← Clinical Trials
Metastatic Breast Cancer Phase 1/2 Randomized Sequential Design NCT04538742

DESTINY-Breast07: Complete Statistical Analysis of T-DXd Combinations in HER2-Positive Metastatic Breast Cancer

An independent statistical analysis of DESTINY-Breast07, a phase 1/2 randomized study evaluating trastuzumab deruxtecan combinations in HER2-positive metastatic breast cancer, with emphasis on its sequential multi-arm design and registered safety endpoints.

ClinicalTrials.gov status: Active, not recruiting  ·  Start date: December 28, 2020  ·  Primary completion: January 31, 2025
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. The registry currently contains registered primary safety endpoints for Parts 1 and 2, but no posted statistical analyses for those endpoints.

1. Trial at a Glance

DESTINY-Breast07 is a phase 1/2 randomized study of trastuzumab deruxtecan combinations in HER2-positive metastatic breast cancer. The registry describes a sequential design with no masking, a treatment purpose, 8 arms, and an enrollment of 245 participants.

245
Enrollment
Registry enrollment
8
Arms
Sequential design
1/2
Phase
Phase 1/2
2020–2025
Trial period
Start to primary completion
FeatureDESTINY-Breast07
PhasePhase 1/2
ConditionMetastatic Breast Cancer
AllocationRandomized
Design modelSequential
MaskingNone
Primary purposeTreatment
Enrollment245
Number of arms8
StatusActive, not recruiting
Lead sponsorAstraZeneca
Sponsor typeIndustry
ClinicalTrials.govNCT04538742

2. Clinical Question

The registry describes DESTINY-Breast07 as a phase 1b/2 study of trastuzumab deruxtecan combinations in HER2-positive metastatic breast cancer. The central statistical question is therefore how treatment combinations can be evaluated across a randomized, sequential multi-arm development program while safety is monitored through prespecified adverse-event endpoints.

Population

Patients with metastatic breast cancer described by the study title as HER2-positive metastatic breast cancer.

Intervention

Trastuzumab deruxtecan, with combinations involving durvalumab, paclitaxel, pertuzumab, and tucatinib.

Comparator

The registry describes a randomized multi-arm study rather than a single two-group comparison. The available record identifies 8 arms but does not provide arm-specific sample sizes or treatment assignments in the information summarized here.

Primary question

How frequently do adverse events and serious adverse events occur in Parts 1 and 2 during the registry-defined follow-up period?

Important design distinction. A multi-arm phase 1/2 study should not automatically be interpreted as though it were a conventional two-arm confirmatory trial. The statistical unit of interest can vary by part, arm, dose or combination, and development stage. The registry identifies randomization and 8 arms, but the available record does not provide enough arm-level information to reconstruct a single treatment-versus-control efficacy comparison.

3. Trial Design

01
Enroll245 participants
02
RandomizeRandomized allocation
03
Sequential8-arm development design
04
Part 1AE and SAE monitoring
05
Part 2AE and SAE monitoring
Allocation
The registry classifies the study as randomized.
Design model
The registry classifies the design model as sequential.
Masking
The study is listed as having no masking.
Primary purpose
Treatment.

Intervention framework

TRASTUZUMAB DERUXTECAN

Core study drug

  • Trastuzumab deruxtecan
COMBINATION PARTNERS

Registered combination drugs

  • Durvalumab
  • Paclitaxel
  • Pertuzumab
  • Tucatinib

The registry therefore describes a development program involving trastuzumab deruxtecan and multiple combination partners rather than a simple intervention-versus-placebo structure. Because arm-specific assignments and sample sizes are not contained in the available registry information, a numerical comparison of individual combinations cannot be constructed from the record alone.

4. Registered Primary Endpoints

All four registered primary endpoints concern safety. They are divided between Part 1 and Part 2 and distinguish adverse events from serious adverse events. Each endpoint uses the same approximately 53-month follow-up horizon and specifies grading according to NCI CTCAE v5.0.

EndpointRegistry definitionTime frame
Occurrence of adverse events (AEs) — Part 1 Occurrence of AEs in Part 1 graded according to NCI CTCAE v5.0 Up to follow-up period, approximately 53 months
Occurrence of serious adverse events (SAEs) — Part 1 Occurrence of SAEs in Part 1 graded according to NCI CTCAE v5.0 Up to follow-up period, approximately 53 months
Occurrence of adverse events (AEs) — Part 2 Occurrence of AEs in Part 2 graded according to NCI CTCAE v5.0 Up to follow-up period, approximately 53 months
Occurrence of serious adverse events (SAEs) — Part 2 Occurrence of SAEs in Part 2 graded according to NCI CTCAE v5.0 Up to follow-up period, approximately 53 months

What these endpoints measure

An adverse-event endpoint describes the occurrence of untoward medical events during the specified observation period. A serious adverse event is a separate regulatory safety classification and should not be treated as simply a more severe numerical value of an ordinary adverse event rate. The registry explicitly distinguishes the two endpoint categories.

