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-Gastric02 was a completed phase 2, single-group clinical trial of trastuzumab deruxtecan in participants with HER2-positive unresectable or metastatic gastric or gastro-esophageal junction (GEJ) adenocarcinoma. The registry reports an enrollment of 79 participants and one intervention arm.
| Feature | DESTINY-Gastric02 |
|---|---|
| Phase | Phase 2 |
| Population | Participants with HER2-positive unresectable or metastatic gastric or gastro-esophageal junction (GEJ) adenocarcinoma |
| Design | Single-group |
| Allocation | NA |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 79 |
| Number of arms | 1 |
| Intervention | Trastuzumab deruxtecan |
| Status | Completed |
| Start date | November 26, 2019 |
| Primary completion date | November 8, 2021 |
| Lead sponsor | Daiichi Sankyo |
2. Clinical Question
The central statistical question is descriptive: among participants with HER2-positive unresectable or metastatic gastric or GEJ adenocarcinoma who received trastuzumab deruxtecan, what proportion achieved a confirmed objective response according to independent central review using RECIST version 1.1?
Population
Participants with HER2-positive unresectable or metastatic gastric or gastro-esophageal junction adenocarcinoma.
Intervention
Trastuzumab deruxtecan.
Comparator
The study has one arm. No randomized comparator arm is reported in the registry record.
Primary question
What percentage of participants achieve a confirmed complete or partial response by independent central review?
3. Trial Design
Trastuzumab deruxtecan
- Trastuzumab deruxtecan
- One study arm
- Objective response assessed using independent central review
- Response assessment based on RECIST version 1.1
No randomized control arm
- The registry identifies the study as single-group
- No comparator intervention is reported
- Between-group randomized treatment effects therefore cannot be estimated from this design
4. Primary Endpoints
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Percentage of Participants With Objective Response Rate (ORR) Based on Independent Central Review Following Treatment With DS8201a in Participants With HER2-Positive Unresectable or Metastatic Gastric or Gastro-Esophageal Junction (GEJ) Adenocarcinoma | Up to 16 months (data cut-off) | The Objective Response Rate (ORR) was the defined as the percentage of participants who achieved a best overall response of confirmed Complete Response (CR) or Partial Response (PR), assessed by independent central review (ICR) committee based on RECIST version 1.1. CR was defined as a disappearance of all target lesions and PR was defined as at least a 30% decrease in the sum of diameters of target lesions. Confirmed ORR based on ICR is reported. |
| Percentage of Participants With Objective Response Rate (ORR) Based on Independent Central Review Following Treatment With DS8201a in Participants With HER2-Positive Unresectable or Metastatic Gastric or Gastro-Esophageal Junction (GEJ) Adenocarcinoma | Up to 23 months (data cut-off) | The Objective Response Rate (ORR) was the defined as the percentage of participants who achieved a best overall response of confirmed Complete Response (CR) or Partial Response (PR), assessed by independent central review (ICR) committee based on RECIST version 1.1. CR was defined as a disappearance of all target lesions and PR was defined as at least a 30% decrease in the sum of diameters of target lesions. Confirmed ORR based on ICR is reported. |
What the endpoint is measuring
ORR is a binary-type efficacy outcome at the participant level: each participant is classified according to whether the best overall response satisfies the protocol's confirmed CR or PR definition. The registry's definition makes the response criterion clinically concrete. CR requires disappearance of all target lesions, while PR requires at least a 30% decrease in the sum of diameters of target lesions.
The two primary endpoint entries use different data cutoffs—up to 16 months and up to 23 months. These are therefore distinct analysis time frames and should not be silently combined into one estimate.
5. Results
The registry indicates that results were posted for both primary ORR endpoint entries. However, the registry record contains no posted statistical-analysis objects for these endpoints, so numerical ORR estimates, confidence intervals, and p-values are not available in the statistical-analysis material represented here. No formal statistical analyses were posted to ClinicalTrials.gov.
Primary endpoint: up to 16 months
Confirmed ORR by independent central review
Time frame: Up to 16 months (data cut-off)
Assessment: Independent central review using RECIST version 1.1
The endpoint counts participants whose best overall response was a confirmed CR or PR. The registry defines CR as disappearance of all target lesions and PR as at least a 30% decrease in the sum of diameters of target lesions.
The estimand is a proportion: the fraction of the analyzed participants meeting the confirmed response definition. It is not a hazard ratio and does not describe a rate of progression or death over time.
Because this is a single-group study, an ORR estimate would describe response within the treated cohort. It would not by itself quantify the treatment effect relative to a concurrent control group.
