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
MRC CR07 was a randomized phase 3 treatment trial in colorectal cancer evaluating two treatment strategies for patients with rectal cancer that could be surgically removed: radiation therapy before surgery versus chemotherapy plus radiation after surgery.
| Feature | MRC CR07 |
|---|---|
| Phase | Phase 3 |
| Condition | Colorectal Cancer |
| Population | Patients with rectal cancer that can be surgically removed |
| Design | Randomized |
| Primary purpose | Treatment |
| Enrollment | 1800 |
| Trial status | Completed |
| Lead sponsor | Medical Research Council |
| Sponsor type | OTHER_GOV |
| ClinicalTrials.gov | NCT00003422 |
2. Clinical Question
The central statistical question is whether the timing and composition of perioperative treatment affect the occurrence of local recurrence in patients with surgically removable rectal cancer.
Population
Patients with rectal cancer that can be surgically removed, within the broader registry condition of colorectal cancer.
Intervention
Radiation therapy before surgery, described in the registry alongside neoadjuvant therapy.
Comparator
Chemotherapy plus radiation after surgery, including fluorouracil, leucovorin calcium, and adjuvant therapy.
Primary question
Does the treatment strategy involving radiation before surgery differ from chemotherapy plus radiation after surgery with respect to local recurrence?
3. Trial Design
The registry identifies MRC CR07 as a randomized phase 3 treatment trial with an enrollment of 1800. Randomization is important because it establishes treatment assignment before outcomes are observed, allowing the primary comparison to be interpreted as a randomized treatment contrast rather than a comparison of preselected treatment groups.
Radiation before surgery
- Radiation therapy
- Timing before surgery
- Registry intervention terminology includes neoadjuvant therapy
- Conventional surgery is included among the registered interventions
Chemotherapy plus radiation after surgery
- Fluorouracil
- Leucovorin calcium
- Adjuvant therapy
- Radiation therapy after surgery
- Conventional surgery
4. Endpoints
The registry identifies one registered primary endpoint:
| Endpoint | Registry definition |
|---|---|
| Local recurrence | Local recurrence by biopsy, imaging, or imaging and carcinoembryonic antigen result |
No additional primary endpoint, secondary endpoint, endpoint time frame, or posted statistical analysis is identified in the registry information summarized here.
5. Statistical Methodology
What type of statistical endpoint is local recurrence?
Local recurrence is naturally analyzed as a binary or time-to-event outcome, depending on how follow-up and the timing of recurrence are incorporated into the prespecified analysis. If the analysis uses only whether recurrence occurred by a fixed assessment point, the comparison can be framed as a difference in proportions. If the time until recurrence and differing follow-up are incorporated, survival-analysis methods are more appropriate.
Why randomization matters statistically
Randomization separates treatment assignment from the factors that might otherwise determine which treatment a patient receives. In an ideal randomized comparison, this reduces systematic differences between treatment groups and supports causal interpretation of an observed difference in the primary endpoint.
For a recurrence endpoint, the treatment contrast could be expressed as a risk difference, risk ratio, odds ratio, or time-to-event effect depending on the prespecified statistical analysis. The registry information does not identify which effect measure was used.
Time-to-event analysis
When recurrence times are available and patients can be followed for different lengths of time, Kaplan-Meier estimation can describe the probability of remaining free of the event over time. Patients who have not experienced recurrence at their last observation can be right-censored rather than treated as though they had experienced no event indefinitely.
Here, di represents events at an observed event time and ni represents patients at risk immediately before that time. The registry does not report a Kaplan-Meier estimate for MRC CR07.
Log-rank comparison
If local recurrence were analyzed as a time-to-event endpoint, a log-rank test would provide a standard way to compare the event-time distributions between randomized treatment groups. The test evaluates whether the observed pattern of events over follow-up differs between groups; it does not itself quantify the magnitude of the treatment effect.
Cox proportional-hazards model
A Cox proportional-hazards model is another standard approach for a time-to-local-recurrence endpoint. It produces a hazard ratio describing the relative instantaneous event rate between treatment groups under the model. A Cox analysis also requires consideration of the proportional-hazards assumption.
A hazard ratio is a relative time-to-event measure. It is not the same as an absolute recurrence reduction, a probability of recurrence, or the proportion of patients who benefit.
6. Planned Analysis
The registry identifies local recurrence by biopsy, imaging, or imaging and carcinoembryonic antigen result as the primary endpoint. Results have not been posted on ClinicalTrials.gov.
