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-PanTumor02 is a phase 2, non-randomized, unmasked, parallel-design study evaluating trastuzumab deruxtecan in patients with selected HER2-expressing tumors. The registry lists confirmed objective response rate as the primary endpoint, assessed over an average time frame of approximately 6 months.
| Feature | DESTINY-PanTumor02 |
|---|---|
| Phase | Phase 2 |
| Study title | A Phase 2 Study of T-DXd in Patients With Selected HER2 Expressing Tumors |
| Allocation | Non-randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 477 |
| Number of arms | 12 |
| Intervention | Trastuzumab deruxtecan |
| Lead sponsor | AstraZeneca |
| Sponsor type | Industry |
| Status | Active, not recruiting |
| NCT ID | NCT04482309 |
2. Clinical Question
The study asks whether trastuzumab deruxtecan produces confirmed tumor responses in patients with selected HER2-expressing tumors. The registry divides the study population into two broad parts, reflecting different disease eligibility definitions.
Population
Part 1 includes bladder, biliary tract, cervical, endometrial, ovarian, and pancreatic cancer, rare tumors, and any tumor type excluding breast, gastric, and colorectal cancer. Part 2 includes HER2-expressing or amplified solid tumors excluding breast, gastric, and colorectal cancer.
Intervention
Trastuzumab deruxtecan, listed in the registry as a drug intervention.
Comparator
The registry describes a non-randomized study and lists no comparator intervention.
Primary question
What percentage of patients achieve a confirmed complete or partial response according to RECIST 1.1?
3. Trial Design
The statistical character of this design is important. A non-randomized study can estimate how frequently a response is observed among treated patients, but it does not create the same randomized treatment-versus-control contrast that would be available in a comparative randomized trial. Consequently, the primary endpoint is naturally interpreted as a response proportion rather than as a treatment-effect estimate such as a hazard ratio or risk ratio comparing randomized groups.
4. Study Population
Part 1
The registry defines Part 1 as including bladder, biliary tract, cervical, endometrial, ovarian, and pancreatic cancer, rare tumors, and any tumor type excluding breast, gastric, and colorectal cancer.
Part 2
The registry defines Part 2 as HER2-expressing or amplified solid tumors excluding breast, gastric, and colorectal cancer.
| Study component | Registry population definition |
|---|---|
| Part 1 | Bladder, Biliary Tract, Cervical, Endometrial, Ovarian, Pancreatic Cancer, Rare Tumors, Any Tumor Type Excluding Breast, Gastric, Colorectal Cancer |
| Part 2 | HER2 Expressing/Amplified Solid Tumors Excluding Breast, Gastric, Colorectal Cancer |
This broad population is statistically relevant because objective response is not being evaluated in a single homogeneous disease setting. The study spans multiple tumor types and therefore combines a treatment question with a population-definition question: the response proportion describes the enrolled population and should not automatically be interpreted as a universal response probability for every HER2-expressing solid tumor.
5. Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Objective Response Rate (ORR) | Confirmed ORR per RECIST 1.1 is the percentage of patients with Complete Response or Partial Response that is subsequently confirmed. | An average of approximately 6 months |
What makes this endpoint "confirmed"?
The registry does not define ORR simply as the percentage of patients with an initial tumor response. It specifies that complete or partial responses must subsequently be confirmed. This distinction is important because a single radiographic observation can be transient or uncertain; requiring confirmation makes the endpoint more stringent than counting every initial response assessment as a final response.
Why RECIST 1.1 matters
RECIST 1.1 supplies the response framework used to classify tumor response. The endpoint therefore has two linked components: a standardized radiologic response definition and a confirmation requirement. The resulting statistic is a proportion of patients meeting that definition during the specified assessment period.
6. Statistical Methodology
The registry identifies confirmed ORR as the primary endpoint but does not post a formal statistical analysis for that endpoint. For a binary response endpoint, the core statistical quantity is the proportion of evaluable patients who meet the prespecified response definition.
The resulting proportion is commonly expressed as a percentage. The denominator is a critical part of the analysis because the interpretation depends on which patients are included in the response-evaluable population.
For a single-arm or non-randomized study, statistical inference around ORR generally focuses on a confidence interval for a binomial proportion and, when specified by the protocol, a hypothesis test against a prespecified response-rate benchmark. The registry record identifies the endpoint and its definition but does not post such a formal benchmark or analysis procedure.
