PD-L1 Testing and Standardization

The discovery that tumors exploit the PD-1/PD-L1 immune checkpoint to evade T-cell killing, recognized by the 2018 Nobel Prize to James Allison and Tasuku Honjo, launched a revolution in cancer immunotherapy. Early trials of anti-PD-1/PD-L1 agents showed that PD-L1 protein expression on tumor cells correlated with treatment response, suggesting its potential as a predictive biomarker. However, each checkpoint inhibitor was developed with its own companion diagnostic IHC assay using different antibody clones, different scoring systems, and different positivity thresholds. Pembrolizumab used the 22C3 clone with tumor proportion score (TPS), nivolumab used 28-8, atezolizumab used SP142 with immune cell scoring, and durvalumab used SP263. This created unprecedented complexity for pathology laboratories.

The Blueprint PD-L1 IHC Assay Comparison Project, an international collaborative effort, systematically compared the four commercial PD-L1 assays on the same tumor samples. The Phase 1 results (2017) showed that three assays (22C3, 28-8, SP263) were largely interchangeable for tumor cell scoring, but SP142 consistently stained fewer tumor cells. Immune cell scoring was poorly concordant across all assays. The Phase 2 study extended these findings to a larger cohort and added pathologist interobserver variability assessment, finding that pathologist agreement was moderate for tumor cell scoring (ICC 0.83) but poor for immune cell scoring (ICC 0.20). These studies revealed that PD-L1 testing was not a single standardized biomarker but a fragmented ecosystem where the choice of assay, scoring method, and cutoff fundamentally influenced patient eligibility for immunotherapy.

As checkpoint inhibitor indications expanded beyond NSCLC to gastric, urothelial, cervical, head and neck, and triple-negative breast cancers, the combined positive score (CPS) emerged as an additional scoring system that counts both tumor cells and immune cells expressing PD-L1. The KEYNOTE-355 trial used CPS>=10 to select triple-negative breast cancer patients for pembrolizumab, while KEYNOTE-811 used CPS>=1 in gastric cancer. This proliferation of scoring systems (TPS, CPS, IC score) across different tumor types with different cutoffs created enormous challenges for pathology practice. Studies explored whether laboratory-developed tests could substitute for FDA-approved companion diagnostics, with mixed regulatory and clinical acceptance. Digital pathology and AI-based PD-L1 scoring emerged as potential solutions to the reproducibility challenge.

The NCCN guidelines require PD-L1 testing for treatment decisions in multiple tumor types, specifying the appropriate assay and scoring system for each indication. The CAP has published guidelines on PD-L1 testing validation and quality assurance. The IASLC Atlas of PD-L1 Testing provides pathologist training resources. European guidelines from ESMO recommend PD-L1 testing while acknowledging its limitations as an imperfect biomarker, noting that some PD-L1-negative patients still benefit from immunotherapy and some PD-L1-positive patients do not respond.

PD-L1 testing remains the most widely used but imperfect biomarker for immunotherapy selection. The field is moving toward multi-biomarker approaches combining PD-L1 with TMB, MSI status, gene expression profiling, and tumor-infiltrating lymphocyte assessment for more accurate prediction of immunotherapy benefit. AI-powered digital pathology tools are being developed to standardize PD-L1 scoring and extract additional predictive features from H&E-stained slides. The regulatory framework for companion diagnostics continues to evolve, with ongoing debate about whether laboratory-developed tests can replace FDA-approved assays. Some experts advocate moving beyond PD-L1 entirely, arguing that combination immunotherapy strategies may render single-biomarker selection obsolete. The fundamental tension between regulatory requirements for specific assays and clinical needs for practical, harmonized testing remains unresolved.