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
ETHOS is a randomized phase 3 parallel-group trial evaluating endoscopic therapy versus surgery for colon cancer. The registry identifies two co-primary endpoints: a 30-day composite of severe adverse events, re-admissions and deaths following assigned treatment, and a three-year endpoint addressing cancer recurrence or colorectal cancer death.
| Feature | ETHOS |
|---|---|
| Trial acronym | ETHOS |
| ClinicalTrials.gov identifier | NCT06940947 |
| Official title | Endoscopic Therapy Or Surgery for Early Colon Cancer |
| Phase | Phase 3 |
| Status | Recruiting |
| Condition | Colon Cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Single |
| Primary purpose | Treatment |
| Enrollment | 434 |
| Number of arms | 2 |
| Interventions | eFTR (procedure); Surgery (procedure) |
| Lead sponsor | Norwegian Department of Health and Social Affairs |
| Sponsor type | OTHER_GOV |
| Trial start | 2025-08-25 |
| Primary completion date | 2033-09-01 |
2. Clinical Question
The central clinical question is whether an endoscopic treatment strategy using eFTR can be evaluated against surgery in patients with colon cancer using both an early safety-related composite endpoint and a longer-term cancer outcome.
Population
Patients with colon cancer, within the population defined by the ETHOS registry record.
Intervention
eFTR, identified in the registry as an endoscopic procedure.
Comparator
Surgery, identified in the registry as a procedure.
Primary question
How do the randomized treatment strategies compare with respect to the 30-day composite of severe adverse events, re-admissions and deaths, and the three-year cancer outcome?
3. Trial Design
eFTR
- eFTR procedure
- Randomized treatment strategy
- Evaluated on the registry's 30-day and three-year primary endpoints
Surgery
- Surgery procedure
- Randomized comparator strategy
- Evaluated on the registry's 30-day and three-year primary endpoints
4. Primary Endpoints
The registry specifies two primary endpoints, one focused on short-term events after treatment and one focused on cancer outcomes after three years. Their different time horizons mean that they address distinct aspects of treatment performance.
| Primary endpoint | Registry definition | Time frame |
|---|---|---|
| Rate of severe adverse events, re-admissions and deaths within 30 days | Co-primary short-term endpoint is the composite rate of rate of severe adverse events, re-admissions and deaths within 30 days following assigned treatment. | 30 days |
| Rate of cancer recurrence or CRC death after three years | Primary long-term endpoint of CRC death or cancer recurrence or sign of lymph nodes or distant metastases after three years. | 3 years |
The short-term composite endpoint
The first primary endpoint combines several clinically distinct outcomes into a single 30-day measure: severe adverse events, re-admissions and deaths. A composite endpoint can increase the number of observed events and therefore provide a more efficient statistical comparison than relying on a single component alone.
However, the components of a composite are not automatically equivalent. A death, a re-admission and a severe adverse event can differ substantially in clinical importance and frequency. Consequently, interpretation of a composite should consider both the overall rate and the contribution of its individual components when those component results become available.
The three-year cancer endpoint
The second primary endpoint evaluates cancer recurrence or CRC death after three years. The registry description also identifies signs involving lymph nodes or distant metastases. This endpoint therefore captures longer-term oncologic outcome rather than the immediate procedural safety addressed by the 30-day endpoint.
5. Planned Analysis
The registry identifies the primary outcomes that will be measured but does not post formal statistical analyses or numerical trial results. The appropriate statistical framework depends on the precise analysis population, endpoint coding, missing-data rules, and prespecified analysis plan used when the trial is analyzed.
Analysis of the 30-day composite rate
The first endpoint is expressed as a rate over a fixed 30-day period. For a participant-level binary analysis, each participant can be classified according to whether at least one qualifying severe adverse event, re-admission or death occurs within 30 days following assigned treatment. The treatment groups can then be compared using an appropriate method for two proportions.
The corresponding control-group rate is defined in the same way. Common comparative measures include a risk difference, risk ratio, or odds ratio, with an appropriate confidence interval. The exact primary model is not specified in the registry information presented here.
