Trastuzumab Deruxtecan (DS-8201a) Versus Investigator's Choice for HER2-low Breast Cancer That Has Spread or Cannot be Surgically Removed [DESTINY-Breast04]: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Trastuzumab Deruxtecan (DS-8201a) Versus Investigator's Choice for HER2-low Breast Cancer That Has Spread or Cannot be Surgically Removed [DESTINY-Breast04] (NCT03734029), a randomized phase 3 study. The primary outcome is Progression-free Survival (PFS) Based on Blinded Independent Central Review (BICR) in the Hormone Receptor-Positive Cohort in Participants With HER2-low Breast Cancer.

2. Clinical Question

3. Trial Design

NCT03734029, titled DESTINY-Breast04, was a phase 3, randomized, parallel-group, open-label, active-controlled trial sponsored by Daiichi Sankyo. The registry lists breast cancer as the condition and total enrollment as 557. The study started on 2018-12-27, reached primary completion on 2022-01-11, and is listed as completed.

The trial compared trastuzumab deruxtecan with physician’s choice of chemotherapy in participants with HER2-low, unresectable and/or metastatic breast cancer previously treated with chemotherapy. The physician’s-choice options were capecitabine, eribulin, gemcitabine, paclitaxel, and nab-paclitaxel. Allocation was randomized. Masking was none, consistent with the registry’s open-label description: participants and treating investigators were not blinded to assigned treatment according to the recorded design.

The primary endpoint was progression-free survival assessed by blinded independent central review in the hormone receptor-positive cohort. For this endpoint, the registry reports 331 participants in the trastuzumab deruxtecan arm and 163 in the physician’s-choice arm. These endpoint-specific arm counts are distinct from the overall enrollment of 557. The registry record does not explain the difference between those counts and overall enrollment, or provide further detail about the subset’s formation in this endpoint entry.

The registry lists 12 outcome measures and 10 posted secondary outcomes. Those include additional progression-free survival assessments, overall survival outcomes, and response outcomes in the hormone receptor-positive cohort and/or all patients. The record lists one analysis for the primary endpoint. Details such as the planned allocation ratio, stratification factors, full protocol analysis plan, and reasons for the endpoint-specific counts are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: All participants had been previously treated with at least 1 and no more than 2 prior lines of chemotherapy in the recurrent or metastatic setting. The treatment chosen for the Physician's Choice arm was based on the label approved in the country of drug administration. The Physician's Choice group was combined to ensure an appropriate sample size for the comparator group.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Progression-free survival (PFS) was assessed in the Hormone Receptor-Positive cohort of Full Analysis Set.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: Baseline characteristics were assessed in the Full Analysis Set. The Full Analysis Set included all participants randomized into the study, including those who did not receive a dose of study treatment. Participants were analyzed according to the treatments and strata assigned at randomization.

6. Primary Endpoint

The registry lists the primary outcome as Progression-free Survival (PFS) Based on Blinded Independent Central Review (BICR) in the Hormone Receptor-Positive Cohort in Participants With HER2-low Breast Cancer, measured over From the date of randomization to the earliest date of the first objective documentation of radiographic disease progression or death due to any cause, up to approximately 3 years.

Progression-free survival (PFS) is a time-to-event endpoint. In this trial, the clock began at randomization and ended at the earliest of the first objective radiographic documentation of disease progression or death due to any cause. The registry describes progression as at least a 20% increase in the sum of the diameters of target lesions. A participant who died before recorded radiographic progression therefore had an event at death under this definition.

The primary endpoint was assessed in the hormone receptor-positive cohort by blinded independent central review, using modified Response Evaluation Criteria in Solid Tumors, version 1.1. Central review can provide an assessment distinct from the treating investigator’s assessment; the registry specifies that the primary result uses the blinded central reviewers’ assessment. The record does not describe the review procedures in further detail.