A confidence interval around a proportion would quantify statistical precision under the chosen binomial framework. A wider interval would indicate greater uncertainty about the underlying response proportion; it would not mean that individual patients' responses vary across that interval.
A p-value, if a prespecified hypothesis test were used, would address evidence against a specified null response proportion. It would not measure the magnitude of ORR or the clinical importance of an observed response proportion.
Primary endpoint: up to 23 months
Confirmed ORR by independent central review
Time frame: Up to 23 months (data cut-off)
Assessment: Independent central review using RECIST version 1.1
The later endpoint uses the same confirmed CR/PR definition but a different data cutoff. The distinction matters because additional follow-up can change response classification and the number of participants contributing information to the analysis.
The up-to-23-month endpoint should be interpreted as a separate time-framed response analysis. It should not automatically be treated as a simple extension of the up-to-16-month result unless the underlying analysis population and censoring or assessment rules are identical.
Independent central review is important because response assessment can otherwise be influenced by knowledge of treatment assignment or expectations. Here, the study itself is unmasked, while tumor-response assessment is specified as independent central review.
The absence of a randomized comparator also limits causal interpretation. Even a precisely estimated response proportion cannot establish how much of that response would differ from an alternative treatment in a randomized comparison.
6. Safety Results
The registry reports serious adverse events by arm for the trastuzumab deruxtecan group. Because there is one arm, the reported safety measure is presented as an affected-versus-at-risk count rather than as a randomized between-group comparison.
| Safety measure | Trastuzumab deruxtecan |
|---|---|
| Serious adverse events | 37/79 |
Serious adverse events
Trastuzumab deruxtecan arm: affected / at risk
The numerator identifies participants affected by serious adverse events and the denominator identifies participants at risk in the reported arm. The ClinicalTrials.gov record does not give a comparator-group rate, so this safety measure should not be interpreted as a relative treatment effect.
A serious-adverse-event proportion is a safety frequency, not an efficacy endpoint. It summarizes how many participants experienced the specified safety outcome among those at risk.
In a single-group study, the observed frequency cannot by itself determine whether the intervention caused more or fewer serious adverse events than another treatment. That question requires a suitable comparison group and careful definition of exposure and follow-up.
7. Statistical Methodology
Analysis of a binary response endpoint
For an ORR endpoint, each participant can be classified as a responder or non-responder according to the prespecified response definition. The fundamental descriptive quantity is therefore a binomial proportion:
The resulting proportion is commonly reported as a percentage. Confidence intervals can be constructed using an appropriate binomial method.
Why independent central review matters
The registry specifies independent central review by an ICR committee and assessment according to RECIST version 1.1. Central review creates a standardized assessment framework and reduces dependence on a single local investigator's interpretation of imaging.
This is especially relevant for response endpoints because ORR depends on categorical decisions about tumor response. The statistical analysis therefore begins with a clinically defined classification rule before the proportion is calculated.
Confirmed response versus an isolated measurement
The endpoint is based on confirmed CR or PR rather than simply any observation of tumor shrinkage. Confirmation is important because a temporary or ambiguous imaging change should not automatically be treated as a durable objective response.
Confidence intervals for ORR
For a single-group proportion, a confidence interval describes uncertainty around the underlying response probability under the selected statistical model. Exact or score-based binomial methods are commonly considered when sample sizes are modest or the observed proportion is near 0% or 100%.
Here, X represents the number of responders, n the number of participants contributing to the response analysis, and p the underlying probability of a confirmed response.
No hazard ratio is required for ORR
ORR is fundamentally different from time-to-event endpoints such as overall survival or progression-free survival. A hazard ratio is designed for comparing event rates over time. A response proportion instead summarizes the frequency of a prespecified categorical outcome.
No randomized treatment contrast
Because the registry identifies the study as single-group, the central inferential limitation is the absence of a concurrent randomized comparator. Statistical precision and causal identification are separate concepts: a single-group proportion can be estimated precisely while still providing limited evidence about the treatment's incremental effect relative to another therapy.
8. Statistical Methods Explained
Why is ORR analysed as a proportion?
ORR classifies participants according to whether they achieved a confirmed CR or PR. The natural summary is therefore the number of responders divided by the number of participants in the analysis population.
What does a confirmed CR mean?
Under the registry definition, CR means disappearance of all target lesions. It is one of the two response categories that qualify a participant as an objective responder.
What does a confirmed PR mean?
PR is defined as at least a 30% decrease in the sum of diameters of target lesions. This threshold converts continuous tumor-size measurements into a prespecified categorical response classification.
Why use independent central review?
Independent central review provides a separate assessment of tumor response using the specified RECIST version 1.1 framework. This can reduce the influence of local assessment differences on the response endpoint.
