Durvalumab With or Without Olaparib as Maintenance Therapy After First-Line Treatment of Advanced and Recurrent Endometrial Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Durvalumab With or Without Olaparib as Maintenance Therapy After First-Line Treatment of Advanced and Recurrent Endometrial Cancer (NCT04269200), a randomized phase 3 study. The primary outcome is Progression-free Survival (PFS) According to RECIST 1.1, Based on Investigator Assessments.

2. Clinical Question

The clinical question is evaluated through time-to-event comparisons rather than by comparing only the median survival times. The reported hazard ratios summarize the relative rate of progression or death over follow-up under the Cox regression analysis. A hazard ratio below 1 is in the direction of a lower event hazard in the treatment arm, but its uncertainty and the analysis context matter when interpreting that direction.

The registry also reports the China cohort separately. Its results should be read as cohort-specific estimates, not as a replacement for the global analysis. The registry record does not report a formal test of whether treatment effects differ between the global and China cohorts.

3. Trial Design

This is a randomized, parallel-group, phase III study. The registry lists quadruple masking and an enrollment of 805. The study is titled DUO-E and concerns first-line carboplatin and paclitaxel with durvalumab, followed by maintenance durvalumab with or without olaparib, in newly diagnosed advanced or recurrent endometrial cancer.

The three global arms distinguish chemotherapy plus placebo from chemotherapy plus durvalumab, and from chemotherapy plus durvalumab followed by maintenance durvalumab and olaparib. The registry describes the control as receiving placebo for both durvalumab and olaparib. The active treatment arms therefore allow comparison of the durvalumab regimen with control and of the durvalumab-plus-olaparib maintenance regimen with control. They do not, on the posted primary comparisons, establish a direct comparison between the two active regimens.

The primary outcome is progression-free survival (PFS), assessed by investigators according to RECIST 1.1. The registry describes assessments at baseline, every 9 weeks up to 18 weeks, then every 12 weeks until objective radiological disease progression. The primary assessment was reported through the 12 Apr 2023 data cut-off for the global analysis and the 08 Jul 2024 data cut-off for the China cohort, with follow-up up to 50 months.

The record lists a study start of 2020-05-05, primary completion of 2024-07-08, and status as active, not recruiting. Enrollment, allocation, masking, interventions, outcome definitions, and reported estimates are available in the registry record. Details such as the trial's original sample-size justification, planned interim analyses, and full statistical analysis plan 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: It was planned that approximately 699 women from the Global cohort and 129 women from the China cohort, with newly diagnosed advanced or recurrent endometrial cancer, were to receive durvalumab/placebo in combination with platinum-based chemotherapy followed by maintenance durvalumab/placebo (same treatment as during chemotherapy phase) and olaparib/placebo in a 1:1:1 ratio.

Global Cohort

Reasons for not completing this period:

China Cohort

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Full Analysis Set (FAS) consisting of all patients randomised as part of global enrolment including patients from China who were randomised before the global recruitment was closed (up to and including 20 Apr 2022). China FAS includes all patients who were randomised at sites in China and Hong Kong.

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: Global cohort: 241 SoC, 238 SoC + Durvalumab, 239 SoC + Durvalumab + Olaparib China cohort: 44 SoC, 43 SoC + Durvalumab, 42 SoC + Durvalumab + Olaparib There were 42 patients included in both Global and China cohorts (15 SOC, 14 SoC + Durvalumab, 13 SoC + Durvalumab + Olaparib).

6. Primary Endpoint

The registry lists the primary outcome as Progression-free Survival (PFS) According to RECIST 1.1, Based on Investigator Assessments, measured over At baseline, every 9 weeks (wks) up to 18 wks, then every 12 wks until objective radiological disease progression. Assessed until 12 Apr 2023 DCO (08 Jul 2024 DCO for China cohort), up to 50 months.

The primary endpoint is progression-free survival according to RECIST 1.1, based on investigator assessments. It measures the time until objective radiological disease progression or death, as defined for the trial. The registry reports the median in months and a 95% confidence interval for each cohort and treatment arm, alongside Cox regression hazard ratios for specified comparisons.

Time-to-event endpoints are useful when participants have different lengths of follow-up or when some have not experienced the event by the analysis cut-off. The median gives a time point in the event-time distribution: it describes when the estimated event-free proportion reaches the midpoint. It does not describe every participant's experience or the full shape of the event-time curves.

