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
NICHE is a randomized, open-label phase 2 trial evaluating neoadjuvant immune checkpoint inhibition and novel immuno-oncology combinations in patients with colon carcinoma. The registry describes 7 study arms and an enrollment of 353 participants.
| Feature | NICHE |
|---|---|
| Clinical trial identifier | NCT03026140 |
| Acronym | NICHE |
| Phase | Phase 2 |
| Condition | Colon Carcinoma |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 353 |
| Number of arms | 7 |
| Status | Recruiting |
| Start date | 2017-03-29 |
| Primary completion date | 2032-03-01 |
| Lead sponsor | The Netherlands Cancer Institute |
2. Clinical Question
The trial addresses whether neoadjuvant immune checkpoint inhibition and related immuno-oncology combinations can provide clinically relevant efficacy and acceptable safety in early-stage colon cancer. The registry identifies disease-free survival and major pathological response as efficacy endpoints and adverse-event incidence as a primary safety endpoint.
Population
Patients enrolled in a phase 2 study of early-stage colon cancer, with the registry condition listed as colon carcinoma.
Interventions
The registry lists nivolumab, ipilimumab, celecoxib 200mg, BMS-986253, and BMS-986016 among the study interventions.
Comparator structure
The study uses randomized allocation across 7 parallel arms. The registry does not provide a single two-arm comparator structure.
Primary questions
How frequently do adverse events occur during treatment and follow-up, and what are the disease-free survival and major pathological response outcomes associated with the registered treatment strategies?
3. Trial Design
NICHE is registered as a randomized, parallel-group phase 2 trial with no masking and a primary purpose of treatment. The 7-arm structure is important statistically because treatment effects are not naturally represented by a single intervention-versus-control comparison.
Immune checkpoint and IO agents
- Nivolumab
- Ipilimumab
- BMS-986253
- BMS-986016
Combination component
- Celecoxib 200mg
- Seven total randomized study arms
- Parallel-group allocation
- No masking
4. Trial Timing and Registry Status
Study start
The registered study start date is March 29, 2017.
Recruiting
The ClinicalTrials.gov record lists the study status as Recruiting.
Primary completion date
The registered primary completion date is March 1, 2032.
The combination of a Recruiting status and a registered primary completion date in 2032 is important when interpreting the statistical maturity of the study. A trial can have registered endpoints and a prespecified analysis framework while still not having posted efficacy estimates.
5. Registered Primary Endpoints
| Primary endpoint | Time frame | Registry definition |
|---|---|---|
| Incidence of adverse events during the treatment and follow-up (safety) | until 100 days after last patient last study drug treatment | Adverse events will be assessed according to CTCAE v4.0 during treatment and follow-up. |
| Disease free survival | until 5 years after diagnosis | To assess efficacy of neoadjuvant ipilimumab plus nivolumab in terms of disease-free survival. |
| Major Pathological Response | From date of randomization until the date of first documented progression, assessed up to 63 months | To assess efficacy of neoadjuvant nivolumab monotherapy and neoadjuvant nivolumab plus relatlimab in terms of major pathologic response. |
6. Endpoint Structure and Statistical Questions
The registered primary endpoints combine a safety endpoint with two efficacy endpoints. That structure creates three distinct statistical questions: the frequency of adverse events, the distribution of time without disease recurrence or progression according to the registered definition, and the frequency or timing of major pathological response.
| Endpoint | Statistical data type | Natural analysis framework |
|---|---|---|
| Incidence of adverse events | Binary or count-based safety outcomes over treatment and follow-up | Arm-specific incidence proportions, event counts, severity summaries, and appropriate comparative methods where prespecified |
| Disease free survival | Time-to-event endpoint | Kaplan-Meier estimation and time-to-event comparison; a Cox model may be used when appropriate |
| Major Pathological Response | Pathological response endpoint | Arm-specific response proportions and an appropriate comparative model or exact method depending on the prespecified analysis plan |
7. Planned Analysis
No formal statistical analyses are posted on ClinicalTrials.gov for NICHE. The registry therefore provides the planned primary endpoints and their definitions, but no posted estimates, confidence intervals, p-values, event counts, or arm-specific efficacy results.
Disease Free Survival
Disease free survival is a time-to-event endpoint. The usual statistical representation is a Kaplan-Meier estimate of the probability of remaining disease-free over time, with treatment groups compared using an appropriate time-to-event test and, when prespecified, a Cox proportional-hazards model to estimate a hazard ratio.
