Biomarkers of Response to Pembrolizumab Combined With Chemotherapy in Non-Small Cell Lung Cancer (KEYNOTE-782, MK-3475-782): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Biomarkers of Response to Pembrolizumab Combined With Chemotherapy in Non-Small Cell Lung Cancer (KEYNOTE-782, MK-3475-782) (NCT03664024), a na phase 2 study. The primary outcome is Tumor Mutation Burden (TMB) in Cell-free Circulating Tumor Deoxyribonucleic Acid (ctDNA).

2. Clinical Question

3. Trial Design

KEYNOTE-782 (NCT03664024) is registered as a Phase II, single-group, unmasked trial of pembrolizumab combined with chemotherapy for non-small cell lung cancer. The record lists enrollment of 118 participants and an overall status of completed. The stated population is participants with Stage IV nonsquamous disease without prior systemic treatment. The registry identifies Merck Sharp & Dohme LLC as sponsor.

The treatment sequence began with pembrolizumab and a platinum doublet for 4 cycles. The doublet was pemetrexed combined with the investigator’s choice of cisplatin or carboplatin. Maintenance treatment was pembrolizumab plus pemetrexed for up to 31 additional cycles. Thus, the treatment description allows a choice of platinum agent within the single treatment group; it does not describe a randomized comparison between cisplatin and carboplatin.

The registry lists a start date of 2018-10-30 and a primary completion date of 2021-11-05. These dates and the completed status describe the trial’s registration history, but do not establish that every planned outcome was reported in the registry. Allocation is listed as not applicable, and masking is listed as none. There is no parallel control arm identified.

Secondary outcomes include progression-free survival, overall survival, the percentage of participants who experienced one or more adverse events, and the percentage who discontinued study intervention because of an adverse event. The registry lists 6 outcome measures, but the record summarized for this analysis does not provide numerical results for the primary endpoint. It also lists no analyses. Protocol details such as a statistical analysis plan, sample-size justification, interim analyses, and prespecified handling of missing data are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All participants who received at least one dose of study intervention.

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 Tumor Mutation Burden (TMB) in Cell-free Circulating Tumor Deoxyribonucleic Acid (ctDNA), measured over Baseline (Day 1).

The registered primary outcome is tumor mutation burden in cell-free circulating tumor deoxyribonucleic acid, measured at baseline on Day 1. TMB is described as the number of somatic mutations per megabase of DNA, expressed in mut/MB. Cell-free circulating tumor DNA in blood can be used to explore features of a tumor; the recorded outcome summarizes participants’ baseline TMB as a mean, with standard deviation as the measure of dispersion.

A mean describes the average of the observed values, while the standard deviation describes how spread out individual values are around that average. Both are useful for characterizing a continuous biomarker: the mean indicates its central level in the analyzed group, and the standard deviation gives context about variation between participants. Neither quantity by itself shows whether a biomarker predicts response.

The stated primary objective concerns the association between baseline TMB and objective response assessed by the investigator according to RECIST 1.1. An association analysis would address whether response patterns vary with TMB; it is distinct from reporting the overall mean TMB. The registry record identifies the mean TMB as the primary outcome, but does not report its numerical value, standard deviation, the response results, or an estimate of the TMB-response association. The intended analysis population consists of participants who received at least one dose and had samples evaluable for TMB analysis.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does the primary measure summarize?

The primary measure is continuous: baseline TMB is expressed in mut/MB, and the registered summary is a mean with standard deviation. For a continuous measure, the mean describes the center of a group’s values. The standard deviation describes the degree of spread around that center. A mean and standard deviation can be informative when the distribution is reasonably represented by these summaries, but they do not reveal the full distribution, the presence of unusual values, or how individual measurements relate to a clinical outcome.

The record defines the relevant population as participants who received at least one dose and had samples evaluable for ctDNA TMB analysis. This is not necessarily identical to all enrolled participants. In general, an analysis restricted to evaluable samples can be affected if sample availability or evaluability is related to participants’ characteristics or outcomes. The registry record does not report the number with evaluable samples or how unevaluable or missing samples were handled.