The hazard ratio uses information across the observed follow-up, rather than comparing only the reported medians. It represents a relative event-rate contrast under the Cox model. A hazard ratio below 1 points toward a lower event hazard in the treatment group relative to control, while the confidence interval indicates the range of estimates compatible with the data and model. Interpretation should consider that interval, follow-up, and the population analyzed.

PFS is used as a measure of disease control in this setting because it incorporates both radiological progression and death. It is distinct from overall survival, which is listed as a secondary outcome. The registry record's primary endpoint description does not itself establish a treatment effect on overall survival or other outcomes.

7. Results Posted in the Registry

• Global Cohort - SoC vs Global Cohort - SoC + Durvalumab: hazard ratio 0.71 (95% CI 0.57–0.89), reported p = 0.003 (as reported in the registry).
• Global Cohort - SoC vs Global Cohort - SoC + Durvalumab + Olaparib: hazard ratio 0.55 (95% CI 0.43–0.69), reported p < 0.001 (as reported in the registry).
• China Cohort - SoC vs China Cohort - SoC + Durvalumab: hazard ratio 0.73 (95% CI 0.42–1.25) (as reported in the registry).
• China Cohort - SoC vs China Cohort - SoC + Durvalumab + Olaparib: hazard ratio 0.96 (95% CI 0.57–1.61) (as reported in the registry).

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Time From Randomisation to Second Progression or Death (PFS2) Based on Local Standard Clinical Practice

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Regression, Cox); it was not recalculated.
• Result:
  • Global Cohort - SoC vs Global Cohort - SoC + Durvalumab: hazard ratio 0.80 (95% CI 0.59–1.07) (as reported in the registry).
  • Global Cohort - SoC vs Global Cohort - SoC + Durvalumab + Olaparib: hazard ratio 0.55 (95% CI 0.40–0.76) (as reported in the registry).
• Interpretation:
  • The 95% confidence interval includes no difference (a ratio of 1), so these data do not show a statistically significant difference between Global Cohort - SoC vs Global Cohort - SoC + Durvalumab and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval excludes no difference (a ratio of 1), so the hazard of the event is statistically lower for Global Cohort - SoC vs Global Cohort - SoC + Durvalumab + Olaparib than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Objective Response Rate (ORR) Based on Investigator Assessment

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Global Cohort - SoC + Durvalumab: 61.9% (125/202) versus 55.1% (109/198) on control; risk difference 6.8 percentage points (95% CI -2.8 to 16.5), risk ratio 1.12 (0.95–1.33); unadjusted two-proportion z-test p = 0.166.
  • Global Cohort - SoC + Durvalumab + Olaparib: 63.6% (117/184) versus 55.1% (109/198) on control; risk difference 8.5 percentage points (95% CI -1.3 to 18.4), risk ratio 1.16 (0.98–1.36); unadjusted two-proportion z-test p = 0.090.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Global Cohort - SoC + Durvalumab and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Global Cohort - SoC + Durvalumab + Olaparib and the control. The interval gives the range of effects that remain compatible with the data.

Change From Baseline in Physical Functioning Score of the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire for Cancer Patients 30 (EORTC QLQ-C30)

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard errors.
• Result:
  • Global Cohort - SoC + Durvalumab: mean -3.60 versus -5.30 on control; difference 1.70 (95% CI -1.14 to 4.54), standardized effect d = 0.12; z-test p = 0.241.
  • Global Cohort - SoC + Durvalumab + Olaparib: mean -5.90 versus -5.30 on control; difference -0.60 (95% CI -3.39 to 2.19), standardized effect d = -0.04; z-test p = 0.673.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Global Cohort - SoC + Durvalumab and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Global Cohort - SoC + Durvalumab + Olaparib and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes span time-to-event measures, response, and quality of life, so they add different perspectives rather than repeating the primary PFS question. The posted results include a further progression endpoint, response rate, and change in physical functioning, alongside other outcomes listed in the registry. Their estimates are not interchangeable: a response measure describes tumor response under its definition, a later progression measure extends the time horizon, and a quality-of-life score concerns a reported domain.

As a broad pattern, the global cohort's posted results do not all provide the same degree or kind of statistical evidence. Some comparisons have estimates and intervals that appear more separated from their null values, while others are less precise or include values compatible with no difference. The China cohort also has its own arm-level summaries, and its estimates should not be treated as a formal test of whether results differ from the global cohort.

