Background
Phase III double-blind RCT (KEYNOTE-412). N=804, locally advanced (unresectable or resectable) HNSCC (oropharynx, oral cavity, larynx, hypopharynx), squamous histology, ECOG PS 0–1. Enrolled 2017–2021, 195 global sites. Pembrolizumab added to cisplatin-based CRT and continued as maintenance, testing whether checkpoint inhibition + CRT improves EFS over CRT alone. Definitive (unresectable) or adjuvant (post-op high-risk) settings.
Interventions and follow up
Arm A: Pembrolizumab 200 mg q3wk + cisplatin 100 mg/m² D1/D22 + IMRT (66–70 Gy definitive; 60–66 Gy adjuvant) → adjuvant pembrolizumab × 14 cycles
Arm B: Placebo + cisplatin + IMRT → placebo maintenance × 14 cycles
Primary endpoint: EFS
mFollow up: 47.7 month
Arm B: Placebo + cisplatin + IMRT → placebo maintenance × 14 cycles
Primary endpoint: EFS
mFollow up: 47.7 month
Results
EFS: HR 0.83 (95% CI 0.68–1.01), one-sided P=.066 — prespecified boundary P<.025 not crossed; primary endpoint not met
OS: not yet mature at analysis
PD-L1 CPS ≥1 subgroup EFS: HR 0.82 (95% CI 0.66–1.02) — not significant
PD-L1 CPS ≥20 subgroup EFS: HR 0.73 (95% CI 0.54–0.99) — exploratory, not formally tested
OS: not yet mature at analysis
PD-L1 CPS ≥1 subgroup EFS: HR 0.82 (95% CI 0.66–1.02) — not significant
PD-L1 CPS ≥20 subgroup EFS: HR 0.73 (95% CI 0.54–0.99) — exploratory, not formally tested
Adverse events
Grade ≥3 AEs: 78.1% (pembrolizumab) vs 73.0% (placebo)
Immune-mediated AEs (any grade): 36.9% vs 14.4%
Grade ≥3 immune AEs: 9.8% vs 1.5%
Discontinuation due to AEs: 18.4% vs 9.0%
Immune-mediated AEs (any grade): 36.9% vs 14.4%
Grade ≥3 immune AEs: 9.8% vs 1.5%
Discontinuation due to AEs: 18.4% vs 9.0%
Conclusions
Adding pembrolizumab to definitive or adjuvant CRT did not significantly improve EFS in unselected LA-HNSCC (HR 0.83, P=.066). The trial was negative for its primary endpoint. Exploratory signals in PD-L1 high (CPS ≥20) patients warrant further investigation but cannot support practice change.
Key Limitations
Pre-specified one-sided alpha of 0.025 not crossed despite a consistent HR of 0.83 — clinically meaningful but statistically non-significant. PD-L1 CPS ≥20 subgroup showed HR 0.73 (exploratory); this population may benefit. Primary endpoint was EFS rather than OS. Concurrent RT + IO may create complex immunologic interactions limiting additive benefit. CheckMate 651 (nivolumab + ipilimumab) also failed in R/M first-line.
Clinical Context
Joins a series of negative or non-definitive trials adding IO to CRT in LA-HNSCC. Standard of care remains cisplatin-based CRT without checkpoint inhibitors for definitive LA-HNSCC and high-risk post-op HNSCC. KEYNOTE-689 (perioperative pembro in resected LA-HNSCC) was positive (2025), a different treatment approach. Ongoing trials test IO in biomarker-enriched populations.