How is a biomarker association different from a mean?

The stated objective is to estimate the association between baseline TMB and investigator-assessed objective response under RECIST 1.1. That question concerns how response varies in relation to TMB, not just the average TMB in the group. A statistical analysis would need to relate biomarker values to response information and define how the association is expressed. The registry record does not report the association estimate, its uncertainty, the analysis procedure, or a numerical response result. It is not appropriate to infer a particular model or test from the objective wording alone.

In general, an estimate is more informative when accompanied by a confidence interval. A confidence interval conveys uncertainty about the value compatible with the data and the analysis assumptions; it is not a guarantee that the true value lies within the interval. No confidence interval for the primary measure or the stated association is reported in the registry record. A p-value, if reported, would address compatibility with a specified null hypothesis under an analysis model; it would not by itself indicate the size or practical importance of an association. The registry record does not provide a primary association p-value.

What does the single-group design allow?

The registry identifies one treatment group and no randomized control arm. The participants received pembrolizumab with pemetrexed and a platinum agent, then pembrolizumab plus pemetrexed maintenance. Cisplatin or carboplatin was selected by the investigator, so these are not described as randomized groups. In general, a single-group study can describe outcomes and explore associations among participants receiving the regimen. Without a concurrent comparator, it cannot isolate the treatment’s effect from other influences or estimate a relative treatment effect against a control.

Likewise, a biomarker that is associated with response among treated participants is not automatically a predictive marker of differential treatment benefit. A marker can be related to outcome for reasons that do not establish that treatment effects differ by marker level. Establishing such a distinction generally calls for a comparison of treatment effects across marker groups. The registry record does not describe such a comparative design or report an analysis of treatment-by-biomarker differences.

What do the planning illustrations represent?

Any sample-size figures associated with planning are illustrations based on an observed effect under stated assumptions, rather than the trial’s own enrollment plan. The stated illustration assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the target effect is a stated fraction of the observed effect. For time-to-event outcomes, the planning quantity is a total number of events, while other planning quantities are participants per arm for 80% power or another stated power level.

Those assumptions do not describe the registered single-group study, which has no two-arm randomized comparison. They must not be interpreted as observed power, as a justification for the enrollment of 118, or as evidence that the trial was designed around the primary biomarker objective. The registry record does not report the trial’s own sample-size justification. Since the primary association estimate is also not reported, the planning illustrations cannot substitute for a reported result or establish the precision achieved in the trial.

Why does over-dispersion matter in general?

Over-dispersion is a general issue for some statistical models when observed variability is greater than the model assumes. If ignored, uncertainty estimates may be too small and inferential conclusions may be overly confident. The registered primary measure is a continuous mean and standard deviation; the registry record does not identify an over-dispersion issue or describe a model for the biomarker-response association. This general concept should not be taken as a claim about the analysis used in this trial.

More broadly, a statistical plan would ordinarily clarify the estimand, analysis population, handling of missing observations, adjustment variables, and approach to uncertainty. These details are not reported in the registry record. Without them, readers cannot determine which analytic choices were prespecified or how sensitive a result might be to alternative reasonable choices.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Pembrolizumab Plus Platinum-doublet Chemotherapy O", "Pembrolizumab Plus Platinum-doublet Chemotherapy R", "Pembrolizumab Plus Platinum-doublet Chemotherapy N")
n   <- c(101, 42, 59)
m   <- c(9, 8, 10)
sp  <- c(11.6, 8.3, 13.5)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Pembrolizumab Plus Platinum-doublet Chemotherapy O  101  9.000  11.600  6.710  11.290
CARM  Pembrolizumab Plus Platinum-doublet Chemotherapy R  42  8.000  8.300  5.414  10.586
CARM  Pembrolizumab Plus Platinum-doublet Chemotherapy N  59  10.000  13.500  6.482  13.518

