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-Gastric04 is a randomized phase 3 treatment trial comparing trastuzumab deruxtecan with ramucirumab plus paclitaxel in subjects with HER2-positive gastric cancer or gastro-esophageal junction adenocarcinoma after progression on or after a trastuzumab-containing regimen.
| Feature | DESTINY-Gastric04 |
|---|---|
| Phase | Phase 3 |
| Conditions | Gastric Cancer, Adenocarcinoma; Gastroesophageal Junction Adenocarcinoma |
| Design | Randomized, parallel, open-label, treatment trial |
| Allocation | Randomized |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 490 |
| Number of arms | 2 |
| Start date | 2021-05-21 |
| Primary completion date | 2026-08-31 |
| Lead sponsor | Daiichi Sankyo |
| Sponsor type | Industry |
| Registry status | Active, not recruiting |
2. Clinical Question
The primary statistical question is whether overall survival differs between subjects administered trastuzumab deruxtecan and subjects receiving ramucirumab in combination with paclitaxel.
Population
Subjects with HER2-positive gastric cancer or gastro-esophageal junction adenocarcinoma after progression on or after a trastuzumab-containing regimen.
Intervention
Trastuzumab deruxtecan.
Comparator
Ramucirumab in combination with paclitaxel.
Primary question
How does overall survival compare between the randomized treatment groups?
3. Trial Design
Trastuzumab deruxtecan
- Trastuzumab deruxtecan is the study intervention.
- The primary endpoint compares this treatment with the randomized comparator.
Ramucirumab + paclitaxel
- Ramucirumab and paclitaxel constitute the comparator intervention.
- The comparison is performed within the randomized parallel design.
4. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Overall Survival in Participants Who Were Administered Trastuzumab Deruxtecan Compared With Ramucirumab in Combination With Paclitaxel | Overall survival (OS) is defined as the time from date of randomization until death from any cause. | Time from date of randomization until death (due to any cause), up to approximately 36 months |
Why overall survival is a time-to-event endpoint
Overall survival is not simply a proportion of subjects who have died. It incorporates the time at which death occurs. A subject who remains alive at the time of the last available follow-up has not necessarily failed; that observation is typically treated as right-censored in a time-to-event analysis.
This distinction is central to the statistical analysis. Two treatment groups can have the same number of observed deaths but different timing of those deaths. Survival analysis uses that timing information rather than reducing follow-up to a single binary outcome.
5. Statistical Methodology
The registry identifies overall survival as the primary endpoint and defines it from randomization until death from any cause, with follow-up extending to approximately 36 months. The ClinicalTrials.gov record does not report posted statistical-analysis results or a detailed statistical-analysis plan for the endpoint.
Typical analysis of overall survival
For a randomized phase 3 trial with an overall-survival endpoint, the principal analysis would ordinarily use a time-to-event framework. A Kaplan-Meier estimator is commonly used to describe the survival distribution within each randomized treatment group, while a treatment comparison may use a log-rank test and a Cox proportional-hazards model to estimate a hazard ratio.
These are descriptions of standard statistical approaches for this endpoint type. They are not presented as methods that the registry specifically reports for DESTINY-Gastric04.
Here, di represents the number of deaths at event time ti, and ni represents the number of subjects at risk immediately before that time.
Hazard ratio
If a Cox proportional-hazards model were used, its principal treatment-effect measure would be a hazard ratio. The hazard ratio compares the estimated instantaneous event rates between randomized groups over follow-up.
A hazard ratio is a relative time-to-event measure. It is not the same as a relative risk, an absolute survival difference, or the percentage of subjects who benefit.
Censoring
Because overall survival is observed over time, some subjects may still be alive when their available follow-up ends. Those observations are typically censored rather than treated as deaths at the date of last contact. Appropriate handling of censoring is therefore an important component of a valid survival analysis.
Randomization
Randomization creates the principal basis for comparing outcomes between the two treatment strategies. In a randomized trial, treatment assignment is determined before the outcome is observed, reducing the role of measured and unmeasured baseline differences as explanations for treatment-group differences, subject to the usual assumptions of randomized-trial inference.
6. Statistical Methods Explained
Why is overall survival analyzed as a time-to-event outcome?
Because both whether death occurs during observation and when it occurs contain information. A binary analysis that simply classifies subjects as alive or dead at a single date would discard much of the follow-up information.
Why is randomization important for the survival comparison?
Randomization establishes the treatment groups before subsequent outcomes occur. This supports a causal comparison of treatment strategies, assuming the randomized groups are analyzed appropriately and follow-up is handled according to the prespecified design.
