A Study of the Efficacy and Safety of Bevacizumab in Chinese Women With Newly Diagnosed, Previously Untreated Stage III or Stage IV Epithelial Ovarian, Fallopian Tube, or Primary Peritoneal Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of the Efficacy and Safety of Bevacizumab in Chinese Women With Newly Diagnosed, Previously Untreated Stage III or Stage IV Epithelial Ovarian, Fallopian Tube, or Primary Peritoneal Cancer (NCT03635489), a randomized phase 3 study. The primary outcome is Progression-Free Survival (PFS).

2. Clinical Question

3. Trial Design

This was a Phase III, randomized, double-blind, parallel-group study sponsored by Hoffmann-La Roche. The registry describes it as multicenter. Enrollment was 100. The study status is completed; the recorded start date is 2018-08-15 and the primary completion date is 2021-05-26.

The two groups received paclitaxel and carboplatin with either bevacizumab or placebo. Both descriptions specify six cycles of chemotherapy, with treatment on Day 1 of each 21-day cycle. Bevacizumab or placebo began at Cycle 2, continued during the combination phase, and then could continue as maintenance up to a maximum of 22 cycles, or until a stopping condition such as disease progression or unacceptable toxicity. The placebo-arm description also names withdrawal as a stopping condition. The registry’s bevacizumab-arm description is cut off after “unacceptable toxic”; any further wording about that arm’s stopping conditions is not reported in the registry record.

The primary efficacy analysis used the intent-to-treat population, defined in the record as all randomized participants regardless of whether they received assigned treatment. Participants were analyzed according to randomized assignment. The reported group sizes for the primary outcome were 49 in the placebo group and 51 in the bevacizumab group. The record also reports safety denominators that differ from these efficacy group sizes, so safety and efficacy counts should be read in their respective populations.

Progression-free survival was the primary outcome. Secondary outcomes included overall survival, objective response rate, duration of response, and patient-reported changes in abdominal pain, bloating, and physical, role, and social function. The registry record does not report the original sample-size justification, the detailed randomization procedure, or a full statistical analysis plan.

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: Participants were randomized with a 1:1 allocation ratio to receive Bevacizumab(B) + Paclitaxel (P) + Carboplatin (C) or Placebo + Paclitaxel + Carboplatin.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The intent-to-treat (ITT) population was defined as all randomized participants regardless of whether the assigned study treatment was received. For efficacy analyses, participants were analyzed according to their randomized treatment assignment.

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.

6. Primary Endpoint

The registry lists the primary outcome as Progression-Free Survival (PFS), measured over Randomization up to disease progression or death from any cause, whichever occurs first (up to approximately 24 months).

Progression-free survival measures the time from randomization until the first of disease progression or death from any cause. In this trial, progression was assessed by the Investigator using RECIST v1.1. The stated observation window was from randomization until progression or death, whichever occurred first, up to approximately 24 months.

As a time-to-event endpoint, progression-free survival incorporates both whether an event occurred and how long participants remained without a recorded event. This can be useful when follow-up durations vary or when some participants have not experienced the event by the end of their observed follow-up. In general, such analyses account for incomplete event times through methods for censored observations; the registry does not describe the detailed handling of censoring in this trial.

The registry reports the median progression-free survival in each group and hazard ratios from Cox regression. A median is the time by which half of the relevant event-time distribution has been reached. It describes a point on that distribution, rather than the experience of every participant. A hazard ratio summarizes the relative event rate over follow-up under the model; it is not a direct comparison of median times. Considering both measures helps distinguish the reported time reached by each group from the relative time-to-event comparison.

Progression-free survival is used in clinical trials because it captures progression as well as death, rather than waiting only for deaths to accumulate. That is a general statistical rationale, not a claim about the suitability of this endpoint for every setting. The registry’s primary endpoint definition and the reported results should be interpreted within the follow-up and assessment procedures documented in the record.

7. Results Posted in the Registry

• Placebo + Paclitaxel + Carboplatin vs Bevacizumab + Paclitaxel + Carboplatin: hazard ratio 0.30 (95% CI 0.17–0.53), reported p < 0.001 (as reported in the registry).
• Placebo + Paclitaxel + Carboplatin vs Bevacizumab + Paclitaxel + Carboplatin: hazard ratio 0.34 (95% CI 0.20–0.58), reported p < 0.001 (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.

