Study of Pembrolizumab (MK-3475) Plus Chemotherapy Versus Placebo Plus Chemotherapy in Participants With Gastric or Gastroesophageal Junction (GEJ) Adenocarcinoma (MK-3475-585/KEYNOTE-585)-China Extension: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Study of Pembrolizumab (MK-3475) Plus Chemotherapy Versus Placebo Plus Chemotherapy in Participants With Gastric or Gastroesophageal Junction (GEJ) Adenocarcinoma (MK-3475-585/KEYNOTE-585)-China Extension (NCT04882241), a randomized phase 3 study. The primary outcome is Event-free Survival (EFS) Per Response Criteria in Solid Tumors Version 1.1 (RECIST 1.1) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms.

2. Clinical Question

3. Trial Design

This was a Phase III, randomized, double-blind, quadruple-masked, parallel-group clinical trial and a China extension of KEYNOTE-585. The registry lists an enrollment of 120 and a completed study status. The recorded primary completion date is 2024-02-16. The study compared pembrolizumab plus chemotherapy with placebo plus chemotherapy, in a neoadjuvant or adjuvant treatment context for gastric and gastroesophageal junction adenocarcinoma.

The primary endpoint analysis was limited to the XP/FP arms: pembrolizumab plus XP/FP and placebo plus XP/FP. XP is cisplatin plus capecitabine; FP is cisplatin plus 5-fluorouracil. The record also describes FLOT cohorts and lists several secondary outcomes involving those cohorts, but they are not part of the stated primary comparison. Further details of how participants were assigned between XP and FP regimens are not reported in the registry record.

The primary analysis population is described as all randomized participants in the two XP/FP arms, as pre-specified per protocol. The arm counts for the primary endpoint are 59 for pembrolizumab plus XP/FP and 61 for placebo plus XP/FP. The registry record gives a primary endpoint timeframe of up to approximately 42 months. Details of the randomization procedure, allocation concealment, masking roles, and protocol deviations are not reported in the registry record.

The brief summary says the study was intended to evaluate efficacy and explicitly states that no formal hypothesis testing would be done. The registry record does not report the trial's original sample-size justification or the assumptions used to plan the trial. Planning calculations presented with the posted results are separate illustrations based on the observed effect, not evidence of the assumptions used to design this trial.

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: 120 Chinese participants were enrolled. 5 participants were randomized in the global portion of the study (NCT03221426) and 115 in the China extension portion. The FLOT cohort was not enrolled in China per specifications in the protocol Supplemental Statistical Analysis Plan (sSAP) amendment and no data were collected for the FLOT cohort in China.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The analysis population consisted of all randomized participants in the Pembrolizumab+XP/FP and Placebo+XP/FP treatment arms, as pre-specified per protocol.

5. Baseline Characteristics

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

Population: The FLOT cohort was not enrolled in China per specifications in the protocol Supplemental Statistical Analysis Plan (sSAP) amendment and no data were collected for the FLOT cohort in China.

6. Primary Endpoint

The registry lists the primary outcome as Event-free Survival (EFS) Per Response Criteria in Solid Tumors Version 1.1 (RECIST 1.1) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms, measured over Up to approximately 42 months.

Event-free survival (EFS) was the primary endpoint. It measures time from randomization until the first qualifying event. Under the registry definition, events include radiographic progression under RECIST 1.1; local or distant recurrence after surgery; clinical progression evidenced by peritoneal carcinomatosis in the specified circumstances; or death from any cause. A second primary malignancy is not counted as an EFS event. Nor is radiographic progression during neoadjuvant treatment counted when it does not prevent successful surgery and the participant is disease free after surgery.

Time-to-event outcomes are useful when the timing of disease events matters and participants may have different lengths of follow-up. The analysis can use information about when events occur rather than reducing follow-up to a simple event/no-event count at one fixed time. A hazard ratio is used to compare event rates over follow-up under the analysis model; it is not a direct comparison of median survival times or a statement about an individual's outcome.

