AttendMe Owl Logo
AttendMe
Evidence Evolution
PathologyPathology

How This Evidence Evolved

PD-L1 Testing and Standardization

From discovery of a biomarker to the challenge of multiple assays and scoring systems

15.3

Timeline

Signal

Early observations and pilot data that first suggested a new direction

PD-L1 expression emerged as a treatment-response biomarker in lung cancer. KEYNOTE-001 studied pembrolizumab in 495 patients with advanced NSCLC. PD-L1 expression on at least 50% of tumor cells was selected in a training group; response was 45.2% in the corresponding validation subgroup. Overall response across all patients was 19.4%. The Blueprint project then compared assays 22C3, 28-8, SP142, and SP263. Assay and cutoff choices could change patient classification, making analytical comparability and clinical interpretation central to pathology practice.
Proof

Landmark RCTs and pivotal trials that established the evidence base

Blueprint Phase 1 compared four PD-L1 assays in NSCLC. Tumor-cell staining was comparable with 22C3, 28-8, and SP263, while SP142 stained fewer tumor cells. Immune-cell staining varied more. Despite similar analytical performance, changing assays and cutoffs changed classification in some patients. Phase 2 evaluated 81 lung cancer specimens with five assays. Pathologist reliability was very strong for tumor-cell scoring (ICC 0.86–0.93) and poor for immune-cell scoring (ICC 0.18–0.19). Glass-slide and digital-image scores were highly concordant (Pearson correlation >0.96). These findings require distinction between analytical agreement and therapy-specific classification.
Extension

Follow-up studies, subgroup analyses, and real-world validation

KEYNOTE-355 assessed pembrolizumab plus chemotherapy in previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer. Its prespecified analyses included CPS ≥10, CPS ≥1, and the intention-to-treat population. CPS ≥10 was an analysis population, not the enrollment threshold. In that group, median PFS was 9.7 versus 5.6 months with placebo-chemotherapy (HR 0.65). KEYNOTE-811 tested pembrolizumab added to trastuzumab and chemotherapy for HER2-positive gastro-oesophageal cancer. Its interim report described improved PFS, specifically in tumors with CPS ≥1; overall-survival follow-up continued. Scoring thresholds must be interpreted within each trial and indication.
Now

Current standard of care and ongoing research directions

PD-L1 assay comparison and scoring assistance address different questions. Blueprint Phase 1 found that changing assays and cutoffs changed classification in 37% of the 38 compared cases. Phase 2 found strong reliability for tumor-cell scoring but poor reliability for immune-cell scoring. In a separate validation study of 543 urothelial carcinoma cases, complete pathologist agreement increased from 82.1% to 93.9% after AI-guided reassessment. These findings concern analytical classification and reader agreement; they do not establish improved treatment outcomes. Can combining PD-L1 with TMB, MSI, gene-expression or lymphocyte measures improve prediction? Can H&E-based methods add useful information? Which alternative assays meet the requirements for a particular treatment? Those questions require their own validation and regulatory evidence.

Landmark Trials in This Story

KEYNOTE-001 NSCLC2015

Pembrolizumab for the treatment of non-small-cell lung cancer.

Validation ORR, PD-L1 ≥50%: 45.2%.

The New England journal of medicinePhase 1 studyN=495
Blueprint Phase 12017Landmark

PD-L1 Immunohistochemistry Assays for Lung Cancer: Results from Phase 1 of the Blueprint PD-L1 IHC Assay Comparison Project

Concordance of PD-L1 scoring across four assays

Journal of thoracic oncology : official publication of the International Association for the Study of Lung CancerDiagnostic comparison studyN=39
Blueprint Phase 22018

PD-L1 Immunohistochemistry Comparability Study in Real-Life Clinical Samples: Results of Blueprint Phase 2 Project

Inter-assay concordance and interobserver agreement

Journal of thoracic oncology : official publication of the International Association for the Study of Lung CancerDiagnostic comparison studyN=81
KEYNOTE-3552020Landmark

Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial

Progression-free survival in PD-L1 CPS>=10 subgroup

LancetRCTN=847
KEYNOTE-8112023

Pembrolizumab plus trastuzumab and chemotherapy for HER2-positive gastric or gastro-oesophageal junction adenocarcinoma: interim analyses from the phase 3 KEYNOTE-811 randomised placebo-controlled trial.

PFS and OS (KEYNOTE-811 interim).

Lancet (London, England)RCTN=698
AI-assisted CPS validation2024

An artificial intelligence-powered PD-L1 combined positive score (CPS) analyser in urothelial carcinoma alleviating interobserver and intersite variability.

Complete pathologist agreement in 543 urothelial cases: 82.1% before versus 93.9% after AI-guided reassessment.

HistopathologyValidation studyN=543

Explore the evidence yourself

Ask AttendMe about any trial, guideline, or clinical question. Evidence-ranked answers from 3.7M+ peer-reviewed articles.

Frequently Asked Questions

Why are there so many different PD-L1 assays?+
PD-L1 assays were independently developed and validated in checkpoint inhibitor trials. Blueprint compared 22C3, 28-8, SP142 and SP263, finding broadly similar tumor-cell staining for three assays but less staining with SP142. Assay and cutoff choices changed classification in some cases. This supports careful attention to the validated assay, scoring method and treatment context.
What is the difference between TPS and CPS scoring?+
Tumor Proportion Score (TPS) counts only PD-L1-positive tumor cells as a percentage of all viable tumor cells.
Can AI improve PD-L1 scoring reproducibility?+
Evidence for AI assistance is assay- and tumor-specific. A study validated an AI-powered PD-L1 22C3 CPS analyser in 543 urothelial carcinoma cases. Complete agreement among three pathologists was 82.1%; after AI-guided reassessment it increased to 93.9%. This supports scoring assistance in that setting and does not establish one performance level for all tumor types or assays.

Medical Disclaimer: This content is for educational purposes only and does not constitute medical advice. Clinical decisions should always be based on individual patient assessment, local guidelines, and professional judgement.

All data sourced from published, peer-reviewed articles and clinical practice guidelines.

Claims checked against cited sources: 10 November 2026