplan-Meier estimation describes the survival distribution directly, while a Cox model can provide a compact relative treatment-effect estimate through the hazard ratio and its confidence interval. The two methods therefore answer related but distinct statistical questions.
9. Understanding the Primary Endpoint
Recurrence
Colon cancer recurrence is one of the qualifying first events in the registered DFS definition.
Second primary cancer
A second primary cancer is also explicitly included as a qualifying first DFS event.
Death
Death from any cause is included when it occurs without prior evidence of colon cancer recurrence or second primary cancer.
Three-year window
The registry specifies that DFS is assessed up to 3 years.
This composite definition means that DFS is not limited to recurrence alone. The statistical endpoint combines several clinically distinct events according to a prespecified first-event rule. Consequently, an analysis of DFS answers a broader question than an analysis restricted only to documented colon cancer recurrence.
10. Interpreting a Hypothetical Hazard Ratio Without Inventing a Result
If a Cox analysis were to produce a hazard ratio below 1 for the oxaliplatin-containing regimen, that would indicate a lower estimated instantaneous rate of a qualifying DFS event relative to fluorouracil plus leucovorin under the fitted model.
It would not mean that the same percentage of participants experienced that proportional reduction in individual risk, nor would it directly state the absolute difference in the 3-year DFS rate.
The confidence interval would indicate the statistical precision of the estimated relative treatment effect. A narrow interval would indicate greater precision than a wide interval, all else equal.
If a formal hypothesis test were reported, its p-value would describe the evidence against the specified null hypothesis under the statistical testing framework. It would not measure how large or clinically important the treatment effect is.
11. Analysis Population and the Role of Randomization
The registry identifies the trial as randomized but does not provide, in the information reported here, a detailed analysis-population definition or a statistical analysis plan specifying how deviations from assigned treatment were handled.
| Design feature | What is established in the registry | Statistical implication |
|---|---|---|
| Randomized allocation | Yes | Supports comparison according to randomized treatment assignment. |
| Parallel design | Yes | Participants are evaluated within concurrent randomized treatment groups. |
| Masking | None | The study is registered as an open, unmasked design. |
| Analysis population details | Not reported in the available registry information | The precise handling of treatment deviations and exclusions cannot be characterized from the registry record alone. |
For an efficacy endpoint, preserving the randomized comparison is generally important because randomization establishes the primary basis for causal comparison. At the same time, the exact analysis population and handling of protocol deviations can affect the interpretation of the estimated treatment effect and should be read from the applicable statistical analysis plan when available.
12. What the Registry Reports — and What It Does Not
The ClinicalTrials.gov record establishes the core design, population, interventions, enrollment, timing, and registered primary endpoint. It does not provide posted statistical analyses for the primary endpoint in the record summarized here.
| Item | Registry record |
|---|---|
| Trial phase | Phase 3 |
| Enrollment | 2472 |
| Number of arms | 2 |
| Primary endpoint | DFS rate |
| Primary endpoint time frame | Assessed up to 3 years |
| Posted statistical analyses | None |
| Numerical primary endpoint results | Not reported in the registry record summarized here |
This distinction is important for statistical interpretation. Trial design information identifies what was intended to be evaluated, whereas an estimated hazard ratio, confidence interval, p-value, or time-specific DFS estimate requires an actual analysis result. The two types of information should not be conflated.
13. Missing Data, Censoring, and Follow-Up
The registry definition establishes a time-to-first-event endpoint, which necessarily creates a role for censoring when a participant has not experienced a qualifying event by the end of observable follow-up. However, the registry information summarized here does not specify a detailed censoring algorithm, rules for loss to follow-up, or a missing-data or imputation strategy.
Censoring
Participants without a qualifying event can contribute event-free follow-up until their censoring time.
Event definition
The first documented recurrence, second primary cancer, or specified death determines the DFS event.
Follow-up window
The registered primary endpoint is assessed up to 3 years.
Missing-data rules
The available registry information does not specify an imputation strategy for missing endpoint information.
14. Proportional-Hazards Considerations
If a Cox proportional-hazards model is used to summarize the randomized treatment comparison, the interpretation of a single hazard ratio depends on the model's proportional-hazards assumption. That assumption concerns the relative hazard between groups remaining sufficiently stable over time for a single hazard ratio to provide an appropriate summary.
A Kaplan-Meier analysis provides an important complementary perspective because it shows the estimated disease-free survival experience over time rather than compressing the comparison into one relative measure. When hazards vary materially over time, additional summaries such as time-specific survival estimates can be more informative than a single hazard ratio.
15. Composite Endpoint Considerations
DFS combines several possible first events: colon cancer recurrence, second primary cancer, and death from any cause without prior evidence of either of the specified cancer events. Composite endpoints can increase the number of observed events, but interpretation depends on what contributes to the composite and whether the treatment has a similar relationship with its individual components.
| Component | Role in registered DFS definition |
|---|---|
| Colon cancer recurrence | Qualifying first documented event |
| Second primary cancer | Qualifying first documented event |
| Death from any cause without prior qualifying cancer event | Qualifying first event |
The registry definition therefore establishes a composite time-to-first-event endpoint rather than a recurrence-only endpoint. Any detailed interpretation of the individual components would require component-specific results, which are not posted in the registry record summarized here.
16. Non-Inferiority, Bayesian Methods, and Multiplicity
The registry information summarized here does not identify a non-inferiority margin, a Bayesian analysis, or a multiplicity-adjustment strategy for the registered primary endpoint.
| Design topic | Information available in the registry record |
|---|---|
| Non-inferiority margin | Not reported |
| Bayesian methods | Not reported |
| Multiplicity strategy | Not reported |
| Interim analysis strategy | Not reported |
| Stratification factors | Not reported |
| Missing-data / imputation strategy | Not reported |
| Crossover | Not reported |
These are important design features when reconstructing a complete statistical analysis, but they should not be inferred from the existence of a randomized phase 3 design alone.
