This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Trial-specific numerical results on this page are restricted to the ClinicalTrials.gov record.
1. Trial at a Glance
DESTINY-Gastric01 was a randomized, parallel-group, open-label phase 2 trial evaluating DS-8201a versus physician's choice in participants with HER2-expressing advanced gastric or gastroesophageal junction adenocarcinoma. The registry reports 233 participants, four arms, two registered primary endpoints, and a formal Cochran-Mantel-Haenszel analysis for the primary objective response rate endpoint.
| Feature | DESTINY-Gastric01 |
|---|---|
| Phase | Phase 2 |
| Population | Participants with HER2-expressing advanced gastric or gastroesophageal junction adenocarcinoma |
| Design | Randomized, parallel-group, open-label |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Interventions | DS-8201a and physician's choice |
| Primary endpoints | Objective response rate and best overall response, both assessed by independent central review |
| Formal statistical analysis posted | Cochran-Mantel-Haenszel test for the primary objective response rate endpoint |
| ClinicalTrials.gov | NCT03329690 |
| Lead sponsor | Daiichi Sankyo Co., Ltd. |
| Sponsor type | Industry |
2. Clinical Question
The principal statistical question was whether participants assigned to DS-8201a had a different objective response rate from participants assigned to physician's choice in the primary cohort, using an intention-to-treat analysis and a Cochran-Mantel-Haenszel test with region as a stratification factor.
Population
Participants with HER2-expressing advanced gastric or gastroesophageal junction adenocarcinoma.
Intervention
DS-8201a.
Comparator
Physician's choice. The registry separately identifies irinotecan and paclitaxel physician's-choice groups in the safety data.
Primary question
Does DS-8201a produce a different objective response rate from physician's choice in the primary cohort under the prespecified superiority analysis?
3. Trial Design
DS-8201a
- Drug intervention.
- Primary objective response rate assessed in the primary cohort.
- Response assessed by independent central imaging review using RECIST version 1.1.
Physician's Choice
- Drug comparator.
- Registry safety data identify irinotecan and paclitaxel physician's-choice groups.
- Formal primary comparison is reported as DS-8201a versus physician's choice overall.
The trial began on November 2, 2017 and reached primary completion on November 8, 2019. The registry reports the trial as completed.
4. Analysis Populations and Stratification
The primary objective response rate analysis was explicitly conducted in the Primary Cohort in the Intent-to-Treat Analysis Set. The reported statistical analysis used a Cochran-Mantel-Haenszel test with region as a stratification factor.
| Feature | Registry information |
|---|---|
| Primary efficacy population | Primary Cohort in the Intent-to-Treat Analysis Set |
| Analysis principle | Intention-to-treat |
| Formal comparison | DS-8201a versus Physician's Choice Overall |
| Statistical method | Cochran-Mantel-Haenszel test |
| Stratification factor | Region |
| Hypothesis type | Superiority |
5. Primary Endpoints
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Objective response rate | The percentage of participants who achieved a best overall response of Complete Response (CR) or Partial Response (PR), assessed by independent central imaging review 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. The registry definition also identifies unconfirmed ORR. | Baseline to date of first documented objective response (CR or PR), up to 36 months postdose |
| Best overall response | The best overall response (BOR) recorded from the start of study treatment until the end of treatment, including complete response (CR), partial response (PR), stable disease (SD), progressive disease (PD), and not evaluable (NE), as assessed by independent central imaging review based on RECIST version 1.1. | Baseline to date of first documented objective response, up to 36 months postdose |
Both registered primary endpoints are binary outcomes in the registry's normalized endpoint classification. The first endpoint directly asks whether a participant achieved an objective response. The second captures the participant's best overall response category.
6. Statistical Methodology
Cochran-Mantel-Haenszel test
The formal statistical analysis posted for the primary objective response rate endpoint used the Cochran-Mantel-Haenszel (CMH) test. The CMH framework is designed for categorical outcomes when the comparison is made across strata. Instead of simply pooling all participants into one unadjusted two-group comparison, the method accounts for the prespecified stratification factor.
