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-CRC01 was a phase 2 treatment study of DS-8201a in participants with HER2-expressing advanced colorectal cancer. The registry describes the study as non-randomized, sequential, unmasked, and consisting of 3 arms, with 86 participants enrolled.
| Feature | DESTINY-CRC01 |
|---|---|
| Study acronym | DESTINY-CRC01 |
| ClinicalTrials.gov identifier | NCT03384940 |
| Phase | Phase 2 |
| Status | Completed |
| Condition | Colorectal Neoplasm |
| Intervention | DS-8201a (drug) |
| Allocation | Non-randomized |
| Design model | Sequential |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 86 |
| Number of arms | 3 |
| Lead sponsor | Daiichi Sankyo Co., Ltd. |
| Sponsor type | Industry |
2. Clinical Question
The central clinical question was whether treatment with DS-8201a produced objective tumor responses in participants with HER2-expressing advanced colorectal cancer, with response assessed by independent central review.
Population
Participants with human epidermal growth factor receptor 2 (HER2)-expressing advanced colorectal cancer.
Intervention
DS-8201a.
Comparator
The registry identifies the study allocation as non-randomized. It does not identify a randomized comparator in the trial record summarized here.
Primary question
How many participants experienced a best objective response, and what was the objective response rate, based on independent central review?
3. Trial Design
DESTINY-CRC01 used a non-randomized, sequential phase 2 design. The registry identifies 3 arms and 86 enrolled participants. Because allocation was non-randomized, the study structure differs fundamentally from a parallel-group randomized controlled trial: there is no randomized treatment contrast from which a between-group causal effect can be estimated.
Study cohorts
DS-8201a Cohort A
The registry identifies Cohort A as one of the 3 DS-8201a cohorts. The serious-adverse-event summary reports 20 affected participants among 53 at risk.
DS-8201a Cohort B
The registry identifies Cohort B as one of the 3 DS-8201a cohorts. The serious-adverse-event summary reports 6 affected participants among 15 at risk.
DS-8201a Cohort C
The registry identifies Cohort C as one of the 3 DS-8201a cohorts. The serious-adverse-event summary reports 9 affected participants among 18 at risk.
4. Endpoints
The registry identifies 2 primary endpoints. Both evaluate tumor response from baseline through withdrawal of consent, progressive disease, or unacceptable toxicity, whichever occurs first, up to 18 months post-dose.
| Primary endpoint | Time frame | Registry definition |
|---|---|---|
| Number of Participants With Best Objective Response Based on Independent Central Review (Confirmed and Unconfirmed) Following DS-8201a Treatment in Participants With Human Epidermal Growth Factor Receptor 2 (HER2)-Expressing Advanced Colorectal Cancer | Baseline up to withdrawal of consent, progressive disease, or unacceptable toxicity (whichever occurs first), up to 18 months post-dose | Best objective response was reported based on independent central review. Under RECIST version 1.1, CR was defined as disappearance of all target lesions; PR as at least a 30% decrease in the sum of diameters of target lesions; and SD as neither sufficient shrinkage for PR nor sufficient increase for PD, with PD defined as at least a 20% increase in the sum of diameters of target lesions. |
| Number of Participants With Objective Response Rate Based on Independent Central Review (Confirmed and Unconfirmed) Following DS-8201a Treatment in Participants With Human Epidermal Growth Factor Receptor 2 (HER2)-Expressing Advanced Colorectal Cancer | Baseline up to withdrawal of consent, progressive disease, or unacceptable toxicity (whichever occurs first), up to 18 months post-dose | Objective response rate was defined as CR+PR and reported based on independent central review. Under RECIST version 1.1, CR was disappearance of all target lesions and PR was at least a 30% decrease in the sum of diameters of target lesions. |
Why two response endpoints?
The two primary endpoints describe closely related but distinct quantities. Best objective response describes the best response category achieved by a participant during the specified assessment period. Objective response rate summarizes the proportion of participants classified as having an objective response, with response defined as complete response or partial response.
The distinction matters statistically because a categorical response endpoint contains more information than a simple responder/non-responder indicator. A participant's best category can distinguish complete response, partial response, stable disease, and progressive disease, whereas ORR collapses complete and partial responses into a single response category.
5. Statistical Methodology
Independent central review
The primary response endpoints were assessed by independent central review. Central review is important because response classification can otherwise be influenced by knowledge of treatment, expectations, or differences in local radiologic interpretation. An independent assessment provides a standardized endpoint determination for the efficacy analysis.
RECIST version 1.1
The registry specifies Response Evaluation Criteria In Solid Tumors (RECIST) version 1.1 as the response framework. The definitions establish prespecified thresholds for classifying tumor response rather than relying on an unrestricted clinical judgment about whether a tumor has improved.