A median summarizes the point in the estimated PFS distribution by which half of the observed population’s estimated event-free time has elapsed. It is one useful summary of a time-to-event distribution, but it does not describe every participant’s outcome or the full pattern over time. The registry reports the median from Kaplan-Meier analysis and a confidence interval calculated using the Brookmeyer-Crowley method.

PFS is used to summarize the time until progression or death, combining these events in one endpoint. It does not, by itself, describe overall survival, symptom experience, or the separate contribution of progression and death. Those questions require other outcomes and analyses.

7. Results Posted in the Registry

• Trastuzumab Deruxtecan (T-DXd) vs Physician's Choice: hazard ratio 0.51 (95% CI 0.40–0.64), reported p < 0.001 (as reported in the registry).

8. Statistical Interpretation

9. Statistical Methods Explained

What does a Kaplan-Meier median represent?

In general, Kaplan-Meier analysis estimates the probability of remaining free of the event over time while accounting for participants whose event time is not observed through the end of follow-up. Such observations are commonly called censored: the participant’s event-free time is known up to a point, but the event is not recorded afterward within the available follow-up. The registry gives the endpoint timeframe as from randomization to progression or death, up to approximately 3 years. It does not provide the detailed censoring rules in this endpoint description.

The median PFS is the time at which the estimated event-free proportion reaches the midpoint of its distribution. It is not the arithmetic average of participants’ observed times. The reported medians are 10.1 months for trastuzumab deruxtecan and 5.4 months for physician’s choice. Each is an estimate, and its confidence interval communicates uncertainty about the median under the analysis method.

What does the hazard ratio describe?

In general, a hazard is the instantaneous event rate among participants who have not yet experienced the event, conditional on having remained event-free up to that time. A hazard ratio compares those rates between groups. A ratio below 1 for the first-named group indicates a lower estimated hazard relative to the comparator in the model’s terms. It is not a ratio of median times, and it is not an absolute risk difference.

The registry identifies the reported parameter as a Cox proportional-hazards estimate and lists Log Rank as the method. The Cox model is a common way to estimate a hazard ratio for time-to-event data; the log-rank test compares event experience across groups over follow-up. The record does not report whether covariates were included, how the model was stratified, or whether proportional hazards were assessed. Those elements should not be inferred from the method labels alone.

Why use blinded central review?

Radiographic progression requires interpretation of lesion measurements. In general, independent central review can separate the endpoint assessment from the treating investigator’s assessment and, when blinded, reduce the possibility that knowledge of assignment influences that review. Here the registry explicitly names blinded independent central review for the primary PFS result. It does not describe reviewer procedures, adjudication, or how disagreements were handled. The registry also lists investigator-assessed PFS as secondary outcomes, which are distinct assessments and should not be conflated with the primary result.

What does the confidence interval add?

A confidence interval gives a range of values compatible with the data and the method at a stated confidence level. It describes statistical uncertainty in an estimate; it is not a range containing a fixed proportion of individual participants’ PFS times. The median estimates have 95% confidence intervals, and the hazard-ratio estimate also has a 95% confidence interval. The registry says the confidence interval for the median was computed using the Brookmeyer-Crowley method. It does not specify a method for the hazard-ratio interval in this endpoint entry.

For the hazard ratio, the reported interval lies below 1. In general, an interval for a ratio that excludes 1 indicates that the data are inconsistent with an equal-hazard value under the corresponding analysis assumptions at the stated confidence level. This statement is about the statistical comparison, not the size or practical importance of an effect. A small p-value similarly measures how incompatible the observed data are with a specified null model under its assumptions; it is not the probability that the null hypothesis is true.

Why can time-to-event methods be useful?

Time-to-event methods use both when events occur and the information available from participants who have not had an event by their last observation. This can be useful when follow-up lengths differ or some participants leave observation before an event. The validity of the methods still depends on assumptions about follow-up and censoring. The registry’s brief endpoint record does not give enough detail to evaluate those assumptions independently.

