Anti-PD-L1/TGF-beta Trap (M7824) Alone and in Combination With TriAd Vaccine and N-803 for Resectable Head and Neck Squamous Cell Carcinoma Not Associated With Human Papillomavirus Infection: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Anti-PD-L1/TGF-beta Trap (M7824) Alone and in Combination With TriAd Vaccine and N-803 for Resectable Head and Neck Squamous Cell Carcinoma Not Associated With Human Papillomavirus Infection (NCT04247282), a non-randomized phase 1, phase 2 study. The primary outcome is Number of Participants Who Experience a Pathologic Complete Response (pCR).

2. Clinical Question

The study also set out to assess imaging response, immune-related adverse events, recurrence-free survival, overall survival, and treatment-related delays to planned surgery. These are distinct questions: a tumor response measured on imaging is not the same outcome as a complete response in examined surgical tissue, and neither alone describes recurrence or survival over time. The registry describes eight outcome measures, but it does not report analyses for them in the record.

3. Trial Design

This is a Phase 1 and Phase 2, non-randomized, unmasked, sequential trial. Enrollment was 21, and the record lists the study as completed. The study dates in the registry are 2020-06-09 for start and 2021-08-24 for primary completion. The population was people with resectable head and neck squamous cell carcinoma not associated with human papillomavirus infection, and the treatments were given before surgery.

The arm descriptions distinguish three regimens: M7824 alone; M7824 with TriAd vaccine; and M7824 with TriAd vaccine plus N-803. The record labels the design sequential, but does not report the rationale or operational details governing movement between regimens. It also does not report random assignment, masking, or a concurrent control group; allocation is explicitly non-randomized and masking is none. Consequently, differences between arms cannot be interpreted as if participants had been randomly assigned to otherwise comparable groups.

The registry describes screening under a separate protocol and assessments including medical history and examinations, imaging, biopsies, electrocardiograms, endoscopies, and laboratory tests. The primary endpoint was assessed after treatment, approximately one month after being on study, using the resected tumor specimen. Surgery is thus part of the clinical context for the primary tissue-based assessment, rather than the endpoint being an imaging-only measure.

The registry reports 20/21 participants analyzed for the primary endpoint because one participant declined before treatment began. It does not report a sample-size justification, an interim-analysis plan, detailed treatment assignment procedures, or a full statistical analysis plan. The planning assumptions described for sample-size illustrations are not a description of the trial's own design or justification.

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: 20/21 participants were analyzed because one participant declined to participate before treatment started.

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: Baseline characteristic data collected for one participant that was enrolled but not treated is reported here.

6. Primary Endpoint

The registry lists the primary outcome as Number of Participants Who Experience a Pathologic Complete Response (pCR), measured over Post treatment after on study, approximately one month.

The primary endpoint is the number of participants who experienced a pathologic complete response (pCR). The registry defines pCR as absence of malignant cells in the resected tumor specimen. A pathologist examines the specimen after surgery, with the endpoint assessed approximately one month after participants were on study. This is a tissue-based endpoint: it concerns the microscopic findings in the removed tumor, not only a change in its size on a scan.

Pathologic response is used in studies of treatment before surgery because the specimen allows direct examination of the treated tumor. It can describe whether malignant cells remain in the resected tissue at the time of assessment. In general, that endpoint answers a different question from an imaging response, which classifies change in measurable lesions using radiographic criteria, and from longer-term outcomes such as recurrence or survival.

The endpoint is recorded as a participant count. The record identifies the population analyzed as 20/21 participants and explains that one participant declined before treatment started. It does not report the number of pCRs or a corresponding estimate of uncertainty. Without that result, the observed frequency of pCR cannot be described from the registry record, and no conclusion about the endpoint's magnitude or precision can be drawn.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

For a binary endpoint such as pCR, each participant is classified according to whether the event occurred. A count of participants with the event describes the observed number, while the analyzed population indicates how many participants contributed to that endpoint. In general, an event count is easier to interpret when accompanied by its denominator and an interval estimate; the interval conveys the uncertainty associated with an estimate. The registry record reports the analyzed population for pCR but does not report the pCR count or an interval estimate.

What does the endpoint measure?