Stable disease occupies the range in which shrinkage is insufficient for PR and increase is insufficient for PD.
Objective response rate
The registry defines ORR as CR + PR. Conceptually, if R denotes the number of participants achieving CR or PR and N denotes the relevant analysis population, then:
The resulting proportion summarizes the fraction of participants who achieved an objective response during the specified assessment period.
Because the study is non-randomized, ORR is principally a measure of response observed under DS-8201a treatment. It does not by itself provide a randomized comparison with another therapy.
Best objective response
Best objective response is a longitudinal categorical outcome. Each participant can be classified according to the best response observed during the specified period, including confirmed and unconfirmed responses as described in the endpoint title. This approach uses repeated tumor assessments to determine the participant's best observed category rather than evaluating only one prespecified time point.
Time frame and censoring of assessment
The response endpoints were followed from baseline until withdrawal of consent, progressive disease, or unacceptable toxicity, whichever occurred first, with the registry specifying a maximum of 18 months post-dose. These stopping rules define the observation window for the endpoint and prevent the assessment period from being treated as indefinitely open-ended.
6. Planned Analysis
The registry reports that results are posted for both primary response endpoints, but no formal statistical analyses are posted in the record summarized here. The appropriate interpretation is therefore centered on the endpoint definitions and the response framework rather than on an independently reconstructed numerical treatment effect.
How these endpoints are typically analysed
For objective response rate, the usual analysis is a binomial proportion: the number of participants with CR or PR is divided by the prespecified analysis population. A confidence interval for that proportion can then describe sampling uncertainty.
For best objective response, the usual descriptive analysis is a frequency distribution across the RECIST categories. The important statistical point is that response is categorical, so the complete distribution can be more informative than ORR alone.
In a single-arm or non-randomized phase 2 study, a response rate may also be compared with a prespecified benchmark or historical control rate when such a hypothesis and threshold are explicitly defined in the statistical analysis plan. The registry information summarized here does not provide such a benchmark, a null response rate, an alpha level, or a formal hypothesis-testing specification.
A response endpoint answers a narrower question than overall survival or progression-free survival. It asks whether tumors met predefined radiologic criteria for response during the specified observation window. It does not by itself establish how long a response lasts, whether survival is prolonged, or how outcomes compare with another treatment in the absence of a randomized comparator.
7. Safety Results
The registry reports serious adverse events by DS-8201a cohort. The available safety measure is the number of participants affected among those at risk in each cohort.
| Cohort | Serious adverse events | At risk |
|---|---|---|
| DS-8201a Cohort A | 20 affected | 53 |
| DS-8201a Cohort B | 6 affected | 15 |
| DS-8201a Cohort C | 9 affected | 18 |
These figures should be read as cohort-specific safety counts. They do not constitute a randomized safety comparison because the study allocation was non-randomized.
Safety denominator matters
The registry expresses the serious-adverse-event results as affected participants divided by the number at risk. The denominator differs across the 3 cohorts.
Safety and efficacy are distinct
Response endpoints describe tumor activity, whereas serious adverse events describe an important component of treatment safety. They should not be combined into a single numerical efficacy-safety measure.
8. Statistical Methods Explained
Why was independent central review used?
Radiologic tumor response requires classification against predefined imaging criteria. Independent central review provides a separate assessment of those images and reduces the influence that treatment knowledge or local assessment practices could otherwise have on endpoint classification.
What does RECIST 1.1 add to the analysis?
RECIST 1.1 turns tumor shrinkage into prespecified categories. A partial response requires at least a 30% decrease in the sum of diameters of target lesions, while progressive disease requires at least a 20% increase. These thresholds create reproducible rules for classifying response.
Why is ORR defined as CR+PR?
ORR focuses on objective tumor shrinkage. Complete response and partial response are both considered objective responses, so they are combined into a single responder category. Stable disease and progressive disease are not counted as objective responses.
Why does a non-randomized design change interpretation?
Without randomization, differences between patients who receive a treatment and an external population can reflect differences in patient characteristics rather than treatment itself. A response proportion can therefore describe activity in the studied population without providing the same causal comparison that a randomized trial can provide.
Why is the analysis population important for ORR?
The denominator determines the meaning of the reported response rate. A response proportion calculated among all enrolled participants answers a different question from one calculated among evaluable participants. The statistical analysis should therefore identify the prespecified analysis population rather than treating every possible denominator as interchangeable.
What does a confidence interval for ORR tell us?
A confidence interval around an observed response proportion describes statistical uncertainty about the underlying response proportion under the specified sampling framework. It does not describe the probability that an individual participant will respond, nor does it establish a comparative treatment effect.