General statistical caution also applies to proportional-hazards summaries: if relative hazards change substantially over time, one hazard ratio may not capture that pattern well. The record does not report whether this occurred or whether alternative summaries were used. The posted result should therefore be read as the registry’s reported model-based comparison, with the stated endpoint and population, rather than as a complete description of every aspect of time-varying outcomes.

The primary endpoint is a composite of progression and death. In general, combining event types can increase the amount of information available for a time-to-event comparison, but it means the overall estimate does not distinguish which event type contributes more to the result. The registry separately lists overall survival and other outcomes, but their numerical results are not included in the primary details described for this analysis.

10. Confidence Intervals and Precision

The median PFS estimates are accompanied by 95% confidence intervals: 9.5 to 11.5 months for trastuzumab deruxtecan and 4.4 to 7.1 months for physician’s choice. These intervals communicate uncertainty around the estimated medians. They do not show the range of individual participants’ outcomes, and they do not mean that 95% of participants had PFS times within the stated interval.

The reported hazard-ratio confidence interval also lies below 1. That indicates that the interval, under the reported analysis, does not include the equal-hazard reference value. The interval still allows a range of relative hazard values; it does not identify one exact effect as certain. Its width reflects statistical precision, which depends in part on the amount of event information and variability in the data. The registry does not provide the event count for this primary analysis in the endpoint description.

Confidence intervals for the two medians should not be read as a formal test of the difference between medians. The registry’s reported comparative analysis is the hazard-ratio analysis, with Log Rank listed as its method. A direct subtraction of medians or a separate test of median differences is not reported in the registry record and is not needed to describe the posted result.

Precision is not the same as generalizability. Even a relatively precise estimate applies to the studied population, treatment options, endpoint definition, and follow-up represented in the analysis. It does not automatically extend to all breast cancer populations, all treatment choices, or outcomes not measured by this endpoint.

11. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are based on the observed time-to-event effect and are expressed as total numbers of events, not participants per arm. For 80% power, the listed event totals are 69 when planning for 100% of the observed effect, 123 when planning for 75% of the observed effect, and 275 when planning for 50% of the observed effect. For 90% power, the corresponding event totals are 92, 164, and 368.

In general, a power target describes the chance that a planned study would detect an effect of the specified size at the chosen significance level, if the assumptions used for planning hold. The 80% column refers to the event total used in the illustration targeting 80% power; the 90% column does the same for a 90% target. These are not observed power calculations and do not describe the probability that the completed trial’s finding is true.

The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is the stated fraction of the observed effect. For time-to-event outcomes, the calculation uses the Schoenfeld approach and reports total events. These planning assumptions are an illustration based on the observed effect, not the trial’s original design assumptions. The trial’s own sample-size justification and original planning assumptions are not reported in the registry record.

Smaller effects require more event information to achieve the same power target under the illustration’s assumptions. This is why the event totals increase when planning for a smaller fraction of the observed effect. Because these figures concern events rather than participants, they should not be mistaken for a required enrollment total; the number of participants needed to accumulate a given number of events also depends on event occurrence and follow-up.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

d is the total number of events needed (Schoenfeld's formula) with 1:1 allocation and proportional hazards.

Worked calculation for this trial

1. Hazard ratio HR = 0.508, so ln(HR) = -0.6763.
2. d = 4 × (1.96 + 0.842)² / [ln(0.508)]² = 68.6, rounded up to 69 events in total for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
hr <- 0.5085                              # observed hazard ratio
events_needed <- function(power, frac = 1) ceiling(round(4 * (z_a + qnorm(power))^2 / log(hr^frac)^2, 6))   # Schoenfeld, 1:1 allocation
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: events = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(events_needed(0.80, f), big.mark = ","), format(events_needed(0.90, f), big.mark = ",")))