The primary endpoint is a count of participants with pCR after treatment. Its definition is based on pathology: no malignant cells are found in the resected tumor specimen. This definition is not interchangeable with a radiographic response. The listed imaging response outcome uses RECIST Version 1.1 and concerns complete or partial response assessed by computed tomography. Tissue examination and imaging rely on different observations and answer related, but distinct, questions.

The primary endpoint is assessed after treatment, approximately one month after being on study. The outcome is therefore tied to the surgical specimen and the stated assessment time. The record does not report further details about pathology review procedures, such as independent review or adjudication, so those procedures cannot be characterized.

Why does the design affect comparison?

Randomization is a method for assigning participants to groups so that, in expectation, measured and unmeasured prognostic factors are balanced. This is a general statistical principle, not a claim about this trial. The registry identifies the allocation as non-randomized and the model as sequential. It also cautions against treating the arms as a like-for-like control comparison. Without random assignment, group differences can reflect participant characteristics, treatment selection, timing, or other factors as well as treatment regimen.

In a sequential design, treatment regimens are considered in a sequence rather than through a conventional randomized concurrent comparison. The registry labels this study sequential but does not report the operational sequence or participant assignment details. Consequently, no assumptions about cohort transitions, escalation rules, or selection criteria should be made. The record also lists no masking. Masking can reduce the influence of knowledge of treatment assignment on some assessments; its absence is a design feature, although the record does not specify how outcome assessment was managed.

What do estimates and tests usually convey?

In general, an estimate summarizes the observed data, while a confidence interval describes a range of values reasonably compatible with those data under the method's assumptions. A wide interval indicates less precision than a narrow one. A hypothesis test and its p-value address compatibility with a specified null model; a p-value is not the probability that a treatment works, nor does it quantify the size or clinical importance of an effect. None of these quantities for the primary endpoint is reported in the registry record.

For a binary outcome, uncertainty depends in part on the number of participants and the event count. With a small study population, estimates can be imprecise, particularly when events are uncommon. This is a general point about binary data and does not establish the precision of this trial's pCR result, which is not reported. A confidence interval cannot be reconstructed from the analyzed population alone without the endpoint count.

How should planning illustrations be understood?

The registry's sample-size assumptions describe planning illustrations based on an observed effect, with two-sided significance level 0.05, equal allocation between two arms, and a normal approximation. The effect to detect is described as a stated fraction of the observed effect. For time-to-event outcomes, the assumptions refer to a total number of events using the Schoenfeld approach; for other outcomes, they refer to participants per arm. Such figures, where presented, are planning illustrations rather than observed trial findings or evidence of achieved power.

These assumptions do not establish the actual design rationale for this trial. The registry does not report a trial-specific sample-size justification, the effect size used in a formal design, or a statistical plan linking those illustrations to the sequential cohorts. In general, power is a design-stage probability under specified assumptions, not a property that can be inferred after observing the results. The registry record does not report an observed-power analysis, and sample-size illustrations should not be interpreted as one.

Why do multiple outcomes need care?

The registry lists multiple outcomes spanning pathology, imaging response, immune-related adverse events, recurrence-free survival, survival, and surgical delay. When several outcomes or comparisons are examined, the chance of at least one apparently unusual result can increase if no multiplicity strategy is used. Statistical plans may address this through a prespecified hierarchy, adjustment, or emphasis on estimation rather than a collection of separate tests. The registry record does not report a multiplicity strategy or outcome hierarchy beyond identifying the primary endpoint.

Time-to-event outcomes such as being alive and recurrence free require methods that account for differing follow-up and censoring. In general, a survival curve estimates the probability of remaining event-free over time, while a hazard-based analysis compares event rates over follow-up under its assumptions. The registry lists probability of being alive and recurrence free and percentage alive, but does not report their numerical estimates, follow-up details, or analysis methods. It is therefore not possible to assess censoring, follow-up completeness, or the precision of those outcomes from this record.

