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
ANCHOR was a phase 2, single-group study evaluating encorafenib, binimetinib and cetuximab in subjects with previously untreated BRAF V600E-mutant metastatic colorectal cancer. The study enrolled 95 subjects and had one treatment arm.
| Feature | ANCHOR |
|---|---|
| Phase | Phase 2 |
| Population | BRAF V600E-mutant metastatic colorectal cancer |
| Design | Single-group, non-masked, treatment study |
| Allocation | NA |
| Number of arms | 1 |
| Primary endpoint | Confirmed Overall Response Rate (cORR) Based on Local Tumor Assessments |
| Enrollment | 95 |
| Lead sponsor | Pierre Fabre Medicament |
| Sponsor type | Industry |
| Study status | Completed |
| ClinicalTrials.gov | NCT03693170 |
2. Clinical Question
The primary statistical question was whether the confirmed overall response rate in subjects treated with encorafenib, binimetinib and cetuximab was sufficiently greater than a prespecified response-rate benchmark of 30%. Because the study had a single treatment arm, the analysis compares the observed response rate with a fixed null value rather than comparing two randomized treatment groups.
Population
Subjects with BRAF V600E-mutant metastatic colorectal cancer who were previously untreated.
Intervention
Encorafenib, binimetinib and cetuximab.
Comparator
There was no comparator arm. The study used a single-group design.
Primary question
Is the true confirmed overall response rate greater than 30% under the prespecified one-sided testing framework?
3. Trial Design
Triplet treatment
- Encorafenib
- Binimetinib
- Cetuximab
Statistical reference
- Null confirmed response rate: 30%
- One-sided alternative
- Exact Clopper-Pearson 95% confidence interval
The single-group structure determines what can and cannot be learned from the design. A response rate above the null benchmark can provide evidence against the prespecified null hypothesis, but the design does not directly estimate a randomized treatment-versus-control difference. In particular, there is no concurrent control group against which differences in response, survival, or safety can be attributed through randomization.
4. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Confirmed Overall Response Rate (cORR) Based on Local Tumor Assessments | The confirmed overall response rate (cORR) is the percentage of confirmed responses, defined as complete response (CR) or partial response (PR), as assessed by local radiologist/investigator review based on the Response Evaluation Criteria in Solid Tumors (RECIST 1.1). Per Response Evaluation Criteria In Solid Tumors Criteria (RECIST v1.0) for target lesions and assessed by MRI: Complete Response (CR): Disappearance of all target lesions; Partiel response (PR): At least a 30% decrease in the sum of diameters of target lesions, taking as reference the baseline sum diameters. Overall Response (OR) = CR + PR. | From initiation of treatment to disease progression up to a maximum of 17.6 months. |
The registry identifies one registered primary endpoint. No additional secondary endpoint results are included in the record information considered here.
5. Analysis Population
The primary efficacy analysis was defined for the Efficacy Set (ES). This population is composed of all included subjects having received at least one dose of study treatment, partial or full, with a centrally confirmed BRAFV600E mutation.
| Population | Definition / role |
|---|---|
| Efficacy Set (ES) | All included subjects having received at least one dose of study treatment, partial or full, with a centrally confirmed BRAFV600E mutation. |
| Enrollment | 95 subjects |
| Number of treatment arms | 1 |
This distinction is statistically important. The enrollment count and the efficacy-analysis population are not automatically interchangeable: the prespecified analysis population is defined by treatment exposure and centrally confirmed mutation status. The registry's statistical description also specifies a decision rule based on 90 treated subjects with a centrally confirmed BRAFV600E mutation, with a provision for using the lower limit of the Clopper-Pearson interval if more than 90 subjects were enrolled.
6. Statistical Methodology
One-sample binomial framework
The primary endpoint is a proportion: the percentage of subjects with a confirmed response. In a single-group trial, the natural inferential question is whether the underlying response probability is compatible with a prespecified null value. The registry specifies a null response rate of 30% and a one-sided alternative in which the true response rate is greater than 30%.
Here, p denotes the true confirmed overall response probability in the efficacy population. The hypothesis is about the underlying response probability, not about an individual patient's probability of response.
Exact Clopper-Pearson confidence interval
The registry specifies a Clopper-Pearson (exact) binomial 95% confidence interval for cORR in the Efficacy Set. Unlike a normal approximation to a binomial proportion, the exact method uses the binomial distribution directly and is particularly natural when the endpoint is a count of responses among treated subjects.
where x is the number of confirmed responses and n is the number of subjects in the efficacy analysis population.