# Simulation: given d total events under proportional hazards, events in the treatment arm are Binomial(d, hr / (1 + hr)).
# The score (log-rank type) statistic is (D1 - d / 2) / sqrt(d / 4).
set.seed(3734029)
d    <- events_needed(0.80)
reps <- 20000
D1 <- rbinom(reps, d, hr / (1 + hr))
z  <- (D1 - d / 2) / sqrt(d / 4)
cat(sprintf("Simulated power with %s events: %.3f (%d replicates)\n", format(d, big.mark = ","), mean(abs(z) > z_a), reps))

Output from running the code

100% of observed effect: events = 69 (80% power), 92 (90% power)
 75% of observed effect: events = 123 (80% power), 164 (90% power)
 50% of observed effect: events = 275 (80% power), 368 (90% power)
Simulated power with 69 events: 0.800 (20000 replicates)

The simulation uses the fact that, with d events in total and proportional hazards, the number of events in the treatment arm is binomial. It is an approximation that conditions on the total number of events, and it should land close to the target power.

13. Group-Sequential Design Analysis

Many time-to-event trials, especially in oncology, stop or report early when interim data are overwhelming, so a single fixed-sample number understates the real design problem. The calculation below is an illustrative group-sequential design built from this trial's posted control median and hazard ratio. The registry record does not state the planned number of looks or the spending function, so the design is a planning illustration, not a reconstruction of the protocol. Assumed: one-sided α = 0.025, 90% power, three equally spaced looks on information, Lan-DeMets O'Brien-Fleming spending, 24 months of accrual, 60 months in total and a 1% monthly dropout hazard.

Under these assumptions the design needs about 105 participants in total, with the boundaries below.

Fixed design versus group-sequential alternatives

Group-sequential designs cost a small increase in maximum sample size and events compared with the fixed design, in exchange for the chance to stop early for overwhelming efficacy. O'Brien-Fleming spending keeps early boundaries very conservative and the final boundary close to the fixed-design value, whereas Pocock spending spends more alpha early.

Sensitivity to a smaller true effect

Planning for exactly the observed effect is optimistic, because observed effects in published trials tend to be somewhat larger than the true effect. The table repeats the main design (three looks, O'Brien-Fleming spending, 90% power) for smaller true hazard ratios.

For context, this trial enrolled 557 participants.

Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993. Agreement between the two packages confirms the boundary calculation.

R code and output

suppressWarnings(suppressMessages({ library(gsDesign); library(rpact) }))
med_control <- 5.4          # control-arm median, months (converted from the posted unit)
hr          <- 0.5085          # hazard ratio used for planning
lambdaC     <- log(2) / med_control
eta         <- 0.01               # dropout hazard per month (assumption)
acc_months  <- 24                 # accrual duration (assumption)
study_months<- 60                 # total study duration (assumption)

design <- function(k, sfu, beta, h = hr) gsSurv(k = k, test.type = 1, alpha = 0.025, beta = beta, sfu = sfu,
  lambdaC = lambdaC, hr = h, hr0 = 1, eta = eta, gamma = 1, R = acc_months, T = study_months, minfup = study_months - acc_months)

# Main design: 3 looks, Lan-DeMets O'Brien-Fleming spending, one-sided alpha 0.025, 90% power
x <- design(3, sfLDOF, 0.10)
cat(sprintf("N_TOTAL|%s\n", format(ceiling(sum(x$gamma * x$R)), big.mark = ",")))
for (i in 1:3) cat(sprintf("GS|%d|%.1f|%.0f|%.3f|%.5f|%.5f\n", i, x$T[i], x$n.I[i], x$upper$bound[i], pnorm(-x$upper$bound[i]), cumsum(x$upper$spend)[i]))
cat("\n"); print(gsBoundSummary(x))

# Comparison of designs
cmp <- list(list("Fixed design, one analysis", 1, sfLDOF), list("Two looks, O'Brien-Fleming", 2, sfLDOF),
            list("Three looks, O'Brien-Fleming", 3, sfLDOF), list("Three looks, Pocock", 3, sfLDPocock))
cat("\n")
for (b in c(0.20, 0.10)) for (d in cmp) {
  y <- design(d[[2]], d[[3]], b)
  cat(sprintf("DES|%s|%d|%.0f|%s|%.1f\n", d[[1]], round(100 * (1 - b)), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ","), tail(y$T, 1)))
}