Why grading matters

The endpoints specify NCI CTCAE v5.0 grading. This means that safety summaries can distinguish not only whether an event occurred, but also its assigned severity grade. In a statistical analysis, this distinction matters because an analysis of any-grade events answers a different question from an analysis restricted to higher-grade events.

5. Statistical Methodology

The registry identifies the endpoints and their follow-up period but does not post statistical analyses for them. Consequently, the most useful statistical interpretation is to distinguish the registered measurement from the standard analytical framework normally used for this type of safety endpoint.

Registry status
Posted statistical analyses: none

No formal statistical analyses are posted for the registered primary endpoints on ClinicalTrials.gov.

Safety-event summaries

For an endpoint defined as occurrence of AEs or SAEs, the most direct descriptive analysis is generally the number and percentage of participants experiencing at least one event during the specified observation period. When CTCAE grades are available, event summaries can additionally be presented by severity grade.

Statistical questionTypical approach for this endpoint
How many participants experienced an event?Count and percentage of participants with at least one qualifying event.
How severe were the events?Summaries by CTCAE grade, using the registry-specified NCI CTCAE v5.0 framework.
How does safety differ across randomized groups?Arm-specific event rates with absolute differences or ratios where an appropriate comparison exists.
Does timing matter?Time-to-first-event methods can be used when event timing and censoring are available.
Are repeated events relevant?Participant-level incidence and recurrent-event analyses answer different questions and should not be conflated.

Participant-level incidence versus event-level counts

A common safety-analysis distinction is whether the denominator is the number of participants or the total number of events. If one participant experiences several adverse events, a simple participant-incidence endpoint can count that participant once, whereas an event-level summary can count multiple events. The registry wording "occurrence of AEs" does not, by itself, establish which detailed summary convention was used for any eventual statistical report.

Part 1 and Part 2 should remain distinct

The four primary endpoints are explicitly separated by study part. Combining Part 1 and Part 2 into a single safety estimate could obscure differences in the populations, interventions, or development objectives represented by the two parts. A formal analysis should preserve the registry's endpoint structure unless the statistical analysis plan specifies a different estimand.

Randomization and safety interpretation

Because the study is randomized, comparisons among randomized groups can benefit from the balance created by random allocation. However, safety analyses also depend on actual exposure to study treatment. A formal report would therefore need to define the safety analysis population, treatment exposure rules, observation window, and handling of participants who discontinue treatment. Those details are not contained in the available registry endpoint information.

6. Planned Analysis

The registry's primary endpoints are safety-occurrence measures rather than time-to-event efficacy endpoints. A typical analysis would therefore begin with participant-level incidence summaries for AEs and SAEs, separately for Part 1 and Part 2, followed by severity-grade summaries using NCI CTCAE v5.0.

Primary safety summary

For each registered endpoint, report the number and percentage of participants experiencing at least one qualifying event during the approximately 53-month follow-up period.

Severity distribution

Because the endpoints specify NCI CTCAE v5.0 grading, an informative analysis would separate events by grade rather than treating all AEs as clinically equivalent.

Part-specific analysis

Part 1 and Part 2 should be analyzed separately because they are separate registered primary endpoints.

Comparative analysis

Where randomized groups are formally compared, absolute event-rate differences and relative measures can supplement descriptive percentages, with confidence intervals used to communicate statistical precision.

No formal statistical analyses were posted to ClinicalTrials.gov. The registry identifies what will be measured and the follow-up period, but it does not provide numerical estimates, confidence intervals, or p-values for these primary endpoints.

Why a p-value alone would not be sufficient

For safety outcomes, a p-value does not describe the magnitude of an adverse-event difference. If one group had a higher observed event rate, the clinically useful quantities would include the event rates themselves and an effect measure such as an absolute risk difference or relative risk, together with an interval describing statistical uncertainty.

Basic safety effect measures
Risk difference = p1 − p0
Risk ratio = p1 / p0

These measures describe different aspects of the same binary safety outcome. The choice depends on the estimand and the randomized comparison being made.

7. Statistical Methods Explained

Why are adverse events and serious adverse events separate endpoints?

