This page separates information reported in the trial registry from statistical interpretation. The registry reports that results have been posted, but the ClinicalTrials.gov record does not include formal statistical-analysis records or numerical efficacy estimates.
1. Trial at a Glance
NRG-GY018 is a randomized, parallel, quadruple-masked phase 3 treatment trial evaluating pembrolizumab in combination with paclitaxel and carboplatin in patients with stage III-IV or recurrent endometrial cancer. The registered primary endpoint is progression-free survival, defined as time from study entry to progression according to RECIST criteria or death due to any cause, whichever occurs.
| Feature | NRG-GY018 |
|---|---|
| Trial | NRG-GY018 |
| ClinicalTrials.gov identifier | NCT03914612 |
| Phase | Phase 3 |
| Status | ACTIVE_NOT_RECRUITING |
| Allocation | RANDOMIZED |
| Design model | PARALLEL |
| Masking | QUADRUPLE |
| Primary purpose | TREATMENT |
| Enrollment | 813 |
| Arms | 2 |
| Lead sponsor | National Cancer Institute (NCI) |
| Sponsor type | NIH |
| Start | 2019-08-22 |
| Primary completion | 2022-12-16 |
| Results posted | Yes |
2. Clinical Question
The registered clinical question concerns whether adding the immunotherapy drug pembrolizumab to the usual chemotherapy treatment of paclitaxel and carboplatin changes progression-free survival in patients with stage III-IV or recurrent endometrial cancer.
Population
The trial's condition list includes endometrial clear cell adenocarcinoma, dedifferentiated carcinoma, endometrioid adenocarcinoma, mixed cell adenocarcinoma, serous adenocarcinoma, undifferentiated carcinoma, recurrent endometrial adenocarcinoma, and recurrent endometrial carcinoma.
Intervention
Pembrolizumab in the context of the trial's chemotherapy regimen of paclitaxel and carboplatin.
Comparator
The trial includes a comparator arm involving placebo administration together with the usual chemotherapy framework.
Primary question
How does the randomized treatment comparison relate to progression-free survival, as defined by the registered time-to-event endpoint?
3. Trial Design
Comparator treatment
- Carboplatin
- Paclitaxel
- Placebo administration
- Computed tomography as a trial procedure
- Quality-of-life assessment
- Questionnaire administration
Pembrolizumab treatment
- Carboplatin
- Paclitaxel
- Pembrolizumab
- Computed tomography as a trial procedure
- Quality-of-life assessment
- Questionnaire administration
The ClinicalTrials.gov record identifies the interventions included in the study but do not provide dosing schedules, treatment-cycle counts, treatment duration, crossover rules, or other operational treatment details.
4. Trial Timeline
Study start
The registered study start date is August 22, 2019.
Primary completion
The registered primary completion date is December 16, 2022.
Active, not recruiting
The ClinicalTrials.gov record lists the study status as ACTIVE_NOT_RECRUITING.
5. Primary Endpoint
| Endpoint | Registered definition | Endpoint type |
|---|---|---|
| Progression-free Survival (PFS) | Duration of time from study entry to time of progression per RECIST criteria or death due to any cause, whichever occurs first. The interquartile range for follow-up was (6.37 months, 30.62 months). The maximum for follow-up was 37.85 months. | Time-to-event |
6. Why PFS Is a Time-to-Event Endpoint
Progression-free survival differs from an ordinary binary outcome because participants can be followed for different lengths of time and may experience the event at different points during follow-up. The registered definition combines two possible events: progression according to RECIST criteria or death from any cause.
This structure creates two important statistical features. First, each participant contributes information over a period of time rather than simply receiving a final yes/no classification. Second, participants who have not experienced progression or death by the time their evaluable follow-up ends may be censored according to the applicable analysis rules.
Here, T represents the time from study entry to the defined event. For a PFS analysis, the event is progression per RECIST criteria or death due to any cause, as specified in the registry definition.
A time-to-event analysis therefore uses both when an event occurs and the information available from participants who have not yet experienced an event at the end of their observed follow-up.
7. Results Status and Planned Analysis
This distinction is important. A registry can contain posted outcome information without the ClinicalTrials.gov record containing the formal statistical comparison needed to reproduce an inferential result.
What would normally be analyzed?
For a randomized trial whose primary endpoint is PFS, a conventional statistical analysis would generally describe the time-to-event distribution in each randomized group, compare the groups using an appropriate survival-analysis test, and estimate a relative treatment effect with a confidence interval using a survival model such as a Cox proportional-hazards model when its assumptions and prespecified analysis plan support that approach.