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("Arm A, Cohort 1 Bintrafusp Alfa (M7824) (Days 1, 1", "Arm B,Cohort 1 M7824 1200mg+TriAdVaccine 5x10e^11v")
n      <- c(14, 6)
events <- c(0, 0)
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  Arm A, Cohort 1 Bintrafusp Alfa (M7824) (Days 1, 1  14  0  0.0  0.0  23.2  0.0  21.5  0.2  21.8
ARM  Arm B,Cohort 1 M7824 1200mg+TriAdVaccine 5x10e^11v  6  0  0.0  0.0  45.9  0.0  39.0  0.4  41.0

11. Multiplicity, Missing Data and Analysis Populations

The record names one primary endpoint, the number of participants who experience pCR, and several secondary outcomes. The secondary outcomes include imaging response, Grade 3 or 4 immune-related adverse events, being alive and recurrence free, being alive, and treatment-related adverse events that cause a delay of at least 4 weeks beyond planned surgery. The registry lists eight outcome measures in total. It does not report a multiplicity adjustment, testing hierarchy, or a prespecified strategy for interpreting multiple endpoints. Without such details, the intended confirmatory status of the secondary outcomes cannot be determined.

The record says 20/21 participants were analyzed for the primary endpoint because one participant declined before treatment began. That explains the difference between enrollment and the stated primary analysis population, but does not establish a general missing-data approach. The registry record does not report whether any other outcome data were missing, how missing values were handled, or whether sensitivity analyses were conducted. It also does not specify whether the primary analysis followed an intention-to-treat, per-protocol, or other named analysis population.

These omissions matter because different outcomes may have different denominators and different patterns of follow-up. In general, excluding participants with unavailable outcome data can affect interpretation if the reasons for missingness relate to outcome status. The record's explanation for the one participant who declined before treatment began is specific to that participant; it does not support assumptions about missingness for other measures. The registry lists 0 analyses, so no reported analysis details clarify these issues.

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: cardiac arrest (16.7%), pneumonitis (16.7%), sinus tachycardia (7.1%), vasculitis (7.1%).

The safety listing separates serious adverse events from other adverse events and reports the number at risk and the number affected for the first two arms. For Arm A, the serious-event listing reports 14 at risk and 2 affected; the other-adverse-event listing reports 14 at risk and 14 affected. For Arm B, the corresponding entries are 6 at risk and 1 affected for serious events, and 6 at risk and 6 affected for other events. Arm C has zero at-risk entries in this listing. These are registry entries, not adjusted comparisons between arms.

The listed deaths are 2 for Arm A and 1 for Arm B; the record lists zero for Arm C. The serious adverse-event terms shown include Cardiac arrest and Pneumonitis at 16.7% each, and Sinus tachycardia and Vasculitis at 7.1% each. The registry record does not state the denominator for these term percentages in the listed information. They should not be treated as arm-specific risks or compared across arms without that context.

In general, a serious adverse event is a regulatory severity category and is not synonymous with every adverse event or with an event judged to be caused by treatment. The counts describe recorded events and do not by themselves establish causation. Zero at-risk entries for Arm C mean that this safety listing contains no at-risk observations for that arm; they should not be reinterpreted as evidence that the regimen had no adverse events. The non-randomized design also limits direct safety comparisons.

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

The registry record establishes that a completed Phase 1 and Phase 2 study evaluated M7824-based regimens before surgery in people with resectable head and neck squamous cell carcinoma not associated with human papillomavirus infection. It identifies pCR in the resected specimen as the primary endpoint and gives a defined assessment timeframe. It also records an analyzed population of 20/21 participants and explains that one participant declined before treatment started.

Those details describe the study's question and conduct; they do not supply the primary outcome result. The number of participants with pCR is not reported in the registry record. Therefore, the record does not establish how often pCR occurred, how precise an estimate would be, or whether any difference between regimens was observed. The completion status does not fill in the missing endpoint result.

The study's non-randomized sequential structure also limits causal interpretation. General statistical reasoning distinguishes a treatment effect from an association between treatment group and outcome. Random allocation can support a more credible causal comparison by balancing factors across groups; this study's allocation is explicitly non-randomized, and the registry cautions that arms are not like-for-like controls. Even if arm-specific outcomes were available, differences could not automatically be attributed to the treatment components.