Here, the reported stratification factor was region. The method therefore evaluates the treatment-group association with response while accounting for the region strata.
Intention-to-treat analysis
The registry identifies the primary objective response rate analysis as an Intent-to-Treat Analysis Set). The practical statistical principle is that the randomized treatment assignment remains the basis for the primary efficacy comparison rather than selectively removing participants because of treatment discontinuation or other post-randomization events.
Stratified analysis
Stratification is useful when a trial deliberately accounts for an important design factor. In this trial, region was used as the stratification factor in the CMH analysis. The purpose is not to create separate trials within each region, but to obtain a treatment comparison that incorporates the stratified structure.
Categorical endpoint analysis
Objective response is fundamentally different from a continuous outcome such as tumor size measured on a numerical scale. Participants are classified according to whether they achieved a response category under the prespecified response criteria. The CMH method therefore operates on categorical response information rather than on a continuous mean difference.
A CMH test evaluates whether the treatment groups differ in the distribution of a categorical outcome after accounting for the specified stratification structure. It does not itself provide a hazard ratio or a median survival time.
7. Primary Objective Response Rate Result
ClinicalTrials.gov reports a formal statistical analysis for the primary objective response rate endpoint. The comparison was between DS-8201a and physician's choice overall, using the primary cohort in the intention-to-treat analysis set.
Cochran-Mantel-Haenszel test
Hypothesis type: Superiority
Analysis: DS-8201a vs Physician's Choice Overall
Stratification factor: Region
| Primary endpoint | Population | Method | Stratification | P-value |
|---|---|---|---|---|
| Objective response rate based on independent central review | Primary Cohort, Intent-to-Treat Analysis Set | Cochran-Mantel-Haenszel | Region | <0.0001 |
The reported P < 0.0001 indicates that the prespecified CMH comparison produced a very small p-value under the statistical testing framework used for the primary objective response endpoint. In descriptive terms, the observed categorical response pattern was sufficiently inconsistent with the null hypothesis to meet the reported superiority analysis criterion.
The p-value does not measure the size of the treatment effect. It does not tell us how many percentage points higher the response rate was, nor does it quantify the probability that DS-8201a is effective.
The analysis is also specific to the stated Primary Cohort and intention-to-treat population. The use of a stratified CMH test means the reported comparison should be understood in the context of the region-stratified analysis rather than as an unadjusted pooled test.
8. Best Overall Response
Best overall response is the second registered primary endpoint. The registry defines it as the best response recorded from the start of study treatment until the end of treatment, including CR, PR, SD, PD, and NE, based on independent central imaging review and RECIST version 1.1.
| Endpoint | Formal statistical analysis posted | Registry information |
|---|---|---|
| Best overall response | No separate statistical analysis is included in the posted results record | Results are posted for the endpoint, but the ClinicalTrials.gov record does not provide a formal comparative test, effect estimate, confidence interval, or p-value |
For a multicategory response endpoint, descriptive response frequencies can be presented across CR, PR, SD, PD, and NE. A formal comparison can use categorical-data methods appropriate to the prespecified analysis, including a stratified approach when the design calls for adjustment by strata. However, because the ClinicalTrials.gov record does not report a separate formal statistical analysis for this endpoint, no comparative test result is attributed to it here.
9. Statistical Methods Explained
Why was a Cochran-Mantel-Haenszel test used?
The primary endpoint is categorical: participants either achieved an objective response according to the registered response definition or did not. The trial also specifies region as a stratification factor. The CMH test provides a way to compare treatment groups while accounting for that stratified structure.
What does intention-to-treat mean here?
The primary objective response analysis was conducted in the intention-to-treat analysis set. The central idea is that randomized participants remain part of the treatment comparison according to their randomized assignment rather than being removed simply because their subsequent treatment experience differed from the intended treatment pathway.