What does a Kaplan-Meier curve show?
A Kaplan-Meier curve estimates the probability of remaining alive over time. It also shows how information accumulates as deaths occur and subjects become censored. The curve therefore provides a graphical description of survival rather than a single summary number.
What does a hazard ratio mean?
A hazard ratio compares the estimated instantaneous rates of death between treatment groups. A value below 1 would indicate a lower estimated hazard in the numerator group, while a value above 1 would indicate a higher estimated hazard. It does not directly state the probability that an individual subject will survive.
Why does the proportional-hazards assumption matter?
A conventional Cox proportional-hazards interpretation assumes that the relative hazard is adequately represented by a common ratio over time. If the treatment effect changes substantially over follow-up, a single hazard ratio may not fully describe the clinical pattern. Kaplan-Meier estimates at clinically relevant time points and other survival summaries can then provide additional context.
Why is censoring important?
Subjects who remain alive at their last known follow-up contribute information through that time even though their eventual survival time is not yet observed. Survival methods are designed to use this partial information without automatically classifying those subjects as deaths.
7. Planned Analysis
The registry identifies overall survival as the primary endpoint. It is defined as the time from randomization until death from any cause, with the specified time frame extending up to approximately 36 months.
Primary statistical estimand
Origin: date of randomization
Registry time frame: up to approximately 36 months
For an endpoint of this type, a conventional analysis would typically estimate survival over time with the Kaplan-Meier method and compare randomized groups using a time-to-event treatment effect such as a hazard ratio, often accompanied by a confidence interval and a formal hypothesis test. The precise inferential procedure, model specification, treatment-effect estimand, censoring rules, and multiplicity strategy should be taken from the prespecified statistical analysis plan when available.
8. Interpreting a Future Overall-Survival Result
When a mature randomized survival result is available, the most informative interpretation should combine several quantities rather than relying on a single p-value.
Hazard ratio
Describes the relative difference in the estimated instantaneous death rate between randomized treatment groups under the chosen survival model.
95% confidence interval
Describes statistical uncertainty around the estimated treatment effect under the specified model and sampling framework.
P-value
Addresses compatibility with the null hypothesis under the specified testing procedure. It does not measure the magnitude or clinical importance of the treatment effect.
Absolute survival
Survival probabilities at clinically relevant times provide an absolute perspective that a relative hazard ratio cannot provide by itself.
Hazard ratio is not absolute risk reduction
A hazard ratio below 1 would indicate a lower estimated rate of death in one randomized group, but it would not mean that the same percentage of subjects experienced that reduction in their individual probability of death.
Confidence intervals and precision
A narrow confidence interval generally indicates greater statistical precision than a wide interval, conditional on the analysis framework. The interval does not describe the range of effects experienced by individual subjects.
P-values and effect size
A p-value is a measure associated with a hypothesis-testing procedure. It does not tell the reader how large the treatment effect is. Effect magnitude and statistical uncertainty are better communicated through an effect estimate and its confidence interval, alongside absolute survival measures.
9. What the Registry Does and Does Not Report
| Element | Registry record |
|---|---|
| Trial phase | Phase 3 |
| Randomized allocation | Reported |
| Parallel design | Reported |
| Masking | None |
| Primary endpoint | Overall survival |
| Primary endpoint definition | Time from date of randomization until death from any cause |
| Primary endpoint time frame | Up to approximately 36 months |
| Enrollment | 490 |
| Posted statistical analyses | None |
| Numerical primary efficacy results | Not reported in the registry record |
| Confidence interval for the primary endpoint | Not reported in the registry record |
| P-value for the primary endpoint | Not reported in the registry record |
The absence of a posted numerical analysis is important statistically. A registry entry can establish the planned endpoint and study structure without supplying the event-level or summary data required to quantify a treatment effect.
10. Analysis Populations and the Meaning of Randomization
The record identifies the study as randomized, but the ClinicalTrials.gov record does not specify detailed analysis-population definitions such as an intention-to-treat population, per-protocol population, or safety population.
For a confirmatory randomized trial, preserving the randomized comparison is generally central to the primary efficacy analysis. The exact population and handling of treatment discontinuation, withdrawal, and protocol deviations should be determined from the prespecified statistical analysis plan rather than assumed from the registry's high-level design description.
11. Missing Data and Censoring
Overall survival has a useful property from a statistical perspective: the event is death from any cause, which is generally objectively defined. Nevertheless, incomplete follow-up creates censoring considerations because a subject may be known to be alive through a particular date without having an observed death time.