An AE endpoint captures the occurrence of adverse events broadly, whereas an SAE endpoint identifies the subset classified as serious according to the applicable safety framework. They answer related but distinct questions. A treatment could have a substantial number of non-serious events without producing a comparable increase in serious events, so the two endpoints should not be collapsed into one measure.

Why is the analysis usually based on participants rather than simply counting events?

Participant-level incidence answers the question, "How many patients experienced at least one event?" This avoids allowing one participant with many repeated events to dominate the rate. Event counts can still be useful, but they address a different question and require a different denominator and statistical framework.

Why are CTCAE grades statistically important?

Severity is part of the clinical meaning of an AE. A summary that combines all grades can obscure whether an observed difference is driven primarily by lower-grade events or by more severe toxicity. Because the registry explicitly specifies NCI CTCAE v5.0, severity-grade classification is part of the endpoint definition.

Why should Part 1 and Part 2 be analyzed separately?

The registry defines separate primary endpoints for each part. Combining the parts without understanding their underlying treatment assignments and populations could produce a pooled rate whose interpretation is unclear. Separate estimates preserve the estimand defined by each registered endpoint.

What does randomization contribute to a safety comparison?

Randomization helps balance measured and unmeasured baseline factors in expectation, making treatment-group comparisons more interpretable than comparisons based solely on non-randomized cohorts. However, safety interpretation still depends on treatment exposure, follow-up, withdrawals, and the definition of the analysis population.

When would a time-to-first-event analysis be useful?

If the registry's underlying data include event dates and censoring information, time-to-first-event methods can account for different durations of observation. Kaplan-Meier methods or related survival techniques can then describe the probability of remaining free of a first event over time. This is conceptually different from a simple proportion experiencing an event by the end of follow-up.

Why is the absence of posted estimates important?

A registered endpoint specifies the question and measurement framework, but it does not provide an observed treatment effect. Without posted event counts, rates, confidence intervals, or p-values, the statistical evidence for the primary endpoints cannot be numerically summarized from the registry record.

8. Understanding the 53-Month Follow-Up Window

Each of the four registered primary endpoints uses the time frame "Up to follow-up period, approximately 53 months." This wording defines a long observation horizon for the safety assessment, but it does not by itself imply that every participant contributed the same amount of follow-up.

Fixed horizon

A specified follow-up horizon gives the analysis a common temporal frame for describing safety occurrence.

Unequal observation

Participants may have different amounts of observed follow-up because of treatment discontinuation, withdrawal, loss to follow-up, or other study events.

Censoring

If time-to-event methods are used, participants without an observed event by their last available assessment can contribute information up to that time.

Interpretation

"Approximately 53 months" describes the registry time frame; it does not establish a median follow-up or a specific amount of follow-up for each participant.

9. Randomized Multi-Arm Design: Statistical Implications

The registry identifies both randomized allocation and 8 arms. This creates a different statistical structure from a conventional two-arm trial. In a multi-arm study, the scientific question can involve several treatment combinations, and each comparison may have its own target population, estimand, and precision.

Design featureStatistical implication
Randomized allocationSupports comparisons based on randomized treatment assignment.
8 armsCreates multiple potential treatment comparisons and requires careful definition of which comparison addresses each scientific question.
Sequential modelSuggests that the study's components or stages are organized in sequence rather than as a single static two-arm comparison.
No maskingTreatment assignment is not masked, which can matter for subjective assessments and treatment-related behavior, although the extent of impact depends on the endpoint.
Phase 1/2Places the study in an early clinical-development framework in which safety and treatment development are important components of the statistical evidence.

Multiplicity in an 8-arm study

When several treatment comparisons are performed, the probability of observing at least one apparently unusual result can increase if each comparison is tested independently at the same nominal significance level. A formal statistical analysis therefore needs to specify which comparisons are confirmatory, which are exploratory, and whether any multiplicity adjustment is used.

The available registry information does not state a multiplicity-adjustment strategy for the four registered primary safety endpoints. It therefore would not be appropriate to assign a particular familywise error procedure to the study based only on the registry record.

10. Safety Analysis: Key Statistical Distinctions

MeasureQuestion answeredImportant distinction
Any AEWhat proportion of participants experienced at least one AE?One participant is generally counted once for a participant-level incidence measure.
SAEWhat proportion experienced at least one serious adverse event?Seriousness is a separate classification from ordinary severity grading.
CTCAE gradeHow severe was the event?Severity grade should not be treated as synonymous with seriousness.
Risk differenceHow many more or fewer participants experienced an event?An absolute measure that is directly interpretable on the probability scale.
Risk ratioHow does event probability compare between groups?A relative measure that can look large even when absolute risks are small.
Time to first eventHow quickly did participants experience an event?Requires event timing and appropriate handling of censoring.