Overall Survival (OS)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor; it was not recalculated.
• Result:
  • Placebo + Paclitaxel + Carboplatin vs Bevacizumab + Paclitaxel + Carboplatin: hazard ratio 0.60 (95% CI 0.30–1.21) (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 Placebo + Paclitaxel + Carboplatin vs Bevacizumab + Paclitaxel + Carboplatin and the control. The interval gives the range of effects that remain compatible with the data.

Objective Response Rate (ORR)

• 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:
  • Bevacizumab + Paclitaxel + Carboplatin: 95.8% (23/24) versus 92.9% (26/28) on control; risk difference 3.0 percentage points (95% CI -9.5 to 15.4), risk ratio 1.03 (0.90–1.18); unadjusted two-proportion z-test p = 0.646.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Bevacizumab + Paclitaxel + Carboplatin and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes add context to the primary progression-free survival result, but they do not all answer the same statistical question. Overall survival concerns time until death from any cause, while objective response rate concerns the proportion meeting a response definition. Duration of response and patient-reported outcomes address further dimensions of outcome. Agreement or disagreement across such endpoints can occur because they capture different events, use different analysis populations, or have different precision. The registry record does not report enough detail to determine which of these considerations explains any particular pattern.

The posted overall-survival estimate is less definitive than the primary Cox results: its confidence interval includes the no-difference value, and a p-value is not reported. The response comparison is also unadjusted and based on smaller reported group sizes than the primary efficacy analysis. The remaining listed secondary outcomes include duration of response, abdominal pain, bloating, and physical, role, and social function. Their results are not reported in the registry record, so no direction or magnitude can be inferred for them.

There are 11 posted secondary outcomes. When several outcomes are examined, the chance of obtaining at least one apparently notable result by chance can increase. In general, multiplicity procedures can control that risk, but they may also affect the interpretation and precision of individual comparisons. The registry record does not report whether a formal multiplicity strategy was used for the secondary outcomes. Consequently, unadjusted secondary results carry less weight as confirmatory evidence than a prespecified primary endpoint analyzed under a documented plan. This does not make secondary findings irrelevant; it means they are best treated as supportive or exploratory unless the record establishes their confirmatory status.

The secondary outcomes should therefore be read as complementary evidence, with attention to their distinct definitions, analysis populations, confidence intervals, and the absence of a reported multiplicity approach. They do not erase or independently validate the primary result, and missing results cannot be interpreted as neutral or favorable.

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

• Overall Survival (OS)
• Objective Response Rate (ORR)
• Duration of Response (DOR)
• Percentage of Participants Who Report a Clinically Meaningful Improvement in Patient-Reported Abdominal Pain
• Percentage of Participants Who Report a Clinically Meaningful Improvement in Patient-Reported Bloating
• Percentage of Participants Who Report a Clinically Meaningful Improvement in Patient-Reported Function (Physical)
• Percentage of Participants Who Report a Clinically Meaningful Improvement in Patient-Reported Function (Role)
• Percentage of Participants Who Report a Clinically Meaningful Improvement in Patient-Reported Function (Social)

9. Statistical Interpretation

10. Statistical Methods Explained

The registry identifies Cox regression as the method for the reported hazard-ratio analyses of progression-free survival. The following is a general explanation of these methods. A Cox model compares the event hazard between groups over follow-up. The hazard is the instantaneous event rate among participants who have not yet had the event. A hazard ratio below the no-difference value indicates a lower estimated hazard in the numerator group relative to the comparison group, given the ordering of the comparison. Here the comparison is placebo versus bevacizumab, and the reported estimates are below that reference value.

What does a Cox hazard ratio represent?

In general, a Cox hazard ratio is a relative, time-to-event measure. It is not the ratio of the two medians, nor does it state the probability that an individual participant will experience progression or death. It also does not directly give the amount of time gained by any participant. The model uses the ordering of event times and the participants still at risk over follow-up. A proportional-hazards interpretation assumes that the relative hazard is adequately represented by a constant comparison over time. Whether that assumption was assessed, and what diagnostics were used, are not reported in the registry record.

