MOUNTAINEER: Complete Statistical Analysis of Tucatinib Plus Trastuzumab in Metastatic Colorectal Adenocarcinoma

1. Trial at a Glance

MOUNTAINEER was a randomized, parallel, unmasked phase 2 study in patients with metastatic colorectal adenocarcinoma. The registry identifies tucatinib and trastuzumab as the study interventions and specifies a primary confirmed objective response rate endpoint based on RECIST v1.1 and blinded independent central review in pooled Cohorts A+B.

2. Clinical Question

The clinical question centers on the antitumor activity of tucatinib plus trastuzumab in patients with HER2-positive metastatic colorectal cancer. The registry defines the primary efficacy endpoint as confirmed objective response rate in pooled Cohorts A+B, using blinded independent central review and RECIST v1.1 criteria.

3. Trial Design

The presence of three randomized arms is important when interpreting the trial statistically. The primary endpoint is specifically defined for pooled Cohorts A+B rather than for all three arms combined. Pooling therefore forms part of the endpoint definition itself, and the denominator used for a response-rate calculation should correspond to the prespecified pooled efficacy population rather than simply the overall enrollment.

4. Study Cohorts and Safety Populations

The registry reports safety experience separately for pooled Cohorts A+B and for Cohort C before and after crossover. These categories describe different exposure periods and should not be treated as independent randomized treatment groups.

The pre-crossover and post-crossover categories for Cohort C are exposure-defined periods rather than two independently randomized cohorts. Consequently, their affected and at-risk counts should be read as longitudinal safety information rather than as a conventional head-to-head comparison.

5. Primary Endpoint

What exactly counts as a response?

The endpoint is a proportion, but it is not simply a proportion of patients with any radiologic improvement. A confirmed complete response requires disappearance of all target lesions together with the specified lymph-node criterion. A partial response requires at least a 30% decrease in the post-baseline sum of diameters relative to baseline. The word confirmed indicates that the response classification is subject to the confirmation requirements associated with the endpoint.

The endpoint also specifies blinded independent central review. This is statistically important because the assessment of tumor response can otherwise be influenced by knowledge of treatment assignment or by differences among local investigators in applying radiologic criteria.

6. Planned Analysis

The registry has results posted for the study, but no formal statistical analyses are posted in the record for the primary endpoint. The endpoint is a binary response outcome summarized as a percentage of participants achieving a confirmed CR or PR.

The important denominator is the population specified for the endpoint: pooled Cohorts A+B. It should not automatically be replaced by the overall enrollment of 117, because enrollment and the primary efficacy analysis population are different concepts.

Why confirmation matters statistically

A response endpoint that requires confirmation is designed to distinguish a sustained response classification from a single assessment that might later fail to meet response criteria. This reduces the extent to which a transient radiologic change is counted as a confirmed response.

Why blinded central review matters

Independent central review provides a common assessment framework across participants. Because the primary endpoint depends on radiologic classification, centralized review can reduce variability associated with local assessment and can make the endpoint more reproducible across treatment groups.

7. Statistical Methodology

Binary endpoint analysis

The primary endpoint is naturally represented as a binary outcome for each participant: confirmed response or no confirmed response. The corresponding population-level quantity is a binomial proportion.

The proportion is easy to interpret, but its precision depends on the number of participants contributing to the analysis and on the number of observed responses. A confidence interval therefore belongs alongside the response estimate when presenting a statistical result.

Confidence intervals for a response rate

For a binomial response endpoint, an exact binomial confidence interval can be used to quantify uncertainty around the estimated response proportion. This is especially useful when the number of participants is not very large or when the response proportion is close to the boundaries of 0% or 100%.

RECIST v1.1 classification

The statistical endpoint depends on a clinical measurement framework. RECIST v1.1 converts longitudinal measurements of target lesions and lymph nodes into categorical response assessments such as complete response and partial response. The resulting categories then form the binary outcome used for cORR.

Blinded independent central review

Because treatment assignment is unmasked, independent blinded review is particularly relevant to the objectivity of the primary response assessment. The statistical analysis begins with the response classifications produced by that review process; the quality of the final proportion therefore depends on consistent application of the radiologic response rules.