# Sensitivity: what if the true effect is smaller than the observed one? (3 looks, O'Brien-Fleming, 90% power)
cat("\n")
for (fr in c(1, 0.75, 0.5)) {
  y <- design(3, sfLDOF, 0.10, exp(log(hr) * fr))
  cat(sprintf("SENS|%.2f|%.3f|%.0f|%s\n", fr, exp(log(hr) * fr), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ",")))
}

# Independent cross-check of the boundaries with rpact
r <- getDesignGroupSequential(kMax = 3, alpha = 0.025, beta = 0.10, sided = 1, typeOfDesign = "asOF", informationRates = x$timing)
cat(sprintf("\nCross-check, upper boundaries: gsDesign %s | rpact %s\n", paste(round(x$upper$bound, 3), collapse = ", "), paste(round(r$criticalValues, 3), collapse = ", ")))

Method: LachinFoulkes 
   Analysis               Value Efficacy
  IA 1: 33%                   Z   3.7103
      N: 68         p (1-sided)   0.0001
 Events: 31        ~HR at bound   0.2619
  Month: 16    P(Cross) if HR=1   0.0001
            P(Cross) if HR=0.51   0.0338
  IA 2: 67%                   Z   2.5114
     N: 106         p (1-sided)   0.0060
 Events: 62        ~HR at bound   0.5267
  Month: 25    P(Cross) if HR=1   0.0060
            P(Cross) if HR=0.51   0.5603
      Final                   Z   1.9930
     N: 106         p (1-sided)   0.0231
 Events: 93        ~HR at bound   0.6600
  Month: 60    P(Cross) if HR=1   0.0250
            P(Cross) if HR=0.51   0.9000



Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993

14. Analyses Reported by the Sponsor

The registry reports the sponsor analysis for the primary endpoint as a Cox proportional-hazards result, with Log Rank listed as the method. The hazard-ratio estimate is accompanied by a 95% confidence interval whose bounds are below 1, and the reported p-value is below < 0.001. The posted result is therefore the sponsor’s own reported comparison, not a calculation reconstructed from the two arm medians.

The record’s recalculation convention concerns simple, unadjusted calculations from posted summary statistics. For this time-to-event endpoint, the applicable reported hazard ratio is used rather than an independently reconstructed ratio. The sponsor’s own analysis takes precedence when a recalculation conflicts with the reported result; the discrepancy indicator for this record is false.

The median estimates and their confidence intervals remain useful descriptive summaries, but they do not contain enough information to recreate the Cox analysis or its uncertainty. Reconstructing a time-to-event model would generally require more information about event times and censoring than the medians alone provide. The registry does not post those participant-level data in this record.

The planning illustrations are a separate exercise based on the observed effect and stated assumptions. They are not a reconstruction of the sponsor’s original design plan. The original sample-size rationale and detailed analysis plan are not reported in the registry record.

15. Multiplicity, Missing Data and Analysis Populations

The registry lists 12 outcome measures, 10 posted secondary outcomes, and one analysis for the primary endpoint. The primary PFS result concerns the hormone receptor-positive cohort and uses blinded independent central review. The listed secondary outcomes include PFS assessments in the full group and hormone receptor-positive cohort, overall survival outcomes, and objective response outcomes.

When a study examines multiple outcomes, populations, or analyses, the chance of obtaining at least one apparently notable result can depend on how many comparisons are made and how they are handled. General statistical practice may use prespecified testing sequences or other multiplicity adjustments to control error rates. The registry record does not report whether such procedures were planned or used, nor does it give a hierarchy or family of hypotheses. No multiplicity adjustment should be assumed.