The record contains two Cox-regression results for the primary comparison: 0.30 with a 95% confidence interval of 0.17 to 0.53, and 0.34 with a 95% confidence interval of 0.20 to 0.58. Both have p < 0.001. The registry does not identify the distinction between the two analyses. Without that information, it is not possible to say whether they reflect different model specifications, analysis populations, or another reason. No such explanation should be inferred from the values alone.

How should medians and confidence intervals be read?

In general, a median time-to-event estimate marks the point at which half of the estimated event-time distribution has occurred. It is not the mean, and it does not summarize the full shape of the distribution. The registry reports medians for each group, but a median alone does not describe how risk changes during follow-up. Confidence intervals express uncertainty around the estimated quantity under the analysis method. They are not ranges containing a stated proportion of individual participants’ event times.

The primary outcome’s confidence interval for the placebo median is 9.53 to 15.05 months. For the bevacizumab median, the lower limit is 18.6 months and the upper limit is not reported in the registry record. These are confidence intervals for group-level median estimates, not individual prediction intervals. The hazard-ratio confidence intervals address a different quantity from the median confidence intervals, so they should not be compared as if they were measuring the same parameter.

What do the p-values add?

In general, a p-value measures how incompatible the observed data are with a specified null model, under the assumptions of the analysis. It is not the probability that the null hypothesis is true, and it does not measure the size or practical importance of an effect. The primary Cox results have p < 0.001, as reported. The registry’s classification of these analyses is superiority. That classification and the p-values do not remove uncertainty about the size of the effect, model assumptions, or the scope of the population to which the results may apply.

For overall survival, the record reports a hazard ratio and confidence interval but no p-value. It is therefore not appropriate to infer or recreate a sponsor-reported p-value from the other posted results. For objective response rate, an unadjusted comparison is reported. General interpretation of an unadjusted p-value should account for the possibility that many outcomes or analyses were examined, especially when no multiplicity adjustment is described.

Why do time-to-event and binary outcomes use different summaries?

Progression-free and overall survival are times until events, so medians and hazard ratios are natural summaries for the reported analyses. Objective response rate is a binary outcome: participants either meet the response definition or do not within the specified assessment context. It is summarized as a percentage, with confidence intervals, and the registry also posts an unadjusted comparison. These outcome types answer different questions. A response percentage does not replace a time-to-event outcome, and a hazard ratio does not describe the proportion responding.

The objective-response comparison includes a risk difference, risk ratio, odds ratio, and unadjusted p-value in the posted analysis data. These measures can differ in scale and precision, particularly when the observed percentages are high and the group sizes are limited. The registry does not report a sponsor analysis entry for this endpoint, so the posted unadjusted comparison should not be confused with a documented adjusted model.

What can be learned from the posted analyses?

The numerical results show the estimated group outcomes and their statistical uncertainty as recorded. They do not reveal all analysis decisions. Details such as stratification factors, handling of protocol deviations, missing-data procedures, censoring rules, model diagnostics, and the rationale for the two Cox estimates are not reported in the registry record. Those omissions limit the ability to reproduce the analyses or assess every assumption. The estimators and results should therefore be read alongside the precise endpoint definitions and the information the record does report, without filling gaps through conjecture.

11. Confidence Intervals and Precision

Confidence intervals describe statistical uncertainty around an estimate under the analysis method. They do not guarantee that the true value lies within the interval, and they are not a range of individual participant outcomes. A narrower interval generally indicates a more precise estimate on the scale being reported; a wider interval indicates more uncertainty. Precision depends on factors such as the amount of information available and, for time-to-event analyses, the number and timing of events. The registry record does not report all details needed to attribute the width of each interval to a specific cause.

For progression-free survival, the placebo-group median has a 95% confidence interval from 9.53 to 15.05 months. The bevacizumab-group median has a lower confidence limit of 18.6 months; its upper confidence limit is not reported in the registry record. The absence of that upper limit means the interval cannot be fully characterized from the posted result. For the hazard ratio, the registry gives two 95% confidence intervals: 0.17 to 0.53 and 0.20 to 0.58. Both are on the lower-than-no-difference side of the hazard-ratio scale, while their range of values still reflects uncertainty about the effect’s magnitude.