The listed secondary outcomes describe further questions about imaging response, immune-related adverse events, recurrence, survival, and delay to surgery. Their registration does not mean numerical results are available. The record lists eight outcome measures and 0 analyses, and does not report numerical secondary findings in the information summarized. Likewise, safety counts and serious-event terms require careful attention to the stated groups, denominators, and event definitions; they do not alone demonstrate that a treatment caused or prevented an event.

The appropriate conclusion is limited: the registry describes an early-phase presurgical study and its intended outcomes, but does not report enough results to characterize efficacy or make a comparative treatment claim. It is not a basis for medical advice or for deciding which regimen should be used.

14. Why This Trial Matters Statistically

This trial raises a useful statistical distinction between measuring a response and demonstrating a comparative treatment effect. Its primary endpoint uses pathology from the resected tumor, while one secondary outcome uses imaging criteria. The measures concern related aspects of tumor response, but they are not interchangeable. Recognizing what an endpoint actually records is essential before interpreting a numerical result.

The sequential, non-randomized design also illustrates why a set of treatment arms is not necessarily a controlled comparison. Regimens may be listed side by side, but without random allocation and a suitable comparison structure, group differences do not isolate the effects of the added treatment components. The registry itself warns that the arms are not a like-for-like control comparison. This is particularly important when more than one agent differs between regimens.

Another statistical lesson is the difference between an outcome being registered and an outcome being reported. The record specifies a pCR endpoint, secondary outcomes, and an analyzed population, yet does not report the pCR count or analyses. Clear endpoint definitions are valuable, but they cannot substitute for event counts, estimates, and uncertainty. A completed study label also does not, by itself, answer the scientific question.

Finally, the study connects short-term presurgical assessment with outcomes that require follow-up, such as recurrence-free and overall survival. Each outcome has its own measurement and analysis needs. Statistical interpretation depends not only on the treatment regimen, but also on the population analyzed, the outcome definition, missing-data handling, and the design used to make comparisons. The registry record documents some of these elements while leaving others unreported, making careful separation of known design details from unavailable results especially important.

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 this record is that it does not report the number of participants who achieved pCR. Although it states that 20/21 participants were analyzed and defines the endpoint, the central outcome result is absent. No primary estimate or measure of uncertainty is reported, so the endpoint's observed frequency and precision cannot be assessed.

The study was non-randomized and sequential, and the record explicitly cautions that the arms are not like-for-like controls. Treatment groups therefore should not be treated as exchangeable. The registry does not report the participant allocation process, details of the sequential design, or adjusted comparative analyses. The absence of those details prevents a fuller appraisal of any between-arm contrast.

The registry record does not report a trial-specific sample-size justification, interim-analysis plan, statistical analysis plan, multiplicity strategy, or missing-data methods. It reports that one participant declined before treatment and that 20/21 participants were analyzed for the primary endpoint, but does not explain the analysis population for every other outcome. Numerical results for the listed secondary outcomes are also not reported in the available record.

The safety listing has zero at-risk entries for Arm C, while the listed serious adverse-event terms have percentages without a stated denominator in the record. These details constrain interpretation of the safety summary and preclude meaningful arm-to-arm safety comparisons. More broadly, the registry information does not include enough outcome, follow-up, or analysis detail to assess the precision and completeness of the full study findings.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Anti-PD-L1/TGF-beta Trap (M7824) Alone and in Combination With TriAd Vaccine and N-803 for Resectable Head and Neck Squamous Cell Carcinoma Not Associated With Human Papillomavirus Infection — NCT04247282.
• Registry record last updated 2025-07-16; status last verified 2025-07.
• Results first posted 2023-08-08.
• 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

NCT04247282 is a completed, non-randomized, unmasked, sequential Phase 1 and Phase 2 trial sponsored by the National Cancer Institute. It enrolled 21 participants with resectable head and neck squamous cell carcinoma not associated with human papillomavirus infection. The study evaluated M7824 alone and in regimens with TriAd vaccine, including a regimen with N-803, before surgery.

The primary endpoint was the number of participants with pCR in the resected tumor specimen, assessed approximately one month after being on study. The registry reports that 20/21 participants were analyzed because one participant declined before treatment. The pCR count and numerical analyses are not reported in the registry record. Eight outcome measures are listed, and the analysis listing reports 0 analyses. The record also cautions that the non-randomized arms are not like-for-like controls.