How a primary recurrence endpoint would typically be analysed
For a recurrence endpoint in a randomized trial, the statistical analysis would ordinarily begin by defining the event precisely, establishing the time origin, specifying how censoring is handled, and comparing the randomized groups using the method prespecified in the statistical analysis plan.
Fixed-time approach
If the endpoint is defined as recurrence by a particular assessment time, each patient can be classified as having or not having the event by that time, allowing comparison of event proportions.
Time-to-event approach
If time to recurrence is incorporated, Kaplan-Meier estimates can describe recurrence-free probabilities and a log-rank or Cox-model framework can compare the randomized groups.
Censoring
Patients without documented recurrence at their final observation can contribute follow-up until that point if the analysis treats them as right-censored.
Effect measure
The appropriate summary might be a risk difference, risk ratio, odds ratio, hazard ratio, or another prespecified measure. The registry information does not identify the planned effect measure.
Why the endpoint definition matters
The inclusion of biopsy, imaging, or imaging together with carcinoembryonic antigen results means that recurrence classification depends on prespecified clinical and diagnostic evidence. The statistical analysis should therefore use the same endpoint definition consistently across randomized groups.
7. Statistical Methods Explained
Why is randomization important?
Randomization determines treatment assignment independently of the patient's subsequent outcome. This creates the basis for comparing outcomes between groups while reducing confounding from measured and unmeasured characteristics, assuming the randomized design is properly implemented and analyzed.
Why might Kaplan-Meier analysis be appropriate?
Local recurrence can occur at different times, and patients may have different amounts of observable follow-up. Kaplan-Meier estimation uses the observed timing of events and accommodates right-censored observations, making it useful when the time to recurrence is part of the estimand.
What does a hazard ratio describe?
A hazard ratio compares instantaneous event rates between groups under a time-to-event model. A value below 1 would indicate a lower modeled instantaneous recurrence rate in the numerator group, while a value above 1 would indicate a higher modeled instantaneous recurrence rate. It does not directly state the absolute difference in recurrence probabilities.
Why is censoring important?
A patient who has not experienced documented recurrence by the last observation is not necessarily known to remain recurrence-free indefinitely. Right-censoring allows that patient's observed follow-up to contribute information without assigning an unobserved future outcome.
Why would the proportional-hazards assumption matter?
The Cox model summarizes relative event rates through a hazard ratio. The conventional interpretation is most straightforward when the relative hazards are reasonably stable over time. If treatment effects change substantially during follow-up, a single hazard ratio can conceal important features of the event-time distributions.
Why distinguish the p-value from the effect estimate?
A p-value addresses evidence against a specified null hypothesis under the statistical model. It does not measure the size or clinical importance of an effect. An effect estimate and its confidence interval provide information about magnitude and precision, while the p-value addresses a different inferential question.
8. Analysis Populations and Estimands
The registry identifies MRC CR07 as a randomized treatment study, but the summarized registry information does not specify an intention-to-treat population, per-protocol population, safety population, or detailed estimand framework.
| Analysis concept | Statistical role | Registry information |
|---|---|---|
| Randomized population | Preserves the treatment assignment created by randomization | Randomized allocation is reported |
| Intention-to-treat | Analyzes patients according to randomized assignment | Not specified in the registry information |
| Per-protocol | Focuses on patients meeting protocol-defined treatment criteria | Not specified in the registry information |
| Safety population | Usually based on treatment exposure | Not specified in the registry information |
| Estimand | Defines the precise treatment effect being targeted | Not specified in the registry information |
These distinctions matter because a treatment effect depends not only on the endpoint but also on the population, treatment strategy, handling of intercurrent events, and summary measure used to define the estimand.
9. Intercurrent Events and Follow-Up
For a local-recurrence endpoint, events occurring after treatment assignment can affect the interpretation of the endpoint and its analysis. Examples include treatment discontinuation, additional therapy, surgery-related events, or loss to follow-up. The registry information does not specify how such intercurrent events were handled statistically.
Event definition
The primary endpoint is based on documented local recurrence using biopsy, imaging, or imaging and carcinoembryonic antigen results.
Censoring rule
The registry information does not specify the censoring convention for patients without documented local recurrence.
Competing events
The registry information does not specify whether death or another event was treated as a competing risk for local recurrence.
Intercurrent treatment
The registry information does not specify an estimand or analysis rule for treatment changes after randomization.
10. Statistical Interpretation of a Recurrence Endpoint
If recurrence proportions were reported, an absolute risk difference would quantify the difference in recurrence probability between randomized groups. This is directly interpretable as a difference in event probability over the specified period.