Secondary outcomes can help assess whether the primary pattern is accompanied by findings in related domains, but they do not replace the primary endpoint. They also involve multiple outcomes and comparisons. When many hypotheses are examined, the chance of at least one apparently notable result can increase unless the analysis plan accounts for multiplicity. The registry record does not report a multiplicity adjustment strategy for these secondary analyses. Unadjusted secondary results therefore carry less confirmatory weight and are best regarded as supporting or exploratory evidence.

Differences in analysis methods and populations also matter. For example, the response analysis uses participants assessed for that outcome, while the primary analysis uses the Full Analysis Set as described in the registry. The record does not establish that every secondary result was prespecified for confirmatory testing. Taken together, secondary outcomes can inform context, but they should be interpreted alongside the primary result, uncertainty, and reporting limitations rather than as independent proof of an overall effect.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Overall Survival (OS) Analysis
• Time From Randomisation to Second Progression or Death (PFS2) Based on Local Standard Clinical Practice
• Objective Response Rate (ORR) Based on Investigator Assessment
• Duration of Response (DoR) Based on Investigator Assessment
• Time From Randomisation to First Subsequent Therapy or Death (TFST)
• Time From Randomisation to Second Subsequent Therapy or Death (TSST)
• Time From Randomisation to Discontinuation of Treatment or Death (TDT)
• Serum Concentration of Durvalumab

9. Statistical Interpretation

10. Statistical Methods Explained

What does a median PFS represent?

For a time-to-event outcome, the median is the time by which half of the estimated population has experienced the event, with the remainder estimated to be event-free beyond that point. It is a compact description of the event-time distribution, not a complete account of it. Participants may have different follow-up durations, and some may still be event-free at the data cut-off. Time-to-event methods can retain information from such participants through censoring, provided the analysis assumptions are appropriate.

The registry reports medians and 95% confidence intervals for the arms. A median is not an average, and comparing two medians alone does not use all the follow-up information in the same way as a time-to-event model. Nor does a difference between medians necessarily represent a constant treatment effect throughout follow-up.

How should a hazard ratio be read?

The reported comparisons use Cox regression and give hazard ratios. A hazard is the rate of experiencing the event among participants who have not yet experienced it, at a given time. A hazard ratio compares these rates between arms over the observed period, under the model. A value below 1 points toward a lower hazard in the treatment arm relative to control; a value of 1 represents no relative difference in the modeled hazard. This is a relative measure, not an absolute probability of progression or death and not a ratio of the median times.

The Cox model is commonly interpreted using a proportional-hazards assumption: the relative hazards are treated as sufficiently stable over time for one hazard ratio to summarize the contrast. Whether this assumption was assessed, and what diagnostics or alternative analyses were used, are not reported in the registry record. If treatment effects vary over time, a single hazard ratio may not fully describe that pattern.

What does the confidence interval add?

A confidence interval describes statistical uncertainty around an estimate under the analysis model. Wider intervals indicate less precision. For a hazard ratio, an interval containing 1 includes the no-relative-difference value; an interval entirely below 1 is consistent with a lower hazard over the range represented by the interval. This interpretation is not a substitute for considering multiplicity, the analysis population, or the clinical context.

The global primary comparisons have intervals below 1, while the China-cohort intervals include 1 and are wider. That contrast is consistent with greater uncertainty in the China estimates, but it does not prove that the true effects differ across cohorts. The record does not report a formal interaction test for that question.

Why distinguish reported from recalculated results?

The primary hazard ratios and their inferential results are sponsor-reported Cox regression analyses. The posted interpretation should rely on those analyses rather than trying to reconstruct a time-to-event model from arm medians and confidence intervals. Medians do not contain the event-time information needed to reproduce the Cox model. The registry record does not flag a discrepancy between the reported primary analysis and any recalculation.

For some non-primary outcomes, the displayed comparison estimates may include simple unadjusted recalculations from posted arm summaries. Such calculations can help describe the scale of a contrast, but they are not equivalent to the sponsor's full model-based analyses. In particular, they may not account for covariates, repeated measures, censoring, or the analysis population used in the sponsor's model. Where a sponsor analysis is reported, it should be distinguished from these simple recalculations.

Why can over-dispersion matter in other trial outcomes?