11. Limitations and Interpretation Issues

12. Why This Trial Matters Statistically

DESTINY-Breast07 is a useful teaching case because its statistical structure differs from the familiar two-arm confirmatory trial. The combination of a phase 1/2 program, randomized allocation, 8 arms, a sequential design, and part-specific safety endpoints illustrates why the estimand and analysis population must be defined before numerical results are interpreted.

ConceptHow it appears in DESTINY-Breast07
RandomizationThe registry classifies the study allocation as randomized.
Multi-arm designThe study contains 8 arms, creating multiple possible treatment comparisons.
Sequential designThe registry classifies the design model as sequential.
Safety endpointsAll four registered primary endpoints concern occurrence of AEs or SAEs.
CTCAE gradingPrimary safety endpoints specify NCI CTCAE v5.0 grading.
Part-specific analysisPrimary AE and SAE endpoints are separately registered for Part 1 and Part 2.
Follow-upEach primary safety endpoint uses a follow-up period of approximately 53 months.
MultiplicityMultiple arms and endpoints make comparison and error-control definitions important.
Missing follow-upDifferent observation times can affect simple incidence summaries and may motivate time-to-event approaches when event timing is available.
Statistical uncertaintyConfidence intervals would quantify uncertainty around observed safety effects when formal comparative estimates are available.

13. Statistical Interpretation of Safety Endpoints

What an event rate would mean

An AE rate expressed as the percentage of participants with at least one event would describe the proportion of the analyzed population experiencing the defined event during the specified observation period. It would not describe the number of total events experienced by those participants.

What a risk difference would mean

If a randomized comparison reported a risk difference, the estimate would describe the absolute difference in the probability of experiencing the specified safety event between the groups being compared. A positive value would indicate a higher observed probability in the first group as defined by the analysis.

What a risk ratio would mean

A risk ratio would compare the probability of experiencing the event between two groups. It would not mean that an individual participant had that same proportional change in personal risk.

Why confidence intervals matter

A confidence interval around an estimated safety effect communicates statistical uncertainty associated with the estimate. A narrow interval indicates greater statistical precision than a wide interval, although precision is not the same thing as clinical importance.

Why a p-value is not an effect size

A p-value addresses evidence against a specified statistical null hypothesis under the model and testing framework. It does not describe how large an adverse-event difference is, how clinically important it is, or how frequently individual patients will experience it.

14. Registry Timeline

December 28, 2020

Study start

The registry lists December 28, 2020 as the study start date.

January 31, 2025

Primary completion

The registry lists January 31, 2025 as the primary completion date.

Current registry status

Active, not recruiting

The ClinicalTrials.gov record lists the study status as active, not recruiting.

15. What the Registry Does and Does Not Establish

Established by the registryNot numerically reported in the available record
Phase 1/2 studyAE estimates by arm
Randomized allocationSAE estimates by arm
Sequential design modelConfidence intervals for safety comparisons
No maskingP-values for primary safety endpoints
8 armsArm-specific sample sizes
245 enrollmentNumerical treatment effects
Four registered primary safety endpointsPosted formal statistical analyses
Approximately 53-month endpoint time frameObserved event counts and rates
NCI CTCAE v5.0 gradingDetailed safety-analysis population definitions

16. Overall Statistical Perspective

DESTINY-Breast07 illustrates an important principle in clinical-trial statistics: the design determines which comparisons are meaningful, while the endpoint definition determines which analysis answers the registered question. Here, the registry identifies a randomized, sequential, 8-arm phase 1/2 study and four primary safety endpoints divided between Parts 1 and 2.

The primary endpoints are all based on occurrence of adverse events or serious adverse events during a follow-up period of approximately 53 months, with NCI CTCAE v5.0 grading. A complete statistical report would ordinarily present participant-level incidence, severity distributions, and appropriately defined comparative estimates where randomized comparisons are relevant. Time-to-event methods could add information about the timing of first events when event dates and censoring information are available.

However, the ClinicalTrials.gov record does not post formal statistical analyses for these endpoints. As a result, the registry establishes the safety questions and their measurement framework but does not provide numerical evidence from which treatment-specific effect estimates can be interpreted.

Bottom line: The statistical importance of DESTINY-Breast07 lies not only in its safety endpoints but also in its randomized, sequential, multi-arm structure. Correct interpretation requires keeping Part 1 and Part 2 distinct, defining the analysis population and exposure window, distinguishing AEs from SAEs and severity from seriousness, and accounting for the multiple comparisons inherent in an 8-arm development program.

17. Sources