Skip to content

KEYNOTE-937 — Adjuvant pembrolizumab post-curative-intent HCC

Headline: Adjuvant pembrolizumab after curative-intent resection or ablation of hepatocellular carcinoma failed to improve recurrence-free survival vs placebo. Negative readout — preserves active-surveillance SOC and is a meaningful counter-signal for the broader "perioperative IO extrapolation" thesis in liver cancer.

At a glance

  • Sponsor / PI / institution: Merck; presented by Pierre Gholam / Masatoshi Kudo (Kindai University) — KEYNOTE-937 investigators
  • NCT ID: NCT03867084
  • Phase / arms / N: Phase 3, randomized double-blind; pembrolizumab vs placebo for up to 17 cycles post-curative-intent therapy; N ~950
  • Primary endpoint(s): RFS (BICR); OS key secondary
  • Indication / line: Adjuvant HCC after surgical resection or local ablation with curative intent, intermediate/high recurrence risk
  • Comparator: Placebo (over active surveillance)
  • Session: Day 2 (Fri Jan 9, 2026), Hepatobiliary General Session, Moscone West

Headline result

KEYNOTE-937 did not meet the primary RFS endpoint — pembrolizumab failed to demonstrate a statistically significant RFS improvement over placebo in the adjuvant HCC population. OS data immature at this analysis; safety consistent with known pembrolizumab profile. The result aligns with the prior IMbrave-050 atezo-bev adjuvant HCC negative readout (RFS positive at interim but did not hold; sponsor withdrew).

Mechanism / class

Pembrolizumab is an anti-PD-1 monoclonal antibody. The adjuvant HCC hypothesis was extrapolated from advanced-HCC IO success (atezo-bev IMbrave150, durvalumab+treme HIMALAYA), where checkpoint blockade clearly improves OS in unresectable disease.

Trial design deep dive

Study architecture. Phase 3 randomized 1:1 double-blind placebo-controlled; pembrolizumab 400 mg Q6W for up to 17 cycles (~2 years) post-curative-intent resection or local ablation. Stratification by recurrence-risk category (intermediate vs high based on tumor size, number, MVI), curative modality (resection vs ablation), and region (Asia vs ex-Asia). N ~950 globally.

Endpoint definitions. Primary endpoint RFS by BICR — defined as time from randomization to first documented recurrence (intra- or extra-hepatic) or death. OS was a key secondary tested hierarchically only if RFS hit. RFS censoring rules included loss to follow-up and start of new anti-cancer therapy. Sensitivity analyses included investigator-assessed RFS and recurrence-risk-stratified RFS. TBD on whether RFS event count was at protocol-specified maturity threshold.

Key subgroup analyses.

  • Intermediate vs high recurrence risk: TBD — both subgroups failed to show benefit per topline; no enrichment for high-risk subset.
  • Curative modality (resection vs ablation): TBD — sample sizes likely too small to declare a directional signal.
  • Etiology (HBV vs HCV vs NASH/alcohol): of high mechanistic interest given different immune microenvironments; TBD on stratified HRs.
  • PD-L1 CPS / TMB: exploratory biomarker analyses pending; no upfront biomarker enrichment in design.

Comparator context. Placebo (effectively active surveillance) — the prevailing post-curative HCC SOC. IMbrave-050 (atezo + bev) was the prior adjuvant HCC Phase 3, with positive interim RFS that did not hold at update; sponsor withdrew the supplementary indication. KEYNOTE-937 is the second consecutive negative adjuvant-IO HCC Phase 3.

FDA / regulatory implications. No filing planned for adjuvant HCC indication. Merck retains pembrolizumab approvals in advanced HCC (post-sorafenib monotherapy and in combination with lenvatinib via LEAP-002 framework). Negative readout reinforces FDA caution on perioperative-IO label extensions in HCC.

Discussant takeaway

Hepatobiliary discussants framed the result as the second consecutive negative adjuvant-IO HCC Phase 3 (after IMbrave-050). The pattern argues that the post-resection HCC immune microenvironment is mechanistically distinct from advanced HCC — adjuvant IO does not extrapolate cleanly. Bears directly on perioperative-IO design across other tumor types where similar extrapolations are being tested.

Practice-changing verdict

No — preserves SOC. Active surveillance with imaging remains the post-curative HCC standard. The negative signal is itself practice-relevant: it removes the empirical basis for off-label adjuvant pembrolizumab in HCC.

  • ASCO 2026 — perioperative IO theme (KEYNOTE-B15 bladder, PROTEUS prostate, LIBRETTO-432 lung)
  • AACR 2026 — HCC immune microenvironment / adjuvant-IO mechanism sessions
  • ESMO 2026 — long-term OS follow-up expected
  • Prior: IMbrave-050 (atezo-bev adjuvant HCC, negative)

Sources