General statistical explanation: binary and count outcomes can vary more than a simple model assumes. This is called over-dispersion. If a model assumes less variability than the data contain, its standard errors can be too small and its confidence intervals too narrow. The primary PFS endpoint is time-to-event, so over-dispersion is not the central issue for the reported Cox comparisons. The registry record does not provide enough detail to assess every model assumption for every outcome.

These methods describe how to interpret the posted estimates, not a claim that an assumption has been satisfied. The record's summaries support an interpretation of reported associations in the analyzed populations; they do not reveal every modeling decision or the full individual-level data.

11. Confidence Intervals and Precision

Precision is reflected in the width of a confidence interval: a narrower interval gives a more concentrated range of estimates, while a wider interval leaves more uncertainty about the underlying effect. The global durvalumab-versus-control hazard ratio is 0.71 with a 95% confidence interval of 0.57 to 0.89. The global durvalumab-plus-olaparib-versus-control estimate is 0.55 with a 95% confidence interval of 0.43 to 0.69. Each interval is entirely below 1, so the reported model-based estimates are compatible with a lower hazard across the interval's range.

The China-cohort estimates have wider intervals: 0.42 to 1.25 for durvalumab versus control, and 0.57 to 1.61 for durvalumab plus olaparib versus control. Both include 1 and encompass values on either side. That means the estimates are comparatively imprecise and do not sharply distinguish a lower hazard from no relative difference or a higher hazard within the interval.

Confidence intervals for the arm medians also show uncertainty. For example, the global control median is 9.6 months with a 95% confidence interval of 9.0 to 9.9 months, whereas the global durvalumab-plus-olaparib median is 15.1 months with a 95% confidence interval of 12.6 to 20.7 months. Median intervals and hazard-ratio intervals answer related but distinct questions; one should not infer the hazard ratio by comparing median intervals.

Precision depends on the amount and distribution of event information, not simply on the number enrolled. Follow-up duration and censoring also affect the information available. The registry provides arm counts and estimates but does not report all details needed to assess how each factor contributed to precision.

12. 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 sample-size displays are planning illustrations based on the observed effects, not the trial's original design calculations. For time-to-event comparisons, the entries are total numbers of events. For other outcomes, the entries are participants per arm. The unit therefore depends on the outcome: event totals should not be read as participant counts, and per-arm participant figures should not be read as totals across arms.

The illustrations assume a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. They show planning for 100%, 75%, and 50% of the observed effect. The trial's own original design assumptions and sample-size justification are not reported in the registry record.

The 80% power column describes the illustrative number of events, or participants per arm as applicable, that would be needed under those assumptions for an 80% chance of detecting the chosen effect if that effect were the true one. The 90% power column gives the corresponding planning requirement for a higher chance of detection. These are planning quantities, not guarantees about any future study.

Smaller effects are harder to distinguish from random variation, so the illustrations generally require more events or participants when planning for a smaller fraction of the observed effect. These figures do not describe the power achieved by this trial after its results were observed. They are not observed power.

Interpret each entry in the context of the endpoint and contrast. A time-to-event illustration is driven by the required event count, whereas a binary or continuous-outcome illustration is expressed as participants per arm. The displayed planning curve should be read in the same units as its associated outcome.

13. 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 = 0.710, so ln(HR) = -0.3425.
2. d = 4 × (1.96 + 0.842)² / [ln(0.710)]² = 267.7, rounded up to 268 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 <- 0.71                              # 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(4269200)
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 = 268 (80% power), 359 (90% power)
 75% of observed effect: events = 476 (80% power), 637 (90% power)
 50% of observed effect: events = 1,071 (80% power), 1,434 (90% power)
Simulated power with 268 events: 0.777 (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.

14. 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 439 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 805 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 <- 9.6          # control-arm median, months (converted from the posted unit)
hr          <- 0.71          # 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: 326         p (1-sided)   0.0001
 Events: 121        ~HR at bound   0.5085
   Month: 18    P(Cross) if HR=1   0.0001
             P(Cross) if HR=0.71   0.0338
   IA 2: 67%                   Z   2.5114
      N: 440         p (1-sided)   0.0060
 Events: 241        ~HR at bound   0.7235
   Month: 28    P(Cross) if HR=1   0.0060
             P(Cross) if HR=0.71   0.5603
       Final                   Z   1.9930
      N: 440         p (1-sided)   0.0231
 Events: 362        ~HR at bound   0.8108
   Month: 60    P(Cross) if HR=1   0.0250
             P(Cross) if HR=0.71   0.9000



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

15. Analyses Reported by the Sponsor

The primary comparisons are reported as Cox regression hazard ratios: 0.71 (95% CI 0.57 to 0.89; p = 0.003) for global durvalumab versus control, and 0.55 (95% CI 0.43 to 0.69; p < 0.001) for global durvalumab plus olaparib versus control. The corresponding China-cohort hazard ratios are 0.73 (95% CI 0.42 to 1.25) and 0.96 (95% CI 0.57 to 1.61); p-values are not reported for these comparisons.

