Epacadostat and Pembrolizumab in Patients With Head and Neck Cancer That Have Failed Prior Immunotherapy: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Epacadostat and Pembrolizumab in Patients With Head and Neck Cancer That Have Failed Prior Immunotherapy (NCT03463161), a non-randomized phase 2 study. The primary outcome is Response Rate.

2. Clinical Question

The question was whether the combination could be associated with tumor response in people whose disease had been treated previously with immunotherapy. The group definitions reflect different patterns of prior benefit, but the record does not report separate outcome results for those groups. With an enrollment of 2 and no reported response result, the registry record does not establish an estimate of response for either group or for the combined population. It also does not provide a comparison against another treatment or against no treatment. The study’s stated question is therefore distinct from a controlled estimate of the combination’s effect.

3. Trial Design

This was a phase II, non-randomized, parallel study sponsored by the University of Chicago. The registry record lists an enrollment of 2 and a status of terminated. The recorded start date is 2018-03-23, and the primary completion date is 2018-12-04. The reason for termination is not reported in the registry record.

The study described two participant groups: Acquired Resistance and Suboptimal Benefit. The former included people who had benefited from prior immunotherapy and then had disease worsening on the most recent imaging. The latter included people described as having had stable disease lasting at least 5 months or a suboptimal response, defined in the record as 11-49% shrinkage, with disease remaining stable on the most recent imaging. Both groups were to receive pembrolizumab and epacadostat.

Non-randomized allocation means participants were not assigned by chance to different treatments. The parallel structure refers to groups followed over the same general study period; it does not make the groups randomized or ensure that they were comparable. Because both groups received the same combination and were distinguished by prior treatment experience, any difference between their outcomes would not isolate a treatment effect. No such group-specific results are reported in the registry record.

The study had no masking. The registry record does not report a control treatment, a protocol for assigning participants between the two groups, or the number enrolled in each group. It also does not report the reason for termination, treatment exposure, or a detailed account of participant follow-up. These omissions limit how much can be concluded about study conduct from 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

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 Response Rate, measured over up to 18 months.

The primary endpoint was Response Rate, measured as a count of participants. It was described as the rate of clinical disease response, meaning tumor shrinkage seen on imaging, with assessment over a timeframe of up to 18 months. A response endpoint is used to describe how many participants meet a defined response criterion during a stated assessment period. Imaging-based tumor change can be a direct way to characterize response, but interpretation depends on how response is defined, when imaging is done, and how missing or unevaluable assessments are handled.

The registry record does not report the response count, a response rate, a confidence interval, or details of the response criteria and assessment procedures. It therefore does not permit a reader to assess the magnitude or precision of the primary endpoint result. The term “rate” in the endpoint name should not be mistaken for a reported result: the registry record identifies the endpoint and its timeframe, but does not report the observed response outcome.

Response rate is also different from progression-free survival and overall survival, which are listed as secondary outcomes. Those outcomes concern time to an event, while response rate is a count-based endpoint in this record. The definitions, analysis methods, and results for the secondary outcomes are not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a response count tell the reader?

A response count records how many participants met a response definition. It is a descriptive result; by itself, it does not show how uncertain that result is. A confidence interval is a range of values compatible with the observed data under a specified statistical method. It helps readers judge precision: a wider interval indicates greater uncertainty. No response count or confidence interval is reported in the registry record, so neither the observed response nor its precision can be assessed.

Response endpoints also depend on the assessment rules. The registry description refers to tumor shrinkage seen on imaging, but the detailed response criteria, imaging schedule, confirmation requirements, and handling of participants without an evaluable scan are not reported in the registry record. Those details matter because different rules can classify the same participant differently. The timeframe of up to 18 months defines the stated assessment window, but does not by itself explain how response timing or follow-up was handled.

Why does the design limit causal interpretation?

Randomization is a method for assigning participants to treatment groups by chance. In a randomized comparison, it can help balance known and unknown participant characteristics across groups. This study was non-randomized, and both described groups were to receive the same combination. Their distinction was based on prior immunotherapy benefit and current disease status. The groups therefore cannot serve as a randomized treatment-versus-control comparison, and outcome differences between them would not isolate the effect of the combination.

Masking concerns whether participants, clinicians, or outcome assessors know the assigned treatment. The registry record lists no masking. It does not report whether imaging assessments were independently reviewed or whether assessors used any procedures intended to limit knowledge-related influence. Those details are not reported in the registry record.

How should the small enrollment be understood?

The enrollment was 2. A study with very few participants can describe what was recorded for those participants, but estimates from such a small group are inherently imprecise and can be strongly affected by one participant’s outcome. The record does not report the number assigned to each of the two groups, the analysis population, or participant-level outcome data. The response result therefore cannot be reconstructed from the registry record.

