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
C-TRAK TN was a randomized, double-masked, parallel phase 2 treatment trial enrolling patients with triple-negative breast cancer. The study evaluated blood-based detection of circulating tumor DNA (ctDNA) after standard treatment and included pembrolizumab as the recorded intervention.
| Feature | C-TRAK TN |
|---|---|
| Phase | Phase 2 |
| Condition | Triple Negative Breast Cancer |
| Population described by the study title | Early-stage triple-negative breast cancer patients |
| Design | Randomized, double-masked, parallel |
| Primary purpose | Treatment |
| Enrollment | 208 |
| Number of arms | 2 |
| Intervention recorded in the registry | Pembrolizumab |
| Lead sponsor | Institute of Cancer Research, United Kingdom |
| Study start | 2017-12-21 |
| Primary completion | 2021-09-30 |
| Registry status | UNKNOWN |
| ClinicalTrials.gov | NCT03145961 |
2. Clinical Question
The trial addresses a treatment-monitoring question in early-stage triple-negative breast cancer: can a blood sample collected after standard treatment identify patients with detectable ctDNA, and can that information be incorporated into additional treatment with pembrolizumab?
Population
Patients with early-stage triple-negative breast cancer, as described by the study title.
Intervention
Pembrolizumab is the intervention recorded in the registry.
Comparator
The study has two randomized arms, but the treatment assigned to the other arm is not specified in the registry information reported here.
Primary question
How frequently is ctDNA detected at prespecified time points, and among patients commencing pembrolizumab, how frequently are both detectable ctDNA and disease recurrence absent at the specified follow-up time?
3. Trial Design
The design is important statistically because randomization creates the framework for comparing outcomes between the two arms. The parallel structure means patients remain associated with their randomized treatment group rather than being intentionally moved through multiple randomized treatment sequences.
4. Trial Timeline
Study start
The C-TRAK TN study began on 2017-12-21.
First primary ctDNA assessment
The registry specifies positive ctDNA detection by 12 months, assessed using the blood sample taken at that timepoint.
Second primary ctDNA assessment
The registry specifies positive ctDNA detection by 24 months, again assessed using the blood sample taken at that timepoint.
Primary completion date
The registry records 2021-09-30 as the primary completion date.
5. Primary Endpoints
| Primary endpoint | Time frame | Registry definition |
|---|---|---|
| Positive ctDNA detection by 12 months | 12 months | The proportion of patients with ctDNA positivity by 12 months as assessed by the blood sample taken at that timepoint. |
| Positive ctDNA detection by 24 months | 24 months | The proportion of patients with ctDNA positivity by 24 months as assessed by the blood sample taken at that timepoint. |
| Absence of detectable ctDNA or disease recurrence 6 months (24 weeks) after commencing pembrolizumab | 6 months (24 weeks) after commencing pembrolizumab | The proportion of patients without either detectable ctDNA or disease recurrence 6 months (24 weeks) after starting pembrolizumab. |
All three registered primary endpoints are fundamentally proportion-based rather than time-to-event endpoints. That distinction drives the statistical analysis. The first two endpoints ask what proportion of patients have positive ctDNA at specified assessment times. The third asks what proportion satisfy a combined absence criterion at a specified interval after pembrolizumab begins.
6. Understanding the Endpoint Structure
Positive ctDNA detection is a binary endpoint
For the 12-month and 24-month endpoints, each patient contributes an outcome that can be classified according to whether ctDNA is positive at the specified assessment. The resulting endpoint is naturally summarized as a proportion:
where x is the number of patients meeting the endpoint definition and n is the number of patients evaluated for that endpoint.
The denominator is particularly important. A proportion is meaningful only when the analysis population is clearly defined. Patients without an evaluable blood sample may not be interchangeable with patients who provide a definitive negative result. Missing assessments therefore have the potential to affect both the numerator and denominator.
The third endpoint combines two clinical states
The endpoint assessed 6 months (24 weeks) after commencing pembrolizumab is defined by the absence of both detectable ctDNA and disease recurrence. Statistically, this is a conjunctive endpoint: a patient must satisfy the complete endpoint definition rather than merely one component.
ctDNA component
The patient must have no detectable ctDNA at the specified assessment.
Disease component
The patient must have no disease recurrence at the specified assessment.
Combined endpoint
The registered endpoint requires absence of either detectable ctDNA or disease recurrence, using the registry's exact wording.
Interpretation
The resulting proportion represents patients satisfying the registered composite absence criterion at that timepoint.
7. Statistical Methodology
Analysis of proportions
The primary endpoints are proportions measured at fixed follow-up timepoints. A natural statistical description is therefore the number and percentage of patients meeting the endpoint definition, together with an interval estimate for the proportion.
For a single proportion, a binomial confidence interval is generally appropriate. The choice between exact and approximate interval methods depends on the prespecified statistical analysis plan and the observed number of events or successes.