These sponsor analyses are the appropriate estimates to use for the primary time-to-event comparisons. A Cox model uses event-time information that cannot be reconstructed from the posted medians alone. The registry record does not flag a discrepancy between the sponsor's primary estimates and a recalculation.

Simple recalculations for other outcomes, where displayed, are unadjusted descriptions derived from posted summary statistics. They may differ from sponsor analyses because they do not reproduce the full model or its assumptions. Such recalculations should not displace a sponsor-reported analysis. The record does not report every detail of the underlying statistical analysis plan, so the posted results should be interpreted with that limitation in mind.

16. Multiplicity, Missing Data and Analysis Populations

The primary endpoint includes more than one treatment-versus-control comparison, and the registry reports multiple secondary outcomes and analyses. Testing multiple hypotheses can increase the chance of a false-positive result if each is interpreted without accounting for the full family of tests. A multiplicity strategy may control that risk by specifying testing order, adjustment, or another prespecified approach.

The registry record does not report whether the primary comparisons were adjusted for multiplicity, what testing hierarchy was planned, or how multiplicity was handled across secondary outcomes. The reported p-values should therefore be understood as the values recorded for those comparisons, not assumed to have a particular family-wise error interpretation. The record also does not report interim-analysis plans or whether any interim testing boundaries were used.

The primary population is described as the Full Analysis Set consisting of patients randomized as part of global enrollment, including patients from China randomized before global recruitment closed. The record gives a separate definition for the China Full Analysis Set. It does not report detailed exclusions, handling of protocol deviations, or the exact analysis-set implementation for every outcome.

Missing-data handling can affect estimates, especially when participants have incomplete follow-up or do not complete a questionnaire. For time-to-event outcomes, censoring is part of the analysis, but the record does not provide enough detail to assess censoring patterns or the assumptions used. For secondary outcomes, the registry record does not report a complete account of imputation, missing-data sensitivity analyses, or how missing observations were handled in each model.

These gaps do not negate the posted results, but they limit independent assessment of the complete inferential strategy. The full statistical analysis plan, detailed missing-data methods, and multiplicity procedures are not reported in the registry record.

17. 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: anaemia (9.8%), alanine aminotransferase increased (4.9%), aspartate aminotransferase increased (4.9%), pyrexia (4.9%), platelet count decreased (4.5%).

The registry lists serious adverse events and other adverse events for global and China cohort arms. In the global cohort, serious adverse events affected 73 participants in the control group, 73 in the durvalumab group, and 85 in the durvalumab-plus-olaparib group. The corresponding numbers at risk for this reporting are 236, 235, and 238. Other adverse events affected 233, 228, and 236 participants, respectively, with 236, 235, and 238 at risk.

In the China cohort, serious adverse events affected 14 participants in control, 8 in the durvalumab group, and 19 in the durvalumab-plus-olaparib group. The listed numbers at risk are 44, 41, and 41. Other adverse events affected 44, 41, and 39 participants, respectively.

The registry also lists anaemia at 9.8% as the most frequent serious adverse event term, followed by alanine aminotransferase increased and aspartate aminotransferase increased at 4.9% each, pyrexia at 4.9%, and platelet count decreased and dizziness at 4.5% each. These summaries do not provide event timing, duration, severity distribution, or a causal assessment for each event. Counts across groups should not be added to infer unique participants, and these registry summaries alone do not establish comparative safety.

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

The global primary results show reported Cox hazard ratios below 1 for both active comparisons with control, with confidence intervals that remain below 1. The posted estimates are consistent with lower progression-or-death hazards under the fitted Cox analyses for those comparisons. The global arm medians also differ, but medians and hazard ratios summarize the data in different ways.