A simple risk difference answers an absolute question: how many more or fewer participants per unit of population experienced the composite. A risk ratio instead expresses the event rate in one randomized group relative to the other. These measures can convey substantially different aspects of the same data.
Analysis of the three-year cancer endpoint
The long-term endpoint has a time horizon of three years and includes cancer recurrence or CRC death. If participants can be followed for different lengths of time or can be censored before experiencing the endpoint, a time-to-event analysis would typically be appropriate. Kaplan-Meier methods can estimate the probability of remaining event-free over time, while a Cox proportional-hazards model can provide a relative hazard estimate.
Here, S(t) represents the estimated probability of remaining free of the defined event beyond time t. A three-year event-free estimate provides an absolute time-specific measure, while a hazard ratio provides a relative comparison over the analyzed follow-up.
The registry does not identify a specific survival-analysis model, stratification factor, hypothesis-testing procedure, or confidence-interval construction method in the information presented for ETHOS. Those details should therefore be treated as analysis-plan questions rather than assumed features of the trial.
Composite endpoint analysis requires component awareness
If the short-term endpoint is analyzed as a single binary composite, a participant experiencing any qualifying component would generally count as having experienced the composite. This approach avoids counting the same participant multiple times in the primary comparison, but it also means that the overall composite can be driven disproportionately by whichever component occurs most frequently.
For example, if re-admissions were much more common than deaths, the overall composite rate could primarily reflect re-admissions. That would not make the composite invalid, but it would affect how the numerical result should be interpreted clinically. Component-specific analyses are therefore valuable for understanding what actually drives the composite.
6. Statistical Methodology
Randomized comparison
The principal statistical structure is the comparison of two treatment strategies after randomized allocation. Randomization is important because it establishes the treatment assignment before the outcomes are observed. In a sufficiently well-conducted randomized trial, this provides the foundation for estimating the causal effect of assignment to one strategy versus the other.
The primary analysis should ordinarily preserve that randomized comparison. Treatment switching, discontinuation, protocol deviations and other post-randomization events can complicate secondary analyses, but they do not change the original treatment assignment.
Fixed-time event rates
The 30-day endpoint is naturally understood as a fixed-time event outcome. Suppose the observed event rate in the eFTR group is \(p_1\) and the corresponding rate in the surgery group is \(p_0\). Three commonly used effect measures are:
These measures answer different questions. None should be substituted for another without explicitly stating the effect measure being reported.
Confidence intervals
A confidence interval quantifies statistical uncertainty around an estimated treatment effect under the chosen statistical model and sampling framework. Wider intervals generally indicate less precise estimation, whereas narrower intervals indicate greater precision.
For a fixed-time binary endpoint, the appropriate confidence interval depends on the effect measure and analysis method. A risk difference, risk ratio and odds ratio each require their own corresponding interval calculation.
Kaplan-Meier estimation
For the three-year cancer endpoint, a time-to-event analysis can accommodate participants who have not experienced recurrence or CRC death by their last known follow-up. Such participants can contribute information up to the time they become censored.
Here, di represents events at an event time and ni represents the number at risk immediately before that time.
Cox proportional-hazards model
A Cox model is a common approach for estimating a relative hazard for a time-to-event endpoint. The resulting hazard ratio compares the estimated instantaneous event rates between treatment groups, subject to the model's assumptions.
A hazard ratio below 1 would correspond to a lower estimated instantaneous event rate under eFTR than under surgery at the modeled comparison. It would not by itself describe an absolute three-year risk difference.
Composite endpoints
The short-term endpoint illustrates an important design principle: combining events can improve statistical efficiency, but the components must be clinically interpretable together. The treatment effect on the composite can differ from the effect on each individual component.
7. Statistical Methods Explained
Why use a composite endpoint for the 30-day outcome?
A composite can capture several forms of short-term harm in one prespecified outcome. This may provide more observed events than a single component and can make the randomized comparison more informative statistically. The trade-off is that the composite's interpretation depends on the relative frequency and clinical importance of its components.
What does "within 30 days" mean statistically?
