DESTINY-Lung03: Complete Statistical Analysis of T-DXd and Immunotherapy in Advanced or Metastatic HER2+ NSCLC

1. Trial at a Glance

DESTINY-Lung03 is a randomized, open-label, parallel phase Ib study evaluating the safety of T-DXd and immunotherapy agents with and without chemotherapy in advanced or metastatic HER2+, non-squamous NSCLC. The registry lists safety and tolerability, including determination of the recommended phase 2 dose (RP2D), as the primary endpoint.

2. Clinical Question

The central statistical question in DESTINY-Lung03 is a safety and tolerability question: how frequently do adverse events (AEs) and serious adverse events (SAEs) occur during treatment, and what safety information is generated for determination of the recommended phase 2 dose?

3. Trial Design

The combination of randomization and a 10-arm parallel structure is important statistically. This is not a simple two-group randomized comparison. The principal registry endpoint is safety, and the study contains multiple treatment configurations involving T-DXd, immunotherapy agents, and chemotherapy.

Study timing

The registry currently describes the study as Recruiting. The difference between the study start date and the primary completion date also indicates that the registry describes an ongoing longitudinal safety assessment rather than a completed efficacy analysis.

4. Study Arms and Interventions

The registry lists 10 study arms but does not provide arm-specific enrollment counts in the available record. It identifies the following interventions:

5. Primary Endpoint

This endpoint has a fundamentally different statistical structure from a time-to-event efficacy endpoint. The registry defines it in terms of the occurrence and frequency of adverse events and serious adverse events, rather than time to disease progression, time to death, tumor response, or another efficacy outcome.

What the endpoint is designed to capture

6. Planned Analysis

The ClinicalTrials.gov record identifies safety and tolerability as the primary endpoint and specifies that AEs and SAEs will be assessed for approximately 20 months from informed consent. Results have not been posted on ClinicalTrials.gov.

For an endpoint based on the frequency of AEs and SAEs, the central statistical summaries would ordinarily be descriptive. The natural quantities are the number of participants experiencing an event and the corresponding proportion within each relevant treatment group, together with the severity categories defined by the applicable grading system.

Because the study has 10 parallel arms, safety summaries would ordinarily preserve the treatment-arm structure rather than collapsing all participants into one overall percentage. The appropriate descriptive unit depends on the specific safety question: overall AEs, SAEs, particular event categories, and severity grades can require separate summaries.

Safety frequency versus event burden

A frequency-based endpoint can answer whether participants experienced an event, but the interpretation can change depending on whether the analysis counts participants, events, or both. For example, one participant with multiple occurrences of an AE and several participants with one occurrence each can produce different conclusions about event burden even when the number of affected participants is the same.

The registry's primary endpoint is expressed in terms of the occurrence of AEs and SAEs. That wording makes participant-level incidence an important descriptive concept, while event counts and severity distributions can provide additional context when available.

7. Statistical Methodology

Descriptive safety analysis

Safety endpoints are commonly summarized using counts and percentages rather than a single model-based effect estimate. For each relevant treatment group, the number of participants experiencing an event can be paired with the number of participants evaluated for safety.

Why denominators matter

In a 10-arm study, the denominator can differ substantially across treatment groups. A reported percentage is therefore incomplete unless the reader knows which participants were included in that percentage.

The distinction is particularly important in an ongoing study. Participants may have different amounts of observation time, and a cumulative proportion of participants with at least one event is not the same quantity as an event rate per unit of exposure.

AE and SAE classification

The primary endpoint explicitly separates AEs and SAEs. Statistically, this means that an analysis should preserve the distinction between the broad adverse-event category and the more specific serious-adverse-event category rather than combining them into a single undifferentiated outcome.

Severity grading

The registry specifies NCI CTCAE v5.0 grading. Severity is therefore an ordered clinical dimension of the safety data. A useful safety summary can distinguish the overall occurrence of events from their severity distribution rather than treating every event as equivalent.

8. Statistical Methods Explained

Why is a safety endpoint different from a survival endpoint?

A safety endpoint based on the occurrence of AEs and SAEs asks whether participants experience specified events during the assessment period. A survival endpoint instead measures time until an event occurs. The statistical summaries are therefore different: safety analyses often emphasize counts, percentages, severity distributions, and exposure, whereas survival analyses require methods such as Kaplan-Meier estimation and hazard models.

Why is the denominator important when reporting an adverse-event percentage?

A percentage is calculated relative to a population. In a 10-arm study, each arm can have its own denominator. If one arm has fewer participants than another, the same number of events can correspond to very different percentages. The treatment group and analysis population therefore need to accompany an incidence percentage.