Why are the 16-month and 23-month cutoffs separate?
Each cutoff defines a different information set. A later cutoff can include additional assessments and follow-up, so the corresponding ORR estimate should be interpreted as its own analysis time frame rather than automatically merged with the earlier one.
Why can't a single-arm ORR establish comparative efficacy?
A response percentage tells us what occurred in the treated cohort. To estimate a treatment effect, the counterfactual question is needed: what would have happened to comparable participants under another treatment? Randomization provides the principal design mechanism for making that comparison credible.
9. Understanding the Evidence From a Single-Group Trial
Descriptive evidence
The study can describe the frequency of confirmed objective response among participants receiving trastuzumab deruxtecan.
Comparative evidence
A randomized comparator is absent, so the design does not directly estimate a randomized treatment difference.
Assessment standardization
Independent central review and RECIST version 1.1 provide a defined framework for response classification.
Safety interpretation
The registry reports serious adverse events as 37/79 in the trastuzumab deruxtecan arm, without a second randomized arm in the record represented here.
This distinction is central to statistical interpretation. A single-arm phase 2 study can be informative for estimating the magnitude and frequency of an observed response, while questions about relative efficacy require a different evidentiary structure.
10. Data Cutoffs and Longitudinal Interpretation
Study start
The registry records the start of DESTINY-Gastric02.
First reported ORR time frame
The registry reports confirmed ORR based on independent central review using this data cutoff.
Primary completion
The registry records November 8, 2021 as the primary completion date.
Later ORR time frame
The registry also reports confirmed ORR based on independent central review using a later data cutoff.
Using separate data-cutoff labels is statistically important. Clinical-trial results are snapshots of an accumulating information set. A later cutoff may provide additional assessments, but it should still be reported with its own time frame and analysis population.
11. Important Limitations and Interpretation Issues
- No randomized comparator: the single-group design does not provide a concurrent randomized estimate of the difference between trastuzumab deruxtecan and another treatment.
- External comparisons: comparing a single-arm ORR with results from another study can be confounded by differences in patient selection, assessment procedures, follow-up, and study conduct.
- Response endpoint: ORR measures confirmed tumor response and is not equivalent to overall survival, progression-free survival, or duration of response.
- Data-cutoff dependence: the registry reports two ORR time frames, up to 16 months and up to 23 months. These should be kept analytically distinct.
- Analysis-population dependence: a percentage is meaningful only with a clearly defined denominator and response-assessment population.
- Confidence-interval interpretation: uncertainty around a response proportion is statistical uncertainty about the population parameter, not a range of outcomes that an individual patient can experience.
- Safety interpretation: 37/79 serious adverse events in the intervention arm is a descriptive frequency and does not provide a randomized comparative safety effect.
- Multiplicity: when multiple endpoints, cutoffs, or analyses are examined, the interpretation of nominal statistical tests depends on the prespecified testing framework.
- Single-arm causal inference: even a large observed response proportion cannot by itself separate treatment effect from the underlying prognosis or other features of the study population.
12. Why This Trial Matters Statistically
DESTINY-Gastric02 is a useful teaching example because it illustrates a different statistical structure from a randomized phase 3 trial. The central endpoint is an objectively defined response proportion, the study has a single treatment group, and tumor response is assessed by independent central review.
| Concept | How it appears in DESTINY-Gastric02 |
|---|---|
| Single-group design | One intervention arm with 79 enrolled participants |
| Binary efficacy endpoint | Participants are classified according to confirmed CR or PR |
| Objective response rate | Percentage of participants achieving confirmed CR or PR |
| Independent central review | Response assessed by an independent central review committee |
| RECIST version 1.1 | Framework used to define the response endpoint |
| Data cutoffs | ORR reported at up to 16 months and up to 23 months |
| Confidence intervals | Would quantify uncertainty around a single-group response proportion |
| Comparative inference | Limited by the absence of a randomized comparator |
| Safety frequency | Serious adverse events reported as 37/79 in the intervention arm |
13. Planned Analysis
For an endpoint of this type, the primary descriptive analysis would ordinarily calculate the proportion of participants whose best overall response is a confirmed CR or PR according to the specified independent central review and RECIST version 1.1 definitions. The corresponding confidence interval would typically use a binomial method appropriate for a single proportion.
A hypothesis test, if prespecified, would compare the observed response proportion with a clinically defined null value. In a single-group study, that null value is not the same thing as a randomized control-group response rate; it represents a benchmark against which the single-group result is evaluated.
The registry reports two primary ORR assessment time frames: up to 16 months and up to 23 months. Each should retain its own data-cutoff designation when interpreting the analysis.