Here, X represents the number of patients meeting the binary endpoint among n evaluable patients, while p is the underlying probability of meeting that endpoint under the statistical model.
Between-arm comparison
Because the study is randomized with two parallel arms, an endpoint proportion can also be compared between treatment groups. Common approaches include a risk difference, risk ratio, or odds ratio, each of which answers a different question.
| Measure | What it describes |
|---|---|
| Risk difference | The arithmetic difference between the two endpoint proportions. |
| Risk ratio | The ratio of the endpoint probability in one arm to that in the other arm. |
| Odds ratio | The ratio of the odds of meeting the endpoint between the two arms. |
The registry information does not identify a specific comparative estimator, confidence-interval method, or hypothesis test for these endpoints. Those choices should therefore be distinguished from the general statistical methods commonly used for randomized binary outcomes.
Fixed-time analysis versus survival analysis
The 12-month and 24-month ctDNA endpoints are defined at specific assessment times. They are therefore not automatically equivalent to a Kaplan-Meier time-to-event endpoint. A patient either meets the registered ctDNA status at the specified blood-sample timepoint or does not, subject to the endpoint's assessment rules.
This distinction matters because survival analysis handles the timing of events and censoring, whereas a fixed-time binary analysis focuses on status at a particular assessment point. Using a time-to-event model without a corresponding endpoint definition would change the statistical question.
8. Planned Analysis
The ClinicalTrials.gov record contains registered primary endpoints but no posted statistical analyses. Results have not been posted on ClinicalTrials.gov.
Positive ctDNA detection by 12 months
The registry defines this endpoint as the proportion of patients with ctDNA positivity by 12 months, assessed from the blood sample taken at that timepoint. A typical analysis would summarize the number and proportion of positive assessments in each randomized arm and provide an appropriate confidence interval. If a formal between-arm hypothesis test is prespecified, the selected binary-outcome method would compare the two randomized groups.
Positive ctDNA detection by 24 months
The same general framework applies at 24 months, but the assessment time is different. The analysis should preserve the registered definition: the outcome is based on the blood sample taken at 24 months rather than an inferred trajectory of ctDNA between assessment points.
Absence of detectable ctDNA or disease recurrence 6 months after commencing pembrolizumab
This endpoint should be treated as a binary outcome defined at 6 months (24 weeks) after starting pembrolizumab. A typical analysis would report the proportion meeting the complete endpoint definition, with uncertainty quantified using an appropriate binomial confidence interval and, where prespecified, a comparison between randomized groups.
9. Statistical Methods Explained
Why are these endpoints analysed as proportions?
The first two primary endpoints explicitly ask for the proportion of patients with positive ctDNA at 12 months and 24 months. The third explicitly asks for the proportion without detectable ctDNA or disease recurrence at a specified time after commencing pembrolizumab. Each therefore has a binary outcome structure.
Why does the denominator matter?
A proportion depends on both the number meeting the endpoint and the population used as the denominator. If some patients do not have an evaluable assessment, the resulting estimate can depend materially on how those observations are handled. A prespecified missing-data rule is therefore important for a rigorous analysis.
What does a confidence interval around a proportion tell us?
A confidence interval describes statistical uncertainty around the estimated population proportion under the selected model and sampling framework. It does not describe the range of outcomes that individual patients can experience, nor does it mean that a specified percentage of future patients must fall inside the interval.
Why is a risk difference different from a risk ratio?
A risk difference describes an absolute separation between two proportions, whereas a risk ratio describes their relative relationship. The same pair of observed proportions can therefore produce a small absolute difference and a substantially different relative measure.
Why is a fixed-time endpoint different from a Kaplan-Meier endpoint?
A fixed-time endpoint asks about patient status at a specified assessment point. Kaplan-Meier analysis instead estimates the distribution of time until an event while accounting for censoring. These methods answer related but different statistical questions.
Why is the third endpoint more restrictive than either component alone?
The registered definition requires absence of detectable ctDNA or disease recurrence at the specified follow-up assessment. A patient satisfying only one component does not necessarily satisfy the complete endpoint definition. Combined endpoints should therefore be interpreted according to their exact logical definition rather than as interchangeable measures of either component.
10. Missing Data and Assessment Timing
Fixed-time biomarker endpoints create an important missing-data issue. The registry defines the 12-month and 24-month endpoints using the blood sample taken at the corresponding timepoint. Consequently, an unavailable or unevaluable sample is analytically different from a documented negative ctDNA result.
Observed negative
A patient has an evaluable blood sample and the ctDNA assessment is negative.
Observed positive
A patient has an evaluable blood sample and the ctDNA assessment is positive.
Missing assessment
The required assessment is unavailable or cannot be classified from the available information.
