REAL-2: Complete Statistical Analysis of Lymph Node Preservation in Rectal Cancer

1. Trial at a Glance

REAL-2 is a randomized, parallel-group clinical trial evaluating preservation versus dissection of the inferior mesenteric artery root lymph node in patients with rectal cancer undergoing robotic resection for mid/low rectal cancer. The registry identifies 3-year disease-free survival rate, assessed 3 years after surgery, as the primary endpoint.

2. Clinical Question

The central question is whether preserving the inferior mesenteric artery root lymph node rather than dissecting it affects the 3-year disease-free survival rate after robotic resection for mid/low rectal cancer.

This is a treatment-purpose randomized comparison. The statistical value of randomization is that, if successfully implemented and followed according to the protocol, treatment assignment provides the basis for comparing outcomes between the two strategies without relying on observational adjustment to create the primary treatment comparison.

3. Trial Design

4. Trial Timeline and Registry Status

The timeline matters statistically because the primary endpoint is defined at a fixed postoperative time point: 3 years after surgery. The primary completion date therefore provides context for when the registry expects the study to reach the stage at which the primary endpoint can be evaluated, although the completion date itself is not an outcome estimate.

5. Primary Endpoint

The endpoint is therefore a composite disease-control outcome evaluated over a fixed 3-year postoperative period. A patient contributes to the 3-year DFS rate if the patient has no death, no locoregional recurrence, and no distant metastases within that period under the registry definition.

This definition is important because "disease-free survival" is sometimes used in different ways across clinical trials. Here, the registry explicitly specifies the three components that prevent a patient from being classified as disease-free through 3 years: death, locoregional recurrence, or distant metastases.

6. Statistical Methodology

The registry identifies a randomized parallel-group treatment comparison with a single primary endpoint. No formal statistical analyses are posted in the ClinicalTrials.gov record. The appropriate statistical framework can therefore be described from the structure of the endpoint without attributing an unposted analysis plan to the registry.

Estimating a 3-year disease-free survival rate

At its simplest, a 3-year DFS rate is a proportion: the number of patients meeting the registry's definition of being disease-free at 3 years divided by the relevant analysis population.

For a randomized comparison, the principal inferential question is not only the estimated DFS rate within each arm but also the difference between the two randomized strategies. A natural absolute measure is the risk difference at 3 years:

Why the endpoint can also be viewed as a time-to-event outcome

Although the registry expresses the primary endpoint as a 3-year rate, DFS is fundamentally related to time until the first qualifying event. Patients can experience death, locoregional recurrence, or distant metastases at different postoperative times, and patients without an event by the end of observation may be censored rather than observed indefinitely.

For that reason, a time-to-event analysis such as Kaplan-Meier estimation is commonly used when the underlying DFS data are available. A 3-year Kaplan-Meier estimate describes the estimated probability of remaining event-free through 3 years while accounting for censoring before that time.

The distinction between a simple observed proportion and a Kaplan-Meier estimate is consequential. If every participant had complete follow-up through exactly 3 years, the two approaches would be closely aligned conceptually. With censoring before 3 years, a survival-analysis approach uses the timing of available information rather than treating every patient without a recorded 3-year assessment as if the patient necessarily remained disease-free.

Comparing the randomized groups

The randomized design permits an arm-to-arm comparison of the primary endpoint. Several measures can describe that comparison:

• Absolute difference: the difference in 3-year DFS rates between preservation and dissection.
• Risk ratio: the ratio of the two fixed-time DFS proportions.
• Time-to-event hazard ratio: if the underlying event times are analyzed with a Cox model, the hazard ratio summarizes the relative instantaneous rate of the first qualifying DFS event under the model.
• Confidence interval: an interval around the estimated treatment effect quantifies statistical uncertainty.

These measures are not interchangeable. A fixed-time risk difference describes an absolute difference at 3 years. A risk ratio describes a relative difference in the probability of being disease-free at that time. A hazard ratio instead concerns the relative event rate over follow-up under a time-to-event model.

Composite endpoint structure

The registry definition combines death, locoregional recurrence, and distant metastases into the DFS endpoint. This has an important statistical consequence: a difference in DFS can arise from differences in any of the qualifying components or from their combined occurrence.

Consequently, an overall DFS result does not by itself identify which component is responsible for a treatment difference. Component-specific analyses can be useful in a complete statistical report, but the ClinicalTrials.gov record summarized here identifies only the composite 3-year DFS rate as the primary endpoint.

7. Planned Analysis

What will be measured

The primary analysis concerns the 3-year disease-free survival rate, with the registry defining disease-free status as no death, no locoregional recurrence, and no distant metastases within 3 years postoperatively.

How this endpoint is typically analyzed

A complete analysis would ordinarily estimate the 3-year DFS rate separately for each randomized treatment strategy and quantify uncertainty around each estimate. The treatment comparison would then be summarized using an appropriate effect measure, such as an absolute difference or relative measure at 3 years. If individual event and censoring times are available, Kaplan-Meier methodology provides a standard way to estimate the probability of remaining disease-free over time.