17. Statistical Interpretation of a Three-Year DFS Rate
A three-year DFS estimate answers a different question from a hazard ratio. The DFS rate at 3 years is an absolute time-specific estimate: the estimated proportion of participants who remain free of the qualifying first event at that point in time.
Absolute measure
The three-year DFS rate describes the estimated event-free probability at a specified time.
Relative measure
A hazard ratio compares event rates between treatment groups within a time-to-event model.
Time dimension
Both measures use follow-up over time, but they summarize that information differently.
Uncertainty
A complete statistical report would normally pair estimates with measures of uncertainty such as confidence intervals.
For readers interpreting a future numerical analysis of NSABP C-07, the most informative presentation would generally distinguish the estimated DFS curves, the three-year DFS rate, and the relative treatment effect rather than treating any one statistic as a complete summary.
18. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record does not contain formal posted statistical analyses for the registered primary endpoint.
- No numerical outcome estimates: the registry information summarized here does not provide a numerical DFS estimate, hazard ratio, confidence interval, or p-value.
- Limited analysis-plan detail: the available registry information does not specify the detailed statistical analysis population, censoring rules, missing-data strategy, stratification factors, multiplicity procedures, or interim-analysis framework.
- Composite endpoint: DFS combines recurrence, second primary cancer, and a specified category of death, so a DFS result would represent the combined endpoint rather than recurrence alone.
- Unmasked design: the registry identifies masking as none. For an endpoint involving documented clinical events, lack of masking can be relevant to how outcome ascertainment and other trial procedures are interpreted, although the extent of any resulting bias requires additional study information.
- Generalizability: the registered population is defined specifically by stage II or stage III colon cancer, so the applicability of results to other populations would require separate consideration.
- Endpoint time frame: the primary endpoint is assessed up to 3 years, so longer-term disease-free outcomes are not defined by this registered primary time frame.
19. Why This Trial Matters Statistically
NSABP C-07 is a useful teaching example because its central question is naturally expressed through randomized time-to-event analysis. The trial combines a clear treatment comparison with a composite DFS endpoint and a prespecified three-year assessment window.
| Concept | How it appears in NSABP C-07 |
|---|---|
| Randomization | Participants were allocated randomly to one of 2 arms. |
| Parallel design | The registry identifies a parallel trial design. |
| Time-to-event analysis | The primary endpoint is explicitly defined as time to first qualifying event. |
| Kaplan-Meier estimation | A standard framework for estimating DFS over time. |
| Hazard ratio | A standard relative measure for a Cox time-to-event comparison, if such a model is used. |
| Confidence interval | Provides uncertainty around a reported treatment-effect estimate. |
| Censoring | Relevant when participants remain event-free when their observable follow-up ends. |
| Composite endpoint | DFS includes recurrence, second primary cancer, and a specified death event. |
| Three-year assessment | The registered primary endpoint is assessed up to 3 years. |
| Open-label design | Masking is listed as none. |
20. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The registered primary endpoint is a time-to-first-event measure. Its analysis would ordinarily use survival-analysis methods that account for event timing and censoring rather than a simple comparison of unadjusted proportions.
Clinical interpretation
The clinical question concerns whether adding oxaliplatin to fluorouracil plus leucovorin changes the disease-free course of patients with stage II or stage III colon cancer. The registry defines the endpoint but does not provide the numerical outcome needed to quantify that effect.
21. What a Complete Numerical DFS Analysis Would Contain
A complete statistical presentation of the registered primary endpoint would ordinarily connect several pieces of evidence rather than relying on a single number.
| Component | Purpose |
|---|---|
| Kaplan-Meier DFS curve | Shows how estimated disease-free survival changes over follow-up. |
| Three-year DFS estimate | Provides the registry-specified time-specific outcome. |
| Confidence interval | Quantifies uncertainty around the DFS estimate. |
| Between-group comparison | Tests whether the randomized treatment groups differ under the prespecified statistical framework. |
| Hazard ratio, if used | Summarizes the relative event rate between treatment groups. |
| Confidence interval for the hazard ratio | Describes statistical precision of the relative treatment estimate. |
| Event counts | Describe how many qualifying DFS events occurred in each treatment group. |
These measures are complementary. The Kaplan-Meier curve provides the temporal pattern, the three-year DFS rate provides an absolute time-specific summary, and a hazard ratio provides a relative comparison under a time-to-event model.
22. Sources
- ClinicalTrials.gov: NCT00004931 — NSABP C-07.
Continue through the Clinical Biostats clinical trial library
Explore additional clinical trial analyses and statistical methods across randomized study designs, endpoints, and therapeutic areas.
23. Record Summary
NSABP C-07 is a randomized phase 3, parallel trial with 2472 participants and 2 treatment arms evaluating fluorouracil plus leucovorin with or without oxaliplatin in stage II or stage III colon cancer. Its registered primary endpoint is DFS rate, defined as time to first documented evidence of colon cancer recurrence, second primary cancer, or death from any cause without prior evidence of colon cancer recurrence or second primary cancer, assessed up to 3 years.
Statistically, the endpoint is naturally suited to time-to-event methods such as Kaplan-Meier estimation and, where prespecified, log-rank and Cox-model comparisons. The key interpretive distinction is between an absolute three-year DFS estimate and a relative treatment-effect measure such as a hazard ratio. The ClinicalTrials.gov record does not post formal statistical analyses or numerical primary endpoint results for this record, so the statistical interpretation here focuses on the registered design and the appropriate framework for analyzing its primary endpoint.