Why does stratification matter?
Randomized trials may use stratification to balance or account for important design factors. In this analysis, region was the stratification factor. The statistical comparison therefore considers response across the region strata rather than treating all participants as if the stratified design had not existed.
What does P < 0.0001 mean?
It means the reported test statistic produced a p-value below 0.0001 under the null hypothesis and the specified CMH testing framework. A small p-value provides evidence against the null hypothesis in that framework. It is not a direct measure of the magnitude or clinical importance of the treatment effect.
Why is a p-value not an effect size?
A p-value depends on both the observed data and the amount of information available for testing. A very small p-value can accompany different effect sizes in different datasets. To understand the magnitude of a treatment effect, one should also examine quantities such as response proportions, risk differences, relative risks, odds ratios, and confidence intervals when those measures are reported.
What does independent central review contribute?
The primary response endpoint was assessed by independent central imaging review. Central review creates a standardized assessment framework for radiologic response and reduces reliance on treatment-site assessments. It does not eliminate all uncertainty in endpoint classification, but it provides a common review process for the response determination.
Why does RECIST version 1.1 matter?
The registry defines objective response using RECIST version 1.1, with complete response requiring disappearance of all target lesions and partial response requiring at least a 30% decrease in the sum of diameters of target lesions. The statistical endpoint therefore depends on these prespecified clinical response classifications rather than an unspecified notion of tumor shrinkage.
10. Endpoint Construction: ORR vs Best Overall Response
The two registered primary endpoints are related but are not identical. Objective response rate reduces the response assessment to whether a participant achieved CR or PR. Best overall response retains the broader response categories recorded during treatment.
| Concept | Objective response rate | Best overall response |
|---|---|---|
| Basic structure | Binary response outcome | Best response category |
| Response categories explicitly used | CR or PR for objective response | CR, PR, SD, PD, NE |
| Assessment | Independent central imaging review | Independent central imaging review |
| Response criteria | RECIST version 1.1 | RECIST version 1.1 |
| Formal analysis reported | CMH test | No separate formal analysis reported |
This distinction is statistically important because collapsing a multicategory outcome into a binary endpoint changes the information being analyzed. ORR focuses specifically on objective tumor response, whereas best overall response describes the broader distribution of response categories.
11. Safety Results
The ClinicalTrials.gov record provides serious adverse event counts by treatment arm. These are reported as affected participants divided by participants at risk. Because the ClinicalTrials.gov record does not provide a separate formal statistical comparison of serious adverse events, the figures below are presented descriptively.
| Group | Serious adverse events affected / at risk |
|---|---|
| DS-8201a | 58/125 |
| Physician's Choice — Irinotecan | 15/55 |
| Physician's Choice — Paclitaxel | 1/7 |
| Physician's Choice — Overall | 16/62 |
| Exploratory: Naïve HER2 IHC 2+/ISH-, DS- | 6/20 |
| Exploratory: Naïve HER2 IHC 1+, DS-8201a | 11/24 |
The exploratory groups should also be kept separate from the primary randomized comparison. Their labels identify specific exploratory populations, and their affected/at-risk counts should not be combined with the main randomized-arm figures.
12. Trial Timeline
Trial start
DESTINY-Gastric01 began enrollment under the randomized phase 2 design.
Primary completion
The registry records primary completion on November 8, 2019.
Registry status
The ClinicalTrials.gov record identifies the study as completed and reports results.
13. Understanding the Primary Statistical Test
The CMH framework preserves the categorical nature of the endpoint while incorporating the region strata specified in the analysis.
What the test establishes
The formal analysis establishes that the reported treatment-group response comparison was statistically significant under the specified CMH framework, with P < 0.0001.
What the test does not establish by itself
- It does not provide the magnitude of the response-rate difference.
- It does not provide an individual-patient probability of response.
- It does not establish how long a response lasted.