The China-cohort estimates are more uncertain: both confidence intervals include 1 and span a range of relative effects. This does not prove that the treatments have no effect in that cohort. It indicates that the registry's estimates do not precisely establish the direction or size of the relative hazard in that group. Nor do differences between the global and China point estimates establish that effects vary by cohort; a formal comparison is not reported.

The results do not establish that every patient will experience the median outcome time, or that the estimated relative hazard applies identically to every participant or at every point in follow-up. They do not by themselves answer questions about overall survival, quality of life, or each secondary outcome. Those outcomes require their own estimates and uncertainty assessments.

The primary findings also do not settle questions about the complete statistical plan, multiplicity, missing data, or model assumptions, because relevant details are not reported in the registry record. Finally, group-level adverse-event summaries describe reported counts and do not provide a complete individualized assessment of risks. The appropriate reading is bounded: these are trial estimates for the specified populations, endpoints, and analyses, with uncertainty described by the posted intervals.

19. Why This Trial Matters Statistically

DUO-E is a randomized phase III study with a three-arm structure that separates the addition of durvalumab from the maintenance regimen that also includes olaparib. That structure makes the global comparisons relevant to distinct treatment questions: one evaluates the durvalumab-containing approach against control, while another evaluates the combination maintenance approach against control.

Statistically, the study illustrates why time-to-event analyses should not be reduced to medians alone. The reported Cox hazard ratios use event-time information across follow-up, while the medians provide an interpretable summary of when events accrue. Their confidence intervals add important context, particularly because the global and China cohorts have different levels of precision.

The registry also demonstrates the distinction between a larger global analysis and a separate cohort estimate. A smaller cohort can yield wide intervals even when its point estimate resembles an estimate elsewhere. Cross-cohort differences in point estimates are not themselves evidence of an interaction, and the record does not report a formal test of treatment-effect differences between cohorts.

Finally, the range of secondary outcomes emphasizes that a trial can ask several related but non-identical questions. Their interpretation depends on multiplicity, analysis populations, and missing-data methods, details that are not fully reported in the registry record. Reading the primary result together with its uncertainty and these limits is central to sound statistical interpretation.

20. 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 reports a primary analysis cut-off of 12 Apr 2023 for the global analysis and a later 08 Jul 2024 cut-off for the China cohort. The different cut-offs mean the cohort summaries do not represent identical data maturity. The reported follow-up is up to 50 months, but the record does not provide complete follow-up distributions or detailed censoring information.

The primary outcome is investigator-assessed according to RECIST 1.1. The registry describes scheduled assessments, but does not report independent central review results or a comparison between investigator and central assessment. It also does not report detailed diagnostics for the Cox model or whether proportional hazards were evaluated.

The global analysis population includes eligible China participants randomized before global recruitment closed, while China cohort results are also presented separately. This distinction should be kept in mind when reading the estimates. The registry does not provide a formal interaction analysis comparing cohort effects, nor does it report detailed information about subgroup-specific power.

Interpretation is further limited by details not reported in the registry record: the original sample-size justification, planned interim analyses, multiplicity procedures, full missing-data methods, and the complete statistical analysis plan. Several secondary outcomes involve different measurement scales and analysis populations; their estimates should not be treated as one unified result. The posted summaries support interpretation of the listed estimates, but cannot answer every question that would require the protocol, analysis plan, or participant-level data.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: Durvalumab With or Without Olaparib as Maintenance Therapy After First-Line Treatment of Advanced and Recurrent Endometrial Cancer — NCT04269200.
• Registry record last updated 2026-11-05; status last verified 2026-11.
• Results first posted 2025-12-23.
• 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.

24. Record Summary

The registry record describes DUO-E, a randomized, quadruple-masked, parallel-group phase III study of first-line platinum-based chemotherapy with durvalumab followed by maintenance durvalumab with or without olaparib in advanced or recurrent endometrial cancer. The record lists enrollment of 805 and status as active, not recruiting.

The primary endpoint is investigator-assessed PFS according to RECIST 1.1. Global cohort sponsor-reported Cox comparisons favor lower estimated progression-or-death hazards for both active regimens relative to control, with 95% confidence intervals below 1. China-cohort estimates are less precise, with intervals that include 1. The registry also reports secondary outcomes and adverse-event summaries.

Multiplicity procedures, detailed missing-data methods, and the original design assumptions are not reported in the registry record. Interpretation should therefore keep the posted estimates, their uncertainty, and those reporting limits in view.