Pooled efficacy population

The primary endpoint is explicitly defined in pooled Cohorts A+B. Pooling can increase the amount of information available for a response estimate, but it also means that the resulting cORR represents the prespecified pooled population rather than a response rate for each cohort separately.

8. Statistical Methods Explained

Why is cORR a binary statistical endpoint?

Each participant ultimately contributes a response classification for the endpoint: a confirmed CR or PR qualifies as a response, while a participant who does not meet the confirmed response definition does not qualify. This permits the primary endpoint to be summarized as a proportion.

What does a response rate actually measure?

A response rate measures the proportion of participants who achieve the prespecified degree of tumor response. It does not measure the magnitude of benefit for every participant and does not describe the duration of that response.

Why use an independent central review?

Radiologic response assessment contains an element of measurement and classification. A blinded independent central review provides an assessment process separated from knowledge of treatment assignment and applies the same response framework across the study.

Why does the confidence interval matter?

An observed response percentage is an estimate rather than a perfectly known population quantity. The confidence interval indicates how much statistical uncertainty surrounds that estimate. Wider intervals generally indicate less precision than narrower intervals.

Why should the overall enrollment not automatically be the cORR denominator?

The registry defines the primary endpoint in pooled Cohorts A+B. A trial can enroll participants across multiple cohorts or arms while defining a particular endpoint for a specific analysis population. Statistical interpretation should therefore follow the endpoint's population definition.

What does crossover change?

Crossover means that treatment exposure can change over time for some participants. For Cohort C, the registry separately reports pre-crossover and post-crossover safety categories. Those periods should not be interpreted as though they were two independently randomized groups.

Why is cORR different from a survival endpoint?

cORR is a fixed categorical response measure. Survival endpoints incorporate the timing of events and censoring. A response rate therefore cannot substitute for an analysis of progression-free or overall survival, and a response rate alone cannot establish the duration of clinical benefit.

9. Results

The ClinicalTrials.gov record indicates that results have been posted for MOUNTAINEER. For the primary endpoint, however, the record represented here does not contain a posted statistical analysis with a response estimate, confidence interval, or p-value.

How the primary result would ordinarily be interpreted

If a cORR estimate were reported, the percentage would represent the proportion of participants in the prespecified pooled Cohorts A+B efficacy population who achieved a confirmed CR or PR according to RECIST v1.1. The accompanying confidence interval would describe uncertainty around that proportion.

A p-value, if used, would address a prespecified hypothesis about the response proportion. It would not measure the size of the response effect. A small p-value and a large response percentage answer different statistical questions: one concerns evidence against a null hypothesis, while the other describes the magnitude of the observed response rate.

10. Safety Results

The registry reports serious adverse events by the specified safety categories. These counts should be interpreted according to their exposure definitions, particularly for Cohort C, where the record separates pre-crossover and post-crossover periods.

The affected/at-risk format provides a direct description of the reported serious adverse-event experience within each specified category. The categories should not be combined into a single overall rate because they represent different cohort and exposure definitions.

11. Crossover and Statistical Interpretation

Cohort C is explicitly represented in the safety record as having pre-crossover and post-crossover periods. Crossover introduces an important distinction between randomized assignment and treatment actually received over time.

For safety, an exposure-based description can be clinically useful because adverse events occur during periods of treatment exposure. For causal efficacy comparisons, however, the randomized treatment assignment remains a fundamental part of the trial design. The statistical meaning of a crossover analysis therefore depends on the exact estimand and analysis population being used.

12. Trial Timeline

13. What the Primary Endpoint Does — and Does Not — Mean

14. Statistical Interpretation of the Safety Counts

The reported serious adverse-event counts are most naturally read as affected participants relative to the participants at risk in each defined safety category.

These are descriptive safety quantities. A formal comparison would require more information about event definitions, timing, follow-up, exposure duration, and the prespecified statistical framework. The raw affected/at-risk counts alone do not establish a comparative treatment effect.

15. Randomization and Unmasked Assessment

Randomization is a major design feature because it establishes treatment assignment through a randomized allocation process rather than through investigator or participant selection. In a randomized study, this helps separate baseline differences caused by chance from systematic differences associated with treatment assignment.