Here, T represents the time from the relevant starting point to the event defined by the disease-free-survival endpoint. Patients without an observed event by the end of available follow-up are typically right-censored.
Because the registry specifies disease free survival until 5 years after diagnosis, a complete analysis would need to respect that endpoint definition and its associated censoring rules. The exact event definition, censoring conventions, stratification factors, and model specification are not reported in the registry information available here.
Major Pathological Response
Major Pathological Response is registered as an efficacy endpoint for neoadjuvant nivolumab monotherapy and neoadjuvant nivolumab plus relatlimab. Because this endpoint is fundamentally a pathological response classification rather than a conventional continuous outcome, its analysis would ordinarily begin with arm-specific response counts and proportions.
If the prespecified analysis compares response rates between randomized groups, methods could include a binomial regression model, risk-ratio or risk-difference estimation, or an exact categorical test, depending on the statistical analysis plan. The appropriate method should be determined from the prespecified estimand and analysis population rather than selected after seeing the response data.
Safety
The safety endpoint measures the incidence of adverse events during treatment and follow-up, with assessment continuing until 100 days after last patient last study drug treatment. Adverse events are assessed according to CTCAE v4.0.
A conventional safety analysis would summarize adverse-event incidence separately by randomized arm and would distinguish the overall occurrence of adverse events from severity categories such as CTCAE grades. With a 7-arm design, the number and nature of pairwise comparisons can become substantial, making prespecification of the safety summaries particularly important.
8. Statistical Methodology
Randomization and causal comparison
Randomization is the central design feature that allows differences between treatment groups to be interpreted as treatment effects, subject to adherence to the protocol and the prespecified analysis. In a randomized parallel trial, treatment assignment is determined before outcome assessment, reducing the role of measured and unmeasured baseline confounding in the comparison of randomized groups.
Seven-arm parallel design
A 7-arm randomized trial produces a vector of potential treatment contrasts rather than one universal treatment effect. For example, a prespecified comparison could contrast one intervention against another, or multiple experimental arms could each be compared with a common reference group. The statistical meaning of the trial therefore depends on the specific contrast being tested.
For two selected randomized arms, the null hypothesis states that their treatment effects are equal on the chosen estimand. With seven arms, several such contrasts can be defined, which is why the prespecified comparison strategy and multiplicity control are important.
Kaplan-Meier estimation for disease-free survival
For a time-to-event endpoint, Kaplan-Meier estimation provides a nonparametric estimate of the survival function while accommodating right censoring.
The estimate is built from the number of events di and the number at risk ni at each observed event time.
The resulting curve can show the estimated probability of remaining disease-free over time. It does not require every participant to have complete follow-up, provided the censoring assumptions and endpoint definitions are appropriately specified.
Hazard ratios and Cox regression
A Cox proportional-hazards model can summarize a relative time-to-event effect as a hazard ratio. For two treatment groups, a hazard ratio below 1 corresponds to a lower estimated instantaneous event rate in one group relative to the reference group under the fitted model.
For a binary treatment indicator, exp(β) represents the model-based hazard ratio comparing the treatment groups, subject to the proportional-hazards assumption when that model is used.
The registry does not report a posted Cox model specification for NICHE. Therefore, whether the final analysis uses an unadjusted model, a stratified model, or a covariate-adjusted model cannot be established from the registry information alone.
Response proportions
Major pathological response is naturally summarized as the proportion of evaluable patients meeting the prespecified pathological-response definition. An estimated response proportion can be accompanied by a confidence interval to describe sampling uncertainty.
For randomized comparisons, a risk difference, risk ratio, odds ratio, or model-based treatment contrast can be used depending on the prespecified estimand. These measures are not interchangeable: a risk difference is an absolute contrast, whereas a risk ratio and odds ratio are relative measures on different scales.
Adverse-event incidence
For the safety endpoint, the basic estimand is the probability or incidence proportion of experiencing an adverse event during the defined observation period. Because the registry specifies CTCAE v4.0, severity grading provides an additional dimension beyond whether an event occurred.
Safety analysis should preserve the randomized-arm structure while also accounting for actual exposure where appropriate. The precise treatment-emergent definition, handling of recurrent events, denominator rules, and treatment of patients with incomplete safety follow-up are not reported in the registry information available here.