Binomial framework
Each patient's response classification can be represented as a binary outcome: confirmed response or no confirmed response. If \(x\) patients among \(n\) evaluated patients have confirmed responses, the observed response rate is \(x/n\).
Here, \(X\) represents the number of patients with a confirmed response, \(n\) represents the number evaluated, and \(p\) represents the underlying response probability for the population represented by the analysis.
Confidence intervals
A confidence interval around ORR quantifies statistical uncertainty in the estimated response proportion. For a binomial endpoint, exact or other binomial-based interval methods can be used. The appropriate method should be taken from the prespecified statistical analysis plan when one is available.
The width of the confidence interval depends strongly on the amount of information available. More response-evaluable patients generally provide greater precision, while smaller analysis populations produce more uncertainty around the observed proportion.
Why a hazard ratio is not the natural primary measure here
The registry's primary endpoint is a response proportion assessed over approximately 6 months, not a time-to-event endpoint. A hazard ratio would therefore answer a different question. Hazard ratios are designed for comparing event rates over time, whereas ORR asks whether patients achieved a defined tumor response.
7. Statistical Methods Explained
Why is ORR analyzed as a proportion?
Because each patient can ultimately be classified according to whether a confirmed complete or partial response occurred. The primary summary is therefore the number of confirmed responders divided by the relevant response-evaluable population.
Why must the response be confirmed?
Confirmation reduces the influence of an isolated initial assessment. A patient is counted as a confirmed responder only when the response subsequently satisfies the registry's confirmation requirement. This makes the endpoint more specific to sustained radiologic evidence than an unconfirmed first response.
What does a confidence interval for ORR mean?
If an observed response rate were accompanied by a 95% confidence interval, that interval would describe uncertainty in the estimated population response proportion under the statistical model used. It would not mean that 95% of individual patients have response probabilities lying inside that interval.
Why is the denominator so important?
A response rate is inseparable from the population used to calculate it. The same number of responders could produce different percentages if the response-evaluable denominator changed. In a multicohort study, eligibility and evaluability definitions are therefore central to interpretation.
What would a hypothesis test against a response benchmark do?
In a single-arm phase 2 setting, a protocol may specify a null response probability and test whether the observed response rate provides sufficient evidence that the underlying response probability exceeds that benchmark. Such a test addresses evidence against a predefined reference value; it does not create the causal interpretation available from randomization.
Why is non-randomized design important statistically?
Without random allocation, the study can characterize response under the treatment being evaluated, but differences in patient characteristics across tumor types or other clinical factors are not balanced through randomization. An observed response proportion therefore should not be interpreted as a randomized comparative treatment effect.
8. Planned Analysis
The registry identifies Objective Response Rate (ORR) as the primary endpoint. Confirmed ORR per RECIST 1.1 is the percentage of patients with complete response or partial response that is subsequently confirmed, with an average time frame of approximately 6 months.
Primary statistical quantity
Percentage of patients with a subsequently confirmed complete or partial response.
Assessment time frame: an average of approximately 6 months.
For this endpoint, the usual statistical analysis would summarize the observed proportion of confirmed responders and accompany it with an appropriate confidence interval for a binomial proportion. In a non-randomized phase 2 study, a formal single-arm hypothesis test may also be prespecified against a clinically or historically defined response benchmark, but the registry does not report such a benchmark or formal testing procedure.
Primary estimand
The central quantity is the proportion of patients meeting the confirmed RECIST 1.1 response definition within the specified assessment framework.
Effect measure
ORR is a proportion rather than a relative treatment effect comparing two randomized groups.
Statistical uncertainty
A binomial confidence interval would describe uncertainty around the estimated response proportion.
Registry reporting
No formal statistical analyses are posted to ClinicalTrials.gov for the primary endpoint.
The distinction between an endpoint definition and an inferential procedure is important. The registry clearly specifies what constitutes a confirmed response and when the endpoint is assessed. It does not, in the record represented here, specify the numerical confidence interval method, a null response-rate benchmark, or a posted hypothesis-test result.
9. Interpreting ORR Correctly
ORR describes the percentage of patients who achieve a confirmed complete or partial response according to the specified RECIST 1.1 framework. It is therefore a direct measure of observed tumor response in the study population.