9. Interpreting Response Endpoints Correctly
Best objective response is a categorical summary of the most favorable response category observed during the specified assessment period. It is useful for showing the distribution of tumor responses across participants.
ORR reduces the response categories to a binary responder definition: CR or PR versus not an objective response. This makes ORR easy to summarize, but it necessarily discards some information contained in the full best-response distribution.
ORR is not a measure of overall survival, progression-free survival, or duration of response. It also does not mean that every responder experienced the same magnitude or durability of benefit.
Because the primary endpoints were based on independent central review, the response classification has a defined external assessment process rather than relying solely on investigator-reported tumor response.
10. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The study is a phase 2, non-randomized, sequential design with 3 arms and 86 enrolled participants. Its primary endpoints are categorical tumor-response measures assessed by independent central review using RECIST version 1.1.
Clinical interpretation
The trial was designed to characterize tumor response to DS-8201a in participants with HER2-expressing advanced colorectal cancer. The registry also reports serious adverse-event counts by cohort.
The distinction between these two interpretations is important. A statistical description focuses on how the endpoint was defined, measured, and analysed. A clinical interpretation considers what those response and safety measures mean in the context of treatment development. Neither response rate nor serious-adverse-event frequency, considered alone, captures the complete clinical experience of treatment.
11. Important Limitations and Interpretation Issues
- Non-randomized allocation: the study does not provide a randomized treatment-versus-control comparison. This limits causal interpretation of observed response relative to another therapy.
- Sequential design: the registry identifies a sequential rather than a conventional parallel randomized design. The sequence and cohort structure therefore need to be considered when interpreting comparisons among cohorts.
- Different cohort denominators: the serious-adverse-event summaries have different numbers at risk across Cohorts A, B, and C.
- Response is not survival: best objective response and ORR describe tumor response and do not directly measure time to progression or death.
- Assessment window: response was evaluated from baseline until withdrawal of consent, progressive disease, or unacceptable toxicity, whichever occurred first, up to 18 months post-dose. The endpoint therefore has a defined observation window.
- Confirmed and unconfirmed responses: the endpoint titles explicitly include both confirmed and unconfirmed responses. These categories should not be silently treated as equivalent to a separately defined confirmed-response analysis.
- Statistical-analysis detail: the registry reports results but does not post a formal statistical analysis in the record summarized here. Details such as a prespecified null response rate, formal hypothesis test, alpha level, and confidence-interval method are not identified in the available registry data.
- Generalizability: the study population consists of participants with HER2-expressing advanced colorectal cancer, so conclusions apply most directly to the population represented by the study eligibility criteria.
12. Why This Trial Matters Statistically
DESTINY-CRC01 is a useful teaching example because it illustrates how a phase 2 non-randomized oncology study can be statistically informative without having the structure of a randomized comparative trial.
| Concept | How it appears in DESTINY-CRC01 |
|---|---|
| Phase 2 development | The study is identified as phase 2 and has treatment as its primary purpose. |
| Non-randomized allocation | Participants were not allocated through randomization. |
| Sequential design | The registry identifies the design model as sequential. |
| Multiple cohorts | The study has 3 arms, identified in the safety data as Cohorts A, B, and C. |
| Independent central review | The primary response endpoints use independent central review. |
| RECIST 1.1 | Response categories are defined using prespecified RECIST version 1.1 criteria. |
| Best objective response | The study evaluates each participant's best observed objective response category. |
| Objective response rate | ORR is defined as CR+PR. |
| Defined observation window | Endpoints are assessed from baseline through specified stopping events, up to 18 months post-dose. |
| Safety denominators | Serious adverse events are reported as affected participants among those at risk within each cohort. |
13. The Statistical Logic of a Single-Arm Response Study
For a non-randomized phase 2 study, the statistical objective is often different from that of a confirmatory randomized trial. The central question can be whether the observed response experience is sufficiently compelling to justify additional clinical development.
That logic places particular importance on three elements: the response definition, the analysis population, and the prespecified benchmark. The response definition determines what counts as success. The analysis population determines who contributes to the denominator. The benchmark determines what observed response level would constitute evidence beyond an assumed background rate.
In DESTINY-CRC01, the registry clearly specifies the first element through independent central review and RECIST version 1.1. It also specifies the response observation period. However, the registry information summarized here does not provide the prespecified statistical benchmark or formal hypothesis-testing parameters. Those details are therefore essential when interpreting any formal statistical claim about response beyond the descriptive endpoint itself.
Here p0 represents a prespecified benchmark response rate. A formal single-arm hypothesis test requires that benchmark and its associated statistical design to be specified in advance.
This framework also shows why an observed response proportion should not automatically be interpreted as statistically significant merely because it appears numerically large. Statistical evidence depends on the prespecified benchmark, sample size, variability, and analysis plan.
