Pembrolizumab (MK-3475) in Hepatocellular Carcinoma: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Pembrolizumab (MK-3475) in Hepatocellular Carcinoma (NCT02940496), a non-randomized phase 2 study. The primary outcome is Correlation of HCC Stiffness to Overall Survival.

2. Clinical Question

The study concerned people with advanced hepatocellular carcinoma (HCC), including those described as hepatitis C virus (HCV) positive and negative, whose disease had progressed on or after sorafenib or who had not tolerated sorafenib. Its primary question was whether HCC stiffness measured by magnetic resonance elastography (MRE) was associated with overall survival (OS). Stiffness was measured in kilopascals, and the registry description says its change was compared with OS.

The clinical question is therefore about a possible association between an imaging-derived measure and survival, not a randomized estimate of pembrolizumab’s effect against another treatment. The record gives the planned statistical approach but does not report the correlation estimate or its p-value. It also does not report survival results by stiffness or enough detail to evaluate a particular stiffness threshold. Consequently, the registry record establishes what association investigators intended to examine, but does not allow readers to determine its direction, magnitude, or precision.

3. Trial Design

This was a Phase II, parallel, non-randomized study sponsored by M.D. Anderson Cancer Center. The registry lists an enrollment of 15 and a status of completed. The study began on 2016-12-09, and the primary completion date was 2023-08-09. Masking was listed as none, meaning the registry does not describe blinded treatment assignment or assessment.

Arm A received pembrolizumab intravenously over 30 minutes on day 1, with cycles repeating every 21 days for up to 35 courses unless disease progressed or unacceptable toxicity occurred. Arm B was planned to receive pembrolizumab on the Arm A schedule, together with daily oral elbasvir/grazoprevir and ribavirin on days 1–28. Its antiviral treatment was described as continuing for 12–16 weeks unless progression or unacceptable toxicity occurred. The registry states that Arm B closed after activation and no patients were enrolled into it. Thus, although two arms were described, the record documents enrollment only in Arm A.

The study description frames pembrolizumab as second-line therapy for people who progressed on or after sorafenib or could not tolerate it. The official title includes HCV-positive and HCV-negative subjects, while the planned antiviral combination was listed only for Arm B. Since Arm B had no enrolled participants, the record cannot support an observed comparison of pembrolizumab alone with pembrolizumab plus antiviral drugs, nor a treatment comparison by HCV status.

The registry does not report a randomized assignment procedure, a concurrent control group, or a masking procedure. It also does not report a detailed sample-size justification or enough information to assess whether the enrollment target was chosen for the primary correlation question. These omissions limit assessment of the study’s planned inferential precision.

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: Eligible patients were over 18 years of age with radiographic disease progression on sorafenib or intolerance to sorafenib treatment, and ECOG 0 or 1. All patients underwent liver MRI with MR Elastography (MRE) and liver biopsy at baseline and at 6 weeks of therapy. Only the Pembrolizumab arm enrolled patients.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Arm B was closed after activation and no patients was enrolled into this arm.

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: Arm B was closed after activation and no patients was enrolled into this arm.

6. Primary Endpoint

The registry lists the primary outcome as Correlation of HCC Stiffness to Overall Survival, measured over up to 2 years.

The primary outcome was titled correlation of HCC stiffness to overall survival. Stiffness was measured using magnetic resonance elastography (MRE), with the result expressed in kilopascals (kPa). The registry says the change in stiffness was compared with OS over a timeframe of up to 2 years. It lists the summary measure as a mean with full range and describes the analysis as descriptive statistics and Spearman correlation.

MRE provides an imaging-based measure of tissue stiffness. Examining whether that measure varies with survival can help characterize whether an imaging feature is associated with the course of disease in the studied group. In general, a correlation describes how two measured quantities vary together; it does not establish that one causes the other. The registry does not report a stiffness summary, a correlation coefficient, a confidence interval, a p-value, or survival estimates for this endpoint. The specific relationship observed in the study is therefore not reported in the registry record.

The endpoint description does not clarify how change in stiffness was defined, which measurement time points were used, or precisely how OS was represented in the correlation analysis. Those details are important for interpreting the analysis and reproducing it. They are not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary analysis was described as descriptive statistics and Spearman correlation. Descriptive statistics summarize the values observed in a study group. They can convey the distribution and range of a measure, but they do not by themselves estimate a treatment effect against a control group. The registry lists the primary outcome summary as a mean with full range. It does not provide the numerical stiffness summary or explain the exact time points and data included in that summary.

What does Spearman correlation measure?

Spearman correlation is a rank-based measure of association. In general, it assesses whether higher values of one variable tend to occur alongside higher or lower values of another, without requiring the relationship to be a straight line. It is useful when the relationship is monotonic but may not follow a linear pattern, or when an analysis is based on ordered values. Like other correlation measures, it does not show that one variable causes another. A correlation may also be affected by the range of values, influential observations, and the way measurements are collected.

