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 R-04 was a randomized, parallel, phase 3 treatment trial in colorectal cancer. The registry describes radiation therapy with either capecitabine or fluorouracil, with or without oxaliplatin, before surgery in patients with resectable rectal cancer. The registered primary endpoint evaluates loco-regional disease control through evidence of tumor at 3 years.
| Feature | NSABP R-04 |
|---|---|
| Phase | Phase 3 |
| Condition | Colorectal Cancer |
| Population | Patients with resectable rectal cancer |
| Design | Randomized, parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1608 |
| Number of arms | 4 |
| Lead sponsor | NSABP Foundation Inc |
| Sponsor type | Network |
| Trial status | Unknown |
| ClinicalTrials.gov | NCT00058474 |
2. Clinical Question
The clinical question is how different preoperative treatment strategies involving radiation therapy, capecitabine or fluorouracil, and oxaliplatin perform in patients with resectable rectal cancer. The registry identifies loco-regional disease control as the primary endpoint, assessed by evidence of tumor at 3 years.
Population
Patients with resectable rectal cancer within the registry's colorectal cancer population.
Interventions
Radiation therapy with capecitabine or fluorouracil, with or without oxaliplatin.
Comparator structure
The study uses four randomized treatment arms. The registry title identifies the chemotherapy and oxaliplatin treatment factors but does not provide arm-specific sample sizes in the record summarized here.
Primary question
How does randomized preoperative treatment affect loco-regional disease control, defined through evidence of tumor at 3 years?
3. Trial Design
The registry describes four arms but does not provide arm-specific enrollment counts in the record summarized here. It also does not provide enough information here to reconstruct the exact assignment structure or treatment sequence for each arm beyond the interventions named in the trial title.
4. Treatment Structure
Capecitabine
- Capecitabine is one of the chemotherapy interventions identified in the registry.
- It is evaluated with radiation therapy before surgery.
- The trial title indicates that oxaliplatin may be included or omitted.
Fluorouracil
- Fluorouracil is the other chemotherapy intervention identified in the registry.
- It is evaluated with radiation therapy before surgery.
- The trial title indicates that oxaliplatin may be included or omitted.
5. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Loco-regional disease control as assessed by evidence of tumor at 3 years | Loco-regional disease control as assessed by evidence of tumor at 3 years | Time from randomization to first local recurrence up to 3 years |
The primary endpoint is therefore a time-to-event endpoint. The event is the first local recurrence, and follow-up is defined from randomization through 3 years. This distinction is important: a time-to-event endpoint contains information not only about whether an event occurred, but also about when it occurred and whether a participant remained event-free through the available observation period.
6. Planned Analysis
No statistical analyses or numerical efficacy results are posted in the ClinicalTrials.gov record summarized here. The registry identifies the primary endpoint as time from randomization to first local recurrence up to 3 years, but it does not provide a posted hazard ratio, confidence interval, p-value, event count, or treatment-specific estimate for that endpoint.
How this endpoint would typically be analysed
For a time-to-first-local-recurrence endpoint, a standard analysis would generally describe the event-time distribution using the Kaplan-Meier method. This approach allows participants who have not experienced local recurrence by the end of their observed follow-up to contribute information without treating them as if they experienced the event.
Here, di is the number of events at time ti, while ni is the number of participants at risk immediately before that time.
A treatment comparison for this type of endpoint could typically use a log-rank test, with a Cox proportional-hazards model providing a relative treatment-effect estimate such as a hazard ratio. The exact inferential model, stratification factors, hypothesis-testing framework, multiplicity procedure, and analysis population are not reported in the registry information summarized here.
What the 3-year time frame means
The phrase "up to 3 years" establishes the registry's endpoint window. It does not mean that every participant necessarily has an observed local-recurrence status for exactly 3 years. In a time-to-event analysis, participants can contribute information until recurrence, another censoring event, or the end of the defined observation period.
Event
The registered event is first local recurrence.
Time origin
Time is measured from randomization.
Administrative horizon
The registered time frame extends up to 3 years.
Statistical form
The endpoint is naturally suited to time-to-event methods rather than a simple cross-sectional proportion.
7. Statistical Methodology
Time-to-event analysis
The primary endpoint is explicitly defined as time from randomization to first local recurrence. That structure makes survival-analysis methodology relevant because participants can have different lengths of observed follow-up and can be censored without experiencing the event.