For a time-to-event analysis, the timing of events matters. A patient developing recurrence early in follow-up contributes a different amount of information from a patient who remains disease-free for nearly 3 years and is then censored. A fixed-time analysis based only on complete 3-year observations would not use that information in the same way.

Confidence intervals and hypothesis testing

A rigorous primary analysis would report confidence intervals alongside the point estimates. The confidence interval describes statistical uncertainty around the estimated DFS rate or treatment effect; it is not a range containing the true outcome for individual patients.

If a formal hypothesis test is prespecified, its p-value addresses the compatibility of the observed data with the null hypothesis under the specified model and testing framework. A p-value is not a measure of the magnitude of the treatment effect. The estimated difference or ratio, together with its confidence interval, provides the more direct description of effect size and precision.

8. Statistical Methods Explained

What exactly counts as a DFS event in REAL-2?

The registry defines a DFS event broadly enough to include three clinically distinct outcomes: death, locoregional recurrence, or distant metastases. Therefore, a patient does not have to experience tumor recurrence specifically to lose disease-free status; death itself is sufficient under the endpoint definition.

This is a classic composite-endpoint structure. The statistical analysis treats the first qualifying event as the endpoint event when DFS is analyzed as a time-to-event outcome.

Why does the registry call the endpoint a 3-year DFS rate?

The endpoint is anchored to a clinically interpretable postoperative time point: 3 years after surgery. A rate at a fixed time allows readers to ask a direct question: what percentage of patients remained free of the specified events through that point?

The fixed time point also makes the treatment groups directly comparable. The important distinction is between a 3-year probability or rate and a median DFS time. They answer different questions, and the registry identifies the former as the primary endpoint.

Why might Kaplan-Meier analysis be useful?

Clinical trials rarely have perfectly complete follow-up for every patient at exactly the same time. Some patients may leave follow-up early, while others may reach the endpoint before 3 years. Kaplan-Meier estimation uses the available event and censoring times to estimate the event-free probability over time.

For REAL-2, this matters because the endpoint is defined over a 3-year period. A time-to-event analysis can incorporate when disease recurrence or death occurs rather than reducing the entire follow-up history to a single yes-or-no observation without considering timing.

What would a hazard ratio mean if one were reported?

A hazard ratio would compare the estimated instantaneous rate of experiencing the DFS event between the two randomized strategies under a Cox proportional-hazards model. A hazard ratio below 1 for preservation versus dissection would indicate a lower estimated event hazard for preservation; a hazard ratio above 1 would indicate a higher estimated event hazard.

A hazard ratio is not the same as the percentage-point difference in 3-year DFS. It also does not mean that every individual patient has exactly that proportional change in risk.

Why is the confidence interval important?

Suppose a future analysis reports a 3-year DFS difference between the two strategies. The point estimate alone does not reveal how precisely that difference has been estimated. A confidence interval adds that information by showing the range of effect estimates compatible with the statistical model and observed data at the stated confidence level.

For example, a narrow interval suggests greater statistical precision than a wide interval, all else equal. Precision is affected by factors such as sample size, number of events, variability, and the amount of censoring.

Why does randomization matter?

Randomization assigns patients to treatment strategies by a study-defined allocation process rather than allowing treatment choice to depend on prognosis or clinician preference. This is central to causal interpretation because, in expectation, randomized groups are comparable with respect to both measured and unmeasured baseline characteristics.

Randomization does not guarantee that every characteristic will be numerically identical between groups. Its principal statistical role is to make treatment assignment independent of baseline prognostic factors in expectation, subject to the integrity of the randomization process.

Why is the composite endpoint important to interpret carefully?

The DFS definition combines death, locoregional recurrence, and distant metastases. A treatment effect on the composite does not establish that all three components changed by the same amount or in the same direction. The overall endpoint summarizes the first occurrence of any qualifying component.

When component outcomes have different clinical importance or frequencies, interpretation of the composite should therefore consider the components separately whenever those analyses become available.

9. Analysis Populations and Randomization

The record identifies REAL-2 as randomized but does not report detailed analysis-population definitions in the information summarized here. The distinction between randomized, treated, and evaluable populations is nevertheless important for interpreting the eventual primary analysis.

A mature statistical report should make these populations explicit. In particular, the denominator used for a 3-year DFS rate should be clearly defined, and any deviations from the randomized population should be justified and prespecified where possible.

10. Censoring and Follow-Up

DFS is naturally subject to censoring because not every participant necessarily experiences death, locoregional recurrence, or distant metastases during the period of observation. Censoring occurs when the event status becomes unknown after the last reliable assessment or when follow-up ends without an event.

For a Kaplan-Meier analysis, censoring is incorporated into the risk set. Patients contribute information while they are under observation and are removed from the risk set at their censoring time if they have not experienced an event.

