KEYNOTE Studies Examining PD-L1 Expression as a Biomarker for Pembrolizumab
The KEYNOTE clinical trial program represents the pivotal series of studies that established PD-L1 expression as a predictive biomarker for pembrolizumab (Keytruda) across multiple cancer types, with KEYNOTE-001, KEYNOTE-012, KEYNOTE-024, KEYNOTE-040, KEYNOTE-042, and KEYNOTE-048 being the most significant trials that examined this relationship.
Key KEYNOTE Studies by Disease Type
Non-Small Cell Lung Cancer (NSCLC)
KEYNOTE-001 was the foundational phase I study that first established the correlation between PD-L1 expression and pembrolizumab response in melanoma and NSCLC 1. This trial demonstrated that patients with PD-L1 tumor proportion score (TPS) ≥50% achieved a 45% response rate, with progression-free survival of 6.3 months 2. The study used the 22C3 antibody assay to assess membranous PD-L1 expression, establishing the MEL scoring scale (0-5) for melanoma 1.
KEYNOTE-024 was the landmark phase III trial that led to FDA approval of pembrolizumab monotherapy for first-line treatment of NSCLC with PD-L1 TPS ≥50% 3, 2. This study demonstrated superior overall survival (30.0 vs 14.2 months, HR 0.60) and progression-free survival (10.3 vs 6.0 months) compared to platinum-based chemotherapy 3. The trial established the 50% TPS threshold as the key biomarker cutoff for pembrolizumab monotherapy 2.
KEYNOTE-042 expanded pembrolizumab's indication to patients with PD-L1 TPS ≥1%, demonstrating improved overall survival (16.7 vs 12.1 months, HR not specified) compared to chemotherapy, though the benefit was primarily driven by the high PD-L1 expression subgroup 3, 4.
Head and Neck Squamous Cell Carcinoma (HNSCC)
KEYNOTE-012 was the phase Ib study that first evaluated pembrolizumab in recurrent/metastatic HNSCC 5. This trial demonstrated that using Combined Positive Score (CPS) rather than TPS significantly improved predictive value: patients with CPS-positive tumors had 22% objective response rate versus 4% in CPS-negative tumors (p=0.021), whereas TPS showed no significant difference (19% vs 16%, p=0.35) 5. This study established CPS as the superior biomarker scoring method for HNSCC 5.
The trial also showed that a gene expression profile (GEP) score based on interferon-gamma and five related genes (CXCL9, CXCL10, IDO1, HLA-DRA, STAT1) significantly correlated with response, with 40% ORR in high-GEP patients versus 5% in low-GEP patients 5.
KEYNOTE-040 was the phase III trial comparing pembrolizumab to standard chemotherapy in previously treated R/M HNSCC 5. Patients with TPS >50% achieved 26.6% ORR with significant improvements in OS and PFS 5. More importantly, patients with CPS >1 showed significant OS improvement (one-year OS 40.1% vs 26.1%, HR 0.74, p=0.0049) 5.
KEYNOTE-048 provided Category 1 evidence for PD-L1 CPS-based patient selection in first-line R/M HNSCC 5. This trial demonstrated that pembrolizumab monotherapy showed superior OS in patients with CPS ≥20 and CPS ≥1, and was non-inferior in the total population 5. Pembrolizumab plus chemotherapy demonstrated superior OS across all PD-L1 expression levels (CPS ≥20, ≥1, and total population) 5. This led to FDA approval with mandated PD-L1 testing using CPS ≥1 as the companion diagnostic threshold 5.
PD-L1 Assay Standardization
The KEYNOTE trials utilized the Dako PD-L1 IHC 22C3 pharmDx assay as the companion diagnostic 5. The FDA expanded the intended use of this assay to include CPS scoring for HNSCC patient selection 5.
Comparative studies showed >90% concordance between the three major assays (22C3, 28-8, and SP263) for tumor cell PD-L1 expression, though the Ventana SP142 assay showed lower tumor cell staining 5, 3.
Critical Biomarker Insights from KEYNOTE Studies
The KEYNOTE program established several key principles:
- CPS is superior to TPS in HNSCC: Including PD-L1 expression on tumor-infiltrating immune cells significantly improves predictive value beyond tumor cells alone 5
- PD-L1 expression is not absolute: Some PD-L1-negative patients still benefit from pembrolizumab, though at lower rates 5, 1
- Higher PD-L1 expression correlates with better outcomes: Response rates increase progressively with higher PD-L1 expression levels 3, 1
- Tumor mutational burden (TMB) and GEP scores provide independent predictive value: High TMB and high GEP scores independently predict response, with highest responses in patients with both 5
Important Caveats
The predictive value of PD-L1 varies by cancer type and scoring methodology. Intra- and inter-tumor heterogeneity can affect results 5. Different cutoff thresholds (≥1%, ≥10%, ≥25%, ≥50%) have been used across trials, making direct comparisons challenging 5. For HNSCC, 94% of experts define PD-L1 positivity as ≥1% TPS or ≥1 CPS, with 81% agreeing that CPS is the optimal biomarker testing method 5.