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
CIRCULATE-US is a randomized, parallel-group phase 2/3 trial in stage III colon cancer. The study evaluates treatment strategies involving circulating tumor DNA assessment and adjuvant chemotherapy, with ctDNA-positive status and disease-free survival specified as primary endpoints.
| Feature | CIRCULATE-US |
|---|---|
| Trial acronym | CIRCULATE-US |
| ClinicalTrials.gov identifier | NCT05174169 |
| Title | Colon Adjuvant Chemotherapy Based on Evaluation of Residual Disease |
| Condition | Stage III Colon Cancer |
| Phase | Phase 2/3 |
| Status | Recruiting |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 1,912 |
| Number of arms | 4 |
| Lead sponsor | NRG Oncology |
2. Clinical Question
The central statistical question is whether evaluation of residual disease using circulating tumor DNA can be incorporated into adjuvant treatment strategies for patients with stage III colon cancer, with treatment effects evaluated using ctDNA-positive status and disease-free survival.
Population
Patients with stage III colon cancer enrolled in a randomized phase 2/3 treatment trial.
Intervention framework
The study includes the Signatera test and adjuvant chemotherapy strategies involving mFOLFOX6, CAPOX, and mFOLFIRINOX.
Comparator framework
The registry identifies four randomized arms and includes chemotherapy regimens with different treatment durations and strategies.
Primary question
How do randomized treatment strategies based on residual-disease assessment compare with respect to ctDNA-positive status and disease-free survival?
3. Trial Design
Intervention framework
The registry lists the Signatera test as a device intervention and identifies chemotherapy interventions including mFOLFOX6 for 3-6 month treatment, CAPOX for 3 month treatment, mFOLFIRINOX, mFOLFOX6 for 6 month treatment, and CAPOX for 6 month treatment.
| Registry-listed intervention | Type |
|---|---|
| Signatera test | Device |
| mFOLFOX6 3-6 month | Drug |
| CAPOX 3 month | Drug |
| mFOLFIRINOX | Drug |
| mFOLFOX6 6 month | Drug |
| CAPOX 6 month | Drug |
4. Endpoints
The registry specifies two primary endpoints. Both are time-to-event outcomes, but they represent different clinical events and therefore answer different statistical questions.
| Primary endpoint | Registry definition | Time frame |
|---|---|---|
| ctDNA positive status (TTPos) | Time from randomization until ctDNA positive event. TTPos events are first ctDNA positive result after randomization for the immediate adjuvant chemo arm (Arm 1), 2nd ctDNA positive result after randomization for the delayed adjuvant chemo (Arm 2), and recurrence without a positive ctDNA result for both arms. | Time from randomization to the first TTPos event, a maximum of 3 years |
| Disease-Free Survival (DFS) | Time from randomization to first disease-free survival event: recurrence, second primary colorectal cancer or death from any cause. | Time from randomization to disease-free survival event, a maximum of 5 years |
Why the two endpoints are statistically different
TTPos is a molecular residual-disease endpoint whose event definition depends on post-randomization ctDNA results and, for the delayed-adjuvant strategy, the second positive result after randomization. DFS is a broader clinical time-to-event endpoint in which recurrence, a second primary colorectal cancer, or death from any cause constitutes an event.
Because the endpoints use different event definitions and maximum follow-up periods, they should not be treated as interchangeable measures of treatment effect. A statistical comparison of TTPos addresses molecular evidence of residual disease, whereas DFS addresses a clinically defined disease-free state.
5. Statistical Methodology
Time-to-event analysis
Both registered primary endpoints are defined as times from randomization to an event. For endpoints of this form, the standard statistical framework is survival analysis rather than a simple comparison of proportions at one fixed time point.
A participant who has not experienced the event by the end of observed follow-up can contribute information up to the censoring time. This allows survival methods to incorporate unequal follow-up durations without treating patients lost to follow-up as if they experienced the event.
Kaplan-Meier estimation
For a time-to-event endpoint such as DFS, Kaplan-Meier estimation is commonly used to describe the probability of remaining event-free over time. The same framework can be used for TTPos when the endpoint is treated as a time-to-first-event outcome.
Here di represents the number of events at event time ti, while ni represents the number at risk immediately before that time.
Hazard ratios and Cox regression
A Cox proportional-hazards model is a common approach for comparing randomized groups on a time-to-event endpoint. It estimates a hazard ratio that summarizes the relative instantaneous event rate between groups under the model.