ORR alone does not describe how long patients live, how long disease remains controlled, or whether treatment is superior to another therapy. Those are different clinical questions requiring different endpoints and analyses.
Two treatments could have the same ORR while producing responses of very different durations. Conversely, a treatment could have a lower response proportion but produce particularly durable responses in a subset of patients. ORR therefore captures response frequency, not the full longitudinal experience of treatment.
An observed response percentage is an estimate rather than an exact measurement of an underlying population probability. A confidence interval communicates the statistical precision of that estimate. A narrow interval indicates greater precision than a wide interval under the same inferential framework.
10. The Role of the 12-Arm Design
The registry lists 12 arms within a non-randomized parallel design. This is statistically different from a conventional two-arm randomized controlled trial. The presence of multiple arms means that the study contains multiple treatment cohorts, but the registry information represented here does not provide separate enrollment counts, response estimates, or inferential results for those arms.
| Design feature | Statistical implication |
|---|---|
| 12 arms | Multiple treatment cohorts are represented within the study. |
| Non-randomized allocation | There is no randomized balancing of patient characteristics between treatment groups. |
| Parallel design | Study groups are structured as parallel cohorts rather than as a crossover sequence. |
| No masking | The registry identifies the study as unmasked. |
Multiple arms can be useful for evaluating a treatment across different tumor populations, but they also create a more complex interpretive environment. A response proportion calculated across a broad population can conceal clinically important differences between individual tumor types. Conversely, response rates from small individual cohorts can be statistically imprecise.
11. What a Non-Randomized ORR Study Can Establish
A non-randomized response study is well suited to estimating the frequency with which a predefined response occurs under the study treatment. Its evidentiary structure differs from a randomized comparative trial because the study does not use random allocation to make treatment groups exchangeable at baseline.
Directly informative
The proportion of treated patients achieving the prespecified confirmed response definition.
Requires caution
Comparisons of response across different tumor populations may reflect differences in the populations as well as differences in response.
Not a randomized contrast
An observed ORR does not by itself quantify the incremental effect of trastuzumab deruxtecan versus another treatment.
Population-specific
The estimated response proportion applies most directly to the population represented by the relevant study cohort and eligibility criteria.
This distinction is particularly important for an umbrella-like tumor population. Part 1 and Part 2 have different registry definitions, and the overall study spans multiple disease categories. Statistical interpretation should therefore remain anchored to the population in which a particular response estimate is calculated.
12. Time Frame and Endpoint Interpretation
The registry specifies an average time frame of approximately 6 months for the primary ORR endpoint. This makes the time window part of the endpoint definition rather than merely an administrative reporting date.
Study start
The registry lists August 18, 2020 as the study start date.
Primary endpoint assessment frame
Confirmed objective response rate is specified over an average time frame of approximately 6 months.
Primary completion date
The registry lists March 23, 2027 as the primary completion date.
It is useful to distinguish the endpoint's assessment frame from the overall study timeline. A study can continue beyond the period used to define or summarize a particular endpoint. The primary completion date therefore should not be interpreted as the time at which the approximately 6-month ORR assessment itself occurs for every participant.
13. Statistical Precision in a Multitumor Study
Precision is a central issue when interpreting a response proportion. Suppose \(n\) patients are evaluated and \(x\) meet the confirmed response definition. The observed ORR is \(x/n\). Even if the observed proportion is numerically straightforward, the underlying population response probability remains uncertain.
This familiar binomial approximation illustrates why sample size matters: with more response-evaluable patients, sampling variability generally decreases.
The formula is an educational approximation rather than a claim about the exact interval procedure used for DESTINY-PanTumor02. For small samples or proportions near the boundaries, exact or alternative binomial interval methods may be preferable.
In a study with multiple tumor populations, this issue becomes even more important. The precision of an overall response estimate depends on the total number evaluated, while the precision of a tumor-specific estimate depends on the number evaluated within that particular population. Pooling populations can increase numerical precision while simultaneously making the clinical interpretation broader.