A time-to-event analysis uses the timing of events rather than reducing each participant to a simple yes/no indicator at a single time point. This is particularly important when some participants remain free of local recurrence for different lengths of time.
Kaplan-Meier estimation
The Kaplan-Meier estimator provides an estimate of the probability of remaining free of the event over time. If local recurrence is the event of interest, the resulting curve can be interpreted as the estimated probability of remaining without first local recurrence through each time point, subject to the censoring assumptions of the analysis.
For an event-time variable T, the survival function represents the probability that the event time exceeds t. In this trial's endpoint, the event is first local recurrence.
Log-rank comparison
A log-rank test is commonly used to compare event-time distributions between randomized treatment groups. Unlike a simple comparison of proportions at 3 years, it can incorporate the timing of recurrences throughout follow-up and appropriately handle right-censored observations.
Cox proportional-hazards model
A Cox model can express the relative difference in the instantaneous event rate between treatment groups as a hazard ratio. A hazard ratio below 1 would indicate a lower estimated instantaneous rate of first local recurrence in the treatment group relative to the comparator, under the fitted model.
A hazard ratio is not the same as a risk ratio, an absolute risk difference, or the probability that an individual patient will avoid recurrence.
Censoring
Time-to-event methods rely on censoring rules. A participant who has not experienced first local recurrence by the end of observed follow-up is not automatically treated as having experienced the event. Instead, the participant contributes event-free follow-up until the censoring time.
Analysis of a randomized trial
Randomization establishes the framework for comparing the treatment groups because treatment assignment is determined independently of the outcome process under the trial design. For efficacy, the randomized comparison is generally the central basis for inference. The registry information available here does not specify the exact analysis population or statistical model used for the primary endpoint.
8. Statistical Methods Explained
Why is this endpoint a survival-analysis problem?
Because the primary endpoint is defined as time from randomization to first local recurrence. The analysis therefore needs to account for both whether recurrence occurred and the amount of time until recurrence or censoring. A simple percentage of participants with recurrence would discard part of that timing information.
What does a Kaplan-Meier curve show?
A Kaplan-Meier curve estimates the probability of remaining free of the event over time. For NSABP R-04, the relevant event is first local recurrence. The curve can therefore describe how the estimated event-free probability changes from randomization through the registered 3-year time frame.
Why not simply compare the percentage with no recurrence at 3 years?
A fixed-time comparison can be useful, but it does not fully use information from participants whose follow-up ends before 3 years. Time-to-event methods can incorporate their observed follow-up through censoring while also using the timing of recurrences that occur before the endpoint horizon.
What would a hazard ratio mean?
A hazard ratio compares the modeled instantaneous event rates between treatment groups. If a hazard ratio were below 1, it would indicate a lower estimated instantaneous rate of first local recurrence for the treatment group relative to the comparator. It would not mean that the same percentage of patients avoided recurrence.
What does a confidence interval add?
A confidence interval describes statistical uncertainty around an estimated treatment effect. A narrow interval indicates greater precision than a wide interval, all else equal. It does not describe the range of effects that individual patients will experience.
Why is the proportional-hazards assumption relevant?
The Cox proportional-hazards model assumes that the relative hazard between comparison groups is stable over time in the form represented by the model. If that assumption is substantially violated, a single hazard ratio can become a less complete summary of the treatment difference. The registry information summarized here does not report a proportional-hazards assessment.
9. Four-Arm Design and Statistical Comparisons
The study has four randomized treatment arms. That structure creates an important statistical distinction between the existence of multiple treatment groups and the particular comparisons used for inference.
| Design feature | Registry information | Statistical implication |
|---|---|---|
| Number of arms | 4 | More than two randomized groups may create multiple possible pairwise comparisons. |
| Allocation | Randomized | Randomization provides the basis for comparing assigned treatment groups. |
| Design model | Parallel | Participants are assigned to treatment arms rather than serving as their own longitudinal treatment controls. |
| Masking | None | The trial is not described as blinded. |
With four arms, the number and nature of formal comparisons matter. If several pairwise hypotheses are tested, the overall probability of at least one false-positive finding can increase unless multiplicity is addressed. However, the registry information summarized here does not specify a multiplicity-adjustment strategy or a hierarchy of hypotheses.