Kaplan-Meier estimates are commonly used to describe the event-free probability over time. A log-rank-type comparison can test whether the survival distributions differ. A Cox model can provide a hazard ratio that summarizes the relative event rate under the model. The exact analysis, stratification factors, censoring rules, hypothesis-testing framework, and multiplicity procedures should come from the prespecified statistical analysis plan.
8. Understanding the Registered PFS Definition
Progression
Progression is defined in the registry using RECIST criteria. The endpoint therefore incorporates a radiologic disease-progression component rather than relying solely on symptoms or treatment discontinuation.
Death
Death due to any cause is also an event. A participant does not need to have documented RECIST progression before death for death to contribute as the PFS event.
Time origin
The registry defines the duration from study entry to the event. The exact operational timing conventions beyond that wording are not reported in the ClinicalTrials.gov record.
Composite endpoint
PFS combines progression and death into one time-to-event endpoint. This allows the analysis to capture both forms of failure of progression-free status.
9. Statistical Methodology
Kaplan-Meier estimation
A Kaplan-Meier analysis is a standard way to estimate the probability of remaining event-free over time when follow-up is incomplete for some participants. Rather than requiring every participant to experience an event, the method incorporates censoring information up to the last relevant observation.
The expression combines the conditional probabilities of remaining event-free at each observed event time. di is the number of events at time ti, while ni is the number at risk immediately before that time.
Log-rank comparison
A log-rank test is commonly used to compare time-to-event distributions between randomized groups. It uses the timing of events throughout follow-up rather than reducing each participant to a simple event/no-event classification.
Cox proportional-hazards model
A Cox model is commonly used to estimate a hazard ratio for a time-to-event endpoint. The hazard ratio compares estimated instantaneous event rates between groups under the fitted model.
A hazard ratio is a relative time-to-event measure. It is not the same as an absolute risk difference, a probability of cure, or a statement that every participant experiences the same proportional change in risk.
Confidence intervals
A confidence interval quantifies statistical uncertainty around an estimated effect under the specified statistical model and sampling framework. A narrow interval indicates greater statistical precision than a wide interval, all else equal.
Analysis of insufficient events
The registry specifically states that in the arm 2 dMMR cohort, the point estimate and upper confidence-limit value cannot be estimated because of insufficient events. This illustrates an important feature of time-to-event analysis: an estimate requires enough observed event information to identify the relevant portion of the statistical model.
10. Statistical Methods Explained
Why is PFS analyzed as a time-to-event endpoint?
Because the endpoint is defined by the duration from study entry until progression or death, both event occurrence and event timing matter. A participant who remains progression-free for a longer period contributes more event-free follow-up than someone who experiences progression earlier.
Why can some participants be censored?
Not every participant will necessarily experience progression or death during the period in which their outcome can be observed. Survival methods allow those participants to contribute the information available up to their censoring time rather than treating them as if they had experienced the event at an arbitrary date.
What would a hazard ratio mean for PFS?
A PFS hazard ratio below 1 would indicate a lower estimated instantaneous rate of progression or death in the numerator treatment group relative to the comparator group, under the fitted model. It would not directly tell us the percentage of participants who are progression-free at a particular calendar time.
Why is the confidence interval important?
The confidence interval shows how precisely the treatment effect has been estimated. Two studies can have similar point estimates but very different levels of precision if their event counts and information differ.
Why does insufficient event information matter?
Time-to-event models depend on observed events to distinguish treatment groups statistically. When too few events occur in a relevant cohort, the model may not be able to estimate a stable effect or one side of its confidence interval. The registry explicitly identifies this issue for the arm 2 dMMR cohort.
Why should the formal statistical analysis not be reconstructed from the endpoint alone?
The endpoint tells us what was measured, but it does not by itself establish the complete inferential procedure. A full analysis can depend on the analysis population, stratification, censoring rules, hypothesis-testing framework, multiplicity adjustment, interim monitoring, and model specifications. Those details should be taken from the trial's statistical analysis plan rather than assumed.
11. Analysis Populations and Missing-Data Issues
The ClinicalTrials.gov record identifies the randomized design and primary endpoint but do not provide the prespecified efficacy analysis population, safety analysis population, censoring conventions, missing-data rules, or imputation strategy.
