Pembro/Carbo/Taxol in Endometrial Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Pembro/Carbo/Taxol in Endometrial Cancer (NCT02549209), a na phase 2 study. The primary outcome is Objective Response Rates (ORR).

2. Clinical Question

In plain terms, the study asked whether objective tumor responses were observed among people receiving this three-drug combination in the specified population. It defined a complete response as disappearance of all lesions in two consecutive observations at least 4 weeks apart. A partial response required a decrease in tumor burden of at least 50% relative to baseline, confirmed by a consecutive assessment at least 4 weeks after the first documentation. Stable disease was defined as not meeting the complete- or partial-response criteria, in the absence of immune-related progressive disease.

Those definitions specify how response was classified; they do not establish how many participants met either response category. The primary endpoint is a proportion, rather than a direct comparison between treatment groups. Because the study has no control group, its response findings would describe outcomes in the enrolled cohort and would not by themselves isolate the contribution of any one drug or establish a comparative treatment effect.

3. Trial Design

This was a phase II, single-group, open-label, multicenter study. The registry record lists an enrollment of 46 and a completed status. The study start date is 2017-08-22, and the primary completion date is 2019-12-12. The brief summary describes the intended population as subjects with measurable advanced or recurrent endometrial cancer. The listed conditions are endometrial cancer and endometrial adenocarcinoma.

Participants received pembrolizumab in combination with carboplatin and paclitaxel. For participants with no prior therapy, the regimen description lists pembrolizumab 200 mg, paclitaxel 175mg/m2, and carboplatin at an AUC of 6. The record says participants with prior external beam radiation therapy and/or platinum-based chemotherapy had to initiate paclitaxel and carboplatin at reduced doses. The available description ends before giving the full dose details, so the reductions and any other dosing rules are not reported in the registry record.

The study was not randomized and had no masking. There is no parallel comparator arm. As a result, it is a cohort description rather than a controlled estimate of the difference between treatments. The open-label design means the registry does not describe masking of participants or study personnel. The registry record does not report further details on site count, treatment duration, dose modifications, or the schedule of response assessments beyond the primary endpoint's stated assessment period and response confirmation rules.

The brief summary says the first six participants enrolled constituted a safety run-in cohort because the combination had not been tested in this subject population. The record does not report the run-in's decision rules, how safety findings affected subsequent enrollment, or whether any protocol changes followed. Those details should not be inferred from the existence of a run-in cohort alone. Similarly, the registry record does not report the rationale for the enrollment target or a formal sample-size justification.

Three participants received one dose and were not evaluable for the primary response endpoint. The record attributes this to transition to hospice care, a severe allergic reaction to therapy, or no further follow-up, without mapping each reason to an individual participant. The remaining 43 participants were described as evaluable for efficacy using the primary endpoint. How the analysis handled people without an evaluable response, beyond this stated evaluability count, is 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.

Overall Study

Analysis population for the primary outcome, as defined in the registry: Forty-three of the enrolled patients were evaluable for efficacy using the primary endpoint of ORR. Three patients received a single dose of therapy and were not evaluable for response due to transition to hospice care, a severe allergic reaction to therapy, or no further follow up.

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 Objective Response Rates (ORR), measured over From start of treatment Day 1 (D1) and assessed for a maximum of 18 months.

The primary endpoint, objective response rate, is the percentage of participants classified as having a partial response or complete response according to immune-related RECIST criteria. It is a binary outcome at the participant level: each participant is classified according to whether the response definition was met, and the rate expresses the proportion meeting that definition in the assessed group. The registry names the endpoint as a percentage of participants and gives a 95% confidence interval as its dispersion measure.

The response rules require more than a single impression of tumor change. A complete response requires disappearance of all lesions across two consecutive observations at least 4 weeks apart. A partial response requires a reduction in tumor burden of at least 50% relative to baseline, confirmed by a subsequent assessment at least 4 weeks after first documentation. Stable disease does not meet the complete- or partial-response definitions and is specified in the absence of immune-related progressive disease. These criteria are part of the endpoint definition, not the result.

General statistical context: a response proportion provides a compact description of how often the prespecified response category occurred in the assessed population. It does not describe the timing or duration of response, and it does not by itself compare outcomes with an alternative treatment. A confidence interval communicates uncertainty around an estimated proportion; its width depends on the amount of information and the observed response pattern. The registry record identifies a 95% confidence interval for this endpoint but does not report the observed response rate or interval limits.

Response was assessed from treatment day 1 for a maximum of 18 months. The record does not report the response count, the number classified as complete versus partial response, or the timing and handling of assessments in participants without a final evaluable response.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Objective response rate is a binary-outcome summary. Each assessed participant either meets the response definition or does not; the rate is the share classified as having complete or partial response. The registry record specifies a percentage of participants and a 95% confidence interval, but it does not report the response count, estimated percentage, or interval limits. Without those values, the endpoint can be explained methodologically but not interpreted as an observed numerical result.