Statistical consequence
The treatment-group proportion can change depending on the prespecified rule for handling patients without an evaluable assessment.
The registry information does not specify a missing-data or imputation method. In a formal statistical analysis, this would be an important component of the analysis plan because assumptions about missing assessments can influence a fixed-time proportion.
11. Randomization and the Causal Question
Randomization is the central design feature that allows a comparison between the two arms to be interpreted as a treatment comparison, subject to adherence, missing data, post-randomization events, and other features of the analysis.
| Design feature | Statistical role |
|---|---|
| Randomized allocation | Creates treatment groups whose differences can be evaluated within a randomized comparison. |
| Two arms | Provides a direct parallel comparison between randomized groups. |
| Parallel design | Patients remain in their randomized treatment pathway rather than contributing observations under multiple randomized sequences. |
| Double masking | Reduces the opportunity for knowledge of treatment assignment to influence treatment delivery or outcome assessment. |
| Phase 2 | Places the study within an earlier clinical-development stage rather than a confirmatory phase 3 framework. |
Randomization does not eliminate all statistical uncertainty. It establishes the design basis for comparison; the resulting estimates still depend on sample size, endpoint ascertainment, missing observations, and the prespecified analysis population.
12. Interpreting ctDNA as a Statistical Endpoint
ctDNA detection is different from a conventional clinical outcome such as overall survival. It is a biomarker measurement collected at defined timepoints. The statistical question is therefore whether the probability of detectable ctDNA differs between groups or changes according to the treatment-monitoring strategy specified by the trial.
A positive ctDNA result is also not itself a measure of treatment effect. It is the endpoint status defined by the registry. The treatment comparison comes from examining how that endpoint is distributed across randomized groups.
A difference in the proportion of patients with positive ctDNA would quantify a difference in the frequency of the registered biomarker state between treatment groups at the specified assessment time. It would not, by itself, establish the magnitude of an effect on survival or another clinical outcome not included in the registered primary endpoint definition.
The 12-month and 24-month endpoints should be interpreted separately because they correspond to different assessment times. A 24-month proportion is not simply a longer version of the 12-month proportion; it represents the ctDNA status observed at a later specified timepoint.
The 6-month (24-week) endpoint after commencing pembrolizumab incorporates both ctDNA status and disease recurrence. Its interpretation therefore depends on both components and on the exact population assessed at that treatment-specific timepoint.
13. Statistical Interpretation of the Study Design
The C-TRAK TN design illustrates a common distinction in modern oncology trials: a biomarker can be used not merely as a baseline characteristic but as a mechanism for identifying patients after initial treatment and informing subsequent therapy.
From a statistical perspective, this creates several layers of analysis. The first layer is the randomized comparison. The second is measurement of ctDNA status at prespecified timepoints. The third is assessment of a combined biomarker-and-disease status after pembrolizumab begins. Each layer must retain the endpoint's precise definition.
| Statistical layer | Question |
|---|---|
| Randomized treatment comparison | How do the two randomized arms differ on the registered endpoint? |
| 12-month ctDNA assessment | What proportion has positive ctDNA at 12 months? |
| 24-month ctDNA assessment | What proportion has positive ctDNA at 24 months? |
| Pembrolizumab follow-up endpoint | What proportion has neither detectable ctDNA nor disease recurrence 6 months (24 weeks) after commencing pembrolizumab? |
14. Multiplicity and Multiple Primary Endpoints
The registry identifies three primary endpoints. That creates a multiplicity issue that is important whenever several endpoints are treated as primary questions.
If each endpoint is subjected to a separate hypothesis test at the same nominal significance level without adjustment, the probability of obtaining at least one statistically significant result under a global null hypothesis can exceed the nominal level. A formal confirmatory analysis therefore normally requires a prespecified multiplicity strategy when multiple primary hypotheses are being tested.
Multiple endpoints
The study has three registered primary endpoints rather than a single primary outcome.
Type I error
Multiple confirmatory tests can require control of the familywise probability of a false-positive conclusion.
Endpoint hierarchy
A prespecified hierarchy can determine whether later hypotheses are tested conditionally on earlier results.
Registry information
The available registry information does not specify a multiplicity-adjustment procedure or endpoint-testing hierarchy.
The existence of three primary endpoints does not by itself determine how they should be tested. The statistical analysis plan would need to specify whether they are co-primary, ordered, or otherwise controlled within a formal testing strategy.
15. Safety and Adverse Events
The registry information identifies the trial as a treatment study and records pembrolizumab as the intervention, but it does not provide serious adverse-event counts by randomized arm in the information reported here.
For a randomized treatment trial, safety is generally summarized separately from the efficacy endpoints. Typical safety summaries include the number and proportion of patients experiencing adverse events, serious adverse events, treatment discontinuations, and clinically important event categories. Those measures require arm-level event data and explicit definitions.