For a randomized efficacy endpoint, the analysis population matters because excluding participants after randomization can potentially compromise the balance created by random assignment. For a time-to-event endpoint, missingness and censoring are also distinct concepts: censoring specifies how incomplete event observation contributes to the analysis, whereas missing data may refer to unavailable measurements or assessments.
| Analysis feature | What the ClinicalTrials.gov record establishes | Statistical implication |
|---|---|---|
| Randomization | Yes | Creates the basis for a randomized treatment comparison. |
| Primary endpoint | PFS | Requires time-to-event methodology. |
| Formal statistical analysis | Not in the posted results | |
| Analysis population | Not reported | Cannot identify an ITT, per-protocol, or other formal efficacy population from the ClinicalTrials.gov record. |
| Missing-data / imputation rules | Not reported | Cannot characterize the prespecified handling strategy. |
| Stratification factors | Not reported | Cannot state whether the formal model was stratified. |
12. Safety Results
The ClinicalTrials.gov record provides serious adverse-event counts by cohort and randomized arm. These figures are reported as affected participants divided by participants at risk.
| Cohort | Arm | Serious adverse events |
|---|---|---|
| dMMR cohort | Arm 1: Active Comparator | 50/106 |
| dMMR cohort | Arm 2: Experimental (Pembrolizumab) | 66/109 |
| pMMR cohort | Arm 1: Active Comparator | 125/274 |
| pMMR cohort | Arm 2: Experimental (Pembrolizumab) | 152/276 |
These are counts of affected participants over the stated at-risk denominators. They should not be converted into a comparative risk estimate here because the ClinicalTrials.gov record does not provide a formal statistical analysis of serious adverse events.
dMMR cohort
The registry reports 50/106 serious adverse events in the active-comparator arm and 66/109 in the pembrolizumab experimental arm.
pMMR cohort
The registry reports 125/274 serious adverse events in the active-comparator arm and 152/276 in the pembrolizumab experimental arm.
13. Results Interpretation: What Can and Cannot Be Concluded From the Supplied Data
The trial is a randomized phase 3 parallel-group study with quadruple masking, 813 enrolled participants, two arms, and a registered primary PFS endpoint defined as time from study entry to progression per RECIST criteria or death due to any cause, whichever occurs.
The ClinicalTrials.gov record contains no numerical primary efficacy estimates. Therefore, this page does not report a PFS hazard ratio, confidence interval, p-value, event count, median PFS, or treatment-specific PFS estimate.
A statistical analysis should distinguish a trial's design and registered endpoint from its inferential results. The fact that results are posted does not permit an absent statistical estimate to be reconstructed without importing additional evidence.
14. Cohort-Specific Estimation Issue
The registered PFS definition contains an explicit statistical qualification for the dMMR cohort: for arm 2, the point estimate and upper limit of the confidence interval are not able to be estimated because of insufficient events.
What insufficient events means statistically
The issue is insufficient event information for the specified quantity, not a numerical estimate of treatment effect.
This distinction is especially important in subgroup or biomarker-defined analyses. A cohort can contain participants while still producing too few endpoint events to estimate a model-based treatment effect with adequate statistical information.
The appropriate description is that the registry definition identifies the estimate as not estimable because of insufficient events.
15. Design Topics Not Specified in the Supplied Data
| Design topic | What can be stated from the ClinicalTrials.gov record |
|---|---|
| Non-inferiority margin | Not reported. The registered primary endpoint is PFS, but no non-inferiority margin is provided. |
| Crossover | Not reported. No crossover procedure is identified in the ClinicalTrials.gov record. |
| Factorial design | Not reported. The design model is explicitly identified as PARALLEL. |
| Multiplicity | Not reported. The registry identifies 1 registered primary endpoint but provides no multiplicity procedure. |
| Interim analysis | Not reported. |
| Missing-data / imputation | Not reported. |
| Stratification | Not reported. |
| Bayesian methods | Not reported. |
These omissions are not evidence that the corresponding methods were absent from the protocol. They mean only that the ClinicalTrials.gov record does not establish them, so they are not attributed to NRG-GY018 on this page.
16. Understanding the Difference Between Endpoint and Effect Estimate
The registry's PFS definition establishes what is being measured. A hazard ratio, median PFS, survival probability at a specified time, or p-value would describe what the comparison found. These are different layers of statistical information.
| Layer | NRG-GY018 the ClinicalTrials.gov record | Interpretation |
|---|---|---|
| Outcome definition | PFS | Progression per RECIST criteria or death due to any cause. |
| Time origin | Study entry | Follow-up duration begins at study entry under the registry definition. |
| Event type | Progression or death | Either event ends progression-free survival. |
| Endpoint type | Time-to-event | Timing of the event is part of the outcome. |
| Formal effect estimate | Not reported | No numerical treatment effect is reported here. |
| Formal hypothesis test | Not reported | No p-value is reported here. |
17. Limitations
- Incomplete statistical-analysis data: no formal statistical analyses were posted even though the registry indicates that results have been posted.
- No numerical efficacy estimates: the ClinicalTrials.gov record does not contain a PFS hazard ratio, confidence interval, p-value, event count, median PFS, or treatment-specific PFS estimate.