What does the response definition require?

The study used immune-related RECIST criteria. A complete response requires disappearance of all lesions at two consecutive observations separated by at least 4 weeks. A partial response requires a decrease in tumor burden of at least 50% relative to baseline, followed by confirmation at a consecutive assessment at least 4 weeks after first documentation. Stable disease is defined as not meeting the complete- or partial-response criteria, in the absence of immune-related progressive disease. The confirmation rules are important because they make a response classification depend on a later assessment, rather than solely on the first observation of change.

General explanation: prespecified response rules help make outcome classification more consistent across participants and assessments. The rules specify what counts as a response; they do not ensure that assessments were complete or that outcome classification was free of measurement variation. The registry record does not report additional information about independent review, assessor training, or adjudication of uncertain classifications.

How should a confidence interval be understood?

General explanation: a confidence interval expresses the statistical uncertainty around an estimated quantity. For a response proportion, it gives a range associated with the estimate under a stated method and its assumptions. A narrower interval generally indicates greater precision, while a wider interval indicates less precision. The interval is not a second estimate of the response rate, and its interpretation depends on the method used. Although the registry names a 95% confidence interval, the interval method and its numerical limits are not reported in the registry record.

The cohort size and number of evaluable participants are relevant to precision, but they are not enough to recover the interval. The observed number of responses is also needed, as is the calculation method. No response count or response proportion is reported, so it would be inappropriate to reconstruct or infer either value. The fact that 43 participants were evaluable does not reveal how outcomes were distributed among them.

What does the single-group design permit?

This study has one treatment group, not randomized treatment groups. Its primary endpoint can describe the proportion of evaluable participants who met the response definition, if that result is reported. It cannot directly estimate a randomized difference between the combination and another intervention. A comparison with an external benchmark would require a benchmark and a justified method, neither of which is reported in the registry record. Because the treatment combines three drugs, a single-group outcome also cannot separate the effect of each drug.

General explanation: random assignment is one way to balance measured and unmeasured participant characteristics between treatment groups on average. Without a randomized comparator, observed outcomes may reflect the treatment, participant selection, disease characteristics, other care, or a combination of influences. This general design consideration does not imply that any particular bias occurred in this trial; it describes what the design can and cannot establish.

How do the planning illustrations differ from observed evidence?

The stated planning assumptions describe an illustration based on an observed effect, using a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The effect to detect is expressed as a stated fraction of the observed effect. Those assumptions are not the trial's own design: the trial is single-group and has no equal allocation between two treatment arms. The planning illustrations are not observed power and should not be read as evidence about the trial's achieved power or outcome.

General explanation: sample-size planning asks how many participants an imagined design would require under specified assumptions. A planning calculation depends on the effect size, variability or event rate, error rate, allocation, and chosen method. Altering an assumption changes the planning result. A calculation based on an observed effect is particularly dependent on that estimate and does not turn the estimate into a prospective design target. The numerical planning estimates are not reported in the registry record, so no participant requirement can be stated from the registry information.

The planning assumptions describe two arms even though the trial has one group. They therefore need to be kept conceptually separate from the enrolled cohort and the primary endpoint result. A planning illustration is not an outcome, a control comparison, or a retrospective measure of how likely the study was to detect a result. The record does not report the protocol's sample-size rationale, target response rate, statistical decision threshold, or a prospective power calculation.

Why can over-dispersion matter?

General explanation: over-dispersion occurs when observed variation is greater than a statistical model assumes. If a model underestimates variation, its uncertainty estimates may be too small and inferential conclusions may appear more precise than warranted. This concept is relevant to some count or clustered-data models, but the registry record does not report an over-dispersion assessment or a model for the primary endpoint. The primary endpoint is a participant-level response proportion, and no numerical model diagnostics or analysis method are reported. Over-dispersion should therefore be treated as a general statistical concept, not as a documented feature of this trial.

For this record, the most important methodological distinction is between outcome definition and outcome estimate. The definition is reported in detail, as are the assessment period and the number evaluable. The estimate, confidence interval, inferential method, and analysis population beyond the evaluability statement are not reported. Statistical interpretation must remain within those limits.

10. Exact Estimates for Each Arm

No two-arm comparison is shown for this trial, but each arm's result can still be summarized with exact interval estimates computed from the posted counts. The Clopper-Pearson interval is exact and conservative; the Wilson interval is a common closer-to-nominal alternative; the Bayesian credible interval uses a uniform Beta(1, 1) prior. For small groups the three can differ noticeably.