9. Statistical Methods Explained
Why is disease free survival analyzed differently from major pathological response?
Disease free survival contains information about when an event occurs and can include patients who remain event-free at the end of observation. That makes it a time-to-event endpoint. Major pathological response is primarily a response classification, so its natural first summary is a proportion of patients meeting the pathological criterion. Applying the same statistical method to both endpoints would ignore their different data structures.
Why is a Kaplan-Meier method useful for disease free survival?
Not every participant necessarily experiences the event during the study. Kaplan-Meier estimation allows participants who are still disease-free when follow-up ends to contribute information up to their censoring time. This is particularly important for long-term endpoints such as disease free survival assessed for up to 5 years after diagnosis.
What does a hazard ratio represent?
A hazard ratio compares the estimated instantaneous event rates between two groups under a time-to-event model. A hazard ratio of 1 would indicate equal hazards under the model, while a value below 1 indicates a lower estimated hazard for the numerator treatment group. It is not a percentage of patients who benefit and is not automatically equivalent to a relative reduction in cumulative risk at a particular time point.
Why does a 7-arm trial require careful choice of contrasts?
With seven randomized groups, there are many possible pairwise comparisons. Testing many contrasts without an appropriate multiplicity strategy increases the chance of observing a statistically significant result simply by chance. A strong analysis therefore starts with the scientific comparisons defined before the outcomes are examined.
Why is the safety endpoint followed for a defined period after treatment?
Adverse events can occur after the last dose of study treatment. The registry therefore defines the safety observation period as continuing until 100 days after last patient last study drug treatment. This creates a common observation rule for the registered primary safety endpoint rather than limiting assessment strictly to the treatment period.
Why does the 5-year disease-free-survival time frame matter?
A time frame defines the period over which the endpoint is observed and interpreted. A 5-year endpoint can capture events occurring long after treatment, but it also creates more opportunity for censoring, competing clinical events, changes in subsequent management, and loss to follow-up. The analysis therefore depends on prespecified rules for these situations.
Why should pathological response not be treated as the same thing as disease-free survival?
A pathological response is an assessment of tumor response based on pathological findings, whereas disease-free survival measures the time until a defined event. A patient can have a major pathological response without that response being equivalent to a specific duration of disease-free survival. The endpoints provide different information about treatment effect.
10. Multiplicity in a Seven-Arm Trial
Multiplicity is especially important when a randomized trial contains 7 arms. If several treatment comparisons are tested independently at the same nominal significance level, the probability of at least one false-positive finding can exceed the intended type I error rate.
Single prespecified contrast
If the scientific question is limited to one prespecified comparison, the multiplicity problem can be much smaller than in a broad set of pairwise tests.
Multiple treatment contrasts
If several experimental arms are compared with a reference arm, the analysis may require a multiplicity-control procedure to preserve the desired familywise error rate.
Multiple primary endpoints
NICHE has three registered primary endpoints spanning safety and efficacy. The relationship among those endpoints must be specified to determine how confirmatory inference should be interpreted.
Exploratory comparisons
Descriptive comparisons can be useful for understanding a multi-arm study, but nominal statistical significance should not automatically be interpreted as confirmatory evidence.
11. Safety Analysis Framework
The first registered primary endpoint is the incidence of adverse events during treatment and follow-up, with follow-up continuing until 100 days after last patient last study drug treatment. Adverse events are assessed according to CTCAE v4.0.
| Safety feature | Registered specification |
|---|---|
| Endpoint | Incidence of adverse events during the treatment and follow-up (safety) |
| Assessment system | CTCAE v4.0 |
| Observation period | Until 100 days after last patient last study drug treatment |
| Randomized structure | 7 parallel arms |
| Posted statistical results | None |
For a multi-arm safety analysis, the most informative presentation would ordinarily show adverse-event incidence separately for each randomized arm, with severity categories and clinically relevant event types reported using consistent denominators. If recurrent events are of interest, counts of events and numbers of affected patients answer different questions and should not be conflated.
Safety interpretation also differs from efficacy interpretation. A treatment can have a favorable efficacy estimate while increasing certain adverse events, or vice versa. The safety endpoint therefore requires its own estimand and analysis rather than being folded into a single numerical benefit-risk statistic.
12. Disease Free Survival: Statistical Interpretation
Disease free survival is registered with a time frame of until 5 years after diagnosis. Its principal statistical feature is that both event occurrence and event timing contribute information.