A hazard ratio is not a risk ratio, an absolute risk difference, or the probability that a particular participant will remain event-free. Its interpretation depends on the time-to-event model and its assumptions.
Log-rank comparison
A log-rank test is another standard method for comparing time-to-event distributions between randomized groups. It evaluates whether the observed pattern of events over follow-up differs between groups, rather than comparing only the percentage of participants with an event at one selected time point.
Analysis population
Because this is a randomized trial, an intention-to-treat analysis is generally the principal framework for estimating the effect of randomized assignment on efficacy outcomes. The registry information available here identifies randomization but does not provide detailed analysis-population definitions or a statistical analysis plan.
6. Planned Analysis
The registry identifies TTPos and DFS as the primary endpoints but does not post formal statistical analyses for either endpoint.
TTPos
The registry will measure time from randomization to the first TTPos event over a maximum of 3 years. Because TTPos is a time-to-first-event endpoint, survival-analysis methods such as Kaplan-Meier estimation, a log-rank comparison, and Cox regression are typical approaches for comparing randomized groups.
DFS
The registry will measure time from randomization to recurrence, second primary colorectal cancer, or death from any cause over a maximum of 5 years. Kaplan-Meier and Cox-model methods are typical approaches for this type of endpoint.
7. Statistical Methods Explained
Why is time from randomization used?
Randomization establishes the starting point for the primary time-to-event comparison. Measuring TTPos and DFS from randomization places participants on a common time scale and preserves the treatment-assignment framework of the randomized trial.
Why use a time-to-event method rather than simply counting events?
Two treatment groups can have the same number of events but very different timing of those events. Survival analysis retains information about when events occur and can appropriately account for participants whose event status is not yet observed at the end of follow-up.
What does a hazard ratio represent?
A hazard ratio compares estimated instantaneous event rates between groups under a time-to-event model. An HR below 1 would indicate a lower modeled event rate in the numerator treatment group, while an HR above 1 would indicate a higher modeled event rate. It does not directly state the proportion of participants who will eventually experience an event.
Why is censoring important?
A participant may reach the end of available follow-up without experiencing the event, or may otherwise have follow-up end before an event is observed. Such observations are censored rather than automatically treated as event-free for all future time. Kaplan-Meier and Cox methods are designed to incorporate this structure.
Why are TTPos and DFS not the same endpoint?
TTPos uses ctDNA positivity and has an event definition that differs between the immediate and delayed adjuvant chemotherapy strategies. DFS uses recurrence, second primary colorectal cancer, or death from any cause. The endpoints therefore capture different biological and clinical dimensions of disease status.
What does a maximum follow-up of 3 or 5 years mean?
The maximum time frame defines the registry's stated observation window for the endpoint. TTPos is specified for a maximum of 3 years, whereas DFS is specified for a maximum of 5 years. It does not mean that every participant will necessarily have exactly that amount of observed follow-up.
8. Interpreting TTPos Statistically
The TTPos endpoint is unusual because its event definition is tied directly to serial molecular testing. The registry defines the event differently for the immediate adjuvant chemotherapy arm and the delayed adjuvant chemotherapy arm: the first post-randomization ctDNA-positive result is the event for Arm 1, while the second post-randomization ctDNA-positive result is the event for Arm 2. Recurrence without a positive ctDNA result is also included in the event definition for both arms.
A time-to-event estimate is only as interpretable as its event definition. Here, TTPos is not simply "time to a positive blood test." The registry's definition incorporates the treatment strategy and specifies different ctDNA-positive event rules for the immediate and delayed chemotherapy arms.
A treatment-group comparison of TTPos would address the timing of the first prespecified molecular residual-disease event under the registry definition. It would not automatically be equivalent to a comparison of overall recurrence risk, because DFS has a broader clinical event definition.
Time-to-event analysis of TTPos depends on when molecular assessments occur and when participants cease to contribute observable follow-up. The timing and availability of ctDNA measurements therefore matter to interpretation in a way that is especially important for a biomarker-defined endpoint.