What does NCI CTCAE v5.0 grading add to the analysis?

Grading provides a structured way to describe the severity of adverse events. A safety analysis can therefore distinguish the overall frequency of events from the frequency of more severe events. This matters because two treatment groups can have similar overall AE incidence while differing in the distribution of event severity.

Why should AEs and SAEs be kept separate?

They represent different safety concepts. AEs are the broad category identified in the primary endpoint, while SAEs are a specifically defined subset of serious events. Combining them can obscure the distinction the registry explicitly makes in the primary endpoint.

Why does a 10-arm design change the statistical presentation?

A two-arm trial can often be summarized with a single treatment-versus-control comparison. A 10-arm study has a more complex comparison structure. For a safety endpoint, arm-specific descriptive summaries can be more informative than reducing the study to one pooled treatment percentage.

Why does the approximately 20-month assessment period matter?

The registry defines the safety and tolerability assessment as occurring for approximately 20 months from informed consent. A safety percentage is therefore tied to an observation window. Changing the observation period can change the number and proportion of participants experiencing an event.

Why is frequency not the same as event rate?

A frequency expressed as the proportion of participants with at least one event gives each participant one contribution to the numerator, regardless of whether that participant experienced one event or multiple events. An event rate can instead account for the number of events relative to an exposure period. These are different statistical quantities and should not be interpreted interchangeably.

9. Randomization and the Multi-Arm Structure

The registry identifies DESTINY-Lung03 as randomized with a parallel design and 10 arms. Randomization provides the design framework for comparing treatment groups, while the parallel structure means participants remain associated with their assigned treatment arm rather than being described as moving sequentially through a crossover design.

The combination of randomization with multiple treatment arms is especially relevant to safety interpretation. Randomization can support comparisons between assigned treatment groups, but the statistical presentation still needs to retain the distinct treatment configurations rather than treating the entire study as a single intervention.

10. Interventions in the Statistical Context

The intervention list contains seven named interventions: T-DXd, durvalumab, cisplatin, carboplatin, pemetrexed, volrustomig, and rilvegostomig. They include drugs and a biological intervention, and the study title specifies treatment combinations with and without chemotherapy.

Because the registry lists interventions rather than a complete arm-by-arm allocation table in the information available here, the statistical interpretation should remain at the level of the registered multi-arm treatment framework. It would be inappropriate to assign specific combinations or sample sizes to individual arms without corresponding registry information.

11. Safety and Tolerability as the Primary Statistical Objective

DESTINY-Lung03 differs from a conventional late-phase efficacy trial because its registered primary endpoint is safety and tolerability, with the assessment also intended to determine the recommended phase 2 dose. That changes what a statistically informative result would look like.

This distinction is important because a safety-focused trial should not be interpreted as though the primary endpoint were an efficacy measure. The statistical question is first about tolerability and adverse-event experience, not about a comparative estimate of survival benefit.

12. The Role of the Recommended Phase 2 Dose

The registry states that safety and tolerability will be assessed both to characterize safety and to determine the recommended phase 2 dose (RP2D). Dose selection therefore forms part of the statistical and clinical decision context surrounding the primary endpoint.

An RP2D decision generally requires more than one overall AE percentage. The relevant evidence can include the occurrence of adverse events, serious adverse events, their severity, and how these observations relate to the treatment configurations under evaluation. The registry's primary endpoint explicitly connects the safety assessment to this dose-selection objective.

13. Analysis Population and Exposure Considerations

The registry information identifies overall enrollment of 304 participants but does not provide an arm-specific safety analysis population in the available record. For safety statistics, the denominator is therefore a critical piece of information that must accompany any future posted result.

Exposure also matters because the registry defines an assessment period of approximately 20 months from informed consent. Two groups with different amounts of observed treatment exposure can have different opportunities for events to occur even when their participant-level incidence proportions are similar.

This distinction becomes increasingly relevant when interpreting safety findings over time. A cumulative proportion is tied to the specified assessment window, whereas an exposure-adjusted measure can provide a different perspective on event burden when follow-up differs between participants.

14. Missing Data and Observation Time

The registry does not specify a formal missing-data or imputation method for the primary safety endpoint in the information available here. For safety analyses, however, the amount of observation available for each participant remains relevant to interpretation.

If a participant has limited follow-up, the absence of a recorded event does not necessarily have the same evidentiary meaning as an absence of an event after a substantially longer observation period. The approximately 20-month assessment frame therefore provides important context for any eventual frequency estimate.

