Recombinant EphB4-HSA Fusion Protein and Pembrolizumab, MK-3475: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Recombinant EphB4-HSA Fusion Protein and Pembrolizumab, MK-3475 (NCT03049618), a non-randomized phase 2 study. The primary outcome is Overall Response Rate (ORR).

2. Clinical Question

3. Trial Design

This phase IIa trial evaluated recombinant EphB4-HSA fusion protein in combination with pembrolizumab. The registry lists University of Southern California as the sponsor and records the study as completed. The recorded start date is 2017-03-10, and the primary completion date is 2024-09-04. Enrollment is listed as 42.

The design is identified as non-randomized, unmasked, and single-group. It includes a Head and Neck Cancer Cohort and a Non-Small Cell Lung Cancer Cohort. Both cohort descriptions give the same treatment schedule: pembrolizumab intravenously over 30 minutes on day 1 and recombinant EphB4-HSA fusion protein intravenously over 30 minutes on days 1, 8, and 15. Courses repeat every 3 weeks for up to 24 months in the absence of disease progression or unacceptable toxicity.

Because there is no randomized control group, the design does not support a like-for-like comparison of the combination against another treatment or against no treatment. The two cancer cohorts also represent different disease groups, so differences between their outcomes could reflect differences in the populations as well as other factors. The registry record does not report a protocol-level statistical plan, sample-size justification, interim analyses, or a randomized treatment comparison.

The primary endpoint is overall response rate, measured in an evaluable population defined as patients who received at least 1 dose of therapy. The stated assessment period is up to 22 months. The registry says no statistical analysis is performed for this endpoint. Secondary outcomes listed are duration of response, number of participants with adverse events, overall survival, and progression free survival. The record lists 20 outcome measures and 12 analyses, but does not describe a multiplicity adjustment.

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: Evaluable population of patients who received at least 1 dose of therapy. No statistical analysis is performed.

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: All enrolled patients are included.

6. Primary Endpoint

The registry lists the primary outcome as Overall Response Rate (ORR), measured over Up to 22 months.

Overall response rate (ORR) counts participants with a complete response or partial response, as defined by RECIST v1.1. The registry describes complete response as disappearance of all target lesions. Partial response requires at least a 30% decrease in the sum of the longest diameters of target lesions relative to the baseline sum. The definition also requires the response assessment to account for non-target lesions and the absence of new lesions: a response of partial response or better assumes at least incomplete response or stable disease for non-target lesions and no new lesions.

The endpoint is based on the evaluable population: patients who received at least 1 dose of therapy. The registry states that responses are recorded from study entry until disease progression, with a timeframe of up to 22 months. In general, a response endpoint provides a way to describe tumor shrinkage under prespecified assessment criteria. It is distinct from outcomes such as duration of response, progression-free survival, and overall survival, which address how long a response lasts or time to later events.

The registry record defines the endpoint but does not report the number of participants with complete or partial response, an ORR estimate, or a confidence interval. It also states that no statistical analysis is performed for this endpoint. As a result, the endpoint definition alone does not establish the observed response rate or its precision.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The trial is listed as phase IIa, non-randomized, unmasked, and single-group. These design labels guide how the results can be interpreted. Randomization is a general method for balancing measured and unmeasured participant characteristics across treatment groups. Since this trial is not randomized and has no control group, a treatment-versus-control effect cannot be estimated from a randomized contrast. The two disease cohorts are descriptive groups, not randomized arms. Comparisons between them would mix treatment exposure with differences between cancer populations.

What does the primary endpoint count?

Overall response rate is a binary outcome at the participant level: a participant is classified according to whether a complete or partial response is recorded under RECIST v1.1. The registry gives detailed response criteria and specifies that the endpoint covers responses from study entry until disease progression. In general, binary outcomes are often described with a count and a proportion. A proportion would require both the number meeting the response definition and the relevant denominator. The registry record does not report those ORR counts or an ORR estimate, so no response proportion can be interpreted from this record.

The record defines the evaluable population as patients who received at least 1 dose of therapy. This is the stated population for the primary endpoint. It is not necessarily the same as the enrollment population: the record does not report the number who received a dose or the number with evaluable response data. No conclusion about the endpoint denominator or completeness of assessments can be drawn from the enrollment entry alone.

What would a confidence interval communicate?

In general, a confidence interval describes the statistical uncertainty around an estimate under the assumptions of the analysis. For a response proportion, an interval would help show how much uncertainty remains around the estimated response rate. A narrow interval generally indicates greater precision than a wide interval, but it does not address systematic biases or the absence of a control group. The registry record does not report an ORR estimate or confidence interval, so the precision of the primary result cannot be assessed.

Why does the absence of a comparator matter?