Secondary estimates show greater uncertainty in some respects. The overall-survival hazard-ratio interval extends from 0.30 to 1.21 and includes the no-difference value. The bevacizumab-group overall-survival median and its confidence limits are not reported in the registry record. The objective-response confidence intervals are broad for both groups, so the percentages should not be read as highly precise estimates of response rates beyond the analyzed participants.

A confidence interval is not a substitute for the analysis description. It cannot indicate whether missing data were handled appropriately, whether model assumptions were checked, or whether multiplicity was addressed. Those details are not reported in the registry record. Precision should therefore be considered together with the endpoint, analysis population, and limitations of the recorded methods.

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 planning figures are illustrations based on the observed effects, not the trial’s original design. For time-to-event endpoints, they give total event counts; for other endpoints, they give participants per arm. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between arms, a normal approximation, and an effect to detect set at the stated fraction of the observed effect. The trial’s own original sample-size assumptions and justification are not reported in the registry record.

In plain language, the 80% power column gives the illustrative amount of information needed for an analysis to have an 80% chance of detecting the assumed effect under those planning assumptions, if that effect and the model assumptions are correct. The 90% power column uses a higher chance of detecting that assumed effect and therefore generally requires more information. These are planning quantities, not a guarantee of a result.

For progression-free survival, the two posted sets of event-count illustrations give 22 and 27 total events for 80% power when planning for 100% of the observed effect, and 29 and 37 for 90% power. Planning for 75% of the observed effect gives 39 and 48 total events for 80% power, and 52 and 65 for 90% power. Planning for 50% gives 87 and 108 total events for 80% power, and 116 and 145 for 90% power. Overall-survival event-count illustrations are also posted. For objective response rate, the figures are participants per arm rather than total events. Across these planning illustrations, smaller effects need more participants or events to detect. These are not observed power calculations.

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 = 0.300, so ln(HR) = -1.2040.
2. d = 4 × (1.96 + 0.842)² / [ln(0.300)]² = 21.7, rounded up to 22 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 <- 0.3                              # 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(3635489)
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 = 22 (80% power), 29 (90% power)
 75% of observed effect: events = 39 (80% power), 52 (90% power)
 50% of observed effect: events = 87 (80% power), 116 (90% power)
Simulated power with 22 events: 0.767 (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 46 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 100 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 <- 12.32          # control-arm median, months (converted from the posted unit)
hr          <- 0.3          # 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: 38        p (1-sided)   0.0001
 Events: 10       ~HR at bound   0.0905
  Month: 20   P(Cross) if HR=1   0.0001
            P(Cross) if HR=0.3   0.0338
  IA 2: 67%                  Z   2.5114
      N: 46        p (1-sided)   0.0060
 Events: 20       ~HR at bound   0.3167
  Month: 32   P(Cross) if HR=1   0.0060
            P(Cross) if HR=0.3   0.5603
      Final                  Z   1.9930
      N: 46        p (1-sided)   0.0231
 Events: 29       ~HR at bound   0.4747
  Month: 60   P(Cross) if HR=1   0.0250
            P(Cross) if HR=0.3   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 registry’s reported Cox analyses are the sponsor-reported results for the primary time-to-event comparison. They give hazard ratios of 0.30 and 0.34, each with p < 0.001, and each with a 95% confidence interval below the no-difference value. The record does not explain why both analyses appear or describe how they differ. They are therefore best presented as two reported analyses, without selecting one as a preferred estimate or assigning an undocumented reason for the difference.

The posted median progression-free survival estimates are 12.32 months for placebo plus paclitaxel and carboplatin and 22.57 months for bevacizumab plus paclitaxel and carboplatin. The hazard ratios and medians are complementary summaries, not interchangeable calculations. The registry record does not contain participant-level event times from which to independently reproduce the Cox analyses. The sponsor’s own reported analysis therefore remains the relevant reported estimate; there is no separate recalculation to displace it.

For overall survival, the registry reports a hazard ratio of 0.60 with a 95% confidence interval of 0.30 to 1.21, but no p-value. The record does not provide a corresponding bevacizumab-group median. These missing details should remain explicit rather than being filled by a derived or assumed sponsor result. For objective response rate, the record posts an unadjusted comparison, but no reported-analysis entry. That distinction matters: a simple unadjusted calculation is not the same as a documented sponsor model or a complete statistical analysis plan.

16. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when a study evaluates more than one endpoint, comparison, subgroup, or analysis. In general, testing many hypotheses can increase the probability of at least one apparently positive finding even when no corresponding effect exists. Prespecifying a primary endpoint and using a multiplicity strategy for other confirmatory comparisons can help control that risk. The registry identifies progression-free survival as the primary outcome and lists 11 secondary outcomes, but a formal adjustment strategy for secondary testing is not reported in the registry record.

The record does not describe whether analyses of overall survival, response, duration of response, or patient-reported outcomes were intended to be confirmatory or exploratory. It also does not report a testing sequence, alpha allocation, or adjustment method. The unadjusted p-value for objective response rate should therefore be interpreted as unadjusted, not as evidence that any multiplicity issue was controlled. No p-value is reported for overall survival. The two reported Cox results for progression-free survival are not explained, so their relationship to a prespecified testing strategy is also unclear.

The primary analysis population is defined as all randomized participants, analyzed according to randomized treatment assignment regardless of whether assigned treatment was received. This is an intent-to-treat approach as described in the registry. That definition is useful for understanding the primary efficacy population, but the record does not report how missing outcome assessments, loss to follow-up, treatment discontinuation, or other incomplete observations were handled. The detailed censoring rules for time-to-event outcomes are also not reported.

The response outcome has posted group sizes of 28 and 24, whereas the primary outcome has group sizes of 49 and 51. The registry does not explain the difference or fully specify the response-analysis population in the information recorded. It is therefore not possible to determine whether the difference reflects eligibility for response assessment, missing assessments, or another criterion. The safety denominators differ again. These are reasons to keep each outcome’s population distinct rather than assuming the same participants contributed to every analysis.

These reporting gaps do not establish that a method was not used; they mean its use cannot be confirmed from the registry record. They limit reproducibility and constrain how strongly secondary findings can be interpreted.

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: myelosuppression (10.0%), platelet count decreased (10.0%), anaemia (6.0%), neutrophil count decreased (6.0%), ileus (4.1%).

The registry reports serious adverse events for 16 of 50 participants at risk in the placebo plus paclitaxel and carboplatin group and 17 of 49 at risk in the bevacizumab plus paclitaxel and carboplatin group. Other adverse events affected 49 of 50 and 49 of 49 participants at risk, respectively. Death counts are 19 and 13, respectively. The record does not state in these figures whether deaths were attributed to treatment, disease, or another cause, so the counts should not be interpreted as treatment-related deaths.

The listed serious adverse-event terms include myelosuppression and decreased platelet count, each reported as 10.0%; anaemia and decreased neutrophil count, each 6.0%; and ileus and intestinal obstruction, each 4.1%. The listing does not assign these percentages to a specific treatment group in the posted information, so no group-specific comparison should be inferred from it.

These counts describe observed safety events in the recorded risk groups. They are not, by themselves, a causal comparison: interpretation also depends on event definitions, follow-up, attribution, severity, and the way events were collected. The registry record does not report all of those details in the information summarized. The safety denominators should not be substituted for the primary efficacy population sizes.

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

The primary results show that the registry’s estimated progression-free survival was longer in the bevacizumab group than in the placebo group, and the reported Cox hazard ratios favor the bevacizumab group in direction. The confidence intervals and p-values quantify statistical evidence and uncertainty under the reported analyses. They do not establish the effect for every participant, show how an individual outcome would change, or determine the clinical importance of the difference.

The results apply to the trial population and regimens described: Chinese women with newly diagnosed, previously untreated Stage III or Stage IV epithelial ovarian, fallopian tube, or primary peritoneal cancer, assigned to the specified combinations. They do not automatically apply to populations, disease settings, or treatment regimens not represented in the registry description. The record reports randomization and double masking, but it does not supply enough detail to assess every implementation feature or analysis assumption.

The primary endpoint is progression-free survival, not overall survival. The overall-survival estimate has a confidence interval that includes the no-difference value, and the record does not report a p-value or the bevacizumab-group median. This means the posted secondary overall-survival evidence is less precise and does not resolve the comparison. Objective response rate is a separate outcome, based on smaller reported analysis groups and an unadjusted comparison. Results for several other listed secondary outcomes are not reported in the registry record.