For this study, the listed variables were HCC stiffness measured by MRE and OS. However, the registry wording says a change in stiffness was compared with OS without specifying how that change was calculated or how survival time entered the correlation. The coefficient, its direction, its uncertainty, and the number of participants included are not reported in the registry record. Those omissions prevent readers from evaluating the observed association or its sensitivity to analytic choices.

How should the p-value criterion be read?

The registry says p-value < 0.05 was considered statistically significant. In general, a p-value describes how compatible the data are with a specified null model, under the assumptions of the analysis. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an association. A threshold is a decision convention rather than a measure of precision. Without the reported p-value, coefficient, and uncertainty interval, the stated criterion cannot tell readers whether the primary result was statistically significant or how precisely it was estimated.

Confidence intervals, in general, express uncertainty around an estimated quantity under a statistical model. A narrower interval indicates greater precision than a wider interval, while the interval’s location helps show which values are compatible with the data and model. No confidence interval for the primary endpoint is reported in the registry record. Therefore, the precision of the planned correlation cannot be assessed from this record.

Why does the study structure matter?

The registry describes a parallel, non-randomized design with no masking. Parallel arms ordinarily involve distinct groups receiving different study regimens over the same period. Randomization, in general, helps make treatment groups comparable by distributing measured and unmeasured characteristics across them. Because this study was not randomized, and because Arm B had no enrolled participants, the record does not contain a populated second group for a like-for-like treatment comparison. The primary analysis is consequently best understood as a planned association analysis within the available study population, not as a randomized test of treatment assignment.

The absence of masking can matter in general when judgments, measurements, or reporting may be influenced by knowledge of treatment. Whether that concern affected MRE measurement or other assessments in this study is not reported in the registry record. The record also does not report whether image assessment was performed independently or whether any procedures were used to reduce measurement variability.

What do planning calculations represent?

In general, a sample-size calculation uses assumptions about an effect, variability, significance level, and desired power to illustrate how much information a future study might need. The registry describes planning illustrations based on a fraction of an observed effect, a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. Such illustrations are not the trial’s own design justification and are not observed power. They also do not establish that the study achieved adequate precision. Equal allocation between two arms is a planning assumption, while no participants were enrolled in Arm B; readers should not interpret the illustration as a description of the enrollment actually achieved.

For time-to-event outcomes, the registry’s calculation note refers to total events using the Schoenfeld approach; for other outcomes, it refers to participants per arm. In general, event-based planning reflects that information in survival analyses depends on observed events as well as participant counts. The registry does not report the actual planning inputs, effect value, or a trial-specific justification in the record. These calculations therefore serve only as illustrations of hypothetical planning assumptions.

Finally, the primary outcome timeframe was up to 2 years, but the registry does not report follow-up completeness, censoring details, or the handling of participants who lacked an evaluable stiffness measurement. Those design and analysis details affect how an association with OS should be interpreted and how reproducible it is. They are not reported in the registry record.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists three outcome measures and three analyses. It identifies correlation of HCC stiffness to OS as the primary outcome, but the information in the registry record does not give enough detail to describe the other outcomes or the full set of analyses. It does not report whether any adjustment for multiple testing was planned. When several outcomes or statistical tests are examined, the chance of at least one apparently unusual result can increase; in general, multiplicity procedures are used to control or describe that issue. Without the analysis details and any stated adjustment strategy, the primary p-value criterion should not be treated as a complete account of how multiple results were handled.

The record does not report how missing stiffness measurements, incomplete survival follow-up, or other missing outcome data were handled. It does not describe imputation, exclusions, or sensitivity analyses for missing data. In general, the implications of missingness depend on how much information is absent and why it is missing. Those details are necessary to understand whether the analyzed participants could differ systematically from those without evaluable measurements.

It is also not reported in the registry record which analysis population was used for the primary outcome, how many participants contributed to the correlation, or whether the analysis included every enrolled participant. The listed enrollment of 15 should not automatically be taken as the endpoint analysis count. The record does not report interim analyses, stopping rules for the study as a whole, or a statistical analysis plan that resolves these questions.

11. 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

For Arm A, the registry lists 15 participants at risk for serious adverse events and 0 with a serious adverse event. It lists 15 at risk for other adverse events and 3 affected. It also records 2 deaths. These are counts in the registry’s adverse-event categories, not estimates of comparative risk against another treatment.

The categories should be kept distinct: the reported count of serious adverse events is not the same as the count of other adverse events or deaths. The registry information does not say that the 2 deaths were caused by pembrolizumab, nor does it report their relationship to treatment in the registry record in the record. Likewise, the absence of recorded serious adverse events does not show that no serious events could occur; it describes the record for this study group.

Arm B has zero participants at risk and zero events listed. Because no patients were enrolled into that arm, its zero counts are not evidence of lower risk and cannot be compared meaningfully with Arm A. The registry record does not provide the event terms, severity details for the other adverse events, or a treatment-relatedness breakdown in the information.