A single-group study can describe outcomes among participants receiving the study regimen. It cannot, on its own, separate the effect of treatment from other explanations for observed outcomes. General sources of uncertainty include differences in participant characteristics, the natural course of disease, and how outcome assessments are made. An unmasked design means masking is listed as none; the registry does not explain whether or how this might have influenced assessment. These are general considerations, not findings that a particular bias occurred in this trial.

The allocation entry is non-randomized and the model is single-group. Therefore, the two cohort labels should not be read as treatment arms with one serving as the other's control. The registry record explicitly notes that these are not a like-for-like control comparison. No adjustment method for differences between cohorts is reported.

What do the planning illustrations mean?

The sample-size assumptions describe planning illustrations based on the observed effect, rather than the trial's own design. They use a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the effect to detect is stated as a fraction of the observed effect. In general, such calculations ask how many participants per arm might be needed under specified assumptions to detect a chosen effect with a chosen power. For time-to-event outcomes, the assumptions describe a total number of events; for other outcomes, they describe participants per arm.

Those are hypothetical planning assumptions, not a retrospective measure of whether this trial had adequate power. Sample-size illustrations are not observed power, and they do not change the single-group design. The registry record does not report its own sample-size justification or the values used to generate particular planning outputs. Any planning figures should be interpreted as illustrations under stated assumptions, not as evidence of an observed treatment effect or as a replacement for a control group.

Why can over-dispersion matter?

In general, over-dispersion occurs when observed variability is greater than a statistical model expects. If a model assumes less variability than the data contain, its uncertainty estimates can be too small and its intervals or tests may be misleading. This concern is relevant to some count-data models, but the primary endpoint is described as a count of participants meeting a binary response definition, and the registry says no statistical analysis is performed for it. The registry record does not report an over-dispersion assessment or a model for the primary endpoint.

The record lists 20 outcome measures and 12 analyses. In general, assessing multiple outcomes or analyses can increase the chance of at least one apparently notable result if no multiplicity strategy is used. The registry record does not report whether multiplicity was controlled, which results were designated for formal testing, or an adjustment method. The listed number of outcomes and analyses alone does not establish that a particular correction was required or applied.

For all endpoints, interpretation depends on knowing the analysis population, outcome definition, denominator, handling of missing data, and uncertainty around estimates. The record gives the primary endpoint definition and population description but does not report an ORR result, missing-data procedure, or primary statistical analysis. These omissions limit statistical interpretation; they should not be filled in by assumptions about standard practice.

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("Head and Neck Cancer Cohort", "Non-Small Cell Lung Cancer Cohort")
n      <- c(25, 17)
events <- c(6, 3)
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  Head and Neck Cancer Cohort  25  6  24.0  9.4  45.1  11.5  43.4  11.6  43.6
ARM  Non-Small Cell Lung Cancer Cohort  17  3  17.6  3.8  43.4  6.2  41.0  6.4  41.4

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 20 outcome measures and 12 analyses. Multiple outcome assessments raise a general multiplicity issue: when many hypotheses are tested, the chance of at least one apparently unusual result can increase unless the analysis strategy accounts for multiple testing. The registry record does not report whether any adjustment was used, which analyses were considered confirmatory, or how the outcomes were prioritized beyond identifying ORR as the primary endpoint.

For the primary endpoint, the registry defines the evaluable population as patients who received at least 1 dose of therapy. It does not report how many participants were evaluable, whether any response assessments were missing, or how missing data were handled. It also does not describe a sensitivity analysis for missing response information. Without those details, the completeness of the response assessment and the denominator used for an ORR calculation cannot be independently interpreted.

The record does not report whether analysis populations for the secondary outcomes differed from the primary evaluable population. It lists duration of response, number of participants with adverse events, overall survival, and progression free survival, but does not give the statistical methods used for those outcomes. Sample-size justification, interim analyses, and a full statistical analysis plan are not reported in the registry record. These omissions leave the treatment of multiplicity, missing observations, and endpoint-specific analysis populations unclear.

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: hypertension (36.0%), headache (11.8%), dyspnea (11.8%), dysphagia (8.0%), aspartate aminotransferase increased (8.0%).

The registry lists serious and other adverse-event counts by cancer cohort. For the Head and Neck Cancer Cohort, it records 25 at risk and 12 affected for serious adverse events, and 25 at risk and 16 affected for other adverse events. For the Non-Small Cell Lung Cancer Cohort, it records 17 at risk and 8 affected for serious adverse events, and 17 at risk and 11 affected for other adverse events. The listed deaths are 12 in the head and neck cohort and 6 in the non-small cell lung cancer cohort.

These are descriptive counts. The registry record does not state in these entries that the events or deaths were caused by treatment, and the figures do not provide a randomized comparison. The record also lists serious adverse-event terms with percentages: Hypertension, 36.0%; Headache, 11.8%; Dyspnea, 11.8%; Dysphagia, 8.0%; Aspartate Aminotransferase Increased, 8.0%; and Anorexia, 8.0%. These percentages should be read as registry-listed values; the record does not explain their denominator alongside the terms. They do not establish comparative risk or causation.