Prespecified decision rule
The registry states that if 37 or more confirmed responses were observed in the 90 treated subjects with a centrally confirmed BRAFV600E mutation, corresponding to a lower limit of Clopper-Pearson (exact) binomial 95% CI exceeding 30%, the study was considered to have met its primary endpoint. If more than 90 subjects were enrolled, the lower limit of Clopper-Pearson was to be used for decision. with a centrally confirmed BRAFV600E mutation, corresponding to a lower limit of the Clopper-Pearson exact binomial 95% confidence interval exceeding 30%, the study was considered to have met its primary endpoint.
Primary decision threshold
Among 90 treated subjects with a centrally confirmed BRAFV600E mutation
The decision criterion was a Clopper-Pearson 95% CI lower limit exceeding 30%.
This is an important feature of the design because the response count is not being interpreted simply by asking whether the observed percentage happens to exceed 30%. The uncertainty around the observed proportion is incorporated into the decision through the lower confidence bound.
Type I error and power
The registry states that the design yields a 1-sided type I error rate equal to 1.6% and power of 80% when the true response rate is 45%.
1-sided type I error: 1.6%
This describes the design's specified probability of rejecting the null hypothesis when the true response rate is at the null value, under the stated one-sided framework.
Power: 80%
This is the design probability of satisfying the prespecified decision criterion when the true response rate is 45%, under the stated design assumptions.
Handling enrollment above 90
The registry specifies that if more than 90 subjects were enrolled, the lower limit of the Clopper-Pearson interval was to be used for the decision. This preserves the confidence-bound approach while allowing the analysis to incorporate an efficacy population larger than the original 90-subject decision framework.
7. Results
The registry identifies results as posted and contains a posted statistical analysis for the primary endpoint. That analysis specifies the null hypothesis, exact confidence-interval method, type I error, power, and response-count decision rule. The posted statistical-analysis entry does not provide an observed cORR estimate, an observed confirmed-response count, a confidence-interval value, or a p-value.
| Primary endpoint | Posted statistical information |
|---|---|
| Confirmed Overall Response Rate (cORR) Based on Local Tumor Assessments | One-sided test of a 30% null response rate; Clopper-Pearson exact binomial 95% CI; 1-sided type I error 1.6%; 80% power when true response rate is 45%; primary endpoint criterion based on the lower CI limit exceeding 30%. |
| Observed cORR | Not reported in the posted statistical-analysis entry. |
| Observed response count | Not reported in the posted statistical-analysis entry. |
| 95% CI | Not reported in the posted statistical-analysis entry. |
| p-value | Not reported in the posted statistical-analysis entry. |
The central result available from the registry's posted analysis is the prespecified inferential rule: evidence against a 30% response-rate null was to be established when the lower limit of the exact 95% confidence interval exceeded 30%, with the specified response-count criterion of 37 or more among 90 treated subjects with a centrally confirmed BRAFV600E mutation.
That rule should not be confused with an observed treatment effect. A response-rate estimate would describe the proportion of subjects who achieved a confirmed response; it would not establish how much better the regimen performs than a randomized comparator because this trial has no comparator arm.
The 95% confidence interval is intended to describe statistical precision around the estimated response probability. It is not a range containing the response probability of individual patients. Likewise, a p-value, when reported, would quantify evidence against the null hypothesis under the specified testing framework; it would not measure the magnitude or clinical importance of the response rate.
The analysis also depends on the definition of a confirmed response and on the efficacy population. Because response is assessed over a period extending from treatment initiation to disease progression up to a maximum of 17.6 months, the endpoint is not simply a snapshot of tumor status at one fixed calendar date.
8. Statistical Methods Explained
Why use an exact binomial analysis?
cORR is a binary-outcome proportion: each subject either meets the confirmed-response definition or does not. With a single treatment group, the resulting response count can be modeled with a binomial distribution. The exact method evaluates that distribution directly rather than relying on an approximate normal distribution.
Why was the Clopper-Pearson method used?
The Clopper-Pearson method provides an exact binomial confidence interval. For a proportion, the width of the interval reflects the amount of information available from the number of subjects and the observed response count. The method is therefore directly aligned with the binary structure of the primary endpoint.
Why is the test one-sided?
The registry specifies a one-sided alternative because the design question is whether the true response rate exceeds the prespecified 30% benchmark. The alternative direction is therefore greater than the null value rather than simply different from it.