The registry specified investigator assessment using RECIST 1.1 and reported the median EFS with a 95% confidence interval as the endpoint summary. The posted comparison also includes a hazard ratio from a log-rank analysis. These summaries address related but different aspects of follow-up: the median describes the time by which half the analyzed group has experienced an event, while the hazard ratio summarizes the relative event experience across follow-up. A median was not reported for the pembrolizumab plus XP/FP arm.

7. Results Posted in the Registry

• Pembrolizumab + XP/FP vs Placebo + XP/FP: hazard ratio 0.92 (95% CI 0.50–1.70), reported p = 0.395 (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.

Disease-free Survival (DFS) Per Response Criteria in Solid Tumors Version 1.1 (RECIST 1.1) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Log Rank); it was not recalculated.
• Result:
  • Pembrolizumab + XP/FP vs Placebo + XP/FP: hazard ratio 0.83 (95% CI 0.34–2.05), reported p = 0.346 (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 Pembrolizumab + XP/FP vs Placebo + XP/FP and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists secondary outcomes spanning disease-free survival, overall survival, event-free survival, and adverse events, including treatment discontinuation due to adverse events. These outcomes cover different aspects of disease course and safety. The reported secondary analyses are not a substitute for the primary EFS comparison; they are complementary measures, some with different populations or definitions.

The numerical secondary comparison available for disease-free survival is also a time-to-event estimate with substantial uncertainty: its reported hazard-ratio interval spans values below and above 1. Other secondary outcomes are listed, but complete numerical results for all of them are not reported in the registry record. The secondary information therefore does not establish a consistent pattern of benefit or harm across outcomes. The posted results cannot support a broad conclusion about all secondary endpoints from the record alone.

Secondary results generally carry less weight than a prespecified primary result when they are numerous or exploratory. Multiple comparisons increase the chance that some results may appear notable by chance, especially when no adjustment is made. The registry summary says no formal hypothesis testing would be done, and a multiplicity adjustment is not reported in the registry record. The secondary findings should consequently be treated as descriptive and interpreted in light of their uncertainty, completeness, and relationship to the primary endpoint.

Differences in endpoint populations also matter. The primary EFS analysis is specified for the XP/FP arms, while some listed secondary outcomes combine those arms with the FLOT cohorts. Such results do not necessarily describe the same treatment comparison or population as the primary analysis. Their role is to broaden the registry's outcome record, not to overturn or confirm the primary result in isolation.

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

• Number of Participants Who Experience One or More Adverse Events (AEs) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms Separately and in Combination With the Pembrolizumab+FLOT and Placebo+FLOT Cohorts
• Number of Participants Who Discontinue Study Treatment Due to an Adverse Event (AE) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms Separately and in Combination With the Pembrolizumab+FLOT and Placebo+FLOT Cohorts
• Disease-free Survival (DFS) Per Response Criteria in Solid Tumors Version 1.1 (RECIST 1.1) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms
• Overall Survival (OS) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms in Combination With the Pembrolizumab+FLOT and Placebo+FLOT Cohorts
• Event-free Survival (EFS) Per Response Criteria in Solid Tumors Version 1.1 (RECIST 1.1) - Pembrolizumab+XP/FP and Placebo+XP/FP Treatment Arms in Combination With the Pembrolizumab+FLOT and Placebo+FLOT Cohorts

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is a time-to-event endpoint. Each participant contributes information from randomization until an EFS event or the end of observed follow-up. When an event has not occurred by the last available assessment, the observation is generally censored: the participant's follow-up contributes information up to that point, but the event time is not known. This is a general explanation of time-to-event analysis. The registry record does not report its detailed censoring rules or how missing assessments were handled.

What does the hazard ratio describe?

A hazard is the instantaneous event rate among participants who have not yet experienced the event, conditional on the model and the observed follow-up. A hazard ratio compares that rate between groups. A value of 1 corresponds to equal hazards under the usual interpretation; a value below 1 indicates a lower estimated hazard in the first named arm, and a value above 1 indicates a higher estimated hazard. This is a relative comparison over follow-up, not an absolute risk difference and not a ratio of medians.

