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
DESTINY-CRC02 was a randomized, double-blind, parallel phase 2 trial evaluating two dose levels of DS-8201a, also identified in the registry as T-DXd, in participants with HER2-overexpressing advanced or metastatic colorectal cancer. The trial enrolled 122 participants into two treatment arms.
| Feature | DESTINY-CRC02 |
|---|---|
| Phase | Phase 2 |
| Condition | Advanced Colorectal Cancer |
| Trial title | Trastuzumab Deruxtecan in Participants With HER2-overexpressing Advanced or Metastatic Colorectal Cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 122 |
| Number of arms | 2 |
| Trial status | Completed |
| Start date | March 5, 2021 |
| Primary completion date | November 1, 2022 |
| Lead sponsor | Daiichi Sankyo |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT04744831 |
2. Clinical Question
The registry defines the primary efficacy question around the objective response rate following intravenous administration of trastuzumab deruxtecan in participants with human epidermal growth factor receptor 2-overexpressing metastatic colorectal cancer. The trial compares two dose levels rather than a drug-versus-placebo treatment strategy.
Population
Participants with HER2-overexpressing advanced or metastatic colorectal cancer.
Intervention
DS-8201a 5.4 mg/kg Q3W or DS-8201a 6.4 mg/kg Q3W.
Comparator
The other randomized T-DXd dose level serves as the parallel treatment comparison.
Primary question
What percentage of participants have a confirmed objective response during the registry-defined assessment period following T-DXd administration?
3. Trial Design
Lower dose level
- DS-8201a 5.4 mg/kg
- Intravenous administration
- Q3W administration schedule
- Primary efficacy assessment based on blinded independent central review
Higher dose level
- DS-8201a 6.4 mg/kg
- Intravenous administration
- Q3W administration schedule
- Primary efficacy assessment based on blinded independent central review
The parallel design means that participants are assigned to one of two treatment groups and the groups are evaluated concurrently. Because both groups receive T-DXd at different dose levels, the statistical comparison is fundamentally a dose-level comparison, not a comparison between active treatment and an untreated control group.
4. Endpoints
| Endpoint | Registry definition and assessment | Time frame |
|---|---|---|
| Objective response rate (ORR) | Percentage of participants with a confirmed objective response, defined as the number (percentage) of participants with complete response (CR) or partial response (PR), assessed by blinded independent central review (BICR) based on Response Evaluation Criteria In Solid Tumors (RECIST) version 1.1. | 6 months post-dose administration to data cut off, up to 20 months |
What counts as a response?
The registry defines complete response as disappearance of all target lesions. It defines partial response as at least a 30% decrease in the sum of diameters of target lesions.
The endpoint is therefore binary at the participant level for purposes of calculating ORR: a participant either satisfies the registry's confirmed CR or PR response definition during the specified assessment framework, or does not. The published percentage is then an aggregate summary of those participant-level classifications.
5. Statistical Methodology
Objective response rate as a proportion
For a binary response endpoint, the basic estimand is the proportion of participants meeting the response definition. If \(R\) participants respond among \(N\) evaluated participants, the observed response proportion is conceptually
The percentage presentation is the same proportion expressed on a percentage scale. The registry's definition specifically requires confirmed complete response or partial response according to BICR and RECIST 1.1.
A confidence interval around a response proportion quantifies statistical uncertainty associated with estimating an underlying response probability from the observed sample. For a phase 2 trial, this precision is especially important because an observed response percentage is itself an estimate rather than a guarantee of the response probability in another population.
Blinded independent central review
The primary endpoint was assessed by blinded independent central review. Blinding the reviewers to treatment assignment helps reduce the possibility that knowledge of dose assignment influences radiologic response classification.
Central review also establishes a common assessment framework across participants. Because ORR depends directly on whether lesions satisfy prespecified response criteria, consistency in image interpretation is important to the statistical validity of the endpoint.
RECIST version 1.1
RECIST 1.1 provides the response framework used for the primary endpoint. The registry specifies the thresholds defining CR and PR, making the response classification more reproducible than an unrestricted clinical judgment about whether a tumor has "improved."
The resulting participant-level classifications form the numerator and denominator used to summarize ORR.
Comparing two response proportions
When two randomized groups are compared on a binary endpoint, a statistical analysis can compare their response proportions and quantify uncertainty around the difference or another prespecified contrast. The registry information identifies ORR as the primary endpoint and identifies the BICR/RECIST assessment framework, but it does not specify a posted formal inferential analysis for the endpoint.
That distinction matters. The observed percentage of responders and an inferential comparison are different statistical objects: the first describes the sample, while the second addresses uncertainty about the underlying treatment-group response probabilities.