9. Interpreting Disease-Free Survival
DFS is defined in the registry as time from randomization to the first disease-free survival event: recurrence, second primary colorectal cancer, or death from any cause. This creates a composite time-to-event endpoint with several clinically distinct ways to leave the disease-free state.
| DFS component | Statistical role |
|---|---|
| Recurrence | Counts as the first DFS event when it occurs before another qualifying event. |
| Second primary colorectal cancer | Counts as a DFS event under the registry definition. |
| Death from any cause | Counts as a DFS event regardless of the immediate cause of death. |
This composite structure is important when interpreting any future DFS estimate. A DFS event does not necessarily mean that the patient experienced documented recurrence; death from any cause and a second primary colorectal cancer also qualify as events.
10. Four-Arm Randomization and Statistical Comparisons
The trial is registered as randomized with four arms in a parallel design. A four-arm trial can support several different statistical comparisons, but the scientific meaning of those comparisons depends on the prespecified contrast structure.
Pairwise comparisons
Two individual randomized strategies can be compared directly if that contrast is part of the prespecified analysis plan.
Multi-arm inference
When more than two randomized groups are compared, the analysis must distinguish the overall question from individual pairwise contrasts.
Multiplicity
Multiple formal comparisons can increase the familywise probability of a false-positive finding unless the statistical design specifies appropriate error control.
Interpretation
The registry information available here identifies four arms but does not provide the prespecified multiplicity strategy or a complete hierarchy of treatment contrasts.
11. Design Features That Affect Statistical Interpretation
Randomization
Randomized allocation is central to causal interpretation. If randomization is properly implemented and analyzed according to assignment, differences in outcomes can be attributed more credibly to the assigned treatment strategy than in an observational comparison.
Parallel-group design
The registry identifies CIRCULATE-US as a parallel trial. Participants remain in their randomized treatment pathway rather than serving as their own control in a crossover design.
Open-label design
Masking is listed as none. This means treatment assignment is not masked. Lack of masking can be relevant to clinical management and assessment, although its impact depends on the endpoint and how the endpoint is measured.
No crossover information is reported
The registry information available here does not specify a crossover design. Therefore, crossover should not be assumed when interpreting the future DFS or TTPos results.
No non-inferiority margin is reported
The registry information available here does not specify a non-inferiority design or margin. A treatment comparison should therefore not be described as a non-inferiority analysis unless the relevant statistical documentation establishes that framework.
No Bayesian methodology is reported
The registry information available here does not identify a Bayesian analysis. The primary endpoints can be understood using conventional frequentist survival-analysis concepts unless a future statistical analysis plan specifies otherwise.
12. Interim Analysis, Missing Data, and Multiplicity
Interim analysis
The registry information available here does not report an interim-analysis schedule, alpha-spending approach, stopping boundary, or other group-sequential monitoring method. Those features should be distinguished from the general concept of monitoring a trial over time.
Missing data and censoring
Time-to-event endpoints naturally involve censoring, but the registry information available here does not specify the detailed censoring rules or missing-data and imputation procedures for either TTPos or DFS. These details can materially affect analysis, particularly for a molecular endpoint whose event definition depends on repeated measurements.
Multiplicity
There are two registered primary endpoints and four randomized arms. The registry information available here does not specify how type I error is allocated across primary endpoints, treatment contrasts, or any interim analyses.
A collection of unadjusted P-values does not automatically provide confirmatory evidence when multiple hypotheses are tested. The interpretation depends on the prespecified testing hierarchy, alpha allocation, and analysis population.
13. Primary Endpoint Structure
| Endpoint | Clinical/statistical role | Maximum time frame |
|---|---|---|
| ctDNA positive status (TTPos) | Molecular residual-disease time-to-event endpoint incorporating prespecified ctDNA-positive and recurrence events. | 3 years |
| Disease-Free Survival (DFS) | Clinical time-to-event endpoint incorporating recurrence, second primary colorectal cancer, and death from any cause. | 5 years |
The distinction between these endpoints is statistically important. TTPos is closely tied to molecular detection of residual disease, whereas DFS reflects a broader clinical state. If both endpoints later produce treatment-effect estimates, they should be interpreted as complementary rather than interchangeable evidence.
14. What a Future DFS Hazard Ratio Would — and Would Not — Mean
If a future Cox analysis reports a DFS hazard ratio below 1 for one randomized group relative to another, the estimate would indicate a lower modeled instantaneous rate of the prespecified DFS event in the numerator group over the analyzed follow-up.
A hazard ratio would not mean that the same percentage of patients avoided recurrence, second primary colorectal cancer, or death. Absolute DFS probabilities at clinically relevant time points would provide a different and complementary perspective.