The 30-day time frame defines the observation window for the short-term primary endpoint. The analysis therefore focuses on whether a qualifying event occurs during that specified period following assigned treatment. A participant without a qualifying event during the window contributes to the denominator of the fixed-time comparison.
Why is the three-year endpoint different from the 30-day endpoint?
The three-year endpoint addresses a substantially longer period and a different clinical process: cancer recurrence or CRC death. It may therefore require methods that account for different follow-up durations and censoring. A time-to-event analysis is often useful for such endpoints because it uses information from the timing of events rather than reducing all follow-up to a single yes-or-no observation.
What is the difference between a risk ratio and a hazard ratio?
A risk ratio compares the probabilities of an event over a defined period. A hazard ratio compares instantaneous event rates over time within a time-to-event framework. They are not interchangeable. A hazard ratio does not directly tell a reader the difference in three-year event probabilities.
Why does the analysis need to distinguish the components of the composite?
A composite endpoint is a combined outcome, but its components can occur at very different frequencies. If one component dominates the event count, the overall treatment effect may largely reflect that component. Reporting component-specific results helps readers understand the clinical meaning behind the composite result.
Why does censoring matter for the three-year endpoint?
If some participants are not observed for the entire follow-up period, their available follow-up still contains information. Survival methods can incorporate their observed time and censor them when follow-up ends. The validity of the resulting estimates depends on assumptions about the censoring process and how follow-up is conducted.
What would a hazard ratio below 1 mean?
For a recurrence-or-death time-to-event analysis, a hazard ratio below 1 would indicate a lower estimated instantaneous event rate in the eFTR group relative to the surgery group under the fitted model. It would not mean that the same percentage of every participant's risk was reduced, nor would it directly equal the difference in three-year event rates.
8. Interpreting the Two Primary Endpoints Together
The ETHOS design separates short-term treatment consequences from longer-term cancer outcomes. This is statistically important because an intervention can affect these domains differently.
Short-term domain
The 30-day endpoint focuses on severe adverse events, re-admissions and deaths following assigned treatment. It captures early consequences of the treatment strategy.
Long-term domain
The three-year endpoint focuses on cancer recurrence or CRC death, with the registry description also referring to signs involving lymph nodes or distant metastases.
Different estimands
The two endpoints describe different outcome processes and should be interpreted separately rather than converted into one overall numerical treatment effect.
Clinical context
A complete assessment requires understanding both early treatment-related events and longer-term cancer outcomes.
This structure also affects statistical interpretation. A result for the 30-day composite cannot be assumed to predict the three-year cancer outcome. Conversely, a long-term cancer endpoint does not fully describe the immediate procedural experience captured by the 30-day endpoint.
9. Multiplicity and Co-Primary Endpoints
ETHOS has two registered primary endpoints. When a trial design includes multiple primary endpoints, the statistical analysis plan normally specifies how they are jointly interpreted and how the overall type I error is controlled.
The registry information presented for ETHOS does not specify an alpha-allocation scheme, testing hierarchy, gatekeeping procedure, or other multiplicity-control method. Those design details are important when interpreting formal hypothesis tests once results are available.
| Endpoint | Role | Statistical issue |
|---|---|---|
| Rate of severe adverse events, re-admissions and deaths within 30 days | Primary | Fixed-time composite endpoint; interpretation depends on component events and the prespecified comparison method. |
| Rate of cancer recurrence or CRC death after three years | Primary | Long-term cancer endpoint; time-to-event methods may be appropriate when follow-up and censoring vary. |
With multiple primary endpoints, the meaning of a statistically significant result depends on the prespecified success criterion. For example, some trials require success on every co-primary endpoint, while others specify a multiplicity-adjusted testing strategy. The registry record does not identify which framework ETHOS will use.
10. Missing Data, Censoring, and Follow-up
Missingness has different implications for the two primary endpoints. For the 30-day composite, incomplete ascertainment of events during the fixed 30-day window can directly affect the estimated event rate. For the three-year endpoint, loss to follow-up can produce censoring and therefore affects the time-to-event analysis.