Right censoring
A subject who has not experienced death by the end of available observation is typically censored at the appropriate follow-up time.
Follow-up completeness
Loss to follow-up can affect the validity of a survival analysis if censoring is related to prognosis in ways not adequately addressed by the analysis.
The registry information does not specify detailed censoring rules, sensitivity analyses, or missing-data and imputation procedures. Those details are important components of a full statistical analysis plan for a time-to-event endpoint.
12. Multiplicity and Interim Analysis
The ClinicalTrials.gov record identifies one registered primary endpoint: overall survival. It does not provide a detailed multiplicity strategy, alpha-spending approach, interim-analysis schedule, stopping boundaries, or hypothesis hierarchy.
| Design topic | What the registry record reports |
|---|---|
| Primary endpoint | Overall survival |
| Multiplicity strategy | Not reported in the ClinicalTrials.gov record |
| Interim efficacy analysis | Not reported in the ClinicalTrials.gov record |
| Alpha-spending method | Not reported in the ClinicalTrials.gov record |
| Stopping boundaries | Not reported in the ClinicalTrials.gov record |
This distinction matters because a p-value cannot be interpreted independently of the hypothesis-testing framework that generated it. If multiple looks at accumulating data or multiple confirmatory hypotheses are part of a design, the type I error framework must account for them.
13. Non-Inferiority, Crossover, and Bayesian Methods
The registry information identifies the trial as a randomized treatment study but does not specify a non-inferiority margin, crossover strategy, factorial structure, or Bayesian analysis.
Non-inferiority margin
No non-inferiority margin is reported in the registry information.
Crossover
No crossover design is reported in the registry information.
Factorial design
The registry describes a parallel design rather than a factorial design.
Bayesian methods
No Bayesian analysis is reported in the registry information.
These distinctions are important because different design structures require different inferential logic. For example, a non-inferiority trial requires interpretation relative to a prespecified margin, whereas a conventional superiority comparison does not use that same decision framework.
14. Safety
The registry information identifies the interventions but does not provide posted numerical serious-adverse-event results by randomized treatment arm in the information summarized here.
| Safety element | Registry information |
|---|---|
| Intervention | Trastuzumab deruxtecan |
| Comparator intervention | Ramucirumab and paclitaxel |
| Serious adverse events by arm | Not reported in the registry information |
| Numerical safety comparison | Not reported in the registry information |
Safety and efficacy answer different statistical questions. The overall-survival endpoint evaluates time to death from any cause, whereas adverse-event analyses generally describe treatment-emergent events, severity, seriousness, exposure, and other safety categories. A treatment effect on overall survival should therefore not be interpreted as a substitute for a separate safety assessment.
15. Trial Timeline
Study start
The registry lists 2021-05-21 as the study start date.
Randomized treatment comparison
The study is registered as a phase 3 randomized, parallel, unmasked treatment trial with 2 arms and planned enrollment of 490 subjects.
Primary completion date
The registry lists 2026-08-31 as the primary completion date.
16. Why This Trial Matters Statistically
DESTINY-Gastric04 is a useful teaching example because the registry design places a classic time-to-event endpoint at the center of a randomized phase 3 comparison. The statistical interpretation therefore depends on understanding how randomization, survival time, censoring, and treatment-effect estimation fit together.
| Concept | How it appears in DESTINY-Gastric04 |
|---|---|
| Randomization | The study uses randomized allocation. |
| Parallel design | The registry identifies a parallel two-arm design. |
| Open-label treatment | The registry reports no masking. |
| Time-to-event analysis | The primary endpoint is time from randomization until death. |
| Kaplan-Meier estimation | A standard method for describing an endpoint of this type. |
| Hazard ratio | A standard relative treatment-effect measure for survival analysis. |
| Censoring | Relevant because subjects may remain alive at the end of observed follow-up. |
| Confidence interval | Needed to describe precision around a future treatment-effect estimate. |
| Hypothesis testing | A formal comparison requires a prespecified inferential framework. |
| Analysis population | The registry information does not provide detailed population definitions. |
17. Limitations and Interpretation Issues
- No posted efficacy analysis: the registry record does not contain a numerical treatment-effect estimate, confidence interval, or p-value for the primary overall-survival endpoint.
- Limited statistical-plan detail: the registry information does not specify the full model, testing procedure, censoring rules, multiplicity strategy, or interim-analysis framework.
- Open-label design: treatment assignment is not masked, which can influence aspects of clinical management and follow-up even though death itself is an objective endpoint.