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

The registry record describes a completed Phase II study of pembrolizumab in advanced HCC after progression on or intolerance to first-line systemic therapy. Its primary scientific question concerned the association between MRE-measured HCC stiffness and OS. The record explains that descriptive statistics and Spearman correlation were planned and that p-value < 0.05 was considered statistically significant. It does not report the correlation estimate, p-value, confidence interval, or supporting outcome summaries. Therefore, the posted information does not establish whether stiffness was associated with survival, in which direction, or with what precision.

The study’s design and enrollment also shape what can be concluded. Allocation was non-randomized, and Arm B closed after activation without enrolling patients. The record therefore does not provide a randomized comparison of pembrolizumab with another treatment or of pembrolizumab alone versus the combination regimen. It cannot establish a causal treatment effect from a between-arm contrast that was not observed.

The adverse-event record lists no serious adverse events among 15 participants at risk in Arm A, 3 participants with other adverse events, and 2 deaths. These counts are descriptive. The record does not identify the deaths as treatment-related, and the information does not support attributing either benefit or harm to the treatment. Arm B’s zero counts reflect its lack of enrollment, not an observed safety experience.

In general, a study can be informative about the feasibility of collecting measurements and about patterns observed in its participants, even when it cannot establish a causal treatment effect. Here, however, the primary association estimate and several analysis details are not reported in the registry record. Readers should distinguish the stated scientific question and methods from an actual result: the record describes what was intended, but does not provide enough primary outcome information to judge the observed stiffness–survival relationship.

13. Why This Trial Matters Statistically

This trial is statistically notable because its primary question concerns an imaging measure and survival, rather than a direct randomized comparison of treatments. Correlation analysis can be appropriate for describing whether two measures vary together, but it answers a different question from whether a therapy changes an outcome. Keeping those questions separate helps prevent an association from being mistaken for a treatment effect or a causal mechanism.

The study also illustrates why the enrolled population and the design actually realized matter more than the planned arm descriptions alone. Two regimens were listed, but the registry says Arm B enrolled no patients. Consequently, the study did not produce an observed comparison between the pembrolizumab regimen and the pembrolizumab-plus-antiviral regimen. The empty arm cannot be interpreted as a control group or as safety evidence.

For statistical interpretation, the absent primary estimate is consequential. The planned use of Spearman correlation and a stated significance threshold do not reveal the association’s size, uncertainty, or direction. Reporting those quantities, along with analysis counts and missing-data handling, would be needed to assess the strength and reliability of the endpoint result. The registry record does not report them.

The safety counts also demonstrate the value of interpreting denominators and event categories together. Zero serious adverse events in the enrolled Arm A group is a descriptive count, while the deaths and other adverse events are separate categories that require their own context. Counts from an arm with no participants cannot be treated as comparative evidence. These distinctions are central to reading small, non-randomized clinical studies without overstating what their data can establish.

14. 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 endpoint is described as a comparison of change in HCC stiffness with OS, but the registry does not report the definition of stiffness change, measurement schedule, survival representation, or the number of observations included in the correlation. The coefficient, p-value, confidence interval, and numerical endpoint summaries are also not reported in the registry record. This prevents assessment of the result’s direction, magnitude, precision, or robustness.

Although two arms were described, Arm B enrolled no patients. This removes the possibility of an observed comparison between the regimens and means that the adverse-event counts for Arm B cannot serve as comparative safety evidence. The study was non-randomized and unmasked, and the registry does not describe procedures that would address possible imbalances or measurement-related influences.

The record does not report the analysis population, the number of participants with evaluable primary endpoint data, or the approach to missing measurements and incomplete follow-up. Nor does it give the full statistical analysis plan, details of the other listed outcomes and analyses, or any multiplicity adjustment. These omissions limit reproducibility and make it difficult to assess whether the primary analysis was prespecified and how it relates to other analyses.

The adverse-event information is limited to counts by broad category in the record. Event terms, details of severity for other adverse events, and attribution of deaths are not reported in the registry record. Finally, the registry does not report the trial’s own sample-size justification. The planned sample-size illustrations are not a substitute for that justification, particularly given the absence of enrollment in one planned arm.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Pembrolizumab (MK-3475) in Hepatocellular Carcinoma — NCT02940496.
• Registry record last updated 2025-07-15; status last verified 2025-07.
• Results first posted 2025-07-15.
• 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.

18. Record Summary

NCT02940496 is a completed Phase II, parallel, non-randomized study sponsored by M.D. Anderson Cancer Center. It enrolled 15 participants and evaluated pembrolizumab for advanced HCC after progression on or intolerance to first-line systemic therapy. Arm A received pembrolizumab; Arm B was planned to add elbasvir/grazoprevir and ribavirin, but closed after activation without enrolling patients.

The primary outcome concerned the relationship between MRE-measured HCC stiffness and OS over up to 2 years. Descriptive statistics and Spearman correlation were specified, with p-value < 0.05 considered statistically significant. The primary estimate and its uncertainty are not reported in the registry record.

For Arm A, the record lists 0 serious adverse events, 3 participants with other adverse events, and 2 deaths. These counts do not establish causation or provide a comparison with Arm B.