Disease free survival describes the distribution of time from the defined starting point until the disease-free-survival event occurs. Kaplan-Meier methods can estimate the probability of remaining event-free over time, while regression models can compare treatment groups.
Disease free survival is not the same as overall survival, pathological response, or a simple proportion of patients who remain disease-free. Each endpoint has its own event definition and follow-up structure.
Patients without an observed event by the end of their available follow-up are typically censored rather than treated as having experienced the event. The validity of the resulting survival estimate depends on the prespecified censoring rules and assumptions about the relationship between censoring and future event risk.
13. Major Pathological Response: Statistical Interpretation
The registry identifies major pathological response as a primary efficacy endpoint for neoadjuvant nivolumab monotherapy and neoadjuvant nivolumab plus relatlimab. The registered time frame runs from randomization until the date of first documented progression, assessed up to 63 months.
| Feature | Registered specification |
|---|---|
| Endpoint | Major Pathological Response |
| Population of interest | Neoadjuvant nivolumab monotherapy and neoadjuvant nivolumab plus relatlimab |
| Starting point | Date of randomization |
| Endpoint time frame | Until the date of first documented progression |
| Assessment horizon | Up to 63 months |
The statistical interpretation of a pathological response proportion depends strongly on the analysis population. If the denominator is all randomized patients, the estimand reflects an intention-to-treat framework. If the denominator is restricted to patients with an evaluable pathological assessment, the estimand answers a different question. The registry information available here does not specify the final denominator rule.
14. What a Future Results Analysis Should Report
When results are posted, the most useful statistical presentation would preserve the randomized 7-arm structure and report each primary endpoint according to its own data type.
| Endpoint | Key quantities to report | Statistical interpretation |
|---|---|---|
| Adverse-event incidence | Number and proportion affected by arm; CTCAE severity; relevant event categories | Describes the observed safety burden over the registered observation period |
| Disease free survival | Events, censoring, Kaplan-Meier estimates, time-specific survival estimates, treatment contrast, confidence interval, and p-value when applicable | Describes relative and absolute differences in time to the registered disease-free-survival event |
| Major Pathological Response | Response counts, denominators, response proportions, confidence intervals, and prespecified treatment contrasts | Describes the frequency of major pathological response in the relevant randomized treatment groups |
For every reported effect estimate, the analysis should distinguish the estimate itself from its uncertainty. A p-value addresses evidence against a specified null hypothesis; it does not describe the magnitude or clinical importance of the observed effect.
15. Confidence Intervals and Effect Size
Confidence intervals are especially important in a phase 2 multi-arm trial because the amount of information available for each individual treatment contrast may be substantially smaller than the total enrollment of 353 participants.
An effect estimate provides the direction and magnitude of the observed contrast, while the confidence interval describes statistical uncertainty under the specified analysis framework.
For a response endpoint, an absolute risk difference can be more directly interpretable than an odds ratio. For disease-free survival, a hazard ratio can summarize relative event rates, while Kaplan-Meier estimates at clinically relevant time points provide an absolute perspective. The two measures answer different questions and should ideally be presented together.
A p-value does not tell the reader how large the treatment difference is. With sufficiently large information, a small effect can produce a small p-value; with limited information, a clinically meaningful effect can have substantial statistical uncertainty. Effect estimates and confidence intervals are therefore essential to interpretation.
16. Limitations and Interpretation Issues
- No posted statistical analyses: the ClinicalTrials.gov record does not currently post formal statistical analyses or numerical results for the registered primary endpoints.
- Seven-arm complexity: multiple randomized arms create multiple possible treatment contrasts, making the prespecified comparison strategy important for interpretation.
- Multiplicity: the registry information available here does not specify a posted multiplicity-adjustment strategy for the 7-arm design and multiple primary endpoints.
- Endpoint heterogeneity: safety incidence, disease free survival, and major pathological response have different data structures and should not be analyzed or interpreted as if they were the same endpoint.
- Analysis population: the registry does not provide the complete final definitions for every analysis population or denominator that would be used for pathological response and safety summaries.
- Censoring: disease-free-survival analysis depends on the timing and rules for censoring observations without a documented event.
- Model assumptions: if a Cox proportional-hazards model is used, the proportional-hazards assumption becomes relevant to interpretation of a single hazard ratio over follow-up.