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

The record describes a phase IIa study of pembrolizumab with recombinant EphB4-HSA fusion protein in separate head and neck cancer and non-small cell lung cancer cohorts. It gives a treatment schedule, an overall response rate definition based on RECIST v1.1, and adverse-event counts. These details help readers understand what was studied and how the primary endpoint was defined.

The primary endpoint result itself is not reported in the registry record. There is no ORR count, estimate, confidence interval, or statistical analysis to interpret. Therefore, the record does not allow a quantitative statement about the observed response rate or its precision. The definition of an outcome is not the same as a result for that outcome.

The trial is non-randomized, unmasked, and single-group. It does not include a randomized control group, so outcomes within the trial cannot establish what would have happened to comparable participants under a different treatment. Nor can the two disease cohorts act as control groups for one another: they describe different cancer populations. This design limitation is separate from the absence of a reported primary result.

The safety entries provide counts and listed serious adverse-event terms, but do not establish that events were caused by treatment or show how event rates compare with another treatment. The listed deaths likewise do not specify cause in these entries. The record does not report missing-data handling, the number evaluable for the primary endpoint, or the statistical plan for the listed analyses.

In general, trial results are best interpreted alongside estimates of effect, uncertainty, a clear analysis population, and a design that supports the intended comparison. Here, the registry information supports a description of the trial and its outcome definitions, but not a conclusion about the size or certainty of an efficacy effect or a comparative safety effect.

14. Why This Trial Matters Statistically

This trial illustrates why statistical interpretation begins with design and endpoint definition, not just a study label or treatment schedule. The registry identifies two clinically distinct cohorts, a single-group non-randomized design, and a primary outcome based on complete or partial response under RECIST v1.1. Those details clarify what the study sought to measure while also setting boundaries on what comparisons it can support.

With no randomized control group, the trial is descriptive with respect to outcomes under the combination regimen. This distinction matters because a response count, even if reported, would not by itself establish how outcomes compare with another treatment. The disease cohorts cannot supply that missing counterfactual, since they are different populations rather than randomized comparison arms.

The record also demonstrates the difference between a prespecified outcome definition and a reported outcome result. ORR is described in detail, including the evaluable population and response criteria, but the result and its uncertainty are not reported. That gap prevents an assessment of both the observed response and the precision of an estimate.

Finally, the listed outcome measures, analyses, and safety entries raise practical questions about populations, missing assessments, and multiplicity. The registry record does not answer those questions with a statistical analysis plan or endpoint-specific methods. For biostatisticians and clinical researchers, this is a useful reminder to distinguish registry design information, endpoint definitions, descriptive safety counts, and inferential results rather than treating them as interchangeable evidence.

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 main limitation for interpreting efficacy is that the registry record does not report the primary ORR result. It provides no response counts, estimate, interval, or statistical analysis. The evaluable population is defined as patients who received at least 1 dose, but its size and the handling of missing response assessments are not reported. The enrollment entry does not establish the ORR denominator.

The study is non-randomized, unmasked, and single-group. There is no like-for-like control comparison, and the head and neck and non-small cell lung cancer cohorts are distinct populations. This limits comparative interpretation and means cohort differences cannot be attributed to treatment differences. The registry does not report adjustment methods for any cohort comparisons.

The stated primary endpoint timeframe is up to 22 months, while the treatment courses are described as repeating for up to 24 months under specified conditions. The record does not provide further details to reconcile these different timeframes. It also does not report the statistical plan, sample-size justification, interim analyses, missing-data methods, or multiplicity strategy.

Safety information is presented as counts and listed terms, without a comparative treatment group. The record does not establish causation for the events or deaths. It also does not explain the denominator associated with the listed serious adverse-event term percentages. These limitations are specific to interpreting the recorded safety entries and do not imply that an event was or was not treatment-related.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Recombinant EphB4-HSA Fusion Protein and Pembrolizumab, MK-3475 — NCT03049618.
• Registry record last updated 2025-10-22; status last verified 2025-10.
• Results first posted 2025-10-22.
• 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

NCT03049618 is a completed phase IIa trial sponsored by University of Southern California. It studied pembrolizumab with recombinant EphB4-HSA fusion protein in head and neck cancer and non-small cell lung cancer cohorts. The registry records enrollment of 42 and describes a non-randomized, unmasked, single-group design.

The primary endpoint is overall response rate, defined using complete and partial response criteria under RECIST v1.1 in patients who received at least 1 dose of therapy. The stated timeframe is up to 22 months. The registry record does not report an ORR result or statistical analysis. Secondary outcomes include duration of response, adverse events, overall survival, and progression free survival. Safety counts and serious adverse-event terms are listed by cohort, but they do not provide a randomized comparison or establish causation.