A confidence interval around a future hazard ratio would quantify statistical uncertainty around the estimated relative effect under the specified model and sampling framework. A narrower interval generally indicates greater statistical precision than a wider interval, but precision and clinical importance are distinct concepts.
A P-value addresses the compatibility of the observed data with a specified null hypothesis under the statistical model. It does not measure the magnitude of the treatment effect, the probability that the treatment works, or the clinical importance of an observed difference.
15. Why the Proportional-Hazards Assumption Matters
The Cox model is commonly interpreted through a hazard ratio that summarizes relative event rates over follow-up. That interpretation is most straightforward when the proportional-hazards assumption is reasonable.
If the relative hazard changes substantially over time, a single hazard ratio can compress a more complicated pattern into one number. For that reason, future CIRCULATE-US results would be most informative when the hazard ratio is considered alongside Kaplan-Meier estimates, confidence intervals, and clinically meaningful time-specific DFS measures.
16. TTPos as a Biomarker-Defined Endpoint
TTPos illustrates an important statistical issue in modern oncology trials: a biomarker endpoint can be both longitudinal and treatment-dependent. The registry defines the event using repeated ctDNA assessment and specifies different positive-event rules for the immediate and delayed adjuvant chemotherapy strategies.
| Feature | Statistical implication |
|---|---|
| Serial ctDNA measurement | Event ascertainment depends on measurements collected over time. |
| Immediate adjuvant chemotherapy | First post-randomization ctDNA-positive result defines the TTPos event. |
| Delayed adjuvant chemotherapy | Second post-randomization ctDNA-positive result defines the TTPos event. |
| Recurrence without positive ctDNA | Counts as a TTPos event for both specified arms. |
| Maximum follow-up | TTPos is evaluated for a maximum of 3 years. |
This construction means that the endpoint is more informative than a single cross-sectional "ctDNA positive versus negative" classification. It is explicitly a time-to-event endpoint, and its event definition is embedded in the treatment strategy being evaluated.
17. Disease-Free Survival as a Composite Endpoint
Composite endpoints are common in clinical trials because combining clinically meaningful events can increase the number of observed events and provide a broader measure of treatment effect. The tradeoff is that the components may differ in clinical frequency and importance.
Recurrence
Represents return of the cancer and is one of the registry-defined DFS events.
Second primary colorectal cancer
Represents a new colorectal cancer and is separately included in the DFS event definition.
Death from any cause
Counts as a DFS event regardless of the immediate cause of death.
First-event principle
DFS ends at the first qualifying event after randomization.
The composite definition should remain visible when reporting a future DFS result. A DFS hazard ratio summarizes the combined endpoint as defined by the registry; it does not separately estimate the treatment effect for each component.
18. Trial Timeline
Trial start
CIRCULATE-US began on July 8, 2022, according to the ClinicalTrials.gov record.
Current registry status
The ClinicalTrials.gov record identifies the study as Recruiting.
Primary completion date
The registry lists March 10, 2029 as the primary completion date.
19. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record does not currently provide formal statistical results for the registered primary endpoints.
- No effect estimates: no hazard ratios, confidence intervals, P-values, event counts, or survival probabilities are reported for TTPos or DFS in the registry information available here.
- Incomplete arm-level mapping: the registry identifies four arms and lists several interventions, but the available information does not provide a complete arm-by-arm mapping of every intervention.
- Multiplicity details: the available registry information does not specify how the two primary endpoints and the four-arm design are incorporated into a formal type I error strategy.
- Interim-analysis details: no interim-analysis schedule, stopping boundary, or alpha-spending method is reported in the available registry information.
- Missing-data rules: detailed missing-data and censoring procedures for TTPos and DFS are not reported in the available registry information.
- Analysis population: detailed definitions of efficacy and safety analysis populations are not reported in the available registry information.
- Generalizability: the trial's applicability to a particular patient depends on its eligibility criteria and the characteristics of the population ultimately enrolled and analyzed.