- Open-label design: the registry identifies masking as none. This can matter for outcomes susceptible to assessment or management decisions, although objective pathological and time-to-event endpoints can have different degrees of susceptibility.
- Recruiting status: the study remains registered as Recruiting, with a primary completion date of 2032-03-01. The absence of posted results must therefore be interpreted in the context of the registered study status and timeline.
17. Why This Trial Matters Statistically
NICHE is a useful statistical teaching case because it combines randomized treatment assignment, a 7-arm parallel structure, multiple primary endpoints, a long-term time-to-event outcome, a pathological response outcome, and a defined post-treatment safety window.
| Concept | How it appears in NICHE |
|---|---|
| Randomization | Participants are allocated randomly across 7 study arms. |
| Parallel-group design | The registry identifies a parallel design model rather than a crossover design. |
| Multi-arm inference | Seven randomized arms create multiple possible treatment contrasts. |
| Multiplicity | Multiple arms and three registered primary endpoints make control of false-positive inference an important design consideration. |
| Time-to-event analysis | Disease free survival is followed until 5 years after diagnosis. |
| Pathological response | Major Pathological Response provides a response-based efficacy endpoint. |
| Safety surveillance | Adverse events are followed until 100 days after last patient last study drug treatment. |
| CTCAE grading | Adverse events are assessed according to CTCAE v4.0. |
| Open-label design | The registry specifies no masking. |
| Statistical maturity | The registry lists the study as Recruiting and does not post formal statistical analyses. |
18. Statistical Concepts in This Trial
Several core statistical ideas can be used to understand the NICHE design even before numerical results are available. The most important distinction is between the design question, the endpoint-specific estimand, and the statistical method used to estimate that estimand.
Design question
Which randomized treatment contrasts are intended to answer the scientific question in a 7-arm study?
Estimand
What treatment effect is being estimated, in which population, over what period, and according to which endpoint definition?
Statistical method
Which model or test appropriately estimates and compares the endpoint while respecting its data structure?
Uncertainty
How precisely can the treatment effect be estimated, and what does the confidence interval imply about plausible values under the analysis framework?
19. Registry Information Summary
| Registry element | Registered information |
|---|---|
| NCT ID | NCT03026140 |
| Official title | Neoadjuvant Immune Checkpoint Inhibition and Novel IO Combinations in Early-stage Colon Cancer |
| Phase | 2 |
| Status | Recruiting |
| Start date | 2017-03-29 |
| Primary completion date | 2032-03-01 |
| Lead sponsor | The Netherlands Cancer Institute |
| Sponsor type | OTHER |
| Condition | Colon Carcinoma |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 353 |
| Number of arms | 7 |
20. Sources
- ClinicalTrials.gov: NCT03026140 — NICHE.
- PubMed: PubMed record 41115454.
- PubMed: PubMed record 39278994.
- PubMed: PubMed record 38838316.
- PubMed: PubMed record 38838311.
- PubMed: PubMed record 34606846.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical-trial analyses and statistical methods for randomized studies, survival endpoints, safety outcomes, and treatment-effect interpretation.
21. Record Summary
NICHE is a randomized phase 2, 7-arm, parallel-group trial in colon carcinoma evaluating neoadjuvant immune checkpoint inhibition and novel immuno-oncology combinations. Its registered primary endpoints span three distinct statistical domains: adverse-event incidence through 100 days after last patient last study drug treatment, disease free survival until 5 years after diagnosis, and Major Pathological Response assessed from randomization through first documented progression up to 63 months.
The principal statistical challenge is the combination of a multi-arm randomized design with multiple primary endpoints of different types. Disease free survival naturally calls for time-to-event methods such as Kaplan-Meier estimation and potentially Cox regression, whereas Major Pathological Response is naturally summarized through response proportions and treatment contrasts. Safety requires arm-specific incidence and CTCAE v4.0 severity summaries over the defined observation period.
The ClinicalTrials.gov record currently lists the study as Recruiting, with 353 participants and 7 arms, and does not post formal statistical analyses or numerical results for the registered primary endpoints. Consequently, the central statistical lesson at this stage is the structure of the planned evidence: randomized treatment contrasts, endpoint-specific estimands, appropriate handling of censoring for disease free survival, response-rate analysis for pathological outcomes, safety surveillance, and careful control of multiplicity in a 7-arm trial.