Planning sample-size illustrations are not the same as the study’s actual design justification or observed power. The record describes illustrations based on an observed effect, with a two-sided significance level of 0.05, equal allocation, a normal approximation, and a target effect stated as a fraction of the observed effect. Those are assumptions for illustrations, not evidence that the trial planned or achieved a particular power. The trial’s own sample-size justification is not reported in the registry record. For time-to-event illustrations, the stated method concerns a total number of events; other illustrations concern participants per arm. Such planning calculations do not replace an analysis of the study’s actual outcomes.

What do time-to-event outcomes require?

Progression-free survival and overall survival are listed as secondary outcomes. Time-to-event analyses generally account for both whether an event occurred and how long participants were followed. Participants who have not had the event by the end of their observed follow-up may be censored, meaning their event time is only known to exceed their observed follow-up. Appropriate interpretation depends on event definitions, follow-up, censoring, and the chosen analysis method. The definitions, event counts, follow-up details, and analysis methods for these outcomes are not reported in the registry record.

For time-to-event studies, methods such as the Schoenfeld approach can be used in planning to relate the required event count to a target effect and other assumptions. A planning illustration of this kind does not establish that the study collected enough events or that its own analysis used that method. No time-to-event analysis is reported in the registry record.

What can be inferred from the absence of analyses?

The registry record lists no analyses. This means it does not present an analysis result for the primary response outcome or the listed secondary outcomes. It does not explain whether analyses were performed outside the registry, whether outcomes were unavailable, or why results are absent. Those possibilities are not reported in the registry record and should not be guessed. In particular, missing analysis results should not be interpreted as a null finding, a lack of response, or evidence of no difference.

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("Acquired Resistance or Suboptimal Benefit Group")
n      <- c(2)
events <- c(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  Acquired Resistance or Suboptimal Benefit Group  2  0  0.0  0.0  84.2  0.0  65.8  0.8  70.8

11. Multiplicity, Missing Data and Analysis Populations

The registry record lists Response Rate as the primary outcome and lists progression-free survival, overall survival, and side effects as secondary outcomes. It also indicates 4 outcome measures, but does not report all their definitions or analysis details. When several outcomes or several statistical comparisons are assessed, the chance of at least one apparently unusual result can increase. Statistical methods for addressing multiplicity may include prioritizing outcomes or adjusting the interpretation of tests. Whether any multiplicity strategy was planned or used is not reported in the registry record.

The record lists no analyses. It does not report a statistical analysis plan, test procedures, confidence intervals, handling of repeated assessments, or any interim analyses. The reason the study was terminated is also not reported in the registry record. These omissions prevent assessment of whether the analysis followed a prespecified plan or how results would have been interpreted across outcomes.

Missing-data procedures and analysis populations are not reported in the registry record. In a response analysis, it matters whether every enrolled participant is included, how participants without a post-baseline assessment are treated, and what happens when an assessment is unevaluable. For time-to-event outcomes, the definition of censoring and the completeness of follow-up matter. The registry record does not provide participant disposition, the number treated, the number assessed for each outcome, or reasons for missing observations.

Without those details, the reader cannot determine whether reported counts use all enrolled participants or a narrower group. There are no reported efficacy analyses to assess for missing-data sensitivity or multiplicity. The absence of these details does not establish that a particular method was or was not used; it means the relevant methods are not reported in the registry record.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent non-serious adverse event terms

The registry record reports serious adverse events for the Acquired Resistance or Suboptimal Benefit Group. Of 2 participants at risk, 0 were recorded as affected by a serious adverse event. Deaths are listed as 0. For other adverse events, 2 participants were at risk and 2 were recorded as affected. These are counts, not estimates of a comparative treatment risk.

The record does not list the terms for the other adverse events, and no top serious adverse-event terms are given. It therefore does not identify the event types, their severity beyond the serious/non-serious reporting categories, their timing, or whether they were considered related to study treatment. The affected count for other events should not be read as meaning that every participant experienced the same event, or as a description of the events’ clinical importance.

With only 2 participants at risk, the counts provide a limited description of safety in the enrolled group. They cannot characterize uncommon events or provide a reliable comparison with another treatment. The registry record does not report a separate safety analysis by prior-benefit group.

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

The record shows that a phase II study was registered to examine pembrolizumab plus epacadostat in people with head and neck cancer after prior immunotherapy. It defines two groups according to prior benefit and subsequent disease status, and states that both groups would receive the combination. The study was non-randomized, had no masking, enrolled 2 participants, and is listed as terminated.