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 6 outcome measures, including the primary baseline TMB measure and secondary outcomes of progression-free survival, overall survival, adverse events, and discontinuation due to adverse events. Considering several outcomes or several analyses can raise multiplicity concerns: when many hypotheses are assessed, the chance of at least one apparently notable result can increase. Statistical plans may address this through a prespecified hierarchy, adjusted significance thresholds, or a clear distinction between confirmatory and exploratory analyses. The registry record does not report a multiplicity strategy, a hierarchy among outcomes, or an analysis plan for the stated biomarker-response objective.

The record lists no analyses and does not provide a numerical primary endpoint result. It also does not report the number of participants with evaluable ctDNA TMB samples, the amount of missing outcome information, or methods for handling missing data. The primary analysis population requires at least one dose and an evaluable TMB sample; this criterion means the analyzed group may be smaller than the enrolled group, but its size is not reported in the registry record.

In general, missingness can affect an estimate when participants with missing information differ systematically from those with observed information. A complete-case result, if that were the approach, would describe those included and might not represent everyone enrolled. No such approach should be attributed to this study: the registry record does not specify how missing samples or outcome data were handled. It likewise does not report whether any sensitivity analyses were planned or conducted. Without these details, the completeness and robustness of the primary analysis cannot be assessed from 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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pneumonia (9.4%), febrile neutropenia (4.3%), diarrhoea (3.4%), respiratory tract infection (3.4%), pancytopenia (2.6%).

For the group titled pembrolizumab plus pemetrexed plus platinum agent, the registry lists 117 participants at risk for serious adverse events and 60 participants with serious adverse events. It lists 117 at risk for other adverse events and 114 with other adverse events. The record also lists 80 deaths. These are descriptive safety figures for the single treatment group; there is no control group in the registry design for a comparative safety estimate.

The registry lists pneumonia at 9.4%, febrile neutropenia at 4.3%, diarrhoea at 3.4%, respiratory tract infection at 3.4%, pancytopenia at 2.6%, and thrombocytopenia at 2.6% among the top serious adverse-event terms. These percentages describe the listed terms, not the overall frequency of every possible event. The registry record does not provide additional event-level context here, such as timing, duration, severity distributions, or a comparison with another regimen.

Counts and term percentages do not establish causality. The safety figures should not be read as proving that treatment caused an event, nor do they establish how the regimen compares with another option. The registry record does not report a control-group rate, and the displayed summary does not identify the cause of the listed deaths.

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

The registry record establishes that KEYNOTE-782 was a completed Phase II, single-group study of pembrolizumab with platinum-doublet chemotherapy followed by maintenance pembrolizumab and pemetrexed in participants with Stage IV nonsquamous non-small cell lung cancer without prior systemic treatment. It also defines a baseline biomarker measure: mean TMB in cell-free circulating tumor DNA among participants with an evaluable sample who received at least one dose.

The study’s stated primary objective was to estimate the association between baseline TMB and investigator-assessed objective response under RECIST 1.1. But the registry record does not report the mean TMB, its standard deviation, the number analyzed, an objective response result, or an association estimate. As a result, the posted information does not show whether TMB was associated with response, how large or precise any association was, or whether the primary objective was met.

Even if an association estimate were available for this treated group, interpretation would depend on the analysis and its uncertainty. An association among participants receiving one regimen would not by itself show that TMB predicts the relative benefit of that regimen compared with another treatment. The single-group design has no randomized comparator to support that kind of treatment-effect comparison.

The safety summaries provide counts and percentages for adverse-event categories in the one treatment group. They are not comparative estimates and do not, by themselves, establish causation. The record also does not give enough information to assess how missing samples affected the biomarker analysis or to judge the precision of the primary endpoint. Consequently, the trial’s registration and safety descriptions are interpretable, while the central numerical biomarker and response findings remain not reported in the registry record.