The posted primary hazard ratio is 0.92, with a 95% confidence interval from 0.50 to 1.70. Because the interval includes 1 and ranges to either side, the estimate is uncertain about direction as well as size. The interval is more informative than the point estimate alone: it shows that the data do not tightly pin down a single relative hazard. The registry gives the hazard ratio as reported by the sponsor; the analysis method listed is Log Rank.

Why use a log-rank analysis?

The log-rank method is a standard approach for comparing time-to-event experience between groups. In general, it uses the timing of events and the number of participants still at risk over follow-up. It is not a comparison of the final number of events alone. The registry lists Log Rank as the method for the reported hazard ratio and p-value. Further implementation details, including assumptions about proportional hazards and any adjustment for covariates, are not reported in the registry record.

The hazard ratio is often interpreted most simply when the relative hazards remain reasonably stable over time. If the relative effect changes over follow-up, a single hazard ratio may compress different patterns into one summary. This is a general limitation of the measure. The registry record does not report an assessment of whether the proportional-hazards assumption was met or whether the effect varied over time, so those properties cannot be evaluated from the posted information.

How does a median differ?

The median event-free time is the time at which the estimated event-free proportion reaches one half, if that point is observed and estimable. It gives an absolute time summary that may be familiar to readers, but it does not describe the entire time-to-event curve. The placebo plus XP/FP median is reported as 36.5 months. The median for pembrolizumab plus XP/FP is not reported in the registry record. The listed confidence limits are incomplete for both arms because an upper limit is not reported in the registry record. Those omissions constrain direct interpretation of the arm-specific medians.

A confidence interval around a median reflects uncertainty in the estimated time at which the survival curve reaches the median threshold. Such an interval can be wide or one-sided when follow-up does not establish the crossing precisely. This is a general property of time-to-event summaries. The registry record does not explain why the median or upper confidence limit is unavailable for the pembrolizumab arm, so no reason should be inferred.

What do the p-value and confidence interval contribute?

A p-value describes how unusual the observed test statistic would be under a specified null model and test procedure. It does not describe the probability that the null model is true, and it does not quantify clinical importance. A confidence interval gives a range of values compatible with the data and analysis under its assumptions. In this trial record, the interval is broad and includes the equal-hazard value of 1. The p-value is reported, but the study summary states that no formal hypothesis testing would be done. The p-value therefore should not be treated as a stand-alone decision rule.

There is no discrepancy flag for the primary comparison. The values used in the interpretation are the sponsor-reported hazard ratio and its confidence interval, rather than a new estimate reconstructed from summary statistics. The registry does not report enough participant-level data to reproduce the time-to-event analysis independently. It also does not report a covariate-adjusted estimate, a model specification beyond the listed method, or an analysis of treatment effects within subgroups.

Several practical aspects of the analysis remain unavailable: the number and timing of EFS events, censoring and missing-data rules, follow-up completeness, and the approach to participants who did not undergo surgery or had protocol deviations. These are important to judge the robustness of a time-to-event analysis. Their absence does not invalidate the posted result, but it means readers cannot fully audit the analysis from the registry record alone.

11. Confidence Intervals and Precision

The 95% confidence interval for the primary hazard ratio runs from 0.50 to 1.70 around a point estimate of 0.92. It includes 1, the value corresponding to equal hazards under the usual interpretation. The interval also includes values below and above 1, so the data are compatible with different directions and a range of magnitudes. The reported point estimate should not be read as a precise estimate of a small difference.

The arm-specific median estimates are less complete. The placebo plus XP/FP median is 36.5 months, with a lower confidence limit of 15.5 months; the upper limit is not reported in the registry record. For pembrolizumab plus XP/FP, the median itself is not reported in the registry record, and the listed lower confidence limit is 15.2 months. The absence of a reported median and upper limits prevents a complete comparison of median EFS precision between arms.

A confidence interval expresses uncertainty in an estimate under the analysis assumptions; it is not a range containing a fixed proportion of individual participant outcomes. Wider intervals indicate less precision. In general, precision in a time-to-event comparison depends substantially on the number and timing of events and on follow-up, not simply on enrollment. The event count and follow-up details needed to judge those contributors are not reported in the registry record.