14. What ORR Does — and Does Not — Mean
ORR is the percentage of participants whose best overall response meets the confirmed CR or PR definition. It is a proportion, not a hazard ratio and not a measure of instantaneous event risk.
A higher ORR would mean that a larger fraction of participants met the prespecified tumor-response criterion. It would not by itself mean that the same fraction of participants experienced prolonged survival or durable disease control.
A confidence interval around ORR describes uncertainty about the underlying response probability under the selected statistical framework. With a single-group sample, it reflects sampling uncertainty; it does not quantify variability in tumor response among individual patients.
If a response hypothesis were tested, its p-value would describe compatibility of the observed data with the prespecified null hypothesis. A small p-value would not automatically imply a large or clinically important response proportion, while a larger p-value would not prove that the treatment has no activity.
For a single-arm study, the observed ORR is inherently a within-study description. A treatment effect requires a comparison with an appropriate counterfactual, and randomization is the strongest standard design for obtaining that comparison.
15. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The principal efficacy endpoint is a confirmed objective response proportion assessed by independent central review at two specified data cutoffs. The appropriate statistical framework is based on estimation of a single binomial proportion rather than a randomized hazard ratio.
Clinical interpretation
The endpoint describes the frequency with which participants met a predefined tumor-response criterion after treatment with trastuzumab deruxtecan. Its clinical meaning should be considered separately from survival outcomes and comparative treatment effects.
This separation prevents a common statistical error: treating an observed response percentage as though it were itself a randomized treatment effect. The endpoint is valuable precisely because it gives a standardized description of tumor response, while the single-group design determines how far causal interpretation can go.
16. Safety and Efficacy Are Different Estimands
The efficacy endpoint asks whether a participant achieved a specified tumor-response state. The safety endpoint reported here asks whether a participant experienced a serious adverse event. These outcomes have different definitions, clinical meanings, and denominators even though both can be summarized as participant-level proportions.
| Outcome type | Statistical quantity | Question answered |
|---|---|---|
| Objective response | Proportion with confirmed CR or PR | How many participants met the response criterion? |
| Serious adverse events | Affected / at risk | How many participants experienced the reported safety outcome? |
Keeping these estimands separate is essential. A treatment cannot be summarized adequately by a single numerical "benefit-risk" statistic unless a specific composite estimand has been prespecified and analysed accordingly.
17. Why the Two Data Cutoffs Should Be Kept Separate
Up to 16 months
The first primary ORR entry uses a data cutoff of up to 16 months and specifies confirmed ORR by independent central review.
Up to 23 months
The second primary ORR entry uses a data cutoff of up to 23 months and applies the same confirmed CR/PR definition.
Two cutoffs can answer related but distinct questions. The earlier analysis reflects the information available by the earlier cutoff; the later analysis incorporates a later information set. Treating them as interchangeable can obscure how the evidence evolved over time.
18. Statistical Interpretation of a Single-Arm Response Study
A useful way to think about the design is to separate three questions:
- What happened? Estimate the proportion of participants who achieved confirmed objective response.
- How precisely was it estimated? Quantify uncertainty with an appropriate confidence interval.
- Compared with what? Determine whether there is a prespecified benchmark or concurrent comparator that supports a treatment-effect interpretation.
DESTINY-Gastric02 directly addresses the first question through its ORR endpoint and provides the framework for addressing the second. The third question is constrained by the single-group design because there is no randomized comparator arm in the registry record.
19. Sources
- ClinicalTrials.gov: NCT04014075 — DESTINY-Gastric02.
- PubMed: PMID 42208007.
- PubMed: PMID 41294268.
- PubMed: PMID 37329891.
Continue through the Clinical Biostats knowledge graph
Explore the broader statistical framework behind clinical-trial endpoints, response analysis, confidence intervals, and single-arm study design.
20. Record Summary
DESTINY-Gastric02 is a phase 2, single-group study of trastuzumab deruxtecan in HER2-positive unresectable or metastatic gastric or gastro-esophageal junction adenocarcinoma. The registry reports 79 enrolled participants, one intervention arm, and confirmed objective response rate as the primary endpoint at two data-cutoff time frames: up to 16 months and up to 23 months.
The statistical structure is centered on estimation of a single-group response proportion using a prespecified clinical response definition. Independent central review and RECIST version 1.1 provide the assessment framework, while the absence of a randomized comparator limits direct causal comparison with another treatment. The registry also reports serious adverse events affecting 37 of 79 participants in the trastuzumab deruxtecan arm.
The most important statistical lesson is that an observed response proportion, its precision, and a comparative treatment effect are three different quantities. A rigorous interpretation keeps those distinctions visible while also respecting the specific data cutoff, response definition, analysis population, and study design.