10. Non-Inferiority, Crossover, Factorial, and Bayesian Methods
The registry information summarized here does not report a non-inferiority margin, crossover procedure, interim-analysis plan, missing-data or imputation strategy, stratification factors, or Bayesian analysis. These features should not be inferred simply from the randomized four-arm design.
Non-inferiority margin
No non-inferiority margin is reported in the registry information summarized here.
Crossover
No crossover procedure is reported in the registry information summarized here.
Factorial structure
The title identifies chemotherapy and oxaliplatin treatment factors, but the registry information does not explicitly describe the statistical design as factorial.
Bayesian methods
No Bayesian analysis is reported in the registry information summarized here.
11. Missing Data and Censoring
For the primary endpoint, censoring is an inherent part of the time-to-event framework. A participant who has not experienced first local recurrence by the end of available follow-up can contribute information without being classified as a recurrence event at that time.
The statistical treatment of other forms of missing information is different from ordinary time-to-event censoring. For example, missing covariates, incomplete assessments, or missing follow-up information may require prespecified rules. The registry information summarized here does not report a missing-data or imputation strategy.
12. What a Primary Analysis Would Report
For a primary time-to-first-local-recurrence endpoint, a statistically informative report would normally include both an absolute description of the event-time distribution and a comparison between randomized groups.
| Statistic | What it would describe |
|---|---|
| Number of first local recurrences | Observed number of primary endpoint events. |
| Kaplan-Meier estimate | Estimated probability of remaining free of first local recurrence over time. |
| Median event-free time | The time at which the estimated event-free probability reaches 50%, if the curve reaches that level. |
| Hazard ratio | Relative treatment effect from a time-to-event model. |
| 95% confidence interval | Uncertainty around the estimated treatment effect. |
| P-value | Evidence against a specified null hypothesis under the chosen testing framework; it is not a measure of effect size. |
None of these numerical primary-endpoint results are posted in the ClinicalTrials.gov record summarized here. The registry provides the endpoint definition and time frame but does not provide a posted statistical analysis for the endpoint.
13. How to Interpret a Future Hazard Ratio
If a future reported analysis produced an OS-like hazard ratio of 0.70, for example, the appropriate interpretation would be that the estimated instantaneous rate of the specified event was approximately 30% lower in the treatment group than in the comparator group under the fitted model. This is a generic illustration of how hazard ratios are interpreted, not a result from NSABP R-04.
It would not mean that 30% of patients avoided recurrence, that every patient experienced a 30% reduction in risk, or that the absolute probability of recurrence was reduced by exactly 30 percentage points.
Time-specific event-free probabilities and other absolute measures provide information that a relative hazard ratio cannot provide by itself. For a three-year endpoint, an absolute estimate at or near the three-year horizon can be particularly useful for communicating the magnitude of the difference in clinically interpretable terms.
14. Statistical Interpretation of the Trial Design
The strongest statistical feature established by the registry is the randomized parallel design. Randomization is important because it is intended to balance measured and unmeasured prognostic factors across treatment assignments in expectation. This makes the between-arm comparison a central component of causal inference.
The four-arm structure also makes the definition of the estimand particularly important. The registry identifies the treatment components but does not provide the formal statistical contrast or hierarchy of hypotheses in the information summarized here. Consequently, the interpretation of any eventual pairwise treatment effect depends on the exact comparison specified in the prespecified statistical analysis plan.
The absence of masking is another design feature with statistical relevance. For an objective time-to-local-recurrence endpoint, outcome ascertainment may be less susceptible to subjective influence than a patient-reported or investigator-assessed symptom endpoint, but lack of masking can still affect other aspects of trial conduct. The registry does not provide enough information here to quantify such effects.
15. Endpoint-Specific Statistical Reasoning
Time origin
Randomization is the starting point for the registered primary endpoint.
Event definition
The first local recurrence is the event that ends event-free follow-up.
Time horizon
The registered endpoint is evaluated up to 3 years.
Primary estimand concept
The central comparison concerns the distribution of time to first local recurrence across randomized treatment groups.
This endpoint structure is different from an endpoint that simply asks whether a patient has recurrence at a fixed visit. The time-to-event formulation retains information about the ordering and timing of recurrences, which is why survival-analysis tools are typically appropriate.