The interval does not prove that the arms are equivalent just because it includes 1. Equivalence requires a suitable design and prespecified margin, neither of which is reported in the registry record. Nor does the interval prove a difference, because it also includes the equal-hazard value. The proper reading is uncertainty: the reported estimate does not resolve the size or direction of the comparative EFS effect with precision.

12. Planning a Similar Trial: Sample Size

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

The planning figures are illustrations based on the observed effect, not the original trial design calculation. For the time-to-event outcomes, they give the total number of events needed, not participants per arm. This distinction matters because event totals depend on how many participants experience the event during follow-up. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect.

The rows describe planning for 100%, 75%, and 50% of the observed effect. The 80% power column gives the event total that the illustration estimates would be needed for an 80% chance of detecting the specified effect under its assumptions. The 90% power column gives the event total for a higher chance of detection under the same assumptions. For these time-to-event calculations, the event totals use the Schoenfeld approach. They are not participant counts per arm.

The general pattern is that detecting a smaller effect requires more information. In a time-to-event design, that means more events; in other outcome settings, the planning figures are expressed as participants per arm. The curve should be read as the relationship between the assumed effect size and the amount of information required, not as observed trial performance. The original trial's own sample-size justification and design assumptions are not reported in the registry record.

These calculations are planning illustrations, not observed power. Observed power is not a useful substitute for interpreting the confidence interval and the actual analysis results. The relevant reading is how the hypothetical information requirement changes under the stated planning assumptions, while recognizing that the original study was not necessarily designed using those assumptions.

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 = 0.920, so ln(HR) = -0.0834.
2. d = 4 × (1.96 + 0.842)² / [ln(0.920)]² = 4,515.7, rounded up to 4,516 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 <- 0.92                              # 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(4882241)
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 = 4,516 (80% power), 6,046 (90% power)
 75% of observed effect: events = 8,028 (80% power), 10,748 (90% power)
 50% of observed effect: events = 18,063 (80% power), 24,182 (90% power)
Simulated power with 4,516 events: 0.798 (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. Analyses Reported by the Sponsor

The primary comparison is reported as a sponsor analysis using the Log Rank method: hazard ratio 0.92, with a 95% confidence interval from 0.50 to 1.70. The registry record also reports a p-value for that analysis. No separate recalculated hazard ratio is listed for the primary endpoint, and the discrepancy flag is false. The interpretation therefore rests on the sponsor-reported analysis rather than a competing estimate reconstructed from the posted summary information.

A time-to-event estimate cannot generally be reproduced from arm medians and confidence limits alone. Reproducing it would require event times, censoring information, or additional sufficient statistics that are not reported in the registry record. The registry does not provide participant-level data or the detailed model specification needed for an independent reanalysis.

The absence of a flagged discrepancy means there is no reported conflict between a sponsor estimate and a recalculation for this comparison. It does not mean that every analysis detail is independently verifiable. The sponsor's own analysis is the relevant reported result, while its precision and limitations should be judged from the confidence interval, endpoint definition, and analysis-population information. The stated plan that no formal hypothesis testing would be done also remains important when interpreting the reported p-value.

15. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when several outcomes, comparisons, time points, or subgroups are examined. Even if each analysis is considered separately, a larger family of tests raises the chance of finding an apparently unusual result by chance. A prespecified approach to controlling multiplicity can help determine which findings are confirmatory and which are descriptive. The registry summary states that no formal hypothesis testing would be done. A detailed multiplicity strategy or adjustment is not reported in the registry record.

The record lists multiple secondary outcomes, including disease-free survival, overall survival, event-free survival, adverse events, and discontinuations due to adverse events. The number of outcome measures in the record is 10, and the record lists 4 analyses. Without a clearly reported testing hierarchy or multiplicity procedure, secondary p-values, where reported, should not be interpreted as independent confirmatory evidence. The available record does not explain whether secondary outcomes were intended to support formal claims.