- It does not provide an overall-survival or progression-free-survival estimate.
- It does not quantify the safety profile by itself.
Those questions require additional endpoint-specific data. The registry extract does not contain those additional efficacy estimates, so they are intentionally not reconstructed here.
14. Multiplicity and Endpoint Hierarchy
The registry identifies two primary endpoints. It also reports one formal statistical analysis, corresponding to the objective response rate endpoint. The ClinicalTrials.gov record does not describe an alpha-allocation scheme, hierarchical testing procedure, multiplicity adjustment across the two primary endpoints, or an interim-analysis plan.
| Endpoint / feature | Registry information reported |
|---|---|
| Objective response rate | Primary endpoint; formal CMH analysis posted |
| Best overall response | Primary endpoint; no separate formal statistical analysis reported |
| Hypothesis type | Superiority |
| Multiplicity adjustment | Not reported in the ClinicalTrials.gov record |
| Interim analysis | Not reported in the ClinicalTrials.gov record |
15. Missing Data and Analysis Considerations
The registry identifies the primary objective response analysis as an intention-to-treat analysis, but the ClinicalTrials.gov record does not specify an imputation method for missing response assessments.
This matters because response endpoints depend on evaluability and assessment timing. A rigorous analysis plan needs to specify how participants without an evaluable assessment are handled, how treatment discontinuation affects assessment, and whether any sensitivity analyses are performed. Those details are not reported here and therefore are not inferred.
Known
The primary objective response rate analysis used the primary cohort in the intention-to-treat analysis set.
Not specified here
The ClinicalTrials.gov record does not state a missing-data imputation method or sensitivity-analysis strategy.
16. Design Features Not Reported in the Supplied Data
Several statistical design questions are important when evaluating a randomized oncology trial, but they should not be filled in from general knowledge when the task is restricted to the ClinicalTrials.gov record.
| Design topic | Status in the ClinicalTrials.gov record |
|---|---|
| Non-inferiority margin | Not applicable to the reported superiority hypothesis; no non-inferiority margin is reported |
| Crossover | Not reported in the ClinicalTrials.gov record |
| Factorial design | Not reported; the registered design model is parallel |
| Interim analysis | Not reported |
| Alpha spending | Not reported |
| Missing-data imputation | Not reported |
| Bayesian methods | Not reported |
| Stratification | Region is explicitly reported for the CMH analysis |
This distinction is important for reproducibility. A statistical analysis page should separate documented methods from methods that would merely be plausible for this type of study.
17. Limitations
- Limited numerical result reporting: the ClinicalTrials.gov record reports the primary CMH p-value but does not provide the corresponding response percentages, event counts, effect estimate, or confidence interval.
- Second primary endpoint: best overall response has posted results, but the ClinicalTrials.gov record does not include a separate formal statistical analysis for that endpoint.
- No survival endpoints in the registry-reported primary endpoint data: the ClinicalTrials.gov record identifies binary primary endpoints rather than time-to-event primary endpoints.
- Safety comparison: serious adverse events are reported as affected/at-risk counts without a formal comparative test or confidence interval.
- Missing-data methods: the ClinicalTrials.gov record does not specify how missing or unevaluable response assessments were handled statistically.
- Multiplicity: the ClinicalTrials.gov record identifies two primary endpoints but do not specify the multiplicity-control procedure used across them.
- Analysis detail: the registry extract does not provide the CMH test statistic, stratum-specific tables, effect-size estimate, or confidence interval.
- Exploratory populations: the registry-reported exploratory serious-adverse-event groups should not be interpreted as substitutes for the primary randomized comparison.