6. Planned Analysis
The registry reports that results have been posted for DESTINY-CRC02, but no formal statistical analyses are posted to ClinicalTrials.gov for the primary endpoint. The registry does not provide a posted ORR estimate, confidence interval, or p-value in the record summarized here.
For an endpoint of this type, the central analysis would normally begin with the number and percentage of participants achieving confirmed CR or PR according to BICR and RECIST 1.1 during the specified time frame. Statistical uncertainty would then ordinarily be summarized with a confidence interval for the response proportion, with a prespecified comparison between the two randomized dose groups if the statistical analysis plan calls for such a comparison.
7. Results
Primary Objective Response Rate
The registry identifies confirmed ORR based on BICR and RECIST 1.1 as the primary endpoint, with an assessment period beginning 6 months post-dose administration and extending to data cut off, up to 20 months. The registry does not report a numerical ORR estimate, confidence interval, or p-value in the posted statistical-analysis information.
Primary endpoint
Assessment: blinded independent central review using RECIST version 1.1
Time frame: 6 months post-dose administration to data cut off, up to 20 months
The appropriate statistical interpretation is therefore centered on the structure of the endpoint rather than on an unreported numerical treatment effect. ORR is a proportion, so its uncertainty depends on both the observed number of responders and the number of participants contributing to the analysis. A larger number of participants generally provides more information about the underlying response probability, while a smaller response-analysis population produces less precise estimation.
The primary endpoint is a proportion of participants who achieve a confirmed CR or PR. It is not a measure of survival time, and it does not by itself describe how long a response lasts.
An ORR estimate would describe the fraction of the analyzed population satisfying the prespecified response definition. It would not mean that every participant has the same probability of response, nor would it establish the durability of that response.
A confidence interval would describe uncertainty around the estimated response probability under the relevant statistical framework. It would not describe the range of responses experienced by individual patients.
A p-value, if reported from a prespecified comparison, would address evidence against a specified null hypothesis. It would not measure the magnitude or clinical importance of the response difference. Effect size and statistical evidence are separate concepts.
Because the endpoint is based on BICR and RECIST 1.1, the response classification also depends on the quality, timing, and interpretation of radiologic assessments. The randomized dose comparison should therefore be interpreted within the defined assessment framework.
Safety
The registry reports serious adverse events by treatment arm. These counts describe participants affected by serious adverse events relative to the participants at risk in each arm.
| Treatment arm | Participants with serious adverse events | Participants at risk |
|---|---|---|
| T-DXd 5.4 mg/kg Q3W | 21 | 83 |
| T-DXd 6.4 mg/kg Q3W | 12 | 39 |
These safety counts should be interpreted separately from the primary efficacy endpoint. A serious adverse event count addresses a safety outcome, whereas ORR addresses tumor response. Neither endpoint can be substituted for the other.
8. Statistical Methods Explained
Why is ORR analyzed as a proportion?
Each participant can be classified according to whether the prespecified criteria for confirmed CR or PR are met. The primary summary is therefore the number of responders divided by the relevant analysis population. Expressing that quantity as a percentage makes the result easier to compare across treatment groups.
Why use blinded independent central review?
Radiologic response is partly an assessment process. If a reviewer knows which dose a participant received, that knowledge could potentially influence classification. BICR separates the assessment of tumor response from knowledge of randomized dose assignment, reducing one important source of assessment bias.
Why does RECIST 1.1 matter statistically?
A statistical analysis requires a reproducible definition of the outcome being counted. RECIST 1.1 provides the framework, while the registry specifies that CR means disappearance of all target lesions and PR means at least a 30% decrease in the sum of diameters of target lesions. Without a consistent outcome definition, the numerator in an ORR calculation would not be well defined.
What does the six-month-to-data-cutoff time frame mean?
The registry defines the endpoint time frame as 6 months post-dose administration to data cut off, up to 20 months. This identifies the observation window used for the registered endpoint. It should not be interpreted as a median follow-up time or as a claim that every participant was observed for the same duration.
Why is a confidence interval important for ORR?
An observed response percentage is estimated from a finite sample. A confidence interval communicates the statistical uncertainty surrounding that estimate. Two response percentages can differ numerically while still having substantial uncertainty, particularly when the number of participants is limited.
Why doesn't a p-value measure treatment benefit?
A p-value is tied to a statistical hypothesis and quantifies how compatible the observed data are with the null hypothesis under the specified test. It does not tell the reader how large the treatment difference is. The magnitude of an effect and the strength of statistical evidence should be reported separately.
