Adjuvant Pembrolizumab Fails to Produce RFS Benefit in HCC

Treatment with adjuvant pembrolizumab (Keytruda) failed to improve recurrence-free survival (RFS) compared with placebo in patients with hepatocellular carcinoma (HCC) who had a complete radiological response after surgical resection or local ablation, according to data from the phase 3 KEYNOTE-937 trial (NCT03867084).¹

Findings presented at the 2026 Gastrointestinal Cancers Symposium demonstrated that patients treated with pembrolizumab (n = 476) experienced a median RFS of 46.7 months (95% CI, 35.6-53.3) compared with 45.5 months (95% CI, 35.6-58.0) for patients treated with placebo (n = 483; HR, 1.06; 95% CI, 0.88-1.26; P = .719). The 24- and 48-month RFS rates in the pembrolizumab arm were 63% and 50%, respectively; these respective rates were 61% and 50% in the placebo arm.

“The study did not proceed to final analysis as the RFS hypothesis was not met, and per multiplicity, overall survival [OS] was not tested statistically,” lead study author Stephen Lam Chan, MBBS, MD, FRCP, FHKCP, FHKAM, said in a presentation of the data.

Chan is a clinical professor in the Department of Clinical Oncology of the Chinese University of Hong Kong.

What was the rationale and design of the KEYNOTE-937 trial?

For patients with newly diagnosed HCC, surgical resection and local ablation represent potential curative strategies; however, tumor recurrence remains frequent among this patient population, with adjuvant therapy representing a potential method to lower disease recurrence rates and improve OS.

KEYNOTE-937 was a randomized, double-blind, placebo-controlled trial that enrolled patients at least 18 years of age with confirmed HCC who achieved complete radiological response following resection or local ablation. Patients needed to have an ECOG performance status of 0 or 1 and a Child-Pugh class A score. Patients with esophageal or gastric variceal bleeding within the last 6 months; those with clinically apparent ascites; and patients with clinically diagnosed hepatic encephalopathy within 6 months of enrollment were excluded from the study.²

Enrolled patients were randomly assigned 1:1 to receive pembrolizumab at 200 mg once every 3 weeks or matching placebo.¹ Treatment continued for up to 1 year, or until disease recurrence, unacceptable toxicity, intercurrent illness, or patient withdrawal.

Patients were stratified by region (Asia [excluding Japan] vs rest of world), prior local therapy (resection vs ablation), risk of recurrence (intermediate risk vs high risk vs very high risk), and alpha fetoprotein (AFP) level at diagnosis prior to resection or ablation (<200 ng/mL vs ≥200 ng/mL).

RFS by blinded independent central review (BICR) and OS served as the trial’s primary end points. Distant metastasis–free survival (DMFS) per BICR or pathology assessment and safety were secondary end points.

What were the baseline characteristics of the enrolled population?

The median age was 62 years (range, 25-85) in the pembrolizumab arm (n = 476) vs 63 years (range, 27-85) in the placebo arm. Most patients in both groups were male (pembrolizumab, 79%; placebo, 80%), Asian (56%; 54%), had an ECOG performance status of 0 (94%; 95%), had hepatitis B (60%; 59%), had AFP levels below 200 ng/mL at initial diagnosis (77%; 77%), had a Child-Pugh score of A5 (93%; 90%), did not undergo prior ablation (87%; 87%), underwent surgical resection as prior local therapy (88%; 88%), had high-risk disease (74%; 71%), had a Barcelona Clinic Liver Cancer staging of A (83%; 83%), and had microvascular invasion (57%; 58%).

What additional data from the study were presented from KEYNOTE-937?

Although OS was a co-primary end point, the statistical design of the study allowed for formal testing of OS if the RFS hypothesis was met. Since it was not, investigators did not test OS statistically.

Findings did show that the median OS was not reached (NR; 95% CI, NR-NR) in both arms (HR, 1.08; 95% CI, 0.81-1.43). The 24- and 48-month OS rates in the pembrolizumab arm were 90% and 79%, respectively; these respective rates were 94% and 81% in the placebo arm.

DMFS outcomes were similar between the 2 groups, with patients treated with pembrolizumab achieving a median DFMS that was NR (95% CI, 58.7-NR) compared with NR (95% CI, 59.0-NR) in the placebo arm (HR, 0.98; 95% CI, 0.77-1.24). The 24-month DMFS rates were 82% for pembrolizumab vs 83% for placebo; these respective rates at 48 months were 71% and 70%.

Regarding safety, any-grade adverse effects (AEs) were reported in 91% of patients in the pembrolizumab arm (n = 471) vs 82% of patients in the placebo group (n = 482). AEs led to death in 1% of patients in both arms.

The rates of treatment-related AEs (TRAEs) of any grade were 60% for pembrolizumab vs 35% for placebo, including rates of grade 3/4 TRAEs of 14% and 5%, respectively. No TRAEs led to death in either group, although 10% of patients in the pembrolizumab arm and 1% of patients in the placebo arm discontinued treatment due to TRAEs.

Any-grade immune-related AEs and infusion reactions occurred in 26% of patients in the pembrolizumab arm vs 7% of patients in the placebo arm; these AEs were reported at grade 3/4 in 7% of patients treated with pembrolizumab vs 1% of patients given placebo.

References

1. Chan SL, Bouattour M, Yau T, et al. Adjuvant pembrolizumab for participants with hepatocellular carcinoma and complete radiologic response after surgical resection or local ablation: The phase 3 KEYNOTE-937 study. J Clin Oncol. 2026;44(suppl 2):477. doi:10.1200/JCO.2026.44.2_suppl.477
2. Safety and efficacy of pembrolizumab (MK-3475) versus placebo as adjuvant therapy in participants with hepatocellular carcinoma (HCC) and complete radiological response after surgical resection or local ablation (MK-3475-937 /​ KEYNOTE-937). ClinicalTrials.gov. Updated August 11, 2025. Accessed January 9, 2026. https://clinicaltrials.gov/study/NCT03867084