16. What This Trial Does Not Establish From the Registered Endpoints Alone
The registered primary endpoints are focused on ctDNA detection and the absence of detectable ctDNA or disease recurrence at a specified follow-up point. They do not, by their definitions alone, provide a direct estimate of overall survival, progression-free survival, or another unregistered time-to-event outcome.
| Question | Relationship to the registered primary endpoints |
|---|---|
| Is ctDNA detectable at 12 months? | Directly addressed by a registered primary endpoint. |
| Is ctDNA detectable at 24 months? | Directly addressed by a registered primary endpoint. |
| Is detectable ctDNA or disease recurrence absent 6 months (24 weeks) after commencing pembrolizumab? | Directly addressed by a registered primary endpoint. |
| What is overall survival? | Not defined among the registered primary endpoints reported here. |
| What is progression-free survival? | Not defined among the registered primary endpoints reported here. |
This distinction is important because biomarker endpoints and clinical time-to-event outcomes can be related without being statistically interchangeable. A rigorous interpretation should preserve the endpoint actually registered.
17. Important Limitations and Interpretation Issues
- No posted statistical analyses: The ClinicalTrials.gov record contains registered endpoints but no posted statistical analyses, so numerical efficacy estimates, confidence intervals, and p-values are not available from the registry record.
- Missing assessment information: Fixed-time ctDNA endpoints depend on the availability and evaluability of blood samples at the specified timepoints.
- Endpoint definition: The third primary endpoint combines ctDNA status with disease recurrence and therefore requires careful interpretation of both components.
- Multiple primary endpoints: Three primary endpoints create a need to understand the prespecified multiplicity strategy before interpreting a collection of hypothesis tests as confirmatory.
- Comparator specification: The registry information identifies two randomized arms and pembrolizumab as the intervention but does not specify the treatment assigned to the second arm in the information reported here.
- Fixed-time interpretation: The 12-month and 24-month endpoints describe status at specified assessment times and should not automatically be interpreted as time-to-event measures.
- Analysis population: The available registry information does not specify the exact analysis population or missing-data rules for the primary endpoints.
- Safety information: Arm-level serious adverse-event counts are not reported in the available registry information.
18. Why This Trial Matters Statistically
C-TRAK TN is a useful statistical teaching case because it moves beyond a conventional endpoint measured once at the end of a study. Its primary endpoints require careful handling of biomarker status, fixed follow-up times, treatment-specific follow-up, and a combined clinical definition.
| Concept | How it appears in C-TRAK TN |
|---|---|
| Randomization | Patients were assigned through a randomized two-arm design. |
| Parallel design | The study used a parallel design. |
| Double masking | The registry identifies the trial as double-masked. |
| Binary endpoints | All three registered primary endpoints are expressed as proportions of patients meeting defined conditions. |
| Fixed-time assessment | ctDNA endpoints are assessed at 12 months and 24 months. |
| Composite condition | The third endpoint combines absence of detectable ctDNA with absence of disease recurrence. |
| Missing data | Availability of blood samples can affect the denominator for fixed-time biomarker analyses. |
| Multiplicity | Three registered primary endpoints require attention to the overall testing framework. |
| Analysis population | The interpretation of proportions depends on which patients are included as evaluable at each timepoint. |
19. Statistical Concepts in This Trial
The statistical structure of C-TRAK TN provides a useful framework for understanding how randomized trials analyse fixed-time binary outcomes and biomarker-based endpoints.
Proportions
The primary endpoints are expressed as the proportion of patients meeting specified binary conditions.
Binomial uncertainty
A binomial framework provides the basic statistical model for a binary outcome observed in a defined analysis population.
Risk differences
An absolute between-arm difference can describe how far apart two endpoint proportions are.
Risk ratios
A relative comparison can describe the endpoint probability in one randomized group relative to the other.
20. Sources
- ClinicalTrials.gov: C-TRAK TN, NCT03145961.
- PubMed: PubMed record associated with C-TRAK TN.
Continue through the Clinical Biostats knowledge graph
Explore statistical methods, clinical-trial concepts, and analysis tools through the broader Clinical Biostats library.
21. Record Summary
C-TRAK TN is a randomized, double-masked, parallel phase 2 treatment trial enrolling 208 patients with triple-negative breast cancer. Its three registered primary endpoints are all proportion-based: positive ctDNA detection by 12 months, positive ctDNA detection by 24 months, and absence of detectable ctDNA or disease recurrence 6 months (24 weeks) after commencing pembrolizumab.
The statistical structure is therefore centered on binary outcome analysis rather than conventional survival analysis. The key analytical considerations are the definition of the evaluable population, the handling of missing blood-sample assessments, the distinction between a fixed-time proportion and a time-to-event measure, the interpretation of the combined third endpoint, and the multiplicity implications of having three registered primary endpoints.