16. ClinicalTrials.gov Record Timeline
Trial start
The registry lists 2004-07 as the study start date.
Randomized treatment evaluation
The study is registered as a phase 3 randomized, parallel treatment trial with four arms and no masking.
Primary completion
The registry lists 2013-08 as the primary completion date.
Current registry status
The ClinicalTrials.gov record summarized here lists the study status as UNKNOWN.
17. Limitations
- No posted statistical analysis: the ClinicalTrials.gov record does not provide numerical results, effect estimates, confidence intervals, p-values, or event counts for the registered primary endpoint.
- Limited arm-level detail: the record identifies four treatment arms and names the interventions, but the information summarized here does not provide arm-specific enrollment counts or detailed treatment assignment procedures.
- Unspecified analysis model: the registry defines the primary endpoint but does not report the exact statistical model, stratification factors, or hypothesis-testing procedure in the information summarized here.
- Unspecified multiplicity strategy: four treatment arms allow several potential comparisons, but the registry information summarized here does not establish which comparisons were confirmatory or how type I error was controlled.
- Unspecified censoring rules: although the endpoint is a time-to-event endpoint, the registry information summarized here does not specify detailed censoring conventions.
- Unspecified missing-data strategy: no imputation or other missing-data methodology is reported in the registry information summarized here.
- Unspecified subgroup analysis: the registry information summarized here does not provide subgroup definitions or subgroup-specific statistical results.
- Unspecified interim monitoring: no interim-analysis framework is reported in the registry information summarized here.
18. Why This Trial Matters Statistically
NSABP R-04 is a useful teaching example because the registry combines a randomized four-arm treatment design with a clinically meaningful time-to-first-recurrence endpoint. That combination raises several core statistical questions: how should event time be represented, how should censoring be handled, how should multiple randomized groups be compared, and how should relative and absolute effects ultimately be communicated?
| Concept | How it appears in NSABP R-04 |
|---|---|
| Randomization | Participants were randomized to four treatment arms. |
| Parallel design | The registry classifies the design model as parallel. |
| Time-to-event analysis | The primary endpoint is time from randomization to first local recurrence. |
| Kaplan-Meier estimation | A standard method for describing the distribution of time to first local recurrence. |
| Hazard ratio | A standard relative-effect measure for a Cox model of time to recurrence. |
| Confidence interval | Would quantify uncertainty around a reported treatment-effect estimate. |
| Multiplicity | Four arms create multiple possible treatment comparisons, making the prespecified hypothesis structure important. |
| Censoring | Participants can contribute event-free follow-up without experiencing first local recurrence during the observed period. |
| Fixed time horizon | The primary endpoint is defined up to 3 years from randomization. |
19. Related Statistical Concepts
The methods most relevant to understanding this trial are survival-analysis concepts: time-to-event endpoints, Kaplan-Meier estimation, hazard ratios, Cox regression, censoring, and interpretation of randomized multi-arm comparisons.
20. Related Statistical Calculators
21. Sources
- ClinicalTrials.gov: NCT00058474 — NSABP R-04.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through the Clinical Biostats knowledge graph
Explore statistical methods for randomized clinical trials, time-to-event endpoints, treatment comparisons, and clinical-trial analysis.
22. Record Summary
NSABP R-04 is a randomized phase 3, four-arm, parallel treatment trial enrolling 1608 participants with colorectal cancer. The registry describes radiation therapy with either capecitabine or fluorouracil, with or without oxaliplatin, before surgery for patients with resectable rectal cancer. Its registered primary endpoint is loco-regional disease control as assessed by evidence of tumor at 3 years, defined as time from randomization to first local recurrence up to 3 years.
Statistically, the primary endpoint is naturally framed as a time-to-event outcome. Kaplan-Meier estimation can describe the event-free distribution, log-rank methods can compare randomized groups, and Cox regression can provide a hazard-ratio summary when its assumptions are appropriate. With four treatment arms, the prespecified comparison structure and multiplicity strategy are important to interpreting any eventual confirmatory findings.
The ClinicalTrials.gov record summarized here does not contain posted statistical analyses or numerical primary-endpoint results. The registry therefore establishes the trial's design, treatment framework, and endpoint definition, while the numerical magnitude and statistical significance of the primary treatment comparison are not reported in the record summarized here.