The primary EFS population is described as all randomized participants in the two XP/FP arms, as pre-specified per protocol. This is useful information about the intended analysis population. It does not by itself answer how missing endpoint assessments, censoring, treatment discontinuation, or protocol deviations were handled. Those procedures are not reported in the registry record.

Missing-data methods are particularly relevant to time-to-event endpoints because participants may stop contributing follow-up before an event is observed. The record does not provide the number of censored participants, reasons for censoring, completeness of follow-up, or sensitivity analyses for missing outcome information. It also does not report whether the analysis included any adjustment for baseline prognostic factors. These omissions limit assessment of the robustness of the primary estimate, but they do not justify assuming that a particular procedure was or was not used.

The registry record identifies a planned population and an endpoint definition, but key details for reproducing or auditing the analyses are absent. Readers should distinguish that reporting limitation from evidence of bias or mishandling: the record does not establish either. Its stated no-formal-hypothesis-testing approach and the lack of reported multiplicity and missing-data methods call for cautious, descriptive interpretation of the posted results.

16. Safety Results

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

Deaths

Other (non-serious) adverse events

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

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: vomiting (13.1%), white blood cell count decreased (10.2%), pneumonia (8.5%), neutrophil count decreased (6.8%), cerebral infarction (5.1%).

The safety summary reports serious adverse events in 32 of 59 participants in the pembrolizumab plus XP/FP arm and 30 of 61 participants in the placebo plus XP/FP arm. It lists other adverse events for 59 of 59 and 61 of 61 participants, respectively. The record also lists 22 deaths in each arm. These counts describe the safety information posted in the registry; they do not by themselves establish that an event was caused by study treatment. The registry record does not provide additional detail about the relationship of the listed deaths to treatment or disease.

The most frequently listed serious adverse-event terms include vomiting (13.1%), decreased white blood cell count (10.2%), pneumonia (8.5%), decreased neutrophil count (6.8%), cerebral infarction (5.1%), and anaemia (4.9%). The record does not assign these term percentages to a specific arm in the information available for this comparison. They should therefore be read as listed terms, not as arm-specific contrasts.

Counts and percentages do not convey severity grades, duration, timing, recurrence, or clinical context. The registry lists treatment discontinuation due to adverse events as a secondary outcome, but the corresponding results are not reported in the registry record. A comparative safety judgment requires more detail than these summary counts provide.

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

The primary result is a sponsor-reported EFS hazard ratio of 0.92, with a 95% confidence interval from 0.50 to 1.70, for pembrolizumab plus XP/FP compared with placebo plus XP/FP. The interval includes 1 and is broad. The result therefore does not establish a clear difference in EFS, but it also does not rule out a range of potentially different relative event experiences. The point estimate alone is not enough to make a definitive claim.

This result does not establish equivalence. A conclusion of equivalence requires a design and prespecified margin suited to that question; such a margin is not reported in the registry record. It also does not establish that either treatment is effective or ineffective. The registry itself states that no formal hypothesis testing would be done, which further limits any confirmatory reading of the reported p-value.

The comparison applies to the randomized XP/FP arms and to the registry's definition of EFS, which includes specified progression, recurrence, and death events. It should not be generalized automatically to other chemotherapy regimens, populations, or clinical settings. The record describes a China extension involving previously untreated Chinese adults, and the primary comparison does not include the FLOT cohorts.

The posted medians are incomplete: the placebo-arm median is reported, while the pembrolizumab-arm median is not reported in the registry record. The confidence limits are also incomplete. These omissions, along with absent event counts and censoring details, limit how fully the time-to-event pattern can be understood from the registry alone.

The most defensible conclusion is about the statistical evidence rather than treatment choice: the reported estimate is uncertain and does not resolve the comparative EFS effect. The result should be interpreted alongside the endpoint definition, analysis population, limited reporting, and the study's stated non-testing purpose. It is not an individual prediction or a recommendation for clinical care.