14. Why This Trial Matters Statistically

This trial raises a useful statistical distinction between a biomarker’s distribution and its potential relationship to a clinical outcome. Reporting mean baseline TMB and its standard deviation would describe the biomarker values in the analyzed population. Assessing whether TMB is associated with objective response requires an analysis that links those values with response. The registry record identifies both the biomarker and the intended response-related objective, but it does not report the numerical results needed to evaluate that link.

The distinction also matters for the word predictive. In general, a marker associated with outcome among people receiving a treatment is not necessarily evidence that the marker identifies who benefits more from that treatment relative to another. A single-group study can explore within-group relationships, but a comparative treatment-effect question needs a suitable comparison. KEYNOTE-782 is registered as a single-group study with no randomized comparator, so its design does not provide that contrast.

The record also illustrates why analysis populations and missing biomarker samples matter. Enrollment is 118, while the primary outcome definition requires both treatment exposure and an evaluable ctDNA sample. Without the number analyzed and information about missingness, readers cannot tell how broadly the primary measure represents the enrolled population or how precise it may be.

Finally, safety summaries need a clear denominator and careful causal interpretation. The record gives a treatment-group denominator and counts for serious and other adverse events, plus percentages for selected serious terms. Those descriptive summaries can characterize reported events, but cannot determine comparative safety in the absence of a control group. Distinguishing what a design can answer from what the registry actually reports is central to a sound statistical reading.

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 most important limitation for interpreting the registered primary question is that the registry record does not report the primary TMB result or the association between TMB and objective response. It gives no numerical mean, standard deviation, association estimate, confidence interval, or p-value for that objective. The record therefore supports discussion of the intended measure and design, but not a conclusion about the observed biomarker-response relationship.

The primary analysis population is defined by treatment receipt and sample evaluability, while the number meeting those conditions is not reported in the registry record. The amount and pattern of missing biomarker data, along with the handling of missing outcomes, are also not reported. These omissions prevent assessment of selection into the evaluable population and of the precision or representativeness of the analysis.

The single-group, unmasked design does not provide a randomized control group. Cisplatin or carboplatin was chosen by the investigator as part of the platinum doublet, rather than assigned as a registered comparative arm. Accordingly, the registry design does not permit a controlled estimate of treatment effect or a direct comparison between the platinum choices. A within-group biomarker association, even if available, would not establish differential treatment benefit.

The registry record does not report a statistical analysis plan, sample-size justification, interim analyses, multiplicity handling, or detailed methods for the intended association analysis. The listed planning assumptions are illustrations and do not document the trial’s own design rationale. Safety reporting is also descriptive: the record gives one group’s event counts and selected serious-event term percentages without a control group or the context needed to establish causation. These limitations arise from the information and design recorded, and should not be resolved by assuming unreported methods or results.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Biomarkers of Response to Pembrolizumab Combined With Chemotherapy in Non-Small Cell Lung Cancer (KEYNOTE-782, MK-3475-782) — NCT03664024.
• Registry record last updated 2024-08-22; status last verified 2024-07.
• Results first posted 2022-11-14.
• 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

NCT03664024, KEYNOTE-782, is registered as a completed Phase II, single-group, unmasked study with enrollment of 118 participants. It evaluated pembrolizumab with pemetrexed and investigator-choice cisplatin or carboplatin, followed by pembrolizumab plus pemetrexed maintenance. The primary outcome is mean baseline TMB in cell-free circulating tumor DNA among treated participants with evaluable samples.

The stated objective was to estimate the association between baseline TMB and investigator-assessed objective response under RECIST 1.1. The registry record does not report the primary endpoint values or that association analysis. It lists secondary outcomes including progression-free survival, overall survival, and adverse-event outcomes. Safety summaries include 60 participants with serious adverse events, 114 with other adverse events, and 80 deaths among a group with 117 at risk. These are descriptive figures, not comparative treatment effects.