9. Understanding the Primary Endpoint
ORR measures response
The endpoint counts participants achieving confirmed complete or partial response under the registry-defined BICR and RECIST 1.1 framework.
ORR is not survival
A response proportion does not measure time to death, time to progression, or duration of response.
Response requires a definition
The registry uses specific lesion-based criteria rather than an informal judgment of clinical improvement.
Precision matters
The response percentage is an estimate. Its uncertainty should be considered alongside the point estimate when formal results are available.
One useful statistical distinction is between response status and time-to-response or durability. DESTINY-CRC02's registered primary endpoint, as described in the registry, asks whether participants achieve confirmed CR or PR during the defined assessment period. It does not, by that definition alone, answer how long a response persists.
10. Dose-Level Comparison and Statistical Interpretation
DESTINY-CRC02 is particularly useful for understanding why the identity of the comparator matters. Both randomized groups receive T-DXd, with the dose level distinguishing the arms. Consequently, a difference in ORR between the groups would be interpreted as evidence about the relative response associated with the two dose regimens studied, rather than evidence for T-DXd versus no T-DXd.
| Statistical question | What it addresses |
|---|---|
| What is ORR in each arm? | Describes the observed response proportion in each dose group. |
| How different are the proportions? | Quantifies the magnitude of the dose-group contrast. |
| How precise is each estimate? | Assesses statistical uncertainty around the response proportions. |
| Is there evidence of a difference? | Addresses a prespecified statistical hypothesis concerning the two groups. |
| What does the endpoint not establish? | ORR alone does not establish duration of response or survival. |
This separation between description, estimation, and inference is fundamental to clinical-trial statistics. A percentage is descriptive. A confidence interval adds estimation of uncertainty. A hypothesis test addresses a prespecified question about compatibility with a null hypothesis. None of these quantities should be treated as interchangeable.
11. Serious Adverse Events
The registry reports serious adverse events separately for the two dose groups. The 5.4 mg/kg Q3W arm has 21 affected participants among 83 at risk, while the 6.4 mg/kg Q3W arm has 12 affected participants among 39 at risk.
| Safety measure | T-DXd 5.4 mg/kg Q3W | T-DXd 6.4 mg/kg Q3W |
|---|---|---|
| Serious adverse events, affected | 21 | 12 |
| Participants at risk | 83 | 39 |
The denominators are important. Counts alone cannot be compared as though the two groups had the same number of participants at risk. The registry's presentation therefore provides both the affected count and the corresponding at-risk count.
A serious-adverse-event count is an event-frequency measure, not an efficacy measure. Comparing safety between the two groups requires attention to the corresponding denominators, the precise safety definition, exposure, follow-up, and the analysis population used for the safety assessment.
The registry information summarized here provides the affected and at-risk counts but does not provide a posted formal statistical comparison of serious adverse events. The counts should therefore be treated as descriptive safety information.
12. Trial Timeline
Trial start
The registry identifies March 5, 2021 as the study start date.
Randomized dose-level study
The completed trial used randomized, double-blind, parallel allocation across two T-DXd dose levels.
Primary completion
The registry identifies November 1, 2022 as the primary completion date.
Results posted
The ClinicalTrials.gov record indicates that results have been posted, including the registered primary endpoint and serious adverse-event information.
13. Limitations
- Limited numerical efficacy reporting in the registry: the registry identifies the primary ORR endpoint and its assessment framework but does not provide a posted ORR estimate, confidence interval, or p-value in the statistical-analysis information summarized here.
- No posted formal statistical analysis: the record contains posted results but does not provide a formal statistical-analysis result for the primary endpoint.
- Binary endpoint: ORR reduces tumor-response assessment to a responder/non-responder classification for the purpose of the primary endpoint. That summary does not describe the full trajectory of tumor measurements.
- Assessment dependence: response classification depends on BICR and RECIST 1.1, so the endpoint is influenced by the specified radiologic assessment framework.
- Dose comparison: both randomized groups receive T-DXd. Any treatment-group contrast is therefore a comparison of dose regimens rather than T-DXd versus a control without T-DXd.
- Different arm sizes: the reported serious-adverse-event denominators are 83 and 39, so raw safety-event counts are not directly comparable without considering the corresponding populations at risk.
- Endpoint scope: the registered primary endpoint is ORR. ORR alone does not establish duration of response, progression-free survival, or overall survival.