18. Why This Trial Matters Statistically
DESTINY-Gastric01 provides a useful teaching example of how a randomized oncology trial can use a categorical primary endpoint rather than a survival endpoint. The central statistical lesson is the connection between the endpoint definition, the analysis population, the stratification factor, and the hypothesis test.
| Concept | How it appears in DESTINY-Gastric01 |
|---|---|
| Randomization | The trial uses randomized allocation in a parallel design. |
| Intention-to-treat analysis | The primary objective response rate analysis uses the Primary Cohort in the ITT Analysis Set. |
| Binary endpoint | Objective response rate is classified as a binary endpoint in the ClinicalTrials.gov record. |
| Cochran-Mantel-Haenszel test | The formal primary ORR analysis uses the CMH test. |
| Stratified analysis | Region is used as the stratification factor in the CMH analysis. |
| Superiority hypothesis | The formal analysis is identified as a superiority test. |
| Independent central review | Objective response and best overall response are assessed by independent central imaging review. |
| RECIST version 1.1 | The response definitions are based on RECIST version 1.1. |
| p-value interpretation | The primary formal result is reported as P < 0.0001 without a registry-reported effect estimate. |
| Safety analysis | Serious adverse events are reported descriptively by randomized treatment group. |
19. Clinical Biostats Interpretation of the Evidence
The primary objective response rate endpoint was formally compared between DS-8201a and physician's choice overall using a Cochran-Mantel-Haenszel test, with region as a stratification factor, in the primary cohort of the intention-to-treat analysis set. The reported p-value was <0.0001, and the registered hypothesis type was superiority.
The registry-reported statistical analysis does not contain the response percentages, response event counts, risk difference, relative risk, odds ratio, confidence interval, or CMH test statistic. Therefore, the size and precision of the treatment effect cannot be numerically reconstructed without adding external trial results.
A p-value addresses evidence against a null hypothesis within a specified statistical framework. It does not describe clinical magnitude. For a binary response endpoint, an informative treatment comparison ordinarily includes the response proportions in each group and an effect measure with a confidence interval. Those numerical quantities are not included in the ClinicalTrials.gov record.
The primary endpoint was analyzed in the primary cohort using the intention-to-treat analysis set. That specification is part of the statistical result. A result from a narrower evaluable population would answer a different question and could be affected differently by post-randomization exclusions.
20. Primary Endpoint Analysis: A Worked Statistical Reading
A useful way to read the reported result is to move through the analysis in the same order as the statistical design.
Step 1 — Define the endpoint. Objective response means achieving CR or PR according to the registered RECIST version 1.1 definitions.
Step 2 — Define the population. The formal analysis uses the primary cohort in the intention-to-treat analysis set.
Step 3 — Define the comparison. The groups compared in the posted analysis are DS-8201a and physician's choice overall.
Step 4 — Account for the design. Region is used as the stratification factor.
Step 5 — Interpret the test. The resulting p-value is reported as <0.0001. This supports rejection of the null hypothesis under the reported testing framework, but it does not provide the treatment-effect magnitude.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Statistical Calculators
23. Sources
- ClinicalTrials.gov: NCT03329690 — DESTINY-Gastric01.
- Publication: PubMed record for PMID 32469182.
- Publication: PubMed record for PMID 42208007.
- Publication: PubMed record for PMID 36379002.
Continue through the Clinical Biostats statistical library
Explore the statistical methods and calculators that connect randomized trial design, categorical endpoints, stratified analysis, and hypothesis testing.
24. Record Summary
DESTINY-Gastric01 is a useful statistical case study because its primary efficacy evidence is based on a binary response endpoint rather than a time-to-event endpoint. The trial used randomized allocation, an intention-to-treat analysis population, independent central imaging review, RECIST version 1.1 response definitions, and a Cochran-Mantel-Haenszel test with region as the stratification factor. The formal primary objective response rate analysis reported P < 0.0001 under a superiority hypothesis.
The most important interpretive limitation is that the ClinicalTrials.gov record does not include the numerical response proportions, effect estimate, or confidence interval. Consequently, the statistical conclusion can be described precisely as a significant stratified categorical comparison, but the magnitude and precision of the treatment effect cannot be reconstructed from the ClinicalTrials.gov record alone.