18. Why This Trial Matters Statistically

This trial is statistically informative as an example of why a point estimate and a p-value should not be read in isolation. The reported hazard ratio is close to 1, but its confidence interval is wide enough to include values on either side of 1. The interval conveys substantial uncertainty that is not captured by repeating the point estimate alone.

It also illustrates the distinction between a planned primary comparison and the broader set of outcomes recorded for a trial. The primary EFS analysis concerns the XP/FP arms, while some secondary outcomes include FLOT cohorts. Results from different endpoint populations should not be treated as interchangeable evidence. The registry's statement that no formal hypothesis testing would be done is important context when interpreting any reported p-value or secondary finding.

Time-to-event outcomes use the timing of events and censoring, so enrollment alone does not tell readers how much statistical information was available. The registry record does not give the event count or detailed follow-up information for the primary endpoint. That makes the confidence interval especially useful as a concise indicator of uncertainty, while also leaving important analysis details unavailable for review.

The planning illustrations make a separate statistical point: sample-size requirements depend on the effect one aims to detect and on the design assumptions. Smaller target effects require more information. Those illustrations are not the trial's original design calculation and are not observed power. Keeping the distinction clear prevents an analysis based on the observed result from being mistaken for the rationale used before the trial began.

19. Limitations

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

The registry record has several limitations specific to interpreting this comparison. Although it gives the primary EFS definition and identifies the analysis population, it does not report the number of EFS events, the distribution of follow-up, or the censoring rules. These details would help readers judge how much information supports the hazard-ratio estimate and how the time-to-event analysis was implemented.

The median reporting is incomplete. A median is listed for placebo plus XP/FP but not for pembrolizumab plus XP/FP, and upper confidence limits are not reported. The record does not explain these omissions. It is therefore not possible to compare the arm-specific medians fully or to infer why a value is unavailable.

The methods description is also limited. The registry lists Log Rank but does not report a full model specification, an assessment of proportional hazards, covariate adjustment, or sensitivity analyses. Randomization details and procedures for handling missing assessments and protocol deviations are not reported in the registry record. These are reporting gaps; they do not establish that the underlying trial lacked such procedures.

The study summary says no formal hypothesis testing would be done, while a p-value is nevertheless listed for the primary comparison. No testing hierarchy or multiplicity adjustment is reported. This makes the p-value less suitable as a confirmatory decision statistic and supports placing more emphasis on the estimate, its confidence interval, and the stated study purpose.

Finally, the primary comparison is restricted to XP/FP arms, while other registry outcomes may combine treatment cohorts. The results should not be extended beyond the analyzed population or endpoints without supporting data. Detailed arm-specific results for all secondary outcomes and fuller safety context are not reported in the registry record.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Study of Pembrolizumab (MK-3475) Plus Chemotherapy Versus Placebo Plus Chemotherapy in Participants With Gastric or Gastroesophageal Junction (GEJ) Adenocarcinoma (MK-3475-585/KEYNOTE-585)-China Extension — NCT04882241.
• Registry record last updated 2026-03-06; status last verified 2026-02.
• Results first posted 2025-02-20.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

23. Record Summary

The registry describes a completed, randomized, double-blind, quadruple-masked Phase III China extension of KEYNOTE-585, comparing pembrolizumab plus XP/FP with placebo plus XP/FP in a neoadjuvant or adjuvant setting for gastric or gastroesophageal junction adenocarcinoma. The primary outcome was investigator-assessed EFS under RECIST 1.1, analyzed in all randomized participants in the two XP/FP arms.

The reported primary hazard ratio is 0.92, with a 95% confidence interval from 0.50 to 1.70. The interval is broad and includes 1; the arm-specific median reporting is incomplete. The registry summary states that no formal hypothesis testing would be done. Event counts, censoring details, missing-data procedures, and the original sample-size justification are not reported in the registry record.

The secondary outcomes cover disease-free survival, overall survival, event-free survival, and safety. The posted safety summary reports serious adverse events and deaths in both primary-comparison arms, but the registry record does not provide all details needed for a full comparative assessment. Overall, the record supports a cautious reading of an uncertain primary estimate rather than a definitive conclusion about comparative efficacy.