The registry record does not report the primary response count or rate, or results for progression-free survival or overall survival. It therefore does not tell readers how many participants had tumor shrinkage, how long participants remained free of progression, or how long they lived. It also does not provide an estimate of treatment effect or its precision. Missing efficacy results do not establish that the combination had no effect; they leave the efficacy question unanswered in this record.

The study’s two groups are not a control comparison. Their prior treatment histories differ by definition, and both were to receive the same intervention. Any outcome contrast between them, if one were available, would not by itself demonstrate a treatment effect. The record also gives no result against which to compare the combination with another therapy or with no treatment.

The adverse-event counts indicate that no serious adverse events and no deaths were recorded among the 2 participants at risk, while other adverse events were recorded for 2 participants. The event types and further safety details are not reported in the registry record. These figures are a limited accounting of recorded events, not proof of safety or a complete account of risk.

Overall, the record supports description of the intended question and the study’s limited enrollment and design, but not a conclusion about efficacy or comparative safety. Any stronger interpretation would require outcome results and additional information about analysis, follow-up, participant disposition, and adverse-event details.

14. Why This Trial Matters Statistically

This trial illustrates why the study design and the completeness of outcome reporting matter as much as an endpoint’s name. A response-rate question can be clear, but without a reported response count and an assessment of uncertainty, readers cannot estimate what the study observed or how precise that estimate is. The registry record identifies the endpoint but does not report its result.

It also illustrates the distinction between clinically different groups and treatment comparison groups. The Acquired Resistance and Suboptimal Benefit groups were defined by prior immunotherapy experience, but both were to receive the same combination. Those categories may organize the study question, yet they do not create a randomized control. A contrast between them could reflect differences in prior disease course rather than a difference caused by treatment.

The enrollment of 2 is central to interpretation. With so few participants, outcome counts would be highly limited as evidence about a broader population, and any estimate would have substantial uncertainty. The record’s terminated status adds context, but the reason for termination is not reported, so its implications cannot be determined from the registry record alone.

Finally, the listed safety counts show why aggregate event totals need context. The record gives counts at risk and affected, but does not describe the other adverse-event terms or their relationship to treatment. A count is informative as a record of what was reported; it cannot substitute for event descriptions, follow-up details, and a comparison group. Together, these features make this study a useful example of the limits of inference when enrollment is small, allocation is non-randomized, and outcome analyses are not reported.

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 immediate limitation is that the primary response outcome is not reported in the registry record. Without the response count, rate, or uncertainty estimate, the primary study question cannot be answered from the posted record. The listed secondary outcomes likewise have no reported results or analysis details.

The study enrolled 2 participants and was terminated. The reason for termination is not reported in the registry record, and the record does not give participant disposition, treatment exposure, or the number assigned to each prior-benefit group. It is not possible to determine from these details how many participants received treatment or contributed data to each outcome.

The non-randomized design and lack of a control group limit causal interpretation. The two groups were defined by prior immunotherapy benefit and both were to receive pembrolizumab and epacadostat. They cannot provide a randomized comparison of treatment strategies. The record also lists no masking, and does not report whether outcome assessments had other safeguards against assessment bias.

Important endpoint and analysis details are not reported in the registry record. These include the full response criteria, imaging schedule, missing-data approach, analysis populations, statistical analysis plan, multiplicity procedures, and methods for the time-to-event outcomes. The registry record lists no analyses and does not explain why outcome analyses are absent.

Safety interpretation is also limited: adverse-event terms beyond the aggregate counts are not reported, and there is no comparative safety group. These limitations mean the record can describe the study’s stated design and a small set of safety counts, but cannot establish an efficacy estimate or a broad characterization of harms.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Epacadostat and Pembrolizumab in Patients With Head and Neck Cancer That Have Failed Prior Immunotherapy — NCT03463161.
• Registry record last updated 2020-01-27; status last verified 2020-01.
• Results first posted 2020-01-27.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

The registry record describes NCT03463161, a phase II, non-randomized, parallel study sponsored by the University of Chicago. It evaluated a planned combination of pembrolizumab and epacadostat in people with head and neck cancer after prior immunotherapy. Two groups were defined by prior benefit and subsequent disease status; both were to receive the same combination.

The recorded enrollment was 2. The study is listed as terminated, with a start date of 2018-03-23 and a primary completion date of 2018-12-04. The primary outcome was response rate, described as tumor shrinkage on imaging over up to 18 months. The response result and analyses are not reported in the registry record.

Progression-free survival, overall survival, and side effects are listed as secondary outcomes. Safety counts show 0 serious adverse events and 0 deaths among 2 participants at risk, and other adverse events in 2 of 2 participants at risk. Further event details are not reported in the registry record.