Why does the lower confidence limit determine success?
A response rate above 30% by itself does not establish that the underlying response probability is convincingly above 30%. The lower confidence limit incorporates sampling uncertainty. If that lower limit exceeds 30%, the complete confidence interval lies above the null benchmark under the stated confidence framework.
What does 80% power at a true response rate of 45% mean?
Power is a property of the trial design under a specified assumed truth. The registry's 80% figure means that, if the true response rate were 45% and the stated design assumptions held, the probability of satisfying the prespecified primary-endpoint criterion would be 80%.
Why does the absence of a comparator matter?
A single-group study can test whether an observed response rate exceeds a predefined benchmark, but it does not provide randomized contemporaneous control data. Historical benchmarks can differ because of patient selection, diagnostic methods, treatment availability, follow-up, and other features of the clinical setting. Consequently, the statistical evidence addresses the prespecified benchmark rather than a randomized treatment contrast.
9. Understanding the Primary Endpoint
The registry defines overall response as complete response plus partial response. Complete response requires disappearance of all target lesions; partial response requires at least a 30% decrease in the sum of diameters of target lesions, using the baseline sum of diameters as the reference.
The word confirmed is important. The primary endpoint is not simply any initial tumor shrinkage. The registry describes cORR as the percentage of confirmed responses assessed by local radiologist/investigator review using RECIST 1.1.
This makes cORR a clinically interpretable binary endpoint while still requiring an explicit assessment framework. A subject's classification depends on whether the observed tumor measurements satisfy the response criteria and the response is confirmed under the study's assessment process.
10. Statistical Interpretation of the Decision Rule
Suppose a single-group trial observes a response proportion above the null value. That observation alone does not establish statistical evidence against the null. The key issue is how much uncertainty surrounds the proportion. The ANCHOR design addresses that uncertainty through an exact 95% confidence interval.
A confidence interval for cORR describes uncertainty about the underlying population response probability under the statistical framework. It does not mean that individual patients have response probabilities restricted to the endpoints of that interval.
A p-value is a measure of evidence against a specified null hypothesis under a statistical model. It is not a measure of how large the response rate is, how clinically important the response is, or how many patients personally benefit.
The 30% benchmark is a fixed reference value in the statistical design. It should not be described as though it were a concurrent control-arm response rate. The distinction is fundamental to interpreting a single-group phase 2 study.
11. Safety Results
The registry reports serious adverse events by arm for the single treatment arm. The affected-to-at-risk count was 49/95.
| Safety measure | Single treatment arm |
|---|---|
| Serious adverse events | 49/95 |
Because there is no comparator arm, the serious-adverse-event count describes the observed experience within the treatment group rather than a randomized between-group safety difference. It should therefore be interpreted separately from the efficacy hypothesis concerning cORR.
12. Limitations and Interpretation Issues
- No comparator arm: the absence of a comparator arm is an explicit limitation of the study. The primary analysis tests the response rate against a fixed 30% benchmark rather than estimating a randomized treatment effect.
- Short follow-up at the data cutoff: the registry states that the short duration of follow-up at the data cut-off date does not allow a robust estimate of OS results.
- Response endpoint: cORR is a tumor-response measure and does not by itself establish an effect on overall survival.
- Analysis population: the primary analysis is defined for the Efficacy Set, which requires at least one dose of treatment and a centrally confirmed BRAFV600E mutation.
- Single-group inference: statistical evidence against the 30% benchmark does not remove the limitations inherent in comparing a single treatment group with a fixed reference value.
- Assessment framework: cORR depends on local radiologist/investigator assessment under the specified RECIST framework, including the definitions of complete and partial response.
- Confidence-interval interpretation: the exact 95% interval reflects uncertainty around the response probability; it is not an interval for individual treatment effects.
