Dr Chan on RFS Outcomes With Adjuvant Pembrolizumab vs Placebo After Complete Radiological Response in HCC

“[These findings], together with other studies, confirm that perhaps the adjuvant PD-1 [inhibition] is not the direction we [should] move [towards] to help our patient with early-stage HCC.”

Stephen Lam Chan, MBBS, MD (CUHK), FRCP, a clinical professor in the Department of Clinical Oncology at the Chinese University of Hong Kong, reviewed interim findings from the phase 3 KEYNOTE-937 study (NCT03867084) evaluating adjuvant pembrolizumab (Keytruda) vs placebo in patients with hepatocellular carcinoma (HCC) who achieved a complete radiologic response following surgical resection or local ablation.

Chan began by noting that KEYNOTE-937 was designed to address the substantial risk of recurrence in early-stage HCC, which persists despite curative-intent local therapy. This ongoing clinical challenge has fueled interest in improved adjuvant systemic strategies.

In the double-blind KEYNOTE-937 trial, patients were randomly assigned to receive 1 year of pembrolizumab or matching placebo. The study’s co-primary end points were recurrence-free survival (RFS)by blinded independent central reviewand overall survival (OS); the prespecified interim analysis focused on RFS outcomes.

During the 2026 Gastrointestinal Cancers Symposium, Chan explained that data from this interim analysis did not demonstrate a superior RFS benefit with adjuvant pembrolizumab over placebo. The median RFS was 46.7 months (95% CI, 35.6-53.3) in the pembrolizumab arm vs 45.5 months (95% CI, 35.6-58.0) in the placebo arm, corresponding to a HR of 1.06 (95% CI, 0.88-1.26; P = .719). He emphasized that the RFS curves were closely overlapping, with comparable landmark outcomes, including 24-month RFS rates of 63% vs 61% in the pembrolizumab and placebo arms, respectively. The 48-month RFS rates were 50% in both arms.

Although individual point estimates varied, no consistent trend in favor of pembrolizumab over placebo was observed across subgroups at this time point, Chan added. Variables included age, sex, geographic region, viral vs nonviral etiology, BCLC stage at diagnosis, Child-Pugh score, AFP level, and type of prior local therapy (resection vs ablation), he highlighted.

Importantly, Chan noted that KEYNOTE-937 used a hierarchical statistical framework in which OS testing was contingent on demonstrating a positive RFS result. Because the interim RFS analysis did not meet the prespecified threshold for benefit, the protocol-specified formal comparison for OS will therefore not proceed. In Chan’s view, this outcome, while disappointing, provides clinically meaningful direction by suggesting that adjuvant PD-1 monotherapy may not be sufficient to meaningfully alter recurrence risk after curative-intent therapy in early-stage HCC.

From a safety standpoint, Chan underscored that no new safety signals were identified with pembrolizumab in this population, corroborating the expected tolerability profile of PD-1 inhibition in appropriately selected patients. He concluded that these findings reinforce the need to prioritize alternative, potentially perioperative approaches, different systemic partners, or novel adjuvant paradigms to improve long-term outcomes for patients at risk for recurrence following resection or ablation.