30-day endpoint
Complete ascertainment of severe adverse events, re-admissions and deaths during the specified 30-day window is important because the endpoint is defined over a fixed period.
Three-year endpoint
Longitudinal follow-up determines how much event and censoring information contributes to the analysis.
The registry information does not specify an imputation method, censoring rules, competing-risk framework, or sensitivity-analysis strategy. These are important components of a statistical analysis plan for a long-term cancer endpoint.
11. Competing Events and Composite Interpretation
The three-year endpoint combines cancer recurrence with CRC death. Death can prevent a participant from subsequently experiencing a recurrence event, creating an important distinction between a composite endpoint and separate component analyses.
When death and recurrence are analyzed separately, the statistical treatment of death can depend on the scientific question. In some settings, death is treated as a competing event for recurrence; in others, recurrence or death is intentionally combined into a composite to capture the first clinically relevant event. The ETHOS registry explicitly defines the primary long-term outcome as CRC death or cancer recurrence, making the composite endpoint itself the central registered outcome.
12. Non-Inferiority, Crossover, and Bayesian Methods
The registry information identifies ETHOS as a randomized phase 3 parallel-group treatment trial but does not report a non-inferiority margin, crossover design, or Bayesian analysis framework.
| Design topic | Registry information | Statistical implication |
|---|---|---|
| Non-inferiority margin | Not specified in the registry information presented. | No non-inferiority margin can be incorporated into interpretation without the prespecified design. |
| Crossover | Not specified in the registry information presented. | No crossover adjustment or dilution analysis is identified. |
| Bayesian methods | Not specified in the registry information presented. | No posterior probability or prior-based analysis is identified. |
| Interim analysis | Not specified in the registry information presented. | No interim efficacy boundary or alpha-spending framework is identified. |
13. Trial Timeline
Trial start
The registry lists August 25, 2025 as the ETHOS study start date.
Recruiting
The ClinicalTrials.gov record identifies the study status as Recruiting.
Planned primary completion
The registry lists September 1, 2033 as the primary completion date.
14. What a Future Result Would Need to Show
Because no formal statistical analyses are posted to ClinicalTrials.gov, the eventual interpretation will depend on the numerical results and the prespecified analysis framework. For the 30-day primary endpoint, readers would ideally see the number of participants experiencing the composite in each randomized group, the corresponding event rates, an effect estimate with a confidence interval, and the relevant hypothesis test.
For the three-year primary endpoint, a complete statistical presentation would ordinarily distinguish the number and timing of cancer recurrences and CRC deaths, describe censoring, present an appropriate time-to-event estimate, and report a treatment-effect measure with its confidence interval.
| Future result element | Why it matters |
|---|---|
| Composite event counts | Shows how many participants contributed events to the short-term primary endpoint. |
| Component event counts | Shows whether severe adverse events, re-admissions or deaths drive the composite. |
| Absolute event rates | Provides direct clinical context for the magnitude of the short-term outcome. |
| Effect estimate and confidence interval | Quantifies the relative or absolute treatment difference and its precision. |
| Three-year cancer outcome | Describes the long-term oncologic endpoint. |
| Time-to-event analysis | Accounts for the timing of recurrence or CRC death and variable follow-up when applicable. |
| Multiplicity framework | Determines how evidence across the two primary endpoints should be interpreted. |
15. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record does not contain formal statistical analyses or numerical outcome results for the registered primary endpoints.
- Endpoint-specific interpretation: the two primary endpoints address different follow-up periods and different clinical domains, so they should not be treated as interchangeable outcomes.
- Composite endpoint: the short-term primary outcome combines severe adverse events, re-admissions and deaths. The contribution of each component can materially affect interpretation of the overall composite.
- Long-term follow-up: the three-year cancer endpoint depends on sustained follow-up and accurate ascertainment of recurrence and CRC death.
- Censoring: time-to-event analysis of the long-term endpoint can depend on assumptions concerning censoring and follow-up.
- Analysis specification: the registry information does not identify the primary effect measure, confidence-interval method, multiplicity procedure, missing-data strategy, stratification factors, or interim-analysis framework.