14. Why This Trial Matters Statistically
DESTINY-CRC02 is a useful statistical teaching case because it illustrates how a randomized clinical trial can focus on a binary tumor-response endpoint while comparing two active dose regimens. It also shows why endpoint definitions, independent review, denominators, and uncertainty are central to interpretation.
| Concept | How it appears in DESTINY-CRC02 |
|---|---|
| Randomization | The trial is identified as randomized. |
| Parallel design | The two treatment groups are evaluated as parallel arms. |
| Double masking | The registry identifies the trial as double-masked. |
| Binary efficacy endpoint | Participants are classified according to confirmed CR or PR for ORR. |
| Independent review | Primary ORR assessment uses blinded independent central review. |
| RECIST 1.1 | Response is defined using a standardized tumor-assessment framework. |
| Proportion estimation | ORR summarizes the fraction of participants meeting the response definition. |
| Confidence intervals | Would quantify uncertainty around a response estimate in a formal analysis. |
| Safety denominators | Serious adverse events are reported with affected and at-risk counts for each arm. |
| Dose comparison | The two randomized groups receive different T-DXd dose levels. |
The most important statistical lesson is that the definition of the endpoint determines what the result can legitimately answer. ORR is useful for quantifying tumor response, but it is not interchangeable with time-to-event outcomes. Likewise, a safety count is not a measure of efficacy. Each statistic has a defined question behind it.
15. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary endpoint is a binary response outcome summarized as a proportion and assessed by blinded independent central review using RECIST 1.1. Formal interpretation would require the observed response counts, analysis population, effect estimate, uncertainty interval, and any prespecified hypothesis test.
Clinical interpretation
The endpoint asks whether participants experienced confirmed tumor response under a standardized radiologic framework. Its clinical meaning depends on the magnitude and precision of the observed response as well as the broader efficacy and safety evidence.
Keeping these interpretations separate prevents a common statistical error: treating a response percentage as if it were a complete summary of treatment benefit. ORR can describe an important aspect of tumor activity, but it does not by itself capture every clinically relevant outcome.
16. Reading the Denominator Correctly
The denominator is one of the most important details in any proportion. If an ORR analysis includes \(N\) participants and \(R\) meet the confirmed-response definition, the observed ORR is \(R/N\). The meaning of the percentage depends on precisely which participants are included in \(N\).
The same numerator can produce different percentages if the denominator changes. A valid interpretation therefore requires the analysis population to be identified alongside the response percentage.
The same principle is visible in the registry's safety reporting. The serious-adverse-event counts are accompanied by their corresponding at-risk populations: 21/83 for the 5.4 mg/kg Q3W group and 12/39 for the 6.4 mg/kg Q3W group.
17. What the Registry Establishes About the Primary Endpoint
The ClinicalTrials.gov record establishes several important features of the primary endpoint. It is a percentage-based ORR measure, it requires confirmed CR or PR, it is assessed by BICR, it uses RECIST version 1.1, and its registered time frame is 6 months post-dose administration to data cut off, up to 20 months.
| Endpoint component | Registry description |
|---|---|
| Outcome | Objective response rate |
| Response categories | Complete response or partial response |
| Complete response | Disappearance of all target lesions |
| Partial response | At least a 30% decrease in the sum of diameters of target lesions |
| Review method | Blinded independent central review |
| Response framework | RECIST version 1.1 |
| Registered time frame | 6 months post-dose administration to data cut off, up to 20 months |
These details are not administrative decoration. They define the outcome being analyzed. Changing the response definition, assessment method, or observation window could change which participants are classified as responders and therefore change the resulting ORR.
18. Sources
- ClinicalTrials.gov: NCT04744831.
- PubMed record: PubMed.
- PubMed record: PubMed.
Continue through Clinical Biostats
Explore statistical methods, clinical-trial concepts, and analysis resources that provide additional context for interpreting randomized clinical-trial evidence.
19. Record Summary
DESTINY-CRC02 is a randomized, double-blind, parallel phase 2 study of two T-DXd dose levels in participants with HER2-overexpressing advanced or metastatic colorectal cancer. Its registered primary endpoint is confirmed objective response rate assessed by blinded independent central review using RECIST version 1.1, with the registered assessment period extending from 6 months post-dose administration to data cut off, up to 20 months.
From a statistical perspective, the trial provides a clear example of how a binary efficacy endpoint is constructed from a clinical assessment rule. The response definition determines the participant-level outcome; those outcomes determine the numerator and denominator; and the resulting proportion becomes the descriptive ORR estimate. Formal statistical interpretation would then require the corresponding uncertainty estimate and, where prespecified, a comparison between the two randomized dose groups.
The safety information illustrates the same principle from another direction: serious adverse events are reported with affected and at-risk counts for each arm. These data should be interpreted as safety information and kept conceptually separate from the efficacy endpoint.