For binary safety summaries, it is also important to distinguish a participant who has no recorded event from a participant for whom the relevant observation information is incomplete. Those situations should not automatically be treated as statistically identical without a prespecified analysis convention.

15. Multiplicity in a 10-Arm Safety Study

Multiplicity becomes a structural issue when a randomized study contains 10 treatment arms. A large number of pairwise comparisons can produce apparently different results even when no single comparison has been designated as the principal confirmatory question.

The primary endpoint in this registry is safety and tolerability rather than a stated efficacy hypothesis involving one treatment versus one control. For that reason, descriptive arm-specific safety summaries can be more directly aligned with the stated endpoint than an indiscriminate collection of pairwise hypothesis tests.

The registry does not identify a formal multiplicity-adjustment strategy for safety comparisons in the information available here. Consequently, future safety results should be interpreted according to the analysis framework actually posted rather than assuming that every observed arm difference represents a formal hypothesis test.

16. Statistical Interpretation of Safety Frequencies

17. What This Study Can and Cannot Answer Statistically

This separation is useful when reading an early-phase study. A statistically detailed safety assessment can be highly informative without requiring the study to answer the same questions as a later-phase efficacy trial.

18. Limitations

• No posted statistical analyses: the ClinicalTrials.gov record contains no posted statistical analyses for the registered primary endpoint.
• No posted results: the registry information does not provide outcome estimates for the frequency of AEs or SAEs.
• Multi-arm complexity: the study contains 10 arms, making treatment-group interpretation more complex than a standard two-arm trial.
• Arm-specific enrollment: the overall enrollment is 304, but arm-specific sample sizes are not provided in the information available here.
• Safety endpoint: the registered primary endpoint concerns AEs and SAEs rather than a time-to-event efficacy outcome.
• Observation period: safety and tolerability are assessed for approximately 20 months from informed consent, so frequency estimates are tied to that assessment framework.
• Ongoing study: the registry status is Recruiting and the primary completion date is 2027-06-30, so the study is not represented as a completed statistical analysis in the registry information summarized here.
• Multiplicity: the presence of 10 arms creates multiple potential comparisons, while the registry information does not specify a formal multiplicity-adjustment strategy for safety comparisons.
• Exposure: participant-level incidence and exposure-adjusted event rates are different quantities, and the registry information does not provide an exposure-adjusted safety analysis.

19. Why This Trial Matters Statistically

DESTINY-Lung03 is a useful statistical teaching case because it illustrates how early-phase randomized trials can have a substantially different analytical objective from conventional confirmatory efficacy trials.

20. Statistical Concepts in This Trial

Several core principles of clinical-trial statistics are particularly relevant to DESTINY-Lung03: randomized allocation, multi-arm design, descriptive safety analysis, event-frequency denominators, severity grading, observation windows, and the distinction between participant incidence and exposure-adjusted event rates.

21. A Statistical Reading Guide for Future Results

When safety results are subsequently reported, several questions should be answered before interpreting any percentage or treatment-group difference.

This framework prevents a common statistical mistake: treating a simple percentage as though it were a complete treatment-effect estimate. Safety interpretation requires the event definition, denominator, treatment arm, observation period, severity, and analysis population to be considered together.

22. Registry Timeline

The timeline reinforces the distinction between a registered study design and a completed results analysis. The registry currently describes a recruiting phase Ib study with a future primary completion date and a safety-focused endpoint assessed over approximately 20 months from informed consent.

23. Statistical Interpretation vs Clinical Interpretation

Keeping these perspectives separate is useful. Statistical analysis describes how frequently safety events occur and how those observations are distributed. Clinical interpretation considers what those safety observations mean for tolerability and dose selection.

24. Summary of the Registered Statistical Framework

25. Record Summary

DESTINY-Lung03 is a randomized, open-label, parallel phase Ib study with 10 arms and registered enrollment of 304 participants. Its primary endpoint is the frequency of adverse events and serious adverse events, with occurrence graded according to NCI CTCAE v5.0 over an assessment period of approximately 20 months from informed consent. The safety assessment is also intended to support determination of the recommended phase 2 dose.

Statistically, the key features are the randomized multi-arm structure, the need for arm-specific safety denominators, the distinction between AEs and SAEs, the importance of severity grading, and the difference between participant-level event frequency and exposure-adjusted event rates. Because the ClinicalTrials.gov record contains no posted statistical analyses or outcome estimates for the primary endpoint, the appropriate statistical interpretation remains centered on how the registered safety endpoint should be measured and understood.

26. Sources

• ClinicalTrials.gov: DESTINY-Lung03, NCT04686305.