A risk ratio would compare recurrence probabilities between groups. It answers a relative question and can look substantially different from the corresponding absolute risk difference when baseline recurrence risk is low or high.
A hazard ratio would compare instantaneous recurrence rates over follow-up under a Cox-model framework. It would not mean that the treatment reduces every patient's probability of recurrence by the same percentage.
A confidence interval around an effect estimate describes statistical uncertainty under the specified analysis framework. A narrower interval generally indicates greater statistical precision than a wider interval, but the interval does not describe the range of effects experienced by individual patients.
11. Limitations
- No posted numerical outcome analysis: the registry information does not provide an estimate, confidence interval, or p-value for the primary local-recurrence endpoint.
- Limited endpoint detail: the registry identifies the primary endpoint definition but does not provide a time frame for the endpoint in the information summarized here.
- Analysis method not specified: the registry information does not identify whether the primary endpoint was analyzed as a fixed-time binary outcome, a time-to-event outcome, or using another statistical framework.
- Effect measure not specified: the registry information does not identify whether the primary treatment effect was expressed as a risk difference, risk ratio, odds ratio, hazard ratio, or another measure.
- Analysis populations not specified: the summarized registry information does not define ITT, per-protocol, or safety populations.
- Censoring and competing risks: the registry information does not specify how censoring, death, or other competing events were handled for local recurrence.
- Stratification not specified: although allocation is randomized, the registry information does not identify any randomization strata or covariates for the primary analysis.
- Multiplicity not specified: the registry information does not identify a multiplicity strategy for the primary endpoint or other possible analyses.
12. Why This Trial Matters Statistically
MRC CR07 provides a useful teaching example of how the statistical analysis of a randomized clinical trial begins with a clearly defined clinical event and then connects that event to an appropriate estimand, analysis population, and effect measure.
| Concept | How it appears in MRC CR07 |
|---|---|
| Randomization | Allocation is identified as randomized. |
| Phase 3 design | The registry identifies the study as phase 3. |
| Primary endpoint | Local recurrence by biopsy, imaging, or imaging and carcinoembryonic antigen result. |
| Event definition | Recurrence is defined using specified diagnostic evidence. |
| Time-to-event methods | Kaplan-Meier, log-rank, and Cox methods are standard approaches when recurrence timing is analyzed. |
| Effect measures | Risk differences, risk ratios, odds ratios, or hazard ratios can answer different statistical questions. |
| Censoring | Requires explicit rules when follow-up differs between patients. |
| Estimands | The population, treatment strategy, endpoint, intercurrent-event handling, and summary measure define the treatment effect being estimated. |
| Multiplicity | Multiple formal hypotheses require a prespecified strategy if they are intended to support confirmatory inference. |
| Statistical precision | Confidence intervals quantify uncertainty around estimated treatment effects. |
13. Statistical Design Questions Worth Asking
What exactly is the event?
The registry defines local recurrence through biopsy, imaging, or imaging and carcinoembryonic antigen results. Consistent event adjudication is central to valid comparison.
When does follow-up begin?
A time-to-event analysis requires a clearly defined time origin. The registry information does not specify that analysis convention.
What happens to censored patients?
The statistical analysis must define when a patient is considered no longer observable for the endpoint and how that observation contributes to the analysis.
What is the estimand?
The treatment effect is not fully defined by the endpoint name alone. The analysis population, treatment strategy, intercurrent events, and summary measure also matter.
14. Sources
- ClinicalTrials.gov: MRC CR07, NCT00003422.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through Clinical Biostats
Explore statistical methods, clinical trial analyses, and educational resources for understanding how randomized evidence is designed and analyzed.
15. Record Summary
MRC CR07 is a randomized phase 3 treatment trial with an enrollment of 1800, conducted under the sponsorship of the Medical Research Council. Its registered primary endpoint is local recurrence by biopsy, imaging, or imaging and carcinoembryonic antigen result. The central statistical task is therefore to compare local recurrence between randomized treatment strategies while defining the appropriate population, follow-up framework, censoring rules, and effect measure.
The registry information establishes the treatment comparison and endpoint but does not provide a posted numerical analysis of the primary endpoint. For a recurrence endpoint, the distinction between a fixed-time risk comparison and a time-to-event analysis is particularly important. Kaplan-Meier estimation, log-rank testing, and Cox proportional-hazards modeling are standard tools when recurrence timing is incorporated, while risk differences and risk ratios provide complementary absolute and relative perspectives when a fixed-time binary endpoint is appropriate.