Statistical significance is not a measure of benefit size, certainty about every analysis choice, or proof of a mechanism. Likewise, a confidence interval does not capture uncertainties from unreported missing-data methods, multiplicity handling, or model diagnostics. The results are best understood as the estimates recorded for this trial, with the primary time-to-event finding distinguished from the less complete secondary evidence and the unreported analysis details.

19. Why This Trial Matters Statistically

This trial is statistically notable for its randomized, double-blind comparison of two regimens that share paclitaxel and carboplatin and differ by bevacizumab versus placebo. That design provides a direct between-group framework for the primary question, while the intent-to-treat definition ties the efficacy analysis to randomized assignment. The registry’s description of the primary endpoint also makes clear that progression and death both count as events.

The posted primary results illustrate why a time-to-event analysis often reports more than one summary. The medians describe points in each group’s event-time distribution, while Cox hazard ratios summarize a relative event hazard over follow-up. The two types of estimate are related to the same clinical endpoint but are not interchangeable. Their confidence intervals communicate uncertainty on different scales.

The record also illustrates the importance of distinguishing primary from secondary evidence. A primary result can be statistically clear while secondary endpoints remain uncertain, use different populations, or have results that are not reported. Multiple secondary comparisons make it important to know whether a multiplicity strategy was planned. That strategy is not reported, and the record does not explain why it contains two Cox estimates for progression-free survival.

Finally, the safety data have separate denominators from the efficacy analyses, and the registry’s reported serious-event terms are not allocated by treatment group in the available listing. Careful statistical reading therefore depends on preserving the endpoint, analysis population, and comparison attached to each number. The record supports a focused account of the posted estimates, but not a complete reconstruction of every design or analytic decision.

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.

Several limitations are specific to the information recorded for this trial. The registry reports two Cox hazard-ratio analyses for progression-free survival but does not explain their differences. Both are statistically significant as reported, yet the absence of an analysis description prevents determining which estimate was primary, whether they represent distinct model specifications, or how they relate to any prespecified testing sequence.

The upper confidence limit for the bevacizumab-group progression-free survival median is not reported. For overall survival, the bevacizumab-group median is not reported, and the record gives no p-value for the hazard-ratio comparison. The primary completion period is stated, but detailed follow-up and event-count information needed to assess the maturity of each time-to-event analysis is not included in the record.

The objective-response analysis reports group sizes of 28 and 24, which differ from the primary endpoint group sizes. The record does not explain the analysis-population difference or report the reason those participants were included. It also does not describe missing-data handling, response assessment completeness, or whether the response comparison was prespecified as confirmatory. The remaining listed patient-reported outcomes and duration of response do not have results reported in the registry record.

The registry record does not report the original sample-size justification, a complete statistical analysis plan, multiplicity handling, detailed censoring rules, Cox-model diagnostics, or full missing-data procedures. Safety counts are available, but the event listing does not assign the named serious adverse-event percentages by arm, and the death counts are not described as treatment-related. These reporting limitations restrict independent reproduction and prevent filling in details that the record does not state.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: A Study of the Efficacy and Safety of Bevacizumab in Chinese Women With Newly Diagnosed, Previously Untreated Stage III or Stage IV Epithelial Ovarian, Fallopian Tube, or Primary Peritoneal Cancer — NCT03635489.
• Registry record last updated 2024-10-03; status last verified 2024-06.
• Results first posted 2023-03-17.
• 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 describes a completed, randomized, double-blind, parallel-group Phase III study sponsored by Hoffmann-La Roche, with enrollment of 100. It compared bevacizumab plus paclitaxel and carboplatin with placebo plus paclitaxel and carboplatin in Chinese women with newly diagnosed, previously untreated Stage III or Stage IV epithelial ovarian, fallopian tube, or primary peritoneal cancer.

Progression-free survival was the primary endpoint, defined as time from randomization to investigator-assessed progression under RECIST v1.1 or death from any cause. The posted medians and Cox hazard ratios favor the bevacizumab group in direction. The registry also reports overall survival and objective response rate, while results for several other listed secondary outcomes are not reported in the registry record.

The record leaves important analytic details unreported, including the reason for two primary Cox estimates, multiplicity handling, and missing-data procedures. Those gaps qualify interpretation of the posted estimates and limit independent reconstruction of the analyses.