The record identifies the primary analysis population as the Full Analysis Set in the hormone receptor-positive cohort. It does not provide a detailed definition of that set, list exclusions, or describe how participants with protocol deviations were handled. The primary endpoint arm counts are 331 and 163, while overall enrollment is 557; the registry record does not explain this difference in the endpoint entry.

Missing outcome data and censoring are important for time-to-event analyses. The endpoint definition and timeframe are reported, but detailed censoring rules, handling of missing radiographic assessments, and sensitivity analyses are not reported in the registry record. It is therefore not possible from this information alone to assess how robust the result is to different missing-data assumptions.

The record names the median confidence-interval method and gives the reported comparative method labels, but does not provide a full statistical analysis plan. Details about interim analyses, stopping rules, covariate adjustment, stratification, and checks of model assumptions are not reported in the registry record. Their absence from this registry entry does not establish whether such work occurred; it means the entry does not document it.

16. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: overdose (2.9%), febrile neutropenia (2.3%), pneumonia (1.9%), interstitial lung disease (1.9%), pneumonitis (1.9%).

The registry lists serious adverse events for 371 participants at risk in the trastuzumab deruxtecan group, of whom 103 were affected, and for 172 participants at risk in the physician’s-choice group, of whom 43 were affected. These are safety-group figures and should not be confused with the primary endpoint arm counts. The record also lists 149 deaths in the trastuzumab deruxtecan group and 90 in the physician’s-choice group. It does not state in this summary whether those deaths were treatment-related or provide a cause-of-death breakdown.

For other adverse events, the registry lists 366 affected among 371 at risk in the trastuzumab deruxtecan group, and 167 affected among 172 at risk in the physician’s-choice group. These counts describe participants recorded as affected, not necessarily the number of distinct events, because a participant may experience more than one event. The registry summary does not provide enough detail to establish severity, duration, or causal attribution for all such events.

The most frequently listed serious adverse-event terms include overdose at 2.9%, febrile neutropenia at 2.3%, pneumonia at 1.9%, interstitial lung disease at 1.9%, pneumonitis at 1.9%, and medication error at 1.7%. The record excerpt does not specify the denominator for these term-specific percentages. These safety figures describe recorded events in the trial; they do not alone establish causality or determine an individual’s risk.

17. What the Result Does, and Does Not, Mean

The primary finding describes a randomized comparison of PFS in the hormone receptor-positive cohort with HER2-low breast cancer. Median PFS by blinded independent central review was 10.1 months in the trastuzumab deruxtecan group and 5.4 months in the physician’s-choice group. The reported hazard-ratio confidence interval lies below 1, and the reported p-value is below < 0.001. Together, these statistics describe the direction and uncertainty of the trial’s reported time-to-event comparison.

The result does not mean every participant assigned to trastuzumab deruxtecan had a longer PFS time than every participant assigned to physician’s choice. Group summaries cannot determine an individual participant’s outcome. The hazard ratio is a relative event-rate summary under the reported model, not a direct estimate of absolute benefit for a particular person or a ratio of median times.

The endpoint combines radiographic progression and death from any cause. It does not separately show the effect on progression alone or on overall survival. Although overall survival and other outcomes appear among the registry’s secondary outcomes, their numerical results are not part of the primary result. The registry’s reported PFS analysis also does not establish effects on quality of life or other unlisted outcomes.

The comparison is against physician’s choice among five listed chemotherapy options, not against each option individually. The registry record does not report the distribution of those choices in the primary analysis description, so the result should not be interpreted as a separate comparison with capecitabine, eribulin, gemcitabine, paclitaxel, or nab-paclitaxel individually.

These findings are specific to the studied population and analysis. They should not automatically be extended to participants outside the hormone receptor-positive cohort, to other disease settings, or to different treatment comparisons. Statistical evidence and clinical importance are related but distinct: interpreting importance also requires considering safety, the broader evidence, and the goals of care. The posted statistics alone do not settle those questions.