- Censoring assumptions: survival analysis depends on appropriate handling of right-censored observations and assumptions concerning the censoring mechanism.
- Hazard-ratio assumptions: if a proportional-hazards model is used, a single hazard ratio may not adequately summarize a treatment effect that varies substantially over time.
- Analysis population: detailed definitions for efficacy and safety populations are not reported in the registry information summarized here.
- Follow-up duration: the primary endpoint is specified for a time frame of up to approximately 36 months, so the maturity of the survival analysis depends on the amount and completeness of follow-up actually available.
- Registry versus statistical analysis plan: the ClinicalTrials.gov record establishes the registered study framework but does not substitute for a complete statistical analysis plan.
18. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The registered primary endpoint is a time-to-event outcome: time from randomization until death from any cause. A valid analysis would account for the timing of deaths and censored observations and would report an appropriate treatment-effect estimate with statistical uncertainty.
Clinical interpretation
The clinical question concerns survival following randomization to trastuzumab deruxtecan versus ramucirumab plus paclitaxel in the specified HER2-positive gastric or gastro-esophageal junction adenocarcinoma population.
Keeping these levels separate is important. Statistical evidence describes the strength, magnitude, and precision of the randomized comparison. Clinical interpretation asks what that evidence means in the context of the disease, treatment setting, follow-up, and patient population.
19. A Framework for Reading the Primary Result
The primary endpoint asks about time from randomization until death from any cause. This defines the event and establishes the time origin.
Kaplan-Meier estimates, if reported, would show how the probability of remaining alive changes over follow-up in each randomized group.
A hazard ratio, if reported, would summarize the relative difference in the instantaneous death rate between the treatment groups under the chosen model.
The confidence interval should be interpreted alongside the point estimate. Precision depends on the information available, including the number and timing of events and the amount of follow-up.
The p-value should be interpreted in the context of the prespecified hypothesis, type I error control, interim analyses, and multiplicity strategy rather than as a standalone measure of treatment importance.
20. What a Complete Survival Analysis Would Contain
A mature statistical report for this trial's primary endpoint would ordinarily allow readers to examine the treatment comparison from several complementary perspectives.
| Component | Statistical purpose |
|---|---|
| Number at risk over time | Shows how much follow-up information remains available at successive time points. |
| Kaplan-Meier survival estimates | Describes the estimated survival distribution in each randomized group. |
| Median survival, if estimable | Provides a time-based summary of the survival distribution. |
| Hazard ratio | Summarizes the relative treatment effect under a time-to-event model. |
| Confidence interval | Quantifies uncertainty around the treatment-effect estimate. |
| Formal hypothesis test | Assesses the treatment comparison under the prespecified inferential framework. |
| Absolute survival probabilities | Shows survival at selected time points in an absolute scale. |
| Sensitivity analyses | Assess whether conclusions are robust to important analytical assumptions. |
The ClinicalTrials.gov record summarized here does not provide these numerical outcome analyses. Their appropriate interpretation therefore depends on the eventual statistical report and its prespecified analytical framework.
21. Statistical Concepts in This Trial
The central statistical concepts in DESTINY-Gastric04 are randomized treatment comparison, time-to-event analysis, survival estimation, censoring, hazard ratios, confidence intervals, and hypothesis testing.
Randomized comparison
The treatment groups are formed through randomized allocation, providing the foundation for a causal efficacy comparison.
Survival function
The survival function describes the probability of remaining alive beyond a given time.
Hazard
The hazard represents an instantaneous event rate conditional on remaining event-free up to that time.
Censoring
Subjects without an observed death during available follow-up contribute partial information through their censoring time.
22. Sources
- ClinicalTrials.gov: NCT04704934 — DESTINY-Gastric04.
Continue through the Clinical Biostats library
Explore statistical tutorials, calculators, and additional clinical-trial analyses covering randomized designs, survival methods, and clinical-trial methodology.
23. Record Summary
DESTINY-Gastric04 is registered as a phase 3 randomized, parallel, unmasked treatment trial with 490 enrolled subjects and two treatment arms. The primary endpoint is overall survival, defined as the time from the date of randomization until death from any cause, with a registry time frame of up to approximately 36 months.
Statistically, the key feature of the study is therefore a randomized time-to-event comparison. A complete interpretation of the primary endpoint would require the observed survival data, an appropriate survival-analysis framework, an effect estimate such as a hazard ratio, its confidence interval, and the prespecified hypothesis-testing procedure. The ClinicalTrials.gov record does not contain posted statistical analyses or numerical results for that endpoint.