14. Statistical Questions Raised by the Design
| Question | Why it matters |
|---|---|
| Who is included in the ORR denominator? | The response percentage depends directly on the response-evaluable population. |
| How are confirmed responses defined? | Confirmation affects which patients qualify as responders. |
| How are multiple tumor populations summarized? | Pooling can improve precision but may reduce disease-specific interpretability. |
| Is there a response-rate benchmark? | A formal single-arm hypothesis test requires a prespecified reference value. |
| Which confidence interval method is used? | Different binomial interval methods can produce different interval limits, particularly with smaller samples. |
| Are individual-arm analyses planned? | The study contains 12 arms, but the registry record represented here does not post arm-specific statistical results. |
These questions demonstrate why a statistical analysis cannot be reduced to a single percentage. The endpoint definition, analysis population, statistical model, confidence interval, and study design all contribute to what an ORR estimate actually means.
15. Limitations
- Non-randomized design: the study does not use random allocation to create balanced treatment groups, so an observed response proportion is not a randomized comparative treatment effect.
- Broad tumor population: the registry includes multiple tumor types and two distinct population definitions, making disease-specific interpretation important.
- Multiple arms: the study contains 12 arms, but the registry information represented here does not provide arm-specific response estimates or enrollment counts.
- Response endpoint: ORR captures confirmed tumor response but does not by itself measure survival or the duration of response.
- Confirmation requirement: the endpoint requires subsequent confirmation, so an initial response observation is not necessarily sufficient to classify a patient as a responder.
- Analysis population: the registry defines the endpoint but does not post the denominator used in a completed ORR calculation.
- Statistical procedure: the registry identifies the primary endpoint but does not post a formal statistical analysis for it.
- Time frame: the endpoint is specified over an average period of approximately 6 months, which should not be conflated with the overall study duration.
- Generalizability: the applicable population is determined by the study's specific tumor and HER2-related eligibility definitions.
16. Why This Trial Matters Statistically
DESTINY-PanTumor02 is a useful statistical teaching case because its primary endpoint is conceptually simpler than a time-to-event endpoint but its study structure is more complex than a conventional single-disease phase 2 study.
| Concept | How it appears in DESTINY-PanTumor02 |
|---|---|
| Phase 2 development | The study is registered as phase 2. |
| Non-randomized design | Allocation is listed as non-randomized. |
| Parallel design | The design model is listed as parallel. |
| Multiple arms | The registry lists 12 arms. |
| Binary endpoint | Confirmed ORR classifies patients according to whether a confirmed complete or partial response occurred. |
| RECIST 1.1 | The primary endpoint uses confirmed ORR per RECIST 1.1. |
| Binomial inference | ORR can naturally be represented as a binomial proportion with confidence-interval methods. |
| Analysis denominator | The percentage depends on the patients included in the response-evaluable analysis. |
| Population heterogeneity | The registry defines multiple tumor populations across Part 1 and Part 2. |
| Confirmation | Complete or partial responses must subsequently be confirmed. |
The broader statistical lesson is that an endpoint can be mathematically straightforward while still requiring careful interpretation. ORR is easy to calculate once the numerator and denominator are known, but deciding what population the proportion describes and what inference can reasonably be drawn from it requires attention to design.
17. Trial Status and Study Timeline
The study's active, not recruiting status and future primary completion date are relevant when interpreting the maturity of registry information. A study can have a substantial enrollment target while still having an incomplete statistical results record. The absence of a posted statistical analysis therefore does not itself establish a treatment outcome in either direction.
18. Sources
- ClinicalTrials.gov: NCT04482309.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through the Clinical Biostats clinical-trial library
Explore additional clinical-trial analyses and statistical resources covering study design, endpoints, estimation, and interpretation.
19. Record Summary
DESTINY-PanTumor02 is a phase 2, non-randomized, unmasked, parallel-design study with 12 arms and an enrollment of 477. The registry identifies trastuzumab deruxtecan as the intervention and defines populations across Part 1 and Part 2 involving selected HER2-expressing or amplified tumors. Its primary endpoint is confirmed objective response rate per RECIST 1.1, defined as the percentage of patients with complete response or partial response that is subsequently confirmed, assessed over an average time frame of approximately 6 months.
From a statistical perspective, the central analysis is a binomial response-proportion problem rather than a randomized time-to-event comparison. The most important interpretive elements are therefore the confirmed-response definition, the response-evaluable denominator, the confidence interval around the observed proportion, and the limitations imposed by non-randomized allocation and a broad multitumor population.