13. Why This Trial Matters Statistically
ANCHOR is a useful teaching case because its primary analysis illustrates a different statistical structure from the randomized two-arm trials that dominate clinical-trial examples. The central inferential task is not estimation of a hazard ratio or a treatment difference. Instead, it is determining whether a single observed response proportion provides sufficient evidence that the underlying response rate exceeds a prespecified benchmark.
| Concept | How it appears in ANCHOR |
|---|---|
| Single-group design | One treatment arm with 95 enrolled subjects |
| Binary endpoint | Confirmed response versus no confirmed response |
| Null benchmark | 30% true response rate |
| One-sided testing | Alternative tests whether the true response rate exceeds 30% |
| Exact binomial inference | Clopper-Pearson exact 95% confidence interval |
| Decision boundary | Lower confidence limit exceeding 30% |
| Response-count rule | 37 or more confirmed responses among 90 treated subjects with centrally confirmed BRAFV600E mutation |
| Type I error | 1-sided type I error rate equal to 1.6% |
| Power | 80% when the true response rate is 45% |
| Analysis population | Efficacy Set defined by treatment exposure and centrally confirmed mutation |
| Safety analysis | Serious adverse events reported as affected / at-risk within the single treatment arm |
The trial therefore provides a compact example of how an endpoint definition, analysis population, hypothesis, confidence interval, and decision threshold fit together. The statistical design is especially instructive because the response-rate benchmark is part of the hypothesis itself: it is not a control-group estimate generated by randomization.
14. Planned Analysis
The registered primary analysis measures confirmed overall response rate from initiation of treatment to disease progression up to a maximum of 17.6 months. Confirmed responses are defined as complete response or partial response under the registry's RECIST-based assessment framework.
Step 1: Define the response
Classify each subject according to whether a confirmed CR or PR is documented under the local tumor-assessment framework.
Step 2: Estimate cORR
Calculate the proportion of confirmed responses in the Efficacy Set.
Step 3: Quantify uncertainty
Construct the Clopper-Pearson exact binomial 95% confidence interval.
Step 4: Apply the decision rule
Determine whether the lower confidence limit exceeds the 30% null benchmark.
The registry's design specifies that 37 or more confirmed responses among 90 treated subjects with a centrally confirmed BRAFV600E mutation correspond to a lower Clopper-Pearson confidence limit exceeding 30%. If more than 90 subjects were enrolled, the registry specifies using the lower limit of the Clopper-Pearson interval for the decision.
15. Trial Timeline
Study start
The ANCHOR study began enrollment.
Primary completion
The registry lists 2020-06-29 as the primary completion date.
Study status
The registry lists the study status as completed.
16. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The study was designed to test whether the true confirmed response rate exceeded 30%, using a one-sided exact binomial framework and a Clopper-Pearson 95% confidence interval.
Clinical interpretation
cORR describes the proportion of subjects achieving a confirmed tumor response under the specified assessment criteria. It does not, by itself, establish comparative treatment benefit or provide a robust estimate of overall survival.
17. What the Exact Confidence Interval Does — and Does Not — Mean
The Clopper-Pearson 95% confidence interval quantifies uncertainty around the estimated underlying response probability using the exact binomial framework. The interval becomes informative only when paired with the observed number of responses and the analysis population.
For the ANCHOR primary hypothesis, the relevant question is directional: is the true response probability greater than 30%? A lower confidence limit above 30% provides a direct confidence-bound expression of that question.
An exact confidence interval above a benchmark would not by itself establish a randomized treatment effect, because the study contains no concurrent comparator group. It also would not establish that every patient benefits or that response translates directly into longer overall survival.
18. Statistical Concepts in This Trial
Learn more about the methods used in this trial:
19. Related Calculators
20. Sources
- ClinicalTrials.gov: NCT03693170.
- PubMed: PubMed record for the ANCHOR publication.
Continue through the Clinical Biostats knowledge graph
Explore the statistical methods behind clinical-trial endpoints and connect them to related tutorials and calculators.
21. Record Summary
ANCHOR illustrates the statistical logic of a single-group phase 2 response-rate study. Its primary endpoint is confirmed overall response rate based on local tumor assessments, with response defined through complete or partial response under the registry's RECIST framework. The prespecified analysis uses a 30% null response rate, a one-sided alternative, and a Clopper-Pearson exact binomial 95% confidence interval. The design specifies a 1-sided type I error rate of 1.6%, 80% power when the true response rate is 45%, and a decision criterion based on the lower confidence limit exceeding 30%, with 37 or more confirmed responses among 90 treated subjects with a centrally confirmed BRAFV600E mutation corresponding to that criterion.
The statistical interpretation must remain tied to the design: this is evidence relative to a fixed response-rate benchmark, not a randomized comparison against another treatment. The registry also identifies the absence of a comparator arm and the short follow-up at the data cut-off date as limitations, including for robust estimation of overall survival. Safety is reported within the single treatment arm, with serious adverse events affecting 49 of 95 subjects.