14. Understanding the Role of the Three Cohorts
The registry identifies 3 arms in a sequential, non-randomized study and reports serious adverse events separately for Cohorts A, B, and C. That structure is different from a randomized three-arm trial in which treatment assignment is randomized among 3 competing interventions.
In a randomized three-arm study, differences in baseline patient characteristics are balanced in expectation through randomization. In a sequential non-randomized study, cohort membership may be connected to the study sequence and eligibility or treatment-development decisions. Consequently, a numerical difference between cohorts should not automatically be interpreted as a treatment effect.
Cohort-level description
The cohort-specific safety counts describe what was observed among participants in each cohort and its corresponding at-risk population.
Comparative caution
Without randomization, numerical differences between cohorts do not have the same causal interpretation as differences between randomized treatment groups.
15. Primary Endpoint Interpretation
The primary endpoint structure emphasizes tumor activity. Best objective response records the most favorable RECIST category achieved during the assessment window, while ORR summarizes the CR and PR categories.
This makes the trial particularly useful for understanding the relationship between an underlying categorical endpoint and a derived binary summary. ORR is easier to communicate because it produces one proportion, but the complete best-response distribution retains information that ORR alone does not.
| Endpoint concept | Statistical meaning |
|---|---|
| CR | All target lesions disappear. |
| PR | The sum of diameters of target lesions decreases by at least 30%. |
| SD | Shrinkage is insufficient for PR and increase is insufficient for PD. |
| PD | The sum of diameters of target lesions increases by at least 20%. |
| ORR | CR + PR. |
| Best objective response | The best response category observed during the specified assessment period. |
16. Limitations of Response-Based Inference
Response endpoints are valuable because they can provide an early signal of antitumor activity. They are nevertheless narrower than time-to-event endpoints. A participant may experience tumor shrinkage without that response necessarily translating into a particular duration of disease control or survival experience.
The use of an independent central review improves consistency of response classification, but it does not transform a non-randomized study into a randomized comparison. Similarly, a precisely measured ORR can still be limited as evidence of comparative treatment efficacy when there is no randomized control group.
The appropriate statistical interpretation therefore depends on the question being asked. For “How frequently did predefined tumor responses occur?”, ORR is directly relevant. For “Did DS-8201a improve outcomes compared with another treatment?”, a non-randomized response proportion cannot answer the question with the same causal strength as a randomized comparative design.
17. Study Timeline
Study start
The registry records 2018-02-23 as the study start date.
Primary completion
The registry records 2019-08-09 as the primary completion date.
Study status
The registry identifies the study status as completed and reports that results have been posted.
18. Statistical Takeaways
Response is the central efficacy measure
The two primary endpoints focus on best objective response and objective response rate, both assessed by independent central review.
RECIST creates reproducibility
Prespecified thresholds for CR, PR, SD, and PD turn imaging observations into standardized categorical outcomes.
Non-randomization limits comparison
The study can characterize response under DS-8201a treatment, but cohort differences do not have the same causal interpretation as randomized treatment contrasts.
Safety needs its own denominator
Serious adverse events are reported separately by cohort using the number affected and the number at risk.
DESTINY-CRC01 illustrates a central principle of clinical-trial statistics: the strength of an inference depends on the design as well as the endpoint. Independent central review and RECIST 1.1 provide a structured way to measure tumor response, while the non-randomized sequential design determines how far those response observations can be taken when making comparative or causal claims.
19. Sources
- ClinicalTrials.gov: DESTINY-CRC01, NCT03384940.
Continue through the Clinical Biostats knowledge graph
Explore additional Clinical Biostats material on clinical-trial design, endpoint definition, statistical analysis, and interpretation.
20. Record Summary
DESTINY-CRC01 is a completed phase 2, non-randomized, sequential study of DS-8201a in participants with HER2-expressing advanced colorectal cancer. The registry identifies 86 enrolled participants, 3 arms, and 2 primary response endpoints. Both primary endpoints use independent central review and RECIST version 1.1, with assessment from baseline through withdrawal of consent, progressive disease, or unacceptable toxicity, whichever occurs first, up to 18 months post-dose.
The statistical structure is important to the interpretation. Best objective response preserves the categorical response information generated during follow-up, while ORR reduces that information to the proportion achieving CR or PR. Because allocation was non-randomized, these endpoints primarily describe observed antitumor activity rather than a randomized comparative treatment effect. The registry also reports serious adverse-event counts separately for the 3 cohorts: 20/53 in Cohort A, 6/15 in Cohort B, and 9/18 in Cohort C.
As a statistical teaching case, the study demonstrates how endpoint definition, independent assessment, RECIST thresholds, analysis denominators, cohort structure, and randomization status all shape the conclusions that can appropriately be drawn from a clinical trial.