20. Why This Trial Matters Statistically
CIRCULATE-US is statistically important as an example of a modern randomized trial in which a molecular residual-disease measure is embedded directly into treatment strategy. The design combines a biomarker-defined time-to-event endpoint with a conventional clinical time-to-event endpoint, creating an opportunity to examine how early molecular evidence and later clinical outcomes relate within a randomized framework.
| Statistical concept | How it appears in CIRCULATE-US |
|---|---|
| Randomization | Four-arm randomized allocation. |
| Parallel-group design | Participants are assigned to parallel treatment strategies. |
| Time-to-event analysis | Both registered primary endpoints are defined as times from randomization to events. |
| Biomarker endpoint | TTPos uses serial ctDNA assessment to define a molecular event. |
| Composite endpoint | DFS includes recurrence, second primary colorectal cancer, and death from any cause. |
| Different follow-up windows | TTPos has a maximum of 3 years; DFS has a maximum of 5 years. |
| Multiple treatment arms | Four randomized arms create the need for a clearly specified contrast structure. |
| Multiplicity | Two primary endpoints and multiple randomized arms make prespecified error control important. |
| Censoring | Time-to-event analyses must account for participants whose event has not been observed by the end of follow-up. |
| Model assumptions | Future Cox-model analyses would require attention to proportional-hazards assumptions. |
21. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial is designed around randomized comparisons of time-to-event endpoints. TTPos measures time to a prespecified molecular residual-disease event, while DFS measures time to recurrence, second primary colorectal cancer, or death from any cause.
Clinical interpretation
The eventual clinical meaning of CIRCULATE-US will depend on the observed relationship between randomized treatment strategies, molecular residual disease, and disease-free survival. Those relationships require the eventual posted outcome data.
22. What to Look for When Results Are Posted
Future results can be interpreted most clearly by examining several pieces of information together rather than focusing on a single P-value.
| Result component | Why it matters |
|---|---|
| Hazard ratio | Summarizes the relative time-to-event effect under the specified model. |
| 95% confidence interval | Shows the statistical uncertainty surrounding the estimated treatment effect. |
| P-value | Addresses compatibility with a prespecified null hypothesis; it does not measure effect size. |
| Kaplan-Meier estimates | Show how event-free probability changes over time. |
| Event counts | Provide the underlying amount of observed information contributing to the analysis. |
| Analysis population | Clarifies which randomized participants were included in the efficacy comparison. |
| Censoring rules | Clarify how incomplete follow-up contributes to the time-to-event analysis. |
| Multiplicity adjustment | Determines how formal inference should be interpreted across multiple endpoints and comparisons. |
| Proportional-hazards assessment | Helps determine whether a single Cox hazard ratio adequately summarizes the treatment effect over time. |
23. Statistical Interpretation Framework
Determine exactly which randomized treatment contrast and which endpoint are being estimated. In CIRCULATE-US, this distinction is particularly important because there are four randomized arms and two primary endpoints.
For TTPos, verify the applicable ctDNA-positive rule and the recurrence rule. For DFS, verify that the analysis uses recurrence, second primary colorectal cancer, and death from any cause as specified.
Review the hazard ratio together with the Kaplan-Meier curves and time-specific event-free probabilities. The hazard ratio alone does not describe the entire temporal pattern.
Use the confidence interval to assess the statistical precision of the estimated effect. A confidence interval should be interpreted in relation to the prespecified clinically meaningful effect, not simply as a binary significant/non-significant label.
Determine how the statistical analysis handles the two primary endpoints, the four randomized arms, any interim looks, and any secondary or exploratory comparisons.
For Cox-model results, assess whether proportional hazards is a reasonable summary of the observed treatment effect. For TTPos, also examine the completeness and timing of molecular assessments.
24. Sources
- ClinicalTrials.gov: CIRCULATE-US, NCT05174169.
Continue through the Clinical Biostats trial library
Clinical trial analyses connect study design, endpoints, survival methods, and statistical interpretation across randomized studies.
25. Record Summary
CIRCULATE-US is a randomized phase 2/3 parallel-group trial enrolling 1,912 participants with stage III colon cancer. The study has four arms, no masking, and a treatment-focused primary purpose. Its two registered primary endpoints are ctDNA positive status (TTPos), measured from randomization to the first prespecified TTPos event over a maximum of 3 years, and disease-free survival, measured from randomization to recurrence, second primary colorectal cancer, or death from any cause over a maximum of 5 years.
Statistically, the trial provides a useful framework for understanding how randomized treatment comparisons can incorporate molecular residual-disease endpoints alongside conventional clinical outcomes. The key interpretive issues include time-to-event analysis, censoring, the distinct event definitions for TTPos and DFS, the four-arm comparison structure, multiplicity across endpoints and treatment contrasts, and the assumptions underlying future survival models.