In particular, informative loss to follow-up can threaten the interpretation of a survival estimate. If patients who become difficult to follow differ systematically from those who remain under observation, standard censoring assumptions may no longer adequately describe the data-generating process.

11. Assessment of Disease-Free Survival

The registry specifies multiple assessment modalities for determining whether patients remain disease-free. These include contrast-enhanced CT/MRI, PET-CT, and histopathological confirmation through colonoscopy/biopsy.

Multiple assessment modalities can strengthen clinical ascertainment because recurrence may require different forms of confirmation depending on its location and presentation. Statistically, however, the important point is that the same endpoint definition should be applied consistently across randomized groups.

Assessment procedures can also affect event timing. If one treatment strategy systematically leads to different opportunities for detecting recurrence, differences in surveillance could potentially influence observed DFS. Randomized trial protocols therefore benefit from standardized assessment schedules and definitions, although the registry information summarized here does not provide a detailed surveillance schedule beyond the primary endpoint time frame.

12. Interpreting the 3-Year DFS Endpoint

For a trial with a fixed-time DFS endpoint, both relative and absolute measures can be useful. A relative measure may facilitate comparison across studies, while an absolute 3-year difference communicates the number of additional patients per 100 who remain disease-free at that time point under the observed trial conditions.

13. Limitations

• No posted efficacy estimates: the ClinicalTrials.gov record does not provide a posted statistical analysis for the primary endpoint, so there are no registry-reported DFS estimates, confidence intervals, or p-values to interpret.
• Future outcome assessment: the record lists the study as NOT_YET_RECRUITING, with a primary completion date of June 1, 2031. The primary endpoint is evaluated 3 years after surgery.
• Endpoint complexity: DFS combines death, locoregional recurrence, and distant metastases. An overall DFS result does not identify the contribution of each component.
• Censoring: a time-to-event analysis requires appropriate treatment of patients whose follow-up ends before a qualifying event is observed.
• Assessment consistency: because DFS can be established through several imaging and histopathological approaches, consistent application of the endpoint definition across randomized groups is important.
• Analysis population: the summarized registry information does not provide detailed definitions for efficacy, safety, or endpoint-evaluable analysis populations.
• Allocation ratio: the record identifies 2 randomized arms and 1596 planned participants but does not specify an allocation ratio in the information summarized here.
• Statistical plan: the record summarized here does not contain posted statistical analyses, so the eventual inferential model, hypothesis-testing framework, and detailed multiplicity procedures are not available for interpretation.

These limitations describe the information available from the registry record and the statistical issues that will matter when the primary results become available. They do not imply a particular future outcome.

14. Why This Trial Matters Statistically

REAL-2 is a useful teaching example because it illustrates how a surgical treatment question can be translated into a randomized statistical comparison with a clinically meaningful time-specific endpoint.

Why a surgical endpoint deserves statistical discipline

In a procedural trial, the treatment contrast may appear straightforward: preserve a structure versus dissect it. The statistical question is more demanding because the clinical consequence of that procedural choice can unfold over time. Recurrence and death are not binary observations occurring at a common instant; they occur during follow-up.

The 3-year DFS endpoint therefore combines a clinically meaningful time horizon with a composite definition. The eventual analysis needs to preserve both features: the fixed 3-year target and the timing of events that occur before it.

Why randomization and endpoint definition work together

Randomization establishes the treatment comparison, while the endpoint definition establishes what outcome is being compared. A statistically valid randomized trial can still be difficult to interpret if the endpoint is ambiguous or inconsistently assessed. REAL-2's registry definition explicitly identifies death, locoregional recurrence, and distant metastases as the events that prevent a patient from being classified as disease-free through 3 years.

15. Clinical Interpretation vs Statistical Interpretation

These two perspectives should remain distinct. Statistical evidence describes the estimated treatment difference and uncertainty. Clinical interpretation asks what that difference means in the context of the surgical strategies and the disease course being studied.

16. Questions to Ask When the Results Become Available

17. Sources

• ClinicalTrials.gov: REAL-2 (NCT06956976).

18. Record Summary

REAL-2 is a randomized, parallel-group treatment trial with 1596 planned participants and 2 study arms, comparing dissection of the inferior mesenteric artery root lymph node with preservation of that lymph node during robotic resection for mid/low rectal cancer. The registry identifies the 3-year disease-free survival rate as the primary endpoint, with assessment 3 years after surgery.

The statistical structure of the trial is centered on a clinically defined composite endpoint. Disease-free status requires the absence of death, locoregional recurrence, and distant metastases within 3 years postoperatively, with assessment through contrast-enhanced CT/MRI, PET-CT, or histopathological confirmation by colonoscopy/biopsy.

For statistical interpretation, the eventual analysis should distinguish the fixed-time 3-year DFS rate from time-to-event measures. An absolute difference directly describes the contrast in 3-year DFS, while a risk ratio provides a relative fixed-time comparison and a hazard ratio, if used, describes relative event rates under a time-to-event model. Confidence intervals are essential for interpreting the precision of those estimates.