The study is described as having no masking. That makes the independent blinded central review particularly relevant for the primary radiologic response endpoint. Treatment assignment was not concealed at the study level, while the endpoint assessment incorporated blinded independent central review.

16. Limits of Response-Rate Interpretation

• Population definition: the primary endpoint applies to pooled Cohorts A+B, so the relevant efficacy denominator must match that defined population.
• Confirmation: cORR requires confirmed CR or PR rather than a single unconfirmed favorable assessment.
• Radiologic classification: the response endpoint depends on applying RECIST v1.1 definitions consistently.
• Central review: the endpoint incorporates blinded independent central review, making the assessment process part of the endpoint's interpretation.
• Response is not survival: cORR does not describe the timing of death or progression.
• Response is not duration: the response percentage does not establish how long individual responses persist.
• Safety populations differ: the reported Cohort C safety categories are defined by pre- and post-crossover exposure periods rather than by two independent randomized groups.
• Statistical precision: a response percentage without an accompanying confidence interval does not fully communicate the uncertainty around the estimate.
• Unmasked study: the absence of masking makes the blinded independent central review component important for objective response assessment.

17. Why This Trial Matters Statistically

MOUNTAINEER is useful as a statistical teaching case because it combines randomized allocation, a three-arm parallel design, an unmasked study structure, centralized blinded response assessment, a pooled binary primary endpoint, and crossover-related safety reporting.

18. Clinical Interpretation vs Statistical Interpretation

This distinction prevents a response percentage from being interpreted as a complete summary of treatment benefit. Objective response is one component of clinical evidence, with its own endpoint definition and statistical interpretation.

19. Limitations and Interpretation Issues

20. A Statistical Walkthrough of the Primary Endpoint

Step 1: Define the analysis population

The registry specifies pooled Cohorts A+B. This establishes the population to which the primary cORR calculation applies.

Step 2: Classify each participant

Each participant is evaluated according to RECIST v1.1. A participant qualifies for the numerator when the assessment meets the confirmed CR or PR definition.

Step 3: Count confirmed responses

The number of participants with confirmed CR or PR becomes the numerator of the response-rate estimate.

Step 4: Identify the applicable denominator

The denominator is the applicable pooled Cohorts A+B efficacy population, not automatically the study's total enrollment of 117.

Step 5: Estimate cORR

Step 6: Quantify uncertainty

A binomial confidence interval can accompany the response estimate. The interval gives readers information about statistical precision that the point estimate alone does not provide.

Step 7: Interpret within the endpoint definition

The final estimate should be described as a confirmed objective response rate according to the specified RECIST v1.1 criteria and blinded independent central review. It should not be generalized automatically to other response definitions or other populations.

21. What a Complete Statistical Result Would Contain

A fully interpretable presentation of this primary endpoint would place several pieces of information together:

Keeping these elements together prevents a response percentage from being detached from the population and measurement framework that produced it.

22. Record Summary

MOUNTAINEER is a randomized phase 2, three-arm, parallel, unmasked study of tucatinib and trastuzumab in metastatic colorectal adenocarcinoma. The registry defines confirmed objective response rate in pooled Cohorts A+B as the primary endpoint, with response assessed according to RECIST v1.1 and blinded independent central review over a time frame of up to 46.6 months.

From a statistical perspective, the primary endpoint is a binary response proportion rather than a time-to-event measure. Its interpretation therefore depends on the precise efficacy population, confirmed response definition, and assessment method. Confidence intervals would provide the key measure of precision around the response estimate. The registry also reports serious adverse events in distinct exposure-defined categories, including separate pre- and post-crossover periods for Cohort C.

The combination of randomized design, pooled response analysis, centralized blinded assessment, and crossover-related safety reporting makes MOUNTAINEER a useful example of how endpoint definitions and analysis populations shape the interpretation of clinical-trial results.

23. Sources

• ClinicalTrials.gov: MOUNTAINEER — NCT03043313.
• PubMed: PubMed record.
• PubMed: PubMed record.
• PubMed: PubMed record.
• PubMed: PubMed record.