- Endpoint-only inference is insufficient: the registered PFS definition does not establish the complete statistical analysis plan.
- Subgroup estimation: the registry explicitly notes insufficient events for estimation of the point estimate and upper confidence limit for arm 2 in the dMMR cohort.
- Safety interpretation: serious-adverse-event counts are available, but the ClinicalTrials.gov record does not include formal comparative safety analyses.
- Unspecified design methods: stratification, interim monitoring, multiplicity, missing-data handling, crossover, non-inferiority margins, and Bayesian methods are not characterized in the ClinicalTrials.gov record.
- Generalizability: the ClinicalTrials.gov record identifies a specific set of endometrial cancer conditions and a phase 3 treatment trial; it does not provide the detailed eligibility and baseline characteristics needed to assess applicability to other populations.
18. Why This Trial Matters Statistically
NRG-GY018 is a useful statistical teaching example because the registry combines a randomized phase 3 design with a time-to-event primary endpoint and explicitly documents a situation in which an effect estimate cannot be calculated because of insufficient events.
| Concept | How it appears in NRG-GY018 |
|---|---|
| Randomization | The registered allocation is RANDOMIZED. |
| Parallel design | The registered design model is PARALLEL. |
| Blinding | The registered masking level is QUADRUPLE. |
| Time-to-event analysis | The primary endpoint is PFS. |
| Composite event definition | PFS includes progression per RECIST criteria or death due to any cause. |
| RECIST-based progression | Progression is explicitly incorporated into the registered PFS definition. |
| Censoring | A time-to-event framework generally requires rules for incomplete event observation, although the ClinicalTrials.gov record does not specify the protocol's censoring rules. |
| Confidence intervals | The registry notes that an upper confidence limit cannot be estimated for a specified dMMR cohort arm because of insufficient events. |
| Insufficient events | The trial provides a concrete example of why event information determines whether a time-to-event effect can be estimated. |
| Safety by cohort | Serious adverse-event counts are reported separately for dMMR and pMMR cohorts and treatment arms. |
19. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary outcome is a time-to-event endpoint. The appropriate analysis must account for the timing of progression or death and for incomplete follow-up. The ClinicalTrials.gov record does not contain the formal inferential estimates needed to quantify the randomized treatment effect.
Clinical interpretation
The trial is designed to evaluate a treatment strategy involving pembrolizumab plus paclitaxel and carboplatin versus the comparator treatment framework in the specified endometrial cancer population. The ClinicalTrials.gov record alone do not establish the magnitude of any efficacy difference.
20. What a Complete PFS Analysis Would Usually Show
For educational purposes, a complete statistical report for this type of endpoint would normally allow the reader to examine several complementary quantities rather than relying on one number.
| Result | Why it matters |
|---|---|
| Number of PFS events | Determines how much event information is available for the time-to-event comparison. |
| Kaplan-Meier curve | Shows how the estimated probability of remaining progression-free changes over time. |
| Median PFS | Provides a time-point summary when the estimated survival curve reaches 0.50. |
| Hazard ratio | Provides a model-based relative comparison of instantaneous event rates. |
| 95% confidence interval | Shows uncertainty around the estimated effect. |
| P-value | Addresses the specified hypothesis-testing framework; it does not measure effect magnitude. |
| Absolute PFS probabilities | Provide time-specific information that can complement a relative hazard ratio. |
The absence of these numerical results from the ClinicalTrials.gov record is why this page stops at the planned-analysis level rather than supplying estimates from an external publication.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Calculators
23. Sources
- ClinicalTrials.gov: NRG-GY018, NCT03914612.
- Linked publication record: PubMed PMID 40044930.
- Linked publication record: PubMed PMID 38431043.
- Linked publication record: PubMed PMID 37411031.
- Linked publication record: PubMed PMID 36972022.
Continue through the Clinical Biostats statistical tutorials
Explore the statistical concepts behind randomized trials, time-to-event endpoints, and clinical-trial analysis.
24. Record Summary
NRG-GY018 is a randomized phase 3, parallel, quadruple-masked treatment trial with 813 enrolled participants and two arms. Its registered primary endpoint is progression-free survival, defined as duration from study entry to progression per RECIST criteria or death due to any cause, whichever occurs. The ClinicalTrials.gov record indicates that results have been posted, but they do not contain formal statistical-analysis records or numerical efficacy estimates. The most important statistical feature that can be examined from the ClinicalTrials.gov record is therefore the structure of the PFS endpoint and the implications of time-to-event analysis, including the explicit note that insufficient events prevent estimation of the point estimate and upper confidence limit for arm 2 in the dMMR cohort.