18. Why This Trial Matters Statistically

This trial illustrates how a randomized comparison can be summarized using both an interpretable time point and a relative time-to-event measure. The median PFS estimates describe points in each group’s estimated event-free time distribution, while the reported hazard ratio summarizes the relative event hazard over follow-up under a proportional-hazards model. Reading both helps distinguish a distribution summary from a model-based comparison.

It also shows why endpoint definition matters. PFS here combined radiographic progression, assessed by blinded independent central review, with death due to any cause. The assessor and event criteria shape what the statistic represents. Investigator-assessed PFS is listed separately among secondary outcomes, so it is important not to treat different assessment methods as interchangeable.

The record makes uncertainty visible through confidence intervals for the medians and hazard ratio. That information is essential: an estimate is not an exact description of the underlying treatment contrast. At the same time, statistical precision does not answer every question about generalizability, clinical importance, or safety. Those require attention to the population studied, comparator choices, endpoint scope, and other trial outcomes.

The planning illustration offers another useful statistical lesson. Event-based planning for time-to-event studies differs from planning based simply on a participant count, and assumptions about the effect to detect matter. The figures are explicitly illustrations based on the observed effect, not the original design calculation and not observed power. Keeping those distinctions clear prevents an after-the-fact planning exercise from being mistaken for a design justification.

19. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The registry description leaves several trial-specific details unresolved. It reports overall enrollment of 557 and primary-endpoint arm counts of 331 and 163, but does not explain the difference in this endpoint entry. The primary population is identified as the Full Analysis Set in the hormone receptor-positive cohort; the detailed population definition and exclusions are not reported in the registry record.

The comparator allowed physician’s choice among five chemotherapy options. The registry does not provide the distribution of those options in this primary endpoint description. The reported comparison therefore summarizes the assigned physician-choice group and cannot be separated into drug-specific estimates from the information presented.

The primary result is based on central-review PFS, a composite of radiographic progression and death. This endpoint does not separately quantify each event type. The registry does not provide the primary analysis event count, detailed censoring rules, missing-assessment procedures, sensitivity analyses, or a description of how disagreements in central review were resolved.

The registry labels the analysis as Cox proportional hazard and Log Rank, but does not document covariate adjustment, stratification, assessment of proportional hazards, or any interim analyses. It also does not report multiplicity procedures across the several outcomes and populations. These points limit independent evaluation of the analysis plan based on the registry entry alone.

The posted serious adverse-event counts and term-specific percentages have limited detail in this record summary. The listed term percentages do not have a stated denominator in the registry record described, and the death counts are not accompanied by causes or attribution. These limitations do not negate the reported statistics; they define what can and cannot be inferred from this registry record.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Trastuzumab Deruxtecan (DS-8201a) Versus Investigator's Choice for HER2-low Breast Cancer That Has Spread or Cannot be Surgically Removed [DESTINY-Breast04] — NCT03734029.
• Registry record last updated 2026-09-28; status last verified 2026-09.
• Results first posted 2023-06-15.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

23. Record Summary

NCT03734029, DESTINY-Breast04, is a completed phase 3 randomized, open-label, active-controlled trial sponsored by Daiichi Sankyo. Total enrollment was 557. It compared trastuzumab deruxtecan with physician’s choice of capecitabine, eribulin, gemcitabine, paclitaxel, or nab-paclitaxel in HER2-low, unresectable and/or metastatic breast cancer previously treated with chemotherapy.

The primary endpoint was PFS by blinded independent central review in the hormone receptor-positive cohort. Median PFS was 10.1 months for trastuzumab deruxtecan and 5.4 months for physician’s choice. The registry reports a hazard-ratio confidence interval below 1 and a p-value below < 0.001. The analysis is listed as Cox proportional hazard, with Log Rank as the method.

The registry also lists secondary PFS, overall survival, and response outcomes, alongside safety counts. Details on multiplicity, missing-data handling, original sample-size justification, and several analysis-plan elements are not reported in the registry record.