R code and output

arm    <- c("Investigational Treatment")
n      <- c(43)
events <- c(32)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  Investigational Treatment  43  32  74.4  58.8  86.5  59.8  85.1  59.7  85.0

11. Multiplicity, Missing Data and Analysis Populations

The registry lists objective response rate as the primary outcome and a toxicity proportion as a secondary outcome. It reports two outcome measures in total. The record does not report whether the outcomes were tested inferentially, whether any adjustment for multiple comparisons was planned, or whether a hierarchy or decision rule governed interpretation. A multiplicity strategy is therefore not reported in the registry record. Without that information, no claim can be made about whether a formal adjustment was used or was needed for the analyses.

The registry record contains no analysis entries. It does not provide a numerical primary response result, a confidence interval, a p-value, or an account of statistical testing. It also does not report an analysis plan describing how outcomes were to be summarized or compared. The lack of these details means the registry information supports description of the endpoint and study population, but not reconstruction of the inferential procedure.

For missing or unevaluable response information, the record states that 43 of 46 enrolled participants were evaluable for efficacy. Three participants received a single dose and were not evaluable because of transition to hospice care, a severe allergic reaction, or no further follow-up. It does not report how missing assessments were handled, whether any missing outcome was imputed, or whether a sensitivity analysis was conducted. It also does not provide a separate result for all enrolled participants. The stated evaluability count should not be treated as proof that missing outcomes had no effect on interpretation.

General statistical context: missing outcome data can affect an estimate when the probability of having an observed outcome is related to the outcome itself or to participant characteristics. Different assumptions about unobserved outcomes can lead to different results. A transparent analysis generally describes who was missing, why data were missing, how the primary analysis handled them, and whether alternative assumptions were explored. The registry record reports some reasons for non-evaluability but does not describe those analytical choices.

The study was not randomized, so there is no intention-to-treat comparison between assigned groups to describe. The record does not report a more specific analysis-population definition beyond the 43 participants evaluable for efficacy. Whether other populations were used for safety or other endpoints, and how those populations were defined, is not reported in the registry record.

12. 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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: fever (6.5%), dehydration (4.3%), hypokalemia (4.3%), syncope (4.3%), acute kidney injury (2.2%).

The registry lists 46 participants at risk for serious adverse events and 16 affected. It also lists 46 participants at risk for other adverse events and 46 affected. These are separate registry summaries; the figures do not establish that every event was caused by treatment. The record also lists 11 deaths. It does not report in this summary whether those deaths were judged related to study treatment, their timing, or their causes.

The listed serious adverse event terms and percentages are fever, 6.5%; dehydration, 4.3%; hypokalemia, 4.3%; syncope, 4.3%; acute kidney injury, 2.2%; and anaphylaxis, 2.2%. The registry's summary does not specify whether each percentage represents participants, events, or another denominator. These terms should therefore be read as the reported serious-event listings, not converted into participant counts or assumed to be mutually exclusive categories.

The record identifies the secondary safety outcome as the proportion of participants experiencing at least Grade 3 toxicity according to CTCAE v4 criteria. It does not report a numerical result for that outcome in the information available. Serious adverse events and Grade 3 toxicity are distinct classifications; the listing of serious events cannot substitute for the missing Grade 3 toxicity result.

General explanation: safety summaries depend on definitions, observation periods, attribution rules, and whether repeated events in one participant are counted as one affected participant or multiple events. The registry summary does not provide all of those details. The reported counts and terms describe safety events recorded for this study, but they do not establish causation or provide a controlled comparison with another treatment.

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

The registry record describes a phase II study of pembrolizumab with carboplatin and paclitaxel in measurable advanced or recurrent endometrial cancer. It identifies objective response rate as the primary endpoint, defines response using immune-related RECIST criteria, and states that 43 of the 46 enrolled participants were evaluable for efficacy. It also identifies three reasons for non-evaluability among participants who received a single dose.

What the record does not report is the primary endpoint result itself. There is no response count, objective response percentage, or confidence interval in the registry information. Accordingly, the posted record cannot establish how often a response occurred in the evaluable cohort or how precise that estimate was. The defined endpoint and the population size are not substitutes for the missing outcome estimate.

The single-group, open-label design also matters. Even if a numerical response result were available, it would describe outcomes among participants who received the combination, rather than estimate a randomized difference against a control treatment. It would not isolate the effect of any one component of the combination. No comparative efficacy conclusion can be drawn from a study without a comparator arm.

General statistical context: an observed response proportion is a description of a study cohort, while a causal or comparative claim requires a design and analysis that support that question. Statistical uncertainty and clinical importance are also different considerations. A confidence interval can characterize uncertainty around an estimate, but neither an estimate nor its interval alone determines clinical value. The necessary response estimate and interval are not reported in the registry record, and no treatment recommendation follows from the available information.

The safety summary gives the number of participants at risk and affected for serious and other adverse events, lists selected serious event terms and percentages, and records deaths. These summaries are important descriptions of reported events, but the record does not establish that the events or deaths were caused by treatment. The listed secondary toxicity outcome lacks a numerical result in the registry information.