- Generalizability: the trial's applicability depends on the eligibility criteria and clinical setting represented by its enrolled population.
16. Why This Trial Matters Statistically
ETHOS is a useful statistical teaching case because it places two very different endpoint structures within the same randomized phase 3 trial. The first primary endpoint is a short-term composite rate, while the second is a longer-term cancer outcome measured after three years.
| Statistical concept | How it appears in ETHOS |
|---|---|
| Randomization | Participants are randomized between two treatment procedures. |
| Parallel-group design | The registry identifies a parallel design with two arms. |
| Composite endpoint | The 30-day primary endpoint combines severe adverse events, re-admissions and deaths. |
| Fixed-time event rate | The short-term primary endpoint is evaluated within 30 days. |
| Time-to-event analysis | The three-year cancer endpoint can naturally be considered in a time-to-event framework when follow-up varies. |
| Censoring | Long-term follow-up may involve participants who are not observed for the entire three-year period. |
| Absolute versus relative effects | Event rates and comparative effect measures answer different questions and should be presented together. |
| Multiplicity | Two registered primary endpoints require a prespecified framework for interpreting multiple primary tests. |
| Component analysis | The meaning of the 30-day composite depends on its severe adverse event, re-admission and death components. |
| Longitudinal outcome assessment | The three-year endpoint requires continued follow-up beyond the initial treatment period. |
17. Statistical Interpretation of the Trial Design
Randomization is the central feature that supports a causal comparison between eFTR and surgery. It separates treatment assignment from the participants' subsequent outcomes, subject to successful implementation of the randomized design.
The 30-day composite increases the scope of the short-term outcome by counting several types of clinically important events. Its interpretation must remain tied to the exact registered components rather than treating it as a generic measure of adverse events.
The three-year endpoint shifts the statistical focus from immediate treatment consequences to cancer recurrence or CRC death. A time-to-event framework can preserve information about when those events occur rather than considering only whether an event occurred by a single final date.
When numerical results become available, confidence intervals will be essential. An effect estimate alone describes the observed treatment comparison, whereas its confidence interval provides information about statistical precision.
If hypothesis tests are reported, a p-value will describe the compatibility of the observed data with the specified null hypothesis under the statistical testing framework. It does not measure the magnitude or clinical importance of an effect.
18. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The ETHOS design creates a randomized comparison of two procedures and prespecifies two primary outcomes with different time horizons. The appropriate analysis must respect the structure of each endpoint.
Clinical interpretation
The short-term endpoint concerns severe adverse events, re-admissions and deaths after treatment, while the long-term endpoint concerns cancer recurrence or CRC death. These outcomes represent distinct aspects of treatment experience.
Statistical significance, when eventually reported, will not by itself determine the clinical importance of either endpoint. Absolute event rates, the individual components of the composite, the precision of the estimates and the duration of follow-up will all contribute to interpretation.
19. Sources
- ClinicalTrials.gov: ETHOS — NCT06940947.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical trial analyses and statistical resources from Clinical Biostats.
20. Record Summary
ETHOS is a randomized phase 3 parallel-group trial of eFTR versus surgery for colon cancer, with planned enrollment of 434 participants. Its two registered primary endpoints deliberately span different aspects of treatment: a 30-day composite rate of severe adverse events, re-admissions and deaths, and a three-year cancer endpoint involving recurrence or CRC death and signs involving lymph nodes or distant metastases.
The statistical structure is therefore best understood endpoint by endpoint. The short-term outcome can be treated as a fixed-time event-rate comparison, while the longer-term outcome naturally raises time-to-event questions involving recurrence, death, censoring and three-year follow-up. The presence of two primary endpoints also makes multiplicity and the prespecified success criterion important considerations.
The registry identifies the study as recruiting, with a start date of August 25, 2025 and a primary completion date of September 1, 2033. No formal statistical analyses or numerical outcome results are posted in the ClinicalTrials.gov record. The statistical interpretation at this stage therefore centers on the randomized design, endpoint construction, appropriate analysis frameworks, and the principles that will govern interpretation when results become available.