Thus the record supports a limited account of trial structure, endpoint definitions, evaluability, and recorded safety summaries. It does not support a numerical account of objective response, a comparative estimate, a full evaluation of missing-response handling, or a definitive attribution of adverse events. Those are gaps in what is reported, not evidence that a particular outcome was absent.

14. Why This Trial Matters Statistically

This trial is statistically notable for the contrast between a clearly defined primary endpoint and the absence of its numerical result in the registry record. The response rules specify the criteria for complete and partial response, including confirmation over time, and the assessment period extends from treatment day 1 for a maximum of 18 months. Yet without the response count, estimated rate, and confidence interval, readers cannot evaluate the endpoint's observed frequency or precision.

The record also makes the analysis population visible: 43 participants were evaluable for efficacy among 46 enrolled. It identifies three reasons for non-evaluability after a single dose. Those details help frame the denominator and raise the relevant question of how unobserved outcomes were handled. The registry record does not report that handling, so the evaluability statement cannot resolve the potential effect of missing response information.

The single-group design is central to statistical interpretation. It permits a cohort-level description if the outcome estimate is reported, but it does not provide a randomized comparator. Consequently, the design does not produce a direct estimate of the difference between the combination and another treatment, and it cannot distinguish the contribution of the three drugs from one another. This is a boundary on the type of inference, not a judgment about the quality of any unreported result.

The listed safety information offers another example of why denominators and definitions matter. The registry gives counts of participants at risk and affected, selected serious adverse event terms and percentages, and a number of deaths. It does not supply all attribution or counting details needed to interpret these as treatment-caused risks or compare them with another group. Safety description and comparative safety inference are not interchangeable.

General statistical context: transparent reporting of estimates, uncertainty, analysis populations, and missing-data procedures allows readers to distinguish what was observed from what can be inferred. In this record, the endpoint definition and study design are described, but key efficacy and analysis details are not. The main statistical lesson is to avoid turning the presence of a well-defined endpoint, a completed status, or a safety listing into a numerical efficacy conclusion that the registry does not report.

15. 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 primary limitation for interpreting efficacy is that the registry record does not report the objective response rate, response count, or 95% confidence interval. Although the endpoint definition is detailed and the evaluable population is stated, the central numerical result is unavailable. This prevents assessment of the response frequency and the precision of its estimate from the registry record.

The study has a single treatment group, no random allocation, and no separate control arm. It therefore cannot provide a randomized comparison with another regimen. Its open-label design is also stated, but the record does not report whether outcome assessment involved independent review or other procedures intended to limit assessment bias. These features constrain interpretation without establishing that bias occurred.

Three of 46 enrolled participants were not evaluable for the primary endpoint after receiving a single dose. The stated reasons were transition to hospice care, a severe allergic reaction, or no further follow-up. The registry record does not map each reason to a participant, describe the missing-data method, or report sensitivity analyses. It also does not report a response analysis for all enrolled participants.

The brief summary describes a six-participant safety run-in, but the registry record does not report its decision criteria or consequences. The available regimen description does not give the full reduced-dose details for participants with prior external beam radiation therapy and/or platinum-based chemotherapy. The rationale for the enrollment target and the trial's own sample-size justification are also not reported in the registry record.

For safety, the record lists affected participants, event terms, percentages, and deaths, but does not provide complete information on attribution, timing, or the denominator used for each listed percentage. It does not give a numerical result for the secondary Grade 3 toxicity outcome in the information available. The record does not report a multiplicity plan or detailed inferential analysis for the outcomes. Together, these omissions limit both quantitative efficacy interpretation and the scope of safety comparisons that can be made.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Pembro/Carbo/Taxol in Endometrial Cancer — NCT02549209.
• Registry record last updated 2022-03-02; status last verified 2022-02.
• Results first posted 2021-11-30.
• 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.

19. Record Summary

The registry record describes a completed phase II, open-label, single-group, multicenter study of pembrolizumab with carboplatin and paclitaxel for measurable advanced or recurrent endometrial cancer. Enrollment was 46. The study start date was 2017-08-22, and the primary completion date was 2019-12-12. The first six participants enrolled were designated as a safety run-in cohort.

The primary endpoint was objective response rate under immune-related RECIST criteria, assessed from treatment day 1 for a maximum of 18 months. Forty-three participants were evaluable for efficacy; three participants who received a single dose were not evaluable for response. The registry record does not report the observed response rate or its confidence interval.

The safety summary lists 16 participants affected by serious adverse events among 46 at risk, 46 affected by other adverse events among 46 at risk, and 11 deaths. Selected serious event terms are listed with percentages, but the record does not provide enough attribution detail to establish causation. A numerical result for the secondary Grade 3 toxicity outcome is not reported in the registry information.